You are on page 1of 84

Six lectures on

relativistic quantum scattering

Thomas Van Riet


Instituut voor Theoretische Fysica, K.U. Leuven,
Celestijnenlaan 200D B-3001 Leuven, Belgium 1

Abstract

These lecture notes introduce basic theoretical concepts in quantum scatter-


ing theory and relativistic quantum mechanics. The aim is to understand the
simplest processes in quantum electrodynamics without the need of quantum
field theory.

1
thomas.vanriet @ fys.kuleuven.be

1
Contents
1 Lecture 1: special relativity 6
1.1 The mathematics of Lorentz transformations . . . . . . . . . . . . . . . . . 6
1.2 The physics of Lorentz transformations . . . . . . . . . . . . . . . . . . . . 9
1.3 Energy, momentum and collisions in relativity . . . . . . . . . . . . . . . . 10
1.4 Maxwell theory in tensor language . . . . . . . . . . . . . . . . . . . . . . . 15

2 Lecture 2: Relativistic Quantum Mechanics (RQM) 21


2.1 Probability currents in quantum mechanics . . . . . . . . . . . . . . . . . . 21
2.2 Non-relativistic quantum mechanics . . . . . . . . . . . . . . . . . . . . . . 22
2.3 A relativistic wave equation: the KleinGordon field. . . . . . . . . . . . . 23
2.4 Remarks on quantum field theory . . . . . . . . . . . . . . . . . . . . . . . 25
2.5 The spin of a field/particle . . . . . . . . . . . . . . . . . . . . . . . . . . . 25
2.6 The Dirac equation. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 27

3 Lecture 3: RQM continued 30


3.1 Transforming a Dirac spinor . . . . . . . . . . . . . . . . . . . . . . . . . . 30
3.2 Probability currents . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 32
3.3 Dirac action and the gauging principle. . . . . . . . . . . . . . . . . . . . . 33
3.4 Plane wave Dirac solutions . . . . . . . . . . . . . . . . . . . . . . . . . . . 34

4 Lecture 4: Decay rates 39


4.1 Decay rates . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 39
4.2 Golden rule . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 39
4.3 density of states . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 40
4.4 Relativistic treatment . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 43

5 Lecture 5: Cross section and amplitudes 47


5.1 Definition of differential cross section . . . . . . . . . . . . . . . . . . . . . 47
5.2 Relativistic treatment . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 49
5.3 Centre-of-mass scattering . . . . . . . . . . . . . . . . . . . . . . . . . . . . 50
5.4 The S-matrix and Feynman diagrams . . . . . . . . . . . . . . . . . . . . . 51
5.5 A simple toy-theory . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 52
5.6 Towards a derivation of Feynman rules in QED . . . . . . . . . . . . . . . 55
5.7 The Feynman rules . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 57

6 Lecture 6: Electron-muon scattering 60


6.1 Workout 1: Summing over spins . . . . . . . . . . . . . . . . . . . . . . . . 60
6.2 Workout 2: Making traces . . . . . . . . . . . . . . . . . . . . . . . . . . . 62
6.3 Workout 3: non-relativistic limit . . . . . . . . . . . . . . . . . . . . . . . . 63
6.4 Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 64

2
A What is a tensor? 65
A.1 Vectors and one-forms . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 65
A.2 (p,q) tensors . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 66

B Photon Helicity 69

C The spinor transformation matrix 69


C.1 Pauli matrices and Lorentz transformations . . . . . . . . . . . . . . . . . 69
C.2 Spatial rotations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 71

D Asymptotic series 72

E The matrix element for electron-muon scattering 73

F Trace over producty of three gamma matrices 74

G Useful formulas 74

3
Introduction
Most information we have gathered from experiments with elementary particles comes from
1) scattering events (in colliders), 2) from particle decays or 3) from studying (excitations
of) bound states. From the point of view of quantum mechanics one can regard this
as transitions between 1) free (continuous) energy eigenstates, 2) discrete (bound) states
and free (continuous) states and 3) discrete (bound) states only. These lecture notes will
primarily focus on scattering.
Since particle physics experiments take place at high velocities most of the scattering
events useful to particle physics are relativistic. These lecture notes will therefore contain
an introduction to relativistic quantum mechanics and (classical and quantum) scattering
theory.
Equipped with these basics we will tackle some processes in quantum electro dynamics
(QED). The latter is a quantum field theory (QFT) and that is beyond the scope of this
text. Nonetheless quite some physics can be derived from relativistic quantum mechanics
without the need of knowing QFT. The interested student is invited to take up courses
on QFT in his master studies. From a reductionist point of view one can say that all
non-gravitational processes in nature seem perfectly described by the theory of quantum
fields (known as Yang-Mills theory).
I have written these lectures notes in a specific form. Instead of working out all the
steps in the derivations, I have cut them in various pieces and let you solve each step
yourselves as an exercise. This is why the lecture notes contain the boldface statements
Ex: Check this
(or something like that) throughout. You should solve all these exercises. If you do so
successfully, you most likely will do well on the exam. If you cannot make (most of) them
that means you are most likely not enough prepared to do the exams.
Since many particle physics experiments involve relativistic speeds and energy scales
much of the physics is relativistic. Since this course will be restricted to quantum electrody-
namics we only introduce the necessary topics to cover that. But it already brings us quite
far since it requires the concept of Dirac spinors, the Dirac equation, the gauging principle,
etc. The course content is organized as follows. Lecture 1 contains a review on special rel-
ativity with an emphasis on what is necessary for doing particle physics computations: the
four-vector notation, tensors, Einstein notation, relativistic collision theory and last but
not least a relativistic formulation of Maxwell theory. Having treated electromagnetism
(photons) relativistically we will need to do the same for electrons and positrons and we
will have to do this quantum mechanically. So lecture 2 and 3 introduce the necessary
basic concepts in relativistic quantum mechanics. In practice this means we need to make
the Schrodinger equation into a Lorentz-covariant (invariant) wave equation. How to do
this strongly depends on the spin of the particle (wave). The easiest are spin 0 particles
described by the KleinGordon equations. This is however just a warm-up for introducing
the Dirac equation, which is the wave equation for spin 1/2 particles such as electrons. The
Dirac equation itself comes with quite some new technical/mathematical concepts such as

4
Dirac spinors, Clifford algebra, etc. But without this one cannot proceed. So roughly 3/5
of this course is an introduction to relativistic quantum mechanics. In lecture 4 we then
finally move towards particle physics concepts such as decay rates, the Golden rule, den-
sity of states and we make these concepts relativistic. Similarly in lecture 5 we present a
relativistic treatment of scattering cross sections and we introduce the concept of Feynman
diagrams. Everything then culminates in lecture 6 where we derive the so-called Feynman
rules for tree-level processes in quantum electrodynamics.
The pedagogical approach I follow is based on the idea that the abstract Feynman rules
for associating a quantum probability (called scattering amplitude) to a Feynman diagram
can actually be understood by bachelor students from their previous knowledge in quantum
mechanics and what is introduced in this course. This does not suffice for understanding all
of particle physics because a full treatment requires quantum field theory (QFT), something
that is beyond the scope of a bachelor education. Nonetheless all tree-level diagrams in
QED can be understood without QED. In case you will follow a course on QFT in your
masters where you will re-derive the Feynman rules using QFT then do not worry that
this course was in vain. Only when you understand how simple quantum mechanics can
reproduce the same results have you reached a genuine understanding.
Finally, these notes are based on chapters in the following textbooks [13] and include
some extras, but not much.

Remark: these notes are under construction. I am interested in all feedback, ranging
from typos to actual content. Feel free to contact me via email for this (thomas.vanriet @
fys.kuleuven.be).

5
1 Lecture 1: special relativity
1.1 The mathematics of Lorentz transformations
Consider a Cartesian coordinate system on R3 , denoted by x, y, z. Distances d between
two points are measured using Pythagoras theorem

d2 = (x1 x2 )2 + (y1 y2 )2 + (z1 z2 )2 . (1.1)

In relativity we extend this to four-dimensional Minkowski space-time by including time t.


In canonical coordinates, denoted x , where = 0, . . . 3:

x0 = ct , x1 = x , x2 = y , x3 = z. (1.2)

The invariant distance between two space-time points is now

d2 = (x01 x02 )2 (x11 x12 )2 (x21 x22 )2 (x31 x32 )2 . (1.3)

We follow the so-named mostly minus conventions in particle physics2 . Different space-
time points that are related causally have d2 > 0. We call them time-like separated. The
vector V = x1 x2 is then called time-like. There are also space-like distances (or vectors)
when d2 < 0, or null-like or light-like distances (or vectors) when d2 = 0. Space-time points
that are light-like separated can be connected with a light ray whereas time-like separated
points can be connected with a curve followed by a massive particle. Space-like separated
points cannot communicate with each other. Let us take x2 in the origin to simplify
notation and denote x1 = x . In the Einstein notation we can write

d2 = x x , (1.4)

where
1 0 0 0
0 1 0 0
=
0 0 1 0 ,
(1.5)
0 0 0 1
and repeated indices are summed over.
Lorentz transformations are those linear transformations that preserve the Minkowski
length d:
x0 = x , x x = x0 x0 . (1.6)
In matrix language we have that Lorentz transformations are those matrices that obey

T = . (1.7)

Ex: show this.

2
Modern texts on general relativity or quantum field theory more often use mostly plus conventions.

6
Mathematically this corresponds to the definition of the group O(3, 1). In other words,
Lorentz transformation are represented by matrices that are in O(3, 1).

Ex: show that the group of rotations of the spatial coordinates x, y, z is part
of the Lorentz group.

A well known Lorentz-transformation is the boost in the x1 -direction



0 0
0 0
= 0
, (1.8)
0 1 0
0 0 0 1

where
1
=p . (1.9)
1 2
Ex: show that this is indeed a Lorentz transformation and write down what
the matrices are that correspond to boosts in the x2 and x3 direction.

Ex(advanced): What is the matrix that corresponds to a boost in the x2 + x3


direction?

If we identify = v/c with v the velocity of an observer moving in the x1 -direction


then we obtain the usual Lorentz transformations that can be found in any text book.

Ex: show this.

In Einstein notation, the bilinear form 3 can be used to raise and lower indices

x = x . (1.10)

Hence
d2 = x x . (1.11)
Quantities that are invariant under Lorentz transformations are then typically such that
all lower-indices are contracted with upper indices.
Indices that are summed over are called dummy indices and they have the property
that their labeling can be changed. For instance

v a a = v b b . (1.12)
3
In the language of tensors , although represented as a matrix, is not a linear map but a bilinear form.
In the appendix I explain these words. The upshot is that a square matrix is typically used for linear maps
from a vector space to itself. These are called (1, 1) tensors. Bilinear forms are instead bilinear maps that
map two vectors to a real number. These are called (0, 2) tensors and since they carry two indices they
can also be represented by a matrix.

7
This propertyalthough obviousis useful in computations since often the certain indices
already occur somewhere else in the same equation and by changing the dummy index
labelling one can remove ambiguities in the notation.
We also define an tensor with indices both upstairs

= . (1.13)

What lies behind this is that the inverse of a bilinear form is denoted with indices upstairs,
but 1 = .

Ex: Check the self-consistency of this notation. For example raise both
indices of with to find again.

This leads to
= , (1.14)
with the Kronecker delta.
Finally the bilinear form can also be used to define the inner-product between two
different four-vectors v, w as follows

v w = v w = v 0 w0 ~v w
~, (1.15)

where we introduced the 3-dimensional Euclidean vector inside a four-vector using the
standard arrow-notation ~v . Also this inner product is invariant under Lorentz transforma-
tions.
Under a Lorentz-transformation a four-vector changes as

v 0 = v . (1.16)

In tensor language a vector is called a (1, 0) tensor or contravariant vector. There is also a
notion of a covariant vector or one-form or (0, 1) tensor. This is an object with one index
down w and transforms as follows

w0 = w . (1.17)

where
. (1.18)
Ex 1.6 : check that, for a given contra-variant vector v the following object
v = v transforms indeed as a covariant vector.
...
Now you are ready for the definition of a general (p, q) tensor T11...qp . This is an object
that transforms as follows

T01 1...
...p
q
= 1 1 . . . p p 11 . . . qq T11...
...p
q
(1.19)

8
This might seem a painful formula to use in practice but it is not. The rule is simple. For
every index up you add a Lorentzmatrix for a contra-variant vector (1.16) and for every
index down you add a Lorentz matrix for a covariant vector (1.18).

Ex: Show that and when regarded as (0, 2) and (2, 0) tensors respec-
tively, have the special property of being invariant.

Ex: For a given (1, 1) tensor T , show that the trace, T is invariant.

Ex: For a given (p, q) tensor T , show that the contraction over the first
upper index and first lower index makes the object a (p 1, q 1)-tensor.

Ex: Demonstrate the consistency of the Einstein notation: if the indices on


a (p, q) tensor are lowered and raised such that now p0 indices are upstairs and q 0
downstairs (with p + q = p0 + q 0 ) then show that this new object is a (p0 , q 0 ) tensor.

1.2 The physics of Lorentz transformations


Consider an observer that moves away from an observer at rest with respect to some
inertial frame S. The moving observers frame is denoted S 0 . The respective coordinates
are denoted x and x0 . We will briefly recall the standard special relativity effects of time
dilation, length contraction and the law for addition of velocities. You should have seen
this before.
If the motion is in the x1 -direction (and if not we rotate the frame), the relation between
the coordinates is (y and z are the same)
v
t0 = (t x) , (1.20)
c2
x0 = (x vt) . (1.21)

Or written differently
v 0
t = (t0 + x ), (1.22)
c2
x = (x0 + vt0 ) . (1.23)

To describe time dilation we envisage a clock at the origin in S 0 and we watch it from S.
The clock has x0 = 0 always and we find t = t0 . Since 1 we conclude that time in
S 0 runs slower (When the clock in S 0 passes 1 second it seem like seconds in the S-frame).
Similarly to describe length contraction we look at a rod positioned along the x axis
which begins at x0 = 0 and ends at x0 = L0 . To measure the distance in the S-frame we
have to measure the beginning and the end position at the same time t. We measure at
t = 0 then we have x0 = x and hence L = 1 L0 .

9
The law for addition of velocities is slightly more work. We consider a particle in the
S 0 -frame moving at velocity w0 along the x0 -axis (x-axis). What is the velocity of this
particle as measured from S? The trajectory of the particle in S 0 -frame is
x0 = w0 t0 . (1.24)
From the Lorentz transformations we find that this straight line in x, t coordinates is given
by
vw0 0
t = (1 + )t , x = (w0 + v)t0 . (1.25)
c2
This gives us that the velocity in the S-frame
x w0 + v
w = 0 . (1.26)
t 1 + vw
c2

In particular when w0 = c we also have w = c as it should.

1.3 Energy, momentum and collisions in relativity


If we define notions as kinetic energy and momentum in special relativity we have to
consider what might change. What counts is that in a closed system energy and momentum
are conserved. This should be our guiding principle to derive the new notions, taking into
account that this conservation law should hold in any inertial frame. In what follows I
closely follow [2] on this.
Classical momentum p~ is velocity times mass, p~ = m~v . There are multiple ways to
extend this to special relativity, at first sight. For instance which clock should one use?
An invariant choice of clock, that all observers can agree on, is to use the clock of the
particle under observation. This is clear by its definition in words. To see mathematically
why the time on the moving clock is an invariant time coordinate, we note that it
can be found as follows4 . Consider the clock motion in Minkowski space and denote its
displacement by x. Then the clock time that has passed is found via the equation:
c2 2 = x x . (1.27)
Of course for particles traveling at the speed of light we cannot use this definition anymore.

Ex: explain why not.

We define proper velocity ~ 5 then as the distance x traveled in the observers frame
divided by the time interval measured in the particles clock. In the limit of infinitesimal
4
Again we rely on some basic postulate of special relativity that states that the proper time, measures
by a clock, equals the length of the curve followed by the clock in 4D Minkowski space-time.
5
I follow a unfortunate notation by Griffiths since the symbol is already used for the Minkowski
bilinear form. The hope is that the reader will not be confused since the bilinear form carries two indices
and the vector carries one index. So from notation and context it should be clear.

