Quantum Field Theory I

Lecture notes by

Jan Louis

II. Institut f
ur Theoretische Physik der Universitat Hamburg, Luruper Chaussee 149,
22761 Hamburg, Germany

ABSTRACT
The lecture notes grew out of a course given at the University of Hamburg in the summer
term 2007 and the winter term 2010/11. A rst version of these lecture notes were written
by Andreas Bick, Lukas Buhne, Karl-Philip Gemmer, Michael Grefe, Jasper Hasenkamp,
Jannes Heinze, Sebastian Jakobs, Thomas Kecker, Tim Ludwig, Ole Niekerken, Alexander Peine, Christoph Piefke, Sebastian Richter, Michael Salz, Bjorn Sarrazin, Hendrik
Spahr, Lars von der Wense.

April 5, 2011

Contents
1 Lecture 1: Special Relativity

1.1

Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

1.2

Covariance of a physical law . . . . . . . . . . . . . . . . . . . . . . . . .

1.3

Lorentz Transformations . . . . . . . . . . . . . . . . . . . . . . . . . . .

1.4

Relativistic wave-equations . . . . . . . . . . . . . . . . . . . . . . . . . .

2 Lecture 2: Classical Scalar Field Theory

2.1

Lagrangian and Hamiltonian formalism . . . . . . . . . . . . . . . . . . .

2.2

Real Scalar Field . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

10

2.3

Noether-Theorem . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

11

3 Lecture 3: Canonical quantization of a scalar field

13

3.1

Solution of Klein-Gordon equation . . . . . . . . . . . . . . . . . . . . . .

13

3.2

Quantizing the scalar eld . . . . . . . . . . . . . . . . . . . . . . . . . .

14

3.3

Fock-Space

15

. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

4 Lecture 4: The Feynman propagator for a scalar field

16

5 Lecture 5: The Dirac equation

20

5.1

-matrices . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

20

5.2

The Dirac equation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

21

5.3

Weyl equation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

22

5.4

Solution of the Dirac equation . . . . . . . . . . . . . . . . . . . . . . . .

22

6 Lecture 6: Covariance of the Dirac equation

24

7 Lecture 7: Quantization of the Dirac field

26

7.1

Dirac propagator . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

8 Lecture 8: Discrete Symmetries

27
29

8.1

Parity . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

29

8.2

Time-reversal . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

29

8.3

Charge conjugation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

30

8.4

Fermion bilinears . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

30

9 Lecture 9: Perturbation theory

31

10 Lecture 10: Wick Theorem

36

11 Lecture 11: Feynman diagrams

39

12 Lecture 12: S-Matrix

41

13 Lecture 13: Feynman rules for QED

43

13.1 The Lagrangian of QED and its gauge invariance . . . . . . . . . . . . .

43

13.2 Quantization of the photon eld . . . . . . . . . . . . . . . . . . . . . . .

44

13.3 The Feynman rules for QED . . . . . . . . . . . . . . . . . . . . . . . . .

44

14 Lecture 14: e+ e +

45

15 Lecture 15: Compton scattering e e

48

15.1 Sum over photon polarisation . . . . . . . . . . . . . . . . . . . . . . . .

48

15.2 Unpolarized Compton scattering . . . . . . . . . . . . . . . . . . . . . . .

49

16 Lecture 16: Dimensional regularization

51

17 Lecture 17: Field strength renormalization

54

17.1 The Kallen-Lehmann-representation of the propagator

17.2 Application: the electron self energy

. . . . . . . . . .

54

. . . . . . . . . . . . . . . . . . . .

55

18 Lecture 18: Renormalization of the electric charge

18.1 General properties . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
18.2 Computation of

. . . . . . . . . . . . . . . . . . . . . . . . . . . . .

19 Lecture 19: The electron vertex

57
57
58
60

19.1 General considerations . . . . . . . . . . . . . . . . . . . . . . . . . . . .

60

19.2 Leading order results . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

61

20 Lecture 20: Renormalization of QED

63

20.1 Supercial degree of divergence of a Feynman diagram . . . . . . . . . .

63

20.2 Application to QED . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

64

21 Lecture 21: Renormalized pertubation theory of QED

66

22 Lecture 22: IR-divergences

69

23 Lecture 23: The LSZ-reduction formalism

72

Lecture 1: Special Relativity

1.1

Introduction

The topic of this course is an introduction to the quantum theory of relativistically invariant eld theories such as scalar eld theories and Quantum-Electrodynamics (QED).
Applications to Particle Physics such as the computation of scattering amplitudes are
also discussed. There are many excellent textbooks on the subjects, for example [2, 6, 1,
3, 4, 7]. This course closely follows [2].
Let us start by recalling various aspects of Special Relativity.

1.2

Covariance of a physical law

A physical phenomenon or process should not depend on the choice of coordinate system which is used to describe it. Therefore the physical laws should be identical in all
coordinate systems. One says that a physical law has to be covariant1 under coordinate
transformations.
As a rst example let us consider Newtons Mechanics which is governed by the equation of motion
d2~x
F~ = m 2 m~x .
(1.1)
dt
P
P
Let us choose Cartesian coordinates F~ = 3i=1 F i~ei , ~x = 3i=1 xi~ei , where ~ei is an orthonormal Cartesian basis of R3 . (1.1) is invariant under the Galilei-transformations
xi xi = xi + v0i t + xi0 ,
t t = t + t0 ,

(1.2)

where v0i , xi0 , t0 are constants. Indeed we can check

d
dt d
d
d
= =
,
dt
dt
dt dt
dt

~x ~x = ~x ,

(1.3)

which shows the invariance of (1.1).

The Newton equation (1.1) is also covariant under rotations which can be expressed
as
i

x x =

3
X

Dki xk .

(1.4)

k=1

The Dki are the constant matrix elements of the 3 3 rotation matrix D which leaves
the length of a vector invariant, i.e.
X
X
X
xi xi =
Dki xk Dli xl =
xj xj .
(1.5)
i

i,k,l

In Mathematics this is often called equivariant.

This implies2
X

Dki Dli = kl ,

DT D = 1 .

or in matrix notation,

(1.6)

Under rotations not only xi transforms as in (1.4) but all vectors (including F i ) do
so that in the rotated coordinate system the Newton equation (1.1) reads
X


Dki F k m
xk = 0 .
(1.7)
F i m
xi =
k

We see that it is not invariant but both sides rotate identically (as vectors) so that
the solutions are unaltered. One therefore calls the Newton equation covariant under
rotations.
In Newtons mechanics all coordinate systems which are related by Galilei-transformations (including rotations) are called inertial reference frames and in these coordinate
systems the Newton equation (or rather their solutions) are identical. (This is sometimes
called the Galilei principle.)
Historically the development of Special Relativity grew out of the observation that the
Maxwell-equations governing Electrodynamics are not covariant under Galilei-transformations. The reason being that the speed of light c was shown to be a constant of
nature in the experiments by Michelson and Moreley. The Maxwell-equations are instead
covariant under Lorentz-transformations and Einstein postulated c being constant and
consequently covariance with respect to Lorentz-transformations for any physical theory.
(This is sometimes called the Einstein principle.) This led to the theory of Special
Relativity.

1.3

Lorentz Transformations

For a particle moving with the speed of light we have

c2 =

X dxi dxi
dt dt

X dxi dxi
i

dt dt

(1.8)

where xi , t are the space and time coordinates of a system which moves with velocity
~v0 with respect to the unprimed coordinate system. Expressed innitesimally, the lineelement
X
X
ds2 = c2 dt2
dxi dxi = c2 dt2
dxi dxi
(1.9)
i

has to be constant in all coordinate systems (with ds = 0 for light rays). It turns out that
Lorentz transformations non-trivially rotate space and time coordinates into each other
and are best described as rotations of a four-dimensional pseudo-Euclidean Minkowskian
space-time M4 . The coordinates of M4 are the 4-tupel (denoting a space-time point)
x = (x0 , xi ) = (x0 , x1 , x2 , x3 ) ,
2

= 0, . . . , 3,

i = 1, 2, 3 ,

(1.10)

Due to the six equations (1.6) the matrix D depends on 9-6=3 independent parameters which are
the three rotation angles. The matrices D form the Group O(3). We will return to the group-theoretical
description in QFT II.

where x0 = ct. Furthermore, on M4 distances are measured by the line-element(1.9)

which expressed in terms of x reads
2

0 2

ds = (dx )

dx dx =

3 X
3
X

dx dx .

(1.11)

=0 =0

is the (pseudo-euclidean) metric on M4 given by

1 0
0
0
0 1 0
0

=
0 0 1 0 .
0 0
0 1
P
The inverse metric is denoted by and obeys = .

(1.12)

Lorentz transformations can now be dened as rotation in M4 which leave its line
element ds2 dened in (1.11) invariant. In the linear coordinate transformations in M4
X

x x =
x
(1.13)

which leave ds2 invariant, i.e. which obey

X
XX
X

dx dx =
dx dx =
dx dx
,

, ,

(1.14)

are Lorentz transformations. (1.14) imply the following ten conditions for the matrix
elements
X
=

T = ,
(1.15)
,

where the latter is the matrix form of the former.3

The matrix has 16 10 = 6 independent matrix elements which turn out to be 3

rotation angles and 3 boost velocities. The rotation angles can be easily identied by
noting that a transformation of the form (1.13) with


1 0
=
(1.16)
0 D
solve (1.15) and precisely correspond to the rotations (1.4). As we already discussed
above D depends on three rotation angles.
In order to discuss the Lorentz boost let us rst look at Lorentz transformations which
only transform t (or rather x0 ) and x1 non-trivially. In this case takes the form

0
0 0 1 0 0
1 0 1 1 0 0
,
(1.17)
=
0
0 1 0
0
0 0 1
3

The corresponding group is the orthogonal group O(1,3), where the numbers indicates the signature
of the underlying metric (1 plus, 3 minuses).

corresponding to the transformations

(ct ) = 00 ct + 0 1 x ,
x = 10 ct + 1 1 x ,

(1.18)

y = y ,
z = z .
Inserted into (1.15) we obtain
(00 )2 (1 0 )2 =

1,

(01 )2 (1 1 )2 = 1 ,

0 0 0 1 1 0 1 1 =
which are solved by

0 0 =

(1.19)

0,

cosh = 1 1 ,

(1.20)

1 0 = sinh = 0 1 ,

or in other words

cosh sinh
sinh cosh
=

0
0
0
0

0
0
1
0

0
0
.
0
1

(1.21)

The parameter is called the rapidity. It is related to the relative velocity v01 v of the
two coordinate systems. Let us choose the primed coordinate system to be at rest while
the unprimed system moves with velocity v in the x1 -direction. Thus x = 1 0 ct+1 1 vt =
0 together with (1.20)implies
v
1 0 = 1 1
c

sinh =

v
cosh
c

tanh =

v
c

(1.22)

Dierently expressed we have

1
cosh = q
1

or in matrix form

,
v 2

sinh =

1
v
q
c
1

v 2
c

(1.23)

0 0

0 0
.
=
0
0
1 0
0
0
0 1

Analogously one can determine

the general solution of (1.15) for the unprimed system
P i
moving with velocity ~v = i v ~ei to be
0 0 = ,

ji

0 j =

vj
= j 0
c

vi vj
+ ( 1) 2 ,
~v
6

1
.
p
1 ~v 2 /c2

(1.24)

On M4 one denes contravariant 4-vectors as

a = (a0 , ai ) ,
with a transformation law

a a =

(1.25)

a .

(1.26)

A covariant 4-vector has a lower index and is dened as

X
a =
a = (a0 , ai ) .

(1.27)

It transforms inversely as
a a =
where

1T
a ,

(1.28)

1T
= . The scalar product of two 4-vectors is dened by

X
X
X
a b =
a b = a0 b0
ai bi

(1.29)

which is indeed a Lorentz scalar.

The partial derivatives can also be formulated as a 4-vector of M4 . One denes

 

1
1

, =
,
,
,
,
=
x
c t
c t x1 x2 x3
(1.30)


X
1

=
=
, .
c
t

In this notation the wave operator  is given by

=

1.4

1 2
.
c2 t2

(1.31)

Relativistic wave-equations

Let us rst recall the denition of the proper time which is the time measured in the
rest-frame of a particle. Here one has
ds2 = c2 d 2 = c2 dt2 d~x2 = (c2 ~v 2 )dt2 ,

(1.32)

where the second equation relates it to the time measured by a moving (with velocity ~v )
observer. Thus they are related by
s
 2
Z t2
~v
d = 1
1 dt
(1.33)
dt ,
=
c
t1
The 4-momentum of a particle is dened as
p = m

dx
dx
= m
,
d
dt
7

(1.34)

where m is the rest-mass, i.e. the mass measured in the rest frame. The three-momentum
thus is given by
~p = m ~v = m~v + O(v 3 /c2 ) for |~v| c ,
(1.35)
cp0 is identied with the energy

cp0 E = mc2 = mc2 + 12 mv 2

for |~v| c

(1.36)

The term mc2 is known as the rest energy of a particle.

Let us compute the Lorentz scalar
X
c2
p p = E 2 c2 ~p2 = m2 c4

(1.37)

This is known as the relativistic energy-momentum relation


p
~2
p
mc2 + 2m
+ . . . for |~p| mc
2
2
2
E = c m c + p~
c|~p| + . . .
for |~p| mc

(1.38)

Inserting the quantum-theoretical correspondences

~p

~
,
i

E i~

(1.39)

into (1.37) using (1.31) yields the dierential operator



mc2
2
2 4
2 2
=0,
E c ~p m c ~ 2 + m c = c ~  + 2
t
~
2

2 2

2 4

(1.40)

Therefore a relativistically invariant wave equation for a scalar eld is given by


 mc 2 
+
=0.
(1.41)
~
This is the massive Klein-Gordon-equation which we study in the next sections. In the
following we also use natural units where we set ~ = c = 1. In this convention the
Klein-Gordon-equation reads


 + m2 = 0 .
(1.42)
Note that for m = 0 we obtain the wave equation  = 0.

Henceforth we also use the summation convention which says that indices which
P occur
twice are automatically summed over (unless stated otherwise). For example p p
will simply be denoted by p p .

Lecture 2: Classical Scalar Field Theory

2.1

Lagrangian and Hamiltonian formalism

A scalar eld (x ) can be viewed as a map from M4 into R if is real or C if is

complex. Thus associated with at every space-time point x is number the value of at
that point. Examples are the scalar potential in Electrodynamics or the Higgs-eld in
Particle Physics.4
As for mechanical systems with a nite number of degrees of freedom it is useful
to develop a Lagrangian and Hamiltonian formalism for elds (systems which have an
innite number of degrees of freedom). Let us be slightly more general and discuss this
formalism immediately for a collection of N scalar elds r (x), r = 1, . . . , N. (From now
on we indicate the dependence of a eld on x simply by x, i.e. we use (x ) (x).)
The Euler-Lagrange equations follow from a stationary condition of an action S given
by
Z
Z
S = dt L =
d4 x L(r , r ) .
(2.1)
V

Here L(r , r ) is a Lagrangian density which depends on the elds and their rst
derivatives. d4 x is the Lorentz invariant volume element and V is the volume of M4 .
Sometimes
R 3 it is useful to consider the Lagrangian L which is related to its density via
L = d x L.

