Feynman diagrams

Contents
1 Aim of the game 2

2 Rules 2
2.1 Vertices . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3
2.2 Anti-particles . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3
2.3 Distinct diagrams . . . . . . . . . . . . . . . . . . . . . . . . . . . 4
2.4 Relativistic propagators . . . . . . . . . . . . . . . . . . . . . . . . 5
2.4.1 Propagator example . . . . . . . . . . . . . . . . . . . . . . 5
2.4.2 Other propagator examples . . . . . . . . . . . . . . . . . . 7
2.5 Trees and loops . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7

3 Key concepts 8

1
 Feynman diagrams

1 Aim of the game

To calculate the probabilities for relativistic scattering processes we need to find

out the Lorentz-invariant scattering amplitude which connects an initial state |Ψi i
containing some particles with well defined momenta to a final state |Ψf i containing
other (often different) particles also with well defined momenta.

We make use of a graphical technique popularised by Richard Feynman1 . Each

graph – known as a Feynman Diagram – represents a contribution to Mf i . This
means that each diagram actually represents a complex number (more generally
a complex function of the external momenta). The diagrams give a pictorial way
to represent the contributions to the amplitude.

In Feynman diagrams, spin- 21 particles such as electrons are indicated with a straight
line with an arrow.

The arrow follows the direction of particle flow, in the same was as in quark-flow Fermion line
diagrams (§??).

Diagrams consist of lines representing particles and vertices where particles are
created or annihilated. I will place the incoming state on the left side and the
outgoing state on the right side. Since the diagrams represent transitions between Photon line
well-defined states in 4-momentum they already include the contributions from all
possible paths in both time and space through which the intermediate particles
might possibly have passed. This means that it is not meaningful to ask about γ
the time-ordering of any of the internal events, since all possible time-orderings are
necessarily included.

e− e−
2 Rules for calculating diagrams Vertex

It turns out that are simple rules for calculating the complex number represented
by each diagram. These are called the Feynman rules. In quantum field theory we
can derive these rules from the Lagrangian density, but in this course we will simply
quote the rules relevant for the Standard Model.
1 American physicist (1918-1988).

April 8, 2013 2
c A.J.Barr 2009-2013.
2.1 Vertices

2.1 Vertices

Vertices are places where particles are created or annihilated. In the case of the
electromagnetic interaction there is only one basic vertex which couples a photon
to a charged particle with strength proportional to its charge.

To calculate the contribution to Mf i , for each vertex we associate a vertex factor.

For interactions of photons with electrons the vertex factor is of size −gEM where
gEM is a dimensionless charge or coupling constant.2 The coupling constant is a
number which represents the strength of the interaction between the particle and the
force carrier at that vertex. For the electromagnetic force the coupling strength must
be proportional to the electric charge of the particle. So for the electromagnetic
vertex we need a dimensionless quantity proportional to the charge. Recall that for
the electromagnetic fine structure constant:

e2 1
αEM ≡ ≈ .
4π0 ~c 137
is dimensionless. It is convenient to choose gEM such that γ
2
gEM
αEM = .
4π
In other words the coupling constant gEM is a dimensionless measure of the |e|
e− e−
where e is the charge of the electron. The size of the coupling between the photon
and the electron is √ The electromagnetic vertex. The
−gEM = − 4παEM . vertex factor is −gEM .

The electromagnetic vertex factor for any other charged particle f with charge Qf
times that of the proton is then
gEM Qf

So, for example, the electromagnetic vertex factor for an electron is of size −gEM
while for the up quark it is of size + 32 gEM .

2.2 Anti-particles

An anti-particle has the same mass as its corresponding particle cousin, but his
charge is the opposite to that of the particle.3 The Feynman diagram for an
2 We are simplifying the situation by ignoring the spin of the electron. If spin is included

the vertex factor becomes −gEM times a matrix, in fact a Dirac gamma matrix, allowing the Anti-fermion.
spin direction of the electron as represented by a 4-component spinor. For now we will ignore
this complication and for the purpose of Feynman diagrams treat all spin 12 fermions, such as
electrons, muons, or quarks, as spinless. The Dirac matrices also distinguish electrons from anti-
electrons. The sign of the vertex factor is well defined when the Dirac representations are used for
the particles.
3 In fact if the particle is charged under under more than one force then the anti-particle has

the opposite values of all of those charges. For example an anti-quark, which has electromag-
netic, strong and weak charges will have the opposite value of each of those compared to the
corresponding quark.