10
displacements we find
d~x
~ = , (1.28)
d
and
d = 1 dt . (1.29)
Interestingly this spatial 3-vector is part of a 4-vector as follows
dx
= . (1.30)
d

Ex: why is this a 4-vector? Is dx dt
also a 4-vector?
The Minkowski norm of this 4-vector is

= c2 . (1.31)

Ex: Show this.

This allows us to define momentum in special relativity. We immediately give the


4-vector definition
p = m , (1.32)
such that
p p = m2 c2 . (1.33)
The spatial 3-vector inside the 4-vector equals

p~ = m~v , (1.34)

and indeed becomes the classical momentum in the classical limit v/c 0.
The temporal component is
p0 = mc . (1.35)
What is its interpretation? For that we Taylor expand in the dimensionless ratio (v/c),
to find:
1 3
p0 c = mc2 + mv 2 + mc2 v 4 + . . . (1.36)
2 8
Ex: Verify this.

The second term we recognize as the kinetic energy. In fact, thought experiments led
Einstein to declare the quantity as the relativistic energy

E = p0 c = mc2 . (1.37)

This would suggest/imply that the first term mc2 is some energy present at zero velocity,
we call it the rest energy, which can indeed be released in nuclear interactions. By now it
has turned into a famous equation

E = mc2 . (1.38)

11
It is the lecturers opinion that this equation is quite unfortunately quoted since 1) it is
only valid at zero velocity and 2) it is a non-covariant equation. Most likely Einstein should
be remembered for his ideas on covariance and paying respect to Einstein would suggest
to write down covariant or invariant equations. Such an equation is6
E
p p = ( )2 p~2 = m2 c2 , (1.39)
c
and still fits on a T-shirt.
The kinetic energy T of a particle is accordingly defined as
T = mc2 ( 1) (1.40)
Finally our definition of momentum allows to take the formal limit m 0. From (1.39)
we then find
E = |~p|c . (1.41)
So all energy is kinetic in some way, but classically a massless particle carries no kinetic
energy. For the case of a photon we can find |~p| from the relation E = ~, with
the frequency of the wave associated with the photon particle (remember wave/particle
duality).
The crucial motivation for this definition of 4-momentum stems from the fact that
it is experimentally observed to be conserved. Mathematically it originates from one of
the postulates of special relativity that states that particles that are not subject to an
external force follow a geodesic curves in four-dimensional space-time. A geodesic curve in
Minkowski space is a straight line. For a massive particle this straight line has to be within
the lightcone, ie, be time-like. A geodesic in Minkowski space-time fulfills the equation
d
= 0. (1.42)
d
This implies the conservation of 4-momentum7 .

What is also useful about this definition of four-momentum is momentum conservation


is valid in all inertial frames. Consider a collision between two particles, A and B, with
initial momenta pA , pB and final momenta p0A , p0B . The conservation equation implies
pA + pB = p0 0
A + pB . (1.43)
6
There are some other famous equations in physics that are interpreted wrongly. A good example is
R = U/I being quoted as a law, whereas it is a definition, not a law. Ohms law should be R = cst. Which
in fact is wrong. Materials that are close to obeying Ohms law are called Ohms resistances. Again a
definition. Most likely Ohm only invented definitions, not laws. For daily use (i.e. the electricity network
in your own home) R = cst is a good first-order approximation. (If you ever find yourself teaching in high
school or university, please get it right.)
7
In curved four-dimensional space (we do not define what curved space-time is here, intuition is
sufficient) geodesics are not straight lines and this is interpreted as the force of gravity. This is the content
of general relativity. The way space-time is curved is dictated by the matter that sits inside space-time
and the so-called Einstein equations relate the curvature to the energy-momentum tensor of the matter
content. Hence, in relativity gravity is not a force, it is a consequence of space-time curvature.

12
Clearly this equation, once valid in one frame, is valid in all frames.

Ex: Explain why.

Now we turn to collisions. We can think of the standard process that two particles, A
and B collide producing two particles C and D. Symbolically

A+B C +D. (1.44)

We can also change the number of particles, and that is left for the reader.
Classically we have the following rules (in absence of an external force field):

1. Mass is conserved: mA + mB = mC + mD .

2. Momentum is conserved p~A + p~B = p~C + p~D .

Collisions are then usually divided into 3 classes depending on whether kinetic energy is
conserved (elastic collisions), decreases (sticky collisions) or increases (explosive collisions).
In relativity we instead have the following single rule

1. The total 4-momentum is conserved. So this implies the conservation of total energy
and spatial 3-momentum.

The same distinction between collisions depending on the rise, fall or constant-ness of the
kinetic energy remains. But we have to keep in mind that total energy (kinetic + rest
energy) cannot change. So to compensate, the masses can change in non-elastic processes.
This is typically not observed in daily, classical, life since the compensation effect at low
energies causes mass differences too small to measure.
Finally, when we perform computations of scattering events there are two typical in-
ertial frames that exist. The laboratory frame and the center of mass (CM) frame. The
first frame is defined such that it is the frame in which the target is at rest. The second
frame is defined as the frame in which the total 3-momentum (sum of target momentum
and beam particle momentum) is zero.

Ex: why does the CM frame always exist?

Exercises on collisions
Questions:

1. A pion at rest decays into a muon and a neutrino. What is the speed of the muon?

2. Consider the collision of two protons into 3 protons and an anti-proton. What is the
threshold kinetic energy in the laboratory frame for the incident proton.

13
3. Consider two identical particles, each with mass m and kinetic energy T 0 (in the CM
frame) collide head on. What is their relative kinetic energy? (Ie in the rest frame
of one of the two particles).
4. All interactions in QED come from gluing the single Feynman diagram depicted in
figure 1, describing (depending on its orientation) an photon materializing into an
electron-positron pair. Demonstrate that this cannot occur as such.

Figure 1: The basis QED vertex. Depending on its orientation it describes various phenomena.
This process, by itself, can never take place, regardless of the orientation.

Brief answers8 :
1. Conservation of 4-momentum implies: p = p p as a 4-vector equation. Now take
the (Lorentz) square and use p2 = m2 c2 to find 2m E = m2 c2 + m2 c2 , where we
used that the neutrino mass is zero to good approximation. Similarly, we find from
squaring p = p p that 2m E = (m2 m2 )c2 . Since E = |~p c| = |~p c| and
m2 m2
2m E = (m2 m2 )c2 we get that |~p | = 2m



c. Now comes an important trick.
In order to compute velocity from relativistic momentum, the easiest is to observe
that p~ = m~v and E = mc, hence division gives ~v = p~c2 /E . In the end we find
m2 m2
v = c. (1.45)
m2 + m2

2. In the CM frame, the problem is easier. Since the threshold energy exactly cor-
responds to the situation in which all of the produced particles are at rest after
production. The total 4-momentum in CM frame is then (p0T ) = (4mc, 0, 0, 0). The
prime denotes the CM frame and absence of prime will denote lab frame. In the lab
frame the total 4-momentum is easiest computed before the collision (and of course
it remains the same after): pT = (mc, 0, 0, 0) + (E/c, |~p|, 0, 0), where we took the mo-
tion along the x-direction, without loss of generality. Since the square of a 4-vector
is frame independent we have
(E/c + mc)2 |~p|2 = 16m2 c2 , (1.46)
which leads to E = 7mc2 .
8
The answers below are done using the fastest way. Typically there are multiple ways to find the
answers. Some can take pages of work. So it is useful to see some tricks to simplify.

14
3. In the CM frame we have (pT )0 = (2E 0 /c, 0, 0, 0). In the lab frame we have pT =
(E/c + mc, p~). Squaring both should be the same, such that 4E 02 /c2 = (E/c + mc)2
|~p|2 . We can eliminate |~p|2 from E 2 /c2 |~p|2 = m2 c2 to find 2E 02 = mc2 (E + mc2 ).
If we then use that T = E mc2 (T 0 = E 0 mc2 ) we find

T0
T = 4T 0 (1 + ). (1.47)
2mc2
This equation shows that the gain of kinetic energy by creating collisions between
beams with opposite velocity, compaired to keeping one target at rest. Classically
the gain is a factor of 4, but relativistically the gain can be huge. Example: a 1GeV
lab energy gives a 4000 GeV relative energy.

4. Consider the reaction e + e+ and move to CM frame before collision. Then


p0T = (2mc, 0, 0, 0). After collision this should be the same, but that is impossible for
a single photon. It can never have a 4-momentum of the form p0T = (2mc, 0, 0, 0),
because it does not square to zero. (In other words a photon has no rest frame).

1.4 Maxwell theory in tensor language


We now provide a quick reminder of Maxwell theory in an explicit covariant language. We
also discuss the free momentum eigenstate solutions and polarisation states. We briefly
touch upon the Lagrangian formalism and gauge invariance which lies at the heart of
modern particle physics.
You have learned that the Maxwell equations are
~ E
~ = 4 , (1.48)
~ B
~ = 0, (1.49)
~ E
~+1B ~ = 0, (1.50)
c t
~ B
~1E ~ = 4 J~ . (1.51)
c t c
~ B
In the Maxwell equations E, ~ are the electric and magnetic field respectively and , J~ are
charge density and current density respectively.
In tensor language these equations can be described using an anti-symmetric (2, 0)-
tensor F
0 E1 E2 E3
E1 0 B3 B2
F = . (1.52)
E2 B3 0 B1
E3 B2 B1 0
We furthermore create a four vector j as follows
~ .
j = (c, J) (1.53)

15
The continuity equation that expresses charge conservation

~ J~ = ,
(1.54)
t
simply becomes
j = 0 . (1.55)
Ex: check this and remind yourselves on why this equation expresses charge
conservation.
Those Maxwell equations with sources on the right hand side can be written as follows
4
F = j . (1.56)
c
Ex: check this.
Ex: Show how, acting with an extra derivative, equation (1.56) leads to the
continuity equation (G.20).

The remaining Maxwell equations can be written as

1 2 3 4 1 F2 3 = 0 , (1.57)

where  is the so-called Levi-Civita symbol. It is the fully antisymmetric symbol, for which
one picks some normalization for one of its components (fixing the rest), for example

0123 = 1 . (1.58)

Ex: check this.


Ex (more advanced): is  a tensor?
A conceptual note: you should wonder in what sense this rewriting is a rewriting or a
proof that the Maxwell equations are Lorentz covariant. It is the lecturers opinion that
this rewriting shows that the Maxwell equations are Lorentz-covariant if we can verify
that F is a two-tensor and j is a vector. The latter has to be verified experimentally.
This is analogues to claiming that E ~ is a vector. There is no derivation of that, you need
to go out in nature and check whether an electric field is indeed a vector.
One can show, using some basic theorems in calculus (you do not need to be able to
do this) that equations (1.57) imply the existence of a vector A , called the Maxwell field,
that obeys:
F = A A . (1.59)
Ex: Verify that this form of the Maxwell tensor automatically obeys half of all
the Maxwell equations, namely the half implied by equation (1.57).

Something important has happened notation-wise in the above equation. The derivative
symbol has its index downstairs because X, with X some scalar is a one-form.

16
Ex: check this.
Hence, if the symbol is used it implies we mean

= , where = . (1.60)
x
The 4-vector A is often written as
~ ,
A = (, A) (1.61)

~ the vector potential.


where is called the scalar potential and A

~ and B
Ex : Write E ~ in terms of A
~ and .

Writing the Maxwell equations in terms of the Maxwell field A naturally leads to the
ultra-important9 concept of gauge invariance. The essence is that the following transfor-
mation (note how the index is down)

A0 = A + , (1.62)

with some function (scalar) leaves the field strength F invariant and hence also the
Maxwell equations. Using gauge invariance we can always impose the so-named Lorenz
gauge
A = 0 . (1.63)
Ex: Verify this.

The Maxwell equations (1.56) then simplify to


4
2A = j , where 2 = . (1.64)
c
The above box operator, 2, is called the dAlembertian and is the relativistic cousin of
the Laplacian. The Lorenz gauge does not uniquely remove the gauge redundancy. This
can easily be seen, since any function that obeys 2 = 0 can still be used for a gauge
transformation such that we preserve the Lorenz gauge.
A useful, but non-covariant gauge (!) that can further be demanded, in absence of any
source (j = 0), is the Coulomb gauge:

A0 = = 0 . (1.65)

Ex: check this.


9
Gauge invariance and the gauging principle uniquely dictates the classical equations of motion for 3
out of the 4 forces in nature.

17
Let us now solve the free-wave equation (1.63) and search for momentum-eigenstates.
So at this point we give a quantum mechanical interpretation to the Maxwell equations
(more later) in the sense that we work with the operator correspondence
p = i~ , (1.66)
which can be derived from the known operator correspondence

~ ,
p~ = i~ E = i~ (1.67)
t
and the relation between energy and p0 (1.37).
Ex: Check (1.66).

This means (complex10 ) solutions of the form


i
A = a exp( p x )  (p) (1.68)
~
if they exist, will be energy and momentum eigenstates and we require p2 = 0.
Ex: why is that so? Why can we assume both momentum and energy eigen-
states? The vector  has four components but the Lorentz and Coulomb gauge together
only leave two independent since we find
0 = 0 , ~ p~ = 0 . (1.69)
This means that the two polarisation states are orthogonal to the direction of motion.
This is a bit counter intuitive since you should have learned that a photon has spin 1 and
that a spin s field has 2s + 1 polarisation (spin) states. So we expect 3 instead of 2. The
catch is that the latter assumption is true for massive fields, not for massless fields. In
fact a massless field/particle has 2 helicity states that are eigenstates of the spin operator.
The helicity eigenstates themselves are complex and the interested student can consult the
Appendix for more. The complexity of the helicity eigenstates is sometimes the reason
that, in the literature, complex polarisation vectors  are used. Whereas in practice one
typically uses the following choices for a photon traveling in the z direction (in the Lorenz
gauge):
~(1) = (1, 0, 0) , ~(2) = (0, 1, 0) . (1.70)
In general, without enforcing the Coulomb gauge one has the following covariant conditions
on the two polarisation vectors
p (1,2)
= 0. (1.71)
They can be chosen to be orthogonal and normalised as follows
(2)
((1)
)  = 0 , ( )  = 1 . (1.72)
10
The Maxwell equations are linear, so if you do not like the complex solutions, you can just take the
real part. The reason complex solutions are considered here has probably to do with the fact that we have
quantum mechanics in mind and interpret A as some quantum mechanical wave function.

18
Once we are in the Coulomb gauge, where  only has non-zero spatial components we can
have the completeness relation
X (s) (s)
i (j ) = ij pi pj , (1.73)
s=1,2

where the indices i, j run over the spatial directions and the vector p is the unit vector
pointing in the direction of travel.
Ex: Prove this relation. Hint: this is a matrix equality. Two matrices are equal
if and only if their maps agree on a set of vectors that form a basis. There is
a natural choice of 3 basis vectors here.
Finally we want to give the Langrangian of Maxwell theory since it turns out useful later.
I expect you are familiar with the EulerLagrange formalism in what follows. Students
are more used to Hamiltonians than Lagrangians. But in particle physics Lagrangians are
more useful for three reasons: 1) symmetries play a central role in particle physics to the
extent that forces are defined by the symmetry group: U (1) for Maxwell theory, SU (2) for
the weak force and SU (3) for the strong force. Symmetries and their conserved quantities
are best seen from the point of view of the Lagrangian. 2) A Hamiltonian is not Lorentz
invariant, (Ex: Explain why), whereas a Lagrangian is. So this gives us constraint on
theory building which would not be visible in the Hamiltonian formalism. 3) Quantum
field theory, which defines particle physics, can be elegantly formulated in terms of path
integrals that rely on the Lagrangian formulation.
In relativistic field theory the action is the integral over whole 4D Minkowski space of
a Lagrangian density L, which can depend on many fields i , where i = . . . N
Z
S = dx0 dx1 dx2 dx3 L(i , i ) . (1.74)

The EL equations of motion are


 L  L
i
= . (1.75)
[ ] i

These are N equations, one for every field i . Summation over is implied. If L is a
Lorentz scalar, then the EOM are Lorentz invariant.
The Lagrangian density for free Maxwell theory is

L = 41 F F . (1.76)

We have
L F
= 12 F Note how a factor of 1/2 got cancelled.
A A
= 12 ( )F
= F . (1.77)

19
Hence, the EL EOM becomes the Maxwell equation (1.56).
Ex: Check all the above equations and statement.
Let us add a source J to Maxwells action:

L = 14 F F 4
c
A j . (1.78)

Then the equation of motion becomes

F = 4
c
j . (1.79)

The continuity equation for the source (G.20) implies gauge invariance of the action, since a
gauge transformation of the Lagrangian density generates a total derivate in the Lagrangian
and a total derivate does not affect the EL equations.
Ex: Check that a gauge transformation indeed gives an extra total derivative
Ex: Check that a total derivate added to a Lagrangian density leaves the EL
equations unchanged.

