2024/2025

PARTICLE
PHYSICS
:: Feynman diagrams

Prof. N. Leonardo
the big picture
• the physics observables
‣ decay rates and cross sections

• how to measure
‣ need particle detectors!

• matrix element
‣ need Dirac equation, interaction by particle exchange, Feynman rules

• phase space
‣ need relativistic QM calculations (Lorentz invariant)
‣ of particle decay rates and cross sections

2
 observable ~ physics x kinematics

started from
quantum mechanics:

relativistic
quantum mechanics
Lorentz invariant expression

initial state dynamics energy & mom. LIPS

inv. amplitude conservation 3
(kinematics)

Decay rate Cross section

CoM:

Matrix element, M 3
 from Feynman diagrams to Physics
10
d2σ/dpTdy × Β(µµ) (nb/(GeV/c))

CMS s = 7 TeV, L = 36 pb-1

1 CMS data
NLO NRQCD
NLO CSM
10-1 NNLO* CSM
PYTHIA (normalized)
CEM
10-2 CASCADE

(a)
10-3

10-4 Υ (1S)
|yΥ | < 2
-5
10

0 5 10 15 20 25 30 35 40 45 50
pYT (GeV/c)

Γfi — rate
number of Tfi — transition matrix element ⍴(Ef) — density
transitions per of final states
unit time from
initial state
to final state
contains the fundamental particle physics just kinematics
observable
4
Recap from last lecture
• we have derived free particle equations of motion
‣ non-relativistic: Schrödinger
‣ relativistic scalars: Klein-Gordon
‣ relativistic fermions: Dirac

• Dirac algebra
‣ achieved a relativistic equation rst-order in both time&space
‣ requiring introduction of 4x1 spinors and 4x4 Ɣ-matrix algebra
‣ leading to description for spin and antiparticles

• achieved solutions to the Dirac equation

‣ without interaction potential (V=0)

• today we like to introduce interactions

5
fi
Solutions of the free Dirac equation

particle and antiparticle helicity eigenstates

particles anti-particles

6
From free to interacting scenarios

• solved free Dirac equation with complete analytical solutions

• now, once we introduce interactions … this may not be even doable
?
↠ ↠

• in such cases, there is a general framework to the rescue

• perturbation theory allow us to obtain approximate solutions
• just as a function can be approximated by a Taylor series expansion

• Approximate? Expansion has in nite terms — which is just ne

• experimental measurements have (increasing but) nite precision
• And it gets even better — expansion can be done with graphs!
fi
fi
fi
 Perturbation theory
• we can derive solution for the free theory
• we like to nd solutions for the full hamiltonian
• expressed in terms of the known solutions of free theory
• introduce perturbation parameter λ
• expand full theory eigenvalues and eigenfunctions in terms of λ

• rst order perturbation

• transition between states

fi
fi
Perturbation Theory

note: this is a starred section, provided for completeness,

material is non-examinable and may be skipped,
recaps QM aspects and derivations used next
 Perturbation theory (time-independent)
• we want to solve the equation of motion for the full Hamiltonian
• start from the unperturbed (solved) Hamiltonian with
• we now add a perturbation H’ re-writing through the parameter
• eigenfunctions & eigenvalues of full Hamiltonian can be written in powers of λ
and
• the equation for the full Hamiltonian can be now re-written as:

• or, collecting like powers of λ:

‣ to 0th order (λ0):

‣ to 1st order (λ1): ➥ take inner product with ψ0n ➥
‣ to 2nd order (λ2):

• to rst order

as H0 is Hermitian and 1st order correction to the energy

= expectation value of perturbation
in the unperturbed state
fi
➡︎
 • to get now the correction to the wave function, re-write the SE using 1st order terms
‣ to 1st order (λ1): ⬌
• express ψ1n in terms of the unperturbed states ({ψ0n}=complete set):

• inserting back into SE

• now take product of SE with ψ0l :

‣ for l=n: (i.e. recovered E1n expression)

‣ for l≠n: ⬌ ⬌

• transition amplitude from state ψi into state ψ0f (up to 1st order, and setting back λ=1)

• or simplifying the notation, we have for the perturbed transition amplitude

for time independent perturbation
theory derivation see e.g.
Grif ths, Intro QM, sec. 6.1
11
fi
Perturbation theory (time-dependent)
• in Lecture #2, we have used a time-dependent calculation to derive Fermi’s golden rule
• we can build up on that derivation, including now a re nement of the approximation
used therein, namely by iteratively extending that calculation to second order
• very similarly to what we have done earlier
➥ ➥