As in classical mechanics one postulate an action principle, i.e. one demands that
the physical system on its trajectory extremizes S. Or in other words one demands that
uctuations of (r r + r ) which vary the action as S S + S are such that
S = 0 holds on the (classical) trajectory subject to the boundary condition r |V = 0,
where V denotes the boundary of V . Applied to (2.1) one therefore obtains
!
Z
N
X
X
L
L
r +
d4 x
0 = S =
( r )
(2.2)
r
( r )
V

r=1
where (in complete analogy to classical mechanics) the elds r and their derivatives
r are treated as independent variables. Using ( r ) = (r ) the variation (2.2)
can be rewritten as



 
Z
N 
X
L
L
L
4
r ,
(2.3)
0=
dx
r +
r
r
( r )
( r )
V
r=1
where we now also used the summation convention introduced at the end of the previous
section. The integral of the second term vanishes by virtue of the Gauss law and the
constraint that r is zero at the boundary

 Z
Z
L
L
4
d x
r =
dF
r = 0 ,
(2.4)
( r )
( r )
V
V
~ x, t), B(~
~ x, t) which associate with
From Electrodynamics you are also familiar with vector fields E(~
with at every space-time point a vector (or rather two vectors). However, for pedagogical reasons we
first discuss in this section a scalar field .
4

where dF is the 3-dimensional normal surface element on V . So from (2.3) we are left
with

Z
N 
X
L
L
4
0=
d x
r
(2.5)

r
( r )
V
r=1

for arbitrary and independent r . Therefore (2.5) can be fullled if and only if the
bracket itself vanishes. This results in the Euler-Lagrange equations for (scalar) elds
L
L

=0
r
( r )

for all r = 1, . . . , N .

(2.6)

These are the exact analogues of the Euler-Lagrange equations in classical mechanics
L
dtd (L
= 0 for the generalized coordinates qr .
qr
qr )

In order to introduce the Hamiltonian formalism for elds one rst denes the canonically conjugated momentum r of r (strictly speaking it is a momentum density)
r :=

L
,
r

with

r 0 r .

The Hamiltonian of a classical eld theory is then given by

Z
Z
N 

X
3
3
H(r , r ) d x H(r , r ) := d x
r r L .

(2.7)

(2.8)

r=1

H is called the Hamiltonian density. Note that L is a Lorentz scalar while H is not!

2.2

Real Scalar Field

We take the Lagrangian density of a real massive scalar eld to be

L(, ) = 21 m2 2 ,

(2.9)

L
= m2 ,

(2.10)

where we have chosen natural units ~ = c = 1. (This convention will be used henceforth.) From the Lagrangian (2.9) we can compute
L
= .
( )

Inserted into the Euler-Lagrange equation (2.6) yields

( + m2 ) = 0

(2.11)

which we immediately recognize as the massive Klein-Gordon equation (1.42) in natural

units. Therefore, (2.9) indeed is the Lagrangian whose Euler-Lagrange equation is the
Klein-Gordon equation.
In order to compute the corresponding Hamiltonian we rst determine the conjugate
momentum from (2.7) to be
L
=
= .
(2.12)

Inserted into (2.8) we arrive at




2
2
2 2
2
2 2
1
1

~
~
~
~
H = L = 2 m = 2 + + m . (2.13)
10

2.3

Noether-Theorem

The Noether-Theorem states that associated with any continues symmetry there is a
conserved (Noether) charge Q. This also applies to the eld theories discussed here. Let
us derive the theorem.
Consider an (arbitrary) innitesimal continues symmetry which transforms the scalar
elds as
r r = r + r ,
(2.14)

where is the innitesimal parameter of the transformation and is the deformation

of the eld conguration. Being a symmetry implies that the action is unchanged under
the transformation (2.14), i.e.
S(r ) = S(r ) .
(2.15)
This implies that Lagrangian density L changes at most by a total divergence or in other
words
L(, ) L( , ) = L(, ) + J ,
(2.16)
where J is constrained to vanish at the boundary V . Inserting the transformation
(2.14) into L and Taylor-expanding to rst order in yields
L(r , r ) = L(r + r , (r + r ))
= L(, ) +
= L(, ) +

X L
X L
r +
r

)
r

r
r
r

X  L
r

L

r +
r
( r )
|
{z
}
=0, EulerLagrange

X
r

= L(, ) + J .

L
r
( r )

(2.17)

If one now denes the Noether current as

X L
r J ,
j
(

)
r
r

(2.18)

(2.17) implies the continuity equation

j = 0 .

(2.19)

The associated Noether charge Q

Q

d3 x j 0 ,

(2.20)

is then conserved, i.e.

dQ
=
dt

d x 0 j =
V

~ ~j =
d x
3

dF~ ~j = 0 ,

(2.21)

whenever ~j|V = 0 holds. Note that in the derivation of the conservation law we used the
current conservation (2.19) which only holds if the Euler-Lagrange equations are satised.
11

Now we can state the Noether-Theorem for eld theories which says that for every
continuous symmetry there is a Noether current j satisfying j = 0 and an associated
conserved Noether charge Q satisfying Q = 0.
As an example consider again the real scalar eld with the Lagrangian (2.9)
L = 12 12 m2 2 .
For m = 0 it has the shift symmetry
= + ,

(2.22)

where is a constant parameter. Comparing with (2.14) we conclude = 1 and we

cvan indeed check that (2.22) is a symmetry
L( ) = 21 = 21 ( + ) ( + ) = L() .

(2.23)

Comparing with (2.16) we see that L is invariant under the transformation (2.22) with
J = 0. Using (2.18) we compute the Noether current to be
j = ,

(2.24)

and check its conservation by explicitly using the Euler-Lagrange equation (i.e. the massless wave equation  = 0)
j = =  = 0 .

(2.25)

As a further application let us considers constant space-time translations of the form

x x = x + a

(2.26)

for four constant and innitesimal parameters a . The scalar eld then changes according to
(x) (x ) = (x + a ) = (x ) + a (x ) + O(a2 ) .
(2.27)
In the previous terminology we thus have four independent transformation (three space
translation and one time translation) so that = a . The corresponding Noether
current is the energy-momentum tensor
T =

L
L .
( )

(2.28)

It obeys the conservation law

T = 0 ,
with the four Noether charges being
Z
H = d3 x T 00 ,

p =

12

d3 x T 0i ,

(2.29)

i = 1, 2, 3 .

(2.30)

Lecture 3: Canonical quantization of a scalar field

Before we quantize the scalar eld let us determine the solution of the Klein-Gordon
equation (2.11)
( + m2 )(x) = 0 .
(3.1)

3.1

Solution of Klein-Gordon equation

This is most easily done with the help of a Fourier-Ansatz

Z
d4 p
(x) =
(p) eip.x ,
(2)4
where we abbreviate p x p x . Inserted into (3.1) we arrive at
Z
d4 p
2
(p)(m2 p2 ) eipx = 0 .
( + m )(x) =
(2)4

(3.2)

(3.3)

This is solved by either setting (p)

= 0 or p2 = m2 with (p)
arbitrary. Or in other
words
ap~
ap~
0

(p Ep~ ) + p
(p0 + Ep~ ) ,
(3.4)
(p) = p
2Ep~
2Ep~
q
where we abbreviate Ep~ |~p |2 + m2 and the coecients in front of the -function are
chosen for later convenience. Inserted into (3.2) we obtain the solution
Z


d3 p

ipx
ipx

1
ap~ e
+ ap~ e
,
(3.5)
(x) =
0
(2)3 2Ep~
p =Ep~
where in the second term we have replaced p~ ~p since the integration region is
symmetric.
The conjugate momentum is computed to be
r 
Z

Ep~

d3 p
ipx
ipx
.
e
a
e

a
(x) = 0 (x) = i
p
~
p
~
x
(2)3
2
(3.5) can be inverted (see problem 2.2) to give
Z
1
p
ap~ =
d3 x eipx (Ep~ (x) + i (x))
2Ep~
Z
1

d3 x eipx (Ep~ (x) i (x)) .

ap~ = p
2Ep~

ap~ and ap~ are time independent, i.e. t ap~ = t ap~ = 0 (see problem 2.2).

13

(3.6)

(3.7)

3.2

Quantizing the scalar field

In canonical quantization (x) and its conjugate momentum are replaced by Hermitian
operators

(x) (x),

(x) = (x);

(x)
(x),

(x) =
(x) .

with canonical (equal-time) commutation relations

h
i

(~x, t),
(~y , t) = i (3) (~x ~y ) ,
h
i
h
i
x, t), (~
y , t) .

(~x, t),
(~y , t) = 0 = (~

(3.8)

(3.9)

Inserting (3.5) and (3.6) we see that also ap~ and ap~ have to be promoted to operators
ap~ a
p~ ,

ap~ a
p~ ,

(3.10)

with commutation relations which follow from (3.9). Using (3.7) and (3.9) one computes
Z
Z

h
i
i
h
h
i
i

d x d3 y ei~p ~x ei~p y~ Ep~

(x), (y)
a
p~, a
p~ = p
Ep~ (x),

(y)
2 Ep~ Ep~
|
{z
}
{z
}
|
i(3) (~
y ~
x)

i
= p
2 Ep~ Ep~

d3 x ei(~p~p

)~
x

i(3) (~
x~
y)

(iEp~ iEp~ ) = (2)3 (3) (~p ~p ) ,

(3.11)
where we evaluated the right hand side at x = y = 0 since we already know that ap~ is
time-independent. Similarly one shows that the a
p~ commute so that altogether one has


a
p~ , a
p~ = (2)3 (3) (~p p~ ) ,
(3.12)


 
a
p~ , a
p~ = 0 = a
p~ , a
p~ .
0

We thus see that a

p~ and a
p~ provide an innite set of harmonic oscillator creation and
annihilation operators for each value of p~
in terms of ap~ , a
We can also express the Hamiltonian H
p~ . Using (2.8), (3.5) and (3.7)
one obtains
!
Z
Z
i
h
3
d
p
= d3 x H
=
.
(3.13)
Ep~ a
p~ a
p~ + 12 a
p~ , a
p~
H
(2)3
| {z }
(2 3 )(3) (0)

(For the explicit computation see problem 2.3.) We see that it is divergent due to the
summation of an innite number of harmonic oscillator ground state energies. Since only
energy dierences can be measured this term is neglected and one has5
Z
d3 p
=
Ep~ a
p~ a
p~ .
(3.14)
H
3
(2)
5

We will return to this issue later.

14

One further has

h
i Z

H, a
p~ =
h
i Z

H, a
p~ =

3.3

i
h
d3 p

p~, a
p~ = Ep~ ap~ ,
Ep~ a
p~ a
3
(2)
| {z }
(2)3 (~
p~
p )

h
i
d3 p

E
a

,
a

p~ .
a
p~ = Ep~ a
p
~
p
~
p
~
(2)3
| {z }

(3.15)

(2)3 (~
p~
p )

Fock-Space

Since we have an innite set of harmonic oscillators we can construct the space of states
algebraically. One rst denes a vacuum state |0i by demanding that it is annihilated
by all ap~ 6
ap~ |0i = 0
~p .
(3.16)

Then n-particle states are constructed by acting with the creation operators ap~ on |0i
p
one-particle-state:
|~pi := 2Ep~ ap~ |0i .
(3.17)
p
N-particle-state:
|~p1 . . . p~N i := 2Ep~1 . . . 2Ep~N ap~1 . . . ap~N |0i .

The corresponding energies are

H |~pi =



2Ep~ H, ap~ |0i = Ep~ |~pi ,

(3.18)

H |~p1 . . . p~N i = . . . = (Ep~1 + . . . + Ep~N )|~p1 . . . ~pN i .

In order to compute the momentum of the states we rst determine P i from (2.30)
to be (for details see problem 2.3)
Z
d3 p i
i
P =
p ap~ ap~ .
(3.19)
(2)3
Acting on a one-particle state yields
Z 3
Z 3
i
h
dp
dp

i
i
i
i

a
|0i
=
p
a
a
p
a
,
a
a
P ap~ |0i =
~
p
~
~ |0i . (3.20)
p
~
p
~ p
p
~
p
~ |0i = p ap
(2)3
(2)3
| {z }
(2)3 (~
p~
p )

Thus, the state |~pi is eigenstate of the momentum operator P i with eigenvalue p i .

Let us compute
Z
Z

d3 p
1 
1
d3 p
ipx
ipx
p
p
|0i =
ap~ e
+ ap~ e
(x) |0i =
eipx ap~ |0i .
3
3
(2)
(2)
2Ep~
2Ep~

This corresponds to the superposition of a single particle momentum eigenstate and thus
corresponds to a particle at position x .
Finally we note that (x) and (x) obey the Heisenberg equations
i t (x) = [(x), H] ,

i t (x) = [(x), H] .

(For the computation see problem 2.4.)

6

We drop the henceforth from all operators.

15

(3.21)

Lecture 4: The Feynman propagator for a scalar

field

Before we turn to the Dirac equation in the next few lectures let us already prepare
for treating an interacting scalar eld and rst solve the inhomogeneous Klein-Gordon
equation with the help of a Greens function.
Let rst recall that an inhomogeneous partial dierential equation can solved with
the help of a Greens function. Let Dx be an arbitrary dierential operator which differentiates with respect to x, (x) the eld under consideration and f (x) an arbitrary
source function. (x) is taken to obey the dierential equation
Dx (x) = f (x) .

(4.1)

The Greens function G(x y) of Dx is dened to be the solution of

Dx G(x y) = i(x y) .
In terms of G(x y) the general solution of (4.1) is given by
Z
(x) = i dy G(x y)f (y) + hom (x) ,

(4.2)

(4.3)

where hom (x) is the homogeneous solution satisfying Dx hom (x) = 0. Indeed inserting
(4.3) into (4.1) we can check
Z
Z
Dx (x) = i dy Dx G(x y) f (y) + Dx hom (x) = dy (x y)f (y) = f (x) . (4.4)
|
| {z }
{z
}
0

i(xy)

In the following we want to determine the Greens function G(x y) for the KleinGordon-operator (x + m2 ), i.e. solve
(x + m2 )G(x y) = i (4) (x y)
As in the last lecture this is most easily done by a Fourier Ansatz
Z
d4 p ip(xy)
e
G(p) .
G(x y) =
(2)4

(4.5)

(4.6)

Inserted into (4.5) yields

Z

d4 p
G(p) (m2 p2 )eip(xy)
(2)4
Z
d4 p ip(xy)
!
= i
e
= i (4) (x y).
(2)4

(x + m )G(x y) =

(4.7)

This implies
=
G

i
,
2
p m2

and

G(x y) =
16

i
d4 p ip(xy)
e
.
4
2
(2)
p m2

(4.8)

The representation (4.8) for G(x y) can be further simplied by performing the
integral in p0 with the help of the residue formula. Let us now see this in more detail
by rst noting that
q the integral in (4.8) is not well dened at the poles of the integrand

p0 = Ep~ = |~p |2 + m2 . Thus we have to give a prescription for how to treat these
poles. The introduces an ambiguity into G(x y) and, as we will see shortly, leads to
four dierent Gs.