20
2 Lecture 2: Relativistic Quantum Mechanics (RQM)
The word classical physics can mean two different things: classical in the sense of non-
relativistic or classical in the sense of non-quantum. In the first lecture we have treated
non-quantum but relativistic description of particles. In this lecture we will be maximally
non-classical and discuss relativistic quantum mechanics; We will introduce the Klein-
Gordon equation and more importantly the Dirac equation and the Dirac spinor. This
allows us to define the coupling of the Maxwell field to charged spin 1/2 particles (think
electrons, quarks, leptons) responsible for all electromagnetic phenomena observed.

2.1 Probability currents in quantum mechanics


We start with a quick reminder of probability currents in quantum mechanics as we need
this later on. The probability density (x, y, z, t)dxdydz of a particle is defined as the
probability to find the particle in the volume element dxdydz around the point (x, y, z) at
time t. Quantum mechanics means this density is the absolute value squared of the wave
function :
(x, y, z, t) = . (2.1)
If the particle does not decay its associated total probability must stay constant. The
conservation of probability is expressed via a continuity equation

~ ~j + = 0 ,
(2.2)
t

where ~j(x, t) is defined as the probability current density. Consider now a static volume
V with boundary V . We have that
Z Z Z
~ ~ ~ ,
~j dS
= j = (2.3)
t V V V

where as usual dS~ denotes ~n dS with ~n


the unit vector along the outward normal and dS
the infinitesimal surface element. From the Schrodinger equation one can verify that

~j = ~ [
~
~ ] . (2.4)
2mi
Ex: check this.
Normalisable wave functions, which are the wave functions for which a probability
interpretation can be given, die off sufficiently fast at infinity for the following to hold
Z
~ = 0,
~j dS (2.5)
V

such that the total probability is always conserved.

21
2.2 Non-relativistic quantum mechanics
The Schrodinger equation
h ~2 2 i
i~ = + V , (2.6)
t 2m
cannot be Lorentz invariant, in the sense that solutions get mapped to solutions, under
Lorentz transformations. This can be understood superficially since the left hand side
contains a single derivate in time and the right hand side a double derivative in spatial
coordinates.
Consider the following Galilean transformation
r~0 = ~r ~v t , (2.7)
t0 = t . (2.8)
So the S frame moves away from the S frame with constant velocity ~v . What happens
with the Schrodinger equation after a Galilean transformation? First observe
(~r, t) = (r~0 + ~v t, t) , (2.9)
(~r, t) (r~0 + ~v t, t) ~ r0 (r~0 + ~v t, t) ,
= ~v (2.10)
t t
~ r (~r, t) =
~ r0 (r~0 + ~v t, t) . (2.11)
The last two identities are a nice illustration of how to make computations with coordinate
transformations. So for any set of coordinates xi (like for instance space-time coordinates)
the effect of a linear mixing X
x0i = M ij xj (2.12)
j

with M some invertible matrix, on the derivatives is


X x0j X X
i = 0=
i j
M ji j 0 = (M T )ij j 0 . (2.13)
j
x j j

Note that the index closest to the matrix symbol M labels the columns and the other one
the rows. Hence for our Galilean transformation we have
0
t 1 0 0 0 t
x0 vx 1 0 0 x
0 = (2.14)
y vy 0 1 0 y
z0 vz 0 0 1 z
such that

t 1 vx vy vz t0
x 0 1 0 0 x0

= (2.15)
y 0 0 1 0 y0
z 0 0 0 1 z0

22
If we apply this to the Schrodinger equation we find a new wave equation
h i h 2 i
i~ ~ r0 (r~0 + ~v t, t) = ~ 2r0 + V (r~0 + ~v t, t) (r~0 + ~v t, t) ,
~v (2.16)
t 2m
which differs from the standard Schrodinger equation. However, after the following unitary
redefinition
i i mv 2
(r~0 + ~v t, t) = 0 (r~0 , t) exp[ m~v ~r0 + t] , (2.17)
~ ~ 2
we do find
0 (r~0 , t) h ~2 2 i
i~ = + V 0 (r~0 , t) 0 (r~0 , t) . (2.18)
t 2m
where V 0 (r~0 , t) = V (~r ~v t, t) so in absence of a potential V (and hence V 0 ) the functions
and 0 satisfy exactly the same equation. This would be Galilean invariance. For Galilean
covariance V does not need to vanish.
Ex: What is the physical reason V has to vanish for Galilean invariance?
Ex: Could you have predicted the unitary transformation? For instance from
thinking about free waves?

2.3 A relativistic wave equation: the KleinGordon field.


The simplest relativistic wave equation is found for particles with no internal structure,
so-called scalars . The easiest way to find a wave equation for a free scalar is to use the
operator correspondence
~ ,
E i~ , p~ i~ (2.19)
t
in the relativistic energy-momentum relation
E
( )2 p~ 2 = m2 c2 , (2.20)
c
to obtain the Klein-Gordon (KG) equation

~2 2 = m2 c2 . (2.21)

The above operator correspondence (2.19) can be neatly written as

p = i~ (2.22)

Ex: check this.


Note that the KG equation (2.21) is second-order in time, whereas the Schrodinger equation
is first-order. To take the non-relativistic limit we have to apply a trick. First define the
following unitary transformation
i
0 = exp[ mc2 t] (2.23)
~
23
Ex: You should see the logic in this.
satisfies the equation
2 0
~2 2
+ 2i~mc2 0 = ~2 c2 2 0 . (2.24)
t t
Now the limit can be taken as follows
~2 2 0 0
<< |~ |. (2.25)
2mc2 t2 t
and this gives indeed gives the free wave Schrodinger equation
0 ~2 2 0
i~ = . (2.26)
t 2m
Ex: Explain exactly how this limit (2.25) works.
If we define the real quantities
i~  
P = , (2.27)
2mc2 t t
 
~j = ~ ~ ~ , (2.28)
2mi
they can be shown to obey the continuity equation
P ~ ~j = 0 .
+ (2.29)
t
Ex: check this.
One can furthermore show that in the non-relativistic limit we obtain the known expressions
for the probability density and current.
Ex: check this.
Hence it must be that the above quantities are the relativistic extension of the classical
probability density and current. However the probability density is not all manifestly
positive, so something is incorrect here. A second problem occurs when we look at the
plane wave solutions:
E,~k = A exp(i~k ~r t) . (2.30)
Unleashing the KG equation on this implies
q
~ = m2 c4 + ~2 c2~k 2 . (2.31)
The interpretation of ~k and are
E = ~ , p~ = ~~k . (2.32)
We have two solutions for every momentum, those with positive and negative energy. We
also had that at the classical level of course, but there we did not have some fundamental
equation whose solutions were physical states. Hence the KG equation predicts a world with
arbitrary negative energy states. This would make the theory unstable since everything
would decay to the minus infinity sending infinite amounts of energy in the form of radiation
into space.

24
2.4 Remarks on quantum field theory
The above two problems are telling us that relativistic quantum mechanics by itself is
incomplete, it has to be interpreted different. This is where quantum field theory comes
into play. Just like the Maxwell equation, the KG equation is a field equation, and the
quantization of the field gives rise to particles whose number can change during interac-
tions. That is the realm of quantum field theory: particles can be created and destroyed,
something that cannot really be described by quantum mechanics. A heuristic way of
understanding this comes from the fact that energy can be transferred to rest mass energy
and vice versa. At the same time quantum mechanics tells us that energy can fluctuate
and since rest mass energy can only take fixed values for elementary particles, it must
be that such a fluctuation creates new particles or destroys particles. So quantum field
theory should allow particle creation and destruction11 . Hence we want wave equations
that describe multiple particle states at once. This occurs when we interpret relativistic
waves as fields and particles as field quanta. I guess you intuitively already did this for
the Maxwell field. Mathematically the transition from quantum mechanics to quantum
field theory occurs through the concept of second quantization, in which the field itself is
regarded as an operator acting on a Hilbert state that described multiple particle states.
QFT also solves the issue of the negative energy states: they correspond to positive energy
anti-particles.
In this course we will nonetheless work with relativistic quantum mechanics and do
not try to resolve these conceptual issues since that is beyond the scope of this text. The
scattering computations we will perform will not be altered in any way by extending RQM
to QFT. Hence we can happily continue with RQM. Since Nature seems described by QFT
I advice you to follow courses on this in your masters.
The KG-field has no internal structure and hence no spin. Spin 0 particles are called
scalars since they are invariant under Lorentz transformations. The tensor nature of fields
is typically denoted by their spin. This is because spin, although defined as a property of
spatial rotations, also fixes the transformation under the full Lorentz group.
We start by reminding how spin 1/2 particles, called spinors, were described to you in
your course on quantum mechanics and then we discuss the relativistic extension called
Dirac spinors. The corresponding wave equation is called the Dirac equation.
A note of warning: Dirac spinors come with their own new technical luggage which can
be hard/cumbersome when you see this for the first time.

2.5 The spin of a field/particle


What is a particle of spin s in quantum mechanics? A spin s particle has a wave function
:
that transforms as follows under spatial rotations with an angle over an axis ~n

0 = U~n () , (2.33)
11
Once you allow that you can also have particle creation and destruction without a temporary fluctu-
ation in the energy. It can occur while preserving the energy.

25
where the unitary transformation is defined as follows
i ~
U~n () = exp[ ~n J] , (2.34)
~
where J~ is known as the total angular momentum.
Ex: derive that this is indeed a unitary operation.
We write
J~ = L
~ +S
~. (2.35)
~ is. The spin angular momentum operator S
I assume I do not need to recall what L ~ obeys
the properties
~ S
S ~ = i~S
~. (2.36)
~ S
S ~ = s(s + 1)~2 . (2.37)
~ = 0. Indeed then the transformation rule (2.33) coincides with
So spin 0 particles have S
the expected result:
0 (~r 0 , t) = (~r, t) (2.38)
Ex: check this.
One can further show that quantum states can be build from a basis of eigenstates of
a component of the spin operator. One typically takes Sz . The eigenstates obey

Sz s,ms = ms ~s,ms , (2.39)

where
ms = s, s + 1, . . . , s 1, s . (2.40)
~ allows a representation that is
Hence there are 2s + 1 eigenstates. The spin operator S
hence 2s + 1 dimensional. Quantum states can be written as follows:
s
X
s = ms (~r, t)s,ms , (2.41)
ms =s

where
0
..
.
0

s,ms = 1 . (2.42)

0

.
..
0
In the above column the non-zero entry appears at the ms spot.
Spin 1 is familiar: it is a 3-vector that rotates the way you have learned a vector rotates.
But for instance spin 1/2 we have no intuition. We call them spinors.

26
Relativistic quantum mechanics slightly alters this picture. This we have already seen
~ they naturally become part of a 4-vector A . So the counting of the com-
for vectors A,
ponents does not quite work out the same. We will also see this with spinors. A spinor
in relativity, called a Dirac spinor, also has four components. This is because the spinor
describes two particle states at the same time: a particle and an anti-particle. In other
words: relativity predicts anti-particles.
So we have that spin 0 particles are scalars. In 2013 CERN has discovered what
seems a fundamental scalar particle: the Brout-Englert-Higgs (BEH) boson. Also modern
cosmology, in particular the theory of inflation invokes the existence of fundamental scalar
fields. This is however untested theoretical guess work. All fermions in nature observed
so far are spin 1/2. Spin 1 particles are responsible for the gauge forces in nature: the
photon for the electromagnetic force, the W +/ and Z 0 boson for the weak force and the
8 gluons for the strong force. The graviton is supposed to be a spin 2 particle12 .
Finally recall that integer spins obey Bose statistics and half-integer spin obey Fermi-
statistics. This is a non-trivial result of quantum field theory whose derivation is rather
advanced.

2.6 The Dirac equation.


Dirac was driven to find a first-order differential wave equation (that was Lorentz-covariant)
instead of the second-order KG equation. There is a semi-heuristic way to arrive at this
equation, which is rather elegant. Consider the energy-momentum relation
p p m2 c2 = 0 . (2.43)
If we change the p by their operator definitions we get the KG equation. So we need a
trick. We could try to factorize equation (2.43)
p p m2 c2 = ( p + mc)( p mc) = 0 . (2.44)
If this can be done we arrive at a first-order equation
p + mc = 0 , (2.45)
since this equation would solve equation (2.43) and can be written as a wave equation
once the ps are replaced by their corresponding operators. (We could also take the other
factorit does not matter). So let us verify what the s or s are. If we work out the
product we get
p p + mc( )p m2 c2 = 0 . (2.46)
To get the (2.43) we require = and
= . (2.47)
12
If supersymmetry is discovered in nature we also expect a spin 3/2 particle called the graviphoton.
Finally if string theory is correct we expect an infinite tower of extremaly massive spin states for all values
of the spin.

27
If the s are ordinary numbers, this equation cannot be satisfied.
Ex: show that.
So to be on the safe side we will not assume that the s mutually commute. Then we can
conclude at best that13
{ , } = 2 . (2.48)
Ex: show that.
It can be shown that the smallest dimension of matrices that can solve this are 4 4. A
well known basis is the following
   
0 1 0 i 0 i
= , = . (2.49)
0 1 i 0

The indices i run from 1 to 3. The i are the Pauli matrices and 1 the 2 2 unity matrix
and 0 the 2 2 zero matrix. We will refer to this basis as the Griffiths basis.
Ex: check that this choice of matrices work!
So the above equation (2.48) should really be written as a matrix equation

+ = 2 144 . (2.50)

The resulting equation


i~ mc = 0 , (2.51)
is the celebrated Dirac equation. Since the -matrices are four by four, the Dirac wave
has four components, denotes . It is crucial that you do not mix up the 4 spinor indices
with the four space-time indices . They are really different.
The above choice of matrices is not unique and any base transformation can be used
since it satisfies (2.50).
Ex: explain that claim.
Such an alternative basis, that is called the Weyl or chiral basis is almost the same, but
not quite:    
0 0 1 i 0 i
= , = (2.52)
1 0 i 0
So only the 0 differs. It is trivial to check that each time the Gamma matrix algebra
(2.50) (called Clifford algebra) holds.
To go from Griffiths basis G to the Weyl basis W

one can use the following transfor-
mation  
1 1 1 1
W = KG K , K= . (2.53)
2 1 1
These transformations do not work on the space-time indices but on the spinor space
indices . Since the act on the spinors also, the above transformations need to acted on
the spinors as well, when bases are changed:

W = KG . (2.54)
13
We used the symbol {} for an anti-commutator: {A, B} = AB + BA.

28
Let us now demonstrate that Dirac spinors have spin 1/2. There are multiple ways this
can be done. We could for instance try to understand how the Dirac spinor transforms
under spatial rotations and compare this with how we defined the rotations of a spin 1/2
state in the above. This is for the next section. Now we proceed differently. We will prove
that the total angular momentum
J~ = L
~ +S~, (2.55)
~ part that corresponds to spin 1/2.
has a S
We start with the observation that the Hamiltonian H derived from the Dirac equation
is
H = c 0 (~ p~ + mc) , where p~ i~(x + y + z ) . (2.56)
Ex: How do we get this?
~ = ~r p~. We can derive it is not conserved
Now consider the orbital angular momentum, L
since
~ = i~c 0 (~ p~) .
[H, L] (2.57)
Ex: prove this.
This already shows that the Dirac field cannot be spin 0, because free spin 0 fields have a
~ We indeed need to compensate this non-conservation by the introduction of
conserved L.
~ It turns out that the following proposal for S
an internal spin S. ~
 
~= ~ ~, ~ = ~ 0
S with .
2 0 ~

is such that
~ + S]
[H, L ~ =0 (2.58)
~ S
and therefore the total angular momentum L+ ~ is conserved. To verify that S
~ corresponds
to spin 1/2 we only need to observe that

~ S
S ~ = i~S~, (2.59)
2
S ~ = 3~ 1 ,
~ S (2.60)
4
Ex: Derive this.

