Gauge Theories

Gauge Theories
of the Strong and Electroweak Interactions

G. Münster, G. Bergner
Summer term 2011
Notes by B. Echtermeyer
Is nature obeying fundamental laws? Does a comprehensive description of

the laws of nature, a kind of theory of everything, exist?
Gauge theories and symmetry principles provide us with a comprehensive
description of the presently known fundamental particles and interactions.
The Standard Model of elementary particle physics is based on gauge theo-
ries, and the interactions between the elementary particles are governed by
a symmetry principle, namely local gauge invariance, which represents an
infinite dimensional symmetry group.
These notes are not free of errors and typos. Please notify us if you find
some.
Contents
1 Introduction 4
1.1 Particles and Interactions . . . . . . . . . . . . . . . . . . . . 4
1.2 Relativistic Field Equations . . . . . . . . . . . . . . . . . . . 13
1.2.1 Klein-Gordon equation . . . . . . . . . . . . . . . . . . 13
1.2.2 Dirac equation . . . . . . . . . . . . . . . . . . . . . . 16
1.2.3 Maxwell’s equations . . . . . . . . . . . . . . . . . . . 19
1.2.4 Lagrangian formalism for fields . . . . . . . . . . . . . 22
1.3 Symmetries . . . . . . . . . . . . . . . . . . . . . . . . . . . . 30
1.3.1 Symmetries and conservation laws . . . . . . . . . . . . 30
1.3.2 U(1) symmetry, electric charge . . . . . . . . . . . . . . 32
1.3.3 SU(2) symmetry, isospin . . . . . . . . . . . . . . . . . 34
1.3.4 SU(3) flavour symmetry . . . . . . . . . . . . . . . . . 42
1.3.5 Some comments about symmetry . . . . . . . . . . . . 44
1
2 CONTENTS
1.4 Field Quantisation . . . . . . . . . . . . . . . . . . . . . . . . 46

1.4.1 Quantisation of the real scalar field . . . . . . . . . . . 47
1.4.2 Quantisation of the complex scalar field . . . . . . . . . 52
1.4.3 Quantisation of the Dirac field . . . . . . . . . . . . . . 54
1.4.4 Quantisation of the Maxwell field . . . . . . . . . . . . 55
1.4.5 Symmetries and Noether charges . . . . . . . . . . . . 57
1.5 Interacting Fields . . . . . . . . . . . . . . . . . . . . . . . . . 58
1.5.1 Interaction picture . . . . . . . . . . . . . . . . . . . . 58
1.5.2 The S-matrix . . . . . . . . . . . . . . . . . . . . . . . 61
1.5.3 Wick’s theorem . . . . . . . . . . . . . . . . . . . . . . 62
1.5.4 Feynman diagrams . . . . . . . . . . . . . . . . . . . . 63
1.5.5 Fermions . . . . . . . . . . . . . . . . . . . . . . . . . . 66
1.5.6 Limitations of the perturbative approach . . . . . . . . 68
2 Quantum Electrodynamics (QED) 69

2.1 Local U(1) Gauge Symmetry . . . . . . . . . . . . . . . . . . . 69
2.2 Quantum Electrodynamics . . . . . . . . . . . . . . . . . . . . 71
3 Non-abelian Gauge Theory 74

3.1 Local Gauge Invariance . . . . . . . . . . . . . . . . . . . . . . 74
3.2 Geometry of Gauge Fields . . . . . . . . . . . . . . . . . . . . 80
3.2.1 Differential geometry . . . . . . . . . . . . . . . . . . . 80
3.2.2 Gauge Theory . . . . . . . . . . . . . . . . . . . . . . . 82
4 Quantum Chromodynamics (QCD) 87

4.1 Lagrangian Density and Symmetries . . . . . . . . . . . . . . 87
4.1.1 Local SU(3) colour symmetry . . . . . . . . . . . . . . 88
4.1.2 Global flavour symmetry . . . . . . . . . . . . . . . . . 90
4.1.3 Chiral symmetry . . . . . . . . . . . . . . . . . . . . . 91
4.1.4 Broken chiral symmetry . . . . . . . . . . . . . . . . . 95
4.2 Running Coupling . . . . . . . . . . . . . . . . . . . . . . . . . 97
4.2.1 Quark-quark scattering . . . . . . . . . . . . . . . . . . 98
4.2.2 Renormalisation . . . . . . . . . . . . . . . . . . . . . . 100
4.2.3 Running coupling . . . . . . . . . . . . . . . . . . . . . 101
4.2.4 Discussion . . . . . . . . . . . . . . . . . . . . . . . . . 102
4.3 Confinement of Quarks and Gluons . . . . . . . . . . . . . . . 105
4.4 Experimental Evidence for QCD . . . . . . . . . . . . . . . . . 108
5 Electroweak Theory 111

5.1 Weak Interactions . . . . . . . . . . . . . . . . . . . . . . . . . 111
5.1.1 Fermi theory of weak interaction . . . . . . . . . . . . 111
CONTENTS 3
5.1.2 Parity violation . . . . . . . . . . . . . . . . . . . . . . 111

5.1.3 V-A theory . . . . . . . . . . . . . . . . . . . . . . . . 113
5.2 Higgs Mechanism . . . . . . . . . . . . . . . . . . . . . . . . . 117
5.2.1 Spontaneous breakdown of a global symmetry . . . . . 117
5.2.2 Higgs mechanism . . . . . . . . . . . . . . . . . . . . . 119
5.3 Glashow-Weinberg-Salam Model . . . . . . . . . . . . . . . . . 121
4 1 INTRODUCTION
1 Introduction
1.1 Particles and Interactions
When reflecting on the constituents of matter, one is lead to the physics of
elementary particles. A classification of elementary particles is done by re-
garding their properties, which are
mass, spin,
(according to the representations of the inhomogeneous Lorentz group)
lifetime,
additional quantum numbers,
(obtained from conservation laws)
participation in interactions.
From these properties the following classification arose.
Leptons
e− , νe electron number
µ− , νµ muon number
τ −, ντ tauon number
Hadrons strongly interacting particles
Mesons integer spin, baryon number = 0
π , π , π , K , K , K , η, ρ+ , ρ− , ρ0 , J/ψ etc.
+ − 0 + − 0
Baryons half integer spin, baryon number = ±1

0
n, p, Λ , Σ, Ξ, ∆, Ω− , Y etc.
Quark model of hadrons (Gell-Mann, SU(3), eightfold way)

The hadrons are build out of two or three quarks.
Mesons
q q̄ (quark, antiquark)
Baryons
qqq
There are six quarks and their antiparticles. They all have spin 1/2.
The six quark types are called “flavours”, which are denoted by
u, c, t
d, s, b
1.1 Particles and Interactions 5
Some baryons some mesons

p = uud π + = ud¯
n = udd K + = us̄
Λ = sud ρ+ = ud¯
Ω− = sss D+ = cd¯
ηc = cc̄
3 Generations of constituents
mass [MeV] Q B
νe ≈0 0 0
e− 0.511 −1 0
u ≈4 2/3 1/3
d ≈7 −1/3 1/3
νµ ≈0 0 0
µ− 105.66 −1 0
c ≈ 1300 2/3 1/3

s ≈ 150 −1/3 1/3
ντ < 18.2 0 0
τ− 1777.0 −1 0
t ≈ 174 000 2/3 1/3

b ≈ 4 200 −1/3 1/3
Table 1: The 3 generations of elementary particles

6 1 INTRODUCTION
Fig. 1 and Fig. 2 show multiplets of mesons and baryons arranged in 3-

dimensional multiplets1 . The coordinates are
(x, y, z) = (Isospin I, Hypercharge Y, Charm C)
Figure 1: SU(4) multiplets of mesons; 16-plets of pseudoscalar (a) and vector

mesons (b). In the central planes the cc̄ states have been added. – From
The Particle Data Group, 2010.
1
The meson multiplets form an Archimedean solid called cubooctahedron
Figure 2: SU(4) multiplets of baryons. (a) The 20-plet with an SU(3) octet.
(b) The 20-plet with an SU(3) decuplet. – From The Particle Data Group,
2010.
Quark confinement
Quarks do not exist as single free particles. There is an additional quantum
number, called “colour”. E.g., Ω− = sss has spin 3/2; therefore the wavefunc-
tion has to be antisymmetric in the spin-coordinates. It is also symmetric
in space coordinates, so the Pauli-principle can only be fulfilled, if the three
charmed quarks are different in some additional quantum number.
All hadrons are colourless combinations of quarks. This phenomenon is called

confinement.
8 1 INTRODUCTION
There is a characteristic feature for each single generation of leptons and

quarks:
X
Qi = 0.
Q
1
u r , u g , ub
0 νe
dr , dg , db
−1 e−
B
0 1
The reason, why (ν, e− ) and (u, d) belong to this same generation and not,
for instance (ν, e− ) and (c, s) will be given later in the chapter on weak
interactions.
Interactions
An important guiding principle in the history of understanding interactions
has been unification. When Newton postulated that the gravitational force
which pulls us down to earth and the force between moon and earth are es-
sentially the same, this was a step towards unification of fundamental forces,
as was the unification of magnetism and electricity by Faraday and Maxwell,
which led to a new understanding of light, or – about a century later – the
unification of electromagnetism and weak interactions.
Nowadays one distinguishes four fundamental interactions:
a) Electromagnetic interactions. They apply to electrically charged par-

ticles only (no to neutrinos, for instance).
Since the electrostatic force is proportional to 1/r2 , one says that the
range of the electromagnetic interactions is infinite. A further charac-
teristic of interactions is their relative strength, when compared with
the strength of other interactions. For electromagnetism it is given by
Sommerfeld’s “Feinstrukturkonstante” α.
range = ∞ (1.1)
e2 1
relative strength = ≈ (1.2)
4πǫ0 ~c 137
b) Weak interactions. They are responsible for the β - decay and other
processes.
range ≈ 10−18 m (1.3)
relative strength ≈ 10−5 (1.4)
c) Strong interactions. They are responsible for the binding of quarks and
for the hadronic interactions. Nuclear forces are also remnants of the
strong interactions.
range ≈ 10−15 m (1.5)
relative strength ≈ 1 (1.6)
d) Gravitation acts on every sort of matter. E.g., it has been shown ex-
perimentally that a neutron falls down through a vacuum tube just like
any other object on earth. The gravitational force is always attractive.
Whereas positive and negative electric charges exist, there are no neg-
ative masses and thus the gravitational force cannot be screened. The
range of this force is infinite like that of electromagnetism. Comparing
the gravitational force between proton and electron in an H-atom with
their electrostatic attraction, one finds that the gravitational force is
extremely weak.
range = ∞ (1.7)
relative strength ≈ 10−39 (1.8)
Forces are mediated by the exchange of bosons.

The range is given by the Compton wavelength of the exchange boson. (But
there is an exception to this law in QCD due to confinement.)
~
range R ≈ (1.9)
mc
Interaction bosons spin mass, range
electromagnetic photon γ 1 m = 0, R = ∞
weak W +, W −, Z 0 1 mW = 80.4 GeV
mZ = 91.2 GeV
strong gluons G 1 m = 0, R 6= 0
gravitation graviton 2 m=0
For gluons the spin 1 is a consequence of gauge theory, and the finite range
R arises from confinement, which holds for gluons as for quarks. The spin
of the exchange boson is related to the possibility of a force being only at-
tractive or both attractive and repulsive. Spin 2 implies that there is only
10 1 INTRODUCTION
attraction. The existence of the graviton with zero mass is predicted theo-
retically and may never be verified by experiment. Measuring gravitational
waves is already very challenging, and to identify the quanta of these waves
would be extremely difficult.
Theories
a) Quantum Electrodynamics originated in 1927, when in an appendix to

the article of Born, Heisenberg and Jordan about matrix mechanics Jor-
dan quantised the free electromagnetic field. It was developed further
by Dirac, Jordan, Pauli, Heisenberg and others and culminated before
1950 in the work of Tomonaga, Schwinger, Feynman and Dyson. The
calculation of the Lamb shift and the exact value of the gyromagnetic
ratio g of the electron are highlights of QED.
Here is an example of a Feynman diagram for the scattering of two

electrons by exchanging a photon.
e− e−
e− e−
The vertex stands for a number, in QED this is α ≈ 1/137. The

propagation of electrons is affected by the emission and absorption of
virtual photons, as shown in the following Feynman diagram.
b) The theory of weak interactions begun in 1932 with Fermi’s theory for
the β − -decay. The Feynman graph for the decay of neutrons involves
a 4–fermion coupling.
p e−
ν̄e
n
Improvements of the theory of β-decay in nucleons were made by the

V-A theory, taking care of parity violation.
Theoretical problems: while in QED perturbation theory in powers of
α works extremely well, it leads to infinities in the Fermi theory of
weak interactions. The problems were overcome in 1961 – 1968 by
Glashow, Weinberg, Salam and others, developing the unified theory
of weak and electromagnetic interactions. The bosons mediating the
electroweak interactions are
Vector bosons W ± , Z 0 and photon γ.
c) Strong interactions between quarks are described by Quantum Chro-

modynamics (QCD), which was formulated by Fritzsch, Gell-Mann and
Leutwyler, and further developed by ’t Hooft and others. There are
three “strong charges”, sources for the forces, named red, green and
blue charge. The gauge bosons which mediate strong interactions are
called gluons.
Unlike the electrically neutral photons in QED, gluons carry colour
charges themselves and interact with each other. Due to their self-
interactions, gluons may form glueballs, and a “theory of pure glue” is
a non-trivial theory.
Feynman diagrams with quarks and gluons

d) Gravitation is described by General Relativity (GRT), a nonlinear the-
ory. A quantum theory of gravitation is not yet known. String theory,
Superstring theory or Loop gravity might be candidates.
12 1 INTRODUCTION
The Standard Model

This means the theory of Glashow, Weinberg and Salam (G.W.S.) plus QCD.
There is no mixing between the Lagrangians for electroweak and strong in-
teractions, therefore, we do not speak of a unification of these interactions.
The theoretical predictions of the Standard Model are so far consistent with
the experimental results.
Common to all parts of the Standard Model are exchange bosons, which are
related to gauge fields showing local gauge symmetries. (Gravitation is also
based on a local symmetry.) Gauge theories are based on gauge groups. The
groups belonging to the Standard Model are
SU(3) ⊗ SU(2) ⊗ U(1) . (1.10)
| {z } | {z }
QCD G.W.S.
The principles of the Standard Model are:

• local gauge symmetry,
• Higgs mechanism giving masses to W ± , Z 0 and quarks.
The Higgs mechanism is due to P. Anderson, F. Englert, R. Brout, P. Higgs,

G. Guralnik, C. R. Hagen and T. Kibble. It uses the Higgs field, associated
with a Higgs-boson. This does not fit into a local gauge theory, so the Higgs
boson might not be a fundamental particle. There is no other reason for the
Higgs field than the mechanism to give the above mentioned masses.
Outlook
A further unification of interactions is attempted in Grand Unified Theories
(GUT). The idea is to extend the semisimple2 Lie group SU(3)⊗SU(2)⊗U(1)
to a simple Lie group as for example SU(5), SO(10) or the exceptional Lie
group E6 . GUTs predict proton decay and several Higgs particles.
A group is called semisimple, if it is the direct product of simple groups. A group is

2
simple, if it has no normal subgroups besides the trivial ones.

1.2 Relativistic Field Equations 13
1.2 Relativistic Field Equations

In classical physics there are two distinct kinds of objects: particles – point
particles or continuous distributions of mass – and secondly fields, like grav-
itational or electromagnetic fields. In quantum mechanics the dichotomy be-
tween particles and fields is upheld, although the wave-particle duality shows
up. But in the relativistic quantum mechanics of particles one encounters
contradictions. These are resolved in Quantum Field Theory (QFT). QFT
deals with quantised fields, functions of space and time, where the values of
the fields f (~r, t) themselves become operators.
field f −→ operator.
QFT is a quantum theory of many particles. In this lecture we consider the

three most prominent relativistic field equations,
• Klein-Gordon equation for spin 0 particles,
• Dirac equation for spin 1/2 particles,
• Maxwell’s equations for massless spin 1 particles.

There are other relativistic equations, too (Proca, etc). In QFT the spin of
fundamental fields does not exceed 2.
1.2.1 Klein-Gordon equation

For a non-relativistic free particle the equation
p~ 2
E= (1.11)
2m
together with de Broglie’s plane wave ansatz
~
ψ = A ei(k·~r−ω t) , E = ~ω, p~ = ~~k (1.12)
leads to the non-relativistic Schrödinger equation

∂ ~2 2
i~ ψ=− ∇ ψ. (1.13)
∂t 2m
For a relativistic particle with E = ~ω = c p0 energy and momentum are
components of a four-vector and our starting point is the equation for the
square of the 4-momentum
E 2 = c2 p~ 2 + m2 c4 , (1.14)
14 1 INTRODUCTION
which leads to
∂2
−~2 φ = −c2 ~2 ∇2 φ + m2 c4 φ
∂t2
or !
∂2 2 m2 c2
− + ∇ − φ = 0. (1.15)
∂(ct)2 ~2
This is the Klein-Gordon equation, invented by Schrödinger, Fock and others,
and rediscovered by Klein and Gordon.
Relativistic notations
x0 = ct, x1 = x, x2 = y, x3 = z
x = (x0 , x1 , x2 , x3 ) = (x0 , ~x) = (xµ )
 
1

 −1 

gµ ν =  
 −1 
−1
x · y = x0 y 0 − ~x · ~y = xµ xµ , xµ = gµ ν xν
!
∂ 1∂
∂µ = = , ∇
∂xµ c ∂t
!
µ ∂ 1∂
∂ = = , −∇
∂xµ c ∂t
∂2
= −∂µ ∂ µ = − +∆
∂(ct)2

E
pµ = , p~
c
E2
p2 = pµ pµ = 2 − p~ 2 = m2 c2 Note3
c
pµ → i~∂ µ de Broglie
!
m2 c2
− φ(x) = 0 Klein-Gordon
~2
4
From now on we use natural units setting ~ = c = 1.
The symbol p2 is ambiguous. Its meaning must be determined from the context.
3
To go back to SI-units in an equation one may analyse the dimension of the terms and
4
insert ~ and/or c to get the right dimension.

Solution of the Klein-Gordon equation

Let φ(x) be a complex scalar field (φ ∈ C), that means, it is not quantised
yet (φ is not operator-valued), Spin = 0 (φ is scalar). It will turn out that
complex scalar fields describe particles with positive and negative charges.
Examples are the mesons π + and π − .
A general solution to the Klein-Gordon equation for free particles, being
linear and of second order, is a superposition of two plane waves
Z
d3 k n
−ikx ∗ ikx
o
φ(x) = a(k) e + b (k) e . (1.16)
(2π)3 2ωk
Here we denoted q
ωk = k = 0 ~k 2 + m2 (1.17)
to be a positive frequency. The solution is verified by
µ 0 x0 −k j xj )
∂µ ∂ µ eikµ x = (∂0 2 − ∂j ∂j ) ei(k
µ
= i2 (k 0 k 0 − k j k j )eikµ x
µ
= −kµ k µ eikµ x
µ µ
( − m2 )eikµ x = (kµ k µ − m2 )eikµ x

2
= (k 0 )2 − (~k + m2 ) eikµ x = 0.
µ
Now let φ(x) be a real scalar field, φ ∈ R, being used for neutral spin 0
particles, like π 0 . Then the general solution is
Z
d3 k n
−ikx ∗ ikx
o
φ(x) = a(k) e + a (k) e . (1.18)
(2π)3 2ωk
The problem with negative frequencies

~
ei(k·~x−ωt) ⇒ Eφ = i~∂t φ = +~ωφ
~
e−i(k·~x−ωt) ⇒ Eφ = i~∂t φ = −~ωφ
So, free particles could have arbitrarily large negative energies, which is un-
physical. In the presence of interactions, e.g. with the electromagnetic field,
this would lead to instabilities, because a particle would jump to lower and
lower states, emitting an unbounded amount of energy. This problem will be
solved by field quantisation.
16 1 INTRODUCTION
1.2.2 Dirac equation

The Dirac equation was found by P. A. M. Dirac in 1928. He was searching
for a covariant version of the Schrödinger equation
i∂t ψ = Hψ. (1.19)
To be manifestly covariant, it has to be of first order in the spatial derivatives,

too.
∂
H linear in (k = 1, 2, 3),
∂xk
3
X 3
X
H= αk P k + βm = − αk i∂k + βm. (1.20)
k=1 k=1
We will now derive conditions for the constant terms αk and β. Squaring
both sides of the equation we get
−(∂ 0 )2 ψ = H 2 ψ (1.21)
3
1 X
= (αj αk + αk αj )Pj Pk ψ
2 j,k=1
3
X
−m (αk β + βαk )Pk ψ
k=1
2 2
+ β m ψ.
Consider a plane wave ψ = eipx , for which one should have
i∂ 0 ψ = Eψ, P~ ψ = p~ ψ, E 2 = p~ 2 + m2 . (1.22)
This wave satisfies the Dirac equation only if
αj αk + αk αj = 2δjk 1 (1.23)
αk β + βαk = 0 (1.24)
β 2 = 1. (1.25)
From this one concludes that αk and β cannot be numbers. The relations
can be satisfied by matrices, which must at least be of size 4 by 4. They can
be composed by blocks of Pauli spin matrices
! !
0 σk 1 0
αk = , β= . (1.26)
σk 0 0 −1
By convention one uses the Dirac matrices γ µ :
γ 0 := β, γ k := β αk , (k = 1, 2, 3) (1.27)
The Hamiltonian can be written

3
X
H = −γ 0 γ k i∂k + γ 0 m,
k=1
and multiplying with γ 0 we get

3
X
γ 0 i∂0 ψ = γ 0 Hψ = − iγ k ∂k ψ + mψ.
k=1
This is the Dirac equation, reading
(iγ µ ∂µ − m)ψ(x) = 0. (1.28)
Equivalent notations of Dirac’s equation are
(γ µ Pµ − m)ψ(x) = (p/ − m)ψ(x) = 0.
The algebra of the γ-symbols is
γ µ γ ν + γ ν γ µ = 2g µ ν 1. (1.29)
The matrices given above are Dirac’s representation of the γ’s. There are
others representations, e.g. by Weyl or by Majorana. The Dirac matrices
written in blocks of Pauli spin matrices are
! !
1 0 0 σk
γ0 = , γk = . (1.30)
0 −1 −σk 0
Solutions of Dirac’s equation will be given by spinor wavefunctions or fields

 
ψ1 (x)
 .. 
ψ(x) =  .  . (1.31)
ψ4 (x)
These are made out of two kinds of plane waves, given by
ψ(x) = u(k)e−ikx , k 0 = ωk > 0 (1.32)
with 2 independent spinors u(r) (k), r = 1, 2, and
ψ(x) = v(k)eikx , k 0 = ωk > 0 (1.33)

18 1 INTRODUCTION
with another 2 independent spinors v (r) (k), r = 1, 2. The general solution

is a superposition
2 n
Z
d3 k X o
~k ) u(r)(~k ) e−ikx + d ∗ (~k ) v (r)(~k ) eikx .
ψ(x) = b r ( r (1.34)
(2π)3 2ωk r=1
Spin
~ =
The Dirac Hamiltonian H and the orbital angular momentum operator L
~ ~
R × P do not commute h i
~ H 6= 0.
L, (1.35)
The angular momentum of free particles should be conserved! So there must
be an additional hidden contribution to the angular momentum, which is
called spin. !
~ ~~ ~ ~σ 0
S= Σ= . (1.36)
2 2 0 ~σ
~ is given by
The algebra of S
[Sk , Sl ] = i Sm (k, l, m) = (1, 2, 3) + cycl. (1.37)
~ 2 = 3 1 = s(s + 1)1 ⇒ s = 1 .
S (1.38)
4 2
~ ~ ~
The total angular momentum J = L + S obeys
~ H] = 0.
[J, (1.39)
Thus the total angular momentum is conserved.

Notice: The last commutator can be verified with the help of
1
Σ1 = [γ2 , γ3 ]. (1.40)
2
Covariant expressions
To write down Lorentz covariant expressions with ψ
 
ψ1 (x)
 .. 

