Introduction to

Particle Physics

Introduction to
Particle Physics
By

Dezső Horváth and Zoltán Trócsányi

Introduction to Particle Physics

By Dezső Horváth and Zoltán Trócsányi

This book first published 2019

Cambridge Scholars Publishing

Lady Stephenson Library, Newcastle upon Tyne, NE6 2PA, UK

British Library Cataloguing in Publication Data

A catalogue record for this book is available from the British Library

Copyright © 2019 by Dezső Horváth and Zoltán Trócsányi

All rights for this book reserved. No part of this book may be reproduced,
stored in a retrieval system, or transmitted, in any form or by any means,
electronic, mechanical, photocopying, recording or otherwise, without
the prior permission of the copyright owner.

ISBN (10): 1-5275-2808-1

ISBN (13): 978-1-5275-2808-6
Contents

Foreword xii

I Particle phenomenology 1
1 Particles and symmetries 3
1.1 Symmetries in particle physics . . . . . . . . . . . . . . . . 3
1.2 Symmetry groups and spin . . . . . . . . . . . . . . . . . . 4
1.3 Fermions and bosons . . . . . . . . . . . . . . . . . . . . . 6
1.4 Mirror reflection: parity . . . . . . . . . . . . . . . . . . . 8
1.5 Charge conjugation . . . . . . . . . . . . . . . . . . . . . . 10
1.6 CPT symmetry . . . . . . . . . . . . . . . . . . . . . . . . 10
1.7 Isospin and strangeness . . . . . . . . . . . . . . . . . . . . 11

2 What is measured in experiment? 15

2.1 Cross section . . . . . . . . . . . . . . . . . . . . . . . . . 15
2.2 Resonance . . . . . . . . . . . . . . . . . . . . . . . . . . . 17

3 Quark model 19
3.1 Coloured quarks . . . . . . . . . . . . . . . . . . . . . . . . 19
3.2 Colour interaction, QCD . . . . . . . . . . . . . . . . . . . 21
3.3 Reminder: summing up spins . . . . . . . . . . . . . . . . . 21
3.4 Lightest mesons . . . . . . . . . . . . . . . . . . . . . . . . 22
3.5 Meson nonet (flavour SU(3)) . . . . . . . . . . . . . . . . . 22
3.6 Ground-state baryons . . . . . . . . . . . . . . . . . . . . . 25
3.7 Baryon multiplets . . . . . . . . . . . . . . . . . . . . . . . 26
3.8 Three families of fermions . . . . . . . . . . . . . . . . . . 26
v
vi Table of Contents

4 Dirac equation 31
4.1 Covariant formalism . . . . . . . . . . . . . . . . . . . . . 31
4.2 Gamma matrices . . . . . . . . . . . . . . . . . . . . . . . 32
4.3 Bilinear products of spinors . . . . . . . . . . . . . . . . . . 33
4.4 Free fermions . . . . . . . . . . . . . . . . . . . . . . . . . 33
4.5 Lagrangians and equations of motion . . . . . . . . . . . . . 34
4.6 Conservation of fermion current . . . . . . . . . . . . . . . 35
4.7 Isospin algebra and conservation . . . . . . . . . . . . . . . 35
4.8 Nucleon as quark atom . . . . . . . . . . . . . . . . . . . . 37

5 Interactions 39
5.1 Three interactions of particle physics . . . . . . . . . . . . . 39
5.2 Electromagnetic interaction . . . . . . . . . . . . . . . . . . 40
5.2.1 Local U(1) invariance . . . . . . . . . . . . . . . . 40
5.2.2 Quantum electrodynamics (QED) . . . . . . . . . . 42
5.2.3 Current-current interaction . . . . . . . . . . . . . . 42
5.2.4 Photon . . . . . . . . . . . . . . . . . . . . . . . . 43
5.3 Mandelstam variables . . . . . . . . . . . . . . . . . . . . . 44
5.4 Strong interaction . . . . . . . . . . . . . . . . . . . . . . . 45
5.4.1 Colour charges . . . . . . . . . . . . . . . . . . . . 45
5.4.2 Nuclear forces . . . . . . . . . . . . . . . . . . . . 47
5.4.3 Local SU(3) invariance . . . . . . . . . . . . . . . . 48
5.4.4 Running coupling . . . . . . . . . . . . . . . . . . . 48
5.4.5 Gluons . . . . . . . . . . . . . . . . . . . . . . . . 50
5.5 Electroweak interaction . . . . . . . . . . . . . . . . . . . . 50
5.5.1 Spontaneous symmetry breaking . . . . . . . . . . . 51
5.5.2 BEH mechanism . . . . . . . . . . . . . . . . . . . 52
5.6 Basic bosons . . . . . . . . . . . . . . . . . . . . . . . . . 53
5.7 Electroweak Lagrangian with interactions . . . . . . . . . . 54

II Experimental methodology 57
6 Accelerators 59
6.1 Magnets: bending and focusing . . . . . . . . . . . . . . . . 60
6.2 Acceleration . . . . . . . . . . . . . . . . . . . . . . . . . . 61
6.3 Colliders . . . . . . . . . . . . . . . . . . . . . . . . . . . . 62
6.4 Flux and luminosity . . . . . . . . . . . . . . . . . . . . . . 62
6.5 Beam cooling . . . . . . . . . . . . . . . . . . . . . . . . . 62
6.6 CERN’s accelerator complex in the LEP era . . . . . . . . . 63
 Introduction to Particle Physics vii

6.6.1 Electrons and positrons . . . . . . . . . . . . . . . . 64

6.6.2 Protons . . . . . . . . . . . . . . . . . . . . . . . . 65
6.6.3 Heavy ions . . . . . . . . . . . . . . . . . . . . . . 66
6.6.4 Antiprotons . . . . . . . . . . . . . . . . . . . . . . 66
6.7 Other accelerators . . . . . . . . . . . . . . . . . . . . . . . 66
6.7.1 Tevatron at Fermilab . . . . . . . . . . . . . . . . . 66
6.7.2 HERA at DESY . . . . . . . . . . . . . . . . . . . 67
6.7.3 RHIC at Brookhaven NL . . . . . . . . . . . . . . . 67
6.8 CERN’s facilities in the LHC era . . . . . . . . . . . . . . . 67
6.8.1 LHC, the Large Hadron Collider . . . . . . . . . . . 67
6.8.2 Neutrinos . . . . . . . . . . . . . . . . . . . . . . . 70
6.8.3 Antiprotons . . . . . . . . . . . . . . . . . . . . . . 70

7 Detectors, calorimetry 73
7.1 Event registration . . . . . . . . . . . . . . . . . . . . . . . 73
7.2 Energy loss in matter . . . . . . . . . . . . . . . . . . . . . 74
7.3 Particle identification . . . . . . . . . . . . . . . . . . . . . 77
7.4 Detector types . . . . . . . . . . . . . . . . . . . . . . . . . 78
7.4.1 Multiwire chambers . . . . . . . . . . . . . . . . . 79
7.4.2 Scintillation counters . . . . . . . . . . . . . . . . . 79
7.4.3 Shower detectors . . . . . . . . . . . . . . . . . . . 80
7.4.4 Cherenkov detectors . . . . . . . . . . . . . . . . . 80
7.4.5 Transition radiation detectors . . . . . . . . . . . . . 81
7.5 The CMS detector . . . . . . . . . . . . . . . . . . . . . . . 82

8 Event registration 85
8.1 LEP events . . . . . . . . . . . . . . . . . . . . . . . . . . 86
8.2 Transverse momentum, pseudorapidity . . . . . . . . . . . . 87
8.3 Observation of the top quark . . . . . . . . . . . . . . . . . 89
8.4 Mysterious events . . . . . . . . . . . . . . . . . . . . . . . 89

9 Data analysis 91
9.1 Statistical concepts of particle physicists . . . . . . . . . . . 92
9.2 Basic concepts of statistical analysis . . . . . . . . . . . . . 93
9.3 Fitting parameters . . . . . . . . . . . . . . . . . . . . . . . 95
9.3.1 Goodness of fit . . . . . . . . . . . . . . . . . . . . 95
9.3.2 Confidence level . . . . . . . . . . . . . . . . . . . 95
9.4 Estimating (fitting) parameters . . . . . . . . . . . . . . . . 96
9.4.1 Arithmetic mean and standard deviation . . . . . . . 96
9.4.2 Linear fitting . . . . . . . . . . . . . . . . . . . . . 97
viii Table of Contents

9.4.3 Non-linear fitting . . . . . . . . . . . . . . . . . . . 99

9.5 Uncertainties . . . . . . . . . . . . . . . . . . . . . . . . . 100
9.6 Lower and upper limits . . . . . . . . . . . . . . . . . . . . 101
9.7 Monte Carlo simulation . . . . . . . . . . . . . . . . . . . . 103
9.8 Event selection . . . . . . . . . . . . . . . . . . . . . . . . 103

III Basic experiments 107

10 Experimental tests of the quark model 109
10.1 Three fermion families . . . . . . . . . . . . . . . . . . . . 109
10.1.1 Z width . . . . . . . . . . . . . . . . . . . . . . . . 109
10.1.2 Invisible width and the families . . . . . . . . . . . 110
10.2 Hadron jets . . . . . . . . . . . . . . . . . . . . . . . . . . 111
10.3 Fractional charges . . . . . . . . . . . . . . . . . . . . . . . 112
10.3.1 Neutral vector mesons . . . . . . . . . . . . . . . . 112
10.3.2 Pion scattering . . . . . . . . . . . . . . . . . . . . 112
10.4 Colours . . . . . . . . . . . . . . . . . . . . . . . . . . . . 113

11 Parity violation, pions and muons 117

11.1 Parity violation . . . . . . . . . . . . . . . . . . . . . . . . 117
11.1.1 τ − θ paradox . . . . . . . . . . . . . . . . . . . . . 117
11.1.2 Wu’s experiment . . . . . . . . . . . . . . . . . . . 118
11.1.3 Parity violation in pion decay . . . . . . . . . . . . 118
11.1.4 Muon spin rotation (μSR) . . . . . . . . . . . . . . 119
11.2 Anomalous magnetic moment of the muon . . . . . . . . . . 120
11.2.1 (g − 2)μ : non-relativistic measurement . . . . . . . 121
11.2.2 (g − 2)μ with relativistic muons . . . . . . . . . . . 122
11.2.3 (g − 2)μ with magic momentum . . . . . . . . . . . 122

12 Kaons and CP violation 125

12.1 Kaons . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 125
12.2 Neutral kaons . . . . . . . . . . . . . . . . . . . . . . . . . 126
12.3 Kaon oscillation . . . . . . . . . . . . . . . . . . . . . . . . 127
12.4 CP violation . . . . . . . . . . . . . . . . . . . . . . . . . . 129

13 Neutrinos 133
13.1 Weak currents . . . . . . . . . . . . . . . . . . . . . . . . . 133
13.2 Neutrino mass . . . . . . . . . . . . . . . . . . . . . . . . . 135
13.3 Early neutrino mysteries . . . . . . . . . . . . . . . . . . . 135
13.4 Neutrino oscillation . . . . . . . . . . . . . . . . . . . . . . 137
 Introduction to Particle Physics ix

13.4.1 Phenomenology . . . . . . . . . . . . . . . . . . . 137

13.4.2 Super-Kamiokande experiment . . . . . . . . . . . . 138
13.4.3 SNO experiment (1999-2003) . . . . . . . . . . . . 139
13.4.4 Neutrino mixing and masses . . . . . . . . . . . . . 140
13.4.5 Neutrino experiments at nuclear reactors . . . . . . 141
13.4.6 Long distance neutrino experiments . . . . . . . . . 142
13.4.7 Sterile neutrinos? . . . . . . . . . . . . . . . . . . . 142
13.5 Present neutrino mysteries . . . . . . . . . . . . . . . . . . 144

14 Higgs boson 147

14.1 Search for the Higgs boson . . . . . . . . . . . . . . . . . . 148
14.1.1 Methodology . . . . . . . . . . . . . . . . . . . . . 148
14.2 Exclusion at LEP . . . . . . . . . . . . . . . . . . . . . . . 150
14.3 Search and observation at LHC . . . . . . . . . . . . . . . . 151
14.3.1 Reactions of the media . . . . . . . . . . . . . . . . 153
14.3.2 Observations . . . . . . . . . . . . . . . . . . . . . 153
14.3.3 Is it really the Higgs boson? . . . . . . . . . . . . . 154
14.4 Vacuum stability . . . . . . . . . . . . . . . . . . . . . . . 155
14.5 BEH field and inflation . . . . . . . . . . . . . . . . . . . . 156

15 Heavy ion physics 159

15.1 Quark gluon plasma . . . . . . . . . . . . . . . . . . . . . . 159
15.2 Hydrodynamics . . . . . . . . . . . . . . . . . . . . . . . . 161
15.3 Experiments . . . . . . . . . . . . . . . . . . . . . . . . . . 163
15.4 Jet quenching . . . . . . . . . . . . . . . . . . . . . . . . . 164
15.5 Heavy ions at LHC . . . . . . . . . . . . . . . . . . . . . . 165
15.6 Big questions . . . . . . . . . . . . . . . . . . . . . . . . . 165

16 Practical applications 167

16.1 Informatics . . . . . . . . . . . . . . . . . . . . . . . . . . 167
16.1.1 World Wide Web . . . . . . . . . . . . . . . . . . . 168
16.1.2 Grid computing . . . . . . . . . . . . . . . . . . . . 168
16.1.3 Computer simulation . . . . . . . . . . . . . . . . . 169
16.2 Radiation . . . . . . . . . . . . . . . . . . . . . . . . . . . 169
16.3 Accelerators . . . . . . . . . . . . . . . . . . . . . . . . . . 169
16.4 Medical diagnostics . . . . . . . . . . . . . . . . . . . . . . 170
16.5 Medical therapy with radiation . . . . . . . . . . . . . . . . 171
16.5.1 Teletherapy . . . . . . . . . . . . . . . . . . . . . . 171
16.5.2 Hadron therapy . . . . . . . . . . . . . . . . . . . . 172
16.5.3 Neutron therapy . . . . . . . . . . . . . . . . . . . . 172
x Table of Contents

16.6 Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . 172

17 Coloured figures 175

Intermezzo 209

IV Standard model of elementary particles 213

18 Gauge theories in the standard model 215
18.1 Underlying gauge group . . . . . . . . . . . . . . . . . . . . 215
18.2 Basics of quantum electrodynamics . . . . . . . . . . . . . 216
18.3 Cross sections . . . . . . . . . . . . . . . . . . . . . . . . . 221
18.4 Quantum chromodynamics . . . . . . . . . . . . . . . . . . 224
18.5 Basics of colour algebra . . . . . . . . . . . . . . . . . . . . 229
18.6 Are we done? . . . . . . . . . . . . . . . . . . . . . . . . . 232
18.7 Symmetries of the classical Lagrangian . . . . . . . . . . . 234
18.8 SU(N)-amplitudes at tree level . . . . . . . . . . . . . . . . 238
18.9 Spinor helicity formalism . . . . . . . . . . . . . . . . . . . 239
18.9.1 Helicity Feynman rules for colour sub-amplitudes
(with massless fermions) . . . . . . . . . . . . . . . 241
18.9.2 A simple application of the helicity formalism . . . . 242

19 Electron-positron annihilation into hadrons 247

19.1 Electron-positron annihilation into hadrons . . . . . . . . . 247
19.2 Ultraviolet renormalization of QCD . . . . . . . . . . . . . 248
19.3 Running coupling . . . . . . . . . . . . . . . . . . . . . . . 255
19.4 Quark masses and massless QCD . . . . . . . . . . . . . . . 260
19.5 Consequences of renormalization and renormalization group
equation . . . . . . . . . . . . . . . . . . . . . . . . . . . . 262
19.6 Re+ e− at NLO . . . . . . . . . . . . . . . . . . . . . . . . . 264
19.6.1 Real corrections in d = 4 dimensions . . . . . . . . 271
19.6.2 Real corrections in arbitrary d dimensions . . . . . . 273
19.7 Origin of the singular behaviour . . . . . . . . . . . . . . . 281
19.7.1 Event shapes . . . . . . . . . . . . . . . . . . . . . 282
19.7.2 Jet algorithms . . . . . . . . . . . . . . . . . . . . . 283
19.8 Factorization of |M n | 2 in the soft limit . . . . . . . . . . . . 288
19.9 Factorization in the collinear limit . . . . . . . . . . . . . . 290
19.9.1 Regularization of real corrections by subtraction . . 293
 Introduction to Particle Physics xi

20 Deeply inelastic lepton-proton scattering 297

20.1 Kinematics . . . . . . . . . . . . . . . . . . . . . . . . . . 297
20.2 Parametrization of the target structure . . . . . . . . . . . . 298
20.3 DIS in the parton model . . . . . . . . . . . . . . . . . . . . 299
20.4 Measuring the proton structure . . . . . . . . . . . . . . . . 302
20.5 Improved parton model: perturbative QCD . . . . . . . . . 303
20.6 Factorization in DIS . . . . . . . . . . . . . . . . . . . . . . 306
20.7 DGLAP equations . . . . . . . . . . . . . . . . . . . . . . . 308

21 Hadron collisions 313

21.1 Factorization theorem . . . . . . . . . . . . . . . . . . . . . 313
21.2 Are we happy? . . . . . . . . . . . . . . . . . . . . . . . . 315
21.3 Modelling events . . . . . . . . . . . . . . . . . . . . . . . 315
21.4 Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . 317

22 Electroweak sector of the standard model 319

22.1 Weinberg mixing . . . . . . . . . . . . . . . . . . . . . . . 319
22.2 U(1) Brout-Englert-Higgs mechanism . . . . . . . . . . . . 324
22.3 Brout-Englert-Higgs mechanism in the standard model . . . 326
22.4 GIM (Glashow, Iliopoulos, Maiani) mechanism . . . . . . . 331
22.5 Fermion masses . . . . . . . . . . . . . . . . . . . . . . . . 332
22.6 Flavour mixing . . . . . . . . . . . . . . . . . . . . . . . . 334
22.7 Parameters and Feynman rules of the
standard model . . . . . . . . . . . . . . . . . . . . . . . . 337
22.8 Neutrino mixing and oscillation . . . . . . . . . . . . . . . 340
22.9 Anomaly cancellation . . . . . . . . . . . . . . . . . . . . . 346

23 Outlook 351

Bibliography 355

Index 36
xii Foreword

Foreword
One of the methods of studying Nature is to penetrate deeper and deeper in
the structure of matter ever increasing the spatial resolution, i.e., studying
smaller and smaller objects. In the history of natural sciences new and new
particles appeared which were thought to be elementary: the four atoms
(a-tom = not divisible) of Anaximenes and Democritus, the elements/atoms
of Dalton and Mendeleev, the atomic nucleus of Rutherford and the so-called
elementary particles of which the proton, the neutron, the electron and the
neutrino are the most well-known. Between 1930 and 1960 hundreds of such
particles were discovered, thus a new level of elementariness was needed
and the quark model appeared. We will see that, in fact, the proton and the
neutron are also composite particles although the electron stays elementary.
This development was crowned by the standard model (SM) in the late
sixties and it is still the uncontested global theory of matter, supported by all
available theoretical and experimental evidence.
In this textbook we summarize the present knowledge of particle physics
at an introductory level. Particle physics is a very broad subject including
many different sub-fields. While we mention many of these, a detailed
account on all is impossible. Our clear focus is on high energy collider
physics that is among the most widely pursued subfields where the threshold
of current research is high. With the Large Hadron Collider in operation new
results appear regularly. Our goal here is to keep the level introductory, yet
help students reach this high threshold making them acquainted with both
the experimental and theoretical minimum needed to comprehend current
research at colliders. Our treatise is detailed on established results of collider
physics while mostly marginal on current developments with the exception
of the discovery of the Higgs particle due to its utmost importance.
The first part (written by D. Horváth) is planned to be accessible for
advanced BSc or freshmen MSc students, while the second part (written
by Z. Trócsányi) on the theory is intended for advanced MSc or freshmen
PhD students in particle physics, with some attempt to go into the rather
complex mathematical formalism of the field. Our aim is to provide concise
but hopefully comprehensive account on the subject and also try to help
students in their decision whether to orient themselves towards experiment
or theory. We assume that the book can be covered during a full academic
year with about 10 hours of serious effort per week. Although the reader
may be confused on several occasions when not all details are given, the
theory is very precisely elaborated and its predictions beautifully agree with
the experimental observations. All present day experimental evidence is
 Introduction to Particle Physics xiii

summarized in the biennial Particle Physics Review of the Particle Data

Group [Beringer et al., 2012]; for a theoretical introduction we recommend
the textbooks of Halzen and Martin [Halzen and Martin, 1984], Collins,
Martin and Squires [Collins et al., 1989], and Perkins [Perkins, 1982]. The
theory provides a nice glance at the key experiments as well.
Experimental particle physics is also called high-energy physics, because
of its basic method of study. Energy is measured in units of electroplate,
eV, the energy gained by a particle of unit charge (e.g. an electron or
proton) when crossing a gap of 1 V voltage. One of the earliest means of
structural studies was the optical microscope. Its resolution is limited by the
wavelength of visible light (corresponding to an energy of ∼ 1 eV) to the size
of bacteria, 10−5 m. The smaller a detail, the shorter wavelength is needed
to see it. For the atoms (10−10 m) we need X-rays or electron beams of
keV (1 keV = 1000 eV) energies, for the atomic nucleus of 10−14 m electrons
or protons between MeV (106 eV) and GeV (109 eV) energies, and for the
quarks (below 10−18 m) up in the TeV (1012 eV) region. Higher energy means
smaller distances and so studying finer details of the structure of matter. As
far as we know at present, the basic particles of the standard model are really
elementary: point-like and structureless.
Throughout this book we use the natural units of particle physics where
the speed of light c and Planck’s reduced constant  = h/(2π) are both equal
to unity and thus both distance and time can be expressed in inverse energy
units, i.e. in GeV−1 (see Table 1).
We shall quote only selected, but not all original publications as this is
intended to be an introductory textbook, not a monograph and we do not
want to overwhelm students. Yet we encourage the reader to consult some
trustworthy sites on the Web (like Wikipedia http://en.wikipedia.org/, Hy-
perphysics http://hyperphysics.phy-astr.gsu.edu/hbase/hph.html and Google
Scholar http://scholar.google.com/) for simple explanations and the inSpire
(http://inspirehep.net/) publication data base for reviews and original pub-
lications. Typing in the author’s name and the date will produce all kind
of information at any depth for the reader. We consider especially use-
ful the web version of Particle Physics Review by the Particle Data Group
(http://pdg.web.cern.ch/pdg/) which provides reliable and comprehensive re-
views. Reliability is important in high energy physics as it is in the very front
of knowledge and as such it is full of speculations, untested ideas and uncon-
firmed experimental findings. At some points we shall only mention some
examples of those: Earth-absorbing black holes produced in high-energy
particle collisions, faster-than-light neutrinos and pentaquarks.
The authors acknowledge the support of the Hungarian Scientific Re-
xiv Foreword

Quantity MKS particle physics natural unit =c=1

Energy 1J 6.24 · 109 GeV GeV GeV
Momentum 1 kg m/s 5.61 · 1026 GeV/c GeV/c GeV
Mass 1 kg 5.61 · 1026 GeV GeV/c2 GeV
Distance 1m 5.07 · 1015 GeV−1 c/GeV 1/GeV
Time 1s 1.52 · 1024 GeV−1 /GeV 1/GeV
Electric √ √ √
charge (e) 0.16 aC 4π0 αc 4π0 cα 4π0 α

Table 1: Natural units of particle physics. α = 1/137 is the fine structure

constant, 0 = 8.8 · 10−12 F/m is the electric permittivity of vacuum. In the
last column only energy units appear, which permits the use of the power of
GeV to characterize the unit as “mass dimension”, e.g., the unit of length,
GeV−1 is referred to as mass dimension minus one. In this book we use the
units of the last column, except for two cases. One is momentum, which
has natural units GeV/c, to become GeV if c = 1. However, in order not to
confuse it with energy, we keep the natural unit for momentum. The other
exception
√ is the electric charge, for which we use the conventional notation
4πα, i.e., we use 0 = 1, which should not cause any confusion.

search Fund and National Research, Development and Innovation Fund of

Hungary (under contracts K101482, K103917, K109703, K125105, K124850
and K128786), and also collaboration with their students especially G. Luisoni
and G. Somogyi. Z.T. is grateful for the hospitality at the CERN Theoretical
Physics Department where this book was completed.
 Part I

Particle phenomenology

1
Chapter 1

Particles and symmetries

MOTTO:
Central to that theory is the concept of sponta-
neously broken gauge symmetry. According
to this concept, the fundamental equations of
physics have more symmetry than the actual
physical world does.