29
3 Lecture 3: RQM continued
3.1 Transforming a Dirac spinor
In what follows we investigate how a Dirac spinor transforms and how it relates to the
spinors we saw in non-relativistic quantum mechanics.
We denote the transformation matrix of a Dirac spinor by S:

0 = S . (3.1)

Now we demand that Diracs equation is Lorentz invariant in the sense that a Lorentz
transformation maps solutions to solutions. The Dirac equation contains a derivative ,
which transforms as
0 = . (3.2)
Diracs equation transforms as

[i~(SS 1 ) mc]S = 0 . (3.3)

(Note the SS 1 insertion for later convenience.) Now we write the matrix S to the left
front of the equation
S[i~S 1 S mc] = 0 . (3.4)
So for this to map equivalent physics we want the expression in front of S to be the same,
hence
S 1 S = , (3.5)
or, after inverting (keep track of the position of the indices)

S 1 S = . (3.6)

In the appendix we explain the explicit construction of S.


The reason I give you the (famous) Weyl basis is that the transformation matrix S is
most easily found there. To find S in the other basis we simply use the relation

SW = KSG K 1 . (3.7)

In the appendix we show that the transformation matrix in the Weyl basis has the following
structure:  
A 0
S= , (3.8)
0 A
where A is an invertible complex matrix, whose explicit expression does not matter at this
point.
Now we search for Lorentz invariant combinations of spinors and to combinations that
transform as a vector and so on. For that we first define the adjoint Dirac spinor (or
Dirac conjugate spinor):
= 0 . (3.9)

30
If we use the following identity
S 0S = 0 . (3.10)
Ex: check this in the Weyl basis.14 we can verify that the adjoint spinor transforms
as follows under a Lorentz rotation:

0 = S 1 , (3.11)

Ex: check this


such that
, (3.12)
is a scalar (ie a real Lorentz invariant number).
One can also make a vector this way, as follows

. (3.13)

Ex: Show that this transforms as a vector.


Therefore the following quantity is Lorentz invariant:

LI = A . (3.14)

This simple interaction term between a spinor and the Maxwell field is what is responsible
for all electromagnetic interactions in Nature. This term appears as a term in the funda-
mental Lagrangian for Quantum Electro Dynamics (QED) and almost all of the physics
resides in this simple expression.
We can also make an anti-symmetric tensor

, (3.15)

where
i
= ( ) . (3.16)
2
Ex: prove this.
For describing the weak interactions it turns out useful to define the following object

5 = 5 = i 0 1 2 3 . (3.17)

In the Griffiths basis this gives  


5 0 1
= . (3.18)
1 0
Ex: what do you get in the Weyl basis?
Independent of the basis we find that

{ 5 , } = 0 . (3.19)
14
In Griffiths basis one of course also finds this, but one has to use the more ugly expression for
SG = K 1 SW K.

31
Ex: show this.
One can show that 5 is a pseudo scalar and that 5 is a pseudo-vector.
Pseudo means that it gets a minus sign under a parity transformation

~x ~x . (3.20)

Such parity transformations can be shown to be represented by the following transformation


matrix
S = 0 (3.21)
Ex: Show that S = 0 indeed corresponds to a parity transformation.
Ex: then show that 5 is a pseudo scalar.
A spinor-bilinear is defined as
1 M 2 , (3.22)
with M a matrix that acts on the spinor indices and 1 and 2 two Dirac spinors. To
understand whether such an object is real or complex it is useful to introduce the Dirac
conjugate for matrices as well. The Dirac conjugate M of a matrix M is defined as

M = 0 M 0 . (3.23)

One can show that


(1 M 2 ) = 2 M 1 . (3.24)
Ex: Show this. Hint: use that ( 0 ) = 0 .
Ex: Show that = . Ex: show that the QED interaction (3.14) is real.
To fully appreciate the link between spinors as defined in Galilean quantum mechanics
and Dirac spinors one should compute how Dirac spinors transform under spatial rotations
and check that it equals the transformation of two-component spinors in Galilean quantum
mechanics. This calculation is left for the appendix.

3.2 Probability currents


The following 4-vector
j = , (3.25)
can be shown to obey the continuity equation j = 0.
Ex: show this.
If we interpret the j 0 as the probability density , we have that probability is conserved.
We can easily verify that is manifestly positive

j 0 = c = c 0 = c = c|1 |2 + c|2 |2 + c|3 |2 + c|4 |2 . (3.26)

32
3.3 Dirac action and the gauging principle.
The Lagrangian density for a free Dirac field is:

L = i~ mc (3.27)

Ex: Do the two terms have the same units?


Since is complex and has components, this is really an action of various degrees of
freedom. We will suppress the spinor index and vary the action w.r.t. both the real and
imaginary parts of the Dirac field, or, equivalently, w.r.t. and . We find
L
= i~ mc = 0 . (3.28)

This is indeed Diracs equation (2.6). We also find
L
= i . (3.29)
[ ]
Such that the EL variation w.r.t. becomes

i~ + mc = 0 . (3.30)

This is nothing but the Dirac conjugate to the Dirac equation (ie the complex conjugated
equation).
Ex: Is the Dirac Lagrangian real? If not how come this does not contradict
the statement that actions have to be real?
We notice that the Dirac equation has global U (1) invariance

eiq , (3.31)

with q some number and an angle between 0 and 2. The gauging principle is the
idea that a global symmetry can be made local by introducing a gauge field B . A global
symmetry means that the symmetry operation does not depend on space-time coordinates,
like above is a constant. A local symmetry means can be an arbitrary function on space-
time: at every point in space-time one should be able to perform a different U (1)-rotation.
How can this be achieved? Under a local transformation we have that
 
eiq + iq , . (3.32)

Now we declare a gauge field B to be a new vector degree of freedom that transforms
under the same local U (1) transformation as

B B . (3.33)

This allows us to write down a gauge invariant Lagrangian by replacing

D , (3.34)

33
where D is called the covariant derivative and is defined as

D = + iqB . (3.35)

that can be made local by introducing D exactly as with the complex scalar. This adds
the term in the action, when D is written out

qB . (3.36)

What is the physics? We will identify the gauge field B with the Maxwell field A . Then
indeed the gauge transformation (3.33) coincides with what we called gauge transforma-
tions earlier when discussing Maxwell theory. We simply called differently, we named
it . Hence the above interaction term is the QED interaction term (3.14) and q is the
charge. If we take into account that B , now called A , has its own dynamics, given by the
Maxwell Lagrangian density we arrive at the following Lagrangian describing the coupling
between a Maxwell field and a charged spinor:

L = 41 F F + i~ mc q~A . (3.37)

This is quite abstract, but it turns out very useful to know that this theory enjoys local U (1)
invariance. Not only does it help in defining the quantum field theory15 , it has predictive
power: if we generalise this idea of gauge invariance to non-Abelian continuous groups
like SU (2) and SU (3) we recover the classical Lagrangians that describe the weak and the
strong force. So Nature seems to be described by this abstract gauging principle!
Ex: how would you describe a charged KleinGordon field? Copy the pro-
cedure used for the Dirac field.
Ex: Consider a coupling of a Maxwell field A to some current 4-vector j
through the following term in the action A j . What is needed for this term to
leave the equations of motion gauge invariant? (Hint: use partial integration
and the fact that total derivates added to an action to do not change the
classical equations of motion.)

3.4 Plane wave Dirac solutions


In this section we use the Griffiths basis for the -matrices.
Let us start by constructing the simplest solutions to the Dirac equation: solutions
which have no dependence on the spatial coordinates. These should correspond to com-
pletely delocalised particles (as always with momentum eigenstates) that are at rest.
If we write the 4-component spinor in terms of two 2-component subspinors A and
B  
A
= , (3.38)
B
15
Again, this is beyond the scope of this course

34
we find:
A mc2 imc2
= i( )A A = A (0) exp( t) , (3.39)
t ~ ~
B mc2 imc2
= +i( )B B = B (0) exp(+ t) . (3.40)
t ~ ~
Ex: verify. The energy should be read of from the phase factor as exp(i E~ t), but this
means that B has mc2 of rest energy. Clearly an impossible result. This is the same
issue of negative energy states we found in the KG equation. Dirac solved this using
the hole theory, which I will not discuss here, due to lack of space-time. The modern
understanding is that B describes anti-particles instead with rest energy +mc2 . So in
total we have four solutions, two particles and two anti-particles since A and B have
two components each. This makes sense since we expect two possible spins each time. So
the 4 solutions are16

1 0
2 2
imc 0 imc 1 ,

(1) = exp( t)
0
, (2)
= exp( t) 0 (3.41)
~ ~
0 0

0 0
2 2
imc 0 imc 0 .

(3) = exp(+ t)
1 , (2)
= exp(+ t) 0 (3.42)
~ ~
0 1

To find the general plane wave solutions to a linear differential equation is now standard.
We start with the usual Ansatz:

(x) = a exp(ik x)u(k) , (3.43)

where k is a 4-vector, so k x = k x . The specific equation of motion then fixes how


the components of the k-vector are related to the mass or energy, etc, and it will fix the
form of u(k). In our case u(k) has four components and we split it into two 2-component
spinors uA , uB :  
uA (k)
u(k) = . (3.44)
uB (k)
We then plug the Ansatz into the Dirac equation and use

= ik , (3.45)

Ex: show this.


to find the following algebraic equation for u(k):

(~ k mc)u(k) = 0 . (3.46)
16
The normalisation is fixed to 1 for no particular reason.

35
To analyse this equation we write it out a bit. The complicated term k becomes, in the
Griffiths basis for the Gamma-matrices:
!
k 0
~
k ~

k = ~ . (3.47)
k ~ k 0

The algebraic Dirac equation (3.46) then becomes



! 0
(~k 0 mc)uA ~~k ~ uB 0
=
0 .
(3.48)
~~k ~ uA (~k 0 + mc)uB
0

This gives us two equations for the unknown uA and uB


1 ~ ~ )uB , 1 ~ ~ )uA
uA = mc (k uB = mc (k (3.49)
k0 ~
k0 + ~

We can further progress by substituting the solution for uB into the expression for uA to
find
(~k ~ )2 uA
uA = 0 2 . (3.50)
(k ) ( mc ~
)2
From (G.65) we then infer that

~k 2
= 1, (3.51)
(k 0 )2 ( mc
~
)2

Ex: check this.


or, written differently
mc 2
k2 = (
) . (3.52)
~
This, together with (3.45), leads us to conclude:

~k = p . (3.53)

However we will be forced to accept that

~k = p , (3.54)

is also allowed, although it conflicts (3.45). The reason we are forced to do this becomes
clear when we derive explicit expressions for u(k). In total we find 4 different solutions as

36
follows:
   
1 1 ~ ) 1
uA =
0
, uB = 0 mc (k ~ 0
, (3.55)
k + ~
   
0 1 ~ ) 0
uA =
1
, uB = 0 mc (k ~ 1
, (3.56)
k + ~
   
1 1 ~ ) 1
uB =
0
, uA = 0 mc (k ~ 0
, (3.57)
k ~
   
0 1 ~ ) 0
uB =
1
, uA = 0 mc (k ~ 1
. (3.58)
k ~

In order for these solutions to go over into the solutions with zero momentum we con-
structed earlier, we are forced to take:

A : p = +~k , (3.59)
B : p = ~k , (3.60)

in line with earlier comments. The reason for this ambiguity resides in the fact that the
Dirac equation, as derived from the heuristic arguments, contained really a sign ambiguity,
which we ignored earlier:
i~ mc = 0 , (3.61)
and now it hits us back. Using quantum field theory gives the answer. We will stick to the
pragmatic attitude that anti-particles would have been described by the other minus sign
and therefore we have to add a minus sign

p = i~ , (3.62)

for anti-particles if we stick to the same Dirac equation. We so far left the normalisation
of the spinors untouched, but in particle physics books the following normalisation is
considered useful:
2E
0 = = . (3.63)
c
Now we can summarise the solutions as follows

1 0
0 1
u(1) = N cpz
E+mc2
, u(2)
= N c(px ipy ) ,
(3.64)
E+mc 2
c(px +ipy ) cpz
E+mc2 E+mc2
c(px ipy ) cpz
E+mc2 E+mc2
pz c 2 c(px +ip2y )
v (1) = N
E+mc
, v (2) = N
E+mc
, (3.65)
0 1
1 0

37
with r
E + mc2
N= . (3.66)
c
Ex: verify how we got to the expressions of the explicit spinors and the nor-
malisation in all detail.
The minus signs in the definition of v (2) is also a convention. So in short we have
ipx
particles : = ae ~ u, (3.67)
ipx
anti-particles : = ae ~ v. (3.68)

and they satisfy:


( p mc1)u = 0 , ( p + mc1)v = 0 . (3.69)
One has the following ortogonality and normalisation relations

u(1) u(2) = v (1) v (2) = 0 , (3.70)


(1) (1) (2) (2)
u u = u u = 2mc , (3.71)
uv = vu = 0 , (3.72)
v (1) v (1) = v (2) v (2) = 2mc . (3.73)

Ex: verify.
and the following completeness relation
X (s)
X (s)
u(s)
u = ( p + mc1) , v(s) v = ( p mc1) . (3.74)
s=1,2 s=1,2

Ex: prove this. It works similar as in the case with the photon completeness
relation

We have found 4 particle states instead of the two we expected. We expected two
since the spin was 1/2 but we got four, which turn out to be so-called anti-particles. So
we find that relativity predicts the existence of anti-particles based on pure mathematical
consistency. That is quite powerful.

38
4 Lecture 4: Decay rates
4.1 Decay rates
Let us recall the concept of a decay rate . Consider an unstable particle that decays
into different particles. It is possible that there are different decay channels each with a
separate decay rate i . If we start out with N unstable particles then their average number
decreases after t time as follows
X
N = N 1 t N 2 t . . . = N t i . (4.1)
i

This equation expresses that the decay is proportional to the number of unstable particles,
the time you wait, and the decay rate. The various decay rates just add and define the
total decay rate X
= i . (4.2)
i

Note that decay rates have the dimension of inverse time, as expected. If we consider the
amount of unstable particles as a function of time, N = N (t), then the above equation
(4.1) can be regarded as a differential equation for N (t) and is easily solved:

N (t) = N (0) exp( t) . (4.3)

From this one can infer that the average lifetime of a particle equals
1
= . (4.4)

Ex: show this.
To express how much a certain decay channel, say the jth channel contributes to the total
decay one defines the notion of branching ratio, denoted BR(j):

j
BR(j) = . (4.5)

4.2 Golden rule


In non-relativistic quantum mechanics decay rates are obtained using Fermis Golden rule
and is derived using perturbation theory. One considers some unperturbed Hamiltonian
H0 for which the initial state of the system |ii is an eigenstate. Then at t = 0 one perturbes
the Hamiltonian with a term H 0 :

t0: H = H0 + H 0 . (4.6)

The final state is called |f i and one is interested in the transition rate f i from the initial
to the final state. How does this relate to particle decays? The latter seems different

39
since there is not really a Hamiltonian that suddenly switches on at some time. It should
be thought of as a trick, not more. In order to study a particle that decays you need to
have one. So one could for instance take the Hamiltonian of the free muon without the
interaction Hamiltonian from the weak force at t < 0 and then at t = 0 one pretends the
weak force suddenly exists. By doing so one can borrow the perturbation theory techniques
from quantum mechanics. Since the decay of a particle follows the Poisson distribution,
the particle has no memory: if you observe it at some time, the probability for it to
decay in the next second is independent of its history. So it could equally have been the
case that there was no weak force.
An essential ingredient is that the interaction Hamiltonian H 0 is small compared to the
free Hamiltonian H0 . Then Fermis golden rule is

2
f i = |Tf i |2 (Ei ) , (4.7)
~

where Tf i is the transition matrix element (see below) and (Ei ) the density of states (also
explained below) evaluated at the initial energy Ei . The dimensions on both sides of the
golden rule (4.7) match since |Tf i |2 has dimensions E 2 and (Ei ) has dimensions E 1 .
Ex: check the dimensions.
The transition matrix element is defined in terms of a perturbation series. The two leading
terms are
X hf |H 0 |jihj|H 0 |ii
Tf i = hf |H 0 |ii + + ... (4.8)
E i Ej
6=i

The summation in the second term proceeds over all eigenstates of H0 different from i.
The perturbation series is what is at the heart of Feynman diagrams. Such diagrams are
literally symbols to denote the terms in the perturbation series. Students that are not
familiar with Fermis golden rule, as stated above, are invited to read sections 9.1 and 9.2
in [3].