ψ(x) =  .  , ψ † = ψ1∗ (x), . . . , ψ4∗ (x) , (1.41)
ψ4 (x),
we define the Dirac conjugate
ψ̄(x) = ψ † (x)γ 0 . (1.42)

Covariant scalar and vector expressions are
ψ̄(x)ψ(x), ψ̄(x)γ µ ψ(x); (1.43)
Objects which transform under an antisymmetric tensor representation of

the Lorentz group are
ψ̄σµν ψ, σµν = [γµ , γν ]. (1.44)
With !
5 0 1
γ5 = γ := iγ1 γ2 γ3 γ4 = (1.45)
1 0
pseudoscalars and pseudovectors are given by
ψ̄γ5 ψ, ψ̄γ5 γ µ ψ. (1.46)
1.2.3 Maxwell’s equations

Maxwell’s equations in the MKSA system read
~ =ρ
∇·E (1.47)
ǫ0
~ =0
∇·B (1.48)
~
~ = − ∂B
∇×E (1.49)
∂t
~
~ = µ0~j + µ0 ǫ0 ∂ E
∇×B (1.50)
∂t
In QFT often the Heaviside-Lorentz unit system is used. Conversion formulae
are:
~ H = √ǫ0 E
E ~ (1.51)
~H = 1 B
B ~ (1.52)
µ0
√
ΦH = ǫ 0 Φ (1.53)
~ H = √1 A
A ~ (1.54)
µ0
1
ρH = √ ρ (1.55)
ǫ0
~jH = √1 ~j. (1.56)
ǫ0
20 1 INTRODUCTION
Now the Maxwell equations in Heaviside-Lorentz units read
∇·E ~ =ρ (1.57)
∇·B ~ =0 (1.58)
~
~ + 1 ∂B = 0
∇×E (1.59)
c ∂t
~
~ − 1 ∂ E = ~j
∇×B (1.60)
c ∂t
The fields can be derived from potentials
~
~ = ∇ × A,
B ~ ~ = −∇Φ − 1 ∂ A
E (1.61)
c ∂t
Equivalently the potentials are written in covariant form
~
Aµ (x) := (Φ(x), A(x)). (1.62)
From these the field strengths are derived by
F µν = ∂ µ Aν − ∂ ν Aµ , (µ, ν = 0, 1, 2, 3) (1.63)
 
0 −Ex −Ey −Ez
E 0 −Bz By 
F µν = x
 
 (1.64)
Ey Bz 0 −Bx 
Ez −By Bx 0
or
1
Ei = F i 0 , Bi = − ǫijk F jk . ( i, j, k ∈ {1, 2, 3}) (1.65)
2
For the the 4-vector current density
j µ := (ρ, ~j) (1.66)
Maxwell’s equations give the continuity equation
∂µ j µ = 0 (1.67)
or
∂
ρ + ∇ · ~j = 0. (1.68)
∂t
Maxwell’s equations themselves may be written in covariant form also:
inhomogeneous equations
∂µ F µν = j ν (1.69)
and homogeneous equations
∂ µ F νρ + ∂ ν F ρµ + ∂ ρ F µν = 0. (1.70)
Gauge freedom, Lorenz gauge

(Ludvig Lorenz, 1867; George F. FitzGerald, 1888)
∂µ Aµ = 0 (1.71)
If one fixes the Lorenz gauge in one inertial frame, then it is fulfilled in all
inertial frames. Let us consider again free fields,
j µ = 0, (1.72)
∂µ F µν = 0. (1.73)
Together with the Lorenz gauge we get
0 = ∂µ F µν = ∂µ (∂ µ Aν − ∂ ν Aµ ) = ∂µ ∂ µ Aν − ∂ ν ∂µ Aµ = ∂µ ∂ µ Aν ,
Aν = 0. (1.74)
To solve the wave equation we take the plane wave ansatz
Aµ (x) = ǫ(λ)
µ e
ikx
. (1.75)
From Lorenz gauge it follows
k · k = 0, k 0 = |~k| = ωk . (1.76)
There are remaining superfluous degrees of freedom. The Coulomb gauge

for a field free of sources fixes
Φ = 0, ~ = 0.
∇·A (1.77)
For the plane wave solutions this implies
ǫ0 = 0, ǫ · k = 0. (1.78)
Thus there are 2 linearly independent solutions, representing the 2 transversal

polarisations of radiation
ǫ(1) (2)
µ (k), ǫµ (k) ⊥ [(1, 0, 0, 0), k]. (1.79)
The two transversal polarisations imply that the photon spin (s = 1) is in

the direction of propagation.
22 1 INTRODUCTION
For a massive particle moving in a certain direction and having its spin
parallel to its velocity, a different inertial frame can be chosen such that in this
frame the particle moves in the opposite direction and its spin is antiparallel
to the velocity. Therefore the projection of its spin on the velocity is not
invariant under Lorentz transformation. On the other hand, for massless
particles travelling with the velocity of light, the projection of the spin on
the velocity is Lorentz-invariant and is called “helicity”:
JS = ±1. (1.80)
The general solution of the electromagnetic wave equation in the Coulomb

gauge is
2
Z
d3 k X
(λ)

(λ) −ikx (λ) ∗ ikx

Aµ (x) = ǫ (k) a (k) e + a (k) e . (1.81)
(2π)3 2ωk λ=1 µ
1.2.4 Lagrangian formalism for fields

Recapitulation: Classical mechanics
m~r¨ = −∇V (~r ), (1.82)
p2
H= + V (~r ) with p~ = m~r˙. (1.83)
2m
Hamilton’s equations give the equation of motion. Hamilton’s principle uses
the action S, build from the Lagrangian L:
Z
S= dt L(~r(t), ~r˙ (t)) (1.84)
m ˙2
L= ~r − V (~r ). (1.85)
2
The realised trajectories ~r(t)
~r1
~r(t)
~r0
are such that the action S is stationary under infinitesimal variations δ~r(t)
provided the endpoints ~r(t0 ) and ~r(t1 ) are fixed:
~r ′ (t) = ~r(t) + δ~r(t) (1.86)

δ~r(t0 ) = δ~r(t1 ) = 0. (1.87)
δS = 0 (1.88)
The calculus of variation leads to the equations of motion:

Z t1 Z t1 !
X ∂L ∂L
δS = dt δL = dt δxi + δ ẋi (1.89)
t0 t0 i ∂xi ∂ ẋi
Z t1 !
X ∂L d ∂L
= dt − δxi = 0. (1.90)
t0 i ∂xi dt ∂ ẋi
The fundamental lemma of the calculus of variation then yields the Euler-
Lagrange equations
∂L d ∂L
− = 0. (1.91)
∂xi dt ∂ ẋi
This procedure can be taken over to field theory. An advantage is that sym-
metries in the action S directly show up as symmetries in the field equations.
The Lagrangian density L in the field variables φa and their derivatives ∂µ φa
shall be denoted by
L (φa (x), ∂µ φa (x)); (1.92)
here φa (x) and ∂µ φa (x) take over the rôle of infinitely many coordinates xi
and velocities ẋi , while the argument x = (xµ ) of φa (x) takes over the rôle
of time t in mechanics. The action is
Z
S= d4 x L (φa (x), ∂µ φa (x)). (1.93)
G
Consider now small variations of φa (x) fixed at the boundary ∂G of the

domain of integration G. Hamilton’s principle δS = 0 leads to
Z !
4 ∂L ∂L
0= dx δφa (x) + δ∂µ φa (x) (1.94)
G ∂φa (x) ∂(∂µ φa (x))
Z !
∂L ∂L
= d4 x − ∂µ δφa (x). (1.95)
G ∂φa (x) ∂(∂µ φa (x))
Here we performed a partial integration and used the fact that the integrated
part vanishes due to δφa = 0 on the boundary ∂G. To see this in detail, let
∂L
B µ := ,
∂(∂µ φa (x))
∂µ (B µ δφa ) = (∂µ B µ )δφa + B µ ∂µ δφa ,
from Leibniz’s product rule. From Stokes theorem we get

Z Z
d4 x ∂µ (B µ δφa ) = dxµ B µ δφa = 0,
G ∂G
24 1 INTRODUCTION
since δφa = 0 on the boundary ∂G. Thus we can replace B µ ∂µ δφa by

−(∂µ B µ )δφa in the integral.
We end up with the Lagrange field equations
∂L ∂L
− ∂µ = 0. (1.96)
∂φa (x) ∂(∂µ φa (x))
Reasons for using Lagrangian densities:

a) There is a single function L instead of many field equations.
b) There are advances when non-Cartesian coordinates are used – similar
as in mechanics.
c) Symmetries can be expressed in a simple manner. Noether theorems
lead to conservation laws.
d) Gauge theories can be quantised in a simpler way.
Real scalar field

1 m2 2
L = ∂µ φ ∂ µ φ − φ (1.97)
2 2
The field equations are linear equations, therefore the Lagrangian has to be
quadratic in the field and its derivatives. Here we have the simplest expres-
sion for L being quadratic and Lorentz invariant. We derive the Lagrange
equations of motion:
!
∂L 1 µ ∂
= ∂ φ + ∂λ φ g λν ∂ν φ
∂(∂µ φ) 2 ∂(∂µ φ)
!
1 µ λν ∂(∂ν φ)
= ∂ φ + g ∂λ φ
2 ∂(∂µ φ)
1
= (∂ µ φ + ∂ ν φ δνµ )
2
= ∂µφ (1.98)
∂L
∂µ = ∂µ ∂ µ φ = −φ (1.99)
∂(∂µ φ)
∂L
= −m2 φ (1.100)
∂φ
This gives the Klein-Gordon-equation

∂µ ∂ µ + m2 φ = 0 (1.101)
Complex scalar field
L = ∂µ φ∗ ∂ µ φ − m2 φ∗ φ (1.102)
φ∗ and φ are not totally independent complex-valued fields – there are not
4 independent real-valued fields. The derivatives ∂φ and ∂φ∗ also do not
give further freedom, as they are connected by Cauchy-Riemann differential
equations. 5
We separate L into independent parts by means of φ = √12 (φ1 +iφ2 ), φ∗ =
√1 (φ1 − iφ2 ):
2
1 1
L = ∂µ φ1 ∂ µ φ1 + ∂µ φ2 ∂ µ φ2
2 2
i i
+ ∂µ φ1 ∂ µ φ2 − ∂µ φ2 ∂ µ φ1
2 2
1 2 2 1 2 2
− m φ1 − m φ2
2 2
1 1
L = ∂µ φ1 ∂ µ φ1 − m2 φ21 + ∂µ φ2 ∂ µ φ2 − m2 φ22 . (1.103)
2 2
This is a Lagrangian density for two real scalar fields φ1 and φ2 , each giving
a Klein-Gordon-equation for the real and imaginary parts of φ.
(∂µ ∂ µ + m2 )φa = 0 (a = 1, 2). (1.104)
Sum and difference of both equations give two identical equations:
(∂µ ∂ µ + m2 )φ = 0 (1.105)
(∂µ ∂ µ + m2 )φ∗ = 0 (1.106)
Dirac field
ψ(x) = (ψ1 (x), . . . , ψ4 (x))⊤ (1.107)
ψ is not a 4-vector like x or Aµ but a spinor with 4 complex-valued compo-
nents. Thus ψ describes 8 real fields.
L (ψ, ∂ψ) = ψ̄(x)(iγ µ ∂µ − m)ψ(x), (1.108)

5
The Cauchy-Riemann differential equations with z = x + iy, φ(z) = φ1 (z) + iφ2 (z)
∂φ∗
read ∂φ ∂x = ∂y
1 ∂φ2
∧ ∂φ ∂φ2
∂y = − ∂x . – Furthermore a term like ∂φ must not be seen as
1
a derivative in the complex plane, for that would not exist. Instead the partial derivative
∂φ should be considered as a vector ∂φ = ( ∂φ1 , ∂iφ2 ).
∂ ∂ ∂ ∂
26 1 INTRODUCTION
where
ψ̄ := ψ † γ 0 = (ψ1∗ , ψ2∗ , −ψ3∗ , −ψ4∗ ). (1.109)
In this form of L the derivatives act on ψ and not on ψ̄. A more symmetric
alternative would be
1 n µ− → ←− o
L = ψ̄ iγ ∂ µ − m ψ + ψ̄ −iγ µ ∂ µ − m ψ , (1.110)
2
where the arrows indicate whether the derivative acts to the right on ψ or to
the left on ψ̄. The two versions for L differ by a total derivative 12 ∂µ (ψ̄iγ µ ψ),
which does not change the field equations. The resulting field equations are
the Dirac equation and its Dirac-conjugate equation
(iγ µ ∂µ − m)ψ = 0, (1.111)
i∂µ ψ̄γ µ + mψ̄ = 0. (1.112)

Taking the first Lagrangian, Eq. (1.108), we have
∂L ∂L
= −mψ̄, ∂µ = ∂µ ψ̄iγ µ ,
∂ψ ∂(∂µ ψ)
∂L ∂L
= (iγ µ ∂µ − m)ψ(x), ∂µ = 0.
∂ψ ∂(∂µ ψ)
This leads to the given Lagrangian equations (1.111) and (1.112).

Maxwell field
The expression of the field strengths in terms of the vector potential Aµ (x),
Fµν = ∂µ Aν − ∂ν Aµ , (1.113)
implies that the homogeneous Maxwell equations
∂ µ F νρ + ∂ ν F ρµ + ∂ ρ F µν = 0 (1.114)
are automatically fulfilled. The inhomogeneous equations
∂µ F µν = j ν (1.115)
should be derived from the Lagrangian. Let us look for a Lorentz invariant
and gauge invariant Lagrangian. The gauge transformations of electrody-
namics, which leave the field strengths invariant,
~ + ∇χ and Φ′ = Φ − 1 χ̇ ,
~′ = A
A
c
read covariantly
A′µ = Aµ − ∂ µ χ .
The Lagrangian cannot contain a mass term m2 Aµ Aµ since this is not gauge
invariant, A′µ A′µ 6= Aµ Aµ . The field strengths are gauge invariant per defini-
tion, therefore the Lagrangian density
1
L (Aµ , ∂ν Aµ ) = − Fµν F µν (1.116)
4
is both Lorentz invariant and gauge invariant. It is the only such choice being
quadratic in the fields.
In the Lagrangian, F µν is to be considered as a function of Aµ (x) and
∂ν Aµ (x), thus
1
L = − (∂µ Aν − ∂ν Aµ )(∂ µ Aν − ∂ ν Aµ ) (1.117)
4
and 6
∂L ∂L
= 0, = F νµ . (1.118)
∂Aµ ∂(∂µ Aν )
From this the field equations are
∂µ F µν = 0. (1.119)
The charges and currents, which enter the inhomogeneous equations, are
themselves fields and should be dealt with by other field equations. On the
6
∂L 1 ∂
− = (∂µ Aν − ∂ν Aµ )g µλ g νρ (∂λ Aρ − ∂ρ Aλ )
∂(∂α Aβ ) 4 ∂(∂α Aβ )

1 µλ νρ ∂
= g g (∂µ Aν − ∂ν Aµ ) (∂λ Aρ − ∂ρ Aλ )
4 ∂(∂α Aβ )

1 µλ νρ ∂
+ g g (∂µ Aν − ∂ν Aµ ) (∂λ Aρ − ∂ρ Aλ )
4 ∂(∂α Aβ )
1 µλ νρ
= g g (δαµ δβν − δαν δβµ )(∂λ Aρ − ∂ρ Aλ )
4
1 µλ νρ
+ g g (∂µ Aν − ∂ν Aµ )(δαλ δβρ − δαρ δβλ )
4
1
= (δαµ δβν − δαν δβµ )F µν + F λρ (δαλ δβρ − δαρ δβλ )
4
1
= (F αβ − F βα + F αβ − F βα )
4
= F αβ
28 1 INTRODUCTION
other hand, it is possible to introduce them as external sources j µ in the

Lagrangian. The sources obey a continuity equation
∂µ j µ = 0. (1.120)
With the Lagrangian

1
L = − Fµν F µν − j µ (x)Aµ (x) (1.121)
4
one gets the inhomogeneous Maxwell equations
∂ν F µν + j µ = 0. (1.122)
The Lagrangian with external currents is not gauge invariant. The action,
however, does not change under a gauge transformation, since
Z Z
d4 x j µ ∂µ χ = − d4 x χ∂µ j µ = 0. (1.123)
Lagrangian for a massive vector field

The field, denoted by
Bµ (x), (1.124)
is a Lorentz vector. In analogy to the Maxwell field we define
Gµν := ∂µ Bν − ∂ν Bµ , (1.125)
and set
1 m2
L = − Gµν Gµν + Bµ B µ . (1.126)
4 2
With
∂L ∂L
= m2 B µ , = Gνµ (1.127)
∂Bµ ∂(∂µ Bν )
we get the field equations
∂ν Gµν − m2 B µ = 0, (1.128)
and with ∂ν Gµν = ∂ν ∂ µ B ν − ∂ν ∂ ν B µ this yields

∂ν ∂ ν + m2 B µ = ∂ µ ∂ν B ν . (1.129)
These are four field equations containing the Klein-Gordon operator. Taking
the 4-divergence we get

∂ν ∂ ν + m2 ∂µ B µ = ∂µ ∂ µ ∂ν B ν ,
⇒ m2 ∂µ B µ = 0. (1.130)
Thus the field equations can be represented equivalently by four Klein-
Gordon equations augmented by an equation which looks like a Lorenz gauge:

∂ν ∂ ν + m2 B µ = 0, (1.131)
∂µ B µ = 0. (1.132)
This field describes massive spin 1 particles, like the ρ±,0 and ω mesons.
30 1 INTRODUCTION
1.3 Symmetries
1.3.1 Symmetries and conservation laws
What does it mean to be symmetric? Hermann Weyl described it as follows.
One needs three things:
• An object, which turns out to be symmetric.
• A procedure, doing something with this object.
• An observer, who states that after the procedure nothing has changed.
In physics we may say, a symmetry is “a mapping, which does not change

the physics.” The meaning of the stated invariance depends on the structure
that defines what “is the same”. For instance, in Euclidean geometry a circle
is symmetric, any closed loop in general not.
circle closed loop

On the other hand, in topology the closed loop is regarded as equivalent
to the circle, the deformation of the circle to a closed loop is a symmetry
transformation in topology, but not in Euclidean geometry.
Symmetry transformations can be concatenated to give a symmetry transfor-
mation again. As with any mapping an associative rule is valid. The inverse
transformation is also a symmetry, so symmetries form groups. (With few
exceptions, when a semigroup is also regarded as a symmetry.) The most
prominent symmetry in everyday life, the mirror reflection symmetry, be-
longs to a very small group of only two elements: reflection and identity. In
this lecture we consider continuous symmetry groups and their associated
conservation laws, which Emmy Noether found in classical mechanics and
field theory. Most of her work can be transferred to quantum mechanics.
In quantum theory also discrete symmetry groups are being considered, for
instance parity. (Here the word “parity” denotes both the symmetry and the
conserved quantity.)
Lagrange formalism
In the Lagrange formalism symmetries can be dealt with by infinitesimal
symmetry transformations, which is simpler than using finite transformations
1.3 Symmetries 31
although this would be possible, too. Let the fields undergo an infinitesimal
transformation, which is very close to the identity.
φa −→ φ′a = φa + δφa . (1.133)
The change of the action is written
S −→ S ′ = S + δS. (1.134)
If the transformation is a symmetry, the system should not change, which

means that the action does not change. Thus symmetry is characterised by
δS = 0, if φa −→ φa + δφa . (1.135)
Here the field equations are not used. In this case one has:
if φ is a solution of the field equations,

then φ′ is also a solution of the field equations.
The Noether theorem states that associated with such a symmetry there
exists a conserved current j µ (x),
∂µ j µ (x) = 0. (1.136)
The conserved quantity is given by

Z
Q= d3 x j 0 (x),
dQ Z Z
~
Z
= 3 0
d x ∂0 j (x) = − 3
d x ∇ · j(x) = − d2~x · ~j(x) = 0.
dt R3 R3 ∂R3
Symmetries might involve space-time transformations, where δx 6= 0. For

example, under translations the fields transform as
φ′a (x) = φa (x − δx). (1.137)
If the system is invariant under translations, the conserved quantity is the

energy-momentum four-vector pµ . Similarly, invariance under rotations gives
the conservation of angular momentum. For the application of Noether’s
theorem to such space-time symmetries we refer to the textbooks.
Here we restrict our discussion to internal symmetries, for which δx = 0.
φa (x) −→ φa (x) + δφa (x). (1.138)

32 1 INTRODUCTION
Symmetry means invariance of the action

Z
S= d4x L (φa (x), ∂φa (x)). (1.139)
We consider the stronger condition that the Lagrangian density is invariant:

δL = 0.
∂L ∂L
δL = δφa + δ∂µ φa
∂φa ∂(∂µ φa )
" # " #
∂L ∂L ∂L
= δφa − ∂µ δφa + ∂µ δφa (1.140)
∂φa ∂(∂µ φa ) ∂(∂µ φa )
= 0.
Here we have used δ∂µ φa = ∂µ δφa and the Leibniz product rule for differ-
entiation. The last term in the above equation already has the form of a
4-divergence ∂µ ( )µ . We write
δφa = ǫfa (1.141)
with infinitesimal small ǫ and finite fa and define
∂L
j µ (x) := fa (1.142)
∂(∂µ φa )
to get ( )
µ ∂L ∂L
∂µ j = ∂µ − fa . (1.143)
∂(∂µ φa ) ∂φa
Consequence: If the field equations hold, the current j µ is conserved
∂µ j µ = 0. (1.144)
1.3.2 U(1) symmetry, electric charge

7
Let φ be a complex scalar field.
φ(x) −→ φ′ (x) = e−iqα φ(x), (q ∈ Z, α ∈ R). (1.145)
The transformation group has the representation
U(1) = {c ∈ C | c∗ c = 1} = {c ∈ C | c = e−iγ , −π ≤ γ < π}. (1.146)

7
It could also be a Dirac field or a field for particles with higher spin.
1.3 Symmetries 33
U(1) is an abelian group generated by the infinitesimal transformation
φ′ (x) = (1 − iqǫ)φ(x)
= φ(x) − iqφ(x)ǫ (1.147)
δφ(x) = −iqφ(x)ǫ (1.148)
δφ∗ (x) = iqφ∗ (x)ǫ. (1.149)
We assume a symmetry under U(1),
δL = 0, (1.150)
as is the case with the Lagrangian for a complex scalar field,
1 m2 ∗
L = ∂µ φ∗ ∂ µ φ − φ φ. (1.151)
2 2
The Lagrangian is in fact invariant under finite U(1)-transformations. Now
we have to find the Noether current.
∂L
jµ = fa , f1 = −iqφ, f2 = iqφ∗
∂(∂µ φa )
∂L ∂L
= (−iqφ) + (iqφ∗ )
∂(∂µ φ) ∂(∂µ φ∗ )
j µ = iq(φ∗ ∂ µ φ − φ ∂ µ φ∗ ). (1.152)
We note that the spatial part of this current has the same form as the prob-
ability current in in Schrödinger theory, which is defined by ~j = 2mi
~
(ψ∇ψ ∗ −
ψ ∗ ∇ψ). Let us check the conservation law
1
∂µ j µ = ∂µ (φ∗ ∂ µ φ − φ ∂ µ φ∗ ) (1.153)
iq
= ∂µ φ∗ ∂ µ φ + φ∗ ∂µ ∂ µ φ − ∂µ φ ∂ µ φ∗ − φ ∂µ ∂ µ φ∗
= φ∗ ∂µ ∂ µ φ − φ ∂µ ∂ µ φ∗
= φ∗ (∂µ ∂ µ + m2 )φ − φ(∂µ ∂ µ + m2 )φ∗ . (1.154)
If the field equation holds, then
∂µ j µ = 0. (1.155)
The conserved charge is given by

Z Z
3 0
Q= d x j = iq d3 x (φ∗ ∂0 φ − φ∂0 φ∗ ), (1.156)
34 1 INTRODUCTION
q may be an integer number.

For the Dirac field we have a similar U(1) symmetry. The transformation is
ψ ′ (x) = e−iqα ψ(x), (1.157)

ψ̄ ′ (x) = eiqα ψ̄(x), (1.158)
and the Lagrangian density
L = ψ̄(iγ µ ∂µ − m)ψ (1.159)
again is invariant under this transformation. For the current and charge we
get
∂L
jµ = (−iqψα ) = q ψ̄γ µ ψ, (1.160)
∂(∂µ ψα )
Z Z
3 0
Q=q d x ψ̄γ ψ = q d3 x ψ † ψ, (1.161)
ψ † ψ is the charge density.
1.3.3 SU(2) symmetry, isospin

Consider neutron and proton, described by two Dirac fields
p(x) = (pα (x)), α = 1, . . . , 4, (1.162)

n(x) = (nα (x)), α = 1, . . . , 4. (1.163)
The nuclear forces are independent of the electric charge. They are the same
for proton and neutron. The idea of isospin (Werner Heisenberg, Dmitri
Ivanenko) is to describe this in terms of a symmetry. The situation is in
analogy with the two spin states of the electron, which form a basis of a
two-dimensional sub-Hilbert space
! !
1 0
| ↑i = , | ↓i = . (1.164)
0 1
In the absence of a magnetic field both states carry the same energy, the
Hamiltonian is invariant under a rotation in spin space, made up of the
Pauli spinors !
ψ+ (x)
ψ(x) = . (1.165)
ψ− (x)
The symmetry transformation shall be denoted by
ψ −→ U (α)ψ (1.166)
1.3 Symmetries 35
with angles α that parameterise the rotation.

