Lecture2017 PDF

Particle Physics 1
Lecture notes for the first year master course on

the electroweak part of the Standard Model
Nikhef - autumn 2017
Wouter Hulsbergen, Marcel Merk, Ivo van Vulpen

Contents
Preliminaries i
Introduction 1
i.1 Quantum fields . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1
i.2 The Yukawa interaction . . . . . . . . . . . . . . . . . . . . . . . . . . . 4
i.3 Feynman diagrams . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7
i.4 The Standard Model . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9
i.5 Units in particle physics . . . . . . . . . . . . . . . . . . . . . . . . . . . 12
i.6 Four-vector notation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 14
1 Wave Equations and Anti-Particles 17

1.1 Particle-wave duality . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 17
1.2 The Schrödinger equation . . . . . . . . . . . . . . . . . . . . . . . . . . 18
1.3 The Klein-Gordon equation . . . . . . . . . . . . . . . . . . . . . . . . . 20
1.4 Interpretation of negative energy solutions . . . . . . . . . . . . . . . . . 23
1.4.1 Dirac’s interpretation . . . . . . . . . . . . . . . . . . . . . . . . . 23
1.4.2 Pauli-Weisskopf interpretation . . . . . . . . . . . . . . . . . . . . 24
1.4.3 Feynman-Stückelberg interpretation . . . . . . . . . . . . . . . . . 24
Exercises . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 27
2 Perturbation Theory and Fermi’s Golden Rule 31

2.1 Decay and scattering observables . . . . . . . . . . . . . . . . . . . . . . 31
2.2 Non-relativistic scattering . . . . . . . . . . . . . . . . . . . . . . . . . . 33
2.3 Relativistic scattering . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 38
2.3.1 Normalisation of the Wave Function . . . . . . . . . . . . . . . . 39
2.3.2 Density of states and phase space factor . . . . . . . . . . . . . . 40
2.3.3 The Flux Factor . . . . . . . . . . . . . . . . . . . . . . . . . . . 42
2.3.4 Golden rules for cross-section and decay . . . . . . . . . . . . . . 43
Exercises . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 44
3 The Electromagnetic Field 47

3.1 The Maxwell Equations . . . . . . . . . . . . . . . . . . . . . . . . . . . 47
3.2 Gauge transformations . . . . . . . . . . . . . . . . . . . . . . . . . . . . 49
3.3 The photon . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 50
i
ii CONTENTS
3.4 Electrodynamics in quantum mechanics . . . . . . . . . . . . . . . . . . . 53

3.5 The Aharanov-Bohm E↵ect . . . . . . . . . . . . . . . . . . . . . . . . . 54
Exercises . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 57
4 Electromagnetic Scattering of Spinless Particles 59

4.1 Electromagnetic current . . . . . . . . . . . . . . . . . . . . . . . . . . . 59
4.2 Coulomb scattering . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 61
4.3 Spinless ⇡ K Scattering . . . . . . . . . . . . . . . . . . . . . . . . . . 63
4.4 Form factors . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 67
4.5 Particles and Anti-Particles . . . . . . . . . . . . . . . . . . . . . . . . . 67
Exercises . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 69
5 The Dirac Equation 71

5.1 Spin, spinors and the gyromagnetic ratio . . . . . . . . . . . . . . . . . . 71
5.2 The Dirac equation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 73
5.3 Covariant form of the Dirac equation . . . . . . . . . . . . . . . . . . . . 75
5.4 Dirac algebra . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 76
5.5 Adjoint spinors and current density . . . . . . . . . . . . . . . . . . . . . 77
5.6 Bilinear covariants . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 79
5.7 Solutions to the Dirac Equation . . . . . . . . . . . . . . . . . . . . . . . 79
5.7.1 Plane waves solutions with p = 0 . . . . . . . . . . . . . . . . . . 79
5.7.2 Plane wave solutions for p 6= 0 . . . . . . . . . . . . . . . . . . . . 80
5.8 Antiparticle spinors . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 82
5.9 Normalization of the wave function . . . . . . . . . . . . . . . . . . . . . 83
5.10 The completeness relation . . . . . . . . . . . . . . . . . . . . . . . . . . 84
5.11 Helicity . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 84
5.12 Charge current and anti-particles . . . . . . . . . . . . . . . . . . . . . . 86
5.13 The charge conjugation operation . . . . . . . . . . . . . . . . . . . . . . 87
Exercises . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 88
6 Spin-1/2 Electrodynamics 93
6.1 Feynman rules for fermion scattering . . . . . . . . . . . . . . . . . . . . 93
6.2 Electron-muon scattering . . . . . . . . . . . . . . . . . . . . . . . . . . . 96
6.3 Crossing: the process e e+ ! µ µ+ . . . . . . . . . . . . . . . . . . . . . 101
6.4 Summary of QED Feynman rules . . . . . . . . . . . . . . . . . . . . . . 104
Exercises . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 105
7 The Weak Interaction 107

7.1 Lifetimes and couplings . . . . . . . . . . . . . . . . . . . . . . . . . . . . 107
7.2 The 4-point interaction . . . . . . . . . . . . . . . . . . . . . . . . . . . . 109
7.3 Parity . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 111
7.4 Covariance of the wave equations under parity . . . . . . . . . . . . . . . 113
7.5 The V A interaction . . . . . . . . . . . . . . . . . . . . . . . . . . . . 114
7.6 The propagator of the weak interaction . . . . . . . . . . . . . . . . . . . 116
7.7 Muon decay . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 117
CONTENTS iii
7.8 Quark mixing . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 120

Exercises . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 125
8 Local Gauge Invariance 127

8.1 Symmetries . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 127
8.2 The principle of least action . . . . . . . . . . . . . . . . . . . . . . . . . 128
8.3 Lagrangian density for fields . . . . . . . . . . . . . . . . . . . . . . . . . 129
8.4 Global phase invariance and Noether’s theorem . . . . . . . . . . . . . . 131
8.5 Local phase invariance . . . . . . . . . . . . . . . . . . . . . . . . . . . . 131
8.6 Application to the Dirac Lagrangian . . . . . . . . . . . . . . . . . . . . 133
8.7 Yang-Mills theory . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 134
8.8 Historical interlude 1: isospin, QCD and weak isospin . . . . . . . . . . . 138
8.9 Historical interlude 2: the origin of the name “gauge theory” . . . . . . . 139
Exercises . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 140
9 Electroweak Theory 143

9.1 SU(2) symmetry for left-handed douplets . . . . . . . . . . . . . . . . . . 144
9.2 The Charged Current . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 146
9.3 The Neutral Current . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 148
9.4 Couplings for Z ! f f . . . . . . . . . . . . . . . . . . . . . . . . . . . . 151
9.5 The mass of the W and Z bosons . . . . . . . . . . . . . . . . . . . . . . 152
Exercises . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 154
10 The Process e e+ ! µ µ+ 157

10.1 Helicity conservation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 157
10.2 The cross section of e e+ ! µ µ+ . . . . . . . . . . . . . . . . . . . . . 158
10.2.1 Photon contribution . . . . . . . . . . . . . . . . . . . . . . . . . 159
10.2.2 Z 0 contribution . . . . . . . . . . . . . . . . . . . . . . . . . . . . 162
10.2.3 Correcting for the finite width of the Z 0 . . . . . . . . . . . . . . 163
10.2.4 Total unpolarized cross-section . . . . . . . . . . . . . . . . . . . 164
10.3 Near the resonance . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 165
10.4 The forward-backward asymmetry . . . . . . . . . . . . . . . . . . . . . . 167
10.5 The Z 0 decay width and the number of light neutrinos . . . . . . . . . . 168
Exercises . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 171
11 Symmetry breaking 173

11.1 Problems in the Electroweak Model . . . . . . . . . . . . . . . . . . . . . 173
11.2 A few basics on Lagrangians . . . . . . . . . . . . . . . . . . . . . . . . . 175
11.3 Simple example of symmetry breaking . . . . . . . . . . . . . . . . . . . 176
11.3.1 µ2 > 0: Free particle with additional interactions . . . . . . . . . 176
11.3.2 µ2 < 0: Introducing a particle with imaginary mass ? . . . . . . . 177
11.4 Breaking a global symmetry . . . . . . . . . . . . . . . . . . . . . . . . . 178
11.4.1 µ2 > 0 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 178
11.4.2 µ2 < 0 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 179
11.5 Breaking a local gauge invariant symmetry: the Higgs mechanism . . . . 180
iv CONTENTS
11.5.1 Lagrangian under small perturbations . . . . . . . . . . . . . . . . 181

11.5.2 Rewriting the Lagrangian in the unitary gauge . . . . . . . . . . . 181
11.5.3 Lagrangian in the unitary gauge: particle spectrum . . . . . . . . 182
11.5.4 A few words on expanding the terms with (v + h)2 . . . . . . . . 182
Exercises . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 183
12 The Higgs mechanism in the Standard Model 185

12.1 Breaking the local gauge invariant SU(2)L ⇥ U(1)Y symmetry . . . . . . 185
12.2 Checking which symmetries are broken in a given vacuum . . . . . . . . 186
12.3 Scalar part of the Lagrangian: gauge boson mass terms . . . . . . . . . . 187
12.3.1 Rewriting (Dµ )† (Dµ ) in terms of physical gauge bosons . . . . 188
12.4 Masses of the gauge bosons . . . . . . . . . . . . . . . . . . . . . . . . . 190
12.4.1 Massive charged and neutral gauge bosons . . . . . . . . . . . . . 190
12.4.2 Massless neutral gauge boson ( ): . . . . . . . . . . . . . . . . . . 191
12.5 Mass of the Higgs boson . . . . . . . . . . . . . . . . . . . . . . . . . . . 191
Exercises . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 191
13 Fermion masses, Higgs decay and limits on mh 195

13.1 Fermion masses . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 195
13.1.1 Lepton masses . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 196
13.1.2 Quark masses . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 197
13.2 Yukawa couplings and the origin of Quark Mixing . . . . . . . . . . . . . 198
13.3 Higgs boson decay . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 202
13.3.1 Higgs boson decay to fermions . . . . . . . . . . . . . . . . . . . . 202
13.3.2 Higgs boson decay to gauge bosons . . . . . . . . . . . . . . . . . 203
13.3.3 Review Higgs boson couplings to fermions and gauge bosons . . . 204
13.3.4 Higgs branching fractions . . . . . . . . . . . . . . . . . . . . . . 205
13.4 Theoretical bounds on the mass of the Higgs boson . . . . . . . . . . . . 205
13.4.1 Unitarity . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 205
13.4.2 Triviality and Vacuum stability . . . . . . . . . . . . . . . . . . . 206
13.4.3 Triviality: g, g0 , ht heavy Higgs boson ! upper limit on mh 207
13.4.4 Vacuum stability ⌧ g, g0 , ht light Higgs boson ! lower limit
on mh . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 208
13.5 Experimental limits on the mass of the Higgs boson . . . . . . . . . . . . 210
13.5.1 Indirect measurements . . . . . . . . . . . . . . . . . . . . . . . . 210
13.5.2 Direct measurements . . . . . . . . . . . . . . . . . . . . . . . . . 211
Exercises . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 213
14 Problems with the Higgs mechanism and Higgs searches 215

14.1 Problems with the Higgs boson . . . . . . . . . . . . . . . . . . . . . . . 215
14.1.1 Problems with the Higgs boson: Higgs self-energy . . . . . . . . . 215
14.1.2 Problems with the Higgs boson: the hierarchy problem . . . . . . 216
14.2 Higgs bosons in models beyond the SM (SUSY) . . . . . . . . . . . . . . 217
Exercises . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 218
CONTENTS v
A Some properties of Dirac matrices ↵i and 221
B Summary of electroweak theory 223

Preliminaries
These are the lecture notes for the Particle Physics 1 (PP1) master course that is
taught at Nikhef in the autumn semester of 2014. These notes contain 14 chapters, each
corresponding to one lecture session. The topics discussed in this course are:
• Lecture 1 - 4: Electrodynamics of spinless particles
• Lecture 5 - 6: Electrodynamics of spin 1/2 particles
• Lecture 7: The weak interaction
• Lecture 8 - 10: Gauge symmetries and the electroweak theory
• Lecture 11-14: Electroweak symmetry breaking
Each lecture of 2 ⇥ 45 minutes is followed by a 1.5 hour problem solving session. The
exercises are included in these notes, at the end of each chapter.
The notes mainly follow the material as discussed in the books of Halzen and Martin.
The first ten chapters have been compiled by Marcel Merk in the period 2000-2011, and
updated by Wouter Hulsbergen for the PP1 courses of 2012 and 2013. The last four
chapters, written by Ivo van Vulpen, were added in 2014.
Literature
The following is a non-exhaustive list of course books on particle physics. (The comments
reflect a personel opinion of your lecturers!)
Thomson: “Modern Particle Physics”:
This is a new book (2013) that covers practically all the material in these lectures. If
you do not have another particle physics book yet, then we recommend that you acquire
this book.
Halzen & Martin: “Quarks & Leptons: an Introductory Course in Modern Particle
Physics ”:
This is the book that your lecturers used when they did their university studies. Though
most of the theory is timeless, it is a bit outdated when it comes to experimental results.
i
ii PRELIMINARIES
The book builds on earlier work of Aitchison (see below). Most of the course follows
this book, but it is no longer in print.
Griffiths: “Introduction to Elementary Particle Physics”, second, revised ed.
The text is somewhat easier to read than H & M and is more up-to-date (2008) (e.g.
neutrino oscillations) but on the other hand has a somewhat less robust treatment in
deriving the equations. The introduction chapter of this book gives a very readable
popular history of particle physics.
Aitchison & Hey: “Gauge Theories in Particle Physics”
Meanwhile in its 4th edition(2012), this 2-volume book provides a thorough theoreti-
cal introduction to particle physics, including field theory. It is excellent (notably its
’comments’ and appendices), but a bit more formal than needed for this course.
Perkins: “Introduction to High Energy Physics”, (1987) 3-rd ed., (2000) 4-th ed.
The first three editions were a standard text for all experimental particle physics. It is
dated, but gives an excellent description of, in particular, the experiments. The fourth
edition is updated with more modern results, while some older material is omitted.
Aitchison: “Relativistic Quantum Mechanics”
(1972) A classical, very good, but old book, often referred to by H & M.
Burcham & Jobes: “Nuclear & Particle Physics”
(1995) An extensive text on nuclear physics and particle physics. It contains more
(modern) material than H & M. Formula’s are explained rather than derived and more
text is spent to explain concepts.
Das & Ferbel: “Introduction to Nuclear and Particle Physics”
(2006) A book that is half on experimental techniques and half on theory. It is more
suitable for a bachelor level course and does not contain a treatment of scattering theory
for particles with spin.
Martin and Shaw: “Particle Physics ”, 2-nd ed.
(1997) A textbook that is somewhere inbetween Perkins and Das & Ferbel. In my
opinion it has the level inbetween bachelor and master.
Particle Data Group: “Review of Particle Physics”
This book appears every two years in two versions: the book and the booklet. Both of
them list all aspects of the known particles and forces. The book also contains concise,
but excellent short reviews of theories, experiments, accellerators, analysis techniques,
statistics etc. There is also a version on the web: http://pdg.lbl.gov
The Internet:
In particular Wikipedia contains a lot of information. However, one should note
that Wikipedia does not contain original articles and they are certainly not re-
viewed! This means that they cannot be used for formal citations.
In addition, have a look at google books, where (parts of) books are online avail-
able.
iii
About Nikhef
Nikhef is the Dutch institute for subatomic physics, where the acronym originates from
”Nationaal Instituut voor Kern en Hoge Energie Fysica”. Nikhef is used to indicate
simultaneously two overlapping organisations:
• Nikhef is a national research lab of the Netherlands Organisation for Scientific
Research (NWO)
• Nikhef is also a collaboration between the Nikhef institute and several Dutch
universities: UvA and VU University (Amsterdam), UU (Utrecht), RU (Nijmegen)
and RUG (Groningen) In this collaboration all Dutch activities related to particle
physics are coordinated.
In addition there are contacts with the Universities of Twente, Leiden and Eindhoven.
For more information see the Nikhef web page: http://www.nikhef.nl.
The research at Nikhef includes both accelerator based particle physics and astro-particle
physics. The accelerator physics research of Nikhef is currently focusing on the LHC
experiments Alice, Atlas and LHCb. Each of these experiments search answers for open
issues in particle physics like the state of matter at high temperature, the origin of
mass, the mechanism behind missing antimatter and hope to discover new phenomena
like supersymmetry, new particles or extra dimensions.
A more recent development is the research field of astro-particle physics. It includes
Antares & KM3NeT (cosmic neutrino sources), Pierre Auger (high energy cosmic rays),
Advanced Virgo & ET (gravitational waves) and Xenon (dark matter). Nikhef houses
a theory departement with research on quantum field theory and gravity, string the-
ory, QCD (perturbative and lattice) and B-physics. Driven by the massive computing
challenge of the LHC, Nikhef also has a scientific computing departement active in the
development of a worldwide computing network to analyze the large datastreams from
the experiments.
Nikhef program leaders/contact persons:
Name office phone email

Nikhef director Stan Bentvelsen H232 5001 [email protected]
Theory Eric Laenen H323 5127 [email protected]
Atlas Wouter Verkerke H241 5134 [email protected]
B-physics Marcel Merk N243 5107 [email protected]
Alice Raimond Snellings N321 2178 r.snellings.nl
KM3Net Paul de Jong H253 2087 [email protected]
Pierre Auger Charles Timmermans - - [email protected]
Xenon Patrick Decowski H349 2145 [email protected]
Detector R&D Niels van Bakel H045 2255 [email protected]
Gravitational waves Frank Linde H044 5140 [email protected]
iv PRELIMINARIES
A very brief history of particle physics
The book of Griffiths starts with a nice historical overview of particle physics in the
previous century. This is a summary of key events:
Atomic Models
1897 Thomson: Discovery of Electron. The atom contains electrons as “plums in
a pudding”.
1911 Rutherford: The atom mainly consists of empty space with a hard and heavy,
positively charged nucleus.
1913 Bohr: First quantum model of the atom in which electrons circled in stable
orbits, quatized as: L = ~ · n
1932 Chadwick: Discovery of the neutron. The atomic nucleus contains both
protons and neutrons. The role of the neutrons is associated with the binding
force between the positively charged protons.
The Photon
1900 Planck: Description blackbody spectrum with quantized radiation. No inter-
pretation.
1905 Einstein: Realization that electromagnetic radiation itself is fundamentally
quantized, explaining the photoelectric e↵ect. His theory received scepticism.
1916 Millikan: Measurement of the photo electric e↵ect agrees with Einstein’s
theory.
1923 Compton: Scattering of photons on particles confirmed corpuscular character
of light: the Compton wavelength.
Mesons
1934 Yukawa: Nuclear binding potential described with the exchange of a quan-
tized field: the pi-meson or pion.
1937 Anderson & Neddermeyer: Search for the pion in cosmic rays but he finds a
weakly interacting particle: the muon. (Rabi: “Who ordered that?”)
1947 Powell: Finds both the pion and the muon in an analysis of cosmic radiation
with photo emulsions.
Anti matter
1927 Dirac interprets negative energy solutions of Klein Gordon equation as energy
levels of holes in an infinite electron sea: “positron”.
1931 Anderson observes the positron.
v
1940-1950 Feynman and Stückelberg interpret negative energy solutions as the positive
energy of the anti-particle: QED.
Neutrino’s
1930 Pauli and Fermi propose neutrino’s to be produced in -decay (m⌫ = 0).
1958 Cowan and Reines observe inverse beta decay.
1962 Lederman and Schwarz showed that ⌫e 6= ⌫µ . Conservation of lepton number.
Strangeness
1947 Rochester and Butler observe V 0 events: K 0 meson.
1950 Anderson observes V 0 events: ⇤ baryon.
The Eightfold Way
1961 Gell-Mann makes particle multiplets and predicts the ⌦ .
1964 ⌦ particle found.
The Quark Model
1964 Gell-Mann and Zweig postulate the existence of quarks
1968 Discovery of quarks in electron-proton collisions (SLAC).
1974 Discovery charm quark (J/ ) in SLAC & Brookhaven.
1977 Discovery bottom quarks (⌥ ) in Fermilab.
1979 Discovery of the gluon in 3-jet events (Desy).
1995 Discovery of top quark (Fermilab).
Broken Symmetry
1956 Lee and Yang postulate parity violation in weak interaction.
1957 Wu et. al. observe parity violation in beta decay.
1964 Christenson, Cronin, Fitch & Turlay observe CP violation in neutral K meson
decays.
The Standard Model
1978 Glashow, Weinberg, Salam formulate Standard Model for electroweak inter-
actions
1983 W-boson has been found at CERN.
1984 Z-boson has been found at CERN.
1989-2000 LEP collider has verified Standard Model to high precision.
vi PRELIMINARIES
Introduction
i.1 Quantum fields

As far as we can tell all ‘ordinary’ matter is made of elementary spin- 12 fermions, which
we call quarks and leptons. These particles interact via four types of interactions,
namely the electromagnetic force, the strong nuclear force, the weak nuclear force and
gravity. Besides gravity, these particles and interactions can be well described by a
relativistic quantum field theory (QFT). In this course we will only consider the electro-
magnetic and weak interaction, leaving quantum chromodynamics (QCD), the theory
of the strong interaction, to the Particle Physics 2 course. The quantum field theory
for electrodynamics is called quantum electrodynamics (QED). If the weak interaction
is included, it is called the ‘electroweak theory’.
Most of our knowledge of the physics of elementary particles comes from scattering
experiments, from decays and from the spectroscopy of bound states. The theory of
bound states in electrodynamics is essentially the hydrogen atom, which (apart from a
few subtle phenomena) is well described by non-relativistic quantum mechanics (QM).
We do not discuss the hydrogen atom: instead we concentrate on processes at high
energies.
To understand when the classical quantum theory breaks down it is useful to look at
typical distance (or energy) scales relevant for electromagnetic interactions. Consider
Coulomb’s law, the well-known expression for the electrostatic force between two elec-
trons in vacuum,
e2 1 ↵~c
F = 2
= , (i.1)
4⇡✏0 r r2
where e is the electron charge, ✏0 is the vacuum permittivity, and
e2
↵ ⌘ (i.2)
4⇡✏0 ~c
is the fine structure constant. The latter is dimensionless and its value is approximately
1/137. It is because of the fact that ↵ ⌧ 1 that perturbation theory works so well in
quantum electrodynamics.
The first typical distance scale is the Bohr radius, the distance at which an electron
circles around an infinitely heavy object (a ‘proton’) of opposite charge. Using just
1
2 PRELIMINARIES
classical mechanics and imposing quantization of angular momentum by requiring that

rp = ~, you will find (try!) that this distance is given by
~
rBohr = . (i.3)
↵me c
(A proper treatment in QM tells you that the expectation value for the radius is not
exactly the Bohr radius, but it comes close.) Hence, the velocity of the electron is
vBohr = p/m = ↵c , (i.4)
which indeed makes the electron in the hydrogen atom notably non-relativistic.
The second distance scale is the Compton wavelength of the electron. Suppose that
you study electrons by shooting photons at zero-velocity electrons. The smaller the
wavelength of the photon, the more precise you look. However, at some point the
energy of the photons becomes large enough that you can create a new electron. (In
our real theory, you can only create pairs, but that factor 2 is not important now.) The
energy at which this happens is when ~! = me c2 , or at a wavelength
2⇡~
e = . (i.5)
me c
Usually, we divide both sides by 2⇡ and speak of the reduced Compton wavelength ¯ e ,
just like ~ is usually called the reduced Planck’s constant. Note that ¯ e = ↵rBohr . In
electromagnetic collisions at this energy, classical quantum mechanics no longer suffices:
as soon as collisions involve the creation of new particles, one needs QFT.
Finally, consider the collisions of two electrons at even higher energy. If the electrons
get close enough, the Coulomb energy is sufficient to create a new electron. (Again,
ignore the factor two required for pair production.) Expressing the Coulomb potential
as V (r) = ↵~c/r, and setting this equal to me c2 , one obtains for the distance
↵~
re = . (i.6)
me c
Note that, taking into account the definition of ↵, this expression does not explicitly
depend on ~: you do not need quantization to compute this distance, which is why it
is usually called the classical radius of the electron. At energies this high lowest order
perturbation theory may not be sufficient to compute a cross-section. The e↵ect of
‘screening’ becomes important, amplitudes described by Feynman diagrams with loops
contribute and QED needs renormalization to provide meaningful answers.
Fortunately for most of us, we will not discuss renormalization in this course. In fact,
we will hardly discuss quantum field theory at all! Do not be disappointed, there are
two pragmatic reasons for this. First, a proper treatment requires a proper course with
some non-trivial math, which would leave insufficient time for other things that we
do need to address. Second, if you accept a little handwaving here and there, then
I.1. QUANTUM FIELDS 3
distance [m] energy [MeV]

rBohr 5.3 ⇥ 10 11 0.0037
¯e 3.9 ⇥ 10 13 0.511
re 2.8 ⇥ 10 15 70
Table i.1: Values for the Bohr radius, the reduced Compton wavelength of the electron and
the classical radius of the electron, and the corresponding energy.
we do not actually need QFT: starting from quantum mechanics and special relativity
we can derive the ’Born level’ — that is, ‘leading order’ — cross-sections, following a
route that allows us to introduce new concepts in a somewhat historical, and hopefully
enlightening, order.
However, before continuing and setting aside the field theory completely until chapter 8,
it is worthwhile to briefly discuss some relevant features of QFT, in particular those that
distinguish it from ordinary quantum mechanics. In QM particles are represented by
waves, or wave packets. Quantization happens through the ‘fundamental postulate’ of
quantum mechanics that says that the operators for space coordinates and momentum
coordinates do not commute,
[x̂, p̂] = i~ (i.7)
The dynamics of the waves is described by the Schrödinger equation. Scattering cross-
sections are derived by solving, in perturbation theory, a Schrödinger equation with a
Hamiltonian operator that includes terms for kinetic and potential energy. Usually we
expand the solution around the solution for a ‘free’ particle and write the solution as a
sum of plane waves. This is exactly what you have learned in your QM course and we
will come back to this in Lecture 2.
In QFT particles are represented as ‘excitations’ (or ‘quanta’) of a field q(x), a function
of space-time coordinates x. There are only a finite number of fields, one for each type
of particle, and one for each force carier. This solves one imminent problem, namely
why all electrons are exactly identical. In its simplest form QED has only two fields: one
for a spin- 12 electron and one for the photon. The dynamics of these field are encoded in
a Lagrangian density L. Equations of motions are obtained with the principle of least
action. Those for the free fields (in a Lagrangian without interaction terms) leads to
wave equations, reminiscent of the Schrödinger equation, but now Lorentz covariant.
Again, solutions are written as superpositions of plane waves. The fields are quantized
by interpreting the fields as operators and imposing a quantization rule similar to that
in ordinary quantum mechanics, namely
[q, p] = i~ (i.8)
where the momentum p = @L/@ q̇ is the so-called adjoint coordinate to q. (You may
remember that you used similar notation to arrive at Hamilton’s principle in your classi-
cal or quantum mechanics course.) The Fourier components of the quantized fields can
be identified as operators that create or destruct field excitations, exactly what we need
4 PRELIMINARIES
for a theory in which the number of particles is not conserved. The relation to classical
QM can be made by identifying the result of a ‘creation’ operator acting on the vacuum
as the QM wave in the Schrödinger equation.
That was a mouth full and you can forget most of it. One last thing, though: one very
important aspect of quantum field theory is the role of symmetries in the Lagrangian. In
fact, as we shall see in Lecture 8 and 9, the concept of phase invariance allows to define
the standard model Lagrangian by specifying only the matter fields and the symmetries:
once the symmetries are defined, the dynamics (the force carriers) come for free.
That said, we leave the formal theory of quantum fields alone. In the remainder of
this chapter we briefly discuss some concepts and the Standard Model. In Lecture 1
we formulate a relativistic wave equation for a spin-0 particle. In Lecture 2 we discuss
classical QM perturbation theory and Fermi’s Golden rule, which allows us to formalize
the computation of a cross-section. In Lecture 3, we show how the Maxwell equations
take a very simple form when expressed in terms of a new spin-1 field, which we identify
as the photon. In lecture 4 we apply the developed tools to compute the scattering of
spin-0 particles. In Lectures 5 and 6 we turn to spin- 12 field, which are considerably more
realistic given that all SM matter fields are indeed fermions. In Lectures 7 through 10,
we introduce the weak interaction, gauge theory and electroweak unification. Finally,
in Lectures 11-14 we look in more detail at electroweak symmetry breaking.
i.2 The Yukawa interaction
After Chadwick had discovered the neutron in 1932, the elementary constituents of
matter were the proton, the neutron and the electron. The force responsible for interac-
tions between charged particles was the electromagnetic force. A ‘weak’ interaction was
responsible for nuclear decays. Moving charges emitted electromagnetic waves, which
happened to be quantized in energy and were called photons. With these constituents
the atomic elements could be described, as well as their chemistry.
However, there were already some signs that there were more elementary particles than
just protons, neutrons, electrons and photons:
• Dirac had postulated in 1927 the existence of anti-matter as a consequence of his
relativistic version of the Schrödinger equation in quantum mechanics. (We will
come back to the Dirac theory later on.) The anti-matter partner of the electron,
the positron, was actually discovered in 1932 by Anderson (see Fig. i.1).
• Pauli had postulated the existence of an invisible particle that was produced in
nuclear beta decay: the neutrino. In a nuclear beta decay process NA ! NB + e
the energy of the emitted electron is determined by the mass di↵erence of the nuclei
NA and NB . It was observed that the kinetic energy of the electrons, however,
showed a broad mass spectrum (see Fig. i.2), of which the maximum was equal
I.2. THE YUKAWA INTERACTION 5
to the expected kinetic energy. It was as if an additional invisible particle of low

mass is produced in the same process: the (anti-) neutrino.
Figure i.1: The discovery of the positron as reported by Anderson in 1932. Knowing the
direction of the B field Anderson deduced that the trace was originating from an anti electron.
Question: how?
Furthermore, though the constituents of atoms were fairly well established, there was
something puzzling about atoms: What was keeping the nucleus together? It clearly
had to be a new force, something beyond electromagnetism. Rutherford’s scattering
experiments had given an estimate of the size of the nucleus, of about 1 fm. With
protons packed this close, the new force had to be very strong to overcome the repulsive
coulomb interaction of the protons. (Being imaginative, physicists simply called it the
strong nuclear force.) Yet, to explain scattering experiments, the range of the force had
to be small, bound just to the nucleus itself.
In an attempt to solve this problem Japanese physicist Yukawa published in 1935 a fun-
damentally new view of interactions. His idea was that forces, like the electromagnetic
force and the nuclear force, could be described by the exchange of virtual particles, as
illustrated in Fig. i.3. These particles (or rather, their field ) would follow a relativistic
wave-equation, just like the electromagnetic field.
In this picture, the massless photon was the carrier of the electromagnetic field. As we
will see in exercise 1.4.3 the relativistic wave equation for a massless particle leads to
an electrostatic potential of the form (in natural units, ~ = c = 1)
1
V (r) = ↵ . (i.9)
r
Because of its 1/r dependence, the force is said to be of ‘infinite range’.
6 PRELIMINARIES
1.0
Relative Decay Probability

0.8
0.00012
0.6 0.00008
Mass = 0
0.00004
0.4 Mass = 30 eV
0
18.45 18.50 18.55 18.60
0.2
0
2 6 10 14 18
Energy (keV)
Figure 1. The Beta Decay Spectrum for Molecular Tritium

The plot on the left shows the probability that the emerging electron has a particular
energy. If the electron were neutral, the spectrum would peak at higher energy and
would be centered roughly on that peak. But because the electron is negatively
charged, the positively charged nucleus exerts a drag on it, pulling the peak to a
lower energy and generating a lopsided spectrum. A close-up of the endpoint
(plot on the right) shows the subtle difference between the expected spectra for
a massless neutrino and for a neutrino with a mass of 30 electron volts.
Figure i.2: The beta spectrum as observed in tritium decay to helium. The endpoint of the
spectrum can be used to set a limit of the neutrino mass. Question: how?
By contrast, in Yukawa’s proposal the strong force were to be carried by a massive

particle, later called the pion. A massive force carrier leads to a potential of the form
r/R
e
U (r) = g2 (i.10)
r
which is called the ’One Pion Exchange Potential’. Since it falls of exponentially, it has
a finite range. The range R is inversely proportional to the mass of the force carrier and
for a massless carrier the expression reduces to that for the electrostatic potential.
In the exercise you will derive the relation between mass and range properly, but it
can also be obtained with a heuristic argument, following the Heisenberg uncertainty
principle. (As you can read in Griffith’s, whenever a physicists refers to the uncertainty
principle to explain something, take all results with a grain of salt.) In some interpreta-
tion, the principle states that we can borrow the energy E = mc2 to create a virtual
particle from the vacuum, as long as we give it back within a time t ⇡ ~/ E. With
the particle traveling at the speed of light, this leads to a range R = c t = c~/mc2 .
From the size of the nucleus, Yukawa estimated the mass of the force carrier to be ap-
proximately 100 MeV/c2 . He called the particle a meson, since its mass was somewhere
in between the mass of the electron and the nucleon.
In 1937 Anderson and Neddermeyer, as well as Street and Stevenson, found that cosmic
rays indeed consist of such a middle weight particle. However, in the years after, it
became clear that this particle could not really be Yukawa’s meson, since it did not
I.3. FEYNMAN DIAGRAMS 7
Figure i.3: Illustration of the interaction between protons and neutrons by charged pion
exchange. (From Aichison and Hey.)
interact strongly, which was very strange for a carrier of the strong force. In fact this
particle turned out to be the muon, the heavier brother of the electron.
Only in 1947 Powell (as well as Perkins) found Yukawa’s pion in cosmic rays. They
took their photographic emulsions to mountain tops to study the contents of cosmic
rays (see Fig. i.4). (In a cosmic ray event a cosmic proton scatters with high energy
on an atmospheric nucleon and produces many secondary particles.) Pions produced in
the atmosphere decay long before they reach sea level, which is why they had not been
observed before.
As a carrier of the strong force Yukawa’s meson did not stand the test of time. We now
know that the pion is a composite particle and that the true carrier for the strong force is
the massless gluon. The range of the strong force is small, not because the force carrier
is massive, but because gluons carry a strong interaction charge themselves. However,
even if Yukawa’s original meson model did not survive, his interpretation of forces as
the exchange of virtual particles is still central to the description of particle interactions
in quantum field theory.
i.3 Feynman diagrams
Figure i.3 is an example of a Feynman diagram. You have probably seen Feynman
diagrams before and already know that they are not just pictures that help us to ‘visu-
alize’ a scattering process: they can actually be translated efficiently into mathematical
expressions for the computation of quantum mechanical transition amplitudes.
In this course, we will always draw Feynman diagrams such that time runs from left to
right. This is just a convention: the diagrams in Fig. i.3 are equally valid if time runs
from right to left, or from top to bottom, etc.
8 PRELIMINARIES
Figure i.4: A pion entering from the left decays into a muon and an invisible neutrino.
I.4. THE STANDARD MODEL 9
With this convention, the two diagrams in Fig. i.3 represent two di↵erent ways of scat-
tering a proton and a neutron via pion exchange: In case (a) a negative virtual pion is
first emitted by the neutron and then absorbed by the proton, while in the case (b) a
positive virtual pion is first emitted by the proton and then absorbed by the neutron.
As usual in quantum mechanics these complex amplitudes need to be added in order
to obtain the total amplitude. It turns out that only if both amplitudes are taken into
account, Lorentz-covariant results can be obtained in a quantum theory.
However, now that we know that both amplitudes must be taken into account, we no
longer need to draw both of them! In fact, in the remainder of this course, we will
always draw only one diagram, with the line that represents the pion exchange drawn
vertically. By convention, Feynman diagrams always present all possible time orderings
for the ‘internal’ lines, the virtual particles.
i.4 The Standard Model
In the Standard Model (SM) of particle physics all matter particles are spin- 12 fermions
and all force carriers are spin-1 bosons. The fermions are the quarks and leptons,
organized in three families (table i.2). The force carriers are the photon, the Z and W
and the gluons (table i.3).
charge Quarks
2 u (up) c (charm) t (top)
3 1.5–4 MeV 1.15–1.35 GeV (174.3 ± 5.1) GeV
1 d (down) s (strange) b (bottom)
3 4–8 MeV 80–130 MeV 4.1–4.4 GeV
charge Leptons
0 ⌫e (e neutrino) ⌫µ (µ neutrino) ⌫⌧ (⌧ neutrino)
< 3 eV < 0.19 MeV < 18.2 MeV
1 e (electron) µ (muon) ⌧ (tau)
0.511 MeV 106 MeV 1.78 GeV
Table i.2: Matter particles in the Standard Model, with their approximate mass.
Force Boson Coupling strength at 1 GeV

Strong g (8 gluons) ↵s ⇠ O(1)
Electromagnetic (photon) ↵ ⇠ O(10 3 )
±
Weak Z ,W (weak bosons) ↵W ⇠ O(10 8 )
Table i.3: Standard Model forces, the mediating bosons, and the associated strength of the
coupling at an energy of about 1 GeV. (The latter are taken from Thomson, 2013.)
10 PRELIMINARIES
In the SM forces originate from a symmetry by a mechanism called local gauge invari-
ance, discussed later on in the course. The strong force (or colour force) is mediated
by gluons, the weak force by the W and Z bosons, and the electromagnetic force by
photons. Only the charged weak interaction can change the flavour of quarks and lep-
tons: it allows for transitions between an up-type quark and a down-type quark, and
between charged leptons and neutrinos. Some of the fundamental diagrams are shown
in figure i.5.
e+ µ ⌫e µ q q
W g
a: b: c:
e µ +
e ⌫µ q q
Figure i.5: Feynman diagrams of fundamental lowest order perturbation theory processes in
a: electromagnetic, b: weak and c: strong interaction.
There is an important di↵erence between the electromagnetic force on one hand, and
the weak and strong force on the other hand. The photon does not carry charge and,
therefore, does not interact with itself. The gluons, however, carry colour and do interact
amongst each other. Also, the weak vector bosons carry weak isospin and undergo this
so-called self-coupling.
The strength of an interaction is determined by the coupling constant as well as the
mass of the vector boson. Contrary to its name the couplings are not constant, but
vary as a function of energy, which is called the running of the coupling constants. At
a momentum transfer of 1015 GeV the couplings of electromagnetic, weak and strong
interaction all obtain approximately the same value. (See figure i.6.) Grand unifica-
tion refers to the hypothesis that at high energy there is actually only a single force,
originating from a single gauge symmetry with a single coupling constant.
Due to the self-coupling of the force carriers the running of the coupling constants of the
weak and strong interaction are opposite to that of electromagnetism. Electromagnetism
becomes weaker at low momentum (i.e. at large distance), the weak and the strong force
become stronger at low momentum or large distance. The strong interaction coupling
becomes so large at momenta less than a few 100 MeV that perturbation theory is
no longer applicable. (The coupling constant is larger than 1.) Although this is not
rigorously proven, it is assumed that the self-coupling of the gluons is also responsible
for confinement: the existence of free coloured objects (i.e. objects with net strong
charge) is forbidden.
Confinement means that free quarks do not exist, at least, not at time-scales longer
than that corresponding to the range of the strong interaction. Quarks always appear
in bound states, either as combinations of three quarks (baryons) or as combinations
of a quark an an anti-quark (mesons). Together these are called hadrons. In the quark
model the various species of hadrons are organized by exploiting quark flavour symmetry,
the fact that equally charged quarks of di↵erent families are indistinguisable except for
I.4. THE STANDARD MODEL 11
their mass. Due to lack of time, we will not discuss the quark model in this course.
For reference table i.4 gives a list of common hadrons, some of which we encounter in
examples in the lectures.
name nickname symbol quark content mass / MeV

proton p uud 938.3
neutron n udd 939.6
charged pion pion ⇡+ ud¯ p 139.6
neutral pion pi-zero ⇡0 (uū dd)/¯ 2 135.0
charged kaon kaon K+ us̄ 493.7
neutral kaon K-zero K0 ds̄ 497.6
charged charmed meson D-plus D+ cd¯ 1869.6
neutral charmed meson D-zero D0 cū 1864.8
strange charmed meson D-sub-s Ds+ cs̄ 1968.3
charged bottom meson B-plus B+ ub̄ 5279.3
neutral bottom meson B-zero B0 db̄ 5279.6
strange bottom meson B-sub-s Bs0 sb̄ 5366.8
Table i.4: Name, quark content and approximate mass of common baryons and mesons. The
complete list of all known hadrons, together with a lot of experimental data, can be found in
the particle data book, http://pdglive.lbl.gov.
Finally, the Standard Model includes a scalar boson field, the Higgs field, which provides
mass to the vector bosons and fermions in the Brout-Englert-Higgs mechanism. The
motivation for the Higgs particle and corresponding precision tests of the SM are the
subject of the last four lectures of this course.
Figure i.6: Running of the coupling constants and possible unification point. On the left:
Standard Model. On the right: Supersymmetric Standard Model.
Despite the success of the standard model in describing all physics at ’low’ energy scale,
there are still many open questions, such as:
• why are the masses of the particles what they are?
• why are there 3 generations of fermions?
12 PRELIMINARIES
• are quarks and leptons truly fundamental?

• is there really only one Higgs particle?
• why is the charge of the electron exactly opposite to that of the proton? Or phrased
di↵erently: why is the total charge of leptons and quarks in one generation exactly
zero?
• is a neutrino its own anti-particle?
• can all forces be described by a single gauge symmetry (unification)?
• why is there no anti matter in the universe?
• what is the source of dark matter?
• what is the source of dark energy?
Particle physicists try to address these questions both with scattering experiments in
the laboratory and by studying high energy phenomena in the cosmos.
i.5 Units in particle physics

In particle physics we often make use of natural units to simplify expressions. In this
system of units the action is expressed in units of Planck’s constant
~ ⇡ 1.055 ⇥ 10 34
Js (i.11)
and velocity is expressed in units of the speed of light in vacuum
c = 2.998 ⇥ 108 m/s. (i.12)
such that all factors ~ and c can be omitted. As a consequence (see textbooks), there is
only one basic unit for length (L), time (T), mass (M), energy and momentum. In high
energy physics this basic unit is often chosen to be the energy in MeV or GeV, where
1 eV is the kinetic energy an electron obtains when it is accelerated over a voltage of
1V. Momentum and mass then get units of energy, while length and time get units of
inverse energy.
To confront the result of calculation with experiments the factors ~ and c usually need to
be reintroduced. There are two ways to do this. First one can take the final expressions
in natural units and then use the table i.5 to convert the quantities for space, time, mass,
energy and momentum back to their original counterparts. (For the positron charge,
see below.) Alternatively, one can express all results in GeV, then use the table i.6 with
conversion factors to translate it into SI units.
Where it concerns electromagnetic interactions, there is also freedom in choosing the
unit of electric charge. Consider again Coulomb’s law,
e2 1
F = , (i.13)
4⇡✏0 r2
I.5. UNITS IN PARTICLE PHYSICS 13
quantity symbol in natural units equivalent symbol in ordinary units

space x x/~c
time t t/~
mass m mc2
momentum p pc
energy E pE
positron charge e e ~c/✏0
Table i.5: Conversion of basic quantities between natural and ordinary units.
quantity conversion factor natural unit normal unit

mass 1 kg = 5.61 ⇥ 1026 GeV GeV GeV/c2
length 1 m = 5.07 ⇥ 1015 GeV 1 GeV 1
~c/ GeV
time 1 s = 1.52 ⇥ 1024 GeV 1 GeV 1
~/ GeV
Table i.6: Conversion factors from natural units to ordinary units.
where ✏0 is the vacuum permittivity. The dimension of the factor e2 /✏0 is fixed — it is
[L3 M/T2 ] — but this still leaves a choice of what to put in the charges and what in the
vacuum permittivity.
In the SI system the unit of charge is the Coulomb. (It is currently defined via the
Ampére, which in turn is defined as the current leading to a particular force between
two current-carrying wires. In the near future, this definition will probably be replaced
by the charge corresponding to a fixed number of particles with the positron charge.)
The positron charge expressed in Coulombs is about
19
e ⇡ 1.6023 ⇥ 10 C (i.14)
while the vacuum permittivity is
12
✏0 ⇡ 8.854 ⇥ 10 C2 s2 kg 1 m 3 . (i.15)
As we shall see in Lecture 3 the Maxwell equations look much more neat if, in addition
to c = 1, we choose ✏0 = 1. This is called the Heaviside-Lorentz system. Obviously, this
choice a↵ects the numerical value of e. However, note that coupling constant ↵, defined
in equation i.2, is dimensionless and hence independent of the system of units. In this
course we will often write e2 , when in fact we mean ↵.
Finally, it is customary to express scattering cross sections in barn: one barn is equal
to 10 24 cm2 .
14 PRELIMINARIES
i.6 Four-vector notation
We define the coordinate four-vector xµ as
xµ = (x0 , x1 , x2 , x3 ) (i.16)
where the first component x0 = ct is the time coordinate and the latter three components
are the spatial coordinates (x1 , x2 , x3 ) = x. Under a Lorentz transformation along the
x1 axis with velocity = v/c, xµ transforms as
0
x0 = (x0 x1 )
0
x1 = (x1 x0 )
0 (i.17)
x2 = x2
0
x3 = x3
p
where = 1/ 1 2.
A general ’contravariant four-vector’ is defined to be any set of four quantities Aµ =

(A0 , A1 , A2 , A3 ) = (A0 , A) which transforms under Lorentz transformations exactly as
the corresponding components of the coordinate four-vector xµ . Note that it is the
transformation property that defines what a ‘contravariant’ vector is.
Lorentz transformations leave the quantity
2
|A|2 = A0 |A|2 (i.18)
invariant. This expression may be regarded as the scalar product of Aµ with a related
‘covariant vector’ Aµ = (A0 , A), such that
X
A · A ⌘ |A|2 = Aµ Aµ . (i.19)
µ
From now on we omit the summation sign and implicitly sum over any index that
appears twice. Defining the metric tensor
0 1
1 0 0 0
B 0 1 0 0 C
gµ⌫ = g µ⌫ = B
@ 0
C (i.20)
0 1 0 A
0 0 0 1
we have Aµ = gµ⌫ A⌫ and Aµ = g µ⌫ A⌫ . A scalar product of two four-vectors Aµ and B µ

can then be written as
A · B = Aµ B µ = gµ⌫ Aµ B ⌫ . (i.21)
One can show that such a scalar product is indeed also a Lorentz invariant.
I.6. FOUR-VECTOR NOTATION 15
You will show in exercise 1.6 that if the contravariant and covariant four-vectors for the
coordinates are defined as above, then the four-vectors of their derivatives are given by
✓ ◆ ✓ ◆
µ 1@ 1@
@ = , r and @µ = ,r . (i.22)
c @t c @t
Note that the position of the minus sign is ‘opposite’ to that of the coordinate four-vector
itself.
16 PRELIMINARIES
Lecture 1
Wave Equations and Anti-Particles
1.1 Particle-wave duality

Ever since Maxwell we know that electromagnetic fields propagating in a vacuum are
described by a wave equation

1 @2
r2 (x, t) = 0 . (1.1)
c2 @t2
The solution to this equation is given by plane waves of the form
(x, t) = eik·x i!t

, (1.2)
where the wave-vector k and the angular frequency ! are related by the dispersion
relation
! = c|k| . (1.3)
(Of course, since the equation above is real, we can restrict ourselves to real solutions.
In fact, the photon field is real. However, it is often more convenient to work with
complex waves.) Maxwell identified propagating electromagnetic fields with light, and
thereby firmly established what everybody already knew: light behaves as a wave.
However, to explain the photo-electric e↵ect Einstein hypothesized in 1904 that light is
also a particle with zero mass. For a given frequency, lights comes in packets (‘quanta’)
with a fixed energy. The energy of a quantum is related to the frequency by
E = h⌫ = ~! , (1.4)
while its momentum is related to the wave-number
p = ~k . (1.5)
In terms of energy and momentum the dispersion relation takes the familiar form E = pc.
The idea of light as a particle was received with much skepticism and only generally
17
18 LECTURE 1. WAVE EQUATIONS AND ANTI-PARTICLES
accepted after Compton showed in 1923 that photons scattering of electrons behave as
one would expect from colliding particles.
So, by 1923 light was a wave and a particle: it satisfied a wave equation, yet it only came
about in packets of discrete energy. That lead De Broglie in 1924 to make another bold
preposition: if light is both a wave and a particle, then why wouldn’t matter particles
be waves as well? It took another few years before physicists established the wave-like
character of electrons in di↵raction experiments, but well before that people took De
Broglie hypothesis seriously and started looking for a suitable wave-equation for massive
particles.
The crucial element is to establish the dispersion relation for the wave. Schrödinger
started with the relativistic equation for the total energy
E 2 = m2 c4 + p2 c2 , (1.6)
but abandoned the idea, for reasons we will see later. He then continued with the
equation for the kinetic energy in the non-relativistic limit
p2
E = , (1.7)
2m
which, as we shall see now, led to his famous equation.
1.2 The Schrödinger equation
One pragmatic way to quantize a classical theory is to take the classical equations
of motion and substitute energy and momentum by their operators in the coordinate
representation,
@
E ! Ê = i~ and p ! p̂ = i~r . (1.8)
@t
Inserting these operators in Eq. (1.7), leads to the Schrödinger equation for a free par-
ticle,
@ ~2 2
i~ = r . (1.9)
@t 2m
In quantum mechanics we interprete the square of the wave function as a probability
density. The probability to find a particle at time t in a box of finite size V is given by
the volume integral
Z
P (particle in volume V , t) = ⇢(x, t) d3 x , (1.10)
V
where the density is

⇢(x, t) = | (x, t)|2 . (1.11)
1.2. THE SCHRÖDINGER EQUATION 19
Since total probability is conserved, the density must satisfy a so-called continuity equa-
tion
@⇢
+r·j =0 (1.12)
@t
where j is the density current or flux. When considering charged particles you can
think of ⇢ as the charge per volume and j as the charge times velocity per volume. The
continuity equation can then be stated in words as “The change of charge in a given
volume equals the current through the surrounding surface”.
What is the current corresponding to a quantum mechanical wave ? It is straightfor-
ward to obtain this current from the continuity equation by writing @⇢/@t = @ ⇤ /@t +
⇤
@ /@t and inserting the Schrödinger equation. However, because this is useful later
on, we follow a slightly di↵erent approach. First, rewrite the Schrödinger equation as
@ i~ 2
= r .
@t 2m
⇤
Now multiply both sides on the left by and add the expression to its complex conju-
gate
✓ ◆
⇤@ ⇤ i~
= r2
@t 2m
✓ ◆
@ ⇤ i~ ⇤
= r2
@t 2m
 +
@ i~
( ⇤ ) = r· ( r ⇤ ⇤
r ) (1.13)
@t | {z } 2m
⇢ | {z }
j
where in the last step we have used that r · ( ⇤ r r ⇤ ) = ⇤ r2 r2 ⇤ . In the

result we can recognize the continuity equation if we interpret the density and current
as indicated.
Plane waves of the form
= N ei(p·x Et)/~
(1.14)
with E = p2 /2m are solutions to the free Schrödinger equation. (In fact, starting
from the idea of particle-wave duality, Schrödinger took the plane wave form above and
‘derived’ his equation as the equation that described its time evolution.) We will leave
the definition of the normalization constant N for the next lecture: as the plane wave
is not localized in space (it has precise momentum, and infinitely imprecise position!),
it can only be normalized on a finite volume.
To get rid of the inconvenient factor ~ in the exponent, we can express energy and
momentum in terms of the wave vector and angular frequency defined above or work
in natural units. Since it is sometimes useful to verify expressions with a dimensional
analysis, we shall keep the factors ~ for now.
For the density of the plane wave we obtain