4.3 density of states


17
We now discuss the density of states (Ei ) . It is defined as

dn
(E) = | |E , (4.9)
dE
where dn is the number of accessible states with energy between E and E + dE. For later
purposes it will turn out to be useful to complicate the expression a bit as follows
Z
dn dn
| |Ei = (Ei E)dE , (4.10)
dE dE
17
In Section 10.1 of [3] we already discussed the concept and derivation of a density of states for spin 0
and spin 1/2 particles. So we will be brief in what follows.

40
such that Fermis Golden rule becomes
Z
2
f i = |Tf i |2 (Ei E)dn . (4.11)
~
In deriving Fermis golden rule and defining the transition matrix element one has used
normalised wave-functions:
hf |f i = hi|ii = 1 . (4.12)
For the case of spin-0 particles in a volume V , this means the following in the position
representation of the quantum state |i:
Z
dx3 (x1 , x2 , x3 )(x1 , x2 , x3 ) = 1 . (4.13)
V

Now consider a particle in a cubic box of width a, such that the volume of the box equals
V = a3 . The free Hamiltonian is simply H0 = p2 /2m and as a basis of quantum states we
can consider the plane-wave momentum eigenstates
i
= N exp( (~p ~x Et)) . (4.14)
~
The normalisation condition now requires
|N |2 = a3 . (4.15)
Since we are looking at a system with a boundary we have to provide boundary conditions,
which we will take to be periodic18 :
(x + a, y, z) = (x, y, z) , etc. (4.16)
This leads to the usual quantisation for the momentum
2~
p~ = ~n , (4.17)
a
where ~n is a 3-vector with integer entries. This is enough information to find the density
of states. In achieving that we need to take the limit a since the density of states is
assumed to be a continuous function. The limit will turn out to work since a will drop out
of all results. In momentum space each state occupies a volume
2~ 3 (2~)3
px py pz = ( ) = . (4.18)
a V
The energy of the free states is given by E = |~p|2 /2m. To count the density of states in a
given energy interval we need to understand the number of states n in a given interval
for the momenta divided by the volume of a single quantum state.
V
n = px py pz . (4.19)
(2~)3
18
One can show that the results are independent of a choice of boundary conditions.

41
In the limit of infinitesimal small intervals we get
V
dn = dpx dpy dpz . (4.20)
(2~)3

In case only the size of the momentum |~p| appears in expressions then it becomes beneficial
to move over to spherical coordinates in momentum space and integrate over the angles in
momentum space. One then finds:
V
dn(|p|) = 4|p|2 dp , (4.21)
(2~)3

where n(|p|) denotes the number of states in the interval [p, |p| + d|p|]. Written in spherical
coordinates one indeed finds that the number of states is the momentum-space volume
of the shell with thickness d|p| and radius |p| divided by the volume of a single quantum
state. Now it is a piece of cake to find the density of states:

dn dn d|p| V d|p|
(E) = | |= | | = 4|p|2 3
| |. (4.22)
dE d|p| dE (2~) dE

One could fill in |p| = 2mE for a free non-relativistic particle in order to work it out
more explicitly but the above equation is valid more generally. There is already some
interesting physics at this point: higher momentum implies more quantum states. This is
why it becomes favorable in particle physics to decay to the lightest states allowed since
there are more quantum states for them and hence they have a higher probability.
As promised we can verify that the volume factor V drops out of the final result in
Fermis golden rule. To see this explicitly note that a decay process

AB+C, (4.23)

implies that
|ii = A |f i = B C . (4.24)
Hence if each wave A,B,C is normalised according to (4.12) then the matrix element
Z
Tf i = B C H 0 A , (4.25)
V

still carries a factor |N | around. Hence the |Tf i |2 part in Fermis golden carries a factor
|N |2 = V 1 (see equation (4.15)).
Next we look at decays to more than 2 particles. The momentum of the decay products
is not uniquely fixed anymore by momentum conservation as it was in the two-particle decay
process. Hence one speaks of the number of quantum states for the jth decay product
dnj . For the decay to M particles one can always find the momentum of the M th particle
in terms of the momenta of the other M 1 particles. Hence the independent momentum

42
quantum-states are those of the M 1 particles, such that one ends up with the following
counting:
1 M 1 V
dn = M (j) (j)
j=1 dnj = j=1 dpx dpy dpz
(j)
. (4.26)
(2~)3
By including a delta-function one can obtain a more symmetric equation that contains the
momenta of all the M decay products:
M
3 1 V X
dn = (2~) V M (i) (i) (i)
i=1 dpx dpy dpz (~
p(A)
p~(i) ) , (4.27)
(2~)3 i=1

where p~A is the momentum of the decaying particle.


Ex: explain now how the factors of V get cancelled in the equation expressing
Fermis golden rule.

4.4 Relativistic treatment


What changes when we go relativistic? One obvious change is that the Hamiltonians
in the game will be Hamiltonians describing relativistic quantum mechanics, such as the
Dirac Hamiltonian, etc. However there is another subtle change. We have normalised the
wave-functions such that they correspond to one particle per volume V :
Z
||2 = 1 (4.28)
V

As is well known in relativity, due to Length contraction, one particle per volume V
corresponds to particles per volume V from the point of view of the boosted reference
frame since the latter reference frame would claim the volume is really V / since it is
contracted in the direction of motion. This is a general fact: densities are always increased
due to length contraction: the same number of particles seems to live in a smaller volume
for someone that is boosted. Alternatively one can say that the probability density is
really the first component of a four-vector j (the probability current and then the same
obviously follows).
Since
E
= , (4.29)
mc2
one finds that a Lorentz-invariant normalisation of wave functions must be proportional to
E particles per volume V .
From now on we take the reference volume V to be equal to unity V = 1 as in most
textbooks. This way the increase in energy compensates for the effect of the Lorentz
contraction. It is customary to take the factor of proportionality equal to 2/c, in accordance
with (3.63): Z
0 0 E
h | i = 0 0 = 2 . (4.30)
V c

43
This way, both sides of the above equation transform in the same way.
I have used the notation (from [1]) that relativistically normalised wave functions get
a prime p
0 = 2E/c . (4.31)
Transition matrix elements in relativistic quantum mechanics are denoted M instead of
T to highlight that both the Hamiltonian and the normalisation of the wave functions is
relativistic. For a generic process

a + b + ... 1 + 2 + ... (4.32)

we write
p
Mf i = h01 02 . . . |H 0 |0a 0b . . .i = 2E1 /c2E2 /c . . . 2Ea /c2Eb /c . . . Tf i . (4.33)

Note again that M itself is Lorentz-invariant.


If we rewrite the golden rule for the process a 1 + 2
Z
2
f i = |Tf i |2 (Ea E1 E2 )dn , (4.34)
~
we can eventually rewrite this as

(2)4 ~2 c3 d3 p~1 d3 p~2


Z
2 3
f i = |Mf i | (Ea E1 E2 ) (~pa p~1 p~2 ) .
2Ea (2~)3 2E1 (2~)3 2E2
(4.35)
For this to transform as an inverse time-interval it should be that the whole integral is
Lorentz-invariant since the prefactor (E a )1 already transforms as an inverse time-interval.
A Lorentz-invariant integral better be an integral over all four coordinates of relativistic
momentum space. This can be done as follows. We start by observing that
Z
c
(p2i m2i c2 )(p0i )dp0i = , (4.36)
2Ei
where is the heaviside step-function.
Ex: derive this using the equations in the appendix. This allows us to rewrite the
expression manifestly Lorentz-invariant:
Z
1
f i = |Mf i |2 4 (pa p1 p2 )(p21 m21 c2 )(p01 )d4 p1 (p22 m22 c2 )(p02 )d4 p2 .
2Ea (2)2 ~4
(4.37)
Ex: verify what happened in detail. Especially the factors of c and ~.19
Note how both sides of the above equation transform in the same way; the time-dilation
on the left hand side is manifest as an energy increase on the right hand side (recall that
the integral itself is Lorentz invariant).
19
If you compare with [2] you will see different factors of c and ~ hanging around and it is the lecturers
opinion that this is due to a different normalisation convention for the matrix element M.

44
Due to the various delta functions in the integral one can show that the integral can be
carried out explicitly. The reason this is the case is that all four-momentum conservation
laws fix the momenta of the outgoing particles completely so there is no real integral
that needs to be done. If this were not the case then the integral cannot be carried out
explicitly since it would require the explicit momentum dependence of the matrix element
M(pa , p1 , p2 ). The explicit calculation is done below and the result is:
p
= 2 4 3
|M|2 , (4.38)
8ma ~ c
where
c p 2
p = [ma (m1 + m2 )2 ][m2a (m1 m2 )2 ] (4.39)
2ma
is the size of the outgoing momenta of either particle 1 or 2 in the CM frame.
Ex: verify the expression for p (a bit lengthy).

The calculation

We aim at performing the integral in (4.35). The integral (with the prefactor omitted) is

d3 p~1 d3 p~2
Z
2 3
|Mf i | (Ea E1 E2 ) (~pa ~p1 ~p2 ) p p . (4.40)
(2~)3 2c p~21 + m21 c2 (2~)3 2c p~22 + m22 c2

Assume that the decaying particle is at rest. Then p~a = 0 and p0a = ma c. The delta
function for the 3-momentum then puts p~1 = ~p2 . Hence the integral simplifies to
q q
2
1
Z (ma c 2
c p~1 + m2 2
1 c c p~1 2 + m22 c2 ) 3
2
|Mf i | p p d p~1 . (4.41)
4c2 (2~)6 p~21 + m21 c2 p~21 + m22 c2

Next we introduce spherical coordinates in momentum space, where r = |~p1 |:


p p
2 2 2 2 2 2 2
2 (ma c c r + m1 c c r + m2 c ) 2
Z
1
|M fi | p p r drd . (4.42)
4c2 (2~)6 r2 + m21 c2 r2 + m22 c2
The following different radial coordinate simplifies the computation:
q q
u = r2 + m21 c2 + r2 + m22 c2 (4.43)

The inspiration for this comes from the need to simplify the delta function. But it turns
out that the denominator also beautifully simplifies since
du ur(u)
=p p . (4.44)
dr r2 + m21 c2 r2 + m22 c2

45
The integral now becomes
Z
1 r(u)
= 2 6
|Mf i |2 (ma c2 uc) dud , (4.45)
4c (2~) u
Z
1 r(u = ma c)
= 3 |Mf i |2 d . (4.46)
4c (2~)6 ma c

If we can assume that M only depends on the size of the momentum then we can integrate
over the angles to get an extra factor of 4. Since r(u = ma c) = p we indeed reproduce
(4.38).

46
5 Lecture 5: Cross section and amplitudes
Apart from the study of bound states, which are amendable to non-relativistic quantum
mechanics, decays and scattering experiments provide information on the interactions be-
tween elementary particles. In the last chapter we studied decay rates and in this chapter
we will study scattering events. For decay processes the main physical quantity measured
is the decay rate, whose value relates to the amplitude Mf i that can be calculated using
Fermis golden rule once the theory of the interaction is known. We have a similar situa-
tion for scattering experiments. A central physical quantity is then the differential cross
section, which can also be found from Fermis golden rule if the amplitude for the process
Mf i can be calculated.
In the last chapter of this course we attempt this for the scattering between two different
charged fermions that only interact through the electromagnetic force.

5.1 Definition of differential cross section


The scattering experiments in particle physics often involve a beam of particles of type
a that is shot towards a target existing of particles of type b. The beam is characterized
by its luminosity or flux, denoted a , and is defined as the amount of particles moving
through a unit area (perpendicular to the beam) in a unit time. The interaction rate per
target particle is denoted rb is proportional to the flux

rb = a , (5.1)

where the factor of proportionality is called the cross section and has units of area.
Ex: check the units.
As we now explain, the cross section can sometimes be thought of as the physical size of
the b-particle.
Consider an a-particle that moves forward with velocity va in a beam of cross-sectional
area A. The density of particles is na . We furthermore assume that the particles of type b
travel at collinear velocity with speed vb . In a time interval t a particle of type a crosses
a region in the beam-tube of volume (va + vb )tA and hence passes

Nb = (va + vb )Anb t , (5.2)

particles of type b. Now define as the effective area around a single particle of type b
that represents its interaction area, meaning that when a particle of type a crosses this
area, it will interact. This is depicted in the figure 2 below.
If the particles are randomly distributed, then the interaction probability Pa for the
particle to interact is given by the total cross section area of the target particles (b-type
particles) divided by the total cross section area A of the beam:

Nb
Pa = = nb (va + vb )t . (5.3)
A

47
.
.
.
.

Figure 2: The beam (lightblue) with the target particles of type b, depicted as a black dot.
The black circle around the dot represents the effective radius in which an interaction with
a particle of type a occurs.

We then derive the interaction rate ra for a single particle of type a to be ra = dPa /dt =
nb (va + vb ). The total interaction rate r for all particles in some total volume V is then

r = ra na V = nb (va + vb )na V = [na (va + vb )][nb V ] = a Nb . (5.4)

Hence the interaction rate per target particle b is rb = r/Nb and coincides with rb given by
(5.1).
However, whenever the target is soft because the interaction is not simply a collision
but rather a process in which particle a feels the presence of a force field that couples
to particle b there is always some non-zero interaction and is infinite. Of course the
interaction becomes negligible quickly, but formally does not vanish. For that reason
we define the differential cross section D, which tends to be finite. The essential new
ingredient is to look at how the particles are scattered around into various directions. One
then singles out a direction and looks at the particles scattered into that direction. This
direction can be defined by isolating a small (infinitesimal) solid angle d. The differential
cross section is then defined as the amount of particles scattered into d per unit time per
target particle divided by the flux of a-particles:
] particles scattered into d
Dd = (5.5)
tNb a
From this definition it should be clear that
Z
= Dd , (5.6)

where as usual d = sin()dd where runs over 0 to 2 and runs from 0 to . Most
often D is therefore written as
d
D= . (5.7)
d
48
Note how D is the sensible thing measured in experiment: the experimentalist has the
flux a under her/his control and can decide to count only those particles hitting a certain
region of the detector (the solid angle).
From this way of stating things it seems that a (differential) cross section can never be
infinite since its definition involves a number of particles measured in experiment which
will always be finite. But this is due to quantisation, if we would instead rephrase this in
terms of probabilities it is not strange that infinities can occur for interactions that have
infinite extend (such as the Coulomb interaction).

5.2 Relativistic treatment


Consider a scattering event
a + b 1 + 2. (5.8)
As explained before we have that the interaction rate r equals (5.4)

r = a nb V . (5.9)

If we normalise the wave functions as one-particle per volume (like we often do in non-
relativistic quantum mechanics) then
1
f i = r , na = nb = . (5.10)
V
Note the way we introduce Fermis golden rule. The latter is defined in ordinary quan-
tum mechanics for which a single wave function, depending on the space-time coordinates
describes a single particle state. This is why we have to take n = 1/V to apply Fermis
golden rule20 . In any case, in any physical quantity, such as a decay rate, the volume drops
out from Fermis Golden rule. So we can equally set V = 1 everywhere.
The relation between and f i is then seen to be (use (5.4) with V = 1):

r = f i = (va + vb ) . (5.11)

Then
f i
= . (5.12)
va + vb
From the explanation of Fermis golden rule and the density of states around equations
(4.25, 4.27) we know that for a scattering process (5.8) is given by

(2)4 ~2 d3 p~1 d3 p~2


Z
= |Tf i |2 (Ea + Eb E1 E2 ) 3 (~pa + p~b p~1 p~2 ) . (5.13)
va + vb (2~)3 (2~)3
20
Of course the density can and will be different in an actual experiment and then you need to adjust
your equations. In this course we derive cross sections and decay rates for particle densities that are
particularly normalised.

