Lecture Notes:

Relativistic Quantum Mechanics

Frank Krauss1
1
Institute for Particle Physics Phenomenology, Physics Department,
Durham University, Durham DH1 3LE, UK

Last Updated: May 4, 2021

1
Disclaimer
These lecture notes accompany the final-year undergraduate lecture course
on “Relativistic Quantum Mechanics”, consisting of 12 lectures, delivered
during the 2020/2021 academic year in an online format. The notes are
by no means original. Instead I shamelessly borrowed from a multitude of
resources, and tried to put material into a coherent form. I also try and
cite online links to books etc., where possible – I am aware that the library
holds some of them, also in electronic form.
The lectures mainly deal with second quantisation, a topic that has been
excellently covered in the literature and on the web. There are many truly
excellent textbooks on the topic, often named “Introduction to Quantum
field Theory” or similar, for example the books by Peskin & Schröder [1],
Griffiths [2], Schwartz, Zee, or Hatfield [3], in addition to a multitude of
freely available lectures notes on the web:

ˆ Mark Srednicki’s notes on Quantum Field Theory [4], which have since
been published as a book;

ˆ David Tong’s lectures on Quantum Field Theory [5];

ˆ lecture notes of the great Sidney Coleman on Quantum Field The-

ory [6];

ˆ Jeff Dror’s summary of practically all relevant relations worked out in

this course, and, in fact, many more beyond it, can be found in [7].

The lecture notes will be continuously updated over the course of the year -
please check the date on the front page to keep track of changes. When you
compare the notes with books you will realise that notation and conventions
differ between different resources. However, quite often these differences boil
down to trivial normalisations. I’ve tried, hopefully successfully, to be at
least self-consistent.
The notes are supplemented with worked examples and problems through-
out, and I cannot overemphasise how important it is to actually calculate
things on your own. Tougher, expert-level problems are identified with an
asterisk. They are outside the scope of examinable material and are solely
geared to helping interested students to develop a deeper understanding of
the subject and to contextualising the material in a wider perspective. I
have also added “extremely unbelievably hard” questions, indicated with
two asterisks. They cover material that is entirely beyond the scope of the
course, but may trigger some further reading and digging by students with
a soft spot for the abyss that is Quantum Field Theory.
Over the course of six weeks we will work through the analogue of 12 lectures
- I will try to highlight and explain crucial concepts in short movies with
me working through things on a white board - however, these movies are

i
 Week Chapters Comments & Suggested Problems
1 1-3 Reminder of important concepts. Introduction
to Classical Field Theory in Lagrange formalism:
real and complex scalar fields and electrodynam-
ics. Conserved current and conserved charge.
2 4 Logic of 2nd quantisation and first example: real
scalar field theory. 2nd quantisation of complex
scalar theory. More on conserved current and
charges, this time in the quantum world.
3 5 Introducing the Dirac equation without quantisa-
tion: Linearising Klein-Gordon equation, spinors,
their properties, and γ matrices. 2nd quantisation
of the Dirac equation: using anti-commutators for
the quantisation conditions on fermions.
4 6 Free electrodynamics fields. Impact of gauge in-
variance: “over-quantising”. 2nd quantisation in
Coulomb and Lorentz gauge.
5 7 Time-ordered products are the Green’s functions
(propagators) of free theories.
6 8 Interacting field theories. A first stab at the S-
matrix and Wick’s theorem. This is extended
reading and will not be subject of the relativis-
tic quantum mechanics part of the exam.

Table 1: Coverage of material during the course

by no means complete and they are mainly meant to structure your own,
self-driven learning. Below a table, Tab 1, of what material would have been
covered week-by-week.

ii
Contents
1 Introduction 1

2 Recapitulation 3
2.1 Natural Units . . . . . . . . . . . . . . . . . . . . . . . . . . . 3
2.2 Some mathematics . . . . . . . . . . . . . . . . . . . . . . . . 3
2.3 Four-Vectors and Minkowski Space . . . . . . . . . . . . . . . 4
2.4 Lorentz Transformations . . . . . . . . . . . . . . . . . . . . . 7
2.5 Lagrange and Hamilton Formalism for Point Particles . . . . 10
2.6 First Quantisation of the Harmonic Oscillator . . . . . . . . . 12
2.7 Problems & Solutions . . . . . . . . . . . . . . . . . . . . . . 15

3 Classical Fields 31
3.1 One-Dimensional Lattice . . . . . . . . . . . . . . . . . . . . . 31
3.2 Scalar Fields: Real Scalars . . . . . . . . . . . . . . . . . . . . 35
3.3 Scalar Fields: Complex Scalars . . . . . . . . . . . . . . . . . 37
3.4 Vector Fields: Maxwell’s Equations . . . . . . . . . . . . . . . 40
3.5 Hamiltonian Formulation . . . . . . . . . . . . . . . . . . . . 43
3.6 Problems & Solutions . . . . . . . . . . . . . . . . . . . . . . 45

4 Second Quantisation 60
4.1 A How-To Guide . . . . . . . . . . . . . . . . . . . . . . . . . 60
4.2 Second Quantisation of the Real Scalar Field . . . . . . . . . 61
4.3 A Little Detour: Causal Structure of the Theory . . . . . . . 67
4.4 Second Quantisation of the Complex Scalar Field . . . . . . . 69
4.5 Problems & Solutions . . . . . . . . . . . . . . . . . . . . . . 72

5 Fermions 87
5.1 The Dirac Equation . . . . . . . . . . . . . . . . . . . . . . . 87
5.2 Second Quantisation . . . . . . . . . . . . . . . . . . . . . . . 94
5.3 Problems & Solutions . . . . . . . . . . . . . . . . . . . . . . 101

6 Electrodynamics 121
6.1 Gauge Invariance as Obstacle . . . . . . . . . . . . . . . . . . 121
6.2 Coulomb Gauge . . . . . . . . . . . . . . . . . . . . . . . . . . 124
6.3 Lorentz Gauge . . . . . . . . . . . . . . . . . . . . . . . . . . 128
6.4 Problems & Solutions . . . . . . . . . . . . . . . . . . . . . . 132

7 Time-Ordered Products 142

7.1 Greens Functions: Non-Relativistic Quantum Mechanics . . . 143
7.2 Propagators in Quantum Field Theory: Scalar Fields . . . . . 146
7.3 Fermion Propagator . . . . . . . . . . . . . . . . . . . . . . . 148
7.4 Photon Propagator . . . . . . . . . . . . . . . . . . . . . . . . 150
7.5 Problems & Solutions . . . . . . . . . . . . . . . . . . . . . . 152

iii
8 Interacting Fields 164
8.1 Perturbative Expansion: Born Series . . . . . . . . . . . . . . 164
8.2 Interacting Field Theory: General Thoughts . . . . . . . . . . 165
8.3 Interacting Field Theory: λφ4 . . . . . . . . . . . . . . . . . . 168

9 List of Problems 178

iv
1 Introduction
In this course, “Relativistic Quantum Mechanics”, we combine Quantum
Mechanics with Special Relativity and develop a formalism to quantise fields
in a Lorentz-invariant way.
We will recapitulate the Lagrange and Hamilton formalism for the treat-
ment of classical point particles as well as the quantisation of the harmonic
oscillator through creation and annihilation operators. Building on the for-
mer, we will briefly analyse the Lagrange formalism and the derivation of
Euler-Lagrange equations of motion for a discrete system, before taking the
continuum limit, resulting in the Lagrange formulation of field dynamics.
We will analyse free real and complex scalar fields in this formalism, and
for the latter, we will find a symmetry – phase shifts of the fields – that
leaves the Lagrangian invariant. We will see that such invariances result in
conserved currents and charges. We will further exemplify the power of the
formalism by constructing a Lagrange density for the electromagnetic fields
and deriving Maxwell’s equations from it.
To quantise fields we will copy the steps known from single-particle systems,
in particular the harmonic oscillator, and adapt it to the case of fields. In
so doing we effectively replace the role of position and momentum of the
particle, and the corresponding operators, with the field and its conjugate
momentum. The resulting logic is to replace the functions describing fields
and their conjugate momenta with field and momentum operators, and to de-
mand suitable commutator relations for them. This is called second quanti-
sation. As a consequence of relativistic invariance, encoded in the quadratic
energy-momentum relation of E 2 − p2 = m2 , solutions with negative energy
become possible. Demanding a Hamiltonian with an energy spectrum that
is bounded from below, i.e. a physically meaningful ground state or vac-
uum, necessitates their interpretation as anti-particles. It also immediately
implies we have arrived at a multi-particle theory, because pairs of parti-
cles and anti-particles with short lifetimes can be produced. We will check,
by explicit calculation, that the resulting theory maintains causality at a
microscopic level, by asserting that commutators of causally disconnected
fields always vanish and that they therefore cannot impact onto each other.
After second quantisation of the simplest possible theory, a single free real
scalar field, we analyse the structure of a free complex scalar field. We will
recover the current and charge stemming from the phase invariance of the
Lagrangian and we will by explicit calculation show that the charge and
the Hamilton operators commute, making charge conservation of the theory
manifest.
After analysing the free scalar or Klein-Gordon fields we will turn our atten-
tion to the treatment of spin-1/2 particles in the celebrated Dirac equation.
We will analyse its structure and ingredients – γ matrices and spinors –
and their properties before second quantisation of the theory. Reflecting

1
the fermionic nature of the particles, we will use anti-commutators {·, ·}
instead of commutators [·, ·] for the quantisation conditions. Similar to the
case of the complex scalar field, also the Lagrangian for free spinors enjoys
invariance under phase transformations of the fields, and again this leads to
a conserved charge.
We then turn our attention to the quantisation of electrodynamics and the
free electromagnetic fields. There, we will encounter an interesting problem:
the vector potential Aµ , on which we build the theory, naively speaking, has
four degrees of freedom in its four-components, but the physical field has
only two degrees of freedom, the well-known linear or circular polarisation
states of the photons, the quanta of electromagnetism. This necessitates the
imposition of additional conditions onto the theory, to correctly reflect its
physical content. In more formalised language, this problem is a result of the
gauge invariance of the underlying theory, electromagnetism, which results
in identical physical fields for different vector potentials. It will become clear
that the problem of the additional content will be fixed by fixing the gauge
of the theory, and we will see how this is shapes the additional conditions
we will impose on the theory.
Having quantised various free field theories and discussing some of their
properties, we will start with developing a framework to analyse their dy-
namical behaviour. To this end we will build on the concept of Green’s
functions and construct the Green’s functions of our quantised theories. It
will turn out that these “propagators” are the vacuum expectation values
of time-ordered products of the field operators.

2
2 Recapitulation
In this section we recapitulate important concepts from previous lectures
and properties of the objects we will use throughout the lecture. The aim
is not to explain in detail how things work or why, but to provide you with
a unified notation and nomenclature. If necessary, please, re-familiarise
yourselves with the concepts in this section.
If you feel you need to read up on
ˆ tensors and indices, please, take a look at the lecture notes of Dulle-
mond and Peeters [8]; also chapter 3 of Griffiths’ book [2] or chapter 7
of Goldstein’s book [9] may be helpful, although the latter keep factors
of c.
ˆ Lagrange and Hamilton formalism and related problems, take a look
at the classical textbooks of Goldstein [9] (chapters 1 and 2 for La-
grange formalism, chapter 8 for Hamilton formalism, chapter 3 for
central force problem, and chapter 6 for oscillations) and Landau and
Lifshitz [10] (chapter 1 for Lagrange formalism, chapter 7 for Hamil-
ton formalism, chapter 3 for motion in a potential and chapter 5 for
oscillations).
ˆ harmonic oscillator in Quantum Mechanics, creation and annihila-
tion operators, maybe you may want to check Sec. 2.3 in Sakurai’s
book [11]?

2.1 Natural Units

Throughout the course we will use “natural units”,
~ = c = 1. (1)
All quantities will be expressed in units of energy, i.e. electron Volts (eV),
or their inverse. One eV is the kinetic energy an electron gains when being
accelerated from rest through an electric potential difference of 1 Volt in the
vacuum. To transform between quantities in different units, we will multiply
or divide by combinations of ~ and c, as in Table 2. In particular this means
we have the electron and proton mass as me ≈ 511 keV = 0.511 MeV and
mp ≈ 938 MeV ≈ 1 GeV.

2.2 Some mathematics

Fourier Transformation Throughout the lecture we will define Fourier
transformations between position x and momentum k in a somewhat asym-
metric form as
Z
dx −ikx
f˜(k) = e f (x)
(2π)

3
 time ←→ length with c ≈ 0.3 · 109 m/s
momentum ←→ energy with c
mass ←→ energy with c2
time ←→ 1/energy with ~ ≈ 6.5 · 10−22 MeV s
length ←→ 1/energy with ~c ≈ 200 MeV fm

Table 2: Transformations between physical quantities

Z
f (x) = dk eikx f˜(k) . (2)

The extension to higher dimensions – for example for the Fourier transfor-
mation of three-vectors – is straightforward:

d3 x −ik·x
Z
˜
f (k) = e f (x)
(2π)3
Z
f (x) = d3 k eik·x f˜(k) . (3)

δ-function The δ-function is defined through an integral relation as

Zb 
f (x0 ) if x0 ∈ [a, b]
dx δ(x − x0 ) f (x) = (4)
0 otherwise.
a

In addition, we have
Z
dx e−ix(k−q) = (2π)δ(k − q) (5)

Again, the extension to more dimensions is straightforward.

2.3 Four-Vectors and Minkowski Space

Four Vectors Throughout the course we will use relativistic notation.
Time t and spatial position x = (x, y, z) are combined into a (contravariant)
four-position

xµ = (t, x, y, z) = (t, x) (6)

and similar, energy E and momentum p = (px , py , pz ) are combined into a

(contravariant) four-momentum

pµ = (E, px , py , pz ) = (E, p) . (7)

We will use Greek indices µ, ν, ρ, . . . to label components of four-objects

and Latin indices i, j, k, . . . to label the spatial or three-components.

4
Einstein Convention When not stated otherwise we will use Einstein’s
convention of summing over repeated indices, for example

p2 = pi pi = p2x + p2y + p2z . (8)

Metric Tensor For four-vectors this is a meaningful operation only when

combining contravariant objects (where the index is a superscript) with co-
variant objects (where the index is a subscript). The two sets of four-objects
– contravariant and covariant – are connected through the metric tensor gµν ,

pµ = gµν pν and pµ = g µν pν , (9)

where the Minkowski metric is given by

 
1 0 0 0
 0 −1 0 0 
gµν = g µν = diag(1, −1) =  . (10)
 0 0 −1 0 
0 0 0 −1

In other words, if pµ = (E, p), pµ = (E, −p).

From pµ = gµν pν we can easily infer that
 
1 0 0 0
 0 1 0 0 
gµν = g µν = diag(1, 1) = 
 0
. (11)
0 1 0 
0 0 0 1

Raising and Lowering Indices We have already seen, how the metric
tensor is used to raise or lower indices of four-vectors, e.g.,

xµ = gµν xν and xµ = g µν xν , (12)

which introduces a sign flip in the spatial coordinates:

if xµ = (t, x) then xµ = (t, −x) . (13)

For tensors with n indices, one metric tensor is necessary to raise or lower
one index. For example, for a tensor F µν of rank two, two metric tensors are
necessary to lower both indices. As an example, consider the field-strength
tensor of electromagnetism, given by
 
0 −Ex −Ey −Ez
 Ex 0 −Bz By 
F µν =  . (14)
 Ey Bz 0 −Bx 
Ez −By Bx 0

5
Therefore,
0 0
Fµν = gµµ0 gν 0 ν F µ ν , (15)

where, making the sequence of matrix multiplications explicit

0 0 0 0
Fνµ = gνν 0 F µ ν = F µ ν gν 0 ν
  
0 −Ex −Ey −Ez 1 0 0 0
 Ex 0 −Bz By   0 −1 0 0 
=   
 Ey Bz 0 −Bx   0 0 −1 0 
Ez −By Bx 0 0 0 0 −1
 
0 Ex Ey Ez
 Ex 0 B z −By 
= 
 Ey −Bz
 (16)
0 Bx 
Ez By −Bx 0
and
0
Fµν = gµµ0 Fνµ0
  
1 0 0 0 0 Ex Ey Ez
 0 −1 0 0  Bz −By 
  Ex 0

= 
 0 0 −1 0   Ey −Bz

0 Bx 
0 0 0 −1 Ez By −Bx 0
 
0 Ex Ey Ez
 −Ex 0 −B z B y 

= 
 −Ey Bz ; (17)
0 −Bx 
−Ez −By Bx 0
in other words, lowering both indices changed the sign in the 0-row and 0-
column of the tensor – the mixed temporal-spatial entries – and left the
temporal-temporal and spatial-spatial entries unchanged.

Scalar Product Scalar products of two four-vectors are then given by

x · p = xµ pµ = xµ pµ = x0 p0 − x · p = x0 p0 − x1 p1 − x2 p2 − x3 p3 (18)

Derivatives Derivatives of a scalar or scalar product with respect to a

vector are given by
∂a · b ∂aµ · bµ
= = bµ , (19)
∂aµ ∂aµ
i.e. derivatives of a scalar quantity w.r.t a covariant vector yield a con-
travariant vector. In particular it is customary to define
∂
∂µ = = (∂/∂t, ∇)
∂xµ

6
 ∂
∂µ = = (∂/∂t, −∇) . (20)
∂xµ
Note that the derivatives have a relative sign in the spatial coordinates!

Relativistic Energy-Momentum Relation In particular, the energy-

momentum relation for a physical particle of (rest) mass m can be written
as

p2 = E 2 − p2 = m2 (21)

Kronecker-δ and Levi-Civita Tensor Two important tensors in three

dimensions are the Kronecker-δ,

ij 1 if i = j
δij = δ = (22)
0 otherwise,
and the anti-symmetric Levi-Civita Tensor, given by

 1 if {ijk} = cyclical permutation of 123
ijk
ijk =  = −1 if {ijk} = anti-cyclical permeation of 123 (23)
0 otherwise.


The latter is generalised to the totally anti-symmetric tensor in four dimen-

sions, µνρσ with

 1 if {µνρσ} = cyclical permutation of 0123
µνρσ
 = −µνρσ = −1 if {µνρσ} = anti-cyclical permutation of 0123
0 otherwise.

(24)

2.4 Lorentz Transformations

General idea Lorentz transformations,

xµ −→ x0µ = Λµν xν (25)

are linear transformations that connect four-vectors with each other. The
Λµν are usually divided into active transformations where the four-vector in
question is moved while the reference system is fixed, and passive transfor-
mations, where the four-vector is fixed, but the reference system is changed.
The difference between active and passive transformations is encoded in a
relative sign of the defining parameters.
In the context of this lecture, the idea of Lorentz transformations is gener-
alised such that they contain both boosts B µν and rotations Rµν , where the
former are defined by three velocities and the latter defined by three angles.
In fact, the rotation are the Galilei transformations, which are superseded
by the Lorentz transformations.

7
Boosts The (active) boosts B µν are defined by the three-velocity1 v; for
example for a boost long the z-axis with velocity v = vz
   
1 0 0 −v cosh η 0 0 − sinh η
 0 1/γ 0 0  0 1 0 0 
B µν (vz ) = γ 

= ,
 0 0 1/γ 0   0 0 1 0 
−v 0 0 1 − sinh η 0 0 cosh η
(26)

where
1
cosh η = γ = √ (27)
1 − v2
is the Lorentz factor and η is the rapidity.
To construct the boost defined by a three-velocity v, B µν (v), it is advan-
tageous to realise that the spatial dimensions can be decomposed into one
component parallel to the boost-vector v, x0k and two perpendicular ones,
~x⊥ . With v = vn and xk = x · n, the transformations read

t0 = γ(t − x · v)
x0k = γ(xk − vt)
~x0⊥ = ~x⊥ , (28)

or, for the spatial components in more compact form

x0 = x + (γ − 1)(n · x)n − γvt (29)

In matrix form this translates to

γ −γvx −γvy −γvz
 
 −γv 1 + (γ − 1) vx2 vx vy
(γ − 1) v2 (γ − 1) vxv2vz 
x v2
B µν (v) =   . (30)
 2

v v v v v
 −γvy (γ − 1) yv2x 1 − (γ − 1) vy2 (γ − 1) yv2 z 
v v 2
−γvz (γ − 1) vzvv2x (γ − 1) zv2y 1 − (γ − 1) vvz2

Rotations Similar to the boosts, the (active) rotations Rµν are defined by
three Euler angles; for example a rotation around the z-axis with angle θ is
mediated by
 
1 0 0 0
 0 cos θ − sin θ 0 
Rµν (θ) = 
 0 sin θ
. (31)
cos θ 0 
0 0 0 1
1
Note that we express the velocity in natural units - in many books the velocity is
given as v = cβ with c the speed of light.

8
Invariance of Norm of Four-Vectors The Lorentz transformations
have been constructed such that the norm of a four vector is invariant under
a boost or rotation. To see how this works look at a four-vector x, boosted
with velocity v. The square of its norm is given by

2 2 2 02
||x0 || = x0 = t0 − ~x02 ⊥ − xk
h i
= γ 2 (1 − v 2 )t2 − (1 − v 2 )x2k − ~x2⊥ = x2 (32)

and therefore invariant.

Time-like vs. Space-like Distances Invariance of the norm of four-

vectors implies that distances of two four-vectors, ∆xµ12 = xµ1 − xµ2 , which of
course are four-vectors themselves, can be decomposed into three cases:
1. Time-like distances: ∆x212 > 0.
A boost can be found in such a way that ∆xµ12 = (∆t, 0), or, in other
words, the spatial positions of x1 and x2 are identical. Events at four-
positions x1 and x2 can be causally connected.
2. Space-like distances: ∆x212 < 0.
A boost can be found in such a way that ∆xµ12 = (0, ∆x), or, in other
words, the temporal positions of x1 and x2 are identical. Events at
four-positions x1 and x2 are not causally connected.
3. Light-like distances: ∆x212 = 0.
A boost can be found such that ∆xµ12 = (0, 0), and events at four-
positions x1 and x2 are on the same light-cone and can be causally
connected through an interaction acting with the speed of light.
The connection of distances with causal structures will become important
at a later stage during the lecture, cf. Section ??.

Inverse Lorentz Transformations Inverse Lorentz transformations are

given by using velocities and rotations with a negative sign with respect to
the originals. This can be used to construct inverse Lorentz Transforma-
tions by expressing the squares of transformed and original four-vectors in
component form:
2 0 0 0 0
x0 = x0µ gµν
0
x0ν = Λµµ0 xµ gµν
0
Λνν 0 xν = xµ gµ0 ν 0 xν (33)
or
Λµµ0 gµν
0
Λνν 0 = gµ0 ν 0 . (34)
Since no system is preferred the metric tensor must be the same in all
0 =g
systems, i.e. gµν µν and therefore

Λµµ0 gµν Λνν 0 = (ΛT )µµ0 gµν Λνν 0 = gµ0 ν 0 . (35)

9
This implies that
0
(Λ−1 )µµ0 = (ΛT )µµ0 ) = Λµµ , (36)

i.e. transposition is the inverse of a Lorentz transformation.

2.5 Lagrange and Hamilton Formalism for Point Particles

Lagrange Function Consider a point particle with kinetic energy T and
a set of generalised coordinates qi (t) and velocities q̇i (t) = dqi /dt that are
suitable to describe its motion in a potential V . The Lagrange function is
given by

L(qi (t), q̇i (t), t) = T − V (37)

and gives rise to the action

Zt1
S(t1 , t0 ) = dtL(qi (t), q̇i (t), t) . (38)
t0

Principle of Least Action Minimising the action by employing virtual

small perturbations of the particle’s path i and ˙i , taken to be zero at
the endpoints t0 and t1 , will yield the Euler-Lagrange Equations of Motion
(E.o.M.). This is also known as Hamilton’s Principle or Principle of Least
Action. Under the usual assumption of an explicitly time-independent La-
grange function and suppressing for a moment the time dependence of the
generalised coordinates and velocities, this yields

Zt1
δS = dt [L(qi + i , q̇i + ˙i ) − L(qi , q̇i )]
t0
Zt1  Zt1 
∂L t1
   
∂L ∂L ∂L d ∂L
= dt i + ˙i = dt i − i + i , (39)
∂qi ∂ q̇i ∂qi dt ∂ q̇i ∂ q̇i t0
t0 t0

where in the last step the term with ˙i has been partially integrated.

Euler-Lagrange Equations of Motion Since the variations i are as-

sumed to vanish at the path’s endpoint the last term vanishes, demanding
that the integral reduces to zero for arbitrary perturbations yields the Euler-
Lagrange E.o.M. for systems without explicit time dependence:

∂L d ∂L
− = 0. (40)
∂qi dt ∂ q̇i

10
Canonical Momentum and Hamilton Function Introducing the canon-
ical momenta
∂L
pi = (41)
∂ q̇i
and expressing the generalised velocities through the canonical momenta pi
allows to construct the Hamilton function as

∂L
H(pi , qi ) = q̇i pi − L(qi , q̇i ) = q̇i − L(qi , q̇i ) = T + V , (42)
∂ q̇i

identical to the energy of the system if it is not explicitly time-dependent.

Hamilton Equations of Motion The Hamilton equations of motions

are given by the two sets of coupled partial differential equations

dpi ∂H
ṗi = =−
dt ∂qi
dqi ∂H
q̇i = =+ . (43)
dt ∂pi

Poisson Brackets Poisson brackets are another possibility to express the

Hamilton E.o.M. they are defined by

∂f ∂g ∂f ∂g
{f, g} = − . (44)
∂qi ∂pi ∂pi ∂qi

They have some interesting properties, for example

ˆ anti-commutativity:

{f, g} = − {g, f } (45)

ˆ bilinearity (a and b constants):

{af + bg, h} = a {f, h} + b {g, h} (46)

ˆ Jacobi identity:

{f, {g, h}} + {g, {h, f }} + {h, {f, g}} = 0 (47)

In particular, the Poisson brackets for the canonical coordinates (positions

and momenta) enjoy the following simple properties:

{qi , qj } = {pi , pj } = 0

11
 {qi , pj } = δij . (48)
Equations of motion can therefore be expressed as

∂H
ṗi = − = {pi , H}
∂qi
∂H
q̇i = + = {qi , H} . (49)
∂pi

The time evolution of any function f (pi , qi , t) can be evaluated using the
chain rule,
df ∂f ∂f ∂f ∂f
= q̇i + ṗi + = {f, H} + . (50)
dt ∂qi ∂pi ∂t ∂t
This translates into explicitly time-independent f are constant of motion, if
their Poisson bracket with the Hamilton function vanishes2 .

2.6 First Quantisation of the Harmonic Oscillator

Hamilton operator In a first step, the Hamilton function is written in
terms of the usual canonical position and momentum, and position, momen-
tum, and Hamilton function are promoted to operators 3 , resulting in
1 2 mω 2 2
Ĥ = p̂ + x̂ . (51)
2m 2
Note that we have used natural units, setting ~ = 1, and in the following we
will also set m = 1 to ease the notation.

Commutator of Position and Momentum Quantisation is achieved by

demanding that the position and momentum operators have a non-vanishing
commutator 4 , namely

[x̂, p̂] ≡ x̂p̂ − p̂x̂ = i . (52)

Creation and Annihilation Operators To cast the Hamilton operator

into a form better suited for analysis, creation and annihilation operators â†
and â are introduced as
1 √
 
i
â = √ ω x̂ + √ p̂
2 ω
2
Note the similarity of the Poisson brackets to the commutator in Quantum Mechanics.
It is, however, important to stress that the functions here are not operators acting on a
Hilbert space, but just functions.
3
Throughout the lecture we will denote operators through aˆsymbol.
4
Remember the Poisson brackets? Of course, as functions, the sequence of their product
is irrelevant, but as operators this is not the case anymore. In this respect the commutator,
although not connected to any derivative, behaves quite similarly to the Poisson brackets.

12
 √
 
† 1 i
â = √ ω x̂ − √ p̂ . (53)
2 ω
Direct calculation shows the following commutation relations:

[â, â] = [↠, ↠] = 0

[â, ↠] = 1 . (54)

Hamilton Operator Expressing the Hamilton operator from Eq, (51)

through the annihilation and creation operators yields

1 2 ω2 2 ω h i
Ĥ = p̂ + x̂ = −(â − ↠)2 + (â + ↠)2
2 2 4  
ω † †
 ω h †
i
†

† 1
= ââ + â â = â, â + 2â â = ω â â + (55)
2 2 2

Number Operator Rewriting the Hamilton operator as

   
1 1
Ĥ = ω ↠â + = ω N̂ + (56)
2 2

with the number operator

N̂ = ↠â . (57)

It has commutator relations

[N̂ , ↠] = ↠and [N̂ , â] = −â (58)

with the creation and annihilation operators.

Eigenstates Denote the energy eigenstates and eigenvalues with

Ĥ |Ei = E |Ei (59)

it is easy to check that â |Ei is also an eigenstate of the Hamilton operator,

 
1
Ĥâ |Ei = ω N̂ + â |Ei
2
   
1 1
= ω N̂ + , â + â N̂ + |Ei = (−ω + E) â |Ei (60)
2 2

with eigenvalue (energy) (E − ω).

Using the fact that eigenvalues of Hermitean operators, such as the position,
momentum, and Hamilton operators, are real numbers and realising that
the Hamilton operator is made up from squares of Hermitean operators with

13
squares of real numbers as eigenvalues, implies that there must be a smallest,
non-negative energy with a corresponding lowest-energy ground state of the
system. Denoting this state as “vacuum”, the only way to guarantee that
there are on lower energy eigenvalues is to demand that the annihilation
operators annihilate this state,

â |0i = 0 , (61)

thereby justifying once more the interpretation of â as annihilation operator.

Conversely, excited states are created by repeated application of the creation
operator,

↠|0i = |1i (62)

and so on. Applying the number operator suggests that the vacuum contains
zero quanta, thereby justifying the notation of |0i and similarly that the first
excited state contains one quantum:

N̂ |0i = 0
h i 
† † † †
N̂ |1i = â ââ |0i = â â, â + â â |0i = 1 · ↠|0i = 1 |1i .
†
(63)

This suggests that the number operator enjoys the eigenvalue equation

N̂ |ni = n |ni (64)

for eigenvectors (eigenkets) |ni.

It is worth commenting here on the states. They populate a Hilbert space -
put in somewhat sloppy terms, this is a vector space with a finite or infinite
number of dimensions, which has a meaningfully defined scalar product.
This scalar product allows to define a measure of distance and the length
of a vector in it. Hilbert spaces are complete as well, which means that we
can safely define limits etc..

Eigenvalues and Eigenstates of the Hamilton Operator Eq. (56)

results in the realisation that the Hamilton and the number operator share
the same eigenvectors/eigenstates, the |ni. Plugging in numbers allows to
directly read off the ground-state energy E0 as
ω
Ĥ |0i = |0i = E0 |0i , (65)
2
and similarly
 
1
Ĥ |ni = ω n + |0i = En |0i (66)
2
with eigenvalues (energies) En = ω(n + 1/2) for the energies of the excited
states.

14
2.7 Problems & Solutions
1. Levi-Civita symbol

(a) Show that for the Levi-Civita symbol in three dimensions,

ijk ilm = δ jl δ km − δ jm δ kl
ijk ijl = 2δ kl
ijk ijk = 6 ,

(b) and that for the Levi-Civita symbol in four dimensions,

0 0 0 0 0 0
µνρσ µν ρ σ = − 2(g ρρ g σσ − g σρ g ρσ )
0 0
µνρσ µνρ σ = − 6g σσ
µνρσ µνρσ0 = − 24 .

Solution

(a) Levi-Civita in three dimensions:

First of all, it is improtant to stress here that we use Einstein’s
convention over repeated indices throughout.
To have a non-vanishing ijk the three indices must be different.
Without any loss of generality this implies that j 6= k and l 6= m
must be fulfilled for the product ijk ilm to be different from 0.
In addition, i has to be different to both j and k and to l and m,
and the sum collapses to only one term (three dimensions, so i, j,
. . . are numbers in {1, 2, 3}), where j and k are identical to l and
m, so either j = l and k = m, or j = m and k = l. These are the
two δ-terms. The first term, with the positive sign, emerges from
ijk and ilm being both either cyclical (ijk = 1) or anticyclical
(ijk = −1), with {ijk} = {ilm}, while the second term, the one
with the negative sign comes from {ijk} = {iml} and one of the
two being cyclical implies that the other is anti-cyclical. This
proves
X
ijk ilm = ijk ilm = δ jl δ km − δ jm δ kl .
i

For the product ijk ijl , similar considerations apply. Demanding

that i, j 6= k and i, j 6= l means that k = l must be fulfilled and
i 6= j means that for a fixed k, there are two combinations possible
for ij, either cyclical or anti-cyclical. Therefore ijk ijl = 2δ kl .
Finally, for ijk ijk , it suffices to count how many permutations
of {ijk} = {123} exist to arrive at ijk ijk = 6.

15
 (b) Levi-Civita in four dimensions:
The identities for the totally antisymmetric tensor in four dimen-
sions follow from using the same logic as before. The realitve
minus sign in front of the expressions is relatively easy to ex-
plain with the signs in the metric tensor, since for the spatial
0 0
components g ρρ = −δ ρρ .

2. Boosts and Rotations

(a) Calculate the effect of two consecutive boosts in z-direction, given

by rapidities η1 and η2 . Do the two operations commute (i.e.
what happens if you reverse the order)?
Hint: Use that

cosh α cosh β ± sinh α sinh β = cosh(α ± β)

cosh α sinh β ± sinh α cosh β = sinh(α ± β)
cos α cos β ± sin α sin β = cosh(α ∓ β)
cos α sin β ± sin α cos β = sinh(α ± β)

(b) Repeat the exercise for two consecutive rotations around the z-
axis with angles θ1 and θ2 .

Solution

(a) Recalling the boost matrices

 
cosh η1,2 0 0 − sinh η1,2
 0 1 0 0 
B12 =  
 0 0 1 0 
− sinh η1,2 0 0 cosh η1,2

and therefore, consecutively applying boost 2 after boost 1

 cosh η2 0 0 − sinh η2   cosh η1 0 0 − sinh η1 
0 1 0 0 0 1 0 0
B2 B1 = 0 0 1 0 0 0 1 0
− sinh η2 0 0 cosh η2 − sinh η1 0 0 cosh η1

cosh(η1 + η2 ) 0 0 − sinh(η1 + η2 )
 
0 1 0 0
= 0 0 1 0 = B1 B2 .
− sinh(η1 + η2 ) 0 0 cosh(η1 + η2 )

(b) Similarly, the rotation matrices

 
1 0 0 0
 0 cos θ1,2 − sin θ1,2 0 
R12 =  0 sin θ1,2

cos θ1,2 0 
0 0 0 1

16
 and therefore, consecutively applying boost 2 after boost 1
 1 0 0 0
 1 0 0 0

0 cos θ2 − sin θ2 0 0 cos θ1 − sin θ1 0
R2 R1 = 0 sin θ2 cos θ2 0 0 sin θ1 cos θ1 0
0 0 0 1 0 0 0 1

1 0 0 0
 
0 cos(θ1 + θ2 ) − sin(θ1 + θ2 ) 0
= 0 sin(θ1 + θ2 ) cos(θ1 + θ2 ) 0 = R1 R2 .
0 0 0 1

3. Inverse Lorentz transformation

(a) use the invariance of distances under Lorentz transformations

2
x0 = [Λµν xν ]2

to show that

[Λµν ]−1 = Λν µ

(b) what is the form of inverse Lorentz boosts and rotations along or
around the z-axis?

Solution

(a)
2
x2 = gµν xµ xν = x0 = gρσ x0ρ x0σ = gρσ Λρ α xα Λσβ xβ
=⇒ gµν = gρσ Λρ µ Λσν
=⇒ δµγ = gµν g νγ = gρσ g νγ Λρ µ Λσν
=⇒ δµγ = gρσ Λσγ Λρ µ
=⇒ δµγ = Iγµ = Λρ γ Λρ µ

Written in matrix notation this implies that

Λ−1 = gΛT g

(b) Using the metric tensor to raise and lower the two indices, yields

Λν µ = gρν Λρ σ g µσ
− sinh u
 1 0 0 0  cosh u 0 0
 1 0 0 0

0 −1 0 0 0 1 0 0 0 −1 0 0
= 0 0 −1 0 0 0 1 0 0 0 −1 0
0 0 0 −1 − sinh u 0 0 cosh u 0 0 0 −1
1 0 0 0 cosh u 0 0 sinh u
  
0 −1 0 0 0 −1 0 0
= 0 0 −1 0 0 0 −1 0
0 0 0 −1 − sinh u 0 0 − cosh u
cosh u 0 0 sinh u
 
0 1 0 0
= 0 0 1 0
sinh u 0 0 cosh u

17
  
cosh(−u) 0 0 sinh(−u)
 0 1 0 0 
=  ,
 0 0 1 0 
sinh(−u) 0 0 cosh(−u)

as expected for boosts along the z-axis. Similarly, for rotations

around the z-axis

Λν µ = gρν Λρ σ g µσ
 1 0 0 0  1 0 0 0
T  1 0 0 0

0 −1 0 0 0 cos θ sin θ 0 0 −1 0 0
= 0 0 −1 0 0 − sin θ cos θ 0 0 0 −1 0
0 0 0 −1 0 0 0 1 0 0 0 −1
1 0 0 0 1 0 0 0
  
0 −1 0 0 0 − cos θ sin θ 0
= 0 0 −1 0 0 − sin θ − cos θ 0
0 0 0 −1 0 0 0 −1
1 0 0 0
 
0 cos θ − sin θ 0
= 0 sin θ cos θ 0 ,
0 0 0 1
 
1 0 0 0
 0 cos(−θ) sin(−θ) 0 
= 
 0 − sin(−θ) cos(−θ) 0  ,


0 0 0 1

again, as expected.

4. ∗ The Generators of the Lorentz group

In this problem we will derive the generators of the Lorentz group and
prove some of their properties.

(a) consider a general, but infinitesimal Lorentz transformation and

write it as
Λµν = δ µν + ω µν .
Show that the infinitesimal parameters ωµν are antisymmetric.
(b) Due to their anti-symmetry, there are only six independent ωµν ,
which shows that the Lorentz-group is a six-parameter group. An
arbitrary Lorentz transformation can be obtained by
 
i µν
U (ω) = exp − M̂µν ω ,
2

where the M̂µν are the generators of the group. We obtain them
by considering infinitesimal transformations and comparing co-
efficients. Chooisng generators from three infinitesimals boosts
along the x, y, and z-axis and the three infinitesimals boosts

18
 around the x, y, and z-axis we arrive at:
   
0 0 0 0 0 1 0 0
 0 0 0 0   Bx = M̂01 = −i  1 0 0 0 
 
Rx = M̂23 = i 
 0 0 0 −1   0 0 0 0 
0 0 1 0 0 0 0 0
   
0 0 0 0 0 0 1 0
 0 0 0 −1   By = M̂02 = −i  0 0 0 0 
 
Ry = M̂13 = i 
 0 0 0 0   1 0 0 0 
1 0 0 0 0 0 0 0
   
0 0 0 0 0 0 0 1
 0 0 −1 0   Bz = M̂03 = −i  0 0 0 0 
 
Rz = M̂12 = i 
 0 1 0 0   0 0 0 0 
0 0 0 0 1 0 0 0

Convince yourself that their commutation relation

h i  
M̂µν , M̂ρσ = i gµσ M̂νρ + gνρ M̂µσ − gµρ M̂νσ − gνσ M̂µρ

holds true. Define

1
M̂i = ijk M̂jk and N̂i = M̂i0
2
and their linear combinations
1 
X̂i± = M̂i ± iN̂i
2
and use the general identity to prove that their commutators are
given by
h i h i
X̂i± , X̂j± = iijk X̂k± and X̂i± , X̂j∓ = 0 .

Solution

(a) Using Eq. (35) we can write

 
gµ0 ν 0 = Λµµ0 gµν Λνν 0 = δ µµ0 + ω µµ0 gµν (δ νν 0 + ω νν 0 )
= gµ0 ν 0 + ων 0 µ0 + ωµ0 ν 0 + O(ω 2 )

and therefore we must demand

ων 0 µ0 + ωµ0 ν 0 = 0 −→ ων 0 µ0 = −ωµ0 ν 0 ,

i.e., anti-symmetric ω.

19
(b) Let us consider a number of cases, namely the commutator of two
two boosts, of two rotations, and of a boost and a rotation.
h i
M̂01 , M̂02 = ig00 M̂12 = iM̂12
h i
M̂12 , M̂13 = −ig11 M̂23 = iM̂23
h i
M̂01 , M̂12 = ig11 M̂02 = −iM̂02

We then have, for the vanishing commutator,

h i 1 1 1

± ∓
X̂i , X̂j = ikl M̂kl ± iM̂i0 , jmn M̂mn ∓ iM̂j0
4 2 2
1 n ikl jmn h i
= M̂kl , M̂mn
4 4
ijmn h i i h i h i
ikl
± M̂i0 , M̂mn ∓ M̂kl , M̂j0 + M̂i0 , M̂j0
2 2
i n ikl jmn  
= gkn M̂lm + glm M̂kn − gkm M̂ln − gln M̂km
4 4
ijmn  
± gin M̂0m + g0m M̂in − gim M̂0n − g0n M̂im
2
iikl  
∓ gk0 M̂lj + glj M̂k0 − gkj M̂l0 − gl0 M̂kj
2 o
− g00 M̂ij

i 1h
= − (δ lj δ im − δ lm δ ij )M̂lm + (δ in δ jk − δ ij δ kn )M̂kn
4 4
i
− (δ ln δ ij − δ lj δ in )M̂ln − (δ ij δ km − δ im δ kj )M̂km

ih i ih i
∓ ijm M̂0m − jin M̂0n ∓ ikj M̂k0 − ijl M̂l0 − M̂ij
2 2

i 1h i
= − M̂ji + M̂ji + M̂ji + M̂ji − M̂ij
4 4
ih i
∓ ijm M̂0m − jim M̂0m − imj M̂m0 + ijm M̂m0
2
 i
i ih
= −M̂ji − M̂ij ∓ ijm M̂0m (1 + 1 − 1 − 1) = 0 ,
4 2
where, in the last step, we have used the anti-symmetry of the
Levi-Civita Tensor and that M̂0m = −M̂m0 . In the treatment of
the terms proportional to only one of the Levi-Civita Tensors we
also realised that terms of the form ijk gk0 vanish – after all ijk
is only defined for permutations of the spatial dimensions, i.e.
{i, j, k} ∈ perm({1, 2, 3}). We also took into account the sign
for space-like components, i.e. gik = −δik .