⇢ ⌘ ⇤
= |N |2 (1.15)
2
i~ |N |
j ⌘ ( r ⇤ ⇤
r )= p (1.16)
2m m
As expected, the density current is equal to the density times the non-relativistic velocity
v = p/m.
Any solution to the free Schrödinger can be written as a superposition of plane waves.
Ignoring boundary conditions (which usually limit the energy to quantized values), the
decomposition is written as the convolution integral
⇣ p ⌘ 3Z
(x, t) = ~ 2⇡ (p)ei(p·x Et)/~d3 p (1.17)
with E = p2 /2m. Note that for t = 0 this is just the usual Fourier transform. For
the exercises, remember that in one dimension the Fourier transform and its inverse are
given by (‘Plancherel’s theorem’),
Z +1 Z +1
1 ikx 1
f (x) = p F (k)e dk () F (k) = p f (x)e ikx dx (1.18)
2⇡ 1 2⇡ 1
If we replace p with p in the plane wave definition

out = N ei( p·x Et)/~
(1.19)
we still have a solution to the Schrödinger equation, since the latter is quadratic in coor-
dinate derivatives. Note that these solutions are already included in the decomposition
in Eq. (1.17).
By convention when describing scattering in terms of plane waves we identify those with
+p · x in the exponent as incoming waves and those with p · x as outgoing waves. In
one dimension, incoming waves travel in the positive x direction and outgoing waves in
the negative x direction.
Note that waves with E ! E are not solutions of the Schrödinger equation, but only
to its complex conjugate. That is di↵erent for solutions to the Klein-Gordon equation,
which we will describe next.
1.3 The Klein-Gordon equation

To find the wave equation for massive particles Schrödinger and others had originally
started from the relativistic relation between energy and momentum, Eq. (1.6). Using
again the operator substitution in Eq. (1.8) one obtains a wave equation
1 @2 m 2 c2
= r2 + (1.20)
c2 @t2 ~2
1.3. THE KLEIN-GORDON EQUATION 21
This equation is called the Klein-Gordon equation. Having seen it with the factors ~
and c included once, we will from now on omit them. The Klein-Gordon equation can
then be efficiently written in four-vector notation as
2 + m2 (x) = 0 , (1.21)
where
1 @2
2 ⌘ @µ @ µ ⌘ r2 (1.22)
c2 @t2
is the so-called d’Alembert operator.
Unlike the Schrödinger equation, the KG equation does not contain factors i. Conse-
quently, it can have both real and complex solutions. These have di↵erent applications.
In chapter 3 we shall see an example of the KG equation for a real field. In this section
we assume that the waves are complex.
Planes waves of the form
ipµ xµ
(x) = N ei(px Et)
=e (1.23)
with pµ = (E, p) are solutions of the KG equation provided that they satisfy the disper-
sion relation E 2 = p2 + m2 . Note that nothing restricts solution to have positive energy:
we discuss the interpretation of negative energy solutions later in this lecture.
Any solution to the KG equation can be written as a superposition of plane waves, like
for the Schrödinger equation. However, in contrast to the classical case, the complex
conjugate of the plane wave above
⇤ µ
(x) = N ei( px+Et)
= eipµ x (1.24)
is also a solution to the KG equation and need to be accounted for in the decomposition.
Note that it is not independent though, since ⇤ (p, E) = ( p, E). Consequently, we
can write the generic decomposition restricting ourselves to positive energy solutions, if
we write
Z
⇥ µ µ⇤
(x) = d3 p A(p) e ipµ x + B(p) eipµ x (1.25)
p
with E = + p2 + m2 . By popular convention, motivated later, we identify the first
exponent as an incoming particle wave, or an outgoing anti-particle wave, and vice-versa
for the second exponent.
In analogy to the procedure applied above for the non-relativistic free particle, we now
derive a continuity equation. We multiply the Klein Gorden equation for from the
left by i ⇤ , then add to the complex conjugate equation:

✓ 2 ◆
⇤ @
i 2
= i ⇤ r2 + m 2
@t
✓ 2 ⇤◆
@
i = i r2 ⇤ + m 2 ⇤
@t2
✓ ◆ +
⇤
@ ⇤ @ @ ⇤
i = r · [i ( r r ⇤ )] (1.26)
@t @t @t | {z }
| {z } j
⇢
where we can recognize again the continuity equation. In four-vector notation the con-
served current becomes
⇤ ⇤
j µ = (⇢, j) = i [ (@ µ ) (@ µ ) ] (1.27)
while the continuity equation is simply
@µ j µ = 0 (1.28)
You may wonder why we introduced the factor i in the current: this is in order to make
the density real.
Substituting the plane wave solution gives
⇢ = 2 |N |2 E
(1.29)
j = 2 |N |2 p
or in four-vector notation
j µ = 2 |N |2 pµ . (1.30)
Like for the the classical Schrödinger equation, the ratio of the current to the density
is still a velocity since v = p/E. However, in contrast to the non-relativistic case,
the density of the Klein-Gordon wave is proportional to the energy. This is a direct
consequence of the Klein-Gordon equation being second order in the time derivative.
We write the conserved current as a four-vector assuming that it transforms under
Lorentz transformation the way four-vector are supposed to do. It is not so hard to show
this by looking at how a volume and velocity change under Lorentz transformations (see
e.g. the discussion in Feynman’s Lectures, Vol. 2, sec. 13.7.) The short argument
is that since is a Lorentz-scalar, and @ µ a Lorentz vector, their product must be a
Lorentz vector.
You may remember that conservation rules in physics are related to symmetries. That
makes you wonder which symmetry leads to the conserved currents for the Schrödinger
and Klein-Gordon equations. In Lecture 8 we discuss Noether’s theorem and show that
it is the phase invariance of the Lagrangian, a so-called U (1) symmetry. The phase of
the wave functions is not a physical observable. For QM wave functions the conserved
current means that probability is conserved. For the QED Lagrangian it implies that
charge is conserved.
1.4. INTERPRETATION OF NEGATIVE ENERGY SOLUTIONS 23
E E
+m +m
−m −m
Figure 1.1: Dirac’s interpretation of negative energy solutions: “holes”
1.4 Interpretation of negative energy solutions
The constraint E 2 = p2 + m2 leaves the sign of the energy ambiguous. p

This leads to an
interpretation problem: what is the meaning of the states with E = p2 + m2 which
have a negative density? We cannot just leave those states away since we need to work
with a complete set of states.
1.4.1 Dirac’s interpretation
In 1927 Dirac o↵ered an interpretation of the negative energy states. To circumvent

the problem of a negative density he developed a wave equation that was linear in time
and space. The ‘Dirac equation’ turned out to describe particles with spin 1/2. (At
this point in the course we consider spinless particles. The wave function is a scalar
quantity as there is no individual spin “up” or spin “down” component. We shall discuss
the Dirac equation later in this course.) Unfortunately, this did not solve the problem
of negative energy states.
In a feat that is illustrative for his ingenuity Dirac turned to Pauli’s exclusion principle.
The exclusion principle states that identical fermions cannot occupy the same quantum
state. Dirac’s picture of the vacuum and of a particle are schematically represented in
Fig. 1.1.
The plot shows all the available energy levels of an electron. Its lowest absolute energy
level is given by |E| = m. Dirac imagined the vacuum to contain an infinite number
of states with negative energy which are all occupied. Since an electron is a spin-1/2
particle each state can only contain one spin “up” electron and one spin-”down” electron.
All the negative energy levels are filled. Such a vacuum (“sea”) is not detectable since
the electrons in it cannot interact, i.e. go to another state.

If energy is added to the system, an electron can be kicked out of the sea. It now gets
a positive energy with E > m. This means this electron becomes visible as it can now
interact. At the same time a “hole” in the sea has appeared. This hole can be interpreted
as a positive charge at that position: an anti-electron! Dirac’s original hope was that he
could describe the proton in such a way, but it is essential that the anti-particle mass
is identical to that of the electron. Thus, Dirac predicted the positron, a particle that
can be created by ’pair production’. The positron was discovered in 1931 by Anderson.
There is one problem with the Dirac interpretation: it only works for fermions!
1.4.2 Pauli-Weisskopf interpretation
Pauli and Weiskopf proposed in 1934 that the density should be regarded as a charge
density. For an electron the charge density is written as
jµ = ie( ⇤ @ µ @µ ⇤
). (1.31)
To describe electromagnetic interactions of charged particles we do not need to consider

anything but the movement of ‘charge’. This motivates the interpretation as a charge
current. Clearly, in this interpretation solutions with a negative density pose no longer
a concern. However, it does not yet solve the issue of negative energies.
1.4.3 Feynman-Stückelberg interpretation
Stückelberg and later Feynman took this approach one step further. Consider the current
for a plane wave describing an electron with momentum p and energy E. Since the
electron has charge e, this current is
j µ ( e) = 2e |N |2 pµ = 2e |N |2 (E, p) . (1.32)
Now consider the current for a positron with momentum p. Its current is
j µ (+e) = +2e |N |2 pµ = 2e |N |2 ( E, p) . (1.33)
Consequently, the current for the positron is identical to the current for the electron
but with negative energy and traveling in the opposite direction. Or, in terms of the
plane waves, to go from the positron current to the electron current, we just need to
µ
change the sign in the exponent of eixµ p . By our earlier convention, this is equivalent
to saying that the incoming plane wave of a positron is identical to the outgoing wave
of an electron.
Now consider what happens to the electron wave function if we change the direction of
time: We will have ct ! ct and p ! p. You immediately notice that this has exactly
the same e↵ect on the plane wave exponent as the transformation (E, p) ! ( E, p).
In other words, we can interprete the negative energy current of the electron as an
electron moving backward in time. This current is identical to that of a positron moving
forward in time.
e+ e−
E>0 E<0
Figure 1.2: A positron travelling forward in time is an electron travelling backwards in time.
This interpretation, illustrated in Fig. 1.2, is very convenient when computing scattering
amplitudes: in our calculations with Feynman diagrams we can now express everything
in terms of particle waves, replacing every anti-particle with momentum pµ by a particle
with momentum pµ , as if it were traveling backward in time. For example, the process
of an absorption of a positron with energy E is the same as the emission of an electron
with energy E (see Fig.1.3). Likewise, the process of an incoming positron scattering
o↵ a potential will be calculated as that of a scattering electron travelling back in time
(see Fig. 1.4).
−e
(+E,p)
absorption
time
emission
+e
(−E,−p)
Figure 1.3: There is no di↵erence between the process of an absorption of a positron with
pµ = ( E, p) and the emission of an electron with pµ = (e, p).
The advantage of this approach becomes more apparent when one considers higher order
corrections to the amplitudes. Consider the scattering of an electron on a localized
potential, illustrated in Fig. 1.5. To first order the interaction of the electron with the
perturbation is described by the exchange of a single photon. When the calculation is
extended to second order the electron interacts twice with the field. It is important to
note that this second order contribution can occur in two time orderings as indicated in
the figure. These two contributions are di↵erent and both of them must be included in
a relativistically covariant computation.
e+ −
e
time x
µ
Figure 1.4: In terms of the charge current density j+(E,p) (+e) ⌘ j µ (E,p) ( e)
The time-ordering on the right can be viewed in two ways:

• The electron scatters at time t2 runs back in time and scatters at t1 .
• First at time t1 “spontaneously” an e e+ pair is created from the vacuum. Later-
on, at time t2 , the produced positron annihilates with the incoming electron, while
the produced electron emerges from the scattering process.
The second interpretation would allow the process to be computed in terms of particles
and anti-particles that travel forward in time. However, the first interpretation is just
more economic. We realize that the vacuum has become a complex environment since
particle pairs can spontaneously emerge from it and dissolve into it!
e−
time
e− e−
t2 x t2 x
t1 t1 x
x
e − e−
Figure 1.5: First and second order scattering.

Exercises
Exercise 1.1 (Conversion factors)
Derive the conversion factors for mass, length and time in table i.6.
Exercise 1.2 (Kinematics of Z production)

The Z-boson has a mass of 91.1 GeV. It can be produced by annihilation of an electron
and a positron. The mass of an electron, as well as that of a positron, is 0.511 MeV.
(a) Draw the (dominant) Feynman diagram for this process.
(b) Assume that an electron and a positron are accelerated in opposite directions with
equal beam energies and collide head-on. Calculate the beam energy required to
produce a Z-boson.
(c) Assume that a beam of positron particles is shot on a target containing electrons.
Calculate the beam energy required for the positron beam to produce Z-bosons.
(d) This experiment was carried out in the 1990’s. Which method (b or c) do you
think was used? Why?
Exercise 1.3 (The Yukawa potential)

(a) The wave equation for an electromagnetic wave in vacuum is given by:
@2
2V =0 ; 2 ⌘ @ µ@ µ ⌘ r2
@t2
which in the static case can be written in the form of Laplace equation:
r2 V = 0
Now consider a point charge in vacuum. Exploiting spherical symmetry, show that
this equation leads to a ‘potential’ V (r) / 1/r.
Hint: look up the expression for the Laplace operator in spherical coordinates.
(b) The wave equation for a massive field is the Klein Gordon equation:
2 U + m2 U = 0
which, again in the static case can be written in the form:
r2 U m2 U = 0
Show, again assuming spherical symmetry, that Yukawa’s potential is a solution

of the equation for a massive force carrier. What is the relation between the mass
m of the force carrier and the range R of the force?
(c) Estimate the mass of the ⇡-meson assuming that the range of the nucleon force is
1.5 ⇥ 10 15 m = 1.5 fm.
Exercise 1.4 (Strangeness conservation (From A&H, chapter 1.))

Using the concept of strangeness conservation, explain why the threshold energy (for
⇡ incident on stationary protons) for
⇡ + p ! K 0 + anything
is less than for

0
⇡ + p ! K + anything
assuming that both processes proceed through the strong interaction.
Hint: Deduce what the minimal quark content of ‘anything’ is.
Exercise 1.5 (KG conserved current)

Verify that the conserved current Eq. (1.27) satisfies the continuity equation: compute
@µ j µ explicitly and make use of the KG equation to show that the result is zero.
Hint: note that that for any fourvectors A and B we have Aµ B µ = Aµ Bµ .
Exercise 1.6 (From A&H, chapter 3. See also Griffiths, exercise 7.1)
In this exercise we derive expression Eq. (i.22) of the Introductory chapter.
(a) Start with the expressions for a Lorentz transformation along the x1 axis in
Eq. (i.17). Write down the inverse transformation ( i.e. express (x0 , x1 ) in
0 0
(x0 , x1 ))
0 0
(b) Use the chain rule to express the derivatives @/@x0 and @/@x1 in the derivatives
@/@x0 and @/@x1 .
(c) Use the result to show that (@/@x0 , @/@x1 ) transforms in the same way as
(x0 , x1 ).
Exercise 1.7 (Wave packets (optional!))

In the coordinate representation a plane wave with momentum p = ~k is infinitely
dislocalised in space. This does not quite correspond to our picture of a particle, which
is why we usually visualize particles as wave packets, superpositions of plane waves that
have a finite spread both in momentum and coordinate space. As we shall see in the
following, the irony is that such wave packets are dispersive in QM: their size increases
as a function of time. So, even wave packets can hardly be thought of as representing
particles.
(a) Consider a one-dimensional Gaussian wave packet that at time t = 0 is given by
ax2 +ik0 x
(x, 0) = Ae
with a real and positive. Compute the normalization constant A such that
Z +1
| (x, 0)|2 dx = 1.
1
Hint: Z 1
y2
p
e dy = ⇡
1
(b) Take the Fourier transform to derive the wave function in momentum space at
t = 0,
✓ ◆1/4
1 2
(k) = e (k k0 ) /4a
2a⇡
Hint: You can write

⇥ ⇤
exp ax2 + bx = exp a(x b/2a)2 + b2 /4a
and then move the integration boundaries by b/2a. (Don’t mind that b is com-
plex.)
(c) Use this result and Eq. (1.17) to show that the solution to the Schrödinger equation
(with E(p) = p2 /2m or !(k) = ~k 2 /2m) is given by
✓ ◆1/4 ✓ ◆
2a 1/2 ax2 + ik0 x i⌫tk02 /4a
(x, t) = (1 + i⌫t) exp
⇡ 1 + i⌫t
with ⌫ ⌘ 2~a/m.
(d) Compute | (x, t)|2 . Qualitatively, what happens to 2
as time goes on?
(e) Now compute the same for a solution to the massless Klein-Gordon equation (! =
ck). Note that the wave packet maintains its size as a function of time.
Exercise 1.8 (The Quark Model (optional!))
(a) Quarks are fermions with spin 1/2. Show that the spin of a meson (2 quarks) can
be either a triplet of spin 1 or a singlet of spin 0.
Hint: Remember the Clebsch Gordon coefficients in adding quantum numbers.
In group theory this is often represented as the product of two doublets leads to
the sum of a triplet and a singlet: 2 ⌦ 2 = 3 1 or, in terms of quantum numbers:
1/2 ⌦ 1/2 = 1 0.
(b) Show that for baryon spin states we can write: 1/2 ⌦ 1/2 ⌦ 1/2 = 3/2 1/2 1/2
or equivalently 2 ⌦ 2 ⌦ 2 = 4 2 2
(c) Let us restrict ourselves to two quark flavours: u and d. We introduce a new
quantum number, called isospin in complete analogy with spin, and we refer to
the u quark as the isospin +1/2 component and the d quark to the isospin -1/2
component (or u= isospin “up” and d=isospin “down”). What are the possible
isospin values for the resulting baryon?
(d) The ++ particle is in the lowest angular momentum state (L = 0) and has
spin J3 = 3/2 and isospin I3 = 3/2. The overall wavefunction (L)space-part,
S)spin-part, I)isospin-part) must be anti-symmetric under exchange of any of
the quarks. The symmetry of the space, spin and isospin part has a consequence
for the required symmetry of the Colour part of the wave function. Write down
the colour part of the wave-function taking into account that the particle is colour
neutral.
(e) In the case that we include the s quark the flavour part of the wave function
becomes: 3 ⌦ 3 ⌦ 3 = 10 8 8 1. In the case that we include all 6 quarks it
becomes: 6 ⌦ 6 ⌦ 6. However, this is not a good symmetry. Why not?
Lecture 2
Perturbation Theory and Fermi’s

Golden Rule
In this chapter we discuss Fermi’s golden rule, which allows us to compute cross-sections
and decay rates. A very readable account of this is given in Griffiths chapter 6 and
Thomson chapter 3.
2.1 Decay and scattering observables
Most species of particles do not live long. This holds for all baryons except the proton
(even the neutron decays, when it is not inside a nucleus), but also for the muon and
the tau. As particles do not age, the probability to decay is independent of time. Given
a large number of particles N0 , the number of surviving particles is hence given by the
exponential law
N (t) = N0 e t/⌧ , (2.1)
where ⌧ is the mean lifetime. For particles that decay via the weak interaction, the mean
lifetime is typically 10 12 10 9 seconds. A notable exception is the neutron which lives
for about 15 minutes.
The mean lifetime is inversely proportional to what is called the decay width
~
= , (2.2)
⌧
which has units of energy. If the particle can decay through di↵erent decay channels
(e.g. a charged pion can decay to µ ⌫¯µ and to e ⌫¯µ ), then the decay width can be
written as the sum of the decay widths to the individual channels
X
= i . (2.3)
i
31
32 LECTURE 2. PERTURBATION THEORY AND FERMI’S GOLDEN RULE
The ratio i / is called the branching fraction. The particle data book is full of branching
fractions of species in the particle zoo.
If the decay is to more than two particles, the distribution of angles and energies of
particles in the final state becomes an observable as well. That is why we often consider
partial or di↵erential decay widths,
d
, (2.4)
dp1 · · · dpN
where p1 , . . . , pN are the momenta of the N particles in the final state.
Besides decay widths we also measure scattering cross-sections. (In fact, in our com-
putations, decays and scattering are quite similar, so we deal with both at once.) In
scattering experiments we collide beams of particles and study the collision rate. Con-
sider an experiment in which we scatter a beam of particles A on a target of particles
B. If nA is the particle number density in the beam, and vA is the particle velocity, the
number of collisions per second per unit volume of B is
dN
= vA nA nB tot . (2.5)
dt
The quantity tot is called the total scattering cross-section. It has units of ‘surface’.
In most cases we do not study the total collision rate, but rather the rate of particular
final states. The total cross-section is a sum of cross-sections for all possible final states,
such that X
tot = i . (2.6)
i
Since the energy and direction of final state particles can be measured as well, we usually
consider di↵erential scattering cross-sections,
d (A + B ! f1 + · · · + fN )
. (2.7)
dp1 · · · dpN
The expression for the calculation of a (di↵erential) cross section can be written schemat-
ically as
Wfi
d = d (2.8)
flux
The ingredients to this expression are:
1. the transition rate Wfi . You can think of this as the probability per unit time and
unit volume to go from an initial state i to a final state f ;
2. a flux factor that accounts for the ‘density’ of the incoming states;
3. the Lorentz invariant phase space factor d , sometimes referred to as ‘dLIPS’.
It accounts for the density of the outgoing states. (It takes care of the fact that
experiments cannot observe individual states but integrate over a number of states
with nearly equal momenta.)
2.2. NON-RELATIVISTIC SCATTERING 33
The ‘physics’ (the dynamics of the interaction) is contained in the transition rate Wfi .
The flux and the phase space factors are just ‘bookkeeping’, required to compare the
result with the measurements.
The rigorous computation of the transition rate requires quantum field theory, which
is outside the scope of this course. However, to illustrate the concepts we discuss non-
relativistic scattering of a single particle in a time-dependent potential and formulate
the result in a Lorentz covariant way. In the next chapter we will derive the lowest order
amplitude for the scattering of A + B ! A + B, which can still be done without field
theory. We can link that result to the ‘Feynman rules’ derived in field theory.
2.2 Non-relativistic scattering
H0
t=−T/2 t=T/2
t=0 ψ
f
H0
ψi V(x,t)
Figure 2.1: Scattering of a single particle in a potential.
Consider the scattering of a particle in a potential as depicted in Fig. 2.1 Assume that
both long before and long after the interaction takes place, the system is described by
the free Schrödinger equation,
@
i~ = H0 (2.9)
@t
where H0 is the unperturbed, time-independent Hamiltonian for a free particle. Let
m (x) be a normalized eigenstate of H0 with eigenvalue Em ,
H0 m (x) = Em m (x). (2.10)
The states m form an orthonormal basis,

Z
⇤ 3
m (x) n (x) d x = mn . (2.11)
We use the Kronecker delta, as if the spectrum of eigenstates is discrete. In chapter 2 we

considered a continuous spectrum of eigenstates for the free Hamiltonian, ‘numbered’
by the wave number k. Eventually, we could do that here, too, replacing the Kronecker
delta by a Dirac delta-function. However, it is trivial to change between the two and
the notation is a bit easier when we work with a discrete set of states.
The time-dependent wave function

iEm t/~
m (x, t) = m (x) e . (2.12)
is a solution to the Schrödinger equation. Since these states form a complete set, any
other wave function can be written as a superposition of the wave functions m .
Now consider a Hamiltonian that includes a time-dependent perturbation,
@
i~ = (H0 + V (x, t)) . (2.13)
@t
Any solution can be written as
1
X
iEn t
= an (t) n (x) e . (2.14)
n=0
where the an (t) are P

time-dependent complex coefficients. We require to be normalized,
which implies that |an (t)|2 = 1. The probability to find in state n at time t is just
|an (t)|2 .
To determine the coefficients an (t) we substitute (2.14) in (2.13) and find
X1 1
X
dan (t) iEn t iEn t
i~ n (x) e = V (x, t) an (t) n (x) e , (2.15)
n=0
dt n=0
where we have used that the m are solutions of the free Schrödinger equation. Multiply
the resulting equation from the left with f⇤ = ⇤f (x) eiEf t and integrate over x to obtain
X1 Z
dan (t)
i~ d3 x ⇤f (x) n (x) e i(En Ef )t~
=
dt
n=0 | {z }
fn
1
X Z
an (t) d3 x ⇤
f (x) V (x, t) n (x) e i(En Ef )t/~
(2.16)
n=0
Using the orthonormality relation for m we then arrive at the following coupled linear
di↵erential equation for ak (t),
1
dak (t) X
i~ = an (t) Vkn ei!kn t , (2.17)
dt n=0
where we have defined

!kn = (Ek En )/~ (2.18)
and what is sometimes called the transition matrix element
Z
Vkn (t) = d3 x ⇤k (x) V (x, t) n (x) . (2.19)
In some cases the set of equations (2.17) can be solved explicitly. A general solution is
obtained in perturbation theory, by expanding in Vkn . The approximation of order p + 1
can be obtained by inserting the p-th order result on the right hand side of Eq. (2.17),
(p+1)
dak (t) X
i~ ⇡ a(p)
n (t)Vkn (t)e
i!kn t
(2.20)
dt n
Without loss in generality we now assume that the incoming wave is prepared in eigen-
state i of the free Hamiltonian, i.e. ak ( 1) = ki . The zeroeth order approximation
(0)
then is ak (t) = ki (no interaction occurs) and the first order result becomes
(1)
dak (t)
i~ = Vki (t)ei!ki t (2.21)
dt
(1)
Using that af ( 1) = 0 and integrating this equation we obtain for the coefficient ak (t)
at time t,
Z t Z t
(1) daf (t0 ) 0 1 0
ak (t) = dt = Vki (t0 )ei!ki t dt0 for k 6= i (2.22)
1 dt i~ 1
Higher order approximations can be obtained by inserting the lowest order solution in
the right side of Eq. (2.20). (See textbooks.) A graphical illustration of the first and
second order perturbation is given in Fig. 2.2. Note that the lowest order approximation
makes one ‘quantum step’ from the initial state i to the final state f , while the second
order approximation includes all amplitudes i ! n ! f .
1−st order 2−nd order
f f
time
Vfn
space Vfi Vni
i i
Figure 2.2: First and second order approximation in scattering.
In the following we only consider the first order approximation (Born approximation).
We define the transition amplitude Tfi as the amplitude to go from a state i to a final
state f at large times,
Z 1 Z
1
Tfi ⌘ af (t ! 1) = dt d3 x f⇤ (x, t) V (x, t) i (x, t) (2.23)
i~ 1
where we substituted the definitions of Vkn and !kn . We can write the result more
compactly as
Z
1
Tfi = d4 x f⇤ (x) V (x) i (x) (2.24)
i~
Somewhat deceptively, the expression for Tfi seems to have a Lorentz covariant form.
However, as we have seen in the previous lecture, the ‘classical’ free particle waves cor-
responds to a density that does not correctly transform under Lorentz transformations.
Therefore, Tfi is actually not yet a proper Lorentz scalar.
We now make a simplification and consider a potential that is time-independent. The
expression for the transition amplitude then becomes
Z
Vfi 1 i!fi t
Tfi = e dt = 2⇡ i Vfi (Ef Ei ) (2.25)
i~ 1
where we have used that the integral is an important representation of the Dirac
function Z +1
1
(x) = eikx dk (2.26)
2⇡ 1
and substituted our definition of !fi . The function expresses conservation of energy.
Note that Tfi is dimensionless.
Can we interprete |Tfi |2 as a probability? Well, there is one conceptual problem and one
pragmatic problem. The conceptual problem is that if the potential is time-independent,
then this probability will just grow with time. The pragmatic problem is that there is
the function. These issues can be solved by considering a potential that is turned on
for a ‘finite time’ T . We define the mean transition rate in the limit for large T as
|Tfi |2
Wfi ⌘ lim . (2.27)
T !1 T
For an interaction that is turned on at time T /2 and turned o↵ at time T /2, the
equation above can be integrated to give for the transition amplitude at T /2,
Z T /2
Vfi 0 2Vfi sin(!fi T /2)
af (T /2) = ei!fi t dt0 = . (2.28)
i~ T /2 i~ !fi
Inserting this in the definition for the transition rate gives
4|Vfi |2 sin2 (!fi T /2)

Wfi ⌘ lim . (2.29)
T !1 ~2 !fi2 T
The function on the right is strongly peaked near !fi = (Ef Ei )/~ = 0, again enforcing
energy conservation. In fact, for T ! 1 it is yet another representation of the Dirac
function,
1 sin2 ↵x
(x) = lim . (2.30)
↵!1 ⇡ ↵x2
Substituting this in the equation, we obtain
2⇡ 2⇡
Wfi = |Vfi |2 (!fi ) = |Vfi |2 (Ef Ei ) . (2.31)
~ 2 ~
You can verify that Wfi is indeed a rate: Vfi is an energy, one of the factors of energy is
canceled by the function and the other one is divided by ~ to turn it into reciprocal
time.
As indicated before we can never actually probe final states with definite energy in a
measurement with finite duration. In general, there will be a number of states with
energy close to Ei that can be reached. Assuming that these states can be numbered by
a continuos variable n, the total transition rate can be written as an integral over these
final states
Z
W fi ⌘ Wfi dn
Z (2.32)
2⇡ 2
= |Vfi | (En Ei ) dn .
~
If ⇢(Ef ) is the density of states per unit energy near Ef , the number of final states with
energy between Ef and Ef + dEf is given by
dn = ⇢(Ef ) dEf . (2.33)
Inserting this in the expression above, we obtain Fermi’s (Second) Golden Rule,
Z
W fi ⌘ Wfi ⇢ (Ef ) dEf
(2.34)
2⇡
= |Vfi |2 ⇢ (Ei ) .
~
Note that in this expression ⇢(Ei ) is really the density of final states at the energy Ei .
Some textbooks therefore write this as ⇢(Ef )|Ef =Ei .
Above, we encountered a function in the transition amplitude. To deal with the square
of that function we considered a finite time interval and went back to the expression
(0)
for ak (T /2) for finite times T , taking the limit T ! 1 only after taking the square. To
make the final step you need to recognize the special representation of the function. For
future applications it is useful to know that one can also solve this problem di↵erently,
namely by taking the limit T ! 1 one integral at a time:
Z ! Z !⇤
1 Vfi T /2 i!fi t Vfi T /2 i!fi t0 0
|Wfi | = lim e dt e dt
T !1 T i~ T /2 i~ T /2
Z T /2
|Vfi |2 1
= 2⇡ (!fi ) lim dt0
~2 T !1 T T /2
| {z }
T
The final result is of course identical. We will encounter this ‘trick’ at various places
when going from a transition amplitude to a transition rate.
You may wonder why we need to consider a finite time interval T . The reason is that
when we assume that the initial state is an eigenstate of the free Hamiltonian with fixed
momentum (or energy), we have lost track of where a particle is in both space and
time. A moving wave packet would see the static potential during a finite time, but the
plane waves do not. Just like we will need to normalize the wave functions on a finite
volume, we will need to normalize the potential to a finite time. A proper treatment is
rather lengthy and relies on the use of wave packets. (See e.g. the book by K.Gottfried,
“Quantum Mechanics” (1966), Volume 1, sections 12, 56.) In the end, we can write
transition probabilities in terms of plane waves, provided that we normalize to T and
V . We discuss the normalization in more detail below.
2.3 Relativistic scattering

Fermi’s golden rule allows us to compute the scattering rate of non-relativistic particles
on a static potential. In scattering experiments at high energies we need to deal with two
scattering particles, rather than single particles scattering on a source. As an example,
consider two spin-less electrons scatter in their mutual electromagnetic field, as depicted
in Fig. 2.3.
C
e−f
Aµ
A
B
e−i
e−i
D e−f
Figure 2.3: Scattering of two electrons in an electromagnetic potential.
Such scattering processes can be described by the exchange of virtual particles, Yukawa’s
force carriers. Even without understanding the details of the interaction, we can readily
identify one place where it should di↵er from the discussion above: the result must
somehow encode four-momentum conservation and not just energy conservation.
Our master formula for the di↵erential cross-section, Eq. (2.8) is essentially a gener-
alization to problems with more than one particle in the initial or final state. We
cannot derive the expressions for a scattering cross section at high energies without
going through the machinery of quantum field theory. (This is not entirely true: see
Thomson, chapter 3 and section 5.1.) Instead, we will sketch the main results, then work
through the electrodynamics of spin-less particles as an example in the next lectures.
2.3. RELATIVISTIC SCATTERING 39
In quantum electrodynamics with scalar particles the transition amplitude Tfi for the
process A + B ! C + D still takes the form in Eq. (2.24). Performing the integral using
incoming and outgoing plane waves = N e ipx the result can be written as
Tfi = i NA NB NC ND (2⇡)4 4
(pA + pB pC pD ) M . (2.35)
where Ni are the plane wave normalization factors, which we will discuss shortly. The
-function takes care of energy and momentum conservation in the process. (Note that
the momentum vectors are four-vectors).
The quantity M is called the (Lorentz) invariant amplitude. It is computed using Feyn-
man diagrams. For topologies with n particles (counting both incident and final state),
the dimension of M is p4 n . Using the convention for the wave function normalization
described below, the invariant amplitude does not depend on arbitrary time intervals T
or normalization volumes V .
To find the transition probability we square the expression for Tfi ,
Z Z
2 2 i(pA +pB pC pD )x0
|Tfi | = |NA NB NC ND | |M| 2 4
d xe i(pA +pB pC pD )x
⇥ d4 x 0 e (2.36)
Z
= |NA NB NC ND |2 |M|2 (2⇡)4 4
(pA + pB pC pD ) ⇥ lim d4 x (2.37)
T,V !1 TV
= |NA NB NC ND |2 |M|2 (2⇡)4 4
(pA + pB pC pD ) ⇥ lim T V (2.38)
T,V !1
Since we now have a -function over 4 dimensions (the four-momentum rather than just
the energy), the integral becomes proportional to both T and V . To get rid of them we
consider a transition probability per unit time and per unit volume:
|Tfi |2
Wfi ⌘ lim
T,V !1 T V
= |NA NB NC ND |2 |M|2 (2⇡)4 (pA + pB pC pD ) (2.39)
To use this result in our master formula, we now need to discuss a few remaining
ingredients, namely the normalization of the wave functions, the flux factor and the
phase space factor.
2.3.1 Normalisation of the Wave Function
Above we defined the eigenstates of the free Hamiltonian to have unit normalization.
As we have seen in lecture 2 the eigenstates for free particles (for both the Schrödinger
equation and the Klein-Gordon equation) are plane waves
i(Et x·p)
(x, t) = N e . (2.40)
In contrast to wave packets the plane waves cannot be normalized over full space x
(which further on leads to problems when computing the square of -functions as above).
The solution is to apply so-called box normalization: we choose a finite volume V and
normalize all wave functions such that
Z
⇤
(x, t) (x, t)d3 x = 1 . (2.41)
V
p
For the plane waves this gives N = 1/ V . Like the time interval T , the volume V is
arbitrary and must drop out once we compute an observable cross-section or decay rate.
For the classical wave function the density ⇢ = | |2 so that the normalization gives one
particle per volume V . This normalization is not Lorentz invariant: under a Lorentz
transformation the volume element d3 x shrinks with a factor = E/m.
For the plane wave solutions of the Klein-Gordon equation, we had ⇢ = 2|N |2 E, which
with the box normalization becomes
⇢ = 2E/V. (2.42)
In other words, in the relativistic case we have 2E particles per volume V . It is customary
to use V = 1 and speak of 2E particles per unit volume. The factor 2E exactly cancels
the contraction of the volume, such that the number of particles in a given volume is
now Lorentz invariant.
Above we have introduced the Lorentz invariant amplitude without an explicit definition,
which is how we have found it in text books that do not derive the formalism with
field theory. Thomson takes an alternative approach: the plane wave functions p in the
classical and relativistic case only di↵er by the normalization p constant 2E. If we
label classical by and relativistic by 0 , then we have 0 = 2E . For a process
A + B + · · · ! 1 + 2 + · · · , we now define the Lorentz-invariant matrix element in terms
of the wave functions with relativistic normalization,
0 0 0 0
Mfi = h 1 2 · · · |V | A B ···i (2.43)
where V is the perturbation to the free Hamiltonian (and not the volume!). As the name
suggests, with this construction M is Lorentz invariant. The non-relativistic transition
element that appears in Fermi’s golden rule is then related to M by
p
Mfi = 2E1 · 2E2 · · · 2EA · 2EB Vfi . (2.44)
We will not further follow this strategy here.
2.3.2 Density of states and phase space factor
In the final step to Fermi’s golden rule we introduced the density of final states ⇢(E).
In the more general expression for the cross-section, it is the phase space factor that ac-
counts for the density of final states. It depends on the volume V and on the momentum
p of each final state particle.
Consider a cross-section measurement in which we measure the 3 components (px , py , pz )

of the momenta of all final state particles. As stated above it is customary to express
the cross-section as a di↵erential cross-section to the final state momenta,
Y
d = ... d3 pf (2.45)
f
where the product runs over all final state particles. To compute an actual number
for our experiment, we now convolute with experimental resolutions and integrate over
eventual particles or momentum components that we do not measure. (For example,
we often just measure the number of particles in a solid angle element d⌦.) For the
di↵erential cross-section the question of the number of accessible states should then be
rephrased as “how many states fit in the ‘momentum-space volume’ V d3 p”.
Assume that our volume V is rectangular with sides Lx , Ly , Lz . Using periodic boundary
conditions to ensure no net particle flow out of the volume we need to require that
Lx px = 2⇡~nx with nx integer. Hence, the total number of states in the range px to
px + dpx is dnx = Lx dpx /2⇡~. Since the total number of available states is n = nx ny nz ,
we find that the number of states with momentum between p and p + dp (i.e. between
(px , py , pz ) and (px + dpx , py + dpy , pz + dpz ) ) is:
V d3 p
dn = . (2.46)
(2⇡~)3
L
n
n= x
λx
Lz 2
1
Ly
Lx
Figure 2.4: Schematic calculation of the number of states in a box of volume V .
As explained above, in the relativistic case the wave functions are normalized such that
the volume V contains 2E particles. Therefore, the number of states per particle is:
V d3 p
# states/particle = (2.47)
(2⇡~)3 2E
If there are more particles in the final state, then the density of states in Fermi’s rule
must account for each of those. Consequently, the phase space factor for a process with
N final state particles becomes
N
Y V d3 pf
d = dLIPS = . (2.48)
f =1
(2⇡~)3 2Ef
In exercise 2.3 you will show that (ignoring V ) the phase space factor is indeed Lorentz
invariant. We will omit the factors ~ in what follows.
2.3.3 The Flux Factor
The flux factor or the initial flux corresponds to the number of particles that pass each
other per unit area and per unit time. It can be most easily computed in a frame in
which one of the particles is not moving. Consider the case that a beam of particles (A)
is shot on a target (B), see Fig. 2.5.
target
beam
A B
Figure 2.5: A beam incident on a target.
The number of beam particles that pass through unit area per unit time is given by
|vA | nA . The number of target particles per unit volume is nB . For relativistic plane
waves the density of particles n is proportional to ⇢ = 2E
V
such that
2|pA | 2mB
flux = |vA | na nb / (2.49)
V V
(Remember that in relativity v = p/E, modulo a factor c. For the KG waves we had
indeed that the current density was j = ⇢ p/E.) In exercise 2.2 you will show that the
kinematic factor |pA |mB is actually Lorentz invariant and that this expression can be
rewritten as
q
flux = 4 (pA,µ pµB )2 m2A m2B / V 2 (2.50)
The volume factor is not Lorentz invariant, but it will drop out later, as explained above.
Note that the incident flux as defined here is not actually a certain number of particles
per unit surface per unit time per unit volume: we need to account for the fact that it
is proportional to the square of an energy. The factors of energy will be accounted for
by the other ingredients to the cross-section formula.
2.3.4 Golden rules for cross-section and decay
Putting this all together, we arrive at the formula to calculate a cross section for the
process Ai + Bi ! Cf + Df + ...:
1
d fi = Wfi d
fluxZ
1
Tfi = d4 x f⇤ (x) V (x) i (x)
i~
|Tfi |2
Wfi = lim (2.51)
V,T !1 T V
YN
V d3 pf
d =
f =1
(2⇡)3 2Ef
q
flux = 4 (pA · pB )2 m2A m2B / V 2
In exercise 2.5 you will show that the cross-section is indeed independent on the volume
V.
Inserting the expression for the transition rate per unit time and volume, Eq. (2.39), we
find for the di↵erential cross-section of the process A + B ! C + D
(2⇡)4 4 (pA + pB pC pD ) d3 pC d3 pD
d = q · |M|2 · (2.52)
4 (pA · pB )2 m2A m2B (2⇡)3 2EC (2⇡)3 2ED
Note that the integrals of the flux factors are only over the spatial part of the outgoing
four-momentum vectors. The energy component has been integrated out, using the fact
that the outgoing particles
q are on the mass shell. Therefore, Ef is not an independent
variable, but equal to |pf |2 + m2f . This is important when performing integrals over
phase space.
In exercise 2.4 we calculate the integrals and flux factors in the centre-of-momentum
system, where pA + pB = pC + pD = 0. The result is
d 1 |pf |
= 2
|M|2 (2.53)
d⌦ cm 64⇡ s |pi |
where we defined pi ⌘ pA = pB , pf ⌘ pC = pD and s = (EA + EB )2 .

The computation of a decay rate for the process A ! C + D follows a similar strategy.
The result for the partial decay rate is
(2⇡)4 4
(pA pC pD ) d3 pC d3 pD
d = · |M|2 · (2.54)
2EA (2⇡)3 2EC (2⇡)3 2ED
p
which after integration of one of the momenta gives (4pi s ! 2EA = 2mA )
d 1
= 2 2
|pf | |M|2 (2.55)
d⌦ cm 32⇡ mA
Exercises
Exercise 2.1 (The Dirac -Function)
infinite
Consider a function defined by the following prescription
⇢
1/ for |x| < /2 surface = 1
(x) = lim
!0 0 otherwise
0
The integral of this function is normalized
Z 1
(x) dx = 1 (2.56)
1
and for any (reasonable) function f (x) we have

Z 1
f (x) (x) dx = f (0). (2.57)
1
These last two properties define the Dirac -function. The prescription above gives an
approximation of the -function. We shall encounter more of those prescriptions which
all have in common that they are the limit of a sequence of functions whose properties
converge to those given here.
(a) Starting from the defining properties of the -function, prove that
1
(kx) = (x) . (2.58)
|k|
(b) Prove that

n
X 1
(g (x)) = (x xi ) , (2.59)
i=1
|g 0 (xi )|
where the sum i runs over the 0-points of g(x), i.e.:g(xi ) = 0.
Hint: make a Taylor expansion of g around the 0-points.
Exercise 2.2 (Lorentz invariance of the flux)

Prove that (ignoring transformations of the volume V ) the flux factor derived in the lab
frame in Eq. (2.49) is indeed Lorentz-invariant by proving the identity
q
(pA · pB )2 m2A m2B = |pA |mB (2.60)
Hint: Since the left hand side is Lorentz invariant, you can compute it in any frame.
Note that pA · pB is an inner product of the four-vectors, not the three-vectors.
Exercise 2.3 (Lorentz invariance of the phase space factor)

Show that for any Lorentz invariant function M (p) of the Lorentz vector p, we have the
identity Z Z
4 2 2 0 d3 p
M (p) 2d p (p m ) ✓(p ) = M (E, p) . (2.61)
E
p
where ✓(p0 ) is the Heavyside function and on the right hand-side E = m2 + |p|2 is a
function of p. (On the left hand side, p0 it is still an independent integration variable.)
Argue that this result implies that
d3 p
(2.62)
2E
is Lorentz invariant.
Exercise 2.4 (AB ! CD cross-section in the c.m.s. See also H&M, Ex. 4.2)
In this exercise we derive a simplified expression for the A + B ! C + D cross-section
in the center-of-momentum frame.
(a) Start with the expression:
Z
d3 pC d3 pD
d = (2⇡)4 4 (pA + pB pC pD ) (2.63)
(2⇡)3 2EC (2⇡)3 2ED
Do the integral over d3 pD using the function and show that we can write:
Z
1 p2f dpf d⌦
d = (EA + EB EC ED ) (2.64)
(2⇡)2 4EC ED
where we have made use of spherical coordinates (i.e. d3 pC = |pC |2 d|pC | d⌦) and
defined pf ⌘ |pC |.
(b) In the C.M. frame
p we have |pA | = |pB | = pi and |pC | = |pD | = pf . Furthermore,
in this frame s ⌘ |pA + pB | = EA + EB ⌘ W . Show that the expression becomes
(hint: calculate dW/dpf ):
Z ✓ ◆
1 pf 1
d = dW d⌦ (W EC ED ) (2.65)
(2⇡)2 4 EC + ED
So that we finally get:
1 pf
d = p d⌦ (2.66)
4⇡ 2 4 s
(c) Show that the flux factor in the C.M. frame is:
p
flux = 4pi s (2.67)
and hence that the di↵erential cross section for a 2 ! 2 process in the center-of-
momentum frame is given by
d 1 pf
= 2
|M|2 (2.68)
d⌦ cm 64⇡ s pi
Exercise 2.5 (Box volume is arbitrary (optional!))