(Frank Wilczek)

1.1 Symmetries in particle physics

Symmetries in particle physics are even more important than in chemistry or
solid state physics. Just like in any theory of matter, the inner structures of
the composite particles are described by symmetries, but in particle physics
everything is deduced from the symmetries (or invariance properties) of
the physical phenomena, or from their violation: the conservation laws, the
interactions and even the masses of the particles. Symmetries that are not
connected to space and time are called internal symmetries. Continuous
symmetries can be global, i.e. independent of the space-time coordinates, or
local. The latter means that we can choose the orientation of the (external or
internal) coordinate axes freely at any space-time point. These symmetries
are also called gauge symmetries or gauge invariance laws.
3
4 Chapter 1

In field theory, according to the theorem of Emmy Noether any con-

tinuous global symmetry leads to a conservation law. Thus the freedom
to choose the origin and orientation of our coordinate system leads to the
conservation of momentum and angular momentum, that of the origin of
time measurement to the conservation of energy. The conserved quantities
due to continuous global internal symmetries are called conserved charges.
An example is the free choice of the phase of the electron wave function
as it is not measurable experimentally. This global phase invariance is also
a continuous symmetry and the emerging conservation law is the conser-
vation of the fermion charge in general. The global gauge invariance of
electrodynamics leads to the conservation of the electric charge.
Gauge symmetries have far reaching consequences as they lead to in-
teractions between particles. In the theory part (Part IV) of this book we
formulate gauge symmetries in a precise mathematical sense to discover the
three fundamental interactions of particles (the strong, weak and electro-
magnetic forces). Local symmetry always implies the existence of a related
global symmetry as well, but a global symmetry may not necessarily imply
a local symmetry, hence an interaction.

1.2 Symmetry groups and spin

The characteristic features of particles are described in terms of symmetry
groups. The language of physics is mathematics: the mathematical formu-
lation makes the difference between theory and speculation in physics as
that allows for making quantitative predictions, which then can be checked
experimentally. A new physical theory is accepted if the predictions agree
with all available experimental information.
As symmetries usually appear at transformations of our coordinate sys-
tems the mathematical apparatus is chosen accordingly. A trivial example
is the rotation of a 2-dimensional coordinate system around its origin by
an angle Θ. As shown by Fig. 1.1 the new (x , y ) coordinates of point P
are obtained in the rotated system from the old (x, y) coordinates by the
transformation

x  = a + b = x cos Θ + y sin Θ ,
y  = y  − c = y cos Θ − x sin Θ .

The point P, just as any two-dimensional vector, undergoes the following

 Particles and symmetries 5

Figure 1.1: Rotation of a coordinate system in two dimensions: coordinate

system [X , Y ] is obtained by rotating system [X, Y ] by an angle Θ.

coordinate transformation:
        
x cos Θ · x + sin Θ · y cos Θ sin Θ x
= = · .
y − sin Θ · x + cos Θ · y − sin Θ cos Θ y
   
x x
This means that the vector is obtained by multiplying the vector
y y
with the matrix in front of it. An important property of these rotation
transformations is that they do not change the length of the vector pointing
to P (its absolute value) as
 
x 2 + y 2 = (x 2 + y 2 ) · cos2 Θ + sin2 Θ = x 2 + y 2 .

The condition that the length of the vector remains unchanged demands
that the complex transformation matrix be unitary:
 ∗ ∗
  
† U11 U21 U11 U12
U U= ∗ ∗ · =
U12 U22 U21 U22
 2 + U2 ∗ U + U∗ U
  
U11 U11 12 21 22 1 0
∗ U + U∗ U
21
2 + U2 =
U12 11 22 21 U12 22
0 1
Rotations of this type have the following mathematical properties:

• they are additive: rotation by Θ1 and then by Θ2 is equivalent to a

rotation by Θ = Θ1 + Θ2 ;
• their addition is associative: (Θ1 + Θ2 ) + Θ3 = Θ1 + (Θ2 + Θ3 );
6 Chapter 1

• they have a unit element: rotation by Θ = 0 which does not do

anything;

• the rotations can be inverted (Θ − Θ = 0 and the inverse elements are

also members of the set.

Sets with operation among its elements obeying these properties are
called groups. Spin is a three-dimensional quantity with the properties of
the rotation group and its mathematical description (representation) is called
the SU(2) group of special (determinant = 1) unitary complex 2 × 2 matrices.
SU(2) can be applied not only for spin, but for any physical quantity with
similar symmetry properties, like for example the isospin to be introduced
later.
When we increase the degrees of freedom we get higher symmetry groups
of similar properties. The next step, SU(3), which is also used in particle
physics, is the symmetry group of special unitary complex 3 × 3 matrices.
It has three possible eigenstates which can be interpreted as three corners
of a triangle with an SU(2) symmetry between any two of its corners (see
Chapter 3).
In case of complex quantities we can also decrease the degrees of freedom
of rotations, then we get the U(1) group of 1 × 1 unitary matrices, i.e. eiφ
complex phases. That is the symmetry group of the gauge transformations
of the electromagnetic interaction. The simplest manifestation of the gauge
symmetry of electromagnetism is that we can freely choose the zero point
of the electrostatic potential as demonstrated by the birds sitting on high-
voltage wires. The global U(1) symmetry of Maxwell’s equations leads
to the conservation of the electric charge. In the more general case, the
U(1) symmetry of the Dirac equation [Dirac, 1931], the general equation
describing the motion of a fermion, causes the conservation of the number
of fermions or fermion charge [Halzen and Martin, 1984].

1.3 Fermions and bosons

Particles are categorized according to various properties. The most important
one is spin, the intrinsic angular momentum. Spin cannot be interpreted as
related to actual rotation, but it is added to the orbital momentum. Its natural
unit is the reduced Planck constant  = h/(2π). Spin is strange: the spin of
the electron is added to its orbital momentum, but it has two eigenstates only:
it is either left or right polarized as compared to its momentum (or points
either up or down in a vertical magnetic field) in any coordinate system. Thus
 Particles and symmetries 7

spin is characterized by two independent quantities in three dimensions: by

its length and one of its vector components.
The particles with half-integer spin (S = 12 , 32 , 52 ...) follow the Fermi–
Dirac statistics and they are called fermions, whereas those with integer
spin (S = 0, 1, 2 ...) follow the Bose–Einstein statistics and called bosons.
They have very different symmetry and other properties. The wave function
describing a system of fermions changes sign when two fermions switch
quantum states whereas in the case of bosons there is no change; all other
differences can be deduced from this property. Fermion number is conserved
whereas one can create and absorb bosons: a lamp can irradiate any number
of photons (S = 1) and an antenna can absorb them, assuming that the energy
and momentum are conserved. At the same time the electrons (S = 12 ) in
order to illuminate a lamp or a cathode tube of an old-fashioned TV set
have to be brought there and then conducted away. Another very important
difference is that any number of bosons can be put in any particular quantum
state, but only one fermion: this is Pauli’s exclusion principle. That is why
the electrons of the atom fill discrete energy levels and this is the force which
prevents the atoms of matter and the nucleons in the nucleus from penetrating
each other; it provides macroscopic forms for our material objects.
Actually, the mathematics behind Pauli’s exclusion principle is extremely
simple. The state function for a two-fermion system changes sign when you
exchange its particle states, ψF (1, 2) = −ψF (2, 1), while that of a bosonic
system does not, ψB (1, 2) = +ψB (2, 1). Thus if the two fermions would have
exactly the same quantum numbers, their state function should be zero as
that is the only function which does not change at changing its sign.
Some basic properties of fermions and bosons are compared in Table 1.1.
The elementary fermions of the standard model are the six leptons and
six quarks, the elementary bosons are the mediators of the three particle
interactions and the Higgs boson.
The standard model assumes that our world is constructed of three
fermion families, each consisting of a pair of quarks and a pair of lep-
tons, and their interactions, deduced on the basis of symmetry principles,
are mediated by vector bosons of spin 1. All fermions have antiparticles
with opposite charges and otherwise the same properties. The interaction
of a particle with its antiparticle leads to their annihilation. When an elec-
tron and its antiparticle, the positron annihilate they produce two or three
photons, whereas the annihilation of a proton with an antiproton releases so
much energy, almost 2 GeV, that half a dozen particles (mostly pions, the
lightest mesons) could be emitted.
The lack of antimatter in the Universe [Cohen et al., 1998] implies a
8 Chapter 1

Property fermion boson

Spin half-integer ( 12 , 32 . . .) integer (0, 1, 2, . . .)
ψ(1, 2) = ±ψ(2, 1) − +
Pauli exclusion yes no
Particle number
conserved yes no

Condensation no yes

Statistics Fermi-Dirac Bose-Einstein

Table 1.1: Comparison of the properties of fermions and bosons

possible asymmetry between particle and its antiparticle, and that is one of
the great mysteries of physics. Were there antimatter galaxies, they would
emit antiparticles and they would be encircled by a halo of annihilation when
meeting the particles emitted by neighbouring galaxies of ordinary matter,
but the astronomers do not see such phenomena anywhere.

1.4 Mirror reflection: parity

Changing signs of all three coordinates, i.e. changing the directions of all
three axes of a rectangular (Cartesian) coordinate system is equivalent to
changing just one sign: we call it going from the usual right-handed coor-
dinates to left-handed ones as in the usual coordinate system rotating the
x-axis to y defines the direction of z according to a right-handed screw. This
is explained in Fig. 1.2.
Parity is a general property of the mathematical functions of physics.
Any function f (x) can be written as a sum of two functions of odd and even
parities, e.g. by separating even and odd terms in its Taylor or trigonometric
expansion or simply by writing f (x) = 12 [ f (x) + f (−x)] + 12 [ f (x) − f (−x)].
The parity operator is of course unitary,
Pψ(r, t) = ψ(−r, t); P2 = 1 .
Any interaction with a spherically symmetric potential, e.g. the Coulomb
 Particles and symmetries 9

Figure 1.2: Mirror reflection and parity change (after D. Kirkby, APS, 2003)
from right-handed to left-handed coordinates.

potential of a point-like electric charge, conserves parity:

V(r) = V(−r)⇒ H(r) = H(−r)⇒ [P, H] ≡ PH − HP = 0 .

As a result, parity is a good quantum number1 of particles. For instance,

the wave function of the hydrogen atom is a parity eigenstate:

PYm (Θ, φ) = Ylm (π − Θ, π + φ) = (−1) Ym (Θ, φ)

with an eigenvalue of Pem = (−1) where is the orbital momentum. As the

minimal electromagnetic transition, E1 means a change of orbital momentum
Δ = ±1, the photon emitted in electromagnetic reactions should have at least
Pγ = −1, the photon has a negative intrinsic parity (it can have an orbital
momentum as well, of course). Anti-fermions have the opposite parities
as fermions, whereas particles and antiparticles of composite bosons2 have
the same parity determined by the parity and angular momentum of its
constituents.
Parity quantum numbers are multiplicative, e.g. for a system of three
particles: P(123) = P1 · P2 · P3 . As mesons are bound states of quarks and
1Quantum number: such a physical quantity which can change in definite quanta
only, like charge or angular momentum, and a set of which can uniquely characterize
a physical state.
2Fundamental bosons do not have antiparticles as the existence of antiparticles is
related to the Dirac equation of the fermions.
10 Chapter 1

anti quarks, their ground state ( = 0) parity is P(qq) = P(q) · P(q) = −1.
By definition for the nucleons Pp = Pn = +1, thus the parities of quarks are
+1, and of anti-quarks –1. For the particles spin J and parity P are denoted
as J P , e.g. for charged pions π ± : 0− .
We shall see later that parity is not conserved by the weak interaction, it
is a broken symmetry. Stephen Weinberg calls such symmetries accidental
symmetries.

1.5 Charge conjugation

Charge conjugation converts a particle into its antiparticle, C|p = ±|p. It
is also a unitary operator, C 2 = 1. It changes the signs of all kinds of charges:
electric, baryon, lepton charges, but not the spin. Only neutral particles could
be C-eigenstates, the eigenvalue is the C-parity. Strong and electromagnetic
interactions conserve it. For instance, in the electromagnetic decay of the
neutral pion 2 photons are emitted, π 0 →γγ and C|π 0  = C|γγ = |γγ =
|π 0 , thus Cπ 0 = +1. As C|γ = −1, the π 0 cannot decay to three photons,
π 0  γγγ.

1.6 CPT symmetry

Antiparticles can mathematically be treated as particles of the same proper-
ties going backward in space and time. This is a very important symmetry
of Nature: the physical laws do not change when charge (C), space (P) and
time (T) are simultaneously inverted:

• charge conjugation (i.e. changing particles into antiparticles),

Cψ(r, t) = ψ(r, t);

• parity change (i.e. mirror reflection), Pψ(r, t) = ψ(−r, t), and

• time reversal, Tψ(r, t) = ψ(r, −t)K where K denotes complex conju-

gation.

This is called CPT invariance. As time reversal is an anti-unitary opera-

tion, CPT is also anti-unitary, it conjugates the phase of the system, but does
not change any measurable properties. The electron-positron annihilation
can be mathematically described as if an electron arrived, irradiated two or
three photons and left backward in space and time. Using an analogy with
 Particles and symmetries 11

the electric current we call this particle current; in the above example the
incoming electron and positron constitute a lepton current.
In the simplest case of particle collision two such particle currents ex-
change a boson. This is made possible by the uncertainty principle of
Heisenberg, as it allows a violation of energy and momentum conservation
for very small time and space intervals: ΔE · Δt ≥ /2 and ΔpΔx ≥ /2,
where Δ indicates a very small change in the quantity behind it and E, p, t, x
the energy, momentum, time and space position. The very small value of the
reduced Planck constant ( 1.055 · 10−34 J·s) ensures that the conservation
laws are fulfilled in the macro-world. The boson mediating the interaction
can be real or virtual depending on whether or not it satisfies the on-shell
condition E 2 = m2 c4 + p2 c2 . Effects of virtual particles can be detected
experimentally: in the inelastic scattering of high energy electrons on each
other quark pairs could be produced when a virtual photon emitted by one
of the electrons is absorbed by a quark of a virtual quark-antiquark pair
produced momentarily by another photon emitted by the other electron.
CPT invariance is supported by ample experimental evidence. Its role
is so important in quantum field theory that according to some theorists it
is impossible to test experimentally: in the case of observing a small de-
viation one should rather suspect the violation of a conservation law than
CPT violation. In spite of this, there are considerable efforts to test it exper-
imentally. The most precise of those tests is the very small possible relative
mass difference between neutral kaon and anti-kaon which is less than 10−18 .
The European Particle Physics Laboratory, CERN has built the Antiproton
Decelerator facility in 1999 with the aim to test CPT invariance using pre-
cision spectroscopy of antihydrogen, the bound state of an antiproton and a
positron and also that of anti-protonic atoms where an electron is replaced
by an antiproton. The latter measures the mass and charge of the antiproton
(antimatter physics).
If CPT invariance is indeed a fundamental symmetry of nature, then
the violation of time reversal is equivalent to the violation of the combined
CP symmetry. We shall see later the weak interaction breaks not only P
(maximally), but also CP (a little). As a result, time invariance is also
violated by the weak interaction, in contrast to classical mechanics.

1.7 Isospin and strangeness

One of the earliest observations implying an inner structure of particles
thought to be elementary is the similarity of the proton and the neutron:
12 Chapter 1

these have almost the same mass and apart from a charge effect the strong
interaction within the atomic nucleus affects them identically. Heisenberg
introduced the concept of the nucleon which has two eigenstates, the proton
and the neutron. This needed a new quantum number characterizing it; as
its symmetry properties are identical to those of the spin he called it the
isospin I from isotopic spin (isobaric would be more precise as isotopes
have different numbers of nucleons whereas isobaric nuclei have the same
numbers of nucleons). The nucleon has an isospin I = 12 , the proton is the
nucleon state with I3 = + 12 and the neutron3 is that of I3 = − 12 .
With the development of experimental methods many strongly interacting
particles, hadrons were observed and all had characteristic isospins, i.e. all
could be arranged in groups of particles of similar properties but different
charges according to their isospins. The nucleon has an isospin I = 12 and
two similar states, with I3 = ± 12 . The lightest hadron, the π-meson or pion
has I = 1 with three eigenstates (I3 = -1, 0 and +1) and three charge states
π + , π 0 and π − . The Δ hyperon has I = 32 :

3 1 1 3
Δ− (I3 = − ) , Δ0 (I3 = − ) , Δ+ (I3 = + ) , Δ++ (I3 = + ) .
2 2 2 2
A unit change of I3 involves a corresponding unit change in charge.
Then a third quantum number, strangeness S was discovered. Pairs of
particles were produced in collisions of energetic protons with probabilities
characteristic of the strong interaction and lived long enough that they must
have decayed via weak interactions. They were called V-particles as their
tracks curved in the magnetic field of the detectors in opposite directions.
To explain this Murray Gell-Mann, Abraham Pais and Kazuhiko Nishijima
introduced strangeness S as a new additive quantum number which is con-
served in strong interactions but not in weak reactions. For instance, the Σ−
hyperon (S = −1, I = 1) created in π − p→ K+ Σ− decays via Σ− → nπ − with
a lifetime of τ ∼ 10−10 s whereas the Δ+ hyperon (S = 0) decays in Δ+ →
nπ + with a lifetime of τ ∼ 10−23 s. It was postulated that only the weak
interaction can change the new quantum number.
Strangeness and isospin made together an SU(3) group which made it
possible to construct a unique frame for all known particles. In order to
explain this, Murray Gell-Mann and George Zweig suggested the quark
model of hadrons. Using three new elementary fermions, three quarks
(Table 1.2), all observed hadrons could be described. Isospin became the

3In nuclear physics often a converse convention is used and I3 = + 12 is associated

with the neutron.
 Particles and symmetries 13

quantum number of the two lightest quarks and because of the analogy to the
spin the I3 = + 12 state was named up quark with the sign u and the I3 = − 12
state down quark, d. The third quark’s quantum number is the strangeness,
so that is the strange (s) quark. The isospin and strangeness, characterizing
the various kinds of quarks are called flavour quantum numbers.
The quark model postulates that quarks can bind together only in two
ways: in quark-antiquark pairs, those are called mesons, and three-quark
states, those are the baryons. As the quarks have spin 12 , naively we expect
that mesons are bosons and baryons are fermions. Quarks have baryon
number 13 and fractional electric charges: in units of elementary charge e
the u quark has charge + 23 while the d and s quarks − 13 . This of course gives
the proper charges to the proton p = [uud] and the neutron n = [udd] or the
pions: π + = [ud], π 0 = √1 [uu + dd], π − = [ud]. Thus the third component
2
of the isospin is directly connected to the charge, as its unit increase means
replacing a d by a u quark, i.e. increasing the total charge by + 23 − (− 13 ) = 1.

quark J eq B I3 S Y = B+S

u 1
2 + 23 1
3
1
2 0 + 13

d 1
2 − 13 1
3 − 12 0 + 13

s 1
2 − 13 1
3 0 −1 − 23

Table 1.2: The first three quarks (up, down and

strange): their spin J, electric charge eq , baryon num-
ber B, isospin third component I3 , strangeness S and
hypercharge Y = B + S.

Strangeness and isospin constitute together an SU(3) group which made

it possible to construct a unique frame for all known particles. Gell-Mann
and Zweig proposed in 1964 the quark model with the first three quarks
(Table 1.2). The quark model postulated that quarks cannot exist free, only
in bound states of two combinations: the three-quark bound states make
baryons and the quark+antiquark states mesons. As the quarks are fermions
(J = 12 ), the baryons are also fermions with baryon number B = 1 while the
mesons are bosons with B = 0. The proton is a [uud], the neutron a [udd]
ground state (i.e. J = 12 with zero orbital momentum) whereas the pions are
the lightest [qq] (q, q’ = u, d) combinations.
14 Chapter 1

Exercise 1.1
What invariance principles are violated by the weak, electromagnetic and
colour interactions?
Exercise 1.2
What are the analogies and differences between spin and isospin?

Exercise 1.3
How can the three-dimensional spin be characterised by two independent
quantities?

Exercise 1.4
What gauge symmetry facilitates the conservation of electric charge and
fermion number?
Chapter 2

What is measured in
experiment?

MOTTO:
I was brought up to look at the atom as a nice
hard fellow, red or grey in colour, according
to taste.

(Ernest Rutherford)

Physics is experimental science, in particle physics every single statement

has to be based on experimental observations. In high energy physics the two
most important experimentally measurable quantities are the cross section
and the resonance with its invariant mass and width.

2.1 Cross section

The probability of an interaction in accelerator experiments of nuclear and
particle physics is usually characterized by the ratio of the measured tran-
sition probability of the reaction in unit time, and the intensity of the bom-
barding beam, the flux:
σ = W/Φ ,
15
16 Chapter 2

measured in units of cross section (see Fig. 2.1).

The flux of a beam, Φ = nb · vb
is the number density times velocity
of the particles in a beam, number
of particles divided by cross sec-
tion and time. The σ cross section
of a reaction is measured in units
of barn (1 barn = 10−28 m2 ). That
unit comes from atomic physics and Figure 2.1: Cross section characteriz-
it is very large in particle physics ing interaction in particle scattering
(that is how it got its name), so large
that in high energy physics we most often express our measured cross sections
in pico-barn (1 pb = 10−12 barn) or femto-barn (1 fb = 10−15 barn).
When the interaction is perturbative, i.e. the Hamilton operator can be
written in the form H = H0 + H  with eigenvalues E  E0 , the transition
probability between an initial and a final state can be written as

2π
W(i→ f ) = |Mi f | 2 ρ f


where ρ f = dEdn is the density of final states in unit centre-of-mass energy

∫ CM
and Mi f ≈ ψ ∗f H ψi dτ is the transition matrix element, the overlap between
the two approximate wave functions. As the strong and electromagnetic
interactions are invariant under CP (charge and parity) reflection, CPT
invariance, which is supposed to be valid in field theory, demands that they
should also be invariant against time reflection T, and so |Mi f | 2 = |M f i | 2
(the principle of detailed balance).
In a centre-of-mass system a non-relativistic a + b → c + d particle
scattering reaction will have the cross section

W 1 (2Sc + 1)(2Sd + 1) 2
σ= = |Mi f | 2 pf
na vi π4 vi v f

where p f = | pc | = | pd | is the final state momentum, na = 1 is the number

of scattered particles, vi = | va − vb | and v f = | vc − vd | are the relative
velocities. The density of available final states will be

dp f Ec E d 1
= ≈ .
dE0 p f E0 vf
 What is measured in experiment? 17

2.2 Resonance
A simple, structureless, unstable system will have an exponential decay law
|ψ(t)| 2 = |ψ(0)| 2 e−Γt (2.1)
with a lifetime τ = Γ−1 where Γ is the decay rate. The state function of the
system will have a time evolution ψ(t) ∼ ψ(0)e−iMt e−Γt/2 . Here we use the
natural units of particle physics . In SI units τ = /Γ 6.582 · 10−22 /Γ s
where the Γ energy is measured in MeV. The time development of the system
(assuming a plain wave initial state) is ψ(t) = ψ(0)e−iMt e−Γt/2 , and its energy
amplitude will be the Fourier transform of the time amplitude:
∫ ∞ ∫ ∞
1
χ(E) = ψ(t)eiEt dt = ei(E−M)t e−Γt/2 dt =
0 0 i(E − M) − Γ/2
The result of the Fourier transformation is a probability distribution,
which appears in the measurement as a Lorentz curve:
1
| χ(E)| 2 = (2.2)
(E − M)2 + Γ2 /4
This is the Breit-Wigner resonance (Fig. 17.2, left): an energy distribution
with a maximum at the M mass of the decaying particle and a Γ width
corresponding to its decay probability. New particles can be discovered by
observing the Breit-Wigner resonances of their decays, those of course have
to be at the same (so-called invariant) mass in the different decay channels.
One of the most important resonances in high energy physics is the Z peak
in electron-positron collisions (Fig. 17.2). Two colliding beam accelerators,
LEP at CERN and SLC at Stanford were especially built to study it, and they
provided an incredible wealth of information for testing and confirming the
standard model, the theory of particle physics.
We have shown above how the lifetime of a decaying particle is connected
to its energy spectrum. In low-energy physics exponential decay curves of
the type (2.1) can be directly measured by measuring the time difference
between the birth and decay of a system. In high energy physics the reactions
are usually much too fast for that, so the lifetimes have to be estimated by
measuring the width of the mass resonance of the decaying particle.

Exercise 2.1
Why can we use cross sections to characterize interactions in particle scat-
tering?
18 Chapter 2

Exercise 2.2
How can we discover a new particle by measuring energy in particle scat-
tering?

Exercise 2.3
How can one estimate the lifetime of a decaying particle by measuring the
energy of its decay products?
Chapter 3

Quark model

MOTTO:
Just because things get a little dingy at the
subatomic level doesn’t mean all bets are off.

(Murray Gell-Mann)

3.1 Coloured quarks

The quark model, although very successful in explaining particle properties,
had serious inconsistencies. Nobody understood, for instance, why only
those postulated two kinds of bound states (3-quark and quark + antiquark
combinations) exist, why there are no free quarks in Nature and why there
could be several quarks in a particle in exactly the same quantum state in
spite of the Pauli principle. An excellent example for the latter is the Δ++
particle: it has double charge and spin 32 and so it consists of three u quarks
in the same spin state.
In the course of the history of particle physics, whenever some new
unexplained behaviour was observed, often new quantum numbers were
introduced. Good examples for those are isospin and strangeness. In order
to explain the magic combination of three identical quarks, a new quantum
number was suggested: colour charge, in analogy with the three basic colours
seen by our eye, with three possible values, red (R), green (G) and blue (B).
19
20 Chapter 3

Figure 3.1:
Colour-SU(3) with its three eigenstates and SU(2)-type stepping operators

The quarks have colour charges whereas anti-quarks have anti-colour ones.
Colour solved all those problems: the new quantum number made each
quark state in the observed particles specific and so the Pauli principle was
fulfilled. Moreover, colour is considered to play the same role for the strong
quark–quark interaction as the electric charge in electrodynamics: it is the
strong charge. The existence of all observed particles is consistent with the
condition that only colourless (white) states are allowed in Nature. This
condition accounts for the non-existence of free quarks (quark confinement),
as only the combinations of colour+anti-colour (those bound states are the
mesons) or that of all three colours (baryons) can exist. This is an algebraic
explanation for quark confinement. The dynamical understanding is far more
complex.
Thus there is a good analogy between the strong colour charge of quarks
and colour vision in humans. The three eigenstates of the quark charges
correspond to the three basic colours and strong anti-charge is the analogy
of the complementary colour. The equilibrium mixture of the three charges
or the mixture of charge + anti-charge will be neutral, colourless in colour
vision.
In order to get free particles we have to construct colourless states.
According to the experimental observations colour makes the state functions
of the composite bosons, the mesons symmetric and of the baryons, as
required for fermions, antisymmetric. For the mesons it is of the form
√1 (RR + GG + BB) and for the baryons √1 (RGB − RBG + BRG − BGR +
3 6
GBR − GRB). Anti-baryons have the colour arrangement √1 (RGB − RBG +
6
 Quark model 21

BRG − BGR + GBR − GRB). Thus when building the state function of a
mesonic or baryonic system, we just have to construct a symmetric (space
⊗ spin ⊗ flavour) function and let the colour part make it symmetric or
asymmetric as needed.
The quark model was very successful: it explained all observed hadrons
and correctly predicted the existence of new hadronic states. There are no
exceptions, like e.g., states with charges Q > 2 or isospin I > 3/2.