In analogy to spin Heisenberg introduced isotopic spin for proton and neu-
tron, sometimes called isobaric spin; today it is mostly called isospin. The
nucleon is represented by !
N1 (x)
N= , (1.167)
N2 (x)
where N1 and N2 are Dirac spinors
Ni,α (x), i = 1, 2; α = 1, . . . , 4. (1.168)
The pure proton or neutron states are (somewhat ambiguously) denoted as

! !
p(x) 0
p(x) = , n(x) = , (1.169)
0 n(x)
and the nucleon field is !

p(x)
N (x) = . (1.170)
n(x)
The corresponding isospin observable is
1
I~ = ~τ . (1.171)
2
!
1 0 1
I1 = τ1 , τ1 =
2 1 0
!
1 0 −i
I2 = τ2 , τ2 =
2 i 0
!
1 1 0
I3 = τ3 , τ3 =
2 0 −1
Thus
1 1
I3 p(x) = + p(x), I3 n(x) = − n(x). (1.172)
2 2
Summary:
Isospin Proton Neutron
1 1
I 2 2
I3 + 21 − 12 .
The symmetry group SU(2)

Symmetries may be approximate symmetries, but here we shall make the hy-
pothesis that the Hamiltonian is invariant under a rotation in 2-dimensional
36 1 INTRODUCTION
isospin space. Because of the form of the free Hamiltonian, the symmetry
transformation must be unitary. We write
N −→ N ′ = U N. (1.173)
The group of such transformations, U(2), is made from 2×2 unitary matrices
U.
U †U = 1 (1.174)
det(U † U ) = det(U † ) det(U ) = |det(U )|2 = 1 (1.175)
| det(U )| = 1 (1.176)
The subgroup that has det U = +1 is called SU(2):
SU(2) = {U ∈ GL2 (C) | U † U = 1, det U = +1}. (1.177)

!
a b
U= (1.178)
c d
! ! ! !
† a∗ c∗ a b a∗ a + c∗ c a∗ b + c∗ d 1 0
U U= ∗ ∗ = ∗ = .
b d c d b a + d∗ c b∗ b + d∗ d 0 1
The unitarity of U implies that the vector norm of the columns has to be 1,
the same holds for the rows of U since U † is also unitary. Together we have
1 = det U = ad − bc
1 = a∗ a + c∗ c = b∗ b + d∗ d = a∗ a + b∗ b = c∗ c + d∗ d
leading to a∗ a = d∗ d, b∗ b = c∗ c. From this we can restrict the general form
of U to be 8 !
a b
U= , a∗ a + b∗ b = 1.
−b∗ eiβ a∗ eiα
The determinant gives
1 = a∗ a eiα + b∗ b eiβ = eiα (a∗ a + b∗ b ei(β−α) ).
Together with a∗ a + b∗ b = 1 this is only possible if α − β = 0. Then α = 0

follows, too, and we have
!
a b
U= , a∗ a + b∗ b = 1. (1.179)
−b a∗
∗
b∗ b a∗ a
8
c= c∗ , |b| = |c| gives c = −b∗ eiα and d = d∗ , |a| = |d| gives d = a∗ eiβ .
1.3 Symmetries 37
Now we may choose four real parameters, together with one condition, to
characterise U 9
! 3
a − ia3 −a2 − ia1 X
U= 0 , a2k = 1. (1.180)
a2 − ia1 a0 + ia3 k=0
P3
This can be written as U = a0 1 − ia1 τ1 − ia2 τ2 − ia3 τ3 . Since k=0 a2k = 1,
one may write
α
a0 = cos( ), 0 ≤ α < 2π, (1.181)
2
α
(a1 , a2 , a3 ) = sin( ) ~n, |~n| = 1. (1.182)
2
Then
α α
U = cos( )1 − sin( )(in1 τ1 + in2 τ2 + in3 τ3 ). (1.183)
2 2
With the definition
α
~ := α ~n (1.184)
we finally write it in a way standard for Lie group representations:
3
αX ~
U = exp(−i nk τk ) = exp(−i~
α · I). (1.185)
2 k=1
That the two expressions for U , (1.183) and (1.185), are equal can easily seen
to first order in α:
α α α α
cos( )1 − i sin( )(~n · ~τ ) ≈ 1 − i ~n · ~τ ≈ exp(−i ~n · ~τ ).
2 2 2 2
It can be shown that they are equal in general.
I~ = ~τ /2 is the Hermitian matrix for the isospin observable. The anti-
Hermitian, traceless exponent is a common feature of Lie algebras.
Once again we state that the situation is similar to quantum mechanics,
where U (~α) in Pauli spinor space describes a rotation by an angle α with a
rotation axis given by ~n. The matrix elements in this representation of SU(2)
are determined by 3 real parameters (α1 , α2 , α3 ) = α
~.
9
In general, the matrices representing the group
SU(N ) = {U ∈ GLn | U † U = 1, det U = 1}
have N 2 − 1 real parameters.

38 1 INTRODUCTION
Lie algebra
Consider an infinitesimal transformation, that means a transformation (1.185)
with infinitesimal small α:
δ~
α = δα ~n, (1.186)
U (δ~
α) = 1 − i δ~ ~
α · I, (1.187)
α · I~ N,
N ′ = N − i δ~ (1.188)
α · I~ N.
δN = −i δ~ (1.189)
The isospin operators (I1 , I2 , I3 ) are called generators of the Lie group SU(2).
From unitarity it follows that they are Hermitian and traceless, 10
Ik† = Ik , tr(Ik ) = 0. (1.190)
Their commutators are the same as those of the usual spin operators:
[Ik , Il ] = iǫklm Im , (1.191)
This is the Lie algebra of SU(2). ǫklm are the structure constants of the Lie
algebra spanned by I1 , I2 , I3 .
Isospin symmetry
Invariant expressions
Expressions which are invariant under isospin symmetry are useful as building
blocks for
an invariant Lagrangian. Here are some expressions containing
p
N= n .
N † N = p† p + n† n : (U N )† U N = N † U † U N = N † N (1.192)
N † γ µ N = p† γ µ p + n† γ µ n (1.193)
N̄ N (1.194)
11
N̄ ∂µ N (1.195)
An invariant Lagrangian for free nucleons is
L = N̄ (x)(iγ µ ∂µ − m)N (x) (1.196)

= p̄(x)(iγ µ ∂µ − m)p(x) + n̄(x)(iγ µ ∂µ − m)n(x).
10
Every anti-Hermitian operator can be decomposed into a traceless part and an imagi-
nary multiple of the identity iH = iJ + i tr(H)1. Then exp(iH) = exp(i tr(H) ) exp(iJ) ∈
U(1) ⊗ SU(n) with n = dimension of the Hilbert(sub)space.
11
The isospin symmetry does not depend on space-time.
1.3 Symmetries 39
Isospin symmetry implies that m = mp = mn . Experimental values for

proton and neutron masses are
mp = 938.272 MeV (1.197)

mn = 939.565 MeV. (1.198)
The isospin symmetry is nearly perfect, it is broken by the electromagnetic

interaction and by differences between the masses of up- and down-quarks.
p n
1
I 1/2 1/2 Q= + I3 . (1.199)
2
I3 1/2 −1/2
Other representations of SU(2)

Similar to higher spin quantum numbers, belonging to operators in higher
dimensional spin subspaces, there are other representations of the isospin
symmetry group. The group is represented by matrices in isospin space with
an arbitrary dimension. The dimension of the isospin multiplets is given by
2I + 1, where I = 0, 21 , 1, . . . denotes the isospin quantum number.
Consider I = 1, I3 = −1, 0, 1, giving an isospin triplet, e.g. the pion triplet
 
π+
π =  π0  .
 
(1.200)
π−
Here the quantum numbers for charge and isospin components are equal
Q = I3 . (1.201)
In this I = 1 representation the generators of SU(2) have the following form

12
 
0 1 0
(1) 1
I1 = 1 0 1, (1.202)
2
0 1 0
 
0 −i 0
(1) 1
I2 =  i 0 −i
, (1.203)
2
0 i 0
This follows in the same wayp
12
as one gets the matrices for angular momentum operators:
L± = Lx ± iLy , L± |l, mi = l(l + 1) ∓ m(m ± 1) |l, m ± 1i. From this the images
Lx,y |l, mi as matrix-columns are found using Lx = 12 (L+ + L− ), Ly = 2i1
(L+ − L− ).
40 1 INTRODUCTION
 
1 0 0
(1) 1
I3 = 0 0 0. (1.204)
2
0 0 −1
Quarks
From the composition of the nucleons or the pions one can deduce the isospin
quantum numbers for the quarks. This is true under the assumption that
the quantum numbers of quark content add up to the total quantum number
for the composed particles. From
π + = ud¯
π 0 = uū − dd¯
π − = dū
p = uud¯
n = ddū
we infer that for quarks

1 1 1
I= , I3 u = u, I3 d = − d. (1.205)
2 2 2
Thus the u, d quarks belong to an isospin doublet and their electric charges
Q obey
1
Q = + I3 . (1.206)
6
In general, one defines a hypercharge Y by 13
1
Q = I3 + Y, Y = 2(Q − I3 ). (1.207)
2
Q I I3 Y
u 2/3 1/2 1/2 1/3
d −1/3 1/2 −1/2 1/3
p 1 1/2 1/2 1
n 0 1/2 −1/2 1
π+ 1 1 1 0
0
π 0 1 0 0
π − −1 1 −1 0
13
Here Y is equal to the baryon number B. In general Y = B + S − 13 C, with S and C
expressing strangeness and charm.
1.3 Symmetries 41
Noether currents
The Noether procedure leads to conservation of currents in the form ∂µ j µ = 0.
For the SU(2) isospin symmetry the field φ(x) has components φa (x), one
for each particle of the multiplet
 +
! ! π
u p  0
(φa ) = or or π  . (1.208)
d n −
π
14
The infinitesimal transformation is
α · I~(I) φ.
δφ = −i δ~ (1.209)
The superscript (I) stands for the isospin quantum number. In components
of the multiplet this reads
(I)
δφa = −i δαk (Ik )ab φb . (1.210)
From this we find the current for each of the three isospin components
∂L δφa ∂L (I)
jkµ = = −i (I )ab φb (k = 1, 2, 3). (1.211)
∂(∂µ φa ) δαk ∂(∂µ φa ) k

u 15
As an example consider a quark isospin doublet q = d with Lagrangian
L = q̄(x)(iγ µ ∂µ − m)q(x). (1.212)
(1/2)
jkµ (x) = q̄(x)γ µ Ik q(x), (1.213)
1 µ
¯ µu ,
j1µ = ūγ d + dγ (1.214)
2
µ i
¯ µu ,
j2 = − ūγ µ d − dγ (1.215)
2
µ 1 µ
¯ µd .
j3 = ūγ u − dγ (1.216)
2
The conserved charge Iˆ3 is

Z
1Z 3 ¯ 0 d).
Iˆ3 = d3 x j30 (x) = d x (ūγ 0 u − dγ (1.217)
2
14
See the procedure for the charge conservation by U(1) symmetry, Eqs. (1.145) –
(1.156).
15
See the Lagrangian for the Dirac field in subsection 1.2.4, Eq. (1.108), page 25.
42 1 INTRODUCTION
1.3.4 SU(3) flavour symmetry

The experimentally found hadrons motivated to extend the approximate
isospin SU(2) symmetry to an approximate SU(3) flavour symmetry. It
should be noted that the SU(3) flavour symmetry is a global symmetry,
in contrast to the SU(3)colour symmetry, which is a local gauge symmetry.
In nature the SU(3)flavour is broken by the electromagnetic interaction and
through the differences between the up-, down- and strange-quark masses.
This leads to considerable mass differences within the SU(3) multiplets. Be-
cause the mass differences to the heavier charm-, top- and bottom-quarks
are even much bigger, one cannot speak of an approximate higher SU(N ),
N ≥ 4, flavour symmetry. 16 Arranging the quarks by their masses, we
depict the symmetries by
SU(2)
z }| {
u, d, s, c, b, t . (1.218)
| {z }
SU(3)
Let the flavour symmetry transformation be
q −→ q ′ = U q, (1.219)
n o
U ∈ SU(3) = U ∈ GL3 (C) | U + U = 1, det U = 1 . (1.220)
The group elements are characterised by n2 − 1 = 8 independent real param-

eters αk , which we write as α
~ = (α1 , . . . , α8 ). For the representation of the
group by its corresponding algebra we note:
Algebra −→ Group by exponentiation

Group −→ Algebra by infinitesimal transformations
8
!
X
U (~
α) = exp −i αk Tk , (1.221)
k=1
8
X
U (δ~
α) = 1 − i δαk Tk . (1.222)
k=1
Tk† = Tk , tr(Tk ) = 0. (1.223)

16
Figs. 1 and 2 on pages 6 – 7 show SU(4) multiplets. There the SU(3) multiplets
have been extended in a third, vertical dimension to incorporate charm quantum numbers
C = 1, 2, 3. That gives 20-plets for baryons and 16-plets for mesons.
1.3 Symmetries 43
The structure constants of the algebra shall be denoted by f
[Tk , Tl ] = fklm Tm . (1.224)
Analogous to the Pauli matrices σk , the Gell-Mann matrices are defined by

λk = 12 Tk for k = 1, . . . , 8. An additional property of the SU(3) generators
is
1
tr(Tk Tl ) = δkl . (1.225)
2
The group SU(2) is a subgroup of SU(3), the elements of SU(2) have the
representation  
0
 U′ 
U =  0 .
 (1.226)
0 0 1
The generators of {U ′ } form a subalgebra of the algebra of SU(3). Using the
isospin operators of SU(2) one can express the corresponding generators of
SU(3) as
     
0 0 0
 I   I2   I3 
T1 =  1
0 , T2 =  0 , T3 =  0 . (1.227)
  
  
0 0 0 0 0 0 0 0 0
A maximum of 2 generators of SU(3) can be chosen such that they commute

with each other and can thus be diagonalised simultaneously. Such a maximal
abelian subalgebra is called Cartan subalgebra. Usually one chooses T3 and
T8 . Explicitly one writes
   
1 1 0 0 1 1 1 0 0
T3 = 0 −1 0
, T8 = √ 0 1 0. (1.228)
2 0 0 0 2 3 0 0 −2
We see that [T3 , T8 ] = 0. The quantum numbers belonging to T3 and T8

are isospin and the additional quantum number Y , called hypercharge. It is
defined by  
√ 1/3 0 0
3
T8 = Y, Y = 0 1/3 0  . (1.229)
2
0 0 −2/3
In this 3-dimensional representation hypercharge Y can have the values 1/3
and −2/3. In addition to the two up- and down-quark states, having isospin
I3 = ±1/2 and hypercharge Y = 1/3, one introduces a third state with
isospin I3 = 0 and hypercharge Y = −2/3, called strange quark. The quan-
tum number strangeness S is defined to be zero for up- and down-quarks and
44 1 INTRODUCTION
S = −1 for strange quarks. Here we list the quantum numbers which are
assigned to the three light quarks.
q Q I I3 Y S B
u 2/3 1/2 1/2 1/3 0 1/3
d -1/3 1/2 -1/2 1/3 0 1/3
s -1/3 0 0 -2/3 -1 1/3
We see that
1
Q = I3 + Y, Y = B + S.17 (1.231)
2
The (partially) conserved quantities now are
Quantity conserved by broken by

interactions interactions
B, Q all none
I strong em, weak
S strong, em weak
The remaining Gell-Mann matrices for SU(3) are

   
0 0 1 0 0 −i
   
λ4 = 0 0 0 λ5 = 0 0 0  (1.232)
1 0 0 i 0 0
   
0 0 0 0 0 1
   
λ6 = 0 0 1 λ7 = 0 0 −i (1.233)
0 1 0 0 i 0
1.3.5 Some comments about symmetry

As we have seen, symmetries play an important rôle in the physics of ele-
mentary particles. The U(1) symmetry leads to charge conservation. The
SU(2) isospin symmetry is a property of the strong nuclear interactions and
is associated with the equality of the masses of proton and neutron or of the
pion masses. This symmetry is only approximately true, but approximate
symmetries give a starting point for perturbative calculations. Symmetries
help to find more or less good quantum numbers and to classify particles.
17
Including the charm quark the hypercharge is
1
Y =B+S− C (1.231)
3
1.3 Symmetries 45
For extensions of the Standard Model, the established symmetries must been
taken into account.
There are two kinds of symmetries: firstly, there are the space-time symme-
tries like translations, rotations, Lorentz transformations, assembled in the
inhomogeneous Lorentz group, also called Poincaré group. Secondly, we have
internal symmetries: U(1), SU(2)flavour , SU(3)flavour , the local gauge symme-
try SU(3)colour and the chiral symmetry.
The integration of these two kinds of symmetries into a unified framework
requires an extension of the concept of symmetry, namely supersymmetry
(SUSY).
46 1 INTRODUCTION
1.4 Field Quantisation

So far we have considered relativistic classical fields, e.g. the Klein-Gordon
field or Maxwell field, which are consistent with special relativity. In this
chapter we develop field theory in the framework of quantum theory.
In classical physics, fields are continuous systems, described by functions of
space and time. Examples are elastic media, electric and magnetic force
fields.
In the context of quantum theory, Schrödinger’s wave function can be con-
sidered as a field. Its interpretation as a probability amplitude is, however,
different from classical fields. The probability density is ρ(x) = ψ ∗ (x)ψ(x).
The attempts to reconcile quantum mechanics and special relativity by intro-
ducing relativistic wave equations, i.e. the Klein-Gordon and Dirac-equation,
led to problems:
• An infinite number of negative energy states arouse.
• When analysing scattering on quantum wells, it turned out that nega-

tive probabilities, as well as probabilities larger than 1, occurred (Klein’s
paradox). This is related to particle creation in strong fields.
The solution of these problems lies in the quantisation of relativistic field

equations. The fields are then not considered as wave functions, but as
physical systems with an infinite number of degrees of freedom, and are
subject to quantisation. It turns out that this leads to quantum theory of
many particles.
Quantisation in quantum mechanics and in quantum field theory
Quantisation of a classical theory implies the replacement of observables by
operators. The fundamental Poisson brackets are replaced by commutators.
In classical mechanics the Poisson brackets are defined by
!
X ∂f ∂g ∂g ∂f
{f, g} = − . (1.234)
i ∂qi ∂pi ∂qi ∂pi
For fields the analogous definition is

Z ( )
3
X δF δG δG δF
{F, G} = d x − ,
a δφa (x0 , ~x ) δπa (x0 , ~x ) δφa (x0 , ~x ) δπa (x0 , ~x )
(1.235)
0
where F and G are functionals of the fields φa (x , ~x ) and their conjugate
momenta πa (x0 , ~x ) at fixed time x0 .
1.4 Field Quantisation 47
Using the functional derivatives
δφa (x0 , ~x )
= δab δ 3 (~x − ~x ′ ) (1.236)
δφb (x0 , ~x ′ )
one gets the fundamental Poisson brackets for fields
{πa (x0 , ~x), φb (x0 , ~x ′ )} = −δab δ 3 (~x − ~x ′ ) (1.237)
Below we summarise variables in the Hamiltonian formalism of classical

physics and their quantum counterparts.
Mechanical description −→ quantum mechanical description

variables operators
qi , p i Qi , Pi
pi = ∂L ∂ q̇i
Poisson bracket commutator/i~
{qi , pk } = δik [Qi , Pk ] = i~ δik
classical field −→ quantum field

field φa (x) field operator φa (x)
π(x) = ∂∂Lφ̇(x)
1
{,} i~
[, ]
states |ψi ∈ H
1.4.1 Quantisation of the real scalar field

With the expression of the Lagrangian density for a real scalar field
1 m2 2
L = ∂µ φ ∂ µ φ − φ (1.238)
2 2
the conjugate momentum density is
∂L
π(x) = = ∂ 0 φ(x) = φ̇(x). (1.239)
∂(∂0 φ(x))
The fundamental Poisson brackets for classical fields are
{π(x0 , ~x ), φ(x0 , ~x ′ )} = −δ 3 (~x − ~x ′ ) (1.240)
{φ(x0 , ~x ), φ(x0 , ~x ′ )} = {π(x0 , ~x ), π(x0 , ~x ′ )} = 0. (1.241)

48 1 INTRODUCTION
Their counterparts in quantum field theory will be

h i ~ 3
π(x0 , ~x ) , φ(x0 , ~x ′ = δ (~x − ~x ′ ), (1.242)
i
h i h i
π(x0 , ~x ) , π(x0 , ~x ′ = φ(x0 , ~x ) , φ(x0 , ~x ′ = 0. (1.243)
A Legendre transformation gives the Hamiltonian density
H = π φ̇ − L (1.244)
1 1 m2 2
= π 2 + (∇φ)2 + φ, (1.245)
2 2 2
and the Hamiltonian
1Z 3 2
H= d x π + (∇φ)2 + m2 φ2 . (1.246)
2
The classical canonical equations of motion then read
φ̇(x) = {φ(x), H} = π(x), (1.247)

π̇(x) = {π(x), H} = ∆φ(x) − m2 φ(x) (1.248)
Combining them yields
φ̈(x) = (∆ − m2 )φ(x), (1.249)
which is the Klein-Gordon equation, (∂µ ∂ µ +m2 )φ = 0. As we see, the result-

ing field equation is Lorentz invariant, although we introduced a distinction
of space and time with the definition of the canonical momentum π, and by
using only equal time Poisson brackets or – after quantisation – equal time
commutators. 18
The corresponding quantum canonical equations read
1
φ̇(x) = [φ(x), H] = π(x), (1.250)
i~
1
π̇(x) = [π(x), H] = ∆φ(x) − m2 φ(x). (1.251)
i~
18
This approach to field quantisation was introduced by Heisenberg and Pauli. They
proceeded from classical mechanics to field theory by dividing space into small cells, to
each of which was associated a pair of conjugate generalised coordinates (qi .pi ). These
undergo time evolution like the many conjugate variables of a many particle mechanical
systems. Going to the limit of a continuum of infinitely small cells one arrives at the time
evolution of fields.
The quantised scalar field obeys the Klein-Gordon equation. Therefore, like
in the classical case it can be decomposed into plane wave solutions
Z
d3 k ~ −ik·x † ~ ik·x

φ(x) = a(k ) e + a (k ) e , (1.252)
(2π)3 2ωk
where
q
k0 = ωk = ~k 2 + m2 . (1.253)
In contrast to the classical case, here the coefficients a(~k ), a† (~k ) are oper-
ators. In the following we just denote them a(k), a† (k), where it is to be
understood that k0 = ωk . From the definition of the canonical conjugate
momentum, Eq. (1.239), we have
Z
d3 k
−ik·x † ik·x

π(x) = (−iωk ) a(k) e − a (k) e . (1.254)
(2π)3 2ωk
We can invert the mode expansions of the field to get an expression for a(k):
Z
d3 k
π(x) − iωk φ(x) = (−2iωk )a(k) e−ikx ,
(2π)3 2ωk
Z
3 ikx

a(k) = i d xe (π(x) − iωk φ(x)) . (1.255)
x0 =0
From these mode decompositions and the commutator rules for the fields,
Eq. (1.242) and (1.243), we find that a(k), a(k † ) satisfy the commutation
rules
[a(k), a† (k ′ )] = (2π)3 2 ωk δ 3 (~k − ~k ′ ) (1.256)
[a(k), a(k ′ )] = [a† (k), a† (k ′ )] = 0. (1.257)
We can express the Hamiltonian (1.246) in these operators,too. Let us do

50 1 INTRODUCTION
this in detail for the three terms. Using (1.252) and (1.254) we obtain
d3 k
Z
~k) a(k) e−ik·x − a† (k) eik·x ,
∇φ(x) = (i
(2π)3 2ωk
1Z 3 2 1Z d3 k Z d3 k ′ Z 3
d xπ = d x(−iωk )(−iωk′ )
2 2 (2π)3 2ωk (2π)3 2ωk′
′ ′
a(k) a(k ′ ) e−i(k+k )·x + a† (k) a† (k ′ ) ei(k+k )·x
′ ′

−a(k) a† (k ′ ) e−i(k−k )·x − a† (k) a(k ′ ) ei(k−k )·x
1 Z d3 k † † † †

= −a(k) a(−k) − a (k) a (−k) + a(k) a (k) + a (k) a(k)
8 (2π)3
1Z 3 1Z d3 k Z d3 k ′ Z 3 ~
2
d x (∇φ) = d x(ik) · (ik~′ )
2 2 (2π)3 2ωk (2π)3 2ωk′
′ ′
′ ′

−a(k) a† (k ′ ) e−i(k−k )·x − a† (k) a(k ′ ) ei(k−k )·x
1 Z d3 k ~k 2 † † † †

= 2
a(k) a(−k) + a (k) a (−k) + a(k) a (k) + a (k) a(k)
8 (2π)3 ωk
m2 Z 3 2 m2 Z d3 k Z d3 k ′ Z 3
d xφ = dx
2 2 (2π)3 2ωk (2π)3 2ωk′
′ ′
′ ′

a(k) a† (k ′ ) e−i(k−k )·x + a† (k) a(k ′ ) ei(k−k )·x
1 Z d3 k m2 † † † †

= a(k) a(−k) + a (k) a (−k) + a(k) a (k) + a (k) a(k)
8 (2π)3 ωk2
(1.258)
Using ωk2 = ~k 2 + m2 and adding the terms up, we finally find
1Z d3 k
† †

H= ωk a (k)a(k) + a(k)a (k) (1.259)
2 (2π)3 2ωk
d3 k
Z
† 1 †
= ω k a (k)a(k) + [a(k), a (k)] (1.260)
(2π)3 2ωk 2
Z 3 Z
dk 1
= 3
ωk a† (k)a(k) + d3 k ωk δ 3 (0). (1.261)
(2π) 2ωk 2
The second integral is an infinite constant. We may interpret it as a divergent

zero point energy. The Hamiltonian can be compared with the Hamiltonian
of an infinite number of harmonic oscillators with creation operators a†i and
annihilation operators ai
X 1
H= ~ωk (a†k ak + ). (1.262)
k 2
The ground state |0i obeys ak |0i = 0 for all k, and the excited states are
given by expressions like
a†j |0i, a†j a†k |0i, a†j a†k a†l |0i, ...
The “quantum number” operator is

X †
N= ak ak . (1.263)
k
In our field theoretical context we interpret
a† (k) as particle creation operator, (1.264)

a(k) as particle annihilation operator. (1.265)
We deal with the zero point energy by requiring that H |0i = 0. This is
achieved by removing the infinite constant from H:
Z
d3 k
H→ ωk a† (k)a(k). (1.266)
(2π)3 2ωk
As this is a fixing of the absolute scale of the energy, physical meaningful
energy differences are not affected by it.
The subtraction of the zero point energy can be expressed in terms of normal
ordering. Normal ordering means, that one has to put each creation operator
to the left of every annihilation operator. Normal ordering is symbolised by
colons, e.g.
: a† a := a† a, : aa† := a† a. (1.267)
Then we define the Hamiltonian to be : H :, such that
: H : |0i = 0. (1.268)
In the following we leave out the colons and understand H to be normal

ordered. Applying creation operators onto the ground state yields excited
states
Ha† (k)|0i = ωk a† |0i, (1.269)

Ha† (k1 )a† (k2 )|0i = (ωk1 + ωk2 ) a† (k1 )a† (k2 )|0i. (1.270)
52 1 INTRODUCTION
From these equations we see that the energies of these states are the energies
of non-interacting relativistic multi-particle states. Each creation operator
creates one particle. The ground state is empty and is the vacuum state.
The representation of the field operators by particle creation and annihilation
operators is called theFock representation; the states created in this way out
of the vacuum state are Fock states. The Hilbert space spanned by all these
multi-particle states is the direct sum of n-particle Hilbert spaces Hn ,
∞
M
H= Hn (1.271)
n=0
and is called Fock space.