Show that the cross-section does not depend on the arbitrary size of the volume V :
identify all places where factors V enters in the summary in Eq. (2.51) and show that
they cancel.
Exercise 2.6 (Important representations of the -function (optional!))
(a) The delta-function can have many forms. One of them is:
1 sin2 ↵x
(x) = lim (2.69)
↵!1 ⇡ ↵x2
Make this plausible by sketching the function sin2 (↵x)/(⇡↵x2 ) for two relevant
values of ↵.
(b) Remember the Fourier transform,
Z +1
1
f (x) = g(k) eikx dk
2⇡ 1
Z +1 (2.70)
ikx
g(k) = f (x) e dx
1
Use this to show that another (important!) representation of the Dirac delta
function is given by Z +1
1
(x) = eikx dk (2.71)
2⇡ 1
Lecture 3
The Electromagnetic Field
3.1 The Maxwell Equations

In classical electrodynamics the movement of a point particle with charge q in an electric
field E and magnetic field B follows from the equation of motion
dp
= q (E + v ⇥ B). (3.1)
dt
The Maxwell equations tell us how electric and magnetic field are induced by static
charges and currents. In vacuum they can be written as:
⇢
Gauss’ law r·E = (3.2)
✏0
No magnetic charges r·B =0 (3.3)
@B
Faraday’s law of induction r⇥E+ =0 (3.4)
@t
1 @E j
Modified Ampére’s law 2
r⇥B = (3.5)
c @t ✏0
where ✏0 is the vacuum permittivity. From the first and the fourth equation we can
‘derive’ the continuity equation for electric charges, r · j = @⇢
@t
. It was the continuity
equation that lead Maxwell to add the time dependent term to Ampère’s law.
The constant c in the Maxwell equations is, of course, the velocity of light. When
Maxwell formulated his laws, he did not anticipate this. He did realize that c is the
velocity of a propagating electromagnetic wave. The value of c2 can be computed from
measurements of ✏0 (e.g. with the force between static charges) and measurements of
✏0 /c2 (e.g. from measurements of the force between static currents). From the fact that
the result was close to the known speed of light Maxwell concluded that electromagnetic
waves and light were closely related. He had, in fact, made one of the great unifications
of physics! For a very readable account, including an explanation of how electromagnetic
47
48 LECTURE 3. THE ELECTROMAGNETIC FIELD
waves travel, see the Feynman lectures, Vol.2, section 18. From now on we choose units
of charge such that we can set ✏0 = 1 and velocities such that c = 1. (That is, we use
so-called ’Heaviside-Lorentz rationalised units’. See section i.5.)
For what follows it is convenient to write the Maxwell equations in a covariant way (i.e.
in a manifestly Lorentz invariant way). As shown below we can formulate them in terms
of a single 4 component vector field, which we denote by Aµ = (V /c, A). As suggested
by our notation, the components of this field transform as a Lorentz vector.
You may prove for yourself that for any vector field A and scalar field V
divergence of rotation is 0: r · (r ⇥ A) = 0 (3.6)

rotation of gradient is 0: r ⇥ (rV ) = 0 . (3.7)
From your electrostatics course you may remember that, because the rotation of E is
zero (which is the same as saying that E is a conservative vector field), all physics can be
derived by considering a scalar potential field V . The electric field becomes the gradient
of the potential, E = rV . The potential V is not unique: we can add an arbitrary
constant and the physics will not change. Likewise, because the divergence of the B
field is zero, we can always find a vector field A such that B is the rotation of A.
So, let’s choose a vector field A such that
B = r⇥A (3.8)
and a scalar field V such that

@A
E= rV (3.9)
@t
Then, by virtue of the vector identities above, the Maxwell equations 3.3 and 3.4 are
automatically satisfied.
What remains is to write the other two equations, those that involve the charge density
and the charge current density, in components of A and A0 = V /c as well. You will
show in exercise 3.1 that these can be written very efficiently as
@µ @ µ A⌫ @ ⌫ @µ Aµ = j ⌫ . (3.10)
The current for electric charge j µ is a conserved current and transforms as a Lorentz
vector. (It is easy to work this out for yourself. See also Feynman, Vol.2, section 13.6.)
The derivative @ µ also transforms as a Lorentz vector. Therefore, if the equation above
is Lorentz covariant, then A⌫ must transform as a Lorentz vector as well. Showing that
the electromagnetic field indeed transform this way is outside the scope of these lecture,
but you may know that the transformation properties of the fields were an important
clue when Einstein formulated his theory of special relativity.
The expressions can be made even more compact by introducing the tensor
F µ⌫ ⌘ @ µ A⌫ @ ⌫ Aµ . (3.11)
3.2. GAUGE TRANSFORMATIONS 49
such that
@µ F µ⌫ = j ⌫ . (3.12)
Just as the potential V in electrostatics was not unique, neither is the field Aµ . Imposing
additional constrains on Aµ is called choosing a gauge. In the next section we shall
discuss this freedom in more detail. Written out in terms of the components E and B
the (4 ⇥ 4) matrix for the electromagnetic field tensor F µ⌫ is given by
0 1
0 Ex Ey Ez
B Ex 0 Bz By C
F µ⌫ = B
@ Ey Bz
C. (3.13)
0 Bx A
Ez By Bx 0
The field tensor is uniquely specified in terms of E and B. In other words, it does not
depend on the choice of the gauge.
3.2 Gauge transformations

For a given E and B field, the field Aµ is not unique. Transformations of the field Aµ
that leave the electric and magnetic fields invariant are called gauge transformations. In
exercise 3.2 you will show that for any scalar field (t, x), the transformations
@
V0 = V +
@t (3.14)
0
A = A r .
or in terms of four-vectors
Aµ ! A0µ = Aµ + @ µ (3.15)
do not change E and B.
If the laws of electrodynamics only involve the electric and magnetic fields, then, when
expressed in terms of the field A, the laws must be gauge ‘invariant’: physical observables
should not depend on . Sometimes we choose a particular gauge in order to make the
expressions in calculations simpler. In other cases, we exploit gauge invariance to impose
constraints on a solution, as with the photon below.
A common gauge choice is the so-called Lorentz gauge 1 . In exercise 3.3 you will show
that it is always possible to choose the gauge field such that Aµ satisfies the condition
Lorentz condition: @µ Aµ = 0. (3.16)

1
It is actually called the Lorenz condition, named after Ludvig Lorenz (without the letter ’t’). It
is a Lorentz invariant condition, and is frequently called the ”Lorentz condition” because of confusion
with Hendrik Lorentz, after whom Lorentz covariance is named. Since almost every reference has this
wrong, we will use ’Lorentz’ as well.
With this choice Aµ becomes a conserved current. In the Lorentz gauge the Maxwell
equations simplify further:
Maxwell equations in the Lorentz gauge: @µ @ µ A⌫ = j ⌫ (3.17)
However, as you will see in the exercise, Aµ still has some freedom since the Lorentz
condition fixes only @µ (@ µ ) and not @ µ itself. In other words a gauge transformation
of the form
Aµ ! A0µ = Aµ + @ µ with 2 = @µ @ µ = 0 (3.18)
is still allowed within the Lorentz gauge @µ Aµ = 0. Consequently, we can in addition
impose the Coulomb condition:
Coulomb condition: A0 = 0 (3.19)
(In combination with the Lorentz condition, also r · A = 0 with this choice of gauge.)
This choice of gauge is not Lorentz invariant. This is allowed since the choice of the
gauge is irrelevant for the physics observables, but it is sometimes considered less elegant.
3.3 The photon

Let us now turn to electromagnetic waves and consider Maxwell’s equations in vacuum
in the Lorentz gauge,
vacuum: j µ = 0 =) 2Aµ = 0. (3.20)
Each of the four components of Aµ satisfies the Klein-Gordon equation of a particle with
zero mass. (See Eq. (1.20).) This particle is the photon. It represents an electromagnetic
wave, a bundle of electric and magnetic field that travels freely through space, no longer
connected to the source. Using results below you can show that the E and B fields
of such a wave are perpendicular to the wave front and perpendicular to each other.
Furthermore, the magnitudes are related by the speed of light, |E| = c|B|.
We have seen before that the following complex plane waves are solutions of the Klein-
Gorden equation,
µ µ
(x) ⇠ e ipµ x and (x) ⇠ eipµ x (3.21)
For a given momentum vector p any solution in the complex plane is a linear combination
of these two plane waves. However, you may have noticed that, in contrast to the
Schrödinger equation, the Klein-Gorden equation is actually real. Since the E and B
fields are real, we restrict ourselves to solutions with a real field Aµ .
We could write down the solution to 2Aµ = 0 considering only the real axis, but it is
customary (and usually more efficient) to form the real solutions by combining the two
complex solutions,
Aµ (x) = aµ (p) e ipx
+ aµ (p)⇤ eipx
(3.22)
= 2 < aµ (p)e ipx
3.3. THE PHOTON 51
(Note that the second term is the complex conjugate of the first.) The four-vector aµ (p)
depends only on the momentum vector. It has four components but due to the gauge
transformation not all of those are physically meaningful. The Lorentz condition gives
0 = @µ Aµ = ipµ aµ e ipx
+ ipµ aµ⇤ eipx , (3.23)
which leads to
pµ aµ = 0. (3.24)
The Lorentz condition therefore reduces the number of independent complex components
to three. However, as explained above, we have not yet exhausted all the gauge freedom:
we are still free to make an additional shift Aµ ! Aµ +@ µ , provided that itself satisfies
the Klein-Gordon equation. If we choose it to be
= i↵e ipx
i↵⇤ e+ipx (3.25)
with ↵ a complex constant, then its derivative is
@ µ = ↵pµ e ipx
+ ↵⇤ pµ eipx . (3.26)
With a bit of algebra we see that the result of the gauge transformation corresponds to
aµ0 = aµ ↵pµ (3.27)
Note that aµ0 still satisfies the Lorentz condition only because p2 = 0 for a massless
photon.
As we have already seen, this additional freedom allows us to apply the Coulomb con-
dition and choose A0 = 0, or equivalently a0 (p) = 0. In combination with the Lorentz
condition this leads to
a·p=0 (3.28)
or p · A = 0.
At this point it is customary to write a(p) as a product of two terms
a(p) ⌘ N (p) ✏(p) (3.29)
where ✏ is a vector of unit length and N (p) is real. The normalization N (p) depends
only on the magnitude of the momentum and corresponds to the energy density of the
wave. The vector ✏ depends only on the direction of p and is called the polarization
vector. Choosing the z axis along the direction of the momentum vector and imposing
the gauge conditions, the latter can be parameterized as
✏ = (c1 ei 1 , c2 ei 2 , 0) . (3.30)
where ci and i are all real and c21 + c22 = 1. We can remove one phase by moving the
origin. (Just look at how a shift of the origin a↵ects the factors e±ipx .) Therefore, only
two parameters of the polarization vector are physically meaningful: these are the two
polarization degrees of freedom of the photon.
Any polarization vector can be written as a (complex) linear combination of the two
transverse polarization vectors
✏1 = (1, 0, 0) ✏2 = (0, 1, 0) . (3.31)
If the phases of the two components are identical, the

plight is said to be linearly polarized.
If the two components have equal size (c1 = c2 = 2) but a phase di↵erence of ±⇡/2,
the light wave is circularly polarized. The corresponding circular polarization vectors
are
✏1 i✏2 +✏1 i✏2
✏+ = p ✏ = p (3.32)
2 2
You will show in exercise 3.4 that the circular polarization vectors ✏+ and ✏ transform
under a rotation with angle ✓ around the z-axis (the momentum direction) as
✏+ ! ✏0+ = e i✓
✏+
0 i✓
(3.33)
✏ !✏ = e ✏
We now show that this means that these polarization states correspond to the two
helicity eigenstates of the photon.
You may remember from your QM course that the z component of the angular momen-
tum operator Jz is the generator of rotations around the z-axis. That means that for a
wavefunction (x) the e↵ect of an infinitesimal rotation around the z axis is given by
(x) ! U (") (x) ⌘ (1 i"J3 ) (x). (3.34)
An arbitrarily large rotation ✓ may be built up from infinitesimal rotations by dividing

it in small steps
U (✓) = lim (U (✓/n))n

n!1
✓ ✓ ◆ ◆n (3.35)
✓ i✓Jz
= lim 1 i Jz ) = e
n!1 n
Consequently, if is an eigen vector of Jz with eigenvalue m, then for any rotation
around the z-axis we have
im✓
(x) ! U (✓) (x) = e (x) (3.36)
Comparing this to the e↵ect of rotations on the polarization states above we now identify
✏+ with an m = +1 state and ✏ with an m = 1 state.
Apparently, the polarization states belong to a representation of the rotation group:
they are spin states. Since we find ±1 for the Jz quantum number the photon must be
a spin-1 representation: it could not be spin zero, because than you would have only
3.4. ELECTRODYNAMICS IN QUANTUM MECHANICS 53
have a state with m = 0. And it could not have higher spin state, because there are no
degrees of freedom in the photon field that could be identified with higher values of m.
Since the photon is spin-1, one could have expected to find 3 spin states, namely for
mz = 1, 0, +1. You may wonder what happened to the mz = 0 component. This
component was removed when we applied the Coulomb gauge condition, exploiting
p2 = 0, leading to A · p = 0. For massive vector fields (or virtual photons!), there is no
corresponding gauge freedom and a component parallel to the momentum (a longitudinal
polarization) remains. Massive vector fields have one spin degree of freedom more.
Another way to look at this is to say that to define spin properly one needs to boost to
the rest frame of the particle. For the massless photon this is not possible. Therefore,
we can talk only about helicity (spin projection on the momentum) and not about spin.
The equivalent of the mz = 0 state does not exist for the photon.
Finally, we compute the electric and magnetic fields. Substituting the generic expression
for Aµ in the definitions of E and B and exploiting the coulomb condition A0 ⌘ V = 0,
we find
E = i a p0 e ipx
+ c.c.
ipx
(3.37)
B= i (p ⇥ a) e + c.c.
Indeed, for the electromagnetic waves, the E and B fields are perpendicular to each
other and to the momentum, while the ratio of their amplitudes is 1 (or rather, c).
3.4 Electrodynamics in quantum mechanics

In classical mechanics an elegant way to introduce electrodynamics is via a method called
minimal substitution. The method states that the equation of motion of a charged par-
ticle under the influence of a vector field Aµ can be obtained by making the substitution
pµ ! pµ qAµ . (3.38)
in the equations of motion of the free particle. Written out in terms of the potential V
and vector potential A, the free Hamiltonian is then replaced by
1
H= (p qA)2 + qV (3.39)
2m
It can be shown (see e.g. Jackson §12.1, page 575) that this indeeds leads to the Lorentz
force law, Eq. (3.1).
Performing the operator substitution, the Schrödinger equation for the Hamiltonian
above becomes
✓ ◆
1 2 @
( ir qA) + qV (x, t) = i (x, t) (3.40)
2m @t
Comparing this to the Schrödinger equation for the free particle, we note that we have
essentially made the substitution
r ! r + iqA
(3.41)
@/@t ! @/@t + iqV
In four-vector notation this can be written as
@ µ ! Dµ ⌘ @ µ + iqAµ (3.42)
The ‘derivative’ D is called the covariant derivative.

We may now wonder what the e↵ect of the gauge transformation in Eq. (3.15) is on
the wave function . We have just established that the classical E and B fields do not
depend on gauge transformations. However, that does not mean that the wave function
is invariant as well. In fact, one can show (see e.g. Aichison and Hey, section 2.4) that
the combined transformation, required to make the Schrödinger equation invariant, is
given by
µ
Aµ ! A0 = Aµ + @ µ
(3.43)
! 0 = exp ( i q/~)
The gauge transformation leads to a change of the phase of the wave function. If
is not constant, then the change in phase is di↵erent at di↵erent points in space-time.
That is why we also call the gauge transformation a local phase transformation.
This result is at the heart of the application of gauge symmetries in quantum field
theory. Because, as we will see in more detail in Lecture 8, one can turn this argument
around: Since the phase of the wave function is not an observable, the equations that
describe the dynamics (a Schrödinger equation, or a Lagrangian) must be invariant
to such arbitrary phase transformations. If we impose this requirement, then we are
forced to introduce an Aµ field in the Hamiltonian via the substitution above and with
transformation properties defined above. In other words, the requirement of local phase
invariance imposes the form of the interaction!
3.5 The Aharanov-Bohm E↵ect

In the classical equations of motion only the E and B fields appear. Therefore, you
may wonder, is the Aµ field ‘real’, or merely a mathematical construct that simplifies
expressions? Or phrased di↵erently, does it contribute anything to our description
of moving charges that the E and B fields do not? The answer to that question is
beautifully illustrated by what is called the Aharanov-Bohm e↵ect.
In quantum mechanics we do not have forces: it is the amplitude of a wave function
that tells us where we are likely to find the particle in space and time. In Feynman’s
3.5. THE AHARANOV-BOHM EFFECT 55
path integral picture, the quantum mechanical particle follows all possible trajectories
to get from point x1 to point x2 , accumulating a phase eiS/~, where S is the action
along the path. Di↵erent paths have di↵erent phases. It is only around the classical
trajectory (obtained by requiring the action to be minimal) that these phases interfere
constructively. The size of deviations along the classical trajectory is determined by ~.
As we have seen above the vector potential appears in the Schrödinger equation and
a↵ects the wave function. In the presence of a magnetic field, the phase of the wave
function is changed along a trajectory according to
Z
q x2
↵(A) = A(r, t) · dr (3.44)
~ x1
where the integral runs along the trajectory. (We do not prove this here. See also
Feynman Vol 2, section 15-5.) Although we do not need it here, for completeness we
also mention that the change in phase due an electric field is given by the integral of
the potential over the time:
Z
q t2
↵(V ) = V (r, t)dt (3.45)
~ t1
This last equation you could easily derive from the SE for a constant electric field.
You will realize that when combined these two equations lead to a Lorentz covariant
formulation if the integral is performed over space and time.
Let us now consider Feynman’s famous two-slit experiment demonstrating the interfer-
ence between two electron trajectories. In the absence of external fields, the intensity at
a detection plate positioned behind the two slits shows an interference pattern. This is
most easily understood by considering the two ‘classical’ trajectories, depicted by 1 and
2 in Fig. 3.1. The relative length of these trajectories di↵ers as a function of the posi-
tion along the detection screen. The resulting phase di↵erence leads to the interference
pattern. For a great description see chapter 1 of the “Feynman Lectures on Physics”
volume 3 (“2-slit experiment”) and pages 15-8 to 15-14 in volume 2 (“Bohm-Aharanov”).
Now consider the presence of a magnetic field in the form of vector field A. (We choose
the electric field zero, so A0 = 0.) Due to the A field, the phases of the two contributions
to the wave functions change,
= 1 ei↵1 (r,t) + 2 ei↵2 (r,t) = 1 ei(↵1 ↵2 )
+ 2 ei↵2 . (3.46)
The extra contribution to the relative phase is given by
✓Z Z ◆ I
q 0 0 q
↵1 ↵2 = dr1 A1 dr2 A2 = dr 0 · A(r 0 , t)
~ ~
Z r1 r2
Z
q q q
= r ⇥ A(r 0 , t) · dS = B · dS = (3.47)
~ S ~ S ~
where we have used Stokes’ theorem to relate the integral around a closed loop to the
magnetic flux through the surface. The magnetic field shifts the interference pattern
detector
slits
Intensity
ψ2
source ψ
1 coil
Figure 3.1: The schematical setup of an experiment that investigates the e↵ect of the presence
of an A field on the phase factor of the electron wave functions.
on the screen. In exercise 3.5 you will show that for a homogenous magnetic field this
leads to the same deflection as the classical force law.
Let us now consider the case that a very long and thin solenoid is positioned in the setup
of the two-slit experiment. Inside the solenoid the B-field is homogeneous and outside
it is zero (or sufficiently small). However, the A field is not zero outside the coil, as
illustrated in Fig. 3.2. The classical trajectories do not pass through the B field, but
they do pass through the A field, leading to a shift in the relative phase. Experimentally
it has been verified (in a technically difficult experiment) that the interference pattern
indeed shifts.
Figure 3.2: Magnetic field and vector potential of a long solenoid.
We introduced the vector potential as a mathematical tool to write Maxwells equations

in a Lorentz covariant form. However, the vector field Aµ is not just an alternative
3.5. THE AHARANOV-BOHM EFFECT 57
formulation, but the only correct way to implement the Maxwell equation in quantum
mechanics. The gauge freedom may seem an undesirable feature now, but will turn out
to be a fundamental concept in our description of interactions.
Exercises
Exercise 3.1 (Maxwell equations)
Using the vector identity
r ⇥ (r ⇥ A) = r2 A + r (r · A) (3.48)
(which one can prove using "ijk "klm = il jm im jl ) show that (with c = 1 and ✏0 = 1)
Maxwell’s equations can be written as:
@µ @ µ A⌫ @ ⌫ @µ Aµ = j ⌫ (3.49)
Hint: Derive the expressions for ⇢ and j explicitly.
Exercise 3.2 (Gauge transformation)

Verify that the transformation in Eq. (3.14) does not change the E and B fields.
Exercise 3.3 (Lorentz gauge)

In this exercise you will show that it is always possible to choose a gauge such that the
field Aµ satisfies the Lorentz condition, Eq. (3.16).
(a) Suppose that for a given Aµ field one has @µ Aµ = g(x), with g(x) = eikx . Find the
function (x) such that Aµ + @ µ satisfies the Lorentz condition.
(b) Now consider an arbitrary function g(x). How does your result in (a) tell you
what (x) is?
Exercise 3.4 (Helicity)

Show that the circular polarization vectors ✏+ and ✏ transform under a rotation of
angle ✓ around the z-axis as
✏± ! ✏0± = e⌥i✓ ✏± (3.50)
Hint: First consider how the rotations transforms ✏1,2 .
Exercise 3.5 (Deflection in magnetic field. Feynman, Vol II, sec. 15-5.)
We have stated above that with the minimal substitution recipe the Schrödinger equation
leads to the Lorentz force law. We have also stated (not proven) how the change of the
phase of a wavefunction due to a vector field A can be obtained by integrating the vector
Figure 3.3: From “The Feynman Lectures in Physics”, volume II.
field along the trajectory. Let’s take these things for given and see if we can reproduce
the deflection of a particle in a magnetic field. Feynman beautifully illustrates that by
looking at the famous two-split experiment.
Consider the setup in Fig. 3.3. Particles with charge q, mass m and momentum p travel
from a source, via two slits, to a photographic plate. The interference of the two paths
leads to a di↵raction pattern. The distance between the slits and the plate is L. Directly
behind the slits is a thin strip of magnetic field. The thickness of the strip is w and
w ⌧ L. The B field is homogenous, coming out of the plane of the figure. We label the
coordinate along the photographic plate by x.
(a) For very small deflections, compute the deflection of the particles of the particles
using the Lorentz force law. Translate this into the displacement x at the pho-
tographic plate. Hint: Assume that the plate is thin enough that direction of the
force is along the x-axis. The force lasts for a time w/v.
(b) Consider two classical (shortest distance) trajectories through the two slits (indi-
cated by 1 and 2 ). For small deflections, compute the phase shift between the
two trajectories as a function of x, in the absence of a magnetic field. Compute
the distance between two maxima in the di↵raction pattern. Hint: The reduced
wavelength of the particles is /2⇡ = ~/p.
(c) Assuming again small deflections use equation (3.47) to compute the increase in
phase shift between the two trajectories as a result of the B field. Translate the
phase shift in a shift x of the di↵raction pattern.
Lecture 4
Electromagnetic Scattering of
Spinless Particles
In this lecture we discuss electromagnetic scattering of spin-0 particles. First we compute

the scattering of a charged particle on a static point charge. We show that in the non-
relativistic limit the result is in agreement with the well known formula for Rutherford
scattering. Subsequently, we derive the cross section for two particles that scatter in
each others field. We end the lecture with a prescription for treating antiparticles.
4.1 Electromagnetic current
As we discussed in section 3.4 the laws of electrodynamics can be introduced in the

equations of motions of free particles by the method of minimal substitution,
pµ ! pµ qAµ , (4.1)
which in terms of operators in coordinate space takes the form
@ µ ! @ µ + iqAµ . (4.2)
Now consider a spinless particle with mass m and charge e scattering in a vector field
Aµ , as in figure 2.1. (It is conventional to consider a charge e as for a hypothetical
spin-0 electron.) The wave equation for the free particle is the Klein-Gordon equation,
@µ @ µ + m2 = 0 (4.3)
Substituting @ µ ! @ µ + iqAµ with q = e, we obtain
(@µ ieAµ ) (@ µ ieAµ ) + m2 = 0. (4.4)
59
60LECTURE 4. ELECTROMAGNETIC SCATTERING OF SPINLESS PARTICLES
Be aware that the operators @ µ act on all field on their right, so both on and Aµ . This
equation can be rewritten as
@µ @ µ + m2 + V (x) =0 (4.5)
with a perturbation potential V (x) given by
V (x) ⌘ ie (@µ Aµ + Aµ @ µ ) e2 A2 . (4.6)
The sign of V is chosen such that compared to the kinetic energy it gets the same sign as
in the Schrödinger equation, Eq. (3.39). Since e2 is small (↵ = e2 /4⇡ = 1/137) and we
only consider the Born level cross-section, we neglect the second order term, e2 A2 ⇡ 0.
From Lecture 2 we take the general expression for the transition amplitude in the Born
approximation and insert the expression for V (x),
Z
Tf i ⌘ i d4 x ⇤f (x) V (x) i (x)
Z
= i d4 x ⇤f (x) ( ie) (Aµ @ µ + @µ Aµ ) i (x). (4.7)
The second @µ operator on the right hand side acts on both Aµ and . However, we can
use integration by parts to write
Z Z
⇤
⇥ ⇤ µ ⇤1
4 µ
d x f @µ (A i ) = f A i 1
@µ ⇤f Aµ i d4 x (4.8)
Requiring the field to be zero at t = ±1, the first term on the left vanishes, such that
the transition amplitude becomes
Z
⇥ ⇤
Tf i = i ( ie) ⇤f (x) (@µ i (x)) @µ ⇤f (x) i (x) Aµ d4 x . (4.9)
In this expression the derivatives no longer act on the field Aµ . Remember the definition
of the charge current density for the Klein-Gordon field of the electron, Eq. (1.31),
⇤ ⇤
jµ = ( ie) [ (@µ ) (@µ ) ] .
In complete analogy we define the “electromagnetic transition current” to go from initial

state i to final state f as
⇥ ⇤
jµf i ⌘ ( ie) ⇤f (@µ i ) @µ ⇤f i . (4.10)
You may verify that if f and i are both solutions to a Klein-Gordon equation with
mass m, then also this current satisfies the continuity equation @ µ jµf i = 0.
The transition amplitude can now be written as
Z
Tf i = i jµf i Aµ d4 x (4.11)
4.2. COULOMB SCATTERING 61
This is the expression for the transition amplitude for going from free particle solution i
to free particle solution f in the presence of a perturbation caused by an electromagnetic
field. Restricting ourselves to plane wave solutions of the unperturbed Klein-Gordon
equation,
ipi x ⇤ ⇤ ipf x
i = Ni e and f = Nf e , (4.12)
we find for the transition current of spinless particles
jµf i = ( e) Ni Nf⇤ piµ + pfµ ei(pf pi )x

. (4.13)
Inserting this in the transition amplitudes gives

Z
Tf i = i ( e) Ni Nf⇤ pµi + pµf Aµ ei(pf pi )x
d4 x (4.14)
4.2 Coulomb scattering

Consider the case that the external field is a static field of a point charge Ze located at
the origin,
Ze
Aµ = (V, A) = (V, 0) with V (x) = . (4.15)
4⇡|x|
With a vector field of this form, we have pµk Aµ = Ek V (x). Consequently, we find for
the transition amplitude
Z
Ze2 4
Tf i = i Ni Nf⇤ (Ei + Ef ) ei(pf pi )x dx (4.16)
4⇡|x|
Since V (x) is time independent, we split the integral over space and time. As we have
seen before, the integral over time turns into a function, expressing energy conserva-
tion, Z
ei(Ef Ei )t dt = 2⇡ (E fE) i (4.17)
For the integral over x we use an important Fourier transform

Z
1 1
2
= d3 x eiqx (4.18)
|q| 4⇡|x|
Using this with q ⌘ (pf pi ) we obtain
Ze2
Tf i = i Ni Nf⇤ (Ei + Ef ) 2⇡ (Ef Ei ) . (4.19)
|pf pi |2
As we have seen before for a time-independent potential, we consider a time-averaged
transition rate,
|Tf i |2
Wf i = lim (4.20)
T !1 T
where the time-averaging e↵ectively takes care of one of the functions when taking the
square of the amplitude. The result is
!2
2
Ze (Ei + Ef )
Wf i = |Ni Nf |2 2⇡ (Ef Ei ) (4.21)
|pf pi |2
Working with normalization of the plane waves over a box with volume V , we have
|Ni Nf |2 = V 2 . The flux factor for a single particle is given by
2Ei 2|pi |
flux = |vi| = , (4.22)
V V
while the phase space factor is
V d3 pf
dLips = . (4.23)
(2⇡)3 2Ef
Inserting these expressions in our master formula for the cross-section, Eq. (2.8), we find
!2
2⇡ (Ef Ei ) Ze2 (Ei + Ef ) d3 p f
d = (4.24)
2|pi | |pf pi |2 (2⇡)3 2Ef
where we have canceled all factors V .

We can still simplify this by integrating over the outgoing momentum. Choose the z-axis
along pi and switch to polar coordinates for pj such that
d3 pf = p2f dpf dcos ✓ d = p2f dpf d⌦ (4.25)
where pf,i now refers to the size of the three-momentum. Since Ef2 = m2 + p2f , we have
pf dpf = Ef dEf , and therefore,
Ef
(Ef Ei ) dpf = (Ef Ei ) dEf (4.26)
pf
Energy conservation will imply that pf = pi , such that
|pf pi |2 = 2 p2i (1 cos ✓) = 4 p2i sin2 ✓/2 . (4.27)
The di↵erential cross-section then becomes

d Z 2 Ei2 e4 Z 2 |Ei |2 ↵2
= = (4.28)
d⌦ 64⇡ 2 |pi |4 sin4 ✓/2 4 |pi |4 sin4 ✓/2
where we defined ↵ ⌘ e2 /4⇡.

In the non-relativistic limit we have E ! m and p2 = 2mEkin , giving
d Z 2 ↵2
= , (4.29)
d⌦ 2
16Ekin sin4 ✓/2
4.3. SPINLESS ⇡ K SCATTERING 63
which is the well-known Rutherford scattering formula.

Above we have not explicitly shown the solution of the wave function itself. Without
deriving it here, we just state that for a spherically symmetric potential, V (r) = V (r),
and an incident wave with momentum along the z-axis, the wave function for large r
takes the form
r!1 eikr
(r) ! eikz + f (✓, ) (4.30)
r
where the first term is the incident ‘unscattered’ wave and the second term the ‘scattered’
wave. Expressed in terms of the f (✓, ), the di↵erential cross-section can be written as
d
= |f (✓, )|2 (4.31)
d⌦
The interference between the scattered and the unscattered wave leads to a ‘shadow’
behind the scattering potential. The flux that is missing in the shadow is exactly the
total scattered flux. This is expressed in the optical theorem, which states that the total
cross-section is proportional to value of f in the forward direction,
k
Imf (0) = . (4.32)
4⇡
See also appendix H of Aichison and Hey, and references therein.
4.3 Spinless ⇡ K Scattering

We now proceed with the electromagnetic scattering of two particles, A + B ! C + D.
As an example we consider the scattering of a ⇡ particle and a K particle. We ignore
the fact that pions and kaons also are subject to the strong interaction, which is fine
as long as the recoil momentum is small compared to the binding energy. We could
equally well consider a process like e µ scattering, provided that we ignore the lepton
spin. For the computation presented here, the essential restrictions are that the incident
particles carry no spin and that they are of di↵erent type.
We have seen above how a particle scatters in an external field. In this case the field is
not external as the particles scatter in each others field. How do we deal with this?
First consider a pion scattering in the vector field Aµ generated by the current of the
kaon. The transition current of the kaon is given by (see Eq. (4.13))
µ
jBD = eNB ND⇤ (pµB + pµD ) ei(pD pB )x
(4.33)
We now assume that the field generated by the kaon can be computed by inserting this
current in the Maxwell equations for the vector potential, i.e.
µ
@⌫ @ ⌫ Aµ = jBD (4.34)
B: K D: K
Aµ
A: ⇡ C: ⇡
Figure 4.1: Leading order diagram for electromagnetic scattering of a charged kaon and a
charged pion.
where we have adopted the Lorentz gauge. (A proof that this indeed works requires the
full theory.) Since @⌫ @ ⌫ eiqx = q 2 eiqx , we can easily verify that the solution is given by
1 µ
Aµ = j , (4.35)
q 2 BD
where we defined q = pD pB . The latter corresponds to the four-momentum transfered
by the photon from the kaon to the pion. The transition probability becomes
Z Z Z
µ 4 µ 1 BD 4 µ gµ⌫ ⌫
Tf i = i jAC Aµ d x = i jAC 2 jµ d x = i jAC j d4 x . (4.36)
q q 2 BD
Four-momentum conservation (which appears as a result of the integral when we sub-
stitute plane waves in the currents) makes that the momentum transfer is also equal to
q = (pC pA ). Therefore, Tf i is indeed symmetric in the two currents. It does not
matter whether we scatter the pion in the field of the kaon or the kaon in the field of
the pion.
The expression has a pole for q 2 = 0, the mass of a ‘real’ photon: zero momentum trans-
fer (non-scattered waves) has ’infinite’ probability. The only contribution to scattering
under non-zero angles comes from photons that are “o↵ the mass-shell”. We call these
virtual photons.
Inserting the plane wave solutions
Z
1
Tf i = ie 2
(NA NC⇤ ) (pµA + pµC ) ei(pC pA )x
· ·(NB ND⇤ ) (pµB + pµD ) ei(pD pB )x 4
d x (4.37)
q2
and performing the integral over x we obtain
1
Tf i = ie2 (NA NC⇤ ) (pµA + pµC ) (NB ND⇤ ) pB D
µ + pµ (2⇡)4 4
(pA + pB pC pD )
q2
(4.38)
where the -function that takes care of four-momentum conservation appears. Usually
this is written in terms of the invariant amplitude M (sometimes called ‘matrix element’)
as
Tf i = i NA NB NC⇤ ND⇤ (2⇡)4 4 (pA + pB pC pD ) · M (4.39)
4.3. SPINLESS ⇡ K SCATTERING 65
with the invariant amplitude given by

igµ⌫
iM = ie (pA + pC )µ · 2
· ie (pB + pD )⌫ . (4.40)
| {z } q | {z }
vertex factor | {z } vertex factor
propagator
The signs and factors i are assigned such that the expressions for vertex factors and
propagator are also appropriate for higher orders. These are, in fact, our first set of
Feynman rules!
B D
ie(pB + pD )µ
igµ⌫
q2
ie(pA + pC )µ
A C
Figure 4.2: Feynman rules for the t-Channel contribution to electromagnetic scattering of
spinless particles.
Feynman rules allow us to specify the amplitude corresponding to a particular Feynman

diagram without going through the explicit computation of the amplitude in quantum
field theory. In figure 4.2 we illustrate the rules for the diagram that we are considering.
The invariant amplitude contains:
a vertex factor: for each vertex we introduce the factor iepµ , where
⇤ e is the intrinsic coupling strength of the particle to the e.m. field
⇤ pµ is the sum of the 4-momenta before and after the scattering (remember
the particle/anti-particle convention).
igµ⌫
a propagator: for each internal line (photon) we introduce a factor q2
, where q is
the 4-momentum of the virtual photon.
The ingoing and outgoing four-momenta, and the four-momenta of internal particles
are free, but we are also required to add a function at each vertex to ensure energy-
momentum conservation. In the end, all internal momentum vectors are integrated
over, and what remains is a single function over ingoing and outgoing momenta. By
convention, the latter does not belong to M. The full set of rules also specify how this
works for higher order diagrams.
The expression of the amplitude in terms of propagators and vertex factors is part
of what we refer to as the Feynman calculus. To prove that this works requires field
theory. However, the attractiveness of this approach is that once you have established
the recipe, you can derive the Feynman rules (the expressions for the propagators and
vertices) directly from the Lagrangian that specifies the dynamics of your favourite
theory: if you insert a new type of particle or interaction in your Lagrangian, you do
not really need field theory anymore to compute cross-sections. We discuss the role of the
Lagrangian in more detail in Lecture 8. In appendix ?? we sketch how the propagators
and vertex factors are obtained from the Lagrangian.
We can now insert the invariant amplitude into the expression for the A + B ! C + D
cross-section that we derived in the previous lecture,
(2⇡)4 4 (pA + pB pC pD ) d3 pC d3 p D
d = q |M|2 3 3 . (4.41)
4 (pA · pB )2 2 2
mA mB (2⇡) 2EC (2⇡) 2E D
In the centre-of-momentum frame (pA = pB ) this expression became

d 1 1 pf
= 2
|M|2 . (4.42)
d⌦ 64⇡ s pi
where s = (pA + pB )2 . As we consider elastic scattering, |pf |/|pi | = 1.
Finally, consider the limit of massless particles. Define p ⌘ pA and p0 ⌘ pC . In the
centre-of-momentum frame the four-vectors are given by
pµA = (|~p|, p)
pµB = (|~p|, p)
pµC = (|p0|, p0)
pµD = (|p0|, p0)
Define p ⌘ |p| which, by four-vector conservation is also equal to |p0|. Define ✓ as the
angle between pA and pC , which means that cos ✓ = pA · pC /|pA ||pC | = pA · pC /p2 . We
then have
(pA + pC )µ gµ⌫ (pB + pD )⌫ = (pA )µ (pB )µ + (pA )µ (pD )µ + (pC )µ (pB )µ + (pC )µ (pD )µ
= 2p2 + p2 (1 + cos ✓) + p2 (1 + cos ✓) + 2p2
= p2 (6 + 2 cos ✓)
Likewise, we get for q 2 ,
q 2 = (pA pC )2
= p2A + p2C 2(pA )µ (pC )µ
= 2p2 (1 cos ✓)
Consequently, we obtain for the invariant amplitude defined above
2
✓ ◆
2 p (6 + 2 cos ✓) 2 3 + cos ✓
M = e =e . (4.43)
2p2 (1 cos ✓) 1 cos ✓
Inserting this in Eq. (4.42) gives (with ↵ = e2 /4⇡),
✓ ◆2 ✓ ◆2
d 1 1 2 2 3 + cos ✓ ↵2 3 + cos ✓
= e = (4.44)
d⌦ 64⇡ 2 s 1 cos ✓ 4s 1 cos ✓
This is the leading order QED cross section for the scattering of massless spin-0 particles
in the centre-of-momentum frame. In the exercises you will derive the formula for
particles with non-zero mass.
4.4. FORM FACTORS 67
4.4 Form factors
In the previous sections we studied ‘point charges’, objects with their charge located in
an infinitely small region. If the charge distribution has a finite size, the di↵erential cross-
section is di↵erent from that of a point source. Consequently, the measured di↵erential
cross-section can tell us important information over the substructure of particles. For
example, most information about the structure of the proton has been obtained in
electron-proton scattering experiments, most notably at the Hera collider in Hamburg.
Consider a static source with a charge distribution Ze⇢(x), normalized so that

Z
⇢(x) d3 x = 1 (4.45)
By following the same procedure as above for the static source, one can show that the
di↵erential cross-section can be written as
d d
= |F (pi pf )|2 (4.46)
d⌦ d⌦ point
where F (q) is called the form factor. It is given by the Fourier transform of the charge
distribution
Z
F (q) = ⇢(x)eiq·xd3 x (4.47)
In real electron-proton scattering we also need to account for the spin and the magnetic
moment of the proton. The form factor will then become more complicated. You will
learn more about this in the Particle Physics II course.
4.5 Particles and Anti-Particles
We have seen that the negative energy state of a particle can be interpreted as the
positive energy state of its anti-particle. How does this e↵ect energy conservation that
we encounter in the -functions? We have seen that the invariant amplitude has the
form of:
Z
⇤
M/ f (x) V (x) i (x) dx
Let us examine four cases.
• Scattering of an electron and a photon:

Z
ipf x ⇤ ikx ipi x
M / e e e dx
pi Z
= e i(pi +k pf )x
dx
pf
= (2⇡)4 (Ei + ! Ef ) 3
(pi + k pf )
k
) Energy and momentum conservation are
enforced by the -function.
• Scattering of a positron and a photon:
Replace the anti-particles always by particles by

reversing (E, p ! E, p) such that now:
pi incoming state = pf , outgoing state = pi :
Z
⇤
pf M / e i( pi )x e ikx e i( pf )x dx
Z
k = e i(pi pf +k)x dx
= (2⇡)4 (Ei + ! Ef ) 3
(pi + k pf )
• Electron positron pair production:
Z
-p+ ip x ⇤ i( p+ +k)x
M / e e dx
Z
k i(k p+ p )x
= e dx
p
= (2⇡)4 (k p p+ )
• Electron positron annihilation:
Z
p i(k p+ )x ⇤ i(p )x
M / e e dx
Z
k i(p +p+ k)x
= e
-p+
= (2⇡)4 (p + p+ k)
4.5. PARTICLES AND ANTI-PARTICLES 69
Exercises
Exercise 4.1 (Scattering o↵ a static source as a limit)

Above we derived the expression for A + B ! C + D scattering in the special case that
A, B, C and D are massless. We will now show that in the limit in which mB mA
and mB p the cross-section for the process A + B ! A + B reduces to the formula
that we derived for scattering of a static source.
Call mA the mass of A and mB the mass of B. We consider scattering in the CMS.
Choose a coordinate system such that the initial momenta are along the z-axis, with A
going in the positive z direction and B in the negative z direction. Label the outgoing
particles with a prime. Call ✓ the polar angle of p0A . Momentum conservation makes
that all momenta have the same size, which we label by p.
0 0 µ
(a) Express (ppA + pA )µ (pB + pB ) inpmA , mB , p and cos ✓. (You may of course also
use EA = mA + p and EB = m2B + p2 .)
2 2
Hint: one method is to first just write down all fourvectors in these symbols. Take
the x, y-coordinates together, because we do not care about the azimuthal angle
( ).Now you can either first add and them take the inner product, or vice versa.
(b) Do the same for q 2 = (pA p0A )2 and for s = (pA + pB )2 .
(c) Write down the di↵erential cross-section d /d⌦ using Eq. (4.42). Note that this
result is more general than our ‘massless particle’ result in Eq. (4.44).
(d) Take the limit mB p and mB mA . Compare to the result for scattering of a
static source, Eq. (4.28).
Exercise 4.2 (Equal particles)

When computing the scattering of two particles A and B above, we explicitly restricted
the computation to the case where the particles were di↵erent.
(a) What changes if B is an anti-A?
(b) What changes if B is an A?
Hint: Look at figure 4.2. Can you imagine additional ‘leading order’ diagrams? If so,
draw them!
Exercise 4.3 (Decay rate of ⇡ 0 ! (see also Griffiths, ex.6.6))

(a) Write down the expression for the total decay width for the decay: A ! C + D
(b) Assume that particle A is a ⇡ 0 particle with a mass of 140 MeV and that particles
C and D are photons. Draw the Feynman diagram for this decay
(i) assuming the pion is a uū state.

(ii) assuming the pion is a dd¯ state.
(c) We do not know the matrix element M. However, we know that it is proportional
p 2
to e2 ⇠ ↵ . Why?
(d) We also know something about the dimension of M: for a two-body decay the
dimension is mc (or p or E/c). The constant with dimension [mc] in the amplitude
is called the decay constant and denoted by f⇡ , such that
M / f⇡ ↵.
where ↵ is the dimensionless coupling constant, and additional factors are dimen-
sionless as well. If you do not know anything else about the ⇡ 0 decay constant but
its dimension, what value would you use?
(e) Assuming that the ⇡ 0 is a uū + dd¯ state
(i) give the expression for the decay width (by adding up the amplitudes);
(ii) calculate the decay width expressed in GeV;
(iii) convert the rate into a mean lifetime in seconds.
(iv) How does the value compare to the Particle Data Group (PDG) value?
Remark: Do not be disappointed if your prediction is completely wrong! It turns
out that the ⇡ 0 lifetime is quite hard to compute.
Lecture 5
The Dirac Equation
In Lecture 1 we have seen how the Klein-Gordon equation leads to solutions with neg-
ative energy and negative ‘probability density’. This is a consequence of the fact that
the wave equation is quadratic in @/@t. In 1928 in an attempt to avoid this problem
Dirac developed a relativistic wave equation that is linear in @/@t. Lorentz invariance
requires that such a wave equation is also linear in @/@x. What Dirac found was an
equation that describes particles with spin- 12 , just what was needed for electrons. We
now think that all fundamental fermions are described by this wave equation. Dirac
also predicted the existence of anti-particles, an idea that was not taken seriously until
1932, when Anderson discovered the positron.
5.1 Spin, spinors and the gyromagnetic ratio

Before we proceed with the derivation of the Dirac equation, we briefly discuss spin
and the Pauli-Schrödinger equation, since this allows us to introduce some important
concepts.
A lump of charge rotating around an axis through its centre-of-gravity is a magnetic
dipole. It has a magnetic moment
µ= S (5.1)
where S is the classical spin, the integral of r ⇥ p over the mass density distribution.
The factor is called the gyromagnetic ratio. If the charge distribution follows the mass
distribution, this ratio is given by
q
classic = . (5.2)
2m
When placed in a magnetic field B, the particle experiences a torque, µ ⇥ B. The
potential energy associated with the torque is
H = µ·B = S·B (5.3)
71
72 LECTURE 5. THE DIRAC EQUATION
So, far this is just classical electrodynamics. The classical spin S is nothing but the total
angular momentum of all bits and pieces that the particle is made up from. However,
as you remember from your quantum mechanics course, elementary particles also carry
intrinsic spin. Though we sometimes imagine it as a result of a charged particle spinning
around an axis, this interpretation actually falls short. In particular, the prediction of
the gyromagnetic ratio that would come out of this picture is wrong.
On the other hand, elementary particles do feel a torque in a magnetic field, as demon-
strated in the Stern-Gerlach experiment in 1922. So, in 1927 Pauli tried to address the
question of how to describe their magnetic moment in quantum mechanics.
Pauli considered a spin- 12 system. As you know, such a system has two values for the
eigenvalue of spin, namely ± 12 ~. An arbitrary spin wave function is a superposition of
the two eigenstates. Pauli represented it as a two-component vector, called a spinor,
✓ ◆
a
= = a (1) + b (2) (5.4)
b
where the basis vectors
✓ ◆ ✓ ◆
(1) 1 (2) 0
⌘ and ⌘ (5.5)
0 1
represent the spin-up and spin-down state respectively. The hermitian operator that
measures spin, the spin operator S, satisfies the same algebra as for orbital angular
momentum in quantum mechanics, namely [Si , Sj ] = i~✏ijk Sk . In the basis above, S is
represented by 2x2 matrices. Choosing the z-axis as the quantization axis, S is given
by
~
S = (5.6)
2
where the Pauli spin matrices are
✓ ◆ ✓ ◆ ✓ ◆
0 1 0 i 1 0
1 = 2 = 3 = . (5.7)
1 0 i 0 0 1
The i all have zero trace, are hermitian, and satisfy
i j = ij + i✏ijk k (5.8)
which implies as well that they anti-commute ({ i , j} =2 ij ).
In a modern interpretation, Pauli then proposed to represent the momentum operator

in spinor space by
p̂ ! · p̂ (5.9)
Take a careful look at what is written on the right-hand-side: it is an inner product of a
vector of 2x2 matrices with the momentum operator. The result is again a 2x2 matrix,
which is more apparent when written out,
✓ ◆
pz px ipy
·p= . (5.10)
px + ipy pz
5.2. THE DIRAC EQUATION 73
You will show in an exercise that ( · p)2 = |p|2 112 , where 112 is the 2x2 identity matrix.
Therefore, the Schrödinger equation for free spinors is just the ordinary Schrödinger
equation, ✓ ◆ ✓ ◆ ✓ ◆
d a 1 2 a p2 a
i~ = ( · p) = (5.11)
dt b 2m b 2m b
and the two spin states are degenerate in energy.
This is no longer the case if we introduce the vector field. Using again minimal substi-
tution, the Hamiltonian (a matrix in spinor space) for a particle in a vector field (V, A)
becomes
1
H= [ · (p qA)]2 + qV (5.12)
2m
It is a not entirely trivial exercise to show1 that this equation can be rewritten as
1 ~q
H= (p qA)2 + qV ·B (5.13)
2m 2m
The Schrödinger equation with this Hamiltonian is called the Pauli-Schrödinger equa-
tion, or simply Pauli equation. Comparing this Hamiltonian to the Hamiltonian of the
classical spin, we find that the gyromagnetic ratio for a spin- 12 particle is
q
spin-1/2 = (5.14)
m
exactly a factor 2 larger than for the classical picture of a spinning charge distribution.
The ratio of the magnetic moment relative to that of the classical case is called the
g-factor. For spin- 12 particles the Schrödinger-Pauli equation predicts g = 2. In QED
the magnetic moment is modified by higher order corrections. The predictions and
measurements of the magnetic moment of the electron and muon are so precise that
they make QED the most precisely tested theory in physics.
Pauli introduced the spin matrices in the Hamiltonian on purely phenomenological
grounds. As we shall see in the rest of this Lecture and the next, Dirac found a the-
oretical motivation: His construction of a wave equation that is linear in space and
time derivatives, leads (in its simplest form) to the description of spin- 12 particles and
anti-particles. As you will prove in exercise 5.8, the Pauli-Schrödinger equation can be
obtained as the low relativistic limit of the equation of motion of Dirac particles in a
vector field Aµ .
5.2 The Dirac equation

In order to appreciate what Dirac discovered we follow (a modern interpretation of) his
approach leading to a linear wave equation. (For a di↵erent approach, which may be
1
Take into account that p = i~r and A do not commute, then use the chain rule to see that in
coordinate space only a term B = r ⇥ A survives.
closer to what Dirac actually did, see Griffiths, §7.1.) Consider the usual form of the
Schrödinger equation,
@
i = H . (5.15)
@t
The classical Hamiltonian is quadratic in the momentum. Dirac searched for a Hamil-
tonian that is linear in the momentum. We start from the following ansatz2 :
H = (↵ · p + m) (5.16)
with coefficients ↵1 , ↵2 , ↵3 , . In order to satisfy the relativistic relation between energy

and momentum, we must have for any eigenvector (p) of H that
H 2 = p2 + m2 (5.17)
where p2 + m2 is the eigenvalue. What should H look like such that these eigenvectors
exist? Squaring Dirac’s ansatz for the Hamiltonian gives
! !
X X
H2 = ↵i pi + m ↵j pj + m
i j
!
X X X
2
= ↵i ↵j pi pj + ↵i pi m + ↵i p i m + m2 (5.18)
i,j i i
!
X X X
= ↵i2 p2i + (↵i ↵j + ↵j ↵i )pi pj + (↵i + ↵i )pi m + 2
m2
i i>j i
where we on purpose did not impose that the coefficients (↵i , ) commute. In fact,
comparing to equation (5.17) we find that the coefficients must satisfy the following
requirements:
• ↵12 = ↵22 = ↵32 = 2
=1
• ↵1 , ↵2 , ↵3 , anti-commute with each other.
With the following notation of the anti-commutator
{A, B} = AB + BA. (5.19)
we can also write these requirements as

2
{↵i , ↵j } = 2 ij {↵i , } = 0 =1. (5.20)
Clearly, the ↵i and cannot be ordinary numbers. At this point Dirac had a brilliant
idea, possibly motivated by Pauli’s picture of fermion wave functions as spinors: what
if the ↵i and are matrices that act on a wave function that is a column vector? As
2
Note that ↵ · p = ↵x px + ↵y py + ↵z pz .
5.3. COVARIANT FORM OF THE DIRAC EQUATION 75
we require the Hamiltonian to be hermitian (such that its eigenvalues are real), the
matrices ↵i and must be hermitian as well,
↵i† = ↵i and †
= . (5.21)
Furthermore, we can show using just the anti-commutation relations and normalization
above that they all have eigenvalues ±1 and zero trace. It then follows that they must
have even dimension.
It can be shown that the lowest dimensional matrices that have the desired behaviour
are 4x4 matrices. (See exercise 5.6 and also Aitchison (1972) §8.1). The choice of the
matrices ↵i and is not unique. Here we choose the Dirac-Pauli representation,
✓ ◆ ✓ ◆
0 11 0
↵= and = . (5.22)
0 0 11
Of course, we may expect that the final expressions for the amplitudes are independent
of the representation: all the physics is in the anti-commutation relations themselves.
Another frequently used choice is the Weyl representation,
✓ ◆ ✓ ◆
0 0 11
↵= and = . (5.23)
0 11 0
5.3 Covariant form of the Dirac equation

With Dirac’s Hamiltonian and the substitution p = i~r we arrive at the following
relativistic Schrödinger-like wave equation,
@
i = ( i↵ · r + m) . (5.24)
@t
Multiplying on the left by and using 2 = 1 we can write this equation as
✓ ◆
@
i +i ↵·r m =0. (5.25)
@t
We now define the four Dirac -matrices by

µ
⌘ ( , ↵) (5.26)
The wave equation then takes the simple form

µ
(i @µ m) = 0 (5.27)
This equation is called the Dirac equation. Note that is a four-element vector. We
call it a bi-spinor or Dirac spinor. We shall see later that the solutions of the Dirac
equation have four degrees of freedom, corresponding to spin-up and spin-down for a
particle and its anti-particle.
The Dirac equation is actually a set of 4 coupled di↵erential equations,
4
" 3 #
for each X X
: i ( µ )jk @µ m jk ( k ) = 0
j=1,2,3,4
k=1 µ=0
2 0 1 3
6 B C 0 1 7 0 1 1 0
6 B . . . . C 1 0 0 0 7 1 0
6 B C B 0 7
6 B . . . . C B 1 0 0 C
C · m7
B
B 2
C B 0 C
C=B C
or : 6i B C·@ @ 0 7
6 B . . . . C µ 0 1 0 A 7 @ 3
A @ 0 A
6 B C 7
4 @ . . . . A 0 0 0 1 5 4 0
| {z }
µ
or even more explicit, in the Dirac-Pauli representation,

0 1 0 1
✓ ◆ ✓ ◆ ✓ ◆ ✓ ◆ ✓ ◆ 1 0
11 0 i@ 0 i@ 0 i@ 0 i@ 11 0 B C B 0 C
+ 1
+ 2
+ 3
m B 2 C=B C
0 11 @t 1 0 @x 2 0 @y 3 0 @z 0 11 @ 3
A @ 0 A
4 0
Take note of the use of the Dirac (or spinor) indices (j, k = 1, 2, 3, 4) simultaneously
with the Lorentz indices (µ = 0, 1, 2, 3). As far as it concerns us, it is a coincidence that
both types of indices assume four di↵erent values.
To simplify notation even further we define the ‘slash’ operator of a four-vector aµ as
µ
6a = µa . (5.28)
The wave equation for spin- 12 particles can then be written very concisely as
(i 6 @ m) =0. (5.29)
Although we write µ like a contra-variant four-vector, it actually is not a four-vector.