If we analytically continue the integral into the complex p0 plane, we can integrate
around the singularities. This can be done in four dierent ways
1. choose Imp0 > 0 to avoid the singularities,
2. choose Imp0 < 0 to avoid the singularities ,
3. avoid p0 = Ep~ by choosing Imp0 < 0 near this point and Imp0 > 0 near p0 = Ep~ ,
4. choose the opposite path of 3.
The strategy now is to close the path in the complex p0 plane and use the residue
formula which states that the value of the integral
I
1
f (z)dz
(4.9)
c1 =
2i
is determined by the coecient of the single pole z0 in the Laurent-series of f (z), i.e.
f (z) =

n=

cn (z z0 )n = . . . +

c1 (z0 )
+ ... .
z z0

(4.10)

Let us see how this can be done for the four dierent path given above.
0

1. For x0 > y 0 the exponent eip (x y ) does not contribute in the integral for p0
i. Therefore we can close the contour in the lower half-plane and both poles
are enclosed. On the other hand, for x0 < y 0 we have to close the contour in the
upper half-plane which then includes no pole and the integral vanishes. Thus we
are left to compute
Z
Z
d3 p
dp0
i2
ip0 (x0 y 0 ) i~
G(x y) =
e
e p(~x~y)
(2)3
2i (p0 + Ep~ )(p0 Ep~ )
Z
d3 p i~p(~x~y)
2
(4.11)
= i
e
[c1 (p0 = Ep~) + c1 (p0 = Ep~ )]
(2)3
Z
d3 p
1 i~p(~x~y) iEp~ (x0 y0 )
=
e
[e
eiEp~ (x0 y0 ) ] ,
3
(2) 2Ep~
where the minus sign in the second line comes from integrating counter-clockwise
in the p0 plane while in (4.9) the integration is clockwise. In the third line we used
c1 (p0 = Ep~ ) =

eiEp~ (x0 y0 )
,
2Ep~

c1 (p0 = Ep~ ) =
17

eiEp~ (x0 y0 )
.
2Ep~

(4.12)

The nal results for case 1. is called the retarded Greens function and can be written
more succinctly as
Z
i
 ip(xy) 
d3 p
1 h ip(xy) 
0
0
Gret = (x y )
. (4.13)
e

e
p0 =E~
p
p0 =E~
p
(2)3 2Ep~
2. In this case only x0 < y 0 contributes where the path is closed in the upper halfplane. One computes analogously the advanced Greens function
Z
i
 ip(xy) 
d3 p
1 h ip(xy) 
0
0
Gav = (y x )
. (4.14)
e

e
p0 =E~
p
p0 =E~
p
(2)3 2Ep~
3. In this case both contours contribute. For x0 > y 0 the contour is closed in the lower
half-plane and only the pole at p0 = Ep~ contributes, i.e.
Z
1  ip(xy) 
d3 p
e
(4.15)
G(x y) =
p0 =Ep~
(2)3 2Ep~
For x0 < y 0 the contour is closed in the upper half-plane and only the pole at
p0 = E~p contributes, i.e.
Z
d3 p
1  ip(xy) 
G(x y) =
e
(4.16)
p0 =Ep~
3
(2) 2Ep~

Together they form the Feynman-propagator

Z
1  ip(xy) 
d3 p
0
0
e
GF = (x y )
p0 =E~
p
(2)3 2Ep~
Z
d3 p
1  ip(xy) 
0
0
+ (y x )
e
.
p0 =E~
p
(2)3 2Ep~

(4.17)

4. In this case only the sign changes compared to case 3.

The physical meaning of these Greens functions becomes more transparent if we
compute the amplitude for a particle to propagate from y to x which is given by
D(x y) := h0|(x)(y)|0i. Using (3.5), (3.12) and (3.16) we compute
Z
Z
i
h
d 3 p
1
d3 p

ipx ip y
q
e
e
h0|
a
,
a
D(x y) =
|0i
p
~
p~
(2)3
(2)3
4Ep~ Ep~
| {z }
(2)3 (~
p~
p )
(4.18)
Z
3
1  ip(xy) 
dp
e
.
=
p0 =Ep~
3
(2) 2Ep~
Analogously one nds

D(y x) =

d3 p
1  ip(xy) 
e
.
p0 =Ep~
(2)3 2Ep~
18

(4.19)

Comparing (4.13), (4.14) and (4.17) with (4.18) and (4.19) we observe
Gret = (x0 y 0) (D(x y) D(y x))

Gav = (y 0 x0 ) (D(y x) D(x y))

GF = (x0 y 0) D(x y) + (y 0 x0 ) D(y x).

Finally, dening the time-ordered product of two operators A, B as

T {A(x)B(y)} := (x0 y 0)A(x)B(y) + (y 0 x0 )B(y)A(x) ,

(4.20)

we see that the Feynman-propagator coincides with

GF = h0|T {(x)(y)}|0i.

19

(4.21)

Lecture 5: The Dirac equation

Historically Dirac suggested the Dirac equation as a relativistic Schrodinger equation in

order to incorporate relativistic eects into quantum mechanics. This approach turned
out to be erroneous and only a fully eld-theoretical description as we develop it here
can properly include relativistic eects. Nevertheless the Dirac equation turned out to
be the appropriate eld equation for fermionic elds.

5.1

-matrices

Since the Schrodinger equation is rst order in time while the Klein-Gordon equation is
second order in time, Dirac took the square root of the wave operator . That is he
dened an operator
X
6 :=

(5.1)

such that 6 2 = . This in turn implies

{ , } = 2 ,

(5.2)

where the curly brackets denote the anti-commutator { , } := + . We

immediately note that the s cannot be numbers but have to be matrices. Furthermore,

they are not unique since any dened as

= M M 1

(5.3)

with M being a non-singular complex matrix also satises (5.2):

{ , } = M { , } M 1 = 2 MM 1 = 2 .

(5.4)

With a little more work one can show that the have to be at least 4 4 matrices.
Two prominent representations are the Dirac representation




1 0
0 i
0
i
, i = 1, 2, 3 ,
(5.5)
,
D =
D =
i 0
0 1
where 1 is a 2 2 identity matrix and i are the standard Pauli matrices which satisfy
i j = ij 1 + iijk k . The so called chiral representation is given by


0

c =
,
where
:= (1, i ), := (1, i ).
(5.6)
0
These two representation are related by
c

MD
M 1

for



1 1
.
M=
1 1

(5.7)

An important role is played by the 5 matrix dened as

5 = i 0 1 2 3
20

(5.8)

which obeys
( 5 )2 = 1 ,

{ , 5 } = 0 ,

In the two representation just discussed they take the form





1 0
0 1
5
5
.
,
c =
D =
0 1
1 0

(5.9)

(5.10)

Finally, one observes

0 = 0 ,

5.2

i = i

= 0 0

(5.11)

The Dirac equation

Now we can write down the Dirac equation

4
X
b=1

(iab
mab ) b = 0 ,

(5.12)

where b is a 4-component complex spinor-eld and m is a again the rest mass. The
Dirac equation can be obtained as the Euler-Lagrange equation of the Lagrangian
L =

4
X

a,b=1

a (i mab )b ,

ab

(5.13)

= 0 . Indeed we can compute

where
L
b ,
= m
b

X
L
a
=i

ab
( b )
a

X
L
=
(iab mab )b ,
a

L
a) = 0 .
(

Inserted into (2.6) we obtain the Euler-Lagrange equations for and

X
X
a + m
b = 0 .
(iab mab )b = 0 ,
i
ab

(5.14)

(5.15)

As promised the rst equation is indeed the Dirac equation while the second equation
can be shown to be its hermitian conjugate using (5.11).

We can also determine the Hamiltonian by rst computing the conjugate momenta
to be
L
L
0 = i ,
=
=0.
(5.16)
= i

The fact that

vanishes is a problem that has to be taken care of when we come to
quantizing the theory. For now we limit ourselves to classical eld theory and obtain for
the Hamiltonian
+
L
H =

(i
0 0 + i i i m)
= i
i i + m
.
= i
21

(5.17)

5.3

Weyl equation

In various application it will turn out to be useful to decompose the 4-component Dirac
spinor into two 2-component Weyl spinors L , R according to
 
L
.
(5.18)
=
R
This relation can be inverted with the help of two projection operators dened as
PL := 21 (1 5 ) ,

PR := 21 (1 + 5 ) ,

(5.19)

which obey
PL2 = PL ,

PR2 = PR ,

PL PR = 0 .

(5.20)

Using 5 in the chiral representation (5.10) we see that

L = P L ,

R = P R

(5.21)

holds.
Continuing in the chiral representation for the -matrices introduced in (5.6) we
obtain from (5.15)
i L mR = 0

i R mL = 0 ,

(5.22)

or in other words two coupled dierential equations for the two Weyl spinors. For m = 0
they decouple and one obtains the two Weyl equations
i R = 0 ,

i L = 0 .

(5.23)

This decoupling is the reason for using the chiral representation.

5.4

Solution of the Dirac equation

Let us rst observe that any solution of the Dirac equation also solves the Klein-Gordon
equation. This can be seen by acting with the operator (i +m) on the Dirac equation
(5.15) which, using (5.2), yields
(i + m)(i m) = ( + m2 ) = 0 .

(5.24)

This motivates the Ansatz:

ipx
+
,
a = ua (p) e

ipx

,
a = va (p) e

(5.25)

where
1
u (p) =
N
s



(p + m) s
,
(p + m) s

1
v (p) =
N
s

(p + m) s
(p + m) s

s = 1, 2 ,

(5.26)

with
0

p = Ep~ 0 ,

N=

2(Ep~ + m) ,
22

 
1
,
=
0
1

 
0
.
=
1
2

(5.27)

The normalization factor N was chosen such that

u = u 0 ,

u us = 2m s ,

v v s = 2m s ,

v us = 0 .

(5.28)

This can be explicitly checked by using the chiral representation and computing

 

u(p)eipx
m p

.
(5.29)

(i m) = ( p m) =
v(p)eipx
p m
Inserting (5.26) we obtain for u(p)

 


m p
(m(p + m) + p(p + m)) s
(p + m) s
=
= 0,
(p + m) s
(p(p + m) m(p + m)) s
p m

(5.30)

where we used p2 = m2 and

i j
pi pj = (p20 ~p2 ) 1 = m2 1 .
p p = 1p20 + p0 pi ( i i )
|{z}

(5.31)

ij 1

A similar computation shows that also v(p) solves the equation.

Before we proceed let us check the non-relativistic limit p0 |~p|. Using the Dirac
representation we obtain

  
i0 + i i i m

=0
(5.32)
=
(i m)
i0 i i i m

0
For p0 m we can split o the time dependence = eimx with slowly varying. The
i
second equations gives us 2m
i i which, when inserted into the rst equation,
yields the non-relativistic Schrodinger equation

1
i0 2m
= H ,

where H is the Hamiltonian.

23

(5.33)

Lecture 6: Covariance of the Dirac equation

In order to make sure that the Dirac equation is a legitimate eld equation in a relativistic
theory we need to check its covariance under Lorentz-transformations. For the case of
the Klein-Gordon equation we observed that  is a Lorentz-invariant operator. However
the Dirac operator is not Lorentz-invariant and as a consequence cannot transform as
a scalar. Let us make the Ansatz for the transformation law of
X

1 ab b ,
a a =
(6.1)
2

where 1 also is a 4 4 matrix but does not coincide with . Let us inspect the
2
transformed Dirac equation

(i m) = (i

1T

m) 1
2

= 1 1 (i
2

1T

1T

= 1 (i1
1 1
2

m) 1
2

(6.2)

m) ,

where we used = 1T and did not transform the -matrices as they have constant entries. So we see that if we can nd a 1 such that
2

1
1 1 =
2

(6.3)

we see that the Dirac equation is covariant, i.e.

(i m) = 1 (i m) = 0 .
2

(6.4)

1 can be systematically constructed using group-theoretical methods. Here we just

2
give the result and check that is satises (6.3). One nds

 i

(6.5)
,
with
Sab
:= 4i [ , ]ab ,
1 ab = e 2 S
2

ab

where the six parameters of the Lorentz-transformation have been expressed in terms of
an anti-symmetric 4 4 matrix = .
It is straightforward to check (6.3) innitesimally (i.e. to rst order in ). Inserting
and 1 expanded to rst order into (6.3) one arrives at
2

[S , ] = i( ) ,

(6.6)

which can be checked by repeated use of (5.2).

Let us also determine the transformation law of the conjugate spinor. From its denition := 0 we compute innitesimally




= 0 = 2i S 0 = ( 0 )2 2i S 0
| {z }
=1
(6.7)




= 0 2i S = 2i S ,
24

where in the second line we used = 0 0 . Exponentiated this yields

= 1
1 .

(6.8)

Thus we see that is a Lorentz-scalar as is the Lagrangian (5.13).

In the chiral representation (5.6) S is block-diagonal and given by


 k
 i
0

0
ij
i
j
0i
i i
i
1 ijk
i 0
.
,
S = 4 [ , ] = 2
S = 4 [ , ] = 2
0 k
0 i

(6.9)

Therefore the transformation law of the two Weyl spinors dened in (5.18) reads
L = (ii i 21 i i )L
1
R = (ii i + i i )R
2

(6.10)

where ij = ijk k , i = 0i. We see that L,R both transform like two-dimensional
spinors under rotations.
Finally there are 16 linearly independent spinor bilinears given by
1:

4:

6:

[ , ]

4:

[ ] 5

1:

[ ] 5

25

(6.11)

Lecture 7: Quantization of the Dirac field

The most general solution of the Dirac equation is a superposition of the solution given
in (5.25), (5.26), i.e.
(x) =

(x) =


d3 p
1 X s s
s s
ipx
ipx
p
a
u
(p)
e
+
b
v
(p)
e
p
~
(2)3 2Ep~ s=1 p~
p0 =Ep~
3

1
dp
p
3
(2)
2Ep~

2 

X
s
ipx
u
(p)
e
bps~ v s (p) eipx + aps
~
s=1

(7.1)

p0 =Ep~

In order to canonically quantize the Dirac eld is promoted to a quantum-theoretical

operator as are the Fourier-coecients aqs~, bqs~. They obey the anti-commutation relation7
{apr~ , aqs~ } = (2)3 (3) (~p ~q) rs = {bpr~ , bqs~ }
{apr~ , aqs~}

{apr~

, aqs~ }

{bpr~ , bqs~}

{bpr~

, bqs~ }

(7.2)
=0,

This in turn implies (see problem 4.3)

{a (~x, t) , b(~y , t)} = (3) (~x ~y ) ab
{a (~x, t) , b (~y , t)} = 0

{a (~x, t) , b(~y , t)}

Inserting (7.2) into (5.17) we obtain

Z
Z


i
3
H = d x i m =i d3 x 0 i
=


d3 p X  s s
s s
Ep~ ap~ ap~ + bp~ bp~ ,
(2)3 s

(7.3)

(7.4)

where the second equation is shown to hold in problem 4.3. In the nal expression again
an innite constant was dropped. We see that H is an innite sum of fermionic harmonic
oscillators.
The space of states (Fock-space) is again constructed from a vacuum state |0i dened

as

The one-particle states are

aps~ |0i = 0 = bps~ |0i .

(7.5)

aps~ |0i ,

(7.6)

bps~ |0i .

There energy can be determined by acting with H given in (7.4)

Z
Z
d3 p X
d3 p X
s s r
r
E
a
a
a
|0i
=
Ep~ aps~ {aps~ , aqr~ }|0i = Eq~aqr~ |0i ,
H aq~ |0i =
p
~
p
~
p
~
q
~
3
3
(2) s
(2) s
H bqr~ |0i = . . . = Eq~bqr~ |0i .
7

(7.7)

The anti-commutator of two operators A, B is defined by {A, B} = AB + BA.

26

Thus we see that both states have the same mass.

We can also compute their charge by rst observing that Dirac Lagrangian (5.13)
L = (i m) is invariant under the transformation
= ei ,

R.

(7.8)

The corresponding Noether current j and Noether charge Q are determined in problem 3.2 to be
Z
Z
Z

d3 p X  s s
s s

3
0
3

ap~ ap~ bp~ bp~ , (7.9)

j = ,
Q = d xj = d x =
(2)3 s
and one can check the conservation law j = 0.8 Now we can compute the charge of
the two one-particle states to be
Z
Z
d3 p X s s r
d3 p X s s r
r
Q aq~ |0i =
a
a
a
|0i
=
ap~ {ap~ , aq~ }|0i = aqr~ |0i ,
p
~
p
~
q~
3
3
(2) s
(2) s
(7.10)
Q bqr~ |0i = . . . = bqr~ |0i .