20
The non-vanising commutators in contrast are given by
h i 1 1 1

± ±
X̂i , X̂j = ikl M̂kl ± iM̂i0 , jmn M̂mn ± iM̂j0
4 2 2
1 ikl jmn
n h i
= M̂kl, M̂mn
4 4
ijmn h i i h i h i
ikl
± M̂i0 , M̂mn ± M̂kl , M̂j0 − M̂i0 , M̂j0
2 2
i n ikl jmn  
= gkn M̂lm + glm M̂kn − gkm M̂ln − gln M̂km
4 4
ijmn  
± gin M̂0m + g0m M̂in − gim M̂0n − g0n M̂im
2
iikl  
± gk0 M̂lj + glj M̂k0 − gkj M̂l0 − gl0 M̂kj
2 o
+ g00 M̂ij

i 1h
= − (δ lj δ im − δ lm δ ij )M̂lm + (δ in δ jk − δ ij δ kn )M̂kn
4 4
i
− (δ ln δ ij − δ lj δ in )M̂ln − (δ ij δ km − δ im δ kj )M̂km

ih i ih i
∓ ijm M̂0m − jin M̂0n ± ikj M̂k0 − ijl M̂l0 + M̂ij
2 2

i 1h i
= − M̂ji + M̂ji + M̂ji + M̂ji + M̂ij
4 4
ih i
∓ ijm M̂0m − jim M̂0m − imj M̂m0 + ijm M̂m0
2
 i
i ih
= −M̂ji + M̂ij ∓ ijm M̂0m (1 + 1 + 1 + 1)
4 2
 
i
= M̂ij ± iijk M̂k0
2
±,
and direct comparison with the definition of X̂m

± iijm  mrs 
iijm X̂m = M̂rs ± iM̂m0
2 2  
i h ijm rsm i i δir δjs − δis δjr
= M̂rs ± iM̂m0 = M̂rs ± iM̂m0
2 2 2 2
 
i 1  ih i
= M̂ij − M̂ji ± iM̂m0 = M̂ij ± iM̂m0 =
2 2 2
yields the desired result.
This proves that the six generators of the Lorentz group fac-
torise into two groups of three generators, where each group has
a commutator structure that is identical to the one enjoyed by the

21
 generators of the angular momentum group, and where the gener-
ators of the two groups do commute. In other words, the Lorentz
group SO(3, 1) decomposes as SO(3, 1) = SU (2)⊗SU (2), hinting
at a deep connenction between the Lorentz group and spin.

5. ∗ Poincare transformations
The Poincare transformation U (Λ, a) is defined by the combination of
a Lorentz transformation, Λµν , and a shift in space-time, aµ , as

xµ → x0µ = Λµν xν + aµ .

(a) Determine the product, the unit and inverse of the resulting
group.
(b) Verify that

U −1 (Λ, 0)U (1, )U (Λ, 0) = U (1, Λ−1 )

and show that this implies that

U −1 (Λ, 0)P̂µ U (Λ, 0) = (Λ−1 )νµ P̂ν .

Use this to determine the commutator [M̂µν , P̂ρ ] of the generators

of the Lorentz group and the momentum operator.
(c) Show that

U −1 (Λ, 0)U (Λ̃, 0)U (Λ, 0) = U (Λ−1 Λ̃Λ, 0)

and use this to prove the commutator relation of the generators

M̂µν from the previous problem, i.e.
h i  
M̂µν , M̂ρσ = i gµρ M̂νσ − gµσ M̂νρ − gνρ M̂µσ + gνσ M̂µρ .

Solution

(a) Let us start by the product of two transformations, U (Λ, a) ⊗

U (Λ̃, ã) i.e.
 
xµ → x0µ = Λµν Λ̃νρ xρ + ãν + aµ ,

and we can read off that

U (Λ, a) ⊗ U (Λ̃, ã) = U (ΛΛ̃, Λã + a)

the product is given by a product of the Lorentz-transformation

with a shift given by the sum of the Lorentz-transformed first
shift and the second shift.

22
 The unit element is obviously given by no Lorentz-transformation,
the unit matrix plus a zero shift, U1 = U (1, 0) and the inverse is
given by
U −1 (Λ, a) = U (Λ−1 , −Λ−1 a) .
To check this explicitly, show that

xµ = U −1 (Λ, a)U (Λ, a)xµ

= (Λ−1 )µν Λνρ xρ + ãν − (Λ−1 )µν aν

= gρµ xρ + (Λ−1 )µν aν − (Λ−1 )µν aν = xµ ,

as expected.
(b)

x0µ = (Λ−1 )µν g νρ Λρσ xσ + ν = xµ + (Λ−1 )µν ν

 

as demanded. Using the fact that the momentum operator is the

generator of infinitesimal translations in space-time, the µ , we
see immediately that the relation above implies that

U −1 (Λ, 0)P̂µ U (Λ, 0) = (Λ−1 )νµ P̂ν .

indeed holds true.

To calculate the commutator of the momentum operator and
the generators of the Lorentz group it is important to remember
that Lorentz transformations parametrised by ωµν are generated
through
 
i µν i
U (ω, 0) = exp − M̂µν ω = 1 − M̂µν ω µν + O(ω 2 ) .
2 2

see the previous problem. Using the transformation law for the
momentum above, and specify it for an infinitesimal Lorentz
transformation we therefore have
   
i µν i µν
1 + M̂µν ω P̂σ 1 − M̂µν ω = (δ µσ − ω µσ ) P̂µ
2 2
iω µν   ω µν  
M̂µν P̂σ − P̂σ M̂µν = − gνσ P̂µ − gνσ P̂ν
2 h i 2
M̂µν , P̂σ = gνσ P̂µ − gνσ P̂ν ,

where in going from the first to the second line we ignored terms
quadratic in ω and we explicitly anti-symmetrised the right-hand
side when lifting the Lorentz index of the ω µσ to reflect its prop-
erty.

23
 (c) Start with

x0µ = U −1 (Λ, 0)U (Λ̃, 0)U (Λ, 0)xµ = (Λ−1 )µν Λ̃νρ Λρσ xσ

as expected and specify it for an infinitesimal Lorentz transfor-

mation for Λ̃ = 1 + ω̃. Then

(Λ−1 Λ̃Λ)µν = δ µν + (Λ−1 )µρ Λσν ω̃ ρσ

and, in analogy to the treatment of the infinitesimal translations

we can use this to deduce the transformation law for the genera-
tors of the Lorentz transformations, namely

U −1 (Λ, 0)M̂ρσ U (Λ, 0) = (Λ−1 )µρ Λνσ M̂µν .

Specfiying this for infinitesimal boosts Λ = 1 + ω we find

   
i µν i µν
1 + M̂µν ω M̂ρσ 1 − M̂µν ω = (δ µρ − ω µρ )(δ νσ + ω νσ )M̂µν
2 2

and ignoring terms of order ω 2 and anti-symmetrising arguments

as before we arrive at
iω µν  
M̂µν M̂ρσ − M̂ρσ M̂µν = δ µρ ω νσ − δ νσ ω µρ M̂µν


2
ω µν  
= − gµρ M̂νσ − gµσ M̂νρ − gνρ M̂µσ + gνσ M̂µρ
2
and therefore we find that the commutator indeed is
h i  
M̂µν , M̂ρσ = i gµρ M̂νσ − gµσ M̂νρ − gνρ M̂µσ + gνσ M̂µρ .

In addition to the known commutator

h i
P̂µ , P̂ν = 0

this fixes the algebra of the Poincare group.

6. Lagrange and Hamilton Formalism: Example Systems

For all of the three systems

(i) Free particle in three dimensions;

(ii) Mathematical pendulum in one dimension, in the small-angle ap-
proximation;
(iii) Particle in two-dimensions in a central potential

analyse the E.o.M. through the following steps:

24
(a) write down the Lagrange function;
(b) derive and solve the Euler-Lagrange E.o.M.;
(c) construct the canonical momenta;
(d) find the Hamilton function;
(e) derive the Hamilton E.o.M.;
(f) try to directly infer constants of motion where possible.

Solution

(a)
m 2
(i) : L = ẋ
2
ml2 2
(ii) : L = θ̇ − mglθ2
2
m 2 2 2

(iii) : L = ṙ + r θ̇ − V (r)
2

(b) For each coordinate q we have

d ∂L ∂L
0= −
dt ∂ q̇ ∂q
and therefore

(i) : 0 = mẍ
(ii) : 0 = ml2 θ̈ + mglθ
(iii) : 0 = m(r2 θ̈ + 2rṙθ̇)
∂V
0 = mr̈ − mrθ̇2 +
∂r

(c) For each coordinate q we have

∂L
p=
∂ q̇
and therefore

(i) : px = mẋ
(ii) : pθ = ml2 θ̇
(iii) : pθ = mr2 θ̇
pr = mṙ

25
 (d) Summing over all coordinates and momenta i
X
H= q̇i pi − L
i

and therefore, replacing generalised velocities with the momenta,

1 2
(i) : H = mẋẋ − L = p
2m
1 2
(ii) : H = ml2 θ̇θ̇ − L = p + mglθ2
2ml2 θ
1 2 1
(iii) : H = mr2 θ̇θ̇ + mṙṙ − L = pr + p2 + V (r)
2m 2mr2 θ

(e) Using

∂H ∂H
ṗi = − and q̇i = +
∂qi ∂pi
we have
p
(i) : ṗ = 0 and ẋ =
m
pθ
(ii) : p˙θ = −mglθ and θ̇ =
ml2
p2θ ∂V pr
(iii) : p˙r = − 3
− and ṙ =
mr ∂r 2m
pθ
p˙θ = 0 and θ̇ = .
mr2

(f) In (i), the momentum is conserved, component-by-component,

i.e. ṗ = 0, and in (iii) the angular momentum pθ is conserved.

7. Conserved energy from Hamilton function

Show that the energy is conserved of a system described by a classical
Hamilton function without explicit time dependence.

Solution

Hamilton equations of motion:

∂H ∂H
q̇i = , ṗi = − , ∂t H = −∂t L = 0
∂pi ∂qi
if L is not explicitly dependent on t.
 
∂L ∂L dL ∂L ∂L dL
0 = −∂t L = q̇ + q̈ − = ∂t q̇ + q̈ −
∂q ∂ q̇ dt ∂ q̇ ∂ q̇ dt

26
    
∂L ∂L ∂L dL d ∂L
= ∂t q̇ − q̈ + q̈ − = q̇ − L
∂ q̇ ∂ q̇ ∂ q̇ dt dt ∂ q̇
 
d ∂L
= q̇ − T + V ,
dt ∂ q̇
where Lagrange E.o.M. and L = T − V have been used with T and
V are the kinetic and potential energy. Using Euler’s theorem for
homogeneous functions,
∂L
q̇ = 2T
∂ q̇
if the kinetic energy is a quadratic function of generalised velocities q
(which is usually the case), and if the potential does - as usual - only
depend on generalised coordinates.
This shows that
d(T + V ) dE
= =0
dt dt
Alternatively:
dH ∂H dp ∂H dq ∂H
= + +
dt ∂p dt ∂q dt ∂t
∂H
= q̇ ṗ − ṗq̇ + (using Hamilton’s equations)
∂t
∂H
=
∂t
= 0 if H not explicitly dependent on t.

8. Quantum Mechanical Harmonic Oscillator

(a) Starting with the Lagrange function
m ẋ2 mω 2 x2
L= − ,
2 2
calculate the conjugate momenta and show that the Hamilton
function reads
1 2 mω 2 2
H= p + x .
2m 2
(b) Promote the Hamilton function to the Hamilton operator of Eq. (51)
and use the commutation relation of the position and momentum
operator to show that the commutator relations
[â, â] = [↠, ↠] = 0
[â, ↠] = 1 .
of Eq. (54) hold true, where the creation and annihilation oper-
ators are defined through Eq. (53).

27
(c) Re-express the Hamilton operator first through the creation and
annihilation operators and then through the number operator.
Evaluate the commutators [N̂ , â] and [N̂ , ↠].
(d) Use the fact that â|0i = 0 to calculate the wave function of the
ground state in position space, i.e.
ψ0 (x) = hx|0i
To do so, you have to express the annihilation operator in position
space and suitably transform hx|â|0i.
(e) Speculate about the spectrum of the fermionic quantum har-
monic oscillator, given by the same Hamiltonian, but where the
creation and annihilation operators ↠and â anti-commute:
{â, ↠} = â↠+ ↠â = 1
{â, â} = {↠, ↠} = 0

Solution

(a) Momentum
∂L
p = = mẋ
∂ ẋ
and Hamilton function
1 2 mω 2 2
H = pẋ − L = p + x .
2m 2
(b) Hamilton operator
1 2 mω 2 2
Ĥ = p̂ + x̂ .
2m 2
With
r  
mω i
â = x̂ + p̂
2 mω
r  
† mω i
â = x̂ − p̂
2 mω
we find
 
mω i
[â, â] = + [x̂, p̂] + [p̂, x̂] =0
2 mω
 
mω i
[â, â] = − [x̂, p̂] + [p̂, x̂] =0
2 mω
   
h
†
i mω i i
â, â = − [x̂, p̂] − [p̂, x̂] = − 2 [x̂, p̂]
2 mω 2
= 1.

28
(c) Express p̂ and x̂ through â and ↠as
r  
1 †
x̂ = + â + â
2mω
r  
mω †
p̂ = − i â − â
2
and therefore
"  2  2 #
ω † †
Ĥ = − â − â + â + â
4
     
ω † † † 1 1
= ââ + â â = ω â â + = ω N̂ + .
2 2 2
Commutators:
[N̂ , â] = ↠ââ − â↠â = ↠ââ − ↠ââ − [â, ↠]â = − â
[N̂ , ↠] = ↠â↠− ↠↠â = ↠ââ − ↠â↠+ [â, ↠]↠= + ↠.

(d) Transform the annihilation relation into x-space as

r    
mω  i 
0 = hx| â |0i = x  x̂ +
 p̂  0
2 mω
Z r       
mω i
dx0 p̂  x0 x0  0
  
= x  x̂ +
2 mω
Realising that the first bracket projects out the eigenvalues of the
position operator,
hx| x̂ x0 = δ(x − x0 )x ,
 

and transforms into a derivative w.r.t. x for the momentum op-

erator,
−i∂
hx| p̂ x0 = δ(x − x0 )
 
,
∂x
and that the second bracket is nothing but the ground-state wave
function in position space, we find that
r r !
mω 1 ∂
0= x+ ψ0 (x) ,
2 mω ∂x

which has the solution

mωx2
 
hx|0i = ψ0 ∼ exp − .
2
Excited states can be obtained in a similar way, by realising that
ψ1 (x) = hx|1i = hx|↠|0i . . . .

29
(e) Writing, in full analogy, the Hamiltonian as
   
1 † 1
Ĥ = ω N̂ − = ω â â −
2 2

will lead to the same ground state,

â|0i = 0 −→ E0 = −ω/2

Repeatedly applying the creation operator ↠will result in

|1i = ↠|0i

but because of
1 n † †o
↠↠= â , â = 0
2
the application of creation operator on the first excited state will
annihilate it:

↠|1i = ↠↠|0i = 0

and therefore the fermionic harmonic oscillator has only

two states with energies E0,1 = ∓ω/2.

30
3 Classical Fields
In this section we re-derive the Lagrange functions for classical fields. For a
more exhaustive explanation of how to make the transition from a discrete to
a continuous system, chapter 13 of Goldstein [9] may be helpful. There, you
will also find a good derivation of the Euler-Lagrange Equations of Motion
for fields. If you are mainly interested in using the formalism in the context
of the course, you may want to consult Sec. 2.2 of Peskin & Schröder [1].

3.1 One-Dimensional Lattice

Setup Consider a system of massive particles with identical mass m, ar-
ranged in a one-dimensional lattice with positions ξi , and with their motion
confined along the lattice direction. The kinetic energy of the system is
given by
m X ˙2
T = ξi (t) (67)
2
i

Coupling the particles with springs with constants k yields the potential
energy
kX
V = (ξi+1 (t) − ξi (t))2 , (68)
2
i

and the Lagrange function

1 X h ˙2 i
L= mξi (t) − k (ξi+1 (t) − ξi (t))2
2
i
 !2 
2 ˙  2
a X
m ξi (t) ξi+1 (t) − ξi (t) 
= −k , (69)
2 a a
i

where a is the equilibrium separation between the particles.

Euler-Lagrange E.o.M. To arrive at the E.o.M. for a specific ηi we have

to take into account that for the same index i the displacement ηi shows up
twice in the sum over the differences, and therefore
¨  
2 ξ(t) 2 ξi+1 (t) − ξi (t) ξi (t) − ξi−1 (t)
0 = ma 2 − ka −
a a2 a2
  
¨ ξi+1 (t) − ξi (t) ξi (t) − ξi−1 (t)
= a µξ(t) − Y − (70)
a2 a2
where the second line was obtained after factoring out one power of a, and
by identifying µ = m/a as the mass density per unit length, and Y = ka as
Young’s modulus of the continuous rod.

31
Continuum Limit Going from discrete lattice distances to a continuum
can be understood as replacing the index i with a position x, ξi (t) → ξ(x, t),
and by taking the limit a → 0 for the lattice spacing. The ξ differences
become
ξi+1 (t) − ξi (t) ξ(x + a, t) − ξ(x, t) ∂ξ(x, t)
lim = lim = (71)
a→0 a a→0 a ∂x
Summation over i translates into an integral over x,
X
a → dx (72)
i

and the discrete Lagrange function of Eq. (69) turns into the Lagrangian
Z   
1 ∂ξ
L= dx µξ˙2 − Y (73)
2 ∂x

for the continuous rod; from now on we suppress the arguments of the ξ.
Going back to the equation of motion, Eq. (69), and taking a closer look at
the second term in the limit of vanishing spacing a
   
ξi+1 − ξi ξi − ξi−1 ξ(x + a) − ξ(x) ξ(x) − ξ(x − a)
− −→ −
a2 a2 a2 a2
a→0 ∂ξ(x + a)/∂x − ∂ξ(x)/∂x ∂ 2 ξ(x)
−→ lim = ,
a→0 a ∂x2
(74)

it is clear that this is a second derivative, and the E.o.M. for the continuous
elastic rod therefore is given by

∂2ξ ∂2ξ
µ − Y = 0, (75)
∂t2 ∂x2
with longitudinal waves as solution.

Lagrange Density The previous considerations suggest that it is sensible

to introduce a Lagrange density
 2  2
µ ∂ξ Y ∂ξ
L(ξ, ∂ξ/∂t, ∂ξ/∂t, x, t) = − (76)
2 ∂t 2 ∂x

which becomes the Lagrange function through integration over the (one-
dimensional) space,
Z
L = dxL(ξ, ∂ξ/∂t, ∂ξ/∂x, x, t) (77)

32
and the action S, as usual, by integrating the Lagrange function over time,
Zt1 Zx1
S(t0 , t1 ) = dt dx L . (78)
t0 x0

In the rest of the lecture course we will assume that Lagrange densities
depend on fields and their derivatives only and do not explicitly depend on
position or time, i.e.
 
∂ξ ∂ξ
L = L ξ, , . (79)
∂t ∂x

Euler-Lagrange E.o.M. To arrive at the Euler-Lagrange equations of

motion we will proceed as before, by minimising the action with respect to
virtual variations of the fields ξ and their derivatives,

ξ(x, t) → ξ 0 (x, t) = ξ(x, t) + αζ(x, t)

∂ξ(x, t) ∂ξ 0 (x, t) ∂ξ(x, t) ∂ζ(x, t)
→ = + α (80)
∂t ∂t ∂t ∂t
∂ξ(x, t) ∂ξ 0 (x, t) ∂ξ(x, t) ∂ζ(x, t)
→ = + α .
∂x ∂x ∂x ∂x
Here α is a parameter that steers the size of the variation, while ζ(x, t) rep-
resents an arbitrary function which vanishes at the endpoints of the integral,
i.e. at times t0 and t1 . Minimising the action with respect to the variations
is achieved by
   
Zt1 Zx1

∂ξ ∂ξ
dS ∂L ∂ξ ∂L ∂ ∂t ∂L ∂ ∂x
0≡ = dt dx  +   +    (81)
dα ∂ξ ∂α ∂ ∂ξ ∂α ∂ ∂ξ ∂α
t0 x0 ∂t ∂x

Because the variation vanishes at the endpoints, integration by parts allows

us to replace the last two terms by
Zt1   Zt1 Zt1
∂L ∂ ∂ξ ∂L ∂ ∂ξ ∂ ∂L ∂ξ
dt   = dt   = − dt  
∂ ∂ξ ∂α ∂t ∂ ∂ξ ∂t ∂α ∂t ∂ ∂ξ ∂α
t0 ∂t t0 ∂t t ∂t 0

Zx1   Zx1 Zx1

∂L ∂ ∂ξ ∂L ∂ ∂ξ ∂ ∂L ∂ξ
dx   = dx   = − dx   (82)
∂ξ ∂α ∂x ∂ξ ∂x ∂α ∂x ∂ ∂ξ ∂α
x0
∂ ∂x x0
∂ ∂x x ∂x 0

Putting it all together results in

Zt1 Zx1
 
∂ξ  ∂L ∂ ∂L ∂ ∂L 
0 ≡ dt dx −  −   (83)
∂α ∂ξ ∂t ∂ ∂ξ ∂x ∂ ∂ξ
t 0 x 0 ∂t ∂x

33
and, finally, the equations of motion

∂ ∂L ∂ ∂L ∂L
 +  − = 0. (84)
∂t ∂ ∂ξ ∂x ∂ ∂ξ ∂ξ
∂t ∂x

This of course is relatively straightforward to extend to the case of two or

three spatial dimensions, by essentially replacing the derivative w.r.t. x with
a gradient, ∂/∂x −→ ∇, and by replacing the
R one-dimensional integral over
3
R
x with an integral over full space, dx −→ d x.

Lorentz-Invariant Formulation The treatment of the fields in the La-

grange formalism until now has not been Lorentz-invariant, and we are going
to rectify this now. The first thing to note is that in a Lorentz-invariant
framework, the integration should not distinguish between time and space,
suggesting to move
µ
Zt1 Zx1 Zx1
dt d3 x −→ d4 x . (85)
t0 x0 xµ
0

A simple calculation will show that the integral over the space-time volume
is boost and hence Lorentz-invariant. In a similar way, the two derivative
terms in the Lagrange density in Eq. (79) will be amalgamated such that
the Lorentz-invariant Lagrange density is given by

L = L (ξ, ∂µ ξ) . (86)

There is one big caveat, however. This Lagrange density must be a Lorentz-
scalar; pictorially speaking, all indices must be contracted. This implies that
terms of the type ∂µ ξ must come at least in squares, like, e.g. (∂µ ξ)(∂ µ ξ)
such that the two Lorentz-indices are contracted off.
To obtain Euler-Lagrange equations of motion from the action
µ
Zx1
S= d4 x L (ξ, ∂µ ξ) , (87)
xµ
0

steps similar to the one before will be necessary. In particular, we will now
demand that the virtual variations of the field vanish on the surface of the
d4 x-integration, leading to
µ
Zx1  
dS 4 ∂L ∂ξ ∂L ∂(∂µ ξ)
0≡ = d x +
dα ∂ξ ∂α ∂(∂µ ξ) ∂α
xµ
0

34
 µ
Zx1     
4 ∂ξ ∂L ∂L ∂L ∂ξ
= d x − ∂µ + ∂µ ∂µ . (88)
∂α ∂ξ ∂(∂µ ξ) ∂(∂µ ξ) ∂α
xµ
0

The last term is a four-dimensional volume integral over a four-dimensional

divergence, which vanishes with the vanishing virtual variations of the fields,
and we are left with the Euler-Lagrange E.o.M. for relativistic fields

∂L ∂L
∂µ − = 0. (89)
∂(∂µ ξ) ∂ξ

3.2 Scalar Fields: Real Scalars

Know Thy Equation of Motion! The desired equations of motion are
a good starting point to construct Lagrange densities for realistic and physi-
cally relevant examples of relativistic field theories. We will first consider the
probably simplest case of a free real scalar field φ(x), i.e. a field that does
not interact with other fields or with an external potential 5 . To see how
this works, let us start with the well-known Schrödinger equation, where
the starting point is the kinetic energy, given by E = p2 /2m. Substituting
derivatives for energy and momentum, E → i∂t and p → −i∇ or pj → −i∂j ,
we arrive at the E.o.M.,
∂ 1 ∂2
i φ(x) + φ(x) = 0 . (90)
∂t 2 ∂x2j
In the same vein, we start with the relativistic energy-momentum relation,
E 2 = p2 + m2 and find the Klein-Gordon Equation

∂2
 
− ∇2 + m2 φ(x) = ∂µ ∂ µ + m2 φ(x) = 0 .


(91)
∂t2

Solutions to the Klein-Gordon Equation The solution to the Klein-

Gordon Equation, Eq. (91) for a fixed momentum is given by
φ(x) = a(k) e−ik·x + a∗ (k) eik·x , (92)
where a(k) and a∗ (k) are the (complex) amplitudes for the plane-wave so-
lution for a fixed wave four-vector k, which satisfies the implicit “on-shell”
condition k 2 = k02 − k 2 = m2 . Of course, we could also sum over many such
waves and we arrive at
d3 k
Z h i
−ik·x ∗ ik·x
φ(x) = a(k) e + a (k) e . (93)
(2π)3 (2k0 )
5
Note that application of external potentials would introduce an explicit dependence
of the Lagrange density on the space-time coordinates x.

35
A few comments are in order here:

1. In Eq. (142) we have directly used the continuum limit. This neces-
sitates the integration over all momenta instead of a summation over
a discrete set of eigenvalues for the momentum. The latter would be
the case for example when second quantising on a lattice with lattice
spacing a, where the eigenvalues for the momentum are discrete and
behave like kn = n/a.

2. The measure of integration, that sums over the different wave vector,
should better be Lorentz-invariant. It is not trivial to see immedi-
ately that d3 k/(2k0 ) fulfils this criterion. To realise that this is indeed
the case, let us start with a manifestly Lorentz-invariant integration
measure,

d4 k d3 k 
Z Z
2 2
dk0 δ(k02 − k 2 − m2 )Θ(k0 )

4
(2π)δ(k − m )Θ(k 0 ) = 3
(2π) (2π)
d3 k
Z  
X 1
= Θ(k0 )
(2π)3 √ 2k0
k0 =± k2 +m2

d3 k
Z
= , (94)
(2π)3 (2k0 )

where the d4 k obviously is a boost and rotation-invariant quantity, the

factor δ(k 2 − m2 ) encodes the (Lorentz-invariant) relativistic energy-
momentum relation necessary to ensure that the quanta behave in a
physically sensible way, and Θ(k0 ) projects on positive-energy solu-
tions. In performing the k0 -integration we have used a property of the
δ-function, namely
Z X δ(x − xi )
dx δ(f (x)) = , (95)
f 0 (xi )
xi : f (xi )=0

which replaces the integral over the δ-function of a function f (x) with
an integral over a sum of its zeroes xi (given by f (xi ) = 0), normalised
by the first derivative of the function at the zero.

Klein-Gordon Lagrange Density It is simple to show that this equation

of motion, cf. Eq. (91), can be obtained from the Lagrange density

1 m2 2
L(∂µ φ, φ) = (∂µ φ)(∂ µ φ) − φ . (96)
2 2

Note that, wherever the dependence is self-evident, we will ignore the ar-
guments of the fields from now on. To see this, let us plug this Lagrange

36
density into Eq. (89), with the obvious replacement ξ → φ.
∂L ∂L
0 = ∂µ −
∂(∂µ φ) ∂φ
2
∂[ 12 (∂µ φ)(∂ µ φ)] ∂[ m2 φ2 ]
= ∂µ − , (97)
∂(∂µ φ) ∂φ
where we have replaced the Lagrange density in the first line with the rel-
evant parts of Eq. (96) in the second one. The first expression looks a bit
tricky and, naively, it seems as if derivation w.r.t. ∂µ φ would only deliver
1 µ
2 ∂ φ – this however is wrong, and it is easy to see why. Rewriting this part
component by component we would arrive at terms like

1 ∂ ∂ φ̇2 1 ∂ φ̇ ∂2φ
=2· = 2 (98)
2 ∂t ∂ φ̇ 2 ∂t ∂t

and similar for the spatial components. Another way to see this is to rewrite
the Lorentz-scalar of the derivatives with other indices – replacing the µ’s
with ν’s in the Lagrangian (it doesn’t matter, they get contracted anyway,
so I can sum over µ’s, ν’s or any other symbol I chose as Lorentz index)
   νρ 
∂ 1 ν ∂ g
(∂ φ)(∂ν φ) = (∂ν φ)(∂ρ φ)
∂(∂µ φ) 2 ∂(∂µ φ) 2
g νρ
 
∂(∂ρ φ) ∂(∂ν φ)
= (∂ν φ) + (∂ρ φ)
2 ∂(∂µ φ) ∂(∂µ φ)
νρ
g 
(∂ν φ)δρµ + (∂ρ φ)δνµ = ∂ µ φ

= (99)
2
Taking into account of this insight, we ultimately arrive at

0 = ∂µ ∂ µ φ + m2 φ , (100)

as requested.

3.3 Scalar Fields: Complex Scalars

Two Real Scalars = One Complex Scalar Consider now two such free
real scalar fields, φ1 and φ2 . The Lagrange density reads
2
X 1 m2i 2
L = (∂µ φi )(∂ µ φi ) − φ (101)
2 2 i
i=1

If both masses are equal, m1 = m2 , the two real fields can be re-arranged
into one complex one,
φ1 + iφ2 φ1 − iφ2
φ= √ and φ∗ = √ , (102)
2 2

37
or
φ + φ∗ −i(φ − φ∗ )
φ1 = √ and φ2 = √ . (103)
2 2
The Lagrange density for the free complex scalar field then becomes

L = (∂µ φ∗ )(∂ µ φ) − m2 φ∗ φ . (104)

It is important to stress here that while the fields φ and φ∗ are connected
through complex conjugation, they still encode two independent degrees
of freedom and therefore must be treated as independent quantities when
analysing the structure of the Lagrange density, or deriving E.o.M..

Equations of Motion The E.o.M. are obtained in the now familiar fash-
ion as
∂L ∂L
0 = ∂µ − = ∂µ ∂ µ φ∗ + m2 φ∗
∂(∂µ φ) ∂φ
∂L ∂L
0 = ∂µ − = ∂µ ∂ µ φ + m2 φ . (105)
∂(∂µ φ∗ ) ∂φ∗

Note that, as we have two independent degrees of freedom (the two fields),
we have two E.o.M., obtained by differentiating the Lagrangian with respect
to each of the two fields.

A Simple Symmetry Inspection of the Lagrangian of Eq. (104) reveals

an interesting invariance under rotations. Transforming the fields as

φ → φ0 = exp(iθ)φ , φ∗ → φ0∗ = exp(−iθ)φ∗ (106)

with a constant angle θ, we have

L → L0 = (∂µ φ0∗ )(∂ µ φ0 ) − m2 φ∗0 φ0

= [∂µ (e−iθ φ∗ )][∂ µ (eiθ φ)] − m2 (e−iθ φ∗ )(eiθ φ) = L . (107)

Clearly, the Lagrangian and therefore the action are invariant under this set
of transformations.

Conserved Current However, let us for a moment look at this from a

different perspective, and demand invariance, by setting

0 ≡ δS
Z  
4 ∂L ∂L ∂L ∗ ∂L ∗
= d x δ(∂µ φ) + δφ + δ(∂µ φ ) + δφ (108)
∂(∂µ φ) ∂φ ∂(∂µ φ∗ ) ∂φ∗

38
Realising that, for example,
δφ = φ0 − φ = (eiθ − 1)φ =⇒ ∂µ (δφ) = δ(∂µ φ) (109)
and using the by now familiar trick of integrating by parts, we arrive at
Z (    
∂L ∂L ∂L ∂L
4
δS = d x iθφ − ∂µ − iθφ∗ − ∂ µ
∂φ ∂(∂µ φ) ∂φ∗ ∂(∂µ φ∗ )
 )
∂L ∂L ∗
+ iθ∂µ φ− φ . (110)
∂(∂µ φ) ∂(∂µ φ∗ )

The first line of the result above equals 0, by virtue of the E.o.M. for both
φ and φ∗ , and in order for the second line to integrate to 0 we must have
   
∂L ∗ ∂L ∗ µ µ ∗
0 ≡ ∂µ φ − φ = ∂µ φ (∂ φ) − (∂ φ )φ . (111)
∂(∂µ φ∗ ) ∂(∂µ φ)
This implies the existence of a conserved current, i.e.
∂µ j µ = 0 (112)

with the current obtained from the equation above

←
→
 
µ ∗ µ µ ∗
j = φ (∂ φ) − (∂ φ )φ ≡ φ∗ ∂ µ φ . (113)

Here, we have introduced the compact shorthand notation

←
→
 
µ µ µ
a ∂ b = a (∂ b) − (∂ a)b . (114)

Conserved Charge The current from Eq. (113) implies the existence
of a conserved charge Q with dQ/dt = 0, constructed by integrating the
temporal component over three-dimensional space,
Z
Q = d3 x j 0 . (115)

In the case of complex scalars this means that

Z  
dQ d 3 1 ∗ ∗
= d x φ (∂t φ) − (∂t φ )φ
dt dt 2
∂j 0
Z Z
= d3 x = d3 x ∇ · j = 0 (116)
∂t
for the three-current vanishing because the fields are assumed to vanish for
|x| → ∞. Here, we have used the fact that the current is conserved,
∂µ j µ = ∂t j t − ∇ · j = 0 −→ ∂t j t − ∇ · j . . (117)

39
3.4 Vector Fields: Maxwell’s Equations
A Little Game of Symmetry Assume you want to introduce two differ-
ent three-vector fields. From a (classical) symmetry point of view, they can
be distinguished through parity, i.e. one of them is parity-odd – a “proper”
vector – while the other one is parity-even – an axial-vector. We call the
parity odd fields (or 1− in spin-parity notation) E, and the parity even ones
(or 1+ ) B. Now let us assume that you only want to allow first derivatives
of the fields, ∂t and ∇ and scalar and pseudo-scalar charge densities ρE,B
and corresponding currents j E,B . Then you can sort resulting quantities by
spin and parity as in Table 3.

name JP allowed terms

scalars 0+ ∇ · E, ρE

pseudo-scalars 0− ∇ · B, ρB

vectors 1− ∂t E, ∇ × B, j E

axial-vectors 1− ∇ × E, ∂t B, j B

Table 3: Terms in Maxwell’s equations, by spin and parity

Symmetry to Dynamics Each of the four rows in Tab. 3 collects possi-

ble terms in one of the four equations defining the system, and this is where
we will introduce data to the game. First of all we identify E and B with
electric and magnetic fields, respectively. Then we realise that to date no
magnetic monopoles have been found, and therefore there is no magnetic
charge density of current, ρB = 0 and j B = 0. Adding lastly that electrody-
namics is a theory of light, and thereby fixing prefactors and signs we arrive
at Maxwell’s equations

∇ · E = 4πρE ∇·B = 0
(118)
∇ × B − ∂t E = 4πj E ∇ × E + ∂t B = 0

Note that we absorbed the usual factors of 0 and µ0 into the definition of
the charge and current, and we have used natural units with c = 1.

40
The Vector Potential The left column in Eq.(118) suggest to use a scalar
potential Φ, which we denote as A0 , and a vector potential A and write

E = −∇A0 − ∂t A and B = ∇ × A . (119)

Of course this now forms a four-vector potential Aµ = (A0 , A), and we will
continue the analysis of electrodynamics mainly based on this object.

Gauge Transformation and Gauge Invariance One of the first bene-

fits of introducing the vector potential is that it is relatively easy to formulate
gauge transformations. To this end we introduce an arbitrary scalar gauge
function, Λ(x), under which A(x) transforms as 6

Aµ → A0µ = Aµ − ∂ µ Λ , (120)

and therefore

E → E 0 = − ∇(A0 − ∂ 0 Λ) − ∂t (A + ∇ · Λ) = −∇A0 − ∂t A = E
B → B 0 = ∇ × (A + ∇ · Λ) = ∇ × A = B , (121)

where we have used that rot·grad of a scalar function vanishes. This suggest
that it would be beneficial to express the theory in terms of gauge invariant
quantities made from Aµ , to directly encode this symmetry.

The Field-Strength Tensor One such gauge-invariant quantity is the

anti-symmetric field-strength tensor
 
0 −Ex −Ey −Ez
 Ex 0 −Bz By 
F µν = ∂ µ Aν − ∂ ν Aµ =  , (122)
 Ey Bz 0 −Bx 
Ez −By Bx 0

and another such tensor is its dual,

 
0 −Bx −By −Bz
1  Bx 0 Ez −Ey 
F̃ µν = µνσρ Fρσ =
 By −Ez
. (123)
2 0 Ex 
Bz Ey −Ex 0

They allows to express the inhomogeneous and homogeneous Maxwell’s

equations, i.e. the left and right column of Eq. (118), as

∂µ F µν = 4πj ν and ∂µ F̃ µν = 0 . (124)

6
Remember that ∂ µ = (∂t , −∇)!

41
Lagrange Density in Terms of the Fields There are various ways
to express the Lagrange density; a version probably familiar from previous
lectures expresses it through the electric and magnetic fields and reads

E2 − B2
L= − ρφ + j · A . (125)
8π

The E.o.M. are obtained in terms of the potential φ and A, using the fact
that the electromagnetic fields are expressed through their derivatives. This
also fixes the two homogeneous Maxwell equations, i.e. the right column of
Esq. (118). This also implies that we are left with the task to check if the
Lagrange density above yields the correct inhomogenous equations – the left
column of Eq. (118).
For example, for φ we have:
∂L
= −ρ
∂φ
∂L Ek ∂Ek Ek
= = − , (126)
∂(∂φ/∂xk ) 4π ∂(∂φ/∂xk ) 4π

where
∂Ek
= −1 (127)
∂(∂φ/∂xk )

follows directly from Eq. (121). Assembling all parts, and making the sum-
mation over repeated indices explicit therefore yields Gauss’ law,
X ∂ ∂L

∂L ∇·E
− = − + ρ = 0. (128)
∂xk ∂(∂φ/∂xk ) ∂φ 4π
k

Similarly, for an arbitrary component of A, Ai we find

∂L
= ji
∂Ai
∂L Ei ∂Ei Ei
= = −
∂(∂Ai /∂t) 4π ∂(∂Ai /∂t) 4π
∂L Bk ∂Bk Bk
= − = −ijk
∂(∂Ai /∂xj ) 4π ∂(∂Ai /∂xj ) 4π
∂L Bj ∂Bj Bj
= − = −ijk , (129)
∂(∂Ai /∂xk ) 4π ∂(∂Ai /∂xk ) 4π

where Eq. (121) has again been used, noting that, expressed in component
notation

B = ∇ × A ←→ Bk = ijk ∂i Aj . (130)

42
specialising i = 1 we are left with Ampere’s law,
 
1 ∂B3 ∂B2 1 ∂E1
− − − j1 = 0 , (131)
4π ∂x2 ∂x3 4π ∂t
or, in vector form,
∂E
∇×B− = 4πj . (132)
∂t

Lagrange Density in Terms of the Field Strength Tensor One ob-

vious short-coming of the form of the The Lagrange density in Eq. (125) is
that it is not manifestly gauge-invariant. This can be overcome by recon-
structing a Lagrange density not in terms of the electromagnetic fields but
rather in terms of the field strength tensor. Rearranging factors of 4π and
introducing an overall sign we arrive at

1
L = − F µν Fµν − 4πj µ Aµ . (133)
4

For the “source” term j µ Aµ , which couples the potentials to charge and cur-
rent densities, we have assumed the so-called “minimal coupling”, typically
of the form source · fields, in a Lorentz-invariant way. This form also fixes
the gauge transformation of the four-vector current j µ . The (E 2 − B 2 )-term
is replaced by a product of field-strength tensors, by realising that

F µν Fµν = −F νµ Fµν
  
0 −Ex −Ey −Ez 0 −Ex −Ey −Ez
 Ex 0 −Bz By   Ex 0 Bz −By 
= Tr    
 Ey Bz 0 −Bx   Ey −Bz 0 Bx 
Ez −By Bx 0 Ez By −Bx 0
 2 
−E • • •
 • −Ex + Bz2 + By2
2 • • 
= Tr  
 • 2 2 2

• −Ey + Bz + Bx • 
• • • 2 2 2
−Ez + By + Bx

= − 2(E 2 − B 2 ) .
(134)

3.5 Hamiltonian Formulation

Hamilton Density In analogy to the case of point particles, momenta πi
conjugate to the fields φi are defined through

∂L(φi , ∂µ φi )
πi (x) = (135)
∂(∂t φi )

43
and a Hamilton density is constructed as

X
H= πi φ̇i − L . (136)
i

The Hamilton function reads

Z Z  
3 3 ∂L
H = d x H(φi , πi ) = d x φ̇i − L . (137)
∂ φ̇i

Equations of Motion Similarly to the case of point particles, the Hamil-

ton E.o.M. read
∂H ∂H
= −π̇i and = φ̇i . (138)
∂φi ∂πi

44
3.6 Problems & Solutions
1. General Solutions for the Klein-Gordon Equation
Consider a real scalar field, given by thr Klein-Gordon Lagrangian,
Eq. (96).

(a) Proof that the solutions to its Equation of Motion, Eq. (91), are
given by the expression in Eq. (93).
(b) Calculate the Hamiltonian and momentum for a free scalar field
using their definitions,
Z
1
d3 x (∂t φ)2 + (∇φ)2 + m2 φ2
 
H =
2
Z
P = − d3 x [(∂t φ)(∇φ)] .