It is a set of four constant matrices that are identical in each Lorentz frame. For a four-
vector aµ , 6 a is a (4 ⇥ 4) matrix, but it is not a Lorentz invariant and still depends on
the frame. The behaviour of Dirac spinors under Lorentz transformations is not entirely
trivial. (See also Griffiths §7.3, Halzen and Martin §5.6 or Thomson appendix B.)
5.4 Dirac algebra

From the definitions of ↵ and we can derive the following relation for the anti-
commutator of two -matrices
µ ⌫ µ ⌫ ⌫ µ
{ , }⌘ + = 2 g µ⌫ 114 (5.30)
5.5. ADJOINT SPINORS AND CURRENT DENSITY 77
where the identity matrix on the right-hand side is the 4 ⇥ 4 identity in bi-spinor space.
Text books usually leave such identity matrices away. However, it is important to realize
that the equation above is a matrix equation for every value of µ and ⌫. In particular,
g µ⌫ is not a matrix in spinor space. (In the equation, it is just a number!)
Using this result we find
0 2 1 2 2 2 3 2
= 114 = = = 114 . (5.31)
The Hermitian conjugates are

0† 0
= (5.32)
i† i † i† † i i
= ↵ =↵ =↵ = . (5.33)
Using the anti-commutation relation once more then gives
µ† 0 µ 0
= (5.34)
µ† 0 0
In words this means that we can undo a hermitian conjugate by moving a
“through it”, µ† 0 = 0 µ . Finally, we define
5 0 1 2 3
= i (5.35)
which has the characteristics

5† 5 5 2 5 µ
= = 114 , =0 (5.36)
5
In the Dirac-Pauli representation is
✓ ◆
5 0 112
= (5.37)
112 0
5.5 Adjoint spinors and current density

Similarly to the case of the Schrödinger and the Klein-Gordon equations we can define a
current density j µ that satisfies a continuity equation. First, we write the Dirac equation
as
@ X
k @
i 0 +i m =0 (5.38)
@t k=1,2,3
@xk
As we now work with matrices, we use hermitian conjugates rather than complex con-
jugates and find for the conjugate equation
@ † X @ †
0 k †
i i m =0 (5.39)
@t k=1,2,3
@xk
k
However, we now see a potential problem: the additional minus sign in dis-
turbs the Lorentz invariant form of the equation. We can restore Lorentz covariance by
multiplying the equation from the right by 0 . Therefore, we define the adjoint Dirac
spinor
† 0
= . (5.40)
Note that the adjoint spinor is a row-vector:
0 1
1
B C
Dirac spinor : B 2 C Adjoint Dirac spinor: 1, 2, 3,
@ 3
A 4
The adjoint Dirac spinor satisfies the equation
@ X @
0 k
i i m =0 (5.41)
@t k=1,2,3
@xk
which can be written in short-hand as

µ
i@µ +m = 0. (5.42)
Now we multiply the Dirac equation from the left by and we multiply the adjoint
Dirac equation from the right by :
µ
i@µ +m = 0
µ
(i@µ m ) = 0
+
µ µ
(@µ ) + @µ = 0
Consequently, we realize that if we define a current as
jµ = µ
(5.43)
then this current satisfies a continuity equation, @µ j µ = 0. The first component of this
current is simply
X4
0 0 †
j = = = | i |2 (5.44)
i=1
which is always positive. This property was the original motivation of Dirac’s work.
The form Eq. (5.43) suggests that the Dirac probability current density transforms as
a contravariant four-vector. In contrast to the Klein-Gordon case, this is not so easy to
show since the Dirac spinors transform non-trivially. We will leave the details to the
textbooks.
5.6. BILINEAR COVARIANTS 79
5.6 Bilinear covariants
The Dirac probability current in Eq. (5.43) is an example of a so-called bilinear covari-
ant: a quantity that is a product of components of and and obeys the standard
transformation properties of Lorentz scalars, vectors or tensors. The bilinear covariants
represent the most general form of currents consistent with Lorentz covariance.
Given that and each have four components, we have 16 independent combinations.
Requiring the currents to be covariant, then leads to the following types of currents:
# of components
scalar 1
µ
vector 4
µ⌫ (5.45)
tensor 6
5 µ
axial vector 4
5
pseudo scalar 1
where the (anti-symmetric) tensor is defined as
µ⌫ i µ ⌫ ⌫ µ
⌘ ( ) (5.46)
2
The names ‘axial’ and ‘pseudo’ refer to the behaviour of these objects under the parity
transformation, x ! x. The scalar is invariant under parity, while the pseudo scalar
changes sign. The space components of the vector change sign under parity, while
those of the axial vector do not. We shall discuss the bilinear covariants and their
transformation properties in more detail in Lecture 7.
5.7 Solutions to the Dirac Equation
5.7.1 Plane waves solutions with p = 0
We now consider explicit expressions for the solutions of the Dirac equation, Eq. 5.27.
In exercise 5.1 you will show that each of the components of the Dirac wave satisfies
the Klein-Gordon equation. Therefore, we try the construct the solutions as plane wave
solutions
(x) = u(p) e ipx (5.47)
where u(p) is a 4-component column-vector that does not depend on x. After substitu-
tion in the Dirac equation we find what is called the Dirac equation in the momentum
representation,
( µ pµ m) u(p) = 0 (5.48)
which, using the ‘slash notation’ can also be written as
(6 p m) u(p) = 0 . (5.49)
In momentum-space the coupled di↵erential equations reduce to a set of coupled linear

equations. In the Pauli-Dirac representation we have
✓ ◆ ✓ ◆ ✓ ◆ ✓ ◆
11 0 0 i 11 0 uA
E pi m =0 (5.50)
0 11 i 0 0 11 uB
where uA and uB are two-component spinors. We can rewrite this as a set of two
equations
( · p) uB = (E m) uA
(5.51)
( · p) uA = (E + m) uB ,
where ⌘ ( 1, 2, 3 ).
Now consider a particle with non-zero mass in its restframe, p = 0. In this case, the two
equations decouple,
E uA = m u A
(5.52)
E uB = m u B .
There are four independent solutions, which we write as

0 1 0 1 0 1 1 0
1 0 0 0
B 0 C B C B 0 C B 0 C
u(1) = N B C , u(2) = N B 1 C , u(3) = N B C ,
u(4) = N B C
@ 0 A @ 0 A @ 1 A @ 0 A
0 0 0 1
(5.53)
The first two have a positive energy eigenvalue E = m and the second two a negative
energy E = m. We discuss the normalization constant N later.
5.7.2 Plane wave solutions for p 6= 0
In order to extend the solution to particles with non-zero momentum, consider two Dirac
spinors for which the two upper coordinates uA (p) of u(p) are given by
(1) (1) (2) (2)
uA = and uA = . (5.54)
with the basis spinors 1,2 defined in Eq. 5.5. Substituting this into the second equation
of (5.51) gives for the lower two components
(1,2) · p (1,2) ·p (1,2)
uB = uA = . (5.55)
E+m E+m
5.7. SOLUTIONS TO THE DIRAC EQUATION 81
To prove that these are indeed solutions of the equations, one can use the identity
( · p) ( · p) = |p|2 112 (5.56)

(1,2) (1,2)
such that uA and uB also satisfy the first equation in (5.51). (See also exercise 5.2.)
In the Pauli-Dirac representation the Hamiltonian is given by
✓ ◆ ✓ ◆
0 ·p m112 0
H = ↵·p+ m = + (5.57)
·p 0 0 m112
With a bit of algebra we obtain for our solution

h i !
p2 (1)
m + u A
Hu(1) = E+m
(1)
, (5.58)
E uB
which illustrates two things: In order that u(1) be a solution we need indeed that
E 2 = p2 + m2 . Furthermore, in the limit that p ! 0, the energy eigenvalue is +m,
such that this is a positive energy solution. The calculation for u(2) is identical. Hence,
two orthogonal positive-energy solutions are
✓ (1)
◆ ✓ (2)
◆
(1) (2)
u =N ·p (1) and u =N ·p (2) (5.59)
E+m E+m
where N is again a normalization constant.

In an exactly analogous manner, we can start for our E < 0 solutions with the lower
component given by (s) ,
(3) (1) (4) (2)
uB = , uB = . (5.60)
Using the first of the equations in Eq. (5.51) gives for the upper coordinates
(3,4) · p (3,4) ·p (1,2)
uA = u = (5.61)
E m B ( E) + m
Note the di↵erence in the enumerator: it has become (E m) rather than (E + m).
Evaluating the energy eigenvalue, we now find e.g.
(3)
!
h E u A i
Hu(3) = 2 (3) , (5.62)
m + Ep m uB
which again requires E 2 = p2 + m2 and in the limit p ! 0 gives E = m, a negative

energy solution. Consequently, two negative-energy orthogonal solutions are given by
✓ ·p (1) ◆ ✓ ·p (2) ◆
(3) E m (4) E m
u =N (1) and u =N (2) (5.63)
To gain slightly more insight, let’s write them out in momentum components. Using
the definition of the Pauli matrices we have
✓ ◆ ✓ ◆ ✓ ◆
0 1 0 i 1 0
·p= px + py + pz (5.64)
1 0 i 0 0 1
to find ✓ ◆✓ ◆ ✓ ◆
(1) pz px ipy 1 pz
( · p) uA = = (5.65)
px + ipy pz 0 px + ipy
(2) (3) (4)
and similar for uA , uB , uB . The solutions can then be written as
0 1 0 1
1 0
B 0 C B 1 C
E > 0 spinors u(1) (p) = N B
@ pz C
A , u(2) (p) = N B
@ px ipy A
C
E+m E+m
px +ipy pz
E+m E+m
0 pz 1 0 px +ipy 1
E+m E+m
B px ipy C B pz C
E < 0 spinors u(3) (p) = N B
@
E+m C
A , u(4) (p) = N @ B E+m C
A
1 0
0 1
You can verify that the u(1) - u(4) solutions are indeed orthogonal, i.e. that
u(i)† u(j) = 0 for i 6= j. (5.66)
5.8 Antiparticle spinors
As for the solutions of the K.-G. equation, we interprete u(1) and u(2) as the positive
energy solutions of a particle (electron, charge e ) and u(3) , u(4) as the positive energy
solutions of the corresponding antiparticle (the positron). We define the antiparticle
components of the wave function as
v (1) (p) ⌘ u(4) ( p)

(5.67)
v (2) (p) ⌘ u(3) ( p) .
Using this definition we can replace the two negative

p energy solutions by the following
anti-particle spinors with positive energy, E = + p2 + m2 ,
0 px ipy 1 0 pz 1
E+m E+m
B pz C B (px +ipy ) C
v (1) (p) = N B
@
E+m C
A , v (2) (p) = N B
@
E+m C .
A
0 1
1 0
5.9. NORMALIZATION OF THE WAVE FUNCTION 83
The spinors u(p) of matter waves are solutions of the Dirac equation in momentum
space, Eq. (5.48). Replacing p with p in the Dirac equation we find that our positive
energy anti-particle spinors satisfy another Dirac equation,
(6 p + m) v(p) = 0 (5.68)
5.9 Normalization of the wave function
As for the Klein-Gordon case we choose a normalization such that there are 2E particles
per unit volume. Remember that we had in the previous lecture for the first component
of the current of the Dirac wave
†
⇢(x) = (x) (x) . (5.69)
Substituting the plane wave solution = u(p) e ipx , and integrating over a volume V
we find Z Z
3
⇢d x = u† (p) eipx u(p) e ipx d3 x = u† (p) u(p) V (5.70)
V V
Consequently, to find 2E particles per unit volume we must normalize such that
u† (p) u(p) = 2E (5.71)
Explicit calculation for the positive energy solutions (s 2 {1, 2}) gives
!
(s) † (s) T (s) T ( · p)† ( · p)
u u(s) = N 2 (s)
+ (s)
(E + m)2
✓ ◆
2 p2 2E
= N 1+ = N2
(E + m)2 E+m
Consequently, in order to have 2E particles per unit volume we choose

p
N = E+m. (5.72)
The computation for the positive energy antiparticle waves v(p) leads to the same nor-
malization. We can now write the orthogonality relations as (with r, s 2 {1, 2})
†
u(r) u(s) = 2E rs
(5.73)
(r) †
v v (s) = 2E rs
5.10 The completeness relation

We now consider the Dirac equation for the adjoint spinor u, v. Taking the hermitian
conjugate of Eq. (5.48) and multiplying on the right by 0 we have
u† µ† 0
pµ u† 0 m = 0 (5.74)
µ† 0 0 µ
Using that = we then find for the Dirac equation of the adjoint spinor
u = u† 0 ,
u (6 p m) = 0 (5.75)
In the same manner we find for the adjoint antiparticle spinors
v (6 p + m) = 0 (5.76)
Using these results you will derive in exercise 5.4 the so-called completeness relations
X
u(s) (p) u(s) (p) = (6 p + m)
s=1,2
X (5.77)
v (s) (p) v (s) (p) = (6 p m)
s=1,2
These relations will be used later on in the calculation of amplitudes with Feynman
diagrams. Note that the left-hand side is not an inner product. Rather, on both sides
we have a (4 ⇥ 4) matrix, or schematically
0 1
. 0 1 0 1
B . C
B C · (....) = @
@ . A
µ A · pµ + @ 11 A·m
P P
(Note: s=3,4 u(s) (p) u(s) (p) = s=1,2 v (s) ( p) v (s) ( p) = (6 p + m) )
5.11 Helicity
The Dirac spinors for a given momentum p have a two-fold degeneracy. This implies that
there must be an additional observable that commutes with H and p and the eigenvalues
of which distinguish between the degenerate states. Could the extra quantum number
be spin? So, eg.: u(1) = spin “up”, and u(2) =spin “down”?
Define the spin operator as S = 12 ~⌃, with
✓ ◆
0
⌃= . (5.78)
0
5.11. HELICITY 85
In exercise 5.3 you will show that ⌃ does not commute with the Hamiltonian in
Eq. (5.57). We can also realize this by looking directly at our Dirac spinor solutions: If
spin is a good quantum number then those solutions should be eigenstates of the spin
operator,
⌃ u(i) = s u(i) ?
where s is the spin eigenvalue. Now insert one of the solutions, for example u(1) ,
0 ✓ ◆ 1 0 ✓ ◆ 1
✓ ◆ 1 1
0 B B ✓ 0 C ?
◆ C
B 0 C
@ A sB
@
✓ ◆ C
A
0 pz / (E + m) = pz / (E + m)
(px + ipy ) / (E + m) (px + ipy ) / (E + m)
and you realize that this could never be true for arbitrary px , py , pz .
The orbital angular momentum operator is defined as usual as
L=r⇥p (5.79)
You will also show in exercise 5.3 that the total angular momentum
J = L + 12 ⌃ (5.80)
does commute with the Hamiltonian. Now, as we can choose an arbitrary axis to get the
spin quantum numbers, we can choose an axis such that the orbital angular momentum
vanishes, namely along the direction of the momentum. Consequently, we define the
helicity operator as ✓ ◆
1 1 · p̂ 0
= ⌃ · p̂ ⌘ (5.81)
2 2 0 · p̂
where p̂ ⌘ p/|p|. We could interpret the helicity as the “spin component in the direction
of movement”. One can verify that indeed commutes with the Hamiltonian in (5.57).
As and H commute, they have a common set of eigenvectors. However, that does
not necessarily mean that our solutions u(i) are indeed also eigenvectors of . In fact,
with our choice above, they are only eigenvectors of if we choose the momentum along
the z-axis. The reason is that the two-component spinors (s) are eigenvectors of 3
only. For other directions of the momentum, we would need to choose a di↵erent linear
combination of the u(i) to form a set of states that are eigenvectors for both H and .
Now, consider a momentum vector p = (0, 0, p). Applying the helicity operator on u(i)
gives
1 1 1
( · p̂) u±
A =
±
3 uA = ± u±
2 2 2 A
1 1 1
( · p̂) u±
B =
±
3 uB = ± u±
2 2 2 B
where the plus sign holds for u(1,3) and the minus sign for u(2,4) . So you see that indeed
u is an eigenvector of with eigenvalues ±1/2. Positive helicity states have spin and
momentum parallel, while negative helicity states have them anti-parallel.
It is not so difficult to derive the spinors that are eigenvectors of both and the Dirac
Hamilitonian for arbitrary momentum p. (See for instance section 4.8.1 in Thomson.)
We save you the algebra and just give the result. To simplify notation we switch to
polar coordinates,
p = (p sin ✓ cos , p sin ✓ sin , p cos ✓) . (5.82)
The particle spinors for helicity +1/2 and helicity 1/2 become, respectively,
0 1 0 1
cos 2✓ sin 2✓
B ei sin ✓ C B ei cos 2✓ C
u" = N B @ p
2
✓
C
A u # = N B
@ p ✓
C ,
A (5.83)
E+m
cos 2 E+m
sin 2
p i ✓ p i ✓
E+m
e sin 2 E+m
e cos 2
while those for the anti-particles are

0 p 1 0 p 1
E+m
sin 2✓ E+m
cos 2✓
B p
ei cos 2✓ C B p
ei sin 2✓ C
v" = N B @
E+m C
A v# = N B
@
E+m C .
A (5.84)
sin 2✓ cos 2✓
✓
i
e cos 2 ei sin 2✓
5.12 Charge current and anti-particles

Once we consider interactions of fermions in QED, we are interested in charge den-
sity rather than probability density. Following the Pauli-Weiskopf prescription for the
complex scalar field, we multiply the current of (negatively charged) particles by e,
µ µ
jem = e . (5.85)
The interaction current density 4-vector takes the form

0 10 1
jfµi = e ( uf )@ µ A @ ui A ei(pf pi )x
(5.86)
We have seen above that although the probability density of the Dirac fields is positive,
the negative energy solutions just remain. Following the Feynman-Stückelberg interpre-
tation the solution with negative energy is again seen as the antiparticle solution with
positive energy. However, when it comes to the Feynman rules, there is an additional
subtlety for fermions.
In the case of Klein-Gordon waves the current of an antiparticle (j µ = 2|N |2 pµ ) gets a
minus sign with respect to the current of the particle, due to reversal of 4-momentum.
5.13. THE CHARGE CONJUGATION OPERATION 87
This cancels the change in the sign of the charge and that is how we came to the nice
property of ‘crossing’: simply replace any anti-particle by a particle with opposite mo-
mentum. For fermions this miracle does not happen: the current does not automatically
change sign when we go to anti-particles. As a result, if we want to keep the convention
that allows us to replace anti-particles by particles, we need an additional ‘ad-hoc’ minus
sign in the Feynman rule for the current of the spin- 12 antiparticle.
This additional minus sign between particles and antiparticles is only required for
fermionic currents and not for bosonic currents. It is related to the spin-statistics con-
nection: bosonic wavefunctions are symmetric, and fermionic wavefunctions are anti-
symmetric. In field theory3 the extra minus sign is a result of the fact that bosonic
field operators follow commutation relations, while fermionic field operators follow anti-
commutation relations. This was realized first by Pauli in 1940.
5.13 The charge conjugation operation
The Dirac equation for a particle in an electromagnetic field is obtained by substituting

@µ ! @µ + iqAµ in the free Dirac equation. For an electron (q = e) this leads to:
µ
[ (i@µ + eAµ ) m] =0. (5.87)
Similarly, there must be a Dirac equation describing the positron (q = +e):

µ
[ (i@µ eAµ ) m] C =0, (5.88)
where the positron wave function C is obtained by a one-to-one correspondence with

the electron wave function .
To find the relation between C and , let’s take the complex conjugate of the electron
equation,
µ⇤ ⇤
[ (i@µ eAµ ) m] =0. (5.89)
Now suppose that there is a matrix M such that
µ⇤ 1 µ
= M M. (5.90)
We can then rewrite the equation above as

1 µ ⇤
M [ (i@µ eAµ ) m] M =0. (5.91)
and we obtain the relation

⇤ 0 T T
C =M =M ⌘C (5.92)
3
See Aitchison & Hey, 3rd edition §7.2
where we have used the definition of the adjoint spinor (see Lecture 6) and defined the
charge conjugation matrix C = M 0 . It can be shown (see Halzen and Martin exercise
5.6) that in the Pauli-Dirac representation a possible choice of M is
0 1
1
B 1 C
M =C 0=i 2=B @
C .
A (5.93)
1
1
Interpreting the probability current as a charge current, we define the electron current
as
jµe = e µ (5.94)
The current of the charge conjugate wave function is then
j µ eC = e C
µ
C = ... = e µ
(5.95)
Exercises
Exercise 5.1 (From Dirac to Klein-Gordon)
Each of the four components of the Dirac equation satisfies the Klein Gordon equation,
(@µ @ µ + m2 ) i = 0. Show this explicitly by operating on the Dirac equation from the
left with (i ⌫ @⌫ + m).
Hint: For any aµ and b⌫ we can write ⌫ µ a⌫ bµ = 12 ( ⌫ µ a⌫ bµ + µ ⌫ aµ b⌫ ) by just
’renaming’ indices’. Now take the special case that b = a and the aµ commute. Then
we can write ⌫ µ a⌫ aµ = 12 ( ⌫ µ + µ ⌫ )aµ a⌫ . Now use the anti-commutation relation
of the -matrices.
Exercise 5.2 (Energy eigenvalue of solutions to Dirac equation)

Starting from the Dirac equation in momentum space, Eq. (5.51), show . . .
(a) . . . by eliminating uB that a solution to the Dirac equation satisfies the relativistic
relation between energy and momentum.
(b) . . . for a non-relativistic particle with velocity , uB is a factor 12 smaller than
uA . (In a non-relativistic description uA and uB are often called respectively the
“large” and “small” components of the wave function.)
Hint: Use the result from exercise 5.7b.
Exercise 5.3 (See also exercise 5.4 of H& M and exercise 7.8 of Griffiths)
The purpose of this problem is to demonstrate that particles described by the Dirac
equation carry “intrinsic” angular momentum (S) in addition to their orbital angular
momentum (L). We will see that L and S = ⌃/2 are not conserved individually but
that their sum is.
(a) Consider the Hamiltonian that leads to the Dirac equation,
H = ↵·p+ m
Use the fundamental commutator [xi , pj ] = i~ ij (with ~ = 1) to show that
[H, L] = i↵⇥p (5.96)
where L = x ⇥ p.
Hint: To do
P this efficiently use the Levi-Civita tensor to write out the cross product
as Li = j,k ✏ijk xj , pk . Now evaluate the commutator [H, Li ].
(b) Show that

X
[↵k , ⌃l ] = 2i ✏klm ↵m
m
where the operator ⌃ (see also Eq. (5.78)) and ↵ in the Pauli-Dirac representation
were ✓ ◆ ✓ ◆
0 0
⌃= and ↵ =
0 0
Hint: Use the commutation relation
P for the Pauli spin matrices [ i , j] = 2i✏ijk k
(which follows from i j = i k ✏ijk k ).
(c) Use the result in (b) to show that
[H, ⌃] = 2 i ↵ ⇥ p (5.97)
We see from (a) and (c) that the Hamiltonian commutes with J = L + 12 ⌃.
Exercise 5.4 (See also H&M p.110-111 and Griffiths p. 242)

The spinors u, v, ū and v̄ are solutions of respectively:
(6 p m) u = 0
(6 p + m) v = 0
ū (6 p m) = 0
v̄ (6 p + m) = 0
(a) Use the orthogonality relations:
u(r)† u(s) = 2E rs
v (r)† v (s) = 2E rs
to show that:
ū(s) u(s) = 2m
v̄ (s) v (s) = 2m
Hint: evaluate the sum of u 0 (6 p m)u and u(6 p m) 0 u and use 0 k
= k 0
(k = 1, 2, 3).
(b) Derive the completeness relations:
X
u(s) (p) ū(s) (p) = 6 p + m
s=1,2
X
v (s) (p) v̄ (s) (p) = 6 p m
s=1,2
Hint: For s = 1, 2 take the solution u(s) from Eq. (5.59) and write out the row-
vector for us using the explicit form of 0 P
in the Dirac-Pauli representation. Then
(s) (s)
write out the matrix u u and use that s=1,2 (s) (s)† = 112 . Finally, note that
✓ ◆
E 112 ·p
6p = (5.98)
·p E 112
Exercise 5.5 (Dirac algebra: traces and products of matrices)

Use the anti-commutator relation for Dirac -matrices in Eq. (5.30) (or anything that
follows from that) to show that:
(a) 6 a 6 b + 6 b 6 a = 2 (a · b) 114
µ
(b) i) µ = 4 114
µ
ii) µ 6a = 2 6a
µ
iii) µ 6a 6b = 4 (a · b) 114
µ
iv) µ 6a 6b 6c = 2 6c 6b 6a
(c) i) Tr (odd number of -matrices) = 0
ii) Tr(6 a 6 b ) = 4 (a · b)
iii) Tr(6 a 6 b 6 c 6 d ) = 4 [ (a · b)(c · d) (a · c)(b · d) + (a · d)(b · c) ]
5 0 1 2 3
(d) i) Tr = Tr i =0
5
ii) Tr 6a 6b = 0
iii) Optional!, for the die-hards:
5
Tr 6a 6b 6c 6d = 4 i "↵ a↵ b c d
where "↵ = +1 ( 1) for an even (odd) permutation of 0,1,2,3; and 0 if two
indices are the same.
Exercise 5.6 (Representations of Dirac matrices)

✓
◆
a b
(a) Write a general Hermitian 2 ⇥ 2 matrix in the form where a and c
b⇤ c
are real. Write then b = s + it and show that the matrix can be written as:
{(a + c) /2} I + s 1 t 2 + {(a c) /2} 3
How can we conclude that ↵ and cannot be 2 ⇥ 2 matrices?
(b) Show that the ↵ and matrices in both the Dirac-Pauli as well as in the Weyl
representation have the required anti-commutation behaviour.
Exercise 5.7 (Pauli vector identity)

(a) Prove the Pauli vector identity
( · a)( · b) = a · b 112 + i · (a ⇥ b) (5.99)
Hint:
P Write the innerPproducts as sums over i and j. Use that i j = ij +
i k ✏i,jk k . Use that i,j ✏ijk ai bj = (a ⇥ b)k .
(b) Prove the identity in Eq. (5.56), i.e.
( · p) ( · p) = |p|2 112
Exercise 5.8 (Pauli equation as non-relativistic limit)

In this exercise you will show that in a non-relativistic approximation, the Dirac equation
combined with minimal substitution leads to the Schrödinger-Pauli equation, and hence
to the prediction of the gyromagnetic ratio of the electron.
Consider again minimal substitution for a charge q and a field Aµ = (V, A):
⇢
µ µ µ E ! E qV
p !p qA =)
p ! p qA
In the following we concentrate on the uA component because in exercise 5.2 you have
shown that in the non-relativistic limit the other component is small.
(a) Starting from the Dirac equation in momentum space Eq. (5.51),write down the
equations for uA and uB after minimal substitution.
(b) In coordinate space p and E are operators. Therefore, they do not commute with
A and V . That means that unlike in the case for free particles (exercise 5.2)
you cannot just eliminate uB ! However, in the non-relativistic limit and assuming
|qV | ⌧ m we can use the approximation E qV + m ⇡ 2m. Use this to eliminate
uB and obtain the Dirac equation for uA
[ · (p qA)] [ · (p qA)] uA = 2m(Ekin qV )uA (5.100)
where Ekin = E m.
(c) Use the Pauli vector identity, Eq. (5.99), to show that
[ · (p qA)] [ · (p qA)] = (p qA)2 iq · (p ⇥ A + A ⇥ p) (5.101)
(d) Show that in coordinate space
· (p ⇥ A + A ⇥ p) uA = i~ · (r ⇥ A) uA (5.102)
where on the right hand side the derivative r works only on A and not on uA .
Therefore, we can replace it with B = r ⇥ A.
(e) Using these results, show that the Dirac equation for an electron with charge q = e
in the non-relativistic limit in an electromagnetic field Aµ = (A0 , A) reduces to
the Schrödinger-Pauli equation
✓ ◆
d 1 2 e 0
i A = (p + eA) + · B eA A , (5.103)
dt 2m 2m
where we have identified Ekin as the classical operator id/dt.

The term with eA0 in (5.103) is a constant potential energy that is of no further impor-
tance. The term with B arises due to the fact that p and A do not commute. In this
term we recognise the magnetic field:
e
µ·B = g S·B . (5.104)
2m
Here g is the gyromagnetic ratio, i.e. the ratio between the magnetic moment of a
particle and its spin. Classically we have g = 1, but according to the Dirac equation
(S = 12 ) one finds g = 2. The current value of (g 2)/2 is according to the Particle
Data Book
(g 2)/2 = 0.001159652193 ± 0.000000000010 (5.105)
This number, and its precision, make QED the most accurate theory in physics. The
deviation from g = 2 is caused by high order corrections in perturbation theory.
Lecture 6
Spin-1/2 Electrodynamics
6.1 Feynman rules for fermion scattering

With the spinor solutions of the Dirac equation we finally have the tools to calculate
cross section for fermions, spin- 12 particles. Analogously to the case of spin-0 particles we
determine the solutions of the Dirac equations in the presence of an electromagnetic field
Aµ by starting from the free equation of motion and applying ’minimal substitution’,
pµ ! pµ qAµ . For a particle with mass m and charge q = e, the perturbed Dirac
equation then becomes
( µ pµ m) + e µ Aµ = 0 . (6.1)
To isolate the perturbation term we write this again in terms of a Hamiltonian,
(H0 + V ) = E (6.2)
One can either start from the Dirac equation in terms of ↵ and , or, work towards
that form by multiplying the Dirac equation on the left by 0 . The result is
0 k k 0
E = p + m e 0 µ Aµ (6.3)
| {z } | {z }
H0 =↵·p+ m V
Consequently, the perturbation potential is

0 µ
V (x) = e µA . (6.4)
In analogy to spinless scattering we now write for the transition amplitude

Z
† 4
Tf i = i f (x) V (x) i (x) d x (6.5)
Note the di↵erences with the case of the KG solutions in spinless scattering: The wave
function has four components and the perturbation potential V (x) becomes a (4 ⇥ 4)
93
94 LECTURE 6. SPIN-1/2 ELECTRODYNAMICS
matrix. We take a hermitian conjugate of the wave , rather than its complex conjugate.
The transition amplitude is still just a scalar.
Substituting the expression for V (x) we obtain
Z
†
Tf i = i f (x) e 0 µ Aµ (x) i (x)d4 x
Z (6.6)
µ 4
= i f (x) ( e) µ i (x)A (x) d x
In Lecture 5 we defined the charge current density of the Dirac wave as

j µ (x) = e (x) µ
(x)
In complete analogy to the spinless particle case we define the electromagnetic transition
current between states i and f as
jfµi (x) = e f (x)
µ
i (x) , (6.7)
such that the transition amplitude can be written as
Z
Tf i = i jµf i Aµ d4 x . (6.8)
ipx
After inserting the plane wave decomposition (x) = u(p)e , the transition current
becomes
jfµi = euf µ ui ei(pf pi )x . (6.9)
Note that the current is a ‘scalar’ in Dirac spinor space, or schematically,

0 10 1
jfµi = ( uf ) @ µ A @ ui A (6.10)
Now consider again the two-body scattering A + B ! C + D:

uB uD
µ
jBD
µ
jAC
uA uC
Just as we did for the scattering of spinless particles, we obtain the vector potential Aµ
by using the Maxwell equation with the transition current of one of the two particles
(say ‘particle AC’) as a source. That is, we take
µ
2Aµ = jAC .
6.1. FEYNMAN RULES FOR FERMION SCATTERING 95
and obtain for the potential

1 µ
Aµ = j .
q 2 AC
where q ⌘ pi pf is the momentum transfer. The transition amplitude becomes
Z Z
BD 1 µ 4 µ gµ⌫ ⌫ 4
Tf i = i jµ jAC d x = i jBD j dx (6.11)
q 2 q 2 AC
which is symmetric in terms of particle BD and AC. Inserting the expressions of the
plane wave currents using Eq. (6.9) we obtain
Z
gµ⌫
Tf i = i e uC µ uA ei(pC pA )x · · e uD ⌫ uB ei(pD pB )x d4 x (6.12)
q2
Performing the integral (and realizing that nothing depends on x except the exponen-
tials) leads us to the expression
Tf i = i (2⇡)4 4
(pD + pC pB pA ) M (6.13)
with the matrix element given by

igµ⌫
iM = ie (uC µ uA ) ie (uD ⌫ uB ) (6.14)
| {z } q2 | {z }
vertex | {z } vertex
propagator
From the matrix element we can now read of the Feynman rules. Again, as for the
spinless case, the various factors are defined such that the rules can also be applied to
higher order diagrams.
without spin: with spin:
1 1 ui uf
ie (pf + pi )µ ie µ
Figure 6.1: Diagrams for a spin-0 (left) and spin- 12 (right) particle with charge e interacting
with the EM field.
The rules for the vertex factors for spin-0 and spin- 12 particles are shown side-by-side
in Fig. 6.1. A spinless electron can interact with Aµ only via its charge. The coupling
is proportional to (pf + pi )µ . However, an electron with spin can also interact with the
magnetic field via its magnetic moment. As you will prove in exercise 6.1, we can rewrite
the Dirac current as
1 ⇥ ⇤
uf µ
ui = uf (pf + pi )µ + i µ⌫
(pf pi )⌫ ui (6.15)
2m
where the tensor µ⌫ was defined in Eq. 5.46. This formulation of the current is called the
‘Gordon decomposition’. We observe that in addition to the contribution that appears
for the spinless wave, there is a new contribution that involves the factor i µ⌫ (pf pi ).
In the non-relativistic limit this leads indeed to a term proportional to the magnetic
field component of Aµ , just as you would expect from a magnetic moment.
6.2 Electron-muon scattering

We will now use the Feynman rules to calculate the cross section of the process e µ !
e µ . The Feynman diagram is drawn in Fig. 6.2.
µ : uB µ : uD
⌫
ieuD uB
iM = igµ⌫
q2
µ
ieuC uA
e : uA e : uC
Figure 6.2: Lowest order Feynman diagram for e µ scattering.
Applying the Feynman rules we find for the lowest-order amplitude

i
iM = e2 uC µ
uA uD µ uB (6.16)
q2
and for its square
  ⇤
2 4 µ 1 ⌫ 1
|M| = e (uC uA ) 2 (uD µ uB ) (uC uA ) 2 (uD ⌫ uB ) (6.17)
q q
For a given value of µ and ⌫ the currents are just complex numbers. (The -matrices
are sandwiched between the bi-spinors.) Therefore, we can reorder them and write the
amplitude as
e4 X
2
|M| = 4 [(uC µ
uA ) (uC ⌫
uA )⇤ ] [(uD µ uB ) (uD ⇤
⌫ uB ) ] (6.18)
q µ⌫
We have factorized the right hand side into two tensors, each of which only depends on
one of the leptons. We call these the polarized lepton tensors.
Up to now we have ignored the fact that the particle spinors come in two flavours,
namely one for positive and one for negative helicity. Assuming that we do not measure
the helicity (or spin) of the incoming and outgoing particles, the cross-section that we
need to compute is a so-called ‘unpolarized cross-section’:
6.2. ELECTRON-MUON SCATTERING 97
• If the incoming beams are unpolarized, we have no knowledge of initial spins.

Therefore, we average over all spin configurations of the initial state;
• If the spin states of the outgoing particles are not measured, we should sum over
spin configurations of the final state.
Performing the summation and averaging leads to the following ‘unpolarized’ matrix
element X
1
|M|2 ! |M|2 = |M|2 (6.19)
(2sA + 1) (2sB + 1) spin
where 2sA + 1 is the number of spin states of particle A and 2sB + 1 for particle B. So
the product (2sA + 1) (2sB + 1) is the number of spin states in the initial state.
Some of you may wonder why in the spin summation we add up the squares of the
amplitudes, rather than square the total amplitude. The reason is that it does not make
a di↵erence since the final states over which we average are orthogonal: there is no
interference between states with di↵erent helicity. It turns out that the math is easier
when we sum over amplitudes squared.
Both the electron and the muon have s = 12 . Inserting the expression for the amplitude
above, we can write the spin averaged amplitude as
1 e4 µ⌫
|M|2 = L Lmuon (6.20)
4 q 4 electron µ⌫
where the unpolarized lepton tensors are defined as
X
Lµ⌫
electron = [uC µ uA ] [uC ⌫
uA ]⇤
e spin
X (6.21)
Lµ⌫
muon = [uD µ
uB ] [uD ⌫
u B ]⇤ .
µ spin
The spin summation is unfortunately rather tedious. The rest of the lecture is basically
just the calculation to do this!
First, take a look at the complex conjugate of the transition current that appears in the
tensor. Since it is just a (four-vector of) numbers, complex conjugation is the same as
hermitian conjugation. Consequently, we have
[uC ⌫
uA ]⇤ = [uC ⌫ uA ]†
h i† h i
† 0 ⌫ † ⌫† 0
= uC uA = u A uC (6.22)
⇥ ⇤
= uA 0 ⌫ † 0 uC = [uA ⌫ uC ]
In other words, by reversing the order of the spinors, we can get rid of the complex
conjugation and find X
Lµ⌫
e = (uC µ uA ) (uA ⌫ uC ) (6.23)
e spin
Next, we apply what is called Casimir’s trick. Write out the matrix multiplications in
the tensors explicitly in terms of the components of the matrices and the incoming spins
s and outgoing spins s0 ,
X X X (s0 ) µ (s) (s) (s0 )
Lµ⌫
e = ⌫
uC,k kl uA,l uA,m mn uC,n (6.24)
s0 s klmn
All of the factors on the right are just complex numbers, so we can manipulate their
order and write this as
X X (s0 ) (s0 ) µ X (s) (s)
Lµ⌫
e = uC,n uC,k kl ⌫
uA,l uA,m mn (6.25)
klmn s0 s
Now remember the completeness relation, Eq. (5.77), that we derived in the previous
lecture1 , X
u(s) u(s) =6 p + m (6.26)
s
Substituting this expression for the spin sums gives

X µ
Lµ⌫
e = (6 pC + me )nk kl (6 pA + me )lm ⌫
mn (6.27)
klmn
where me is the electron mass.

Let’s look more carefully at this expression: the right hand side contains products of
components of (4 ⇥ 4) matrices. Call the product of these matrices ‘A’. We could obtain
the components of A by summing P over the indices k, l and m. The final expression
for the tensor would then be L = n Ann , which is nothing else but the trace of A.
Consequently, we can write the expression for the lepton tensor also as
Lµ⌫
e = Tr [(6 pC + m)
µ
(6 pA + m) ⌫
] (6.28)
You now realize why we made you compute the traces of products of -matrices in
exercise 5.5. We briefly repeat here the properties that we need:
• In general, for matrices A, B and C and any complex number z
– Tr(zA) = z Tr(A)
– Tr (A + B) = Tr(A) + Tr(B)
– Tr (ABC) = Tr (CAB) = Tr (BCA)
µ ⌫ ⌫ µ
• For -matrices (from the anti-commutator + = 2g µ⌫ ):
– Tr(odd number of -matrices = 0)
µ ⌫
– Tr ( ) = 4 g µ⌫
for anti-fermions this gives an overall “ ” sign in the tensor: Lµ⌫
1
e ! Lµ⌫
e for each particle !
anti-particle.
6.2. ELECTRON-MUON SCATTERING 99
↵ µ ⌫
– Tr( ) = 4 g ↵ g µ⌫ g ↵µ g ⌫
+ g ↵⌫ g µ
Using the first rule we can write out the tensor as a sum of traces,
Lµ⌫
e = Tr [(6 pC + m)
µ
(6 pA + m) ⌫
]
= Tr [6 pC µ 6 pA ⌫
] + Tr [m µ m ⌫ ] + Tr [6 pC µ m ⌫ ] + Tr [m µ 6 pA ⌫
] (6.29)
| {z } | {z } | {z } | {z }
case 1 case 2 3 0 s)0 3 0 s)0
The last two terms vanish because they contain an odd number of -matrices. For the
second term (‘case 2’) we find
µ
Tr [m m ⌫ ] = m2 Tr [ µ ⌫
] = 4 m2 g µ⌫ . (6.30)
Finally, for the first term (‘case 1’) we have
⇥ ⇤
Tr [6 pC µ 6 pA ⌫ ] ⌘ Tr ↵ pC,↵ µ
pA, ⌫
⇥ ⇤
= Tr ↵ µ ⌫
pC,↵ pA,
(6.31)
= 4 g ↵µ g ⌫ g ↵ g µ⌫ + g ↵⌫ g µ pC,↵ pA,
= 4 (pµC p⌫A + p⌫C pµA g µ⌫ (pA · pC )) ,
where we used the trace formula for four -matrices in the third step. Adding the two
contributions gives for the lepton tensor
⇥ ⇤
Lµ⌫
e = 4 pµC p⌫A + p⌫C pµA + m2e pC · pA g µ⌫ (6.32)
The expression for the muon tensor is obtained with the substitution (pA , pC , me ) !
(pB , pD , mµ ), ⇥ ⇤
Lµ⌫
µ = 4 pµD p⌫B + p⌫D pµB + m2µ pD · pB g µ⌫ (6.33)
To compute the contraction of the two tensors, which appears in the amplitude, we just
write everything out
⇥ µ ⌫ ⌫ µ
⇤ ⇥ ⇤
Lµ⌫ µ
e Lµ⌫ = 4 pC pA + pC pA + me
2
pC · pA g µ⌫ · 4 pDµ pB⌫ + pD⌫ pBµ + m2µ pD · pB gµ⌫
⇥
= 16 (pC · pD ) (pA · pB ) + (pC · pB ) (pA · pD ) (pC · pA ) (pD · pB ) + (pC · pA ) m2µ
+ (pC · pB ) (pA · pD ) + (pC · pD ) (pA · pB ) (pC · pA ) (pD · pB ) + (pC · pA ) m2µ
(pC · pA ) (pD · pB ) (pC · pA ) (pD · pB ) + 4 (pC · pA ) (pD · pB ) 4 (pC · pA ) m2µ
⇤
+m2e (pD · pB ) + m2e (pD · pB ) 4m2e (pD · pB ) + 4m2e m2µ
⇥ ⇤
= 32 (pA · pB ) (pC · pD ) + (pA · pD ) (pC · pB ) m2e (pD · pB ) m2µ (pA · pC ) + 2m2e m2µ
Combining everything we obtain for the square of the unpolarized amplitude for electron-
muon scattering
e4 h
|M|2 = 8 (pC · pD ) (pA · pB ) +
q4
i
(pC · pB ) (pA · pD ) m2e (pD · pB ) m2µ (pA · pC ) + 2m2e m2µ (6.34)
We now consider the ultra-relativistic limit and ignore the rest masses of the particles.
The amplitude squared then becomes
e4 h i
|M|2 ' 8 4 (pC · pD ) (pA · pB ) + (pC · pB ) (pA · pD ) (6.35)
q
Furthermore, we define the Mandelstam variables
s ⌘ (pA + pB )2 = p2A + p2B + 2 (pA · pB ) ' 2 (pA · pB )
t ⌘ (pD pB )2 ⌘ q 2 ' 2 (pD · pB ) (6.36)
u ⌘ (pA pD )2 ' 2 (pA · pD )
where the approximation on the right follows in the ultra-relativistic limit (m ⇡ 0).
From energy-momentum conservation (pµA + pµB = pµC + pµD ) we have
(pA + pB )2 = (pC + pD )2 pA · pB = pC · pD
(pD pB )2 = (pC pA )2 =) pD · pB = pC · pA (6.37)
(pA pD )2 = (pB pC )2 pA · pD = pB · pC
which gives
1 1 1 2
(pA · pB ) (pC · pD ) = s s= s (6.38)
2 2 ◆
✓ 4
✓ ◆
1 1 1
(pA · pD ) (pC · pB ) = u u = u2 (6.39)
2 2 4
(6.40)
Inserting this in the amplitude, we find
✓ ◆
2 4 s2 + u 2
|M| ' 2 e (6.41)
t2
Finally, as we did for the spinless scattering in Lecture 4, consider again the scattering
process in the centre-of-momentum system. The four-vectors can then be written as
pµA = (|pA |, pA ) pµB = (|pA |, pA )
pµC = (|pC |, pC ) pµD = (|pC |, pC )
Define p ⌘ |pA | which, by four-vector conservation is also equal to |pB,C,D |. Define ✓ as

the angle between pA and pC (see Fig. 6.3), such that
pA · pC = pB · pD = p2 cos ✓
We then find for the Mandelstam variables
s = 4 p2
t = 2 p2 (1 cos ✓) (6.42)
u = 2 p2 (1 + cos ✓)
6.3. CROSSING: THE PROCESS E E+ ! µ µ+ 101
which gives for the amplitude squared
2 4 + (1 + cos ✓)2
|M| ' 8 e4 . (6.43)
(1 cos ✓)2
Inserting this in the expression for the di↵erential cross-section (which we obtained after
integrating over the final state momenta in exercise 2.4) we find
d 1 1 2 ↵2 4 + (1 + cos ✓)2
= |M| ' (6.44)
d⌦ c.m. 64⇡ 2 s 2s (1 cos ✓)2
with ↵ ⌘ e2 /4⇡.
µ µ pC
ieuD ⌫
uB pA
θ
iM = q2
q2
pD pB
µ
ieuC uA
e e
Figure 6.3: e µ ! e µ scattering. Left: the Feynman diagram. Right: definition of
scattering angle in C.M. frame.
6.3 Crossing: the process e e+ ! µ µ+

We now introduce the method of “crossing” by computing the amplitude for e e+ !
µ µ+ scattering from the amplitude of e µ ! e µ scattering. The rules are the
following:
1. assuming that you had computed the original amplitude in terms of particles,
replace p ! p for every anti-particle in the diagram
2. now relabel the momenta such that the ingoing and outgoing lines correspond to
those in the original diagram
3. for every crossed fermion line, i.e. for every outgoing fermion that became incom-
ing or vice-versa, multiply the amplitude squared by a factor ( 1). (This has to
do with the sign of the current which we discussed in section 5.12.)
The procedure is illustrated in Fig. 6.4. Labeling the momenta of the ‘target’ process
with primes, we use crossing to derive for the amplitude
⇥ ⇤ ⇥ ⇤
M e (p0A )e+ (p0B ) ! µ+ (p0C )µ (p0D ) = M e (p0A )µ ( p0C ) ! e ( p0B )µ (p0D )
(6.45)
e (pA ) e (pC ) e (pA ) µ (pB )
)
µ (pB ) µ (pD ) e (pC ) µ (pD )
)
e (pA ) µ+ ( pB )
p0A = pA
p0B = pC (
p0C = pB
p0D = pD
e + ( pC ) µ (pD )
Figure 6.4: Illustration of crossing. Use the anti-particle interpretation of a particle with the
4-momentum reversed in order to related the Matrix element of the “crossed” reaction to the
original one.
In other words, we can use the original computation of the amplitude provide that we
relabel the momenta as follows:
pA = p0A pB = p0C pC = p0B pD = p0D
Consequently, the Mandelstam variables of the ’original’ particle diagram are
s ⌘ (pA + pB )2 = (p0A p0C )2 ⌘ t0
t ⌘ (pD pB )2 = (p0C + p0D )2 = s0 (6.46)
u ⌘ (pA pD )2 = (p0A p0D )2 = u0
Using the result in Eq. (6.41) the amplitude squared for the two processes are then
2 s2 + u2
|M|e µ !e µ = 2 e4 ”t-channel”: q2 = t
t2
2 u02 + t02
|M|e e+ !µ µ+ = 2 e4 ”s-channel”:
s02 q2 = s
It is customary to label these as the t-channel and the s-channel process, because we
have q 2 = t and q 2 = s, respectively.
We can express the momenta in the centre-of-momentum frame in terms of an initial
momentum p and a scattering angle ✓, where ✓ is now the angle between the incoming
6.3. CROSSING: THE PROCESS E E+ ! µ µ+ 103
e (p0A ) and the outgoing µ (p0C ). The expressions for u0 , s0 and t0 are identical to those
in (6.42).
We immediately get for the matrix element:
2 t02 + u02
4
|M|c.m.
=2e 02
= e4 1 + cos2 ✓ (6.47)
s
The di↵erential cross-section becomes
d ↵2
= 1 + cos2 ✓ (6.48)
d⌦ 4s
Finally, to calculate the total cross section for the process we integrate over the azimuthal
angle and the polar angle ✓:
4⇡ ↵2
e+ e ! µ+ µ = (6.49)
3 s
Note that the ‘shape’ of the angular distribution does not depend on the available energy,
but that the total cross-section scales as 1/s: the higher the cms energy, the smaller the
cross-section. If you look back to our original formulation of the golden rule, you’ll find
that the 1/s dependence comes from the density of the incoming waves. The faster the
relative velocity, the shorter the particles are in each others vicinity!
Figure 6.5 shows a comparison of the kinematic factors in the di↵erential cross-section
of the t-channel process e µ ! e µ and the s-channel process e e+ ! µ µ+ for
spin-0 and spin- 12 leptons. For the t-channel process the di↵erence is only visible in the
very backward region, while in the s-channel process there is a constant o↵set.
t-channel s-channel
α 2 dΩ
α 2 dΩ
4s dσ
4s dσ
3
103
spin-1/2
2 spin-0
10 2
10
1
1
0
-1 -0.5 0 0.5 1 -1 -0.5 0 0.5 1
cos(θ) cos(θ)
Figure 6.5: Leading order QED di↵erential cross-section d /d⌦ divided by ↵2 /4s as function
of cos ✓ for the t-channel process e µ ! e µ (left) and the s-channel process e e+ ! µ µ+
(right) in the ultra-relativistic limit (me = mµ = 0).
Figure 6.6 shows a table copied from Halzen and Martin with the kinematic factors
for important leading order QED processes. These processes are related by crossing.
The interference terms follow via crossing procedure as well, provided that you add up
amplitudes (not amplitudes squared).
Figure 6.6: Leading order QED processes and their relations via crossing. From Halzen and
Martin, “Quarks and Leptons”.
6.4 Summary of QED Feynman rules
In the computation of the e µ above we have seen only a subset of the Feynman
rules for QED. As an example of things we missed, consider the annihilation process
e+ e ! . (Draw it!) To compute the cross-section for this process we need more
Feynman rules, namely those for the electron propagator and those for external photon
lines. We now briefly summarize the rules for QED. You can find these in more detail in
the textbooks, e.g. in appendix D of Griffiths and on the inside of the cover of Halzen
and Martin, or Thomson.
For the external lines, we have in the matrix element:

6.4. SUMMARY OF QED FEYNMAN RULES 105
spin-0: nothing
8
>
> incoming particle: u
<
outgoing particle: u
spin- 12 :
>
> incoming anti-particle: v (6.50)
:
outgoing anti-particle: v
⇢
incoming: ✏µ
spin-1:
outgoing: ✏⇤µ
We have seen the photon polarization vectors in Lecture 3. Both the spin- 12 and spin-1
external lines carry also an index for the helicity. In calculations for cross-sections or
decays in which we measure the spin, we need explicit forms of the Dirac spinors and
the photon polarization vectors. However, often we sum over all incoming and outgoing
spins (’spin averaging’) and we can use the completeness relations.
For the internal lines (the propagators) we have
i
spin-0:
q2 m2
i(6 q + m)
spin- 12 :
q 2 m2 (6.51)
8 ig µ⌫
>
< massless:
q2
spin-1: µ⌫ µ ⌫ 2
: masssive: i [ g + q q /m ]
>
q 2 m2
Finally, the QED vertex factors are

spin-0: ige (pµin + pµout )
(6.52)
spin- 12 : ige µ
with ge the charge of the particle in the vertex. Section 7.6 of Griffiths contains worked
out examples of several key QED processes, both with and without spin averaging.
Exercises
Exercise 6.1 (The Gordon decomposition)
A spinless electron can interact with Aµ only via its charge; the coupling is proportional
to (pf + pi )µ . An electron with spin, on the other hand, can also interact with the
magnetic field via its magnetic moment. This coupling involves the factor i µ⌫ (pf pi ).
The relation between the Dirac current and the Klein-Gordon current can be studied as
follows:
(a) Show that the Dirac current can be written as

1 ⇥ ⇤
uf µ
ui = uf (pf + pi )µ + i µ⌫
(pf pi )⌫ ui
2m
where the tensor is defined as
µ⌫ i µ ⌫ ⌫ µ
= ( )
2
Hint: Start with the term proportional to µ⌫ and use: µ ⌫ + ⌫ µ
= 2g µ⌫ and
use the Dirac equations: ⌫ pi⌫ ui = mui and uf ⌫ pf ⌫ = muf .
(b) (optional!) Make exercise 6.2 on page 119 of H&M which shows that the Gordon
decomposition in the non-relativistic limit leads to an electric and a magnetic
interaction. (Compare also exercise 5.8.)
Exercise 6.2
Can you easily obtain the cross section of the process e+ e ! e+ e from the result of
e+ e ! µ+ µ ? If yes: give the result, if no: why not?
Exercise 6.3 (The process e+ e ! ⇡ + ⇡ )