April 8, 2013 3
c A.J.Barr 2009-2013.
2.3 Distinct diagrams

anti-particle shows the arrow going the ‘wrong’ way (here right to left), since the
particle flow is opposite to that of anti-particle.

The same basic electromagnetic vertex is responsible for many different reactions.
Consider each of the partial reactions
e− → e− + γ
e− + γ → e−
e+ → e+ + γ
+
e +γ → e+
e− + e+ → γ
γ → e− + e+ . (1)

Each of these is just a different time ordering of the same fundamental vertex that
e+
couples an electron to a photon. ∗
γ
e−

The electromagnetic vertex with

2.3 Distinct diagrams particle-antiparticle final state

A Feynman diagram represents all possible time orderings of the possible vertices,
so the positions of the vertices within the graph are arbitrary. Consider the following
two diagrams for e+ + e− →µ+ + µ− :

e− µ−
µ−

γ e− µ+

e+ µ+ e+

In the left diagram it appears that the incoming particles annihilated to form a virtual
photon, which then split to produce the outgoing particles. On the right diagram
it appears that the muons and the photon appeared out of the vacuum together,
and that the photon subsequently collided with the electron and positron, leaving
nothing. Changing the position of the internal vertices does not affect the Feynman
diagram – it still represents the same contribution to the amplitude. The left side
and right side just represent different time-orderings, so each is just a different way
of writing the same Feynman diagram.

On the other hand, changing the way in which the lines in a diagram are connected
to one another does however result in a new diagram. Consider for example the
process e+ + e− →γ + γ

In the two diagrams above the outgoing photons have been swapped. There is no
way to move around the vertices in the second diagram so that it is the same as
the first. The two diagrams therefore provide separate contributions to Mf i , and
must be added.

April 8, 2013 4
c A.J.Barr 2009-2013.
2.4 Relativistic propagators

e+ (P2 ) γ (P4 ) e+

γ (P4 )

γ (P3 )

e− (P1 ) γ (P3 ) e−

2.4 Relativistic propagators

For each internal line – that is each virtual particle – we associate a propagator
factor. The propagator tells us about the contribution to the amplitude from a
particle travelling through space and time (integrated over all space and time). For
a particle with no spin, the Feynman propagator is a factor

1
Q · Q − m2

where Q · Q = EQ2 − q · q is the four-momentum-squared of the internal virtual

particle4 .

These intermediate particles are called virtual particles. They do not satisfy the
usual relativistic energy-momentum constraint Q · Q = m2 . For an intermediate
virtual particle,
Q · Q = EQ2 − q · q 6= m2 .

Such particles are said to be off their mass-shell.

If this inequality worries you, it might help you if you consider that their energy and
momentum cannot be measured without hitting something against them. So you
will never “see” off-mass-shell particles, you will only see the effect they have on
other objects they interact with.

External particles in Feynman diagrams do always individually satisfy the relativis-

tic energy-momentum constraint E 2 − p2 = m2 , and for these particles we should
therefor not include any propagator factor. The external lines are included in the
diagram purely to show which kinds of particles are in the initial and final states.

2.4.1 Propagator example

Consider the annihilation-creation process e+ + e− → γ ∗ → µ+ + µ− proceeding

via a virtual photon γ ∗ . (The star on the particle name can be added to help remind
us that it is off mass shell and virtual). We will ignore the spin of all the particles,
so that we can concentrate on the vertex factors and propagators. The Feynman
diagram is:
4 This propagator is the relativistic equivalent of the non-relativistic version of the Lippmann-

Schwinger propagator (E − H + i)−1 that we found in non-relativistic scattering theory. Why

are the forms different? Non-relativistic propagators are Greens functions for integration over all
space. Relativistic propagators by contrast are Greens functions for integrations over both space
and time.

April 8, 2013 5
c A.J.Barr 2009-2013.
2.4 Relativistic propagators

e− (P1 ) µ− (P3 )

γ(P1 + P2 )

e+ (P2 ) µ+ (P4 )

where we have labelled the four-momenta of the external legs. The diagram shows
two vertices, and requires one propagator for the internal photon line. We can
calculate the photon’s energy-momentum four-vector Qγ from that of the electron
P1 and the positron P2 . Four momentum is conserved at each vertex so the
photon four-vector is Qγ = P1 + P2 . Calculating the momentum components in
the zero momentum frame:
P1 = (E, p), P2 = (E, −p). (2)
Conserving energy and momentum at the first vertex, the energy-momentum vector
of the internal photon is
Qγ = (2E, 0).
So this virtual photon has more energy than momentum.