Solution

(a) Inserting the solution for the Klein-Gordon equation from Eq. (93)
into the E.o.M. yields
d3 k
 Z  
2 −ik·x ∗ ik·x
+m a(k)e + a (k)e
(2π)3 2k0
d3 k
Z  
2 −ik·x ∗ 2 ik·x
= a(k)( + m )e + a (k)( + m )e
(2π)3 2k0
d3 k
Z  
2 2 −ik·x ∗ ik·x
= (−k + m ) a(k)e + a (k)e
(2π)3 2k0
d3 k
Z  
2 2 2 −ik·x ∗ ik·x
= (−k0 + k + m ) a(k)e + a (k)e =0
(2π)3 2k0

because of the relativistic energy-momentum relation k02 = k 2 +

m2 .
(b) Inserting the solutions into the two expressions for the Hamilto-
nian and the momentum results in
Z
1
d3 x (∂t φ)2 + (∇φ)2 + m2 φ2
 
H=
2
d3 k d3 q
Z 
1 3
= d x
2 (2π)3 2k0 (2π)3 2q0
a(k)a(q)e−i(k+q)·x −k0 q0 − k · q + m2
 

+ a(k)a∗ (q)e−i(k−q)·x k0 q0 − k · q + m2
 

+ a∗ (k)a(q)e+i(k−q)·x k0 q0 − k · q + m2
 

45
 
∗ ∗ +i(k+q)·x 2
 
+ a (k)a (q)e −k0 q0 − k · q + m

d3 k d3 q
Z 
1
=
2 (2π)3 2k0 (2π)3 2q0
a(k)a(q)(2π)3 δ 3 (k + q)e−i(k0 +q0 )x0 −k0 q0 − k · q + m2
 

+ a(k)a∗ (q)(2π)3 δ 3 (k − q)e−i(k0 −q0 )x0 k0 q0 + k · q + m2

 

+ a∗ (k)a(q)(2π)3 δ 3 (k − q)e+i(k0 −q0 )x0 k0 q0 + k · q + m2

 


∗ ∗ 3 3 +i(k0 +q0 )x0 2
 
+ a (k)a (q)(2π) δ (k + q)e −k0 q0 − k · q + m

d3 k
Z 
1
=
2 (2π)3 (2k0 )2
a(k)a(−k)e−i(k0 +k0 )x0 −k02 + k 2 + m2
 

+ a(k)a∗ (k)e−i(k0 −k0 )x0 k02 + k 2 + m2

 

+ a∗ (k)a(k)e+i(k0 −k0 )x0 k02 + k 2 + m2

 


∗ ∗ +i(k0 +k0 )x0
−k02 2 2
 
+ a (k)a (−k)e +k +m

d3 k
Z  
1 2 ∗ ∗
= 2k a(k)a (k) + a(k)a (k)
2 (2π)3 (2k0 )2 0
d3 k
Z
= k0 a(k)a∗ (k)
(2π)3 (2k0 )

and
Z
P =− d3 x [(∂t φ)(∇φ)]

d3 k d3 q
Z 
3
= − d x
(2π)3 2k0 (2π)3 2q0
a(k)a(q)e−i(k+q)·x (k0 q) + a(k)a∗ (q)e−i(k−q)·x (−k0 q)

+ a∗ (k)a(q)e+i(k−q)·x (−k0 q) + a∗ (k)a∗ (q)e+i(k+q)·x (k0 q)

d3 k d3 q
Z 
= − k0 q
(2π)3 2k0 (2π)3 2q0
a(k)a(q)(2π)3 δ 3 (k + q)e−i(k0 +q0 )x0

46
 − a(k)a∗ (q)(2π)3 δ 3 (k − q)e−i(k0 −q0 )x0

− a∗ (k)a(q)(2π)3 δ 3 (k − q)e+i(k0 −q0 )x0


∗ ∗ 3 3 +i(k0 +q0 )x0
+ a (k)a (q)(2π) δ (k + q)e

d3 k
Z 
= k0 k
(2π)3 (2k0 )2
a(k)a(−k)e−i(k0 +k0 )x0 + a(k)a∗ (k)e−i(k0 −k0 )x0

∗ +i(k0 −k0 )x0 ∗ ∗ +i(k0 +k0 )x0
+ a (k)a(k)e + a (k)a (−k)e

d3 k
Z  
= k0 k a(k)a∗ (k) + a∗ (k)a(k)
(2π)3 (2k0 )2
d3 k
Z
= k a(k)a∗ (k)
(2π)3 2k0

where the terms proportional to ka(k)a(−k) and ka∗ (k)a∗ (−k)

vanish due to the symmetry of the integration.

2. ∗ Klein-Gordon Equation in Two-Component Form

Introduce a two-component form of the real scalar (Klein-Gordon) field
as χ = (χ+ , χ− )T , where
 
1 i ∂φ
χ± = φ± .
2 m ∂t

(a) Rewrite the Klein-Gordon E.o.M. in the Schrödinger form, i.e.

as
i∂χ
= Hχ
∂t
and construct the Hamiltonian as a 2×2 matrix. Solve the energy
eigenvalue equation Hχ = Eχ.
(b) Consider the non-relativistic limit where the solution of the E.o.M.
has the form
χ(t) = e−i(m+T )t χ(0)
for the time-evolution of the fields and where T is the kinetic
energy. What does this imply for the relative sizes of χ+ and χ− ?
Expand the product T χ+ (0) to second order in T and deduce the
first relativistic correction to the Hamiltonian.

47
Solution

(a) The trick is to construct a Hamiltonian which makes sure that

unwtend terms such as single derivatives of φ w.r.t time vanish.
So, looking at

∂t φ + i/m∂t2 φ
 
i∂χ i
= = Hχ
∂t 2 ∂t φ − i/m∂t2 φ

we realise that the Hamitonian must assume the form

∇2
   
1 1 1 0
H=− +m
2m −1 −1 0 −1

because then
∇2
   
φ m φ + i/m∂t φ
Hχ = − +
2m −φ 2 −φ + i/m∂t φ

∇2
 
m i i 1 2
 
−
 2m φ + 2 φ + ∂t φ  ∂ φ− ∂t φ
2  2 t 2m 
=   = 
   
 ∇2 m i  i 1 2

+ φ− φ+ ∂t φ ∂t φ + ∂t φ
2m 2 2 2 2m
This gives rise to two two identical equations, namely the Klein-
Gordon E.o.M.,

∂t2 ψ − ∇2 φ + m2 φ = 0 .

To solve them, let us look again at the Hamiltonian, given by

∇2
   
1 1 1 0
H = − +m
2m −1 −1 0 −1
  2
p2
 2 2

∇ ∇ p
− 2m +m − 2m + m
=   =  2m 2m 
∇2 ∇2 p2 p2
2m 2m − m − 2m − 2m − m

in momentum space and solve the equation

q Hχ = Eχ. The
energy eigenvalues are given by E± = ± p + m2 , as expected.
2

(b) From i∂/∂t χ(t) = Hχ(t) we find that

    
∇2 ∇2
χ+ (t) − +m − 2m χ+ (t)
(m + T )   =  2m  
∇ 2 ∇ 2
χ− (t) 2m 2m − m χ− (t)

48
 and therefore
∇2 ∇2
 
(m + T )χ+ = − − m χ+ − χ−
 2m2 2m
∇2

∇
(m + T )χ− = − m χ− + χ+ .
2m 2m

From the equation for χ− we see that we can approximate

∇2 ∇2
χ− = χ+ · ≈ χ + ·
2m(2m + T ) − ∇2 4m2

in the non-relativistic limit where m  T, ∇2 /(2m). This sug-

gests that χ− is the smaller of the two components. Using the
approximate result for χ− in the equation for χ+ results in

∇2 ∇2
 
T χ+ = − 1+ χ+
2m 4m2

and the first relativistic correction therefore is −∇4 /(8m3 ).

3. Euler-Lagrange Equations of Motion

Find the Euler-Lagrange Equations of Motion for the following La-
grangians

(a) real scalar field φ:

1 m2 2 λ 4
L = (∂µ φ)(∂ µ φ) − φ − φ ;
2 2 4!

(b) “funny” vector field Aµ

m2 λ
L = −(∂µ Aν )(∂ν Aµ ) + Aµ Aµ + (∂µ Aµ )2 ;
2 2

(c) massive vector field Aµ :

1 m2
L = − Fµν F µν + Aµ Aµ ;
4 2

(d) complex scalar fields φ and φ∗ plus the electromagnetic Aµ :

1
L = − Fµν F µν + (∂µ φ∗ + ieAµ φ∗ )(∂µ φ − ieAµ φ) − m2 φ∗ φ ;
4
where in all cases
Fµν = ∂µ Aν − ∂ν Aµ .

49
Solution

Before starting to derive E.o.M.’s let us first look at some relevant

derivatives that appear more than once:

∂Fµν ∂(∂µ Aν − ∂ν Aµ )
= = gµρ gνσ − gνρ gµσ
∂(∂ρ Aσ ) ∂(∂ρ Aσ )

and therefore
1 ∂Fµν F µν 1 ∂Fµν 1
= − F µν = − F µν gµρ gνσ − gνρ gµσ


−
4 ∂(∂ρ Aσ ) 2 ∂(∂ρ Aσ ) 2
1
= − (F ρσ − F σρ ) = −F ρσ = F σρ
2
and
1 ∂Fµν F µν
− ∂ρ = −∂ρ F ρσ = Aσ − ∂ σ (∂ · A) .
4 ∂(∂ρ Aσ )

(a) We quickly arrive at

 
λ 2
+m + φ = 0.
3!

(b) There are three terms to evaluate:

1 ∂(m2 A2 )
= m 2 Aρ
2 ∂Aρ

λ ∂(∂ · A)2
∂ρ = λgµρ ∂ρ g σµ (∂ · A) = λ∂ σ (∂ · A)
2 ∂(∂ρ Aσ )

∂[−(∂µ Aν )(∂ν Aµ )] µ ν 

ν ∂(∂ν A ) µ ∂(∂µ A )
∂ρ = −∂ρ (∂µ A ) + (∂ν A )
∂(∂ρ Aσ ) ∂(∂ρ Aσ ) ∂(∂ρ Aσ )

= −∂ρ (∂µ Aν )g ρν gµσ + (∂ρ ∂ν Aµ )g ρµ gνσ

 

= −2∂ρ ∂ σ Aρ = −2∂ σ (∂ · A)

and therefore we arrive at

(λ − 2)∂ σ (∂ · A) − m2 Aσ = 0

(c) Previous results mean that we only have to put terms together
and arrive at

∂σ F σρ + m2 Aρ = [gσρ ( + m2 ) − ∂σ ∂ ρ ]Aσ = 0 .

50
 (d) Here we have three active fields and arrive at:

(− + ∂σ ∂ µ ) Aσ = ie [φ∗ ∂ µ φ − φ∂ µ φ∗ ] + 2e2 φ∗ φAµ

( + m2 )φ = 2ieAρ ∂ρ φ + ieφ∗ ∂ρ Aρ − e2 A2 φ∗
( + m2 )φ∗ = −2ieAρ ∂ρ φ∗ − ieφ∗ ∂ρ Aρ − e2 A2 φ∗ .

4. Massive Vector Field

The Lagrangian of a massive vector field Vµ is given by

1 m2
L = − Vµν V µν + Vµ V µ ,
4 2
where the field strength tensor Vµν assumes the usual form

Vµν = ∂µ Vν − ∂ν Vµ .

(a) derive the Euler-Lagrange equations of motion from the Lagrangian.

(b) show that the condition

∂µ V µ = 0

is a consequence of the equations of motion.

(c) use this condition and construct three linearly independent po-
(λ)
larisation vectors µ (k) that satisfy this condition, or after Fourier
transformation
kµ µ (k) = 0 .

Solution

(a) For the various derivatives we find

∂L 1 ∂Vµν
= − V µν
∂(∂ρ Vσ ) 2 ∂(∂ρ Vσ )
1 µν ρ σ 1
gµ gν − gνρ gµσ = − (V ρσ − V σρ ) = −V ρσ


= − V
2 2
∂L
= mV σ ,
∂Vσ
and therefore
∂L ∂L
∂ρ − = −∂ρ V ρσ − mV σ = 0.
∂(∂ρ Vσ ) ∂Vσ

(b) Forming the derivative (four-dimensional divergence) of the E.o.M.

of part (a) of the problem yields

∂σ (∂ρ V ρσ + mV σ ) = m∂ · V = 0 .

51
 (c) Assuming the momentum being oriented along the z-axis, we have
p
k µ = (ω, 0, 0, κ) with ω = κ2 + m2

and therefore the following three polarisation vectors are orthonor-

mal:
     
0 0 κ
 1   0  1   0 
(1) =  (2) (3)

 0  ,  =  1  ,  = √−m2  0  .
  

0 0 −ω

5. Electrodynamics with gauge-fixing term

An alternative form for the Lagrange density of the electromagnetic
fields reads
1 1
L = − F µν Fµν − (∂µ Aµ )2 − 4πjµ Aµ ,
4 2
where

F µν = ∂ µ Aν − ∂ ν Aµ

is the field strength tensor for the vector potential Aµ . The additional
term 1/2(∂ · A)2 is also known as “gaug-fixing” term, and in this case
corresponds to the Lorentz gauge. We will come back to this in Section
6 of the notes.
Show that the Euler-Lagange E.o.M. lead directly to a wave equation
of the form

Aµ = ∂ν ∂ ν Aµ = 4πj µ .

Solution

Using Aµ as the dynamical variable, the Euler-Lagrang E.o.M. read

∂L ∂L
0 = ∂ν − .
∂(∂ A ) ∂Aµ
ν µ

Let us first rewrite the product of field strength tensors,

1 1 h κ λ  i
− F κλ Fκλ = − ∂ A − ∂ λ Aκ (∂κ Aλ − ∂λ Aκ )
4 4
1h κ λ i
= − ∂ A ∂κ Aλ − ∂ κ Aλ ∂λ Aκ − ∂ λ Aκ ∂κ Aλ + ∂ λ Aκ ∂λ Aκ
4
1h κ λ i
= − ∂ A ∂κ Aλ − ∂ κ Aλ ∂λ Aκ ,
2

52
 where in the last step we have used that the first and last term are
identical when swapping λ and κ in the last term – which is allowed, as
both are just repeated indices summed over from 0 to 3, and similarly
for the second and third term.
Using the fact that
∂(∂ ρ Aσ )
= g ρξ g σχ
∂(∂ ξ Aχ )
we can write
 
∂ 1 κλ 1 ∂ h
κ λ κ λ
i
− F Fκλ = − ∂ A ∂ κ A λ − ∂ A ∂ λ Aκ
∂(∂ ν Aµ ) 4 2 ∂(∂ ν Aµ )
∂(∂ κ Aλ )

1 ∂(∂κ Aλ )
= − ∂ κ Aλ + ∂ κ Aλ
2 ∂(∂ ν Aµ ) ∂(∂ ν Aµ )
∂(∂ λ Aκ )

κ λ ∂(∂λ Aκ )
−∂κ Aλ − ∂ A
∂(∂ ν Aµ ) ∂(∂ ν Aµ )
1h i
= − ∂κ Aλ g κν g λµ + ∂ κ Aλ gκν gλµ − ∂κ Aλ g λν g κµ − ∂ κ Aλ gλν gκµ
2
= ∂µ Aν − ∂ν Aµ
In addition
 
∂ 1 κ 2
− (∂κ A ) = −∂κ Aκ gλν g λµ = −∂κ Aκ gµν .
∂(∂ ν Aµ ) 2
Therefore
∂L
∂ν = ∂ ν ∂µ Aν − Aµ − ∂µ (∂ · A) = −Aν .
∂(∂ ν Aµ )
Combining this with
∂L
= −4πjµ
∂Aµ
yields the desired result.
6. Free Schrödinger field The Lagrangian of the free Schrödinger Field
is given by
i ∗ 1
L = (φ ∂t φ − φ∂t φ∗ ) − (∇φ∗ ) · (∇φ)
2 2
(a) write down the equations of motion for the Schrödinger field.
(b) show that the conserved current is given by
i ∗
j 0 = φ∗ φ , j= (φ ∇φ − φ∇φ∗ ) ,
2
i.e. show that ∂µ j µ = 0

53
 Solution

(a)
∂L ∂L ∂L
0 = ∂t − ∇ − = −i∂t φ∗ + 1 2 ∗
2∇ φ
∂(∂t φ) ∂(∇φ) ∂φ
∂L ∂L ∂L 1 2
0 = ∂t − ∇ − = +i∂t φ + 2∇ φ .
∂(∂t φ∗ ) ∂(∇φ∗ ) ∂φ∗

(b)
i
∂µ j µ = ∂t (φ∗ φ) − ∇(φ∗ ∇φ − φ∇φ∗ )
 2
= φ∗ (∂t φ) + φ(∂t φ∗ )
 
i ∗ ∗ 2 ∗ 2 ∗
− (∇φ )(∇φ) + φ ∇ φ − (∇φ)(∇φ ) − φ∇ φ
2
     
∗ i 2 i 2
=φ ∂t − ∇ φ + φ ∂t + ∇ φ∗ ,
2 2
and both terms vanish under the E.o.M. above.

7. ∗ Equations of Motion with Boundary Conditions

Consider a real scalar field in 1 + 1 (time+space) dimensions, with an
action given by

+∞ Z+L
m2 2
Z  
1 µ
S= dt dx (∂µ φ)(∂ φ) − φ .
−∞ 2 2
−L

Find the Equation of Motion for the field φ, and discuss the importance
of the boundary terms.

Solution

The E.o.M. are obtained by minimizing the action, i.e.

∂S ∂S
0 = δS == δ(∂µ φ) + δφ
∂(∂µ φ) ∂φ
Z +∞ Z+L
dx (∂ µ φ)δ(∂µ φ) − m2 φδφ
 
= dt
−∞
−L
Z +∞ Z+L
dx (∂t φ)δ(∂t φ) − (∂x φ)δ(∂x φ) − m2 φδφ
 
= dt
−∞
−L

54
 Z +∞ Z+L
dx (∂t φ)∂t (δφ) − (∂x φ)∂x (δφ) − m2 φδφ
 
= dt
−∞
−L
Z+L +∞
 
t=+∞ Z
dt (∂t2 φ)δφ

= dx (∂t φ)δφ  −
t=−∞
−L −∞
+∞ x=+L Z+L
 
Z
dx (∂x2 φ)δφ

− dt (∂x φ)δφ  −
x=−L
−∞ −L
+∞
Z Z+L
− dt dx m2 φδφ
−∞ −L
Z+L t=+∞ Z+∞ Z+L
dx ( + m2 )φδφ ,

= dx (∂t φ)δφ  − dt
t=−∞
−L −∞ −L

where we have assumed, as usual, that the variations of the field vanish
for t → ±∞, δφ(t → ±∞, x) = 0.
This leaves us with the equation of motion

( + m2 )φ = 0 .

This however holds true only either if the (Dirichlet) boundary condi-
tions
δφ(t, , x = ±L) = 0
or if the (Neumann) boundary conditions

∂x φ(t, , x = ±L) = 0

are fulfilled. The latter are better suited for the solution of our prob-
lem, since they are formulated as conditions on the field φ or its deriva-
tive and not on its – in principle arbitrary – variation δφ.
8. Symmetry and Conserved Current
Consider the Lagrangian density for two real scalars φ1, 2 given by
2
m2 2
    
1 µ µ 2 λ 2 2
L= (∂µ φ1 )(∂ φ1 ) + (∂µ φ2 )(∂ φ2 ) − φ1 + φ2 − φ + φ2 .
2 2 4! 1
ˆ Show that it invariant under the transformation

φ1 → φ01 = cos θ φ1 − sin θ φ2

φ2 → φ02 = sin θ φ1 + cos θ φ2 .

ˆ Construct the corresponding conserved current and charge,

55
 Solution

ˆ To see the invariance of the Lagrangian it suffices to realise

1. that the angle θ does not depend on space-time, and therefore
∂µ φ01,2 = cos θ∂µ φ1,2 ∓ sin θφ2,1 ,
2. and that the mixed terms ∝ cos θ sin θ vanish in the sum of
squares (∂µ φ01 )2 + (∂µ φ02 )2 and φ02 02
1 + φ2 , leaving only terms
∝ (cos2 θ + sin2 θ) = 1 behind.
ˆ To construct the conserved current we use that
2
X ∂L
jµ = δφi = ∂µ φ1 δφ1 + ∂µ φ2 δφ2
∂(∂ µ φi )
i=1

and that to first order in the angle

δφ1,2 = φ01,2 − φ1,2 = ∓θφ2,1 .
Therefore the current
jµ = ∂µ φ1 δφ1 + ∂µ φ2 δφ2 = θ (φ1 ∂µ φ2 − φ2 ∂µ φ1 )
is conserved, ∂ µ jµ = 0 – easy to see when using the E.o.M. on
the evaluation of ∂ µ jµ . The conserved charge is given by
Z
Q = d3 x(φ1 φ̇2 − φ̇1 φ2 ) .

9. ∗ A SU (2) Symmetry
Consider a doublet of conplex scalars


Φ= φ1 , φ 2
with dynamics defined by the free Lagrangian
L = (∂µ Φ)† (∂ µ Φ) − m2 Φ† Φ .
(a) Show that this Lagrangian is invariant under the three-parameter
transformations
 
† i
Φ → Φ = exp θa σa Φ ,
2
with the three constant real angles θ1, 2, 3 and where the σa are
the three Pauli matrices,
     
0 1 0 −i 1 0
σ1 = , σ2 = , σ3 =
1 0 i 0 −1 0
which enjoy the commutation relation
[σi , σj ] = iijk σk

56
 Solution

(a) Because the angles are constant, the derivatives do not act on
them, and we only have to evaluate terms of the form
        
iθa † iθb
exp − σa exp + σb = exp X Y = exp Z ,
2 2
where Z is given by the Baker–Hausdorff formula as
1 1 1
Z = X + Y + [X, Y ] + [X, [X, Y ]] − [X, [X, Y ]] + . . . .
2 12 12

Making the summation over a and b explicit and using that σa† =
σa , the commutator is given by
3 3
X θa θ b X iabc θa θb σc
[X, Y ] = [σa , σb ] = = 0,
4 4
a,b=1 a,b=1

because we multiply a symmetric tensor with an anti-symmetric

one. This means that we have
    " 3 3
#
iθa † iθb X iθa
†
X iθb
exp − σa exp + σb = exp − σ + σb = 1 .
2 2 2 a 2
a=1 b=1

Therefore, the Lagrangian density is invariant under the SU (2)

symmetry.

10. Energy-Momentum Tensor

The energy-momentum tensor for a Lagrangian given as a function of
a set of fields {φα } and their first derivatives {∂µ φα } is defined as

∂L
T µν = ∂ ν φα − g µν L .
∂(∂µ φα )

(a) Can you identify the T 00 component?

(b) Can you identify the T 0j component?
(c) Using the Euler-Lagrange E.o.M., show that it is a conserved
quantity, i.e.

∂µ T µν = 0 .

(d) Show that for the free real scalar field, the energy-momentum
tensor is symmetric, T µν = T νµ .
(e) Verify that the interpretations of T 00 and T 0i from questions (a)
and (b) are correct for the free electromagentic field.

57
Solution

(a)
∂L
T 00 = φ̇α − L = πα φ̇α − L = H ,
∂ φ̇α
the Hamiltonian or energy density of the system.
(b)
∂L j
T 0j = ∂ φα − g 0j L = πα ∂ j φα ,
∂ φ̇α
the components of the three-momentum density of the system.
(c) Direct calculation yields
 
µν ∂L ν µν
∂µ T = ∂µ ∂ φα − g L
∂(∂µ φα )
∂L ∂L
= ∂µ · ∂ ν φα + · ∂µ ∂ ν φα − ∂ ν L
∂(∂µ φα ) ∂(∂µ φα )
∂L ∂L
= ∂µ · ∂ ν φα + · ∂µ ∂ ν φα
∂(∂µ φα ) ∂(∂µ φα )
∂L ∂L
− · ∂ ν φα − · ∂ ν ∂µ φα
∂φα ∂(∂µ φα )
 
∂L ∂L
= ∂µ − ∂ ν φα ≡ 0 ,
∂(∂µ φα ) ∂φα
as demanded.
(d) Direct calculation:
∂L
T µν = ∂ ν φ − g µν L
∂(∂µ φ)
g µν 
= − (∂ µ φ)(∂ ν φ) (∂µ φ)(∂ µ φ) − m2 φ2 = T νµ ,

2
as expected. However, there may be more complicated Lagrangians,
where the energy-momentum tensor is not symmetric.
(e) The free field Lagrangian is given by
1 1
L = − Fµν F µν = − (∂µ Aν − ∂ν Aµ )(∂ µ Aν − ∂ ν Aµ ) .
4 4
Energy density: obviously there are no term Ȧ0 due to the sym-
metry of the F µν , this means in the summing over field compo-
nents we can concentrate on the Ȧi
∂L E2 − B2 E2 + B2
T 00 = Ȧi − L = E 2 − = ,
∂ Ȧi 2 2

58
 the energy density of the field, as expected. This, of course, was
clear from the beginning, since we already explicitly calculated
T 00 = H in (a).
Momentum density: the momenta conjugate to Aµ , πµ are given
by

∂L π0 = 0
πµ = =
∂ Ȧµ πi = Ei

This implies that we only have to look at the spatial components

of Aµ and

T 0j = Ei ∂ j Ai = (E × B)j ,

the jth component of the Poynting vector, which encapsulates

the three-momentum density of the fields.

To see that this identification actually holds true, consider T 01 . It is

given by
   
01 ∂A2 ∂A1 ∂A1 ∂A3
T = E2 B3 − E3 B2 = E2 − − E3 −
∂x1 ∂x2 ∂x3 ∂x1
3  
X ∂ ∂A1 ∂A1 ∂A1
= Ei Ai − E1 + E2 + E3
∂x1 ∂x1 ∂x2 ∂x3
i=1
3 3
X ∂ X ∂
= Ei Ai − E · ∇A1 = Ei Ai − ∇ · (A1 E) ,
∂x1 ∂x1
i=1 i=1

where we have used that due to ∇ · E = 0 in the absence of charges,

E · ∇A1 = ∇ · (A1 E) − A1 ∇ · E = ∇ · (A1 E) .

The last term in our expression for T 01 , ∇·(A1 E), is a total derivative.
This means that, when we integrate over all space, this term will vanish
and we are therefore free to ignore it.

59
4 Second Quantisation
At the beginning of this part of the course you may have read the title and
asked yourself: What does second quantisation actually mean? Haven’t we
already quantised the theory? The answer is that in “first quantisation”
we quantised the position and momentum of point particles. This led to
important properties related to our ability to measure them – the uncer-
tainty principle – and to a crucial reassessment of the inner working of the
world around us, replacing Laplace’s demon of deterministic physics with
a determinism of probabilities. So while in this first quantisation we re-
placed the real numbers x and p with operators x̂ and p̂ and constructed
wave functions for the emerging states, in “second quantisation” we quantise
something else, namely the fields. Consequently, x and p become “ordinary”
numbers again, which serve as arguments of the field operators φ̂ and π̂.
This step necessitates the introduction of a new state. While, formally
speaking, the states of Quantum Mechanics constitute a Hilbert space, the
field operators act on objects in a more complicated Fock space, which is not
labelled by eigen-positions or momenta, but by the number of field quanta
of a given momentum. We will, however, not discuss the properties of these
vector spaces in the lecture.
Simply put: while first quantisation quantised the point dynamics of Clas-
sical Mechanics, leading to Quantum Mechanics, second quantisation pro-
duces a Quantum Field Theory.
If you like to do some additional reading, I would recommend to take a closer
look at Chapter 3 of Hatfield’s book [3], in particular sections 3.1-3.4 or at
Sections 2.3-2.4 in Peskin & Schroeder [1].

4.1 A How-To Guide

Process Summary The process of second or field quantisation follows the
logic of the familiar first quantisation programme, with suitably replacing
position and momentum with the field and its conjugate momentum, and by
replacing the δ’s of the commutators with δ-functions of the positions. This
proceeds in a relatively straight-forward “algorithmic” fashion, as outlined
in Fig. 1. The crucial part in it is to demand equal-time commutator rela-
tions between fields and their momenta, which also fixes a Lorentz frame in
which the field quantisation is performed. Obviously, there are other choices
for such a programme, for example a quantisation on the light-cone, which
however is beyond the scope of the lecture here. It is, nevertheless, im-
portant to stress that despite the implicit choice of a Lorentz frame during
quantisation, the resulting theory has the correct causal properties. This
will be shown towards the end of this section.

60
 How-to: Second Quantisation

1. determine conjugate momenta of fields:

π = ∂L/∂(∂t φ) = ∂L/∂ φ̇

2. construct Hamiltonian as function of fields φ and their mo-

menta π Z  
H = d3 x φ̇π − L

3. promote fields and momenta to operators, φ −→ φ̂, π −→ π̂

4. demand equal time commutators of fields and momenta

h i
φ̂(t, x), π̂(t, y) = iδ 3 (x − y)
h i  
φ̂(t, x), φ̂(t, y) = π̂(t, x), π̂(t, y) = 0

5. express fields as linear combination of plane waves and anni-

hilation and creation operators (which will “inherit” com-
mutator relations)
Xh i
φ̂(x) = â(k)e−ik·x + ↠(k)eik·x ,
k

where summation is replaced with integration for continuous

momenta k.

Figure 1: The steps performed during second quantisation

in form of an “algorithm”. Details will be worked out and
highlighted through examples during the course.

4.2 Second Quantisation of the Real Scalar Field

Lagrangian: Fields and Conjugate Momenta Starting with the La-
grangian of Eq. (96),

1 m2 2
L(∂µ φ, φ) = (∂µ φ)(∂ µ φ) − φ ,
2 2
the conjugate momentum reads
∂L
π= = φ̇ . (139)
∂ φ̇

61
Hamiltonian The Hamilton function therefore is given by
Z   Z  
3 3 1 2 2 2 2
H = d x π φ̇ − L = d x π + (∇φ) + m φ . (140)
2

Field Operators and Commutators Promoting the field and its con-
jugate momentum to operators, φ(x) → φ̂(x) and π(x) → π̂(x), we demand
the equal-time commutators,
h i
φ̂(t, x), π̂(t, y) = iδ 3 (x − y)
h i  
φ̂(t, x), φ̂(t, y) = π̂(t, x), π̂(t, y) = 0 (141)

Creation and Annihilation Operators The field and the conjugate

momentum is expressed through creation and annihilation operators as

d3 k
Z h i
φ̂(x) = â(k) e−ik·x + ↠(k) eik·x
(2π)3 2k0
d3 k
Z h i
π̂(x) = −ik0 â(k) e−ik·x + ik0 ↠(k) eik·x , (142)
(2π)3 2k0

where we have obtained the momentum operator through straightforward

calculation of the derivative π̂(x) = ∂t φ̂(x). Comparing the expression for
the field operator φ̂(x) in the equation above with the solution to the Klein-
Gordon equation for the classical field φ(x), Eq. (93), we recognise the same
pattern of an expansion in amplitude factors and plane waves. But while the
amplitude factors for the classical field are merely numbers a(k) and their
complex conjugate a∗ (k), they become operators for the quantised fields,
and the complex conjugation turns into an Hermitean conjugate7 . The in-
terpretation is clear. While for classical fields every value of the amplitude
is allowed, in quantised fields the amplitude is composed by adding finite
quanta. This “amplitude quantisation” is reflected by using creation and
annihilation operators from which the field “inherits” its quantised proper-
ties. We will build on this in the following by expressing the Hamiltonian
through these operators, by creating a number operator, and by analysing
their inherent properties.

Commutators of the Creation and Annihilation Operators To cal-

culate the commutators it is necessary to express, in a first step, the creation
and annihilation operators through the field and conjugate momentum op-
erators. To see how this works, let us first try to combine φ̂ and π̂ in such
7
As a result the field operator is Hermitean φ̂ = φ̂† , which guarantees that it has real
eigenvalues – as you would expect from a real scalar field.

62
a way that the annihilation operator â drops out. Multiplying, inside the
integral, the expression for φ̂ with k0 and π̂ with i and adding the expression
for π̂ we arrive at
d3 k 1 h
h i Z i
† ik·x
“ k0 φ̂(x) − iπ̂(x) ” = 2ik 0 â (k)e
(2π)3 2k0
d3 k †
Z
= â (k)e−ik·x eik0 x0 , (143)
(2π)3

which looks suspiciously like the Fourier transform of ↠times a factor.
Therefore, Fourier-back-transforming yields

d3 q †
Z h i Z Z
d3 xeik·x k0 φ̂(x) − iπ̂(x) = d3 xeik·x â (q)e−iq·x eik0 x0
(2π)3
d3 q 3
Z
= δ (k − q)↠(q)eik0 x0 = ↠(k)eik0 x0 .
(2π)3
(144)

After rearranging and repeating similar steps to extract the annihilation

operator â,
Z h i
â(k) = d3 xeik·x k0 φ̂(x) + iπ̂(x)
Z h i
†
â (k) = d3 xe−ik·x k0 φ̂(x) − iπ̂(x) . (145)

The equal-time commutators of the creation and annihilation operators can

be readily calculated as
Z  
3 3 ik·x+iq·y
 
â(k), â(q) = d xd ye k0 φ̂(x) + iπ̂(x), q0 φ̂(y) + iπ̂(y)
Z     
= d3 xd3 yeik·x+iq·y k0 q0 φ̂(x), φ̂(y) − π̂(x), π̂(y)
   
+ik0 φ̂(x), π̂(y) + iq0 π̂(x), φ̂(y)
Z  
3 3 ik·x+iq·y 3 3
= d xd ye 0 − 0 − k0 δ (x − y) + q0 δ (y − x)
Z  
3 i(k+q)·x
= d xe (q0 − k0 )
 
i(k0 +q0 )x0
=e (q0 − k0 ) (2π)3 δ 3 (k + q) = 0 (146)

because k = −q from the δ function implies that k 2 = q 2 and therefore k0 =

q0 . Similarly, with the only difference being an ultimately inconsequential

63
relative sign in front of both momentum operators and in both exponential
factors,
h i
↠(k), ↠(q) = 0 . (147)

This leaves us with calculating

h i Z  
† 3 3 ik·x−iq·y
â(k), â (q) = d xd ye k0 φ̂(x) + iπ̂(x), q0 φ̂(y) − iπ̂(y)
Z     
3 3 ik·x−iq·y
= d xd ye k0 q0 φ̂(x), φ̂(y) + π̂(x), π̂(y)
   
−ik0 φ̂(x), π̂(y) + iq0 π̂(x), φ̂(y)
Z  
3 3 ik·x−iq·y 3 3
= d xd ye 0 − 0 + k0 δ (x − y) + q0 δ (y − x)
Z  
3 i(k−q)·x
= d xe (k0 + q0 )
 
i(k0 −q0 )x0
=e k0 + q0 ) (2π)3 δ 3 (k − q) = 2k0 (2π)3 δ 3 (k − q)
(148)

Putting it all together, we arrive at the following set of commutation rela-

tions between the creation and annihilation operators
h i
â(k), ↠(q) = 2k0 (2π)3 δ 3 (k − q)
h i
â(k), â(q) = ↠(k), ↠(q) = 0 .
 
(149)

Hamilton Operator In a next step in our analysis of the theory we ex-

press the Hamilton operator of Eq. (140) through the creation and annihila-
tion operators. A somewhat tricky part are the quadratic terms such as φ2
and similar. For them, we need to use the expansion of Eq. (142) for each
factor, leading to two integrals over three-momenta k and k 0 :
Z  
3 1 2 2 2 2
Ĥ = d x π̂ + (∇φ̂) + m φ̂
2
d3 k d3 k 0
Z Z 
1 3 1
= d x
2 2 (2π)3 2k0 (2π)3 2k00
 
0
â(k)â(k ) −k0 k0 − kk + m e−i(k+k )·x
0 0 0 2

 
0
+â(k)â (k ) +k0 k0 kk + m e−i(k−k )·x
† 0 0 0 2

64
  
† 0 0 0 0
+â (k)â(k ) +k0 k0 kk + m e+i(k−k )·x
2

  
† † 0 0 0 2 +i(k+k0 )·x
+â (k)â (k ) −k0 k0 − kk + m e

d3 k d3 k 0
Z 
1
=
2 (2π)3 2k0 (2π)3 2k00
 
3 3 0 −i(k0 +k00 )x0 0 0 0 2
(2π) δ (+k + k ) e â(k)â(k ) −k0 k0 − kk + m
 
3 3 0 −i(k0 −k00 )x0 † 0 0 0 2
+ (2π) δ (+k − k ) e â(k)â (k ) +k0 k0 + kk + m
 
3 3 0 +i(k0 −k00 )x0 † 0 0 0 2
+ (2π) δ (−k + k ) e â (k)â(k ) +k0 k0 + kk + m
 
3 3 0 +i(k0 +k00 )x0 † † 0 0 0 2
+ (2π) δ (−k − k ) e â (k)â (k ) −k0 k0 − kk + m

d3 k
Z   
1 −2ik0 x0 2 2 2
= â(k)â(−k)e −k0 + k + m
2 (2π)3 4k02
   
† 2 2 2 † 2 2 2
+â(k)â (k) +k0 + k + m + â (k)â(k) +k0 + k + m
 
† † 2ik0 x0 2 2 2
+â (k)â (−k)e −k0 + k + m

d3 k
Z    
1 2 † † ˆ
= 2k0 â(k)â (k) + â (k)(k)
2 (2π)3 4k02
d3 k
Z  
1 † † ˆ
= k0 â(k)â (k) + â (k)(k) , (150)
2 (2π)3 2k0
where the δ functions in the first step emerge from the integral over x and
where we have eliminated the terms ââ and ↠↠by realising that due to
the relativistic energy-momentum relation k02 = k 2 + m2 . Therefore, the
Hamilton operator for the real scalar field is given by

d3 k
Z  
1 † † ˆ
Ĥ = k0 â(k)â (k) + â (k)(k) (151)
2 (2π)3 2k0

It suggests that the Quantum Field Theory for a real scalar field can be in-
terpreted as a continuous sum of harmonic oscillators, permeating all space.

Simple States: Ground State and First Excited State Following the
same logic already present in the harmonic oscillator in Quantum Mechanics
we introduce a ground state – the “vacuum” – |0i which is annihilated by
any annihilation operator,

â(k) |0i = 0 ∀k . (152)

65
States containing fields (or particles) with momenta k i are generated by
repeated application of the corresponding creation operators
↠(k 1 ) |0i = |k 1 i
↠(k 1 )↠(k 2 ) |0i = |k 1 k 2 i . . . . (153)

Normalisation of States But here’s a new problem. Let us take a look

at the norm of a one-field (one-particle) state |ki:
 2  †  
 2  † † †
hk|ki = |ki = â (k) |0i  = â (k) |0i â (k) |0i
  
D h i E
= h0|â(k)↠(k)|0i = 0  â(k)↠(k) − ↠(k)â(k)  0
 

= 0 2k0 (2π)3 δ 3 (k − k) 0 = 0 2k0 (2π)3 δ 3 (0) 0

  
  

= 2k0 (2π)3 δ 3 (0) , (154)

where in the second line we have subtracted a 0 – remember that â |0i = 0,
and where in the last step we used our normalisation of the ground state,
h0|0i = 1.
Using that
Z Z
3 3 ik·x
(2π) δ(k) = d xe −→ (2π) δ(0) = d3 x
3
(155)

suggests that the normalisation of the state equals the (infinite) spatial vol-
ume – a veritable divergence. This is actually not a surprising finding, after
all, the uncertainty principle tells you that a particle with completely fixed
momentum has no localisation. Our particle here, with its fixed momentum
represents a plane wave, filling all volume. If the volume is infinite – which it
is for us to have a continuous spectrum – such states have no normalisation.
The solution to this conundrum is to “smear” the state with a modulating
function f (k), and to define
|ki −→ |kif = f (k)↠(k) |0i (156)
which will lead to perfectly normalisable states, if
Z
d3 k |f (k)|2 < ∞ . (157)

In this respect, states obtained through application of the creation operator

on the vacuum, ↠(k) |0i are physically sensible only, if used together with
a test function that smear them out.
A natural question to ask is: in what space to these new states live? The
answer is that they populate a Fock space, which is the sum of all n-particle
Hilbert spaces plus, possibly, some symmetrisation that takes care of the fact
that identical particles are indistinguishable. It also has a meaningful scalar
product, again allowing the definition of a distance, which it “inherits” from
the underlying Hilbert space.

66
Ground-State Energy The ground state |0i is an eigenstate of the Hamil-
tonian; calculating its energy E0 we arrive at

d3 k
  Z i 
1 h
† †
E0 = h0| Ĥ |0i = 0   k0 â(k)â (k) + â (k)â(k)  0
2 (2π)3 2k0
d3 k
Z
1 D 
†
 E
= k 0 â(k)â (k)0
 
0
2 (2π)3 2k0

Z
1
= d3 kk0 δ 3 (0) = ∞ , (158)
2
the product of the volume in both position and momentum space, infinity
for a Quantum Field Theory in an infinite volume. A simple solution is to
just subtract the ground state energy, by redefining the Hamiltonian as

Ĥ −→ Ĥ 0 = Ĥ − h0| Ĥ |0i . (159)

Normal Ordering Alternatively, we can define normal-ordering of the

operators, indicated by colons around the operators as

:â(k)↠(k): = :↠(k)â(k): = ↠(k)â(k) , (160)

and therefore we replace

d3 k
Z
1 h
† †
i
Ĥ −→ :Ĥ: = k0 : â(k)â (k) + â (k)â(k) : (161)
2 (2π)3 2k0
and, finally,

d3 k
Z
:Ĥ: = k0 ↠(k)â(k) . (162)
(2π)3 2k0

This obviously cures the divergence stemming from h0| â(k)↠(k) |0i in the
ground state energy and similar observables. In the remainder of this lecture
we will always assume implicit normal ordering, if not stared otherwise.

4.3 A Little Detour: Causal Structure of the Theory

Commutator of Field Operators To guarantee the correct causal struc-
ture of the theory, we need to convince ourselves that fields in space-like
distances cannot influence each other: they must commute for space-like
distances. To see this, define the commutator of two field operators at arbi-
trary four-positions x and y as
h i
i∆(x − y) = φ̂(x), φ̂(y)
d3 k d3 k 0 nh
Z i
† 0 0
= 3 3 0 â(k), â (k ) e−ik·x+ik ·y
(2π) 2k0 (2π) 2k0

67
 h i 0
o
+ ↠(k), â(k 0 ) eik·x−ik ·y
d3 k
Z n o
= e−ik·(x−y) − eik·(x−y)
(2π)3 2k0
= ∆+ (x − y) − ∆− (x − y) , (163)

where we have used the commutator relations for the creation and annihi-
lation operators to arrive at two δ-functions that allowed us to perform the
k 0 -integration, and where we have also introduced the two terms ∆± (x − y).

Properties of the Commutator The commutator has a number of prop-

erties, which are worthwhile to discuss:

1. it is manifestly Lorentz-invariant, i.e. its value will not change under

Lorentz transformations such as boosts, rotations, or combinations of
both. This is because it is given by a Lorentz-invariant integral over
a function that only depends on scalar products k(x − y).

2. it manifestly encodes micro-causality, as it vanishes for all space-like

distances of x and y. The easiest way to see this is by looking directly
at the first line of the equation above, Eq. (163), where the argument
of the integration depends on [φ̂(x), φ̂(y)]. We know that this commu-
tator vanishes for equal times, cf. Eq. (141). Since for every space-like
distance of four-positions (x − y) a Lorentz-boost can be found that
reduces it to a space-like distance at equal times, (x − y), the commu-
tator vanishes for all space-like distances. Hence, the theory is causal
in the sense that fields at space-like distances are decoupled.

3. direct calculation reveals that ∆(x − y) is a solution of the Klein-

Gordon equation,

0 = ∂ µ ∂µ + m2 ∆(x − y)

d3 k h −ik·(x−y)
Z i
= ∂ µ ∂µ + m2 ik·(x−y)


e − e
(2π)3 2k0
Z 3
d k
h −ik·(x−y) i
µ 2 ik·(x−y)
= ∂ ∂ µ + m e − e
(2π)3 2k0
d3 k
Z
2 2

h −ik·(x−y) ik·(x−y)
i
= −k + m e − e (164)
(2π)3 2k0

where the term k 2 − m2 in the last line guarantees that the overall
expression vanishes.