We thus see that aqr~ |0i and bqr~ |0i have the same mass but opposite charges. These states
are therefore identied with the electron and its anti-particle the positron. In fact, the
existence of anti-particle (same mass but opposite charge) is a generic prediction of the
Dirac theory. Finally, using
Z
Z

d3 p X  s s
s s
3

~
p~ = d x (i) =
(7.11)
p~ ap~ ap~ + bp~ bp~ ,
(2)3 s
we can check that aqr~ |0i and bqr~ |0i carry momentum ~q.

7.1

Dirac propagator

Let us compute
h0|a (x)b (y)|0i =
!
Z
X
d 3 p d 3 p
1
p
aps~ usa (p) eipx
h0|
(2)3 (2)3 4Ep~ Eq~
s
Z
d3 p 1 X s
=
u (p)usb (p) eip(xy)
(2)3 2Ep~ s a
In order to proceed we observe
X

aps~

ub (p ) e

|0i

(7.12)

usa (p)usb (p) = ab

p + mab ,

ip y

vas (p)v sb (p) = ab

p mab ,

We will see later that Q is indeed the electromagnetic charge operator.

27

(7.13)

which are proved in exercise 4.2. Inserted into (7.12) we arrive at

Z
d3 p 1

h0|a (x)b (y)|0i =

(ab
p + mab ) eip(xy)
3
(2) 2Ep~
Z


d3 p 1 ip(xy)
x
= iab + mab
e
.
(2)3 2Ep~

(7.14)

Analogously one obtains

h0|b (y)a (x)|0i =

x
iab

+ mab

d3 p 1 ip(yx)
e
.
(2)3 2Ep~

(7.15)

Let us dene
SR ab := (x0 y 0 ) h0|{a (x) , b (y)}|0i

(7.16)

SR ab = (i x + m) Gret (x y) ,

(7.17)

and observe
where Gret (x y) was computed in (4.13). We can now easily check that SR ab is the
Greens function of the Dirac operator in that it satises
(i x m)SR = (i x m)(i x + m) Gret (x y) = i 4 (x y)
{z
}
|

(7.18)

(+m2 )

Alternatively one can consider the Fourier-transformation of SR

Z
d4 p ip(xy)
e
SR (p)
SR =
(2)4

(7.19)

and demand (7.18) to hold. This yields

i( p + m)
i
SR =
=
2
2
p m
p m

(7.20)

Now one has to perform the p0 integral analogously to lecture 4 which indeed conrms
(7.17).
The Feynman propagator is dened by

h0|(x)(y)|0i for x0 > y 0
SF (x y) =
h0|(y)(x)|0i for x0 < y 0

(7.21)

One generalizes the time-ordered product to include fermionic operators by

B(y)

A(x)

T {A(x),
B(y)}
:= (x0 y 0)A(x)
(y 0 x0 )B(y)
,

(7.22)

where the +-sign holds for bosonic operators while the -sign is taken for fermionic
operators. We see that


(7.23)
SF (x y) = h0|T (x)(y) |0i
holds.

28

Lecture 8: Discrete Symmetries

In this section we discuss three discrete symmetries:

parity P : (t, ~x) (t, ~x) ,
time reversal T : (t, ~x) (t, ~x) ,

(8.1)

charge conjugation C : particles anti-particles .

Quantenelectrodynamics (QED) and Quantenchromodynamics (QCD) preserve individually P, C, T while the weak interaction violate them and only preserve the combination
P CT . Let us start with parity.

8.1

Parity

Parity acts on the operators aps~ , bps~ via

P aps~ P = a as~p ,

P bps~ P = b bs~p ,

(8.2)

i.e. it reverses p~ without changing the spin s. Since two parity transformations should
give back the original operators up to a phase, one needs to demand a2 = 1, b2 = 1.
For (x) we compute from (7.1)
P (x)P =

2

d3 p
1 X s
s
s
ipx
s
ipx
p
v
(p)
e
P
a
P
u
(p)
e
+
P
b
P
p
~
p0 =Ep~
(2)3 2Ep~ s=1 | {zp~ }
| {z }
a as~p

(8.3)

b bs
~
p

Changing variables p p = (p0 , ~p) and using

p x = p (t, ~x) ,

p = p
,

u(p) = 0 u(
p) ,

p
=p ,

v(p) = 0 v(
p) ,

(8.4)

we nd
P (x)P =


1 X
d3 p
s 0 s
i
p(t,~
x)
s 0 s
i
p(t,~
x)
p
a ap~ u (
p) e
b bp~ v (
p) e
(2)3 2Ep~ s=1
p0 =Ep~

= a 0 (t, ~x)

(8.5)
for b = a or a b = 1 respectively. Note that P acts dierently on particles and
anti-particles.

8.2

Time-reversal

~ = ~x ~p L).
~
Time-reversal acts as in (8.1) and thus also inverts p~ and the spin (since L
Thus together we have
t t ,

p~ ~p ,
29

1 2

(8.6)

Let us denote the spin-ipped by s . Then we have

s = i 2 s .

(8.7)

Let us also compute the spin-ipped u


 2

0
s
us (p) = 1 3 us (p) .
u (
p) = i
2
0
Analogously one nds

v s (
p) = 1 3 v s (p)

Now we demand

T aps~ T = a as
~
p ,

T bps~ T = b bs
~
p .

(8.8)

(8.9)
(8.10)

T acts as an anti-unitary operator, i.e.

T = T 1 ,

T c = c T ,

for c C .

(8.11)

For the action on we get

T (x)T =

2

1 X  s s
d3 p
s s
ipx
ipx
p
T
a
u
(p)
e
+
b
v
(p)
e
T
p
~
p
~
(2)3 2Ep~ s=1
p0 =Ep~

2

ipx 
1 X  s s
ipx
d3 p
s
s
p
a
u (p) e + b~p v (p) e
(2)3 2Ep~ s=1 ~p
p0 =Ep~

1 3


d3 p
1 X  s s
s s
i
p(t,~
x)
i
p(t,~
x)
p
a
u
(
p
)
e
+
b
v
(
p
)
e
~
p
(2)3 2Ep~ s=1 p~
p0 =Ep~

= 1 3 (t, ~x) ,

(8.12)
where in third equation we have again changed variables to p.

8.3

Charge conjugation

Charge conjugation C exchanges particles with anti-particles while keeping the spin unipped. This amounts to
C aps~ C = bps~ ,

C bps~ C = aps~ .

Repeating analogous steps as in the previous two sections one nds




02 T .
C (x) C = i 2 = i

8.4

(8.13)

(8.14)

Fermion bilinears

One can now compute the transformation properties of the 16 fermion bilinears introduced in (6.11). The result is summarized in table 8.4 using the convention (1) = 1
for = 0 and (1) = 1 for = 1, 2, 3.
30

P
T
C
CPT

[ , ]
+1
1
(1)
+1
1
(1)
+1
+1
1
+1
+1
1

5
5
(1)
(1) (1)
(1)
(1) (1)
+1
1
1
+1

Table 8.0: Transformation properties of fermion bilinears

Lecture 9: Perturbation theory

Let us start with a simple example of an interacting eld theory

R,

L(, ) = 21 12 m2 2 4!1 4 ,

(9.1)

where is again a real scalar eld. Compared to the Lagrangian (2.9) a self-interaction
term 4 has been added and the strength of the interaction is governed by a real dimensionless coupling constant . We can repeat the steps of section 2 and compute
L
= m2 3!1 3 ,

L
= ,
( )

L
= .

(9.2)

Inserted into the Euler-Lagrange equation (2.6) yields

( + m2 ) = 3!1 3

(9.3)

while from (2.8) we obtain

H = L =

1
2

~
~ + m2 2 ) + 1 4 .
( 2 +
4!

(9.4)

Our nal goal is to compute scattering cross sections and decay rates in an interacting
quantum eld theory. Let us start by computing the Greens function h|T {(x)(y)}|i
where |i is the ground state auf the interacting theory which does not coincide with
the ground state |0i of the free theory, i.e. |i =
6 |0i. For arbitrary this is a dicult
computation. What we can do is evaluate the Greens function perturbatively in which
gives good results for 1. We split the Hamiltonian
H = H0 + Hint
Z


1
~
~ + m2 2
H0 = 2 d3 x 2 +
Hint =

4!

(9.5)

d3 x4 .

Before we proceed let us recall the relation of operators in the Heisenberg versus
Schrodinger picture. They are related by9
H (t, ~x) = eiH(tt0 ) S (t0 , ~x)eiH(tt0 ) ,
9

In the previous sections we always used a time-dependent (t, ~x), i.e. a Heisenberg operator.

31

(9.6)

where H is the time-dependent Heisenberg operator while S is the time-independent

Schrodinger operator. H solves the Heisenberg equation
it H = HH + H H = [H , H] .

(9.7)

An operator in the interaction picture is dened by

Z


d3 p
ipx 
iH0 (tt0 )
iH0 (tt0 )
ipx
1

I (t, ~x) = e
S (t0 , ~x)e
=
ap~ e
+ ap~ e
,
0
(2)3 2Ep~
p =Ep~
(9.8)
where the second equation holds since we have already solved the free theory. Inverting
(9.8) and inserting into (9.6) we arrive at
H (t, ~x) = U (t, t0 )I U(t, t0 ) ,

U(t, t0 ) = eiH0 (tt0 ) eiH(tt0 ) ,

with

(9.9)

where one can easily check U U = 1 and U(t0 , t0 ) = 1. In order to determine U we

compute
it U = eiH0 (tt0 ) H0 eiH(tt0 ) + eiH0 (tt0 ) HeiH(tt0 )
= eiH0 (tt0 ) Hint eiH(tt0 )
= eiH0 (tt0 ) Hint eiH0 (tt0 ) eiH0 (tt0 ) eiH(tt0 )

(9.10)

= HI U ,
where
HI = eiH0 (tt0 ) Hint eiH0 (tt0 ) .

(9.11)

This can be turned into the integral equation

Z t
Z t

dt HI (t )U(t , t0 ) ,
dt t U(t , t0 ) = U(t, t0 ) U(t0 , t0 ) = i

(9.12)

t0

t0

and thus
U(t, t0 ) = 1 i

dt HI (t )U(t , t0 ) .

(9.13)

t0

(9.13) can now be solved iteratively by inserting (9.13) repeatedly into itself
U(t, t0 ) = 1 i

t0

dt HI (t ) 1 i

= 1 + (i)

dt HI (t )U(tt , t0 )
t0

dt HI (t ) + (i)

+ (i)

t0

t0

dt
t0

t0

dt

dt

dt HI (t )HI (t )

(9.14)

t0

dt HI (t )HI (t )HI (t ) + . . . .
t0

(9.14) can be written as a time-ordered product which for an arbitrary number of operators is dened as
T {H(t1 ) . . . H(tn )} :=(t1 t2 )(t2 t3 ) . . . (tn1 tn ) H(t1 ) . . . H(tn )
+ n! permutations
32

(9.15)

Let us rst consider

Z t
Z t
1
dt1
dt2 T {H(t1)H(t2 )}
2
t0

=
=
=

t0

1
2

dt1

t0

1
2

t0

dt1

t0

dt1

dt2 (t1 t2 )H(t1 )H(t2 ) +

t1

dt2 H(t1 )H(t2 ) +

t0

t0

1
2

dt2
t0

1
2

dt1

t0

t
t0

dt2 (t2 t1 )H(t2 )H(t1 )

t2

dt1 H(t2 )H(t1 )

t0

t1

dt2 H(t1 )H(t2 ) ,

t0

(9.16)
where in the last step we changed the integration variables t1 t2 . An analogous
computation shows
Z t
Z t
Z t
Z t1
Z tn1
1
dt1 . . .
dtn T {H(t1 ) . . . H(tn )} =
dt1
dt2 . . .
H(t1 ) . . . H(tn ) .
n!
t0

t0

t0

t0

t0

(9.17)

Thus we can write (9.14) in terms of time-ordered products as

i

U(t, t0 ) = T {e

Rt

t0

dt HI (t )

}.

(9.18)

}.

(9.19)

In order to proceed we need the generalization

U(t, t ) = T {ei

Rt

dt HI (t )

which satises the properties of a time-evolution operator

U(t, t ) = U(t, t0 )U(t0 , t ) .

(9.20)

However, this requires a modication of (9.9)

U(t, t ) = eiH0 (tt0 ) eiH(tt ) eiH0 (t t0 ) ,

(9.21)

which agrees with (9.9) for t = t0 and obeys

U(t, t )U(t , t ) = U(t, t ) .

(9.22)

So far we rewrote H in terms of the known I and HI . We are left with expressing
the ground state |i of the interacting theory in terms of the ground state |0i of the free
theory. The two ground states are dened by
H0 |0i = E00 |0i = 0 ,

H|i = E0 |i ,

where we inserted our choice E00 = 0. Let us compute

X
X
eiHT |0i =
eiHT |nihn|0i =
eiEn T |nihn|0i ,
n

33

(9.23)

(9.24)

where we inserted a complete set of states. Now we take the somewhat unconventional
limit T (1 i) which allows us to single out the rst term in the sum
lim

T (1i)

eiHT |0i = eiE0 T |ih|0i + subleading ,

(9.25)

where we used E0 < En6=0 . (9.25) implies

|i =

lim

T (1i)

eiHT |0i
eiE0 T h|0i

(9.26)

In order to express |i in terms of U we shift T T + t0 and insert eiH0 (T +t0 ) to obtain

|i =

U(t0 , T )|0i
eiH(T +t0 ) eiH0 (T +t0 ) |0i
=
lim
.
iE
(T
+t
)
iE
0
0
0 (T +t0 ) h|0i
T (1i) e
e
h|0i

lim

T (1i)

(9.27)

Analogously one derives

h| =

h0|U(T, t0 )
iE
0 (T t0 ) h0|i
T (1i) e
lim

(9.28)

The normalization h|i = 1 implies

lim

T (1i)

e2iE0 T |h0|i|2 =

lim

T (1i)

h0|U(T, T )|0i .

(9.29)

Now we are prepared to compute the correlator h|(x)(y)|i. Let us rst choose
x > y 0 > t0 and use
0

(x)(y) = U (x0 , t0 )I (x)U(x0 , t0 )U (y 0 , t0 )I (y)U(y 0, t0 )

= U (x0 , t0 )I (x)U(x0 , y 0)I (y)U(y 0, t0 ) ,

(9.30)

where in the last step we used (9.9). Inserting (9.27), (9.28), and (9.30) we arrive at
h|(x)(y)|i =
=

h0|U(T, t0 )U (x0 , t0 )I (x)U(x0 , y 0)I (y)U(y 0, t0 )U(t0 , T )|0i

T (1i)
h0|U(T, T )|0i
lim

h0|U(T, x0 )I (x)U(x0 , y 0)I (y)U(y 0, T )|0i

.
T (1i)
h0|U(T, T )|0i
lim

(9.31)

Note that the numerator is time-ordered so that together with (9.18) we arrive at
h|(x)(y)|i =

h0|T {I (x)I (y)U(T, x0 )U(x0 , y 0)U(y 0 , T )}|0i

T (1i)
h0|U(T, T )|0i
lim

h0|T {I (x)I (y)U(T, T )}|0i

T (1i)
h0|U(T, T )|0i

h0|T {I (x)I (y)ei

lim
lim

T (1i)

h0|T {ei
34

RT
T

RT
T

dtHI (t)

dtHI (t)

}|0i

}|0i

(9.32)
.

For y 0 > x0 one derives a similar fomula so that altogether one has
h|T {(x)(y)}|i =

lim

T (1i)

h0|T {I (x)I (y)ei

h0|T {ei

RT
T

RT
T

dtHI (t)

dtHI (t)

}|0i

}|0i

(9.33)

Thus we see that the correlator can indeed be compactly expressed in terms of I , HI
and |0i. For an arbitrary product of operators one derives analogously
h|T {(x1 ) . . . (xn )}|i =

lim

T (1i)

h0|T {I (x1 ) . . . I (xn )ei

h0|T {ei

RT
T

dtHI (t)

RT
T

dtHI (t)

}|0i

}|0i

This formula will now be used as the starting point of perturbation theory.