We consider scattering of spin 1/2 electrons with spin-0 pions. We assume point-
particles; i.e. we forget that the pions have a substructure consisting of quarks. Also we
only consider electromagnetic interaction and we assume that the particle masses can
be neglected.
(a) Consider the process of electron - pion scattering: e ⇡ ! e ⇡ . Draw the Feyn-
man diagram and write down the expression for the iM using the Feynman
rules.
2
(b) Perform the spin averaging of the electron and compute |M| .
2
(c) Use the principle of crossing to find |M| for e+ e ! ⇡ + ⇡
(Note the extra minus sign that appears from the 3rd crossing rule.)
(d) Determine the di↵erential cross section d /d⌦ for e+ e ! ⇡ + ⇡ in the centre-of-
momentum of the e+ e -system.
Lecture 7
The Weak Interaction
In 1896 Henri Becquerel studied the e↵ect of fluorescence, which he thought was related
to X-rays that had been discovered by Wilhelm Röntgen. To test his hypothesis he
wrapped a photographic plate in black paper and placed various phosphorescent salts
on it. All results were negative until he used uranium salts. These a↵ected photographic
plates even when put in the dark, such that the e↵ects clearly had nothing to do with
fluorescence. Henri Becquerel had discovered natural radioactivity, and thereby the weak
interaction.
We know now that the most nuclear decays are the result of the transition of a neutron
to an electron, a proton and an anti-neutrino,
p+
n
e−
νe
or in a formula,
n ! p + e + ⌫e . (7.1)
A ‘free’ neutron has a lifetime of about 15 minutes, but the lifetime of various weakly
decaying isotopes spans a very wide range.
7.1 Lifetimes and couplings
Compare the lifetime of the following particles:
107
108 LECTURE 7. THE WEAK INTERACTION
particle lifetime [sec] dominant decay mode

⇢0 4.4 · 10 23 ⇢0 ! ⇡ + ⇡
⇡0 8.4 · 10 17 ⇡0 !
⇡ 2.6 · 10 8 ⇡ ! µ ⌫µ
µ 2.2 · 10 6 µ ! e ⌫ e ⌫µ
The listed decay modes are the dominant decay modes. Other decay modes exist, but
they contribute marginally to the total decay width. As we have seen before, the lifetime
of a particle is inversely proportional to the total decay width,
1
⌧ = . (7.2)
We have also seen that the decay width to a particular final state is proportional to the
matrix element squared. For example, for the two-body decay A ! B + C we had (in
particle A’s rest frame)
Z
|M|2 pB
(A ! B + C) = d = |M|2 (7.3)
2EA 8⇡m2A
In Yukawa’s picture of scattering by particle exchange, the leading order contribution to

the matrix element is proportional to the square of the coupling constant. Consequently,
the lifetime of particles tells us something about the strength of the interaction that is
responsible for the decay.
All fundamental fermions in the standard model ‘feel’ the weak interaction. However,
in processes that can also occur via the strong or electromagnetic interaction, those
interactions will dominate. The reason that we still see the e↵ects of the weak interaction
is because the strong and electromagnetic interaction do not change quark and lepton
flavour. Consequently, if a particle cannot decay to a lighter state obeying the ’flavour-
conservation’ rule, then it can only decay through the weak interaction. In contrast to
quarks and charged leptons, neutrinos feel only the weak interaction. That is the reason
why they are so hard to detect!
We can now understand the hierarchy of the lifetimes above as follows:
• The ⇢0 particle (which is an excited meson consisting of u and d quarks and their
anti-quarks) decays via the strong interaction to two pions.
• The ⇡ 0 is the lightest neutral hadron such that it cannot decay to hadrons. It
decays via the electromagnetic interaction to two photons, as we have seen in
exercise 4.3.
• The ⇡ + is the lightest charged hadron. Because it is charged, it cannot decay two
photons. Instead, it decays via the weak interaction to a µ+ and a neutrino. (It
could also decay to an e+ and a neutrino, but for reasons explained later that
mode is kinematically suppressed, despite the larger ‘phase space’.)
7.2. THE 4-POINT INTERACTION 109
• The µ+ is a lepton and therefore does not couple to the strong interaction. It can-
not decay to an electron and photon, as the electromagnetic interaction conserves
lepton flavour. Its dominant decay is via the weak interaction to an electron and
neutrinos.
Considerations like these explain the gross features in the hierarchy of lifetimes. How-
ever, as you can also judge from the wide range in lifetimes of particles that decay
weakly, kinematic e↵ects must be important as well.
Besides the fact proper that the weak interaction unlike the electromagnetic and strong
interaction does not ‘honour’ the quantum numbers for quark and lepton flavour, the
weak interaction is special in at least two more ways:
• it violates parity symmetry P . Until 1956, when the parity violating aspects of
the weak interaction were demonstrated, physicists were convinced that at least
at the level of fundamental interactions our world was left-right symmetric;
• in the quark sector, it even violates CP symmetry. That means, because of CP T
invariance, that it also violates T (time-reversal) symmetry. As we shall see, the
existence of a third quark family was predicted from the observation that neutral
Kaon decays exhibit CP violation.
7.2 The 4-point interaction
In 1932 Fermi tried to formulate a theory to describe nuclear decays quantitatively. He

proposed a so-called 4-point interaction, introducing the Fermi constant as the strength
of the interaction: GF ⇡ 1.166 · 10 5 GeV 2 .
n p+ The “Feynman diagram” of the 4-point in-

teraction “neutrino scattering on a neutron”
has the following matrix element:
µ
M = GF (up un ) (ue µ u⌫ ) (7.4)
⌫e e
This is to be compared to the electromagnetic diagram for electron proton scattering:
p+ p+ Here the matrix element was:

4⇡↵ µ
M= (up up ) (ue µ ue ) (7.5)
q 2 q2
1. e2 = 4⇡↵ is replaced by GF
e e 2. 1/q 2 is removed
We take note of the following properties of the weak interaction diagram:

1. The matrix element involves a hadronic current and a leptonic current. In contrast
to electromagnetic scattering, these currents change the charge of the particles
involved. In this particular process we have Q = 1 for the hadronic current and
Q = 1 for the leptonic curent. Since there is a net charge from the hadron to
the lepton current we refer to this process as a charged current interaction. We
will see later that there also exists a neutral current weak interaction.
2. There is no propagator; ie. a “4-point interaction”.
3. There is a coupling constant GF , which plays a similar rôle as ↵ in QED. Since
there is no propagator, the coupling constant is not dimensionless.
4. The currents have what is called a “vector character” similar as in QED. This
µ
means that the currents are of the form .
The vector character of the interaction was just a guess that turned out successful
to describe many aspects of -decay. There was no reason for this choice apart from
similarity with quantum electrodynamics. In QED the reason that the interaction has
a vector behaviour is because the force mediator, the foton, is a spin-1 (or ‘vector’)
particle.
In the most general case the matrix element of the 4-point interaction can be written as
M = GF p (4 ⇥ 4) n e (4 ⇥ 4) ⌫ , (7.6)
where the (4 ⇥ 4) is a matrix. Lorentz invariance puts restrictions on the form of these
matrices. We have seen these already in lecture 5: Any such matrix needs to be a
so-called bilinear covariant. The bilinear covariants all involve 4 ⇥ 4 matrices that are
products of matrices:
current # components # -matrices spin
Scalar 1 0 0
µ
Vector 4 1 1
µ⌫
Tensor 6 2 2
µ 5
Axial vector 4 3 1
5
Pseudo scalar 1 4 0
The last column in the table shows the spin of the mediator in the interaction, if the
4-point interaction is interpreted as the result of a force-carrier exchange.
The most general 4-point interaction now takes the following form:
S,P,V,A,T
X
M = GF Cij (up Oi un ) (ue Oj u⌫ ) (7.7)
i,j
where Oi , Oj are operators of the form S, V , T , A, P . The kinematics of a decay

depend on the type of operator involved. For example, it can be shown (see eg. Perkins:
7.3. PARITY 111
“Introduction to High Energy Physics”, 3rd edition, appendix D) that for the decay
n ! pe ⌫ e
• S, P and T interactions imply that the helicity of the e and the ⌫ e have the same
sign;
• V and A interactions imply that they have opposite sign.
Fermi had assumed that the weak interaction was of the V type. In a number of
experiments performed in the late fifties it was established that the weak interaction
was a combination of V and A. Before we look at that in more detail, we need to discuss
the concept of parity.
7.3 Parity
Parity, or (space) inversion, is the operation that multiplies all spatial coordinates by
1, so x ! x. It is closely related to reflection in a mirror: the parity operation
is identical to a reflection in a plane through the origin, followed by a rotation under
180 degrees around an axis through the origin perpendicular to the mirror. Therefore,
for systems that are rotation and translation invariant, the two are equivalent. When
illustrating parity violation in pictures, we usually use an image with a reflection in a
mirror. Yet, when formulating the e↵ect of parity in a physics theory, we work with
space inversion.
Now, consider a process i ! f for some initial state i and final state f . The relation
between i and f given by an operator that describes the time evolution,
f = Ûf i i (7.8)
(We can look at the process at any time scale. So Û can just be a continuous function
of time.) Denoting the parity operation by P̂ we can also consider the mirror process,
characterised by 0i = P̂ i and 0f = P̂ f . We define the process to be ‘symmetric under
parity’ when it does not make any di↵erence whether we first transform i to its mirror
image and then look at its time evolution, i ! 0i ! 0f , or first wait for the system
to evolve and then reflect it, i ! f ! 0f . Or, in terms more common in quantum
mechanics, the process is symmetric under parity when P̂ and Û commute,
[Û , P̂ ] = 0 (7.9)
Because for small times t we have Û (t) = e iHt/~ ⇡ 1 iHt/~, it follows that such P̂
also commutes with the Hamiltonian. This definition of a symmetry is not limited to
mirror symmetry, but holds for any operator: if an operator Q̂ commutes with H then
it is called a symmetry operator.
If P̂ and H commute, then they have a common set of eigenvectors. If we consider
eigenvectors with energy E that are not degenerate (that is, there is no other state with
equal energy) then this immediately implies that these states have definite parity: they
are eigenstates of the parity operator and there is an observable property (a quantum
number) associated with the parity operation.
If we apply the parity operator twice, then we put the system back in its original state.
Consequently, if p is the eigenvalue of our state under the parity operator, then p2 = 1.
(Strictly speaking, the system would be in the same state even if we had changed the
wave function by an arbitrary phase. However, for simplicity we will not deal with the
minor complications that this introduces.) Therefore, the eigenvalue is either +1 or 1.
We call such states states of even and odd parity respectively.
Until 1956 all the known laws of physics were invariant under inversion symmetry. At
the scale of elementary particles our world was perfectly left-right symmetric. This
symmetry was well tested for the electromagnetic and strong interaction and it was
generally assumed that it held for the weak interaction as well.
Since all all our leptons, mesons, and baryons are characterised by di↵erent masses (e.g.
by di↵erent eigenvalues of the total Hamiltonian) they all have definite parity: they are
either odd or even under the parity operation. (You will find their quantum numbers for
parity in the PDG.) These facts, definite parity for all quasi-stable particle and parity
conservation in known interactions, is exactly what lead in the early fifties to what was
called the ‘theta-tau puzzle’.
The ✓ and the ⌧ were charged particles with strangeness one that decayed through the
weak interaction to two and three pions respectively,
✓+ ! ⇡ + + ⇡ 0
(7.10)
⌧ + ! ⇡ + + 2⇡ 0 or 2⇡ + + ⇡
The pions were all known to have parity 1. Then, assuming parity to be conserved
in these processes the theta had even parity and the tau odd parity. However, what
was truly strange is that the theta and tau were otherwise seemingly identical particles:
they had the same mass and same lifetime.
After verifying that there had never been any experimental tests of parity conservation
in the weak interaction, Lee and Yang hypothesized in 1956, that the tau and theta
were actually the same particle, and that the weak interaction was responsible for the
apparent violation of parity. They also proposed a number of experiments that could
establish parity violation in weak decays directly. Within half a year two of these
experiments were performed (Wu et al. (1957), Garmin, Lederman and Weinrich (1957))
and the parity violating character of the weak interaction was firmly established.
7.4. COVARIANCE OF THE WAVE EQUATIONS UNDER PARITY 113
7.4 Covariance of the wave equations under parity

Let us now consider the parity transformation of the solutions to the wave equations in
more detail. In our notation the parity operation transforms
x ! x0 = x
t ! t0 = t (7.11)
(x) ! 0 (x0 )
We would now like to establish the relation between 0 (x0 ) and (x). For the wave
equation to be ‘covariant’ (that is, the same form in every frame) this relation is to be
chosen such that if (x) satisfies the wave equation, then 0 (x0 ) satisfies the same wave
equation.
Now remember the Klein-Gordon equation for a free particle:
(@µ @ µ + m2 ) (x) = 0 (7.12)
Because this equation is quadratic in @µ , the equation itself does not change under
parity. Hence, one possible solution is simply
0 0
KG (x ) = KG (x) = KG ( x0 ) (7.13)
Note that any global change in the phase for 0 is allowed as well: such a phase change
can be considered part of our definition of parity and we can choose it to be zero. If we
take to be a positive energy plane-wave solution with momentum p, then 0 is also a
positive energy solution but with momentum p. (Try!)
Now consider the more complicated case of the Dirac equation. Split in its time and
space part, the Dirac equation in coordinate space can be written as
@ (x, t)
i = ( i↵ · r + m) (x, t) (7.14)
@t
In order for this equation to be covariant under parity transformations we must find the
field 0 (x0, t) such that it satisfies the transformed Dirac equation,
@ 0 (x0 , t)
i = ( i↵ · r0 + m) 0
(x0 , t) (7.15)
@t
Applying our definition of space inversion, which implies r0 = r, we find
@ 0 ( x, t) 0
i = (i↵ · r + m) ( x, t) (7.16)
@t
Note that 0 ( x, t) does not satisfy the Dirac equation due to the additional minus sign
in front of the spatial derivative. However, we now multiply the equation on the left by
and use that fact that ↵ and ant-commute. The result is
0
@ ( x, t) 0
i = ( i↵ · r + m) ( x, t) (7.17)
@t
0
Hence ( x, t) does satisfy the Dirac equation. Consequently, one choice for the
parity operation is
0
(x0 , t) = ( x0 , t) (7.18)
Note, again, that we could insert any constant phase factor in the transformation. By
convention, we choose that factor to be one.
We now look at the solutions to the Dirac equation in the Pauli-Dirac representation.
In this representation, we have for the = 0 matrix:
✓ ◆
0 11 0
= (7.19)
0 11
Consequently, the parity operator has opposite sign for the positive and negative energy
solutions. In other words, fermions and anti-fermions have opposite parity. With our
choice of the phase of the parity transformation, fermions have positive parity and anti-
fermions have negative parity.
What does this mean for the currents in the interactions? Under the parity operation
we find
0 0
S: ! = Scalar
5 0 5 0 5
P : ! = Pseudo Scalar
⇢ 0
µ 0 µ 0
V : ! = k Vector
⇢ 0
µ 5 0 µ 5 0
A: ! = k Axial Vector.
Experiments in the fifties had shown that the weak interaction was of the type vector or
axial vector. However, if only a single bi-linear covariant contributes to the interaction,
a parity transformation does not a↵ect the cross-section or decay width as these are
always proportional to the amplitude squared. Consequently, the experiments by Wu
and others implied that the weak interaction received contributions from both the vector
and the axial vector covariants,
V,A
X
M = GF Cij (up Oi up ) (ue Oj u⌫ ) (7.20)
i,j
Which combination of V and A appears in the weak interaction was established with a
famous experiment by Goldhaber.
7.5 The V A interaction

Goldhaber and collaborators studied in 1958 the decay of Europium-152 to Samorium
via a so-called electron capture reaction, 152 Eu + e !152 Sm⇤ (J = 1) + ⌫:
7.5. THE V A INTERACTION 115
152 152
Eu + e ! Sm⇤ + ⌫e
direction of travel: !
1/2 1 +1/2
spin configuration A: ( (= )
+1/2 +1 1/2
spin configuration B: ) =) (
The excited Samorium nucleus is in a J = 1 state. To conserve angular momentum,

J must be parallel to the spin of the electron, but opposite to that of the electron-
neutrino. The neutrino in this decay cannot be observed. However, the spin of the
Samorium nucleus can be probed with the photon that is emitted in its decay to the
ground state,
152
Sm⇤ !152 Sm + .
If we imagine the photon traveling left and project spins onto the photon direction, then
situations A and B in the picture correspond to two di↵erent helicity configurations for
the photon and the neutrino. Therefore, a measurement of the photon helicity is also
a measurement of the neutrino helicity. Measuring the photon spin is a work-of-art by
itself (for a good description of the experiment, see Perkins ed 3, §7.5.), but assuming
it can be done, this allows to distinguish topologies A and B. The measurement by
Goldhaber’s group showed that only case ‘B’ actually occurs: the neutrinos in this
decay always have helicity 12 . It is therefore said that neutrinos are left-handed.
The cumulative evidence from Goldhaber’s and other experiments involving weak in-
teractions led to the conclusion that the weak interaction violates parity maximally.
Rather than the vector form assumed by Fermi, the charge-lowering lepton current is
actually
J µ = ue µ 12 1 5
u⌫ . (7.21)
The current for quarks looks identical, for example for a u ! d transition
J µ = uu µ 1
2
1 5
ud . (7.22)
We call this the “V-A” form.
For massless particles (or in the ultra-relativistic limit), the projection operator
1 5
PL ⌘ 2
1 (7.23)
selects the ‘left-handed’ helicity state of a particle spinor and the right-handed helicity
state of an anti-particle spinor. As a result, only left handed neutrinos (⌫L ) and right-
handed anti-neutrinos (⌫ R ) are involved in weak interactions.
For decays of nuclei the structure is more complicated since the constituents are not
free elementary particles. The matrix element for neutron decays can be written as
GF µ 5 5
M = p up CV CA un ue µ 1 u⌫ . (7.24)
2
For neutron decay, the measured vector and axial vector couplings are CV = 1.000 ±
0.003, CA = 1.260 ± 0.002
7.6 The propagator of the weak interaction
The Fermi theory has a 4-point interaction, unlike the Yukawa theory: there is no
propagator to ‘transmit’ the interaction from the lepton current to the hadron current.
However, we know now that forces are carried by bosons:
• the electromagnetic interaction is carried by the massless photon with a propagator
ig µ⌫
q2
• the weak interaction is carried by the massive W , Z bosons, for which we have
the propagators:
i g µ⌫ q µ q ⌫ /MZ,W
2
2
.
MZ,W q2
At low energies, i.e. when q 2 ⌧ MZ,W

2
, the q 2 dependence of the propagator vanishes and
the interaction looks like a four-point interaction. Therefore, Fermi’s coupling constant
is related to the actual coupling constant ‘g’ of the weak interaction:
W
g g
G g2
strength: ⇠ pF
2
⇠ 2
8MW
It is an experimental fact that the strength of the coupling of the weak interaction, the
coupling constant “g”, is identical for quarks and leptons of all flavours. For leptons
this is sometimes called ‘lepton-universality’.
e 2
How “weak” is the weak interaction? For the electromagnetic coupling we have ↵ = 4⇡ ⇡
g 2
1/137. It turns out that the weak coupling is equal to ↵w = 4⇡ ⇡ 1/29. We see that at
low energies, the weak interaction is ‘weak’ compared to the electromagnetic interaction
not because the coupling is small, but because the propagator mass is large! At high
energies q 2 & MW2
the weak interaction is comparable in strength to the electromagnetic
interaction.
7.7. MUON DECAY 117
7.7 Muon decay
Similar to the process e+ e ! µ+ µ in QED, the muon decay process µ ! e ⌫ e ⌫µ is

the standard example of a weak interaction process. The Feynman diagram is shown in
Fig. 7.1.
e−(p’) νµ (k)
µ−(p) µ−
νe(k) (p)
W e−(p’)
νµ (k’) νe(−k’)
Figure 7.1: Muon decay: left: Labelling of the momenta, right: Feynman diagram. Note
that for the spinor of the outgoing antiparticle we use: u⌫e ( k 0 ) = v⌫e (k 0 ).
Using the Feynman rules we can write for the matrix element:
0 1 0 1
g B µ1 C 1 g 1
M = p @ u(k) 1 5
u(p) A 2
p @ u(p0 ) µ 1 5
v(k 0 ) A
2 |{z} 2 |{z} MW 2 | {z } 2 | {z }
outgoing ⌫µ incoming µ |{z} outgoing e outgoing ⌫ e
propagator
(7.25)
Next we square the matrix element and sum over the spin states, just like we did for
e+ e ! µ+ µ . Then we use again Casimir’s tric, as well as the completeness relations,
to convert the sum over spins into a trace. The result is:
✓ 2 ◆2
2 1 X 2 1 g
|M| = |M| = 2
· Tr µ
1 5
(6 p 0 + me ) ⌫
1 5
6k 0
2 Spin 2 8MW
5 5
· Tr µ 1 6k ⌫ 1 (6 p + mµ )
G g2
Now we use some more trace theorems (see below) and also pF
2
= 8MW2 to find the result:
2
|M| = 64 G2F (k · p0 ) (k 0 · p) (7.26)
Intermezzo: Trace theorems used (see also Halzen & Martin p 261):
µ ⌫
Tr ( 6a 6 b ) · Tr ( µ 6 c ⌫ 6 d ) = 32 [(a · c) (b · d) + (a · d) (b · c)]
µ ⌫ 5
Tr 6a 6 b · Tr µ 6 c ⌫ 5 6 d = 32 [(a · c) (b · d) (a · d) (b · c)]
µ 5
Tr 1 6a ⌫ 1 5
6 b · Tr µ 1 5
6c ⌫ 1 5
6 d = 256 (a · c) (b · d)
The decay width is computed with Fermi’s golden rule:

1 2
d = |M| dQ (7.27)
2E
where
d3 p 0 d3 k d3 k 0
dQ = · · · (2⇡)4 4
(p p0 k0 k) (7.28)
(2⇡)3 2E 0 (2⇡)3 2! (2⇡)3 2! 0
with
E = muon energy
E0 = electron energy
!0 = electron neutrino energy
! = muon neutrino energy
First we evaluate the expression for the matrix element. Working in the rest frame of
the muon and ignoring the mass of the electron and the neutrinos, we find (Eq. 7.48 in
exercise 7.3),
2
|M| = 64 G2F (k · p0 ) (k 0 · p) = 32 G2F m2 2m! 0 m! 0 (7.29)
where m is the muon mass. Inserting this in the expression for the di↵erential decay
width, we obtain
1 2 16G2F
d = |M| dQ = (m2 2m! 0 m! 0 dQ (7.30)
2E m
where we used that E = m in the muon rest frame. To obtain the total decay width we
must integrate over the phase space,
Z Z
1 2 16G2F
= |M| dQ = (m2 2m! 0 m! 0 dQ (7.31)
2E m
The integrand only depends on the neutrino energy ! 0 . So, let us first perform the
integral in dQ over the other energies and momenta:
Z Z
1 0 0 3 0 0 d3 p0 d3 k0 d3 k
dQ = (m E ! !) (p + k + k)
other 8 (2⇡)5 E 0 !0 !
Z
1 0 0 d3 p0 d3 k0
= (m E ! !)
8 (2⇡)5 E 0!0!
since the -function gives 1 for the integral over k.

We also have the relation:
p
! = |k| = |p0 + k0 | = E 02 + ! 02 + 2E 0 ! 0 cos ✓ (7.32)
7.7. MUON DECAY 119
where ✓ is the angle between the electron and the electron neutrino. We choose the
z-axis along k0 , the direction of the electron neutrino. From the equation for ! we
derive:
2E 0 ! 0 sin ✓ ! d!
d! = p d✓ , d✓ = (7.33)
2 E 02 + ! 02 + 2E 0 ! 0 cos ✓ E 0 ! 0 sin ✓
| {z }
!
Next we integrate over d3 p0 = E 02 sin ✓ dE 0 d✓ d with d✓ as above:

Z
1 0 0 E 02 sin ✓ 0 d3 k0 1
dQ = (m E ! !) dE d✓ d
8 (2⇡)5 E0 !0 !
Z
1 0 0 0 d3 k 0
= 2⇡ (m E ! !) dE d!
8 (2⇡)5 ! 02
!
(using the relation: E 0 sin ✓ d✓ = !0
d!).
Since we integrate over !, the -function will cancel:
Z 3 0
1 0 d k
dQ = dE (7.34)
8 (2⇡)4 ! 02
such that the full expression for becomes:

Z 3 0
2G2F 2 0 0 0 d k
= m 2m! ! dE (7.35)
(2⇡)4 ! 02
Next we do the integral over k 0 as far as possible with:

Z Z Z
d k = ! sin ✓ d! d✓ d = 4⇡ ! 02 d!
3 0 02 0 0 0 0
(7.36)
so that we get: Z
G2 m
= F 3 (m 2! 0 ) ! 0 d! 0 dE 0 (7.37)
(2⇡)
Before we do the integral over ! 0 we have to determine the limits:
νe e−
• maximum electron neutrino energy:
! 0 = 12 m νµ
• minimum electron neutrino energy:
! 0 = 12 m E 0 νµ e−
νe
Therefore, we obtain for the distribution of the electron energy in the muon rest frame
Z 1m ✓ ◆
d G2F m 2
0 0 0 G2F m2 02 E0
= (m 2! ) ! d! = E 3 4 . (7.38)
dE 0 (2⇡)3 12 m E 0 12⇡ 3 m
VOLUME 14, NUMBER 12 PHYSICAL REVIE%' LETTERS 22 MARCH 1965
of internal consistency of the data over various mus for fabricating the chambers; J. Williams
momentum and angular regions lead to larger for computer programming; and C. Carlson,
uncertainty and a preliminary result of p =0.747 S. Herzka, B. Palatnick, and S. Stein for gen-
+ 0.005. eral assistance in the experiment.
120 %'e wish to thank Dr. G. Sutter for help in LECTURE 7. THE WEAK INTERACTION
the early phases of the experiment; F. Sippach *Work supported in part by the U. S. Office of Naval
for the design of the electronic system', G. Dore- Research under Contract No. Nonr-266(72}.
X
LA
CU
I-
IO 20 30 40 50
POS I TRON MOME NTU M M eV/c
Figure 7.2: Experimental —

measurement
300 of the energy spectrum of a positron in the decay of
+ —
µ . The superimposed curve is the prediction. From Bardon et al., PRL 14, 449 (1965).
200
—IOO
X p =. 737 r
Z —0 p =.747
I II
0.
x —-i00 p =. 757
z'
Figure 7.2 shows a comparison
—-200 between this prediction and an actual measurement.
—-300
Finally, integrating the expression over the electron energy we find for the total decay
I I l
30
width of the muon 40 50
FIG. 4. {a}Experimental points for magnetic-field settings, normalized to the overlap region. The solid line is
the theoretical spectrum for p= 0.75. The Michel spectrum, 24
1 GF m 5
p(x)dx =⌘
zjl2x —=12x +p[(32/3)x —8x ])dx, (7.39)
where x is the positron momentum
⌧ 192 ⇡ 3
divided by its maximum value, has been corrected for internal radiation,
bremsstrahlung, and ionization loss. (b) The deviation of experimental points from the best-fit theoretical curve
for p= 0.747, showing typical experimental errors for four points. Curves for p= 0.737 and 0.757 are shown for
The measurement of the muon lifetime is the standard method to determine the coupling
comparison.
constant of
452 the weak interaction. The muon lifetime has been measured to be ⌧ =
2.19703 ± 0.00004 µs. From this we derive for the Fermi coupling constant GF =
(1.16639 ± 0.00002) · 10 5 GeV 2 .
7.8 Quark mixing
The strong and electromagnetic interaction do not couple to currents that connect lep-
tons or quarks of di↵erent flavour: These interactions conserve the type of lepton or
quark at the interaction vertex.
This is di↵erent for the weak interaction: As the W is charged, it necessarily couples to
a current that contains two particles that di↵er one unit in charge. For the quarks and
7.8. QUARK MIXING 121
leptons in the standard model, the Feynman diagrams for the interactions are
⌫e e ⌫µ µ ⌫⌧ ⌧
g g g
W W W
u d c s t b
g g g
W W W
Leptons and quarks are usually ordered in three ‘generations’ to show how the weak
interaction couples:
✓ ◆ ✓ ◆ ✓ ◆ ✓ ◆ ✓ ◆ ✓ ◆
⌫e ⌫µ ⌫⌧ u c t
Leptons: Quarks: (7.40)
e µ ⌧ d s b
If the only couplings of the W are those shown in the Feynman diagrams, then the
lightest hadrons with a strange quark (such as the K which is a sū bound state) would
be stable. However, K mesons do decay, for instance to a muon and a muon neutrino:
u
?? g µ−
−
K W νµ
s
This decay looks a lot like that of the ⇡ , which is a dū bound state:
u
g g µ−
π−
W νµ
d
Experimentally the K decay is found to have an much smaller decay width than the
pion decay.
In 1963 Nicola Cabibbo provided a solution that explained most available data on strange
hadron decay by presenting the d quark in the current that couples to the W as a linear
combination of a d quark and an s quark:
d ! d0 = d cos ✓c + s sin ✓c
(7.41)
s ! s0 = d sin ✓c + s cos ✓c
where ✓c is a ‘mixing’ angle now known as the Cabibbo angle. In matrix representation
the mixing can be written as
✓ ◆ ✓ ◆✓ ◆
d0 cos ✓c sin ✓c d
= (7.42)
s0 sin ✓c cos ✓c s
while in terms of the diagrams the replacement looks like:
u d u d0 u d u s
g g g cos ✓c g sin ✓c
) = +
W W W W
Due to the mixing the amplitudes for pion and kaon decay contain factors cos ✓c and
sin ✓c :
1. Pion decay
u gcos θ g µ−
⇡ ! µ ⌫µ π−
2 2
⇡ / GF cos ✓c
W νµ
d
2. Kaon decay
u gsin θ g µ−
K ! µ ⌫µ −
2 2
K W νµ
K / GF sin ✓c
s
A proper calculation gives for the ratio of the decay rates
✓ ◆3 ✓ ◆2
(K ) m⇡ m2K m2µ
⇡ tan2 ✓c · (7.43)
(⇡ ) mK m2⇡ m2µ
From the experimental result on the lifetime ratios, the Cabibbo angle is then found to
be
✓C ⇡ 13.0 (7.44)
Even though Cabibbo’s theory explained strange decays, it did not quite get everything
right. The proposed quark mixing would allow neutral kaons (sd¯ combinations) to decay
to muons, via the amplitude represented by this Feynman diagram:
µ+
d W+
K 0 u ⌫µ
s W
µ
According to Cabibbo’s calculation this decay should have an appreciable rate, but it
was never found! An explanation was provided by Glashow, Iliopoulis and Maiani in
1970: They hypothesised the existence of the charm (c) quark, contributing with a
diagram
µ+
d +
W
K 0 c ⌫µ
s W
µ
The up and charm quark amplitudes have opposite sign, which leads to a nearly van-
ishing decay rate. This mechanism, which is now known as the GIM mechanism, was
the first well-motivated prediction for a fourth quark. The charm quark was discovered
3 years later.
Including charm quarks, the couplings for the first two generations are:
u d c s u s c d
g cos ✓c g cos ✓c g sin ✓c g sin ✓c
W W W W
| {z } | {z }
Cabibbo “favoured00 decay Cabibbo “suppressed00 decay
The flavour eigenstates u, d, ✓s, c ◆are✓the ◆ mass eigenstates of the total Hamiltonian
u c
describing quarks. The states , are the eigenstates of the weak interaction
d0 s0
Hamiltonian, which a↵ects the decay of the particles. By convention mixing is presented
for ’down’ quarks, but in fact that choice is arbitrary: We could also consider the mixing
matrix to mix the u and c quarks.
Of course, the story of quarks did not stop with the discovery of the charm quark. In
1964 Cronin and Fitch had shown in experiments that CP symmetry is violated in
neutral kaon decays. Kobayashi and Maskawa found a solution in 1973: They extended
Cabibbo’s picture of quark mixing with a third family of quarks,
0 1 0 10 1
d0 Vud Vus Vub d
@ s0 A = @ Vcd Vcs Vcb A @ s A (7.45)
b0 Vtd Vts Vtb b
| {z }
CKM matrix
The three-generation mixing matrix is nowadays called the “Cabibbo-Kobayashi-Maskawa”

matrix, or simply the CKM matrix. The couplings in the Feynman diagram for charged
current interactions all get modified by elements of the mixing matrix:
u Vud d u Vus s u Vub b

W W W
c Vcd d c Vcs s c Vcb b

W W W
t Vtd d t Vts s t Vtb b

W W W
The mixing matrix VCKM is a 3 ⇥ 3 unitary matrix. This matrix is not uniquely defined
since the phases of the quark field can be chosen arbitrarily. If the phases are ’‘absorbed’
in the quark fields, the matrix can be parametrized by four real parameters, which are
usually chosen to be three mixing angles between the quark generations ✓12 , ✓13 , ✓23 ,
and one complex phase ,
0 1
c12 c13 s12 s13 s13 e i
VCKM = @ s12 c23 c12 s23 s13 ei c12 c23 s12 s23 s13 ei s23 c13 A (7.46)
i i
s12 s23 c12 c23 s13 e c12 s23 s12 c23 s13 e c23 c13
where sij = sin ✓ij and cij = cos ✓ij .
Kobayashi and Maskawa realized that the fact elements of VCKM can have a non-trivial
complex phase — i.e. a phase that can not be removed by redefining the phase of the
quark fields — leads to CP violation in charged current decays. For CP violation to
occur this way, at least three generations of quarks are required. The bottom and top
quark were eventually discovered in 1977 and 1994, respectively.
In case neutrino particles have a non-zero mass, mixing occurs in the lepton sector as
well. Just like the down-type quarks were chosen to describe mixing in the quark sector,
the neutrinos are chosen for the lepton sector:
0 1 0 10 1
⌫e U11 U12 U13 ⌫1
@ ⌫µ A = @ U21 U22 U23 A @ ⌫2 A (7.47)
⌫⌧ U31 U32 U33 ⌫3
| {z }
PMNS-matrix
The lepton mixing matrix is called the Pontecorvo-Maki-Nakagawa-Sakata matrix. This

matrix relates the mass eigenstates of the neutrinos (⌫1 , ⌫2 , ⌫3 ) to the weak interaction
eigenstates (⌫e , ⌫µ , ⌫⌧ ). There is an interesting open question whether neutrino’s are
their own anti-particles (”Majorana” neutrino’s) or not (”Dirac” neutrino’s). In case
neutrinos are of the Dirac type, the UP M N S matrix has one complex phase, similar to the
quark mixing matrix. If neutrinos are Majorana particles, the UP M N S matrix includes
three complex phases.
Violation of CP in fundamental interactions is needed to understand why we live in a
‘matter-dominated’ universe — rather than universe with equal amounts of matter and
anti-matter. It is still unknown if the matter-dominance is a result of CP violation in

the quark sector (”baryogenesis”) or in the lepton sector (”leptogenesis”).
Exercises
Exercise 7.1 (Helicity versus chirality. See also H&M exercise 5.15)
5
(a) Write out the chirality operator in the Dirac-Pauli representation.
(b) The helicity operator is defined as = 12 ⌃ · p̂, where p̂ is a unit vector along the
momentum and ⌃ is ✓ ◆
0
⌃= .
0
Show that in the ultra-relativistic limit (E m) the helicity operator and the
chirality operator have the same e↵ect on a spinor solution, i.e.
✓ (s)
◆ ✓ (s)
◆
5 5
= ·p (s) ⇡ 2 ·p (s) =2
E+m E+m
(c) Explain why the weak interaction is called left-handed.
Exercise 7.2 (Pion Decay)

Usually at this point the student is asked to calculate pion decay, which requires again
quite some calculations. The ambitious student is encourage to try and do it (using some
help from the literature). However, the exercise below requires little or no calculation
but instead insight in the formalism.
(a) Draw the Feynman diagram for the decay of a pion to a muon and an anti-neutrino:
⇡ ! µ ⌫ µ.
(b) Compute the momentum of the muon in the rest frame of the pion. You may
ignore the neutrino mass, but not the muon mass.
Due to the fact that the quarks in the pion are not free particles we cannot just apply
the Dirac formalism for free particle waves. However, we know that the interaction is
transmitted by a W and therefore the coupling must be of the type: V or A. (Also,
the matrix element must be a Lorentz scalar.) It turns out the decay amplitude has the
form:
GF
M = p (q µ f⇡ ) u(p) µ 1 5
v(k)
2
where pµ and k µ are the 4-momenta of the muon and the neutrino respectively, and q is
the 4-momentum carried by the W boson. f⇡ is called the decay constant.
(c) Can the pion also decay to an electron and an electron-neutrino? Write down the
Matrix element for this decay.
Would you expect the decay width of the decay to electrons to be larger, smaller,
or similar to the decay width to the muon and muon-neutrino?
Base your argument on the available phase space in each of the two cases.
The decay width to a muon and muon-neutrino is found to be:

✓ ◆2
G2 m2⇡ m2µ
= F f⇡2 m⇡ m2µ
8⇡ m2⇡
The measured lifetime of the pion is ⌧⇡ = 2.6 · 10 8 s which means that f⇡ ⇡ m⇡ . An

interesting observation is to compare the decay width to the muon and to the electron:
✓ ◆2 ✓ ◆2
(⇡ ! e ⌫ e ) me m2⇡ m2e 4
= ⇡ 1.2 · 10 !!
(⇡ ! µ ⌫ µ ) mµ m2⇡ m2µ
(d) Can you give a reason why the decay rate into an electron and an electron-neutrino
is strongly suppressed in comparison to the decay to a muon and a muon-neutrino.
Consider the spin of the pion, the handedness of the W coupling and the helicity
of the leptons involved.
Exercise 7.3 (Kinematics of muon decay)

We consider the decay in Fig. 7.1.
(a) Starting from four-momentum conservation (p = p0 + k 0 + k) and by ignoring the
electron and neutrino masses, show that
2k · p0 = m2 + 2p · k 0
where m is the muon mass.

Hint: write the equation as p k 0 = p0 + k, then square both sides.
(b) Use this result to show that in the muon rest frame
2(k · p0 )(k 0 · p) = (m2 2m! 0 ) m! 0 (7.48)

Lecture 8
Local Gauge Invariance
In the next two lectures we discuss the theory of the electroweak interaction, the so-
called “Glashow-Salam-Weinberg model”. This theory can be formulated starting from
the principle of local gauge invariance.
8.1 Symmetries
Symmetries play a fundamental role in particle physics. There is a theorem stating
that a symmetry is always related to a quantity that is fundamentally unobservable. In
general one can distinguish1 four types of symmetries:
• permutation symmetries: These lead to Bose-Einstein statistics for particles with
integer spin (bosons) and to Fermi-Dirac statistics for particles with half integer
spin (fermions). The unobservable is the absolute identity of a particle;
• continuous space-time symmetries: translation, rotation, acceleration, etc. The re-
lated unobservables are respectively: absolute position in space, absolute direction
and the equivalence between gravity and acceleration;
• discrete symmetries: space inversion, time inversion, charge inversion. The unob-
servables are absolute left/right handedness, the direction of time and an absolute
definition of the sign of charge;
• unitary symmetries or internal symmetries, also called ‘gauge invariance’: These
are the symmetries discussed in this lecture. As an example of an unobservable
quantity think of the phase of a complex wave function in quantum mechanics.
The relation between symmetries and conservation laws is expressed in a fundamental
theorem by Emmy Noether: each continuous symmetry transformation under which the
Lagrangian is invariant in form leads to a conservation law. Invariances under external
1
T.D. Lee: “Particle Physics and Introduction to Field Theory”
127
128 LECTURE 8. LOCAL GAUGE INVARIANCE
operations as time and space translation lead to conservation of energy and momentum,
and invariance under rotation to conservation of angular momentum. Invariances under
internal operations, like the shift of the complex phase of wave functions, lead to con-
served currents, or more specific, conservation of charge. We discuss the application of
Noether’s theorem to phase transformations in section 8.4.
In our current understanding elementary interactions of the quarks and leptons (elec-
tromagnetic, weak and strong) are all the result of gauge symmetries. Starting from a
Lagrangian that describes free quarks and leptons, the interactions can be constructed
by requiring the Lagrangian to be symmetric under particular transformations. The
idea of local gauge invariance will be discussed in this lecture and will be applied in the
unified electroweak theory in the next lecture.
8.2 The principle of least action
In classical mechanics the equations of motion can be derived using the variational prin-
ciple of Hamilton. This principle states that the action integral S should be stationary
under arbitrary variations of the so-called generalized coordinates qi . For a pedagogical
discussion of the principle of least action read the Feynman lectures, Vol.2, chapter 19.
Generalized coordinates are coordinates that correspond to the actual degrees of freedom
of a system. As an example, consider a swinging pendulum in two dimensions. We could
describe the movement of the weight of the pendulum in terms of both its horizontal
coordinate x and its vertical coordinate y. However, only one of those is independent
since the length of the pendulum is fixed. Therefore, we say that the movement of the
pendulum can be described by one ‘generalized’ coordinate. We could choose x or y,
but also the angle of the pendulum with the vertical axis (usually called the amplitude).
We denote generalized coordinates with the symbol q and call the evolution of q with
time a trajectory or path.
The Lagrangian of the system can be defined as the kinetic energy minus the potential
energy,
L(q, q̇, t) = T (q̇) V (q) , (8.1)
where the potential energy only depends on q (and eventually t) and the kinetic energy
only on the generalized velocity q̇ = dq/dt. We denote the action (or ‘action integral’)
of a path that starts at t1 and ends at t2 with
Z t1
S(q) = L(q, q̇, t) dt . (8.2)
t0
Hamilton’s principle now states that the actual trajectory q(t) followed by the system is
the trajectory q(t) that minimizes the action. (It is said that the action is ’stationary’
around this trajectory.) This is equivalent to requiring that for a given point q, q̇ on this
8.3. LAGRANGIAN DENSITY FOR FIELDS 129
trajectory, the change in the action following from a small deviation q, q̇ is zero:
S(q, q, q̇, q̇) = 0 . (8.3)
You will show in exercise 8.1 that for each of the coordinates qi , this leads to the so-called
Euler-Lagrange equation of motion
d @L @L
= . (8.4)
dt @ q˙i @qi
This may be written more symmetrically as
@L @L
ṗi = with pi = , (8.5)
@qi @ q˙i
where pi is called the generalized momentum, or the momentum canonical to qi . In

terms of these coordinates, the Hamiltonian takes the form
X
H(p, q, t) = pi q̇i L. (8.6)
i
and the equation of motion can also be written in the form of Hamilton’s equations,
@H @H
ṗi = and q˙i = . (8.7)
@qi @pi
Finally, the classical system can be quantized by imposing the fundamental postulate
of quantum mechanics,
[qi , pj ] = i~ ij . (8.8)
8.3 Lagrangian density for fields

The classical theory does not treat space and time symmetrically as the Lagrangian
might depend on the parameter t. This causes a problem if we want to make a Lorentz
covariant theory. The solution is to go to field theory: Rather than a finite set of degrees
of freedom we consider an infinite set of degrees of freedom, represented by the values of
a field , that is a function of the space-time coordinates xµ . The Lagrangian is replaced
by a Lagrangian density L (usually just called Lagrangian),
L(q, q̇, t) ! L( , @ µ , xµ ) (8.9)
such that the action becomes

Z x2
S = d4 x L (x), @ µ (x), xµ (8.10)
x1
Following the principle of least action we obtain the Euler-Lagrange equation for the
fields:
@L @L
= @µ (8.11)
@ (x) @ (@µ (x))
(If at this point you are confused about the position of Lorentz indices on the right-
hand-side, then remember that what is meant is
@L @ @L @ @L @ @L @ @L
= + + + (8.12)
@ (x) @x0 @(@ /@x0 ) @x1 @(@ /@x1 ) @x2 @(@ /@x2 ) @x3 @(@ /@x3 )
You could also use upper indices as long as you are consistent.)
To create a Lorentz covariant theory, we require the Lagrangian to be a Lorentz scalar.
(This also means that in the expression for L above the ‘loose’ Lorentz indices must
somehow be contracted with others.) This requirement imposes certain conditions on
the Lorentz transformation properties of the fields. (We have not discussed these in
detail. See textbooks.) Furthermore, although we consider complex fields, we always
require the Lagrangian to be real.
In quantum field theory, the coordinates become operators that obey the standard
quantum mechanical commutation relation with their associated generalized momenta.
The wave functions that we have considered before can be viewed as single particle
excitations that occur when the creation and annihilation operators of the field act on
the vacuum. For the discussions here we do not need field theory. What is important to
know is that field theory tells us that, given a Lagrangian, we can find a set of Feynman
rules that can be used to draw diagrams and compute amplitudes.
Now consider the following Lagrangian for a complex scalar field:
⇤ ⇤
L = (@µ )(@ µ ) m2 (8.13)
You will show in an exercise that the equation of motion corresponding to this La-
grangian is the Klein-Gordon equation. Because the field is complex, it has two separate
components. We could choose these to be the real and imaginary part of the field, such
that = 1 + i 2 with 1,2 real. It is easy to see what the Lagrangian looks like and
what the equations of motion become. However, rather than choosing 1 and 2 we can
also choose and ⇤ to represent the ‘independent’ components of the field.
A similar argument can be made for the bi-spinor and the adjoint bi-spinor in the
Lagrangian of the Dirac field. The latter is given by
µ
L = (i @µ m) (8.14)
and its equation of motion (treating and as independent components of the field)
is the Dirac equation.
8.4. GLOBAL PHASE INVARIANCE AND NOETHER’S THEOREM 131
8.4 Global phase invariance and Noether’s theorem

Lagrangians for complex fields are constructed such that the Lagrangian is real (and the
Hamiltonian hermitian). As a result, the Lagrangian is not sensitive to a global shift in
the complex phase of the field. Such a global phase change is called a U (1) symmetry.
(The group U (1) is the group of unitary matrices of dimension 1, e.g. complex numbers
with unit modulus.) Noether’s theorem tells us that there must be a conserved quantity
associated with such a phase invariance. For ‘internal’ symmetries of the Lagrangian
this works as follows.
Consider a transformation of the field components with a small variation ✏,
i ! i + ✏i (x) (8.15)
The resulting change in the Lagrangian is

!
X ✓ @L @L
◆ X @L
L = ✏i + @µ ✏i = @µ ✏i (8.16)
i
@ i @(@µ i ) i
@(@µ i )
where in the second step we have used the Euler-Lagrange equation to remove @L/@ .
Consequently, if the Lagrangian is insensitive to the transformation, then the quantity
X @L
jµ = ✏i (8.17)
i
@(@µ i )
is a conserved current.
Let’s now apply this to the complex scalar field for a (small) U (1) phase translation.
The two independent field components are and ⇤ . Under the phase translation these
change as
! ei↵ ⇡ (1 + i↵)
⇤ ⇤ i↵ ⇤
(8.18)
! e ⇡ (1 i↵)
Consequently, we have ✏ = i↵ and ✏ ⇤ = i↵ ⇤ . Inserting these into the expression
for the Noether current, Eq. (8.17), we find
⇤ ⇤
jµ = ↵ i (@ µ ) (@ µ ) (8.19)
Since ↵ is an arbitrary constant, we omit it from the current. We have obtained exactly
the current that we constructed for the Klein-Gordon wave in Lecture 1.
8.5 Local phase invariance

We can also look at the U (1) symmetry from a slightly more general perspective. The
expectation value of a quantum mechanical observable (such as the Hamiltonian) is
typically of the form: Z

⇤
hOi = O (8.20)
If we make the replacement (x) ! ei↵ (x) the expectation value of the observable
remains the same. We say that we cannot measure the absolute phase of the wave
function. (We can only measure relative phases between wave functions in interference
experiments.)
However, this only holds for a phase that is constant in space and time. Are we allowed
to choose a di↵erent phase convention on, say, the moon and on earth, for a wave
function (x)? In other words, can we choose a phase that depends on space-time,
0
(x) ! (x) = ei↵(x) (x)? (8.21)
In general, we cannot do this without breaking the symmetry. The problem is that the
Lagrangian density L ( (x), @µ (x)) depends on both the fields (x) and the derivatives
@µ (x). The derivative term yields:
@µ (x) ! @µ 0 (x) = ei↵(x) @µ (x) + i (@µ ↵(x)) (x) (8.22)
and therefore the Lagrangian is not invariant.