• for a time-independent perturbing Hamiltonian, we assumed , yielding

➥ ➥

• but now, instead of taking , we can take this estimate of

• obtaining

• same integration procedure that led us to the expression of Fermi’s golden rule

• leads as now to the re ned expression

that includes the second-order term Thomson: 2.3.6

12
fi
fi
Feynman diagrams
Perturbation theory & Feynman diagrams
• a Feynman diagram represents a term in the perturbation theory expansion
of the matrix element for a transition
• normally, a full matrix element contains an in nite number of diagrams
‣ total amplitude:

+ + +…

α α2 α3
• each diagram corresponds to a term in the perturbation expansion

• the observable involves the modulo square of the matrix element

‣ Γ ~ |M |^2 ~

14
fi
Interaction by particle exchange
• Transition rates calculated from Fermi’s Golden Rule

• T is perturbation expansion for the Transition Matrix Element

‣ for particle scattering rst two terms in perturbation series represented as

f
f
“scattering in “scattering via an
a potential” j intermediate state”

i
i

‣ “Classical picture” – particles act as sources for fields which give rise to a
potential in which other particles scatter – “action at a distance”
‣ “QFT picture” – forces arise due to the exchange of virtual particles
— no action at a distance + forces between particles now due to particles
15
fi
 Feynman diagrams and time ordering
• interactions by particle exchange are represent by Feynman diagrams
• that are different time orderings possible
e– νe e– νe

W– W+
νµ µ– νµ µ–

• a Feynman diagram represents the effect of all time orderings

• Feynman rules will allow to compute the associated matrix element M

inspection of time ordering will allow us to derive a core aspect, aka rule
16
the propagator: time ordering
• consider particle interaction a+b→c+d via exchange of particle x
• one possible space-time picture is a c
Vji

space
‣ intermediate state j: c+b+x x
‣ this time ordering: a emits x then b absorbs x
Vfj
b d
i j f
time
• the ‘LI’ transition amplitude M is related T as

• is the Lorentz invariant matrix element for a → c + x

• the simplest LI quantity (from a scalar coupling) is taken here
ga is a measure of the strength
of the interaction a  c + x
17
the propagator: time ordering
a c x
ga

x gb
b d

• ‘LI’ matrix element for entire process

18
the propagator: time ordering
a c a c
ga x

space
x gb
b d

b d
i j f
time

• need to consider also the

other time ordering for
the process
• ‘LI’ matrix element for entire process ‣ can similarly get the ‘LI’
matrix element

19
 the propagator: time ordering
a c a c
ga x

space
x gb
b d

b d
i j f
time

• need to consider also the

other time ordering for
the process
• ‘LI’ matrix element for entire process ‣ can similarly get the ‘LI’
matrix element

Remarks: these partial amplitudes M are not Lorentz invariant; order of events in time depends on
frame; momentum is conserved at each vertex, but energy is not; particle x is “on-mass
20 shell”.
the propagator: time ordering
• in QM need to sum the amplitudes
‣ ie need to sum over the matrix elements corresponding to same nal state

a c
ga

is the propagator

propagator arises naturally from

After summing over all possible time interaction by particle exchange
orderings, M is Lorentz invariant ! fi
21
Feynman diagrams
• a Feynman diagram represents the sum over all possible time-orderings

c c

space
space

a a a c

b d b d
b d
time time

a c
‣ what's between the initial and nal state
is ‘how interaction happened’
‣ the relative position of the vertices with
respect to time does not matter
‣ Energy and momentum are conserved at
b d
each interaction vertex in the diagram
• The matrix element depends on
‣ the fundamental strength of the interaction at the vertices
‣ the four-momentum q2 carried by the (virtual) propagator particle
22
fi
 Virtual particles
“Time-ordered QM” Feynman diagram

a c a c a c

space
space

b d b d
b d
time time

• momentum conserved at vertices •momentum and energy conserved

• energy not conserved at vertices at interaction vertices
• exchanged particle “on mass shell” •exchanged particle “off mass shell”

VIRTUAL PARTICLE

•Can think of observable “on mass shell” particles as propagating waves and
unobservable virtual particles as normal modes between the source particles