35

(9.34)

10

Lecture 10: Wick Theorem

As stated at the end of the

last section, the goal is to compute amplitudes of the form
R
i dtHI
h0|T {I (x1 ) . . . I (xn )e
}|0i. We already did compute h0|T {I (x1 )I (x2 )}|0i =
GF (x1 x2 ) in section 4 which is the Feynman propagator(cf. (4.21)). The strategy will
be to express in terms of the operators a, a and commute them appropriately. Wicks
Theorem provides a systematic way to do this.

It is useful to introduce the notation I = +

I + I with
Z
Z
d3 p
1
1
d3 p
+
ipx

p
p
I =
e
a
,

eipx ap~ ,
=
p
~
I
3
3
(2)
(2)
2Ep~
2Ep~

(10.1)

which obey

+
I |0i = 0 ,

and

h0|
I = 0 .

(10.2)

Now we dene the normal order N{a} by arranging all a -operators on the left and
all a-operators on the right, i.e.
arbitrary order

z }| {
N{ap~1 ak~ ak~ ap~r } := ak~ ak~ ap~1 ap~r
n
1
1
| {z }
|
{z n }

(10.3)

arbitrary order

arbitrary order

This implies immediately

h0|N{I (x1 ) . . . I (xn )}|0i = 0 .

(10.4)

Next we dene the Wick contraction by





(x)(y) := (x0 y 0) + (x), (y) + (y 0 x0 ) + (y), (x)

(10.5)

By inserting (10.1) and using (4.17) one shows

(x)(x) = GF (x y) .

(10.6)

The Wick contraction is precisely the dierence between time ordering and normal ordering and one has
T {(x)(y)} = N{(x)(y) + (x)(y)}

(10.7)

In order to prove (10.7) we insert (10.1) into the denition (10.3) and obtain



0
0
T {(x)(y)} = (x y ) + (x)+ (y) + (x) (y) + + (x), (y)



+ (x)+ (y) + (y)+ (x) + (y 0 x0 ) x y



= (x0 y 0) N{(x)(y)} + + (x) (y)



+ (y 0 x0 ) N{(x)(y)} + + (y) (x)
= N{(x)(y) + (x)(y)} ,

(10.8)
36

where in the last step we used (10.5) and the fact that GF is a c-number and thus can
be moved inside the normal order.
The Wick Theorem generalizes (10.7) for an arbitrary product of eld operators and
states:
T {(x1 )(x2 ) (xn )} = N{(x1 )(x2 ) (xn ) + all Wick contractions} . (10.9)
As example consider
T {1 2 3 4 } = N{1 2 3 4 + 1 2 3 4 + 1 2 3 4 + 1 2 3 4

(10.10)

+ 1 2 3 4 + 1 2 3 4 + 1 2 3 4 + 1 2 3 4 } ,
where we abbreviated (xi ) = i .
Wicks Theorem is particularly useful for computing vacuum expectation values (VEVs)
of products of operators since all normal ordered terms drop out due to the property
(10.4). For example, taking the VEV of (10.10) and using (10.6) one obtains
h0|T {12 3 4 }|0i = h0| 1 2 3 4 + 1 2 3 4 + 1 2 3 4 |0i
= GF (x1 x2 )GF (x3 x4 ) + GF (x1 x3 )GF (x2 x4 ) + GF (x1 x4 )GF (x2 x3 ) ,
(10.11)
and everything is expressed in terms of GF only.
One can prove Wicks theorem by induction. Assume that it holds for n 1 operators
and then prove that it also holds for n operators
T {(x1 ) . . . (xn )} = (x01 x02 ) (x02 x03 ) . . . (x0n1 x0n ) 1 . . . n
+ all permutations

for x01 largest

1 T {2 . . . n }
2 T {1 3 . . . n } for x02 largest
=

...

For each possibility we can now use the assumption that the theorem holds for n 1
operators, i.e.
1 T {2 . . . n } = 1 N{2 . . . n + all contractionen without1 }

= N{
1 ((2 . . . n ) + all contractionen without1 ) +

+
+
N{[+
1 , 2 ] 3 . . . n + 2 [1 , 3 ] . . . n + 2 3 [1 , 4 ] . . . n + . . .}
| {z }
| {z }

1 2

1 3

= N{1 2 . . . n + all contractionen}

(10.12)
Before we continue let us also state the Wick-theorem for fermions. We already dened
the time-ordered product for fermions in (7.22) and expressed the Feynman propagator
37

in terms of a time-ordered product in (7.21) and (7.23). The crucial dierence compared
to bosonic operators is an extra minus sign steming from the anti-commutation property
of fermions. Analogously one denes for n fermionic operators an extra minus sign for
each permutation
T {(x1 ) . . . (xn )} = (x01 x02 )(x02 x03 ) . . . (x0n1 x0n ) (x1 ) . . . (xn )

(x02 x01 )(x01 x03 ) . . . (x0n1 x0n ) (x2 )(x1 ) . . . (xn )

+ ...
+ (1)ni for ni permutations + . . .
(10.13)

Similarly, the normal ordering is dened with a analogous minus sign, i.e.
N(ap~1 ap~2 aq~1 . . .) = (1)k aq~1 . . . aq~r ap~1 . . . ap~n = (1)k+1aq~1 . . . aq~r ap~2 ap~1 . . . ap~n , (10.14)
for r raiusing and n fermionic lowering operators.
The Wick-theorem for fermions again states
2 3 . . .} = N(1
2 3 . . . + all Wick contractions) ,
T {1
where a Wick contraction now is dend by

(y)} for x0 > y 0
{+ (x),

1 2 :=
+ (y), (x)} for y 0 > x0
{

(10.15)

(10.16)

where = + + analogously to = + + . Comparing to (7.21) we see

2 |0i = SF (x y) holds.
h0|1

38

11

Lecture 11: Feynman diagrams

Let us compute
1
A = h|T {(x)(y)}|i = h0|T {(x)(y) exp (i
N
where
N = h0|T {exp (i

HI (t)dt)}|0i

(11.1)

HI (t)dt)}|0i .

(11.2)

Since H is time independent it is a Schrodinger operator and thus for a 4 -theory we

have
Z
Z
Z
3

(11.3)
dt d zHI (t) = 4! d4 z4I (z) .

Expanding in yields

h|T {(x)(y)}|i =

1
h0|T {(x)(y)}|0i
N

i
4!N

d4 zh0|T {x y 4I }|0i + O(2)

(11.4)
Using Wicks theorem one obtains three distinct terms plus permutations which merely
result in multiplication factor
h|T {(x)(y)}|i =

1
h0|I (x)I (y)|0i
N

+3

i
4!N

+43

i
4!N

+ O(2 )

d4 zh0|T {I (x)I (y)I (z)I (z)I (z)I (z)}|0i

d4 zh0|T {I (x)I (z)I (z)I (z)I (z)I (y)}|0i

(11.5)

In terms of Feynman propagators this corresponds to

Z

1
i
A = h|T {(x)(y)}|i = N GF (x y) 8
d4 z(GF (x y)GF (z z)GF (z z)

i
2

d4 zGF (x z)GF (y z)GF (z z)) + O(2)

(11.6)

Now one associates a set diagrammatic rules (the Feynman rules) with such expressions:
1. For each propagator GF (x y) =
R
2. For each vertex i dz =
3. For each external point 1 =

4. Finally one needs to compute the symmetry factor (which is not encoded in the
diagram)
39



In terms of such diagrams the expression (11.6) corresponds to

A=

1
N

(11.7)

Before we continue recall that in expressions like (9.34) we encounter the limit
lim

T (1i)

dz

d3 z ei(p1 +p2 +p3 p4 )z .

(11.8)

The integrand diverges at either unless p0 is also taken to be slightly imaginary

p0 p0 (1 + i). This precisely corresponds to taking Feynman propagator.

The second diagramR in (11.7) is called a disconnected diagram. Such diagrams arise
when (x), (y) and e HI are contracted seperately among themselves. Thus one has
schematically




(11.9)
A = N1 all connected diagrams e(disconnected pieces)
Since N is given by (11.2) we see that is also is proportional to (e(disconnteced pieces) and
thus cancel the second factor in (11.9). We are left with
A = sum of all connected diagrams

(11.10)

This observation is readily generalized to an arbitrary n-point amplitude

An = h|T {(x1 ) . . . (xn )}|i
= sum of all connected diagrams with n external points

40

(11.11)

12

Lecture 12: S-Matrix

A typical experiment in particle physics consists of a collision of two particle beams and
the subsequent detection of the decay products. The measured and computed quantity
is the cross section dened as
=

number of scattering events

,
A lA B lB A

(12.1)

where the two colliding particle bunches A and B have densities A and B and length
lA and lB . A is the common area.
In order to compute one assumes that at t = the particles are non-interacting
and can be described by the the state
Z
d3 k
1
p
(~k)|~ki .
(12.2)
|i =
3
(2)
2E~k

Thus the intial state is given by

Z 3
Z 3
d kA
d kB A (~kA )B (~kB ) i~b~kB ~ ~
p
|A B iin =
e
|kA kB iin
(2)3
(2)3 (2EA )(2EB )
where ~b is the impact parameter. The nal state reads
Y Z d3 pf f (~pf )
p
p1 ~p2 . . . | .
out h~
out h1 2 | =
(2)3 2Ef
f

The probability P that the intial state |A B iin evolves into the nal state
is given by
P = |out h1 2 . . . |A B iin |2 .

(12.3)

(12.4)
out h1 2

...|

(12.5)

Thus we have to compute

p1 ~p2
out h~

. . . |~kA~kB iin = lim h~p1 p~2 . . . |eiH(2T ) |~kA~kB i =: h~p1 p~2 . . . |S|~kA~kB i ,
T

(12.6)

where the last step denes the S-matrix. If there is no interaction S = 1 holds and
thus the non-trivial scattering is captured by the T -matrix dened as S = 1 + iT (the
factor i is convention). Computing out h~p1 p~2 ...|iT |~kA~kB iin one realizes that momentum
conservation is garuanteed and therefore one denes the matrix M via
X
h~p1 p~2 . . . |iT |~kA~kB i =: (2)4 (4) (kA kB
pf )iM(kA , kB pf ) .
(12.7)
f

The relation bewteen and M is computed elsewhere [2] and is not repeated here. One
nds for the dierential cross section
!
P
(2)4 (4) (pA + pB f pf ) Y d3 pf 1
d =
|M(kA , kB pf )|2 ,
(12.8)
3
2EA 2EB |~vA ~vB |
(2) 2Ef
f

where |~vA ~vB | denotes the relative velocity of the incoming beams in the laboratory
frame and pA , pB are the central values of the incoming momenta.
41

Later on we often consider the situation where only two particles are in the nal state.
In the center of mass frame (12.8) then simplies to
 
d
1
1
|~p1 |
=
|M(pA, pB p1 , p2 )|2 .
(12.9)
2
d CM
2EA 2EB |~vA ~vB | (2) 4Ecm

Furthermore, if all four particles involved have the same mass, (12.9) further simplies
 
d
|M|2
.
(12.10)
=
2
d CM
64 2 Ecm
For the decay rate of a particle one nds
!
P
(2)4 (4) (pA + pB f pf ) Y d3 pf 1
|M(mA {pf })|2 ,
d =
2mA
(2)3 2Ef

(12.11)

The matrix M can be computed by the LSZ-formalism (the derivation will be given
later) with the result


RT
dtHI (t)
i T
h~p1 . . . ~pn |iT |~pA ~pB i =
lim
}|~pA p~B i0 fully connected,
p1 . . . ~pn |T {e
0 h~
T (1i)
amputated
(12.12)
The notation fully connected, amputated refers to the contributing Feynman-diagrams.
A Feynman-diagram is called amputated if an external point cannot be removed by
cutting a single propagator. Fully connected diagrams are diagrams where all external
points are connected to each other.
In order to evaluate ampltudes or Feynman diagrams we also need the expressions
Z
p
d3 k
1
+
ikx
p
(x)|pi0 =
(12.13)
2Ep~ ap~ |0i) = eipx |0i ,
e
a
(
~
k
3
(2)
2E~k

where in the last step we used the commutator (3.12). Thus we conclude
I (x)|pi0 = eipx |0i

and analogously

0 hp|I (x)

= h0|eipx

(12.14)

As an example let us compute

out hp1 p2 |pA pB iin

= hp1 p2 |pA pB i + hp1 p2 |iT |pA pB i ,

(12.15)

where the rst term represents no scattering while the second term contains the interaction. Using (12.7) and (12.12) it evaluates at leading order to
Z
i
hp1 p2 |iT |pA pB i =
4! d4 zhp1 p2 |(z)(z) (z)(z)|pA pB i
4!
Z
= i d4 zei(pA +pB p1 p2 )
(12.16)
= i(2)4 4 (pA + pB p1 p2 )

which implies
M=,

= (2)4 4 (pA + pB p1 p2 )iM ,

and

d
d

=
CM

42

|M|2
2
=
.
2
2
64 2 ECM
64 2 ECM

(12.17)

13
13.1

Lecture 13: Feynman rules for QED

The Lagrangian of QED and its gauge invariance

The Lagragian for QED is given by

L=

N
X
j=1

j i D j mj j 1 F F ,

(13.1)

D j := j + iqj A j .

(13.2)

where the covariant derivative D is dened by

A is the photon eld and the index j runs over all known elementary particles, i.e.
leptons: e, , , e , , ,
quarks: u, d, c, s, t, b ,

(13.3)

and qj is the electric charge of the respective eld.

An alternative way to display the Lagrangian (13.1) reads
L = LDirac + LMaxwell + Lint
where
LDirac =

N
X
j=1

(13.4)

j i j mj j ,

LMaxwell = 41 F F ,

(13.5)

Lint = A j ,

j =

N
X

j j .

j=1

Note that j is a conserved Noether current satisfying j .

The Lagrangian (13.1) is invariant under the gauge transformations
j j = eiqj (x) j ,
A A = A (x) .

(13.6)

This is a generalization of the phase rotation discussed earlier in that the parameter of
the transformation (x) is an arbitrary real function of the space-time coordinates x.
This situation is called a local symmtry or a gauge symmetry. Using (13.6) it is easy to
check that
D j D j = eiqj (x) D j ,
(13.7)

F F
= F .

The fact that D j transforms exactly like j is the dening feature of a covariant
derivative.
Finally, the Euler-Lagrange equations are given by


i D mj j = 0 ,
F = j .
43

(13.8)

13.2

Quantization of the photon field

From (13.8) we immediatly nd that the free eld equation for the photon eld reads
F = A A = 0 ,

(13.9)

while the canonically conjugated momentum is = F 0 (i.e. 0 = 0). The quantization

procedure for A is complicated by the gauge invariance which renders one d.o.f. of
A unphysical. As a consequence the rules for canonical quatization cannot be applied
naively in that a gauge has to be chosen. Common gauge choice are the Coulomb gauge
~A
~ = 0, the temporal gauge A0 = 0 or the Lorentz gauge A = 0. Each gauge has

its advantages and disadvantages; following [2] we continue in the Lorentz gauge which
maintains Lorentz invariance.
In the Lorentz gauge (13.9) simplies to
A = 0 ,
whose classical solutions we can give immediately as
Z
3


d3 k
1 X r r
r r
ikx
ikx
p
A =
a

(k)e
+
a

(k)e
,
~

~k
(2)3 2E~k r=0 k

(13.10)

(13.11)

with k k = 0. r is a basis of polarization vectors which, due to A = 0, are transversal

and obey k r = k r
= 0. In scattering processes one imposes the additional boundfary
condition that the initial and nal photons are transversely polarized, i.e.
= (0,~) ,

with p~ ~ = 0 .