3.2 Colour interaction, QCD

The strong interaction is interpreted as an interaction of colour charges. As
this means an exchange of colour between quarks, its mediators should have
the form colour+anti-colour. There are 9 such combinations, but one of
them does not exchange colour, drop out as colourless, so there are eight
independent ones, called gluons. The existence of 8 gluons corresponds
to the 8 generators (denoted as λ1 . . . λ8 in Fig. 3.1) of the related SU(3)
group. Comparison of the colour interaction with electromagnetism shows
similarities and differences as well. The gluons and the photon are similar as
they carry neither electric charge nor isospin, have zero mass and their spin-
parity is J P = 1− . However, while the photon does not carry the charge of
its mediated electromagnetic interaction, the gluon carries the strong charge,
colour. As a result, there is a strong gluon-gluon interaction. Photons can
freely propagate whereas gluons cannot: they must convert into colourless
free hadrons. When we try to separate two coloured objects, quarks or
gluons, more and more energy is consumed and more and more gluons are
produced until the energy is sufficient to break up and unite all coloured
particles into hadrons. This fragmentation or hadronization goes on until no
free colour remains. An electric charge when watched from a large distance
is partially screened due to vacuum polarization as different charges attract
and identical ones repulse each other, whereas the colour grows with distance
as all colours attract each other the same way. Colour interaction strengthens
with increasing distance while the electromagnetic one weakens.

3.3 Reminder: summing up spins

When constructing a meson we add up 12 spins: S1 = S2 = 12 ; S = S1 +S2 = 0
or 1: 2 ⊗ 2 = 3 ⊕ 1, thus the result is 4 states, a triplet and a singlet. The
eigenstates of the angular momentum operator J = (J1, J2, J3 ) are of the form
|S, M > where S is the spin and M is its projection.
22 Chapter 3

Triplet states:

|S = 1, M = +1 >= ↑↑ ⎫
⎪
⎬
⎪
|S = 1, M = 0 >= √1 (↑↓ + ↓↑) (3.1)
2 ⎪
⎪
|S = 1, M = −1 >= ↓↓ ⎭
Singlet state:
1
|S = 0, M = 0 >= √ (↑↓ − ↓↑) (3.2)
2
The conserved quantities and eigenvalues of angular momentum are:
J 2 |S, M >= S(S + 1)|S, M > and J3 |S, M >= M |S, M >.
The projections are increased and decreased with the stepping operators:
J± = J1 ± iJ2 such that J± | j, m >= j( j + 1) − m(m ± 1)| j, m ± 1 >.

3.4 Lightest mesons

From the two light quarks, u and d one can construct 4 J P = 0− pseudo-
scalar (zero spin, negative parity) mesons (2 ⊗ 2 ⇒ 3 ⊕ 1): I = I1 + I2 =
2 + 2 = (0, 1) (Table 3.1).
1 1

I I3 meson state charge mass (MeV

1 +1 ud = π+ +1 139.6
1 -1 −ud = π− −1 139.6

2 (uu − dd) = π 0
1
1 0 0 135.0

2 (uu + dd) = η0
1
0 0 0 547.9

Table 3.1: The lightest mesons: the three pions, and the η0 .

3.5 Meson nonet (flavour SU(3))

Figure 3.2 shows how the three lightest quarks can combine into the pseudo-
scalar mesons. uu, dd and ss can be combined in three different ways to get
symmetric bound states as shown in Table 3.2. This scheme also shows that
the Y hypercharge is a real charge as it changes sign when going from K0 to
K0 , i.e. between a neutral particle and its antiparticle.
 Quark model 23

Figure 3.2: Construction of the meson nonet of the three light quarks (3 ⊗ 3 =
8 ⊕ 1). A, B and C are the three possible symmetric combinations of uu, dd
and ss.

(I = 0, I3 = 0) C= √1 (uu + dd + ss) (singlet: η1 )

3
(I = 1, I3 = 0) A= √1 (uu − dd) (triplet: π 0 )
2
(B ⊥ A, B ⊥ C) B= √1 (uu + dd − 2ss) (rest: η8 )
6

Table 3.2: Symmetric combinations of uu, dd and ss.

The quark flavour states of the meson nonet have to be combined with
spin states. If the spins of the quarks are anti-parallel they will make J P = 0−
pseudo-scalar mesons, parallel spins make J P = 1− vector mesons. Orbital
momentum L between the quarks will change the state parity P = −(−1) L .
The lightest, ground-state, neutral pseudo-scalar mesons are listed in
Table 3.2. States B and C have the same quantum numbers (I = 0), they will
mix and create two mass eigenstates: η1 , η8 ⇒ η, η  and the latter can be
observed in experiment:

η = η8 cos Θ p − η1 sin Θ p
η  = η8 sin Θ p + η1 cos Θ p
where Θ p ≈ 10◦ is the mixing angle. In particle physics all states that may
mix happen to mix with mixing angles to be determined experimentally. We
shall see more such cases later.
As the SU(3) group corresponds to 3 SU(2) subgroups, in the scheme
drawn in Fig. 3.2 in addition to the usual SU(2) isospin connected to the u↔d
exchange one can define similar symmetries for u↔s and d↔s; historically,
the u↔d SU(2) was called I−spin, the d↔s SU(2) U−spin and the u↔s
SU(2) V−spin, but these obsolete quantities are not used any more.
24 Chapter 3

Figure 3.3: Construction of the lightest hadrons using the first four quarks. Left:
pseudo-scalar (a) and vector (b) mesons ; right: baryons of octet (a) and decuplet (b)
structure [Beringer et al., 2012]). Note the nucleons at the back of the baryon octet.

Adding the fourth quark, charm to the set we get an SU(4) scheme.
In the left panel of Fig. 3.3 each horizontal plane corresponds to the Y
vs. I3 SU(3) plane of Fig. 3.2, the vertical axis shows the charm quark
content. In addition to the quark compositions also the historic names1 of
the mesons are noted. The u+d SU(2) is quite good symmetry as those
quarks have very little mass, of the order of 5 MeV. Flavour SU(3) for u+d+s
is already broken as the s quark is much heavier, around 100 MeV. The c
quark is even heavier, Mc 1275 MeV, so flavour-SU(4) cannot really be
used for quantitative predictions. We have to point out here that the mass
of our macroscopic world is predominantly due to the energy content of the
nucleons, the masses of the u and d quarks contribute very little.
There are quite a few meson nonets observed (Table 3.3). Generally,
when we speak of a particle, we always write down its mass to make the
situation unambiguous. For instance, the ground-state vector mesons (J P =
1− ) are ρ(770), K∗ (892), ω(782) and φ(1020) (in Table 3.3 average mass
values are listed).
1Enrico Fermi: Young man, if I could remember the names of these particles I
would have been a botanist.
 Quark model 25

L Spin J PC Nonet <mass>

(qq) (qq) I=1 I= 1
2 I=0 MeV
0 0 0−+ π K η, η  500
1 1−− ρ K∗ ω, φ 900
1 0 1+− b1 K1B h1 , h1 1250
1 0++ a0 K∗0 f0 , f0 1150
1 1++ a1 K1A f1 , f1 1300
1 2++ a2 K∗2 f2 , f2 1400
2 .. .. .. .. .. ..
.. .. .. .. .. ..

Table 3.3: Some of the observed meson nonets. ω ≈ √1 (uu − dd) and φ ≈ ss
2

flavour−SU(3) ⊗ spin−SU(2)
(qqq) 1
2 + 1
2 + 1
2 (L = 0)
3⊗3⊗3 2⊗2⊗2
(10 ⊕ 8 ⊕ 8 ⊕ 1) ⊗ (4 ⊕ 2 ⊕ 2)
TS MS MA TA TS MS MA

Table 3.4: Three-quark combinations to create ground-state baryons.

3.6 Ground-state baryons

The construction of baryons is similar to that of the mesons, but using three
quarks. They can be composed of symmetric flavour-SU(3) ⊗ spin−SU(2)
combinations (see Table 3.4).
When we switch the positions of the first two quarks in the bound states
of three, we shall have the following symmetry situations:

Total symmetry (TS) (123) = (213) = (132)

Mixed symmetry (MS) (123) = (213) = −(132)
Mixed antisymmetry (MA) (123) = −(213) = (132)
Total antisymmetry (TA) (123) = −(213) = −(132)

As colour makes the baryon states antisymmetric, we have to choose

symmetric flavour ⊗ spin states:
26 Chapter 3

TS⊗TS = (10 ⊗ 4): 4 decuplets, J = 32

√1 (MS⊗MS + MA⊗MA) = (8 ⊗ 2): 2 octets, J = 1
2
2

3.7 Baryon multiplets

Figure 3.3 (right) shows the construction of the lightest baryons of the first
four quarks [Patrignani et al., 2016] for the cases of an octet and a decuplet.
Generally, as a rule, we denote mesons with lower-case, baryons with upper-
case Greek letters (except the first observed ones). The masses of baryons
increase by about 200 MeV when adding an s quark, and that is an indication
of the mass of the s quark. Changing the spin of a baryon from 12 to 32
i.e. making the quark spins all parallel adds Δm ∼ +300 MeV to the mass.
Baryons, of course, can be excited. Spin states up to J = L + S = 92 were
observed.

3.8 Three families of fermions

A big step from the simple quark model to the standard model of particle
physics was made by Sheldon Glashow, Jon Iliopoulos and Luciano Maiani
in 1970 [Glashow et al., 1970] by introducing what became known as the
GIM mechanism. On the basis of various experimental observations they
postulated that the weak interaction arranges the quarks and leptons in pairs,
weak isospin doublets and they predicted the existence of a fourth quark,
the charm (c) quark (Table 3.5), with a charge of + 23 . In 1974 two groups,
[Augustin et al., 1974] and [Aubert et al., 1974], succeeded at the same time
in discovering the (cc) bound state (they named it differently so it is now
called J/ψ meson) and for that Burton Richter and Samuel Ting received the
Nobel Prize in 1976.
According to the GIM mechanism there should be as many lepton pairs as
quark pairs (these groups are called families), because in order to prevent the
theory from having anomalies2 the sum of electric charges of all particles has
to be zero. This is fulfilled for each family (see Table 3.5) as the charge sum
from the top down is 0 − 1 + 3 · ( 23 − 13 ) = 0, where the factor 3 stands for the
number of quark colours. Of course, as soon as Martin Perl’s group observed
a third lepton, the τ (discovery in 1976, Nobel Prize in 1995) another pair of
quarks were immediately predicted and much later discovered. Thus we have
the menagerie of basic fermions listed in Table 3.5. The place of a fermion
2Their definition is given in Section 22.9.
 Quark model 27

Family 1 Family 2 Family 3 charge T3

ν ν ν 0 + 12
 e   μ   τ 
Leptons      
 e L  μ L  τ L −1 − 12

u c t + 23 + 12
     
Quarks      
  
 d L  s L  b L − 13 − 12

Table 3.5: Leptons and quarks, the three families of basic fermions. T3 is
the third component of the weak isospin, the rest of the notation is explained
in the text step by step.

in the table is determined by its weak isospin (T) introduced in analogy of

the isospin: for the upper particles its third component has an eigenvalue
of T3 = + 12 , for the lower ones it is T3 = − 12 . Thus the pairs are weak
isospin doublets. Note that the weak isospin is quite different from isospin:
the latter is a property of the two lightest quarks only, whereas all fermions
have a weak isospin. Moreover, isospin is a quantum number of the strong
interaction whereas weak isospin is related to the weak interaction.
At this point one could ask how many more such families are hidden by
Nature at higher energies. The answer was given by experiments performed
on LEP of CERN: none, there are three families only. The proof is that
the Z-boson produced in those high-energy collisions can decay in various
ways and the standard model quantitatively describes all of them; detecting
all products the only unknown is the number of different neutrinos which
cannot be seen there. By comparing the total decay width with those of the
detected modes one can deduce the contribution of the invisible ones and
it turned out that we have just three types of light neutrinos (with masses
lower than half of the mass of the Z-boson) and so three families only (a
possible, unobservable heavy neutrino would not carry a charged lepton and
quarks, because those should have been observed already from their low–
energy effects). The number of light neutrino flavours is limited to three by
cosmology as well.
Only the fermions of the first family are stable and therefore those build
our observable world; the heavier leptons and quarks of the second and third
28 Chapter 3

families decay quickly to lighter ones. For the quarks the eigenstates of
weak and strong interactions are different. The bound states of quarks, the
hadrons are produced in strong interaction. These particles are identified by
their masses, so if we speak of a given quark, we mean their mass eigen-
states. However, the weak isospin doublets should contain weak eigenstates,
mixtures of mass eigenstates, symbolized by the prime behind their symbols
in Table 3.5. It was shown by Makoto Kobayashi and Toshihide Maskawa
(discovery in 1973, Nobel Prize in 2008) that in order to explain CP-violation
(see Section 12.4) there have to be at least 3 families of quarks. It is sufficient
to mix one row of quarks, the lower ones are chosen.
According to our present knowledge Table 3.5 contains all of the basic
fermions of the standard model. The reader should not be frightened by the
extraordinary caution of the above statement. Theoretically, the standard
model was extended in many-many ways and all those extensions predicted
numerous hypothetical new basic particles [Collins et al., 1989]. Although
till now no real evidence was found against the standard model, there are
many possible extensions that do not contradict to the present experimental
data, and so one cannot a priori exclude their validity.

Exercise 3.1
How many different quarks are in the standard model? Explain in what
sense 3, 6, 12, 18 and 36 could be all correct answers.

Exercise 3.2
How could electromagnetic interaction explain the fractional charges of the
quarks?

Exercise 3.3
How can the bound state of 3 spin- 12 quarks have a spin 92 ?

Exercise 3.4
Is there any direct evidence for the 3 quark colours?

Exercise 3.5
Quark confinement is explained by gluon exchange. How could the 983 MeV
proton mass created by three quarks of 5–15 MeV masses?

Exercise 3.6
The existence of gluons is shown by detecting 3-jet events in lepton collisions.
Why must the 3 jets be in the same plane?
 Quark model 29

Exercise 3.7
Compare and explain the signs of the terms in Eqs. (3.1, 3.2) and Table 3.1.
Hint: consider symmetries.

Exercise 3.8
How can the quark model explain the non-zero magnetic moment of the
neutron?
Chapter 4

Dirac equation

MOTTO:
When I was a young man, Dirac was my hero.
He made a breakthrough, a new method of
doing physics. He had the courage to simply
guess at the form of an equation, the equation
we now call the Dirac equation, and to try to
interpret it afterwards.

(Richard P. Feynman)

4.1 Covariant formalism

Paul Dirac composed his famous equation for fermions when he tried to
find a linear equation of motion for the electron. The Dirac equation is
based on the covariant formalism: it works with covariant and contravariant
4-vectors where component 0 is related to time and energy and the other
three components make the usual three-dimensional vectors of space and
momentum. The covariant 4-vectors have their indices as subscripts Aμ =
(A0, − A), whereas the contravariant ones as superscripts: Aμ = (A0, + A).
∂
The symbol of derivation has the opposite rule: ∂μ = ( ∂t , +∇) and ∂ μ =
∂
( ∂t , −∇) where ∇ = (∂1, ∂2, ∂3 ) is the vector operator of spatial derivation in
31
32 Chapter 4

three space dimensions.

The same Greek index repeated as super- and subscript means a scalar
product of two 4-vectors (with an implied sum 3μ=0 ) and it can be written in
different ways: A·B = A0 B0 − A· B = Aμ Bμ = Aμ Bμ = gμν Aμ Bν = g μν Aμ Bν
where
1 0 0 0
 
 0 −1 0 0 
gμν = g μν =   (4.1)
 0 0 −1 0 
 0 0 0 −1 
is the metric tensor that converts contravariant and covariant vectors into
each other and defines the length of a vector in Minkowski space-time.
The eigenfunctions of the Dirac equation are four-component spinors
ψ. The components are in the sequence: particle with spin up, particle
with spin down, antiparticle with spin up, antiparticle with spin down1.
Antiparticles have negative rest energies (masses). As the concept of spinors
was originally connected to the SU(2) eigenvectors of spin, the Dirac spinors
are often called bispinors.
1 0 0 0 ↑ m
    
 0  1  0  0  ↓ m
    
 0  0  1  0  ↑ −m
 0  0  0  1  ↓ −m

4.2 Gamma matrices

The Dirac–Pauli formalism builds the Dirac equation using the 4 × 4 γ-
I 0
matrices. They are defined as γμ = (β, βα) where β = and
0 −I
   
0 σ 1 0
α = . Here I = is the 2 × 2 unit matrix and σi are
σ 0 0 1
Pauli’s spin matrices:
     
0 1 0 −i 1 0
σ1 = σ2 = σ3 = .
1 0 i 0 0 −1
γμ (μ = 0 . . . 3) is treated as a space-time 4-vector. There is an additional
gamma
 matrix
 which plays a key role in this formalism: γ 5 ≡ iγ 0 γ 1 γ 2 γ 3 =
0 I
. γ 4 is not used here; it replaces γ 0 in other notation systems.
I 0
1The four components of the spinor have absolutely no relation to the four-vector
of space-time.
 Dirac equation 33

4.3 Bilinear products of spinors

In the physical quantities the following bilinear products of spinors can take
place (Table 4.1):

Type form components effect of P reflection

Scalar ψψ 1 sign +
Vector ψγ μ ψ 4 spatial comp. −
Tensor ψσ μν ψ 6
Axial vector ψγ μ γ 5 ψ 4 spatial comp. +
Pseudo-scalar ψγ 5 ψ 1 sign −

Table 4.1: Bilinear products of spinors, component numbers and the effect
of coordinate reflection.

In Table 4.1 ψ = ψ † γ 0 is Dirac’s adjoint spinor and in the definition of the

tensor σ μν = 2i (γ μ γν − γν γ μ ).
The theory of weak interactions is called V-A (vector minus axial-vector)
theory as the weak charged current has the form ψγ μ (1 − γ 5 )ψ.

4.4 Free fermions

The standard model uses the Lagrange-Hamilton formalism of field theory.
The Lagrangian is L = T − V where T and V are the kinetic and poten-
tial energy densities of the particle. The classical Lagrangian depends on
 t) whereas in field theory it
the general coordinates of the particle, L(q, q,
∂Φ
becomes a Lagrangian energy density L(Φ, ∂x μ
, xμ ) with the Φ(xμ ) field
∂Φ
depending on space-time four-vector xμ and on ∂μ Φ ≡ ∂x μ.

The Lagrangian is completely equivalent to the equation of motion, but

it shows the symmetry properties of the system better. The Dirac Lagrangian
of the free fermion is
L = iψγ μ ∂μ ψ − mψψ (4.2)

where ψ ≡ ψ † γ 0 is the Dirac-adjoint spinor. Applying the Euler-Lagrange

equation,
 
∂L ∂L
δL = ∂μ − = 0, (4.3)
∂(∂μ ψ) ∂ψ
34 Chapter 4

to Lagrangian (4.2) one immediately obtains the Dirac equation for the
adjoint spinor:
i∂μ ψγ μ + mψ = 0 .

Hermitian conjugation transforms it into the Dirac equation in its usual,

covariant form:
† †
[i∂μ ψγ μ + mψ]† = −iγ μ† γ 0 ∂μ ψ + mγ 0 ψ = −iγ 0 γ μ ∂μ ψ + mγ 0 ψ
(4.4)
= −γ 0 (iγ μ ∂μ ψ − mψ) = 0 .

Thus the Dirac equation of the free fermion in its final form is

(iγ μ ∂μ − m)ψ ≡ (i∂/ − m)ψ = 0 (4.5)

where we introduced the popular notation ∂/ ≡ γ μ ∂μ .

4.5 Lagrangians and equations of motion

In particle physics all equations of motion can be written in terms of the
Lagrange formalism.

• A spinless, massive (J = 0, m > 0) real scalar particle has the

Lagrangian L = 12 (∂μ φ)(∂ μ φ) − 12 m2 φ2 which gives the Klein-Gordon
equation: ∂μ ∂ μ φ + m2 φ = 0.

• The Lagrangian of quantum electrodynamics for a jμ current in in-

teraction with a zero-mass Aμ vector field (m A = 0, J A = 1 spin)
is
1
L = Efield − current × field = − Fμν F μν − jμ Aμ
4
which yields Maxwell’s equation in its covariant form: ∂μ F μν = jν
where Fμν = ∂μ Aν − ∂ν Aμ is the field strength tensor of the vector
field.

• The Lagrangian of a vector field Aμ of non-zero mass m A > 0 and

spin J = 1 would be L = − 14 Fμν F μν − jμ Aμ − 12 m2 Aμ Aμ , but such a
field was never observed.
 Dirac equation 35

4.6 Conservation of fermion current

Let us add Dirac’s equation multiplied from left by ψ and its adjoint equation
multiplied from right by ψ:
ψ[iγ μ ∂μ ψ − mψ] + [i∂μ ψγ μ + mψ]ψ = 0 .
Using the rules of derivation we find ∂μ (ψγ μ ψ) = 0 which states that the
total derivative of the four-vector j μ = ψγ μ ψ is zero, i.e. it is a conserved
 ∂ji
quantity. In 3-vector notation it has the form ∂j ∂ρ
∂t − i ∂xi = ∂t − div j = 0.
0

Thus it is a continuity equation. We saw earlier that ψγ μ ψ is the only

vector constructed from the ψ spinor of the Dirac fermion, so that must be
the fermion current that is conserved. The matter density of the fermion
2 
is j 0 = ψγ 0 ψ = ψ † γ 0 ψ = ψ † Iψ = 4i=1 |ψi | 2 and continuity means the
change in matter density balances the current.
It can be shown that the same law can be deduced from the fact that the
Dirac Lagrangian of the free fermion has a global U(1) gauge invariance.
U(1) is the group of unitary (U †U = 1) 1 × 1 matrices i.e. of U = eiα
scalars. Indeed, the Dirac Lagrangian is invariant to the transformation
ψ→eiα ψ where α is a real number. This is, of course, in accordance with
Noether’s theorem that any invariance of a Lagrangian with respect to a
global continuous transformation leads to a conserved current. Thus the
fermion charge is conserved. For instance, a lamp can emit any number of
bosons (photons), but for that one has to bring fermions (electrons) there and
carry them away. Neutron decay is a good example as well: n → p + e− + ν e ,
where both the baryon and the lepton currents are separately conserved.

4.7 Isospin algebra and conservation

Isospin helps to perform two-nucleon calculations. As we saw before, the
SU(2) symmetry of isospin makes a triplet and a singlet isospin states of two
nucleons, (I = 12 ) + (I = 12 ); 2 ⊗ 2 = 3 ⊕ 1 (Table 4.2).
Figure 4.1 illustrates how isospin helps to calculate two-nucleon corre-
lations. When corrected for the Coulomb-interaction differences between
proton and neutron, the 4 He + 2N systems have very similar masses. How-
ever, the whole isospin is also a good quantum number. Let us compare the
production rates of deuterons and pions in the inelastic scattering of protons
and neutrons on protons:
σ(pp→π + d)
R= =?
σ(np→π 0 d)
36 Chapter 4

|I = 1, I3 = +1 >= pp ⎫
⎪
⎬
⎪
|I = 1, I3 = 0 >= √1 (pn + np) triplet states
2 ⎪
⎪
|I = 1, I3 = −1 >= nn ⎭

|I = 0, I3 = 0 >= √1 (pn − np) singlet state
2

Table 4.2: Isospin eigenstates in two-nucleon systems

Figure 4.1: Nuclear masses of 4 He + NN isotopes as a function of isospin. The

values are corrected for the proton-neutron differences.

where the σ cross sections2 of the reactions are similar:


σ∝ | < I , I3 | A|I, I3 > | 2 .
I

Here I, I3 and I , I3 are the whole isospin and its 3rd component in the
initial and final states. Conservation laws dictate that I  = I; I3 = I3 . We
know the measured values Iπ = 1 and Id = 0. The first reaction:

pp→ π + d : |I = 1, I3 = 1 > →|I = 1, I3 = 1 >

and the second one (see Table 4.2):

1
np→ π 0 d : √ (|I = 1, I3 = 0 > − |I = 0, I3 = 0 >)→|I = 1, I3 = 0 >
2
√
and thus R = 12 /(1/ 2)2 = 2 in agreement with the experimental data.
Similar considerations can explain that the deuteron, the bound state of
a neutron and a proton exists whereas that of two neutrons does not.
2In high-energy physics the probability of a reaction is characterized by the
effective cross section of the involved collision process, see in detail in Ch. 2
 Dirac equation 37

4.8 Nucleon as quark atom

How could the neutral neutron have a non-zero magnetic moment? Let us
calculate the magnetic moments of the nucleons assuming that they consist
of three quarks only and nothing else (we shall see later that this is a very
rough approximation). We have to construct a symmetric spatial distribution
function as colour will make it antisymmetric and that has no role in the
magnetic moments. To make the state function symmetric we have to couple
wave functions with symmetric flavour (pS ) to symmetric spin parts (MS ),
or those with antisymmetric flavour (p A) to antisymmetric spin ones (M A).
Here symmetry and antisymmetry are related to the exchange of the first two
quarks:
1
|p >= √ [pS · χ(MS ) + p A · χ(M A)] .
2
 
pS = √1 [(ud + du)u − 2uud]
Flavour (I − SU(2)): 6
p A = √1 (ud − du)u
 2 
χ(MS ) = √1 [(↑↓ + ↓↑)↑ − 2↑↑↓]
Spin (J − SU(2)): 6
χ(M A) = √1 (↑↓ − ↓↑)↑
2
where the arrows stand for spin-up and spin-down states.
After some cancellations we get for the state function of the proton three
permutations:
1
|p↑ >= − √ (u↑u↓d↑ + u↓u↑d↑ − 2u↑u↑d↓+
18
d↑u↑u↓ + d↑u↓u↑ − 2d↓u↑u↑+
u↓d↑u↑ + u↑d↑u↓ − 2u↑d↓u↑) .