However, suppose that we now introduce ‘local U (1) symmetry’ as a requirement. Is
it possible to modify the Lagrangian such that it obeys this symmetry? The answer
is ‘yes’, provided that we introduce a new field, the so-called gauge field. The recipe
consists of two steps.
First, we replace the derivative @µ by the so-called gauge-covariant derivative:
@µ ! Dµ ⌘ @µ + iqAµ (x) , (8.23)
where Aµ is a new field and q is (for now) an arbitrary constant. Second, we require
that the field Aµ transforms as
1
Aµ (x) ! A0µ (x) = Aµ (x) @µ ↵(x) . (8.24)
q
By inserting the expression for A in the covariant derivative, we find that it just trans-
forms with the local phase ↵(x):
✓ ◆
0 0 i↵(x) 1
Dµ (x) ! Dµ (x) = e @µ (x) + i@µ ↵(x) (x) + iqAµ (x) (x) iq @µ ↵(x) (x)
q
= ei↵(x) Dµ (x) (8.25)
As a consequence, terms in the derivative that look like ⇤ Dµ are phase invariant.
With the substitution @µ ! Dµ the Klein-Gordon and Dirac Lagrangians (and any
other real Lagrangian that we can construct with 2nd order terms from a complex field
and its derivatives) satisfy the local phase symmetry.
8.6. APPLICATION TO THE DIRAC LAGRANGIAN 133
8.6 Application to the Dirac Lagrangian

Now consider the e↵ect of the substitution of the derivative with the covariant derivative
in Eq. (8.23) in the Lagrangian for the Dirac field (Eq. (8.14)),
µ
L= (i Dµ m)
= (i µ @µ m) qAµ µ (8.26)
⌘ Lfree + Lint ,
where we defined the interaction term
Lint = J µ Aµ (8.27)
with the familiar Dirac current

Jµ = q µ
. (8.28)
This is exactly the form of the electromagnetic interaction that we discussed in the
previous lectures. Furthermore, since we require the Lagrangian to be real and since
the conserved current is real, the field Aµ must be a real as well. We can now identify
q with the charge and the gauge field Aµ with the electromagnetic vector potential, i.e.
the photon field. The transformation of the field in Eq. (8.24) is just the gauge freedom
that we identified in the electromagnetic field in Lecture 3 (with = q↵).
The picture is not entirely complete yet, though. We know that the photon field satisfies
its own ‘free’ Lagrangian. This is the Lagrangian that leads to the Maxwell equations
in vacuum. It is given by
1
Lfree
A = Fµ⌫ F µ⌫ (8.29)
4
with F µ⌫ = @ ⌫ Aµ @ µ A⌫ . We call this the ‘kinetic’ term of the gauge field Lagrangian
and we simply add it to the total Lagrangian. The full Lagrangian for a theory that has
one Dirac field and obeys local U (1) symmetry is then given by
µ µ 1
LQED = (i @µ m) qAµ Fµ⌫ F µ⌫ (8.30)
4
This is called the QED Lagrangian.
At this point you may wonder if we could also add a mass term for the photon field. If
the photon would have a mass, the corresponding term in the Lagrangian would be
1 2 µ
L = m A Aµ . (8.31)
2
However, this term violates local gauge invariance, since:
⇣ ⌘⇣ ⌘
Aµ Aµ ! Aµ 1q @ µ ↵ Aµ 1q @µ ↵ 6= Aµ Aµ (8.32)
Therefore, the requirement of local U (1) invariance automatically implies that the pho-
ton is massless. This actually holds for other gauge symmetries as well. In chapters 11
to 14 we discuss how masses of vector bosons can be generated in the Higgs mechanism
by ‘breaking’ the symmetry.
You may wonder why we put so much emphasis on the principle of local gauge invariance.
After all, it looks like all we have done is find a di↵erent way of arriving at the equations
of motions of electrodynamics: is it really so attractive to formulate QED as a symmetry?
The reason that local gauge symmetries are so important is because of what is called
‘renormalizability’. By way of the Feynman rules, the Lagrangian encodes the infor-
mation to compute scattering and decay processes to arbitrary order. However, if you
compute anything beyond leading order you will quickly find that the result is not finite.
This can be solved by a number of di↵erent techniques, called collectively ‘renormaliza-
tion’. It was shown by ’t Hooft and Veldman in the early seventies that only Lagrangians
with interaction terms generated by local gauge symmetries are renormalizable. In other
words, if we want to have a theory in which we can compute something, then we cannot
have any other interactions than those derived from internal symmetries.
8.7 Yang-Mills theory

The U (1) symmetry discussed above is the simplest local gauge symmetry. To extend
the gauge principle to the weak and strong interaction, we need to consider more compli-
cated symmetries. We introduce these so-called ‘non-abelian’ symmetries in a somewhat
historical fashion as this helps to understand the origin of the term ‘weak isospin’ and
the relation to (strong-) isospin.
In the 1950s Yang and Mills tried to describe the strong interaction in the proton-
neutron system in terms of a gauge symmetry. Ignoring the electric charge, the free
Lagrangian for the nucleons can be written as
µ µ
L = p (i @µ m) p + n (i @µ m) n . (8.33)
In terms of the bi-spinor doublet ✓ ◆

p
= , (8.34)
n
the Lagrangian becomes
✓ µ
◆
i @µ m 0
L = µ (8.35)
0 i @µ m
Have a careful look at what is written here: The doublet is a 2-component column
vector with a Dirac spinor for each component. Each of the entries in the matrix in the
Lagrangian is again a 4x4 matrix.
Note that we have taken the two components to have identical mass m. Because they
have identical mass and no charge the nucleons are indistinguishable. Therefore, we
8.7. YANG-MILLS THEORY 135
consider a global transformation of the field with a complex unitary (2 ⇥ 2) matrix U

that ‘rotates’ the proton-neutron system,
!U and ! U†
Since U † U = 1, our Lagrangian is invariant to this transformation.
Any complex unitary (2 ⇥ 2) matrix can be written in the form
✓ ◆
i✓ i
U = e exp ↵·⌧ , (8.36)
2
where ↵ and ✓ are real and ⌧ = (⌧1 , ⌧2 , ⌧3 ) are the Pauli spin matrices 2 . We have already
considered the e↵ects of a phase transformation ei✓ , which was the U (1) symmetry.
Therefore, we concentrate on the case where ✓ = 0. Since the matrices ⌧ all have zero
trace, the matrices U with this property all have determinant 1. They form the group
SU (2) and the matrices ⌧ are the generators of this group. (In group theory language
we say that SU (2) is an irriducible subgroup of U (2) and U (2) = U (1) ⌦ SU (2).)
Note that members of SU (2) do in general not commute. This holds in particular for the
generators. We call such groups “non-abelian”. In contrast, the U (1) group is abelian
since complex numbers just commute.
Using the same prescription as for the U (1) symmetry, we can derive a conserved current.
Consider a small SU (2) transformation in doublet space
✓ ◆
0 i
! = 11 + ↵ · ⌧ (8.37)
2
and similar for . The Lagrangian transforms as
@L @L @L @L
L = + (@µ ) + + @µ
@ @ (@µ ) @ @ @µ
 
@L @L
= @µ + @µ (8.38)
@(@µ ) @(@µ )
where in the second line we have used the Euler-Lagrange equation to eliminate @L/@ ,
just as we did in Eq. (8.16). Computing the derivatives of the Lagrangian, we find that
the right term vanishes, while the left term gives

1 µ
L = @µ ↵·⌧ (8.39)
2
Since ↵ is a constant and since the requirement of phase invariance must hold for any ↵,
we can drop ↵ and obtain three continuity equations @µ J µ = 0 for the three conserved
currents
µ ⌧
Jµ = . (8.40)
2
2
We had labeled these by , but for some obscure
✓ reason
◆ the✓textbooks
◆ also switch
✓ from ◆ to ⌧ at
0 1 0 i 1 0
this point. Our default representation is: ⌧1 = , ⌧2 = , ⌧3 = ,
1 0 i 0 0 1
As for the U (1) symmetry, we now try to promote the global symmetry to a local
symmetry. The strategy is similar to that for U (1), but because the group is non-abelian,
the implementation is more complicated. The first step is to make the parameters ↵
depend on space time. To simplify the notation we define the gauge transformation as
follows,
0
(x) ! (x) = G(x) (x)
✓ ◆
i (8.41)
with G(x) = exp ⌧ · ↵(x)
2
We have again, as in the case of QED, that the derivative transforms non-trivially
@µ (x) ! G (@µ ) + (@µ G) (8.42)
such that the Lagrangian is not phase invariant. To restore phase invariance, we intro-
duce the 2 ⇥ 2 covariant derivative
Dµ = 11@µ + igBµ (8.43)
where g is a (so far arbitrary) coupling constant and Bµ a gauge field. In spinor space
the latter is a 2 ⇥ 2 unitary matrix with determinant 1. It is customary to parametrize
it in terms of three new real vector fields b1 , b2 and b3 ,
✓ ◆
1 1X k k 1 b3 b1 ib2
Bµ = ⌧ · bµ = ⌧ bµ = . (8.44)
2 2 k
2 b1 + ib2 b3
We call the fields bi the gauge fields of the SU (2) symmetry. We need three fields rather
than one, because SU (2) has three generators.
In terms of the covariant derivative the Lagrangian is
µ
L= (i Dµ 11m) (8.45)
In order to obtain a Lagrangian that is invariant, we need the gauge transformation to

take the form
Dµ ! Dµ0 0 = G (Dµ ) (8.46)
Inserting the expression for the covariant derivative, we find for the term on the left side
of the equality
Dµ0 0
= @µ + igBµ0 0
= G (@µ ) + (@µ G) + igBµ0 G (8.47)
while the right side is

G (Dµ ) = G@µ + igGBµ (8.48)
8.7. YANG-MILLS THEORY 137
Comparing these two expressions we find that invariance implies that
igBµ0 (G ) = igG (Bµ ) (@µ G) (8.49)
Since this expression must hold for all values of the field , we can omit the field from
this expression. If we subsequently multiply both sides of the equation on the right by
G 1 we find for the transformation of the gauge field
i
Bµ0 = GBµ G 1
+ (@µ G) G 1
. (8.50)
g
Although this looks rather complicated we can again try to interpret this by comparing
to the case of electromagnetism. For Gem = ei↵(x) we have
i
A0µ = Gem Aµ Gem1 + (@µ Gem ) Gem1
q
1
= Aµ @µ ↵ (8.51)
q
which is exactly the transformation rule that we had before.
We see that for an SU (2) symmetry the transformation of the gauge field Bµ involves
both a rotation and a gradient. The gradient term was already present in QED. The
rotation term is new. It arises due to the non-commutativity of the elements of SU (2).
If we write out the gauge field transformation formula in the components of the real
vector fields
0 1
bkµ = bkµ ✏klm ↵l bm @µ ↵k (8.52)
g
we can see that there is a coupling between the di↵erent components of the field. We call
this the self-coupling. (To derive this start from Eq. 8.50, consider infinitesimal small
↵(x) and use the commutation relation of the SU(2) generators, [⌧i , ⌧j ] = 2✏ijk ⌧k .)
The e↵ect of the self-coupling becomes clear if one considers the kinetic term of the
SU (2) gauge field. Analogous to the QED case, the three new fields require their own
free Lagrangian, which we write as
1 X µ⌫ 1 µ⌫
Lfree
b = F Fµ⌫,l = F · Fµ⌫ . (8.53)
4 l l 4
Mass terms like m2 b⌫ b⌫ are again excluded by gauge invariance: as for the U (1) symme-
try, the gauge fields must be massless. However, while for the photon the field tensor in
the kinetic term was given by F µ⌫ = @ µ A⌫ @ ⌫ Aµ , this form does not work here because
it would break the symmetry. Rather, the individual components of the field tensor are
given by
Flµ⌫ = @ ⌫ bµl @ µ b⌫l + g ✏jkl bµj b⌫k (8.54)
or in vector notation
F µ⌫ = @ µ b⌫ @ ⌫ bµ g bµ ⇥ b⌫ (8.55)
As a consequence of the last term the Lagrangian contains contributions with 2, 3 and
4 factors of the b-field. These couplings are respectively referred to as bilinear, trilinear
and quadrilinear couplings. In QED there is only the bilinear photon propagator term.
In the SU (2) theory there are self interactions by a 3-gauge boson vertex and a 4 gauge
boson vertex.
✓ ◆
p
Summarizing, we started from the free Lagrangian for a doublet = of two
n
fields with equal mass,
Lf ree = (i µ @µ m)
This Lagrangian has a global SU (2) symmetry. We then hypothesized a local SU (2)
phase invariance which we could implement by making the replacement @µ ! Dµ =
@µ + igBµ with Bµ = 12 ⌧ · bµ . The full Lagrangian of the theory (which is called the
Yang-Mills theory) is then given by
µ 1 µ⌫
LSU (2) = (i Dµ m) F · Fµ⌫
4
1 µ⌫ (8.56)
= (i µ @µ m) gJ µ bµ F · Fµ⌫
4
⌘ Lfree + Linteraction + Lfreeb
where we now absorbed the coupling constant g in the definition of the conserved current,
g
Jµ = µ
⌧ (8.57)
2
Comparing this to the QED Lagrangian
1 µ⌫
LU (1) = Lf ree Aµ · J µ F Fµ⌫ (8.58)
4
(with the electromagnetic current J µ = q µ ), we see that instead of one field, we now
have three new fields. Furthermore, the kinetic term is more complicated and gives rise
to self-coupling vertices with three and four b-field lines.
8.8 Historical interlude 1: isospin, QCD and weak

isospin
Yang and Mills tried to extend the isospin symmetry in the proton-neutron system to a
local isospin symmetry in the hope of formulating a viable theory of strong interactions.
In order to do so, they needed three new gauge fields, and obviously they wondered
what those were. They had to be massless vector bosons that couple to the proton and
neutron. Clearly, it could not be the three pions, since those are pseudo-scalar particles
rather then vector bosons.
8.9. HISTORICAL INTERLUDE 2: THE ORIGIN OF THE NAME “GAUGE THEORY”139
As we know now, the SU (2) theory cannot describe the strong interaction. Rather the
strong interactions follow from an SU (3) symmetry. The implementation is a carbon
copy of the Yang-Mills theory for SU (2) symmetry. The mediators of the force are the
eight massless gluons, corresponding to the 8 generators of the fundamental representa-
tion of SU (3), namely
0 1 0 1 0 1
0 1 0 0 i 0 1 0 0
1 =@ 1 0 0 A 2 =
@ i 0 0 A 3 =
@ 0 1 0 A
0 0 0 0 0 0 0 0 0
0 1 0 1 0 1
0 0 1 0 0 i 0 0 0
4 =@ 0 0 0 A 5 =
@ 0 0 0 A 6 =
@ 0 0 1 A
1 0 0 i 0 0 0 1 0
0 1 0 1
0 0 0 1 0 0
1 @
7 =@ 0 0 i A 8 = p 0 1 0 A
0 i 0 3 0 0 2
In this case, too, the Lagrangian contains self-coupling terms for the gauge fields. The
strong interaction is discusses in the Particle Physics II course.
The isospin symmetry in the proton-neutron is a flavour symmetry. Extended to the
system of all other hadrons it is essentially just the symmetry between u and d quarks.
We know that such a symmetry only exists if we ignore electromagnetic interactions,
and the small di↵erence in mass between the u and the d quark (or the proton and the
neutron). Since the symmetry is not exact, we call it an approximate symmetry.
Although the Yang-Mills isospin theory is of no real use to the proton-neutron system, it
turns out to be exactly what is needed to describe the weak interactions. For historical
reasons the local SU (2) symmetry applied to the Lagrangian of Dirac fermion doublets,
discussed in the next lecture, is sometimes called ’weak isospin’. It should certainly not
be confused with the u d flavour symmetry. In contrast with the flavour symmetry,
gauge symmetries are exact symmetries of the Lagrangian.
8.9 Historical interlude 2: the origin of the name

“gauge theory”
The idea of gauge invariance as a dynamical principle is due to Hermann Weyl. He

called it “eichinvarianz” (“gauge” = “calibration”). Weyl tried to find a geometrical
basis for both gravitation and electromagnetism.3 Although his e↵ort was unsuccessful
the terminology survived. His idea is summarized here.
Consider a change in a function f (x) between point xµ and point xµ + xµ . If the space
3
H. Weyl, Z. Phys. 56, 330 (1929)
has a uniform scale we expect simply:
f (x + x) = f (x) + @ µ f (x) xµ (8.59)
But if in addition the scale, or the unit of measure, for f changes by a factor (1 + S µ xµ )
between x and x + x, then the value of f becomes:
f (x + x) = (f (x) + @ µ f (x) xµ ) (1 + S ⌫ x⌫ )
(8.60)
= f (x) + (@ µ f (x) + f (x)S µ ) xµ + O( x)2
So, to first order, the increment is:
f = (@ µ + S µ ) f xµ (8.61)
In other words Weyl introduced a modified di↵erential operator by the replacement:

@ µ ! @ µ + S µ.
One can see this in analogy in electrodynamics in the replacement of the momentum by
the canonical momentum parameter: pµ ! pµ qAµ in the Lagrangian, or in Quantum
Mechanics: @ µ ! @ µ + iqAµ , as was discussed in the earlier lectures. In this case the
“scale” is S µ = iqAµ . If we now require that the laws of physics are invariant under a
change:
(1 + S µ xµ ) ! (1 + iq Aµ xµ ) ⇡ exp (iq Aµ xµ ) (8.62)
then we see that the change of scale gets the form of a change of a phase. When Weyl
later studied the invariance under phase transformations, he kept using the term “gauge
invariance”.
Exercises
Exercise 8.1 (Classical Euler-Lagrange equations)

Use Hamilton’s principle in Eq. (8.3) to derive the Euler-Lagrange equation of motion
(Eq. (8.4)).
Hint: Write down the change in the action S for an arbitrary small change in the
trajectory ( q, q̇). Now use integration by parts to replace q̇ by q. The bound-
ary condition (the fact that we know where the trajectory starts and ends) makes that
q(t0 ) = q(t1 ) = 0. The Euler-Lagrange equations then follow from the fact that for the
trajectory that minimizes the action, the change in the action should be zero, independent
of q.
Exercise 8.2 (Lagrangians versus equations of motion)

8.9. HISTORICAL INTERLUDE 2: THE ORIGIN OF THE NAME “GAUGE THEORY”141
(a) Show that the Euler-Lagrange equations of the Lagrangian

1 1 2
Lfree
KG = (@µ ) (@ µ ) m 2
(8.63)
2 2
of a real scalar field leads to the Klein-Gordon equation.
(b) Show the same for a complex scaler field starting from the Lagrangian
⇤ ⇤
Lfree
KG = (@µ ) (@
µ
) m2 (8.64)
taking and ⇤ as the (two) independent fields. (Alternatively, you can take the
real and imaginary part of . Note that you obtain two equations of motion, one
for and one for ⇤ .)
(c) Show that the Euler-Lagrange equations for the Lagrangian
Lfree
Dirac = i µ@
µ
m (8.65)
leads to the Dirac equations for and for . Note again that you need to consider
and as independent fields.
(d) Show that the Lagrangian
1 µ ⌫ 1 µ⌫
L = LEM = (@ A @ ⌫ Aµ ) (@µ A⌫ @⌫ Aµ ) j µ Aµ = F Fµ⌫ j µ Aµ
4 4
(8.66)
leads to the Maxwell equations:
@µ (@ µ A⌫ @ ⌫ Aµ ) = j ⌫ (8.67)
Hence the current is conserved (@⌫ j ⌫ = 0), since F µ⌫ is antisymmetric.
Exercise 8.3 (Global phase invariance)

(a) Show that the Dirac Lagrangian remains invariant under the global phase trans-
formation
(x) ! eiq↵ (x) ; (x) ! e iq↵

(x) (8.68)
(b) Noether’s Theorem: consider an infinitesimal transformation: ! 0 = ei↵ ⇡

(1 + i↵) . Show that the requirement of invariance of the Dirac Lagrangian
( L( , @µ , , @µ ) = 0) leads to the conserved current
!
@L L
jµ = i = µ
(8.69)
@ (@µ ) @ @µ
Consequently, the charge carried by the current is conserved.

Exercise 8.4 (Local phase invariance)

(a) (i) Start with the Lagrange density for a complex Klein-Gordon field
L = (@µ )⇤ (@ µ ) m2 ⇤
(8.70)
and show that a local field transformation:

⇤ iq↵(x) ⇤
(x) ! eiq↵(x) (x) ; (x) ! e (x) (8.71)
does not leave the Lagrangian invariant.

(ii) Substitute the covariant derivative @µ ! Dµ = @µ + iqAµ in the Lagragian.
Show that the Lagrangian now remains invariant under the gauge transfor-
mation, provided that the additional field transforms as
Aµ (x) ! A0µ (x) = Aµ (x) @µ ↵(x) . (8.72)
(b) (i) Start with the Lagrange density for a Dirac field
µ
L=i @µ m (8.73)
and show that a local field transformation:
(x) ! eiq↵(x) (x) ; (x) ! e iq↵(x)

(x) (8.74)
also does not leave the Lagrangian invariant.

(ii) Again make the replacement @µ ! Dµ = @µ + iqAµ and show that the
Lagrangian remains invariant provided that Aµ transforms are above.
Exercise 8.5 (Extra exercise, not obligatory!)

Consider an infinitesimal gauge transformation:
i
G=1+ ⌧ ·↵ |↵| ⌧ 1 (8.75)
2
Use the general transformation rule for Bµ0 and use Bµ = 12 ⌧ · bµ to demonstrate that
the fields transform as:
1
b0µ = bµ ↵ ⇥ bµ @µ ↵ (8.76)
g
(Hint: use the Pauli vector identity (⌧ · a)(⌧ · b) = a · b + i⌧ · (a ⇥ b).)
Lecture 9
Electroweak Theory
In the previous lecture we have seen how imposing a local gauge symmetry requires a
modification of the free Lagrangian in such a way that a theory with interactions is
obtained. We studied two symmetries, namely
• local U (1) gauge invariance:
µ µ µ
(i Dµ m) = (i @µ m) q Aµ (9.1)
| {z }
Jµ
• local SU (2) gauge invariance for a Dirac spinor douplet :

g
(i µ Dµ m) = (i µ @µ m) µ
⌧ bµ . (9.2)
2 | {z }
Jµ
For the U (1) symmetry we can identify the Aµ field as the photon. The Feynman rules
for QED, as we discussed them in previous lectures, follow automatically.
Yang and Mills implemented the SU (2) local gauge symmetry, hoping that they could
derive the strong interaction from proton-neutron isospin symmetry. Although that did
not work, we now show that the SU (2) gauge symmetry is still useful, but then to explain
the weak interaction. (The strong interaction follows from SU (3) gauge invariance.)
Before we continue with SU (2), we make a small modification to the interaction terms
above. First consider the U (1) symmetry. Every fermion field has its own charge.
Within the Standard Model we cannot explain why the charge of an up quark is two-
thirds of the charge of an electron. This is why the symbol q appears in the interaction
term above: it is a dimensionless parameter that signifies the strength of the interaction
and it can be di↵erent for di↵erent fields.
At this point it is customary to introduce a charge operator Q which acts as the generator
of the U (1) symmetry group for electromagnetic interactions. It appears in the field
transformation rule as
0
= ei↵(x)Q . (9.3)
143
144 LECTURE 9. ELECTROWEAK THEORY
We define the conserved current as

µ µ
JEM = Q (9.4)
and write the interaction term in the Hamiltonian as

µ
Lint = gEM JEM Aµ , (9.5)
where we now use the same coupling gEM for all fields. The fields are eigenstates of
the charge operator Q with an eigenvalue equal to the charge in units of the positron
charge. (Why we do this will become clear later.)
A similar strategy is taken for the isospin symmetry. Rather than ⌧ as the generator
we consider an operator T which for unit isospin charge is given by T = ⌧ /2. It enters
into the douplet transformation rule as
0
= ei↵(x)T (9.6)
and finds its way into the conserved current as
JTµ = µ
T , (9.7)
while the interaction terms is given by
Lint = g JTµ bµ . (9.8)
When, in the following sections, we consider SU (2) symmetry to generate the weak
interaction, the coupling constant g is taken to be the same for all douplets, but the
physical fields (which are eigenstates of T3 ) each have their own value of ‘weak-isospin
charge’, the eigenvalue for T3 . In the Standard Model, this eigenvalue is always ±1/2.
The coupling constants gEM and g in the interaction terms are dimensionless. For the
electromagnetic interaction the coupling is related to the unit charge as
e
gEM = p = 4⇡↵ . (9.9)
✏0 ~c
In our system of units ✏0 ~c = 1. Therefore, in the following we substitute gEM = e, as

you will find in most textbooks.
9.1 SU(2) symmetry for left-handed douplets

We define for any Dirac field the left- and right-handed chiral projections,
1 5 1 5
L ⌘ 1 and R ⌘ 1+ . (9.10)
2 2
9.1. SU(2) SYMMETRY FOR LEFT-HANDED DOUPLETS 145
As we have seen in Lecture 7, for particles with E m these correspond to the negative
and positive helicity states, respectively. Using the fact that (see exercise 9.1)
µ µ µ
= L L + R R (9.11)
where the chiral projections of the adjoint spinors are given by

1 5 1 5
L ⌘ ( L) = 1+ and R ⌘ ( R) = 1 (9.12)
2 2
we can rewrite the Dirac Lagrangian for as
µ µ
L = R (i @µ ) R + L (i @µ ) L R m L L m R (9.13)
The mass terms ‘mix’ the left- and right-handed components. That is incovenient for
what we are going to do next. Therefore, in the following we consider only massless
fields and deal with non-zero mass later.
Let us now introduce the following doublets for the left-handed chirality states of the
leptons and quarks in the first family:
✓ ◆ ✓ ◆
⌫L uL
L = and L = (9.14)
eL dL
We call these “weak isospin” doublets. Again, is not a Dirac spinor, but a doublet of
Dirac spinors. Consider the Lagrangian for the electron and neutrino and verify that it
can be written as (c.f. Eq. 8.35)
µ
L = eR i @µ eR + ⌫ R i µ @µ ⌫ R +
✓ µ ◆✓ ◆
i @µ 0 ⌫L (9.15)
+ (⌫ L eL )
0 i µ @µ eL
Now it comes: We impose the SU (2) gauge symmetry on the left-handed doublets only.
That is, we require that the Lagrangian be invariant for local rotations of the doublet.
To do this we need to ignore that the two components of a doublet have di↵erent charge,
a problems that we will clearly need to deal with later. As in the Yang-Mills theory, we
also need to ignore that they have di↵erent mass, which is another motivation for only
considering massless fields.
The fact that we only impose the gauge symmetry on left-handed states leads to a weak
interaction that is completely left-right asymmetric. This is why it is referred to as
maximal violation of parity.
To construct the weak SU (2)L theory1 we start again with the free Dirac Lagrangian
and we impose SU (2) symmetry on the weak isospin doublets:
µ
Lf ree = L i @µ L (9.16)
1
The subscript L is used to indicate that we only consider SU (2) transformations of the left-handed
doublet.
After introducing the covariant derivative
@µ ! Dµ = @µ + igBµ with Bµ = T · bµ (9.17)
the Dirac equation obtains an interaction term,

µ
Lfree ! Lfree g bµ · Jweak (9.18)
where the weak current is

µ µ
Jweak = L T L (9.19)
This is just a carbon copy of the Yang-Mills theory for “strong isospin” in the previous
lecture.2
The gauge fields (b1 , b2 , b3 ) couple to the left-handed doublets defined above. However,
the particles in our real world do not appear as doublets: we scatter electrons, not
electron-neutrino doublets. We now show, as Glashow first did in 1961, how these
gauge fields can be recast into the ‘physical’ fields of the 3 vector bosons W + , W , Z 0
in order to have them interact with currents of the physical electrons and neutrinos.
9.2 The Charged Current

We choose for the representation of the SU (2) generators the Pauli spin matrices,
✓ ◆ ✓ ◆ ✓ ◆
0 1 0 i 1 0
⌧1 = ⌧2 = ⌧3 = . (9.20)
1 0 i 0 0 1
The generators ⌧1 and ⌧2 mix the components of a doublet, while ⌧3 does not. We define
the fields W ± as
1
Wµ± ⌘ p b1µ ⌥ i b2µ (9.21)
2
The ± index on the W refers to the electric charge. However, at this point we have not
yet shown that these fields are indeed electrically charged: That would require us to
look at the coupling of the W fields to the photon, which we will not do as part of these
lectures. As an alternative, we now show that these W fields couple to charge-lowering
and charge-raising currents. Charge conservation at each Feynman diagram vertex then
implies the charge of the gauge boson.
We define the charged current term of the interaction Lagrangian as
LCC ⌘ g b1µ J 1µ g b2µ J 2µ (9.22)
with
⌧1 ⌧2
J 1µ = L
µ
L J 2µ = L
µ
L (9.23)
2 2
2
Note that in terms of physics strong and weak isospin have nothing to do with one another. It is
just that we use the same math!
9.2. THE CHARGED CURRENT 147
As you will show in exercise 9.2 we can rewrite the charged current Lagrangian as
LCC = g Wµ+ J +µ g Wµ J µ
(9.24)
with
1 µ ±
J µ,± = p L ⌧ L (9.25)
2
and ⌧ ± = 12 (⌧1 ± i⌧2 ), or in our representation
✓ ◆ ✓ ◆
+ 0 1 0 0
⌧ = and ⌧ = . (9.26)
0 0 1 0
The leptonic currents can then be written as
1 1
J +µ = p ⌫ L µ
eL and J µ
= p eL µ
⌫L (9.27)
2 2
or written out with the left-handed projection operators:
1 1 1
J +µ = p ⌫ 1+ 5 µ
1 5
e (9.28)
2 2 2
µ
and similar for J . Verify for yourself that
5 µ 5 µ 5
1+ 1 = 2 1 (9.29)
such that we can rewrite the leptonic charge raising current as
1
J +µ = p ⌫ µ
1 5
e (9.30)
2 2
and the leptonic charge lowering current as
µ 1 µ 5
J = p e 1 ⌫ . (9.31)
2 2
Remembering that a vector interaction has an operator µ in the current and an axial
vector interaction a term µ 5 , we recognize in the charged weak interaction the famous
“V-A” interaction. The story for the quark doublet is identical. Drawn as diagrams,
the charged currents then look as follows:
⌫e u
Charge raising: W+ W+
e d
⌫e u
Charge lowering: W W
e d
9.3 The Neutral Current
The third component of the weak isospin gauge field leads to a neutral current interac-
tion,
Lint = g b3µ J3µ (9.32)
with b3µ the third gauge boson (another real vector field) and the conserved current given
by
⌧3
J3µ = L µ L . (9.33)
2
It is now tempting to identify this third component as the Z 0 boson and simply add
the electromagnetic interaction term that we had previously constructed with a U (1)
symmetry with the electromagnetic charge operator Q as generator.
However, this is not a valid way to extend the symmetry of the Lagrangian: the left-
handed doublets that we have constructed are not eigenfunctions of Q since they mix
fields with di↵erent charge. Therefore, our SU (2)L invariant Lagrangian cannot be
symmetric under a transformation with Q as generator.
The solution is to start from another U (1) gauge symmetry, called ‘weak hypercharge’.
We denote its generator with the symbol Y and require that it commutes with the
SUL (2) generators. The di↵erent members of the isospin multiplet then by construction
obtain the same value of hypercharge.
We denote the combined symmetry by SU (2)L ⌦ U (1)Y . Under this symmetry a left-
handed doublet transform as
0
⇥ ⇤
L ! L = exp i ↵(x) T + i (x) Y L , (9.34)
where T = ⌧ /2 are the SU (2) generators and Y is the generator for U (1)Y . At the
same time, the right-handed components of the fields in the doublet transform only
under hypercharge,
0 i (x)Y
R ! R = e R . (9.35)
The conserved current corresponding to the U (1)Y symmetry is
JYµ = µ
Y . (9.36)
The Lagrangian following from local SU (2)L ⌦ U (1)Y symmetry takes the form (see e.g.
Halzen and Martin, Chapter 13)
g0 µ
LEW = Lf ree g JTµ · bµ J aµ , (9.37)
2 Y
where aµ is the gauge field corresponding to U (1)Y and g 0 /2 is its coupling strength.
The factor 2 appears just because of a convention.
9.3. THE NEUTRAL CURRENT 149
The transformations corresponding to T3 and Y both lead to neutral current interactions.

As a result the gauge boson fields can actually ‘mix’. Neither of them couples specifically
to the electromagnetic charge. Therefore, an important question is whether we can
recast these fields such that one becomes a physical ‘photon’ field Aµ that couples to
the fermion fields via the charge operator Q and the other one becomes the Z 0 boson.
Weinburg and Salam showed indepently that the answer to this has everything to do
with the Higgs mechanism. The gauge fields bµ and aµ in the formalism above are all
massless. An explicit mass term (LM = Kbµ bµ ) would break the gauge invariance of the
theory. Their masses and the masses of all fermions can be generated in a mechanism
that is called spontaneous symmetry breaking and involves a new scalar field, the Higgs
field. The main idea of the Higgs mechanism is that the Lagrangian retains the full
gauge symmetry, but that the ground state (or ‘vacuum’), i.e. the state from which we
start perturbation theory, is not symmetric. We discuss this in detail in Lectures 11 and
Lectures 12. For now, you will have to do with some conjectures.
The Higgs mechanism leads to a solution in which three out of four vector fields (bµ
and aµ ) acquire a mass. Two of the massive fields can be written as charge raising and
charge lowering fields W ± as we did before,
1
Wµ± = p b1µ ⌥ i b2µ . (9.38)
2
The physical neutral fields are linear combinations of the T3 and Y gauge fields, written
as
Aµ = aµ cos ✓W + b3µ sin ✓W
(9.39)
Zµ = aµ sin ✓W + b3µ cos ✓W
where ✓W is called the weak mixing angle.

The Higgs mechanism predicts that in order that the massless field becomes the pho-
ton, the quantum number for the charge operator is related to the SU (2)L and U (1)Y
quantum numbers by
Y
Q = T3 + (9.40)
2
This relation is also called the Gell-Mann-Nishima relation. Interpreted in terms of
quantum numbers, Q is the electromagnetic charge, Y is the hypercharge and T3 is the
charge associated to the third generator of SU (2) (i.e. ⌧3 /2). The quantum number for
the hypercharge of the various fermion fields is not predicted by the theory, just as the
electromagnetic charges are not predicted.
As an example, consider the neutrino-electron doublet. A left-handed neutrino state
has T3 = 1/2 while the left-handed electron has T3 = 1/2. (If you don’t understand
this, consider the eigenvalues of the eigenvectors
✓ ◆ ✓ ◆
⌫ 0
L,⌫ = and L,e =
0 e
for the ⌧3 /2 generator.) The right-handed electron is a singlet under SU (2)L and has
T3 = 0. Given a coupling constant e, the observed electromagnetic charge of the elec-
tron is 1. Therefore, the hypercharge of the right-handed electron is 2 while the
hypercharge of the left-handed electron and neutrino are both 1. (The latter two must
be equal, since the SU (2)L doublet is a singlet under U (1)Y .)
Expressing the interactions terms of the bµ3 and aµ fields in the Lagrangian above in
terms of the physical fields, we find
✓ ◆
µ 3 g0 µ µ 0 JYµ
gJ3 bµ J aµ = g sin ✓W J3 + g cos ✓W Aµ
2 Y 2
✓ ◆
µ 0 JYµ
g cos ✓W J3 g sin ✓W Zµ
2
µ
⌘ eJEM Aµ gZ JNµ C Zµ (9.41)
where in the last line we defined the currents and coupling constants associated to the
physical fields.
A direct consequence of Eq. (9.40) is that also the currents are related, namely by
µ 1
JEM = J3µ + JYµ . (9.42)
2
Comparing this to Eq. (9.41) we find that
e = g sin ✓W = g 0 cos ✓W (9.43)
This relation implies that there are only two independent parameters. In particular, the
weak mixing angle is related to the SU (2)L and U (1)Y coupling constants by
g 0 /g = tan ✓W (9.44)
Analogously, we find for the Z 0 interaction term,

✓ ◆
µ g0 µ µ
g cos ✓W J3 sin ✓W · 2 (JEM J3 ) Zµ
2
= ...
e
= J3µ sin2 ✓W JEM µ
Zµ
cos ✓W sin ✓W
Consequently, expressed in J3 and JEM the neutral current is
JNµ C = J3µ µ
sin2 ✓W JEM (9.45)
and its coupling constant becomes

e g
gZ = = . (9.46)
cos ✓W sin ✓W cos ✓W
9.4. COUPLINGS FOR Z ! F F 151
9.4 Couplings for Z ! f f

Let us now take a closer look at the neutral current in Eq. (9.45). The J3 current only
involves the left-handed doublets, while the electromagnetic current couples both to
left- and right-handed fields. Therefore, we conclude that the neutral current cannot be
purely left-handed.
We defined the left-handed current in terms of a left-handed neutrino-electron (or up-
down quark) douplet L as
J3µ = L µ T3 L . (9.47)
In analogy with the procedure that we applied for the charged current, we can now
rewrite this in terms of the two fermion fields in the douplet using the explicit represen-
tation of T3 as
J3µ = ⌫ L µ T3⌫ ⌫L + eL µ T3e eL , (9.48)
where the neutrino and electron weak isospin charges are T3⌫ = 1/2 and T3e = 1/2.
Since we scatter particles rather than douplets, we consider the currents for the neutrino
and electron separately.
We can generalize this and write for any fermion field f
µ
J3,f = L
µ
T3f L
1 (9.49)
= µ
(1 5
)T3f
2
Adding the contribution from the electromagnetic current, the neutral current for fermion
f then becomes
µ µ 5
JNC,f = 1
2
(1 )T3,f sin2 ✓W Qf (9.50)
It is customary to write the term on the right in terms of a vector and an axial vector
coupling such that ⇣ ⌘
µ µ 1
JNC,f = CVf CAf 5 (9.51)
2
which implies that
CVf = T3f 2Qf sin2 ✓W
(9.52)
CAf = T3f
Alternatively, we can write the current in terms of left- and right-handed fields as
1 ⇣ f f µ f f µ f
⌘
JNµ C,f = CL L L + C f
R R R (9.53)
2
with the left- and right-handed couplings given by
CLf ⌘ CVf + CAf = T3f Qf sin2 ✓W

(9.54)
CRf ⌘ CVf CAf = Qf sin2 ✓W
As stated before the values of the charge of the di↵erent fermion fields is not predicted.
Table 9.1 lists the quantum numbers and resulting vector and axial-vector couplings
for all fermions in the Standard Model. The model can be experimentally tested by
measuring these couplings in di↵erent processes.
fermion T3 Y T3 Y Q CAf CVf

left-handed right-handed
⌫e , ⌫µ , ⌫⌧ + 12 1 0 0 0 + 12 + 12
e, µ, ⌧ 1
2
1 0 2 1 1
2
1
2
+ 2 sin2 ✓W
u, c, t + 12 + 13 0 + 43 + 23 1
+2 1
+2 4
3
sin2 ✓W
d, s, b 1
2
+ 13 0 2
3
1
3
1
2
1
2
+ 23 sin2 ✓W
Table 9.1: Gauge interaction quantum numbers and corresponding vector and axial vector
couplings for the fermions in the Standard Model.
Finally, expressed in terms of the left- and right-handed couplings, the Feynman rule
corresponding to Z vertex becomes
f ⇣ ⌘
g µ1
Z 0 i CVf CAf 5
cos ✓W 2
f
9.5 The mass of the W and Z bosons

In Lecture 7 we expressed the charged current coupling for processes with momentum
transfer q ⌧ MW as a four-point interaction. Comparing the expressions to those in
this lecture, we can show that the Fermi coupling constant is related to the gauge field
couplings as
GF g2
p = 2
(9.55)
2 8MW
For neutral current processes we can also compute the coupling-constant of the four-
point interaction. It is given by
GF g2
⇢p = (9.56)
2 8MZ2 cos2 ✓W
The parameter ⇢ specifies the relative strength between the charged and neutral current
weak interactions. Comparing the two expressions, we have
2
MW
⇢= (9.57)
Mz2 cos2 ✓W
The masses of the W ± and Z 0 can be precisely measured, for instance by reconstructing
a ‘two-jet’ invariant mass distribution in high-energy e+ e collisions: Provided that
9.5. THE MASS OF THE W AND Z BOSONS 153
the collision energy is large enough, the di-jet mass will show mass peaks for ‘on-shell’
produced W ± and Z 0 . The most precise measurement of the four-point coupling for
the charged current comes from the measurement of the muon lifetime. The ratio
of the charged and neutral current couplings was first measured by the Gargamelle
experiment, which exploited an intense neutrino beam to measure the cross-section for
a neutral current process ⌫µ + nucleus ! ⌫µ + hadrons and a charged current process
⌫µ + nucleus ! µ + hadrons.
Upon combination of measurements for the couplings an the masses it is found that the
experimental value for ⇢ is 1 within small uncertainties. This is actually a prediction
of the Higgs mechanism. In the Higgs mechanism the mass generated for the W and Z
are respectively
q
1 1
MW = v g and MZ = v g 2 + g 0 2 , (9.58)
2 2
where v is the so-called vacuum expectation value of the Higgs field. With g 0 /g = tan ✓W
we find that ⇢ = 1. Therefore, in the Standard Model the masses of the massive vectors
bosons are related by
MW = MZ cos ✓W . (9.59)
The best fit of the Standard Model to all experimental data gives approximately
sin2 ✓W = 0.231 (9.60)
sp
2 e
MW = = 80.4 GeV (9.61)
8GF sin ✓W
MZ = MW (gz /g) = MW /cos ✓ = 91.2 GeV (9.62)
Summary
We have introduced a local gauge symmetry SU (2)L ⌦ U (1)Y to obtain a Lagrangian
for electroweak interactions,
✓ ◆
µ g0 µ
g JL · bµ + JY · aµ (9.63)
2
The coupling constants g and g 0 are free parameters. We can also take e and sin2 ✓W .
The electromagnetic and neutral weak currents are then given by:
µ 1
JEM = J3µ + JYµ
2
JYµ
JNµ C = J3µ sin 2 µ
✓W JEM 2
= cos ✓W J3µ 2
sin ✓W
2
and the interaction term in the Lagrangian becomes:
✓ ◆
µ e µ
eJEM · Aµ + J · Zµ (9.64)
cos ✓W sin ✓W N C
in terms of the physical fields Aµ and Zµ .

Note that we still have two independent coupling constants (be it e and ✓W or g and
g 0 ). Therefore, it is sometimes said that we have not really ’unified’ the electromagnetic
and the weak interaction, but rather just put them under the same umbrella. Still,
there are clear predictions in this model, such as all the relations between the couplings
to the di↵erent fermion field and the W to Z mass ratio. Up to now the Glashow-
Salam-Weinberg theory, which predicted the W and Z more than a decade before their
experimental confirmation, has gloriously passed all experimental tests.
Exercises
Exercise 9.1 (Currents for left and right-handed chirality)
We define the chiral projection operators as PL ⌘ 12 (1 5
) and PR ⌘ 12 (1 + 5
) =
1 PL and the left- and right-handed chirality bi-spinor states as L ⌘ PL and
R ⌘ PR .
(a) Show that the left-handed adjoint spinor, defined as L ⌘( L ), is given by
L = PR
(b) Show that the vector current can be decomposed as

µ µ µ
= L L + R R
(c) Show that the scalar current can be decomposed as
= R L + L R
Exercise 9.2 (Charged current interaction)

Show how we get from Eq. (9.22) to Eq. (9.24).
Exercise 9.3 (Symmetries (optional))

For each of the symmetries below indicate what the symmetry is about, and motivate
whether it is an exact or a broken symmetry in nature:
(a) U1(Q) symmetry
(b) SU2(u-d-flavour) symmetry
(c) SU3(u-d-s-flavour) symmetry
(d) SU3(colour) symmetry
(e) SU2(weak-isospin) symmetry
9.5. THE MASS OF THE W AND Z BOSONS 155
(f ) SU5(Grand unified) symmetry

(g) Super-symmetry (SUSY)
Lecture 10
The Process e e+ ! µ µ+
10.1 Helicity conservation

In Lecture 6 we performed computations of the cross-section of several scattering pro-
cesses in QED. In Lecture 7 we also looked at the decay of the µ and the charged ⇡
via the weak interaction. The results of these computations must of course obey con-
servation of total angular momentum. In the processes that we look at, orbital angular
momentum plays no role, such that we really talk about conservation of spin. (See also
section 5.11.) The conservation rule is built into the computations, but sometimes it is
still useful to understand how it comes about.
In the previous lecture we have defined the chiral projections,
5 5
L = 12 (114 ) R = 12 (114 + )
1 5 1 5
(10.1)
L = 2
1 +
(1 4 ) R = 2
1
(1 4 )
The states L and L are respectively the in- and outgoing left-handed particle chiral
states, while R and R are the in- and outgoing right-handed particle chiral states.
We have argued before that in the ultra-relativistic limit there is a correspondence
between helicity states and chiral states. You can formalize this by defining the helicity
projection operators as
P" = 12 (114 + ⌃ · p̂) P# = 12 (114 ⌃ · p̂) (10.2)
and subsequently show that for any positive energy spinor with momentum pz along the
z-axis
uR = x u" + (1 x) u#
(10.3)
uL = x u# + (1 x) u"
where we defined
r
(1 + ↵)2 |p| 1
x⌘ with ↵⌘ = . (10.4)
2(1 + ↵2 ) E+m +1
157
158 LECTURE 10. THE PROCESS e e+ ! µ µ+
In the ultra-relativistic limit, we have x ! 1 and the correspondence between the chiral
states and helicity states is obtained.
As a consequence, for any process that would violate helicity conservation in the ultra-
relativistic limit, such as the ⇡ + ! e+ ⌫e decay via the weak interaction, a helicity
suppressing factor (1 x) appears in the amplitude. Simply said, this is because the
interaction can only couple to a ‘fraction’ (1 x) of the lepton wave function.
You have shown in the previous lecture that for a vector coupling, we can decompose
the current in the vertex factor as follows
µ µ µ
= R R + L L (10.5)
This means that a right-handed state only couples to a right-handed state, and a left-
handed state only to a left-handed state. This results holds equally well for an axial
vector coupling ( µ 5 ). It is graphically illustrated in Fig. 10.1. Note that ‘crossing’ a
particle flips its chirality.
R R L L R L
or
L R
Figure 10.1: Helicity conservation in vector and axial-vector couplings. left: A right-handed
incoming electron scatters into a right-handed outgoing electron and vice versa in a vector
or axial vector interaction . right: In the crossed reaction the energy and momentum of one
electron is reversed: i.e. in the e+ e pair production a right-handed electron and a left-handed
positron (or vice versa) are produced. This is the consequence of a spin=1 force carrier. (In
all diagrams time increases from left to right.)
For scalar couplings the situation is exactly opposite, as its decomposition would read
= R L + L R. (10.6)
As we shall see in the remainder of this chapter, conservation of helicity has interesting
consequences for the e e+ ! µ µ+ process as well. For example, it allows us to under-
stand the angular dependence of a the polarized cross-sections without going in detail
through the kinematics.
10.2 The cross section of e e+ ! µ µ+

Equipped with the Feynman rules of the electroweak theory we now proceed with the
calculation of the cross-section of the electroweak process e e+ ! ( , Z) ! µ µ+ . We
study the process in the centre-of-momentum frame,
10.2. THE CROSS SECTION OF e e+ ! µ µ+ 159
µ+
θ −
e+ e
µ−
with pi the momentum of an incoming electron, pf the momentum of an outgoing muon

and cos ✓ the angle between the e+ and the µ+ .
e+ µ e+ µ
M : MZ :
Z
+
e µ e µ+
Figure 10.2: Leading order Feynman diagrams contributing to e e+ ! µ µ+ .
In lecture 6 we considered this process in QED. At leading order there was only one
contribution to the amplitude, namely via an intermediate photon. In the electroweak
theory also the amplitude with an intermediate Z 0 boson contributes. The corresponding
Feynman diagrams are shown in Fig. 10.2.
Once we have computed the relevant amplitudes, the di↵erential cross-section follows
as usual from the golden rule,
d (e e+ ! µ µ+ ) 1 1 pf 2
= 2
|M| (10.7)
d⌦ 64⇡ s pi
where the invariant amplitude is the sum of the photon and Z 0 contributions.
In Lecture 6 we computed the spin-averaged amplitude via a rather lengthy procedure,
involving Casimir’s track and the trace theorems. Because it is actually a nice illustration
of the concept of helicity conservation, we will here follow a di↵erent approach.
10.2.1 Photon contribution
Consider first only the matrix element of the photon contribution (evaluated using the
Feynman rules, see e.g. appendix B),
gµ⌫
M = e2 m
µ
m · · e
⌫
e (10.8)
q2
where the subscript ‘m’ referes to the muon and the subscript ‘e’ to the electron. We
now decompose the spinors in left- and right-handed chirality states, as we did in lecture
11,
µ µ µ
m m = Lm Lm + Rm Rm
e µ e = Le µ Le + Re µ Re .
The total amplitude then becomes
e2 ⇥ µ µ
⇤
M = Lm Lm + Rm rm ·
s ⇥ ⇤ (10.9)
Le µ Le + Re µ Re
where we have also substituted q 2 = s.

The matrix element thus consists of four contributions with definite chirality for the
four particles in the process. For high energies the chiral projections are helicity states
and the four contributions correspond exactly to four polarized amplitudes. Since we
can choose a basis with helicity eigenstates, the four contributions do not interfere, and
the total cross-section becomes the sum of the amplitudes squared:

d unpolarized 1 d d
= eL e + +
R ! µL µR + e e+ ! µR µ+ L +
d⌦ 4 d⌦ d⌦ L R
(10.10)
d + + d + +
e e ! µL µR + e e ! µR µL
d⌦ R L d⌦ R L
where we average over the incoming spins and sum over the final state spins. Note that
e+
R ⌘ Le etc.
Let us now look in more detail at the helicity dependence (H&M §6.6):
Initial state:
z−axis In the initial state the e and

e+ e− e+ have opposite helicity (as they
produce a spin 1 ).
z’−axis
Final state: µ− The same is true for the final state

θ µ and µ+ .
µ+
In the center of mass frame, scattering proceeds from an initial state with JZ = +1 or
1 along axis ẑ into a final state with JZ0 = +1 or 1 along axis ẑ 0 . Since the interaction
proceeds via a photon with spin J = 1 the amplitude for scattering over an angle ✓ is
given by the rotation matrices1
⌦ ↵
djm0 m (✓) ⌘ jm0 |e i✓Jy
|jm (10.12)
where Jy is the y component of the angular momentum operator (which is also the
generator for rotations around the y-axis). The coefficients djm,m0 are sometimes called
‘Wigner d-matrices’. Computing them for the spin-1 system is not so hard (see e.g.
exercise 10.2, or H&M exercise 2.6) and gives
1
d1+1,+1 (✓) = d1 1, 1 (✓) = (1 + cos ✓)
2 (10.13)
1
d1+1, 1
1 (✓) = d 1,+1 (✓) = (1 cos ✓)
2
For the kinematic factors in the four amplitudes we then get
µ µ 1
Lm Lm Le Le = d 1, 1 (✓) = 2
(1 + cos ✓)
µ µ 1
Rm Rm Re Re = d+1,+1 (✓) = 2
(1 + cos ✓)
µ µ 1
(10.14)
Lm Lm Re Re = d+1, 1 (✓) = 2
(1 cos ✓)
µ µ 1
Rm Rm Le Le = d+1,+1 (✓) = 2
(1 cos ✓)
As a consequence of the helicity conservation, scattering is suppressed in the direction

in which the spin of the µ is not aligned with the spin of the e .
Using this result the polarized cross-sections become
d d ↵2
e L e+ +
R ! µL µR = e e+ ! µR µ+ = (1 + cos ✓)2
d⌦ d⌦ R L L
4s (10.15)
d d ↵2
e L e+ +
R ! µR µL = eR e + +
L ! µL µR = (1 cos ✓)2
d⌦ d⌦ 4s
The unpolarised cross section is obtained as the spin-averaged sum,
d unpol 1 ↵2 ⇥ ⇤ ↵2
= 2 (1 + cos ✓)2 + (1 cos ✓)2 = 1 + cos2 ✓ . (10.16)
d⌦ 4 4s 4s
in agreement with our computation in Lecture 6.