(13.12)

With these prerequisites one imposes the commutation relations

 r r 

a~k , a~k = (2)3 rr (3) (~k k~ ) ,

 r r 


a~k , a
~k = 0 = a~kr , a~kr .

(13.13)

d4 k i ik(xy)
e
(2)4 k 2

(13.14)

As a consequence the propagator is given by

Z
h0|T {A(x)A (y)}|0i =

Before we continue let us note that the sign of h0|T {A0(x)A0 (y)}|0i is wrong. This
is related to the unphysical gauge degree of freedom which, as we will see, decouples from
physical amplitudes.

13.3

The Feynman rules for QED

Let us give the Feynman rules directly in momentum space. One has

Note that the indicates where the diagram continues. The arrow on the fermion lines
indicates particle number ow and not the ow of the momentum.
44

fermion propagator

i( p +m)
p2 m2

photon propagator

A A

i
k2

vertex

iq

incoming fermion

|p, si

us (p)

incoming anti-fermion

si
|p,

vs (p)

outgoing fermion

hs, p|

us (p)

outgoing anti-fermion

hs, p|

v s (p)

incoming photon

A |k, ri

r (k)

outgoing photon

hr, k|A

r
(k)

Table 13.0: Feynman rules for QED

14

Lecture 14: e+e +

In this lecture we consider e+ e -annihilation into a + pair. The relevant Feynmandiagram at lowest order is given in g. 14.

Figure 14.1: Feynman diagram for e+ e + .

The corresponding M-matrix is given by

iM = (ie)2 ur (k) v r (k )
which arises from the Wick-contraction
h~k, r; ~k , r |

d x() (x) () (x)A (x)

4

i

vs (p ) us (p) (e) ,

2
+p)

() (p

d4 y (e) (y) (e) (y)A (y)|~p, s; ~p, s i

(14.1)

(14.2)

Note that the spins of the incoming and outgoing particles are xed in (14.1). In order
to compute |M|2 we need
(
v u) = (
v u) = u 0 v = u v ,

45

(14.3)

where in the last step we used (5.11). Together with (14.1) this yields
|M|2 =

e4
r
r

s
s
r
r
s
s
u

(k)
v
(k
)
v

(p
)
u
(p)
v

(k
)
u
(k)
u

(p)
v
(p
)

()
(e)
()
(e)
(p + p )4
(14.4)

This expression can be further simplied if one sums over the external spins which
experimentally corresponds to incoming and outgoing beams which are unpolarized. Using
X
X
(14.5)
usa (p) u
sb(p) = ( p + m)ab ,
vas (p) vbs (p) = ( p m)ab ,
s

which were shown in problem 4.1 and the notation /p p , we arrive at

1
|M|2
4

e4
Tr[( (k/ m ) (k/ + m )]Tr[( (/p + me ) (p/ me )]
4(p + p )4

(14.6)

To simplify this expression further one uses the following properties of the -matrices10
Tr(odd number of ) = 0 ,

(14.7)

Tr( ) = 4 ,

(14.8)

Tr( ) = 4( + ) .

(14.9)

Inserted into (14.6) together with the property m me 0 we arrive at

1X
8e4

|M|2 =
(p

k)(p

k
)
+
(p

k
)(p

k)
+
m
p

p
.
()
4 Spin
(p + p )4

(14.10)

We can further simplify this expression by going to the center-of-mass (CM) frame.
Let us assume that the electrons come in along the z-axis, i.e. we choose p = (E, 0, 0, E),
p = (E, 0, 0, E) with p2 = m2(e) 0. The -pair is scattered in an angle relative to
the z-axis, i.e. k = (E, ~k), k = (E, ~k) with k 2 = k 2 = m2() (which implies |k~ | = |~k| =
q
E 2 m2() and ~k ~ez = k~ ~ez = |~k| cos ). This further implies
p p = 2E 2 ,

p k = p k = E 2 E|~k| cos ,

p k = p k = E + E|~k| cos ,

(p + p ) = 4E

(14.11)

Inserted into (14.10) we arrive at

1
4

X
Spin

10


8e4  2
~k cos )2 + E 2 (E + ~k cos )2 + 2m2 E 2
E
(E

()
16E 4
!
2 
2 


m
m
()
()
1 + 2 + 1 2 cos2 .
= e4
E
E

|M|2 =

(14.12)

The first identity was shown in problem 6.3, the two latter in the problem 2 of the first exam.

46

Inserted into (12.9) we obtain



d
d

CM

|~k|
1

|M(pp kk )|2 .
4EA EB |~vA ~vB | 16 2 ECM

Using |~vA ~vB | 2c, EA = EB = E =



d
d

CM

ECM
2

and =

e2
4

(14.13)

yields

 X

|~k|
2
1
=
4
|M|
3
32 2 ECM
Spin
2

4E 2

m2()
E2

1+

m2() 
E2

+ 1

m2()  2
cos
E2

The total cross section is obtained by integration

  Z 2
 
Z
Z 1
d
d
= d
=
d
d(cos )
d
d
0
1
s
m2() 
m2() 
42
=
1
+
1

2
3ECM
E2
2E 2
In the high energy limit E or

m()
E

(14.14)
.

(14.15)

<< 1 one obtains

42
2
3ECM

(14.16)

Let us close this section with some further remarks. First of all the derived theoretical
result is conrmed experimentally [2]. It can also be used to dertermine m() or m( )
for the analogous process e+ e + . For the process e+ e hadrons one replaces
+ by a quark pair. This changes on the one hand the charges (2/3 for u,c,t and 1/3
for d,s,b) and adds an overall factor of 3 for the three colors. Thus in the high energy
limit one has
 q 2
X
f
+
(e e Hadronen) 3R
.
(14.17)
e
f {u,d,s,c,b,t}

47

15

Lecture 15: Compton scattering e e

In this lecture we discuss aspects of Compton scattering. The leading order Feynman
diagrams are depicted in g. 15.1.
p,s

k,r

p,s

p+k

k,r

k,r

pk

p,s

k,r

p,s

Figure 15.1: Feyman diagrams for Compton scattering

The corresponding M-matrix reads

(p/ + k/) + m s

s
u (p) r (k)
u (p )
iM = ie r
(k )
2
2
(p + k) m
2


(p/ k/ ) + m s
r
s
r

+ (k)
u (p )
u (p) (k )
(p k )2 m2

(15.1)

This can be further simplied due to the properties

p2 = p2 = m2 ,

k 2 = k 2 = 0 ,

(p + k)2 m2 = 2p k ,

(p k )2 m2 = 2p k ,

(15.2)

and
(/p + m) us (p) = 2p us (p) (/p m)us (p) = 2p us (p) ,

(15.3)

where in the last step we used (5.29) and (5.30). Inserted into (15.1) we arrive at
iM =

r
ie2 r
us (p )
(k ) (k)

 ( k/ + 2 p )
( k/ ) 2 p )  s
u (p)
+
2pk
2pk

(15.4)

This expression can be further simplied by computing |M|2 and summing over spins
and photon polarisations. Therefore let us pause and determine the sum over photon
polarisation.

15.1

Sum over photon polarisation

Consinder an arbitrary Feyman diagram of the type depicted in 15.1.

The details do not matter but let us focus on one outgoing photon line with momentum
k. The corresponding M-matrix reads
X
X

r
2
r

iM = ir
|
r
(k)M

(k)M
|
=
.

(k) (k)M M
(15.5)
r

48

Here M represents the rest of the diagram and due to the A coupling necessarily
has the structure
Z

M = d4 xeikx hf |j (x) . . . |ii .

(15.6)
Since j is a conserved Noether current we can derive the Ward identity
Z
Z

4
ikx

k M = i d x( e )hf |j (x) . . . |ii = i d4 xeikx hf | j (x) . . . |ii = 0 ,

(15.7)

where in the second step we used partial integration. Without loss of generality we can
choose k = (k, 0, 0, k) and transverse polarisations 1 = (0, 1, 0, 0), 2 = (0, 0, 1, 0). In
this basis (15.7) implies M0 = M3 and thus
X
2
1 2
2 2
0 2
1 2
2 2
3 2

|r
.
(k)M | = |M | + |M | = |M | + |M | |M | + |M | = M M
r

(15.8)

Comparing with (15.5) we conclude

X
r
r
= + k X + k X ,

(15.9)

where X is anarbitrary four-vector. Another way of saying this is that in physical

amplitudes we can replace
X
r
r
(15.10)
.
r

15.2

Unpolarized Compton scattering

Let us return to the Compton scattering for unpolarized beams. However, in the follwing
we will not give the details of the computation but merely state the results.11 Using
(14.5) and (15.10) we compute from (15.4)
1
4

r,r ,s,s

|M|2 = 14 e4 tr

 ( k/ + 2 p )


( k/ ) 2 p )

+
p/ + m

2pk
2pk

 (15.11)
( k/ ) 2 p )

( k/ + 2 p )
+
p
+
m
/
2pk
2pk

1
1
2
1

pk

1
pk
= 2e4 m4
.
+ + 2m2
+

pk pk
pk pk
pk
pk

11

Some details can be found in [2].

49

The partial cross section is conveniently given in the lab-frame where

k = (, 0, 0, ) ,

p = (m, ~0) ,

k = ( , sin , 0, cos ) ,

p = (E , ~p ) ,

(15.12)

where is again the scattering angle. One then nds the Klein-Nishina formula

d
2  2 

2
=
+

sin

,
d cos
m2

where =

1+

.
cos )

(1
m

(15.13)

In the limit 0 one has / 1 and

d
2
2
=
1
+
cos

,
d cos
m2

(15.14)

which results in the total Thomson cross section

tot =

82
.
3m2

50

(15.15)

16

Lecture 16: Dimensional regularization

In this lecture we compute the leading order correction to the fermion propagator. This
amounts to evaluating the diagram in g. 16.1. Truncating the two external fermion lines
and using the momentum space Feynamn rules one arrives at
Z
d4 k i(k/ + m)
i
2
i2 (p) = (ie)

,
(16.1)

(2)4
k 2 m2
(p k)2 2
where a IR regulator is included for later convenience. As a next step one simplies
the denominator using Feynman integrals.
pk

Figure 16.1: Leading order correction for the fermion propagator

An expression 1/AB can be given an integral representation of the form
Z 1
Z 1 Z 1
1
(x + y 1)
1
=
=
,
dx
dx
dy
AB
[xA + (1 x)B]2
[xA + yB]2
0
0
0

(16.2)

which can be proven by elementary integration. It is a special case of the more general
representation
Z 1
n
X

1
(n 1)!
=
dx1 . . . dxn
xi 1
,
(16.3)
n
A1 . . . An
[x
A
+
.
.
.
+
x
A
]
1
1
n
n
0
i=1
which is proven for example by induction.
Applied to 2 one derives
1
=
2
2
[k m ][(p k)2 2 ]
=

dx

[k 2

1
(1 x)2pk + (1 x)(p2 2 ) xm2 ]2

1
dx 2
,
[l ]2

(16.4)

where
l = k (1 x)p ,

= xm2 + (1 x)2 + x(1 x)p2 .

Using = 2 and = 4 we have altogether

Z 1
Z
(4m 2k/)
d4 k
2
dx 2
.
i2 (p) = e
4
(2) o
[l ]2
Changing the integration variables to l we arrive at
Z 1
Z


2
dx 4m 2(1 x)p/
i2 (p) = e
o

51

d4 l
1
,
4
2
(2) [l ]2

(16.5)

(16.6)

(16.7)

where the term linear in l vanishes when integrated over a symmetric domain. From this
expression we see that 2 is not well dened. The l-integral does not exist in that it is
logarithmically divergent for large l, i.e. in the UV.
In order to proceed the integral has to be regulated. For the case at hand a simple
momentum cut-o would do but for later use we introduce the idea of t Hooft and
Veltman to regulate such integrals by going away from four space-time dimensions. One
rst Wick rotates to Euclidean momenta12
l0 ilE0 ,

~l ~lE ,

l2 lE2 ,

d4 l id4 lE .

(16.8)

Now one evaluates the integral by considering it as an analytic function of the space-time
dimension d. That is one replaces
Z 4
Z d
d lE
d lE
1
1
I1 = i
= lim I1 (d) = i lim
.
(16.9)
2
2
4
2
d
d4
d4
(2) [lE + ]
(2) [lE + ]2
Choosing spherical coordinates with a measure dd lE = lEd1 dldd and an area element
dd of the d-dimensional unit sphere normalized with
Z
Z
d
2 2
dd = d ,
() :=
dz z 1 ez ,
(16.10)
( 2 )
0
one arrives at
I1 (d) =

i
d

(4) 2 ( d2 )

dlE

lEd1
.
[lE2 + ]2

Substituting y = [lE2 + ]1 into I1 and using the integral representation

Z 1
()()
dyy 1(1 y)1 =
,
( + )
0
one obtains
I1 (d) =

(2 d2 ) d 2
2 .
d
(4) 2 ( d2 ) (2)
i

(16.11)

(16.12)

(16.13)

() has isolated poles at = 0, 1, 2, . . . and as a consequence I1 has poles at d =

4, 6, 8, . . .. The pole at d = 4 precisely corresponds to the logarithmic divergence we
already observed earlier. However in d = 4 dimension the integral I1 is nite and thus
regulated by going away from d = 4. The behavior of the -function near the pole is

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

(16.14)

This implies

12

( ) 2

lim ( )e 2 ln(/4)
I1 = lim I1 () = i lim 2 2 =
2
0
0 (4)
2
(4) 0 2
2

i
lim

ln(/4)
+
O()
.
=
(4)2 0

The rotation is done such that no pole is encountered in the complex l0 -plane.

52

(16.15)

In order to give a consistent formula for 2 we need to reevaluate the numerate in

d-dimensions. One insists that the Dirac algebra is unchanged with 4 4 -matrices
{ , } = 2 144 ,

, = 0, . . . , d 1 .

(16.16)

This in turn implies

= d144 ,

= ( 2) .

Inserted into 2 one obtains

Z 1 

2
i2 (p) = e
dx ( 2)(1 x)/p + (4 )m I1

(16.17)

ie2
=
lim
(4)2 0


 2

dx ( 2)(1 x)p/ + (4 )m
ln(/4) + O() .

o
(16.18)
For the moment we leave this as an divergent expression and return to its interpretation
when we discuss renormalization in section 20.

53

17
17.1

Lecture 17: Field strength renormalization

The Kall
en-Lehmann-representation of the propagator

In the last lecture we found an UV-divergence in the leading order perturbative correction

of the electron or rather fermion propagator h|T {(x)(y)}|i.

A similar divergence
4
occurs for the scalar propagator in a -theory, i.e. in h|T {(x)(y)}|i. Therefore we
need to reconsider the structure of these quantities independent of perturbation theory.
For simplicity we will focus on h|T {(x)(y)}|i and then only give the result for

h|T {(x)(y)}|i.