68
4.4 Second Quantisation of the Complex Scalar Field
Lagrangian and Hamilton and Field Operators Starting with the
Lagrangian of Eq. (104),

L = (∂µ φ∗ )(∂ µ φ) − m2 φ∗ φ ,

the conjugate momenta to the two fields φ and φ∗ are given by

∂L ∂L
π = = φ̇∗ and π ∗ = = φ̇ (165)
∂ φ̇ ∂ φ̇∗

and the Hamilton operator density reads

Ĥ = π̂ ∗ π̂ + ∇φ̂∗ · ∇φ̂ + m2 φ̂∗ φ̂ , (166)

after promoting fields and momenta to operators. We demand the equal-

time commutation relations
h i h i
φ̂(t, x), π̂(t, y) = φ̂∗ (t, x), π̂ ∗ (t, y) = iδ 3 (x − y) (167)

with all other equal-time commutators vanishing.

Creation and Annihilation Operators The field operators can be ex-

panded, as before, as products of plane waves and creation and annihilation
operators

d3 k
Z h i
φ(x) = â(k) e−ik·x + b̂† (k) eik·x
(2π)3 2k0
d3 k
Z h i
φ∗ (x) = b̂(k) e−ik·x + ↠(k) eik·x . (168)
(2π)3 2k0

As before, momentum operators are obtained through straightforward deriva-

tion with respect to time. Expressing the fields and operators through the
annihilation and creation operators,
Z h i
â(k) = d3 xe+ik·x k0 φ̂(x) + iπ̂ ∗ (x)
Z h i
†
b̂ (k) = d3 xe−ik·x k0 φ̂(x) − iπ̂ ∗ (x)
Z h i
b̂(k) = d3 xe+ik·x k0 φ̂∗ (x) + iπ̂(x)
Z h i
†
â (k) = d3 xe−ik·x k0 φ̂∗ (x) − iπ̂(x) , (169)

69
we arrive at commutator relations, for example,
h i Z
0 0
â(k), ↠(k 0 ) = d3 xd3 x0 e+ik·x−ik ·x
h i
× k0 φ̂(t, x) + iπ̂ ∗ (t, x), k00 φ̂∗ (t, x0 ) − iπ̂(t, x0 )
Z n h i
0 0
= d3 xd3 x0 e+ik·x−ik ·x −ik0 φ̂(t, x), π̂(t, x0 )
h io
+ ik00 π̂ ∗ (t, x), φ̂∗ (t, x0 )
Z
0 0 
d3 xd3 x0 e+ik·x−ik ·x (k0 + k00 )δ 3 (x − x0 )

=
Z
0
= d3 xe+i(k−k )·x (k0 + k00 ) = 2k0 (2π)3 δ 3 (k − k 0 ) , (170)

where we have used the definition of the δ function, as usual. Therefore,

h i h i
â(k), ↠(k 0 ) = b̂(k), b̂† (k 0 ) = 2k0 (2π)3 δ 3 (k − k 0 ) (171)

and all other commutators vanishing.

Hamilton and Number Operators The normal-ordered Hamilton op-

erator is given by

d3 k
Z h i
† †
:Ĥ: = k 0 â (k)â(k) + b̂ (k)b̂(k) , (172)
(2π)3 2k0

and it looks like the Hamilton operator for the sum of two free real scalar
fields. This further fortifies the idea that we are presented by two kinds of
particles – those created and annihilated by ↠and â, and those created and
annihilated by b̂† and b̂, and that the vacuum is annihilated by both â and
b̂,
â(k) |0i = b̂(k) |0i = 0 . (173)
It is therefore natural to introduce two number operators for the two kinds
of particles,
d3 k
Z
N̂a = ↠(k)â(k)
(2π)3 2k0
d3 k
Z
N̂b = b̂† (k)b̂(k) . (174)
(2π)3 2k0
It is easy to check that they are indeed number operators counting the
number of a and b fields in a given state |ψi. Denoting
 E Yna h nb h
iY i
 (a) (a) (a) (b) (b) (b) † †
k
 1 2k . . . k k
na 1 k 2 . . . k nb = â (k i ) b̂ (k i |0i ,
) (175)
i=1 i=1

70
it is easy to show that
D   E
(a) (a) (a) (b) (b)  (a) (a) (b) (b)
k 1 k 2 . . . k na k 1 k 2 . . . k nb  N̂a k 1 k 2 . . . k (a)
(b)  (b)
na k 1 k 2 . . . k nb = na .
(176)

We leave this as part of a problem.

Current and Charge As noted in Sec. 3.3, the Lagrangian for the com-
plex scalar field enjoys invariance under the “gauge transformation”

φ → φ0 = exp(iθ)φ , φ∗ → φ0∗ = exp(−iθ)φ∗ ,

cf. Eq. (106). This leads to a conserved current given by Eq. (113)

←→
 
j µ = i φ∗ (∂ µ φ) − (∂ µ φ∗ )φ ≡ iφ∗ ∂ µ φ ,

where we added a factor i to ensure that the current is a real number. This
factor, obviously, does not change the fact that ∂µ j µ = 0. Of course the
current can be promoted to a current operator by replacing the fields with
field operators,
←→
ĵ µ = iφ̂∗ ∂ µ φ̂ . (177)

The spatial integral over the (normal-ordered) time-component of the cur-

rent is the charge, given in operator form by
Z Z
:Q̂: = d3 x :ĵ0: = i d3 x :φ̂∗ (∂t φ̂) − (∂t φ̂∗ )φ̂:

d3 k h †
Z i
†
= â (k)â(k) − b̂ (k) b̂(k) = N̂a − N̂b . (178)
(2π)3 2k0

This suggests that our two particle types a and b have opposite charged with

qa,b = ±1 . (179)

Conserved Charge To show that the charge is conserved, we need to

verify that the charge operator

d3 k h †
Z i
†
:Q̂: = â (k)â(k) − b̂ (k)b̂(k) = N̂a − N̂b (180)
(2π)3 2k0

commutes with the Hamiltonian, i.e. [:Ĥ:, :Q̂:] = 0. This is indeed the case,
and we leave this proof for the problems below.

71
4.5 Problems & Solutions
1. States and Operators of the Real Scalar Field

(a) one- and two particle states

i. create one- and two-particle states of particles with momenta
k 1,2 , |k 1 i and |k 1 k 2 i
ii. show that the two-particle state |k 1 k 2 i is symmetric, i.e.
|k 1 k 2 i = |k 2 k 1 i.
(b) calculate the energy of the two-particle state above, i.e.

E12 |k 1 k 2 i = Ĥ |k 1 k 2 i .

(c) show that the number operator N̂ counts the number of quanta:

N̂ |k 1 k 2 . . . k n i = n|k 1 k 2 . . . k n i (181)

Solution

(a) In real scalar field theory,

|k 1 i = ↠(k 1 )|0i
|k 1 k 2 i = ↠(k 1 )|k 2 i = ↠(k 1 )↠(k 2 )|0i = ↠(k 2 )↠(k 1 )|0i

The last equality shows the symmetry of the state.

(b) With the (normal-ordered) Hamilton operator given by

d3 k
Z
1 h 
† †
i
:Ĥ: = k 0 : â (k)â(k) + ââ (k)(k) :
2 (2π)3 (2k0 )
d3 k
Z h i
†
= k 0 â (k)â(k) ,
(2π)3 (2k0 )

the energy of the state |k 1 k 2 i is given by

Ek1 k2 |k 1 k 2 i = Ĥ|k 1 k 2 i
d3 k  †
Z  
1 
† †

= 3
â (k)â(k) â (k 1 )â (k 2 ) 0
2 (2π)
d3 k
Z 
1 †
h
†
i
†
= â (k ) â(k), â (k 1 â (k 2 )
)
2 (2π)3
 
+ ↠(k)↠(k 1 )â(k)↠(k 2 )  0


d3 k
Z   
1
q
† † 3 2 2
= â (k)â (k 2 ) (2π) δ(k − k 2 ) 2 k + m
2 (2π)3

72
 h i
+ ↠(k)↠(k 1 ) â(k), ↠(k 2 )
 
† † †

− â (k)â (k 1 )â (k 2 )â(k)  0

d3 k
Z   
1
q
† † 3 2 2
= â (k)â (k 2 ) (2π) δ(k − k 1 ) 2 k + m
2 (2π)3
 q   
+ ↠(k)↠(k 1 ) (2π)3 δ(k − k 2 ) 2 k 2 + m2 − 0  0


 q 
1
q  E
= 2 k 1 + m + 2 k 2 + m ↠(k 1 )↠(k 2 ) 0
2 2 2 2 
2
= (E1 + E2 ) |k 1 k 2 i

as anticipated.
(c) With the number operator given by

d3 k
Z
N̂ = ↠(k)â(k)
(2π)3 (2k0 )

and satisfying the commutator

d3 k
Z h i
† † † † †
[N̂ , â (q)] = â (k)â(k)â (q) − â (q)â (k)â(k)
(2π)3 (2k0 )
d3 k
Z h
= ↠(k)(2π)3 (2q0 )δ 3 (k − q)
(2π)3 (2k0 )
i
+↠(k)↠(q)â(k) − ↠(q)↠(k)â(k)
= ↠(q) .

we can calculate the result of it acting on a multi-particle state:

N̂ |k 1 k 2 . . . k n i = N̂ ↠(k 1 )|k 2 . . . k n i
= ↠(k 1 )(N̂ + 1)|k 2 k 3 . . . k n i = ↠(k 1 )↠(k 2 )(N̂ + 2)|k 3 . . . k n i = . . .
= ↠(k 1 )↠(k 2 ) . . . ↠(k n )n|0i = n|k 1 k 2 . . . k n i

as anticipated.

2. Wave Functional from State Vectors

Show that the wave functional of the field with fixed momentum k

φk (x) = hk| φ̂(x) |0i

is a solution of the Klein-gordon Equation.

73
 Solution

0 = ( + m2 )φk (x) = ( + m2 ) hk| φ̂(x) |0i

d3 q h −iq·x
  Z i 
2 iq·x †

= ( + m ) 0 â(k)
 e â(q) + e â (q)  0
(2π)3 2q0
d3 q
 Z h i 
2 iq·x †

= ( + m ) 0   e â(k), â (q)  0
(2π)3 2q0
3q
 Z  
2
 d iq·x 3 3

= ( + m ) 0  3
e (2π) 2q0 δ (q − k) 0
(2π) 2q0
= ( + m2 )eik·x = −k 2 + m2 = 0 ,

which vanishes due to the relativistic energy-momentum relation.

3. Two Real Scalar Fields Equal One Complex Scalar Field

(a) write down the Lagrangian two real scalar fields φ1,2 of equal
mass and determine their canonical momenta π1,2
(b) construct the Hamiltonian from the fields and their momenta
(c) demand suitable commutators for fields and momenta
(d) express the fields in terms of creation and annihilation operators
and determine their commutation relations
(e) introduce the complex scalar fields φ and φ∗ as linear combi-
nations of φ1,2 , express them through creation and annihilation
operators. Fix the commutators of the creation and annihilation
operators that have not been explicitly calculated so far.
(f) calculate the commutator of the number operators with the cre-
ation and annihilation operators of the two fields, the commuta-
tor between the two number operators and with the charge and
Hamilton operator. Show that the charge is conserved by showing
that indeed [:Ĥ:, :Q̂:] = 0.

Solution

(a) Lagrangian for two fields as sum of two Lagrangians for single
fields
2
X 1
(∂µ φi )(∂ µ φi ) − m2i φ2i

L=
2
i=1
assume m1 = m2 .
∂L ∂L
πi = = = φ̇i
∂(∂t φi ) ∂ φ̇i

74
(b) As usual, Hamiltonian density given by
2 h 2
X i X 1h i
H= πi φ̇i − L = (φ̇i )(φ̇i ) + (∇φi )(∇φi ) + m2i φ2i
2
i=1 i=1

(c) Commutators in the usual way: field operators and and their
conjugate momentum operators do not commute, everything else
does:
h i
φ̂i (x, t), π̂j (y, t) = iδij δ 3 (x − y)
h i  
φ̂i (x, t), φ̂j (y, t) = π̂i (x, t), π̂j (y, t) = 0

(d) Employ the usual plane-wave expansion with factors exp[±ik · x]

and operators:

d3 k
Z h i
−ik·x † ik·x
φ̂i (x) = âi (k)e + â i (k)e
(2π)3 (2k0 )
d3 k
Z h i
−ik·x † ik·x
π̂i (x) = −ik â
0 i (k)e + ik â
0 i (k)e
(2π)3 (2k0 )

and therefore
Z h i
0
d3 xe−ik ·x k00 φ̂i (x) + iπ̂i (x)

d3 k
Z
0
= d3 xe−ik ·x
(2π)3 (2k0 )
h i
âi (k)e−ik·x (k00 + k0 ) + â†i (k)eik·x (k00 − k0 )
d3 k
Z h 0
= 3
d 3
x âi (k)e−i(k+k )·x−ik0 t (k00 + k0 )
(2π) (2k0 )
0
i
+â†i (k)ei(k−k )·x+ik0 t (k00 − k0 )
d3 k
Z h
= âi (k)(2π)3 δ 3 (k + k 0 )(k00 + k0 )e−ik0 t
(2π)3 (2k0 )
i
+â†i (k)(2π)3 δ 3 (k − k 0 )(k00 − k0 )eik0 t
1 h 0 0 −ik00 t
i 0
= 0 âi (k )(2k 0 )e + 0 = âi (k 0 )e−ik0 t ,
(2k0 )
p
where we have used that k0 = k 2 + m2 = (−k)2 + m2 = k00
p

and the Fourier transform of the δ function,

Z
d3 xeip·x = (2π)3 δ 3 (p) .

75
 0
Multiplying on both sides with eik0 t and replacing k 0 → k yields
Z h i
âi (k) = d3 xeik·x k0 φ̂i (x) + iπ̂i (x)

Taking the Hermitean conjugate and using that φ̂†i = φ̂i and
π̂i† = π̂i for real fields implies that
Z h i
† 3 −ik·x
âi (k) = d xe k0 φ̂i (x) − iπ̂i (x)

This allows us to directly calculate the commutators, for example

h i
âi (k), â†j (q)
Z Z h i
= 3 ik·x
d xe d3 ye−iq·y k0 φ̂i (x) + iπ̂i (x), q0 φ̂j (y) − iπ̂j (y)
Z Z
= 3 ik·x
d xe d3 ye−iq·y
n h i h io
−ik0 φ̂i (x), π̂j (y) + iq0 π̂i (x), φ̂j (y)
Z Z
= 3 ik·x
d xe d3 ye−iq·y

−ik0 · iδij δ 3 (x − y) + iq0 · (−i)δij δ 3 (y − x)


Z
= d3 xei(k−q)·x {(k0 + q0 )δij } = (2π)3 2k0 δij δ 3 (k − q) ,

where we have used that we discussed equal time communtators,

i.e. x0 = y0 = t. Commutators of two creation or annihilation
operators will vanish, because in the last line the term (k0 + q0 )
will become ±(k0 − q0 ) → 0.
(e) Write the two complex fields as linear combinations of the two
real fields:
1 1
φ = √ (φ1 + iφ2 ) and φ∗ = √ (φ1 − iφ2 ) ,
2 2

therefore8
π = φ̇∗ and π ∗ = φ̇ .
8
Expressing the fields φ1,2 through φ and φ∗ ,
1 i
φ1 = √ (φ + φ∗ ) and φ2 = − √ (φ − φ∗ ) and
2 2
yields the Lagrangian
L = (∂µ φ)(∂ µ φ∗ ) − m2 φ∗ φ
and the conjugate momenta are given, as before, by π = ∂L/∂ φ̇ and π ∗ = ∂L/∂ φ̇∗ .

76
Suitably combining the expansions for φ1 and φ2 yields

d3 k
Z 
â1 (k) + iâ2 (k) −ik·x
φ̂(x) = √ e
(2π)3 (2k0 ) 2
#
↠(k) + i↠(k)
+ 1 √ 2 eik·x
2
d3 k
Z h i
−ik·x † ik·x
= â + (k)e + â − (k)e
(2π)3 (2k0 )
d3 k

â1 (k) − iâ2 (k) −ik·x
Z
φ̂∗ (x) = 3
√ e
(2π) (2k0 ) 2
#
â†1 (k) − iâ†2 (k) ik·x
+ √ e
2
d3 k
Z h i
−ik·x † ik·x
= â − (k)e + â + (k)e ,
(2π)3 (2k0 )

where

â1 (k) ± iâ2 (k) ↠(k) ∓ i↠(k)

â± (k) = √ and ↱ (k) = 1 √ 2
2 2

˙
Using that π̂i = φ̂i and that [φ1 , π2 ] = 0 and similar, the equal-
time commutators read
h i
φ̂(x, t), π̂(y, t)
1 h ˙ ˙
i
= φ̂1 (x, t) + iφ̂2 (x, t), φ̂1 (x, t) − iφ̂2 (x, t)
2
1 nh i h io
= φ̂1 (x, t), π̂1 (y, t) + φ̂2 (x, t), π̂2 (y, t) = iδ 3 (x − y)
h 2 i
φ̂ (x, t), π̂ ∗ (y, t)
∗

1 h ˙ ˙
i
= φ̂1 (x, t) − iφ̂2 (x, t), φ̂1 (x, t) + iφ̂2 (x, t)
2
1 nh i h io
= φ̂1 (x, t), π̂1 (y, t) + φ̂2 (x, t), π̂2 (y, t) = iδ 3 (x − y)
h 2 i
φ̂(x, t), π̂ ∗ (y, t)
1 h ˙ ˙
i
= φ̂1 (x, t) + iφ̂2 (x, t), φ̂1 (x, t) + iφ̂2 (x, t)
2
1 nh i h io
= φ̂1 (x, t), π̂1 (y, t) − φ̂2 (x, t), π̂2 (y, t) = 0
2
and similarly for all commutators of fields with fields and mo-
menta with momenta.

77
 The non-vanishing commutators of the creation and annihilation
operators read
h i 1h i
â+ (k), â†+ (q) = â1 (k) + iâ2 (k), â†1 (k) − iâ†2 (k)
2
1 nh i h io
= â1 (k), â†1 (q) + â2 (k), â†2 (q) = (2k0 )(2π)3 δ 3 (k − q)
2
h i 1h i
â− (k), â†− (q) = â1 (k) − iâ2 (k), â†1 (k) + iâ†2 (k)
2
1 nh i h io
= â1 (k), â†1 (q) + â2 (k), â†2 (q) = (2k0 )(2π)3 δ 3 (k − q)
2

while for example

h i 1h i
â+ (k), â†− (q) = â1 (k) + iâ2 (k), â†1 (k) + iâ†2 (k)
2
1 nh i h io
= â1 (k), â†1 (q) − â2 (k), â†2 (q) = 0
2
and similar for all other commutators.
(f) Commutators of number operators (replacing types 1 and 2 with
±),
d3 k
Z
N̂± = ↠(k)â± (k) ,
(2π)3 (2k0 ) ±
with the annihilation and creation operators:
h i Z d3 k h
†
i
N̂± , â± (q) = â (k)â ± (k), â ± (q)
(2π)3 (2k0 ) ±
d3 k
Z h i
† †
= â ± (k)â± (k)â ± (q) − â ± (q)â ± (k)â ± (k)
(2π)3 (2k0 )
d3 k
Z nh i o
† †
= â± (k), â ± (q) â ± (k)
(2π)3 (2k0 )
d3 k
Z
= − (2π)3 δ 3 (k − q)↱ (k) = −â±
(2π)3 (2k0 )
and
d3 k
h i Z h i
N̂± , ↱ (q) = â†
(k)â ± (k), â†
(q) = ↱ (q)
(2π)3 (2k0 ) ± ±

(†) (†)
As the a± commute with the a± we also have
h i h i
N̂± , â†∓ (q) = N̂± , â∓ (q) = 0 .

Commutator of the number operators:

h i
N̂+ , N̂−

78
 d3 k d3 q
Z
=
(2π)3 (2k0 ) (2π)3 (2q0 )
h i
â†+ (k)â+ (k)â†− (q)â− (q) − â†− (q)â− (q)â†+ (k)â+ (k)
= 0,

because the positive charge â+ and â†+ commute with their neg-
ative charge counterparts, as seen above.
This implies that the commutator of Hamilton or Charge operator
with the number operators also vanish. For example, with the
Hamilton operator from Eq. (172):

h i Z d3 k d3 q
Ĥ, N̂+ = k0
(2π)3 2k0 (2π)3 2q0
 
† † †
â+ (k)â+ (k) + â− (k)â− (k), â+ (q)â+ (q)

d3 k d3 q
Z
= k0
(2π)3 2k0 (2π)3 2q0

â†+ (k)â+ (k)â†+ (q)â+ (q) − â†+ (q)â+ (q)â†+ (k)â+ (k)

† † † †
+ â− (k)â− (k)â+ (q)â+ (q) − â+ (q)â+ (q)â− (k)â− (k)

d3 k d3 q
Z
= k0
(2π)3 2k0 (2π)3 2q0

(2π)3 2q0 δ 3 (k − q)â†+ (k)â+ (q) + â†+ (k)â†+ (q)â+ (k)â+ (q)

− (2π)3 2q0 δ 3 (k − q)â†+ (q)â+ (k) − â†+ (q)â†+ (k)â+ (q)â+ (k)

+0 = 0,

since [â+ , â+ ] = [â†+ , â†+ ] = 0.

The same is also true for the commtator [Q̂, N̂± ] - the only dif-
ference between Ĥ and Q̂ being the exgra factor of energy in the
integration, while, of course, the algebra is identical up to trivial
relative signs. With the Hamilton and Charge operators being
effectively composed of number operators this also proves that
they commute, and, hence, the charge is a conserved quantity of
the theory.

4. Momentum Operator

79
The total four-momentum operator of a real scalar field is given by

d3 k
Z
:P̂ µ:= k µ ↠(k)â(k) .
(2π)3 2k0

(a) Show that P̂ µ can be expressed in terms of the field operator φ̂(x)
and the conjugate momentum operator π̂(x) as
Z
:P̂ := d3 x :π̂(x)∂ µ φ̂(x):
µ

(b) show that

h i
P̂ µ , φ̂(x) = −i∂ µ φ̂(x) .

Solution

(a) Inserting the expansion of the field operator and its conjugate
momentum through plane waves and creation and annihilation
operators we have

d3 k d3 q
Z h  
µ −ik·x †
:P̂ : = d3 x : −ik 0 â(k)e − â (k)eik·x
(2π 3 )2k0 (2π 3 )2q0
 i
iq µ −â(q)e−iq·x + ↠(q)eiq·x :
d3 k d3 q q µ
Z
= d3 x
(2π 3 ) (2π 3 ) 4q0
h
: −â(k)â(q)e−i(k+q)·x + ↠(k)↠(q)ei(k+q)·x
i
+â(k)↠(q)e−i(k−q)·x + ↠(k)â(q)ei(k−q)·x :
d3 q q µ h
Z
= : −â(−q)â(q)e−2iq0 t + ↠(−q)↠(q)e2iq0 t
(2π 3 ) 4q0
i
+â(q)↠(q) + ↠(q)â(q) :
d3 q q µ †
Z
= â (q)â(q)
(2π 3 ) 2q0

The first two terms in the second-to last line vanishes because â(q)
commutes with â(−q), and similarly for the “daggered” operators.
We can therefore replace

q µ e−2iq0 t
Z
d3 q â(−q)â(q)
q0
q µ e−2iq0 t 1 
Z
d3 q

= â(−q)â(q) + â(q)â(−q) ,
q0 2

80
 showing for each component of q µ that this is an integration of an
odd function over an even integration space. The same reasoning
holds also true for the daggered operators. Therefore only the
two last terms survive and we have shown that indeed
d3 k
Z Z
µ µ †
:P̂ : = k â (k)â(k) = d3 x :π̂(x)∂ µ φ̂(x):
(2π)3 2k0

(b) Direct calculation shows that

[P̂ µ , φ̂(x)]
d3 k d3 q
Z h i
µ † −iq·x † iq·x
= k â (k)â(k), â(q)e + â (q)e
(2π)3 2k0 (2π)3 2q0
d3 k d3 q
Z n h i
µ −iq·x †
= k e â (k), â(q) â(k)
(2π)3 2k0 (2π)3 2q0
h io
+ eiq·x ↠(k) â(k), ↠(q)
d3 k
Z n o
µ −ik·x ik·x †
= k −e â(k) + e â (k)
(2π)3 2k0
d3 k
Z n o
µ −ik·x ik·x †
= (−i∂ ) e â(k) + e â (k) = −i∂ µ φ̂(x) ,
(2π)3 2k0
as demanded.

5. Causality and anti-commutators (real scalars)

Consider real scalar fields and define

∆1 (x − y) = ∆+ (x − y) + ∆− (x − y)

(a) show that ∆1 is equal to the vacuum expectation value of the

anti-commutator of the field operators φ̂(x) and φ̂(y)

∆1 (x − y) = h0|{φ̂(x), φ̂(y)}|0i = h0|[φ̂(x)φ̂(y) + φ̂(y)φ̂(x)]|0i

(b) show that ∆1 (x − y) does not vanish outside the light-cone, i.e.
that ∆1 (x − y) 6= 0 for (x − y)2 < 0.

Solution

(a) Remember definitions for ∆± , expansion of fields in terms of cre-

ation and annihilation operators and â(k)|0i = 0, then

∆+ (x − y) = h0|φ̂(x)φ̂(y)|0i
d3 k d3 k 0
Z D   E
−ikx+ik0 y † 0 
= e 0 â(k)â (k ) 0

(2π)3 (2k0 ) (2π)3 (2k00 )


81
 d3 k d3 k 0
Z D h i E
−ikx+ik0 y † 0 
= e 0 â(k), â (k ) 0

(2π)3 (2k0 ) (2π)3 (2k00 )


d3 k d3 k 0
Z
0

e−ikx+ik y 0 (2π)3 (2k0 )δ 3 (k − k 0 ) 0
 
= 3 3 0
(2π) (2k0 ) (2π) (2k0 )
d3 k
Z
= e−ik(x−y)
(2π)3 (2k0 )

In a similar way, we can write

d3 k
Z
∆− (x − y) = eik(x−y)
(2π)3 (2k0 )
d3 k d3 k 0
Z
−ik(y−x) (2π)3 (2k0 )δ 3 (k − k 0 ) 0

  
= e 0
(2π)3 (2k0 ) (2π)3 (2k00 )
d3 k d3 k 0
Z D h i E
−ik0 y+ikx 0 †
= e 0 â(k ), â (k) 0
 
(2π)3 (2k0 ) (2π)3 (2k00 )


= h0|φ̂(y)φ̂(x)|0i

Therefore, as demanded
D h i E
∆(x−y) = ∆+ (x−y)+∆− (x−y) = 0  φ̂(x)φ̂(y) + φ̂(y)φ̂(x)  0
 

(b)

d3 k
Z  
−ik(x−y) ik(x−y)
∆1 (x − y) = e + e
(2π)3 (2k0 )
d3 k
Z
x0 →y0
 
−ik(x−y) ik(x−y)
−→ e + e
(2π)3 (2k0 )
d3 k
Z
x→y
−→ 2 →∞
(2π)3 (2k0 )

6. Commutators for free reals scalar fields

Calculate the equal time commutators for
h i
(a) P̂ µ , φ̂(x) , where the momentum operator is given by

d3 k
Z Z
P̂ µ = k µ ↠(k)â(k) = d3 xπ̂(x)∂ µ φ̂(x) ;
(2π 3 )2k0
h i
(b) Ĥ, ↠(k)â(q) ;

82
Solution

(a) Through field operators

h i Z h i
µ
P̂ , φ̂(x) = d3 y π̂(y)∂yµ φ̂(y), φ̂(x)
x0 =y0 =t
Z nh i h io
= d3 y π̂(y), φ̂(x) ∂ µ φ̂(y) + π̂(y) ∂ µ φ̂(y), φ̂(x)
x0 =y0
Z n o
= d3 y −iδ(x − y)∂yµ φ̂(y) + π̂(y) · 0 = −i∂ µ φ̂(x) ,

and through the expansion in creation and annihilation operators

h i
P̂ µ , φ̂(x)
d3 k d3 q
Z h i
µ † −iq·x † iq·x
= k â (k)â(k), â(q)e + â (q)e
(2π 3 )2k0 (2π 3 )2q0
d3 k d3 q
Z nh i
†
= k µ
â (k)â(k), â(q) e−iq·x
(2π 3 )2k0 (2π 3 )2q0
h i o
+ ↠(k)â(k), ↠(q) eiq·x
d3 k d3 q
Z nh i
†
= k µ
â (k), â(q) â(k)e−iq·x
(2π 3 )2k0 (2π 3 )2q0
h i o
+↠(k) â(k), ↠(q) eiq·x
d3 k d3 q
Z
k µ −2q0 (2π 3 )δ 3 (k − q) â(k)e−iq·x

= 3 3
(2π )2k0 (2π )2q0
o
+↠(k) 2q0 (2π 3 )δ 3 (k − q)eiq·x
d3 k
Z n o
µ −ik·x † ik·x
= k −â(k)e + â (k)e
(2π 3 )2k0
d3 k n
Z o
−ik·x †
= −i∂ µ â(k)e + â (k)e ik·x
= −∂ µ φ̂(x) .
(2π 3 )2k0
(b) Expand the Hamilton operator in creation and annihilation op-
erators and use their commutation relations
d3 p
h i Z h i
Ĥ, ↠(k)â(q) = p0 â †
(p)â(p), â†
(k)â(q)
(2π)3 2p0
d3 p h †
Z
1 † † †
i
= â (p)â(p)â (k)â(q) − â (k)â(q)â (p)â(p)
2 (2π)3
d3 p h †
Z
1
= â (k)↠(p)â(q)â(p) + 2k0 (2π 3 )δ 3 (k − p)↠(p)â(q)
2 (2π)3
i
−↠(k)↠(p)â(q)â(p) − 2q0 (2π 3 )δ 3 (q − p)↠(p)â(q)
2(k0 − q0 ) †
= â (k)â(q) = (k0 − q0 )↠(k)â(q) .
2

83
7. ∗ Properties of the Charge Operator In the following, consider a
free complex scalar field φ.
(a) Show that the (normal-ordered) charge operator is given by
Z h i
:Q̂: = d3 x :φ̂∗ (x)π̂ ∗ (x) − φ̂(x)π̂(x):

d3 k h †
Z i
†
= â (k)â(k) − b̂ (k)b̂(k)
(2π)3 2k0

(b) Show that [:Q̂:, :P̂ µ:] = 0.

Solution

(a) Using that π̂ ∗ = ∂t φ̂ and π̂ = ∂t φ̂∗ we arrive at

Z
i h i
:Q̂: = d3 x :φ̂∗ (x)π̂ ∗ (x) − φ̂(x)π̂(x):
2
d3 k d3 q
Z
i
= d3 x (−iq0 )
2 (2π)3 2k0 (2π)3 2q0
h ih i
: ↠(k)eik·x + b̂(k)e−ik·x â(q)e−iq·x − b̂† (q)eiq·x
h ih i
−ik·x † ik·x † iq·x −iq·x
− â(k)e + b̂ (k)e −â (q)e + b̂(q)e :

d3 k d3 q
Z
1
= d3 x q0
2 (2π)3 2k0 (2π)3 2q0
 h i
: e+i(k−q)·x ↠(k)â(q) − b̂† (k)b̂(q)
h i
− e−i(k−q)·x b̂(k)b̂† (q) − â(k)↠(q)
h i
− e+i(k+q)·x ↠(k)b̂† (q) − b̂† (k)↠(q)
h i
−i(k+q)·x
+e b̂(k)â(q) − â(k)b̂(q) :

d3 k d3 q
Z
1
= q0
2 (2π) 2k0 (2π)3 2q0
3
 h i
: e+i(k0 −q0 )x0 δ 3 (k − q) ↠(k)â(q) − b̂† (k)b̂(q)
h i
− e−i(k0 −q0 )x0 δ 3 (k − q) b̂(k)b̂† (q) − â(k)↠(q)
h i
− e+i(k0 +q0 )x0 δ 3 (k + q) ↠(k)b̂† (q) − b̂† (k)↠(q)
h i
+ e−i(k0 +q0 )x0 δ 3 (k + q) b̂(k)â(q) − â(k)b̂(q) :

84
 d3 k h †
Z
1 †
i
= â (k)â(k) − b̂ (k)b̂(k) .
2 (2π)3 2k0

(b) To calculate the commutator let us first take a look at one typical
term, namely

[φ̂(x)π̂(x), P̂ µ ] = φ̂(x)π̂(x)P̂ µ − P̂ µ φ̂(x)π̂(x)

= φ̂(x)P̂ µ π̂(x) − P̂ µ φ̂(x)π̂(x) + φ̂(x)π̂(x)P̂ µ − φ̂(x)P̂ µ π̂(x)
= [φ̂(x), P̂ µ ]π̂(x) + φ̂(x)[π̂(x), P̂ µ ] .

Using the result from one of the previous problems that

[P̂ µ , φ̂(x)] = −i∂ µ φ̂(x)

and similarly (because the derivatives commute) that

[P̂ µ , π̂(x)] = [P̂ µ , ∂t φ̂∗ (x)] = −i∂t ∂ µ φ̂∗ (x) = −i∂ µ π̂(x)

we see that
h i Z h i h i
µ
:Q:, :P̂ : = − d x 3
P̂ µ , φ̂∗ (x) π̂ ∗ (x) + φ̂∗ (x) P̂ µ , π̂ ∗ (x)
h i h i
µ µ
− P̂ , φ̂(x) π̂(x) + φ̂(x) P̂ , π̂(x)
Z 
= i d x [∂ µ φ̂∗ (x)]π̂ ∗ (x) + φ̂∗ (x)[∂ µ π̂ ∗ (x)]
3


µ µ
− [∂ φ̂(x)]π̂(x) − φ̂(x)[∂ π̂(x)]
Z  h i
3 µ ∗ ∗
= i d x ∂ φ̂ (x)π̂ (x) − φ̂(x)π̂(x) = 0

8. ∗ Parity of a Scalar Field

The parity operator P̂ for a real scalar field is given by

d3 k h †
 Z i
iπ †
P̂ = exp − â (k)â(k) − ηP â (k)â(−k) ,
2 (2π)3 2k0
where the phase ηP = ±1 is the intrinsic parity of the field. Fields
with ηP = 1 are scalars and those with ηP = −1 are pseudoscalars.
Prove that [P̂, Ĥ] = 0.

Solution

To prove this, we need to realise that any function of an operator

Ô commutes with another operator X̂, if the operators commute, i.e.

85
[f (Ô), X̂] = 0 if [Ô, X̂] = 0, because functions of operators are defined
through their series expansion. This means that we have to show that

d3 k
Z  
† †
0 = â (k)â(k) − ηP â (k)â(−k) ,
(2π)3 2k0
d3 q
Z 
†
q0 â (q)â(q)
(2π)3 2q0
d3 k d3 q
Z Z h i
† †
= q 0 â (k)â(k), â (q)â(q)
(2π)3 2k0 (2π)3 2q0
h i
† †
− ηP â (k)â(−k), â (q)â(q)

d3 k d3 q
Z Z 
= q0
(2π)3 2k0 (2π)3 2q0
(2π 3 )2q0 δ 3 (k − q)↠(k)â(q) − ↠(k)↠(q)â(k)â(q)
−(2π 3 )2q0 δ 3 (k − q)↠(q)â(k) + ↠(q)↠(k)â(q)â(k)
h
− ηP (2π 3 )2q0 δ 3 (k + q)↠(k)â(q) − ↠(k)↠(q)â(−k)â(q)

− (2π 3 )2q0 δ 3 (k − q)↠(q)â(−k) + ↠(q)↠(k)â(q)â(−k)

d3 k d3 q
Z Z  
† †
= −ηP q0 â (k)â(−k) − â (k)â(−k) = 0
(2π)3 2k0 (2π)3 2q0

and therefore the parity operator commutes with the Hamiltonian.

This implies that parity is a conserved quantity for the free scalar
field.

86
5 Fermions
In this section we will get acquainted with the Dirac equation, which intro-
duces not only fermions, but also provides insight into anti-particles.
The Dirac equation emerges through linearisation of the Klein-Gordon equa-
tion, after realising that its quadratic form yields negative energy solutions.
Such a linearised form, however, only satisfies the original energy-momentum
relation – essentially the kernel of the free Klein-Gordon equation – if the
fields have an even number of components, at least two. This proves to be a
blessing, as it allows us to describe spin-1/2 particles, and the corresponding
fields are dubbed “spinors”. Insisting on maintaining that the spinors sat-
isfy the Klein-Gordon equation for massive particles leads to spinors with
four components - two more than necessary for spin-1/2 particles. These
additional degrees of freedom are identified with negative energy solutions
and interpreted as anti-particles. As before, in the case of the scalar fields,
this implies that the energy spectrum of the theory is unbounded from be-
low. Consequently the vacuum is not empty, and i fact it contains short-
lived quantum fluctuations of particle+anti-particle with opposite energy,
momentum and spin.
The Dirac equation has been covered ubiquitously in the literature. Keeping
in mind that we use a somewhat different (and in my opinion, more mod-
ern) normalisation, it would maybe be a good idea to also take a look at
Sections 4.1 and 4.2 of Hatfield [3] or Sections 3.1-3.4 and 3.6 in Peskin &
Schroeder [1], the latter section more of some extended reading. It is also
worthwhile to check out Chapter 2 of Itzykson & Zuber [?], if you can find
it somewhere.

5.1 The Dirac Equation

Short-comings of the Klein-Gordon Lagrangian Consider, again, the
Klein-Gordon equations of motion, Eq. (91) in Sec. 3.2,
 2 
∂ 2 2 µ 2


− ∇ + m φ(x) = ∂µ ∂ + m φ(x) = 0 .
∂t2

Fourier-transforming it into

(E 2 − p2 − m2 )φ = 0 −→ E 2 = p2 + m2 (182)

we realise that, due to its quadratic form, nothing prevents us from con-
structing solutions with negative energies. Assuming plane wave solutions
for the fields, φ(x) ∼ exp(ikx) the charge or probability density for the
complex scalar field is given by

ρ = j0 = (∂t φ∗ )φ − φ∗ (∂t φ) = −2ik0 (183)

87
which can be translated into a real, unit-free quantity, that is more ap-
propriate for a probability density. This is achieved through a suitable
normalisation, for example
i µ
j µ → j̃ µ = j , (184)
2m
such that ρ = k0 /m. For negative-energy solutions, though, this would
result in negative probability densities, which are extremely hard to inter-
pret. Ultimately, the appearance of these solutions mean that the energy
spectrum of the theory is not bound from below and there is no lowest en-
ergy ground-state. In other words, there is nothing that prevents us from
producing more and more particles, by pairing positive and negative energy
solutions – clearly an unacceptable problem for the interpretation of the the-
ory. Ultimately this shows that it is impossible to produce a single-particle
theory when imposing Lorentz-invariance as a construction paradigm.
Of course, we know by now that this issue can be completely circumnav-
igated by identifying the negative energy-solutions as anti-particles, parti-
cles with positive energy but opposite charge that propagate backwards in
time. However, when Dirac introduced his famous equation in 1928 this
anti-particles were not discovered yet, and it was in fact his work that in-
troduced anti-particles as a meaningful theoretical concept that emerges
naturally when combining Quantum Mechanics and Special Relativity into
Quantum Field Theory.

Linearising the Klein-Gordon Equation Dirac’s aim was to construct

a linearised version of the Klein-Gordon equation such that the resulting
E.o.M. are linear in ∂t , and being Lorentz-invariant exhibit solutions that
still satisfy the original equation. Choosing an ansatz for the field ψ, that
is first order in ∂t and first order in ∇
∂ψ(x, t)
i = −iα · ∇ψ(x, t) + βmψ(x, t) , (185)
∂t
it becomes obvious that αi and β must be matrices, and that ψ has at
least two components. The latter property in fact was seen as a nice bonus,
because they could be identified with the two spin states (spin up and spin
down) of the electrons that Dirac wanted to describe. This identification of
the components of the field ψ with spin states has led to the name for ψ(x):
spinor or spinor field.

Properties of the αi and β matrices To guarantee that the equation

above, Eq. (185), reduces to the Klein-Gordon E.o.M. when squaring the
kernel,
 2 
∂ !
− ∇ 2
+ m2
= [i∂t + iα · ∇ − βm]2 (186)
∂t2

88
αi and β must satisfy the following relations

{αi , αj } = αi αj + αj αi = 2δij
{αi , β} = 0
β 2 = αi2 = 1
Tr(αi ) = Tr(β) = 0 . (187)

This implies that the eigenvalues of αi and β are ±1, and the combination of
them being traceless and having these eigenvalues suggests that they must
be of an even dimension, i.e. dim(αi , β) = 2, 4, . . . . Focusing on the case
of lowest dimension, 2×2 matrices, we can see straightaway that the αi can
be identified with the Pauli matrices, αi = σi , where
     
0 1 0 −i 1 0
σ1 = , σ2 = , and σ3 = . (188)
1 0 i 0 0 −1

This however won’t work for massive theories where m 6= 0: There is just no
fourth candidate matrix for β that satisfies all the properties of Eq. (187).
This has two implications: First of all, for massless theories, we could stick
with two-component fields ψ, also known as Weyl spinors. And secondly,
for massive theories like the ones we’re going to pursue, we must use four-
component fields – the Dirac spinors – and have four-dimensional αi and β
matrices:
   
0 σi 1 0
αi = and β = . (189)
σi 0 0 −1

Here – and later in γ0 – the 1 denote 2 × 2 identity matrices.

γ Matrices and Their Properties For practical purposes, the α and

β matrices proved a bit cumbersome, and they are usually replaced by the
γ-matrices, defined by
   
0 1 0 i 0 σi
γ =β= and γ = βαi = . (190)
0 −1 −σi 0

Direct calculation shows that they enjoy the following anti-commutator re-
lation

{γ µ , γ ν } = γ µ γ ν + γ ν γ µ = 2g µν . (191)

2
In addition, γ 0 = γ 0† is Hermitean with γ 0 = 1, while the γ i = −γ i† are
anti-Hermitean, with (γ i )2 = −1.

89
Dirac Equation Multiplying the Dirac equation, expressed through the
α and β matrices, Eq. (185), from the left with γ 0 we arrive at
(iγ µ ∂µ − m1)ηξ ψξ = (i∂/ − m)ψ = 0 . (192)
In the equation above, Eq. (192) the components of the Dirac equation in
“spinor space” have been made explicit, indicated by the indices η and ξ. It
is important to stress that this exhibits the fact that there are two spaces
in the Dirac equation, namely the “normal” Minkowski space with index µ,
incorporating the external Lorentz symmetry of space-time, and this spinor
space. The γ matrices and the spinor ψ have multiple components in this
space, and the mass term is diagonal in this space, indicated by the 1-
symbol. As before, the Lorentz indices µ etc. run from 0 to 3, while the
Dirac or spinor indices run from 1 to 4.