1
See H&M§2.2:
X
e i✓J2
|j mi = djm m0 (✓) |j m0 i (10.11)
m0
and also appendix H in Burcham & Jobes

10.2.2 Z 0 contribution
Having written the total cross-section as a sum of polarized amplitudes we are ready to
include the contribution from the Z 0 boson amplitude. Using the Feynman rules (see
e.g. appendix B) we find for the invariant amplitude
g2 ⇥ µ
⇤ gµ⌫ qµ q⌫ /MZ2 ⇥ ⇤
MZ = m CVm CAm 5
m · · e
⌫
CVe CAe 5
e
4 cos2 ✓w q 2 MZ2
(10.17)
p
We can simplify the Z 0 propagator if we ignore the lepton masses (m` ⌧ s). In that
case the Dirac equation becomes:
µ µ
e (i@µ m) = 0 ) e ( pµ,e ) = 0 (10.18)
Since pe = 12 q we also have:

1
( µ qµ ) = 0 (10.19)
2 e
As a result the qµ q⌫ term in the propagator vanishes and we obtain for the matrix
element
g2 1 ⇥ µ
⇤⇥ ⇤
MZ = 2 2 2
· m CVm CAm 5
m e µ CVe CAe 5
e .
4 cos ✓w q MZ
(10.20)
We have shown in the previous lecture how the (CVe CAe 5 ) terms can be written in
terms of left- and right-handed couplings. Defining
CR ⌘ CV CA and CL ⌘ CV + CA (10.21)
one finds
5
CV CA = CR R + CL L . (10.22)
Consequently, the Z 0 amplitude can be written as
g2 1 ⇥ m µ
⇤
MZ = 2 2
CL Lm Lm + CRm Rm
µ
Rm ·
4 cos ✓w s MZ (10.23)
⇥ e ⇤
CL Le µ Le + CRe Re µ Re
Comparing this the expression to Eq. (10.9) we realize that we can obtain the polarized
cross-sections directly from the results obtained for the QED process. For two of the
four contributions we then obtain
d ↵2
e L e+ +
R ! µL µR = (1 + cos ✓)2 · |1 + r CLm CLe |2
d⌦ ,Z 4s
(10.24)
d ↵2
e L e+
R ! µR µ+
L = (1 cos ✓)2 · |1 + r CRm CLe |2
d⌦ ,Z 4s
with the relative contribution of the Z 0 and parameterized as
g2 1 s
r= 2 . (10.25)
e 4 cos ✓w s Mz2
2
The other two helicity configuration follow using the relation in Eq. (10.15) and replacing
CL by CR etc . Using the relation between the coupling constants
GF g2 g2
p = 2
= . (10.26)
2 8MW 8MZ2 cos2 ✓w
we can also write r in terms of GF as

p
2GF MZ2 s
r= . (10.27)
e2 s MZ2
10.2.3 Correcting for the finite width of the Z 0
The propagator for the massive vectors bosons has a ‘pole’ at the boson mass: it becomes
infinitely large for an ‘on-shell’ (p2 = m2 ) boson. As you can readily see from the
expression above,
p this would lead to an infinite cross-section when we tune the beam
energies to s = MZ . The problem is that the propagator does not take into account
the finite decay width of the Z 0 . The Z 0 boson is not a stable particle and hence the ’on-
shell’ Z 0 is actually something with a rather broad mass distribution. We can account
for the width by replacing the mass in the propagator with
i
MZ ! MZ Z (10.28)
2
where is the total decay width of ‘on-shell’ (i.e. not virtual) Z 0 -bosons.
A heuristic explanation (Halzen and Martin, §2.10) is as follows. The decay of an
unstable particle follows the exponential law
| (t)|2 = | (0)|2 e t
(10.29)
where | (0)| is the probability (density) at t = 0 and 1/ is the lifetime. Therefore, the
time-dependencepof the wave function, which already involves the rest mass, must also
include a factor e t/2 , or
imt t/2
(t) = (0)e e (10.30)
Consequently, with the substitution above we can ‘correct’ the propagator mass for the
finite decay width. The lineshape that results from such a propagator is usually called
a (spin-1) Breit-Wigner.
With the replacement MZ ! MZ i Z /2 in the propagator, the expression for r becomes

p
2GF MZ2 s
r= 2 2
e s MZ i 2Z
p (10.31)
2GF MZ2 s
= ⇣ 2
⌘
e2 s Mz2 Z
+ iMZ Z
4
p
Note that r is now complex and that the phase of r depends on s.
10.2.4 Total unpolarized cross-section
As a final step we add up the unpolarized cross-sections and find

d ↵2 ⇥ ⇤
= A0 1 + cos2 ✓ + A1 cos ✓ (10.32)
d⌦ 4s
with A0 and A1 given by
⇣ 2 2
⌘
A0 = 1 + 2Re(r) CVe CVf + |r|2 CVe 2 + CAe 2 CVf + CAf
(10.33)
A1 = 4Re(r) CAe CAf + 8|r|2 CVe CVf CAe CAf
In this expression we replaced CR and CL by CV and CA using the definitions given
earlier.
Since we have not actually used that the final state fermions are muons, our result is
valid for any process e e+ ! , Z 0 ! f f¯, provided that we insert the correct weak
couplings.2
Assuming “lepton universality”, we have CVe = CVµ and CAe = CAµ . The expressions for
A0 and A1 then become
2
A0 = 1 + 2 Re(r) CV2 + |r|2 CV2 + CA2
A1 = 4 Re(r) CA2 + 8|r|2 CV2 CA2
In the Standard Model the coefficients for leptons are CA = 1
2
and CV = 1
2
+2 sin2 ✓w .
Finally, we consider the total cross-section. In the integrated cross-section the term
proportional to cos ✓ vanishes and for the other term we use
Z
16⇡
(1 + cos2 ✓) d⌦ = (10.34)
3
Consequently, the total cross-section is
4⇡↵2
(s) = A0 (s) (10.35)
3s
2
This is not entirely true: For quarks we also need to insert the charge in the coupling to the photon.
10.3. NEAR THE RESONANCE 165
To summarize, on the amplitude level there are two diagrams that contribute:
e µ e µ
M : MZ :
Z (10.36)
e+ µ+ e+ µ+
Using the following notation
d
[Z, Z] = Z · Z / |r|2
d⌦
d
[ Z] = · Z / Re (r)
d⌦
d
[ , ] = · /1
d⌦
the expression for the di↵erential cross-section is

d d d d
= [ , ] + [Z, Z] + [ , Z] (10.37)
d⌦ d⌦ d⌦ d⌦
with
d ↵2
[ , ] = 1 + cos2 ✓
d⌦ 4s
d ↵2 2 h e 2 ⇣ 2 2
⌘ i
[Z, Z] = |r| CV + CAe 2 CVf + CAf 1 + cos2 ✓ + 8CVe CVf CAe CAf cos ✓
d⌦ 4s
d ↵2 h i
[ , Z] = Re|r| CVe CVf 1 + cos2 ✓ + 2CAe CAf cos ✓
d⌦ 4s
10.3 Near the resonance

Let us take a closer look at the cross-section for beam energies close to the Z 0 mass.
One can see from Eq. (10.31) that |r| is maximal for
2
Z
s0 = MZ2 . (10.38)
4
Since the denominator in Eq. (10.31) is purely imaginary for s = s0 , the interference
term, which is proportional to Re(r), vanishes at the peak. Defining the kinematic
factor in r as
s
(s) = ⇣ 2
⌘ (10.39)
s Mz2 Z
4
+ iMZ Z
you will show in exercise 10.1(b) that

⇣ s0 ⌘ 2 s2
Re() = 1 || with ||2 = . (10.40)
s (s s0 )2 + MZ2 2
Z
You will also show that at the resonance |r|2 1 such that we can ignore the photon
contribution entirely. With neither the interference nor the photon contribution, we
have
p !2
2GF MZ2 s2 ⇣ 2 ⌘
e2 e2 f f2
A0 (s) ⇡ C V + C A CV + CA (10.41)
e2 (s s0 )2 + Mz2 2Z
Exactly at the resonance, this gives for the total cross-section to the final state f f¯:
G2 s0 MZ2 ⇣ e 2 ⌘⇣ 2
f2
⌘
(e+ e ! f f¯)|s=s0 = F 2
CV + CA
e2
CV
f
+ CA (10.42)
6⇡ Z
σhad [nb]
0
σ
40
ALEPH
DELPHI
L3
OPAL
30
ΓZ
20
measurements, error bars
increased by factor 10
10 σ from fit
QED unfolded
MZ
86 88 90 92 94
Ecm [GeV]
Figure 10.3: left: The Z 0 -lineshape: the cross-section for e+ e ! hadrons as a function of
p
s. right: Same but now near the resonance. The dashed line represents the leading order
computation,while the continuous gray line includes higher order corrections.
For quark-antiquark final states (f = q) we need to take into account that there are
three distinct colour configurations, namely blue-anti-blue, red-anti-red and green-anti-
green. Therefore, for a quark-anti-quark park, the cross-section involves another factor
Nc = 3,
+ G2F s0 MZ2 ⇣ e 2 e2
⌘⇣
q2 q2
⌘
(e e ! q q̄)|s=s0 = Nc · 2
CV + CA CV + CA (10.43)
6⇡ Z
Figure 10.3 shows the measured cross-section in hadronic final states as function of the
collision energy.
10.4. THE FORWARD-BACKWARD ASYMMETRY 167
p
At collision energies well above the typical QCD binding energy ( s 2m⇡ ), the q q̄
state is observed as two ‘jets’, collimated showers of light mesons. The ratio between
the hadronic and leptonic event yields at the Z 0 resonance,
(e+ e ! hadrons)
Rl = (10.44)
(e+ e ! µ+ µ )
provides an important test of the standard model, as shown in Fig. 10.4.
0.022
68% CL
0.018
0,l
mt
Afb
αs
0.014 mH
+−
ll
+ −
ee
+ −
µµ
+ −
ττ
0.01
20.6 20.7 20.8 20.9
Rl
combined in plots with SLD results
Figure 10.4: left: Tests of the standard model. The leptonic Af b vs. Rl . The contours show
the measurements while the arrows show the dependency on Standard Model parameters.
right: Determination of the vector and axial vector couplings.
10.4 The forward-backward asymmetry

A direct consequence of the photon-Z 0 interference is that the angular distribution is not
symmetric. Figure 10.5 shows the cos ✓ distribution observed at the Jade experiment,
which operated at the PETRA collider in Hamburg. The beam energy in this experiment
was not yet sufficient to directly produce Z 0 bosons. Still, the e↵ect of the interference
was clearly visible long before the direct discovery of the Z 0 resonance.
At the peak and ignoring the pure photon exchange (because it is negligibly small), the
polar angle distribution is given by
d 8
/ 1 + cos2 ✓ + AFB cos ✓ , (10.45)
d cos ✓ 3
where we defined the ‘forward-backward asymmetry’,
3 2CVf CAf
A0,f
FB = Ae Af with Af = . (10.46)
4 CV2 + CA2
Precise measurements of the forward-backward asymmetry can be used to determine
the couplings CV and CA .
p
Figure 10.5: Angular distribution for e+ e ! µ+ µ for s > 25 GeV at the JADE experi-
ment. ✓ is the angle between the outgoing µ+ and the incoming e+ . The curves show fits to
the data p(1 + cos2 ✓) + q cos ✓ (full curve) and p(1 + cos2 ✓) (dashed curve). (Source: JADE
collaboration, PLB, Vol108B, p108, 1981.)
10.5 The Z 0 decay width and the number of light

neutrinos
Using the Feynman rules we can also compute the Z 0 ! f f¯ decay width, represented
by the diagram f
f
You cannot easily do this computation yourself, since we have not discussed the external
line for the Z 0 in this course. (The computation needs to take into account the three
polarization states of the massive vector boson.) The result of the computation is
1 1 2
Z ! ff = M
16⇡ MZ
g2 Mz ⇣ f 2 f2
⌘
= C V + C A (10.47)
48⇡ cos2 ✓w
GF MZ3 ⇣ f 2 2
⌘
= p CV + CAf
6 2 ⇡
For quark-antiquark final states (f = q) we again need to multiply by the colour factor
Nc = 3,
GF MZ3 ⇣ q 2 q2
⌘
(Z ! qq) = p CV + CA · NC . (10.48)
6 2 ⇡
The total decay width of the Z 0 is the sum of all partial widths to all accessible final
10.5. THE Z 0 DECAY WIDTH AND THE NUMBER OF LIGHT NEUTRINOS 169
states,
Z = ee + µµ + ⌧⌧ +3 uu +3 dd +3 ss +3 cc +3 bb + N⌫ · ⌫⌫ , (10.49)
where N⌫ is the number of neutrino species, which is equal to three in the standard
model.
Using all available data to extract information on the couplings we can compute the
decay widths to all final states within the standard model,
1
ee ⇡ µµ ⇡ ⌧⌧ = 84 MeV CV ⇡ 0 CA =
2
1 1
⌫⌫ = 167 MeV CV = CA =
2 2
1
uu ⇡ cc = 276 MeV CV ⇡ 0.19 CA =
2
1
dd ⇡ ss ⇡ bb = 360 MeV CV ⇡ 0.35 CA =
2
p
A measurement of the lineshape (the cross-section as function of s) gives for the total
decay width of the Z 0 ,
Z ⇡ 2490 MeV
So, even though we cannot see the neutrino contribution, we can estimate the number
of neutrinos from the total width of the Z 0 . The result is
Z 3 l had
N⌫ = = 2.984 ± 0.008 . (10.50)
⌫⌫
Figure 10.6 shows the predicted lineshape for di↵erent values of N⌫ . This results put
strong constraints on extra generations: if there is a fourth generation, then either it
has a very heavy neutrino, or its neutrino does not couple to the Z 0 . In either case, this
generation would be very di↵erent from the known generations of quarks and leptons.
Figure 10.6: The Z lineshape fit for N⌫ = 2, 3, 4.

10.5. THE Z 0 DECAY WIDTH AND THE NUMBER OF LIGHT NEUTRINOS 171
Exercises
Exercise 10.1 (Z 0 production and decay)
(a) Derive the expression for Re() in Eq. (10.40).
(b) Calculate the relative contribution of the Z 0 -exchange and the exchange to the
cross section at the Z 0 peak. Use sin2 ✓W = 0.23, Mz = 91GeV and Z = 2.5GeV .
(c) Show also that at the peak
12⇡ e µ
peak (e e+ ! µ µ+ ) ⇡ 2
(10.51)
Mz2 Z
(d) Why does the top quark not contribute to the decay width of the Z 0 ?
(e) Calculate the value of Rl = had / lep at the resonance s = s0 . Ignore the masses
of the fermions, as we did in the lecture. You may also ignore the contribution of
the photon, as it is very small at the resonance.
p
(f ) The actual line shape of the Z 0 -boson is not a pure Breit Wigner: at the high s
side of the peak the cross section is higher then expected from the formula derived
in the lectures. Can you think of a reason why this would be the case?
(g) The number of light neutrino generations is determined from the “invisible width”
of the Z 0 -boson as follows:
Z 3 l had
N⌫ =
⌫
Can you think of another way to determine the decay rate of Z 0 ! ⌫ ⌫¯ directly?
Do you think this method is more precise or less precise?
Exercise 10.2 (Spin projections (optional!) (From Thomson, ex. 6.6))

Consider a spin-1 system. The eigenstates of the operator Sn = n · S correspond to
the spin projections in the direction n. These can be written in terms of the eigenstates
of the operator Sz , for instance as
|1, +1in = ↵ |1, +1i + |1, 0i + |1, 1i
Taking n = (sin ✓, 0, cos ✓) show that

1 p1
|1, +1in = 2
(1 cos ✓) |1, +1i + 2
sin ✓ |1, 0i + 12 (1 + cos ✓) |1, 1i (10.52)
Hint: Write Sx in terms of the spin ladder operators and use that all states are normal-
ized to 1.
Lecture 11
Symmetry breaking
After a review of the shortcomings of the model of electroweak interactions in the Stan-
dard Model, in this section we study the consequences of spontaneous symmetry break-
ing of (gauge) symmetries. We will do this in three steps of increasing complexity and
focus on the principles of how symmetry breaking can be used to obtain massive gauge
bosons by working out in full detail the breaking of a local U(1) gauge invariant model
(QED) and give the photon a mass.
11.1 Problems in the Electroweak Model

The electroweak model, beautiful as it is, has some serious shortcomings.
1] Local SU(2)L ⇥ U(1)Y gauge invariance forbids massive gauge

bosons
In the theory of Quantum ElectroDynamics (QED) the requirement of local gauge in-
0
variance, i.e. the invariance of the Lagrangian under the transformation ! ei↵(x)
plays a fundamental rôle. Invariance was achieved by replacing the partial derivative by
a covariant derivative, @µ ! Dµ = @µ ieAµ and the introduction of a new vector field
0
A with very specific transformation properties: Aµ ! Aµ + 1e @µ ↵. This Lagrangian for
a free particle then changed to:
1
LQED = Lfree + Lint Fµ⌫ F µ⌫ ,
4
which not only ’explained’ the presence of a vector field in nature (the photon), but also
automatically yields an interaction term Lint = eJ µ Aµ between the vector field and the
particle as explained in detail in the lectures on the electroweak model. Under these
173
174 LECTURE 11. SYMMETRY BREAKING
symmetry requirements it is unfortunately not possible for a gauge boson to acquire a

mass. In QED for example, a mass term for the photon, would not be allowed as such
a term breaks gauge invariance:
1 2 1 1 1 1
m Aµ Aµ = m2 (Aµ + @µ ↵)(Aµ + @ µ ↵) 6= m2 Aµ Aµ
2 2 e e 2
The example using only U(1) and the mass of the photon might sounds strange as the
photon is actually massless, but a similar argument holds in the electroweak model for
the W and Z bosons, particles that we know are massive and make the weak force only
present at very small distances.
2] Local SU(2)L ⇥ U(1)Y gauge invariance forbids massive fermions
Just like in QED, invariance under local gauge transformations in the electroweak model
~ µ + ig 0 1 Y Bµ
requires introducing a covariant derivative of the form Dµ = @µ + ig 12 ~⌧ · W 2
introducing a weak current, J weak and a di↵erent transformation for isospin singlets and
doublets. A mass term for a fermion in the Lagrangian would be of the form mf ¯ ,
but such terms in the Lagrangian are not allowed as they are not gauge invariant. This
is clear when we decompose the expression in helicity states:
mf ¯ = mf ¯R + ¯L ( L + R)
⇥ ⇤
= mf ¯R L + ¯L R , since ¯R R = ¯L L =0
Since L (left-handed, member of an isospin doublet, I = 12 ) and R (right-handed,

isospin singlet, I = 0) behave di↵erently under rotations these terms are not gauge
invariant:
0
L ! L = ei↵(x)T +i (x)Y L
0
R ! R = ei (x)Y R
3] Violating unitarity
Several Standard Model scattering cross-sections,

like WW-scattering (some Feynman graphs are W Z W Z
shown in the picture on the right) violate unitar- +
W
ity at high energy as (WW ! ZZ) / E2 . This
W Z W Z
energy dependency clearly makes the theory non-
renormalizable.
11.2. A FEW BASICS ON LAGRANGIANS 175
How to solve the problems: a way out
To keep the theory renormalizable, we need a very high degree of symmetry (local
gauge invariance) in the model. Dropping the requirement of the local SU(2)L ⇥ U(1)Y
gauge invariance is therefore not a wise decision. Fortunately there is a way out of this
situation:
Introduce a new field with a very specific potential that keeps the full Lagrangian
invariant under SU(2)L ⇥ U(1)Y , but will make the vacuum not invariant under
this symmetry. We will explore this idea, spontaneous symmetry breaking of a
local gauge invariant theory (or Higgs mechanism), in detail in this section.
The Higgs mechanism: - Solves all the above problems
- Introduces a fundamental scalar ! the Higgs boson !
11.2 A few basics on Lagrangians

A short recap of the basics on Lagrangians we’ll be using later.
L = T(kinetic) V(potential)
The Euler-Lagrange equation then give you the equations of motion:

✓ ◆
d @L @L
=0
dt @ q̇i @qi
For a real scalar field for example:

1 1 2
Lscalar = (@µ ) (@ µ ) m 2
! Euler-Lagrange ! (@µ @ µ + m2 ) = 0
2 2 | {z }
Klein-Gordon equation
In electroweak theory, kinematics of fermions, i.e. spin-1/2 particles is described by:
Lfermion = i ¯ µ @ µ m ¯ ! Euler-Lagrange ! (i µ @ µ m) = 0
| {z }
Dirac equation
In general, the Lagrangian for a real scalar particle ( ) is given by:
L= (@µ )2 + |{z}C + ↵ + 2
+ 3
+ 4
+ ...
| {z } |{z} |{z} |{z} |{z}
kinetic term constant ? mass term 3-point int. 4-point int.
(11.1)
We can interpret the particle spectrum of the theory when studying the Lagrangian
under small perturbations. In expression (11.1), the constant (potential) term is for
most purposes of no importance as it does not appear in the equation of motion, the
term linear in the field has no direct interpretation (and should not be present as we will
explain later), the quadratic term in the fields represents the mass of the field/particle
and higher order terms describe interaction terms.
11.3 Simple example of symmetry breaking
To describe the main idea of symmetry breaking we start with a simple model for a real
scalar field (or a theory to which we add a new field ), with a specific potential term:
1
L = (@µ )2 V( )
2
1 1 2 1
= (@µ )2 µ 2 4
(11.2)
2 2 4
Note that L is symmetric under ! and that is positive to ensure an absolute

minimum in the Lagrangian. We can investigate in some detail the two possibilities for
the sign of µ2 : positive or negative.
11.3.1 µ2 > 0: Free particle with additional interactions
To investigate the particle spectrum we look at the Lagrangian for

small perturbations around the minimum (vacuum). The vacuum
is at = 0 and is symmetric in . Using expression (11.1) we see
V(φ) that the Lagrangian describes a free particle with mass µ that has
an additional four-point self-interaction:
1 1 2 2 1 4
L= (@µ )2 µ
|2 {z 2 } | 4{z }
φ free particle, mass µ interaction
11.3. SIMPLE EXAMPLE OF SYMMETRY BREAKING 177
11.3.2 µ2 < 0: Introducing a particle with imaginary mass ?
The situation with µ2 < 0 looks strange since at first glance it

would appear to describe a particle with an imaginary mass.
However, if we take a closer look at the potential, we see that it
V(φ) v does not make sense to interpret the particle spectrum using the
φ field since perturbation theory around = 0 will not converge
(not a stable minimum) as the vacuum is located at:
r
µ2
η 0 = = v or µ2 = v2 (11.3)
As before, to investigate the particle spectrum in the theory, we have to look at small
perturbations around this minimum. To do this it is more natural to introduce a field
⌘ (simply a shift of the field) that is centered at the vacuum: ⌘ = v.
Rewriting the Lagrangian in terms of ⌘
Expressing the Lagrangian in terms of the shifted field ⌘ is done by replacing by ⌘ + v

in the original Lagrangian from equation (11.2):
1
Kinetic term: Lkin (⌘) = (@µ (⌘ + v)@ µ (⌘ + v))
2
1
= (@µ ⌘)(@ µ ⌘) , since @µ v = 0.
2
1 1
Potential term: V(⌘) = + µ2 (⌘ + v)2 + (⌘ + v)4
2 4
1 1 4
= v 2 ⌘ 2 + v⌘ 3 + ⌘ 4 v ,
4 4
where we used µ2 = v 2 from equation (11.3). Although the Lagrangian is still
symmetric in , the perturbations around the minimum are not symmetric in ⌘, i.e.
V( ⌘) 6= V(⌘). Neglecting the irrelevant 14 v 4 constant term and neglecting terms or
order ⌘ 2 we have as Lagrangian:
1 1 4 1 4
Full Lagrangian: L(⌘) = (@µ ⌘)(@ µ ⌘) v2⌘2 v⌘ 3 ⌘ v
2 4 4
1
= (@µ ⌘)(@ µ ⌘) v2⌘2
2
From section 11.2 we see that this describes the kinematics for a massive scalar particle:
1 2 p ⇣ p ⌘
m⌘ = v 2 ! m⌘ = 2 v 2 = 2µ2 Note: m⌘ > 0.
2
Executive summary on µ2 < 0 scenario
At first glance, adding a V ( ) term as in equation (11.2) to the Lagrangian implies

adding a particle with imaginary mass with a four-point self-interaction. However,
when examining the particle spectrum using perturbations around the vacuum, we see
that it actually describes a massive scalar particle (real, positive mass) with three-
and four-point self-interactions. Although the Lagrangian retains its original symmetry
(symmetric in ), the vacuum is not symmetric in the field ⌘: spontaneous symmetry
breaking. Note that we have added a single degree of freedom to the theory: a scalar
particle.
11.4 Breaking a global symmetry

In an existing theory we are free to introduce an additional complex scalar field: =
p1 ( 1 + i 2 ) (two degrees of freedom):
2
L = (@µ )⇤ (@ µ ) V( ) , with V( ) = µ2 ( ⇤
)+ ( ⇤
)2
0
Note that the Lagrangian is invariant under a U(1) global symmetry, i.e. under ! ei↵
since 0⇤ 0 ! ⇤ e i↵ e+i↵ = ⇤ .
The Lagrangian in terms of 1 and 2 is given by:

1 2 1 2 1 2 2 2 1 2 2 2
L( 1 , 2) = (@µ 1) + (@µ 2) µ( 1 + 2) ( 1 + 2)
2 2 2 4
There are again two distinct cases: µ2 > 0 and µ2 < 0. As in the previous section, we
investigate the particle spectrum by studying the Lagrangian under small perturbations
around the vacuum.
11.4.1 µ2 > 0
V(Φ)
This situation simply describes two massive scalar par-
ticles, each with a mass µ with additional interactions:
1 1 2 2 1 1 2 2
L( 1 , 2) = (@µ 1 )2 µ 1 + (@µ 2 )2 µ 2
φ2 |2 {z 2 } |2 {z 2 }
particle 1 , mass µ particle 2 , mass µ
φ1
+ interaction terms
11.4. BREAKING A GLOBAL SYMMETRY 179
11.4.2 µ2 < 0
When µ2 < 0 there is not a single vacuum located at

„ «
V(Φ)
0
0
, but an infinite number of vacua that satisfy:
q r
2 2 µ2
1 + 2 = =v
φ2
−v From the infinite number we choose 0 as 1 = v and

φ1
2 = 0. To see what particles are present in this model,
ξ the behaviour of the Lagrangian is studied under small
η
oscillations around the vacuum.
Looking at the symmetry we would use a ↵ei . When

looking at perturbations around this minimum it is nat-
ural to define the shifted fields ⌘ and ⇠, with: ⌘ = 1 v φ2 ξ
and ⇠ = 2 , which means that the (perturbations around [2]
the) vacuum are described by (see section 11.5.2): φ1 [1] η
circle of vacua
1
0 = p (⌘ + v + i⇠)
2
Using 2 = ⇤ = 12 [(v + ⌘)2 + ⇠ 2 ] and µ2 = v 2 we can rewrite the Lagrangian in

terms of the shifted fields.
1
Kinetic term: Lkin (⌘, ⇠) = @µ (⌘ + v i⇠)@ µ (⌘ + v + i⇠)
2
1 1
= (@µ ⌘)2 + (@µ ⇠)2 , since @µ v = 0.
2 2
Potential term: V(⌘, ⇠) = µ2 2 + 4

1 2 1
= v [(v + ⌘)2 + ⇠ 2 ] + [(v + ⌘)2 + ⇠ 2 ]2
2 4
1 4 1 1 1
= v + v 2 ⌘ 2 + v⌘ 3 + ⌘ 4 + ⇠ 4 + v⌘⇠ 2 + ⌘ 2 ⇠ 2
4 4 4 2
Neglecting the constant and higher order terms, the full Lagrangian can be written as:
1 1
L(⌘, ⇠) = (@µ ⌘)2 ( v 2 )⌘ 2 + (@µ ⇠)2 + 0 · ⇠ 2 + higher order terms
2
| {z } 2
| {z }
massive scalar particle ⌘ massless scalar particle ⇠
We can identify this as a massive ⌘ particle and a massless ⇠ particle:

p p
m⌘ = 2 v 2 = 2µ2 > 0 and m⇠ = 0
Unlike the ⌘-field, describing radial excitations, there is no ’force’ acting on oscillations
along the ⇠-field. This is a direct consequence of the U(1) symmetry of the Lagrangian
and the massless particle ⇠ is the so-called Goldstone boson.
Goldstone theorem:
For each broken generator of the original symmetry group, i.e. for each generator that
connects the vacuum states one massless spin-zero particle will appear.
Executive summary on breaking a global gauge invariant sym-

metry
Spontaneously breaking a continuous global symmetry gives rise to a massless (Gold-

stone) boson. When we break a local gauge invariance something special happens and
the Goldstone boson will disappear.
11.5 Breaking a local gauge invariant symmetry: the

Higgs mechanism
In this section we will take the final step and study what happens if we break a local
gauge invariant theory. As promised in the introduction, we will explore its consequences
using a local U(1) gauge invariant theory we know (QED). As we will see, this will allow
to add a mass-term for the gauge boson (the photon).
Local U(1) gauge invariance is the requirement that the Lagrangian is invariant under
0
! ei↵(x) . From the lectures on electroweak theory we know that this can be achieved
by switching to a covariant derivative with a special transformation rule for the vector
field. In QED:
@µ ! Dµ = @µ ieAµ [covariant derivatives]

0 1
Aµ = Aµ + @µ ↵ [Aµ transformation] (11.4)
e
The local U(1) gauge invariant Lagrangian for a complex scalar field is then given by:
1
L = (Dµ )† (Dµ ) Fµ⌫ F µ⌫ V( )
4
The term 14 Fµ⌫ F µ⌫ is the kinetic term for the gauge field (photon) and V ( ) is the extra
term in the Lagrangian we have seen before: V ( ⇤ ) = µ2 ( ⇤ ) + ( ⇤ )2 .
11.5. BREAKING A LOCAL GAUGE INVARIANT SYMMETRY: THE HIGGS MECHANISM181
11.5.1 Lagrangian under small perturbations

„ «
The situation µ2 > 0: we have a vacuum at 00 . The exact symmetry of the La-
grangian is preserved in the vacuum: we have QED with a massless photon and two
massive scalar particles 1 and 2 each with a mass µ.
In the situation µ2 < 0 we have an infinite number of vacua, each satisfying 21 + 22 =

µ2 / = v 2 . The particle spectrum is obtained by studying the Lagrangian under small
oscillations using the same procedure as for the continuous global symmetry from section
(11.4.2). Because of local gauge invariance some important di↵erences appear. Extra
terms will appear in the kinetic part of the Lagrangian through the covariant derivatives.
Using again the shifted fields ⌘ and ⇠ we define the vacuum as 0 = p12 [(v + ⌘) + i⇠].
Kinetic term: Lkin (⌘, ⇠) = (Dµ )† (Dµ )

= (@ µ + ieAµ ) ⇤ (@µ ieAµ )
= ... see Exercise 1
Potential term: V (⌘, ⇠) = v 2 ⌘ 2 , up to second order in the fields. See section 11.4.2.
The full Lagrangian can be written as:
1 1 1 1
L(⌘, ⇠) = (@µ ⌘)2 v 2 ⌘ 2 + (@µ ⇠)2 Fµ⌫ F µ⌫ + e2 v 2 A2µ evAµ (@ µ ⇠) +int.-terms
|2 {z } |2 {z } |4 {z 2 } | {z }
⌘-particle ⇠-particle photon field ?
(11.5)
At first glance: massive ⌘, massless ⇠ (as before) and also a mass term for the photon.
However, the Lagrangian also contains strange terms that we cannot easily interpret:
evAµ (@ µ ⇠). This prevents making an easy interpretation.
11.5.2 Rewriting the Lagrangian in the unitary gauge
In a local gauge invariance theory we see that Aµ is fixed up to a term @µ ↵ as can be

seen from equation (11.4). In general, Aµ and change simultaneously. We can exploit
this freedom, to redefine Aµ and remove all terms involving the ⇠ field.
Looking at the terms involving the ⇠-field, we see that we can rewrite them as:
 2
1 1 1 2 2 1 1 2 2 0 2
(@µ ⇠)2 evA (@µ ⇠) + e2 v 2 A2µ =
µ
e v Aµ (@µ ⇠) = e v (Aµ )
2 2 2 ev 2
This specific choice, i.e. taking ↵ = ⇠/v, is called the unitary gauge. Of course, when
choosing this gauge (phase of rotation ↵) the field changes accordingly (see first part
of section 11.1 and dropping terms of O(⇠ 2 , ⌘ 2 , ⇠⌘) ):
0 i ⇠/v i ⇠/v 1 i ⇠/v 1 1

!e = e p (v + ⌘ + i⇠) = e p (v + ⌘)e+i ⇠/v
= p (v + h)
2 2 2
Here we have introduced the real h-field. When writing down the full Lagrangian in
this specific gauge, we will see that all terms involving the ⇠-field will disappear and
that the additional degree of freedom will appear as the mass term for the gauge boson
associated to the broken symmetry.
11.5.3 Lagrangian in the unitary gauge: particle spectrum
Lscalar = (Dµ )† (Dµ ) V ( † )

1 1
= (@ µ + ieAµ ) p (v + h) (@µ ieAµ ) p (v + h) V ( † )
2 2
1 1 1 4 1 4
= (@µ h)2 + e2 A2µ (v + h)2 v 2 h2 vh3 h + v
2 2 4 4
1
Expanding (v + h)2 into 3 terms (and ignoring 4
v 4 ) we end up with:
1 1 2 2 2 1 1 4
= (@µ h)2 v 2 h2 + e v Aµ + e2 vA2µ h + e2 A2µ h2 vh3 h
|2 {z } |2 {z } | {z2 } | {z 4 }
massive scalar gauge field ( ) interaction Higgs Higgs self-
particle h with mass and gauge fields interactions
11.5.4 A few words on expanding the terms with (v + h)2
Expanding the terms in the Lagrangian associated to the vector field we see that we do
not only get terms proportional to A2µ , i.e. a mass term for the gauge field (photon), but
also automatically terms that describe the interaction of the Higgs field with the gauge
field. These interactions, related to the mass of the gauge boson, are a consequence of
the Higgs mechanism.
In our model, QED with a massive photon, when expanding 12 e2 A2µ (v + h)2 we get:
11.5. BREAKING A LOCAL GAUGE INVARIANT SYMMETRY: THE HIGGS MECHANISM183
1] 12 e2 v 2 A2µ : the mass term for the gauge field (photon)

Given equation (11.1) we see that m = ev.
γ
2] e 2
vA2µ h: photon-Higgs three-point interaction h
γ
3] 12 e2 A2µ h2 : photon-Higgs four-point interaction h
h
γ
Executive summary: breaking a local gauge invariant symmetry
We added a complex scalar field (2 degrees of freedom) to our existing theory and broke
the original symmetry by using a ’strange’ potential that yielded a large number of
vacua. The additional degrees of freedom appear in the theory as a mass term for the
gauge boson connected to the broken symmetry (m ) and a massive scalar particle (mh ).
Exercises
Exercise 11.1 (interaction terms)
(a) Compute the ’interaction terms’ as given in equation (11.5).

(b) Are the interaction terms symmetric in ⌘ and ⇠ ?
Exercise 11.2 (Toy-model with a massive photon)
(a) Derive expression (14.58) in Halzen & Martin.

Hint: you can either do the full computation or, much less work, just insert =
p1 (v + h) in the Lagrangian and keep Aµ unchanged.
2
(b) Show that in this model the Higgs boson can decay into two photons and that the
coupling h ! is proportional to m .
(c) Draw all Feynman vertices that are present in this model and show that Higgs
three-point (self-)coupling, or h ! hh, is proportional to mh .
(d) Higgs boson properties: how can you see from the Lagrangian that the Higgs boson
is a scalar (spin 0) particle ? What defines the ’charge’ of the Higgs boson ?
Exercise 11.3 (the potential part: V( † ))

Use in this exercise = p12 (v + h) and that is real (1 dimension).
(a) The normal Higgs potential: V ( † ) = µ2 2 + 4 .

Show that 12 m2h = v 2 , where ( 0 = v). How many vacua are there?
(b) Why is V ( † ) = µ2 2 + 3 not possible ?
How many vacua are there?
Terms / 6 are allowed since they introduce additional interactions that are not can-
celled by gauge boson interactions, making the model non-renormalizable. Just ignore
this little detail for the moment and compute the ’prediction’ for the Higgs boson mass.
† 4 2 2
(c) Use V ( ) = µ2 2 4
+ 3
6
, with µ2 < 0, > 0 and = µ2
.
q
3
Show that mh (new) = m (old),
2 h
with ’old’: mh for the normal Higgs potential.
Lecture 12
The Higgs mechanism in the

Standard Model
In this section we will apply the idea of spontaneous symmetry breaking from section
11 to the model of electroweak interactions. With a specific choice of parameters we
can obtain massive Z and W bosons while keeping the photon massless.
12.1 Breaking the local gauge invariant SU(2)L ⇥ U(1)Y

symmetry
To break the SU(2)L ⇥ U(1)Y symmetry we follow the ingredients of the Higgs mecha-
nism:
1) Add an isospin doublet:
✓ +
◆ ✓ ◆
1 1+i 2
= 0 =p
2 3+i 4
Since we would like the Lagrangian to retain all its symmetries, we can only add
SU(2)L ⇥ U(1)Y multiplets. Here we add a left-handed doublet (like the electron
neutrino doublet) with weak Isospin 12 . The electric charges of the upper and lower
component of the doublet are chosen to ensure that the hypercharge Y=+1. This
requirement is vital for reasons that will become more evident later.
2) Add a potential V( ) for the field that will break (spontaneously) the symmetry:
† † 2
V ( ) = µ2 ( )+ ( ) , with µ2 < 0
185
186 LECTURE 12. THE HIGGS MECHANISM IN THE STANDARD MODEL
The part added to the Lagrangian for the scalar field
Lscalar = (Dµ )† (Dµ ) V ( ),
where Dµ is the covariant derivative associated to SU(2)L ⇥ U(1)Y :
1 ~ µ + ig 0 1 Y Bµ
Dµ = @µ + ig ~⌧ · W
2 2
3) Choose a vacuum:
We have seen that any choice of the vacuum that breaks a symmetry will generate a
mass for the corresponding gauge boson. The vacuum we choose has 1 = 2 = 4 =0
and 3 = v: ✓ ◆
1 0
Vacuum = 0 = p
2 v+h
This vacuum as defined above is neutral since I = 12 , I3 = 12 and with our

choice of Y = +1 we have Q = I3 + 12 Y =0. We will see that this choice of
the vacuum breaks SU(2)L ⇥ U(1)Y ,but leaves U(1)EM invariant, leaving only the
photon massless. In writing down this vacuum we immediately went to the unitary
gauge (see section 11.5).
12.2 Checking which symmetries are broken in a

given vacuum
How do we check if the symmetries associated to the gauge bosons are broken ? In-
variance implies that ei↵Z 0 = 0 , with Z the associated ’rotation’. Under infinitesimal
rotations this means (1 + i↵Z) 0 = 0 ! Z 0 = 0.
What about the SU(2)L , U(1)Y and U(1)EM generators:
✓ ◆ ✓ ◆ ✓ ◆
0 1 1 0 1 v+h
SU(2)L : ⌧1 0 = p = +p 6= 0 ! broken
1 0 2 v+h 2 0
✓ ◆ ✓ ◆ ✓ ◆
0 i 1 0 i v+h
⌧2 0 = p = p 6= 0 ! broken
i 0 2 v+h 2 0
✓ ◆ ✓ ◆ ✓ ◆
1 0 1 0 1 0
⌧3 0 = p = p 6= 0 ! broken
0 1 2 v+h 2 v+h
✓ ◆ ✓ ◆
1 0 1 0
U(1)Y : Y 0 = Y 0 p = +p 6= 0 ! broken
2 v+h 2 v+h
12.3. SCALAR PART OF THE LAGRANGIAN: GAUGE BOSON MASS TERMS187
This means that all 4 gauge bosons (W1 , W2 , W3 and B) acquire a mass through the
Higgs mechanism. In the lecture on electroweak theory we have seen that the W1 and
W2 fields mix to form the charged W + and W bosons and that the W3 and B field will
mix to form the neutral Z-boson and photon.
W1 W2 W3 B
| {z } | {z }
W+ and W bosons Z-boson and
When computing the masses of these mixed physical states in the next sections, we
will see that one of these combinations (the photon) remains massless. Looking at the
symmetries we can already predict this is the case. For the photon to remain massless
the U(1)EM symmetry should leave the vacuum invariant. And indeed:
✓ ◆ ✓ ◆
1 1 0 1 0
U(1)EM : Q 0 = (⌧3 + Y ) 0 = p = 0 ! unbroken
2 0 0 2 v+h
It is not so strange that U(1)EM is conserved as the vacuum is neutral and we have:
0
0 ! ei↵Q 0
0 = 0
Breaking of SU(2)L ⇥ U(1)Y : looking a bit ahead
1) W1 and W2 mix and will form the massive a W+ and W bosons.

2) W3 and B mix to form massive Z and massless .
3) Remaining degree of freedom will form the mass of the scalar particle (Higgs boson).
12.3 Scalar part of the Lagrangian: gauge boson

mass terms
Studying the scalar part of the Lagrangian
To obtain the masses for the gauge bosons we will only need to study the scalar part of
the Lagrangian:
Lscalar = (Dµ )† (Dµ ) V ( ) (12.1)
The V ( ) term will again give the mass term for the Higgs boson and the Higgs self-
interactions. The (Dµ )† (Dµ ) terms:
 ✓ ◆
1 1
~ µ + ig Y Bµ p
0 1 0
Dµ = @µ + ig ~⌧ · W
2 2 2 v+h
will give rise to the masses of the gauge bosons (and the interaction of the gauge bosons
with the Higgs boson) since, as we discussed in section 11.5.4, working out the (v + h)2 -
terms from equation (12.1) will give us three terms:
1) Masses for the gauge bosons (/ v 2 )

2) Interactions gauge bosons and the Higgs (/ vh) and (/ h2 )
In the exercises we will study the interactions of the Higgs boson and the gauge boson
(the terms in 2)) in detail, but since we are here primarily interested in the masses of
the vector bosons we will only focus on 1):
 ✓ ◆
1 1 ~ 01 0
(Dµ ) = p ig ~⌧ · Wµ + ig Y Bµ
2 2 2 v
i h 0
i✓ 0 ◆
= p g(⌧1 W1 + ⌧2 W2 + ⌧3 W3 ) + g Y Bµ
8 v
i h 0
i✓ 0 ◆
Y 0 Bµ 0
= p g 0
W1
W1
0 + iW2 0 iW2
0 + 0 W3 0
W3 +g 0 Y 0 Bµ
8 v
✓ 0 ◆ ✓ ◆
i gW3 + g Y 0 Bµ g(W1 iW2 ) 0
= p 0
8 g(W 1 + iW 2 ) gW 3 + g Y 0 Bµ v
✓ ◆
iv g(W1 iW2 )
= p 0
8 gW3 + g Y 0 Bµ
We can then also easily compute (Dµ )† : (Dµ )† = piv8 g(W1 + iW2 ) , ( gW3 + g Y 0 Bµ )
0
and we get the following expression for the kinetic part of the Lagrangian:
µ † 1 2h 2 2 2 0 2
i
(D ) (Dµ ) = v g (W1 + W2 ) + ( gW3 + g Y 0 Bµ ) (12.2)
8
12.3.1 Rewriting (Dµ )† (Dµ ) in terms of physical gauge bosons
Before we can interpret this we need to rewrite this in terms of W+ , W , Z and since
that are the gauge bosons that are observed in nature.
1] Rewriting terms with W1 and W2 terms: charged gauge

bosons W+ and W
When discussing the charged current interaction on SU(2)L doublets we saw that the
charge raising and lowering operators connecting the members of isospin doublets were
⌧+ and ⌧ , linear combinations of ⌧1 and ⌧2 and that each had an associated gauge
boson: the W+ and W .
✓ ◆ W
+ ν
1 0 1
⌧+ = (⌧1 + i⌧2 ) =
2 0 0 e−
✓ ◆
1 0 0 − ν
⌧ = (⌧1 i⌧2 ) = W
2 1 0
e−
12.3. SCALAR PART OF THE LAGRANGIAN: GAUGE BOSON MASS TERMS189
We can rewrite W1 , W2 terms as W + , W using W ± = p1 (W1 ⌥ iW2 ). In particu-

2
lar, 12 (⌧1 W1 + ⌧2 W2 ) = p12 (⌧+ W + + ⌧ W ).
Looking at the terms involving W1 and W2 in the Lagrangian in equation (12.2), we see
that:
2 2
g 2 (W12 + W22 ) = g 2 (W + + W ) or, alternatively, 2g 2 W + W (12.3)
2] Rewriting terms with W3 and Bµ terms: neutral gauge bosons

Z and
✓ ◆✓ ◆
0 2 g2 gg 0 Y 0 W3
( gW3 + g Y 0 Bµ ) = (W3 , Bµ )
gg 0 Y 0 g02 Bµ
When looking at this expression there are some important things to note, especially
related to the role of the hypercharge of the vacuum, Y 0 :
1 Only if Y 0 6= 0, the W3 and Bµ fields mix.
2 If Y 0 = ± 1, the determinant of the mixing matrix vanishes and one of the

combinations will be massless (the coefficient for that gauge field squared is 0). In
our choice of vacuum we have Y 0 = +1 (see Exercise 4 why that is a good idea).
In the rest of our discussion we will drop the term Y 0 and simply use its value of
1.
The two eigenvalues and eigenvectors are given by [see Exercise 3]:
eigenvalue eigenvector
✓ ◆
1 g0 1
=0 ! p = p (g 0 W3 + gBµ ) = Aµ photon( )
2
g +g 0 2 g2
g +g 0 2
✓ ◆
2 1 g 1
= (g 2 + g 0 ) ! p 0 = p (gW3 g 0 Bµ ) = Zµ Z-boson (Z)
2
g +g 0 2 g 2
g +g 0 2
Looking at the terms involving W3 and B in the Lagrangian we see that:
2
( gW3 + g 0 Y 0 Bµ )2 = (g 2 + g 0 )Zµ2 + 0 · A2µ (12.4)
3] Rewriting Lagrangian in terms of physical fields: masses of

the gauge bosons
Finally, by combining equation (12.3) and (12.4) we can rewrite the Lagrangian from
equation (12.2) in terms of the physical gauge bosons:
1 2
(Dµ )† (Dµ ) = v 2 [g 2 (W + )2 + g 2 (W )2 + (g 2 + g 0 )Zµ2 + 0 · A2µ ] (12.5)
8
12.4 Masses of the gauge bosons
12.4.1 Massive charged and neutral gauge bosons
As a general mass term for a massive gauge boson V has the form 12 MV2 Vµ2 , from equation
(12.5) we see that:
1
MW + = MW = vg
2 q
1
MZ = v (g 2 + g 0 2 )
2
Although since g and g 0 are free parameters, the SM makes no absolute predictions for
MW and MZ , it has been possible to set a lower limit before the W - and Z-boson were
discovered (see Exercise 2). The measured values are MW = 80.4 GeV and MZ = 91.2
GeV.
Mass relation W and Z boson:
Although there is no absolute prediction for the mass of the W- and Z-boson, there is
a clear prediction on the ratio between the two masses. From discussions in QED we
know the photon couples to charge, which allowed us to relate e, g and g 0 (see Exercise
3):
e = g sin(✓W ) = g 0 cos(✓W ) (12.6)
In this expression ✓W is the Weinberg angle, often used to describe the mixing of the
W3 and Bµ -fields to form the physical Z boson and photon. From equation (12.6) we
see that g 0 /g = tan(✓W ) and therefore:
1
MW vg
= 1 p2 = cos(✓W )
MZ v g 2 + g02
2
12.5. MASS OF THE HIGGS BOSON 191
This predicted ratio is often expressed as the so-called ⇢-(Veltman) parameter:

2
MW
⇢= =1
MZ2 cos2 (✓W )
The current measurements of the MW , MZ and ✓W confirm this relation.
12.4.2 Massless neutral gauge boson ( ):
Similar to the Z boson we have now a mass for the photon: 12 M 2 = 0, so:
M = 0. (12.7)
12.5 Mass of the Higgs boson

Looking at the mass term for the scalar particle, the mass of the Higgs boson is given
by: p
mh = 2 v 2
Although v is known (v ⇡ 246 GeV, see below), since is a free parameter, the mass of
the Higgs boson is not predicted in the Standard Model.
νµ
Extra: how do we know v ?: µ GF
s νe
2 G
g GF 1 Fermi: / pF e
Muon decay: 2
= p !v= p 2
νµ
8MW 2 2GF
µ g
1 5
We used MW = 2
vg.Given GF = 1.166 · 10 , we νe
g2 W g
see that v = 246 GeV. This energy scale is known EW: / 8MW2 e
as the electroweak scale.
Exercises
Exercise 12.1 (Higgs - Vector boson couplings)
In the lecture notes we focussed on the masses of the gauge bosons, i.e. part 1) when
expanding the ((v + h)2 )-terms as discussed in Section 11.5.4 and 12.3. Looking now at
the terms in the Lagrangian that describe the interaction between the gauge fields and
the Higgs field, show that the four vertex factors describing the interaction between the
Higgs boson and gauge bosons: hWW, hhWW, hZZ, hhZZ are given by:
MV2 MV2
3-point: 2i v
g µ⌫ and 4-point: 2i v2
g µ⌫ , with (V = W,Z).
Note: A vertex factor is obtained by multiplying the term involving the interacting
fields in the Lagrangian by a factor i and a factor n! for n identical particles in the
vertex.
Exercise 12.2 (History: lower limits onpMW and MZ )

Use the relations e = g sin ✓W and GF = (v 2 2) 1 to obtain lower limits for the masses
of the W and Z boson assuming that you do not know the value of the weak mixing
angle.
Exercise 12.3 (Electroweak mixing: (Wµ3 , Bµ ) ! (Aµ , Zµ ))

The mix between the Wµ3 and Bµ fields in the lagrangian can be written in a matrix
notation: ✓ 2 ◆✓ 3 ◆
3 g gg 0 Wµ
(Wµ , Bµ ) 0 02
gg g Bµ
(a) Show that the eigenvalues of the matrix are 1 = 0 and 2 = (g 2 + g 0 2 ).
(b) Show that these eigenvalues correspond to the two eigenvectors:
1 1
V1 = p 2
(g 0 Wµ3 + gBµ ) ⌘ Aµ and V2 = p (gWµ3 g 0 Bµ ) ⌘ Zµ
2
g +g 0 g2 + g02
(c) bonus: Imagine that we would have chosen Y 0 = 1. What, in that scenario,
0
0 0
would be the (mass-)eigenvectors Aµ and Zµ , the ’photon’ and ’Z-boson’ ? In such
a model, what would be their masses ? Compare them to those in the Standard
Model.
Exercise 12.4 (A closer look at the covariant derivative)

The covariant derivative in the electroweak theory is given by:
Y
Dµ = @µ + ig 0 Bµ + ig T~ · W
~µ
2
(a) Looking only at the part involving Wµ3 and Bµ show that:
✓ ◆ ✓ ◆
gg 0 Y 1 2 02 Y
Dµ = @µ + iAµ p T3 + + iZµ p g T3 g
g 02 + g 2 2 g 02 + g 2 2
12.5. MASS OF THE HIGGS BOSON 193
(b) Make also a final interpretation step for the Aµ part and show that:
gg 0 Y
p = e and T3 + = Q, the electric charge.
g 02 + g 2 2
(c) bonus: Imagine that we would have chosen Y 0 = 1. Show explicitly that in
0
that case the photon does not couple to the electric charge.
Exercise 12.5 (Gauge bosons in a model with an SU(2)L symmetry)

Imagine a system described by a local SU(2)L gauge symmetry (iso-spin only) in which
all gauge bosons are be massive. Note that this is di↵erent from the SU(2)L ⇥U(1)Y
symmetry of the SM involving also hypercharge. In this alternative model:
(a) Explain why the Higgs field needs to be an SU(2)L doublet.
(b) How many gauge bosons are there and how many degrees of freedom does have?
(c) Determine the masses of the gauge bosons in this model.
(d) What property of the particles do the gauge bosons couple to and what defines
the ’charge’ of the gauge bosons themselves?
Lecture 13
Fermion masses, Higgs decay and

limits on mh
In this section we discuss how fermions acquire a mass and use our knowledge on the
Higgs coupling to fermions and gauge bosons to predict how the Higgs boson decays as a
function of its mass. Even though the Higgs boson has been discovered, we also discuss
what theoretical information we have on the mass of the Higgs boson as it reveals the
impact on the Higgs boson at higher energy scales (evolution of the universe).
13.1 Fermion masses
In section 11 we saw that terms like 12 Bµ B µ and m ¯ were not gauge invariant. Since
these terms are not allowed in the Lagrangian, both gauge bosons and fermions are
massless. In the previous section we have seen how the Higgs mechanism can be used
to accommodate massive gauge bosons in our theory while keeping the local gauge in-
variance. As we will now see, the Higgs mechanism can also give fermions a mass: ’twee
vliegen in een klap’.
Chirality and a closer look at terms like m¯
A term like m ¯ = m[ ¯L R + ¯R L ], i.e. a decomposition in chiral states (see

exercise 1). Such a term in the Lagrangian is not gauge invariant since the left handed
„ «
fermions form an isospin doublet (for example ⌫e ) and the right handed fermions
L
form isospin singlets like eR . They transform di↵erently under SU(2)L ⇥ U(1)Y .
0 iW~ ·T~ +i↵Y
left handed doublet = L ! L = Le
0 i↵Y
right handed singlet = R ! R = Re
195
196 LECTURE 13. FERMION MASSES, HIGGS DECAY AND LIMITS ON MH
This means that the term is not invariant under all SU(2)L ⇥ U(1)Y ’rotations’.
Constructing an SU(2)L ⇥ U(1)Y invariant term for fermions
If we could make a term in the Lagrangian that is a singlet under SU(2)L and U(1)Y
, it would remain invariant. This can be done using the complex (Higgs) doublet we
introduced in the previous section. It can be shown that the Higgs has exactly the right
quantum numbers to form an SU (2)L and U (1)Y singlet in the vertex: ¯
f L R,
where f is a so-called Yukawa coupling.
Executive summary: - a term: / ¯L R is not invariant under SU(2)L ⇥ U(1)Y