49
To employ a relativistic expression we simply change the normalisation of the wave-
functions as explained before, to obtain:

(2)2 ~4 c4 d3 p~1 d3 p~2


Z
2 3
= |Mf i | (Ea + Eb E1 E2 ) (~pa + p~b p~1 p~2 ) . (5.14)
4Ea Eb (va + vb ) 2E1 2E2
or
(2)2 ~4 c
=
4Ea Eb (va + vb )
Z
|Mf i |2 4 (pa + pb p1 p2 )(p21 m21 c2 )(p01 )(p22 m22 c2 )(p02 )d4 p1 d4 p2 . (5.15)

Since the cross section can be thought of as the interaction radius perpendicular to the
direction of motion one does not expect any Lorentz contraction and hence a fully Lorentz-
invariant expression. The integral is already invariant, but not yet the pre-factor. In fact,
with some work one can show that
q
4Ea Eb (va + vb ) = 4c3 (pa pb )2 m2a c2 m2b c2 , (5.16)

which is manifest Lorentz-invariant.


Ex: check.
The end result is
(2)2 ~4 c2
= p
4 (pa pb )2 m2a c2 m2b c2
Z
|Mf i |2 4 (pa + pb p1 p2 )(p21 m21 c2 )(p01 )(p22 m22 c2 )(p02 )d4 p1 d4 p2 . (5.17)

5.3 Centre-of-mass scattering


Since the expression for is Lorentz invariant we can compute it in any frame. The easiest
frame turns out to be the centre-of-mass frame. There we have

p~a = ~pb p~i p~1 = ~p2 p~f . (5.18)

If we use
E
p~ = m~v = ~v , (5.19)
c2
we can rewrite (5.16) as follows

|~pa | |~pb |
4Ea Eb (va + vb ) = 4Ea EB c2 ( + ) = 4c2 |~pi |(Ea + Eb ) . (5.20)
Ea Eb
If we use the more practical, non-manifest invariant expression for the cross section we find

50
(2)2 ~4 c2 d3 p~1 d3 p~2
Z
2 3
= |Mf i | (Ea + Eb E1 E2 ) (~p1 + p~2 ) . (5.21)
4(Ea + Eb )|~pi | 2E1 2E2

This integral is exactly the same as the one we carried out earlier to find the decay rate
and the result is
|~pf |
Z
1
= |Mf i |2 d , (5.22)
64 2 ~4 |~pi |(Ea + Eb )2
Ex: check this.
or, written differently

d 1 |~pf |
D= = |Mf i |2 (5.23)
d 64 ~ |~pi |(Ea + Eb )2
2 4

5.4 The S-matrix and Feynman diagrams


In more formal treatments of scattering processes in quantum mechanics it is beneficial to
define the so-called S-matrix. It is defined as the operator that maps an initial state to a
final state:
f = Si . (5.24)
Preservation of probability implies that S is unitary. It should be clear that
iHt
S = lim exp( ). (5.25)
t ~
In the context of scattering in a weak potential, or through a weak interaction there is a
natural expansion of the S-matrix called the Dyson expansion. In our setting of relativistic
quantum mechanics this expansion coincides with the expansion in quantum perturbation
theory (5.26). We will use this from now on, a more elaborate and precise treatment of
quantum scattering should entail a precise definition of the Dyson expansion and things
like the Lippmann-Schwinger equation, time-ordering, the optical theorem, etc. We will
brush over all of this and go straight to the concept of Feynman diagrams.
Feynman diagrams are literally pictures that symbolise the various terms in the Dyson
expansion. So in practice the matrix element Mf i can only be calculated perturbatively,
as in a series expansion, and each term is associated to a Feynman diagram. Feynman
diagrams are hence tools to find the series expansion. The Feynman rules are a set of
rules that tell us which diagrams are allowed and how we exactly associate a mathematical
expression to each diagram. The resulting series expansion is most likely a non-convergent
series, its radius of convergence is zero. But still, when the expansion parameter is small,
like in QED, where it is roughly 1/137, the series is very useful as an asymptotic series 21 .
The student that feels uncomfortable with this seemingly bad approximation might feel
21
The concept of asymptotic series is discussed in the appendix.

51
comfort in the fact that QED is the most precise scientific theory that exists or might feel
comfort in studying more closely the concept of asymptotic series.
The essential insight from quantum field theory is that all interactions in nature pro-
ceed via particle interchange instead of interaction through some external potential which
implies action at a distance and seems to conflict special relativity since nothing can
go faster than the speed of light. In practice this means that in the perturbation theory
expansion (5.26)
X hf |H 0 |jihj|H 0 |ii
Tf i = hf |H 0 |ii + + ... (5.26)
6=i
Ei Ej
the first term is absent since it corresponds to scattering in a potential. The second term
however corresponds to interaction through particle/state exchange. The exchanged state
is then |ji. The same holds for the higher order terms. We will just include the terms that
can be interpreted as interaction through particle exchange, since that can be derived from
QFT, or even the relativistic quantum mechanics done in this course. For now we assume
this without derivation and we will get back to this point in the next chapter.

5.5 A simple toy-theory


Let us consider some process in which we have 2 particle scattering:

a + b c + d, (5.27)

where a, b, c and d can be different, but do not need to be. We assume that the interaction
is carried out by the exchange of a boson force carrier X. At leading order we can have
two interactions:
1. First we have a X + c and then b + X d. This means that a emits the force
carrier and it is absorbed by b.
2. First we have b X + d and then a + X c. This means that b emits the force
carrier and it is absorbed by a.
The total amplitude for this process will be the sum of these two time ordered processes.
Let us consider the first process as shown in figure 3. Then
(1) hf |H 0 |jihj|H 0 |ii hd|H 0 |X + bihX + c|H 0 |ai
Tf i = = (5.28)
Ei Ej (Ea + Eb ) (Ec + Eb + Ex )
Note that the intermediate state is really |ji = |X + b + ci, but it is not written like that
in the second equality on the right hand side since the interaction occurs at vertices and
at the first vertex b is not taking part and at the second c is not taking part.
Now let us focus on the separate matrix elements hd|H 0 |X + bi and hX + c|H 0 |ai. In
terms of Lorentz-invariant matrix elements M we can write them as
MaX+c
hX + c|H 0 |ai = , (5.29)
2c1 Ea c1 2Ec c1 2EX

52
a c

b
d
i j f

Figure 3: The diagram that gives M(1) .

similarly
MX+bd
hd|H 0 |X + bi = 1 . (5.30)
2c Eb c1 2Ed c1 2EX
Now we will make a simplification, which defines our toy-theory. To create the simplest
Lorentz-invariant matrix element we will simply assume that the Ms are constants:

MaX+c = ga , MX+bd = gb . (5.31)

If we then evaluate p
(1)
Mf i 2c1 Ea 2c1 Eb 2c1 Ec 2c1 Ed Tf1i , (5.32)
we find:
(1) cga gb
Mf i = , (5.33)
2Ex (Ea Ec Ex )
Now we consider the second process shown in picture 4. An analogous computation

a c

b d
i j f

Figure 4: The diagram that gives M(2) .

53
gives:
(2) cga gb
Mf i = , (5.34)
2Ex (Eb Ed Ex )
Ex: check. What now matters is that the full probability is given by the sum of the above
two. If we take the sum and use energy conservation Eb Ed = Ec Ea we find
ga gb
Mf i = c . (5.35)
(Ea Ec )2 Ex2
For both time-ordered processes the energy of the exchanged particle can be written as

Ex2 = c2 ((~pa p~c )2 + m2X c2 ) , (5.36)

Ex: check. If we substitute this into (5.35) we eventually obtain:

ga gb /c ga gb /c
Mf i = 2 2
2 , (5.37)
2
(pa pc ) mX c q m2X c2

where q = pa pb is the four-momentum of the virtual force carrier X. The diagram that
represents M is the Feynman diagram depicted in figure 5

a c

b d

Figure 5: The diagram that gives M(1) .

There are some crucial lessons to be learned from this computation which turn out to
be general
The sum of the probability amplitudes associated to the separate time-ordered di-
agrams/processes led to a Lorentz-invariant expression (which had to be since we
normalized the wave functions in an invariant way and have chosen invariant ampli-
tudes at each vertex.).
At the vertices of the time-ordered diagram energy is not conserved. (Check). But
this energy conservation violation only takes place in a (small) time interval since
the time-ordered diagrams themselves preserve energy.

54
If the two diagrams are symbolized by a single diagram then this can be seen as
the exchange of a virtual particle. It is virtual in the sense that it is not on-shell,
q 2 6= m2X c2 , which is necessary to preserve both momentum and energy at the separate
vertices. Also note that, if it would be on-shell, the amplitude would formally be
infinite.

So the consequence of summing the two time-ordered processes leads to a diagram


for which both energy and momentum are preserved at each vertex and for which the
amplitude is manifestly Lorentz-invariant at the expense of having a fiducial particle
that is not on its mass shell.
Feynman diagrams are then a construct to write down the contributions to the transi-
tion amplitude at each order, which implies the sum over the time-ordered processes. The
order in some expansion parameters (ga , gb ) is then counted by the number of vertices.
One can furthermore show (we do not) that as long as there is no closed loop inside
the Feynman diagram one is effectively doing classical field theory, in the sense that one
perturbatively solves the classical field equation (eg Maxwell, Klein-Gordon). The only
quantum-ness sits in the interpretation of the field equations and the Feynman diagrams
as particle exchange and transitions between particle states22 .

5.6 Towards a derivation of Feynman rules in QED


To develop the same perturbation theory for the case of quantum electrodynamics we
first need to identify the interaction Hamiltonian H 0 . To know the Dirac equation in the
presence of the interaction, we can use the Lagrangian we derived based on the gauging
principle. Alternatively we note that the essence of the gauging principle was the replace-
ment:
+ iqe A . (5.38)
with qe electric charge. The Dirac equation then becomes

i~ ( + iqe A ) = mc , (5.39)

We now recall that 0 = c1 t and we multiply the Dirac equation with 0 to obtain:
h i
0 i 2
i~t = i~c i + mc + qe ~c A H . (5.40)

From this we verify that the interaction Hamiltonian is given by

H 0 = qe ~c 0 A . (5.41)

Let us now look at the Feynman diagram for the scattering between two different23 fermions,
say a muon and an electron:
22
http://www.staff.science.uu.nl/ hooft101/lectures/basisqft.pdf
23
We take different fermions such that we do not need to worry about Fermi statistics.

55
p2 p4
muon

Electron

p1 p3

As before we keep in mind that there are two time-ordered processes and that we sum
over them. The Feynman diagram represents this sum. There are two interaction vertices.
The reasoning of the previous section goes through where now the role of the X-particle
is played by the (virtual) photon. There are only two differences: 1) the matrix element
of an interaction vertex is not a simple constant and 2) the photon has two polarisation
states and we need to sum over them.
Let us first look at the time-ordered process in which the muon sends out the photon
which is then captured by the electron.
The interaction at the electron vertex gives a matrix element of the form

he (p3 )|H 0 |e (p1 )i . (5.42)

The photon is represented by the usual plane wave (since we consider momentum eigen-
states)
i
A = ( )(s) exp( q x ) . (5.43)
~
where we do not specify the polarisation state s of the photon. We then have:
i
he (p3 )|H 0 |e (p1 )i = exp( q x )qe ~cue (p3 ) 0 (s)
ue (p1 ) . (5.44)
~
whereas at the second vertex, the photon is now conjugated (Ex: explain why):
i (s)
h(p4 )|H 0 |(p2 )i = exp( q x )qmuon ~cumuon (p4 )q 0 ( ) umuon (p2 ) . (5.45)
~
The combination u 0 will be replaced by u. If we use this to compute the matrix element,
and do not forget to sum over the two possible time-orderings, we as before generate the
term 1/(q 2 m2X c2 ), but now mX = 0.
We also need to sum over the possible polarisation states:
(s) (s)
X  ( )
M = qe qmuon ~2 c2 ue (p3 ) 0 ue (p1 ) 2
umuon (p4 ) 0 umuon (p2 ) . (5.46)
s
q

Ex: why do we sum the Ms for every spin and not the |M|2 ?

56
It turns out that we can take
X (s)
(s)
( ) = , (5.47)
s

which is explained in appendix E. In the end the result is the following expression

M = qe qmuon ~2 c2 ue (p3 ) ue (p1 ) umuon (p4 ) umuon (p2 ) . (5.48)
q2

which is nicely manifest Lorenz invariant.


In lecture 6 you will learn how to get an actual explicit function in terms of the mo-
menta from this and verify that in the non-relativistic limit the cross-section is the one
you know from Rutherford scattering.

5.7 The Feynman rules


If one would work out in all detail how the perturbation series looks like, given the above
logic one would arrive at the so-called Feynman rules. These rules simply give you an
algorithm to associate a matrix element contribution to each Feynman diagram. This way
you do not need to repeat each time the whole reasoning. You check it for once in your life
and then you blindly follow the rules. Apart from the Feynman rules one also needs a rule
about the possible diagrams that exist. We will not go into the latter issue but it turns
out that the Feynman diagrams are build from simple vertex diagrams that can be glued
together in all possible ways one would like to. For QED there is only one such vertex :

Figure 6: The QED vertex.

To make the various Feynman diagrams one can glue all of these diagrams together,
including rotations of the single diagram.
A diagram with n vertices contributes at order n in perturbation theort. For example
the possible, leading order diagrams in QED are the following set:
In this course we only explain how to compute the matrix elements for the first process.
But the derivation of the matrix elements for the other processes goes through similar

57
Figure 7: The various QED processes to second order in the coupling. Table taken from [2].

reasonings. Instead of each time going through the whole derivation a simple algorithm
has been found that immediately allows you to associate a contribution to M for a given
diagram. This algorithm goes by the name of Feynman rules and is beyond the scope of
this course, but the interested reader can for instance consult [1, 2]. The end result is that
observables, like decay rates and cross sections are given by a perturbative expansion in
the dimensionless fine-structure constant

e2
= 1/137 . (5.49)
~c
where e denotes qe from here onwards. The last equality is only true at sufficiently low

58
energies since it can be shown that e effectively depends on the energy scale of the exper-
iment in such a way that it rises for higher energies. The perturbative expansion has zero
radius of convergence but provides nonetheless a very accurate description of QED since
the expansion should be thought of as an asymptotic series (see appendix D).

59
6 Lecture 6: Electron-muon scattering
In this lecture we explain how one obtains the actual cross section to leading order in
the fine structure constant for electron-muon scattering. The result holds for all scattering
between different fermions. The result for same fermions or fermion-anti-fermion scattering
is very similar up to small changes.
So we are after computing explicitly the amplitude (5.48) in the sense that we want
to get a number depending on the external momenta (p1 , p2 , p3 , p4 ). Since the amplitude
is Lorentz invariant we most likely will get an answer that involves scalar products p1 p2
etc.
We start by first rewriting (5.48) in a simpler way, using that the charge of a muon
equals that of an electron qe = qmuon e and we put ~ = c = 1:
e2
M= u(3) u(1) u(4) u(2) . (6.1)
q2
where we also left out the subscript of the spinors and labelled the spinors using the number
on their momentum label: 1, 2, 3, 4.
Computing the actual value is quite a workout and fills this whole chapter.