The propagator factor for the photon in this example is then

1 1
= .
(2E)2 2
− mγ 4E 2
The contribution to Mf i from this diagram is obtained my multiplying this propa-
gator by two vectex factors each of size gEM . The modulus-squared of the matrix
element is then  2 2
2
g 
|Mf i | =  EM2  .
4E e
We can get the differential scattering cross section by inserting this |Mf i |2 into
Fermi’s Golden Rule with the appropriate density of states
dN p2µ dΩ
= ,
dpµ (2π)3
and divide by an incoming flux factor 2ve . The differential cross section is then
1 p2µ dpµ
dσ = 2π |Mf i |2 dΩ.
2ve (2π)3 d(E0 )
A little care is necessary in evaluating the density of states. Overall momentum
conservation means that only one of the two outgoing particles is free to contribute
to the density of states. The muon energy in the ZMF, for Eµ  mµ is Eµ = 12 E0 ,
so
dpµ 1 dpµ
= ,
dE0 2 dEµ
where pµ and Eµ are the momentum and energy of one of the outgoing muons.
Since those muons are external legs they are on-shell so that
p2µ + m2µ = Eµ2 .

April 8, 2013 6
c A.J.Barr 2009-2013.
2.5 Trees and loops

Taking a derivative pµ dpµ = Eµ dEµ . Inserting this into the F.G.R. we get

dpµ 1 dpµ 1 Eµ 1 1 1
= = = ≈ .
dE0 2 dEµ 2 pµ 2 vµ 2

We then integrate over all possible outgoing angles to gain a factor of 4π and note
p e− (P1 ) e− (P3 )
that g 2 /4π = α, and that Eµµ = vµ . Gathering all the parts together, and taking
the limit v → c we find we have a total cross-section for e+ + e− → µ+ + µ− of 5 γ
2
α
σ=π µ− (P2 ) µ− (P4 )
s
A Feynman diagram for
where s = (2E)2 is the square of the center-of-mass energy. electron–muon elastic scat-
tering, via photon (γ) ex-
A quick check of dimensions is in order. The dimensions of s are [E]2 , while those change.
of σ should be [L]2 = [E]−2 . The fine structure constant α is dimensionless, so the
equation is dimensionally consistent.

2.4.2 Other propagator examples

In the previous example the virtual photon’s four-momentum vector (E, 0) was
Feynman diagram for Compton
time-like. scattering, with a virtual internal
electron
In the electron–muon scattering case e− + µ− → e− + µ− the virtual photon
(γ ∗ ) is exchanged between the electron and the muon. The virtual photon carries
momentum and not energy, so the propagator is space-like.

To see this, transform to in the zero-momentum frame. In the ZMF the electron is
kicked out with the same energy as it came in with, so it has received no energy
from the photon, and conserving energy at the vertex Eγ = 0. The direction of
the electron momentum vector has changed so it has received momentum from the
photon, pγ 6= 0. Therefore Eγ2 − |p|2γ < 0 and the propagator is space-like.

An internal line requires a propagator regardless of the type of particle. An example

of a process in which an electron is the virtual particle is the Compton process in
which an electron scatters a photon

e− + γ → e− + γ.

2.5 Trees and loops

In principle to calculate |Mf i | we are supposed to draw and calculate all of the
infinite number of possible Feynman diagrams. Then we have to add up all those Some of the more complicated
complex numbers to get the total amplitude Mf i . loop diagrams for electron–muon
scattering.
5 Neglecting spin and relativistic normalization and flux factor issues – see ‘caveats’.

April 8, 2013 7
c A.J.Barr 2009-2013.
However in practice we can get away with just summing the simplest diagram(s).
To see why, we first note that the electromagnetic fine structure constant is small
(αEM  1)

The simplest “tree level” scattering diagram has two vertices so contains two factors
of gEM . The diagrams with the loops contain four vertices and hence four factors of
2
gEM . Since gEM /4π = αEM  1, we can see that the more complicated diagrams
with with more vertices will (all other things being equal) contribute much less to
the amplitude than the simplest ones since they contain higher powers of αEM . This
process of truncating the sum of diagrams is a form of perturbation theory.