Let us try to estimate the magnetic moments of the nucleons considering

eq
the three valence quarks only. The magnetic moment of a quark is μq = 2m
q
where eq and mq are the charge and mass of quark q. Assuming that the two
valence quarks have approximately the same mass,
μu eu 2/3
≈ = = −2 .
μd ed −1/3
The magnetic moment of the proton is roughly

 
3
μp = μq = < p↑| μi σ3 |p↑ >
i=1
38 Chapter 4

     
1 0 1 0
where σ3 = , u= , d= .
0 −1 0 1

1 1
μp = [(μu − μu + μd ) + (−μu + μu + μd ) + 4(2μu − μd )] × 3 = (4μu − μd ) .
18 3
By switching u and d in the formula we get the magnetic moment of the
neutron, neutron = proton (u ↔d): μn = 13 (4μd − μu ) . Their ratio is

μn 4 − μu /μd 2
= − .
μp 4μu /μd − 1 3

The experimental observation μμnp = −0.68497934(16) (with the exper-

imental uncertainty of the last digits in parentheses) [Mohr et al., 2016]
agrees very well with this rough estimation in spite of the fact that the
mass of the nucleon is mostly energy (in the form of gluons and virtual
quark-antiquark pairs), the valence quarks contribute very little to its mass.
This agreement is a supporting argument for the validity of the quark model.

Exercise 4.1
Considering the Dirac equation in what sense can we speak of anti-bosons?

Exercise 4.2
What property of the Dirac equation leads to the conservation of the lepton
and baryon numbers?

Exercise 4.3
2 †
Prove Eq. (4.4) using γ 0 = I, γ 0 = γ 0 and γ μ† = γ 0 γ μ γ 0 ).

Exercise 4.4
Explain the signs of the terms when summing up the isospins in Table 4.2.
Hint: consider symmetries.

Exercise 4.5
Why is it a rough approximation to consider the three quarks only for calcu-
lating the magnetic moment of the nucleon?
Chapter 5

Interactions

MOTTO:
When you’ve exhausted all possibilities, re-
member that you haven’t.

(Robert H. Schuller)

5.1 Three interactions of particle physics

As mentioned before, ignoring gravity (which is many orders of magnitude
weaker than the others), there are three fundamental interactions in nature.
Their basic properties are summarized in Table 5.1.
According to the standard model the interactions are deducible from
local gauge symmetries, their source is some charge and they are mediated
by bosons characteristic for the given interaction. These bosons are real
particles in the sense that they exist not only as virtual particles mediating
interactions, but they can also be emitted and observed experimentally. A
fermion enters into an interaction if it possesses the corresponding charge;
the weak interaction operates on all fermions, the electromagnetic one on
those having electric charges or magnetic moments and the strong interaction
on the coloured fermions, i.e., on the quarks.
The photon (γ) mediates the electromagnetic interaction, and the three
weak bosons (W+ , W− and Z0 ) the weak interaction. In the strong interaction
39
40 Chapter 5

Interaction relative potential lifetime med. mass

strength boson GeV
10−23 s
Strong 1 ∝r 8 gluons 0
(Δ→pπ)
Electro- 10−20 − 10−16 s
10−2 ∝ 1
r photon 0
magnetic (π 0 →γγ)
∝ r1 e− R
r
> 10−12 s W± 80
Weak 10−7

R∼ MW c (π − →μ− ν) Z0 91

Table 5.1: Three fundamental interactions: strength, potential, lifetime,

mediating boson and its mass. In column 3 r is the distance from the source
and R is the range of interaction. In column 4 under the typical lifetime a
characteristic reaction is presented in parentheses

two quarks exchange colour so the mediating boson, the gluon (from glue)
should carry a colour and an anti-colour. This means 8 different gluons: the
3 × 3 = 9 combinations have one less degree of freedom as the combination
RR + GG + BB changes white to white and so does not carry any colour.
Pion decay provides an excellent example to compare the strengths of
the electromagnetic and weak interactions. The decay of the neutral pion to
two photons, π 0 →γγ, is a typical electromagnetic reaction with a lifetime
of 8 · 10−17 s. The charged pion can decay via weak interaction only to a
muon and its neutrino: π − →μ− + ν μ and its lifetime is 26 ns = 2.6 · 10−8 s, 8
orders of magnitude longer than that of its neutral brother. Please note that
in the above reaction a boson disappeared and a lepton was created together
with an anti-lepton: the fermion number is conserved, whereas the boson
number is not.

5.2 Electromagnetic interaction

5.2.1 Local U(1) invariance
The properties of the electromagnetic interaction have been well known for a
long time: its source is the electric charge, its mediator is the photon and it is
 Interactions 41

described by Maxwell’s equations. The Lagrangian of the latter in covariant

form is L = Efield + current×field = − 14 Fμν F μν − jμ Aμ . The Maxwell
equations themselves are of the form ∂μ F μν = jν , where F μν = ∂ μ Aν −∂ν Aμ .
The standard model derives electromagnetism from the local version
of the U(1) invariance of the Dirac equation. The requirement of local
symmetry (also called gauge invariance) means that the equation of motion
has to be invariant under a gauge transformation with space-time variation,
i.e. choosing the phase of the electron-field freely at any point in space and
time. For a free ψ(x) particle field the global U(1) symmetry that provides
the conversation of fermion charge is trivial: L(eiα ψ) = L(ψ) for any real
constant α as L contains both ψ and its adjoint spinor, ψ ≡ ψ † γ 0 with the
complex conjugate. On the other hand, the gauge transformation will be
ψ (x) = eiα(x) ψ(x), where α(x) is an arbitrary space-time function.
In the case of the classical Maxwell equation the electromagnetic vector
potential A can be included in the general momentum: p→p + Q f A, where p
is the momentum and Q f is the particle charge. In transition from classical
to quantum treatment the momentum is replaced by derivation, thus electro-
magnetism should involve a general derivative i∂μ →iDμ = i∂μ + Q f Aμ .
Let us define a covariant derivation for the local U(1) gauge symmetry
where the A vector field transforms according to the α(x) function: Aμ →Aμ +
e ∂μ α, with the e electric charge unit (we assume to have a charged lepton).
1

The new Lagrangian,

L  = i ψ γ μ Dμ ψ − mψψ
can be brought to the following form:
ψ(i γ μ ∂μ − m)ψ + eψ γ μ ψ Aμ = L − j μ Aμ .

We saw previously that the vector j μ = Q f ψ γ μ ψ is a conserved current

density of particle f and that should be multiplied by the new A field of the
interaction. We have to account for the energy density of the field itself and
with that we complete the Lagrangian of electromagnetism:
1
L  = ψ(i γ μ ∂μ − m)ψ + eψ γ μ ψ Aμ − F μν Fμν
4
Thus we have an interaction of a massive fermion with a field of zero-mass
field. Local U(1) gauge symmetry does not allow for a massive gauge field
, adding a mass term of 12 mγ2 Aμ Aμ should violate the symmetry. In order
to create the massive vector fields of the weak interaction the related SU(2)
gauge symmetry has to be violated (see below in Section 5.5.1 and in the
Theory part).
42 Chapter 5

5.2.2 Quantum electrodynamics (QED)

The free phase gives rise to the electromagnetic field with zero mass of the
mediating boson after some mathematical manipulations. As the photon
has zero rest mass, the electromagnetic interaction has an infinite range
and its Coulomb potential is inversely proportional to the distance. Freely
propagating photons are detected by our eyes (visible light) or by our TV
sets (radio waves) and so their real existence is evident.
Let us consider the interaction of spinless point charges. The Klein-
Gordon equation describes them as free particles: (∂μ ∂ μ + m2 )φ = 0.
Introducing the U(1) covariant derivation one gets

Dμ Dμ φ = (∂μ − ieAμ )(∂ μ − ieAμ )φ =

∂μ ∂ μ φ − ie(∂μ Aμ + Aμ ∂ μ )φ − e2 Aμ Aμ φ = ∂μ ∂ μ φ + V φ
where V = −ie(∂μ Aμ + Aμ ∂ μ ) − e2 Aμ Aμ is the scalar potential of the inter-
action derived from the Aμ vector potential.
The transition probability from an i initial to an f final state is propor-
tional to the overlap of their state functions in space-time,
∫ the matrix element,
when embracing the V potential: M f i = −i φ∗f V φi d4 x. As the coupling
α ≡ 4π
2
e 1 2 μ
137 is small, the e A Aμ term can be neglected in first approxima-
tion (first order perturbation calculation), and then V ≈ −ie(∂μ Aμ + Aμ ∂ μ ).
Assuming that the potential disappears at infinity, i.e. in the initial and final
states, ∫
∂μ (φ∗f Aμ φi )d4 x = [φ∗f Aμ φi ]+∞
(0,−∞) = 0,

after an integration by parts, the matrix element becomes

∫ ∫
∗ ∗ μ 4
jμ Aμ d4 x.
if
M f i = −i (−ie)[φ f (∂μ φi ) − (∂μ φ f )φi ]A d x = −i

5.2.3 Current-current interaction

Before and after the interaction we have free particles with plain wave state
functions of the form φ(x) = Ne−ip ·x where N is a normalization factor. The
fi
current is jμ = −eNi N f (pi + p f )μ exp{i(p f − pi ) · x}.
We see that in the QED Lagrangian a term of current×field strength
accounts for the interaction of the fermion with the field. In the following we
shall show in a simple example that interaction between two charged particles
will be similar: the field raised by the current of one particle interacts with
 Interactions 43

pC pD

1 q 2
t

pA pB

Figure 5.1: Scattering of point charges, A + B → C + D (AB)

the charge of the other one. Thus, it can be treated as a current-current

interaction. The probability amplitude for the transition is proportional to
∫ ∫
φ∗f (x)V(x)φi (x)d4 x = −i jμ Aμ d4 x
fi
M f i = −i

where jμ = −ie[φ∗f (∂μ φi )−(∂μ φ∗f )φi ] is the electromagnetic current density.
fi

5.2.4 Photon
The scattering of point charges A + B → C + D (AB) involves a momentum
exchange q ≡ pD − pB = pA − pC and the matrix element ∫ (1) μdetermining
the reaction probability will have the form M (1)
fi = −i jμ A(2) d4 x. The
μ μ μ
potential is the solution of D’Alembert’s equation: 2 A(2) = j(2) with j(2) =
μ μ
−eNB ND (pD + pB )μ e−iqx , giving A(2) = − q12 j(2) . Thus we have for the
transition matrix element
∫
1 μ
M f i = −i jμ(1) (− 2 ) j(2) d4 x
q
g
and in the given case −iM = [ie(p A + pC )μ ](−i qμν2 )[ie(pB + pD )ν ] where
gμν is the metric tensor (4.1).
Thus the interaction is described in a symmetric current × current form.
It is mediated by the photon which is obviously virtual as it transfers energy
and momentum and so it has a finite mass, q2 > 0. That, however, is allowed
by Heisenberg’s uncertainty principle. The photon appears between the two
currents as 1/q2 : that corresponds to the photon propagator. It has finite
mass and lifetime and is denoted by a wavy line in the Feynman diagrams
(Fig. 5.1).
44 Chapter 5

5.3 Mandelstam variables

Particle physics widely uses the Lorentz-invariant variables introduced by
Stanley Mandelstam in 1958:

s = (p A + pB )2 ,
t = (p A − pC )2 ,
u = (p A − pD )2 .

The use of these variables makes the handling of the Feynman diagrams
extremely simple as it facilitates the interrelation of different reactions via
diagram rotations, i.e. space-time reflections, called crossing. The scattering
reaction of Fig. 5.1 the original AB→CD reaction is called s channel, a
(s↔ − t) rotation gives the DB→C A t channel, and (s↔ − u) the u channel.
There are many sum rules for the Mandelstam variables, some of them are
tested in the Exercises.
As the simplest possible case, let us consider electron-positron collisions.
The momentum is p = (E, k) and its square is p2 ≡ m2 = E 2 − k 2 where
m is the electron mass. In the centre-of-mass system (like at the LEP
collider, CERN, 1989-2000) p A = (E, ki ), pB = (E, − ki ), pC = (E, k f ), and
pD = (E, − k f ). Then s = ECM 2 (E
CM = 2E is the total collision energy),
t = −2 k 2 (1 − cos θ) and u = −2 k 2 (1 + cos θ), where θ is the scattering angle
between the incoming and outgoing particles.
Without going into details, let us list a few examples for diagram rotation,
or crossing. The square of invariant amplitudes averaged for initial and
summed over final state spins for the e− e− →e− e− Møller scattering is
 
e4 s2 + u2 s2 + t 2 2s2
|M| 2 = + +
2 t2 u2 tu

where the terms describe the forward and the backward scattering, and
their interference. This interference term has a negative sign from direct
computation, but there is a subtle point in computing the square of the
amplitude: as there are identical fermions in the final state and under particle
interchange fermionic systems are odd, we have to square the difference of
the two contributions shown in the left panel of Fig. 5.2, which turns the sign
of the interference term positive. If we want to obtain the squared amplitude
for the e+ e− → e+ e− Bhabha scattering via an s↔ − u crossing, we should
use the sum of the two diagrams, i.e. the Møller formula with minus sign for
the interference term as this time the fermions are different in the final state.
 Interactions 45

Thus, we find
 
e4 s2 + u2 u2 + t 2 2u2
|M| 2 = + + .
2 t2 s2 ts
Another example is the connection between the γe→ γe Compton scat-
tering and positron annihilation. From the invariant amplitude of the former,
e4  u s 
|M | 2 = − −
2 s u
(there is no interference term in this case), one obtains that of annihilation
(shown in Fig. 5.3):
e4  u t 
|M | 2 = +
2 t u
by an s↔ − t crossing.
Another interesting case is electron scattering on muon, described with
a photon exchange (shown in Fig. 5.4 left). Its invariant amplitude is
 
e4 s 2 + u2
|M| 2 = ,
2 t2
which gives after an s↔ −t crossing the amplitude for the creation of a muon
pair in an electron-positron annihilation. At high energies the lepton masses
can be neglected as compared to their energies and we obtain
 
e4 u2 + t 2 e4 8k 4 (1 + cos2 θ) e4
|M | 2 = = = (1 + cos2 θ)
2 s2 2 (4k 2 )2 4
for the invariant squared amplitude of the process e− e+ →μ− μ+ . Its differ-
 α2
dΩ CM = 4s (1 + cos θ)
ential cross section in the centre-of-mass system is dσ 2

where α is the fine structure constant. Integrating over the polar and azimuth
angles, we find its total cross section σ(e+ e− →μ+ μ− ) = 4πα
2
3s in very good
agreement with the experimental result. √
In experimental papers very frequently s is used to denote the centre-
of-mass energy of particle collisions.

5.4 Strong interaction

5.4.1 Colour charges
The source of the strong interaction is the colour charge. It is mediated
by the 8 gluons and according to the three colours, its origin is the local
46 Chapter 5

e− e− e− e−
e− e−
A C A D
e− e−

+
e+ e+
B D B C
e− e− e− e− +
e e+

Figure 5.2: From Møller scattering on the left we get via s↔ − u crossing
(diagram rotation) Bhabha scattering

e−
γ
γ γ e−
p k p
γ γ k k’ k
γ
k k’ p−k’
p+k
p+k
γ
p−k’ k’
e+ p’
e− p p’ e− e− p p’ e− p’ k’ γ e+

Figure 5.3: From Compton scattering (left) s↔ − t crossing we get electron-

positron annihilation (diagram rotation)

e− e−

e− μ−

μ− μ− e+ μ+

Figure 5.4: From electron-muon scattering (left) one gets by s↔ − t cross-

ing the production of muon pairs in electron-positron scattering (diagram
rotation)
 Interactions 47

SU(3) symmetry. The usual representation of SU(3) transformations is


U = exp(i α · T) ≡ exp(i 8a=1 αa Ta ) where αa are real constants for a
global SU(3) symmetry and real space-time functions for the local one.

The eight generators have the commutation rule [Ta, Tb ] = i 8c=1 fabc√Tc .
The structure constants are the following: f123 = 1; f458 = f678 = 23 ;
f147 = f165 = f246 = f257 = f345 = f376 = 12 with the permutations fabc =
− facb = − fbac = − fcba . The rest with the possible indices are zero. The
generators can be realized as Ta ≡ λa /2 with the λa 3×3 Gell-Mann matrices
built via extending
 the 2 × 2 Pauli matrices of SU(2) with zeroes:
σi
λi = (i = 1, 2, 3).
0
0 0 1 0 0 −i 0 0 0
     
λ4 =  0 0 0  λ5 =  0 0 0  λ6 =  0 0 1 
 1 0 0   i 0 0   0 1 0 
0 0 0 1 0 0
   
λ7 =  0 0 −i  λ8 = √1  0 1 0 .
3
 0 i 0   0 0 −2 
Note that λ3 and λ8 are diagonal. As mentioned before in some sense
SU(3) ∼ 3 × SU(2), see Fig. 3.1 and the stepping operator between any two,
k and of the three SU(3) states are 12 (λk ± iλ ).

5.4.2 Nuclear forces

Before the colour charge was invented and quantum chromodynamics (QCD)
developed, Hideki Yukawa interpreted the short range of nuclear forces as
a sign of a massive mediator and introduced the Yukawa potential in the
form of V(r) = −g 2 e−Mr /r (in energy units).1 In V(r) g is the coupling, r
is the distance to the source, and M is the mass of the mediating particle.
Assuming that the proton radius (0.8 fm) should be close to the range of
the interaction, in 1935 Yukawa predicted the existence of the meson with
a mass of M ∼ 100 − 300 MeV. In cosmic rays C. D. Anderson found the
muon of the mass in the right range (first called μ-meson) in 1936, but that
turned out to be a lepton, the heavy brother of the electron. However, in
1947 C. Powell observed the π-meson at M 140 MeV and it seemed to
be already correct for mediating the strong interaction. Pion exchange is a
good approximation for the nuclear forces at low energies even today. All
three received Nobel Prizes in physics: Anderson in 1936 (for the positron),
1In atomic and sub-atomic physics, the term ‘potential’ is often used sloppily
instead of the correct term ‘potential energy’.
48 Chapter 5

Yukawa in 1949 and Powell in 1950.

5.4.3 Local SU(3) invariance

A free quark (please note that it cannot exist) should move according to
the Dirac equation L0 = 3j=1 q j (iγ μ ∂μ − m)q j where the summation is
performed over the colour degrees of freedom, which is usually not shown,
but meant implicitly (over repeating indices).
Global SU(3) gauge invariance of the Dirac Lagrangian is trivial. Local
SU(3) requires invariance under the transformation

q(x)→Uq(x) = eiαa (x)Ta q(x)

(this time the summation over the repeating indices is over the 8 generators).
Here, similarly to the QED case, αa (x) are real space-time functions. The
SU(3) gauge field is introduced via the covariant derivation

Dμ = ∂μ + igs Ta G μa

with the transformation rule

1
G μa → G μa − ∂μ αa − fabc αb G μc .
g
The field strength will be
a
G μν = ∂μ Gνa − ∂ν G μa − gs fabc G μb Gνc

with a coupling gs . Note that the first two terms are exact analogies of the
QED field. Similarly to QED the QCD Lagrangian is
1 a μν
LQCD = q j (iγ μ ∂μ − m)q j − gs (q j γ μ Ta q j )G μa − G μν Ga .
4
The QCD Lagrangian in its expanded form (Fig. 17.1), in addition to a linear
term similar to that in the QED case, has terms of higher order interactions.

5.4.4 Running coupling

The colour interaction is mediated by eight gluons. As local gauge invariance
cannot include masses for the mediating bosons, the mass of the gluons is
also zero and so the range of the colour interaction is also infinite; its potential
at short distances is of the Coulomb form, but at long distances it is roughly
 Interactions 49

QED QCD
Elementary fermions leptons quarks
Charge electric colour
Gauge boson photon (γ) 8 gluons (g)
no charge have colours
Coupling α(Q2 = 0) = 137
1
αs (Q2 = m2Z ) = 0.12
Q2 dependence weak strong
Free particles leptons hadrons
Calculation precision < 10−8 5 − 20%

Table 5.2: Comparison of the basic properties of quantum electrodynamics

(QED) and quantum chromodynamics (QCD)

proportional to the distance of the coloured objects. This the consequence

of the fact that–as opposed to the photons–the gluons themselves also carry
colour, the charge of the interaction, and so they interact with each other.
When we try to separate two quarks, the energy of the field will grow with
the distance as the gluons will produce more gluons and quark–antiquark
pairs, and the quarks will combine into hadrons so that we see colourless
objects only: that is quark confinement. Presently it is not known how this
hadronization can be deduced from first principles and it is one of the main
sources of uncertainty of simulations in particle physics.
The strengths of interactions as reflected by their coupling depend on
the distance of the charges involved. This dependence is weak for electro-
magnetism: the fine structure constant at low energies is α(E = 0) 1/137,
whereas at the energy corresponding to the mass of the Z boson it is
α(E = 91 GeV) 1/128. The strong coupling is close to unity at low
energies (as manifested by quark confinement) and converges to zero at ex-
treme high energies (asymptotic freedom). Table 5.2 compares the basic
properties of the two interactions.
Because of quark confinement the characteristic particles of strong in-
teraction do not appear free in Nature, but can be studied via their particular
decay properties. The most convincing proof for the existence of quarks
is the detection of events with two hadron showers, jets, in high-energy
electron-positron collisions . In electromagnetic or weak processes the pro-
duction of an additional particle in scattering or decay reactions involves
one more coupling, decreasing the reaction probability by at least an order
of magnitude. In the electron+positron atomic bound state, the positronium
50 Chapter 5

the two particles can have parallel or opposite spins; the former state, with
spin S = 1 called ortho-positronium decays by emitting three photons be-
cause of parity conservation and has a lifetime of 140 ns, whereas the other
one, the S = 0 para-positronium, decays to two photons with a lifetime of
125 ps. A high-energy hadron jet contains dozens of particles and so cannot
be produced by other interactions, just hadronization.

5.4.5 Gluons
Gluons are produced most frequently in hadron-hadron collisions. In the
proton-proton collisions of the Large Hadron Collider that is the most fre-
quent phenomenon. Higgs bosons are produced most frequently in gluon-
gluon collisions (gluon fusion). Gluon production is manifested best in 3-jet
events (Fig. 17.25); at the production of quark pairs in electron-positron
collisions, the quarks can interact with each other emitting a gluon and we
see an event of three jets in a common plane. The third jet must be from a
boson because of the conservation of fermion charge and that boson must be
coloured to be able to form a third jet.
The properties of gluons were established via studying such 3-jet events
in electron-positron collisions where two of the jets were identified as b
quarks (due to its longer lifetime and increased lepton emission), so that the
third jet had to be a gluon. The unit spin and zero mass of the gluons were
this way confirmed experimentally.

5.5 Electroweak interaction

The weak interaction is weak not because of a small coupling, but because
of its very heavy mediating bosons. When the neutron decays,

n→p + W− →p + e− + ν e ,

the 1.3 MeV mass difference have to be transferred by the MW = 80 GeV

gauge boson, which is possible due to the uncertainty relation allowing for a
low decay probability and a relatively long lifetime for the neutron.
The birth of the standard model is usually dated to the early seventies, the
time when the theory of the electroweak interaction was developed, for which
Sheldon Lee Glashow, Abdus Salam and Steven Weinberg was awarded the
Nobel Prize in physics in 1979. The theory had to solve two problems:
 Interactions 51

• One had to introduce the masses of elementary particles, those for the
weak bosons and the elementary fermion, the quarks and leptons. As
gauge invariance prohibited that, it had to be somehow violated.

• The equations of electromagnetic and weak scattering had infinite

terms. In order to help to eliminate them a scalar boson (having zero
quantum numbers) was needed.