A remark about the zero-point energy
The zero-point energy changes, if there are boundary conditions. An example
is the Casimir effect, where two parallel conducting plates feel a force pulling
them together, although they are not electrically charged. Between the con-
ductors the possible “cavity modes” are restricted by boundary conditions.
Their number increases, when the distance a of the plates is increased. Thus
the zero-point energy in between the plates grows with distance, while out-
side there is no restriction on the possible modes. The increasing zero-point
energy leads to an attractive force on the plates,
∂ 1
Energy = attractive force ∼ 4 . (1.272)
∂a a
1.4.2 Quantisation of the complex scalar field

The complex scalar field
Z
d3 k −ikx † ikx

φ(x) = a(k)e + b (k)e (1.273)
(2π)3 2ωk
can be written as a combination of two real scalar fields
1
φ = √ (φ1 + iφ2 ), (1.274)
2
so that
Z
d3 k −ikx † ikx

φj = aj (k)e + aj (k)e , (j = 1, 2.)
(2π)3 2ωk
d3 k
Z
1 −ikx 1 † † ikx
φ= √ (a1 (k) + ia2 (k)) e + √ (a1 (k) + ia2 (k)) e .
(2π)3 2ωk 2 2
| {z } | {z }
=a(k) =b† (k)
This leads to creation and annihilation operators

1
a(k) = √ (a1 (k) + ia2 (k)), (1.275)
2
1
b† (k) = √ (a†1 (k) + ia†2 (k)), (1.276)
2
† †
b (k) 6= a (k). (1.277)
The complex scalar field is equivalent to a pair of two independent real scalar
fields. The commutation relations are
[a(k), a† (k ′ )] = [b(k), b† (k ′ )] = (2π)3 2ωk δ 3 (~k − ~k ′ ), (1.278)
[a, b] = [a† , b† ] = [a, b† ] = [a† , b] = 0. (1.279)
The field describes two sorts of particles, which can be created from the
vacuum state |0i, obeying
a(k)|0i = b(k)|0i = 0. (1.280)
For example, one gets Fock states like
one-particle states a† (k)|0i, b† (k)|0i,
two-particle states a† (k)a† (k ′ )|0i, a† (k)b† (k ′ )|0i.
The conjugate momentum and the Hamiltonian density are
∂
π= (∂µ φ† )(∂ µ φ) − m2 φ† φ) = ∂ 0 φ† = φ̇† , π † = φ̇, (1.281)
∂(∂0 φ)
H = ππ † − L . (1.282)
The Hamiltonian looks like the classic Hamiltonian with the fields replaced
by field operators,
Z
H= d3 x (ππ † + (∇φ† )(∇φ) + m2 φ† φ). (1.283)
In the Fock representation after normal ordering this is

Z
d3 k
† †

:H : = ωk a (k)a(k) + b (k)b(k) . (1.284)
(2π)3 2ωk
In a similar way one finds an expression for the charge. From the U(1)
symmetry and the Noether theorem a conserved charge Q follows, which can
be expressed by field operators
Z
Q = iq d3 x φ† φ̇ − φφ̇† . (1.285)
54 1 INTRODUCTION
In the Fock representation this is

Z
d3 k
† †

Q=q a (k)a(k) − b (k)b(k) . (1.286)
(2π)3 2ωk
This immediately gives
Q a† (k)|0i = q a† (k)|0i for a particle state of sort a, (1.287)
Q b† (k)|0i = −q b† (k)|0i for a particle state of sort b. (1.288)
We see that the state b† (k)|0i represents a particle with opposite charge
than the particle represented by a† (k)|0i, and we call it the corresponding
antiparticle. We define a charge conjugation operator C, which exchanges a
and b:
Ca† C = b† , Cb† C = a† . (1.289)
Then, since CC = 1,
C a† (k)|0i = C a† (k)CC|0i = b† (k)C|0i = b† (k)|0i. (1.290)
1.4.3 Quantisation of the Dirac field

Similar as for the complex scalar field, the expansion of the Dirac field in
terms of plane waves, Eq. (1.34), contains two sorts of coefficients, labelled
b and d. Quantisation turns them into creation and annihilation operators,
such that
b†r (k) creates particles, e.g. electrons,
br (k) annihilates particles,
d†r (k) creates antiparticles, e.g. positrons,
dr (k) annihilates antiparticles.
Starting from the Lagrangian (1.110)
L = ψ̄(iγ µ ∂µ − m)ψ
one defines the conjugate momentum field
∂L
Π(x) = = iψ † . (1.291)
∂(∂0 ψ)
Then the Hamiltonian is
Z Z
H= d3 x (Πψ̇ − L ) = d3 x (ψ † i∂0 ψ − L ) (1.292)
4
Z
d3 k X
† †

= ωk b r (k)b r (k) − dr (k)dr (k) . (1.293)
(2π)3 2ωk r=1
In contrast to the scalar field, we have to impose anticommutation rules for

the Dirac field, because it describes Fermions:
h i
ψα (x0 , ~x) , ψβ† (x0 , ~x ′ ) = δαβ δ 3 (~x − ~x ′ ). (1.294)
+
This leads to
h i h i
br (k) , b†r′ (k ′ ) = dr (k) , d†r′ (k ′ ) = (2π)3 2ωk δr,r′ δ 3 (~k − ~k ′ ), (1.295)
+ +
h i
[br (k) , br′ (k ′ )]+ = b†r (k) , b†r′ (k ′ ) = 0, (1.296)
+
h i
[dr (k) , dr′ (k ′ )]+ = d†r (k) , d†r′ (k ′ ) = 0, (1.297)
+
[br (k) , dr′ (k ′ )]+ = mixed anticommutators = 0, (1.298)
The definition of normal ordering contains a minus sign, when two operators
are interchanged. Therefore the normal ordered Hamiltonian
4
Z
d3 k X
† †

H= ωk b r (k)b r (k) + dr (k)dr (k) (1.299)
(2π)3 2ωk r=1
is positive for all states in the Fock space.
1.4.4 Quantisation of the Maxwell field

The vector potential Aµ (x) is only determined up to gauge transformations.
Therefore it contains redundant, unphysical degrees of freedom, and the
quantisation procedure is non-trivial. There are two possibilities, which are
considered in this context. The first possibility is to impose the Coulomb
(or radiation) gauge on the fields. In this way, however, manifest Lorentz
invariance is lost. To show the relativistic invariance of the physical results
of the theory is a non-trivial task. The second possibility is to keep mani-
fest Lorentz invariance and to quantise the theory covariantly. But then the
gauge freedom in the field variables leads to unphysical states, which must
be removed afterwards. This second possibility is called the Gupta-Bleuler
quantisation.
Coulomb or radiation gauge
When external currents are absent, in the radiation gauge the field only
contains transversal modes with amplitudes a(λ) (k) and a(λ) ∗ (k), where λ =
1, 2 labels two polarisation states or two helicity modes of the photon, k 0 = ωk
and ~k ⊥ A,
~ see Eq. (1.81). Quantisation promotes the amplitudes to operators
a (k), a(λ) † (k). They are photon annihilation and creation operators, and
(λ)
we have
Ha(λ) † (k)|0i = k0 a(λ) † (k)|0i. (1.300)
56 1 INTRODUCTION
Let us study the conjugate momenta of the field. We have

∂L
= −F µν , (1.301)
∂(∂µ Aν )
so that
∂L
Π0 (x) = = 0, (1.302)
∂ Ȧ0 (x)
∂L
Πi (x) = = −F 0i = E i . (1.303)
∂ Ȧi (x)
Canonical commutation relations
[Ai (x0 , ~x ), Πj (x0 , ~x ′ )] = −i δij δ 3 (~x − ~x ′ ) (1.304)
would be in contradiction with the transversality ∂i Ai (x) = 0, since
0 = [∂i Ai (x0 , ~x ), Πj (x0 , ~x ′ )] = −i ∂j δ 3 (~x − ~x ′ ) 6= 0. (1.305)
Instead, the standard commutators for a(λ) (k) and a(λ) † (k) lead to
[Ai (x0 , ~x ), Πj (x0 , ~x ′ )] = −i δ̃ij (~x − ~x ′ ), (1.306)
where the transverse δ-function is defined by
!
′
Z
d3 k i~k·(~x−~x ′ ) ki kj
δ̃ij (~x − ~x ) := e δij − . (1.307)
(2π) 3 ~k 2
Covariant quantisation
Covariant quantisation starts with a Lagrangian
1 1
L = − Fµν F µν − (∂µ Aµ )2 , (1.308)
4 2
which is manifestly Lorentz invariant, but not gauge invariant. Covariant
commutators are imposed on the fields and momenta. There are four types
of annihilation and creation operators a(λ) (k), a(λ) † (k) for λ = 0, . . . , 3. The
Fock space contains unphysical states, e.g. with negative norm. The physical
states are restricted by
(∂µ Aµ )+ |physical statei = 0, (1.309)
where (∂µ Aµ )+ is the positive frequency part of ∂µ Aµ . With this formalism
all physical, gauge invariant results are the same as with the radiation gauge.
To summarise
• explicit Lorentz invariance is kept,
• unphysical states in the Fock space have to be removed by constraints.
1.4.5 Symmetries and Noether charges

According to the Noether theorem, to each continuous symmetry belongs a
conserved charge. Quantisation turns it into a charge operator. Consider,
for example, the isospin for up and down quarks:
Z
1Z 3 ¯

Iˆ3 := d3 x q̄(x)γ 0 I3 q(x) = d x ū(x)γ 0 u(x) − d(x)γ 0
d(x) . (1.310)
2
In the quantised field theory the quark fields ū, u, d¯ and d are field operators.
In Fock space with creation and annihilation operators b(u)† , b(u) , b(d)† , b(d)
and states
|ui = b(u)† |0i, |di = b(d)† |0i, (1.311)
the charge operator acts on these states as
1 1
Iˆ3 |ui = |ui, Iˆ3 |di = − |di. (1.312)
2 2
compare Eqs. (1.286) ff.
58 1 INTRODUCTION
1.5 Interacting Fields

In the previous sections free field theories have been considered. They de-
scribe particles that don’t interact with each other. Now we turn to the
consideration of interactions. Consider a scattering process between parti-
cles, as indicated in the picture.
p1 p′1
|ini |outi
pn p′m
interaction
A number of n particle approach each other and interact with each other.
They form the ingoing state. After the interaction there are particles in an
outgoing state. During the scattering process the particles are in a highly
complicated state. But in the far past and in the far future they are far
away from each other and can be considered as non-interacting (we neglect
self-interactions here). The corresponding asymptotic states describe free
particles.
The transition probability from the ingoing state |ini to an outgoing state
|outi will be described by the matrix element of an unitary time evolution
operator U (t1 , t0 ),
hout|U (+∞, −∞)|ini. (1.313)
1.5.1 Interaction picture

Here is a short reminder about the interaction picture. In the Schrödinger
picture the states are time dependent, their time evolution is given by an
unitary operator:
|ψS (t)i = U (t, t0 )|ψS (t0 i = e−iH(t−t0 ) |ψS (t0 )i. (1.314)
This holds, provided the Hamiltonian has no explicit time dependence. In

the Heisenberg picture the time dependence is shifted from the states to the
operators:
OH (t) = U † (t, t0 )OS U (t, t0 ). (1.315)
In the free field theories considered so far, the field operators and the opera-
tors formed out of them are understood to be in the Heisenberg picture, e.g.
1.5 Interacting Fields 59
the Hamiltonian
1Z 3 Z
d3 k
: H0 : = 2 2 2 2
d x : [π + (∇φ) + m φ ] := 3
ωk a†k ak . (1.316)
2 (2π) 2ωk
For free fields the Hamiltonians in the different pictures are identical:
(H) (S)
H0 = H0 = H0 . (1.317)
In free theories there are no non-trivial transition probabilities
S hout, t|in, t0 iS =H hout|U (t, t0 |iniH . (1.318)
As for the harmonic oscillator, where the time dependence of the ladder
operators in the Heisenberg picture is given by
ȧH (t) = i[H, aH ] = iω[a† a, aH ] = −iω aH (t), (1.319)
resulting in
aH (t) = exp(iω a† a t) a exp(−iω a† a t) = a exp(−iωt)

a†H (t) = exp(iω a† a t) a† exp(−iω a† a t) = a†H exp(iωt),
the time dependence of annihilation and creation operators in field theory is

given by
ak,H (t) = exp(iH0 t)ak,S exp(−iH0 t) = ak,S exp(−iωk t)

a†k,H (t) = exp(iH0 t)a†k,S exp(−iH0 t) = a†k,S exp(iωk t),
and we find that in the transition probabilities only terms of the type
e−iωk (t−t0 ) δin ,out (1.320)
survive. So there are no transitions in the free theory.

Now we include interactions. In the Schrödinger picture we have an interac-
(I)
tion Hamiltonian HS ,

(I)
i∂t |ψ(t)i = H0 + HS |ψ(t)i. (1.321)
If one thinks of small interactions one may define slowly varying states |φ(t)i
|φ(t)i = eiH0 t |ψ(t)i = U0−1 |ψ(t)i, (1.322)

60 1 INTRODUCTION
and with i∂t U0−1 = −H0 U0−1 one finds

(I)
i∂t U0−1 |ψ(t)i = −H0 U0−1 |ψ(t)i + U0−1 (H0 + HS )ψ(t)i (1.323)
(I)
= U0−1 HS U0 U0−1 |ψ(t)i. (1.324)
(I)
With HI (t) := U0−1 (t)HS U0 (t) we get the time dependence of states in the
interaction picture
i∂t |φ(t)i = HI (t)|φ(t)i. (1.325)
Operators in the interaction picture are consistently defined to be
OI (t) = U0† (t)OS U0 (t). (1.326)
They evolve like Heisenberg operators in the free theory,
ȮI (t) = i[H0 , OI (t)]. (1.327)
The advantages of the interaction picture are

• one keeps the formulations of the free theory for the operators,
• the |ini, |outi states can be prepared as states of the free theory.
While at times t = ±∞ the states are simple eigenstates of H0 , the time evo-
lution during the interacting becomes complicated, since it involves HI (t).
Therefore, one has to find approximations. Integrating the Schrödinger equa-
tion in the interaction picture (1.325) and iterating the result leads to
Z t
|φ(t)i = φ(t0 )i + (−i) dt1 HI (t1 )|φ(t1 )i
t0
Z t
= φ(t0 )i + (−i) dt1 HI (t1 )|φ(t0 )i
t0
Z t Z t1
2
+ (−i) dt1 dt2 HI (t1 )HI (t2 )|φ(t0 )i
t0 t0
Z t Z t1 Z t2
+ (−i)3 dt1 dt2 dt3 HI (t1 )HI (t2 )HI (t3 )|φ(t0 )i
t0 t0 t0
......... .
The limits of integration obviously underlie the restriction t > t1 > t2 > t3 >
. . . > t0 . The restriction can be implemented by defining an operator T that
generates time ordered products from arbitrary products of operators, i.e.
(
O(t1 )O(t2 ), if t1 > t2 ,
T [O(t1 )O(t2 )] = (1.328)
O(t2 )O(t1 ) else.
In the same manner T orders higher operator products with respect to their
time arguments. With the aid of T the integral over the simplex t > t1 >
t2 > t3 > . . . > t0 can be transformed into an integral over a hypercube
∞ t t Zt
X (−i)n Z Z
|φ(t)i = . . . dt1 dt2 . . . dtn T [HI (t1 )HI (t2 ) . . . HI (tn )] |φ(t0 )i
i=0 n!
t0 t0 t0
(1.329)
Z t
= T exp −i dt′ HI (t′ ) |φ(t0 )i. (1.330)
t0
This series is called Dyson series. It represents the unitary time evolution
operator in the interaction picture,
|φ(t)i = UI (t, t0 )|φ(t0 )i. (1.331)
1.5.2 The S-matrix

In the limit t0 → −∞, t → ∞, the time evolution operator contains the
description of scattering processes. The S-matrix is defined by
Z ∞
S := UI (+∞, −∞) = T exp −i dt HI (t) . (1.332)
−∞
It gives the transition probability hout|S|ini. In a free theory the S-matrix

is the identity.
In general the interaction Hamiltonian contains field operators
Z
d3 k Z
d3 k
φI (x) = ak e −ikx
+ a† e−ikx , (1.333)
(2π)3 2ωk (2π)3 2ωk k
| {z } | {z }
φ+ φ−
where φ+ and φ− are the positive and negative frequency parts, respectively.
R
Let us assume that HI = − d3 xLI arises from a Lagrangian density of the
form LI = λF (φ(x)) with small λ. Then one can try to expand S in powers
of λ,
S = 1 + S (1) + S (2) + . . . (1.334)
S (0) = 1 (1.335)
Z
S (1) = iλ d4 x [F (φ(x))] (1.336)
(iλ)2 ZZ 4
S (2) = d x1 d4 x2 T [F (φ(x1 )) F (φ(x2 ))] . (1.337)
2
62 1 INTRODUCTION
For the calculation of matrix elements hout|S (n) |ini, it is convenient to bring
this into normal ordered form by use of the commutation relations. The
terms in the matrix element are of the form
(constants, integrals)
h0|(creation operators for out states)†
T [creation and annihilation operators from HI ]
(creation operators for in states)|0i.
To evaluate this one uses commutation or anticommutation relations to arrive

at
(constants, integrals)functions(x1 , . . . , xn )h0|0i

+(constants, integrals)functions(x1 , . . . , xn )from [ , ]±
×h0|N [creation and annihilation operators ]|0i.
Since the vacuum expectation values of all terms, which contain normal or-
dered operators vanish, this transformation of time ordered operator products
to normal ordered products is especially useful. It can be accomplished with
the help of Wick’s theorem.
1.5.3 Wick’s theorem
Define contractions φ(x1 )φ(x2 ) as the difference between time ordered prod-
ucts and normal ordered products:
φ(x1 )φ(x2 ) = T [φ(x1 )φ(x2 )] − N [φ(x1 )φ(x2 )] (1.338)

= h0|T [φ(x1 )φ(x2 )]|0i (1.339)
=: i∆F (x1 − x2 ). (1.340)
∆F (x) is called Feynman propagator
Z
d4 k 1
∆F (x) = lim+ e−ikx . (1.341)
ǫ→0 (2π) k − m2 + iǫ
4 2
Wick’s theorem expresses time ordered products in terms of normal ordered

products and contractions. Some examples are
T [φ1 φ2 φ3 ] = N [φ1 φ2 φ3 ] +φ1 φ2 φ3 + φ1 φ2 φ3 + φ̌1 φ2 φ̌3

T [φ1 φ2 φ3 φ4 ] = N [φ1 φ2 φ3 φ4 ]
+φ1 φ2 N [φ3 φ4 ] +φ1 φ3 N [φ2 φ4 ]
+φ1 φ4 N [φ2 φ3 ] +φ2 φ3 N [φ1 φ4 ]
+φ2 φ4 N [φ1 φ3 ] +φ3 φ4 N [φ1 φ2 ]
+φ1 φ2 φ3 φ4 + φ̌1 φ2 φ3 φ̌4 +φ1φ̌2 φ3φ̌4 ,
where we write φi for φ(xi ). The general rule is
X
T [fields] = N [subset of fields]×(all possible contractions of other fields).
subsets
of fields
(1.342)
From Wick’s theorem it follows that in h0| . . . |0i only the complete contrac-
tions survive.
1.5.4 Feynman diagrams

--------
The contractions resulting from Wick’s the-
orem are represented by lines. The lines for
bosonic fields are drawn dotted or dashed.
R
The integral −iλ d4 x in the interaction Hamiltonian gives a vertex, with
lines connected to it like the following
or
x x
Sometimes the integration variable (x) is denoted near the vertex. It remains
to take into account the operators that generate the in and out states out of
the vacuum. The contraction of these operators leads to diagram lines which
come from outside. These lines are called external lines, while internal lines
arise from contractions of field operators in the interaction Hamiltonian part.
By means of Fourier transform the x-integrations can be evaluated in mo-
mentum space. External lines are associated with solutions of the free field
equations and are subject to the restriction
q
k0 = ~k 2 + m2 ; (1.343)
64 1 INTRODUCTION
the corresponding propagators are denoted to be on the mass shell. At each

vertex the sum of incoming momenta minus the sum of outgoing momenta
has to be zero.
We illustrate the Feynman diagrams for the case of a real scalar field with
quartic self-interaction. Let the Lagrangian be
1
L0 = (∂µ φ ∂ µ φ − m2 φ2 ), (1.344)
2
LI = −gφ4 . (1.345)
Here g is for a small, real coupling constant. Other powers of φ in the

interaction terms are not useful, for
φ1 would give only a constant shift,

φ2 gives an extra mass,
φ3 would lead to instabilities.
The Hamiltonian density for the interaction is
HI = gφ4 (x) (1.346)
and for two-particle scattering to first order in g one has to evaluate a term
of the type
Z
hout| d4 x gφ4 (x)|ini. (1.347)
This is represented by
Z
1 vertex for d4 x g,
4 external lines.
The graphs for the lowest contributions to two-particle scattering are

S (0) +
S (1)
S (2) + +
Classification of graphs
Let us comment on some particular types of graphs.
• Graphs with no external lines, the so called “vacuum bubbles”. It can

be shown that they can be neglected in S-matrix elements.
• Graphs with 2 external lines are called self-energy graphs.
Tadpole (Kaulquappe) Sunset

the full particle propagator and lead to corrections to the mass of the
They contribute to
particle.
• Scattering graphs with more than 2 external lines.
• Connected graphs, on which it is possible to go from each vertex to

each other vertex by moving on lines, like the two previous ones.
66 1 INTRODUCTION
• One-particle-irreducible graphs, which cannot be decomposed into 2

disconnected graphs by removing a single line. All graphs can be com-
posed of these.
1.5.5 Fermions
For fermions the calculations explained above can be performed, too. It
has to be observed that commutators between the basic fields have to be
replaced by anticommutators. The resulting Feynman rules are analogous to
the bosonic case. Due to the anticommuting nature of fermions, closed loops
of fermion lines get an extra factor of (-1).
Examples for graphs including fermions are
inelastic fermion scattering in a field
fermion
fermion
boson
t
e+ –e− annihilation
fermion
boson
antifermion
t
pair production
fermion
boson
antifermion
t
Yukawa coupling
The Lagrangian for the simplest Yukawa model contains a real scalar field
φ(x) for a π 0 -meson, a Dirac field for a proton, and an interaction term:
L = L0φ + L0ψ + LIφψ , (1.348)
1
L0φ = ∂µ φ∂ µ φ − M 2 φ2 , (1.349)
2
ψ
L0 = ψ̄ (iγ µ ∂µ − m) ψ, (1.350)
LIφψ = −G : ψ̄(x)ψ(x)φ(x) = −HI . (1.351)
A self-interaction for the pions, like 4!λ φ4 can also be included.