Dirac Equation for ψ † The nature of the equation above suggest that
the Hermitean conjugate spinor ψ † represents a second, independent field,
similar to φ∗ and φ. Straightforward Hermitean conjugation of Eq. (185)
results in
∂ψ † (x, t)
−i = i∇ψ † (x, t) · α† + mψ † (x, t)β † , (193)
∂t
and multiplying from the right with β † = β = γ 0 yields
←−
−iψ † (x, t)∂µ γ †µ = mψ † (x, t) . (194)
2
Using γ 0 = 1 and γ † = (βα)† = αβ = β(βα)β = γ 0 γγ 0 while defining the
“barred” spinor ψ̄ = ψ † γ 0 allows to find the E.o.M. for the barred spinor as
←−
ψ̄(i ∂/ + m) = 0 . (195)

Lagrangian It is easy to check that the two E.o.M. for the spinors ψ and
ψ † can be obtained from the free Dirac Lagrangian
←→ 
L = ψ̄(x) i ∂/ − m ψ(x) , (196)

where
←→ 1
a ∂ b = [a(∂b) − (∂a)b] . (197)
2
The E.o.M. for ψ (ψ̄) are obtained,as usual, by varying the Lagrangian with
respect to ψ̄ (ψ):
∂L ∂L 1 →− 
− ∂µ = − mψ + [i∂/ψ − ∂µ (−iγ µ ψ)] = i ∂/ − m ψ = 0
∂ ψ̄ ∂(∂µ ψ̄) 2
∂L ∂L 1h ← − i ←− 
− ∂µ = − mψ̄ − ψ̄(i ∂/ ) + ∂µ (iψ̄γ µ ) = −ψ̄ i ∂/ + m = 0 .
∂ψ ∂(∂µ ψ) 2
(198)

90
Conserved Current It is relatively straightforward to construct a con-
served current from the two E.o.M. Eqs (192) and (195): multiply the former
from the left with ψ̄ and the latter from the right with ψ and add. This
results in
→
− ←− →
− ← −
0 = ψ̄ · (i ∂/ − m)ψ + ψ̄ · (i ∂/ + m) · ψ = iψ̄( ∂/ + ∂/ )ψ (199)

and we arrive at the conserved current

∂µ j µ = ∂µ iψ̄γ µ ψ .
 
(200)

Solutions to the Dirac E.o.M.: Spinors at Rest To construct solu-

tions for the Dirac equation, it is important to keep in mind that the ψ and
ψ̄ are objects with four components9 . Let us for the moment describe the ψ
as a product of advanced and retarded plane wave factors and polarisation
eigenstates u(p) and v(p),

d3 p  −ip·x
Z
ip·x

ψη (x) = e uη (p) + e v η (p) , (201)
(2π)3
where we have made explicit the spinor index η. This expansion moves the
spinor index to the u and v spinors, i.e. they are objects with four entries,
and the Dirac matrices act on these indices10 . To construct them, it is
sufficient to realise that the E.o.M. become a system of linear equations for
the eigenstates u(p) and v(p). Let us first solve this equation for a particle
at rest, p = 0, p0 = E = m, leading to

(Eγ 0 − m)u(0) = m(γ 0 − 1)u(0) = 0 and, similarly, (γ 0 + 1)v(0) = 0 ,

(202)

Inserting the (diagonal) form of

 
1 0 0 0
0 1 0 0 
γ0 = 

 (203)
 0 0 −1 0 
0 0 0 −1

implies that the third and fourth component of u and the first and second
component of v must be zero. Both u and v therefore have two independent
9
But, although they look like vectors because of the four-components, they differ from
four-vectors in how they behave under Lorentz transformations. Simply put: spinor index
6= Lorentz index)
10
Positive and negative energy solutions ψ± are of course related to the wave factors
such that

ψ+ = e−ip·x u(p) and ψ− = eip·x v(p) ,

and we will recycle them later when quantising the Dirac fields.

91
solutions each, and the corresponding eigenstates can be readily identified
with the two spin states: u(1/2) describe positive-energy particles with spin
up/down, and v (1/2) decibel negative-energy particles with spin up/down.
Choosing normalised “eigenvectors” then results in
   
1 0
0  , u(2) (0) =  1  ,
u(1) (0) = 
   
 0   0 
0 0
   
0 0
0   0 
v (1) (0) =  (2)
   
 1  , v (0) =  0  . (204)
0 1

Solutions to the Dirac E.o.M.: General Momenta To obtain so-

lutions for general momenta, we use the fact that suitable multiplication
of the kernels of the E.o.M. for u and v with terms (p/ ± m) encodes the
energy-momentum relation for a massive particle,
/ − m)(p/ + m) = p2 − m2 = 0 .
(p (205)
This means that, including normalisation factors η(p), the transformed eigen-
states
u(i) (p) = η i (p/ + m)u(i) (0)
v (i) (p) = η i (−p/ + m)v (i) (0) (206)
/ − m)u = 0 and (p/ + m)v = 0. Introducing
will satisfy the E.o.M. (p p± =
px ± ipy of the momentum components the spinors and using
 
E±m 0 −pz −px + ipy
0 E ± m −px − ipy pz
/ ± m = pµ γ µ ± m = 
 
p 
 pz px − ipy −E ± m 0 
px + ipy −pz 0 −E ± m
(207)
we arrive at
1 0
   
 0   1 
u(1) (p) = η   , u(2) (p) = η   ,
   
pz p−
E+m E+m
   
p+ −pz
E+m E+m
 pz   p− 
E+m E+m
 p+   −pz 
v (1) (p) = η  E+m
 , v (2) (p) = η  E+m
 , (208)
   
 1   0 
0 1

92
 q
where the energy E = p2 + m2 > 0 and the normalisation is given by
√
η= E + m. (209)

Note that we have normalised the spinors such that, apart from the norm η
the first component of the spinors equals 1.
What is left to do now is to explictly check that the spinors indeed satisfy
their equations of motion, i.e. that (p/ − m)u(1,2) (p) and (p/ + m)v (1,2) (p)
vanish. For example, for u(1) and v (1) we find

/ − m)u(1) (p)
(p
E−m −pz −p− 1
  
0
 0 E−m −p+ pz  0 
=
  
pz
−(E + m)
 
 pz p− 0  E+m

p+
p+ −pz 0 −(E + m) E+m
p2
 
E − m − E+m
 
 −p+ pz +p+ pz 
E+m
=  = 0;
 
 −pz + (E+m)pz 
 E+m 
(E+m)p+
−p+ + E+m

/ + m)v (1) (p)

(p
pz
−pz −p−
  
E+m 0 E+m
p+
 0 E+m −p+ pz 
E+m

=
  
−(E − m)
 
 pz p− 0  1 
p+ −pz 0 −(E − m) 0
 
− (E+m)p z
E+m + pz
− (E+m)p
 
+

E+m + p+ 
=  = 0. (210)
 
p2
E+m − (E − m)
 
 
p+ pz −p+ pz
E+m

Similar calculations for u(2) and v (2) prove that the spinors indeed satisfy
the equations of motion.

Spinor Products in Components The normalisation has been chosen

such that the spinors form a “nearly” ortho-normal basis,

ū(i) (p)u(j) (p) = 2mδij = −v̄ (i) (p)v (j) (p) . (211)

93
A simple calculation exemplifies how to calculate such spinor products. For
example for i = j = 1 we find

p2z
 
(i) (j) (i)† 0 (j) 2 p+ p−
ū u = u γ u = η 1 + 0 − −
(E + m)2 (E + m)2
E 2 + 2Em + m2 − p2 2m(E + m) 2m
= η2 2
= η2 2
= η2 , (212)
(E + m) (E + m) E+m

and plugging in our chosen normalisation leads to the anticipated product

of Eq. (211) A similar calculation for the “daggered” instead of the “barred”
spinors, i.e. ignoring the γ 0 yields

† E 2 + 2Em + m2 + p2 2E(E + m) 2E
u(i) u(j) = η 2 2
= η2 2
= η2 = 2E .
(E + m) (E + m) E+m
(213)

Therefore,
u(i)† (p)u(j) (p) = v (i)† (p)v (j) (p) = 2p0 δij

v̄ (i) (p)u(j) (p) = ū(i) (p)v (j) (p) = 0 . (214)

Completeness Relations Let us now reverse the order of multiplication

and instead of calculating scalar products of a “row” spinor times a “column”
spinor, ūu, let us calculate the product of a “column” spinor times a “row”
spinor, uū. This leads to the completeness relations

2 2
(i) (i)
X X
u(i)
α ūβ = (p
/ + m)αβ , vα(i) v̄β = (p/ − m)αβ . (215)
i=1 i=1

Using Eq. (207) and directly calculate the spinor products, i.e. the terms
uū we see that this holds in fact true. It is important to stress that the
product of “column vector” and “row vector” is not a scalar product but
generates a matrix.

5.2 Second Quantisation

Some Interpretations Before second quantising Dirac theory, it is worth
to first analyse and interpret the structure of the solutions obtained above.
As before for the case of scalar fields we have plane waves moving in the
“wrong direction” - the states that come with the v-spinors. They can be
interpreted either as states of negative energy moving forward in time or
as states of positive energy moving backwards in time. In any case, they
describe anti-particles. Of course, as before, their existence indicates that
the energy states of the theory are not bound from below, so there is a

94
priori no well-defined ground state. Dirac circumnavigated this problem
by demanding that the negative energy solutions are all fill, and that the
v-states are “holes” in this otherwise full “sea” of negative energy solutions.
This obviously abandons any notion of the resulting Quantum Field The-
ory describing just one particle – which is possible in Quantum Mechanics.
Adding Special Relativity to the mix implies that the resulting Quantum
Field Theory indeed can only be realised as a multi-particle theory. It is
then not surprising that the vacuum is not “empty”; instead it can have
short-time quantum fluctuations of particle+anti-particle (hole), with op-
posite energy, momentum, and spin such that the overall quantum numbers
(all 0) are conserved. We will now move on to quantise this theory.

Lagrangian and Conjugate Momenta Derivation of the Lagrange den-

sity of Eq. (196), L = ψ̄(x)(i∂/ − m)ψ(x), with respect to the time-derivative
of the two independent spinor fields ψ and ψ † yields
i
π = ∂L/∂ ψ̇ = ψ̄iγ 0 = ψ †
2
† † i 0 0 i
π = ∂L/∂ ψ̇ = − γ γ ψ = − ψ . (216)
2 2
The Hamiltonian density then reads
i † 
H = π ψ̇ + π † ψ̇ † − L = ψ (∂0 ψ) − ψ(∂0 ψ † ) − L
 ← 2
→ ←→ ←
→   ←
→ 
= ψ̄ iγ0 ∂0 − iγ0 ∂0 + iγ · ∇ + m ψ = ψ̄ iγ · ∇ + m ψ (217)

It is worth noting here that our conjugate momenta differ from the usual
form in textbooks by a factor of 1/2, stemming from our vey literal inter-
pretation of the derivative of Eq. (197) in the Lagrangian, Eq. (196).

Anti-Commutators Quantisation is achieved by promoting fields, mo-

menta etc. to field operators and by demanding suitable commutation rela-
tions for them. However, we know that spin-1/2 particles are fermions so we
need to encapsulate Fermi-statistics into the quantisation condition. This
necessitates to replace the equal-time commutators of fields and momenta
with equal-time anti-commutators. Using the relationship between fields
and momenta from Eq. (216) they therefore read
n o in o
ψ̂α (t, x), π̂β† (t, y) = ψ̂α (t, x), ψ̂β† (t, y) = iδαβ δ 3 (x − y)
n o 2n o
ψ̂α (t, x), ψ̂β (t, y) = ψ̂α† (t, x), ψ̂β† (t, y) = 0 , (218)

where the anti-commutator of two operators is defined by

n o
Â, B̂ = ÂB̂ + B̂ Â , (219)

95
and where we used that π̂β = ψ̂β† .

Creation and Annihilation Operators As before, we expand the fields

in plane waves multiplied with creation and annihilation operators. As we
already have such plane waves for the “classical” fields, multiplied with
the eigen-spinors u and v, we merely need to add one creation/annihilation
operator for each such state and arrive at

2
d3 p X h −ip·x
Z i
(i) ip·x ˆ† (i)
ψ(t, x) = e b̂i (p)u (p) + e d i (p)v (p)
(2π)3 2p0
i=1
2 h
d3 p
Z i
e−ip·x dˆi (p)v̄ (i) (p) + eip·x b̂†i (p)ū(i) (p) γ 0
X
ψ † (t, x) =
(2π)3 2p0
i=1

(220)

With the following anti-commutation relations of the creation and annihila-

tion operators,
n o n o
b̂α (p), b̂†β (q) = dˆα (p), dˆ†β (q) = 2p0 (2π)3 δ 3 (p − q)δαβ , (221)

and all others vanishing, the anti-commutators of Eq. (218) are fulfilled.
For example:
n o
ψ̂α (t, x), ψ̂β† (t, y)
d3 p d3 q
Z   n o
−ip·x−iq·y (β) 0 (α) ˆβ (q)
= e v̄ (q)γ u (p) b̂α (p), d
(2π)3 2p0 (2π)3 2q0
 n o
+e−ip·x+iq·y ū(β) (q)γ 0 u(α) (p) b̂α (p), b̂†β (q)
 n o
+e+ip·x−iq·y v̄ (β) (q)γ 0 v (α) (p) dˆ†α (p), dˆβ (q)
 n o
+e +ip·x+iq·y (β)
ū 0 (α)
(q)γ v (p) dˆ†α (p), b̂†β (q)

d3 p d3 q
Z 
3 3
= 2p0 δαβ (2π) δ (p − q)
(2π)3 2p0 (2π)3 2q0
   
−ip·x+iq·y (β)† (α) +ip·x−iq·y (β)† (α)
e u (q)u (p) + e v (q)v (p)

96
 d3 p
Z  
−ip·(x−y) (β)† (α) +ip·(x−y) (β)† (α)
= δαβ e u (p)u (p) + e v (p)v (p)
(2π)3 2p0
d3 p
Z  
−ip0 ·(t−t)+ip·(x−y) +ip0 ·(t−t)−ip·(x−y)
= δαβ 2p0 δαβ e +e
(2π)3 2p0
d3 p
Z h i
−ip·(x−y) +ip·(x−y)
= δαβ e + e = 2δαβ δ 3 (x − y) , (222)
(2π)3
in agreement with Eq. (218). We realise that due to the equal times, the
exponentials of the time differences vanish; in addition, because α and β
are external parameters, we cannot use Einstein’s convention of summing
over repeated indices, since this would eliminate these parameters and the
right-hand side of the anti-commutator would not depend on them. Simply
put, the α and β are not indices in some space but label the spin-states of
the fermions and cannot be summed over. Finally, we used that δ 3 (x − y) =
δ 3 (y − x).

States To construct states with one and more particle states, we first
realise that
ˆ b̂†1,2 (p)/b̂1,2 (p) creates/annihilates positive-energy electrons with spin
up/down and momentum p;

ˆ dˆ1,2 (p)/dˆ†1,2 (p) creates/annihilates negative-energy electrons – positrons

with – spin up/down and momentum p.
For example, a one-electron (positron) state with positive (negative) energy,
spin-up (down) and momentum p is created by

+, p, ↑ = b† (p)|0i
 
1
−, p, ↓ = d† (p)|0i .
 
2 (223)

While this looks straightforward, things become more interesting when con-
sidering two-electron states, both with positive energy, momentum p, and
one spin up and one spin down:
+, p, ↑; +, p, ↓ = b† (p)b† (p) |0i = −b† (p)b† (p) |0i ,
 
1 2 2 1 (224)

where the sign is a reflection of the quantisation through anti-commutators.

But if both electrons populate the same space in energy, momentum, and
spin, for example
+, p, ↑; +, p, ↑ = b† (p)b† (p) |0i = −b† (p)b† (p) |0i = 0 ,
 
1 1 1 1 (225)

i.e. such states cannot be produced. In fact double application of fermionic

creation operators with identical momenta, energies and spins will annihilate
any state.

97
Hamilton Operator To promote the Hamilton density of Eq. (217) to an
operator it is sufficient to replace the fields with field operators. Plugging in
the expansion in terms of creation and annihilation operators, using ūγ 0 =
u† and v̄γ 0 = v † , and integrating over space, we find
2 
d3 p d3 q
Z X
3
Ĥ = d x
(2π)3 2p0 (2π)3 2q0
i,j=1
  ←
→ 
e−ip·x dˆi (p) v̄ (i) (p) + eip·x b̂†i (p) ū(i) (p) iγ · ∇ + m
 
× −iq·x
e (j)
b̂j (q)u (q) + eiq·x dˆ†j (q)v (j) (q)
2 
d3 p d3 q
Z X
3
= d x
(2π)3 2p0 (2π)3 2q0
i,j=1
  
1
 
−i(p+q)·x ˆ (i) (j)
e di (p)b̂j (q) v̄ (p) γ · (−p + q) + m u (q)
2
  
1
 
−i(p−q)·x ˆ ˆ† (i) (j)
+ e di (p)dj (q) v̄ (p) γ · (−p − q) + m v (q)
2
  
1
 
+ e+i(p−q)·x b̂†i (p)b̂j (q) ū(i) (p) γ · (+p + q) + m u(j) (q)
2
  
1
 
† †
+ e +i(p+q)·x
b̂i (p)dˆj (q) ū (p)
(i) (j)
γ · (+p − q) + m v (q)
2
2 
d3 q
Z X
=
(2π)3 4q02 i,j=1
 h i
dˆi (−q)b̂j (q) v̄ (i) (−q) +γ · q + m u(j) (q) e−2iq0 x0

 h i
dˆi (q)dˆ†j (q) v̄ (i) (q) −γ · q + m v (j) (q)


+
 h i
b̂†i (q)b̂j (q) ū(i) (q) +γ · q + m u(j) (q)


+
 h i 
b̂†i (−q)dˆ†j (q) ū(i) (−q) −γ · q + m v (j) (q) e+2iq0 x0 ,


+
(226)

where we the x-integration over space resulted in a δ-function, δ 3 (p − q),

which in turn enabled the integration over p. Using the E.o.M. for the u
and v spinors,

(q/ − m)u(q) = 0 −→ q0 γ0 u(q) = (q · γ + m)u(q)

/ + m)v(p) = 0 −→ q0 γ0 v(q) = (q · γ − m)v(q)
(p (227)

98
and ūγ0 = u† and v̄γ0 = v † ,
2 n
d3 q
Z
1 X h i
ˆi (−q)b̂j (q) v (i)† (−q)u(j) (q) e−2iq0 x0
Ĥ = d
2 (2π)3 2q0
i,j=1
 h i  h i
− dˆi (q)dˆ†j (q) v (i)† (q)v (j) (q) + b̂†i (q)b̂j (q) u(i)† (q)u(j) (q)
 h i 
† ˆ† (i)† (j) +2iq0 x0
+ b̂i (−q)dj (q) u (q)v (−q) e

With the orthogonality relations of Eq. (214) and their counterparts for
terms u(i)† (−q)v (j) (q) and v (i)† (−q)u(j) (q), the first and the last term in the
bracket above vanish. We finally arrive at the Hamiltonian
2 h
d3 q
Z i
† †
X
Ĥ = q 0 b̂ (q) b̂i (q) − ˆ
di (q)dˆ (q) , (228)
i i
(2π)3 2q0
i=1

which exhibits the same problems with infinite ground state energy as its
counterpart of the Klein-Gordon field, cf.. Sec. 4.2. We cure this, again,
by applying normal-ordering, Eq. (160) for bosons, which for fermion fields,
however, comes with an extra minus sign to encode the Pauli exclusion
principle,
:dˆ†i (q)dˆi (q):= − :dˆi (q)dˆ†i (q):= dˆ†i (q)dˆi (q) . (229)
Therefore, the normal-ordered Hamiltonian is given by

2 h
d3 q
Z i
† ˆ† (q)dˆi (q) .
X
Ĥ = q 0 b̂i (q) b̂i (q) + di (230)
(2π)3 2q0
i=1

Introducing number operators N̂± for particles with positive and negative
energy, electrons and positrons,
2 2
b̂†i (q)b̂i (q) and N̂− (q) = dˆ†i (q)dˆi (q) ,
X X
N̂+ (q) = (231)
i=1 i=1

we see that the Hamiltonian merely sums the energies of these particles
d3 q
Z h i
:Ĥ: = q 0 N̂+ (q) − N̂− (q) . (232)
(2π)3 2q0

Conserved Charge In a similar way, we can construct the (normal-

ordered) charge operator : Q̂ :. Promoting the fields in the 0-component
of the current density Eq. (200) to field operators
Z h i
:Q̂:= d3 x : iψ̄ˆ(x)γ 0 ψ̂(x) : , (233)

99
we arrive at
2 h
d3 q
Z i
† ˆ† (q)dˆi (q) .
X
:Q̂:= b̂i (q)b̂i (q) − di (234)
(2π)3 2q0
i=1

Expressed through the number operator this becomes

d3 q
Z h i
:Q̂: = N̂+ (q) − N̂− (q) , (235)
(2π)3 2q0

and the overall charge of the system is given by the difference of the total
numbers of positively and negatively charged particles. It is a straightfor-
ward exercise to show that the charge is conserved, by asserting that the
commutator of the charge and Hamilton operator vanishes; we leave this as
an exercise.

100
5.3 Problems & Solutions
1. Dirac equation and Anti-Commutators
Show how the anti-commutation relations for the α and β matrices
follow from the requirement that the solutions to the Dirac E.o.M.
∂
i ψ(x, t) = [−iα · ∇ + βm] ψ(x, t)
∂t
also satisfy the KG equation.

Solution

To show that ψ also satisfies the Klein-Gordon equation, consider the

square of the equation and demand that it reduces to the differential
operator of the Klein-Gordon equation, i.e.

∂2
− ψ(x, t) = [−iα · ∇ + βm]2 ψ(x, t)
∂t2 
= −(α · ∇)2 + β 2 m2 − im(α · ∇ · β + βα · ∇) ψ(x, t)

 
X 3 3
X
= − αi ∂i αj ∂j + β 2 m2 − im (αi β + βαi ) ∂i  ψ(x, t)
i,j=1 i=1
3
" #
!
X
= − ∂i2 + m2 ψ(x, t) = [−∇2 + m2 ]ψ(x, t)
i=1

where we have imposed equality with the relevant part of the Klein-
Gordon equation in the final line. Direct comparison with individual
terms shows that:
3
X 3
X
− αi αj ∂i ∂j = − ∂i2
i,j=1 i=1
2 2 2
+β m = m
3
X
−im (αi β + βαi ) ∂i = 0
i=1

and therefore

αi αj + αj αi = {αi , αj } = 0 fori 6= j and αi2 = 1

β2 = 1
αi β + βαi = {αi , β} = 0

101
2. Commutators with the Dirac Hamilton Operator
Calculate the following commutators:

(a) [Ĥ, p̂];

(b) [Ĥ, L̂] with the orbital angular momentum operator L̂ = r̂ × p̂;
2
(c) [Ĥ, L̂ ]
i
(d) [Ĥ, Ŝ] with the spin operator Ŝ = − α̂ × α̂;
4
(e) [Ĥ, Ĵ] with the total angular momentum operator Ĵ = L̂ + Ŝ

Hint: To alleviate the calculation, express the Hamilton op-

erator with the α and β matrices.

Solution

Remember that the Hamilton operator expressed through the α and

β matrices is given by

Ĥ = α · p̂ + βm ,

and use the commutator [x̂i , p̂j ] = iδij

(a) [Ĥ, p̂]:

[Ĥ, p̂] = [α · p̂ + βm, p̂] = 0 .

(b) [Ĥ, L̂]: we calculate this commutator component-wise,

[Ĥ, L̂i ] = ijk [α · p̂ + βm, x̂j p̂k ]

i
= ijk p̂k αl [p̂l , x̂j ] = −iijkαj p̂k = ip̂ × α


and therefore [Ĥ, L̂] = ip̂ × α.

2
(c) [Ĥ, L̂ ]:
2
[Ĥ, L̂ ] = [Ĥ, L̂i L̂i ]
 
= αj [p̂j , L̂i L̂i ] = αj [p̂j , L̂i ]L̂i + L̂i p̂j , L̂i ]
 
ijk k i i k
= −i αj p̂ L̂ + L̂ p̂ 6= 0 .

(d) [Ĥ, Ŝ]: we calculate this commutator component-wise,

i iijk
[Ĥ, Ŝ ] = − [α · p̂ + βm, αj αk ]
4

102
 iijk l iijk l
= − p̂ [αl , αj αk ] = − p̂ [(2δlj − αj αl )αk − αj (2δlk − αl αk )]
4 4
iijk  j  i
p̂ αk − p̂ αj = −iijk p̂j αk = − ip̂ × α
k

= −
2
and therefore
(e) [Ĥ, Ĵ] = −p̂ × α:

[Ĥ, Ĵ] = [Ĥ, L̂] + [Ĥ, Ŝ] = 0 .

This proves that neither orbital nor spin angular momentum are
conserved quantities for the free fermions described by the Dirac
equation, and only their total angular momentum is conserved.

3. ∗ Direct Solution of the Dirac Equation Solve the Dirac equation

directly, by using the specific form of the γ matrices in the Dirac form.

Solution

We express the Dirac spinor in the equation of Eq. (??), (iγ µ ∂µ −

m)ψ = 0 by decomposing it into a plane-wave factor multiplying two
two-component spinors,
 
−ip·x ψ+
ψ=e .
ψ−

Using the Dirac γ matrices of Eq. (??) we then obtain an equation for
the two components as
  
E − m −σ · p ψ+
= 0,
σ · p −E − m ψ−

where pµ = (E, p). To solve this system, its determinant must vanish
and we arrive at
 
 E − m −σ · p 
0 =   = −(E 2 − m2 ) + (σ · p)2 = −E 2 + p2 + m2 ,
σ · p −E − m 

and we recover theqwell-known energy-momentum relation leading to

solutions if E = ± p2 + m2 .
For the positive energy solution, the system hs the form

(E − m)ψ+ − (σ · p)ψ− = 0
(σ · p)ψ+ − (E + m)ψ+ = 0

103
 implying that
σ·p
ψ− = ψ+
E+m
and therefore he positive energy solutions are given by
 
ψ+
u(p) = u+ (E, p) =  σ · p
 

ψ+
E+m
with two basic spinors ψ+ for spinup and spin-down solutions given by
   
(±) 1 0
ψ+ = and .
0 1

Similarly, for the negative energy-solutions we find

 σ·p 
− ψ−
u− (−E, p) =  E + m
 

ψ−

with two basic spinors ψ+ for spinup and spin-down solutions given by
   
(±) 1 0
ψ− = and .
0 1

The last thing to note is that for the negative-energy solutions there
emerged a relative sign between energy and momentum, which makes
the assignment of a plane-wave factor tricky. Therefore the v-spinors
where introduced such that
 σ·p 
ψ−
v(p) = u− (−E, −p) =  E + m
 

ψ−

4. Dirac spinor relations

(a) prove, by explicit calculation, that

ū(i) (p)u(j) (p) = 2mδij = −v̄ (i) (p)v (j) (p)

for all combinations of i and j and that

v̄ (i) (p)u(j) (p) = ū(i) (p)v (j) (p) = 0 .

For all scalar products use that ū = u† γ 0 and thar p∗± = p∓ .

104
(b) prove, by explicit calculation, that
2 2
(i) (i)
X X
u(i)
α ūβ = (p/ + m)αβ , vα(i) v̄β = (p/ − m)αβ
i=1 i=1

(c) can you find a normalisation constant η such that

2   2  
X (i) p/ + m X (i) p/ − m
u(i)
α ūβ = , vα(i) v̄β = ?
2m αβ 2m αβ
i=1 i=1

This is another often usec normalisation. What does it imply for

the scalar products?

Solution
√
(a) Scalar products ūu and v̄v, for η = E + m:
 T   
1 0 0 1
0 1
0 1 0   0
0  0
ū(1) (p)u(1) (p) = η 2 
  
 pz pz
   
E+m 0 −1 0
  0  
E+m

p− p+
E+m 0 0 −1 0 E+m
!
2 p2 + p2
p z x y
= η2 1 + 0 − −
(E + m)2 (E + m)2
E 2 + m2 + 2Em − p2 2m
= η2 2
= η2 = 2m
(E + m) E+m

 T   
1 0 0 1
0 0
0 1 0   0
0  1
ū(1) (p)u(2) (p) = η 2 
  
 pz
   p− 
E+m 0 −1 0
  0  
E+m

p− −pz
E+m 0 0 −1 0 E+m
 
2 pz p− pz p−
= η 0+0− + =0
(E + m)2 (E + m)2

 T   
0 0 0 10 1
1 1 0  00  0
ū(2) (p)u(1) (p) = η 2 
  
p +
   pz


E+m 0 −1 0  
  0
E+m

−pz p+
E+m 0 0 −1 0 E+m
 
2 pz p+ pz p+
= η 0+0− + =0
(E + m)2 (E + m)2

105
  T   
0 1 0 0 0 0
1   0 1 0 0  1
ū(2) (p)u(2) (p) = η 2 
  
p+    p− 

E+m
  0 0 −1 0  
E+m

−pz −pz
E+m 0 0 0 −1 E+m
E2 + 2
m + 2Em − p2 2m
= η2 = η2 = 2m
(E + m)2 E+m

 pz T   pz 
E+m 1 0 0 0 E+m
p−   0 p+
1 0 0 
v̄ (1) (p)v (1) (p) = η 2 
  
E+m    E+m 
 1   0 0 −1 0   1 
0 0 0 0 −1 0
!
p2z p2x + p2y
= η2 + −1−0
(E + m)2 (E + m)2
p2 − E 2 − m2 − 2Em 2m
= η2 2
= −η 2 = −2m
(E + m) E+m

 pz T   p− 
E+m 1 0 0 0 E+m
p−   0 −pz
1 0 0 
v̄ (1) (p)v (2) (p) = η 2 
  
E+m    E+m 
 1   0 0 −1 0   0 
0 0 0 0 −1 1
 
2 pz p− pz p−
= η − −0−0 =0
(E + m)2 (E + m) 2

 p+ T   pz 
E+m 1 0 0 0 E+m
−pz   0 p+
1 0 0 
v̄ (2) (p)v (1) (p) = η 2 
  
E+m    E+m 
 1   0 0 −1 0   1 
0 0 0 0 −1 0
 
2 pz p+ pz p+
= η 2
− 2
−0−0 =0
(E + m) (E + m)

 p+ T   p− 
E+m 1 0 0 0 E+m
−pz   0 −pz
1 0 0 
v̄ (2) (p)v (2) (p) = η 2 
  
E+m    E+m 
 1   0 0 −1 0   0 
0 0 0 0 −1 1
!
p2x + p2y p2z
= η2 + −1−0
(E + m)2 (E + m)2
p2 − E 2 − m2 − 2Em 2m
= η2 2
= −η 2 = −2m
(E + m) E+m

106
Scalar products of v̄u and ūv:
 pz T   
E+m 10 0 0 1
p−   01 0 0  0
v̄ (1) (p)u(1) (p) = η 2 
  
E+m
pz
   
 1   00 −1 0  
E+m

p+
0 00 0 −1 E+m
 
pz pz
= η2 − =0
E+m E+m

 pz T   
E+m 1 0 0 0 0
p−   0 1 0 0  1
v̄ (1) (p)u(2) (p) = η 2 
  
E+m    p− 
 1   0 0 −1 0   E+m

−pz
0 0 0 0 −1 E+m
 
2 p− p−
= η − =0
E+m E+m

 p+ T   
E+m 10 0 0 1
−pz   01 0 0  0
v̄ (2) (p)u(1) (p) = η 2 
  
E+m
pz
   
 1   00 −1 0  
E+m

p+
0 00 0 −1 E+m
 
p+ p+
= η2 − =0
E+m E+m

 p+ T   
E+m 1 0 0 0 0
−pz   0 1 0 0  1
v̄ (2) (p)u(2) (p) = η 2 
  
E+m    p− 
 1   0 0 −1 0   E+m

−pz
0 0 0 0 −1 E+m
 
2 −pz −pz
= η − =0
E+m E+m

 T   pz 
1 1 0 0 0 E+m
p+
0   0 1 0 0 
ū(1) (p)v (1) (p) = η 2 
  
E+m
 pz
   
E+m
  0 0 −1 0   1 
p−
E+m 0 0 0 −1 0
 
−pz −pz
= η2 − =0
E+m E+m

107
  T   p− 
1 0 0 1
0 E+m
−pz
0   0
1 0 0 
ū(1) (p)v (2) (p) = η 2 
  
E+m
 pz
   
E+m 0 −1 0
  0   0 
p−
E+m 0 0 −10 1
 
2 p− p−
= η − =0
E+m E+m

 T   pz 
0 0 0 01 E+m
p+
1 1 0  0
0 
ū(2) (p)v (1) (p) = η 2 
  
E+m
 p+
   
E+m 0 −1 0
  0  1 
−pz
E+m 0 0 −1 0 0
 
2 p+ p+
= η − =0
E+m E+m

 T   p− 
0 0 0 10 E+m
−pz
1   0
1 0 0 
ū(2) (p)v (2) (p) = η 2 
  
E+m
 p+
   
E+m 0 −1 0
  0   0 
−pz
E+m 0 0 −1
0 1
 
2 −pz −pz
= η − =0
E+m E+m

(b) We will use that p2 = E 2 − m2 = (E + m)(E − m) and that the

product of a “column-vector” and a “row-vector”, v · v T , yields a
matrix-object.
2
(i)
X
u(i) / + m]αβ = Eγ 0 − p~ · ~γ + m αβ
 
α ūβ = [p
i=1

−pz −p−
 
E+m 0
 0 E+m −p+ pz 
= 
 
−E + m

 pz p− 0 
p+ −pz 0 −E + m αβ

  T  
1 1 1 0 0 0
    

2 
0  0   0 1 0 0 
= η 
   
 pz 
 pz 


 0 0 −1 0


 E+m  E+m   
p+ p−
E+m E+m 0 0 0 −1

108
   T  
0 0 1 0 0 0
    
 1  1   0 1 0 0 
+
    
 p− 
 p+ 


 0 0 −1 0


 E+m  E+m   
−pz −pz
E+m E+m 0 0 0 −1 αβ

  T   T 
1 1 0 0
      
0 0 1 1
      
2 
  
= η   +
   
pz  −pz p−  −p+  
      
E+m E+m E+m E+m  
     
p+ −p− −pz pz
E+m E+m E+m E+m
αβ

pz p−
− E+m − E+m
 
1 0
 
 0 0 0 0 
= (E + m) 
 
pz 2 pz p−
pz


 E+m 0 − (E+m)2 − (E+m) 2


p+ pz p+ p+ p−
E+m 0 − (E+m) 2 − (E+m) 2

 
0 0 0 0
p+ pz
− E+m
 
 0 1 E+m

+
 
p− p+ p− pz p−
 0 E+m − (E+m)

2 (E+m)2

 
pz p+ pz 2
pz
0 − E+m (E+m)2
− (E+m)2

 
E+m 0 −pz −p−
 
 0 E+m −p+ pz 
= 
 


 pz p− −E + m 0 

p+ −pz 0 −E + m

The calculation for the v’s follows the same pattern.

(c) The normalisation then would be η 2 = (E + m)/(2m) and result
in

ū(i) (p)u(j) (p) = δij = −v̄ (i) (p)v (j) (p)

u†(i) (p)u(j) (p) = E
m δij = −v †(i) (p)v (j) (p)

5. γ Algebra

109
Prove the following identities
γ µ γµ = 4
γ µ γ ρ γµ = −2γ ρ
Tr(γ µ γ ν ) = 4g µν
Tr(γ µ γ ν γ ρ γ σ ) = 4 (g µν g ρσ + g µσ g νρ − g µρ g νσ )

Solution

Straightforward matrix multiplcation shows that

 
1 0 0 0
0 1 0 0 
γ 0 γ0 = γ 1 γ1 = γ 2 γ2 = γ 3 γ3 = 

=1
 0 0 1 0 
0 0 0 1
and summing them thus yields 4
Alternatively, without using the explicit representation as matrices:
1 1
γ µ γµ = {γ µ , γµ } = 2g µµ = 4
2 2
because g µµ = δ µµ and we thus sum over the 4 entries of the four-
dimensional unit matrix.
Using the anti-commutator of the gamma-matrices and cyclicity of
matrix multiplication under the trace operation yields the next two
desired results
γ µ γ ρ γµ = (2g µρ − γ ρ γ µ ) γµ = 2γ ρ − 4γ ρ
1 1
Tr(γ µ γ ν ) = Tr(γ µ γ ν + γ ν γ µ ) = Tr(2g µν ) = g µν Tr(1) = 4g µν ,
2 2
where in the last step we realise that the trace is over the 4 Dirac
(spinor) indices.
Finally, with the same steps,
Tr [γ µ γ ν γ ρ γ σ ] = Tr [γ µ γ ν (2g ρσ − γ σ γ ρ )]
= 2g ρσ Tr(γ µ γ ν ) − Tr [γ µ γ ν γ σ γ ρ ]
= 8g ρσ g µν − 8g νσ g µρ + Tr [γ µ γ σ γ ν γ ρ ]
= 8g ρσ g µν − 8g νσ g µρ + 8g µσ g νρ − Tr [γ σ γ µ γ ν γ ρ ]
= 8g µν g ρσ + 8g µσ g νρ − 8g µρ g νσ − Tr [γ µ γ ν γ ρ γ σ ]
and therefore
2Tr[γ µ γ ν γ ρ γ σ ] = 8 (g µν g ρσ + 8g µσ g νρ − 8g µρ g νσ )

110
 6. Dirac Hamiltonian from Creation and Annihilation Opera-
tors
Show that for the Dirac field indeed the Hamiltonian (not! normal-
ordered) is given by
2 Z
d3 p
Z h i
† ˆi (p)dˆ† (p)
X
3
Ĥ = d xĤ = p 0 b̂i (p)b̂i (p) − d i
(2π)3 2p0
i=1

Solution

Plugging expansions of fields in terms of creation and annihilation

operators,
2
d3 p X h −ipx (i)
Z i
ipx (i) ˆ†
ψ = e u (p)b̂i (p) + e v (p)di (p)
(2π)3 2p0
i=1
2 h
d3 p
Z i
e−ipx v̄ (i) (p)γ 0 dˆi (p) + eipx ū(i) (p)γ 0 b̂†i (p) ,
X
ψ† =
(2π)3 2p0
i=1

where multiplying from the right with γ 0 for the latter gives (remember
(γ 0 )2 = 1)
2
d3 p X h −ipx (i)
Z i
ψ̄ = e v̄ (p) ˆi (p) + eipx ū(i) (p)b̂† (p) ,
d i
(2π)3 2p0
i=1

into Hamiltonian from lecture, expressed in ψ and ψ † , gives11

←
→ ←
→
Ĥ = ψ † (−iα · ∇ + βm)ψ = ψ̄(−iγ · ∇ + m)ψ

and therefore
←
→
Z
Ĥ = d3 xψ̄(−iγ · ∇ + m)ψ

We use a lot of spinor identies, such as ū = u† γ 0 etc. in the following and try to make
11

them explicit through lots of intermediate steps. Remember that

←→ 1 1 − → ←−
ψ̄ ∂ ψ = ψ̄(∂ψ) − (∂ ψ̄)ψ = ψ̄( ∂ ψ) − (ψ̄ ∂ )ψ
2 2
Two other identities we will use arise from the E.o.M.:

/ − m)u(p) = 0
(p ←→ (p · γ + m)u(p) = Eγ 0 u(p)
(p
/ + m)v(p) = 0 ←→ (p · γ − m)u(p) = Eγ 0 u(p)

111
 2 
d3 p d3 q
Z X
3
= d x
(2π)3 2p0 (2π)3 2q0
i,j=1
h i ←
→
e−ipx v̄ (i) (p)dˆi (p) + eipx ū(i) (p)b̂†i (p) (−iγ · ∇ + m)
h i
−iqx (j) iqx (j) ˆ†
e u (q)b̂j (q) + e v (q)dj (q)
2 
d3 p d3 q
Z
1 3
X
= d x
2 (2π)3 2p0 (2π)3 2q0
i,j=1
h i
e−ipx v̄ (i) (p)dˆi (p) + eipx ū(i) (p)b̂†i (p)
h i
× (γ · q + m)e−iqx u(j) (q)b̂j (q) + (−γ · q + m)eiqx v (j) (q)dˆ†j (q)
h i
− e−ipx v̄ (i) (p)dˆi (p)(−γ · p + m) + eipx ū(i) (p)b̂†i (p)(γ · p + m)
h i
× e−ipx u(j) (p)b̂j (p) + eipx v (j) (p)dˆ†j (p)
2 
d3 p d3 q
Z
1 3
X
= d x
2 (2π)3 2p0 (2π)3 2q0
i,j=1
h i
e−ipx v̄ (i) (p)dˆi (p) + eipx ū(i) (p)b̂†i (p) Eq γ 0
h i
e−iqx u(j) (q)b̂j (q) − eiqx v (j) (q)dˆ†j (q)
h i
− e−ipx v̄ (i) (p)dˆi (p) − eipx ū(i) (p)b̂†i (p) Ep γ 0
h i
−iqx (j) iqx (j) ˆ†
e u (q)b̂j (q) + e v (q)dj (q)
2 
d3 p d3 q
Z
1 X
= d3 x
2 (2π)3 2p0 (2π)3 2q0
i,j=1
h
−i(p+q)x †(i)
Eq e v (p)u (q)dˆi (p)b̂j (q)
(j)

+ei(p−q)x u†(i) (p)u(j) (q)b̂†i (q)b̂j (p)

− e−i(p−q)x v †(i) (p)v (j) (q)dˆi (p)dˆ†j (q)

i
−ei(p+q)x u†(i)(p)v (j) (q)b̂†i (q)dˆ†j (q)
h
−Ep e−i(p+q)x v †(i) (p)u(j) (q)dˆi (p)b̂j (q)

−ei(p−q)x u†(i) (p)u(j) (q)b̂†i (q)b̂j (p)

112
 + e−i(p−q)x v †(i) (p)v (j) (q)dˆi (p)dˆ†j (q)
io
−ei(p+q)x u†(i)(p)v (j) (q)b̂†i (q)dˆ†j (q)
2 
d3 p d3 q
Z
1 X
=
2 (2π)3 2p0 (2π)3 2q0
i,j=1
h
Eq (2π)3 δ 3 (p + q)v †(i) (p)u(j) (q)dˆi (p)b̂j (q)

+(2π)3 δ 3 (p − q)u†(i) (p)u(j) (q)b̂†i (q)b̂j (p)

− (2π)3 δ 3 (p − q)v †(i) (p)v (j) (q)dˆi (p)dˆ†j (q)

i
−(2π)3 δ 3 (p + q)u†(i)(p)v (j) (q)b̂†i (q)dˆ†j (q)
h
−Ep (2π)3 δ 3 (p + q)v †(i) (p)u(j) (q)dˆi (p)b̂j (q)

−(2π)3 δ 3 (p − q)u†(i) (p)u(j) (q)b̂†i (q)b̂j (p)

+ (2π)3 δ 3 (p − q)v †(i) (p)v (j) (q)dˆi (p)dˆ†j (q)

io
−(2π)3 δ 3 (p + q)u†(i)(p)v (j) (q)b̂†i (q)dˆ†j (q)
2 
d3 p
Z
1 1 X
=
2 (2π)3 2p0 2p0
i,j=1
h
p0 v (p)u (−p)dˆi (p)b̂j (−p) − u†(i)(p)v (j) (−p)b̂†i (q)dˆ†j (−p)
†(i) (j)

i
+u†(i) (p)u(j) (p)b̂†i (p)b̂j (p) − v †(i) (p)v (j) (p)dˆi (p)dˆ†j (p)
h
−p0 v †(i) (p)u(j) (−p)dˆi (p)b̂j (−p) − u†(i)(p)v (j) (−p)b̂†i (q)dˆ†j (−p)
i
†(i) † †(i) †
−u (p)u (p)b̂i (p)b̂j (p) + v (p)v (p)dˆi (p)dˆj (p)
(j) (j)

2 
d3 p
Z
1
u†(i) (p)u(j) (p)b̂†i (p)b̂j (p)
X
=
2 (2π)3 2p0
i,j=1

†(i) (j) ˆ ˆ †
− v (p)v (p)di (p)dj (p)
2 h
d3 p
Z i
† †
X
= p0 b̂ (q)b̂i (p) − ˆ
di (p) ˆ
d (q)
i i
(2π)3 2p0
i=1

7. Gordon identities

113
(a) Prove the Gordon identities
 
ν
2mū(p1 )γµ u(p2 ) = ū(p1 ) (p1 + p2 )µ + iσµν (p1 − p2 ) u(p2 )
 
ν
2mv̄(p1 )γµ v(p2 ) = −v̄(p1 ) (p1 + p2 )µ + iσµν (p1 − p2 ) u(p2 ),

where
i
σµν = [γµ , γν ]
2
(b) Prove that

ū(p1 ) [σµν (p1 + p2 )ν ] u(p2 ) = iū(p1 )(p1 − p2 )ν u(p2 )

(c) Write the current Jµ = ū(p2 )p/1 γµ p/2 u(p1 ) as

 
2 2 ν
Jµ = ū(p2 ) F1 (m, q )γµ + F1 (m, q )σµν q u(p1 )

with q µ = pµ2 − pµ1 and determine the functions F1,2 (m, q 2 )

Solution

(a) Let us evaluate the terms proportional to σµν , by repeatedly using

the Dirac E.o.M.’s
/ − m)u(p) = 0 ,
(p ū(p)(p/ − m)
(p
/ + m)v(p) = 0 , v̄(p)(p/ + m)

and the anti-commutation relation of the γ-matrices, {γµ , γν } =

2gµν :
 
ū(p1 ) iσµν (p1 − p2 )ν u(p2 )
 
1
= ū(p1 ) γµ (p/1 − p/2 ) − (p/1 − p/2 )γµ u(p2 )
2
 
1
= ū(p1 ) γµ (p/1 − m) − (m − p/2 )γµ u(p2 )
2
 
1
= −mū(p1 )γµ u(p2 ) + ū(p1 ) γµ p/1 + p/2 γµ u(p2 )
2
= −mū(p1 )γµ u(p2 )
 
1 ν ν
+ ū(p1 ) 2gµν p1 − p/1 γµ + 2gµν p2 − γµ p/2 u(p2 )
2
 
= −ū(p1 ) 2m + (p1 + p2 )ν ) γµ u(p2 ) ,

114
 which proves the Gordon identity for the u-spinors. For the ex-
pressions v-spinors the proof is completely analogous, the only
difference is the sign in front of the terms proportional to mass.
(b)
 
ν
ū(p1 ) σµν (p1 + p2 ) u(p2 )
 
i
= ū(p1 ) γµ (p/1 + p/2 ) − (p/1 + p/2 )γµ u(p2 )
2
 
i
= imū(p1 )u(p2 ) + ū(p1 ) γµ p/1 − p/2 γµ u(p2 )
2
= imū(p1 )u(p2 )
 
i ν ν
+ ū(p1 ) 2gµν p1 − p/1 γµ − 2gµν p2 + γµ p/2 u(p2 )
2
= iū(p1 )(p1 − p2 )ν u(p2 ) .