5.5.1 Spontaneous symmetry breaking

As we could see, the electromagnetic and strong interactions are successfully
described by local gauge symmetries with massless mediating bosons. For
the weak interaction local SU(2) symmetry seems to be adequate except for
the masses: we have to introduce a violation to get finite masses for the
gauge bosons. How can one violate local SU(2) without violating the other
two invariances, to keep zero masses for the gluons and the photon?
The Brout–Englert–Higgs (BEH) mechanism solves this problem the
following way. It unites the electromagnetic U(1) and weak-like SU(2)
gauge symmetries in a vacuum filled with the four-component, scalar SU(2)
BEH field:    
φα φ1 + iφ2
φ= = 2 1
φβ φ3 + iφ4
with the corresponding Lagrangian

L = (∂ν φ)† (∂ν φ) − μ2 φ† φ − λ(φ† φ)2

where (λ > 0) is a real constant. If μ2 > 0 this is a scalar field with a finite
mass; however, if μ2 < 0 its stable vacuum state is not at φ = 0, but at a
different vacuum expectation value.
The Mexican hat illustrates the BEH potential in two dimensions: it has
a perfect axial symmetry (Fig. 5.5) and that is not violated by putting a ball
on the top of the potential at point (0,0). However, the symmetry will be
spontaneously violated when the ball will roll down. In the valley it can
move without spending energy, so we can put one axis of our coordinate
system on the position of the ball. The new system will be characterized
by one parameter, the distance of the minimum from zero, the vacuum
expectation value (generally abbreviated as vev). Then we parametrize the
BEH field around  the displaced minimum and write it in the form φ(x) =
1 0
.
2 v + h(x)
52 Chapter 5

Figure 5.5: Spontaneous symmetry breaking. The perfect axial symmetry

of the BEH potential is violated when we place the ball on top, because its
stable state is somewhere in the valley; as it is indifferent where, we can
adjust our coordinate system accordingly.

5.5.2 BEH mechanism

The U(1)Y local gauge invariance, related to the Y hypercharge generates
a B field and the SU(2)L (the L index means that it acts on left-handed
fermion currents only) a 3-component W = (W 1, W 2, W 3 ) field. Together
they correspond to a U(1) ⊗ SU(2) 4-component local gauge invariance field
which is then violated by the 4-component (complex doublet) BEH field
. We know that the photon is massless, so we have to restore the U(1)
local gauge invariance, i.e. to separate an electromagnetic vector potential
A from a mixture of the B field with the neutral component of W, W3 . The
charged currents will be W ± = √1 (W1 ± W2 ) and the rest of the neutral part
2
of U(1) ⊗ SU(2) will give the neutral current with the Z0 boson of mass
mZ 91 GeV as mediator.
To calculate the electromagnetic current we use the definition of the
hypercharge as Y = 2(Q − T 3 ) where Q is the electric charge and T 3 is
the third component of the weak isospin related to the SU(2) doublets.
The current of the hypercharge is connected to U(1)Y as jμY = ψγμY ψ and
the electromagnetic current will be jμem = Jμ3 + 12 jμY (in units of the electric
charge unit). Identifying the coupling of the electromagnetic current with the
charge one immediately gets the coupling for electromagnetism as the mixing
of U(1) with the neutral part of SU(2) and it is characterized–as usual in
particle physics–with the weak or Weinberg mixing angle θ W 28.20 . In the
measurable quantities it generally appears as sin2 θ W with a slightly energy-
dependent value: 0.221 at low energies and 0.231 at E = mZ c2 91.2 GeV.
 Interactions 53

The BEH field introduces four new degrees of freedom in the system,
three of them will give masses to the weak bosons and the fourth one the
Higgs boson, a heavy particle with all quantum numbers zero. From the
known couplings one can calculate the masses of the weak bosons and the
vacuum expectation value of the BEH field, v 246 GeV.
The BEH mechanism was independently published in 1964 by Peter
Higgs and two research groups, but at that time it was received by a great
deal of scepticism. It actually introduces an artificial new field, a force field
with no source, which fills vacuum and creates a new scalar particle with all
zero quantum numbers except its mass which is not predicted by the theory.
It is very hard to detect such a particle. It was clarified in the early seventies
only that the BEH mechanism, in addition to the production of masses for
the weak bosons, solves a whole bunch of other problems: facilitates to
introduce the fermion masses and creates the badly needed scalar boson to
eliminate some mathematical difficulties. As the theory and experiment of
particle physics developed, the belief of physicists increased that the BEH
mechanism should be correct: the predicted neutral currents were observed,
the properties of the weak bosons determined, and all experimental data
seemed to agree with the calculations of the standard model. Finally, 48
years after its prediction, the Higgs boson was observed at the LHC in 2012.

5.6 Basic bosons

Thus the interactions are derived from local gauge symmetries: exact SU(3)
symmetry in the case of QCD and a spontaneously broken U(1)Y ⊗ SU(2) L
symmetry in the case of the electroweak interaction. The strong and the
electromagnetic interactions are mediated by zero-mass gauge bosons, the
8 gluons and the photon. The gauge bosons of the weak interaction are
heavy: MW = (80.385 ± 0.015) GeV and MZ = (91.1876 ± 0.0021) GeV
[Patrignani et al., 2016]. The Higgs boson was the last particle of the
standard model discovered, it was observed in 2012 only by the ATLAS
and CMS experiments at the Large Hadron Collider of CERN at the mass
MH = (125.09 ± 0.24) GeV [Patrignani et al., 2016] where the uncertainty
includes both statistical and systematic ones.
With the observation of the Higgs boson the standard model became
complete: all of its predicted particles were found experimentally, all mea-
surements agree with its predictions and no exception or deviation was
observed. Of course, since its foundation in the early seventies, many exper-
iments gave results seemingly contradicting the standard model, but those
54 Chapter 5

Figure 5.6: CERN coffee mug with the electroweak Lagrangian

were either refuted later or stayed unconfirmed by independent new data.

Thus at the moment we have to accept the standard model as the theory of
matter, of particle physics.

5.7 Electroweak Lagrangian with interactions

The rather complicated electroweak Lagrangian of the standard model in-
cluding the interactions and the BEH mechanism (derived and explained in
detail in Section 22.2 of this book) can be written in an extremely concise
form which is widely distributed on T-shirts and coffee mugs (Fig. 5.6) at
CERN:
1
L = − F μν Fμν + iψ Dψ
/ + ψi yi j ψ j φ + h.c. + (Dμ φ)2 − V(φ)
4

The first two terms are similar to those of QED: the first one is the field
energy density, the second term stands for the interaction of the fermion,
/ = γ μ Dμ where D denotes the covariant derivation. The third term is
D
a possible gauge-invariant coupling between the fermions and the φ BEH
field that leads to fermion masses, the term (Dμ φ)2 generates the masses
of the gauge bosons and the last term is the self-interacting BEH-potential.
The fermion interaction terms need to be complemented by their Hermitian
conjugates. This equation is nicely explained in a generally comprehensible
way in [Woithe et al., 2017].
 Interactions 55

Exercise 5.1
What is common in the three interactions described by the Standard Model?
Why is gravity different from them?

Exercise 5.2
To what extent does the magic number 3 connect the 3 fermion families with
the 3 quark colours and 1/3 quark charges?

Exercise 5.3
Conservation of the electric charge and colour charge is connected to which
Noether theorem? What are the two physical quantities whose conservation
is violated by the weak interaction?

Exercise 5.4
Why do quark states mix in ground state? Why is the number of families
limited to 3 in the standard model?
Exercise 5.5
Why does the number of quarks equal that of the leptons?

Exercise 5.6
Explain in what sense it is correct to consider for the numbers of different
quarks 3, 6, 9, 12, 18 and 36.

Exercise 5.7
Explain in what sense it is correct to consider in the standard model for
the numbers of different elementary particles (i.e. of the basic fermions and
bosons) 17, 21, 44, 52, 61 and 109.

Exercise 5.8
How did we determine the number of fermion families? Why cannot quarks
heavier than the top quark exist within the standard model?

Exercise 5.9
Why are the masses of the charged and neutral weak bosons different?
 Part II

Experimental methodology

57
Chapter 6

Accelerators

MOTTO:
I sometimes think about the tower at Pisa as the
first particle accelerator, a (nearly) vertical linear
accelerator that Galileo used in his studies.

Leon Lederman in “The God Particle: If the

Universe is the Answer, What is the Question”
(by Leon Lederman and Dick Teresi)

One needs higher and higher energies to get deeper and deeper in the
structure of matter. Cosmic rays can have very high energies, but at extremely
low intensity and with no choice of parameters. At the times of Columbus
the front line of discovery in Europe was the shore of the Atlantic Ocean:
he used ships to reach India and found America. Nowadays on the front line
of physics we see the Higgs boson and the quark-gluon plasma, the state of
matter right after the Big Bang. We use particle accelerators, now the Large
Hadron Collider (LHC) of CERN to study them, and in the 21st century after
many decades of hard work we see them both. Of course, the devil is in the
details, one needs more and more precise measurements as that is the only
way to find deviations from the predictions of the standard model, i.e. new
physics.
The word accelerator is a bit confusing in contemporary particle physics
as all machines produce particles with velocities extremely close to c, the
59
60 Chapter 6

Figure 6.1: Left: forces in a quadrupolar magnetic field. Right: focusing

charged particles in a quadrupole triplet: each magnet focuses in its focusing
plane to the centre of the defocusing plane of the next magnet.

speed of light in vacuum. We increase rather energies than velocities. For

instance, in 2012 the protons in the LHC ring moved at velocities v
0.99998c.
The electromagnetic field in an accelerator has to serve both to increase
the energy of the particles and to keep them together on the required orbit.
As a general rule, an electric field accelerates and a magnetic fields bends
the trajectories of charged particles. Using potentials one can write for
the magnetic field strength B = ∇ × A and for that of the electric field
E = −∇Φ − ∂ A/∂t. Here ∇ = (∂ x, ∂ y, ∂z) is the 3-dimensional derivation
vector, A is the vector potential and Φ is the scalar potential.

6.1 Magnets: bending and focusing

The bending strength of a magnetic field B on a particle of charge e and
velocity v is F = ev × B; | F | = mv 2 /r where r is the bending radius. The
magnetic rigidity is Br = p/e ≈ 3.34p for a particle with momentum p
orthogonal to B in units of [p] = GeV/c, [B] = T and [r] = m. In length L
the bending angle (Fig. 17.3) is sin Θ2 = 2r L
= 2(Br)
LB
. At high energies the
bending angle is small, so sin Θ Θ and Θ (Br) . LB

As shown in Fig. 6.1 a quadrupole magnet focuses in one plane and

defocuses in the other one, while it does not affect the particles in its centre.
They are set in doublets and triplets and tuned so that the first magnet focuses
the particles to the centre of the defocusing plane of the second one, which
ensures an overall focusing effect.
The high-energy accelerators use superconducting technologies. Fig-
 Accelerators 61

Figure 6.2: Left: The original drawing of Lawrence’s cyclotron patent.

Right: Lawrence with his first cyclotron in hand

ure 17.4 shows the cross section of a dipole magnet of the Large Hadron
Collider. The LHC keeps two proton beams in orbit in opposite directions,
and so it has two storage rings with two sets of dipole magnets.

6.2 Acceleration
At low energies sufficient particle acceleration was achieved using electro-
static fields. The largest electrostatic accelerators are the Van de Graaff
generators which can provide about 35 MeV energies. The contemporary
accelerators use resonance acceleration in microwave cavities rather than
electrostatic fields: E = −∂ A/∂t. The modern accelerator cavities can pro-
vide as high as 45 MeV/m acceleration. Figure 17.4 shows such a resonator
cavity built for the TESLA project at DESY, Hamburg.
The first cyclotron was built by E. Lawrence in Berkeley in 1929. He
spent 20 $ on it and was awarded the Nobel Prize for it in 1939. It is
a circular machine (Fig. 6.2) consisting of two half-cylinders (D’s): the
particles circulate in orbits of increasing radii as gaining energy in the gaps
between the D’s where the direction of the field changes in phase with the
circulation of the particles. In the synchrotron the radius of the particles is
fixed and the magnetic fields are tuned to keep the particles in the same orbit.
At CERN all circular machines are synchrotrons.
The linear accelerator (nicknamed linac) developed parallel with the
circular ones. It was first built by Rolf Widerøe in 1928; now usually the
first stage of any accelerator system is a linear one. The accelerator cavity is
in principle similar in the circular and linear accelerators, but in the circular
62 Chapter 6

ones the same cavities are used again and again in every turn of the beam,
whereas in the linear accelerators they are passed only once.

6.3 Colliders
In an ordinary accelerator the beam hits a fixed target. For the collision of
identical particles the centre-of-mass energy is ECM 2 = s ≡ (p1 + p2 )2 =
√
2MX Eb + 2MX , i.e. ECM ∼ Eb . With increasing beam energy ECM , the
2

useful energy of the collision that can produce new particles grows with the
square root of the beam energy. If both particles are accelerated, the energy
released in the collision is the sum of the beam energies in the CM system.
Another advantage of the colliders with identical particles is the centre-
of-mass system in the laboratory. This means not only an easier interpretation
of the data, but also the possibility to encircle the interaction point by a com-
pletely symmetric detector system with which one can detect all produced
particles.

6.4 Flux and luminosity

Flux is defined as the number density of particles hitting a unit surface of
the target in unit time. It is naturally connected to the cross section, and so
physicists like to give it in units of 1/(barn s).
Luminosity is the rate of collecting data for colliders, similar to the flux
of fixed-target experiments. It is defined as L = f n N1AN2 where f is the
circulation frequency of the colliding beams; n is the number of particle
bunches in the ring; N1, N2 are the numbers of particles in the two kinds of
bunches; A is the spatial overlap of the colliding bunches. The∫ t2 total number
of collisions is characterized by the integrated luminosity: t Ldt which is
1
usually measured in units of inverse cross section, at LHC in [pb−1, fb−1 ].
The expected detection rate of a reaction with cross section σ at  detection
efficiency is R =  σL.

6.5 Beam cooling

In order to store particles in storage ring, they have to be focused, both
in space and in momentum. The trajectories of particles are focused with
magnets, and in momentum space by stochastic and electron cooling. It is
 Accelerators 63

called cooling in analogy to gases where higher temperature means larger

momentum spread among the molecules.
Stochastic cooling was invented together with many other things in ac-
celerator physics by Simon van der Meer (Nobel Prize, 1984). It is used to
reduce the transverse momentum spread within a bunch of charged particles
in a storage ring by detecting fluctuations in the momentum of the bunches
and applying corrections to reduce them. While acceleration (or in some
cases, deceleration) is done to particles in small, dense bunches, cooling is
most efficient when they are stretched out, debunched. This is technically
done via bridges of microwave conductors (Fig. 17.6) across the storage
ring to time the correction kick to reach the same particle whose wayward
momentum was detected.
Electron cooling, invented by G. I. Budker can be used to low-energy or
slow particles (Fig. 17.6) to reduce their relative momenta. A dense beam of
electrons having the same velocity as the average one of the beam to be cooled
is mixed to the stored particles. The stored heavier particles loose their
momentum spread via Coulomb scattering on the highly monochromatic
electrons.

6.6 CERN’s accelerator complex in the LEP era

For more than 10 years, from 1989 to 2000, two storage rings, the Large
Electron Positron (LEP) collider at CERN and the Tevatron proton-antiproton
collider at the American Fermilab dominated high energy physics. Tevatron
observed the top quark first. Right at the beginning of its work LEP has shown
that there are only three kinds of light neutrinos, and consequently, there are
three fermion families only. After this an incredible amount of precise
LEP measurements tested and confirmed the predictions of the standard
model of particle interactions. At the same time two other large colliders
worked as well: the Stanford linear electron-positron collider at the American
SLAC laboratory and the Hadron-Electron Ring Accelerator (HERA) at the
DESY laboratory in Hamburg. The former–similarly to LEP–studied the
production and decay of the Z boson, while HERA delivered extremely
important information on the quark structure of the proton. At present there
is just one large collider, the LHC.
The scheme of the accelerator complex of CERN in the LEP era is shown
in Fig. 17.5. It was based on the old (but apparently forever living) Proton
Synchrotron (PS) which in every repetition cycle of 14.5 s duration went
through the following phases: proton acceleration for antiproton production,
64 Chapter 6

∫
Year E(e+ e− ), GeV Ldt/4, pb−1 main study
1989–94 91 140 properties of Z
1995 130–136 5
1996 161–172 20 properties of W±
1997 184 60 W+ W− , ZZ production
1998 189 190 W+ W− , ZZ production
1999 192–202 220 search for the Higgs boson
2000 204–209 220 search for the Higgs boson

Table 6.1: The Large Electron-Positron (LEP) collider: energies, integral

luminosities for each experiment and main goal of study.

proton and heavy ion acceleration for the Super Proton Synchrotron (SPS),
electron and positron acceleration for LEP and antiproton deceleration for
LEAR, the Low Energy Antiproton Ring. Let us study these phases.

6.6.1 Electrons and positrons

Electrons of 180 MeV energy from a linac were injected into a tungsten target
to produce positrons via pair production. Both the electrons and positrons
were accelerated in another linac to 500 MeV and then stored in EPA, the
Electron-Positron Accumulator. When the luminosity of LEP was already
low, they were accelerated by the PS to 3.5 GeV and transferred to the SPS
for further acceleration and finally fed into the LEP ring. The electrons
and positrons collided at four interaction points of LEP equipped with huge
experimental detector systems called ALEPH, DELPHI, L3 and OPAL. The
collision energies and luminosities of LEP are summarized in Table 6.1. In
its last two years of study, while searching for the Higgs boson, LEP provided
more data than in the previous 10 years together, in spite of the fact that its
collision energy was well over the enormous yield of the Z peak (Fig. 17.2).
LEP was a very precise machine, its 27 km ring was sensitive to all kinds
of macroscopic and regional factors such as the position of the Moon or
seasonal changes of the rainfall in the area (tide effect of underground water).
In 1995 a strange noise was registered in its magnets and champagne was
promised to the person who finds its source. Somebody found a correlation
with the schedule of the fast train between Geneva and Paris. A parasitic
flow of electricity, originating from the trains, travelled along underground
rivers to the grounding cable and the vacuum chamber of LEP, disturbing
the current of the bending magnets.
 Accelerators 65

LEP energy was severely limited by synchrotron radiation. A relativistic

particle of charge ep in a circular machine of radius r loses its energy due to
synchrotron radiation. The energy loss is

4π ep β γ
2 2 4
ΔE = − (6.1)
3 r

per turn where β = v/c is its relative velocity and γ = 1/ 1 − β2 is the

relativistic factor. This means that the energy loss from synchrotron radiation
sensitively depends on the mass of the charged particle, roughly proportional
to the negative fourth power of its mass. It prevented LEP’s further energy
increase, while for the much heavier proton it is negligible at LHC energies.
In the last two years of LEP running workers started to dig the huge caves
for ATLAS and CMS, the two LHC experiments too large to fit in the place of
a former LEP detector. Simulations predicted that the LEP tunnel will relax,
elevate by several millimetres during this work because of the removal of the
thousands of tons of earth, possibly ruin the extremely precise alignment of
the LEP magnets. The accelerator experts at CERN solved this problem by
continuously correcting the LEP magnets so that the work of LEP was not
disturbed by the underground work.

6.6.2 Protons
The protons started from a duo-plasmotron ion source where the electrons
were shaken off from hydrogen molecules, then accelerated in a proton linac
up to 50 MeV. They were injected into the PS Booster where they got to
1 GeV and injected into the PS. The PS provided:

• proton beams up to 25 GeV/c momentum to the PS experiments in a

continuous extraction mode;

• injected bunches of protons into the Super Proton Synchrotron (SPS)

for further acceleration to 450 GeV/c;

• shot bunches of protons onto an antiproton production target for the

Antiproton Accumulator / Collector storage ring.

The PS Booster also provided 1 GeV protons for the ISOLDE radioactive
beam facility of CERN to prepare radioactive atomic beams.
66 Chapter 6

6.6.3 Heavy ions

The SPS accelerated also ions with atomic mass up to that of lead for heavy
ion physics. Lead ions, Pb53+ are accelerated in a dedicated linac, in the PS
Booster, in the PS and injected into SPS where they are completely stripped
to Pb82+ and after further acceleration ejected for the SPS experiments. The
SPS remains in operation in the regime even in the LHC era.

6.6.4 Antiprotons
25 GeV protons from PS were injected into an iridium target to produce
proton-antiproton pairs; antiprotons of momentum 3.75 GeV/c were col-
lected in the AA-AC (antiproton accumulator - antiproton collector) double
ring to be cooled and stored until the SPS in its SppS collider regime or later
LEAR, the Low Energy Antiproton Ring received them. The PS decelerated
the antiprotons to 600 MeV/c before ejecting them into LEAR that further
accelerated or decelerated them for the LEAR experiments.

6.7 Other accelerators

In addition to CERN there are several particle-physics oriented accelerator
laboratories on Earth, most of them are in the USA, Germany, Japan and Rus-
sia. Here we shall mention only a few of them. Particle and nuclear physics
uses about a hundred accelerators whereas medicine and solid-state technol-
ogy uses a hundred times more, of course, at much lower energies. Later we
shall mostly concentrate on CERN experiments, but here we mention some
of others.

6.7.1 Tevatron at Fermilab

Before LHC Tevatron at Fermilab, the Fermi National Accelerator Labo-
ratory (Batavia near Chicago, USA) was the highest energy accelerator on
Earth. It collided accelerated protons and antiprotons at almost 2 TeV en-
ergy. It had two major experiments, CDF and D0. The last quark with the
highest mass, the t quark was discovered at the Tevatron, although its mass
was estimated much earlier from radiative corrections. The Tevatron was
stopped when LHC started to collect p-p collision data at 7 TeV, later in 2012
at 8 TeV, at orders of magnitude higher luminosities.
 Accelerators 67

6.7.2 HERA at DESY

The DESY (Deutsches Elektronen-Synchrotron) laboratory is located in
Hamburg, Germany. It operated several accelerators, including HERA,
the first accelerator using superconducting magnets. HERA was the only
accelerator in the world that was able to collide protons with either electrons
or positrons. It operated between 1990 and 2007. HERA’s tunnels run 10
to 25 meters below ground level and contained two accelerator rings. Its
two major experiments, ZEUS and H1 watched the collisions of 27.5 GeV
electrons or positrons with protons of 920 GeV, they increased the available
information on the structure of the proton by an order of magnitude.

6.7.3 RHIC at Brookhaven NL

The accelerator physicists at Brookhaven National Laboratory played a pi-
oneering role by designing and building AGS, the Alternating Gradient
Synchrotron. The Proton Synchrotron at CERN was built along the line
laid by the AGS. BNL is a general research laboratory, not only for particle
physics, but also for solid state physics, chemistry and material sciences.
Among other facilities BNL built another ground-breaking machine, the
Relativistic Heavy Ion Collider (RHIC) where ions are collided inside the
detectors of several experiments, the largest ones being STAR and PHENIX.
The types of colliding particles are p + p, d + Au, Cu + Cu, Cu + Au, Au
+ Au and U + U. The typical Au+Au collision energy is 200 GeV/nucleon
pair (that is the usual energy unit of heavy ion collisions) at a luminosity of
3 × 1027 cm−2 s−1 .

6.8 CERN’s facilities in the LHC era

CERN’s accelerator complex after 2008 is shown in Fig. 17.7. The com-
plexity of the system decreased by eliminating EPA, the electron-positron
ring, but made more complicated by several additions, the greatest ones are
the proton injectors from two sides and the neutrino beam.

6.8.1 LHC, the Large Hadron Collider

We shall devote special emphasis on the description of the LHC, today’s
flagship of high energy physics .
Using protons means two advantages against electrons in a collider. In
addition to the much less energy loss due to synchrotron radiation, a proton
68 Chapter 6

machine has much larger discovery potential. As the proton is a composite

particle with quarks and gluons inside, a high-energy p-p collision means
many different possible collisions among quarks and gluons. We shall see
later that the 125 GeV Higgs boson was discovered at LHC in 8 TeV p-p
collisions where the predominant production mechanism was gluon fusion.
When LEP was stopped, the electron-positron system was also dis-
mounted. The LEP tunnel was filled with 9300 superconducting magnets,
the most important ones being the 1232 double dipoles (Fig. 17.8, left) keep-
ing the two circulating beams in orbit opposite to each other. Figure 17.8
(right) shows the radio-frequency cavities accelerating the particles in the
LHC ring.