The propagator for fermion fields is given by
Z
d4 k γ µ kµ + m −ikx
SF (x) := lim+ i e . (1.352)
ǫ→0 (2π)4 k 2 − m2 + iǫ
In SF , the time ordering for fermions is defined by
(
ψ(x)ψ̄(y), if x0 > y 0 ,
T [ψ(x), ψ̄(y)] = (1.353)
−ψ̄(y)ψ(x), if x0 < y 0 .
The contractions are defined by
ψ(x)ψ̄(y) = T [ψ(x)ψ̄(y)] − N [ψ(x)ψ̄(y)] = SF (x − y). (1.354)

Based on this we can draw Feynman
graphs for the perturbative contributions.
The fermion contraction is represented by
a line with an arrow.
The arrow gives the direction of charge flow. If charge flows in the same
direction as momentum, the particle is a fermion, else it is the antiparticle.
Sometimes the flow of momentum or the direction of time is indicated addi-
tionally as in these graphs for an incoming antifermion.
or
~k t
To distinguish fermions from bosons one draws dotted or dashed lines for
bosons.
68 1 INTRODUCTION
1.5.6 Limitations of the perturbative approach

• The coupling constant, e.g. λ, should be small, higher orders in the
Dyson series should be less relevant. In various interesting cases, e.g.
in strong interactions, this is not the case.
• At higher orders the number of graphs increases rapidly. They can only
be managed by computer programs.
• Complicated integrals need special technical tricks.
• Some integrals are divergent. Renormalisation is needed.
• In general the perturbative series are not convergent, but only asymp-
totic. For practical purposes they have to be truncated.
69
2 Quantum Electrodynamics (QED)

2.1 Local U(1) Gauge Symmetry
QED deals with the interactions of electrons and positrons with the Maxwell
field. Let us begin with the Dirac Lagrangian
L = ψ̄(x)(iγ µ ∂µ − m)ψ(x), (2.1)
which has a global U(1) symmetry:
ψ(x) −→ ψ ′ (x) = e−iqα ψ(x). (2.2)
If instead of a constant α, one considers α(x) to depend on x, one has local

transformations
ψ(x) −→ ψ ′ (x) = e−iqα(x) ψ(x). (2.3)
But since ∂µ ψ does not transform like ψ,
∂µ ψ ′ (x) = e−iqα(x) ∂µ ψ(x) − iq (∂µ α(x)) e−iqα(x) ψ(x),
the Lagrangian is not any more invariant under this transformation:
L ′ = L + ψ̄iγ µ (−iq∂µ α)ψ = L + q ψ̄γ µ ψ ∂µ α = L + j µ ∂µ α. (2.4)
The invariance of the Lagrangian can be restored by coupling the Dirac field
to the Maxwell field:
L = ψ̄(iγ µ ∂µ − m)ψ − j µ Aµ
= ψ̄(iγ µ ∂µ − m)ψ − q ψ̄γ µ ψAµ
= ψ̄(iγ µ (∂µ + iqAµ ) − m)ψ.
Under gauge transformations the Maxwell field transforms as
Aµ (x) −→ A′µ (x) = Aµ (x) + ∂µ α(x). (2.5)
Thus we find that L is invariant:
L ′ = L + q ψ̄γ µ ψ ∂µ α − q ψ̄γ µ ψ ∂µ α = L . (2.6)
The Lagrangian contains the covariant derivative
Dµ := ∂µ + iqAµ . (2.7)
70 2 QUANTUM ELECTRODYNAMICS (QED)
The covariant derivative of the fermion field transforms as
Dµ ψ −→D ′µ ψ ′ (x)
=(∂µ + iqA′µ )ψ ′ (x)
=(∂µ + iq(Aµ (x) + ∂µ α(x))) e−iqα(x) ψ(x)
=e−iqα(x) (∂µ − iq(∂µ α(x)) + iq(Aµ (x) + ∂µ α(x)))ψ(x)
=e−iqα(x) (∂µ + iqAµ (x))ψ(x)
=e−iqα(x) (∂µ + iqAµ (x))eiqα(x) ψ ′ (x)
=e−iqα(x) Dµ eiqα(x) ψ ′ (x).
So we get the following transformation of the covariant derivative operator
D ′µ = e−iqα(x) Dµ eiqα(x) . (2.8)
The case considered here is the simplest example of a more general situation,
where a gauge field is coupled to matter fields.
The general picture
We start with a global symmetry. (U(1))

This gives a Noether current. (j µ = q ψ̄γ µ ψ)
We generalise to a local symmetry. (α → α(x))
We restore the invariance of the action by adding
a gauge field coupled to this Noether current.
Thus a symmetry determines the interaction. (−q ψ̄γ µ ψAµ )
To complete the Lagrangian, we have to add the dynamics of the free Maxwell
field
1
L(Aµ ) = − Fµν F µν , (2.9)
4
with the field strength
i
Fµν = ∂µ Aν − ∂ν Aµ = − [Dµ , Dν ]. (2.10)
q
The complete Lagrangian density of QED thus reads
1
L = ψ̄(iγ µ Dµ − m)ψ − Fµν F µν . (2.11)
4
2.2 Quantum Electrodynamics 71
2.2 Quantum Electrodynamics

Now we consider quantisation of the theory. The Hamiltonian contains the
free parts for fermions and photons and an interaction term
HI = −e0 ψ̄γ µ ψAµ . (2.12)
The free parts give
a fermion (electron or positron) propagator
and a photon propagator µ ν
The photon propagator in momentum space is given by

−ig µν
. (2.13)
k 2 + iǫ
R µ
The interaction vertex is ie0 γ µ ( d4 x).
The external photon lines get polarisation vectors

ǫµ (p), ǫ∗µ (p) ⊥ p~. (2.14)
Some QED processes

e− e− scattering
This graph represents the leading order contri-
bution. In the non-relativistic limit this gives
the Coulomb interaction with potential19
α e2
V =− , α=
r 4π~c
t
Compton scattering
There are two graphs in leading order (O(α)), resulting in two interaction
processes.
e− e− e− e−
γ γ γ γ
19
In SI units α = e2 /4πǫ0 ~c = 1/137.036
72 2 QUANTUM ELECTRODYNAMICS (QED)
e+ e− scattering
The leading order graphs are
e− e− e− e−
e+ e+ e+ e+
e+ e− annihilation
The leading order graphs describe e+ + e− → 2γ.
γ γ
e− e−
e+ γ e+ γ
The graphs considered so far do not contain closed loops of lines – these
are tree level graphs. In higher orders graphs with loops give corrections to
the leading order, or they may correspond to new processes. For example,
corrections to e+ e− scattering are
The following loop graph contributes to photon–photon scattering by ex-

change of virtual electrons. This nonlinear breaking of the superposition
principle is also called vacuum birefringence.
2.2 Quantum Electrodynamics 73
74 3 NON-ABELIAN GAUGE THEORY
3 Non-abelian Gauge Theory

3.1 Local Gauge Invariance
As in the case of QED, we start by considering a global symmetry. This
time, it is given by the non-abelian group SU(N ). Consider a field having N
complex components:
 
φ1 (x)
 . 
 ..  ,
φ(x) =   φ+ (x) = φ∗1 (x), . . . φ∗N (x) . (3.1)
φN (x)
The scalar product between two complex n-component vectors φ and φ′ is

denoted
N
X
φ+ · φ′ := φ∗a φ′a . (3.2)
a=1
Let the Lagrangian for the free field be
L = ∂µ φ+ (x) · ∂ µ φ(x) − m2 φ+ (x) · φ(x). (3.3)
It is invariant under the transformation
φ −→ φ′ = U · φ, (3.4)
where U is a N × N matrix obeying U + · U = 1, namely
∂µ φ′+ · ∂ µ φ′ = ∂µ (U · φ)+ · ∂ µ (U · φ) = ∂µ φ+ · U + · U ∂ µ φ = ∂µ φ+ · ∂ µ φ (3.5)
and analogously
φ′+ · φ′ = φ+ · φ. (3.6)
In particular, matrices U which are in the group SU(N ) obey the condition,
so that SU(N ) is a symmetry group of the Lagrangian.
Let us now consider local transformations U (x) ∈ SU(N )
 2 −1

NX
U (x) = exp −i αa (x)Ta  , (3.7)
a=1
where Ta are the N 2 − 1 generators of SU(N ). The spacetime dependency

of the local transformations U (x) gives an additional inhomogeneous term in
the derivative of U (x) · φ(x),
∂µ φ′ (x) = U (x) · ∂µ φ(x) + (∂µ U (x)) φ(x), (3.8)

3.1 Local Gauge Invariance 75
such that the Lagrangian is no longer invariant. Again, the invariance can
be restored by replacing the derivative by the covariant derivative
 2 −1

NX
Dµ φ(x) := ∂µ − ig Aaµ (x)Ta  φ(x). (3.9)
a=1
This introduces N 2 −1 gauge fields Aa (x), one for each generator Ta of SU(N ).
The coupling constant, denoted by −g, replaces the coupling constant +q of
QED.
A more compact notation is
2 −1
NX
Aµ := Aaµ Ta := Aaµ Ta . (3.10)
a=1
The fields Aµ (x) are elements of the Lie algebra of SU(N ). Then the covariant
derivation reads
Dµ = ∂µ − igAµ . (3.11)
In terms of the covariant derivatives the Lagrangian is written
L = Dµ φ+ (x) · Dµ φ(x) − m2 φ+ (x) · φ(x). (3.12)
Requiring the invariance of the Lagrangian, we can determine the transfor-

mation law for the gauge fields. The covariant derivative should obey
!
Dµ′ φ′ (x) = U (x) · Dµ φ(x), (3.13)
which is
!
Dµ′ φ′ (x) = U (x) · Dµ U −1 (x)φ(x). (3.14)
As this should hold for all φ′ (x), we get
!
Dµ′ = U (x) · Dµ U −1 (x). (3.15)
With the explicit expression for Dµ , Eq. (3.14) reads

!
(∂µ − igA′µ (x)) · φ′ (x) = U (x) · (∂µ − igAµ (x)) · U −1 (x)φ′ (x). (3.16)
which implies
−igA′µ (x) = U (x) · ∂µ U −1 (x) − igU (x)Aµ (x)U −1 (x).
or
i
A′µ (x) = U (x)Aµ (x)U −1 (x) + U (x)∂µ U −1 (x). (3.17)
g
This is the transformation law for the gauge fields. It generalises the trans-
formation law for the potentials Aµ (x) in QED.
Let us compare the formulae for gauge groups U(1) and SU(N ):
U(1) : U (x) = exp(−iqα(x)1)

Aµ (x) = Maxwell field
Dµ = ∂µ + iqAµ (x)
A′µ (x) = Aµ (x) + ∂µ α(x)
SU(N ) : U (x) = exp (−iαa (x)Ta ), and [Ta , Tb ] 6= 0 in general
Aµ (x) = Aaµ (x)Ta
Dµ = ∂µ − igAµ (x)
i
A′µ (x) = U (x)Aµ (x)U −1 (x) + U (x)∂µ U −1 (x).
g
Infinitesimal transformations
It is also instructive to consider the formulae for the case of infinitesimal
transformations
U (x) = 1 − iδαa (x)Ta , (3.18)

φ′ (x) = φ(x) − iδαa (x)Ta φ(x), (3.19)
1
A′µ (x) = Aµ (x) − iδαa (x)[Ta , Aµ (x)] − ∂µ δαa (x)Ta + O((δα)2 ). (3.20)
g
Remembering that Aµ = Aaµ Ta , we see that the commutators of the genera-
tors Ta are involved. These commutators are given by the structure constants
of the Lie algebra associated with the group SU(N )
c
[Ta , Tb ] = ifab Tc . (3.21)
With this, the transformation rule for the gauge fields are
a 1
A′µ Ta = Aaµ Ta − iδαa Abµ [Ta , Tb ] − ∂µ δαa (x)Ta
g
1
= Aaµ Ta + δαa Abµ fab
c
Tc − ∂µ δαa (x)Ta
g
1
= (Aaµ + δαb Acµ fbc
a
− ∂µ δαa (x))Ta ,
g
a 1
A′µ (x) = Aaµ (x) + fbc
a
δαb (x)Acµ (x) − ∂µ δαa (x). (3.22)
g
Field strengths and Yang-Mills Lagrangian

The missing piece is now the Lagrangian for the gauge fields. Motivated by
QED we attempt to express it in terms of field strengths. We have to find
appropriate expressions for the field strengths. Just defining them by Fµν =
∂µ Aν − ∂ν Aµ does not work, because this expression does not transform in
such a way that a gauge invariant expression can be built from it. Considering
the commutators of the covariant derivatives leads to the proper expression:
[Dµ , Dν ]φ(x) = [(∂µ − igAµ ), (∂ν − igAν )]φ(x)

= (−ig[Aµ , ∂ν ] − ig[∂µ , Aν ] + (−ig)2 [Aµ , Aν ])φ(x)
= −ig{Aµ ∂ν − ∂ν Aµ + ∂µ Aν − Aν ∂µ − ig[Aµ , Aν ])}φ(x)
= −ig{(∂µ Aν − ∂ν Aµ ) − ig[Aµ , Aν ]}φ(x)
In the last line, the derivatives act on the Aµ only, and dividing by φ(x) we
define the field strengths through
i
Fµν = [Dµ , Dν ] = (∂µ Aν − ∂ν Aµ ) − ig[Aµ , Aν ]. (3.23)
g
Expanding the Aµ in the generators yields
Fµν = ∂µ Aaν Ta − ∂ν Aaµ Ta + gAaµ Abν fab

c
Tc

= ∂µ Aaν − ∂ν Aaµ + gAbµ Acν fbc
a
Ta
a
=: Fµν Ta . (3.24)
So the components of Fµν are
a
Fµν = ∂µ Aaν − ∂ν Aaµ + gfbc
a b c
Aµ Aν . (3.25)
How does the field strength transform under gauge transformations? From
Eq. (3.15) we obtain
′ i i i
Fµν = [Dµ′ , Dν′ ] = [U Dµ U −1 , U Dν U −1 ] = U [Dµ , Dν ]U −1 = U Fµν U −1 .
g g g
(3.26)
This homogeneous transformation law allows us to form gauge invariant ex-
pressions from Fµν . In particular, we can write down a gauge invariant La-
grangian for the gauge fields, which parallels the Lagrangian of the Maxwell
theory. This is the Yang-Mills Lagrangian: LY M

1
LY M = − tr(Fµν F µν ) (3.27)
2
1 a
= − tr(Fµν F a,µν Ta Tb )
2
1 a 1
= − tr(Fµν F a,µν δab )
2 2
1 a a,µν
= − Fµν F . (3.28)
4
In fact, it is invariant, as can be checked by using the cyclicity of the trace.
Remarks:
• A mass term like m2 Aaµ Aa,µ is forbidden by gauge invariance.
• LY M contains cubic (Aaλ Abµ Acν ) and quartic (Aaκ Abλ Acµ Adν ) terms. They
represent self-interactions of the gauge field.
Field equations
We can now formulate the complete Lagrangian for an N -component complex
scalar field interacting with non-abelian gauge fields, including an optional
φ4 -term:
2 1 a a,µν
L = (Dµ φ)+ ·Dµ φ − m2 φ+ ·φ − λ φ+ ·φ − Fµν F . (3.29)
4
From this we may derive field equations for the field φ(x) and for the gauge
field strength F a,µν (x):

Dµ Dµ + m2 φ(x) = 0, (3.30)
∂µ F µν − ig[Aµ , F µν ] = j ν , (3.31)
where
j ν := j a,ν Ta (3.32)
is the current formed out of the scalar fields. The first equation is a gauge
covariant generalisation of the Klein-Gordon equation. The second equation
is the non-abelian generalisation of the inhomogeneous Maxwell equations.
The analogue of the homogeneous Maxwell equations is
[Dρ , Fµν ] + [Dµ , Fνρ ] + [Dν , Fρµ ] = 0. (3.33)
It holds identically due to the definition of the field strengths. This equa-
tion is called Bianchi identity, because it has the same structure as the first
Bianchi identity, which expresses a symmetry of the curvature tensor in dif-

ferential geometry. (Since Fµν is a Lie bracket, the Bianchi identity can also
be considered as a Jacobi identity.)
The second field equation can also be written as
[Dµ , F µν ] = j ν , (3.34)
because
[Dµ , F µν ] = [∂µ − igAµ , F µν ]
= ∂µ F µν − F µν ∂µ − ig[Aµ , F µν ]
= (∂µ F µν ) − ig[Aµ , F µν ]. (3.35)
The current j µ is not conserved, ∂µ j µ 6= 0, since the gauge field itself is
charged and contributes to the total current.
Historical remarks on gauge theory
The concept of local gauge invariance has its origin in the work of Hermann
Weyl in 1918. He tried to extend general relativity by allowing spacetime
dependent scale changes, “Umeichungen”, of the meter stick and the time-
normal. Thus the metric tensor undergoes a local gauge transformation
gµν (x) −→ Ω(x)gµν (x). (3.36)
In Weyl’s approach 0 < Ω(x) = eα(x) . Einstein pointed out, however, that
the length of a meter stick or the frequency of a clock could change if it is
moved around a closed path and returned to its origin, which is physically
unacceptable. After the introduction of Schrödinger’s wave function, London
and others found that by considering purely imaginary α(x) in Weyl’s gauge
transformations, and transforming the wave function correspondingly instead
of the metric, the coupling to the electromagnetic field is recovered.
Here are some cornerstones in the development of gauge theory.
H. Weyl 1918
V. Fock 1926 Gauge covariance of the Schrödinger equation.
F. London 1927 Ω(x) = eiα(x) .
H. Weyl 1929 Local gauge theory for the Maxwell and matter fields.
O. Klein 1938 Conference in Warsaw. Non-abelian gauge field with
covariant derivative, no Lagrangian.
W. Pauli 1953 Non-abelian gauge theory, based on non-abelian
Kaluza-Klein theory. Unpublished, because there is
no mass term in the theory.
C. N. Yang,
and R. Mills 1954 Phys. Rev. 96 (1954) 191. Gauge theory for SU(2).
R. Shaw 1955 PhD thesis as student of A. Salam, Cambridge (UK).
R. Utiyama 1957 Phys. Rev. 101 (1957) 1597. Gauge model of gravity.
3.2 Geometry of Gauge Fields

3.2.1 Differential geometry
A manifold M is, roughly spoken, a space

with a curvilinear coordinate system. The
coordinates are sufficiently smooth. To each
point x of the manifold is associated a tan-
gent space. Its elements are tangent vectors.
A vector field associates a particular vector
v(x) to each point x in a smooth way.
Vectors can be represented by their components vµ (x) relative to a basis in
the tangent space at x.
Now we would like to define how to take the derivative of a vector field along
a particular curve. One is tempted to define the derivative via
dvµ (x) = vµ (x + dx) − vµ (x), (3.37)
where dx is an infinitesimal line element along the curve.
The problem with this is, however, that

the direction of the basis vectors, which de-
termine the components vµ , in general can
change from point to point. So, we need an
appropriate way to compare vectors at differ-
ent points. This can be done by taking the
vector at a point x, v(x), and first shift it
by parallel transport to the point x + dx and
then compare it with the vector v(x + dx).
In order to put this into practice, a notion of

parallel transport for vectors in the surface
is needed. Parallel transport does not mean
that the transported vectors are parallel in
the embedding space. This is illustrated in
the picture, where vectors are parallel trans-
ported along great circles on a sphere. The
embedding space is not constitutive for the
vector fields. We have introduced it here
to illustrate the concepts. Parallel transport
must be defined without referring to an em-
bedding space.
3.2 Geometry of Gauge Fields 81
Thus we must specify a rule for parallel transport. A vector v(x) which is
parallel transported from x to the point x′ = x + dx is a vector v p (x + dx),
v(x) −→ v p (x + dx). (3.38)
Although v p (x + dx) is parallelly transported from v(x), its components

change due to, say a rotation of the vector space basis, when going from
x to x + dx.
v(x) v p (x + dx)
Therefore, associated with an infinitesimal parallel transportation is an in-

finitesimal rotation.
ν
vµp (x + dx) = 1 − Γλ dxλ vν (x). (3.39)
µ
| {z }
infinitesimal rotation matrix
The elements of the infinitesimal rotation matrix are known as Christoffel

symbols. With
vµp (x + dx) = vµ (x) + δvµ (x) (3.40)
one writes
δvµ (x) = −Γνµλ (x)vν (x)dxλ . (3.41)
Differentiation
Because v(x) and v(x + dx) belong to different vector spaces, there is a priori
no way to add or subtract these vectors. To define differentiation, one may
instead use the difference between vµ (x + dx) and the components of the
parallel transported v whose coordinates are denoted by vµp (x + dx).
v(x + dx) Dv
v(x)
v p (x + dx)
⇒
In order to distinguish it from the naive difference dvµ (x) = vµ (x+dx)−vµ (x),
it is denoted by a capital D:
Dvµ (x) := vµ (x + dx) − vµp (x + dx). (3.42)

Using δvµ (x), introduced above, we find
Dvµ (x) = dvµ (x) − δvµ (x)

= ∂λ vµ (x)dxλ + Γνµλ (x)vν (x)dxλ
h i
= ∂λ vµ (x) + Γνµλ (x)vν (x) dxλ
=: Dλ vµ (x)dxλ , (3.43)
with the covariant derivative
Dλ vµ = ∂λ vµ + Γνµλ vν . (3.44)
Dλ vµ transforms covariantly under coordinate transformations. It is a tensor,

while ∂λ vµ , Γνµλ vν , and Γνλν are not tensors.
3.2.2 Gauge Theory

In contrast to differential geometry, the vector fields φ(x) of gauge theory are
not vectors in the tangent space of a manifold. They belong to an “internal”
vector space Vx like isospin, flavour, and colour vector space 20 .
 
φ1 (x)
 . 
φ(x) =  .. 

. (3.45)
φN (x)
The local gauge transformation with gauge group elements U (x), denoted by
φ(x) −→ φ′ (x) = U (x)φ(x), (3.46)
can be considered as a spacetime dependent change of the basis in Vx . It is a

passive transformation, the components φa change because the basis vectors
change.
Physics should not depend on the local choice of the basis. Therefore, diffe-
rentiation has to be defined based on the change of φ relative to the parallel
transported φp . Similar to the geometric case of the last section we write
δφa (x) = φpa (x + dx) − φa (x). (3.47)
In this case we do not have Christoffel symbols, however, we can write the
equation, which fixes the meaning of parallel transport, in an analogous way:
δφ(x) = igAµ (x) · φ(x)dxµ . (3.48)

20
The mathematics of gauge theory and fiber bundles was developed by Élie Cartan,
and Charles Ehresmann.
Here Aµ (x) is a Hermitian matrix, and igAµ (x)dxµ describes an infinitesimal

rotation in the internal vector space. Aµ (x) is an element of the Lie algebra
belonging to the gauge group element U (x). We define
Dφ(x) :=φ(x + dx) − (φ(x) + δφ(x))

=dφ(x) − δφ(x)
=∂µ φ(x)dxµ − igAµ (x)φ(x)dxµ
=(∂µ − igAµ (x))φ(x)dxµ . (3.49)
| {z }
=:Dµ
The gauge covariant derivative
Dµ = ∂µ − igAµ (x) (3.50)
corresponds to the covariant derivative of geometry.