(c) To evaluate the current let us take a look at the argument first
(and keep in mind that we can replace ū(p2 )p/2 → ū(p2 )m and
/1 u(p1 ) → mu(p1 ).
p
 
p /2 = 2pν1 gµν − γµ p/1 p/2
/ 1 γµ p
 
= 2p1µ p/2 − γµ 2p1 · p2 − p/2 p/1
 
= 2 p1µ p/2 + p2µ p/1 − p1 · p2 γµ − p/2 γµ p/1

and therefore, with q 2 = p21 + p22 − 2p1 · p2 = 2m2 − 2p1 · p2 ,

Jµ = ū(p2 )p
/1 γµ p
/2 u(p1 )
   
= ū(p2 ) 2 p1µ p/2 + p2µ p/1 − p1 · p2 γµ − p/2 γµ p/1 u(p1 )
 
2
= ū(p2 ) 2m(p1µ + p2µ ) − (2p1 · p2 + m )γµ u(p1 )
 
ν 2 2
= ū(p2 ) − 2miσµν q + (q + m )γµ u(p1 )
 
2 2 ν
= ū(p2 ) F1 (m, q )γµ + F1 (m, q )σµν q u(p1 )

where we have used the Gordon identity from part (a) in the last
step. Comparing coefficients we arrive at

F1 = (q 2 + m2 ) and F2 = −2im .

115
8. ∗ Dealing with γ5
We introduce γ5 , given by
 
i 0 1
5 0 1 2 3
γ5 = γ = iγ γ γ γ = − µνρσ γ µ γ ν γ ρ γ σ =
4! 1 0

(a) Confirm, by direct calculation and comparison that Hermitian

conjugates of the γ-matrices are given by

γ µ† = γ 0 γ µ γ 0 .

Use the definition of γ5 to show that

γ5† = γ5

and that
{γ µ , γ5 } = 0 .
(b) Show that
exp[−iθγ5 ] = cos θ + iγ5 sin θ .
(c) Analyse the behaviour of the free Dirac field Lagrangian under
chiral phase transformations given by

ψ → ψ 0 = exp [iθγ5 ] ψ
h i
ψ † → ψ 0† = ψ † exp −iθγ5† ,

and keep in mind that the Hermitian conjugate of γ5 , γ5† = γ5 .

Under which condition is the Lagrangian invariant under this
transformation, i.e. which condition must be fulfilled for L0 = L
to hold true.

Solution

(a) Direct calculation yields

γ 0† = γ 0 γ 0 γ 0 = γ0
   
1 0 0 +σi 1 0
γ i† = 0 i 0
γ γγ =
0 −1 −σi 0 0 −1
   †
0 −σi 0 σi
= = ,
+σi 0 −σi 0

because σi† = σi .

†
γ5† = iγ 0 γ 1 γ 2 γ 3 = −iγ 3† γ 2† γ 1† γ 0†

116
 = −iγ 0 γ 3 γ 0 γ 0 γ 2 γ 0 γ 0 γ 1 γ 0 γ 0 γ 0 γ 0 = −iγ 0 γ 3 γ 2 γ 1
= iγ 0 γ 3 γ 1 γ 2 = −iγ 0 γ 1 γ 3 γ 2 = iγ 0 γ 1 γ 2 γ 3 = γ5 ,

where we have used that γ0 γ0 = γ 0 γ 0 = 1 and the fact that the

different γ-matrices anti-commute.
For the anti-commutator, we use that squaring the γ µ results in
plus or minus the unit matrix, (γ 0 )2 = −(γ i )2 = 1, and that
the γ-matrices anti-commute, which yields a minus sign for ev-
ery “swap” of γ-matrices with explicitly different indices. We
therefore have

{γ µ , γ5 } = i γ µ γ 0 γ 1 γ 2 γ 3 + γ 0 γ 1 γ 2 γ 3 γ µ

0γ0γ1γ2γ3 + γ0γ1γ2γ3γ0 


 µ = 0 : i γ 
1γ2γ3 − γ0γ0γ1γ2γ3 = 0 
 


 = γ 

1 0 1 2 3 0 1 2 3 1




 µ = 1 : i γ γ γ γ γ + γ γ γ γ γ 


0 2 3 0 2 3
 
=γ γ γ −γ γ γ =0

 
= =0

 µ = 2 : i γ2γ0γ1γ2γ3 + γ0γ1γ2γ3γ2  
= −γ 0 γ 1 γ 3 + γ 0 γ 1 γ 3 = 0


 


 


 3 0 1
µ=3 : i γ γ γ γ γ +γ γ γ γ γ  2 3 0 1 2 3 3 

 
= γ0γ2γ3 − γ0γ1γ2 = 0
 

(b) To see how this works, remember that fnctions with matrices as
arguments can be defined by their Taylor series, and therefore
∞
X
exp[iθγ5 ] = (iθγ5 )k .
k=0

Let us first calculate powers of γ5 ,

 
0 1 0
γ5 = 1 =
0 1
γ51 = γ5
 2  
0 1 1 0
γ52 = = ,
1 0 0 1

and we see that even powers of γ5 are the unit matrix, while odd
powers yield the γ5 . Thus

exp[iθγ5 ] = cos θ + iγ5 sin θ .

(c) Reminding ourselves that the “barred” spinor is given by ψ̄ =

ψ † γ0 0 and therefore

ψ̄ → ψ̄ 0 = ψ † exp[iγ5† θ]† γ0 .

117
 ←→
Ignoring for a moment the ∂ notation, the Lagrangian trans-
forms as
L → L0 = ψ̄ 0 i∂/ − m ψ 0 = ψ † exp[iγ5† θ]† γ0 i∂/ − m exp[iθγ5 ]ψ



h i
h i
= ψ † cos θ − iγ5† sin θ γ0 i∂/ − m cos θ + iγ5 sin θγ5 ψ
  
= ψ † (cos2 θ)γ0 (i∂/) + (cos θ sin θ) γ5† γ 0 ∂/ − γ 0 ∂/γ5
i
+(sin2 θ)γ5† γ 0 (i∂/)γ5 ψ
  
† 2 0 † 0
−mψ γ0 cos θ + i cos θ sin θ γ γ5 − γ5 γ

+γ5† γ0 γ5 sin2 θ ψ
  
† 2 0 0
= ψ (cos θ)γ0 (i∂/) + cos θ sin θ γ5 γ ∂/ − γ ∂/γ5

2
+ sin θγ5 γ0 (i∂/)γ5 ψ
  
† 2 0 0
−mψ γ0 cos θ + i cos θ sin θ γ γ5 − γ5 γ

2
+ γ5 γ0 γ5 sin θ ψ
   
† 2 2
= ψ γ0 cos θ + sin θ (i∂/) ψ
    
† 2 2
−ψ γ0 cos θ − sin θ + γ0 iγ5 cos θ sin θ mψ
 
= iψ̄∂/ψ − mψ̄ cos(2θ) + iγ5 sin(2θ) ψ

where we have used that γ5 anti-commutes with every other γ-

matrix and that its square equals to 1.
This shows that the free Dirac field Lagrangian is invariant if the
fermions are massless, i.e. if m ≡ 0.

9. ∗ Spin Operator The spin-operator for Dirac fermions,expressed by

the γ-matrices is given by
i
Ŝ = γ × γ .
4
(a) Show that it can also written as
1
Ŝ = γ5 γ0 γ ,
2
where γ5 = iγ0 γ1 γ2 γ3 .

118
(b) Prove that the spin operator indeed satisfies the spin/angular
momentum algebra (a SU (2) algebra), given by

[Ŝ i , Ŝ j ] = iijk Ŝ k .

(c) Prove that

2 3
Ŝ = .
4

Solution

(a) Using the anti-commutator relation for the γ matrices and γ0 γ 0 =

1 and γi γ i = −1 we find that
i 1 i i
Ŝ i = ijk γj γk = γ5 γ0 γi = γ0 γ1 γ2 γ3 γ0 γ i = γ1 γ2 γ3 γi .
4 2 2 2
For example, for i = 1 we then have
i i i i
Ŝ 1 = 1jk γj γk = γ 2 γ 3 = γ1 γ2 γ3 γ 1 = γ 2 γ 3
4 2 2 2
and similar logic gives us the results for i = 2 and i = 3.
(b) To calculate the commutator, we will use the anti-commutator of
the γ-matrices and the anti-symmetry of the Levi-Civita tensor.
We need to compare the result for the commutator with
1 1
iijk Ŝ k = − ijk klm γ l γ m = − γ i γ j − γ j γ i


4 4
and therefore
h i 1
Ŝ i , Ŝ j = iijk Ŝ k = − γ i γ j − γ j γ i


4
1 ikl jmn
= −   [γk γl , γm γn ]
16  
1 ikl jmn
= −   [γk γl , γm ] γn + γm [γk γl , γn ]
16

1 ikl jmn
= −   γk {γl , γm } γn − {γk , γm } γl γn
16

+γm γk {γl , γn } − γm {γk , γn } γl
 
1 ikl jmn
= −   glm γk γn − gkm γl γn + gln γm γk − gkn γm γl
8
 
1 ikl jnl ikl jnl ikl jnl ikl jnl
= −   γk γn +   γk γn −   γn γk −   γn γk
8

119
  
1 ikl jnl
= −   γk γn − γn γk
4
  
1 ij kn in jk
= − g g −g g γk γn − γn γk
4
   
1 ij j i ij i j 1 i j j i
= − −3g − γ γ + 3g + γ γ = − γ γ −γ γ
4 4

(c) Let’s now calculate Ŝ 2 :

2 1 ijk ilm
Ŝ =   γj γk γl γm
16
1  jl km  1  
= g g − g jm g kl γj γk γl γm = γj γk γ j γ k − γj γk γ k γ j
16 16
1 h j k j
i 1 h i
= (2gjk − γk γj )γ γ + 3γj γ = −2γk γ k + 3γk γ k + 3γj γ j
16 16
12 3
= − =− .
16 4

120
6 Electrodynamics
In this section we will quantise electrodynamics, by quantising the free vec-
tor potential that gives rise to the (free) electromagnetic fields. It turns
out that this results in a somewhat more involved procedure; while the vec-
tor potential has four components, which we would naively treat as four
independent quantities – scalar fields – and quantise them accordingly, the
gauge invariance of the fields implies that in fact there are only two physically
meaningful degrees of freedom. This means that, naively exercised, our algo-
rithm of second quantisation would lead to a degree of “over-quantisation”,
i.e. trying to quantise objects that cannot and should not be quantised in a
consistent and physically meaningful way. The solution to this is to fix the
gauge before quantising the fields, which is nothing but the imposition of
additional external conditions.
The quantisation of the four potential is, as indicaed, a somewhat tricky
business. In my opinion, the best explannations of the procedure can be
found in Chapter 5 of Hatfield’s book [3], and in Section 3.2 of Itzykson &
Zuber [12].

6.1 Gauge Invariance as Obstacle

Lagrangian and Gauge Invariance, once more Remember the (free)
Lagrangian of Eqs. (125) and (133),

E2 − B2 1
L = = − F µν Fµν ,
2 4
where we have set the current to zero, j µ = 0 and moved a factor of 4π into
the vanishing j µ Aµ term in the first expression. It is simple to show that
under the gauge transformations of Eq. (120) ,

Aµ → A0µ = Aµ − ∂ µ Λ ,

the field strength tensor F µν is a gauge-invariant quantity. In fact it is a

constant,

F µν → F 0µν = ∂ µ A0ν − ∂ ν A0µ = ∂ µ (Aν − ∂ ν Λ) − ∂ µ (Aµ − ∂ µ Λ)

= ∂ µ Aν − ∂ ν Aµ = F µν . (236)

Reminding ourselves of the connection of the field strength tensor with the
electric and magnetic fields E and B, Eq. (122), invariance of the fields
under gauge transformations is manifest.
This has two implications, which are worth making explicit: First of all, al-
though we will explicitly quantise the vector potential Aµ and only indirectly,
through it, the fields, the latter are the physical quantities, measurable in

121
every day life 12 . Secondly, and in the context of what follows more impor-
tantly, we may use special forms of the gauge transformation, Eq. (120), to
eliminate some components of Aµ without impacting on the physics. But
this also implies that there are less than four physically meaningful degrees
of freedom encoded in the vector potential, and we will have to deal with
the problem of how to quantise a system that has less physical degrees of
freedom than the field that is used for its description.

Fixing the Gauge Let us discuss now some of the conditions that can
be imposed on Aµ , which effectively fix the gauge. Looking at the form
of the field strength tensor it is worth noting that F 00 = 0, which implies
that there is no conjugate momentum for the temporal component of Aµ .
Defining them, as before, through
∂L
πµ = (237)
∂ Ȧµ
and specialising on µ = 0 yields
∂L
π0 = = 0. (238)
∂ Ȧ0
This motivates us to use a temporal gauge defined by
Zt
Λ(t, x) = dt0 A0 (t0 , x) (239)
−∞

which results in A0Λ = 0.

Coulomb vs. Lorentz vs. Axial Gauge It turns out, however, that
this does not yet entirely fix the gauge and an additional condition has to
be applied. Three types of gauge, with different calculational advantages
and disadvantages in different situations are frequently found:
ˆ Coulomb gauge, defined through

∇ · A = 0. (240)

ˆ Lorentz gauge, defined through

∂µ Aµ = 0 . (241)

ˆ Axial gauge, defined through, e.g.

Az = 0 . (242)
12
The impact of a finite vector potential in regions where the fields vanish is subject of
the Aharonov-Bohm effect, which is discussed, for example, in Chapter 2.6 of Sakurai’s
book [11].

122
Polarisation Vectors and Degrees of Freedom To build on this idea
of gauge fixing, let us analyse in some more detail what this actually implies.
Most transparently this can be done in Coulomb gauge. For free fields, i.e.
with j = 0 and, in particular the charge density ρ = j 0 = 0, A0 is not
a dynamical degree of freedom: its derivative w.r.t. time is not present
in the free field Lagrangian and hence its conjugate momentum vanishes.
The temporal part of the gauge in Eq. (239) fixes this constant then to
A0 = 0, making the lack of dynamical relevance explicit. This leaves only
the spatial components of A, A, and the field strength tensor is composed
of the components of ∇ × A, as F ij = ∂ i Aj − ∂ j Ai .
But imposing the Coulomb gauge condition by demanding that the diver-
gence of A vanishes, ∇ · A = 0 we exposed that there is a longitudinal com-
ponent of A, AL . By definition of it being longitudinal, ∇ × AL = 013 . This
implies that yet another component of F vanishes, or, differently put, we
see that also AL is not dynamically relevant. This shows that the Coulomb
gauge is the one where the longitudinal degree of freedom vanishes, AL = 0.
Not surprisingly, imposing two conditions on the four-vector Aµ eliminates
two of its components, and we are left with two degrees of freedom. The
logic above, eliminating the temporal and longitudinal parts of A from the
dynamical degrees of freedom means that we are left with two transverse
degrees of freedom AT .
To make the physics of this more explicit, let us see how this works out in
practice. Assume we want to describe a quantum of electromagnetism, a
photon, with momentum k. It’s four-momentum of course is given by
q
k µ = (ω, k) with ω = k0 = k 2 . (243)

The relevant degrees of freedom for the photon are its two remaining po-
larisations. They are usually denoted by λ = {1, 2} and represented by
polarisation vectors µ(λ) (k). Fourier transformation of the gauge conditions
above then become conditions on products of the three-momentum and the
polarisation three-vector; while the temporal gauge condition implies 0 = 0
we have:

k · (λ) (k) = 0 (no longitudinal polarisation) . (244)

Demanding additionally that the polarisation vectors are real and ortho-
normal we have

(λ) (k) · (κ) (k) = δλκ . (245)

13
The simplest way to see this is to assume a fixed longitudinal axis, for example the
z-axis. Then the photon momentum k is parallel to the z-axis, but, in addition, also Az ,
the logitudinal component, is parallel to the z-axis. Fourier-transforming the condition
then yields k × AL = 0 and therefore ∇ × AL = 0.

123
A simple way to guarantee this is to orient k along the z-axis. Then

(1) = (1, 0, 0) and (2) = (0, 1, 0) . (246)

6.2 Coulomb Gauge

Logic of the Procedure We will try to explicitly follow the algorithm
for the second quantisation of a field as summarised in Fig. 1, and highlight
specifically, where this algorithm starts to crash. Identifying the components
of Aµ as the fields to be quantised, we have:

1. determine conjugate momenta π ν

∂L

0
 π = = 0


ν ∂L ∂ Ȧ0
π = =⇒ (247)
∂ Ȧν  πi
 ∂L
 = = −Ei .
∂ Ȧi
This makes the anticipated problem of vanishing conjugate momen-
tum for A0 manifest. Further down the line it will prevent us from
quantising it, because we will not be able to produce a non-vanishing
commutator between this field component and its conjugate momen-
tum: for our choice of Lagrangian, it is guaranteed that [A0 , π 0 ] = 0
irrespective of what we try to do and therefore quantisation of A0 is
bound to fail.

2. construct the Hamiltonian

As before, the Hamiltonian density expressed through the electric and
magnetic fields is given by

E2 + B2
H = Ȧµ πµ − L = + E · ∇A0 , (248)
2
where the last term obviously vanishes if we set A0 = 0.

3. promote fields to field operators

4. demand equal-time commutators of fields and conjugate momenta

Due to π 0 = 0 we have only non-vanishing equal-time commutators
for spatial components, namely
h i h i
Âi (t, x), π̂j (t, y) = iδij δ 3 (x − y) = − Âi (t, x), Êj (t, y) .
(249)

124
Dealing with A0 : Gauss’ law To re-iterate: the fact that π 0 = 0 means
that also the field operator vanishes and hence commutes with every field
operator. Therefore A0 is not a dynamical variable, and Ȧ0 = 0. This
means that A0 is not an operator but an ultimately inconsequential number
in our construction of a quantum field theory. However, there is a direct
consequence of it being not dynamical:
∂L
= 0 −→ ∇ · E = 0, (250)
∂A0
Gauss’ law in the absence of sources14
We would of course be tempted to implement this as a wonderfully physical
constraint on the field operators. But this would lead to yet another way to
see the problem with the procedure. Going back to the commutation rela-
tions, and forming a divergence we would arrive at, somewhat schematically,
X ∂ h i
 i (t, x), Êj (t, y)
∂y j
j
h i X ∂
= Âi (t, x), ∇ · Ê j (t, y) = −i δij j δ 3 (x − y) . (251)
∂y
j

This is difficult, because while the left hand side of the second line vanishes,
due to Gauss’ law, the right hand side doesn’t. This implies that we cannot
implement Gauss’ law as an operator equation.

Dealing with A0 : Conditions on the States Realising that we cannot

implement Gauss’ law as a direct constraint on the field operators, we could
try and rephrase it as a condition on the allowed states |ψi forming the Fock
space on which the operators then act. We would proceed by demanding
that all physical states |ψi satisfy

∇ · Ê |ψi = 0 (252)

and would classify all states that do not fulfil this criterion as unphysical
and ignore them. It is a bit cumbersome to show that this doesn’t work
either and in fact would also violate the commutation relations.
The next weaker constraint, however, works. Demanding that for physical
states Gauss’ law is satisfied as expectation value,

hψ| ∇Ê |ψi = 0 (253)

14
To see this, let us go back to Eq. (119), which encodes the connection of the vector
potential to the electric field, and form its divergence, i.e.




∇ · E = ∇ · −∇A0 − ∂t A = −∇ · ∂t A = −∂t ∇ · A = 0 ,
where we have first used that A0 = 0 and then employed the Coulomb gauge condition
after switching the sequence of derviatives.

125
encapsulates this part of Maxwell’s equation as an average. We will come
back to its implications at a somewhat later state.

Solving the Crisis: Transverse δ-function The solution to the prob-

lem of Gauss’ law is to modify the commutation relation in such a way that
they automatically encode it. This is done by replacing the δ-function on the
right hand side of the commutation relations of Eq. (255) with a transverse
tr defined as
δ-function, δij

d3 k ik·(x−y)
Z  
3 tr ki kj
δij δ (x − y) −→ δij (x − y) = e δij − 2 .
(2π)3 k
(254)

The modified commutators then read

h i h i
tr
Âi (t, x), π̂j (t, y) = iδij (x − y) = − Âi (t, x), Êj (t, y) . (255)

It is easy to show that the gradient of the transverse δ-function with respect
to x or y vanishes, because derivatives will produce a term ±ki multiplying
the rounded bracket, and
X  kj k 2

ki kj
ki δij − 2 = kj − 2 = 0. (256)
i
k k

This means that, with the modified commutator relation, ∇ · E now com-
mutes with every meaningful operator, and in particular
h i
Âi (t, x), ∇ · Ê(t, y) = 0 . (257)

We can therefore safely set it to 0, asserting the validity of Gauss’ law.

More Benefits of δ tr As a byproduct, forming a divergence w.r.t to the

x-position yields
h i
∇ · Â(t, x), Êj (t, y) = 0 , (258)

and we recover the Coulomb gauge condition ∇ · A = 0.

Non-Vanishing Commutator at Space-like Distances But there is a

little snag. Replacing the δ-function with its transverse modification implies
that it is not guaranteed any more that the commutators [Âi , Êj ] vanish for
space-like distances. This looks like a severe problem with the causality
structure of the theory. However, there are two answers to the problem.

126
 1. Â is a gauge-dependent quantity and therefore essentially unphysical.
It cannot directly be measured, and therefore, any potentially harmful
a-causal behaviour may not have physical implications.
2. careful calculations reveals that while [Âi , Êj ] may not vanish for
space-like distances, the commutators of the physical E and B fields
and their components do vanish, irrespective of the use of the trans-
verse δ function.

Creation and Annihilation Operators Reminding ourselves that we

have set A0 = 0, the field is expanded in terms of plane waves and creation
and annihilation operators as
2 
d3 k
Z X 
(λ) −ik·x ∗(λ) † ik·x
Â(x, t) =  (k)â(k, λ)e +  (k)â (k, λ)e ,
(2π)3 (2k0 )
λ=1
(259)
where the sum is over the two polarisations of the photons and, similar to
the case of the Dirac spinors, we have “scalar” creation and annihilation
operators for each of the polarisation states. As before, the creation and
annihilation operators enjoy commutation relations, namely
h i
â(k, λ), ↠(q, κ) = (2π)3 2k0 δ 3 (k − q) δλκ (260)

with all other commutators vanishing.

More on Polarisations Note that we now also allow complex polarisation

vectors, to capture, for example circular polarisations. Assuming that the
photon momentum is oriented along the positive z-axis, k = kez , we could
use real polarisation vectors for linear polarisations, as
   
0 0
1 0 
µ(λ=1) (k) =  µ(λ=2)
  
 0  and  (k) = 
 1 , (261)
 

0 0
while for circular polarisations we could write
 
0
1 1
µ(λ=1,2) (k) = √ 
 
. (262)
2  ±i 
0
Using four-vectors instead of three vectors means that we replace the transver-
sality condition with k µ · µ = 0, keeping the ortho-normality condition of
Eq. (245).

127
6.3 Lorentz Gauge
Modifying the Lagrangian One of the issues with the Coulomb gauge
is that it is not Lorentz-invariant. To achieve this invariance, though, we
will need to demand commutator relations that fully reflect this symmetry,
h i
µ (t, x), π̂ν (t, y) = igµν δ 3 (x − y) . (263)

This implies, obviously, that all four components of the vector potential have
a conjugate momentum, and, in particular, that π 0 doesn’t vanish. Since
π 0 emerges by differentiation of the Lagrangian w.r.t Ȧ0 , we must modify
the Lagrangian such that this derivative does not vanish any more. This is
achieved by modifying the free-field Lagrange density,

1 1 α
L = − F µν Fµν −→ L = − F µν Fµν − (∂µ Aµ )2 . (264)
4 4 2

Here, α is the, in principle, arbitrary gauge parameter, and physical results

should not depend on its actual choice. This kind of modification is not un-
known from classical mechanics, where external conditions on the dynamics
are often encoded through the method of Lagrange multipliers 15 .

Modified Maxwell Equations and Feynman Gauge Adding a source

term 4πjµ Aµ , the resulting Maxwell equations read

∂µ ∂ µ Aν − (1 − α)∂ ν (∂ · A) = 4πj ν , (265)

and it is suggestive to set α = 1 to recover their original form. This gauge

choice is commonly referred to as Feynman gauge.

Conjugate Momenta As usual, the conjugate momenta are calculated

by differentiation, and we have
 0
µ ∂L µ0 µ0 π = −α(∂ · A)
π = = F − αg (∂ · A) −→ (266)
∂ Ȧµ π i = −E i .

Clearly, the modification of the Lagrange density only modified π 0 , which

now is proportional to the gauge parameter α.

Imposing Lorentz Gauge We now have to decide how to impose the

constraint ∂ · A = 0 which defines the Lorentz gauge that we chose at the
beginning of this discussion. Adjusting the commutators, like in the case of
the Coulomb gauge, is not viable, because we have already postulated the
15
See for example the discussion in Goldstein’s book [9].

128
commutation relations we would like to use, namely the ones in Eq. (263).
We also cannot impose the constraint ∂ · A = 0 as an operator equation,
because this would imply that π 0 = 0, and we would not be able to recover
our postulated commutator relations.
This means that we are forced down an avenue that we briefly considered in
the case of the Coulomb gauge, by demanding that we implement the gauge
condition as a condition on physical states. We realise very quickly that it
cannot be realised as a condition on physical states |ψi,

∂ · Â |ψi = 0 , (267)

for the following reason. Consider the expectation value of the commutator
relation Eq. (263), specified for µ = ν = 0:
h i
hψ| Â0 (t, x), π̂ 0 (t, y) |ψi = iδ 3 (x − y) hψ|ψi . (268)

But at the same time

π̂ 0 |ψi = (∂ · Â) |ψi = 0 (269)

enforces that the l.h.s. of Eq. (268) must vanish, while the r.h.s. does not.
This rules out the weaker constraint as a viable, consistent option.
This leaves us the only option to encode the gauge condition by demand-
ing that it holds only true for expectation values of physical states, i.e.,
demanding that

hψ| ∂ · Â |ψi = 0 (270)

for physical states ψ. To implement this, it is sufficient to demand that for

the positive energy/frequency Â+ part of the field operator we have

∂ · Â+ |ψi = 0 , (271)

because we can write

 
hψ| ∂ · Â |ψi = hψ| ∂ · Â− + ∂ · Â+ |ψi
 †
= ∂ · Â− |ψi |ψi + hψ| ∂ · Â+ |ψi = 0 . (272)

We will use this after we defined polarisation vectors and expanded the field
operators in plane waves and creation and annihilation operators.

Field Operators The field operators are expanded as

3 h
d3 k
Z i
†
X
(λ) −ik·x ∗(λ) ik·x
µ (x) = µ (k)âλ (k)e + µ (k)âλ (k)e ,
(2π)3 2k0
λ=0
(273)

129
where we have chosen four linearly independent polarisation vectors as
       
1 0 0 0
0  1  (2)  0  (3)  0 
(0) (k) = 
  (1)
 ,  (k) =   ,  (k) =   ,  (k) =   .
 0   0   1   0 
0 0 0 1
(274)

For simplicity we assumed that the photon momentum is oriented along the
positive z-axis, k k ez It is easy to check that the polarisation vectors satisfy

(λ) (k) · ∗(κ) (k) = (λ)µ (k)µ∗(κ) (k)gλκ , (275)

where the differnence between labels (λ) for the polarisation vectors and
their components - the Lorentz index µ has been made explicit. A simple
calculation shows that the commutators of Eq. (263) are satisfied, if the
only non-vanishing commutator of the creation and annihilation operators
is given by
h i
âλ (k), â†κ (q) = −(2π 3 )(2k0 )gλκ δ 3 (k − q) . (276)

Hamiltonian The resulting Hamiltonian density is given by

" 3 #
d3 k
Z X † 
†
:Ĥ: = k0 âλ (k)âλ (k) − â0 (k)â0 (k) . (277)
(2π)3 2k0
λ=1

The form of the Hamilton operator exhibits a potential problem: clearly,

scalar photons, i.e. those with λ = 0, come with a negative sign, opposite
to what we want and what we know how to deal with. At first sight, this
seems to signal that our attempt at quantising the electromagnetic fields
in Lorentz gauge failed, and that we arrived at a Hamiltonian describing
an energy spectrum that is not bounded from below, despite the normal
ordering. The reason for this, of course, can be traced back to the use of the
metric tensor in the quantisation conditions, which enforces a state with a
“wrong” sign. However, careful inspection below will reveals that this is not
a real problem and that the corresponding states are unphysical, motivating
us to call them “ghosts”.

Physical States So, let us now take a closer look at some of the states
and their energies. Start with the by now familiar assertion that the vacuum
reduces to zero when the one of the annihilation operators is applied,

âλ (k) |0i = 0 ∀λ . (278)

130
Now, let us analyse one of the more tricky states: a scalar photon, modulated
by some well-behaved function f (k),

d3 k
Z
|1S i = f (k)â†0 (k) |0i . (279)
(2π)3 2k0

As already anticipated, the norm of this states is negative,

d3 k d3 k 0
Z Z
h1S |1S i = f (k)f ∗ (k 0 )h0|â(k 0 , 0)↠(k, 0)|0i
(2π)3 2k0 (2π)3 2k0
d3 k
Z
= − h0|0i |f (k)|2 < 0 . (280)
(2π)3 2k0

The minus sign of course stems from the relative sign in the metric, or,
when followed through, from the “-”-sign in front of the right-hand side
of the commutator in Eq. (276). Phrased differently, the combinations of
positive and negative energy solutions that are still allowed destroys the
positive definiteness of the norm.
So let us impose the gauge constraint ∂ · Â+ |ψi = 0. Evaluating ∂ · Â+ we
of course only take into account the positive energy solutions and arrive at
3 h
d3 k
Z X i
µ (λ) −ik·x
∂ · Â+ = −i k µ (k)â λ (k) e . (281)
(2π)3 2k0
λ=0

We can simplify this further by realising that for λ = {1, 2} the polarisations
are orthogonal to the momentum,  ⊥ k and therefore k ·  = 0. This leaves
us with two surviving polarisations, scalar (λ = 0) and longitudinal (λ = 3).
Our constraint on physical states |ψi therefore becomes

0 = ∂ · Â+ |ψi
d3 k
Z h i
= −i k · (0)
(k)â0 (k) − k · (3)
(k)â3 (k) e−ik·x . (282)
(2π)3 2k0

For massless four-momenta – k 2 = 0 – longitudinal momentum and energy

coincide, kk = k = k0 and therefore k · (0) = −k · (3) . This implies that the
gauge constraint can be satisfied if
 
â0 (k) − â3 (k) |ψi = 0 (283)

for all physical states |ψi

131
6.4 Problems & Solutions
1. Polarisation vectors in Coulomb gauge
Assume a momentum parallel k to the z-axis √ and introduce left- and
right-circular polarisations  (L),(R) (k) = 1/ 2(0, 1, ±i, 0). Show by
explicit calculation that for the spatial components of the polarisation
vectors
X (λ) ∗(λ) kj ki
i (k)j (k) = δij − 2
λ=1,2
k

Solution

In Coulomb gauge we assume the two polarisation vectors to be trans-

verse to the axis of motion, and that for linear polarisations they are
real-valued, ∗ = , which of course is not true for the circular polari-
sations. For k parallel to the z-axis we have
   
0 0
1  0 
(1) (k) =  (2)
   
 0  and  (k) =  1 


0 0

We need the product

2  
X (λ) (λ) δij if i, j ∈ {1, 2} 16 ki kj
i (k)j (k) = = δij −
λ=1
0 if i, j = 3 k2

Choosing instead circular polarisations yields the following products,

1·1+1·1


 i = j = 1 : =1
2



i · i∗ + (−i) · (−i∗ )



2  i=j=2:
 =1
(λ) ∗(λ) 2
X
i (k)j (k) =
1 · i + 1 · (−i)
i = 1, j = 2 : =0

λ=1 



 2
 i = 2, j = 1 : 1 · (−i) + 1 · i = 0



2
This is the same result as before - the sum over polarisations therefore
is independent of your choice of basis (linear vs. circular, as in this
example).
16
To see the first equation set, i = j = 1. Then the only relevant entry is the x-
component of the λ = 1 polarisation vector (=1), squaring it yields a 1. Similarly, for
i = j = 2 the only contributor is the y-component of the λ = 2 polarisation vector. If i
and j are any different combination, one or both entries will be zero, for each λ.

132
 Expert level: To write this in more convenient form, introduce a time-
like unit-length four-vector η (for simplicity we can assume η = (1, 0, 0, 0)),
and
k µ − η µ (k · η)
 
µ 1 0
k̃ = p −→ .
(k · η)2 − k 2 |k| k

Thus, for our choice of η, k̃ just becomes the direction of the three-
momentum of k, with no temporal component.
Expressed through these vectors,
2
X
(λ) (λ)
µ (k)ν (k) = −gµν + ηµ ην − k̃µ k̃ν .
λ=1

2. Equal-time commutators of E and B

Compute the equal-time commutators

(a) [Ei (x, t), Ej (y, t)],

(b) [Ei (x, t), Bj (y, t)], and
(c) [Bi (x, t), Bj (y, t)]

using both of the equal-time commutators

[Ai (x, t), Ej (y, t)] = −iδij δ 3 (x − y)

and
tr
[Ai (x, t), Ej (y, t)] = −iδij (x − y)

and show with explicit calculation that the modification does not affect
the physically observable fields.

Solution

(a) The commutator of the electric fields is trivial; because Êi = −π̂i ,
h i  
Êi (x, t), Êj (y, t) = π̂i (x, t), π̂j (y, t) = 0

according to the quantisation condition.

(b) Let us turn now to the commutator of an electric and a magnetic
field, and denote derivatives w.r.t. the y coordinates as ∂ (y) , ∇(y) ,
etc.. Inserting B̂ = ∇ × Â, we find for the “reg.” case
h i h i
Êi (x, t), B̂j (y, t) = − π̂i (x, t), (∇(y) × Â)j (y, t)
h i
(y) (y)
= −jkl ∂k π̂i (x, t), Âl (y, t) = iδil jkl ∂k δ 3 (x − y)

133
 d3 k ik(x−y)
Z
(y) (y)
= iijk ∂k δ 3 (x − y) = iijk ∂k e
(2π)3
d3 k
Z
= −ijk kk eik(x−y)
(2π)3
while for the “trans.” case
h i h i
Êi (x, t), B̂j (y, t) = − π̂i (x, t), (∇(y) × Â(y, t))j
h i
(y) (y) (tr)
= −jkl ∂k π̂i (x, t), Âl (y, t) = ijkl ∂k δil (x − y)
d3 k
Z  
ki kl (y)
= ijkl 3
δil − 2 ∂k eik(x−y)
(2π) k
3
Z  
d k ki kl kk
= −jkl δ k
il k − eik(x−y)
(2π)3 k2
d3 k
Z
= −ijk kk eik(x−y) ,
(2π)3
identical to the “reg.” case, as advertised. In going from the
second-to-last to the last line we have used that the product of
a symmetric combination (kj kk ) with the anti-symmetric Levi-
Civita tensor (the jkl ) vanishes.
(c) Let us finally calculate the commutator of the magnetic fields.
h i
B̂i (x, t), B̂j (y, t)
h i
= − (∇(y) × Â(x, t))i (∇(y) × Â(y, t))j
h i
(x) (y)
= ikl imn ∂k ∂m Âl (x, t), Ân (y, t) = 0

because of the quantisation condition on the components of the

vector potential.

3. Momentum Operator P̂ µ
The four-momentum operator is given by

P̂ µ = (Ĥ, P̂ ),

where
Z
1 h i
Ĥ = − d3 x π̂ µ π̂µ + ∇µ ∇µ
2
Z h i
P̂ i = − d3 x ∂t µ ∂ i µ

are the Hamilton and ith component of the three-momentum operator,

respectively.

134
 To show that this represents indeed the right structure, show that
[P̂ µ , Aν ] = −i∂ µ Aν .

Hint: You may want to use the Lorentz gauge with α = 1

(the Feynman gauge) to keep your calculation simple.

Solution
Z
1 h i
Ĥ = − d3 x π̂ µ π̂µ + ∇µ ∇µ
2
Z h i
P̂ i = − d3 x ∂t µ ∂ i µ

Let us start with the commutator of the Hamiltonian and the vector
potential,
Z
ν 1 h i
[Ĥ,  ] = − d3 y π̂ µ (y)π̂µ (y) + ∇µ (y)∇µ (y), Âν (x)
2
Z
1 n h i h i o
= − d3 y π̂ µ (y) π̂µ (y), Âν (x) + π̂ µ (y), Âν (x) π̂µ (y)
2
Z
i
d3 yδ(x − y) π̂ µ (y)gµν + g µν π̂µ (y) = iπ̂ ν (x) = −i∂ 0 Âν (x) ,

=
2
and turn now to the ith component of the momentum operator. To
use this let us remind ourselves that
π µ = F µ0 − αg µ0 (∂ · A) = ∂ µ A0 − ∂ 0 Aµ − αg µ0 (∂ · A)
This suggests that, in the commutator, we can replace ∂0 Aν with −π ν ,
and we find
Z h i
[P̂ i , Âν ] = − d3 y (∂y0 µ (y))(∂ i µ (y)), Âν (x)
Z h i
= + d3 y π̂ µ (y), Âν (x) ∂ i µ (y)
Z
µν
= −ig d3 yδ 3 (x − y)∂ i µ (y) = −i∂ i Aν (x)

Putting it all together proves that

[P̂ µ , Aν ] = −i∂ µ Aν .
and that therefore P̂ µ indeed is the momentum operator.
4. ∗∗ Casimir Effect
Consider the quantization of the electromagentic field in space between
two parallel large square plates of size L located at z = 0 and z = a.
The plates are perfect conductors.

135
(a) Find a general solution for the vector potential inside the capac-
itor made by the two plates, ignore the effect of the limited size
L.
(b) Quantise the electromagentic field.
(c) Find the Hamiltonian and show that the vacuum energy is given
by
" ∞ r #
L2
Z
d2 k X  nπ 2 r
E= 2 2 2
k1 + k2 + 2 2
+ k1 + k2 .
2 (2π)2 a
n=1

(d) Define the quantity

E − E0
∆ε = ,
L2
the energy difference per unit area in the presence and absence of
the plates. This quantity is divergent and needs to be regularised;
we achieve this by introducing a function f (k), for example

f (k) = Θ(Λ − k)

with a cut-off parameter Λ for the high-momentum modes of the

field. Calculate the attractive force between the plates.

Hint: You may want to use the Coulomb gauge.