Protons in LHC

The Large Hadron Collider operates two counter-propagating proton beams,

so it contains two accelerator systems and two storage rings. The fact
that it collides protons instead of protons with antiprotons like the Tevatron
increased its luminosity by many orders of magnitude. As gluon fusion is the
dominant process of new particle production (like that of the Higgs boson),
this did not reduce the discovery potential.
The work of LHC is well illustrated by the official beam diagnostics
page of CERN (Fig. 17.9) taken by one of authors on 12 June 2016 at 18:28
hours while supervising data collection by the CMS detector. The previous
evening at 18:00 the LHC beam was dumped, as the proton intensities in
the two beams (red and blue curves) went already down to 6 × 1013 protons.
The magnet currents were brought to zero at 19:00 hours for a short time (as
shown by the black curve corresponding to the right axis) and shortly later
brought back to 450 GeV SPS energy. Around 22:30 the two LHC beam
pipes were refilled by protons in an hour (in 24 steps by 96 bunches in each)
by 2.4 × 1014 protons in each direction. After cooling and focusing, while
the magnet currents were increased step-by-step, in about a quarter hour
the two beams were accelerated up to 6500 GeV. After further cooling and
squeezing the beams and adjusting them to collide in the geometric centres
of the detectors, stable collisions were reached at about midnight and the
detectors could be turned on fully. At the moment of taking the snapshot the
LHC data were collected already since 18 hours in stable collision regime,
but the beam intensities were down at 70% of the initial values. All five
experiments, the huge ATLAS and CMS, the somewhat smaller ALICE and
LHCb and the much smaller TOTEM were working in their stable PHYSICS
mode.
 Accelerators 69

The LHC detectors are quite different from each other. ATLAS and
CMS can handle 15 times higher luminosities than LHCb, and 1500 times
higher than ALICE. Depending on the circumstances the LHC could deliver
stable collisions for several days with one fill, although they were frequently
interrupted by glitches in the electric network. The accelerator has many
information pages, for instance we could learn that at the time of taking
Fig. 17.9 the two proton beams met at the collision points at an angle of
170 μrad different from the head-on 180 degrees, their overlap was 25 mm
along the beam direction and 20 × 16 μm in the x − y plane orthogonal to
the beam. The proton bunches followed each other by 25 ns and each ring
contained 2220 proton bunches. In ATLAS and CMS there were 2208 bunch
crossings, in LHCb 2036, and in ALICE 1940.
LHC was started with protons in 2009 and went through an incredible
development. During 2010, its first full data taking year, the luminosity
went through a gradual increase of 8 orders of magnitude in 8 months, it
delivered 46 pb−1 integrated luminosity of proton-proton collisions at 7 TeV
energy. In 2011 it worked at 7 TeV and provided 4.5 fb−1 . In 2012 the
integrated luminosity was 23 fb−1 at 8 TeV. That was the time of observation
of the Higgs boson. After a long shutdown in 2015, the LHC restarted
with 6.5 + 6.5 GeV p-p collisions: it provided 4 fb−1 in 2015, 40 fb−1 in
2016 and 51 fb−1 integrated luminosity in 2017. Of course, there are always
inefficiencies, things sometimes break down and the inner detector part of the
detector is always delayed a bit behind the announcement of “Stable beams”.
Nevertheless, the experiments did very well, with beam usage efficiencies
well above 90%.

Heavy ions in LHC

Lead ions are prepared and stored in LEIR, the Low Energy Ion Ring that
was built using the elements of LEAR (Fig. 17.6). The lead ions are then
injected to PSB, PS, SPS and LHC for acceleration. At the end of 2010 and
2011 LHC worked a few weeks with Pb-Pb collisions, and after its 2012
proton run, it collided Pb ions with protons. The beam energy of the LHC in
its heavy ion regime was 2.76 TeV / nucleon, so for reference LHC was also
running for a short while colliding protons at 2.76 TeV. For the same reason,
after the 13 TeV proton runs similarly reduced collision energies were also
studied: 5.02 and 8.16 TeV/nucleon p-Pb collisions in 2016, and 5.02 TeV
p-p collisions in 2017. Heavy ion physics will be treated in Chapter 15.
70 Chapter 6

6.8.2 Neutrinos
SPS has a special facility that provides a neutrino beam to study neutrino
oscillations: CERN Neutrinos to Gran Sasso (CNGS). The Italian Laboratori
Nazionali del Gran Sasso (LNGS) is the largest underground laboratory on
Earth devoted to particle and nuclear physics. It is located at 120 km from
Rome near a highway tunnel through the Gran Sasso mountains, under 1.4 km
of rock. It hosts 15 experiments of 900 scientists from 29 countries. We shall
describe the neutrino beam in Chapter 13 devoted to neutrino experiments.

6.8.3 Antiprotons
CPT invariance, the equivalence of matter and antimatter is deeply embedded
in particle physics. However, there are no antimatter galaxies in the Universe,
although during the Big Bang particles and antiparticles should have been
produced in identical quantities. There are also CPT violating extensions of
the standard model. All this demands the precise check of CPT invariance.
LEAR, the Low Energy Antiproton Ring was built mostly for meson spec-
troscopy and it was stopped when the major LEAR experiments finished data
taking. LEAR was converted into LEIR, but the CPT experiments asked
for a low-energy antiproton source. Thus CERN built the Antiproton Decel-
erator facility using the space and elements of the dismounted Antiproton
Accumulator / Collector rings (Fig. 17.10).
The Antiproton Decelerator (AD) works the following way. An extracted
beam from the Proton Synchrotron of 1.5 × 1013 protons at p = 26 GeV/c
momentum is produced in 5 pulses with a total duration of 500 ns. These
protons hit an iridium target where they produce proton-antiproton pairs.
Antiprotons of 3.57 GeV/c momentum are collected and focused using the
magnetic horn technique developed by Simon van der Meer (Nobel Prize,
1984). The antiprotons are injected into the AD ring where they are decel-
erated to 100 MeV/c in four steps, in the first two steps with stochastic and
then electron cooling (Fig. 6.3).
The AD delivers 3 × 107 antiprotons at 100 MeV/c momentum to several
(seven in 2017) experiments, which trap them in electromagnetic fields and
using slow positrons make antihydrogen (pe+ ) atoms. ALPHA and ATRAP
prepares spectroscopy on trapped antihydrogen, ASACUSA and BASE com-
pares the properties (mass, charge and magnetic moment) of protons and
antiprotons at high precision, AEGIS and GBAR tries to measure the gravi-
tational mass of antihydrogen, and ACE studies the effects of antiprotons on
living tissue.
 Accelerators 71

Figure 6.3: Deceleration and cooling of antiprotons in the Antiproton De-

celerator.

Viewing the success of the AD experiments CERN decided to increase

their efficiency by building a small storage ring ELENA (Extreme Low
Energy Antiprotons) which will supply an order of magnitude higher number
of slow antiprotons for trapping.

Exercise 6.1
What is the energy loss per turn of 100 GeV electrons and 7 TeV protons in
the LEP/LHC tunnel of 27 km circumference due to synchrotron radiation?

Exercise 6.2
What is the energy limit one can reach with protons in a ring with a circum-
ference of 27 km filled with 8 T magnets?

Exercise 6.3
Why is it impossible to keep the LHC detectors ready to take data during
beam acceleration and adjustment?

Exercise 6.4
LHC was designed for 7 + 7 TeV collisions, but so far that was impossible
to reach. According to Fig. 17.9 the 6.5 TeV beam energy could be reached
72 Chapter 6

very fast. What could be the reason why 7 TeV beams were not delivered in
2016-17?
Exercise 6.5
How do you get the 2.76 TeV / nucleon energy for the Pb + Pb run of LHC?
Hint: To what collision energy does it correpond for protons and lead ions?
Chapter 7

Detectors, calorimetry

MOTTO:
Experimenter: A physicist who does experi-
ments. Theorist: A physicist who does not do
experiments.

(Leon M. Lederman)

7.1 Event registration

In high energy physics one tries to detect and identify as many of the par-
ticles produced in the beam collisions as possible. The main method is
event registration (see Ch. 8). It means that after a suitable trigger signal
we record all relevant information. The trigger may be physics or apparatus
related. Physics triggers can be many things: observation of a muon or an
isolated electron, large missing momentum (from a high-energy particle that
escaped detection) or many particles in jets. In the high-energy experiments
the various physics analysis groups define the triggers related to their in-
terest, sometimes hundreds of different conditions. Other triggers could be
apparatus-related, for instance laser pulses lighting up various parts of the
detector for testing, or reading environmental measurements like the temper-
atures of the various parts of the equipment or the magnetic field strength.
Of course, only those parts of the detector are read out that fired and are
73
74 Chapter 7

related to the given trigger.

It is very important to collect and record all (but at least as much as
possible) energy produced in the collisions. In analogy with thermal mea-
surements that is called calorimetry and the detector elements involved are
calorimeters. The high energy experiments generally have electromagnetic
and hadron calorimeters, the former to catch electrons and photons, the latter
for the hadrons.
In 2012 the LHC provided 20 million collision events per second, of
which about 400 could be stored by CMS and ATLAS. With the development
of the techniques in 2016-17 already 1000 were stored of the 40 million
events per second of the LHC. Thus a very careful selection was made by
the experiments to choose which events to store. That was done in two steps
(two trigger levels) by CMS and three by ATLAS. For CMS the first level
reduced the event rate to 100 kHz, while the second one to 1000 Hz. Of
course, signals belonging to the same event are automatically in coincidence
with each other. Every collision event contains many p-p collisions (called
pileup, see Fig. 17.11). In 2016-17 that was already above 50 at average.
This overwhelming amount of signals increases the discovery potential, but
also the difficulties of data analysis. The detected signals in an event are
automatically in coincidence, which is therefore unnecessary to prescribe
in the analysis, but the particle tracks and calorimeter hits must originate
from the same interaction point, called vertex, to belong to the same physical
phenomenon. In low-rate experiments of nuclear studies one prescribed time
coincidence for the events. In present-day high-energy experiments we have
to demand precise space-time coincidence for the origins of particle tracks
in the events.

7.2 Energy loss in matter

Fast particles lose their energy in collisions with the atoms of the medium.
At lower energies, around 1 MeV Coulomb scattering dominates. Let us
consider a particle of mass M and charge z × e moving at velocity v = cβ
along the x axis in a medium of atomic number Z and atomic weight A. The
forces along its trajectory compensate
∫ each other, there will be no exchange
2
of momentum along x, px = Fx (t)dt ≈ 0. It applies a force Fy = x 2ze+b2 on
an atomic electron (assumed to be at rest) located at longitudinal coordinate
x in a cylinder at a radius b (Fig. 7.1) and thickness db. The momentum loss
 Detectors, calorimetry 75

Figure 7.1: Derivation of the Bethe equation: a particle of mass M and charge
ze moving at velocity v interacts with the atomic electrons in a cylinder of
radius b of a medium

of the particle along an infinitesimal distance dx will be

∫
py = Fy (t)dt =

∫ ∫
ze2 b dx ze2 b dx 2ze2
√ = =
x 2 + b2 x 2 + b2 v v (x 2 + b2 )3/2 vb
The energy transfer from the incoming particle to a single electron is

p2e 2z2 e4
= .
2me me v 2 b2
We have nZ(2πbdbdx) electrons in unit length of the cylinder, so in order
to calculate the average energy loss per unit path length we have to integrate
over all impact parameters b between its minimal and maximal values:
∫  2  
dT bmax
2 ze2 4π(ze2 )2 nZ bmax
− = nZ 2πb db = ln
dx bmin me vb me v 2 bmin
The minimum distance cannot be shorter than the de Broglie wavelength,
bmin ≈ mhe v , whereas the maximal distance is limited by the ionization
γhcβ
potential of the medium, ie. the minimal transferable energy: bmax ≈ I .
76 Chapter 7

Here γ = 1/ 1 − v 2 /c2 and I is the average ionization potential of the

2
medium, which can be roughly estimated by I ≈ 10Z eV. Thus bbmax min
≈ meI v
and the total energy loss with relativistic and atomic shell corrections is
 
dE 2Z 1 1 2me c2 β2 γ 2Tmax δ C
− = Kz ln −β − −
2
(7.1)
dx A β2 2 I2 2 Z
This is called the Bethe–Bloch equation. Here x is the thickness of the
medium in units of g/cm2 , K = 0, 307 MeV · cm2 /g a constant, and Tmax is
the maximal transferable energy to an electron. The Bethe–Bloch equation
can be written in a simplified form using the approximation
2me c2 β2 γ 2
Tmax = ≈ 2me c2 β2 γ 2 (2γme /M 1)
1 + 2γme /M + (me /M)2
as  
dE z2 Z 2me c2 β2
− =K 2 ln − β 2
.
dx β A I(1 − β2 )
Note that here the energy loss does not depend on the mass m of the
particle, only on its velocity. Figure 7.2 shows the energy loss of a muon in
copper as a function of the muon momentum. There is a wide minimum-
ionization region between 0.1 and 100 GeV/c momentum which practically
encompasses the energy range of the muon in high-energy experiments and
that is why the energetic muons cannot be contained in the detector volumes.
Nuclear losses are negligible for muons, but they dominate for the strongly
interacting particles. That is why pions and muons have so vastly different
penetration depths in matter although they have quite similar masses.
For high-energy electrons radiation  is the most
 important means of en-
ergy loss. The total loss is dx = dx
dE dE
+ dE dx rad . Critical energy is
ion
the energy at which the radiation
  energy
  loss becomes equal to the ioniza-
tion one for an electron, dx dE
= dx dE
. In rough approximation it is
ion rad
Ec ≈ 800MeV
Z+1.2 . Radiation length X0 is defined as the thickness of matter at
−) 
which dE(e
dx rad = − X0 .
E

For composite materials (element j is present at weight fraction w j ) we

can use Bragg’s rule
   
dE dE 
= wj
dx j
dx  j
and    
Z Zj j nj Zj
= wj = .
A j
Aj j nj Aj
 Detectors, calorimetry 77

Figure 7.2: Energy loss of a positive muon in copper. Up to 500 GeV the
Bethe-Bloch mechanism works well for muons. Below the minimal ioniza-
tion region, 100 MeV/c, for different energy regions different approximative
calculation methods are used. Above the Bethe-Bloch region radiative losses
dominate for muons

Bragg’s rule can be used for calculating the radiation length as well, but it
overestimates the average ionization potential for molecules as the electrons
have tighter bonds there than in the atoms.
Electrons lose energy via radiation and photons via bremsstrahlung1 and
pair creation. As a result, both produce electron-photon showers, and thus
electrons and photons behave very similarly at high energies [Leo, 1987].

7.3 Particle identification

Practically all detectors of high-energy physics have strong magnetic fields
inside. It is usually a solenoid, but ATLAS combines a solenoid encircled
by a huge toroidal field (Fig. 17.12).
There are several methods to identify the particles in an event. From
1In English the original German word is used for the radiation emitted in the
deceleration of charged particles in matter; other languages use the simple mirror
translation of the German word: braking radiation
78 Chapter 7

Det. type space res., rms (μm) time res. dead time
Photo-emulsion 1
Bubble chamber 10 − 150 1 ms 50 ms
Streamer chamber 300 2 μs 100 ms
√
Prop. chamber 50 − 300 (d/ 12) 2 ns 200 ns
Drift chamber 50 − 300 2 ns 100 ns
Scintillator 100 ps/n 10 ns
LAr drift chamber 170 − 450 200 ns 2 μs
Microstrip ch. 30 − 40 < 10 ns
Resistive plate ch. ≤ 10 1 − 2 ns
Silicon strip d/(3 − 7) readout readout
Silicon pixel 2 readout readout

Table 7.1: Detector types used in high-energy physics. The time resolution
of the various scintillators depends on the material and size of the detectors.
For the silicon detectors the read-out speed defines the timing

the trajectory of a charged particle in a magnetic field one can determine its
charge and momentum. A mixed beam of different particles with a given
momentum can be separated by their time of flight between two detectors.
Different particles lose their energy in collisions with the atoms of a medium
at different rates approximately at dE/dx ∼ z 2 /β2 where x is the thickness
of matter, z is the charge of the particle in units of the electron charge and
β = v/c is its relative velocity. Cherenkov radiation is also characteristic of
a particle velocity: when a particle flies in a medium faster than the light, it
emits Cherenkov light at an angle θ c = arccos nβ 1
where n is the refraction
index of the medium.

7.4 Detector types

Some properties of the detector elements used in high-energy physics are

summarized in Table 7.1
 Detectors, calorimetry 79

7.4.1 Multiwire chambers

The invention of multiwire chambers by Georges Charpak in the sixties
revolutionized experimental particle physics. It made the visualization of the
events electronic, the experiments fully automated and orders of magnitude
faster than the previous method of bubble chambers with photographing
events and later processing by hand. He published the method in 1968 and
was awarded the Nobel Prize in 1992.
In a multiwire chamber the particles ionize a gas, the electrons are
collected by anodes of very thin wires (typically gold-coated tungsten of 10-
20 μm thickness) with cathodes of metal plates, semiconductor pads or thick
steel wires. The gas should have an electron-collector quencher admixture
in order to prevent a self-supporting long discharge. Typical such mixture is
argon + isobutane with some freon added.
In the classical chambers there are many layers of wires placed in or-
thogonal directions with both ends read out: the position of the particle
trajectory is determined by the position of the wire and the charge ratio at
the two ends. This was changed as the detector volumes increased. The drift
chambers gather one of the coordinates by the drift time of the electrons to
the wire. The time projection chambers (TPC) combine the drift times with
silicon pads on the bottom, which detect the mirror charges of the trajec-
tories (Fig. 17.13). Drift chambers and TPC’s sacrifice detection speed for
spatial resolution: because of its TPC tracker, ALICE is better at particle
identification, but it can take much less events than ATLAS or CMS that
have semiconductor trackers. In the case of heavy ion collisions this is not a
problem as they happen much less frequently than p-p collisions.

7.4.2 Scintillation counters

The scintillation detectors (called counters although they rarely count) are
the most widely used detectors in nuclear and particle physics: they are fast
and have an excellent energy resolution. Table 7.2 summarizes the most
important types of scintillators. NaI(Tl) has the highest light yield, but
it is highly hygroscopic, so hard to handle. Polystyrene is the fastest and
PbWO4 has the highest specific weight while almost as fast as polystyrene.
That is why ALICE and CMS use PbWO4 single crystal scintillators in their
electromagnetic calorimeters (Fig. 17.15).
The light from scintillators is converted to electric pulses using photo-
multipliers or avalanche photo-diodes (APD). Photo-multipliers give higher
gains than APD-s, but occupy more space and are sensitive to magnetic
fields.
80 Chapter 7

Scintillator photon- τ λmax ρ dE

dx X0
yield ns nm g/cm3 MeV/cm cm
NaI(Tl) 1 250 410 3.7 4.8 2.59
polystyrene 0.12 3–5 360–480 0.9 12
CsI(Tl) 0.4 1000 565 4.5 5.6 1.85
BGO 0.15 300 410 7.1 9.0 1.12
PbWO4 0.01 5–15 420–440 8.3 10.2 0.9

Table 7.2: Basic properties of various scintillators: name, light yield relative
to NaI(Tl), decay time of the signal, maximum of wavelength distribution,
density, stopping power and radiation length.

7.4.3 Shower detectors

If the particles to be detected deposit little energy in unit volume they are
forced to penetrate dense materials (some heavy metal) interleaved with
scintillators or other active detector elements to detect secondary particles
from the interaction of the primary particles with the dense medium. Very
frequently the return yoke of the main magnet of the detector serves as the
dense matter. The muon chambers of CMS are such shower detectors in
between thick layers of steel.
In high energy calorimetry a special requirement is the complete coverage
of the interaction point which leaves very little space for photon detectors.
The solution is the scintillation tile: wavelength-shifting (green) fibres are
embedded in the (blue) scintillators, which collect the light and transport it to
the photon detectors (Fig. 17.16). This technique makes it possible to build
walls of flat scintillators with no gap between them. The hadron calorimeter
of CMS is a shower detector system, consisting of 4 mm thick scintillator
tiles between 50 mm thick brass plates. For the latter over a million brass
shell casements were melted, left behind by the Russian navy after World
War II. (It was not enough, the USA added 1 million dollars worth of copper
to finish them.)

7.4.4 Cherenkov detectors

Cherenkov detectors are widely used at LHC. Both ATLAS and CMS uses
them to detect the forward scattered particles, which provides important
information to measure the instantaneous luminosity of the experiment. Fig-
ure 17.17 shows the structure of the forward hadron calorimeter of CMS,
 Detectors, calorimetry 81

consisting of quartz fibres in steel to collect Cherenkov light from secondary

electrons. Here the Cherenkov radiator is quartz, very resistant to radiation.
The electromagnetic calorimeter of the OPAL detector at LEP consisted of
large blocks of lead glass read by photo-multiplier tubes. The ATLAS for-
ward detector is filled with freon gas (C4 F10 ) as Cherenkov radiator just like
the ring-imaging Cherenkov counters of LHCb (CF4 and C4 F10 ).

7.4.5 Transition radiation detectors

In a medium containing ne quasi-free electrons in unit volume an electro-
magnetic radiation with ω frequency will have a phase velocity of
c c
vf = = (7.2)
n ne e2
1− ε0 me ω 2

where n is the refraction index. It can propagate if v f > 0, i.e.


ne e2
ω > ωp = (7.3)
ε0 me

where ω p is the plasma frequency of the medium. Of course, in condensed

matter the plasma frequency can deviate from Eq. 7.3 as there the electrons
are not free.
Transition radiation is emitted when a fast charged particle passes the
boundary of two media with different dielectric constants. It is caused by
the continuous change of the electric displacement, D = εε0 E through a
boundary while E jumps. When a charged particle approaches a boundary,
with its mirror charge on the boundary they make a dipole which changes
in time, that is why the radiation is emitted. This is significant for extreme
relativistic particles only with γ = 1/ 1 − β2  1.
When such a particle passes a thin absorber (foil) with plasma frequency
ω p1 and enters a medium with ω p2 , it radiates photons of frequency ω  ω p1
with a double differential energy distribution [Leo, 1987]
  −1
dEtr d1 ω −1 ω2p1
= γ + θ c2 + 2 (7.4)
dωdΩ 2c ω

where d1 is the thickness of the foil and Ω is the solid angle. This is a radiation
in the X–ray region, which is strongly forward peaked with a cone angle of
θ c ≈ 1/γ. The total energy radiated for ω p1  ω p2 is Et r = αz 2 γω p1 /3.
82 Chapter 7

A transition radiation detector (TRD) usually consists of hundreds of

thin foils and a chamber to detect X-rays. ATLAS has a transition radiation
tracker system made of straw tubes, small, thin multilayered tubes with
proportional counter wires inside.

7.5 The CMS detector

All detector systems of high-energy collider experiments have similar struc-
tures. They consist of a barrel and two endcaps, both kinds are built of
layers of different tasks. For the sake of presentation we consider the CMS
(Compact Muon Solenoid) detector as a typical example (Fig. 17.14). The
innermost layer, closest to the interaction point, right against the beam pipe is
a 3-layer silicon pixel detector of high modularity, to make possible to iden-
tify the interaction point from which the given particle track originates. It is
surrounded by many layers of silicon strip detectors, the tracker to trace the
trajectories of charged particles curving in the magnetic field. The tracker is
surrounded by the electromagnetic calorimeter, consisting of 75,848 PbWO4
single crystal scintillators (Fig. 17.15), which absorbs all photons and elec-
trons measuring their energies. The next layer is the hadron calorimeter for
absorbing all energies carried by hadrons: pions, kaons, protons, neutrons,
with its brass plates and scintillator tiles (Fig. 17.16).
All CMS calorimeters of the barrel are inside the largest superconducting
solenoid on Earth: in its volume of 6 m inner diameter and 13 m length
(Fig. 17.18) it can maintain a magnetic field of B = 4 Tesla, although it was
up to now used at 3.8 T only to ensure its longevity. The outermost layer of
the CMS detector around the magnet is the muon system: steel rings (return
yoke) of the magnet interleaved with chambers to detect the muons emitted
in high-energy collisions. The endcaps have a somewhat similar structure
with elements close to the beam pipe to detect forward-scattered particles.

Exercise 7.1
What are the advantages and disadvantages of drift chambers against ordi-
nary multiwire chambers and scintillation counters?

Exercise 7.2
Arrange the following particles according to their penetration depths in
matter: muons, pions, protons, electrons, photons, neutrinos.
 Detectors, calorimetry 83

Exercise 7.3
Why does the muon has so much longer free flight path in matter than the
pion that has almost the same mass?

Exercise 7.4
Arrange the following detectors in order of detection speed: scintillator,
bubble chamber, emulsion, multiwire chamber, Cherenkov detector, drift
chamber.
Exercise 7.5
Arrange the following detectors in order of energy resolution: scintillator,
bubble chamber, multiwire chamber, Cherenkov detector, drift chamber.
Chapter 8

Event registration

MOTTO:
For those who want some proof that physicists
are human, the proof is in the idiocy of all the
different units which they use for measuring
energy.

(Richard P. Feynman)

As mentioned in the previous chapter, the basic experimental method of

high energy physics is event registration: particle collisions are received and
studied one-by-one, and whenever a pre-defined trigger condition is fulfilled,
the relevant parts of the detector (i.e. those which fired and/or related to the
given trigger) are read out and the data are stored. There are physics and
detector related events. In a physics event we have particle tracks in the
tracking system and energy deposits in the calorimeters, and from all that
one can reconstruct what happened in the detector.
Events are detected and simulated. The analysis of events is usually made
by comparing the detected events with the results of simulations using the
known processes. Events are simulated via creating particles with the Monte
Carlo method and then letting them go through the detector simulation. This
way we make sure that the simulated events are as close to detected ones as
possible. New phenomena are searched for (and possibly observed) above
the background of known processes when all possible reactions are taken
85
86 Chapter 8

into account weighted by their cross sections. Of course one has to check
first how well the simulation reproduces the known processes. At the end
that has to be accounted for in the final uncertainty of the result.
Let us view a few typical event pictures recorded at LEP, Tevatron and
LHC.