There is another object of geometry that can be taken over to gauge theory,
the curvature. As the picture about parallel displacement of vectors on a
sphere showed, curvature is related to the differing results of parallel trans-
portation, when the vectors are transported along different paths. Consider
two different infinitesimal paths from a point to a neighbouring point, along
which vectors are parallelly transported.
Path 2
x+b x+a+b
Path 1
x x+a
A vector v(x), which is transported from x to x + a + b along one path P1

differs from the vector, which is transported along path P2 .
Along path 1 the coordinates change according to
vµ (x) → vµ (x) − Γνµα (x)vν (x)aα

→ vµ (x) − Γνµα (x)vν (x)aα − Γλµβ (x + a) {vλ (x) − Γνλα (x)vν (x)aα } bβ
= vµ (x) − Γνµα (x)vν (x)aα − Γλµβ (x)vλ (x)bβ − ∂α Γλµβ (x)aα vλ (x)bβ
+ Γλµβ (x)Γνλα (x)vν (x)aα bβ + O(a2 b).
Along path 2, after having exchanged a and b, the parallel transported vector
is
vµ (x) → vµ (x) − Γνµα (x)vν (x)bα − Γλµβ (x)vλ (x)aβ − ∂α Γλµβ (x)bα vλ (x)aβ
+ Γλµβ (x)Γνλα (x)vν (x)bα aβ + O(a2 b).
Then both displaced vectors differ by
∆vµ = δ(path 2) − δ(path 1)

= ∂α Γλµβ (x)aα vλ (x)bβ − ∂α Γλµβ (x)bα vλ (x)aβ
+ Γλµβ (x)Γνλα (x)vν (x)bα aβ − Γλµβ (x)Γνλα (x)vν (x)aα bβ .
The result is
ν
∆vµ = Rµαβ vν aα bβ , (3.51)
ν
where we introduced the Riemann-Christoffel curvature tensor Rµαβ
ν
Rµαβ = ∂α Γνµβ − ∂β Γνµα + Γλµα Γνλβ − Γλµβ Γνλα . (3.52)
In a flat manifold both parallel transports must give the same result ∆v = 0,
ν
and thus Rµαβ = 0. Manifolds can be flat locally, i.e. in an infinitesimal
neighbourhood of a point, as is the case with a saddle point on a two dimen-
sional surface.
Curvature of gauge fields
In gauge theory we have the same algebraic structure for parallel transporting
the vector field φ
δφ(x) = igAµ (x) · φ(x)dxµ . (3.53)
Writing the matrix elements of Aµ as Adcµ , we have
δφc (x) = igAdcµ (x)φd (x)dxµ . (3.54)
Comparing this with
δvν (x) = −Γλνµ (x)vλ (x)dxµ , (3.55)
we see that the analogue of the Christoffel symbol Γ is −igA. The difference
of the parallel transport along two infinitesimally paths is then
h i
∆φc = −ig ∂α Adcβ − ∂β Adcα − igAλcα Adλβ + igAλcβ Adλα φd aα bβ , (3.56)
or using matrix notation
∆φ = −ig {∂α Aβ − ∂β Aα − igAα ·Aβ + igAβ ·Aα }·φ aα bβ

= {[∂α , ∂β ] −[∂α , igAβ ] + [∂β , igAα ] + [igAα , igAβ ]}·φ aα bβ
| {z }
=0
= [∂α − igAα , ∂β − igAβ ]·φ aα bβ
= [Dα , Dβ ]·φ aα bβ .
As we have already seen, the commutator of the D’s gives the field strengths
[Dµ , Dν ] = −igFµν , therefore,
∆φ(x) = −igFµν (x)φ(x) |aµ{zbν} , (3.57)

:=f µν
where f µν is a surface element. Based on the analogy with differential geom-

etry, on calls
Fµν = “Curvature of the gauge field”, (3.58)
and if Fµν ≡ 0, one says “the gauge field is flat.” In this case Aµ can be
transformed to 0 by a special choice of gauge transformation.
Let us consider parallel transport at finite distances. To see, how this works,
we will look at the abelian case with symmetry group U (1).
φ(x) ∈ C φp (x′ )
x x′
Then an infinitesimal parallel transport along a curve is given by
δφ(x) = −iqAµ (x)dxµ , (3.59)
which formally may be integrated

Z
p ′ µ
φ (x ) = exp −iq Aµ (z)dz φ(x). (3.60)
We check this by returning to infinitesimal steps
φp (x + dx) = {1 − iqAµ (x)dxµ } φ(x)

= φ(x) − iqAµ (x)dxµ φ(x)
= φ(x) + δφ(x).
R ′
In the non-abelian case the exponentiated line integral xx Aµ (z)dz µ has to
be path ordered in much the same way as the time ordered expressions for
the S-matrix.
For a closed curve C the change ∆φ in the abelian case is given by

I
φ(x) + ∆φ(x) = exp{−iq Aµ (z)dz µ }φ(x). (3.61)
C
From Stokes theorem

I Z
Aµ (z)dz µ = Fµν df µν = flux through area A 21
(3.62)
C A
we have Z
φ(x) + ∆φ(x) = exp −iq Fµν df µν φ(x). (3.63)
This is a gauge invariant expression and consequently the line integral of Aµ

along a closed loop is gauge invariant.
In electrodynamics the phase of the elec- double
slit screen
tron wavefunction changes along a curve due thin
solenoid magnetic
to the gauge dependent vector potential. flux
But the phase is only measurable through
interference, which means that one needs
two different paths to the point of measure-
ment. The resulting phase difference equals B=0
the phase of ∆φ for a closed curve. The
Aharonov-Bohm interference relies on this
gauge invariant phase difference.
The infinitesimal version of Eq. (3.63) is
∆φ(x) = −iqFµν (x)f µν φ(x). (3.64)
21
Remind e.g. the magnetic flux
Z Z I
~ ~
B · df = ∇ × A · df = A~ · d~r.
87
4 Quantum Chromodynamics (QCD)

4.1 Lagrangian Density and Symmetries
Quarks are represented by Dirac fields, which we now denote q instead of ψ.
 
q1
 
q = (qα ) = · · · . (4.1)
q4
Today, six “flavours” of quarks are known:

! ! !
u c t
, , ,
d s b
and we write
qf = (qα,f ), f = 1, 2, . . . Nf , Nf = 6. (4.2)
Finally, there are three “colours” for each quark, which have been introduced
originally to solve the problem of quark statistics:
qi = (qα,i,f ), i = 1, 2, 3, or i = red, green, blue. (4.3)
Thus in total we characterise a quark state by the three quantum numbers

α, i, f , which gives 4 × 3 × Nf possibilities.
Let us begin with the Lagrangian density of free quarks
X
L = q̄if (iγ µ ∂µ − mf )qif . (4.4)
i,f
For clarity we have explicitly written out the summation over the colour and
flavour indices. The spinor indices are hidden. From this Lagrangian one
gets 3 × Nf Dirac equations, one for each colour and flavour.
Symmetries
Symmetries put very strong restrictions on the structure of physical objects
or relations. For instance, symmetry around one axis of rotation restricts
objects to look as being made on a turning lathe like some chess figures,
and symmetry under the full rotation group allows only spherical objects.
Symmetries also restrict interactions, which to a large extent are fixed by
symmetry requirements. An example is the term Fµν F µν in the Lagrangian.
Higher powers of Fµν F µν would show the same symmetry, but they must be
excluded, because they lead to non-renormalisable theories.
88 4 QUANTUM CHROMODYNAMICS (QCD)
Global SU(3) colour symmetry:

X
′
qif (x) −→ qif (x) = (U q)if = Uij qjf (x) (4.5)
i,f
U is an unitary 3 × 3 matrix. Since there is another U(1) symmetry, U

can be restricted to det U = +1, so we have U ∈ SU(3). U commutes with
operators on flavour and spinor subspace, therefore, the symmetry of the
Lagrangian is proven by
′
q̄if (iγ µ ∂µ − mf )qif
′
= q̄if (U † )ij (iγ µ ∂µ − mf )Ujk qkf
= q̄if (U † )ij Ujk (iγ µ ∂µ − mf )qkf
= q̄if (iγ µ ∂µ − mf )qif .
U(1) baryon symmetry
′
qif (x) −→ qif (x) = e−iα qjf (x) (4.6)
holds for the same reason. The corresponding Noether current is
X
j µ (x) = q̄(x)γ µ q(x) = q̄if (x)γ µ qif (x). (4.7)
i,f
The conserved charge is the integral of the zero component of the Noether
current, which counts the number of quarks. By convention the baryon
number is
1Z 3 0
B= d x j (x). (4.8)
3
Approximate flavour symmetry SU(Nf )

This symmetry would be exact, if all quark masses mf were equal. We shall
give the transformation laws later, when discussing interactions.
4.1.1 Local SU(3) colour symmetry

This symmetry was introduced in the early 1970’s. The idea that colour is
the source of a SU(3) gauge field, was formulated by Fritzsch, Gell-Mann and
Leutwyler. Local SU(3) gauge symmetry is introduced in the following way.
1. Replace the derivative

∂µ −→ Dµ = ∂µ − igAaµ (x)Ta , (4.9)
1
Ta = λa , (a = 1, . . . 8). (4.10)
2
4.1 Lagrangian Density and Symmetries 89
2. This introduces 8 gauge fields, called gluon fields. In the quantised theory
they are associated with 8 new particles, the gluons. The field strengths are
denoted Gaµν ,
Gaµν = ∂µ Aaν − ∂ν Aaµ + gfabc Abµ Acν . (4.11)
3. The Lagrangian of QCD now gets an extra Yang-Mills-action part
1
LY M = − Gaµν Gµν,a , (4.12)
4
X 1 a µν,a
LQCD = q̄(iγµ Dµ − mf )q − G G . (4.13)
4 µν
The gluon fields lead to new Feynman

graphs. The quark propagators are fermionic
lines, and the gluon propagators are similar
to photon lines.
The term q̄ iγ µ Dµ q leads to quark-gluon ver-

tices, similar to QED.
The quadratic term in the gluon field strengths contains self-interaction of

gluons.
Gaµν Gµν,a
1
= − (∂µ Aaν − ∂ν Aaµ + gfabc Abµ Acν )(∂ µ Aν,a − ∂ ν Aµ,a + gfade Aµ,d Aν,e )
4
1
= − (∂µ Aaν − ∂ν Aaµ )(∂ µ Aν,a − ∂ ν Aµ,a )
4
1
− gfabc Abµ Acν (∂ µ Aν,a − ∂ ν Aµ,a ), 3-gluon vertex
2
1
− g 2 fabc fade Abµ Acν Aµ,d Aν,e . 4-gluon vertex
4
⊳
Quantisation of gluon fields in perturbation
theory requires gauge fixing. This introduces
new fields, the ghost fields. Ghost propaga-
tors are represented by dotted lines. There
are also ghost-gluon vertices.22
4.1.2 Global flavour symmetry
While the masses of up and down quarks are nearly the same, the other
quarks are much more massive. As a consequence the flavour symmetry
SU(Nf ) is strongly broken. In spite of this, we can look not only for approx-
imate symmetries but also for symmetries in subspaces of the quark states.
For U ∈ SU(Nf ) we write
qf −→ qf′ ′ (x) = Uf ′ f qf (x). (4.14)
Here, as elsewhere, we hide indices, which are unimportant in the context.

Some examples are:
for quarks q means (qαif (x)),

µ
(γ µ q)αif means γαβ qβif ,
if U ∈ SU(Nf ), U ·q means (U · q)αif = Uf f ′ qαif ′ (x),
if U ∈ SU(3)colour , U · q means (U · q)αif = Uij qαjf (x).
The term q̄(iγ µ Dµ − m)q in the Lagrangian LQCD transforms under U ∈

SU(Nf ) as
q̄(iγ µ Dµ − m)q −→ q̄ ′ (iγ µ Dµ − m)q ′

=q̄(U † iγ µ Dµ U − U † mU )q
=q̄(U † U iγ µ Dµ − mU † U )q
=q̄(iγ µ Dµ − m)q.
Since the quark masses of different flavours are not the same, the mass term
in the Lagrangian is
X
mf q̄f qf = q̄M q, (4.15)
f
22
See Faddeev-Popov ghosts.
with the quark mass matrix

 
mu 0 0 ... 0
 
 0 md 0 ... 0 
 
 0
M = 0 ms ... 0 
. (4.16)

 0 ... .. 
 0 0 . 

0 0 0 ... mb
A total SU(Nf ) flavour symmetry would arise, if
q̄ ′ M q ′ = q̄U † M U q = q̄U † U M q = q̄M q. (4.17)
This would be the case, if all U commute with M , which means M = m1.
Then all quark masses are equal. If some quark masses are approximately
equal, like mu ≈ md , one has an approximate symmetry in their subspace,
e.g. the SU(2)-isospin symmetry or the SU(3)-flavour symmetry for mu ≈
md ≈ ms . In these cases the unitary transformation has the form
!
V 0
U= , V ∈ SU(n). (4.18)
0 1
4.1.3 Chiral symmetry

For processes at very high energies the masses of the quarks are negligible.
This motivates another symmetry, which holds, if all quark masses vanish,
mf = 0. The transformations, which act in flavour and Dirac space, are
called axial transformations and read
qf −→ qf′ = [exp(−iω a Ta γ5 )]f f ′ qf ′ . (4.19)
The invariance of the kinetic term relies on the relation
[γ µ , γ5 ]+ = 0. (4.20)
Let us check this:

! ! ! ! !
1 0 0 1 0 1 0 1 1 0
γ 0 γ5 = = =− = −γ5 γ 0
0 −1 1 0 −1 0 1 0 0 −1
! ! ! ! !
0 σk 0 1 σ 0 0 1 0 σk
k
γ γ5 = = k =− = −γ5 γ k
−σk 0 1 0 0 −σk 1 0 −σk 0
The axial transformations do not form a group:
aT γ ′b T γ ′′a T γ
e−iω a 5
e−iω b 5
6= e−iω a 5
. (4.21)
They are elements of a larger group, the chiral symmetry group, which we
shall discuss now.
In order to find the chiral symmetry group, we introduce the notion of chi-
rality. The chiral projections are defined by
1
PL : q(x) −→ qL (x) := (1 − γ5 )q, (4.22)
2
1
PR : q(x) −→ qR (x) := (1 + γ5 )q. (4.23)
2
We speak of left-handed fields qL or right-handed fields qR .
We see immediately that
q(x) = qL (x) + qR (x) or PL + PR = 1. (4.24)
PL and PR are projection operators
2 1 1
PR,L = (1 ± γ5 ± γ5 + γ5 γ5 ) = (2 ± 2γ5 ) = PR,L . (4.25)
4 | {z } 4
=1
One also has

PL PR = PR PL = 0, (4.26)
from
(1 − γ5 )(1 + γ5 ) = 1 − γ52 = 0. (4.27)
For high energetic particles described by a Dirac wavefunction, PL ψ = ψ or
PR ψ = ψ means that the momentum of the particle is parallel (right-handed)
or antiparallel (left-handed) to the direction of motion.
We shall make use of the following relations.
1
q̄L = q̄ · (1 + γ5 ), (4.28)
2
1
q̄R = q̄ · (1 − γ5 ), (4.29)
2
Check: since 1 − γ5 is Hermitian and γ5 γ 0 = −γ 0 γ5 , one has
1 1 1 1
q̄L = ( (1 − γ5 )q)† γ 0 = q † (1 − γ5 )γ 0 = q † γ 0 (1 + γ5 ) = q̄ (1 + γ5 ),
2 2 2 2
and similarly for q̄R . Also
q̄L qL = q̄R qR = 0 (4.30)

holds due to (1 + γ5 )(1 − γ5 ) = 0. The mass term can thus be decomposed

as
q̄q = (q̄L + q̄R )(qL + qR ) = q̄L qR + q̄R qL . (4.31)
On the other hand, for the terms in the kinetic part of the Lagrangian we
find
1 1
q̄L γ µ qR = q̄(1 + γ5 )γ µ (1 + γ5 )q = q̄(1 + γ5 )(1 − γ5 )γ µ q = 0,
4 4
1 1
q̄R γ qL = q̄(1 − γ5 )γ µ (1 − γ5 )q =
µ
q̄(1 − γ5 )(1 + γ5 )γ µ q = 0.
4 4
Using these relations the Lagrangian density for QCD can be written in terms
of the left-handed and right-handed fields as
X 1
L = {q̄f L iγ µ Dµ qf L + q̄f R iγ µ Dµ qf R − mf (q̄f L qf R + q̄f R qf L )} − Gaµν Gµν,a .
f 4
(4.32)
The mass terms couple the left-handed to the right-handed fields. If all
quark masses vanish, the left-handed and right-handed fields decouple. In
this case there is a larger symmetry, which acts on the left-handed and the
right-handed fields independently. The symmetry transformations are
a a
qL −→ qL′ = e−iωL Ta qL , e−iωL Ta ∈ SU(Nf )L , (4.33)
aT aT
qR −→ qR′ =e −iωR a
qR , e −iωR a
∈ SU(Nf )R . (4.34)
Both groups SU(Nf )L and SU(Nf )R are isomorphic to SU(Nf ). The indices L
or R just indicate that the symmetry transformations act on the left-handed
and the right-handed fields, respectively. The invariance of the kinetic part
of the Lagrangian holds because of
q̄L′ iγ µ Dµ qL′ = q̄L iγ µ Dµ qL , (4.35)

q̄R′ iγ µ Dµ qR′ = q̄R iγ µ Dµ qR , (4.36)
whereas in the mass terms the exponentials do not cancel, i.e.

a a
q̄L′ qR′ = q̄L eiωL Ta e−iωR Ta qR 6= q̄L qR .
So, for mf = 0 there is a symmetry of the Lagrangian under the group
SU(Nf )L ⊗ SU(Nf )R , (4.37)
which has 2(Nf2 − 1) parameters. This symmetry is called chiral symmetry.

The currents, belonging to this symmetry according to Noether’s theorem

are
µ
jLa = q̄L γ µ Ta qL , (4.38)
µ
jRa = q̄R γ µ Ta qR . (4.39)
For example, for the SU(Nf )L transformations we find
µ ∂L δqLf
jLa = = q̄Lf iγ µ (−iTa )qLf = q̄L γ µ Ta qL ,
∂(∂µ qLf ) δωLa
and similarly for the right handed transformations. Using
1 1 1 1
PR γ µ PL = (1 + γ5 )γ µ (1 − γ5 ) = γ µ (1 − γ5 ) (1 − γ5 ) = γ µ PL2 = γ µ PL ,
2 2 2 2
the currents can be written
µ 1
jLa = q̄γ µ Ta (1 − γ5 )q, (4.40)
2
µ 1
jRa = q̄γ µ Ta (1 + γ5 )q. (4.41)
2
Relation to flavour symmetry
The flavour and chiral symmetries are related,
flavour SU(Nf ) ←→ chiral SU(Nf )L ⊗ SU(Nf )R . (4.42)
To see this, consider the infinitesimal transformations
qL′ = (1L − iωLa Ta )qL ,
qR′ = (1R − iωRa Ta )qR , (4.43)
q ′ = (qL′ + qR′ ) = (1 − iTa [ωLa PL + ωRa PR ])q.
If we restrict the chiral transformations to
ωLa = ωRa ≡ ω a , (4.44)
we obtain the infinitesimal flavour transformations
q ′ = (1 − iω a Ta )q. (4.45)
Therefore, the flavour symmetry group is a subgroup of the chiral symmetry
group SU(Nf )L ⊗ SU(Nf )R . It is called the diagonal subgroup, because it
consists of those group elements UL ⊗ UR which obey UL = UR .
flavour group = {UL ⊗ UR | UL , UR ∈ SU(Nf ), UL = UR }
(4.46)
≃ SU(Nf ) ⊏ SU(Nf ) ⊗ SU(Nf )
The flavour current belonging to the flavour symmetry is

µ µ
jaµ = q̄γ µ Ta q = jLa + jRa . (4.47)
The flavour current is also called vector current, since it transforms as a

Lorentz vector.
Infinitesimal axial transformations are defined by
ωLa = −ωRa ≡ −Ωa (4.48)
and read (PL − PR = −γ5 )
q ′ = (1 − iTa [ωL PL + ωR PR ])q

= (1 − iTa (−Ωa )(PL − PR )q (4.49)
= (1 − iΩa γ5 Ta )q.
The corresponding axial current or axial vector current is given by

µ µ µ
j5a := q̄γ µ γ5 Ta q = jRa − jLa . (4.50)
It transforms as an axial vector under Lorentz transformations, i.e. if the

axial vector current is looked at through a mirror, symbolised by the space
reflection (parity) operator P , then left and right would be interchanged, and
µ µ µ µ
the current changes sign: P (jRa − jLa ) = −jRa + jLa .
4.1.4 Broken chiral symmetry

Symmetries of a physical theory are not just mathematical curiosities. In
general they give strong restrictions on the structure and behaviour of the
corresponding models. In case of a field theory, if the Lagrangian possesses a
certain symmetry and if this symmetry is also realised by the physical states,
then as a consequence various relations between observables emerge.
Now we shall consider the question, in which way chiral symmetry of the
QCD Lagrangian is realised in nature. If chiral symmetry were realised on
the physical states, then e. g. the mesons and baryons should form multiplets
of the symmetry group, such that the members of a multiplet have the same
mass.
Consider the case Nf = 2 with the chiral SU(2)L ⊗ SU(2)R symmetry. The
flavour symmetry subgroup, the SU(2) isospin symmetry, is an approximate
symmetry in nature, as is shown by the up and down quarks, by the nucleons
p, n, and by the pions π ±, 0 .
On the other hand the chiral SU(2)⊗SU(2) group is not found as a symmetry
group in nature. One can show that this symmetry would imply that to each
isospin multiplet of particles with definite parity there should belong another
isospin multiplet with the same mass but opposite parity. This is, however,
not observed in nature. For instance, the pions form an isospin triplet of
pseudoscalar particles, but there is no isospin triplet of scalar particles with
masses near the pion masses. The same holds for other SU(N)-multiplets.
If the Lagrangian L possesses a symmetry that is not realised in nature, one
speaks of a “spontaneously broken symmetry”, in contrast to an explicitly
broken symmetry, where the breaking results from symmetry breaking terms
in the Lagrangian.
Spontaneously broken symmetries are already familiar from everyday life and
classical physics, as the following examples show.
a) Swimming
If one puts a rectangular piece of wood in water, it can take two positions, in
which one end of the wood is immersed deeper than the other (supposed the
density of the wood is not too different from the density of water). There is
obviously a reflection symmetry in the Lagrangian, which is broken by the
outcome of the experiment.
b) Ferromagnet
a2
a1
A ferromagnet shows spontaneous magnetisation below the Curie point, T <
TC . The directions of the molecular magnetic moments m(~
~ r) give up their
random distribution in favour of one macroscopic direction, although the
4.2 Running Coupling 97
Hamiltonian X
H = −J m(~
~ r ) · m(~
~ r + ~aµ ) (4.51)
~
r,µ=1,2,3
obeys rotation symmetry according to
m(~
~ r) −→ R m(~
~ r ), R ∈ SO(3). (4.52)
The average values of the molecular magnetic moments turn out to be non-
zero:
hm(~
~ r )i = m
~ 0 6= 0. (4.53)
In addition to explicitly and spontaneously broken symmetries, there is a

another kind of symmetry breaking in quantum theory, which originates from
a pure quantum effect. An example of such a broken symmetry is given by
the axial U(1)A symmetry
q(x) −→ q ′ (x) = e−iαγ5 q(x). (4.54)
The corresponding current density
j5µ (x) = q̄(x)γ µ γ5 q(x) (4.55)
is an axial vector current. If mf = 0 then U(1)A is a symmetry of the

Lagrangian. Classically the axial vector current is conserved. But this sym-
metry is not realised in nature. (We shall not discuss the arguments for this
fact here.) It turns out that in the quantum theory the conservation of the
axial vector current is violated,
∂µ j5µ 6= 0. (4.56)
A situation, where a classically conserved current is not conserved in the

quantum theory, is called an “anomaly”. So, anomalies are different from
spontaneous symmetry breaking, they are quantum effects.
4.2 Running Coupling

The running of the QCD coupling is very important for phenomenology and
experiments. A thorough discussion would require to treat renormalisation
and the Callan-Symanzik equations. Here we just give a short outline of
this topic, presenting some results and experimental evidence for the varying
coupling “constants” of QCD.
4.2.1 Quark-quark scattering