For the free theory h0|T {(x)(y)}|0i gives the amplitude for a particle propagating
from y to x. Now we want to understand its meaning in the interacting theory. Let us
insert a complete set of eigenstates of H and p~. States with ~p 6= 0 can be viewed as
Lorentz transformation of states with ~p = 0, i.e.
X Z d3 p
1
1 = |ih| +
|p~ihp~| ,
(17.1)
3
(2) 2Ep~ ()

~
0

where the sum is over all zero-momentum states and |p~i denotes all boasts of the zeroR d3 p 1
momentum states |~0 i . (Note that (2)
3 2E () (...) is Lorentz invariant).
p
~
Choosing x0 > y 0 we obtain

h|T {(x)(y)}|i = h|(x)|ih|(y)|i

X Z d3 p 1
h|(x)|p~ ihp~|(y)|i.
+
(2)3 2Ep~

(17.2)

~
0

For simplicity let us consider theories where the vacuum expectation value of vanishes,
i.e. theories with h|(x)|i = 0. Furthermore let us use the space-time translation
P i) to write (x) = eiP x (0)eiP x . Using the invariance of the
operator P = (H,

ground state h|eiP x = h| we obtain

h|(x)|p~ i = h|(0)|p~ ieipx |p0 =Ep~
= h|U 1 U(0)U 1 U|p~ ieipx |p0 =Ep~

(17.3)

= h|(0)|0 ieipx |p0 =Ep~ ,

where in the second line we inserted an operator U. One can choose U to be the unitary
operator which represents Lorentz-boasts in the Fock-space and connects U|p~ i = |0 i.
Since is a scalar one also has U(0)U 1 = (0) and these two properties then lead to
the third line (17.3). Inserting (17.3) into (17.2) we obtain
X Z d3 p 1
h|(x)(y)|i =
eip(xy) |p0 =Ep~ |h|(0)|0 i|2
3 2E
(2)
p
~
0
(17.4)
X Z d4 p
1
=
eip(xy) |p0 =Ep~ |h|(0)|0 i|2 .
(2)4 p2 m2 + i
0

54

An analogous computation for y 0 > x0 similarly yields

Z
dM 2
h|T {(x)(y)}|i =
(M 2 )GF (x y, M 2 ),
2
0

(17.5)

which is called Kallen-Lehmann-representation of the propagator. The spectral density

is given by
X
(2)(M 2 m2 )|h|(0)|p~i|2
(M 2 ) =
0

= 2(M 2 m2 )Z + possibly bound states

(17.6)

+ multiparticle excitations for M 2 4m2 .

Z = |h|(0)|p~i|2 is called eld strength renormalization and m is the physical mass.

It is important to note that m does not coincide with the so called bare parameter m0
which appears in the Lagrangian.
It is also useful to consider the Fourier transform
Z
Z
dM 2
i
4
ipx
d xe h|T {(x)(0)}|i =
(M 2 ) 2
2
p M 2 + i
0
Z
dM 2
i
iZ
+
(M 2 ) 2
.
= 2
2
p m + i
p M 2 + i
4m2 2
(17.7)
Comparison with the free theory
Z
i
(17.8)
d4 xeipx h0|T {(x)(0)}|0i = 2
p m20 + i
reveals the following dierences:
(i) the factor Z which in the free theory becomes Z = |h0|(0)|pi|2 = 1,
(ii) the dierence between the physical mass m and the bare mass m0 6= m,
(iii) the existence of multi-particle states for M 2 4m2 .
For fermions one similarly determines
Z
iZ2 (/p + m)
+ multi-particle states ,
d4 xeipx h|T {(x)(y)}|i = 2
p m2 + i
where
h|(0)|p, si =

17.2

p
Z2 us (p) .

(17.9)

(17.10)

Application: the electron self energy

Let us now apply these consideration to 2 computed in (16.18).13 We immediately

observe that there is no simple pole at p2 = m20 . This is an artefact of perturbation theory
13

To be consistent with the notation of this section we should replace m m0 in (16.18).

55

as can be seen as follows. Let us dene one-particle irreducible Feynman diagrams (1PI)
as diagrams which cannot be split by removing a single line. For the case at hand the
1PI diagrams are depicted in g. 17.1. Let us denote this sum by i(p) and the rst
diagram in the sum by i2 (p).
1PI

Figure 17.1: 1PI diagrams

+

1PI

1PI

1PI

+ . . .

Figure 17.2: Diagramatic expression for exact fermion propagator

R
Let us further denote the full propagator d4 xeipx h|T {(x)(y)}|i by a blob
without any label. This full propagator can then by given by a sum of 1PI diagrams as
depicted in g. 17.2. This sum corresponds to

i

i

i
i
i
i
+
i(p)
+
i(p)
i(p)
+ ...
p/ m0 p/ m0
p/ m0 p/ m0
p/ m0
p/ m0
n

X
(p)
i
i
.
=
=
/p m0 n=0 /p m0
/p m0
(17.11)
Now we see that there is a simple pole at


(17.12)
/p m0 (p) (/p=m) = 0 ,
which denes the physical mass m = m0 + (p/ = m). Taylor expansion around the pole
yields


d 
(/p m) + O (/p m)2 .
(17.13)
= (/p = m) +
(
p
=m)
/
d/p

which implies







d 
2
+
O
(
p

m)
.
/p m0 (p) = (/p m) 1
/
(
p
=m)
dp/ /

(17.14)

Comparing with (17.9) we conclude

Z21 = 1

d 
.
d/p (/p=m)

56

(17.15)

18

Lecture 18: Renormalization of the electric charge

In the previous lecture we discussed the corrections to the fermion propagator. Let us
now repeat a similar analysis for the photon propagator. The exact photon propagator is
diagramatically depicted in g. 18.1. It can be expressed in terms of the 1 PI diagramms
which we denote by i (q) and which are depicted in g. 18.2. As before we denote
by i
2 (q) the leading contribution of the 1PI diagrams which is the rst diagram of
g. 18.2.
+

=
q

1PI

1PI

1PI

+ . . .

Figure 18.1: Diagramatic expression for the exact photon propagator

1PI

+ ...

q
Figure 18.2: 1PI diagrams

18.1

General properties

Before computing i
2 (q) we can constrain the structure of the exact expression i (q).
The Ward identity q = 0 implies


= q 2 q q (q 2 ) ,
(18.1)

where has to be regular at q = 0. Furthermore, we can write the expansion of g. 18.1

as

i
i
i
+
i(q 2 q q )(q 2 )
+ ...
2
2
q
q
q2
(18.2)
i
i
i 2 2
2
= 2 +
(q ) +
(q ) + . . . ,
q
q2
q2
where
= i(q 2 q q )

q q

i

q2
q2

(18.3)

is a projection operator in that it satises

= .

(18.4)

Inserted into (18.2) one obtains

q q
i
i i
2

q
q
i 4 ,
+

+
.
.
.
=

2
2
2
q
q
q (1 )
q

57

(18.5)

where used the geometric series. The last term in (18.5) vanishes in amplitudes due to
the Ward identity and thus we have

iZ3
i
q q =
+ ... ,
)
q2

q 2 (1
where

Z3 =

1
1 (0)

(18.6)

(18.7)

is the rst term in the Taylor expansion of the exact photon propagator and Z3 the
residue of the pole at q = 0.

18.2

Computation of
2

Let us now compute the rst diagram in g. 18.2. Using momentum-space Feynman rules
one obtains
Z
d4 k  i(k/ + m) i(k/ + /q + m) 

2
tr 2

.
(18.8)
i2 = (ie)
(2)4
k m2 (k + q)2 m2
Using properties of the -matrices the numerator N can be simplied as14


N = 4e2 k (k + q) + (k + q) k (k(k + q) m2 ) .

Expressed in terms of Feynman integrals the denominator D reads

Z 1
1
1
1
=
dx
D= 2
2
2
2
2
2
k m (k + q) m
[(1 x)(k m ) + x((k + q)2 m2 )]2
0
Z 1
1
=
dx 2
,
(l )2
0

(18.9)

(18.10)

where
l = k + xq ,

= m2 x(1 x)q 2 .

(18.11)

Inserting (18.9) and (18.10) into (18.8) and dropping terms linear in l one arrives at
Z 1
Z
d4 l (2l l l2 2x(1 x)q q + (m2 + x(1 x)q 2 )

2
i2 = 4e
.
dx
(2)4
(l2 )2
0
(18.12)
This integral appears to be quadratically divergent in the UV and has to dimensionally
regulated. Following the same steps as before we Wick rotate l0 ilE0 , go to d-dimension
and use that under the integral l l = d1 l2 holds to obtain
i
2

14

= 4ie

dx

dd lE lE2 (1 2/d) 2x(1 x)q q + (m2 + x(1 x)q 2 )

.
(2)d
(lE2 + )2
(18.13)

Note that the expression also holds in d-dimensions.

58

Before we proceed note that regulated the divergent integral by a simple momentum
cut-o we would get a quadratic divergence
Z
d4 l
l2
2 ,
(18.14)
(2)4 (lE2 + )2
which violates the Ward identity in that q
2 6= 0. As we will see shortly this does not
happen in dimensionally regularization which is one of the reason for ist usefullness.
In problem 9.2 we showed
Z d
d lE
1 d 1 (1 d/2)
lE2
=
2
(2)d (lE + )2
(4)d/2 2 1d/2
(2)
1

d
=
(2 d/2) ,
d/2
2d/2
(4) 2(1 d/2)

(18.15)

where in the second step we used (u + 1) = u(u). In problem 9.2 we also showed
Z d
1
1
1 (2 d/2)
d lE
=
.
(18.16)
2
(2)d (lE + )2
(4)d/2 2d/2
(2)
Inserting both integrals into (18.13) we arrive at
Z 1


(2 d/2) 1

2
i2 = 4ie lim
dx
(m2 + x(1 x)q 2 ) 2x(1 x)q q
d/2
2d/2
d4 0
(4)

Z 1


(2 d/2) 1
2
2

,
dx x(1 x)
= 4ie q q q lim
d4 0
(4)d/2 2d/2
(18.17)
where in the second step we merely used (18.11). We see that, as promised,
2 is
proportional to the projection operator and thus the Ward identity is fulllled. The
quadratic divergence cancelled and the nal expression is only logarithmically divergent.
Comparing with (18.1) we conclude
Z 1
(2 d/2) 1
2
(18.18)
dx x(1 x)
lim
,
2 =
d4 0
(4)d/22 2d/2
which indeed has no pole at q = 0.

59

19
19.1

Lecture 19: The electron vertex

General considerations

So far we considered electron scattering at lowest order (see g.19.1). In this lecture we
compute a rst set of loop corrections to this process.


Figure 19.1: electron scattering at lowest order

We insist that the initial and nal state are unchanged and thus the loops occur
inside the diagramm. One set of corrections termed vertex corrections are depicted
in g.19.2.

+ ...

Figure 19.2: Leading correction to the electron vertex

In the notation we amputate the photon line and dene the vertex correction as
iM = ieu(p ) (p, p)u(p) . . .

(19.1)

where ie (p, p ) stands for the sum of all vertex diagrams, i.e. for the diagrams given
in g.19.2. We also see that at lowest order = holds. The corrections can be
constrained by the following considerations:
1. Lorentz invariance constrains the corrections to be of the form
= A + (p + p )B + (p p )C ,

(19.2)

where A, B, C can only depend on the Lorentz-scalars m, e or q 2 = (p p )2 =

2m 2p p .15
2. Current conservation j = 0 implies the Ward-identity
q = 0 .
15

(19.3)

Dependence on p/ und /
p can be neglected since we can always use u(p )p/ = u(p )m and /pu(p) =
mu(p). Thus the only non-trivial scalar is q 2 .

60

Eq. (19.3) implies

q = (p p ) A + (p p )(p + p ) B + (p p )(p p ) C = 0 .
|
{z
}
{z
}
|
p2 p2 =0

(19.4)

(p p )2

The rst term drops out of any amplitude since

u(p )(p/ p/ )u(p) = u(p )u(p)(m m) = 0 ,

(19.5)

as a consequence of u(p )(p/ m) = (/p m)u(p) = 0. Thus in (19.4) only the last term
proportional to C survives and therefore C = 0 has to hold. Finally, one can use the
Gordon-identity (proven in problem 10.1)
u(p ) u(p) =

1
u(p ) (p + p ) + 2iS q u(p) ,
2m

(19.6)

where S = 4i [ , ]. Eliminating the (p + p ) dependence in (19.2) via the Gordon

identity we have altogether the structure
(p , p) = F1 (q 2 ) +

i
m

S q F2 (q 2 ) .

(19.7)

F1 and F2 are called form factors which at lowest order are given by F1 = 1, F2 = 0.

19.2

Leading order results

Let us now consider the rst corrections = + given by the diagram of g. 19.3.
It corresponds to
ieu(p ) (p, p )u(p) =
Z

i(k/ + m)
d4 k
i i(k/ + m)

(ie ) 2
(ie )u(p) ,
u(p )(ie )
4
2
2
2
(2)
(k p) k m
k m2

(19.8)

where k = k + q, q = p p. Note that also this correction is logarithmically divergent

in the UV.

Figure 19.3: Leading order correction of electron vertex

Expressions of the type given in (19.8) we have already computed a few times in the
previous lecture. Therefore we only state the result here and refer for more details to
the literature [2]. Using the identies for -matrices derived in problem 9.3 the numerator
can be simplied as

u(p ) 2k/ k/ + k/ k/ + 4m(k + k ) m( k/ + k/ ) + ( 2)2 m2 u(p) , (19.9)

61

where we give the results straight away in 4 dimensions. The denominator is again
rewritten in terms of Feynman integrals as
Z 1
1
2
=
(19.10)
dxdydz
(x
+
y
+
z

1)
[(k p)2 2 ][k 2 m2 ][k 2 m2 ]
D3
0
where
D = x[k 2 m2 ] + y[k 2 m2 ] + z[(k p)2 2 ] = l2

(19.11)

= xyq 2 + (1 z)2 m2 + z2 .

(19.12)

with
l = k + 4yq zp ,

The next steps are well known by now:

1. change the integration variables k l,
2. go to d-dimensions,
3. Wick rotate,
4. express the d-dimensional integral in terms of -function according to problem 9.2.
In addition we can use the Gordon identity (19.6) to write the anl result in the form
(19.7). One nds
Z
 (2 )2 (2 d/2)
1
dxdydz (x + y + z 1)
F1 =
2 0
2(4)d/22 2d/2


(3 d/2) 2
2
2
2
q
[(1

x)(1

y)

xy]
+
m
[2(1

4z
+
z
)
+
(1

z)
]
,
+
3d/2
Z
 2m2 z(1 z) 
1
F2 =
,
dxdydz (x + y + z 1)
2 0
m2 (1 z)2 xyq 2
(19.13)
e2
2 2
2
2
where = 4 and = (1 z) m yxq + z . Note that F1 is UV divergent while
F2 is nite.

62

20

Lecture 20: Renormalization of QED

20.1

Superficial degree of divergence of a Feynman diagram

In this section we introduce a quantity D which represents the supercial degree of

divergence of a Feynman diagram. As we will see it is a useful guide in classifying the
divergent diagrams in QED. Let us rst introduce some notation:
Ne
N
Pe
P
V
L

=
=
=
=
=
=

number
number
number
number
number
number

of
of
of
of
of
of

external fermion lines

external photon lines
fermion propagators
photon propagators
vertices
loops

As an example consider the diagram in g. 20.1. It has

Ne = 4 ,

N = 0 ,

Pe = 6 ,

P = 5 ,

V = 10 ,

L=4.

(20.1)

Figure 20.1:
Now we dene the supercial degree of divergence D as
D = power of ks in the numerator power of ks in the denominator
= 4L Pe 2P ,
which holds since each loop adds a d4 k while a fermion progators adds a k/
propagator a k 2 .

(20.2)

and a photon

The UV-divergence can now estimated to be

D

for D > 0 ,

ln for D = 0 ,

(20.3)

none for D < 0 .

This counting is useful but often too naive. For example, the rst diagram in g. 19.2
has D = 0 but is nite. Similarly, the diagram 18.2 has D = 4 2 = 2 but is only
63

ln-divergent and the rst diagram of g. 17.1 has D = 4 1 2 = 1 and is also only
ln-divergent. Nevertheless D is a usefull quantity since for QED it can be expressed
entirely in terms of the number of external lines. This holds since the quatities dened
above enjoy further relations. First of all one has
L = Pe + P V + 1 ,

(20.4)

which holds since each propagator has a momentum integral but each vertex has a function. (The +1 expresses the overall momentum conservation.) For the example of
g. 20.1 one indeed has L = 8 + 5 10 + 1 = 4.
A second relation is

V = 2P + N = 21 (2Pe + Ne ) ,

(20.5)

which holds since out of each vertex comes one -line and two e-lines. The factor of
two accounts for the fact that a photon or electron propagator always connects two
vertices while an external line does not. For the example of g. 20.1 one has indeed
V = 2 5 = 10 = 12 (2 8 + 4). As a consequence of (20.4) and (20.5) one derives
D = 4L Pe 2P = 2P Pe 2Ne + 4 = 4 N 23 Ne .