8.1 LEP events

OPAL (Omni-Purpose Apparatus for LEP, Fig. 17.19) was a typical high-
energy calorimeter at LEP (the authors were its participants). Its main layers
were from inside out: a vertex detector to identify the interaction point, a
tracker made of multiwire drift chambers, the magnet, the electromagnetic
calorimeter made of lead glass Cherenkov counters, the return yoke of the
magnet instrumented with drift chambers serving as the hadron calorimeter,
and the outermost layer, the muon chambers.
In the first years of LEP’s operations (LEP1) the main goal was to study
electron-photon processes at high-energy collisions (Fig. 17.20) and the
properties of the Z-boson. It can decay to leptons (Fig. 17.21) and to quarks
(Fig. 17.22).
Figures 17.20 and 17.21 show how electrons and photons are detected
in the tracking system and in the electromagnetic calorimeter. The energy
deposited in the calorimeter is characterized by the size of the yellow hit
cluster. High energy electrons barely bend in the magnetic field and photons
do not leave any trace in the tracker elements. They are both absorbed in the
electromagnetic calorimeter, depositing all their energy there. Figure 17.20
shows electron-positron annihilation to two photons and a decay of a Z boson
to two neutrinos accompanied by emission of a photon in the initial state by
one of the charged particles. Neutrinos are not detected at all in accelerator
calorimeters. Their creation and escape is observed by missing momentum
only. In Fig. 17.21, middle and right the electron-positron collisions form Z
bosons that subsequently decay to muon pairs or tau pairs. The tau lepton
is very interesting as it can decay quite various ways: to its neutrino and
electron + neutrino or muon + neutrino pairs and also to its neutrino and
hadrons.
In electron-positron collisions at high energy quark production is very
frequent and it is easy to identify by the presence of jets, sets of collimated
particles (like a disperse beam). Fig. 17.22 shows events with two, three and
four jets. The 2-jet events confirmed the existence of quarks, as without the
requirement of fragmentation (colour neutrality of all detected particles), it
 Event registration 87

Figure 8.1: Dependence of pseudorapidity η = − ln tan θ2 on the polar angle

is impossible to explain the emission of dozens of charged particles. The

existence of gluons was proven by the 3-jet events: fermions can be produced
in pairs only and the third jet must belong to a boson with colour chargeThe
4-jet event demonstrates the pair production of W-bosons.

8.2 Transverse momentum, pseudorapidity

Hadron collisions are complicated because we collide composite systems
with many constituent particles. In a proton-proton collision most of the
constituents will scatter in the forward direction; as the decay fragments of
new particles will be emitted in all directions, whatever is orthogonal to
the beam direction is interesting. There are several quantities characteriz-
ing this feature, the most important ones being transverse momentum and
pseudorapidity.
The transverse momentum is defined as the projection of the particle
momentum on the plane orthogonal to the beam direction. As in accelerator
experiments the beam is usually considered to be along the z axis, the
transverse momentum is pT = px + py .
As the initial state is cylindrically symmetric with respect to the beam
axis, therefore the two direction coordinates are conveniently chosen as the
88 Chapter 8

φ azimuthal and θ polar angles. Note that at beam crossing, where the
collisions occur, the beams actually intersect each other at a very small angle
different from zero (so the beams indeed cross each other), the transverse
momentum corresponding to this can be neglected. While the transverse
momentum and azimuthal angle are invariant under Lorentz boost along the
z axis, the polar angle is not so. Thus instead of the polar angle we can use
rapidity that is invariant under Lorentz-boost:
 
E + pL
y = 12 ln (8.1)
E − pL
where E and pL are the energy and longitudinal momentum of the particle.
In high-energy collisions the mass of the detected particles (mostly leptons,
pions and kaons) is negligibly small as compared to their energy and their
rapidity is simplified to pseudorapidity, η = − ln tan θ2 (Fig. 8.1), which is
much easier to calculate as for that one does not need to determine the energy
of the particle. Furthermore, the kinematics of the events are such that most
of the secondary particles fly out at small angles and pseudorapidity gives a
more uniform scattering distribution than the polar angle.
Figure 8.1 shows the dependence of pseudorapidity on the polar angle.
In the CMS experiment pseudorapidities |η| < 2.1 − 2.5 (the exact value
depends on the subdetector) belong to the barrel region, outside that is the
forward region.
At hadron colliders the number of events is frequently plotted on an η − φ
picture which corresponds to the plane that you get if you cut the cylindrical
part of the detector at a given radius along the z axis and fold it out. The
angular distance of two tracks is expressed at hadron colliders in the rather
peculiar angle-like quantity

ΔR = Δφ2 + Δη2 = (φ1 − φ2 )2 + (η1 − η2 )2 .

Another very important quantity is the missing (transverse) momentum,
which means an imbalance in the momentum distribution of the event due
to a neutrino or other undetected particle. In the case when leptons collide
the total centre-of-mass energy is known, and adding up the momenta of
the detected particles we get what is called visible momentum (energy).
The missing momentum (sometimes called missing energy) is the difference
between the collisional and visible momenta (if particles of the same mass
and energy collide, the collisional momentum is, of course, zero). Depending
on how it is calculated, even at hadron colliders we use the quantities of
missing energy or even missing mass, calculated from the missing transverse
momentum.
 Event registration 89

8.3 Observation of the top quark

The t–often called top–quark is the heaviest among the quarks and after
formation immediately decays to a b quark: t→bW+ . The b quark is easy to
identify as it has a long lifetime, therefore if a reaction produces a b quark,
the event may have a secondary vertex from such a “long-lived” particle.
Another characteristic feature of the b quark is that it decays predominantly
through two or three virtual W bosons and so together with jets several
leptons could also be produced. Identifying the b quark is called b-tagging
and it is quite important in high energy physics.
At the Tevatron the t quark was produced in the pp→tt reaction and the
t quarks decayed to b quarks. The event reconstructed in Fig. 17.23 (left)
is a tt created in pp collision: we required two identified b quarks from the
t quark decays, 2 jets from one W-boson and muon + missing momentum
from the other.

8.4 Mysterious events

There are several events which apparently contradict the standard model, but
there are too few of them to draw conclusions. The most famous among
them is the CDF event where a proton+antiproton collision produced an
electron, a positron, two photons and some missing energy (Fig. 17.23,
right). Thus all the coloured content–quarks and gluons–of the colliding
hadrons simply disappeared. When this event was published by the CDF
collaboration an avalanche of theoretical papers tried to interpret it in terms
of various extensions of the standard model. Later the Tevatron experiments
tried to find similar typhoon events for several years, in vain.
As we shall describe it in detail in Chapter 9 of data analysis, in particle
physics we accept a new discovery, if (1) its signal is statistically significantly
above background, (2) it is reliably and convincingly presented, and (3)
confirmed by another, independent experiment. These conditions were all
fulfilled when the observation of the Higgs boson was announced in 2012 at
the LHC. ATLAS and CMS reported the observation of a new boson of the
same mass, about 125 GeV, at more than five times the total experimental
uncertainty above the background. That it had indeed been the Higgs boson
as predicted by the standard model was confirmed only later, on the basis of
a 4 times larger data set. The observation of this pp→e+ e− γγ + Emiss event
clearly does not fulfil the conditions of discovery.
90 Chapter 8

Exercise 8.1
What is the expected number of events of a signal with the cross section of
2.5 pb at a detection efficiency of 20% if the collected luminosity is 10 fb−1 ?

Exercise 8.2
Group the following data whether they must or may be written as part of
the registered events: amplitude and timing of detector hits, temperatures
of detector elements, temperatures of parts of the electronics, temperature
of the counting room, atmospheric pressure, field strengths of the detector
magnet, mains voltage, comments of the physicists on shift.

Exercise 8.3
How do multi-jet events like those in Fig. 17.22 prove the existence of quarks
and gluons?

Exercise 8.4
Why do transverse momentum and pseudorapidity play more important roles
at hadron collisions than at leptonic ones?

Exercise 8.5
In what conditions is pseudorapidity invariant under Lorentz boost?

Exercise 8.6
Why do we expect the t quark to decay predominantly to a b quark?
Chapter 9

Data analysis

MOTTO:
If you are receptive and humble, mathematics
will lead you by the hand.

(Paul A. M. Dirac)

In high energy physics data analysis is very critical for several reasons.

• The experiments are extremely expensive in many senses: the acceler-

ators and detectors are designed and built for decades, their operation
needs the continuous works of thousands of physicists, engineers and
technicians, and the accelerator time itself is extremely expensive.
Thus one does not want to spare the effort to extract the maximum
possible information from the data.

• The discovery potential of these experiments is very high and that

makes it important to minimize the possibility of biased analysis, of
hasty or even false discoveries.

• The huge collaborations (CMS and ATLAS each have more than 3000
participating researchers) can afford to let several groups analyse the
same data, and that makes a strong competition for the best analysis
that can make it to the paper at the end. The competing groups are the
best testers for each others’ work quality.
91
92 Chapter 9

9.1 Statistical concepts of particle physicists

Those are as different from the official mathematical statistics as mechanical
engineering from the Lagrangian or Hamiltonian formulation of theoretical
mechanics. At the same time statistics is extremely important for data
analysis in particle physics: every few years international workshops are
organized by particle physicists working at the Large Hadron Collider to
exchange ideas on the newest developments of the statistical methods, the last
one in 2011 and its proceedings was published in [Prosper and Lyons, 2011].
In the Appendix of that thick volume Eilam Gross defines the aim of his paper
entitled LHC Statistics for Pedestrians in a rather witty way: ”A pedestrian’s
guide . . . to help the confused physicist to understand the jargon and methods
used by HEP phystatisticians. . . . A phystatistician is a physicist who knows
his way in statistics and knows how Kendall’s advanced theory of statistics
book looks like.”
Every high-energy collaboration has phystatistician experts and they all
have quite different ideas how to analyse data. In order to avoid confusion, the
large LHC collaborations have Statistics Committees which maintain home
pages of recommendations how to do things. The Statistics Committees of
both CMS and ATLAS have several members who published text books of
the type Statistics for Physicists and the two collaborations have a joint such
committee as well.
In high energy physics the experimental data are events with many dif-
ferent variables as characteristics: tracking properties, energies deposited
in various subdetectors, places and timing of detector hits. The results of
event selection are counts collected in spectra, often equidistant intervals of
a variable, called bins and the probability that we have ni events collected in
bin i follows the Poisson distribution
μini e−μi
P(ni | μi ) = ,
ni !
with μi being the average value (or mean value) of ni . There are many differ-
ent ways of treating these spectra, we can build e.g., a likelihood function to
be optimized in order to get the desired physics result, an optimal estimation
of the physical parameters to be determined. On the basis of the Poisson
distribution one can build a Poisson likelihood: L = Πi P(ni | μi ) where the

expected number of events is μ̃i = j Lσj  ji , L is the integral luminosity
collected, σj is the cross section of source j and  ji is the efficiency (usu-
ally determined by Monte Carlo simulation) of detecting the contribution of
source j in bin i. Note that μ̃i is an estimation of μi .
 Data analysis 93

According to the general convention in accelerator experiments a given

new phenomenon is excluded if we can show it not appearing at a confidence
level above 95 % and observed if it exceeds 5σ above background where
now, for a change, σ is the experimental uncertainty of the excess according
to the best honest guess of the experimentalist. These two conditions will be
duly explained and refined later in this chapter.
Another important feature of high-energy data analysis is the blind anal-
ysis. According to [Klein and Roodman, 2005]: “A blind analysis is a mea-
surement which is performed without looking at the answer. Blind analyses
are the optimal way to reduce or eliminate experimenter’s bias, the unin-
tended biasing of a result in a particular direction.” It came from med-
ical research and in high-energy physics it means to optimize, prove and
publish your analysis technique using simulations and earlier data only be-
fore touching new data in the critical region. For instance, in spring and
early summer 2012, the new CMS data were blinded in the mass region
110 < mH < 140 GeV (where mH is the simulated Higgs mass) because
of the 3σ excess observed in 2011. The same procedure was used again in
autumn 2012, after the announcement of the 5σ observation. The analysing
methods had to be fixed, described in voluminous analysis notes and de-
fended before the whole collaboration before the simultaneous unblinding
for all analysis channels. These analysis notes, sometimes hundreds of pages
long, are rarely read by other people, but frequently checked for certain plots
and results.

9.2 Basic concepts of statistical analysis

Given a set of n data points yi (e.g. event counts) as a function of a variable
xi , a spectrum is {xi ; yi ±σi }i=1
n where σ is the uncertainty of determination
i
of yi (sometimes misleadingly called experimental error). If yi follow the
y
Poisson distribution, i.e. P(y) = e−a ay! with a mean value of y = a, then the
uncertainties can be estimated as σi2 = yi ≈ yi .
A model function is a theoretical description of the data:

y = f (x; p1, p2, ...pm ) = f (x; p)

with the parameter vector p. The aim of the analysis is to determine whether
or not the model function describes the data well and to estimate its parameter
values. When the values of p are determined on the basis of the data (via
fitting the parameters of the model function to the data) we can also determine
94 Chapter 9

the covariance matrix of the estimations for parameters p as

Qik = Δpi Δpk  = (pi − pi )(pk − pk )

where x and x both denote the estimations of the expectation values of
x. The diagonal elements are the variances (the squares of the statistical
uncertainties) and the off-diagonal ones the covariances:

σp21 Δp1 Δp2  ···

 
Q =  Δp2 Δp1  σp22 ··· 
 ··· ··· ··· 

The correlation matrix is obtained by the nor-

malization of the covariance matrix,
Qik
Cik ≡ C( p̃i, p̃k ) = ,
σp2i σp2k

−1 ≤ C( p̃i, p̃k ) ≤ +1
so that its values fall in the range [−1, +1].
The correlation between two parameters is +1 Figure 9.1: Uncor-
if they are proportional, and –1 if they are inversely related and correlated
proportional to each other. High correlation means coordinates for the
that the model function is poor, or we have chosen same point.
the wrong parametrization. For instance in Fig. 9.1
the same point P can be described in the uncorrelated (x, y) and the highly
correlated (x , y ) coordinate systems. If P changes in the x direction, its y
coordinate can remain the same, but if P changes in the x  direction, its y 
coordinate must also change.
If a secondary variable z = z( p) is calculated as a function of p, its
uncertainty according to the law of error propagation (here the misleading
name error has stayed on) is

m
∂z ∂z
σz2 = Δpi Δpk . (9.1)
i,k=1
∂pi ∂pk

If the parameters are independent of each other (which never happens in

practice except when orthogonal functions are fitted), then Qik = σp2i δik and
m  ∂z  2 2
σz2 = i=1 ∂pi σpi .
 Data analysis 95

9.3 Fitting parameters

9.3.1 Goodness of fit
We have a model function f (x; p) and a set of data {xi ; yi ± σi }i=1
n . Once we

have an estimation of the parameters, the question is how well the function
fits the data. There are several ways to check that, the most common being
k
the χ−square test. The χ−square function is defined as χk2 = i=1 Xi2 where
Xi are independent variables following a Gaussian distribution

1 X2
P(X) = √ exp(− )
2π 2

with a mean value Xi = 0 and variance σx2i = 1. χk2 follows the Γ distribution,
thus it has a mean value of χk2 = k and variance σ 2 ( χk2 ) = 2k where k is its
degree of freedom.
When fitting a spectrum by, e.g., the least-squares fitting method to a
set of n independent data and m parameters, the χk2 will have a degree of
freedom k = n − m − 1:
n
1
χn−m−1
2
= [y − f (xi ; p1 · · · pm )]2 .
2 i
σ
i=1 i

χ2
The relative or reduced χk2 is defined as χr2 = n−m−1n−m−1
with a mean χr2 = 1.
χr2 1 means a too good fit, that is too little information in the data for the
given model, whereas χr2  1 means a bad fit: a model function of poorly
fitting parameters.

9.3.2 Confidence level

Let us assume that variable x follows a Gaussian distribution:

1 (x − μ)2
P(x) = √ exp(− )
2πσ 2 2σ 2

and the estimation of the mean μ = x is x̃ ± σ. The confidence level of x

∫b
being in between a and b is W(a ≤ x ≤ b) = a P(x)dx. For instance, the
∫∞
confidence that x > 0 is 0 P(x)dx.
96 Chapter 9

The full width at half max-

imum (FWHM) T is defined as
P( x̃ ± T/2) = P( x̃)/2. For a
Gaussian distribution
√ (Fig. 9.2)
it is T = 2 2 ln 2 σ ≈ 2.355 σ.
In practice we use the standard
library functions erf and erfc to
calculate
∫ x these values: erf(x) =
√2 e −t 2 dt, erfc = 1 − erf.
π 0
Sometimes the value of erf is
so close to unity that you bet-
ter calculate erfc that omits the
first term 1 from the expansion
of erf. Figure 9.2: Gaussian distribution with
mean x=0 and width σ=1. The confidence
level of having x̃ within the unshaded re-
9.4 Estimating (fit- gion is 1 − α.
ting) parameters
We have a set of measured data
and some model function with
adjustable parameters. The final result will be the best estimation of some
of the parameters. How can we get it?

9.4.1 Arithmetic mean and standard deviation

We shall now treat the various ways to estimate parameters in analysing
data. For the examples we use the method of least squares, which for just
one parameter means solving the following equation:


n
S(a) = wi (yi − ỹ)2 = min , (9.2)
i=1

with the weights wi = 1

σi2
providing the minimal uncertainty σy for the
arithmetic average ỹ of the n measured values {yi ± σi }i=1
n . Then the

weighted average is
 1
i σ 2 yi
!
1
ỹ =  i1 ±  1 . (9.3)
i σ2 i σ2
i i
 Data analysis 97

In some sense the real estimated uncertainty is the standard deviation

(Fig 9.3):

1
n
S
σSD = √ with S 2 = (yi − ỹ)2
n−1 n i=1

where ỹ is the estimated average. For large n this simplifies to σSD ≈

1 n
i=1 (yi − ỹ) .
2
n
Of course, in all data analyses
the most important factor is the rea-
sonable mind of the experimental-
ist. There is a nice old tale about
a committee which wanted to solve
the problem of estimating the height
of the Chinese Emperor by asking a
million people (as measuring him Figure 9.3: Gaussian (normal) distri-
was out of question). They rea- bution with its standard deviation re-
soned that everybody can guess the gions around the centroid μ.
height of people by the uncertainty of ±10 cm and thus the result will be as
precise as σstat = ± 10cm
√ = ±0.1 mm ≈ σSD . Of course, nobody asked had
106
ever seen the Emperor standing!
Let us estimate the average of the measurements y1 = 4 ± 1; y2 = 2 ±
1; y3 = 9 ± 1. The average is 5. Its statistical uncertainty σstat = 1
3 = 0.58
is obviously too small, but correcting it with the reduced χr2
is an overkill:
√
σstat · χr = 0.58 · 26 = 3.0. The standard deviation, however, gives a
2

more reasonable error: σSD ≈ 26 6 = 2.1. If we are not sure about the
statistical nature of the measurements, we should use the standard deviation
for the estimation of the uncertainty.

9.4.2 Linear fitting

We call the model function f (x; p) linear in parameters p = (p1, p2, . . . pm )
if it can be written like


m
∂ f (x; p) 
m
f (x; p) = pk = pk Φk (x),
k=1
∂pk k=1
98 Chapter 9

where Φk (x) is independent of the parameters. Fitting parameters p using

the least-squares method means solving the following minimization:
n
1
S( p) = (y − f (xi ; p))2 = min ,
2 i
σ
i=1 i

∂S( p)
i.e. the system of equations ∂pk = 0; (k = 1, ...m). Using the notation
fi ≡ f (xi ; p) we get

m n
1 ∂ fi ∂ fi  1
n
∂ fi
pl = yi ,
l=1
σ
i=1 i
2 ∂p k ∂pl σ
i=1 i
2 ∂p k

m
equivalent to l=1 pl Mkl = bk . We can solve this if the matrix Mkl is
symmetric and positive definite:

Mp = b ⇒ p = M−1 b ≡ Q b.

It can be easily shown that the covariance matrix of p will be just the inverse
of the Mkl matrix.
For fitting a straight line f (x; p) = p1 x + p2 to points {xi ; yi ± σi }i=1
n :

⎡  xi2  xi ⎤ (  x i yi )
⎢ 2 ⎥ σ2
M = ⎢⎢  xiiσ
 1i ⎥⎥ ; b =  yi i
σ
2
.
⎢ σ σ 2 ⎥ σi2
⎣ i ⎦
2
i

 
p˜1
The solution is = M −1 b with the covariance matrix Q = M −1 .
p˜2
 Data analysis 99

9.4.3 Non-linear fitting

If ∂∂pf depends on p,
then our model function
y = f (x; p1, . . . pm ) is
non-linear in the parame-
ters p. Trivial example
is y = p1 p2 x which must
be linearized to y = ax,
with a = p1 p2 . One
could also linearize, say
y = p1 exp(−p2 x) as ln y =
ln p1 − p2 x = a − bx, but it
is not worth it. Lineariza-
tion in most of the practi-
Figure 9.4: Lorentzian functions
cal cases cannot be done and
even it can be, it gives a biased estimation with distorted uncertainties.
The most frequent non-linear functions (with their usual parameter hand-
ling) are the Gaussian function (or normal distribution): p1 exp[−p2 (x−p3 )2 ]
(Fig. 9.2) and the Lorentzian (Cauchy or Breit-Wigner) function p1 [p22 + (x −
p3 )2 ]−1 (Fig. 9.4). Non-linear functions can be fitted using least-squares,
maximum likelihood or other methods. In the case of least-squares fitting as
an example one looks for a local minimum of the function
n
1
S(P) = [y − f (xi ; p)]2
2 i
σ
i=1 i

but in all other methods one has to find a local minimum or maximum.
In particle experiments the MINUIT program [James and Roos, 1975] has
been the most popular for fitting for almost half a century.
The Monte Carlo method can be used if we have just a vague idea of
the location of the minimum: one makes random probes in the given region
and tries to find a minimal value of the function. It is extremely slow and
inefficient.
The simplex method is very popular. The value of the function S( p) to
be minimized is calculated in m + 1 points in the m-dimensional space of the
parameters p making a simplex of m + 1 edges R1 . . . Rm+1 . Take the worst
point (where e.g. S( p) is maximal in the case of minimization) and send it to
a point in the m-dimensional space on opposite side of centroid calculated
from remaining m points (Fig. 9.5). The speed of convergence will depend
on the distance you send the point away from the centroid: the usual value is
100 Chapter 9

once or twice the original distance. This method is somewhat slow, but does
not require calculating derivatives.
In particle physics the most popular minimiza-
tion methods are the gradient ones. The gradient
of S( p) in the point of p = p0 is
 T
∂S ∂S( p) ∂S( p)
g= = ,...
∂p ∂p1 ∂pm p= p0

The gradient methods search for a minimum along

the derivatives of the function, so its convergence
is much faster than that of the simplex method.
One can calculate the derivatives analytically or Figure 9.5: Simplex
numerically, but the numerical derivatives are gen- minimization for two
erally more reliable, stable and fast. Sometimes it parameters.
is worth to start the minimization with the simplex
method and when the minimum is close, change to the gradient methods.

9.5 Uncertainties
The uncertainty of the estimated pa-
rameter value is usually obtained by
changing the χ2 of the fit by unity
at the two sides of the minimum.
This way one can get asymmetric
uncertainty values. A quite popu-
lar method to gain error estimation
of a single parameter is the MINOS
method (Fig. 9.6). It changes the
value of the given parameter and op-
timizes the rest of the parameters of
the fitting so that χnew
2 2 + 1. Figure 9.6: Estimating MINOS uncer-
= χmin
This will account for part of the cor- tainties by changing the parameter in
relations as well. question while optimizing the rest so
All uncertainties of the estima- that the χ of the fit would increase by
2

tion which do not come from the unity [James and Roos, 1975].
number of detected events are con-
sidered to be systematic. They come from various sources, like Monte Carlo
statistics and inputs, experimental calibration factors, detection efficiencies
etc, with the common name nuisance parameters. They can be estimated
 Data analysis 101

by changing the input information of the analysis by reasonable amounts.

Most of the time this is just the experimental or theoretical uncertainty of
the input parameters, but it could also be a different Monte Carlo algorithm
or fit range. There are systematic uncertainties which can be corrected later
on, like results of earlier measurements or theoretical calculations or some
global calibration of the detector or accelerator beam. However, one can-
not correct later the results of the simulations or uncertainties related to
elements of the measuring apparatus, like the calibration of a subdetector.
What amount of change is reasonable largely depends on the problem and
has to be decided by the physicist. For instance, one can check the effect
on the result of shifting a simulated spectrum by the amount of its deviation
from data. Estimating the systematic uncertainties frequently takes much
more effort than the pure data analysis itself.
One cannot just add statistical and systematic uncertainties to get a total
uncertainty as one has to include the possible correlations, which is usually
quite difficult. What is done at LHC is to eliminate the nuisance parameters
Θ from the estimation of parameters P by marginalizing them (integrating
out) [Cousins and Highland, 1992] from the analysis using the related W
distributions:
∫
L(P; x) = W(x|P) = W(x|P, Θ)W(Θ|P)dΘ .

One cannot be careful enough considering systematic uncertainties. The

Particle Physics Review [Patrignani et al., 2016] analyses the time depen-
dence of certain experimental data and the result is quite shocking: some
data do not change for many years, different groups get similar results until
somebody invents a more precise measuring method and the value undergoes
a big change, sometimes much larger than the estimated total uncertainty of
the previous measurements. A very difficult part of performing a measure-
ment is to provide a correct estimation of the systematic uncertainties which
should be neither too large nor underestimated.