Solution

(a) Let us decompose the spatial components of the vector potential

into a part A⊥ perpendicular to the plates (i.e. along the z-axis),
and a parallel part Ak . In Coulomb gauge, A0 = 0 and ∇ · A, the
electrical field is given by
∂A
E=− .
∂t
Since the plates are ideal conductors, the parallel component of
the electric field and the normal component of the magnetic fields
vanish on the plates,
∂Ak 
 

E k  = −  =0
z=0,a ∂t z=0,a


Bz  = 0.
z=0,a

Keeping in mind that the vector potential satisfies the wave equa-
tion
∂t2 − ∇2 A = 0 ,

136
we assume a solution factorising into the form

A = F (t, x, y) [Z1 (z)e1 + Z2 (z)e2 + Z3 (z)e3 ]

with unit polarisation vectors e1,2,3 , where 1,2 live in the xy-
plane and 3 = (0, 0, 1). pointing into the x, y, and z-direction.
We therefore arrive at

0 = ∂t2 − ∇2 {F (t, x, y) [Z1 (z)e1 + Z2 (z)e2 + Z3 (z)e3 ]}

[Z1 (z)e1 + Z2 (z)e2 + Z3 (z)e3 ] ∂t2 − ∂x2 − ∂y2 F (t, x, y)

=
+∂z2 [Z1 (z)e1 + Z2 (z)e2 + Z3 (z)e3 ] F (t, x, y)

A typical solution for this equation demands that the derivatives

lead to a constant, let’s call it k32 , times the functions, i.e.

k32 Zi (z) = ∂z2 Zi (z)

k32 F (t, x, y) = ∂t2 − ∂x2 − ∂y2 F (t, x, y) .

A solution for the Zi is given by

Zi = ai sin(k3 z) + bi cos(k3 z)

and the boundary conditions demand that b1 = b2 = 0 and k3 =

nπ/a with n ∈ {0, 1, 2, . . . }. For the function F we make the
ansatz
F = exp [−iωt + ik1 x + ik2 y]
with r  nπ 2
ω = ±ωk,n = ± k12 + k22 + .
a
The Coulomb gauge condition demands that
nπ
ia1 k1 + ia2 k2 − b3 = 0 ,
a
and we see that for n 6= 0 we can choose two independent a1 and
a2 , translating into two independent polarsiation in the xy-plane,
while for n = 0 we only have a mode along the z axis. This means
we have particular solutions of the form
h nπz nπz i
A = F k sin + b3 3 cos
a a
and the general solution reads
∞ Z 2 
X d2 k X
A =
(2π)2 2ωk,n
n=1 λ=1

137
 aλ (k1 , k2 , n)e−iωk,n t+ik1 x+ik2 y

(λ) nπz (λ) nπz 
× k (k, n) sin + 3 (k, n) cos
a a
a†λ (k1 , k2 , n)eiωk,n t−ik1 x−ik2 y


∗(λ) nπz (λ) nπz 
× k (k, n) sin + 3 (k, n) cos
a a
d2 k
Z 
+ a3 (k1 , k2 , 0)e−iωk,n t+ik1 x+ik2 y
(2π)2 2ωk,0

† iωk,n t−ik1 x−ik2 y
+ a3 (k1 , k2 , 0)e 3 .

(b) Quantization now proceeds similar to the case of the photon

field in the absence of the plates. We promote the amplitude
factors aλ (k1 , k2 , n) and their complex conjugates to operators
âλ (k1 , k2 , n) their Hermitian conjugates and demand that the
operators have suitable commutator relations. In the case of the
setup at hand, all commutators vanish, apart from

âλ (k1 , k2 , n), âλ0 (k10 , k20 , n0 )

 

= (2π)2 2ωk, n δ(k1 − k10 )δ(k2 − k20 )δnn0 δλλ0

It is worth noting that while the form is different from, e.g.,

Eq. (260),
h i
â(k, λ), ↠(q, κ) = (2π)3 2k0 δ 3 (k − q) δλκ ,

and this difference stems from the fact that the momentum in
z-direction only takes discrete values, encoded in the n and n0 .
(c) As before the (not! normal-ordered, therefore a factor of 1/2
in front) Hamiltonian is given by a sum over all modes, and we
arrive at
(∞ Z 2 
1 X dk1 dk2 X †
Ĥ = âλ (k, k2 , n)âλ (k, k2 , n)
2 (2π)2 2ωk, n
n=1 λ=1

†
+ âλ (k, k2 , n)âλ (k, k2 , n)
Z 
dk1 dk2
+ ↠(k, k2 , 0)â3 (k, k2 , 0)
(2π)2 2ωk, 0 3

†
+ â3 (k, k2 , 0)â3 (k, k2 , 0)

138
 For the calculation of the ground-state energy we realise that â |0i
vanishes and that we use the commutator for terms like h0| â↠|0i.
We arrive at

E = h0| Ĥ |0i
*  (∞ Z 2  
1 X dk1 dk2 †
X
= 0 ωk, n âλ (k, k2 , n), âλ (k, k2 , n)

2 (2π)2 2ωk, n
n=1 λ=1
Z   
dk1 dk2 † 
+ ωk, 0 â3 (k , k2 , 0)â3 (k, k 2 , 0) 0
(2π)2 2ωk, 0 
*  (∞ Z 2
1 X dk1 dk2 X
= 0 ω (2π)2 2ωk, n δ 2 (0)

k, n
2 (2π)2 2ωk, n
n=1 λ=1
Z  
dk1 dk2 2 2

+ 2
ωk, 0 (2π) 2ωk, 0 δ (0)  0
(2π) 2ωk, 0
Z " ∞ #
1 2 2
X
= d kδ (0) 2 ωk, n + ωk,0
2
n=1

This leaves us with a first task to evaluate δ 2 (0). Remember-

ing that this comes from the integration over the xy-plane (and
ignoring boundary effects) we have

L2
Z
dxdy ik1 x+ik2 y 
δ 2 (0) = e = ,
(2π)2 (2π)2

kk =0

the volume of the plane, normalised by factors of 2π. Therefore

the energy is given by
" ∞ r #
L2
Z
d2 k X  nπ 2 r
E= 2 2 2
k1 + k2 + 2
+ k1 + k2 2
2 (2π)2 a
n=1

(d) To calculate the difference in energy to the system without the

plates we have to take the continuum limit of the summation
over the discrete modes in the expression for the ground state,
and ignore the “0” mode. For the ground state energy of the
system in the volume V = L2 a, but without the plates, we thus
have
L2 a d3 k
Z  q 
E0 = 2 2 2
2 k1 + k2 + k3
2 (2π)3
Z∞ r
d2 k  nπ 
Z
2 2 2
= L dn k1 + k2 + ,
(2π)2 a
0

where we used that k3 = nπ/a with n now taking continuously.

139
The normalised difference reads
E − E0
∆ε =
L2
∞ r
" r
d2 k 1
Z
2 + k 2 + nπ
X  2
= k 2 + k 2 + k
1 2 1 2
(2π)2 2 a
n=1
Z∞ r

 nπ 
− dn k12 + k22 + 
a
0
∞ Z∞ p
 
2 ∞ p
1√
Z
π X
= 3
du  u+ u + n2 − dn u + n2  ,
4a 2
0 n=1 0

after substituting u = a(k12 +k22 )/π 2 .

Including the regularisation
function f (k) we have
Z∞  √  X  √ 

2 ∞ p
π √
∆ε = du  1 uf π u + u + n 2f
π u
4a3 2 a a
0 n=1

Z∞ p  √ 

π u 
− dn u + n2 f
a
0
Z∞
 
∞
π2  1 X
= F (0) + F (n) − 2 F (n) ,
4a3 2
n=1 0
where we have introduced
Z∞ p  √ 
2
π u
F (n) = du u + n f .
a
0

To evaluate this we use the somewhat obscure Euler-McLaurin

formula and by realising that F (n) and all of its derivative F (k) (n)
vanish for n → ∞, we arrive at
∞ 
( )
π2 1 X Bk (k−1)
∆ε = − F (0) − F (0)
4a3 2 k!
k=1
π2
  
1 B2 0 B4 000
= − F (0) + B1 F (0) − F (0) + F (0) + . . .
4a3 2 2! 4!
π 2 B2 0
 
B4 000
= − 3 F (0) + F (0) + . . . ,
4a 2! 4!
where we have used that B1 = 1/2 and that all other odd Bernoulli
numbers are zero. A quick calculation reveals that
−dn3 
 
0 dF (n) 
F (0) = = =0
dn  dn 
n=0 n=0

140
 d3 F (n) 

000
F (0) = = −4
dn3 n=0

and with B2 = 1/6 and B4 = −1/30 we arrive at

π2 π2
 
4
∆ε ≈ − 3 =−
4a 30 ∗ 4! 720a3

The vacuum energy of the lectromagnetic field between the two

conducting plates therefore produces a weak force f ,

∂∆ε π3
f =− = ,
∂a 240a4
which for a = 1µm and L = 1 cm is approximately f ≈ 10−8 N.
This is the Casimir effect, measured for the first time in 1958.
17

17
It states that the difference of a sum and an integral of the same function can be
expressed by the Bernoulli numbers bk and (multiple) derivatives of the function as
m Zm
X X Bk h (k−1) i
f (n) − dxf (x) = ∞ f (m) − f (k−1) (n) .
n
k!
n k=1

141
7 Time-Ordered Products
Until now we have quantised various free elementary fields: real and complex
scalars, Dirac-spinors, and the vector fields of electrodynamics. The result-
ing structure in each case can be condensed into a sequence of algorithmic
steps, which, starting from a Lagrange density, resulted in the expansion
of field operators as products of plane waves and creation and annihilation
operators, and we succeeded in expressing “static” global quantities such as
the Hamilton or charge operators through the latter.
This implies an underlying causal structure if the theory: in non-relativistic
field theories, which we do not discuss here, evolution is forward in time,
and for the analysis of causality it is usually sufficient to concentrate on
the positive–energy solutions only. This changes in relativistic field the-
ories, where both forward and backward evolution, and therefore positive
and negative energy solutions, are included. It is important to realise the
interplay with causality requirements of the theory - the simplest one is that
the commutator of two fields must vanish for space-like distances. It turns
out that in non-relativistic theory this cannot be achieved, which actually
should not come as a surprise. If you do not embed relativity in your formal-
ism you cannot expect to obtain relativisticly sensible results from it. In the
relativistic field theories we have discussed here, the positive and negative
energy solutions could be arranged such that the commutator of two fields
vanishes outside the light-cone, i.e. for space-like distances, but remains
finite inside the light-cone.
In this chapter we will build further on the logic and discussion started in
Sec. 4.3, and we will analyse the propagation of particles. This first step
towards a dynamic picture of quantised field theories is deeply connected
to the notion of Green’s functions, which will return to us in this chapter,
and called propagators. They will fortify the notion of the negative-energy
solutions as anti-particles, which travel backwards in time, with opposite
charge. We will also see how the wrong commutator (or anti-commutator)
for a given statistics (Bose–Einstein vs. Fermi–Dirac) destroys the causality
structure of the theory.
In this chapter of the lecture notes I have amalgamated time-ordered prod-
ucts for the three Quantum Field Theories we discussed so far – for some
additional reading, I’d like to refer you to Sections 3.5, 4.3, and Chapter 5
of Hatfield’s book [3], or maybe take a look at Sections 2.4, 3.5, and possibly
4.8 of Peskin & Schroeder [1].

142
7.1 Greens Functions: Non-Relativistic Quantum Mechanics
What is the Green’s Function? Consider the time-dependent Schrödinger
equation for a point particle,
∇2
   
i∂ i∂
− Ĥ |ψ(t)i = + − V̂ |ψ(t)i = 0 . (284)
∂t ∂t 2m
It is formally solved through the introduction of Green’s function, G(t, x; t0 , x0 ):
Z
ψ(t, x) = hx|ψ(t)i = d3 x0 G(t, x; t0 , x0 )ψ(t0 , x0 ) . (285)

The interpretation is clear: the wave function ψ(t, x) at time t and position
x in position space is constructed as the superposition of all wave functions
at all positions x0 at an earlier time t0 , and the Green’s function parame-
terises the “strength” of the connection. Because it this has been couched in
the framework of non-relativistic Quantum Mechanics the maximal velocity
of causation (speed of light in relativistic physics) is infinite, and the connec-
tion is instantaneous. The Green’s function G is also called the (retarded)
propagator.

Construction of the Green’s Function The interpretation of the Green’s

function above suggests that it is the solution of 18
 
i∂
− Ĥ G(t, x; t0 , x0 ) = δ(t − t0 )δ 3 (x − x0 ) , (286)
∂t

with the boundary condition that it vanishes for t0 > t. This allows to
rewrite it as

G(t, x; t0 , x0 ) = K(t, x; t0 , x0 ) Θ(t − t0 ) , (287)

where K(t, x; t0 , x0 ) is the transition amplitude

hx, t|x0 , t0 i = hx| Û (t, t0 ) x0

 
(288)

and
Zt
 
h i
Û (t, t0 ) = exp −i dτ Ĥ(τ ) −→ exp −iĤ(t − t0 ) (289)
t0

is the unitary time-evolution operator, well-known from Quantum Mechan-

ics, which reduces to the second expression if Ĥ does not explicitly depend
on time.
18
You may be reminded of the definition of Green’s function in (classical) electrodynam-
ics; in fact for each differential operator O the corresponding Green’s function is defined
as solution of OG = δ, with a product of δ-functions of pairs of arguments.

143
Free-Particle Propagator: Direct Solution in Momentum Space
As a simple example consider a free point particle in Quantum Mechanics.
Its propagator (Green’s function) G0 is a solution to
 
i∂
− Ĥ0 G0 (t, x; t0 , x0 ) =
∂t
∇2
 
i∂
+ G0 (t, x; t0 , x0 ) = δ(t − t0 )δ 3 (x − x0 ) . (290)
∂t 2m
A simple way to solve this equation is by Fourier-transforming on, resulting
in
!
p2
ω− G0 (ω, p) = 1 (291)
2m

and therefore
1
G0 (ω, p) = . (292)
ω − p2 /2m

Position Space The back-transformation into position space is formally

achieved by

d3 p dω exp[ip · (x − x0 ) − iω(t − t0 )]
Z
0 0
G0 (t, x; t , x ) = . (293)
(2π)3 2π ω − p2 /(2m)

This however does come with two interesting problems:

1. the integral obviously diverges for ω = p2 /2m, and

2. the propagator G0 does not satisfy the (causal) boundary condition,

i.e. it does not vanish for t0 > t.

There is a way, however, to solve simultaneously both problems. And this

is how it works, we deform the energy integration by shifting the pole on
ω = p2 /2m by a minimal amount of −i+ into the imaginary plane – here
and in the following, + represents an infinitesimal positive number. This
yields the retarded Greens function

d3 p dω exp[ip · (x − x0 ) − iω(t − t0 )]
Z
(R) 0 0
G0 (t, x; t , x ) = . (294)
(2π)3 2π ω − p2 /(2m) − i+

Cauchy’s theorem asserts that the energy integral yields

" #
dω exp[−iω(t − t0 )] ip2 (t − t0 )
Z
0
= Θ(t − t ) exp − . (295)
2π ω − p2 /(2m) − i+ 2m

144
Therefore the overall result is given by
" #
d3 p ip2 (t − t0 )
Z
(R) 0 0 0 0
G0 (t, x; t , x ) = Θ(t − t ) exp − + ip · (x − x )
(2π)3 2m
3 Z
ip2i (t − t0 )
 
Y dpi 0 0
= Θ(t − t ) exp − + ipi (xi − xi )
(2π)3 2m
i=1
3 Z
"
i(t − t0 ) m(xi − x0i ) 2
 
0
Y dp i
= Θ(t − t ) exp − pi −
2π 2m t − t0
i=1
im(xi − x0i )2

+
2(t − t0 )
3
s
im(xi − x0i )2
 
0
Y 1 2mπ
= Θ(t − t ) exp
2π i(t − t0 ) 2(t − t0 )
i=1
s 3 " 3
#
0 −im im X
0 2
= Θ(t − t ) exp (xi − xi )
2π(t − t0 ) 2(t − t0 )
i=1
3
im(x − x0 )2
  
0 −im 2
= Θ(t − t ) exp . (296)
2π(t − t0 ) 2(t − t0 )

We have made use of the fact that we can write this integral as a product
of three integrals, one for each spatial component of p, then completed the
squares in each component of p, rendering this a product of three Gaussian
integrals.

Propagator from Position Space Transition Amplitude An alter-

native way to arrive at the same result rests on the identification of the
propagator with the transition amplitude times the boundary condition,
Eq. (287). Using the free particle Hamiltonian in Eq. (288) we have
*  " # +
0 0
 ip̂2 
0  0
K(t, x; t , x ) = x exp − (t − t )  x

 2m 
*  " #  +
d3 p d3 p0 ip̂2
Z    
0 0 0  0
= p exp ip · x exp − (t − t ) exp −ip · x  p

(2π)3 (2π)3  2m 
" #
d3 p d3 p0 ip2
Z       
 0 0 0 0
= p  p exp ip · x exp −
 (t − t ) exp −ip · x
(2π)3 (2π)3 2m
" #
d3 p ip2
Z
= exp − (t − t0 ) + ip · (x − x0 ) . (297)
(2π)3 2m

145
In going from the first to the second line we have use the fact that momentum
and position space kets are connected through a Fourier transform,

d3 p −ip·x  
Z
|xi = e p , (298)
(2π)3
2
and in going to the third line we have replaced   the operator p̂ with the
eigenvalues p corresponding to its eigenkets p . Of course, they form an
ortho-normal base, such that their scalar product is a δ function,

 0
p p i = (2π)3 δ 3 (p − p0 ) . (299)

Including the Θ-function which connects the transition amplitude with the
restarted propagator, this is exactly the same result we already obtained
with the more direct method in momentum space.

7.2 Propagators in Quantum Field Theory: Scalar Fields

Scalar Theories The Green’s function for the Klein-Gordon equation,
i.e. the propagator of the free scalar field is defined by

 + m2 G0 (x, x0 ) = iδ 4 (x − x0 ) .


(300)

Fourier transformation results in

−i
−p2 + m2 G0 (p) = i −→


G0 (p) = . (301)
p2 − m2

As before, the Green’s function exhibits a pole when the energy-momentum

relation is satisfied, that is, when the particle goes “on its mass-shell” or
“on-shell”, and as before this is repaired by shifting the pole in the complex
plane by an infinitesimally small amount of i+ away from the real axis.

Time-Ordered Products and Green’s Functions In what follows we

will show that the Green’s function is also given by the vacuum expectation
value of a time-ordered product,

G0 (x, x0 ) = h0| T φ̂(x)φ̂(x0 ) |0i

= Θ(t − t0 ) h0| φ̂(x)φ̂(x0 ) |0i + Θ(t0 − t) h0| φ̂(x0 )φ̂(x) |0i .

(302)
When expanding these products in terms of the creation and annihilation
operators, it is worth noting that â |0i = h0| ↠= 0 and that we therefore

146
can replace products â↠with the commutators of the two operators, when
they are sandwiched between vacuum states. This leads to

G0 (x, x0 ) = h0| T φ̂(x)φ̂(x0 ) |0i

d3 k d3 k 0
Z
0 0 0
= Θ(t − t ) 3 3 0 e−ik·x+ik ·x h0| â(k)↠(k 0 ) |0i
(2π )2k0 (2π )2k0
d3 k d3 k 0
Z
0 0
+ Θ(t0 − t) 3 3 0 e+ik·x−ik ·x h0| â(k 0 )↠(k) |0i
(2π )2k0 (2π )2k0
d3 k d3 k 0
Z h i
−ik·x+ik0 ·x0
= Θ(t − t0 ) e h0| â(k), â † 0
(k ) |0i
(2π 3 )2k0 (2π 3 )2k00
d3 k d3 k 0
Z h i
+ik·x−ik0 ·x0
+ Θ(t0 − t) e h0| â(k 0
), â †
(k) |0i
(2π 3 )2k0 (2π 3 )2k00
d3 k
Z h i
0 −ik·(x−x0 ) 0 +ik·(x−x0 )
= Θ(t − t ) e + Θ(t − t) e
(2π 3 )2k0
= Θ(t − t0 ) ∆+ (x − x0 ) + Θ(t0 − t) ∆− (x − x0 ) . (303)

The similarity with the vacuum expectation value is striking – it is given by

d3 k
Z h i
−ik·(x−x0 ) +ik·(x−x0 )
∆(x − y) = e − e
(2π 3 )2k0
= ∆+ (x − x0 ) − ∆− (x − x0 ) , (304)

cf. Eq. (163), where in both cases the sign indicates whether the energies,
i.e. the k0 come with the correct or wrong sign for a wave the evolves
in positive or negative time direction. This means that the propagator is
composed of two components: a +-component of forward propagation, and
a − component of backward propagation of a particle with four-momentum
k.

i+ -Prescription To finish the discussion of how to arrive at the correct

Green’s function, let us remember the property
dω −ieiωt
Z
Θ(t) = lim . (305)
→0+ 2π ω − i
We plug this into the result for the time-ordered product above and close the
0 0
p for t − t > 0 and t − t < 0.
contour integral above or below the real k0 -axis
2
We also introduce a “dummy” energy ωk = k + m2 which we will identify
with the “proper” k0 at some convenient point of the calculation. With all
these steps we arrive at
dωd3 k −i h i(ω−k0 )(t−t0 )+ik(x−x0 )
Z
0
G(x, x ) = lim e
→0+ (2π)4 2k0 ω + i
0 0
i
+ e−i(ω−k0 )(t−t )−ik(x−x )

147
 0
dωk d3 k −ie−ik·(x−x )
Z  
1 1
= lim +
→0+ (2π)4 2k0 k0 − ωk − i k0 + ωk − i
0
dωk d3 k −ie−ik·(x−x )
Z
= lim
→0+ (2π)4 k02 − ωk2 + i
0
dk0 d3 k −ie−ik·(x−x )
Z
= lim
→0+ (2π)4 k02 − (k 2 + m2 ) + i
d4 k −ik·(x−x0 ) −i
Z
= 4
e
(2π) k − m2 + i+
2

d4 k −ik·(x−x0 )
Z
= e ∆F (k) = ∆F (x − x0 ) . (306)
(2π)4
This is obviously the Fourier transform of our propagator from Eq. (301),
and it confirms that indeed propagators are time-ordered products. It is
also called the Feynman propagator of the theory.

7.3 Fermion Propagator

Direct Solution As before for the Klein-Gordon equation, the propagator
for the free Dirac field is defined by

(i∂/ − m) G0 (x, x0 ) = iδ 4 (x − x0 ) , (307)

or, in momentum space,

i p/ + m
/ − m) G0 (p) = i
(p −→ G0 (p) = = i 2 . (308)
p/ − m p − m2

In the last step we have used that p/p/ = p2 . There are a couple of things
worth noting of this propagator. As before, it exhibits a pole for on-shell
particles, where p2 = m2 , and, as before, we will cure this by shifting the
pole in the complex plane by i+ . This is in complete analogy to the case
of scalar particles. In addition we realise that the numerator, (p/ + m),
represents a matrix in Dirac space. This is not a surprise, as the propagator
connects two Dirac spinors and their components. What is structurally more
interesting is that this matrix is the completeness relation for the u-spinors
from Eq. (215), and we will see the emergence of analogous terms later
when we discuss the propagator of the photon field. However, to build more
confidence into our interpretation of the propagator we will now check if we
can recover it as a time-ordered product of two spinor fields.

Time-Ordered Product The starting point to building a time-ordered

product for fermions is to construct the transition amplitude for a positive-
energy fermion to move from x to y, naively

hf (+) (y)|f (+) (x)i = h0| ψ̂(y)ψ̂ † (x) |0i . (309)

148
But we must also include the opposite case of a negative-energy fermion to
go from y to x, taking into account Fermi statistics. Making time-ordering
explicit through Θ-functions we therefore arrive at
iSF (y, x)γ 0 = h0| ψ̂(y)ψ̂ † (x) |0i Θ(y0 − x0 ) − h0| ψ̂ † (x)ψ̂(y) |0i Θ(x0 − y0 ) ,
(310)
or, in a more compact form

iSF (y, x) = h0| T ψ̂(y)ψ̄ˆ(x) |0i . (311)

Since ψ̂ and ψ̄ˆ are spinors, SF is a matrix in spinor space, as already antici-
pated. Let us now include the expansion of the fermion fields in place waves
and creation and annihilation operators. Making spinor indices explicit, us-
ing the fact that b̂ |0i = dˆ|0i = 0 and 0̄b̂† = h0| dˆ† , and taking into account
ˆ = [b̂† , dˆ† ] = 0, which makes their products vanish, we arrive at
that [b̂, d]

iSF (y, x)βα = h0| T ψ̂β (y)ψ̄ˆα (x) |0i

* Z 2
 d3 k d3 q X
= 0 
3 3
 (2π) 2k0 (2π) 2q0 i,j=1
h
e−ik·y+iq·x b̂i (k)b̂†j (q)uβ (k)ū(j)
(i)
α (q)Θ(y0 − x0 )

 +
i 
− e+ik·y−iq·x dˆj (q)dˆ†i (k)vβ (k)v̄α(j) (q)Θ(x0 − y0 )  0
(i)


* Z 2 h
d3 k X (i)
e−ik·(y−x) uβ (k)ū(j)

= 0  3 2k α (k)Θ(y0 − x0 )
 (2π) 0 i,j=1
 +
i 
(i)
− e+ik·(y−x) vβ (k)v̄α(j) (k)Θ(x0 − y0 )  0

d3 k h −ik·(y−x)
 Z

= 0  e (k/ + m)βα Θ(y0 − x0 )
(2π)3 2k0
 +
i 
− e+ik·(y−x) (k/ − m)βα Θ(x0 − y0 )  0 . (312)


Repeating the same steps of replacing the Θ functions with integrals, suit-
ably closing the contours and keeping track of the relative signs in the re-
placement of the “dummy” energy with the real energy, we arrive at

d4 k −ik·(y−x)
Z  
k/ + m
iSF (y, x)βα = e i 2 , (313)
(2π)4 k − m2 + i+ βα

149
the Fourier transform of Eq. (311), modified by the now familiar i prescrip-
tion.

7.4 Photon Propagator

Feynman (Lorentz) gauge As before, the E.o.M. for the field will pro-
vide the kernel for the Green’s function. In the case of the free electromag-
netic field we have to realise that
ˆ the propagator will have two Lorentz indices, to become a matrix in
Minkowski space. This is necessary, because it connects two photon
fields and their components, which are labelled by Lorentz indices.
ˆ we need to cast the homogeneous (j = 0 )Maxwell’s equations for the
vector potential from Eq. (265) into a suitable form such that it has
two Lorentz indices that can be contracted with the two indices from
the propagator.
We therefore arrive at
[∂ ρ ∂ρ gµν − (1 − α)∂ν ∂µ ] Gνρ 0 4 0 ρ
0 (x, x ) = iδ (x − x )gµ , (314)
and Fourier transformation results in
p gµν − (1 − α)pµ pν Gνρ
 2  ρ
0 (p) = −igµ , (315)
or
igµρ
 
(1 − α)pµ pν
gµν − Gνρ
0 (p) = − 2 . (316)
p2 p
To arrive at a solution we realise that there are only two possible tensors
without any mass dimension and two Lorentz indices and make the ansatz
ν ρ
g νρ − κ pp2p
Gνρ
0 (p) = −i . (317)
p2
We solve this by realising that this implies that
pν pρ
  
(1 − α)pµ pν νρ
gµν − g −κ 2 = gµρ (318)
p2 p
and therefore κ = (1 − α)/α. Therefore the photon propagator, including
the i+ term reads
1 − α kµ kν
g µν −
Dµν (k) = −i α k2 , (319)
2
k + i+

which in Feynman gauge reduces to the simpler form

−ig µν
Dµν (k) = . (320)
k 2 + i+

150
Time-Ordered Product To arrive at the same expression using time-
ordered products we employ the Feynman gauge from the beginning, where
the completeness relation for the polarisation vectors is given by
3
X
∗(λ)
(λ)
µ (k)ν (k) = gµν (321)
λ=0

Using the by now usual â |0i = 0 relation allows to simplify the time-ordered
product and the propagator reads

Dµν (x, y) = −i h0| T Aµ (x)Aν (y) |0i

d3 k d3 q
Z
=i
(2π)3 2k0 (2π)3 2q0
X3 D h
0  Θ(x0 − y0 )e−ik·x+iq·y âλ (k)â†κ (q)(λ) ∗(κ)
µ (k)ν (q)

λ,κ=0
i E
+ Θ(y0 − x0 )e+ik·x−iq·y âκ (q)â†λ (k)∗(λ) (k)(κ)
(q) 0

µ ν
3
d3 k X D h
Z
= −i 0  Θ(x0 − y0 )e−ik·(x−y) (λ) ∗(λ)
µ (k)ν (k)
(2π)3 2k0
λ=0
i E
+ Θ(y0 − x0 )e+ik·(x−y) ∗(λ) (λ)
µ (k)ν (k)  0


d3 k h
Z i
−ik·(x−y) +ik·(x−y)
= − igµν Θ(x0 − y0 )e + Θ(y0 − x0 )e
(2π)3 2k0
= − igµν [Θ(x0 − y0 )∆+ (x − y) + Θ(y0 − x0 )∆− (x − y)]
d4 k e−ik·(x−y)
Z
= − igµν . (322)
(2π)4 k 2 + i+

This is, of course, the Fourier transform of the propagator of Eq. (320)

151
7.5 Problems & Solutions
1. Green’s function for a free particle in Quantum Mechanics
Consider, once again, the free Hamiltonian, Ĥ0 = p̂2 /(2m) with eigen-
states labelled by their momentum, |pi.

(a) give the (time-dependent) Schrödinger picture wave functions of

the eigenstates in position space,

ψp (x, t) = hx|pi

(Hint: don’t forget the time dependence!)

(b) show that the Green’s function is given by

d3 p
Z
G(x, t; , x0 , t0 ) = Θ(t − t0 ) ψp (x, t)ψp∗ (x0 , t0 )
(2π)3

(c) show, by Fourier transformation from times t to energies ω, that

for time-independent Hamiltonians the Green’s function can be
expressed by the energy eigenstates En as
Z∞
0 0 dω iω(t−t0 ) X ψn∗ (x)ψn (x0 )
G(x, t; , x , t ) = e
2π ω − En − i+
−∞ En

when discrete energy eigenstates are assumed.

(d) (this is “expert level”) using principal value decomposition
1 1
lim +
= P + iπδ(x)
+ →0 x − i x
show that the density of states ρ(ω) is given by the imaginary
part of the retarded Green’s function

Solution

(a) To construct the wave function add the time evolution to a set
R 3 basis kets |pi = |p(t = 0)i and insert a 1 in
of momentum-space
the form of 1 = d p|pihp|
*  
Zt

 +
0 0
 
ψp (x, t) = hx|p(t)i = x exp −i dt Ĥ(t ) p(0)
 
0
*  " # + *  " # +
   p̂2 
= x exp −iEp t  p(0) = x exp −i t  p(0)
  
   2m 

152
 *  +*  " # +
2
d3 p0 p̂
Z   
 0
= x p p0 exp −i t  p(0)
 
(2π)3   2m 
d3 p0 ix·p0 −ip2 t/(2m)
Z
2
= e e (2π)3 δ 3 (p0 − p) = eix·p−ip t/(2m)
(2π)3

(b) Inserting the wave-functions above,

d3 p
Z
0 0 0
G(x, t; , x , t ) = Θ(t − t ) ψ(x, t)ψ ∗ (x0 , t0 )
(2π)3
d3 p ip·(x−x0 )−ip2 (t−t0 )/(2m)
Z
0
= Θ(t − t ) e ,
(2π)3
exactly the form found in the lecture.
(c) For discrete energy eigenstates, the wave functions can be labelled
as ψn (x, t), and the integral over momentum eigenstates collapses
to a sum over energy eigenstates – in the end this is only a change
of basis from one orthonormal set to another. In this case,
X
G(x, t; , x0 , t0 ) = Θ(t − t0 ) ψn (x, t)ψn∗ (x0 , t0 )
En

For time-independent Hamiltonians, the time-dependence of the

basis kets factorises such that

ψn (x, t) = e−iEn t ψn (x)

and therefore
0
Θ(t − t0 )ψn (x, t)ψn∗ (x0 , t0 ) = Θ(t − t0 )e−iEn (t−t ) ψn (x)ψn∗ (x0 )

Simple Fourier transform with respect to the time-difference, and

using the trick from the lecture results in
+∞
Z
dω 0 0 1
Θ(t − t0 )eiω(t−t ) e−iEn (t−t ) = .
2π ω − En − i+
−∞

This allows us to recover the expression quoted above.

(d) Assuming an isolated system where the Hamiltonian is time-
independent the retarded Green function is obtained from Fourier
transforming with respect to time, assuming the correct (i.e., re-
tarded) time ordering:
Z∞
0 0 dω iω(t−t0 ) (ret)
G (ret)
(x, t; , x , t ) = e G (x, x0 , ω) ,
2π
−∞

153
 where
X ψ ∗ (x)ψn (x0 )
G(ret) (x, x0 , ω) = n

n
ω − En − i+
Using the prinicpal value decomposition above, we see that
h i X  ψ ∗ (x)ψn (x0 ) 
(ret) 0 n
Re G (x, x , ω) = P
n
ω − En
h i X
Im G(ret) (x, x0 , ω) = π ψn∗ (x)ψn (x0 )δ(ω − En )
n

As a consequence the (discrete) density of states ρ(ω) is given by

Z
X 1
ρ(ω) δ(ω − En ) = Im d3 xG(ret) (x, x0 , ω)
n
π

2. Feynman propagator for the Dirac Field

Use the completeness relations for the u and v spinors, the integral rep-
resentation of the Θ-function and suitable substitutions for the energy
integral to show that indeed
d4 p −ip·(y−x)
Z  
p/ + m
SF (y, x)βα = i e
(2π)4 p2 − m2 + i+ βα

Solution

ˆ † ˆ†
Puse b̂|0i = d|0i = 0 = h0|
In the following we will Pb̂ = h0|d and the
completeness relations u(p)ū(p) = (p/+m) and v(p)v̄(p) = (p/−m)
 
iSF (y, x)βα = h0|T ψβ (y)ψ̄α (x) |0i
= h0|ψ(y)ψ̄(x)|0iΘ(y0 − x0 ) − h0|ψ̄(x)ψ(y)|0iΘ(x0 − y0 )
d3 p d3 q
 Z

= 0 
(2π)3 2p0 (2π)3 2q0
2 h i
e−ip·y b̂i (p)u(i) (p) + eip·y dˆ†i (p)v (i) (p)
X
β
i,j=1
h i  
× eiq·x b̂†j (q)ū(j) (q)
+e −iq·x
dˆj (q)v̄ (j)
(q)  0 Θ(y0 − x0 )
α
d3 p d3 q
 Z

− 0 
(2π)3 2p0 (2π)3 2q0
2 h i
eiq·x b̂†j (q)ū(j) (q) + e−iq·x dˆj (q)v̄ (j) (q)
X
×
α
i,j=1
h i  
−ip·y (i) ip·y ˆ† (i)
× e b̂i (p)u (p) + e di (p)v (p)  0 Θ(x0 − y0 )
β

154
 d3 p d3 q
 Z

= 0 
(2π)3 2p0 (2π)3 2q0
2 
e−i(p·y−q·x) b̂i (p)b̂†j (q)uβ (p)ū(j)
(i)
X
α (q)Θ(y0 − x0 )
i,j=1
 
i(p·y−q·x) ˆ ˆ† (i) (j)

−e dj (q)di (p)vβ (p)v̄α (q)Θ(x0 − y0 )  0

d3 p d3 q
 Z

= 0 
(2π)3 2p0 (2π)3 2q0
2  n o
e−i(p·y−q·x) b̂i (p), b̂†j (q) uβ (p)ū(j)
(i)
X
α (q)Θ(y0 − x0 )
i,j=1
n o  
i(p·y−q·x) ˆ ˆ† (i) (j)

−e dj (q), di (p) vβ (p)v̄α (q)Θ(x0 − y0 )  0

d3 p d3 q
 Z
2q0 (2π)3 δ 3 (p − q)δij

= 0 
(2π)3 2p0 (2π)3 2q0
2 
(i)
X
e−i(p·y−q·x) uβ (p)ū(j)
α (q)Θ(y0 − x0 )
i,j=1
 
(i)
ei(p·y−q·x) vβ (p)v̄α(j) (q)Θ(x0

− − y0 )  0

d3 p
 Z 
e−ip·(y−x) (p/ + m)βα Θ(y0 − x0 )

= 0

(2π)3 2p0
i E
−eip·(y−x) (p/ − m)βα Θ(x0 − y0 )  0


Using the representation of the Θ-function,

dω eiωt
Z
Θ(t) = ,
2π ω − i+
substituting p0 = ±(Ep − ω) in the first and second term, respectively,
and flipping the sign in the second term (in the third step below),
yields
d3 p
Z h i
−ip(y−x) ip(y−x)
e (p
/ + m)Θ(y 0 − x 0 ) − e (p
/ − m)Θ(x 0 − y 0 )
(2π)3 (2Ep )

d3 p dω e−iEp (y0 −x0 )+ip(y−x)+iω(y0 −x0 )
Z Z 
0
= (p0 γ − p · γ + m)

(2π)3 (2Ep ) 2π (ω − i+ ) 
ω=Ep −p0

iE (y −x )−ip(y−x)−iω(y −x )
d3 p
Z Z
dω e p 0 0 0 0 
− (p0 γ 0 − p · γ − m)

(2π)3 (2Ep ) 2π (ω − i+ ) 
ω=Ep +p0
−ip0 (y0 −x0 )+ip(y−x) 0
 
Z 3
d p
Z
dp0 e 2p0 p0 γ − 2Ep p · γ + 2Ep m
= −
(2π)3 (2Ep ) 2π (Ep − p0 − i+ )(Ep + p0 − i+ )

155
 d4 p e−ip0 (y0 −x0 )+ip(y−x) (p/ + m) d4 p e−ip(y−x) (p/ + m)
Z Z
= =
(2π)4 Ep2 − p2 − m2 + i+ (2π)4 p2 − m2 + i+ )

In going from the secon to the third line we have put both terms onto
one denominator, replaced the integral over ω with an integral over
p0 , and we shifted the sign of the spatial integration, going from −p
to +p in the second term. In the second to last step we have first
replaced p20 = p2 + m2 and then identified Ep with p0 . Therefore, the
propagator reads

d4 p −ip·(y−x)
Z
p/ + m
iSF (y − x) = 4
e
(2π) p − m2 + i+
2

3. Propagator in general Lorentz gauge Show that the photon prop-

agator in general Lorentz gauge (arbitrary α) is given by
k k
d4 k −ik(x−y) gµν + 1−α
Z µ ν
α k2
h0|T [µ (x)Âν (y)]|0i = 4
e 2 +
(2π) k + i

Solution

Let us start with the propagator in Coulomb gauge,

iGµν (x − y) = h0|T [µ (x)Âν (y)]|0i

Plugging in the expansion of the  and realising that â(k, λ)|0i = 0

this results in

iGµν (x − y) = h0|T [µ (x)Âν (y)]|0i

2
d3 k d3 q
Z Z X
= µ (k, λ)ν (q, κ)
(2π)3 (2k0 ) (2π)3 (2q0 )
λ, κ=1
 
× 0  Θ(x0 − y0 )e−ikx+iqy â(k, λ)↠(q, κ)


 
−iqy+ikx †

+ Θ(y0 − x0 )e â(q, κ)â (k, λ)  0
2
d3 k d3 q
Z Z X
= µ (k, λ)ν (q, κ)
(2π)3 (2k0 ) (2π)3 (2q0 )
λ, κ=1
3 3
 
× −(2π) δ (k − q)(2k0 )δκλ
   
−ikx+iqy −iqy+ikx
 
× 0  Θ(x0 − y0 )e
 + Θ(y0 − x0 )e 0


156
 2
d3 k
Z X
= − 3
µ (k, λ)ν (k, λ)
(2π) (2k0 )
λ=1
 
ik(y−x) ik(x−y)
× Θ(x0 − y0 )e + Θ(y0 − x0 )e

d4 k gµν − nµ nν + k̃µ k̃ν

Z
= ,
(2π)4 k 2 + i+
where we have used the polarisation sum in Coulomb gauge from the
previous problem, and the usual trick of replacing the Θ functions with
an integral.
Obviously, the polarisation sum defines the numerator of the propa-
gator, while the denominator stems from the integral representation
of the Θ functions and the implied causality structure. Naively, we
would like to directly use the polarisation sum for the electromagnetic
field in Lorenz gauge, but this is not entirely straightforward due to
the implied gauge constraint. In Coulomb gauge we could directly
produce a set of polarisation vectors that satisfy this condition. Un-
fortunately, this is not straightforward in Lorenz gauge, because the
gauge condition ∂µ Aµ = 0 does not allow the same simple identifica-
tion of vanishing longitudinal polarisations. We will therefore have to
resort to yet another trick.
To see how this works, let us take a little detour. We have already
produced propagators for scalar and Dirac fields, given by

iD(x − y) = h0|T [φ(x), φ(y)]|0i

d4 p e−ip·(x−y) d4 p −ip·(x−y)
Z Z
= = e D(p)
(2π)4 p2 − m2 + i+ (2π)4
iSF (x − y) = h0|T [ψ(x), ψ̄(y)]|0i
d4 p −ip·(x−y) d4 p −ip·(x−y)
Z Z
p/ + m
= e = e SF (p)
(2π)4 p2 − m2 + i+ (2π)4
In both cases they can be obtained from the solution of the classical
E.o.M.. For example, for a scalar field with Lagrangian
1
L = (∂µ φ)(∂ µ φ) − m2 φ2
2
we have the equation of motion
∂L ∂L
∂µ − = (∂µ ∂ µ + m2 )φ(x) = 0
∂(∂µ φ) ∂φ
and, therefore, the Green’s function is defined by

(∂µ ∂ µ + m2 )G(x − y) = iδ 4 (x − y) .

157
A solution is readily obtained by Fourier transforming and inverting
this equation:
−i
(−pµ pµ + m2 )G̃(p) = i −→ G̃(p) =
p2 − m2 + i+

where the + takes care of the causality structure of the theory. Equat-
ing the propagator with the classical Green’s function yields the de-
sired result from above. The same also works for the Dirac equation,
where the Lagrangian
L = ψ̄(i∂/ − m)ψ
gives rise to the E.o.M.

(i∂/ − m)ψ = 0 ,

and, consequently, the Green’s function is defined by

(i∂/ − m)G(x − y) = iδ 4 (x − y) .

As before the solution is given by Fourier transforming and inverting,

i p/ + m
/ − m)G̃(p) = i −→ G̃(p) =
(p = 2
p/ − m + i+ p − m2 + i+

where we have used that (p/ − m)(p/ + m) = p2 − m2 .