- a term: / ¯L R is invariant under SU(2)L ⇥ U(1)Y
We have constructed a term in the Lagrangian that couples the Higgs doublet to the
fermion fields:
Lfermion-mass = ¯
f[ L
¯ ¯ L]
R + R (13.1)
When we write out this term we’ll see that this does not only describe an interaction
between the Higgs field and fermion, but that the fermions will acquire a finite mass if
„ «
the -doublet has a non-zero expectation value. This is the case as 0 = p12 v +0 h as
before.
13.1.1 Lepton masses
 ✓ ◆ ✓ ◆
1 0 ⌫
Le = ep (¯
⌫ , ē)L eR + ēR (0, v + h)
2 v + h e L
e (v + h)
= p [ēL eR + ēR eL ]
2
e (v + h)
= p ēe
2
v
= pe ēe pe hēe
2 2
| {z } | {z }
electron mass term electron-higgs interaction
ev
me = p pe / me
2 2
A few side-remarks:
13.1. FERMION MASSES 197
p m
1) The Yukawa coupling is often expressed as f = 2 vf and the coupling of the
m
fermion to the Higgs field is pf2 = vf , so proportional to the mass of the fermion.
2) The mass of the electron is not predicted since e is a free parameter. In that
sense the Higgs mechanism does not say anything about the electron mass itself.
3) The coupling of the Higgs boson to electrons is very small:

gme
The coupling of the Higgs boson to an electron-pair (/ mve = 2M W
) is very small
compared to the coupling of the Higgs boson to a pair of W-bosons (/ gMW ).
2
✓ ◆2
(h ! ee) eeh gme /2MW m2e 21
/ 2
= = 4
⇡ 1.5 · 10
(h ! W W ) WWh gMW 4MW
13.1.2 Quark masses
The fermion mass term Ldown = f ¯L R (leaving out the hermitian conjugate term
¯R ¯ L for clarity) only gives mass to ’down’ type fermions, i.e. only to one of the isospin
doublet components. To give the neutrino a mass and give mass to the ’up’ type quarks
(u, c, t), we need another term in the Lagrangian. Luckily it is possible to compose a
new term in the Lagrangian, using again the complex (Higgs) doublet in combination
with the fermion fields, that is gauge invariant under SU(2)L ⇥ U(1)Y and gives a mass
to the up-type quarks. The mass-term for the up-type fermions takes the form:
Lup = ¯L ˜c R + h.c., with
✓ ◆
˜c = ⇤ 1 (v + h)
i⌧2 = p (13.2)
2 0
Mass terms for fermions (leaving out h.c. term):

✓ ◆
¯ ¯ 0
down-type: d (ūL , dL ) dR = d (ūL , dL ) dR = dv d¯L dR
v
✓ ◆
¯ ˜c ¯ v
up-type: u (ūL , dL ) dR = u (ūL , dL ) uR = uv ūL uR
0
As we will discuss now, this is not the whole story. If we look more closely we’ll see that
we can construct more fermion-mass-type terms in the Lagrangian that cannot easily
be interpreted. Getting rid of these terms is at the origin of quark mixing.
13.2 Yukawa couplings and the origin of Quark Mix-

ing
This section will discuss in full detail the consequences of all possible allowed quark
’mass-like’ terms and study the link between the Yukawa couplings and quark mixing in
the Standard Model: the di↵erence between mass eigenstates and flavour eigenstates.
If we focus on the part of the SM Lagrangian that describes the dynamics of spinor
(fermion) fields , the kinetic terms, we see that:
Lkinetic = i ¯(@ µ µ) ,
where ¯ ⌘ † 0 and the spinor fields . It is instructive to realise that the spinor fields
are the three fermion generations can be written in the following five (interaction)
representations:
I
general spinor field (color, weak iso-spin, hypercharge)
1) left handed quarks QILi (3, 2, +1/3)

2) right handed up-type quarks uIRi (3, 1, +4/3)
3) right handed down-type quarks dIRi (3, 1, +1/3)
4) left handed fermions LILi (1, 2, 1)
I
5) right handed fermions lRi (1, 1, 2)
In this notation, QILi (3, 2, +1/3) describes an SU(3)C triplet, SU(2)L doublet, with hy-
percharge Y = 1/3. The superscript I implies that the fermion fields are expressed in the
interaction (flavour) basis. The subscript i stands for the three generations (families).
Explicitly, QILi (3, 2, +1/3) is therefor a shorthand notation for:
✓ ◆ ✓ ◆ ✓ I I I ◆ ✓ I I I ◆
uIg , uIr , uIb uIg , uIr , uIb c g , cr , cb tg , tr , tb
QILi (3, 2, +1/3) = = , , .
dIg , dIr , dIb i
dIg , dIr , dIb I I I
sg , sr , sb bIg , bIr , bIb
We saw that using the Higgs field we could construct terms in the Lagrangian of the
form given in equation (13.1). For up and down type fermions (leaving out the hermitian
conjugate term) that would allow us to write for example:
Lquarks = ⇤down ¯L R ⇤up ¯L ˜c R

v v
= ⇤down p d¯I dI ⇤up p u¯I uI ,
2 2
= ¯I
md d d I ¯I I
mu u u ,
where the strength of the interactions between the Higgs and the fermions, the so-called
Yukawa couplings, had again to be added by hand.
13.2. YUKAWA COUPLINGS AND THE ORIGIN OF QUARK MIXING 199
This looks straightforward, but there is an additional complication when you realize
that in the most general realization the ⇤’s are matrices. This will introduce mixing
between di↵erent flavours as we will see a little bit later. In the most general case, again
leaving out the h.c., the expression for the fermion masses is written as:
LYukawa = Yij Li Rj
= Yijd QILi dIRj + Yiju QILi ˜c uIRj + Yijl LILi I

lRj , (13.3)
where the last term is the mass term for the charged leptons. The matrices Yijd , Yiju and
Yijl are arbitrary complex matrices that connect the flavour eigenstate since also terms
like Yuc will appear. These terms have no easy interpretation:
v v v
LYukawa = ... + ⇤dd p u¯I uI ⇤us p u¯I sI ⇤ss p s¯I sI + . . . (13.4)
2 2 2
| {z } | {z } | {z }
mass term down quark ?? mass term strange quark
To interpret the fields in the theory as physical particles, the fields in our model should
have a well-defined mass. This is not the case in equation (13.4). If we write out all
Yukawa terms in the Lagrangian we realize that it is possible to re-write them in terms
of mixed fields that do have a well-defined mass. These states are the physical particles
in the theory
Writing out the full Yukawa terms:
Since this is the crucial part of flavour physics, we spell out the term Yijd QILi dIRj
explicitly and forget about the other 2 terms in expression (13.3):
✓ +
◆
Yijd QILi dIRj = Yijd (up-type down-type)IiL (down-type)IRj =
0 ✓ +
◆ ✓ + ◆ ✓ +
◆ 1
B Y11 (u d)IL 0 Y12 (u d)LI
0 Y13 (u d)LI
0 C 0 I 1
B ✓ ◆ ✓ + ◆ ✓ ◆ C dR
B + + C
B Y21 (c s) I
Y22 (c s)LI
Y23 (c s)LI C · @ sIR A
B L 0 0 0 C
B ✓ ◆ ✓ + ◆ ✓ ◆ C bIR
@ + + A
Y31 (t b)IL 0 Y32 (t b)IL 0 Y33 (t b)IL 0
After symmetry breaking we get the following mass terms for the fermion fields:
Lquarks d I
Yukawa = Yij QLi dIRj + Yiju QILi ˜ uIRj
v v
= Yijd dILi p dIRj + Yiju uILi p uIRj + ...
2 2
d I I u I I
= Mij dLi dRj + Mij uLi uRj + , (13.5)
where we omitted the corresponding interaction terms of the fermion fields to the Higgs
field, q̄qh(x) and the hermitian conjugate terms. Note that the d’s and u’s in equation
(13.5) still each represent the three down-type and up-type quarks respectively, so the
’mixed’-terms are still there. To obtain mass eigenstates, i.e. states with proper mass
terms, we should diagonalize the matrices M d and M u . We do this with unitary matrices
V d as follows:
d
Mdiag = VLd M d VRd†
u
Mdiag = VLu M d VRu†
Using the requirement that the matrices V are unitary (VLd† VLd = ) and leaving out
again the hermitian conjugate terms the Lagrangian can now be expressed as follows:
Lquarks I d I I u I
Yukawa = dLi Mij dRj + uLi Mij uRj + ...
= dILi VLd† VLd Mijd VRd† VRd dIRj + uILi VLu† VLu Miju VRu† VRu uIRj + ...
= dLi (Mijd )diag dRj + uLi (Miju )diag uRj + ...,
where in the last line the matrices V have been absorbed in the quark states. Note that
the up-type and down-type fields are now no longer the interaction states uI and dI , but
are now ’simply’ u and d. A bit more explicit, we now have the following quark mass
eigenstates:
dLi = (VLd )ij dILj dRi = (VRd )ij dIRj

uLi = (VLu )ij uILj uRi = (VRu )ij uIRj ,
which allowed us to express the quark interaction eigenstates dI , uI as quark mass

eigenstates d, u. It is now interesting to see how various parts of the Standard Model
Langrangian change when you write them either in the mass or the interaction eigen-
states.
Rewriting interaction terms using quark mass eigenstates
The interaction terms are obtained by imposing gauge invariance by replacing the partial
derivative by the covariant derivate
Lkinetic = i ¯(Dµ µ) , (13.6)
~ µ . The ⌧ ’s are the Pauli

with the covariant derivative defined as Dµ = @ µ + ig 12 ~⌧ · W
µ µ
matrices and Wi and B are the three weak interaction bosons and the single hyper-
charge boson, respectively. It is very natural to write the charged current interaction
between the (left-handed) iso-spin doublet interaction eigenstates that are connected by
13.2. YUKAWA COUPLINGS AND THE ORIGIN OF QUARK MIXING 201
W-bosons:
i
Lkinetic, weak (QL ) = iQILi µ @ µ + gWiµ ⌧i QILi
2
✓ ◆I
I µ i µ u
= i(u d)iL µ @ + gWi ⌧i
2 d iL
g g
= iuIiL µ @ µ uIiL + idIiL µ @ µ dIiL p uIiL µW
µ I
diL p dIiL µW
+µ I
uiL + ...
2 2
, where we used W ± = p1 (W1 ⌥iW2 ), see Section 12.

2
If we now express the Lagrangian in terms of the quark mass eigenstates d, u instead
of the weak interaction eigenstates dI , uI , the ’price’ to pay is that the quark mixing
between families (i.e. the o↵-diagonal elements) appear in the charged current inter-
action as each of the interaction fields is now replaced by a combination of the mass
eigenstates:
g g
Lkinetic, cc (QL ) = p uIiL µ W µ dIiL + p dIiL µ W +µ uIiL + ...
2 2
g g
= p uiL (VLu VL )ij µ W µ diL + p diL (VLd VLu† )ij
d†
µW
+µ
uiL + ...
2 2
The CKM matrix
The combination of matrices (VLd VLu† )ij , a unitary 3⇥3 matrix is known under the short-
hand notation VCKM , the famous Cabibbo-Kobayashi-Maskawa (CKM) mixing matrix.
By convention, the interaction eigenstates and the mass eigenstates are chosen to be
equal for the up-type quarks, whereas the down-type quarks are chosen to be rotated,
going from the interaction basis to the mass basis:
uIi = uj
dIi = VCKM dj
or explicitly:
0 1 0 10 1
dI Vud Vus Vub d
@ sI A = @ Vcd Vcs Vcb A @ s A (13.7)
bI Vtd Vts Vtb b
From the definition of VCKM it follows that the

transition from a down-type quark to an up-type u u
quark is described by Vud , whereas the transition d V d V
from an up type quark to a down-type quark is
⇤
described by Vud . A separate lecture describes in W
− W
detail how VCKM allows for CP-violation in the SM.
Note on lepton masses
We should note here that in principle a similar matrix exists that connects the lep-
ton flavour and mass eigenstates. In this case, contrary to the quarks, the down-type
interaction doublet-states (charged leptons) are chosen to be the same as the mass eigen-
states. The rotation between mass and interaction eigenstates is in the neutrino sector.
This matrix is known as the Pontecorvo-Maki-Nakagawa-Sakata (PMNS) matrix and
has a completely di↵erent structure than the one for quarks. Just like for the CMS
matrix, the origin of the observed patterns are completely unknown. A last thing to
remember: neutrino interaction eigenstates are known as ⌫e , ⌫µ and ⌫⌧ , whereas the
physical particles, the mass eigenstates, are ⌫1 , ⌫2 and ⌫3 .
13.3 Higgs boson decay

It is interesting to study details of the Higgs boson properties like its coupling to fermions
and gauge bosons as that determines if and how the Higgs boson is produced in exper-
iments and what the event topology will be. In Section 13.3.3 we list all couplings and
as an example we’ll compute the decay rate fractions of a Higgs boson into fermions as
a function of it’s unknown mass in Section 13.3.1.
13.3.1 Higgs boson decay to fermions
Now that we have derived the coupling of fermions and gauge bosons to the Higgs field,
we can look in more detail at the decay of the Higgs boson.
The general expression for the two-body decay rate:
d |M|2
= |pf | S, (13.8)
d⌦ 32⇡ 2 s
with M the matrix element, |pf | the momentum of thepproduced particles andpS = n!1
for n identical particles. In a two-body decay we have s = mh and |pf | = 12 s (see
exercise 2). Since the Higgs boson is a scalar particle, the Matrix element takes a simple
form:
imf
iM = ū(p1 ) v( p2 ) h
v
imf
iM† = v̄( p2 ) u(p1 )
v
13.3. HIGGS BOSON DECAY 203
Since there are no polarizations for the scalar Higgs boson, computing the Matrix ele-
ment squared is ’easy’:
⇣ m ⌘2 X
2 f
M = (v̄)s2 ( p2 )us1 (p1 )(ū)s1 (p1 )vs2 ( p2)
v s1 ,s2
⇣ m ⌘2 X X
f
= us1 (p1 )(ū)s1 (p1 ) v̄s2 ( p2 )vs2 ( p2 )
v s1 s1
⇣ m ⌘2
f
= Tr ((6 p1 + mf )( 6 p2 mf ))
v
⇣ m ⌘2 ⇥ ⇤
f
= Tr(6 p1 6 p2 ) m2f Tr( ))
v
⇣ m ⌘2 ⇥ ⇤
f
= 4p1 · p2 4m2f
v
use: s = (p1 p2 )2 = p21 + p22 2p1 · p2 and since p21 = p22 = m2f
and s = m2h we have m2h = 2m2f 2p1 · p2
⇣ m ⌘2 ⇥ ⇤
f
= 2m2h 8m2f
v s
⇣ m ⌘2 4m2f
f
= 2m2h 2 , with = 1
v m2h
Including the number of colours (for quarks) we finally have:

⇣ m ⌘2
f
M2 = 2m2h 2 Nc
v
Decay rate:
p p
Starting from equation (13.8) and using M2 (above), |pf | = 12 s, S=1 and s = mh
we get:
d |M|2 Nc m h ⇣ m f ⌘ 2 3
= |p f | S =
d⌦ 32⇡ 2 s 32⇡ 2 v
R
Doing the angular integration d⌦ = 4⇡ we finally end up with:
Nc 2
(h ! f f¯) = m mh 3
f.
8⇡v 2 f
13.3.2 Higgs boson decay to gauge bosons
The decay ratio to gauge bosons is a bit more tricky, but is explained in great detail in
Exercise 5.
13.3.3 Review Higgs boson couplings to fermions and gauge

bosons
A summary of the Higgs boson couplings to fermions and gauge bosons.
h p 4m2f
(h ! f f¯) = Nc
8⇡v 2
m2f mh 1 x , with x = m2h
g2
p
h (h ! V V ) = 64⇡MW2 m3h SV V (1 x + 34 x2 ) 1 x
4MV2
, with x = m2h
and SW W,ZZ = 1, 12 .
The decay of the Higgs boson to two o↵-shell gauge bosons is given by:
3M 4 0
(h ! V V ⇤ ) = 32⇡2Vv4 mh V R(x) , with
h 0 0
W = 1, Z = 12
7 10
9
sin2 ✓W + 40
27
sin4 ✓W , with
2)
R(x) = 3(1 p8x+20x
4x 1
acos 2x3x 1
3/2
1 x
2x
(2 13x + 47x2 )
3 2
2
(1 6x + 4x ) ln(x)
Since the coupling of the Higgs boson to gauge bosons is so much larger than that to
fermions, the Higgs boson decays to o↵-shell gauge bosons even though MV ⇤ + MV <
2MV . The increase in coupling ’wins’ from the Breit-Wigner suppression. For example:
at mh = 140 GeV, the h ! W W ⇤ is already larger than h ! bb̄.
γ P 2
↵2 4 (f )
h (h ! )= 256⇡ 3 v 2
m3h 3
2
f Nc ef 7
γ
γ , where ef is the fermion’s electromagnetic charge.
h Note: - WW contribution ⇡ 5 times top contribution
γ
- Some computation also gives h ! Z
 ✓ ◆ 2
↵s2 95 7Nf ↵s
(h ! gluons) = m3 1+ + ...
72⇡ 3 v 2 h 4 6 ⇡
h
Note: - The QCD higher order terms are large.
- Reading the diagram from right to left you see the dominant
production mechanism of the Higgs boson at the LHC.
13.4. THEORETICAL BOUNDS ON THE MASS OF THE HIGGS BOSON 205
13.3.4 Higgs branching fractions
2
10
Branching fraction (%)

--
bb WW
Having computed the branching ratios to
fermions and gauge bosons in Section 13.3.1 ZZ
+ -
and Section 13.3.2 we can compute the rel- 10 τ τ
--
ative branching fractions for the decay of a cc
Higgs boson as a function of its mass. The

gg
distribution is shown here. 1
60 80 100 120 140 160 180 200 220

Higgs mass (GeV/c2)
13.4 Theoretical bounds on the mass of the Higgs

boson
Although the Higgs mass is not predicted within the minimal SM, there are theoretical
upper and lower bounds on the mass of the Higgs boson if we assume there is no new
physics between the electroweak scale and some higher scale called ⇤. In this section
we present a quick sketch of the various arguments and present the obtained limits.
As the Higgs boson mass is now known to quite some precision this section might
feel strange and unnecessary to revisit. Since similar arguments are used to obtain
theoretical limits on the mass of hypothetical particles that are predicted in models
that go beyond the Standard Model it is good to understand the various elements that
enter in such a discussion.
13.4.1 Unitarity
In the absence of a scalar field the amplitude for elastic scattering of longitudinally
polarised massive gauge bosons (e.g. WL+ WL ! WL+ WL ) diverges quadratically with
the centre-of-mass energy when calculated in perturbation theory and at an energy of
1.2 TeV this process violates unitarity. In the Standard Model, the Higgs boson plays
an important role in the cancellation of these high-energy divergences. Once diagrams
involving a scalar particle (the Higgs boson) are introduced in the gauge boson scattering
mentioned above, these divergences are no longer present and the theory remains unitary
and renormalizable. Focusing on solving these divergences alone also yields most of the
Higgs bosons properties. This cancellation only works however if the Higgs boson is not
too heavy. By requiring that perturbation theory remains valid an upper limit on the
Higgs mass can be extracted. With the requirement of unitarity and using all (coupled)
gauge boson scattering processes it can be shown that:

s p
4⇡ 2
mh < ⇠ 700 GeV/c2 .
3GF
It is important to note that this does not mean that the Higgs boson can not be heavier
than 700 GeV/c2 . It only means that for heavier Higgs masses, perturbation theory is
not valid and the theory is not renormalisable.
This number comes from an analysis that uses a partial wave decomposition for the
matrix element M, i.e.:
l=1
X
d 1
= M2 , with M = 16⇡ (2l + 1)Pl (cos ✓)al ,
d⌦ 64⇡s l=0
where Pl are Legendre polynomials and al are spin-l partial waves. Since (WL+ WL +
ZL + ZL + HH)2 is well behaved, it must respect unitarity, i.e. |ai | < 1 or |Re(ai )|  0.5.
As the largest amplitude is given by:
GF m2h 3
amax
0 = p ·
4⇡ 2 2
This can then be transformed into an upper limit on mh :

p ✓ ◆
1 2 8⇡ 2 8 2 1
|a0 | < ! mh < = ⇡v using GF = p
2 6GF 3 2v 2
mh < 700 GeV using v= 246 GeV.
This limit is soft, i.e. it means that for Higgs boson masses > 700 GeV perturbation
theory breaks down.
13.4.2 Triviality and Vacuum stability
In this section, the running of the Higgs self-coupling with the renormalisation scale
µ is used to put both a theoretical upper and a lower limit on the mass of the Higgs
boson as a function of the energy scale ⇤.
Running Higgs coupling constant
Similar to the gauge coupling constants, the coupling ’runs’ with energy.
d
= , where t = ln(Q2 ).
dt
Although these evolution functions (called -functions) have been calculated for all SM
couplings up to two loops, to focus on the physics, we sketch the arguments to obtain
these mass limits by using only the one-loop results. At one-loop the quartic coupling
runs with the renormalisation scale as:

d 3 1 2 1 4
⌘ = 2
2
+ ht ht + B(g, g 0 ) (13.9)
dt 4⇡ 2 4
, where ht is the top-Higgs Yukawa coupling as given in equation (13.1). The dominant
terms in the expression are the terms involving the Higgs self-coupling and the top
quark Yukawa coupling ht . The contribution from the gauge bosons is small and explic-
itly given by B(g, g 0 ) = 18 (3g 2 + g 02 ) + 64
1
(3g 4 + 2g 2 g 02 + g 04 ). The terms involving the
mass of the Higgs boson, top quark and gauge bosons can be understood from looking
in more detail at the e↵ective coupling at higher energy scales, where contributions from
higher order diagrams enter:
= + + + + ...
This expression allows to evaluate the value of (⇤) relative to the coupling at a reference
scale which is taken to be (v).
If we study the -function in 2 special regimes: g, g 0 , ht or ⌧ g, g 0 , ht , we’ll see
that we can set both a lower and an upper limit on the mass of the Higgs boson as a
function of the energy-scale cut-o↵ in our theory (⇤):
Triviality and Vacuum stability

| {z } | {z }
upper bound on mh lower bound on mh
mmax
h (⇤) mmin
h (⇤)
13.4.3 Triviality: g, g0 , ht heavy Higgs boson ! upper limit

on mh
For large values of (heavy Higgs boson since m2h = 2 v 2 ) and neglecting the e↵ects
from gauge interactions and the top quark, the evolution of is given by the dominant
term in equation (13.9) that can be easily solved for (⇤):
d 3 2 (v)
= ) (⇤) = 3 (v)
(13.10)
dt 4⇡ 2 1 ln ⇤2
4⇡ 2 v2
Note:
- We now have related at a scale v to at a higher scale ⇤. We see that

pas ⇤ grows,
(⇤) grows. We should remember that (v) is related to mh : mh = 2 v2.
- There is a scale ⇤ at which (⇤) is infinite. As ⇤ increases, (⇤) increases until

at ⇤=v exp(2⇡ 2 /3 (v)) there is a singularity, known as the Landau pole.
✓ ◆
3 (v) ⇤2 2 /3
ln = 1 ! At a scale ⇤ = ve2⇡ (v)
(⇤) is infinite.
4⇡ 2 v2
If the SM is required to remain valid up to some cut-o↵ scale ⇤, i.e. if we require

(Q) < 1 for all Q < ⇤ this puts a constraint (a maximum value) on the value of the
Higgs self-coupling at the electroweak
p scale (v): (v)max and therefore on the maximum
max
Higgs mass since mh = 2 (v) v 2 . Taking (⇤) = 1 and ’evolving the coupling
max
downwards’, i.e. find (v) for which (⇤) = 1 (the Landau pole) we find:
s
max 4⇡ 2 8⇡ 2 v 2
(v) = 2 ) mh < 2 (13.11)
3 ln ⇤v2 3 ln ⇤v2
For ⇤=1016 GeV the upper limit on the Higgs mass is 160 GeV/c2 . This limit gets less
restrictive as ⇤ decreases. The upper limit on the Higgs mass as a function of ⇤ from a
computation that uses the two-loop function and takes into account the contributions
from top-quark and gauge couplings is shown in the Figure at the end of Section 13.4.4.
13.4.4 Vacuum stability ⌧ g, g0 , ht light Higgs boson ! lower

limit on mh
For small (light Higgs boson since m2h = 2 v 2 ), a lower limit on the Higgs mass is
found by the requirement that the minimum of the potential be lower than that of the
unbroken theory and that the electroweak vacuum is stable. In equation (13.9) it is
clear that for small the dominant contribution comes from the top quark through the
Yukawa coupling ( h4t ).

1 3 2
= 3h4t + (2g 4 + (g 2 + g 0 )2 )
16⇡ 2 16
3 ⇥ 4
⇤
= 2 4
2MW + MZ4 4m4t
16⇡ v
< 0.
λ < 0 (not ok) Since this contribution is negative, there is

λ>0 a scale ⇤ for which (⇤) becomes negative.
ok If this happens, i.e. when (µ) < 0 the
potential is unbounded from below. As there
is no minimum, no consistent theory can be
constructed.
The requirement that remains positive up to a scale ⇤, such that the Higgs vac-
uum is the global minimum below some cut-o↵ scale, puts a lower limit on (v) and
therefore on the Higgs mass:
✓ ◆
d ⇤2
= ! (⇤) (v) = ln and require (⇤) > 0.
dt v2
✓ ◆
⇤2
(v) > ln and min (v) ! (mmin 2
h ) > 2
min
(v)v 2 , so
v2
✓ 2◆
⇤
m2h > 2v 2
ln
v2
3 ⇥ ⇤
(mmin
h )
2
= 2 2
2MW4
+ MZ4 4m4t
8⇡ v ✓ ◆
⇤2
> 493 ln
v2
Note: This result makes no sense, but is meant to describe the logic. If we go to the
2-loop beta-function we get a new limit: mh > 130 140 GeV if ⇤ = 1019 GeV. A
detailed evaluation taking into account these considerations has been performed. The
region of excluded Higgs masses as a function of the scale ⇤ from this analysis is also
shown in the Figure at the end of Section 13.4.4 by the lower excluded region.
Summary of the theoretical bounds on the Higgs mass
800
M H (GeV/c 2 )
In the Figure on the right the theoretically
mt = 175 GeV/c2
allowed range of Higgs masses is shown as a 600
function of ⇤.
400
For a small window of Higgs masses around
Landau pole
160 GeV/c2 the Standard Model is valid up
to the Planck scale (⇠ 1019 GeV). For other 200 Allowed
values of the Higgs mass the Standard Model
Vacuum instability
is only an e↵ective theory at low energy and
0
new physics has to set in at some scale ⇤. 10
3 6
10 10
9 12
10 10
15
10
18
Λ (GeV)
13.5 Experimental limits on the mass of the Higgs

boson
13.5.1 Indirect measurements
The electroweak gauge sector of the SM is described by only three independent pa-
rameters: g, g 0 and v. The predictions for electroweak observables, are often presented
using three (related) variables that are known to high precision: GF , MZ and ↵QED .
To obtain predictions to a precision better than the experimental uncertainties (often
at the per mill level) higher order loop corrections have to be computed. These higher
order radiative corrections contain, among others, contributions from the mass of the
top quark and the Higgs boson. Via the precision measurements one is sensitive to these
small contributions and thereby to the masses of these particles.
Radiative corrections
An illustration of the possibility to estimate

the mass of a heavy particle entering loop
corrections is the very good agreement be-
tween the estimate of the top quark mass us-
ing only indirect measurements and the di- Estimate: mt = 177.2+2.9
3.1 GeV/c
2
rect observation. Measurement: mt = 173.2 ± 0.9 GeV/c2

13.5. EXPERIMENTAL LIMITS ON THE MASS OF THE HIGGS BOSON 211
Sensitivity to Higgs boson mass through loop corrections
Apart from the mass of the W -boson, there are more measurements that provide sensi-
tivity to the mass of the Higgs boson. A summary of the measurements of several SM
measurements is given in the left plot of Figures 13.1.
While the corrections connected to the top quark behave as m2t , the sensitivity to the
mass of the Higgs boson is unfortunately only logarithmic (⇠ ln mh ):
M2W ⇥ ⇤
⇢ = 2
1 + quarks
⇢ + higgs
⇢ + ...
MZ cos✓W
 ✓ ◆
M2W 3 ⇣ m t ⌘2 11 tan ✓W 2 mh
= 1+ +1 g ln + ...
M2Z cos ✓W 16⇡ 2 v 96⇡ 2 MW
The results from a global fit to the electroweak data with only the Higgs mass as a free
2
parameter is shown in the right plot of Figure 13.1. The plot shows the distribution
as a function of mh . The green band indicates the remaining theoretical uncertainty
in the fit. The result of the fit suggested a rather light Higgs boson and it could be
10
∆ χ2
AUG 11
MZ G fitter SM
0.1
Tevatron 95% CL
LEP 95% CL
AUG 11
ΓZ 0.1 G fitter SM
9 3σ
σ0had -1.7
0
Rlep -1.0
0,l
8
AFB -0.9
Al(LEP) 0.2
7
Al(SLD) -2.0
sin2Θlept (Q ) -0.7
eff FB
0,c
6
AFB 0.9
0,b
AFB 2.5
Ac
5
-0.1
Ab 0.6
0
4 2σ
Rc 0.1
0
Rb -0.8
3 Theory uncertainty
(5)
∆αhad(M2) -0.1
Z
Fit including theory errors
MW -1.3
ΓW
2 Fit excluding theory errors
0.2
mc -0.0
1 1σ
mb -0.0
mt 0.3
0
-3 -2 -1 0 1 2 3 50 100 150 200 250 300
(O - Omeas) / σmeas
fit
MH [GeV]
Figure 13.1: Status of various SM measurements (left) and the 2 distribution as a function
of mh from a global fit with only mh as a free parameter (right). Before the discovery.
summarised by the central value with its one standard deviation and the one-sided (95%
CL) upper limit:
mh = 95+30
24
+74
43 GeV/c2 and mh < 162 GeV/c2 (at 95% CL).
13.5.2 Direct measurements
In July 2012 the ATLAS and CMS experiments at the Large Hadron Collider at CERN
announced the discovery of the Higgs boson. We will discuss the details of the search
for the Higgs boson and its discovery in a separate lecture, but we since we cannot have
a lecture note on the Higgs boson without proof of its discovery I include here 4 plots
that were in the discovery paper of the ATLAS experiment.
Events/5 GeV
Data ATLAS
(*)
25 Background ZZ
Σ weights / 2 GeV
(*)
ATLAS Data S/B Weighted H→ZZ →4l
100 Background Z+jets, tt
Sig+Bkg Fit (mH=126.5 GeV)
Bkg (4th order polynomial)
Signal (m =125 GeV)
80 20 H
Syst.Unc.
60
-1
40 15 s = 7 TeV: ∫Ldt = 4.8 fb
-1
s=7 TeV, ∫ Ldt=4.8fb
20 s = 8 TeV: ∫ Ldt = 5.8 fb-1
s=8 TeV, ∫ Ldt=5.9fb-1 H→γ γ
8100
10
Σ weights - Bkg
110 120 130 140 150 160

4
0
-4 5
-8
100 110 120 130 140 150 160
mγ γ [GeV]
0
100 150 200 250
m4l [GeV]
ATLAS 2011 - 2012 mH = 126.0 GeV

0
Local p
ATLAS 2011 - 2012 Obs.

W,Z H → bb -1
s = 7 TeV: ∫ Ldt = 4.7 fb
s = 7 TeV: ∫ Ldt = 4.6-4.8 fb-1 Exp.
s = 8 TeV: ∫ Ldt = 5.8-5.9 fb-1
±1 σ
H → ττ -1
s = 7 TeV: ∫ Ldt = 4.6-4.7 fb
1 0σ (*)
10-1 1σ
H → WW → lν lν
s = 7 TeV: ∫ Ldt = 4.7 fb-1
10-2 2σ
s = 8 TeV: ∫ Ldt = 5.8 fb-1
10-3 3σ
10-4
H → γγ -1
4σ s = 7 TeV: ∫ Ldt = 4.8 fb
10-5 s = 8 TeV: ∫ Ldt = 5.9 fb
-1
10-6 (*)
10-7
5σ H → ZZ → 4l
s = 7 TeV: ∫ Ldt = 4.8 fb-1
10-8 s = 8 TeV: ∫ Ldt = 5.8 fb-1
10-9 6σ
10-10
Combined
10-11 s = 7 TeV: ∫ Ldt = 4.6 - 4.8 fb
-1
µ = 1.4 ± 0.3
110 115 120 125 130 135 140 145 150 s = 8 TeV: ∫ Ldt = 5.8 - 5.9 fb
-1
mH [GeV]
-1 0 1
Signal strength (µ)
Figure 13.2: Plots from the Higgs discovery paper from ATLAS. Two-photon invariant mass
distribution (top left), the 4-lepton invariant mass distribution (top right), the p-value as a
function of the Higgs mass (bottom left) and the measurement of the coupling strength of the
Higgs boson to gauge bosons and fermions (bottom right).
All results on the Higgs boson from the ATLAS and CMS experiments at the LHC can
be found on these locations:
ATLAS: https://twiki.cern.ch/twiki/bin/view/AtlasPublic/HiggsPublicResults
CMS: http://cms.web.cern.ch/org/cms-higgs-results
13.5. EXPERIMENTAL LIMITS ON THE MASS OF THE HIGGS BOSON 213
Exercises
Exercise 13.1
Show that ūu = (ūL uR + ūR uL )
Exercise 13.2
Show that in a two body decay (a heavy particle M decaying into two particles with
mass m) the momentum of the decay particles can be written as:
p
s p 4m2
|pf | = , with = 1 x and x =
2 M2
Exercise 13.3
Higgs decay into fermions for mh = 100 GeV.
Use mb = 4.5 GeV, m⌧ = 1.8 GeV, mc = 1.25 GeV
(a) Compute (H! bb̄).
(b) Compute (H! all) assuming only decay into the three heaviest fermions.
(c) What is the lifetime of the Higgs boson. Compare it to that of the Z boson.
Exercise 13.4 (H&M exercise 6.16)

The helicity states of a massive vector particle can be described by polarization vectors.
Show that: X ⇤ pµ p⌫
✏(µ ) ✏(⌫ ) = gµ⌫ +
M2
Exercise 13.5 (Higgs decay to vector bosons)

Computing the Higgs boson decay into gauge bosons (W/Z = V), with boson momenta
p, q and helicities , is a bit more tricky. Let’s go through it step by step.
(a) Draw the Feynman diagram and use the vertex factor you computed last week to
show that the matrix element squared is given by:
✓ ◆2 X
gMV2
M =2
gµ⌫ (✏µ )⇤ (✏⌫ )⇤ g↵ (✏↵ )(✏ ),
MW ,
where and are the helicity states of the Z bosons.

(b) Use your results of exercise 13.4 and work out to show that:
✓ ◆2 
2 gMV2 (p · q)2
M = 2+ ,
MW MV4
where p and q are are the momenta of the two Z bosons.

(c) Show that the matrix element can finally be written as:
g2 3 4M2V
M2 = m4 (1 x + x2 ), with x =
4M2W h 4 m2h
(d) Show that the Higgs decay into vector bosons can be written as:
g 2 SV V 3 3 p
(h ! V V ) = m (1 x + x2 ) 1 x,
64⇡M2W h 4
4MV2
with x = m2h
and SWW,ZZ = 1, 12 .
(e) Compute (h! WW) for mh = 200 GeV.

What is the total width (only WW and ZZ decays)? And the lifetime ?
Lecture 14
Problems with the Higgs

mechanism and Higgs searches
Although the Higgs mechanism cures many of the problems in the Standard Model,
there are also several ’problems’ associated to the Higgs mechanism. We will explore
these problems in this section and very briefly discuss the properties of non-SM Higgs
bosons.
14.1 Problems with the Higgs boson
14.1.1 Problems with the Higgs boson: Higgs self-energy
Since the Higgs field occupies all of space, the non-zero vacuum expectation value of
the Higgs field (v) will contribute to the vacuum energy, i.e. it will contribute to the
cosmological constant in Einstein’s equations: ⇤ = 8⇡G
c4
N
⇢vac .
Energy density Higgs field:

With V ( † ) = µ2 2 + 4 , The ’depth’ of the potential is:
1 2 2 1 4
Vmin = V (v) = µ v + v use µ2 = v2
2 4
1 4
= v use m2h = 2 v 2
4
1 2 2 Vmin
= m v
8 h
Note that we cannot simply redefine Vmin to be 0, or any arbitrary number since quantum
corrections will always yield a value like the one (order of magnitude) given above. The
Higgs mass is unknown, but since we have a lower limit on the (Standard Model) Higgs
215
216LECTURE 14. PROBLEMS WITH THE HIGGS MECHANISM AND HIGGS SEARCHES
boson mass from direct searches at LEP (mh > 114.4 GeV/c2 ) we can compute the
contribution of the Higgs field to ⇢vac .
1 2 2
⇢Higgs
vac = m v
8 h
1
> 1 · 108 GeV4 and since GeV =
r
> 1 · 108 GeV/r3 (energy density)
Measured vacuum energy density:

An experiment to measure the energy density in vacuum and the energy density in
matter has shown:
46
⌦m ⇡ 30% and ⌦⇤ ⇡ 70% ⇠ 10 GeV4 ! empty space is really quite empty.
Problem: • 1054 orders of magnitude mismatch.

• Why is the universe larger than a football ?
14.1.2 Problems with the Higgs boson: the hierarchy problem
In the electroweak theory of the SM, loop corrections are small. In the loops the inte-
gration is done over momenta up to a cut-o↵ value ⇤.
Success of radiative corrections:

When we discussed the sensitivity of the electroweak measurements
to the mass of the Higgs boson through the radiative corrections,
the example of the prediction of the top quark mass was mentioned:
Indirect estimate: mt = 178+9.8

4.2 GeV/c
2
Direct result: mt = 172.4 ± 1.2 GeV/c2
Failure of radiative corrections:

Also the Higgs propagator receives quantum corrections.
mh = mbare
h + mh
ferm.
+ mgauge
h + mHiggs
h +... The corrections from the fermions (mainly
from the top quark) are large. Expressed in
terms of the loop-momentum cut-o↵ ⇤ given
h
t
h h
W/Z
h h
h
h
by:
and +
top 3 2 2
m2h = ⇤
t W/Z h
8⇡ 2 t
14.2. HIGGS BOSONS IN MODELS BEYOND THE SM (SUSY) 217
Correction to Higgs mass

The corrections from the top quark are not small at all,
but huge and of order ⇤. If ⇤ is chosen as 1016 (GUT) or ∆m gauge
h m SM
h
1019 (Planck), and taking the corrections into account
∆m Higgs
h m bare
(same order of magnitude), it is unnatural for mh to be h
of order of MEW (⇡ v).

∆mtop
h
The hierarchy problem: why is MEW ⌧ MPL ?
Most popular theoretical solution to the hierarchy problem is the concept of Supersym-
metry, where for every fermion/boson there is a boson/fermion as partner. For example,
the top and stop (supersymmetric bosonic partner of the top quark) contributions (al-
most) cancel. The quadratic divergences have disappeared and we are left with
✓ ◆
2 2 2 ⇤
mh / (mf mS ) ln .
mS
14.2 Higgs bosons in models beyond the SM (SUSY)
When moving to a supersymmetric description of nature we can no longer use a single

Higgs doublet, but will need to introduce at least two, because:
A) In the SM we used / ˜c to give mass to down/up-type particles in SU(2)L doublets.
In susy models these two terms cannot appear together in the Lagrangian. We
need an additional Higgs doublet to give mass to the up-type particles.
P
B) Anomalies disappear only if in a loop f Yf = 0. In SUSY there is an additional
fermion in the model: the partner for the Higgs boson, the Higgsino. This will
introduce an anomaly unless there is a second Higgsino with opposite hypercharge.
✓ + ◆ ✓ ◆ ✓ 0 ◆ ✓ ◆
1 0 2 v2
1 = 0 ! and 2 = !
1 v1 2 0
| {z } | {z }
Y 1
=+1 Y 2
= 1
Number of degrees of freedom in SUSY models:
SM: Add 4 degrees of freedom ! 3 massive gauge bosons ! 1 Higgs boson (h)
SUSY: Add 8 degrees of freedom ! 3 massive gauge bosons ! 5 Higgs boson (h, H, A, H+ , H )
v2
parameters:tan( ) = v1
and MA .
Note: - Sometimes people choose ↵ = mixing angle to give h,A, similar to

W3 /Bµp-mixing to give Z-boson and photon.
1
- MW = 2 v12 + v22 g ! v12 + v22 = v2 (246 GeV).
Di↵erences SM and SUSY Higgses:
With the new parameters, all couplings to gauge bosons and fermions change:
SUSY SM
ghV V = ghV V sin( ↵)
SUSY SM sin ↵ (h ! bb̄)SUSY sin2 (↵)
ghb b̄ = ghbb̄ ! =
cos (h ! bb̄)SM cos2 ( )
SUSY SM cos ↵ (h ! tt̄)SUSY cos2 (↵)
ghtt̄ = ghtt̄ ! =
sin (h ! tt̄)SM sin2 ( )
To determine if an observed Higgs sparticle is a SM or SUSY Higgs a detailed investi-

gation of the branching fraction is required. Unfortunately, also SUSY does not give a
prediction for the lightest Higgs boson mass:
m2h < MZ2 + 2 mtop + 2
mX + ...
 130 GeV.
Exercises
Exercise 14.1 (b-tagging at LEP)
A Higgs boson of 100 GeV decays at LEP: given a lifetime of a B mesons of roughly 1.6
picoseconds, what distance does it travel in the detector before decaying ? What is the
most likely decay distance ?
Exercise 14.2 (H ! ZZ ! 4 leptons at the LHC (lepton = e/µ))
(a) Why is there a ’dip’ in te fraction of Higgs bosons that decays to 2 Z bosons
(between 160 and 180 GeV)?
(b) How many events H ! ZZ ! e+ e µ+ µ muons are produced in 1 fb 1 of data for
mh = 140, 160, 180 and 200 GeV ? The expected number of evets is the product
of the luminosity and the cross-section: N = L ·
On the LHC slides, one of the LHC experiments shows its expectation for an analysis
aimed at trying to find the Higgs boson in the channel with 2 electrons and 2 muons.
We concentrate on mh =140 GeV.
14.2. HIGGS BOSONS IN MODELS BEYOND THE SM (SUSY) 219
(c) What is the fraction of events in which all 4 leptons have been well reconstructed
in the detector ? What is the single (high-energy) lepton detection efficiency ?
Name reasons why not all leptons are detected.
We do a counting experiment using the two bins around the expected Higgs boson mass
(we assume for the moment that the background is extremely well known and does not
fluctuate). In a counting experiment a Poisson distribution describes the probabilities
to observe x events when are expected:
x
e
P(x| ) =
x!
(d) Does this experiment expect to be able to discover the mh =140 GeV hypothesis
after 9.3 fb 1 .
(e) Imagine the data points was the actual measurement after 9.3 fb 1 . Can this
experiment claim to have discovered the Higgs boson at mh =140 GeV?
Appendix A
Some properties of Dirac matrices

↵i and
This appendix lists some properties of the operators ↵i and in the Dirac Hamiltonian:
@ ⇣ ⌘
E =i = i~ ~ + m
↵·r
@t
1. ↵i and are hermitian.
They have real eigenvalues because the operators E and p~ are hermitian. (Think
µ
of a plane wave equation: = N e ipµ x .)
2. T r(↵i ) = T r( ) = 0.
Since ↵i = ↵i , we have also: ↵i 2 = ↵i . Since 2
= 1, this implies:
2
↵i = ↵i and therefore T r(↵i ) = T r( ↵i ) = T r(↵i ) = T r(↵i ), where
we used that T r(A · B) = T r(B · A).
3. The eigenvalues of ↵i and are ±1.
To find the eigenvalues bring ↵i , to diagonal form and since (↵i )2 = 1, the square
of the diagonal elements are 1. Therefore the eigenvalues are ±1. The same is
true for .
4. The dimension of ↵i and matrices is even.
The T r(↵i ) = 0. Make ↵i diagonal with a unitary rotation: U ↵i U 1 . Then, using
again T r(AB) = T r(BA), we find: T r(U ↵i U 1 ) = T r(↵i U 1 U ) = T r(↵i ). Since
U ↵i U 1 has only +1 and 1 on the diagonal (see 3.) we have: T r(U ↵i U 1 ) =
j(+1) + (n j)( 1) = 0. Therefore j = n j or n = 2j. In other words: n is
even.
221
222 APPENDIX A. SOME PROPERTIES OF DIRAC MATRICES ↵I AND
Appendix B
Summary of electroweak theory
Take the Lagrangian of free fermions (leptons and quarks)

X
µ
L = f (i @µ m) f (B.1)
f
Arrange the left-handed projections of the lepton and quark fields in doublets
✓ ◆ ✓ ◆
⌫L uL
L = or L = (B.2)
eL dL
Ignore their masses (or choose them equal within the doublet). Now consider that the
Lagrangian remains invariant under
U (1)Y :
0
! = eiY (x)
(B.3)
SU (2)L :
0
L ! L ~ (x) · ~⌧ ]
= exp [iY ↵ L (B.4)
To keep the Lagrangian invariant compensating gauge fields must be introduced. These
transform simultaneously with the Dirac spinors in the doublet:
U (1)Y : hypercharge field aµ
Y
@µ ! Dµ = @µ + ig 0 aµ (B.5)
2
SU (2)L : weak isospin fields b1µ , b2µ , b3µ (only couple to left-handed doublet):
@µ ! Dµ = @µ + ig~⌧ · ~bµ (B.6)
Ignoring the kinetic and self-coupling terms of the gauge fields, the Lagrangian becomes
g0
L = Lfree i JYµ aµ ig J~Lµ · ~bµ (B.7)
2
223
224 APPENDIX B. SUMMARY OF ELECTROWEAK THEORY
For the generators of SU (2) we choose the Pauli spin matrices. The first field in a
left-handed doublet has T3 = +1/2 and the second field T3 = 1/2. By construction
the right-handed projections are singlets under SU (3)L and therefore have T3 = 0.
The physical gauge fields (connecting the particle fields) become
“charged currents”
b1µ ⌥ ib2µ
Wµ± = p (B.8)
2
“neutral currents”
Zµ = aµ sin ✓w + b3µ cos ✓w
(B.9)
Aµ = aµ cos ✓w + b3µ sin ✓w
The Higgs mechanism takes care that 3 out of 4 gauge bosons get mass. For the field
Aµ (the photon) to be massless, we need
g0
tan ✓w = (B.10)
g
The coupling of the massless field becomes proportional to a charge
Q = T3 + 12 Y (B.11)
Furthermore, the W and Z masses obey the relation
MW = MZ cos ✓ (B.12)
The interaction Lagrangian for the doublet can now be written as

g
Lint = p u
µ1
2
1 5
d Wµ+
2
g µ1 5
p d 2
1 u Wµ
2
" #
X (B.13)
µ
e Qf f f Aµ
f =u,d
" #
X ⇣ ⌘
gz f
µ1
2
CVf CAf 5 f Zµ
f =u,d
with
e = g sin ✓w gz = g/ cos ✓w tan ✓w = g 0 /g (B.14)
and
CVf = T3f 2Qf sin2 ✓w CAf = T3f (B.15)
The relevant quantum numbers for our fields are
225
f Q TL3 TR3
u, c, t + 23 + 12 0
1 1
d, s, b 3 2
0
1
⌫e , ⌫µ , ⌫⌧ 0 +2 –
1
e ,µ ,⌧ 1 2
0
Till now we have ignored that the weak interaction mixes the quark fields. Inserting the
CKM matrix we get for the charged currents,
g
LC.C. = i p Vud u
µ1
2
1 5
d Wµ+
2
g (B.16)
⇤ µ1 5
i p Vud d 2
1 u Wµ
2
226 APPENDIX B. SUMMARY OF ELECTROWEAK THEORY
The Feynman rules for the vertex factors are then as follows
f f
i e Qf µ
Z0 i cosg✓w µ 1
2
(CVf CAf 5 )
f f
⌫ d
g µ 1 5
W+ ip 2
(1 ) W+ i pg2 Vud µ 1
(1 5
)
2 2
` u
while those for the propagators are

f W, Z
i ig µ⌫ i(g µ⌫ pµ p⌫ /m2 )
6p m p 2
p2 m2
The photon propagator is not unique: the form above holds in the Lorentz gauge.

Lecture2017 PDF

Uploaded by

Copyright:

Available Formats

Lecture2017 PDF

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Lecture2017 PDF

Uploaded by

Copyright:

Available Formats

Particle Physics 1

Lecture notes for the first year master course on

Nikhef - autumn 2017

Wouter Hulsbergen, Marcel Merk, Ivo van Vulpen

1 Wave Equations and Anti-Particles 17

2 Perturbation Theory and Fermi’s Golden Rule 31

3 The Electromagnetic Field 47

3.4 Electrodynamics in quantum mechanics . . . . . . . . . . . . . . . . . . . 53

4 Electromagnetic Scattering of Spinless Particles 59

5 The Dirac Equation 71

7 The Weak Interaction 107

7.8 Quark mixing . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 120

8 Local Gauge Invariance 127

9 Electroweak Theory 143

10 The Process e e+ ! µ µ+ 157

11 Symmetry breaking 173

11.5.1 Lagrangian under small perturbations . . . . . . . . . . . . . . . . 181

12 The Higgs mechanism in the Standard Model 185

13 Fermion masses, Higgs decay and limits on mh 195

14 Problems with the Higgs mechanism and Higgs searches 215

A Some properties of Dirac matrices ↵i and 221

B Summary of electroweak theory 223

Nikhef program leaders/contact persons:

Name office phone email

A very brief history of particle physics

i.1 Quantum fields

classical mechanics and imposing quantization of angular momentum by requiring that

vBohr = p/m = ↵c , (i.4)

distance [m] energy [MeV]

i.2 The Yukawa interaction

to the expected kinetic energy. It was as if an additional invisible particle of low

Relative Decay Probability

Figure 1. The Beta Decay Spectrum for Molecular Tritium

By contrast, in Yukawa’s proposal the strong force were to be carried by a massive

i.3 Feynman diagrams

i.4 The Standard Model

Force Boson Coupling strength at 1 GeV

name nickname symbol quark content mass / MeV

• are quarks and leptons truly fundamental?

i.5 Units in particle physics

quantity symbol in natural units equivalent symbol in ordinary units

quantity conversion factor natural unit normal unit

Table i.6: Conversion factors from natural units to ordinary units.

while the vacuum permittivity is

i.6 Four-vector notation

We define the coordinate four-vector xµ as

A general ’contravariant four-vector’ is defined to be any set of four quantities Aµ =

we have Aµ = gµ⌫ A⌫ and Aµ = g µ⌫ A⌫ . A scalar product of two four-vectors Aµ and B µ

Wave Equations and Anti-Particles

1.1 Particle-wave duality

(x, t) = eik·x i!t

while its momentum is related to the wave-number

1.2 The Schrödinger equation

where the density is

where in the last step we have used that r · ( ⇤ r r ⇤ ) = ⇤ r2 r2 ⇤ . In the

For the density of the plane wave we obtain

If we replace p with p in the plane wave definition

1.3 The Klein-Gordon equation

Planes waves of the form