6.1 Workout 1: Summing over spins


Casimir observed a trick that simplifies the computation of cross sections and decay rates.
The idea is that in any experiment you typically have a beam of particles with mixed
spin-up and spin-down of equal amounts. The particles after interaction are not treated
differently depending on their spin. So an amplitude, or probability, should really average
over the initial spins and sum over the outgoing ones. Average because you do not measure
the incoming spins and sum because, regardless of the outgoing spin, a scattering event is
measured and adds to the amplitude for the process. This implies that our final expres-
sions will not contain traces of the spins and eventually the expression for the amplitude
drastically simplifies.
Also note that we really want the absolute value squared
|M|2 = MM . (6.2)
So in the end we require
e2 2 1 X
h|M|2 i = ( ) |us3 (3) us1 (1) us4 (4) u(2)s2 |2 . (6.3)
q 2 4 s s ,s ,s
1 2 3 4

The factor of 1/4 comes from averaging over 4 initial spins, two for each ingoing particle.
We rewrite this more conveniently as a product of 4 factors
e4 1 X
h|M|2 i = 4 [us3 (3) us1 (1)][us4 (4) u(2)s2 ][us3 (3) us1 (1)] [us4 (4) u(2)s2 ] .
q 4 s s ,s ,s
1 2 3 4
(6.4)

60
Ex: check how the rewriting happened, especially the indices on the gamma
matrices.
The first simplification we use is in computing the square, which involves computing a
complex conjugation. Consider any spinor bilinear

u(a)u(b) (6.5)

with some matrix (typically made out of gamma matrices). Its complex conjugation can
be shown to be given by
(u(a)u(b)) = u(b)u(a) (6.6)
where is defined as
= 0 0 . (6.7)
Ex: Prove this. (Use that ( 0 )2 = 1 and ( 0 ) = 0 ).
Since = (Ex: prove that) we can rewrite the right hand side of equation (6.3)
as
1 X s3 X
[u (3) us1 (1)][us1 (1) us3 (3)] [us4 (4) u(2)s2 ][us2 (2) u(4)s4 ] . (6.8)
4ss s s
1 3 2 4

The main idea behind the Casimir trick is that a sum like:
X
[us3 (3) us1 (1)][us1 (1) us3 (3)] (6.9)
s1 s3

can be written as as simple trace involving gamma matrices and momenta, but no spinors
any more. This hinges essentially on the fact that we sum over the spins. It works as
follows. Consider some general product of spinor bilinears G

G = u(a)1 u(b)u(b)2 u(a) (6.10)

Let us now sum over the b-spins:


X X
G = u(a)1 ( u(sb ) (pb )u(sb ) (pb ))2 u(a) , (6.11)
bspins sb =1,2

= u(a)1 ( pb + mb 1)2 u(a) . (6.12)

Now we also wish to sum over the a-spins. For simplicity we temporary introduce the

61
matrix Q = 1 ( pb + mb )2
X X
G= u(a)sa Qusa (a) ,
a-spins b-spins sa
XX
= u(a)sa Q u(a)sa ,
sa ,
XX
= u(a)sa u(a)sa Q ,
, sa
X
= ( pa + ma 1) Q ,
,

= Tr( pa + ma 1)Q ,
h i
a b
= Tr ( p + ma 1)1 ( p + mb )2 . (6.13)

6.2 Workout 2: Making traces


From here on we use a famous short-hand notation

p/ = p . (6.14)

Furthermore m denotes electron mass and M muon mass.


If we apply Casimirs trick twice to (6.4) we then find

e4 h i h i
h|M|2 i = Tr
( p
/ + m)
(p/ + m) Tr (p/ + M ) ( p
/ + M ) . (6.15)
4(p1 p3 )4 1 3 2 4

where we used momentum conservation q 2 = (p1 p3 )2


Let us work out the first factor with the traces. The second is identical up to relabelling.
Let us first rewrite it
h i
Tr (p/1 + m) (p/3 + m) =
h i
2
Tr p/1 p/3 + m p/1 + m p/3 + m . (6.16)

The first term is the most involved one since it contains a product of 4 gamma matrices.
The second and third term are zero since a product of three gamma matrices vanishes as
explained below. The last term can be rewritten immediately since

Tr[ ] = 4 . (6.17)

Ex: prove this. (Use the anti-commutation relation of the gamma matrices
and the fact that a trace is cyclic.
In appendix F we prove that the trace of a product of three gamma matrices is zero.

62
Then we use the identity

Tr[ ] = 4( + ) . (6.18)

Ex: prove this. (Use the anti-commutation relation of the gamma matrices
three times.
Putting this all together we find
h i n o
2
Tr (p/1 + m) (p/3 + m) = 4 p1 p3 + p3 p1 + (m (p1 p3 ])) . (6.19)

Hence the whole amplitude becomes

4e4 n

o
h|M|2 i = p p + p p +
(m 2
(p 1 p 3 ]))
(p1 p3 )4 1 3 3 1
n o
(p2 ) (p4 ) + (p4 ) (p2 ) + (M 2 [p2 p4 ]) , (6.20)
8e4 n
2 2 2 2
o
= (p 1 p 2 )(p 3 p 4 ) + (p 1 p 4 )(p 2 p 3 ) (p 1 p 3 )M (p 2 p 4 )m + 2m M .
(p1 p3 )4

6.3 Workout 3: non-relativistic limit


We want to know the differential cross section in Lab frame. In the CM frame (5.23) we
have:
d 1 |~pf |
D= = |Mf i |2 . (6.21)
d 64 ~ |~pi |(Ea + Eb )2
2 2

Since the collision is elastic (ex: why?) it must be that: |~pf | = |~pi |. We will assume
that we work in the limit where the recoil of the muon, which is much heavier than the
electron, can be ignored. Then the lab and CM frame coincide. Furthermore in the same
limit Emu = M Eel (Ex: why?). Hence
d 1
D= = h|M|2 i . (6.22)
d 64 2 M 2
Concretely the momenta can be taken as follows

p1 = (Eel , p~1 ) , p2 = (M, ~0) , p3 = (Eel , p~3 ) , p4 = (M, ~0) , (6.23)

where we used that the recoil is negligible and the interaction is elastic. In this case we
have that the angle between p~1 and p~2 equals the scattering angle . Hence

p~1 p~3 = |~p|2 cos() . (6.24)

Using the above to compute all the scalar products appearing in h|M|2 i in the end leads
to  e2 M 2
h|M|2 i = [m2 + |~p|2 cos2 (/2)] , (6.25)
|~p|2 sin2 (/2)

63
Ex: check this. Use that 1 cos = 2 sin2 (/2). such that M drops out of the
differential cross section:
 e2 2
D= [m2 + |~p|2 cos2 (/2)] . (6.26)
8|~p|2 sin2 (/2)

This is called the Mott formula and is a good approximation whenever the recoil can
be neglected. Otherwise one can redo the above exercise to obtain a much longer but
precise expression. If the incident electron is non-relativistic one should obtain Rutherfords
equation. The non-relativistic limit is taken by (recall that c = 1 here)

|~p|2
<< 1 (6.27)
m2
such that  e2 m 2
D= . (6.28)
8|~p|2 sin2 (/2)
We nicely recover Rutherfords formula which you should have seen in a course on classical
mechanics. If we would consider higher order diagrams with virtual particles in loops then
we would find quantum corrections to the classical result.

6.4 Conclusion
Note how amazing this is: we went all the way to the relativistic Dirac equation, coupled to
an electromagnetic field, used quantum perturbation theory and finally recover the classical
limit of scattering charged point particles. This is physics at its best.

64
A What is a tensor?
Here is a brief introduction to the concept of tensor(field)s. An essential tool is a vector
space and its dual. So make sure you first recall the concept of dual vector spaces before
you read this.

A.1 Vectors and one-forms


Consider a real vector space V and its dual vector space V . The elements of V are called
vectors and we can call the elements of V one-forms.
The dual vector space is the vector space of linear maps that map a vector in V to a
real number. Consider V , v, w V and , R arbitrary, then

(v + w) = (v) + (w) R. (A.1)

What we are most interested in is the transformation properties of the one forms when
we transform the base of the vector space. For that we go to components. Since the are
real linear maps we can write them as follows, using Einstein notation

(v) = a v a , (A.2)

where the v a are the components of v in a basis ea of V

v = v a ea , (A.3)

and the wa must be real numbers, that we consider to be the components of in the dual
basis f a of V
= a f a . (A.4)
We see that the basis and dual basis obey the following relation

f a (eb ) = ba . (A.5)

Now consider the following base transformation in V

e0a = Mab eb (A.6)

where the matrix M is invertible. From now on our notation is as follows. If an object has
two indices and one wants to think in terms of matrix(multiplication) then the index closest
to the object enumerates the different rows and the second index the different columns.
It is clear that the induced transformation on the dual vector spaces base is obtained
by maintaining the fundamental relation (A.5). So denote

f 0a = N ab f b . (A.7)

We find
Ma b N cb = ac (A.8)

65
or written differently
Ma b [N T ]b c = ac (A.9)
So I added a transpose because the indices on N where switched from left to right (ie
columns were interchanged with rows). The above identity says

N T = M 1 , (A.10)

or, put differently, the transformation on the dual space obeys

N = M T . (A.11)

Let us now check what this implies for the components of vectors and one forms. Since
we have
v = v a ea = v 0a e0a , = a f a = a0 f 0a , (A.12)
we deduce that the components transform as

v 0a = [M T ]ab v b = [N ]ab v b , a0 = Ma b b . (A.13)

Note that this is nicely consistent with the fact that

(v) = v a a = v 0a a0 . (A.14)

A.2 (p,q) tensors


A real number (or in general an element of the field K the vector space is defined over)
can be called a scalar. It does not change under base transformations.
Consider a bilinear form, ie an operator B that acts on two vectors and gives a real
number
B : V V R : v, w V : B(v, w) R. (A.15)
where the action is taken to be linear in both arguments of B. Then we call B a (0,2)-
tensor. This means that the action of B can be written in components as follows

B(v, w) = Bab v a wb . (A.16)

If B is invertible and symmetric we call B a quadratic form or metric. If B is furthermore


positive definite we call B a Riemannian metric, otherwise it is called pseudo-Riemannian.
Such a bilinear form or metric, can be used to raise and lower indices (physics slang).
By this we mean something simple. Take a vector v, act with B and you get a one-form
= B(v, .); ie
Bab v b a . (A.17)
(Vice versa for raising the index.) This is what you have been doing in special relativity.
You had the bilinear form, given by the Lorentz metric and you raised and lowered
indices with it.

66
Consider now a linear map A from V to V

A(v) = w , (A.18)

or in components
Aab v b = wa . (A.19)
This is how you came to the concept of a matrix in your earlier studies. Ie, this square
with numbers in it is naturally associated with a linear map from a vector space to a
vector space. Of course a bilinear form is also associated with a matrix. By acting with a
one-form on A(v) we get a real number

(A(v)) = a Aab v b R (A.20)

Note that this is linear in . Hence a linear map from a vector to a vector can also be
defined as an operator that acts on a one-form and a vector to give a real number, which
is linear in the arguments:

A : V V R : v V and V : A(v, ) R. (A.21)

Such an operator is called a (1,1)-tensor.


Let us now define a (p,q)-tensor, named T , in general. T will be an operator, linear
in all its arguments that maps p one-forms and q vectors to a real number. Depending on
the tensor the first argument acts on a vector or a one-form, and so on. In components
this object has p indices upstairs and q indices downstairs. For instance a (1, 2) tensor can
look like
T abc , or Ta bc , or Ta b c . (A.22)
In this language a vector can be seen as a (1,0) tensor since it indeed acts on a one form
to give a real number in a linear way as follows

v() = (v) = va a . (A.23)

Similar, a one-form is a (0,1) tensor. So all the tensors you already encountered in your
studies where (in order of typical appearance)

1. vectors (1,0)

2. linear maps (1,1)

3. dual vectors (0,1)

4. bilinear forms (0,2) (for instance Hookes law or the Lorentz metric in special rela-
tivity.)

67
Concerning the transformation rules, induced by the change of base in V (and hence
V ) we can follow the previous logic. For every index upstairs there will be transformation
matrix N and for every index down there will be a matrix M as follows
0a ...a d c ...c
1
Tb1 ...b q
p
= N ac11 . . . N acpp Mb1d1 . . . Mbq q Td11...dqp , (A.24)
Einstein notation is developed such that you do not need to think anymore about how
to compute the transformation rule. You just match the height of the indices and you
remember that indices, that have to be summed over, need to have different height. For
instance the transformation of a vector has to be
v 0a = N ab v b . (A.25)
So the height of the indices on N is obvious. The orientation, on the other hand, is fixed
by demanding that the index M carries, used to sum over, has to stand next to the object
it will be summed with. Similar for a one form
a0 = Ma b b . (A.26)
Summing over indices is called contraction. As mentioned, this can only happen when
indices have different height. For example, for a (1,1) tensor (a matrix)
Aaa = Trace[A] (A.27)
Another beauty of Einstein notation is that we can read from an object what kind of
tensor it becomes, after having contracted indices. One simply looks at the structure of
the FREE indices. You can immediately see this from the transformation properties (Ex:
show that.) example above there is no free index after having taken the trace of the
matrix and hence the trace should be a scalarindeed, in any good introduction to linear
algebra you have learned that the trace is invariant under base transformations. In the
tensor language this is obvious. Ex: is the trace of a bilinear form an invariant?
Another example, take a (1,2) tensor, then
T aab , (A.28)
is a one-form, since it has 1 free index down.
Summing over indices can also be done by multiplying tensors. For example the action
of a one-form on a vector gives a scalar, by definition, and this can be seen from that
notation, since
v a a , (A.29)
has no free index.
Finally, a tensor, like the example of a vector and a one-form can be expressed in terms
of a basis. This is why we actually wrote the indices, this gives the components of the
tensor expanded in a basis. For example the basis for (p,q) tensors is given by the following
set (IF the first p indices are up and the last q indices are down. When the heights are
mixed, the generalisation is obvious)
ea1 . . . eap f b1 . . . f bq , (A.30)
The symbol denotes a tensor product

68
B Photon Helicity
How to define the spin operator for a photon? Well, a spinor transforms under spatial
rotations in the 2-dimensional representation (recall SU (2) = SO(3) up to a Z2 factor).
Whereas photons, who are described by vectors, transform under the 3-dimensional repre-
sentation. Such representations are described by exponentials of anti-symmetric matrixes
~ = exp(1 J 1 + 2 J 2 + 3 J 3 )
O() (B.1)
where i corresponds to a rotation around the xi axis. The generators are

0 0 0 0 0 1 0 1 0
J1 = 0 0
1 ,
J2 = 0 0 0 , J3 = 1 0 0 . (B.2)
0 1 0 1 0 0 0 0 0
The helicity operator then becomes
p~ J~ . (B.3)
Consider a photon moving along the z-axis. Then the helicity will be eigenstates of J3 . This
is an anti-symmetric matrix and therefore has imaginary eigenvalues i. The eigenvectors
are    
1 1
Vi = , Vi = . (B.4)
i i
Now the second footnote on page 241 (section 7.5) of Griffiths should be clear. Note: in
what follows I put ~ = c = 1 and I am also not careful with the coefficients in front of
terms in the action. This is often convention anyway.

C The spinor transformation matrix


C.1 Pauli matrices and Lorentz transformations
First some notation
= = (1, i ) , (C.1)
and accordingly we must have (by acting with to lower or higher the indices)
= = (1, i ) . (C.2)
The essential trick to derive the transformation comes from the following observation.
Consider the four-vector x and define a complex 2 2 matrix X out of it as follows
 0 
x + x3 x1 ix2
X = x = . (C.3)
x1 + ix2 x0 x3
Notice that X is hermitian X = X . In reverse, we can say that any 22 hermitian matrix
can define four real numbers x as follows
x = 21 Tr( X) . (C.4)

69
Where I used that
1
2
Tr = . (C.5)
The central observation is that
detX = x x . (C.6)
Consider the following linear transformation of X, preserving its hermiticity property

X 0 = AXA . (C.7)

As long as the complex 2 2 matrix A has unit determinant, det(A) = 1, then detX is
unchanged, and by equation (C.6), and linearity we know that A must relate to a Lorentz
transformation as follows:

x0 = 21 Tr( X 0 ) = (A)x . (C.8)

(One can show that this defines an isomorphism between the Lorentz group and SL(2, C).);
let us make the link between A and a bit more explicit. We have that for any x

x0 = (A)x = 21 Tr[A AX] , (C.9)

where we used that the trace is cyclic in the last step. Since x was arbitrary, we must have

A A = . (C.10)

(Note that can be pulled in or out of the trace since it is a number from the point
of view of the spinor indices.) This starts looking like (3.6), but we are not there yet; In
particular S, are 4 4 matrices whereas A, are 2 2 matrices. To complete our task
we need one more transformation rule

A A = [1 ] . (C.11)

This also implies


A1 A = [] . (C.12)
Here is one way to derive this. Start from
1
2
Tr[ ] = . (C.13)

and note that the left hand side equals


1
2
Tr[A AA1 A ] , (C.14)

by cyclicity of the trace. Now write

A1 A = T , (C.15)

70
with T a real matrix (the reasoning for this is similar to arguments given above). Together
with equation (C.10) and (C.13) we deduce that

T = , (C.16)

which proves that T is the Lorentz matrix and therefore reproduces (C.12).
In the Weyl basis (2.52) it becomes straightforward to check that the following choice
for SW  
A 0
S= . (C.17)
0 A
satisfies equation (3.6) using the identities (C.10) and (C.11) (just write it out).