We start with a perturbative theory along
the lines of QED. There is no 4-quark di- p1 q
agram, so we expand the graph for quark-
quark scattering in increasing order of the
coupling. The simplest term corresponds
to the Coulomb scattering of electrically
q q
charged particles, i. e. Rutherford scattering.
The simplest, tree level diagram describes a

one gluon exchange. The mediating gluon p1 p′1
carries 4-momentum
q = p′1 − p1 = −(p′2 − p2 ). (4.57) q
q is a spacelike vector, p2 p′2

q 2 < 0. (4.58)
The scattering amplitude is determined by the gluon propagator and two

vertices
(
2 1 1 1 rep. 6
M = −g ū(p′1 )γµ u(p1 ) 2 ū(p′2 )γµ u(p2 ) , , (4.59)
q 3 −2 rep. 3∗
where u(pi ) are spinors, denoting solutions of the Dirac equation, and the
simple form of the gluon propagator 1/q 2 is due to Feynman gauge. This
factor is dominant for the behaviour of the scattering.
Let us compare this with the e− µ− scattering in QED, where the muon
mass is about 200 times the electron mass, so we are near the static limit of
Rutherford scattering. Therefore, |p1 | = |p′1 | =: k and
1 1
M ∼ 2
≈− 2 2 θ . (4.60)
q 4k sin ( 2 )
The differential cross-section then obeys
dσ 1
∼ |M |2 ∼ . (4.61)
dΩ sin ( 2θ )
4
Quantum mechanics tells us that in the Born-approximation M is propor-

tional to the Fourier transformation (F T ) of the potential, thus
1 1
∼ F T (V (r)) ⇒ V (r) ∼ . (4.62)
q2 r
Thus, in lowest order of perturbation theory the quark-quark potential is a

Coulomb potential. Since we only considered one-gluon exchange, we should
not be surprised to get the same result as in first order of perturbative QED.
In QCD, higher orders of perturbation theory add corrections to 1/q 2 , which
will modify this result. The 1-loop corrections are the following:
(a) (b) (c) (d)
(e) (f) (g)
The diagrams are denoted as: quark-loop (a), gluon-loop (b), gluon-tadpole
(c), ghost-loop (d), quark-propagator correction (e), and vertex corrections
(f), (g). Additional graphs, which are similar to (e) – (f) are not shown.
The result of these calculations in the limit of large momentum transfer,
Q2 := −q 2 large, (4.63)
is essentially a modification of the one-gluon exchange case, with a corrected
coupling constant
( !)
2 2 g2 Q2
g −→ g 1 − b0 ln , (4.64)
16π 2 M2
with
1 2
b0 = (11Nc − 2Nf ) = 11 − Nf > 0 (for Nc = 3 colours). (4.65)
3 3
b0 is positive as long as the number of flavours is not larger than 16, and
there is no experimental evidence for that. The number 11 comes from the
graphs (b) to (g), while 32 Nf stems from graph (a).
The expression contains a mass M . It enters the result due to renormali-

sation. The loops in the Feynman diagrams involve divergent integrals over
the momenta of the type
Z 4
dp
. (4.66)
p4
The renormalisation procedure removes the divergencies in two steps,
1. Regularisation: introduces a cutoff for the momentum integration, or
analogous divergent integrals. Typical regularisations introduce a cut-
off mass M , which characterises the largest allowed momenta. Loop
integrals are then finite, but depend on M .
2. Renormalisation.
4.2.2 Renormalisation
The coupling g is a parameter in the Lagrangian and is not directly mea-
surable. We call it “bare” coupling and from now on denote it by g0 . A
measurable, physical relevant coupling should be defined or fixed through
some process. Here, for example, we fix it by choosing a fixed mass scale µ
and considering quark-quark scattering at a momentum transfer Q2 = µ2 .
The modified coupling at this momentum transfer is defined to be the renor-
malised coupling
( ! )
g2 µ2
gR2 (µ) = g02 1 − b0 0 2 ln + O(g04 ) . (4.67)
16π M2
Renormalisation now means

• express everything in terms of gR in place of g0 ,
• then remove the cutoff by taking the limit M −→ ∞.

Renormalisable theories are those which give finite physical results by this
procedure. (Eventually other parameters like quark masses, etc., have to be
renormalised, too.)
In our case, expressing g0 in terms of gR (µ) and expanding in powers of gR (µ),
one gets the scattering amplitude
( ! )
1 g 2 (µ) Q2
M ∼ 2 gR2 (µ) 1 − b0 R 2 ln + O(gR4 ) . (4.68)
Q 16π µ2
Indeed, M does not longer contain the cutoff M , and is finite.

4.2.3 Running coupling

The result for M suggests to define an effective coupling for this process,
( ! )
g 2 (µ) Q2
gR2 (Q) := gR2 (µ) 1 − b0 R 2 ln + O(gR4 ) , (4.69)
16π µ2
which indicates that the scattering at momentum transfer Q2 is like Ruther-
ford scattering with an effective coupling gR2 (Q). We see that the effective
coupling is not constant. Therefore it is called “running coupling”. More-
over, in the one-loop expression above the corrections get large for large
Q2 . (Throughout our calculation we have made the assumption of large Q2 .)
Therefore, we have to extend the perturbation series. Calculating higher or-
ders of perturbation, the dominant contributions can be found to lead to a
geometric series, so
" !#−1
g 2 (µ) Q2
gR2 (Q) ∼ gR2 (µ) 1 + b0 R 2 ln . (4.70)
16π µ2
The figure shows the dependence of the running coupling on Q2 .
gR2 (Q)
gR2 (µ)
running coupling
perturbative regime
Q2
2 2
Λ µ
The coupling gets weaker for higher momentum transfer, supporting the
suggestion that perturbation theory is reliable at large Q.
A more solid consideration is based on the renormalisation group, in par-
ticular the Callan-Symanzik equations, which are beyond the scope of this
lecture. The renormalisation group equation for the running coupling is
d
Q gR = β(gR ) (4.71)
dQ
g3
= −b0 R 2 + . . . (odd powers), (4.72)
16π
where β(gR ) is the famous beta-function of QCD. Thus we have

Q increases −→ gR decreases. (4.73)
So, perturbation theory is reliable at large Q.
Here, a remark is in order. The perturbation series of QCD or QED are not
convergent in the mathematical sense. However, the error made by truncating
the series at some finite order can be controlled due to Borel summability.
This allows to get useful results from finite orders of perturbation theory in
a suitable regime of applicability, which in QCD is the high energy regime.
Let us now consider the region of small Q, or the low energy range. In QCD,
b0 > 0, therefore, for small enough Q, ln(Q2 /µ2 ) becomes negative, and in
the considered approximation, gR (Q) gets a pole at a certain point Q = Λ.
The QCD Λ-parameter is (roughly) in one-loop order defined by
!
g 2 (µ) Λ2
b0 R 2 ln = −1. (4.74)
16π µ2
Its value (in a particular scheme, the M S scheme) is about ΛM S ≈ 200 MeV.
The apparent pole at Q = Λ is actually an artefact of the approximation

and is not present for the full gR (Q). A more precise definition of Λ, which
requires two-loop terms, can nevertheless be given.
Using the Λ-parameter, for µ ≈ Λ and Q2 ≫ Λ2 , the high energy behaviour
of the coupling constant can be written as
16π 2
gR2 (µ) =
Q2
. (4.75)
b0 ln Λ2
This decrease of the running coupling towards zero is called “asymptotic

freedom”,
lim gR (Q) = 0. (4.76)
Q→∞
Asymptotic freedom in QCD was first realised by Gerard ’t Hooft in the work
on his thesis as a student of Martinus Veltman, but was not published. In
1973 asymptotic freedom in field theories has been discovered by D. J. Gross,
F. Wilczek and H. D. Politzer, who were awarded the Nobel prize in 2004.
4.2.4 Discussion
a) Asymptotic freedom guaranties a good applicability of perturbation theory
on processes at high energies or momenta. At low energies perturbation
calculation fails.
b) In QED the renormalisation group parameter is b0 < 0. This comes from

the 1-loop calculation of e− e+ scattering. Taking the electric charge for the
coupling constant, one has
" !#−1
e2 Q2
e2R (Q) = e2R 1 − |b0 | R 2 ln . (4.77)
16π m2e
In QED the running fine-structure constant
e2 (Q) e2R (Q) 1

α= = ≈
4πǫ0 ~c 4π 137.036
shows a tiny increase with momentum.
2
gR (Q)
e2R
4π
Q2
Measurements at 100 GeV have given a value of α−1 ≈ 120.

The Λ-point (“Landau pole”) belongs to a momentum already beyond the
Planck-scale.
c) Interpretation
Since Q ∼ 1r and b0 < 0 in QED, the electric charge eR (1/r) increases as
r → 0. The decrease of the measured charge, when going away from r = 0,
has been interpreted as a screening of charges due to vacuum polarisation.
In QCD on the other hand, where b0 > 0, there are also gluon loops and the
gluons have self-interactions. This leads to an anti screening effect.
d) Experiments
2 (Q)
gR
The strong “fine-structure constant” αS = 4π
of QCD has been measured
in several experiments:
• deep inelastic e− p scattering,

• τ -lepton decay,
• e+ e− scattering, which has contributions of virtual q q̄ processes,
• Ψ, Υ (heavy quarkonia) spectroscopy,
• Z 0 -decay.
The results are in very good agreement with the prediction from asymptotic
freedom.
Figure 3: Summary of measurements of αs (Q), from S. Bethke, Prog. Part.

Nucl. Phys. 58 (2007) 351–386, hep-ex/0606035.
4.3 Confinement of Quarks and Gluons 105
4.3 Confinement of Quarks and Gluons

In the QCD Lagrangian the quarks and gluons carry colour charge, but these
particles do not exist as free particles. Accessible particles are the hadrons:
mesons (q q̄) as quark-antiquark states and baryons (qqq), made of three
quarks. All these particles are colour neutral. Therefore, one states the
Confinement Hypothesis:
Physical states are colour neutral;
in particular, quarks and gluons do not exist as free particles.
Since perturbation theory works with quark and gluon fields, and assumes
these constituents as ingoing and outgoing particles, confinement cannot be
proven in the framework of perturbation theory. To demonstrate confine-
ment is a non-perturbative problem. A rigorous proof of confinement is still
missing.
Instead, there are some phenomenological descriptions.
• The bag model

is based on antiscreening effects. The particles are assumed to be con-
fined in a region with boundary conditions similar to boundary con-
ditions of electrodynamics. In electrodynamics or optics one has the
polarisability of the matter, which in the bag model is replaced by the
polarisability of vacuum.
-
+
- -
+ +
- + + + -
ǫ
+ + ~r
ǫvac < ǫ - -
+
-
• In the string picture

the quark-antiquark potential at large distances grows proportional to
the distance 23 ,
V (r) = kr. (4.78)
23
For the static QCD potential see G.S. Ball, Phys. Rept. 343 (2001) 1.
The force, therefore, remains constant, and when one tries to separate
the quarks, energy grows until it suffices to generate a new quark-
antiquark pair. The situation may depicted by lines of force, similar to
electrodynamics, with the exception that the vacuum compresses the
lines to flux tubes, which eventually split. Thus the chromoelectric flux
tubes are squeezed by being “repelled” from the vacuum.
q q̄
The bag model and the string model are based on postulated dielectric
properties of the vacuum with a (relative) dielectric constant
ǫr < 0 ←→ antiscreening. (4.79)
Lattice QCD
is able to provide the spectrum of particles and many other physical quan-
tities from first principles. The calculated hadron masses are in very good
agreement with the experimental values. Practical limitations are the num-
ber of lattice points and/or the lower limit to the lattice spacing, which
typically is of the order of 0.1 fm. This also puts an upper limit to momenta.
643 × 128
lattice points
ïż£ In pure gauge theory, for a

gluon field with fixed quarks, the static quark-antiquark potential can rather
well be described by
c
V (r) = + k · r + const. (4.80)
r
Here the c/r term comes from perturbation theory, which is valid at small dis-
tances. Lattice calculations at large r lead to the static confinement V (r) ∼ r.
4.3 Confinement of Quarks and Gluons 107
Lattice QCD with quarks supports the string picture as it results in string
breaking at larger quark distance.
q q̄
q q̄
q q̄ q q̄
In this case the quark-quark potential is of the form
V (r)
V = const.
string breaking →
∼ kr (+const.)
1
∼ r
r
Lattice simulations in pure gauge theory reveal pure gluonic, colour neutral
particles, the so-called glue-balls. These also exist in full QCD with quarks.
Jets
In perturbation theory, scattering processes in QCD are based on the scat-
tering of quarks and gluons. On the other hand, all experimental information
stems from the interaction of observable particles, namely hadrons. There-
fore, the perturbative results for scattering processes in QCD have to be
“hadronised”. Consider for example proton-antiproton scattering.
−→ ? ←−
q −→ ←− q̄
Perturbation theory gives cross-sections for quark-quark scattering. Hadro-
nisation turns the outgoing quarks and antiquarks into hadrons. This leads
to the formation of jets of hadrons (S. Weinberg, G. Sterman). At high
momenta the primary quarks and antiquarks combine with newly created
quarks and antiquarks and form jets.
creation of q q̄g
q −→ ←− q̄
The formation of jets is a non-perturbative process.
4.4 Experimental Evidence for QCD

In general, there is a good consistency between the theoretical and experi-
mental results for QCD. In the following we list some examples, where ex-
periments agree well with theoretical modelling and/or predictions.
• Running coupling αS (Q).
• Jet distribution.
The statistical angular distribution and the momentum distributions
of the particles in jets were quantitatively predicted by QCD.
• Evidence for gluons.
– Deep inelastic e− p+ scattering at HERA.
e− p+
4.4 Experimental Evidence for QCD 109
It is found that quarks carry only about 50 % of the momentum

of the nucleon. The other half has to come from particles, which
are neutral with respect to electroweak interactions. This is an
indication for gluons.
– Three-jet events.
Figure 4: A three-jet event observed at PETRA. The figure is from:

F. Halzen, A. D. Martin, Quarks & Leptons: An Introductory Course in
Modern Particle Physics, J. Wiley & Sons (1984).
They are explained by additional particles. The angular distribu-

tion of the jets leads to spin 1 particles, which are identified with
the gluons, in agreement with the QCD prediction.
• Lattice QCD.
The masses of stable hadrons have been calculated in lattice QCD.
The masses mu = md , ms of up, down, and strange quarks, which
are needed as input parameters, have been obtained by fitting the
proton and Kaon masses, respectively. The lattice simulations from
the Budapest-Marseille-Wuppertal collaboration have given the baryon
masses for N, Λ, Σ, ∆, Σ∗ , Ξ, Ω, and the masses of the mesons

ρ and K ∗ . The results are within the experimental errors.
There have been many other experiments that support QCD theory. Let us
mention one more example: Calculating the anomalous magnetic moment
of the muon in QED in the same way, as it was done for the electron, the
result does not agree with the extremely precise experimental value. How-
ever, taking hadronic contributions from QCD into account, the agreement
is perfect.
111
5 Electroweak Theory
5.1 Weak Interactions
The weak interactions of elementary particles are distinguished from other
interactions by some characteristic properties like lifetimes, strength of cou-
pling, cross-sections, and violation of symmetries. We refer to the introduc-
tory lectures on particle physics for details.
Some typical processes of weak interactions are the following.
a) Leptonic processes
Muon decay: µ− −→ e− + ν̄e + νµ
eν-scattering: e− νµ −→ µ− + νe
b) Semi-leptonic processes. They involve hadrons.
β-decay n −→ p+ + e− + ν̄e
β-decay in quark picture
d −→ u + e− + ν̄e
Pion decay24 π − −→ µ− + ν̄µ (dū −→ µ− + ν̄µ )
c) Non-leptonic weak interaction processes
Λ decay Λ0 −→ p+ + π − (uds −→ uud + dū)
Kaon decay −
K −→ π + π− 0 ¯
(sū −→ dū + √12 (−uū + dd))
These weak interaction processes violate isospin symmetry.
5.1.1 Fermi theory of weak interaction

As mentioned in the introduction, in 1932 Enrico
Fermi formulated a theory for the β-decay of the p e−
neutron as a four fermion process. The Lagrangian
consists of the free parts for n, p+ , ν, and e− , plus ν̄e
the four-fermion interaction term n
Lw = G (ē(x)γ µ νe (x)) (p̄(x)γµ n(x)) . (5.1)
Lw is invariant under space reflections. (Parity operator P.) In Fermi’s theory

parity is conserved.
5.1.2 Parity violation

In 1957 parity violation, occurring in the β-decay of 60 Co, was experimentally
established by Wu. At that time, when the neutrinos were regarded to be
There also is a dominant hadronic decay mode giving the same particles. The leptonic
24
mode is distinguished by its timescale.

112 5 ELECTROWEAK THEORY
massless particles, the νe was always considered to be left-handed, and the

ν̄e right handed. Madame Wu found that the e− arising in this decay is
dominantly left-handed, violating parity conservation. Parity violation in
Kaon-decay was proposed by Lee and Yang in 1966.
In order to describe parity violation and handedness of neutrinos, let us recall
some facts about chirality from Sec. 4.1.3. The projection operators on left-
and right-handed spinors are, respectively
1 1
PL = (1 − γ5 ), PR = (1 + γ5 ). (5.2)
2 2
With ψL = PL ψ, ψR = PR ψ, the Dirac Lagrangian is
L = ψ̄(iγ µ ∂µ − m)ψ (5.3)

µ µ
= ψ̄L iγ ∂µ ψL + ψ̄R iγ ∂µ ψR − m(ψ̄L ψR + ψ̄R ψL ). (5.4)
If m = 0, L separates into a left-handed and a right-handed part.

Consider a left-handed Dirac spinor plane wave solution
1
u(k) = uL (k) = (1 − γ5 )u(k), (5.5)
2
1 1
γ5 uL (k) = γ5 (1 − γ5 )uL (k) = (γ5 − 1)uL (k) = −uL (k), (5.6)
2 2
2
since (γ5 ) = 1.
With the spin-operator S ~ = (1/2)Σ ~ we define he-
licity λ to be the projection of spin onto the direc-
tion of momentum. In the massless case one can ~k
show that helicity and chirality are equal:
~
~ · k uL (k) = γ5 uL (k) = −uL (k).
Σ (5.7)
|~k| ~
Σ
The helicity in this case is λ = − 21 .

Similarly γ5 uR (~k) = +uR (~k), the particle is right-handed, and the helicity is
λ = + 21 .
Since all neutrinos were found to be left-handed and all antineutrinos to be
right-handed, we have
1
ν = νL = (1 − γ5 )ν, (5.8)
2
1
ν̄ = ν̄R = (1 + γ5 )ν̄. (5.9)
2
5.1 Weak Interactions 113
The non-existence of right-handed neutrinos implies that parity is violated.

Nowadays, experiments have shown that neutrinos are massive particles. Due
to their very small mass, they almost always move with a speed very close to
the speed of light. Nevertheless, in a moving frame that moves even faster,
they appear to be right-handed. So, helicity (and chirality) can take both
values.
5.1.3 V-A theory
The V-A theory was developed by Feynman and Gell-Mann in 1958 and
independently by Marshak and Sudarshan in the same year. It assumes
massless neutrinos and takes chirality and parity violation into account.
The neutrino part in the Lagrangian is replaced by25
1
ē(x)γµ νe (x) → ē(x)γµ (1 − γ5 )νe (x) (5.10)
2
= ēL (x)γµ νe,L (x)
1 1
= ē(x)γµ νe (x) − ē(x)γµ γ5 νe (x) (5.11)
|2 {z } |2 {z }
vector current axial vector current
1 (e)
= Vµ (x) − A(e)
µ (x) . (5.12)
2
Thus, the V-A theory modifies the Fermi theory, represented by the vector
current V , by subtracting the axial vector current term A.
There are further contributions, which come from the muon and tau lepton.
The total weak leptonic current then reads
Jµ(l) = 2ēL (x)γµ νe,L (x) + 2µ̄L (x)γµ νµ.L (x) + 2τ̄L (x)γµ ντ,L e(x). (5.13)
Hadronic Currents
Hadronic currents enter in the same way as leptonic currents. For the process
that a down quark is converted to an up quark, the current is
Jµ(h) = 2ūL (x)γµ dL (x). (5.14)
(For nucleons one would write 2p̄L γµ nL .)

25
Compare the calculations for vector currents on page 94.
In total we have
Jµ (x) = Jµ(l) (x) + Jµh) (x). (5.15)

GF
Lw = √ Jµ (x)J µ (x)+ (5.16)
2
GF +
= √ Jµ(l) (x)J µ(l) (x) leptonic interactions
2
+ +
+ Jµ(l) (x)J µ(h) (x) + Jµ(h) (x)J µ(l) (x) semi-leptonic interactions
+
+ Jµ(h) (x)J µ(h) (x) . hadronic interactions
The Fermi coupling constant GF is the same for all weak interaction pro-
cesses. Its value is
GF = 1.03 · 10−5 m−2

p = 4.51 · 10
−33
cm2 = 1.17 · 10−5 GeV−2 . (5.17)
Modifications of the V-A theory

a) For nucleons (and other hadrons) the hadronic current is replaced by
Jµ(h) = gV V (h) − gA A(h) . (5.18)
While gV = 1 always, the value of the other parameter is in the case of

nucleons
gA ≈ 1.24, (5.19)
whereas for quarks gA = 1.
b) For the u, d, s and c-quarks, forming the doublets
! !
u c
, , (5.20)
d s
the Lagrangian of V-A theory only allows transitions u ↔ d and c ↔ s

between quarks. At times, when the charmed quark was not yet known,
weak processes had been observed that change strangeness, like the weak
Kaon decay
K + −→ µ+ + ν̄µ . (5.21)
Numerical relations between strangeness-conserving and strangeness-changing
processes motivate to incorporate strangeness-changing processes by postu-
lating that mixtures of s and d states
d′ =d cos θC + s sin θC (5.22)

s′ =s cos θC − d sin θC . (5.23)
5.1 Weak Interactions 115
take part in the weak interactions. Including charmed particles, the hadronic
part of the current is
Jµ(h) = 2ūL γµ d′L + 2c̄L γµ s′L . (5.24)
The angle θC is called Cabibbo angle, its value is
θC ≈ 13◦ . (5.25)
The V-A theory very successfully describes many weak processes at relatively
low energies. Owing to the structure of the currents Jµ , in the processes
described by the V-A theory, the electric charge does always change in the
leptonic sector. One speaks of “charged currents”.
Problems of the V-A theory
• It is not renormalisable. We have outlined the theory based on tree
level diagrams. Adding loop-corrections will give non-renormalisable
infinities.
• It shows bad behaviour at high energies. Scattering cross-section be-

have like
σ ∼ G2F · s, s = (p + q)2 = ECM
2
S (5.26)
At high energies this violates the rigorous “unitary bound”
4π
σ≤ . (5.27)
s
This happens at s ≈ 1 TeV2 .
• The discovery of neutral currents at CERN, 1973. The following pro-

cesses have been found. (X stands for a hadron.)
ν̄µ + e− −→ ν̄µ + e− ,
νµ + N −→ νµ + X,
ν̄µ + N −→ ν̄µ + X.
Intermediate vector bosons

In order to improve the high-energy behaviour, it was proposed that the weak
interactions are mediated by bosons, analogous to the photon in QED. For
the charged currents two charged spin 1 particles, the vector bosons Wµ(+)
and Wµ(−) have been postulated. If these particles were massless like photons,
their range would be infinite in contrast to the short range nature of weak
interactions. Therefore, the W-bosons are massive.
The coupling to other fields are given by
Lw = gW Jµ W (−)µ + gW Jµ+ W (+)µ . (5.28)
In β-decay, for example, the 4-fermion-vertex of the Fermi theory is replaced

by a graph containing a W (+) propagator.
u e− u W (+) e−
−→
d ν̄e d ν̄e
For low energies the V-A theory should be reproduced. This is the case if
2
gW GF
2
=√ , (5.29)
mW 2
and consequently the mass mW has to be large.
To describe the neutral currents within this approach, there should in addi-
tion exist a neutral intermediate vector boson, denoted Z (0) .
In fact, in 1983 the W and Z bosons have been found with the Super-Proton-
Synchrotron (SPS) at CERN. Nevertheless, there remains a serious problem:
The theory of intermediate vector bosons is not renormalisable. This fact is
related to the mass of the intermediate particles. (The renormalisability of
QED is connected with the fact that photons are massless.)
So we are faced with the question:
Is there a theory with intermediate vector bosons,

which is renormalisable ?
One can think of two different approaches to answer this question. On the
one hand, one could start with the most general Lagrangian, having all imag-
inable kinds of intermediate particles and appropriate symmetries, and then
sort out all those theories, which are non-renormalisable. The other path,
which we shall follow, is to try to find a gauge theory that describes the weak
interactions. The immediate problem with this attempt is that no mass term
is allowed for gauge bosons, the gauge bosons are massless.
The solution to this problem is the “Higgs mechanism” 26
26
Here, the word “mechanism” does not mean a physical process but a theoretical con-
struction.
5.2 Higgs Mechanism 117
5.2 Higgs Mechanism

5.2.1 Spontaneous breakdown of a global symmetry
We shall now consider a concept in field theory, which is also known in other
areas of physics like phase transitions, e.g. ferromagnetism, spin waves or
even laser theory. Consider a complex scalar field with Lagrangian
L = ∂µ φ∗ ∂ µ φ − m2 φ∗ φ − λ(φ∗ φ)2 . (5.30)
The coefficient m2 should be thought of as a real parameter, which can be

negative, too. L has a global U(1) symmetry
φ(x) −→ eiα φ(x). (5.31)
Let us look at the potential part of the Lagrangian
V (φ) = m2 φ∗ φ + λ(φ∗ φ)2 (5.32)
as a real valued function over the complex plane with

1
φ = √ (φ1 + iφ2 ). (5.33)
2
Case a) m2 > 0
V (φ)
iφ2
φ1
The symmetry is unbroken, the minimum of

the potential is not degenerate, and the mean value for the field is
hφi = 0. (5.34)
Case b) m2 < 0
V (φ)
iφ2
φ1
Now the minima of the potential are lo-
cated on a circle with radius
s
−m2 v
|φ| = =: √ . (5.35)
2λ 2
In the classical theory, a groundstate corresponds to a field configuration
with minimal energy, which in this case is a constant field φ0 with a value in
one of the potential minima. Without loss of generality we choose it to be
real,
v
φ0 = √ . (5.36)
2
Let us assume that in the quantum theory the corresponding situation holds
and the mean value of the field is
v
hφi = √ . (5.37)
2
(More precisely there will be corrections to this value.) The symmetry is
then spontaneously broken.
We decompose φ(x) into its mean value and a complex remainder,
1
φ(x) = √ (v + ρ(x) + iϕ(x)). (5.38)
2
iφ2
i √ϕ2
−φ1 √ρ
√v 2
2
5.2 Higgs Mechanism 119
The Lagrangian density, written in terms of the real fields ρ(x) and ϕ(x),
becomes up to a constant
1 1 λ
L = (∂µ ρ)2 + (∂µ ϕ)2 − λv 2 ρ2 − λv(ρ3 + ρϕ2 ) − (ρ2 + ϕ2 )2 . (5.39)
2 2 4
We see that ρ(x) is a massive field coupled to the massless field ϕ(x). The
mass mρ is given by
m2ρ = 2λv 2 = 2|m2 |. (5.40)
ρ(x) describes radial excitations of the original field. The massless field ϕ(x)
corresponds to tangential excitations. These are called Goldstone bosons.
A general statement about the existence of Goldstone bosons is made by the
Goldstone theorem:
Spontaneous breakdown of a continuous, global symmetry
leads to one massless spin 0 particle for each generator of
the group that is spontaneously broken.
In the above situation, the U(1) group has one generator, and one Goldstone
boson field arises.
5.2.2 Higgs mechanism

Now we turn to a seemingly similar situation, where the symmetry is, how-
ever, a local gauge symmetry. Will there be massless Goldstone bosons?
Let us consider the abelian Higgs model, describing a complex scalar field
coupled to gauge field.
1
L = (Dµ φ)∗ (Dµ φ) − V (φ) − Fµν F µν , (5.41)
4
Dµ = ∂µ + iqAµ . (5.42)
In this “scalar QED”, Aµ (x) is the gauge field and the gauge group is U(1).
As before the potential is a quartic one,
V (φ) = m2 φ∗ φ + λ(φ∗ φ)2 , (5.43)
and we consider the case m2 < 0.