We are now in a position to apply this to the electromagnetic fields in
Lorenz gauge, starting with the Lagrangian
1 α
L = − Fµν F µν − (∂µ Aµ )2
4 2
1
= − [(∂µ Aν )(∂ µ Aν ) + (∂ν Aµ )(∂ ν Aµ )
4
α
−(∂µ Aν )(∂ ν Aµ ) − (∂ν Aµ )(∂ µ Aν )] − (∂µ Aµ )(∂ν Aν )
2
Realising that these are just scalar products (all indices contracted)
allows to “swap” the “names” of indices as long as the structure of the
products is conserved. Therefore
1 α
L = − [(∂µ Aν )(∂ µ Aν ) − (∂µ Aν )(∂ ν Aµ )] − (∂µ Aµ )2
2 2
The equations of motion are given by
∂L ∂L
0 = ∂ρ −
∂(∂ρ Aσ ) ∂Aσ
1 
= − ∂ρ gρµ gσν ∂ µ Aν + gρµ gσν ∂µ Aν − gρµ gσν ∂ ν Aν − gρν gσµ ∂µ Aν

2

158
 −α∂ρ g ρσ ∂ν Aν

= −∂ρ [∂ ρ Aσ − ∂ σ Aρ + α∂ ρ ∂σ Aσ ] = −gρσ Aρ − (1 − α)∂σ (∂ρ Aρ )

and therefore
[−gρσ  − (1 − α)∂σ ∂ρ ] Gρµ (x, x0 ) = −igσµ δ 4 (x − x0 )

Fourier transforming the resulting definition of the Green’s function

results in  2
p gρσ − (1 − α)pσ pρ G̃ρµ (p) = −igσµ


To invert the kernel (the object in the square brackets) , multiply it

with a suitable ansatz and demand that the result equals 1. Realising
that the only tensors with two indices ρ and µ that can be constructed
are the metric tensor g ρµ and pρ pµ /p2 justifies to try
ρ µ
g µρ − λ p pp2
p2
Therefore we solve the following equation for λ
pρ pµ
!  g ρµ − λ p2
gσµ =
 2
p gρσ − (1 − α)pρ pσ
p2
pσ pµ
= gσµ − [1 − α + λ − λ(1 − α)] 2
p
which yields
1−α
1 − α + λα = 0 −→ λ=−
α
The Green’s function is therefore given by
1−α pµ pν
µν
g µν + α · p2
G̃ (p) = −i ·
p2
In Feynman gauge (α = 1) this conveniently reduces to
−ig µν
G̃µν (p) = ,
p2
the result from the lecture.
4. ∗ Propagator for the Schrödinger field
The Lagrangian for the non-relativistic spin-less Schrödinger field ψ is
given by
←→ 1
L = iψ † ∂ t ψ − (∇ψ † ) · (∇ψ) − V (r)ψ † ψ
2m
In the following we will quantise this field and construct its propagator:

159
(a) derive the Euler-Lagrange equations of motion for ψ and its con-
jugate ψ † ;
(b) find the canonical momenta π and π † ;
(c) promote the fields and momenta to operators and demand equal
time commutation relations;
(d) expand the fields in plane waves and creation and annihilation
operators, taking into account that this is a non-relativistic field
theory in which negative-energy solutions are absent;
(e) calculate the commutators for the annihilation and creation op-
erators;
(f) express the Hamilton operator through the creation and annihi-
lation operators;
(g) calculate the free-field propagator

G0 (x0 , x; y0 , y) = −iΘ(x0 − y0 ) h0| ψ(x0 , x)ψ † (y0 , y) |0i

and show that it satisfies

∇2
 
i∂
+ G(t, x; 0, 0) = δ(t)δ 3 (x) .
∂t 2m

For this proof you will have to use that the δ-function can be
represented by r
a −ax2
δ(x) = lim e .
|a|→∞ π

Solution

(a) equations of motion:

∂L ∂L ∂L
0 = ∂t †
+∇ −
∂(∂t ψ ) ∂(∇ψ ) ∂ψ †
†

i 1 2 i
= − ∂t ψ − ∇ ψ − ∂t ψ + V (r)ψ
2 2m 2
1 2
= − i∂t + ∇ − V (r) ψ
2m
∂L ∂L ∂L
0 = ∂t +∇ −
∂(∂t ψ) ∂(∇ψ) ∂ψ
i 1 i
= ∂t ψ † − ∇2 ψ † + ∂t ψ † + V (r)ψ †
2 2m 2
1 2
= + i∂t − ∇ + V (r) ψ † .
2m

160
(b) conjugate momenta:
∂L i
π = = ψ†
∂ ψ̇ 2
∂L i
π† = = − ψ.
∂ ψ̇ † 2
(c) we will demand that the non-vanishing commutator is
h i
ψ̂(t, x), iψ̂ † (t, y) = iδ 3 (x − y)

and all others vanish.

(d) expansion of the field operators in plane waves and annihilation
and creation operators:
d3 k
Z
ψ̂(x) = â(k)e−ik·x
(2π 3 )
d3 k †
Z
†
ψ̂ (x) = â (k)eik·x
(2π 3 )
and we allow only positive energy solutions for the particle field
(ψ), and for the expansion of the field ψ † we form the Hermitian
conjugate of ψ.
Note that we have not included the term 1/(2E) for the wave
expansion – it emerged in relativistic field theory from integra-
tion over the field’s energy when using the δ-function encoding
relativistic energy-momentum-relation. This term, obviously, is
not present here, in the case of non-relativistic fields.
(e) commutators for â and ↠. Assume we have
h i
â(k), ↠(q) = Aδ 3 (k − q)

and we have to fix the constant A:

h i
iδ 3 (x − y) = ψ̂(t, x), π̂(t, y)
h i
= i ψ̂(t, x), ψ̂ † (t, y)
d3 k d3 q h
Z Z i 
† −ik·x+iq·y 
= i â(k), â (q) e
(2π 3 ) (2π 3 )

x0 −y0 =t
Z 3
d k
Z 3
d q 3
= iA 3
δ (k − q)e−ik·x+iq·y
(2π) (2π)3
d3 k −ik·(x−y)
Z
iA 3
= iA 6
e = δ (x − y)
(2π) (2π)3
and therefore A = (2π)3 and
h i
â(k), ↠(q) = (2π)3 δ 3 (k − q)

161
(f) The Hamiltonian is given by
H = π ψ̇ + π † ψ̇ † − L
  ← 
i † † †→ 1 † †
= ψ ψ̇ − ψ ψ̇ − iψ ∂ t ψ − (∇ψ ) · (∇ψ) − V (r)ψ ψ
2 2m
1
= (∇ψ † ) · (∇ψ) + V (r)ψ † ψ ,
2m
the sum of kinetic and potential energies, as expected.
(g) To evaluate the free propagator we will use the non-relativistic
energy-momentum relation Ek = k0 = k 2 /(2m) and the fact that
â |0i = 0. This allows to calculate free propagator as
G0 (x0 , x; y0 , y) = −iΘ(x0 − y0 ) h0| ψ(x0 , x)ψ † (y0 , y) |0i
d3 k d3 q −i(k0 x0 −q0 y0 ) i(k·x−iq·y)
Z
= −iΘ(x0 − y0 ) e e
(2π)3 (2π)3
D   E
× 0 â(k)↠(q) 0
 

d3 k d3 q −i(k0 x0 −q0 y0 ) i(k·x−iq·y)

Z
= −iΘ(x0 − y0 ) e e
(2π)3 (2π)3
×(2π)3 δ 3 (k − q)
d3 k ik 2 (x0 − y0 )
Z  
= −iΘ(x0 − y0 ) exp − + ik · (x − y)
(2π)3 2m
To evaluate the last integral, it is useful to write the exponent in
components; completing the squares reveals that this is nothing
but a (shifted) Gaussian integral,
d3 k ik 2 (x0 − y0 )
Z  
exp − + ik · (x − y)
(2π)3 2m
3 Z∞ " #
Y dkj ikj2 (x0 − y0 )
= exp − + ikj (xj − yj )
2π 2m
i=1−∞

3 Z∞
"
m(xj − yj ) 2
 
Y dkj i(x0 − y0 )
= exp − kj −
2π 2m x0 − y0
i=1−∞

im(xj − yj )2

+
2(x0 − y0 )
3
im(xj − yj )2
r  
Y m
= exp
2πi(x0 − y0 ) 2(x0 − y0 )
i=1

and therefore
3/2 " #
im(x − y)2

m
G0 (x0 , x; y0 , y) = −iΘ(x0 − y0 ) exp
2πi(x0 − y0 ) 2(x0 − y0 )

162
To show that this satisfies the definition of a Green’s function we
apply the differential kernel of the free E.o.M., and using ∇·x = 3
we find:
∇2
 
i∂
+ G0 (t, x; 0, 0)
∂t 2m
∇2 imx2
   
i∂ h m i3/2
= −i + Θ(t) exp
∂t 2m 2πit 2t
2
 
3i mx
= δ(t) − Θ(t) − 2
2t 2t
3im m2 x2 imx2
  h  
1 m i3/2
− − 2 exp
2m t t 2πit 2t
2
h m i3/2  
imx
= δ(t) exp .
2πit 2t

Using now the following property of th δ-function

r
a −ax2
δ(x) = lim e .
|a|→∞ π

with
t→0
a = −im/(2t) = m/(2it) −→ ∞
due to the δ(t) we see that indeed

∇2
 
i∂
+ G0 (t, x; 0, 0) = δ(t)δ 3 (x) .
∂t 2m

163
8 Interacting Fields
8.1 Perturbative Expansion: Born Series

Green’s Function for Full Theory In the previous section we have

analysed propagators, and identified them with the Green’s functions of free
theories. We will now extend the treatment to also include potentials such
that the Hamilton operator can be written as the sum of a free Hamilton
operator plus some interactions,

Ĥ = Ĥ0 + V̂ , (323)

in the simplest case a potential. Going back to Eq. (286), where we have
defined the Green’s function this means that we now have
 
i∂t − Ĥ G(t, x; t0 , x0 )
 
= i∂t − Ĥ0 − V̂ G(t, x; t0 , x0 ) = δ(t − t0 )δ 3 (x − x0 ) . (324)

A formal solution can be obtained by starting with the Fourier transform of

the free Green’s function, signified with a ˜ symbol

ˆ G̃ = 1 −→ G̃ =
 1
ω − H̃0 0 0 , (325)
ˆ
ω − H̃ 0

where we have for the moment suppressed the i prescription. The Fourier
transform of Eq. (324) can therefore be rewritten as
 

ˆ
 1 ˆ
ω − H̃ G̃ = − Ṽ G̃ = 1 . (326)
G̃0
This can be formally solved, and
1 1
= − Ṽˆ (327)
G̃ G̃0
or
1
G̃ = . (328)
1
G̃0
− Ṽˆ

Born Series After Fourier back-transformation we arrive at the implicit

equation

G(t, x; t0 , x0 ) = G0 (t, x; t0 , x0 )
Z
+ dτ d3 ξG0 (t, x; τ, ξ)V̂ (τ, ξ)G(τ, ξ; t0 , x0 ) , (329)

164
which can now be expanded in powers of interactions with the potential.
This is called the Born series or the perturbative expansion of the Green’s
function. For it to converge we implicitly assume that interactions with the
potential are sufficiently small. Replacing explicit time and space coordi-
nates with four-positions t, x → xi , the Born series therefore reads

G(xN ; x0 ) = G0 (xN ; x0 )
Z
+ dx1 G0 (xN ; x1 ) V̂ (x1 ) G(x1 ; x0 )
Z
+ dx1 dx2 G0 (xN ; x2 ) V̂ (x2 ) G(x2 ; x1 ) V̂ (x1 ) G(x1 ; x0 ) . . . , (330)

where in non-relativistic theory we assume a strict time ordering,

tN ≥ tN −1 ≥ tN −2 · · · ≥ t2 ≥ t1 ≥ t0 . (331)

Truncating this series after the first non-trivial term, i.e. after one interac-
tion with the potential is called the Born approximation.

8.2 Interacting Field Theory: General Thoughts

Quantisation and Particle Interpretation In principle we could try
and quantise interacting theory as before, by promoting fields to field oper-
ators and by demanding suitable equal-time commutation relations. How-
ever, the equations of motion for interacting fields are usually not linear
any more, due to the potential terms responsible for the interactions and
featuring more than two fields. This prevents us from being able to solve
them in closed form and we therefore lose the ability to expand the field
operators in products of creation and annihilation operators and some wave
that captures the solution of the E.o.M..
But this means that it is not entirely obvious any more how we arrive at a
meaningful particle interpretation for our fields. One way to answer this is
to realise that ultimately we want to be able to compute numbers that we
can compare with the experiment. In particle physics, we usually have two
colliding particles, which produce more particles in their interaction. This
means that we are mainly interested in being able to calculate cross sections.

Transition Amplitudes between Asymptotic States The cross sec-

tion for a process is proportional for this process to occur. In quantum
mechanics this probability is given by the absolute square of the amplitude
Mf ←i , |Mf ←i |2 , for the transition of an initial state |ii to a final state |f i,

Mf ←i = hf |ii . (332)

The definition of these states is subject on how they are being prepared (for
this initial state) or measure (the final state). For perturbation theory to

165
work, this means that they must be prepared or measured infinitely far away,
both in space and time, from the point where they collider – this assumes,
of course, that the interaction between the states vanishes with increasing
distance. This assumption of asymptotic states is crucial for us to be able
to calculate in a quantum field theory19 .

The S matrix There is yet another problem, while the states |ISi span-
ning the possible initial states of our collision are eigenstates in the initial-
state Fock space of the theory, the corresponding final states |FSi live in
the final-state Fock space. These two sets of states are related to each other
through the Ŝ-matrix such that |FSi = Ŝ |ISi. Therefore the transition
amplitude within the same Fock space is given by
Mf ←i = hf | Ŝ |ii . (333)
In this chapter we will discuss first steps on how to calculate the S-matrix
elements, i.e.
Ŝf i = Mf ←i = hf | Ŝ |ii . (334)

Operators and Pictures From Quantum Mechanics we know that there

is a dichotomy between fields and operators and how they evolve over time,
and it has become customary to distinguish between three pictures:
1. in the Schrödinger picture, the operators Ô(S) (x) are time-independent,
and it is the states that carry the time-dependence,
h i
|ψ(t)i = exp −iĤ(t − t0 ) |ψ(t0 )i ; (335)

2. in the Heisenberg picture, the states that carry the time-independent

and the operators Ô(H) (x, t) are time-independent,
h i h i
Ô(H) (t, x) = exp iĤ(t − t0 ) Ô(H) (t0 , x) exp −iĤ(t − t0 ) ; (336)

3. in the interaction picture, the Hamiltonian is split into a “free” part,

Ĥ0 , and an “interaction” part, Ĥint such that
h i h i
Ô(I) (t, x) = exp iĤo (t − t0 ) Ô(I) (t0 , x) exp −iĤ0 (t − t0 ) , (337)

and the time evolution is distributed over both operators and states.
19
This is because the interacting fields are not identical to the free fields: the vacuum
of interacting and free theories is potentially different, and we only know how to quantise
the latter. This implies immediately that the states |ii and |f i are eigenstates of the free
field theory but usually not eigenstates of the interacting field theory. Their interactions
with a cloud of virtual particles around them, from the surrounding interacting vacuum,
will ultimately force us to renormalise the external field, a topic well beyond this lecture
course.

166
The exponential terms exp[−iĤ(t − t0 )] are known as the time evolution
operator,

Û (t, t0 ) = exp[−iĤ(t − t0 )] . (338)

Time Evolution and Perturbation Theory Let us now take a closer

look at the field operators in both the Heisenberg and the interaction picture.
It is important to stress that in the following we will only sketch the logic of
how, starting from the interaction picture, we arrive at an expression that
can be perturbatively evaluated.
Assuming an explicitly time-independent Hamiltonian, and identifying the
operators at time t = t0 with the Schrödinger-picture operators, φ̂(S) , their
relationship is given by

φ̂(H) (t, x) = eiĤ(t−t0 ) φ̂(S) (x)e−iĤ(t−t0 )

= eiĤ(t−t0 ) e−iĤ0 (t−t0 ) φ̂(I) (t0 , x)eiĤ0 (t−t0 ) e−iĤ(t−t0 ) . (339)

Similar equations naturally also hold true for other operators in the inter-
action picture.
We now redefine the time-evolution operator in the interaction picture such
that the “free” time-evolution is factored out:

Û (I) (t, t0 ) = exp[iĤ0 (t − t0 )] exp[−iĤ(t − t0 )] . (340)

Although it looks as if the two exponentials could be directly multiplied,

to result in an exponential of the interaction Hamiltonian alone, this de-
ceptively simple picture is misleading and hold true only, if Ĥ0 and Ĥint
commute. This, however, is usually not the case and one would have to
resort to the Baker-Hausdorff formula to directly calculate this operator.
Instead, let us construct a differential equation to determine Û (I) . Differen-
tiation with respect to time yields

∂ Û (I) (t, t0 )
i = eiĤ0 (t−t0 ) Ĥ0 (t0 )e−iĤ(t−t0 ) − eiĤ0 (t−t0 ) Ĥ(t0 )e−iĤ(t−t0 )
∂t h i
= −eiĤ0 (t−t0 ) Ĥ(t0 ) − Ĥ0 (t0 ) e−iĤ(t−t0 ) = −eiĤ0 (t−t0 ) Ĥint (t0 )e−iĤ(t−t0 )

= −eiĤ0 (t−t0 ) Ĥint (t0 )eiĤ0 (t−t0 ) eiĤ0 (t−t0 ) e−iĤ(t−t0 )

(I)
= Ĥint (t)Û (I) (t, t0 ) (341)

The formal solution to this differential equation is given by

Û (I) (t, t0 )
Zt

Zt Zt1 tZn−1 
∞
(I) (I) (I)
X
= dt1 (−i)n dt1 dt2 . . . dtn Ĥint (t1 )Ĥint (t2 ) . . . Ĥint (tn )
t0 n=0 t0 t0 t0

167
 Zt
  
(I)
= T exp −i dt0 Ĥint (t0 ) , (342)
t0

where we have used the time-ordering symbol T , that we already encoun-

tered when we constructed propagators for the fields.

Connection to the S-Matrix Recalling that the S-matrix describes the

transition from the initial to the final state, with the former in the infinite
past and the latter in the infinite future, we can connect it to the interaction
Hamiltonian and write
+∞
  
Z
(I)
Ŝ = lim Û (I) (t+ , t− ) = T exp −i dt Ĥint (t) . (343)
t± →±∞
−∞

To evaluate it we will usually go back and expand the exponential to arrive

at an expression like the first line of Eq. (342), and we would truncate
this series after the first few terms. This is well justified if there is a small
parameter – usually a coupling constant – in the relevant parts of interaction
Hamiltonian that steer its size.

8.3 Interacting Field Theory: λφ4

Lagrangian & S-Matrix We will now specify the results of the previ-
ous chapter to a Klein-Gordon field with quartic interactions. This theory,
specified by its Lagrangian

1 m2 2 λ 4
L = (∂µ φ)(∂ µ φ) − φ − φ , (344)
2 2 4!
is probably the most used example on how to construct and evaluate inter-
acting field theories. The Taylor expansion of its S-matrix elements is thus
given by
 Z
−iλ D  h i E
hf |Ŝ|ii = hf |1̂|ii + d4 x f T φ̂4 (x)  i
 
4!
−iλ 2
  Z D  h i E
+ d4 xd4 y f T φ̂4 (x)φ̂4 (y)  i + . . . .
 
4!
(345)

This perturbative expansion will succeed, if λ is sufficiently small. It is

worth noting that the first term, hf |1̂|ii, reduces to a δ in initial and final
states.

168
S-Matrix vs. Creation and Annihilation Operators Let us now see,
how we can evaluate this expression. We will discuss a 2 → 2 scattering
process, where two φ-particles with momenta p1 and p2 scatter to become
two φ-particles with momenta q1 and q2 , p1 + p2 → q1 + q2 . This means
we will have to manipulate expressions like hq 1 q 2 ; out|p1 p2 ; ini between the
in-space and the out-space. For the sake of clarity we will keep a notation,
where we make it explicit to which space the states and operators belong.
Let us start by using creation and annihilation operators to move one of the
in-particles, p1 , from the state-ket into operators:
D  E D   E
q 1 q 2 ; out  p1 p2 ; in = q 1 q 2 ; out ↠(p1 ; in) p2 ; in
  
D   E
= q 1 q 2 ; out ↠(p1 ; out) p2 ; in
 
D   E
+ q 1 q 2 ; out  ↠(p1 ; in) − ↠(p1 ; out)  p2 ; in .
 

(346)
The first term vanishes, unless one of the two momenta q 1,2 = p1 . But this
would mean that one particle would not really participate in the scattering,
something that is usually called a “disconnected diagram”. In such cases
we wouldn’t calculate an amplitude that contributes to a scattering cross
section, and we ignore contributions like this. This leaves us with the second
term. Here, it is important to realise that the in-operator lives at times
t = −∞, while the out-operator is positioned at time t = +∞. This will
help us when we re-express the creation and annihilation operators ↠and
â with the field operators φ̂.

External Particles through Field Operators Starting from Eq. (145)

to write the creation operator as
Z h i
â(k) = d3 x eik·x k0 φ̂(t, x) + iπ̂(t, x)
  
−i∂ ik·x
Z
3 ik·x i∂
= d x e φ̂(t, x) + e π̂(t, x)
∂t ∂t
←→
Z h i
= i d3 x eik·x ∂t φ(t, x)
←
→
Z h i
↠(k) = − i d3 x e−ik·x ∂t φ(t, x) , (347)

where we have redefined, for this chapter,

←→
a ∂ b = a(∂b) − (∂a)b . (348)
This allows us to replace the in-space and out-space creation operators in
Eq. (346) with expressions for the field operators from Eq. (347). Using that
Z Z Z Z
3 3
d x f (t, x) = dt∂t d x f (t, x) = d4 x∂t f (t, x) (349)

169
allows us to replace the in-space and out-space field operators φ̂in and φ̂rmout
with the field operators in the limits t → −∞ and t → +∞ resulting ulti-
mately in
D  E D   E
q 1 q 2 ; out  p1 p2 ; in = q 1 q 2 ; out  ↠(p1 ; in) − ↠(p1 ; out)  p2 ; in
  
  
−ip1 ·x ←
→
Z
3

= −i lim d x e ∂ q 1 q 2 ; out  φ̂(tf , x; out)
ti → −∞
tf → +∞
 

−φ̂(ti , x; in)  p2 ; in

Ztf     
−ip1 ·x ← →
Z
3 ∂  
= −i lim dt d x e ∂ q 1 q 2 ; out φ̂(t, x) p2 ; in
 
ti → −∞ ∂t
tf → +∞ t
i
Z     
4 2 −ip1 ·x
 
= −i d x E1 e q 1 q 2 ; out φ̂(t, x) p2 ; in
 
   
−ip1 ·x 2
 
+e ∂t q 1 q 2 ; out φ̂(t, x) p2 ; in
 
Z     
4 2 2

−ip1 ·x  
= −i d x ∇ + m e q 1 q 2 ; out φ̂(t, x) p2 ; in
 
   
−ip1 ·x 2
 
+e ∂t q 1 q 2 ; out φ̂(t, x) p2 ; in
Z    
4 −ip1 ·x 2

 
= −i d x e x + m q 1 q 2 ; out φ̂(x) p2 ; in .
  (350)

In going from the third to the fourth line we have used that e−ip1 ·x is a
solution for the Klein-Gordon equation, which allowed us to replace the
energy square E12 with (p21 + m2 ), and in going from the fourth to the fifth
line we have integrated by parts, which shifts the ∇2 from the plane wave
to the field operator. This step is possibly only because the interaction is
localised, φ4 (x) and we assume that the fields vanish fast enough for x → ∞
such that the surface terms equal zero.
In a similar way, we can “pull” a state from the final-sate bra through
annihilation operators into a field operator, and we arrive at
D   E
q 1 q 2 ; out φ̂(x) p2 ; in
 
  
−iq1 ·y ←
→
Z
3

= −i lim d y e ∂ q 2 ; out  φ̂out (tf , y; out)φ̂(x)
ti → −∞
tf → +∞
 

−φ̂(x)φ̂(ti , y; in)  p2 ; in

Z   h i  
4 −iq1 ·y 2


= −i d ye y + m q 2 ; out T φ̂(y)φ̂(x)  p2 ; in , (351)

170
where the time-ordering results from the limits for the temporal integration.

Lehmann-Symanzik-Zimmermann Pulling all particles from the bras

and kets into the fields we arrive at the Lehmann-Symanzik-Zimmermann
(LSZ) formula; for our case of four particles it reads
   Z 
2 2
d x1 d x2 d y1 d y2 e−i(p1 ·x1 +p2 ·x2 −q1 ·y1 −q2 ·y2 )
4 4 4 4

q 1 q 2  p1 p2 = (−i) (i)


× x1 + m2 x2 + m2 y1 + m2 y2 + m2

  h i 

× 0 T φ̂(y1 )φ̂(y2 )φ̂(x1 )φ̂(x2 )  0
 (352)

The pattern is clear: each external particle with momentum k results in an

integral over all space, d4 x, and obtains a plane-wave factor exp(±ik · x),
an inverse propagator term20 , and is represented by a corresponding field
operator in the vacuum expectation value of a time-ordered product of such
operators. The inverse propagator terms ultimately reduce the S-matrix
to the normalised residue of this vacuum expectation value by, pictorially
speaking, “truncating” (cutting off) the effect of the external particles prop-
agating to the interaction zone.

Wick’s Theorem To evaluate the perturbative series encoded in the LSZ

formula, Eq. (352) we will use Wick’s theorem. It connects time-ordered
products of field operators with their normal ordered products and products
of Feynman propagators. Without any attempt at proving it we will just
state it for some examples below.

1. For two field operators we have

h i
T φ̂(x)φ̂(y) = :φ̂(x)φ̂(y): +∆F (x − y) , (353)

2. and for four field operators it reads

h i
T φ̂(x1 )φ̂(x2 )φ̂(x3 )φ̂(x4 ) = :φ̂(x1 )φ̂(x2 )φ̂(x3 )φ̂(x4 ):
X
+ :φ̂(xi )φ̂(xj ): ∆F (xk − xl )
i<j;k<l
X
+ ∆F (xi − xj )∆F (xk − xl )
i<j;i<k;k<l
(354)
20 2 2 2
It is easy to see, by Fourier transformation, that ( + m ), indeed is (p − m ), up to
some phase factor.

171
By using that the vacuum expectation value of any normal-ordered product
vanishes when sandwiched between vacua,
D   E
0 :φ̂(x1 )φ̂(x2 ) . . . φ̂(xn ): 0 = 0 , (355)
 

and by realising that the Feynman propagators are just numbers, for ex-
ample Eq. (301), and that therefore the vacuum expectation number of any
product of them just equals their product,
h0 |∆F (x1 − x2 )∆F (x3 − x4 ) . . . | 0i = ∆F (x1 − x2 )∆F (x3 − x4 ) . . . (356)
we see that the vacuum expectation value of the time-ordered product of
fields reduces to a product of Feynman propagators and, possibly, “vertex
factors” related to interaction points, where three or more of these fields
interact.

0th -Order Let us now see how this plays out for the 0th -order term, where
we merely have the four field operators. This is equivalent to the term
hf |1̂|ii, the first term in the perturbative expansion of Eq. (345). Going
back to Eq. (352) we therefore end up with
   
 
q 1 q 2  1̂  p1 p2
 
Z 
= d x1 d x2 d y1 d y2 e−i(p1 ·x1 +p2 ·x2 −q1 ·y1 −q2 ·y2 )
4 4 4 4

× x1 + m2 x2 + m2 y1 + m2 y2 + m2


× ∆F (x1 − x2 )∆F (y1 − y2 )

+∆F (x1 − y1 )∆F (x2 − y2 )


+∆F (x1 − y2 )∆F (x2 − y1 )
Z 
= d x1 d x2 d y1 d y2 e−i(p1 ·x1 +p2 ·x2 −q1 ·y1 −q2 ·y2 )
4 4 4 4

× x1 + m2 x2 + m2 y1 + m2 y2 + m2

"Z
d4 k1 d4 k2 e−ik1 (x1 −x2 ) e−ik2 (y1 −y2 )
×
(2π)4 (2π)4 k12 − m2 − i+ k22 − m2 − i+
d k1 d4 k2 e−ik1 (x1 −y1 ) e−ik2 (x2 −y2 )
Z 4
+
(2π)4 (2π)4 k12 − m2 − i+ k22 − m2 − i+
#)
d k1 d4 k2 e−ik1 (x1 −y2 ) e−ik2 (x2 −y1 )
Z 4
+
(2π)4 (2π)4 k12 − m2 − i+ k22 − m2 − i+

172
 Z 
= d x1 d x2 d y1 d y2 e−i(p1 ·x1 +p2 ·x2 −q1 ·y1 −q2 ·y2 )
4 4 4 4

Z 4
d k1 d4 k2
2 2
2
× 4 4
k12 − m2 k2 − m2
(2π) (2π)
"
e−ik1 (x1 −x2 ) e−ik2 (y1 −y2 )
×
k12 − m2 − i+ k22 − m2 − i+
e−ik1 (y1 −x1) e−ik2 (y2 −x2)
+
k12 − m2 − i+ k22 − m2 − i+
#)
e−ik1 (y2 −x1 ) e−ik2 (y1 −x2 )
+ 2
k1 − m2 − i+ k22 − m2 − i+
d4 k1 d4 k2
Z
k12 − m2 k22 − m2



= 4 4
(2π) (2π)

× δ 4 (k1 + p1 )δ 4 (k1 − p2 )δ 4 (k2 + q1 )δ 4 (k2 − q2 )

+δ 4 (k1 + p1 )δ 4 (k1 + q1 )δ 4 (k2 + p2 )δ 4 (k2 + q2 )


4 4 4 4
+δ (k1 + p1 )δ (k1 + q2 )δ (k2 + p1 )δ (k2 + q1 )

δ 4 (p1 − p2 )δ 4 (q1 − q2 ) p21 − m2 q12 − m2



=

+ δ 4 (p1 − q1 )δ 4 (p2 − q2 ) p21 − m2 p22 − m2


4 4
p21 2
p22 2



+ δ (p1 − q1 )δ (p2 − q2 ) −m −m . (357)

Looking at this, we realise that the first term will vanish if all external
particles have positive energies - which they should as we want to calculate
a physical cross section. This leaves us the last two terms where particle p1,2
transit directly, without interaction, into particles q1,2 or vice versa. The
absence of an interaction should not come as a surprise: as a starting point
we have only sandwiched the free theory between initial and final state.

1st -Order This however changes, when we go to the first order of pertur-
bation theory, or the first term with an interaction Hamiltonian sandwiched
between hf | and |ii, i.e. the second term on the right-hand side of Eq. (345).
In this case, and in order to arrive at connected diagram, i.e. those where
all external lines are connected through propagators, we will have to con-
nect the four outgoing particles with the interaction vertex. Integrating over
all possible permutations of possible connections and over all space for the

173
vertex position we arrive at
Z    
iλ
d z q 1 q 2  : − φ̂4 (z) :
4
 
p p
4!  1 2
Z 
iλ X
=− d x1 d x2 d y1 d y2 d z e−i(p1 ·x1 +p2 ·x2 −q1 ·y1 −q2 ·y2 )
4 4 4 4 4
4!
{x1 ,x2 ,y1 ,y2 }

× x1 + m2 x2 + m2 y1 + m2 y2 + m2

 
× ∆F (x1 − z)∆F (x2 − z)∆F (z − y1 )∆F (z − y2 )
Z 
= − iλ d x1 d x2 d y1 d y2 d z e−i(p1 ·x1 +p2 ·x2 −q1 ·y1 −q2 ·y2 )
4 4 4 4 4

Z 4
d k1 d4 k3 d4 k3 d4 k4
×
(2π)4 (2π)4 (2π)4 (2π)4
× x1 + m2 x2 + m2 y1 + m2 y2 + m2

"
e−ik1 ·(z−x1) e−ik2 ·(z−x2)
×
k12 − m2 + i+ k22 − m2 + i+
#)
e−ik3 ·(y1−z) e−ik4 ·(y2−z)
× 2
k3 − m2 + i+ k42 − m2 + i+
Z 
= − iλ d x1 d x2 d y1 d y2 d z e−i(p1 ·x1 +p2 ·x2 −q1 ·y1 −q2 ·y2 )
4 4 4 4 4

Z 4
d k1 d4 k3 d4 k3 d4 k4
×
(2π)4 (2π)4 (2π)4 (2π)4
× k12 − m2 k22 − m2 k32 − m2 k42 − m2

"
e−ik1 ·(z−x1) e−ik2 ·(z−x2)
× 2 2
k1 − m2 + i+ k2 − m2 + i+
#)
e−ik3 ·(y1−z) e−ik4 ·(y2−z)
× 2
k3 − m2 + i+ k42 − m2 + i+
Z 4
d k1 d4 k3 d4 k3 d4 k4

= − iλ (2π)4 δ 4 (k1 + k2 − k3 − k4 )
(2π)4 (2π)4 (2π)4 (2π)4
× (2π)4 δ 4 (k1 − p1 )(2π)4 δ 4 (k2 − p2 )

4 4 4 4
× (2π) δ (k3 − q1 )(2π) δ (k4 − q2 )

= (2π)4 δ 4 (p1 + p2 − q1 − q2 )iλ . (358)

We realise that, after this long calculation, the amplitude for the 2 → 2-
scattering including one interaction vertex is given by the value λ of the

174
interaction vertex, when taking into account the 4! combinations of combin-
ing the four external legs with the vertex.

Feynman Rules This finding allows us to formulate simpler rules for the
construction of amplitudes. The LSZ formula above guarantees that we only
have to take into account interaction vertices connecting the internal lines
for particles, and we know that they are given by the time-ordered products
– or commutators – of the fields. This gives rise to the Feynman rules for
the λφ4 theory, namely

−i
p =
p2 − m2 + i+

@
@t −iλ
@ = (359)
@ 4!

2nd -Order Amplitude Let us now construct a second order amplitude for
the 2 → 2-scattering, using the Feynman rules from Eq. (359). Labelling in-
coming particles as 1, 2 and outgoing particles as 3, 4, we find three different
diagrams, namely

(a) (b) (c)

1 3 1 HH
t
 3 1 HH
t
 4
A k

At t



A

q A t t
2 4 2 HH 4 2 HH 3 (360)

Let us focus now on diagram (a) and translate it into an expression for the
amplitude. We have
2 Z
d4 k d4 q
 
(a) iλ −i −i
Ŝ = −
4! (2π)4 (2π)4 k2 − m + i q − m2 + i+
2 + 2

(4!)2

× (2π)4 δ(p1 + p2 − q − k)(2π)4 δ(q + k − p3 − p4 )
2
λ2
= (2π)4 δ(p1 + p2 − p3 − p4 )
2!
d4 k
Z
1
× , (361)
(2π) [k − m + i ][(P − k)2 − m2 + i+ ]
4 2 2 +

where we have introduced P = p1 + p2 . The two factorials 4! from the

interaction vertex are compensated by similar factors from attaching lines
to the vertices, but modified by 1/2. This “symmetry factor” stems from the

175
fact that there are two internal lines connecting the two vertices at positions
y1 and y2 , taking out a combinatorial factor of 2!. For diagrams (b) and (c)
we arrive at similar expressions, where P is modified to become P = p1 − p3
and P = p1 − p4 , respectively.
Closer inspection of the k-integration reveals that this diagram gives rise to
a logarithmic divergence. To see this, consider a limit where k becomes in-
finitely large, k → ∞. In this limit the integral assumes the asymptotic form
of d4 k/k 4 , and using polar coordinates in four dimensions, we can write this
as k 3 d3 Ωdk/k 4 , where d3 Ω takes care of the finite angular integrals. This
leaves us with a final integral dk/k which diverges for k → ∞. This consti-
tutes yet another divergence in Quantum Field Theory, and, similar to the
treatment before, it is cured by subtracting suitable terms, this time directly
in the Lagrangian. These terms are constructed after “regularising” the inte-
grals, i.e., after quantifying the degree of their divergence and its prefactors.
The overall procedure of dealing with these ultraviolet divergences is known
as “renormalisation”.

Cross Section To arrive at a cross section σi→f for a process i → f , we

have to
ˆ absolute-square the transition amplitude, |Sf i |2

ˆ sum or average over all outgoing or incoming unobserved internal de-

grees of freedom such as spins, polarisations, or colours, indicated by
the symbol ¯
P

ˆ integrate over Lorentz invariant phase space given by the outgoing

momenta, qi

ˆ and multiply the result with the Lorentz-invariant flux that describes
√
the phase space density of the incoming particle beam (the term 1/(4 . . .)
in front of the overall expression).
Expressed as an equation and using that p21,2 = m21,2 , and making four-
momentum conservation explicit this therefore reads
1
σi→f = p
4 (p1 · p2 )2 − p21 p22
n n
!
d3 qi
Z Y
¯
X X
× 3
|Ŝf i |2 (2π)4 δ 4 p1 + p2 − qi (362)
(2π) 2Ei
i=1 d.o.f. i=1

For the case at hand, we have the first-order amplitude from Eq. (358).
Stripping out the overall four-momentum conservation it is given by Ŝf i =
iλ. Assuming incident momenta
p
p1,2 = (E, 0, 0, ± E 2 − m2 ) , (363)

176
we arrive at
λ2
σi→f = p
4 (2E 2 − m2 )2 − m4
d3 q1 d3 q2
Z
× (2π)4 δ 4 (p1 + p2 − q1 − q2 )
(2π)3 2E1 (2π)3 2E2
λ2 d3 q1
Z 

= √ 3
(2π)δ(2E − E1 − E2 ) √
8E E − m2 2 (2π) 4E1 E2 E2 = q 21 +m2
Z 2
λ2 q 1 d|q 1 |d2 Ω1 q
= √ δ(2E − 2 q 21 + m2 )
32π 2 E E 2 − m2 4(q 21 + m2 )
λ2 (E12 − m2 )dE1
Z
= √ δ(2E − 2E1 )
32πE E 2 − m2 E12
r
λ2 m2
= 1 − (364)
32πE 2 E2
for the cross section at the lowest order in the couplnig constant, O(λ2 )
where we have used polar coordinates for the q1 -integration and realised
that d|q 1 | = dE1 . The cross section has units of inverse energy squared or
area and is usually given in units of “barn”, where

1 barn = 1 b = 10−28 m2 . (365)

177
9 List of Problems

Section Page Title

2 15 Levi-Civita symbol
16 Boosts and rotations
17 Inverse Lorentz transformation
18 ∗ The Generators of the Lorentz group

22 ∗ Poincare transformations

24 Lagrange and Hamilton Formalism: Example Systems

26 Conserved Energy from Hamilton function
27 Quantum Mechanical Harmonic Oscillator
3 45 General Solutions for the Klein-Gordon Equation
47 ∗ Klein-Gordon equation in Two-Component Form

49 Euler-Lagrangr Equation of Motion

51 Massive Vector Field
52 Electrodynamics with Gauge-Fixing Term
53 Free Schrödinger Field
54 ∗ Equations of Motion with Boundary Conditions

55 Symmetry and Conserved Current

56 ∗ ASU (2) Symmetry

57 Energy-Momentum Tensor
4 72 States and Operators of the Real Scalar Field
73 Wave Functional from State Vectors
74 Two Real Scalar Fields Equal One Complex Scalar Field
79 Momentum Operator
81 Causality and Anti-Commutators (Real Scalars)
82 Commutators for Free Real Scalar Fields
84 ∗ Properties of the Charge Operator

85 ∗ Parity of a Scalar Field

5 101 Dirac Equation and Anti-Commutators

101 Commutators with the Dirac Hamilton Operator
103 ∗ Direct Solution of the Dirac Equation

104 Dirac Spinor Relations

109 γ Algebra
110 Dirac Hamiltonian from Creation and Annihilation Operators
116 ∗ Dealing with γ
5
118 ∗ Spin Operator

178
Section Page Title
6 132 Polarisation Vectors in Coulomb Gauge
133 Equal-Time Commutators of E and B
134 Momentum Operator P̂ µ
135 ∗∗ Casimir Effect

7 152 Green’s Function for a Free Particle in Quantum Mechanics

154 Feynman Propagator for the Dirac Field
156 Propagator in General Lorentz Gauge
159 ∗ Propagator for the Schrödinger field

179
References
[1] M.E. Peskin and D.V. Schroeder. ”An Introduction to Quantum Field
Theory”. http://www.physicsbook.ir/book/An%20Introduction%
20To%20Quantum%20Field%20Theory%20-%20M.%20Peskin,%20D.
%20Schroeder%20(Perseus,%201995).pdf. (probably the standard
reference for quantum field theory. The book is full of insight and
definitely worth a read; be aware that they use a somewhat different
normalisation convention for their fields.).
[2] D.J. Griffiths. ”Introduction to Elementary Particles”. http://
nuclphys.sinp.msu.ru/books/b/Griffiths.pdf. (a light-touch in-
troduction, especially he first few chapters. Recommended for addi-
tional reading to develop intuition and to contextualise some of the
material of the course.).
[3] B. Hatfield. ”Quantum Field Theory of Point Particles and Strings”.
http://www.fulviofrisone.com/attachments/article/483/
Hatfield%20QFT%20of%20particles%20and%20strings(T).pdf.
(this book is eminently readable. What I like best about it is that it
shows field quantisation in three parallel formalisms: Heisenberg oper-
ators, Schr”odinger wave functionals, and Feynman’s path integrals.
This is where – for me – the penny finally dropped. It took a quite
some time . . . .).
[4] M. Srednicki. ”Quantum Field Theory”. http://web.physics.ucsb.
edu/~mark/ms-qft-DRAFT.pdf. (a great resource and an excellent
book – this is the pre-publication version of one of the best recent
books on QFT, and I recommend it full-heartedly!).
[5] D. Tong. ”Quantum Field Theory”. https://www.damtp.cam.ac.uk/
user/tong/qft/qft.pdf. (great lecture notes on Quantum field The-
ory with some overlap with our course).
[6] S Coleman. ”Notes from Sidney Coleman’s Physics 253a: Quantum
Field Theory”. https://arxiv.org/pdf/1110.5013v4.pdf. (arguably
one of the best lectures of all time on Quantum Field Theory, by one
of the great minds in the field.).
[7] J.A. Dror. ”Useful Relations in Quantum Field Theory”. http://
pages.physics.cornell.edu/~ajd268/Notes/UsefulFormulas.pdf.
(quite a summary of things that we will work out during this lecture
course).
[8] K. Dullemond and K. Peeters. ”Introduction to Tensor Calcu-
lus”. http://www.ita.uni-heidelberg.de/~dullemond/lectures/
tensor/tensor.pdf. (exhaustive lecture notes on tensor calculus).

180
 [9] H. Goldstein. ”Classical Mechanics”. https://detritus.
fundacioace.com/pub/books/Classical_Mechanics_Goldstein_
3ed.pdf. (standard book on classical mechanics).

[10] L. D. Landau and E. M. Lifshitz. ”Mechanics, Third Edition: Volume

1 (Course of Theoretical Physics)”. https://archive.org/details/
Mechanics_541, 1976. (at times quite short, but very concise; excellent
and super-challenging problems and solutions).

[11] J.J. Sakurai. ”Modern Quantum Mechanics”. http://kgut.ac.ir/

useruploads/1505647831850hcd.pdf. (in my opinion the best book
ever on Quantum Mechanics. In fact, this was the book where the penny
dropped for me – before that I was quite hopeless.).

[12] C. Itzykson and J.B. Zuber. ”Quantum Field Theory”. McGraw-Hill.

(this is a very old book, from 1980. At times a bit cumbersome it is still
the go-to reference for me with lots of insights, identities, etc., and my
printed version suffered a lot in the past 25 or so years. Unfortunately
I haven’t found a good source for it on the web - please, let me know if
you find one!).

181