(20.6)

The last expression is useful as it expresses D solely in terms of the external legs and
furthermore it shows that only diagrams with a small number of external legs can have
a UV divergence i.e. D 0.

20.2

Application to QED

As a consequence of (20.6) the supercially divergent diagrams can simply be listed

in g.20.2. The diagram a) is the vacuum energy which in theories without gravity is
unobservable. The diagramms b) and d) vanish due to Furrys theorem which states that
one cannot have an odd number of currents j in an amplitude. To show that one inserts
the unitary charge conjugation operators C and uses Cj C = j together with the
C|i = |i to arrive at
h|j 1 . . . j n |i = h|CC j 1 CC . . . CC j n |i = (1)n h|j 1 . . . j n |i ,

(20.7)

which thus vanishes for n odd. The diagram f) vanishes as it does not respect fermion
number. All other diagrams we look at in slightly more detail now.
The diagram g) can be Taylor-expanded around p = 0 to obtain
A0 + A1 p/ + A2 p2 + . . . ,

(20.8)

where in the coecient An terms like dd/pn (/k+1/pm ) appear. As a consequence D is lower
for higher terms in the expansion (20.8). Thus we conclude
A0 (D = 1) ,

A1 ln (D = 0) ,

An2 nite .

(20.9)

However in the chiral limit m 0 one can show A0 0 and hence one nds
A0 m ln ,
64

(20.10)

a)

D=4

b)

D=3

c)

f)

D=5/2

g)

D=1

h)

D=0

D=2

d)

D=1

e)
D=0

Figure 20.2: Diagrams with D 0.

i.e. only a ln-divergence occurs.
The diagram h) we already computed at leading order and did indeed see the lndivergence. Finally, the diagrams c) and e) are less divergent that esitimated by D.
We already argued argued in lecture 18 that due to the Ward identity = ( q 2
q q )(q 2 ). As a consequence can only be ln-divergent. The same argument implies
that the diagram e) has to be nite. Thus we are left with three only ln-divergent
diagrams which are the diagrams c), g) and h). In the next lecture we will see that these
divergence can be absorbed by redening the elds and coupling.

65

21

Lecture 21: Renormalized pertubation theory of

QED

In this lecture we discuss how the UV-divergence of QED can be absorbed in a redenition
of elds and couplings. This systematic procedure is sometimes called renormalized
perturbation theory.
So far we started from the bare Lagrangian
L = 14 F F + (i m0 ) e0 A ,

(21.1)

with the bare parameters m0 , e0 . In lecture 17 we argued that the fermion propagator
has a single pole at /p = m where m is the physical or renormalized mass with a residue
Z2 (cf. (17.9)). Similarly we argued in lecture 18 that the photon propagator has a single
pole at q 2 = 0 with a residue Z3 .
Z2 and Z3 can be absorbed in a new set of renormalized eld variables r , Ar by the
redenition
p
p
= Z 2 r ,
A = Z3 Ar .
(21.2)

Inserted into (21.1) one obtains

L = 14 Z3 Fr Fr + Z2 r (i m0 ) r e0 Z2

p
Z3 r r Ar .

(21.3)

In addition one denes physical or renormalized parameters m, e as

Z2 m0 = m + m ,

e0 Z2

and splits

p
Z3 = eZ1 ,

Z1,2,3 = 1 + 1,2,3 .

(21.4)
(21.5)

This split is at the moment arbitrary but as we will see shortly it will be dened such
that the UV-divergence disappears from the physical quantities. Inserted into (21.3) we
arrive at
L = 14 Fr Fr + r (i m) r er r Ar
(21.6)
1
3 Fr Fr + r (i2 m ) r e1 r r Ar .
4
The prescription now is to do perturbation in e (instead of e0 ) with
Lagrangian (21.6) instead of (21.1). For this we need new Feynman rules.
in (21.6) leads to the exact same Feynman rules as given in table 13.0.
second line leads to new vertices which are called counterterms and which
=

i( q 2 q q )3 ,

i(/p2 m ) ,

ie 1 .

66

the redened
The rst line
However the
are given by

(21.7)

Now one denes the split given in (21.4) and (21.5) by the requirement that the
renormalized elds r , Ar have propagators like a free eld with the renormalized mass
m being the position of the pole. In other words
Z
i
r (0)}|i =
d4 xeipx h|T {r (x)
=
+ ...
(21.8)
/p m
Now we need to redo the perturbation theory for the Lagrangian given in (21.6). This
leads to same diagrams as before plus new diagrams involving the counterterms. One
thus has
= i = i( q 2 q q )(q 2 ) ,
1PI

1PI

i(p/) ,

ie (p, p ) ,

(21.9)

where in this notation now all contribution of counterterms are included.

In lecture 17 we already computed the full fermion propagator in terms of the 1PI
contribution with the result
=

i
i

=

p/ m (p/ = m) d
(p/ m) + . . .
/p m
d/
p p
/=m

(21.10)

Comparing to (21.8) we obtain the rst two renormalization conditions

(i) (/p = m) = 0 ,
(ii))

d 
=0,
d/p /p=m

(21.11)

which x the physical mass m and the residue of the electron-propagator. Similarly we
computed in lecture18
i
1
q
= 2
+ ...
(21.12)
q 1 (0)
This implies the third renormalization condition

(iii) (q 2 = 0) = 0 ,

(21.13)

which xes the -propagator. Finally in lecture 19 we computed

=

ie (p, p) .

(21.14)

In order to x the electric charge e we demand

(iv) (p p = 0) = ,

(21.15)

which is the forth renormalization condition. Thus we managed to derive four conditions
for the four counterterms 1,2,3,m .

67

Let us see how this works explicitly at one-loop. The one-loop contribution to was
computed in lecture 16 and termed 2 . Now we need to add the contribution of the
second counterterm in (21.7). Inserted into (21.11) we obtain
(i)
(ii)

i2 (p/ = m) + i(m2 m ) = 0 ,
i

d 
+ i2 = 0 ,
d/p /p=m

(21.16)

2 (0) 3 = 0 .

(21.17)

The one-loop contribution of was computed in lecture 18 and termed

2 . Here we
need to add the rst counterterm in (21.7) and inserted it into (21.13) we obtain

Finally, was computed in lecture 19. Now we need to add the contribution of the
third counterterm in (21.7) and inserted it into (21.15) we obtain
F1 (p p = 0) + 1 = 1 .

(21.18)

Using the explicit expressions computed previously one sees that all UV-divergent
contributions are absorbed into the counterterms 1,2,3,m . This means that the physical
parameters are nite (and small) while the bare parameters carry the innities.
Physically this implies that a theory depends on the distance scale it is looked at
or rather being probed at. Due to vacuum polarization where virtual e+ e pairs screen
a bare electric charge the physical charge is measured at a dierent value. Such scale
dependent couplings have indeed been measured and will be subject to further studies in
the second part of this course.

68

22

Lecture 22: IR-divergences

In this section we discuss IR-divergences but for a more exhaustive treatment we have
to refer the reader to the literature (see for example [2]).
In lecture 19 we computed the leading order correction to the electron vertex. The
result for the two form-factors F1,2 was given in (19.13). In the previous lecture we
learned that in order to properly treat the UV-divergence we have to add 1 to F1 which
is determined by (21.18). So altogether we have
F1 = 1 + F1 (q 2 ) + 1 = 1 + F1 (q 2 ) F1 (0) .

(22.1)

In this lecture we study a completely dierent divergence of the Feynman diagram 19.3
which occurs when the momentum of the photon in the loop vanishes. This divergence is
called an IR divergence and can be seen in the expression (19.13) in the limit z 1, x
0, y 0, 0 such that 0. The parameter was introduced as a small mass of
the internal photon precisely to regulate this divergence. More precisely we substituted
1
1

2
(k p)
(k p)2 2

corresponding to

P hoton P hoton + z2 . (22.2)

In the following we only keep the leading singular terms to arrive at


Z
Z 1z  2
2m2
q 2m2
1

,
dz

dy
F1 (q) F1 (q) F1 (0)
2 0
(q)
(0)
0

(22.3)

where (q) = (1 z)2 m2 + 2 q 2 y(1 z y). Substituting

z =1w ,

y = (1 z) = w ,

dz = dw ,

dy = wd ,

w [0, 1] ,

[0, 1]
(22.4)

we obtain



q 2 2m2
2m2
d(w )
d

(m2 q 2 (1 ))w 2 + 2 m2 w 2 + 2
0
0
 2

 2

Z
q 2 2m2
m q 2 (1 ) + 2
m + 2
1
d 2
ln
+ 2 ln
.
=
4 0
m q 2 (1 )
2
2

F1
2

1
2

In the limit 0 the details of the numerator inside the ln do not matter and we can
take it eectively out of the integral to obtain


2
2

q
or
m

2
F1 = fIR (q ) ln
,
2
2
(22.5)

Z 1 
2
2
m

q
/2
1 .
with
fIR (q 2 )
d
m2 q 2 (1 )
0
For later use let us compute
1
f
lim
IR (q ) =
q 2
2
2

1
0



1
m2
d 2 2
ln 2
m /q (1 )
q
69

(22.6)



Figure 22.1: Feynman diagrams for Bremsstrahlung in leading order

In this limit the Sudakov double logs appear and we have

  2
q2
q

F1 (q ) ln 2 ln 2 .
2
m

(22.7)

In the following it is argued that the IR divergence are canceled by soft photons in
the nal state (Bremsstrahlung) depicted in g. 22. The contribution to the iM from
these diagrams is


(/p k/ + m)

(k) u(p)
iM = e
u(p ) M0 (p , p k)
(p k)2 m2
(22.8)



/
/
(
p
+
k
+
m)
e
u(p ) (k)
M0 (p + k, p) u(p)
(p + k)2 m2
where M0 is the matrix element where the external photon line has been cut, i.e. exactly
the diagram which has been computed in the last lectures. Assuming that the external
photon is soft corresponds to the limit
|~k| |~p ~p|

(22.9)

In this limit we also have

1. M0 (p , p k) M0 (p , p) M0 (p + k, p),
2. we can ignore k/ in the numerator,
3. (p/ + m) u(p) = 2p u(p)
4. u(p ) (p/ + m) = u(p )2p
5. (p k)2 m2 = 2pk, (p + k)2 m2 = 2p k
Inserted in (22.8)) one obtains

iM = e
u(p )M0 u(p)

p p

p k
pk

and thus the dierential cross section becomes


2
Z |~k|max 3
d k 1 X 2  p p 

e 
d(p p + ) = d(p p )
(2)3 2k 0
pk
pk 
0
70

(22.10)

which is divergent for k 0.

The sum in (22.10) can be simplied as




 2


X   p

p
p
p X p
p



 p k pk  =
k
k
p
pk
p
pk

2p p
m2
m2
=

(p k)(pk) (p k)2 (pk)2

=
where we used

1
I(p, p , cos()) ,
2
k
=

(22.11)

(22.12)

I(p, p , cos()) is an angle-dependent contribution whose precise form we do not need

(see [2] for an explicit expression). Inserting (22.11) into (22.10) yields

d(p p + )

q 2

Z
d3
k2
I(p, p , cos())
dk 3
k
4

 2Z
q
d3

I(p, p , cos())
d(p p )0 ln 2

4
 2 

q
q2

d(p p )0 ln 2 ln 2 ,
(22.13)

e2
d(p p )0 2
4

|~
q|

where again the photon mass was introduced. We see that the same Sudakov double
log appears albeit with opposite sign.
Now the claim ist that this cancellation persists for arbitrary q and to all orders (see
[2]). Thus Feynman diagrams can well be IR-divergent but the physical observables are
not.

71

23

Lecture 23: The LSZ-reduction formalism

In this lecture we sketch the proof for the earlier promised LSZ-relation. For 2 n body
scattering it states
Z
d4 x1 . . . d4 xn dyA dyB eip1 x1 . . . eip1 xn eikA yA eikB yB h|T {(x1 ) . . . (xn )(yA )(yB )}|i
pi Ep~i

n
2
kA E~kA  Y
i Z  Y i Z 

hp1 . . . pn |S|kA kB iamputated ,

p2 m2
k 2 m2
i=1 i
A=1 A

(23.1)

which relates the time-ordered product of eld operators to the S-matrix.

Let us sketch the proof in the following. Start with Fourier-transforming one leg and
consider
Z
A = d4 x1 eip1 x1 h|T {(x1 ) . . .}|i
(23.2)
In order to take of the time ordering split the time integration into
Z T
Z
Z T+
Z
0
0
0
dx0 ,
dx +
dx +
dx =
T

T+

(23.3)

and choose T+ x0i6=1 T . In the rst integral x0 is the latest time and one can insert
a complete set of states.
X Z d3 q
1
|q~ihq~| ,
(23.4)
1 = |ih| +
3
(2) 2Eq~()

~
0

where the sum is over all zero-momentum states and |q~i denotes all boasts of the zeromomentum states |~0 i . Using
h||q~i = h|(0)|0 ieiqx |q0 =E~q ,

h|(x)|i = 0 ,

(23.5)

one obtains
A=

XZ
0

dx
T+

d3 q
1
ip0 x0 iE~q x0
e
e
h|(0)|0i
(2)3 2Eq~()
Z
d3 x1 ei(~p~q)x hq~|T {(x2 ) . . .}|i

(23.6)


X 1 Z
0 i(p0 Ep~ )x0
=
dx e
h|(0)|0ihp~|T {(x2 ) . . .}|i .
2Ep~ T+
0

For the integral we have

Z

T+

i(p0 E

dx0 e

p
~

)x0


0
0
ei(p Ep~ )x 
=
 ,
i(p0 Ep~ ) 
T+

72

(23.7)

which has a pole at p0 = Ep~ . Therefore we obtain

p0 Ep~
i Z
A
hp~|T {(x2 ) . . .}|i .
p2 m2
We can repeat the same analysis for region III with the result

p0 Ep~
i Z
h|T {(x2 ) . . .}|p
A

~ i .
p2 m2

(23.8)

(23.9)

In region II there is no pole since the integral is taken over a bounded region.
One can now also repeat the analysis for all particles using sharply concentrated wave
packets around xi in order to separate them in space. For details we refer to [2].

73

Acknowledgements
These lecture notes were written with the help of Andreas Bick, Lukas Buhne, KarlPhilip Gemmer, Michael Grefe, Jasper Hasenkamp, Jannes Heinze, Sebastian Jakobs,
Thomas Kecker, Tim Ludwig, Ole Niekerken, Alexander Peine, Christoph Piefke, Sebastian Richter, Michael Salz, Bjorn Sarrazin, Hendrik Spahr und Lars von der Wense.

References
[1] C. Itzykson, J.-B. Zuber, Quantum Field Theory, McGraw Hill
[2] M. Peskin, D. Schroder, An Introduction to Quantum Field Theory, Westwood.
[3] S. Pokorski, Gauge Field Theories, Cambridge University Press
[4] P. Ramond, Field Theory, Benjamin/Cummings
[5] M. Sredniki, Quantum Field Theory, Cambridge University Press
[6] S. Weinberg, Quantum Theory of Fields I, Cambridge University Press
[7] A. Zee, Quantum Field Theory in a Nutshell, Princeton University Press

74