As in the previous section, we expand the field φ(x) near the real minimum
of V (φ), but this time we choose a different parameterisation in terms of a
radial variable ρ(x) and an angular variable ξ(x):

1
φ(x) = √ (v + ρ(x))eiξ(x)/v (5.44)
2
1
= √ (v + ρ(x) + iξ(x) + . . . ). (5.45)
2
After a few lines of algebra the Lagrangian becomes
1
L = (∂µ ρ)2 − λv 2 ρ2
2
1
+ (∂µ ξ)2 + qvAµ ∂ µ ξ
2 (5.46)
1 2 2 1
+ q v Aµ Aµ − Fµν F µν
2 4
+ . . . (interaction terms).
As before, the mass belonging to the radial excitations is
m2ρ = 2λv 2 . (5.47)
There is also term quadratic in the gauge field, namely 21 q 2 v 2 Aµ Aµ , from

which we read off a mass for the field Aµ , which is
mA = qv. (5.48)
There is no mass term for the field ξ(x) and it looks like a massless Goldstone
field. However, there is a mixing term qvAµ ∂ µ ξ, quadratic in the fields, and
we have to clarify the situation.
Looking at the parameterisation of φ(x), we recognise that the factor

i
exp ξ(x) , (5.49)
v
has precisely the form of a local gauge transformation. Therefore the field
ξ(x) can be transformed away by means of a compensating gauge transfor-
mation:

i 1
φ(x) −→ φ′ (x) := exp − ξ(x) φ(x) = √ (v + ρ(x)). (5.50)
v 2
At the same time the gauge field Aµ is transformed as
1
Aµ −→ A′µ (x) = Aµ (x) + ∂µ ξ(x) =: Bµ (x). (5.51)
qv
5.3 Glashow-Weinberg-Salam Model 121
This gauge is called “unitary gauge”. The Lagrangian now becomes
1 m2 λ
L = (∂µ ρ)2 − ρ ρ2 − λvρ3 − ρ4
2 2 4
1 1
− Fµν F µν + q 2 v 2 Bµ B µ (5.52)
4 2
1
+ vq 2 ρBµ B µ + q 2 ρ2 Bµ B µ ,
2
where Fµν is now formed from Bµ (x). The mixing term has disappeared.
The remaining physical fields are the following.
• ρ(x) is a massive scalar field with m2ρ = 2λv 2 .
• Bµ (x) is a massive vector field with mass mv = qv. It describes “mas-

sive photons”.
The usual massless photons have two transverse degrees of freedom (com-
ponents), whereas the massive field Bµ (x) has three degrees of freedom. Its
additional “longitudinal” degree of freedom originates from the alleged Gold-
stone field ξ(x). This is sometimes expressed by saying, “the Goldstone boson
is eaten by the vector boson.”
The last two terms in the Lagrangian are interaction terms between the two
fields.
In the unitary gauge the local gauge symmetry is hidden. In the literature
one sometimes finds the formulation of a “spontaneous breaking of local
gauge symmetry”. This is not quite correct. Local gauge symmetries cannot
be broken spontaneously (“Elitzur’s theorem”). Instead, the fields ρ(x) and
Bµ (x) are gauge invariant combinations of the original fields.
5.3 Glashow-Weinberg-Salam Model

The Glashow-Weinberg-Salam model is a gauge theory describing the elec-
troweak interactions. The gauge bosons, mediating the weak and electromag-
netic interactions, are the massive W + , W − , Z 0 , and the massless photon
Aµ . The crucial piece in the formulation of a gauge theory with massive
vector bosons is the Higgs mechanism. The gauge group is
gauge group = SU(2) ⊗ U(1)Y (5.53)

| {z } | {z }
weak isospin weak hypercharge
and the model uses the Higgs mechanism to create the masses mW and mZ .
Historically, the basic structure was formulated by Sheldon Glashow (1961),

but without the Higgs mechanism, and the complete form was found by
Steven Weinberg (1967) and Abdus Salam (1968).
Let us start by concentrating on leptons. In the weak interactions only the
left handed components couple to the SU(2) gauge field. Therefore the group
is denoted SU(2)L . It acts on the doublets
! !
νe 1 ν
= (1 − γ5 ) −e , (5.54)
e− L
2 e
and on the doublets of the remaining two families

! !
νµ ντ
, . (5.55)
µ− L
τ− L
This SU(2) symmetry is called weak isospin, its generators are denoted T1 ,
T2 , T3 , and the corresponding quantum numbers of the lepton doublets are
1 1
t= , t3 = ± . (5.56)
2 2
The right-handed parts of the leptons, (e− )R , (µ− )R , (τ − )R , do not have
right-handed neutrino-partners, and are singlets (t = 0) under weak isospin.
The other symmetry group U(1)Y is associated with the weak hypercharge
Y , which is related to the electric charge Q by
y
Q = t3 + . (5.57)
2
ν e− − −
L , µL , τ L e− − −
R , µR , τ R
weak isospin t3 + 12 − 12 0
hypercharge y −1 −1 −2
electric charge Q 0 −1 −1
Gauge fields
The symmetry groups are now gauged by introducing corresponding gauge
fields.
symmetry group generators gauge field field strength
SU(2) Ta = 12 τa Wµa (x) Wµνa
(x)
U(1) Y = y1 Bµ (x) Bµν (x)
The gauge field dynamics is given by the gauge part Lagrangian

1 a aµν 1
Lg = − Wµν W − Bµν B µν . (5.58)
2 4
The fermionic part of the Lagrangian is
! !
µ g′ g νe
Lf =(νe , e− )L iγ ∂µ + i Y Bµ + i τa Wµa
2 2 e− L
!
µ g′ (5.59)
+ e−
R iγ ∂µ + i Y Bµ e−
R
2
+ the same terms for muon and tauon fields.
Higgs field
Finally, we have to add the Higgs field part of the Lagrangian. For the Higgs
field we take a complex SU(2) doublet
! !
φ+ 1 φ1 + iφ2
φ(x) = 0 =√ (5.60)
φ 2 φ3 + iφ4
The Higgs field has weak isospin t = 1/2 and weak hypercharge yφ = 1. The
upper component φ+ has t3 = +1/2, yφ = 1 and therefore charge Q = +1,
whereas the lower component φ0 has t3 = −1/2, yφ = 1 and charge Q = 0.
In matrix form this reads
! !!
Y 1 1 0 1 1 0
Qφ = T3 + φ= + φ
2 2 0 −1 2 0 1
! (5.61)
1 0
= φ.
0 0
The covariant derivative of the Higgs field is
g′ g
Dµ φ = (∂µ + i Bµ + i τa Wµa )φ . (5.62)
2 2
Its Lagrangian is
Lh = (Dµ φ)† (Dµ φ) − µ2 φ† φ − λ(φ† φ)2 , (µ2 < 0, λ > 0). (5.63)
The potential has minima at
† −µ2
φφ= . (5.64)
2λ
The vacuum state is chosen to be

!
1 0 µ2
φ0 = √ , v2 = − . (5.65)
2 v λ
It is chosen such that Qφ0 = 0. This guarantees that the electromagnetic
gauge group U(1), which is generated by Q, is unaffected by the Higgs mech-
anism (it is “unbroken”), so that the photon remains massless. We shall see
this explicitly below.
Without the gauge interactions we would have a spontaneously broken global
symmetry, and there would be
3 massless Goldstone bosons,
1 massive Higgs field.
In the presence of the gauge fields, the fields are transformed into the unitary
gauge as in Sec. 5.2.2, so that we have
!
1 0
φ(x) −→ √ . (5.66)
2 v + ρ(x)
The Lagrangian of the Higgs and gauge fields then becomes
1 1 λ
Lh = (∂µ ρ)2 − m2ρ ρ2 − λvρ3 − ρ4
2 2 4
1 2
+ mW (Wµ1 W µ 1 + Wµ2 W µ 2 )
2 (5.67)
v2
+ (g ′ Bµ − gWµ3 )(g ′ B µ − gW µ 3 )
8
+ interaction terms.
As a result one gets a massive neutral Higgs field ρ with m2ρ = 2λv 2 , and
massive vector fields W 1 and W 2 with mW = 21 gv. The charged bosons W ±
are defined as the complex linear combinations
1
W ± = √ (Wµ1 ∓ iWµ2 ) (5.68)
2
and we have
Wµ1 W µ 1 + Wµ2 W µ 2 = 2Wµ+ W µ − . (5.69)
The remaining fields Bµ and Wµ3 appear mixed in the quadratic terms. They
can be diagonalised by means of
gW 3 − g ′ Bµ
Zµ := √ µ 2 , (5.70)
g + g ′2
g ′ W 3 + gB µ
Aµ := √ µ2 . (5.71)
g + g ′2
The quadratic term then reads
1 2
m Z Zµ Z µ , (5.72)
2
and we see that the masses belonging to the fields Zµ and Aµ are
1 q 2
mZ = v g + g ′2 , mA = 0. (5.73)
2
In this way fields for the massive Z 0 boson and the massless photon emerge.
Defining an angle θW through
g′ mW
tan θW = , or = cos θW , (5.74)
g mZ
the fields can be represented as

! ! !
Zµ cos θW − sin θW Wµ3
= . (5.75)
Aµ sin θW cos θW Bµ
θW is called weak mixing angle or “Weinberg angle”. We also have
g g′
cos θW = q , sin θW = q . (5.76)
g2 + g′2 g2 + g′2
To identify the charges of the fields, we consider the covariant derivative
g′
Dµ = ∂µ + i Y Bµ + igTa Wµa (5.77)
2
and recall
1
Q = T3 + Y. (5.78)
2
With
T± := T1 ± iT2 (5.79)
we have
[Q, T± ] = [T3 , T± ] = ±T± , [Q, Y ] = 0. (5.80)
Then we get the charged field components as parts of
g
gTa Wµa = √ (T+ Wµ+ + T− Wµ− ) + gT3 Wµ3 . (5.81)
2 | {z }
± charged fields
For the fields Wµ3 and Bµ we find

Y Y
gT3 Wµ3 + g ′ Bµ = (g sin θW T3 + g ′ cos θW )Aµ
2 2
Y
+ (g cos θW T3 − g ′ sin θW )Zµ
2
′
gg Y
=q (T3 + )Aµ
g2 + g′ 2 2
(5.82)
q
2 Y
+ g2 + g′2 T3 − sin θW (T3 + ) Zµ
2
gg ′
=q QAµ
g2 + g′2
q
+ g 2 + g ′ 2 T3 − sin2 θW Q Zµ .
The photon field Aµ couples to the charge Q, as it should. From the expres-
sion we can identify the coupling constant of the photon, the electric charge
unit
gg ′
e0 = q = g sin θW = g ′ cos θW . (5.83)
2
g2 + g′
Altogether we can write the fields appearing in the covariant derivative in

the form
Y e0
g ′ Bµ + gTa Wµa = √ (T+ Wµ+ + T− Wµ− )
2 2 sin θW
+ e0 QAµ (5.84)
e0
+ (T3 − sin2 θW Q)Zµ .
sin θW cos θW
From the discussion above it is already clear that the masses of the W and
Z bosons stem from the non-vanishing vacuum value of the Higgs field. To
make it manifest, let us nevertheless show this explicitly again. The covariant
derivative acts on the constant vacuum value of the Higgs field as
ig ig
Dµ φ0 = √ (Wµ+ T+ + Wµ− T− )φ0 + Zµ T3 φ0
2 cos θW
! ! (5.85)
ig + v ig 0
= Wµ − √ Zµ .
2 0 2 2 cos θw v
In the vacuum state the kinetic term of the Higgs field then yields
g2v2 1
(Dµ φ0 )† (Dµ φ0 ) = − (2Wµ+ W µ− + 2
Zµ Z µ ), (5.86)
8 cos θW
and we get the masses of the gauge bosons

1 e0 v mw
mW = gv = , mZ = . (5.87)
2 2 cos θW cos θW
Fermion masses
Under the local gauge symmetry group
SU(2)L ⊗ U(1)Y (5.88)
the right- and left-handed fields eR and eL transform differently. Therefore,

the mass term
mψ̄ψ = m(ψ̄L ψR + ψ̄R ψL ) (5.89)
cannot be invariant, and the bare fermion masses have to be zero. How can
we get fermion masses?
The solution is that the fermion masses come from the Higgs field, like the
masses of the vector bosons. This can be achieved by coupling the Higgs
field to the fermion fields through a Yukawa coupling
!
φ+ −
LY = −Ge (νe , e− )L e + Hermitian conjugate. (5.90)
φ0 R
The same terms are added for muon and tauon fields. These Yukawa terms
are invariant under SU(2)L . They are also invariant under U(1)Y because
the hypercharges add to zero:
(νe , e− )L φ e−
R (5.91)
| {z }
Yφ = 1 +1 −2 = 0
In the unitary gauge ! !

φ+ 1 0
0 =√ (5.92)
φ 2 v + ρ(x)
the Lagrangian becomes
Ge v Ge
LY = − √ e− e− − √ ρ e− e− (5.93)
2
| {z } |
2 {z }
mass remaining Yukawa coupling
The electron mass is

Ge v
me = √ . (5.94)
2
In this way all masses (of W , Z, leptons, quarks) are generated by the Higgs
mechanism.
Reduction to the Fermi theory

As discussed in Sec. 5.1.3, the reduction to the V-A theory at low energies
follows the scheme
κ
κ2 ≪ m2W,Z
where the vector boson propagator is replaced by igµν /m2 . The resulting
effective theory has a four-fermion vertex with coupling constant
GF g2 1
√ = 2
= 2. (5.95)
2 8mW 2v
From the known value of the Fermi coupling GF one derives
√ 1
v = ( 2 GF )− 2 = 246 GeV. (5.96)
From independent measurements the values
sin2 θW = 0.231, (5.97)
mW = 80.42 GeV, (5.98)
mZ = 91.19 GeV (5.99)
have been obtained. Recent experiments indicate the value
mρ ≈ 125 GeV (5.100)
for the Higgs mass.
Quarks
The quarks couple to the gauge fields of SU(2)L ⊗ U)(1)Y in a similar way
as leptons. The left-handed quarks form weak iso-doublets, i.e. t = 1/2,
! ! !
u c t
, , . (5.101)
d′ L
s′ L
b′ L
Their weak hypercharge is y = 1/3 in order to yield the correct electric

charges. The d′ , s′ and b′ components have been denoted with a prime,
because mixing among them will occur, generalising the Cabibbo angle. To
remind you, for two generations of quarks the mixing of the d and s quarks
is
d′ = d cos θC + s sin θC (5.102)
s′ = − d sin θC + s cos θC . (5.103)
The right-handed quarks
uR , d′R , cR , s′R , tR , b′R (5.104)
form weak iso-singlets with t = 0 and y = 4/3 and y = −2/3, respectively.

Quark masses are generated through Yukawa couplings to the Higgs field.
The most general form of the resulting mass terms for the three generations
is
   
uL d′L
Mu  cL  + d¯R , s̄R , b̄R Md s′L  + h.c.
′ ′ ′
   
ūR , c̄R , t̄R (5.105)
tL b′L
with two 3 × 3 mass matrices Mu and Md . The mass matrices cannot be

diagonalised simultaneously and one ends up with a diagonal Mu and
 
md 0 0
Md = V  0 m s 0  V + .
 
(5.106)
0 0 mb
The relation between the mass eigenstates d, s, b and the mixed states d′ , s′ ,
b′ , entering the weak interactions, is therefore
 ′  
d d
 ′  
s  = V s (5.107)
b′ b
with the unitary Cabibbo-Kobayashi-Maskawa (CKM) matrix. In a theory

with only 2 generations one has the Cabibbo mixing
!
cos θc sin θc
V = . (5.108)
− sin θc cos θc
The Glashow-Weinberg-Salam model gives a consistent theory of the weak

and electromagnetic interactions. A detailed study of the interactions de-
scribed by the theory shows
• there are charged currents as in the V-A theory,
• it predicts neutral currents coupling to left-handed and right-handed

fermions: Jµ(neutral) = Jµ3 − sin2 θW Jµ(e.m.) ,
• it contains further interactions of Higgs and gauge boson fields.

Properties of the GWS theory are

• Renormalisability.
Although the Lagrangian does not look renormalisable in the unitary
gauge, it is! The proof (G. ’t Hooft and others) is complicated. The
basic idea is: without unitary gauge the theory is renormalisable, and
it has been shown that this property is independent of the gauge.
• Phenomenologically the theory is extremely successful.
• It provides a unified description of the electromagnetic and weak inter-
actions.
However, because the gauge group is the direct product of SU(2) for
weak isospin and U(1) for electromagnetism, this is not really a full
unification. This motivates to search for a unified theory with a sim-
ple Lie group, which cannot be decomposed into the direct product of
subgroups.
• The theory is free of anomalies.
In field theory an anomaly occurs, when the classical theory has a sym-
metry and an associated conserved Noether current, ∂µ j µ = 0, which
is not conserved in the quantum theory: ∂µ j µ 6= 0. The symmetry is
then broken by a quantum effect. In the GWS theory there are possible
anomalies, but they cancel each other. The condition for the absence
of anomalies is that the electric charges of all left-handed particles and
all right-handed particles in a generation cancel against each other:
X X
Q− Q = 0. (5.109)
L R
Here the quark charges are counted with a factor of 3 due to the three
colours. The condition is satisfied for leptons and quarks within each
generation. As a consequence the number of colours is related to the
cancellation of anomalies.
• The Glashow-Iliopoulos-Maiani (GIM) mechanism
explains the suppression of flavour-changing neutral currents, if there
is an even number Nf of flavours. It was suggested at a time, when
only three quark flavours had been known, and led to the prediction of
the charm quark.
Adding the strong interactions of quarks via QCD leads to the complete
Standard Model. The total gauge group then is
SU(3)c ⊗ SU(2)L ⊗ U(1)Y . (5.110)

Gauge Theories

Uploaded by

Copyright:

Available Formats

Gauge Theories

Uploaded by

Document Information

Original Description:

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Gauge Theories

Uploaded by

Copyright:

Available Formats

Gauge Theories

of the Strong and Electroweak Interactions

Is nature obeying fundamental laws? Does a comprehensive description of

1.4 Field Quantisation . . . . . . . . . . . . . . . . . . . . . . . . 46

2 Quantum Electrodynamics (QED) 69

3 Non-abelian Gauge Theory 74

4 Quantum Chromodynamics (QCD) 87

5 Electroweak Theory 111

5.1.2 Parity violation . . . . . . . . . . . . . . . . . . . . . . 111

From these properties the following classiﬁcation arose.

Baryons half integer spin, baryon number = ±1

Quark model of hadrons (Gell-Mann, SU(3), eightfold way)

Some baryons some mesons

c ≈ 1300 2/3 1/3

t ≈ 174 000 2/3 1/3

Table 1: The 3 generations of elementary particles

Fig. 1 and Fig. 2 show multiplets of mesons and baryons arranged in 3-

Figure 1: SU(4) multiplets of mesons; 16-plets of pseudoscalar (a) and vector

All hadrons are colourless combinations of quarks. This phenomenon is called

There is a characteristic feature for each single generation of leptons and

a) Electromagnetic interactions. They apply to electrically charged par-

Forces are mediated by the exchange of bosons.

a) Quantum Electrodynamics originated in 1927, when in an appendix to

Here is an example of a Feynman diagram for the scattering of two

The vertex stands for a number, in QED this is α ≈ 1/137. The

Improvements of the theory of β-decay in nucleons were made by the

c) Strong interactions between quarks are described by Quantum Chro-

Feynman diagrams with quarks and gluons

The Standard Model

The principles of the Standard Model are:

The Higgs mechanism is due to P. Anderson, F. Englert, R. Brout, P. Higgs,

A group is called semisimple, if it is the direct product of simple groups. A group is

simple, if it has no normal subgroups besides the trivial ones.

1.2 Relativistic Field Equations

QFT is a quantum theory of many particles. In this lecture we consider the

• Dirac equation for spin 1/2 particles,

• Maxwell’s equations for massless spin 1 particles.

1.2.1 Klein-Gordon equation

leads to the non-relativistic Schrödinger equation

insert ~ and/or c to get the right dimension.

Solution of the Klein-Gordon equation

The problem with negative frequencies

1.2.2 Dirac equation

i∂t ψ = Hψ. (1.19)

To be manifestly covariant, it has to be of ﬁrst order in the spatial derivatives,

Consider a plane wave ψ = eipx , for which one should have

This wave satisﬁes the Dirac equation only if

By convention one uses the Dirac matrices γ µ :

The Hamiltonian can be written

and multiplying with γ 0 we get

This is the Dirac equation, reading

(iγ µ ∂µ − m)ψ(x) = 0. (1.28)

Equivalent notations of Dirac’s equation are

(γ µ Pµ − m)ψ(x) = (p/ − m)ψ(x) = 0.

The algebra of the γ-symbols is

Solutions of Dirac’s equation will be given by spinor wavefunctions or ﬁelds

These are made out of two kinds of plane waves, given by

ψ(x) = u(k)e−ikx , k 0 = ωk > 0 (1.32)

with 2 independent spinors u(r) (k), r = 1, 2, and