C.2 Spatial rotations


To verify that Dirac spinors are really the relativistic extension of the notion of a spinor
as seen in Galilean quantum mechanics we need to verify that under spatial rotations the
Dirac spinor transforms as in equation (G.41). Let us focus purely on the part of the
transformation that acts on the spinor indices. That means that a Galilean spinor, that
carries two indices only, transforms as follows
i ~ ~ = ~ ~ .
U~n () = exp[ ~n S] , S (C.18)
~ 2
under a rotation.
In the Weyl basis we had a concrete form for the transformation matrix of a Dirac
spinor:  
A 0
S= . (C.19)
0 A
For a spatial rotation we must have the property that

S 1 0 S = 0 , S 1 i S = Oji j , (C.20)

with O the usual three-dimensional rotation (orthogonal) matrix. In the Weyl basis, the
first equation above implies:

AA = A A = 1 A = A1 . (C.21)

Then the other equations become

A i A = Oji j . (C.22)

This matrix A is then exactly the unitary transformation we were looking for
i ~
A = exp[ ~n S] . (C.23)
~

71
Indeed, consider a rotation around the z-axis with angle . Then
i
A = exp[ z ] , (C.24)
2
and
cos sin 0
Oji = sin cos 0 (C.25)
0 0 1
Ex: verify!
Since it works for the rotations around the z axis, and the z-axis was chosen arbitrarily, it
should work for a rotation around any axis.
Ex: verify explicitly (long computation, but tricks exist.)

So what we get is that a single Dirac spinor is made up out of two Galilean spinors
u , uB
A
 A 
1 u1
2 uA
2 
= 3 u3 .
=  B (C.26)
4 uB
4

D Asymptotic series
You can read up here:
http://en.wikipedia.org/wiki/Asymptotic_expansion
http://mathworld.wolfram.com/AsymptoticSeries.html

72
E The matrix element for electron-muon scattering
Consider the matrix elements hj|H 0 |ii or hj|H 0 |ii at the vertices of the time-ordered dia-
grams, such that the photons are on their mass-shell.
The matrix elements can be written as inner products of the form
j  (E.1)
or complex conjugations thereof. The current j denotes the electron or muon current. Now
we insist on our theory to be gauge invariant and this implies that the matrix elements do
not change under a gauge transformation. Let us look at the following gauge transformation
A A + (E.2)
such that we remain in Lorenz gauge. This means 2 = 0 then such a could be
= a exp(ip x) (E.3)
with p the on-shell photon momentum: p2 = 0. This means that the polarisation vectors
of the plane wave solutions are altered by
  + aip (E.4)
for any a. Then matrix elements can then only be gauge-invariant if
j p = 0 . (E.5)
Now if we go back to writing the Feynman diagram as a sum of two time-ordered
diagrams then the exchanged photon is indeed on its mass-shell and we can use that
j p = 0 in every diagram. So if we encounter
X
( )  (E.6)
s

and write it in Coulomb gauge, we get


X
(i ) j = ij pi pj . (E.7)
s

Since j p = 0 we deduce that


p0
ji pj = j0 = j0 . (E.8)
|~p|
If we put all of this together we find that
j j ( )  = ji jj (ij pi pj ) ,
= ji jj ij j02 ,
= j j . (E.9)

73
F Trace over producty of three gamma matrices
We prove that the product of three gamma matrices is necessarily traceless. In fact one
can show that the product of any odd number is traceless. First we use that

Tr[ ] (F.1)

vanishes whenever one of the indices are equals. Take for instance that = then the
cyclicity of the trace implies

Tr[ ] = Tr[( )2 ] = Tr[ ] = 0 . (F.2)

So now take all three indices in (F.3) different. Then using the fact that they anti-commute
we can always order them, for example

Tr[ 0 2 3 ] , Tr[ 0 1 2 ] (F.3)

and so on. Then we insert the lacking gamma matrix squared since the square is either
plus or minus one. So we have

Tr[ ] = Tr[ 0 1 2 3 ] , (F.4)

where was the lacking gamma matrix in the product of (F.3). Then we observe that, in
the Griffiths basis  
0 1 2 3 0 1
5 = i . (F.5)
1 0
Since now    
0 5 0 1 j 5 j 0
= i = i (F.6)
1 0 1 j
(where j = 1, 2, 3), we find that it is indeed traceless.
Ex: now extend to a general product of odd number of matrices.

G Useful formulas
Special relativity
In Minkowski space we have
x0 = ct . (G.1)
Distance in Minkowski space
d2 = x x , (G.2)
where
1 0 0 0
0 1 0 0
=
0 0 1 0 ,
(G.3)
0 0 0 1

74
Lorentz transformations are those linear transformations that preserve the Minkowski
length d:
x0 = x , x x = x0 x0 . (G.4)
The bilinear form can be used to raise and lower indices
x = x . (G.5)
Hence
d2 = x x . (G.6)
...
The inverse of is the same matrix but has indices upstairs. A (p, q) tensor T11...qp
transforms as follows
T01 1...
...p
q
= 1 1 . . . p p 11 . . . qq T11...
...p
q
(G.7)
Boost in the x1 -direction

0 0
0 0
=
0
, (G.8)
0 1 0
0 0 0 1
where
1
=p . (G.9)
1 2
Addition of velocities
x w0 + v
w = 0 . (G.10)
t 1 + vw
c2
Proper time is defined through
c2 d 2 = dx dx . (G.11)
Momentum 4-vector
dx
p = m = m . (G.12)
d
The spatial 3-vector inside the 4-vector equals
p~ = m~v . (G.13)
The temporal component is
p0 = mc = E/c . (G.14)
We have:
E
p p = ( )2 p~2 = m2 c2 . (G.15)
c
The kinetic energy T of a particle is
T = mc2 ( 1) (G.16)
Massless particles obey:
E = |~p|c . (G.17)

75
Maxwell theory

0 E1 E2 E3
E1 0 B3 B2
F = . (G.18)
E2 B3 0 B1
E3 B2 B1 0
The current four-vector j
~ .
j = (c, J) (G.19)
The continuity equation
j = 0 . (G.20)
Maxwell equations
4
F = j , (G.21)
c
1 2 3 4 1 F2 3 = 0 . (G.22)

 is the Levi-Civita symbol. It is the fully antisymmetric symbol, for which one picks some
normalization for one of its components (fixing the rest), for example

0123 = 1 . (G.23)

The Maxwell four-vector


F = A A . (G.24)
Gauge transformation:
A0 = A + , (G.25)
The Lorenz gauge
A = 0 . (G.26)
Coulomb gauge:
A0 = = 0 . (G.27)
Complex plane wave solutions:
i
A = a exp( p x )  (p) (G.28)
~
with p2 = 0.
Photon traveling in the z direction (Lorenz gauge):

~(1) = (1, 0, 0) , ~(2) = (0, 1, 0) . (G.29)

One has the following covariant conditions on the two polarisation vectors

p (1,2)
= 0. (G.30)

76
They can be chosen to be orthogonal and normalised as follows
(2)
((1)
)  = 0 , ( )  = 1 . (G.31)

Once we are in the Coulomb gauge, where  only has non -zero spatial components we
can have the completeness relation
X (s) (s)
i (j ) = ij pi pj . (G.32)
s=1,2

Euler-Lagrange
Z
S= dx0 dx1 dx2 dx3 L(i , i ) . (G.33)

The EL equations of motion are


 L  L
i
= . (G.34)
[ ] i
The Lagrangian density for free Maxwell theory is

L = 41 F F . (G.35)

Probability currents in ordinary QM


The Schrodinger equation
h ~2 2 i
i~ = + V , (G.36)
t 2m
Density:
(x, y, z, t) = . (G.37)
Continuity equation
~ ~j + = 0 ,
(G.38)
t
where ~j(x, t) is the probability current density:

~j = ~ [
~
~ ] . (G.39)
2mi

Spin
A spin s wave function that transforms under a spatial rotations of an angle over an axis
as follows:
~n
0 = U~n () , (G.40)
where
i ~
U~n () = exp[ ~n J] , (G.41)
~
77
with J~ the total angular momentum:

J~ = L
~ +S
~. (G.42)

~ obeys the properties


The spin angular momentum operator S
~ S
S ~ = i~S
~. (G.43)

~ S
S ~ = s(s + 1)~2 . (G.44)
The quantum states can be build from a basis of eigenstates of a component of Sz and
obey
Sz s,ms = ms ~s,ms , (G.45)
where
ms = s, s + 1, . . . , s 1, s . (G.46)
~ allows a representation that is
Hence there are 2s + 1 eigenstates. The spin operator S
hence 2s + 1 dimensional. Quantum states can be written as follows:
s
X
s = ms (~r, t)s,ms , (G.47)
ms =s

where
0
..
.
0

s,ms = 1 . (G.48)

0

.
..
0
In the above column the non-zero entry appears at the ms spot.

Relativistic QM
Operator correspondence
p = i~ (G.49)
KleinGordon
~2 2 = m2 c2 . (G.50)
Dirac:
i~ mc = 0 . (G.51)
with probability four-vector
j = , (G.52)

78
The Lagrangian density for a free Dirac field is:

L = i~ mc (G.53)

Transformation matrix of a Dirac spinor:

0 = S , (G.54)

with
S 1 S = . (G.55)
Dirac conjugate spinor:
= 0 . (G.56)
From
S 0S = 0 . (G.57)
one finds:
0 = S 1 , (G.58)
A spinor-bilinear is defined as
1 M 2 , (G.59)
with M a matrix that acts on the spinor indices and 1 and 2 two Dirac spinors. The
Dirac conjugate M of a matrix M is defined as

M = 0 M 0 . (G.60)

Such that
(1 M 2 ) = 2 M 1 . (G.61)

Pauli matrices
The Pauli matrices are:
     
0 1 0 i 1 0
x = , y = , z = . (G.62)
1 0 i 0 0 1
One can verify that
i = i = i1 . (G.63)
They obey the following algebraic relations:

i j = ij 1 + iijk k , (G.64)
(~a ~ )(~b ~ ) = ~a ~b1 + i~ (~a ~b) , (G.65)

exp[i~ ~ ] = cos 1 + i(~ ~ ) sin . (G.66)

In the above equation ~ is defined as a vector and it its norm and ~ is the normalised
vector.

79
Gamma matrix algebra

( ) = 0 0 ,
+ = 2g ,
5 = i 0 1 2 3 , = i 5 , 0123 = 1 ,
= 0 . (G.67)
Representation of gamma matrices
     
0 1 0 i 0 ~ 0 1
= , ~ = = , 5 = ,
0 1 ~ 0 1 0
     
1 0 1 2 0 i 3 1 0
= , = , = . (G.68)
1 0 i 0 0 1

Alternatively on can use the Weyl basis which has a different 0 :


 
0 0 1
= . (G.69)
1 0
Spin of the electrons:  
~= ~ ~ 0
S . (G.70)
2 0 ~
Notation: for a one-form w we have

/ w .
w (G.71)

Dirac plane waves


We take ~ = c = 1.
Electrons Positrons
ipx (s)
(x) = ae u (p) , (x) = aeipx v (s) (p) ,
(p/ m)u = 0 , (p/ + m)v = 0 ,
u(p/ m) = 0 , v(p/ + m) = 0 ,
(s) (s0 ) ss0 (s) (s0 ) 0

Pu u(s) (s)= 2m , v v = 2m ss ,
(s) (s)
P
s u u = p/ 
+m , s v v = p/  m ,
1 ~
~ p 0
0  E+m (G.72)
 1 ,

u(1) = E + m ~~p 1 , v (1)
= E + m 0
E+m 0 1
   
0 ~
~ p 1
1  E+m
 0 .

u(2) = E + m ~~p 0 , v (2)
= E + m 1
E+m 1 0

80
Fermis Golden rule
If we start out with N unstable particles then their average number decreases after t time
as follows
N = N t . (G.73)
The transition/decay rate is:
2
f i = |Tf i |2 (Ei ) , (G.74)
~
where Tf i is the transition matrix element and (Ei ) the density of states evaluated at the
initial energy Ei . To second order the matrix element is:
X hf |H 0 |jihj|H 0 |ii
Tf i = hf |H 0 |ii + + ... (G.75)
6=i
Ei Ej

The density of states is:


Z
dn dn
(E) = | |E = (Ei E)dE , (G.76)
dE dE
In deriving Fermis golden rule and defining the transition matrix element one has used
normalised wave-functions:
hf |f i = hi|ii = 1 . (G.77)
For spinless particles one then has;

dn dn d|p| V d|p|
(E) = | |= | | = 4|p|2 3
| |. (G.78)
dE d|p| dE (2~) dE

A relativistic normalisation of wave functions is


p
0 = 2E/c . (G.79)

Hence
p
Mf i = h01 02 . . . |H 0 |0a 0b . . .i = 2E1 /c2E2 /c . . . 2Ea /c2Eb /c . . . Tf i . (G.80)

itself is Lorentz-invariant.
The interaction rate per target particle is denoted rb and is proportional to the flux

rb = a , (G.81)

where the factor of proportionality is called the cross section and has units of area. The
flux, denoted a , is the amount of particles moving through a unit area (perpendicular to
the beam) in a unit time.

81
Properties of delta functions
The multi-variable delta function is simply a product of single variable delta-functions:

(~x) = N i
i=1 (x ) . (G.82)

The delta function of a function works as follows


X 1
(g(x)) = (x xi ) , (G.83)
i
|g 0 (xi )|

where the xi are all the zeros of g, ie, g(xi ) = 0.


The heaviside step-function is defined as

(x) = 1 , x > 0 , (x) = 0 , x < 0. (G.84)

The derivative of the heaviside function is the delta-function:

0 = . (G.85)

Feynman calculus
We take ~ = c = 1.
Golden rules (S = 1/j! for j identical particles) (spacetime metric (+ ))

decay
: : 1 2 + 3 + ... + n
d3 p~2 d3 p~3 d3 p~n
 
2 S
d = |M| ...
2m1 (2)3 2E2 (2)3 2E3 (2)3 2En
(2)4 4 (p1 p2 p3 . . . pn ) ,
S|p~2 |
1, 2 + 3 : = |M|2
8m21
scattering
: : 1 + 2 3 + ... + n
d3 p~3 d3 p~4 d3 p~n
 
2 S
d = |M| p ...
4 (p1 p2 )2 m21 m22 (2)3 2E3 (2)3 2E4 (2)3 2En
(2)4 4 (p1 + p2 p3 . . . pn ) ,
d S |p~f |
1+23+4 : = |M|2 2 . (G.86)
d [8(E1 + E2 )] |~ pi |

electrons positrons photons


External lines: incoming arrow IN : u(s) (p) arrow OUT : v (s) (p)  (p)
outgoing arrow OUT : u(s) (p) arrow IN : v (s) (p) ( (p))

vertices: ige (2)4 4 (k1 + k2 + k3 ) , ge2 = 4.

82
d4 q
R
propagators: always a factor (2)4
and then
i i(/
q +m) i
for electrons and positrons: q m
/
= q 2 m2
, for photons: q2
.

Antisymmetrization : minus sign for diagrams that differ in the exchange of two exter-
nal fermion lines with identical arrows (for the same particles or antiparticles)

Matrix element: combining all the previous factors gives (2)4 4 (p1 + p2 . . . pn )M.

83
References
[1] M. Thompson, Modern Particle Physics,.

[2] D. Griffiths, Introduction to elementary particles,.

[3] B. Bransden and C. Joachain, Quantum mechanics,.

84

You might also like