23
Anatomy of a diagram

24
a toy model
a Scalar model
• computation of the transition matrix element M will be obtained
from the relevant Feynman diagram with the associated rules
• the Feynman rules will depend on the theory, i.e. the type of
interaction and the kinds of particles involved in the transition
‣ most elementary particles e.g. all of the leptons and quarks are spin 1/2
fermions — that obey the Dirac equation
‣ the SM gauge bosons have spin 1
‣ only one (known) elementary particle with no spin in Nature (the Higgs!)
‣ we have not yet derived the SM gauge interactions (QED,QCD,Weak)

• however a baseline set of rules does not depend on the theory

• for a rst application, to illustrate calculation of matrix elements,
we will introduce a toy model involving only particles without spin
‣ the model will consist of 3 distinct neutral spin 0 (scalar) particles
(ABC model, following Grif ths)
26
fi
fi
a Scalar model
• the model contains particles A,B,C
‣ each has spin 0 and is its own antiparticle
‣ masses satisfy mA > mB + mC
• the model has only one primitive vertex (by which particles interact)
‣ the associated coupling constant is g
• this is also the lowest order diagram describing decay
‣ with higher order processes involving closed loops
A→B+C
g g3

• scattering processes (examples, at lowest order)

A+B→A+B A+A→B+B

27
Feynman rules (scalar model)
1. draw all distinct ways of connecting the initial state to the final state,
at a given order in the interaction
2. for each external (scalar) line multiply by 1
3. for each vertex multiply by the factor −i g
4. for each internal line with momentum q multiply by the propagator

6. apply energy-momentum conservation at each vertex

7. for each loop choose one momentum q for one of the internal lines
and multiply by the factor

9. the result of the previous rules gives -iM, i.e. multiply result by i

28
A lifetime
• which particle?

• which diagrams?

• external lines
• vertices
• internal lines
• loops/integration

29
A lifetime
• as a rst application, we wish the to calculate the lifetime of A
‣ A→B+C
‣ B and C are stable, as their decay is kinematically not allowed (mA>mB,mC)
• the Feynman diagram coincides with the vertex
• application of Feynman rules gives
‣ M = 1 x 1 x 1 x (-ig) x i = g
• decay rate
recall decay
rate expression
➡

• lifetime recall
momentum in the rest frame
of decaying particle

30
fi
 Scattering A+A →B+B
• consider the scattering process
A+A → B+B
• energy momentum conservation gives

• Feynman rules give for each diagram

• where q2 stands for t (1st) and u (2nd)

• the total amplitude is

• cross section
x
➡

should express the Mandelstam variables in the CoM

β are the velocities in CoM,
is scattering angle particles 1,3
1/2 symmetry factor for identical particles
𝜽
[higher order processes]
• we’ve calculated processes at lowest order
• when we go to higher orders we start to run into problems
‣ we're not going to study how this is solved, but let’s illustrate
the issue
• consider, in the ABC model, corrections to the propagator of
particle A, also called the self-energy.
• applying the Feynman rules we get

• the integral is to be done from/to ∞, but it is logarithmically

divergent (!)

32
[renormalization]
• this problem (disaster!) of divergences has held up the development of
QED for a long time, taking more than 30 years to be fully understood
‣ this is achieved with a procedure known as renormalisation
• one starts by regularising the integral, via a cut-off procedure,
rendering it finite, without spoiling e.g. Lorentz invariance
•

• solving the integration, one realizes the M-depend terms (divergent)

appear in the form of additions to masses and coupling constants

• what we measure, are the physical quantities, not the bare mass/coupling
• also as a result coupling constants, depend on energy, they run (but slowly!)
• a non renormalisable theory would be meaningless
• the ABC theory is renormalizable, and so is QED
• in fact, all Gauge theories are, including the SM! (uff…)
33
Matrix Element Feynman diagrams + rules
Yukawa model
Yukawa: scalar ⊗ spinor
• we have derived the equations of motion for scalar (Klein-Gordon) and
spin-1/2 (Dirac) particles
• let’s now consider a theory with scalar bosons and spin-1/2 fermions

• interacting via a Yukawa coupling

36
 Yukawa: scalar ⊗ spinor
• we have derived the equations of motion for scalar (Klein-Gordon) and
spin-1/2 (Dirac) particles
• let’s now consider a theory with scalar bosons and spin-1/2 fermions