In general tree diagrams are those without closed loops. Loop diagrams – those with
internal closed loops – tend to have larger powers of the coupling constant. A good
approximation to Mf i can usually be obtained from the sum of the amplitudes for
the ‘leading order’ diagrams – those with the smallest power of αEM that make a
non-zero contribution to Mf i .

The other forces also have coupling constants, which have different strengths. The
strong force is so-called because it has a fine structure constant close to 1 which
is about a hundred times larger than αEM . In fact the weak force actually has
a larger coupling constant ≈ 1/29 than the electromagnetic force ≈ 1/137. The
reason why this force appears weak is because the force is transmitted by very heavy
particles (the W and Z bosons) so it is very short-range.

3 Key concepts
• Feynman (momentum-space) diagrams help us calculate relativistic, Lorentz-
invariant scattering amplitudes.

• Vertices are associated with dimensionless coupling constants g with vertex

factors that depend on the charge Qg

• Internal lines are integrated over all time and space so include all internal
time orderings.

• Intermediate/virtual/off-mass-shell particles have Q2 6= m2 and have prop-

1
agators .
Q2 − m2

• For fermions, arrows show the sense of particle flow. Anti-particles have
arrows pointing the “wrong way”.

Caveats
• Sometimes you will see books define a propagator with a plus sign on the
bottom line: 1/(q 2 + m2 ). One of two things is going on. Either (a) q 2

April 8, 2013 8
c A.J.Barr 2009-2013.
 is their notation for a four-vector squared, but they have defined the metric
(−, +, +, +) in the opposite sense to us so that q 2 = −m2 is their condi-
tion for being on-mass-shell or (b) q 2 is acually intended to mean the three-
momentum squared. A bit of context may be necessary, but regardless of the
convention used the propagator should diverge in the case when the virtual
particle approaches its mass-shell.

• We have not attempted to consider what the effects of spins would be. This
is done in the fourth year after the introduction of the Dirac equation – the
relativistic wave equation for spin-half particles. The full treatment is done in
e.g. Griffiths Chs. 6 & 7.

• We have played fast and loose with phase factors (at vertices and overall
phase factors). You can see that this will not be a problem so long as only one
diagram is contributing to Mf i , but clearly relative phases become important
when adding diagrams together.

• Extra rules are needed for diagrams containing loops, because the momenta in
the loops are not fully constrained. In fact one must integrate over all possible
momenta for such diagrams. We will not need to consider such diagrams in
this course.

• The normalization of the incoming and ougtoing states needs to be considered

more carefully. The statement “I normalize to one particle per unit volume”
is not Lorentz invariant. The volume of any box at rest will compress by a
factor of 1/γ due to length contraction along the boost axis when we Lorentz
transform it. For relativistic problems we want to normalize to a Lorentz
invariant number of particles per unit volume. To achieve this we convention-
ally normalize to 1/(2E) particles per unit volume. Since 1/(2E) also scales
like 1/γ it transforms in the same manner as V . Therefore the statement “I
normalize to 1/(2E) particles per unit volume” is Lorentz invariant.

Terminology

Mf i . . . . . . . . . . . . . . . . Lorentz invariant amplitude for |Ψi i → |Ψf i transition

Feynman diagram . . . Graphical representation of part of the scattering amplitude
Vertex . . . . . . . . . . . . . . Point where lines join together on such a graph
Constant coupling (g) Dimensionless measure of strength of the force
Vertex factor (Qg) . . The contribution of the vertex to the diagram
Propagator . . . . . . . . . . Factor of 1/(Q · Q − m2 ) associated with an internal line

Tree level / leading . Simplest diagrams for any process with the smallest
order . . . . . . . . . . . . . . number of g factors. Contain no closed loops.

April 8, 2013 9
c A.J.Barr 2009-2013.
References and further reading
• “Introduction to Elementary Particles” D. Griffiths Chapters 6 and 7 does
the full relativistic treatment, including spins, relativistic normalization and
relativistic flux factor.

• “Femptophysics”, M.G. Bowler – contains a nice description of the connection

between Feynman propagators and non-relativistic propagators.

• “Quarks and Leptons”, Halzen and Martin – introduction to the Dirac equa-
tion and full Feynman rules for QED including spin.

• “QED - The Strange Theory of Light and Matter”, Richard Feynman. Popular
book with almost no maths. Even a PPE student could understand it – if you
explained it slowly to him. In fact it has a lot to recommend it, not least that
you can buy it for about five points.

April 8, 2013 10
c A.J.Barr 2009-2013.