• interacting via a Yukawa coupling

• the matrix elements may be inferred by analogy with the previous model
a c

ABC ga gb
b d

ψ1 ψ2

Yukawa φ

ψ3 ψ4
37
Feynman rules (Yukawa model)

External lines

spin 0 1

Vertices -ig

spin 0
Internal lines
(propagators)
scattering: e e →e e (w/ Yukawa)
- - - -
e– e–
• consider the scattering process e-e-→e-e-
occurring in this model via φ-exchange
φ
• at leading order, there are two diagrams
contributing to this process e– e–

39
 scattering: e e →e e (w/ Yukawa)
- - - -
e– e–
• consider the scattering process e-e-→e-e-
occurring in this model via φ-exchange
φ
• at leading order, there are two diagrams
contributing to this process e– e–

t-channel

u-channel

40
 towards the cross section
• for each diagram: if initial and nal spins are unknown, need to
form average of the amplitude over initial state spins and sum over
nal state spins

• the amplitudes of each contributing diagram need to be summed

• and squared

• decay rates and cross sections can be nally computed using the
previously derived formulae
for a decay for scattering in CoM for scattering in Lab frame

41
fi
fi
fi
cross section: e e →e e (w/ Yukawa)
- - - -

• the total amplitude squared for the process is

• with 16 spin combinations!

we’re not doing these

spin calculations today!

• yielding eventually to the cross section

➠ … ➠

• this illustrates an (intentionally fairly complex) amplitude calculation with spin:

we’ll later explore calculation techniques, properties of the underlying interactions,
and suitable regimes — that will allow to readily compute our observables. 42
[remarks]
• we have now managed to close the loop — congrats!
‣ we’ve derived from rst principles (quantum mechanics & special relativity)
the procedures that allow us to compute the particle physics observables
‣ derived complete expressions, starting from phase space calculations
(kinematics), and the procedure to compute the transition matrix element

• we have set up the language and baseline framework we’ll be using

‣ Feynman diagrams and rules, Dirac spinors, …

• what's next?
‣ how to measure the observables: particle detection and identi cation
‣ derive the SM interactions — start by exploring processes in QED

43
fi
fi
the big picture
• the physics observables
‣ decay rates and cross sections

• how to measure
‣ need particle detectors!

• matrix element
‣ need Dirac equation, interaction by particle exchange, Feynman rules

• phase space
‣ need relativistic QM calculations (Lorentz invariant)
‣ of particle decay rates and cross sections

44
 • Thomson
‣ chapter 5 (sections 5.1-5.2)
• Grif ths
‣ chapter 6
• Romão
‣ section 2.5
• Barr et al
‣ sections 2.5, 3.4-3.7
• Angelis&Pimenta
‣ sections 2.5–2.7

• plan for next lecture

‣ Thomson: sections 1.3-1.4
‣ Barr et all: chapter 4
‣ Angelis&Pimenta: chapter 4

45
fi
Answers to some selected questions
some of You have asked recently
and that may be interesting to share
on 4-index contraction
Question: why do ɣμ∂μ and ɣμpμ contract di erently?

⬌
➥ ➥

Clearly, it must be so!

And that’s because, under a Lorentz transformation,

unlike pμ or xμ that transform as contravariant 4-vectors,
the four-derivative ∂μ transforms as a covariant 4-vector

Thomson
sec. 2.2.2

47
ff
 on spinor orthogonality
from
Thomson

Question: are all these

spinors really orthogonal?

obvious but important remark: spinors are not 4-vectors.

48
 on spinor orthogonality {u1,u2} are orthogonal
from {v1,v2} are orthogonal
Thomson
{ui} are not orthogonal
❌ {vi} are not orthogonal
_i,vi} are not orthogonal
{u
{ui,vi} are orthogonal

but:

also:

obvious but important remark: spinors are not 4-vectors. ie {ui,vj} are orthogonal (w/appropriate metric..)
49
on the negative energy spinor interpretation
Question: is the positive energy solution really an antiparticle?
e+ (E>0)
e– (E<0) γ
γ

e– (E>0)
e– (E>0)
charge conservation at vertex charge conservation at vertex
• electron (E>0) in: -e • electron (E>0) in: -e
• electron (E<0) out: +e • positron (E>0) in: +e
• photon out: 0e • photon out: 0e
• total: 0e ✓ • total: 0e ✓

We’ve also derived last time the charge conjugation operator for spinors

anti-particles have opposing charge: particle spinors transform into

anti-particle spinors — and with same spin
 Charge conjugation for spinors
• the Dirac equation including interaction with the e.m. eld
‣ with the covariant derivative →
‣ is
• taking the complex conjugate and pre-multiplying by
➡

• de ning ➡
ψ’ has same mass
• conclude particle spinor  anti-particle spinor but opposite charge

• wave function ➡

‣ charge conjugation transforms particle

spinors u into antiparticle spinors v
51
fi
fi
measurements
 measurement: lifetime
• measure: number of decays as a function of (proper) (decay) time

— lifetime (particle)
t — proper time (event)
L — distance in detector

Lorentz boost factor

↦go to particle’s rest frame

need to:
- decay distance, i.e. vertex separation
- momentum
- luminosity ↪ precision detector!
measure charged particles and vertices
rst, need to reconstruct the events 53
fi
𝛕
measurement: lifetime
• measure: number of decays as a function of (proper) (decay) time

+ L(Background) experiment
theory

54
measurement: cross section
• measure: rate of particles produced in collision
• example: Higgs production cross-section at LHC: σ(pp
• start from the de nition of cross section
N=Lσ
• need to measure yields N (event counts) and the luminosity L
• but, we do not see Higgses, what we detect are its decay products
‣ particles are detected by reconstructing their decays, e.g. H→ɣɣ,WW,…
‣ thus what’s directly measured from data is the product σ x B
‣ i.e. need to account for the decay branching fraction

• always, need to account for backgrounds and detector ef ciencies

fi
fi
 measurement: cross section
• measure: number of particles produced as function of momentum
‣ differential cross section

σxB
σ(pp→H) σ(pp→Y)
σ(√s)
10

d2σ/dpTdy × Β(µµ) (nb/(GeV/c))

CMS s = 7 TeV, L = 36 pb-1
1 CMS data
NLO NRQCD
NLO CSM
10-1 NNLO* CSM
PYTHIA (normalized)
CEM
10-2 CASCADE

(a)
units: fb/GeV 10-3

10-4 Υ (1S)
|yΥ | < 2
10-5

0 5 10 15 20 25 30 35 40 45 50
pYT (GeV/c)

56
 FCC

Roughly one Higgs for every

10,000,000,000 pp interactions

much of the current focus is on reaching and studying rare events

Resonances
Resonances
• scattering can occur through a resonance A+B → R → C+D
‣ consider e.g. e+e-→Z0→μ+μ- (s-channel)

ɣ/Z

• process occurs both through ɣ and Z propagators

‣ 4-momentum of virtual particle = center of mass energy √s
‣ low √s << mZ: the QED process dominates due to mZ2 term in propagator
‣ high √s >> mZ: photon and Z-mediated processes are of same order
‣ √s ~ mZ: the process via Z boson exchange dominates (3 orders of magnitude)
• indeed it would appear propagator diverges at √s ~ mZ
‣ apparent issue due to Z-propagator not accounting for Z boson being unstable particle

59
 Resonances
• in QM, time dependence of wave function of stable particle rest frame: e-imt
• for unstable particle with total width Γ, it must be modi ed
→ →
• nite lifetime may be accounted for in propagator

→
• or in LI form the modulus gives the BW shape

→
• the cross section can be approximated by a Breit-Wigner line shape

make it LI by
multiplying by (E+ER)2
approximate E~ER=M0

60
fi
fi
Resonances

can measure resonance’s mass and width

recall: a model describing data
ought to include:
- in addition to the theoretical shape
(e.g. resonance Breit Wigner shape)
- backgrounds (other processes)
- detector e ciency
- detector resolution: BW(m,Γ)⊗G(σ)
61
ffi
3-body decays
• for a three-body decay, there are no longer enough constraints from 4-
momentum conservation to determine the energies of the decay products
• situation can be suitably described considering one decay product at rest
with the other two particles sharing the available energy as in 2-body decay
• most elegant approach to three-body decays is the Dalitz plot
• — a 2D plot of either 2 decay particle invariant mass pairs
•

clear π π
resonance
structures

62
Dalitz plots
• recent results (LHCb’2021) reporting discovery of new tetraquarks
• studied in decays

2103.01803

ccus

ccss
63
measurement tools

(next week!)