9.6 Lower and upper limits

When searching for new physics one wants either to observe certain phe-
nomena, or to exclude them in given parameter regions. According to the
convention in accelerator particle physics an observation (discovery) has to
have a significance of at least 5σ. Thus one can declare the observation of a
new particle if one detects a mass peak at 5 times its uncertainty above the
background. Still, it is not acknowledged by the community before another
102 Chapter 9

n0 0 1 2 3
. nmax (90%) 2.30 3.89 5.32 6.68
nmax (95%) 3.00 4.74 6.30 7.75

Table 9.1: Numbers of excluded events at 90 % and 95 % confidence levels

if we detected 0, 1, 2 or 3 events assuming a Poisson distribution

experiment reproduces the result. This convention is based on the experience

that we do not know any such > 5σ observation yet, that was later refuted
by other experiments.
Exclusion needs a 95 % confidence level in accelerator particle physics.
However, at certain areas where much less rates are detected, like in neutrino
physics or astrophysics, 90 % exclusion is the accepted limit. For a Gaussian
distribution, as shown in Fig. 9.2, the confidence limit that the real value of
the estimated parameter is within the ±δ region of the estimated μ value is
∫ μ+δ
1−α = f (x; μ, σ)dx .
μ−δ

whereas the confidence that it is below or above that region is α/2.

In the case of Poisson distribution if we detect n0 events and we do
not know the mean μ, what maximal event number can be excluded at a
confidence of CL? In other words, at what μ will the probability
 of having
more that n0 events less than CL: P(n > n0 ) > CL = 1 − α = ∞ n=n0 +1 f (n; μ)?
Table 9.1 shows what one can say of possible exclusion for various confidence
levels and detected numbers of events. For instance, if we did not detect any
event for a given phenomenon then at a 95% confidence level we can say
that we have less than 3.0 events.
In some cases we get non-physical measured data. In order to be able to
state a limit in such cases one has to renormalize the obtained distribution to
the physically meaningful region (Bayes-limit):
∫
f (x)dx = 1
phys. area

For instance, there was a case when a fitted mass-square was negative:
π + →μ+ νμ ⇒ m(νμ )2 = (−0.016 ± 0.023)MeV2,
which is of course non-physical. The distribution was renormalized and the
limit m(νμ ) < 0.17 MeV (90%CL) accepted.
 Data analysis 103

9.7 Monte Carlo simulation

Any data analysis of high energy physics starts with simulations: that of the
expected background and signal. It is done these days using the GEANT4
[Agostinelli et al., 2003] simulation package developed mostly at CERN.
The basic idea is to produce simulated events which can be directly compared
to the detected ones: the observed event features should not be manipulated.
For that one simulates all possible physics processes which means using
various algorithms for the various processes with cross checks and drives
them all through the simulation of the whole detector system. The compo-
nents of the background are summed up according to their cross sections
and then the event selection is optimized to maximize the NS signal at a
minimal NB background. The optimization
√ means to maximize
√ a figure of
merit like in the simplest case NS / NB = max√or NS / NB √ + NS = max
although these days we prefer the condition 2 · ( NB + NS − NB ) = max
[Bityukov and Krasnikov, 2000] or using the approximate formula of Ref.
[Cowan et al., 2011].

9.8 Event selection

Both background and signal have to be checked against the previous mea-
surements and the theoretical calculations for correctness. The steps of
the event selection can be simple cuts, such as conditions on variables like
pT > 20 GeV/c or complicated multi-variate procedures: likelihood or arti-
ficial neural network optimization.
A concrete example for event selection is the search for charged Higgs
bosons in the four-jet decay channel (e+ e− →H+ H− →cs cs) from the practice
of the authors in the OPAL experiment. Before likelihood optimization we
applied a preselection with the following cuts (Fig 17.24):

(a) Eliminate those events where a Z-boson was emitted together with a
photon (radiative return to Z). If an isolated
√ photon is found, remove
it from the total centre-of-mass
√ energy s and apply a cut on the
remaining energy s  > 170 GeV.
√
(b) Cut on the total visible energy of the event by Evis / s > 0.7.

(c) Choose 4-jet events by cutting on the variable y34 of the Durham jet-
determining algorithm, which is the threshold track distance where
the event becomes 4-jet-like from 3-jet-like.
104 Chapter 9

(d) Select those events where all jets have at least two charged tracks.

One can see that the simulation describes the data very well. We tried to
apply cuts with maximum efficiency, i.e. eliminate much of the background
and little of the signal. Of course the total number of events went down as
we applied the cuts consecutively from (a) to (d).
The next step was a likelihood optimization with the basic aim to reduce
the main background of our searched signal e+ e− →H+ H− →cs cs. The only
difference between W± and H± of the same mass is spin as the W is a
J = 1 vector boson whereas the charged Higgs bosons are scalar with J = 0.
That means we should try to find variables to select between them like
angular distributions and jet structure. Using the simulations we created
reference histograms for several such variables for signal and background
and determined the probability of signal likeness for each event. The analysis
was mass-dependent, i.e. for each assumed mass of the hypothetical charged
Higgs boson a different optimization was made with different signal but the
same background. As Fig. 17.24 shows the signal was obviously rejected
for M(H± ) = 60 GeV, but that cannot be told of a signal at 75 GeV. Actually,
after collecting all information we obtained M(H± ) > 75.5 GeV at a 95%
confidence level. We got close to the mass of W± , but could not get over it.

Exercise 9.1
What are the main sources of the statistical and systematic uncertainties?

Exercise 9.2
What is the use of the covariance matrix in the interpretation of the ana-
lyzed data? How can it be used to estimate the statistical and systematic
uncertainties of derived physical parameters?

Exercise 9.3
What is the meaning of χ2 ? What shall we do if it is too high or too low?

Exercise 9.4
Deduce Eq. 9.3 for the weighted average from Eq. 9.2 using the error prop-
agation law, Eq. 9.1.

Exercise 9.5
How can one include former results in the analysis of new data?
 Data analysis 105

Exercise 9.6
How can one optimize the event selection for data analysis?
 Part III

Basic experiments

107
Chapter 10

Experimental tests of the

quark model

MOTTO:
If I could remember the names of all these
particles, I’d be a botanist.

(Enrico Fermi)

10.1 Three fermion families

10.1.1 Z width
The Z boson can decay into particle-antiparticle pairs of leptons and quarks
(it cannot decay into photons). In experiment the quark pair will appear
as hadron showers, jets, and only the charged leptons will be detected, the
neutrino pairs will escape the detector. Thus the total Z decay width will be

ΓZ = Γe+ e− + Γμ+ μ− + Γτ + τ − + Γhad + Γinv . (10.1)

According to the standard model the production of Z bosons in electron-

positron collisions and its decay to fermion pairs should have the cross
109
110 Chapter 10

section (the * superscript of Z means that it can be virtual as well):

12πMZ2 Γ(Z→e+ e− )Γ(Z→X)
σ(e+ e− →Z(∗) →X) = . (10.2)
2
ECM (ECM
2 − M 2 )2 + M 2 Γ2
Z Z Z
The parameters of Eq. (10.2) were determined by fitting the peak shape of
Fig. 17.2. The three charged leptons have the same cross section (lepton
universality principle) with the average
Γ +  − /ΓZ = 3.3658(23)%
and that of the hadronic decays:
Γhadr /ΓZ = 69.91(6)%
where the numbers in parentheses mean uncertainties in the last digits of the
results.
Thus the Z decay to quark pairs is 21 times more probable than that to a
lepton pair. This much higher probability is not surprising as there are 3 × 5
pairs of quarks available for Z decay, three colours and 5 flavours.

10.1.2 Invisible width and the families

According to the standard model each fermion carries a loop correction,
proportional to the charge of the fermion, and they appear as annoying
anomalies (cf. Section 22.9) in the equations with no trace of them in
experiment. However, they are cancelled if the sum of the electric charges of
all fermions in a family is zero. That means that the number of quark pairs
should be the same as that of leptons as the fractional charge of the quarks
is evened out by the three colours.
The invisible decay of the Z boson is related to neutrinos and it constitutes
Γinv /ΓZ = 1 − Γhadr /ΓZ − 3 × Γ +  − /ΓZ = 20.0(6)%
of the total. According to the standard model the contribution of neutrino
decay to the total Z width is
2Γ +  −
Γνν = = 1.979Γ +  −
[1 + (1 − 4 sin2 ΘW )2 ]
where ΘW is the weak mixing angle, the mixing angle of U(1)Y and SU(2) L
in the weak neutral current, sin2 ΘW = 0.23120(15). Thus the number of
light neutrinos (neutrinos with masses less than half the Z mass)
Nν = Γinv /Γνν = 2.9840(82),
meaning there are three fermion families (Fig. 10.1).
 Experimental tests of the quark model 111

10.2 Hadron jets

The quark model, at its
birth, was met by gen-
eral doubts: the frac-
tional charges and baryon
number and the postu-
lated non-existence of free
quarks made the validity of
the model very improba-
ble. Even when some of the
predictions calculated with
the quark model were ex-
perimentally confirmed, the
quarks could have thought
to be mathematical tricks
to explain things. The ob-
servation of jets however,
started to make them real. Figure 10.1: Z-width and three fermion fam-
Fig. 17.25 shows an event ilies. Hadronic decay cross section against
where an e+ e− collision pro- centre-of-mass energy of electron-positron
duces 75 charged particles collisions.
in 4 jets due to the produc-
tion of a W+ W− pair. As any additional particle production decreases the
cross section of the reaction by several orders of magnitude, such an event
can be explained by quark hadronization only.
Quarks are fermions, they can be produced in particle-antiparticle pairs
only. Thus the 3-jet events copiously produced in high-energy electron-
positron collisions have to contain at least one boson: in order to explain the
jets all three have to be produced by coloured partons (coloured particles,
quarks or gluons), and at least one of them has to be a gluon. Fig. 17.25
shows such an event. According to quantum chromodynamics, the theory
of the strong interaction, the quarks interact with each other also via gluon
exchange. As quarks exchange colours, gluons must carry colour and anti-
colour and so they strongly interact with each other as well. As a gluon has
a larger colour charge than a quark, they must have higher jet multiplicity.
This was shown experimentally at LEP by studying 3-jet events where two
jets were identified as belonging to b quarks due to their longer lifetimes,
secondary vertices. Similarly, studying 3-jet events at LEP helped to identify
the gluon’s spin as 1.
112 Chapter 10

ρ(770) = √1 (uu − dd)

2
ω(782) = √1 (uu + dd)
2

Table 10.1: Quark composition of the lightest neutral vector mesons

10.3 Fractional charges

As the electric charge is quantized, and the charges of the electron and proton
compensate each other to the level of 10−22 judging from the neutrality of
matter, it is hard to accept the fractional charges of the quarks. There are,
however, many experimental proofs for the quark charges.

10.3.1 Neutral vector mesons

The neutral vector mesons1 in ground state, V0 = (qq) J PC = 1−− can
decay via electromagnetic interaction into lepton + anti-lepton pairs. The
decay width will depend on the electric charges involved:

+ −
Γ(V0 → ) ∝ eq2 |ψ(0)| 2 /MV2 .
Two vector mesons, ρ(770) and ω(782) have very similar masses, but
due to their different quark charges ρ(770) should decay to electrons with 9
times higher probability than ω(782) (Table 10.1): Γe (ρ) : Γe (ω) 9 : 1.
In spite of the very rough approximation when we neglected their different
masses and state functions, the experimental observation is quite close to it,
[Γe (ρ) : Γe (ω)]exp = 7.02(11)
0.60(2) 12. Of course, if the u and d quarks had
different charges, the ratio would be quite different.

10.3.2 Pion scattering

The carbon-12 nucleus has 6 protons and 6 neutrons, that is 18 u and 18
d quarks. Let us study muon production in the scattering of positive and
negative pions on 12 C. A negative pion has a d quark and a u anti-quark,
π − = (ud), its quark will scatter off the quarks of the nucleons, while the u
can annihilate with one of the u quarks of the nucleons producing a μ+ μ−
pair. The positive pion, π + = (du), will do the same with the other kinds of
1Particles of unit spin are called vector bosons because of their 3 spin sub-states
J3 = −1, 0, +1
 Experimental tests of the quark model 113

Figure 10.2: Pion scattering on nucleon.

quarks of the nucleons (Fig. 10.2). The reaction rates will be proportional
to the charges of the annihilating quarks:

π − = (ud) ⇒ σ ∝ 18Q2u = 18 · 4
9
π + = (ud) ⇒ σ ∝ 18Q2d = 18 · 1
9

The cross section ratio of the two reactions is

σ(π − C→μ+ μ− . . .)
4,
σ(π + C→μ+ μ− . . .)

confirmed by experiment.

10.4 Colours
The quark model postulates that quarks come with three colour charges and
colour makes the state function of baryon systems antisymmetric against
baryon exchange. Of course, the three colours make possible the existence
of such particles like the Δ++ = (u↑u↑u↑): J = 3/2; I3 = 3/2 hyperon.
Also, the J = 3/2 ground state needs at least three quarks, and no impossible
baryon state (like one having I > 3/2 isospin) was discovered. The simplest
solution is a new SU(3) quantum number, the three colours.
The standard model assumes that only 3-quark and quark + anti-quark
states exist. The requirement of colourlessness of the free hadrons could also
be fulfilled by e.g. bound states of 2 quarks + 2 anti-quarks (tetra-quark),
of 4 quarks + 1 antiquark (penta-quark), or 3 + 3 quarks and/or anti-quarks
(hexa-quark or dibaryon). Previously, all sightings of such states were
subsequently refuted, but recently the LHCb Collaboration reported the
observation of such states.
The existence of three colours were proven, e.g. by measuring the total
cross section of hadron production as compared to that of muon pair emission.
114 Chapter 10

Figure 10.3: Hadron production cross section in electron-positron scattering

divided by the cross section of muon emission. The high-energy results are
well reproduced by assuming 3 quark colours and fractional charges

Again, those are electromagnetic reactions and their ratio is:

 + − 
σ(e+ e− →hadrons) i σ(e e →qi qi )
R= = ∝ Q2qi
σ(e+ e− →μ+ μ− ) σ(e+ e− →μ+ μ− ) i

If there is no colour, the ratio will be R0 = q Q2q , the sum of charge-squared
of the kinematically available quarks. In the case of three colours it will be
R3 = 3R0 . Depending on EC M (e+ e− ):

{u, d, s}: R0 = (2/3)2 + 2 · (1/3)2 = 2/3; R3 = 2

{u, d, s, c}: R0 = 2 · (2/3)2 + 2 · (1/3)2 = 10/9; R3 = 10/3
{u, d, s, c, b}: R0 = 2 · (2/3)2 + 3 · (1/3)2 = 11/9; R3 = 11/3

As Fig. 10.3 shows the experimental results confirm the existence of

three colours. The sharp peaks are meson resonances, the base lines show
the hadron production rates related to the numbers of available quarks. This
measurement also demonstrates the validity of fractional quark charges.

Exercise 10.1
How can we discover a new particle by measuring energy in particle scat-
tering?
 Experimental tests of the quark model 115

Exercise 10.2
What is the significance of having 3 kinds of light neutrinos for the rest of
fermions? How is the number of lepton families connected to those of the
quarks?

Exercise 10.3
How could the total Z-width be determined by studying the resonances of
individual decay channels in electron-positron scattering?

Exercise 10.4
Why does the observation of hadron jets prove the existence of coloured
particles? Why can 3 jets originate from gluons only?

Exercise 10.5
How was the idea of fractional charges of the quarks proven experimentally?
And that of the three colours?

Exercise 10.6
Using the quark compositions of the lightest vector mesons, deduce the cross
section ratio of electromagnetic decay widths Γe (ρ) : Γe (ω) 9 : 1 (Hint:
sum up the charges in Table 10.1).
Chapter 11

Parity violation, pions and

muons

MOTTO:
I cannot believe God is a weak left-hander.

(Wolfgang Pauli)

11.1 Parity violation

11.1.1 τ − θ paradox
There was a mystery: two mesons were observed having the same mass
and properties except parity: θ + decayed to two pions, θ + →2π, whereas the
other one to three: τ + →3π. The pion is a pseudo-scalar state, J P = 0− , so
τ + and θ + had to have opposite parities unless the weak interaction violates
the conservation of parity.
Tsung-Dao Lee and Chen-Ning Yang at the University of Columbia stud-
ied the question and concluded in 1956 that all proofs of parity conservation
were made in electromagnetism. They assumed that τ + and θ + are the same
particle (indeed, today it is called the K+ meson) and the weak interaction
does not conserve parity. They also proposed several experiments to check
it. Their theory was experimentally confirmed, and they were awarded the
117
118 Chapter 11

60 Co → 60 Ni∗ +e− +ν e
J=5 J=4 + 12 + 12
−→ −→ → →

Figure 11.1: The Wu experiment: decay of magnetically oriented cobalt nuclei

Nobel Prize in physics in 1957.

11.1.2 Wu’s experiment

Chien-Shiung Wu also at Columbia together with a group at Washington
made a thorough test of this idea. They oriented the spins of 60 Co nuclei in a
magnetic field at very low temperatures (T < 0.1 K) to minimize precession
and detected the electrons from beta decay. 60 Co has a nuclear spin J = 5,
so when it decays to 60 Ni∗ whose spin is J = 4 the emitted electron and anti-
neutrino must have spins oriented to that of 60 Co, i.e. along the magnetic
field (Fig. 11.1).
The angular distribution of emitted electrons was found to be completely
antisymmetric:
v
I(θ) = 1 + α cos θ
c
with the direction coefficient α = −1. Later, when similar β+ reactions were
studied, α = +1 was found for the e+ emission.
Thus the weak interaction maximally violates parity conservation as
it prefers left-polarized particles and right-polarized antiparticles. When
Wolfgang Pauli, the father of neutrinos, the exclusion principle and spin
matrices learned about the Wu experiment, he exclaimed: “I cannot believe
God is a weak left-hander!”

11.1.3 Parity violation in pion decay

In the weak decay of pions to muons: π + →μ+ νμ there are only two possible
emission directions, A and B in Fig. 11.2, but only B is realized in Nature:
this means a maximal possible parity violation. As muons also decay in weak
 Parity violation, pions and muons 119

Figure 11.2: Parity violation in pion decay to muon and neutrino: only
process B is realized in Nature

Figure 11.3: Positive pion decay produces right-polarized positive muons and those
emit positrons in the direction of polarization

interaction parity violation is manifested in that as well, in the μ+ →e+ νe ν μ

reaction. The zero-spin pions decay to polarized muons, and those will emit
the electrons in the direction of their polarization.
R.L. Garwin, L.M. Lederman and M. Weinrich (also at Columbia Uni-
versity) used the reaction of pion decay to test parity conservation in weak
decays: they stopped positive pions in a carbon target placed in a mag-
netic field and measured the number of positrons from muon decay in a
given direction as a function of the field strength (in the given experiment
as a function of the magnet current). The number of detected positrons has
shown the precession of the muons in the magnetic field, the violation of mir-
ror symmetry i.e. parity violation. This experiment was much simpler than
Madame Wu’s, but the results were published simultaneously: Lederman’s
group had waited for the Wu group to finish their work before submitting
their result for publication and acknowledged the precedence of Wu’s result.

11.1.4 Muon spin rotation (μSR)

When positive pions are stopped in matter they decay to muons, π + →μ+ νμ
with a lifetime of τπ = 26 ns. As the pion has zero spin and the neutrino is
120 Chapter 11

Figure 11.4: Corrections to the magnetic moment of the muon: electromagnetic,

weak, hadronic terms and possibly new physical phenomena beyond the standard
model.

forced to be left polarized, the emitted positive muons are also left polarized,
and they can be used to study local magnetic fields in solid, liquid or gaseous
matter just by detecting the direction of the positrons emitted at muon decay.
The muons first slow down via Coulomb scattering (not losing their polar-
ization) and then get trapped somewhere in the sample. In a magnetic field
the slow muons decay with a lifetime of τμ = 2.2μs while precessing with a
frequency ω proportional to the magnetic field strength B (Fig. 11.3):
qB qB
ω = gμ S≈
mc mc
where q = ±e is the charge and m is the mass of the muon. For several
decades μSR was used to study the behaviour and measure the field strength
of local magnetic fields in solid state physics and chemistry, even meson
factories were built to facilitate this method. Today it is much less cultivated.

11.2 Anomalous magnetic moment of the muon

The magnetic moment of the muon,
q
μμ = g S (11.1)
2m
is a very important physical quantity as it can detect the effects of heavier
particles, i.e. possible new physical phenomena. In Eq. (11.1) g, e and m
are the giro-magnetic factor, charge and mass of the muon and S is its spin
vector. The Dirac equation assumes a g factor of exactly 2, but other heavy
particles via Feynman loops may add sizeable contributions to it. As shown
in Fig. 11.4, the charged muon can emit and absorb virtual photons before
and after the interaction with the external field, the virtual photons, in their
turn, can be converted into fermion-anti-fermion pairs (all that as long as the
uncertainty principle allows it).
 Parity violation, pions and muons 121

g−2
The anomalous part of the magnetic moment, a = 2 is due to all kinds
of corrections. The standard model predicts it to be

aμ (SM) = aμ (QED) + aμ (weak) + aμ (hadr).

However, if there is physics beyond the standard model with new particles,
they could also contribute to the corrections, that is why the anomalous
magnetic moment of the muon is one of those important quantities which
could make it possible to discover new physics even at low energies. Thus
it is continuously probed at higher and higher precision. If we find a devi-
ation from the standard model prediction, it can be due to new physics or
incomplete calculation, and thus useful in any case. Let us add that precise
measurements are always useful in physics, many discoveries were made just
by trying to make a measurement more precise. Two bright examples are
cosmic background radiation and CP violation: both discoveries changed
our view of Nature and earned Nobel Prizes for the researchers.

11.2.1 (g − 2) μ : non-relativistic measurement

q
Recall that the magnetic moment of the muon is1 μ = g 2mc S where the Dirac
equation predicts g = 2. In a non-relativistic case the Larmor precession of
the muon in a magnetic field of strength B will have the frequency
   
g q qB g qB g−2 qB
ωS = B= · = 1+ = (1 + a).
 2mc mc 2 mc 2 mc

A particle with spin S = 12 will precess around its momentum in field

B at a frequency which is the difference between its spin and cyclotron
frequencies:
qB qB
ω a = ωS − ω c = ωS − =a
mc mc
if it is polarized along its momentum (which is natural for the muon emitted
in pion decay). The stored muons decay to positrons, μ+ →e+ νe ν μ with
velocity into the direction of the muon spin, at a rate of the time dependence
N(t) = N0 e− τ [1 − A cos(ωa t + φ)] where τ = 2.2μs is the muon lifetime
t

and φ is the angle of measurement.

Such an experiment was made first at CERN in a muon storage ring in
1965 (with the leadership of Georges Charpak, who was awarded the Nobel
Prize in physics 30 years later for constructing the multiwire proportional
1We use SI units in this section.
122 Chapter 11

chamber ). The experiment was made with muons of pμ = 90 MeV/c

momentum, in a field of B=1.6 T and it gave the result aμ = 1162(5) × 10−6 ,
i.e. a large anomaly.

11.2.2 (g − 2) μ with relativistic muons

One of the obvious ways to make the (g − 2)μ measurement more precise is
to extend the lifetime of the muons by making them relativistic. The spin
precession frequency of relativistic muons will have a contribution from
Thomas precession:

eB eB
ωs = g + (1 − γ) ,
2mc mcγ

their cyclotron frequency is ωc = eB

mcγ where γ = √ 1
is the usual
1−v 2 /c 2
Lorentz factor. Thus the measurable difference will be the same as in the
non-relativistic case:
eB
ω a = ωs − ωc = a (11.2)
mc
This experiment was performed also at CERN in 1972 by J. Bailey et
al. They used muons at pμ = 1.9 GeV/c which meant a relativistic factor
of γ = 12 and the muon lifetime extended by an order of magnitude to
γτμ = 26μs. The B = 1.7 T inhomogeneous magnetic field helped to
focus the muons and keep them on orbit, but it also smeared the result.
Nevertheless it was still more precise than before: aμ = 116616(31) × 10−8 .

11.2.3 (g − 2) μ with magic momentum

The inhomogeneous magnetic field focuses the stored muons, but it also
causes a serious systematic smearing by broadening of the muon momentum
distribution, thereby reducing the experimental precision. The solution is to
separate focusing from the magnetic field of storage: electrostatic focusing.
In an electric field E for a muon of velocity v the frequency connected to the
anomalous magnetic moment will be
(   )
e 1 v×E
ωa = aB − a − 2 (11.3)
mc γ −1 | v|

One can see from Eq. (11.3) that for muons having a magic momentum
(hence speed) for which a − γ21−1 = 0 the electric field effect is completely
 Parity violation, pions and muons 123

Figure 11.5: The storage ring of the muon g-2 experiment at Brookhaven
National Laboratory, 1999-2006

eliminated from the measured frequency. This experiment was first made
at CERN in 1979 by J. Bailey et al. in a homogeneous magnetic field
of strength B = 1.5 T with the following parameters: pμ = 3.094 GeV/c
magic momentum; γμ = 29.37; γτμ = 64.4μs. Using the method of magic
momentum increased the precision of the measurement by another factor of
3 to aμ = 1165924(85) × 10−9 .
The last such experiment was made at the AGS accelerator of Brookhaven
National Laboratory using the same magic momentum method. The stor-
age ring (Fig. 11.5) with 24 detectors collecting the electrons from μ−
and the positrons from μ+ decay was operated in 1999–2002, and mea-
sured separately the anomalous magnetic moments for μ+ and μ− . The
results agreed with each other and their average [Patrignani et al., 2016] was
(116592089 ± 54stat ± 33syst) × 10−11 , two orders of magnitude better than
the CERN result. There is a 3.5σ (σ is the experimental uncertainty) differ-
ence between theory and experiment. This could mean some new physics,
but may also be due to the incomplete calculation of hadronic contributions,
so the muon (g −2) value is worth further improvement. The experiment will
continue at Fermilab and for that the storage ring was transported from New
York to Chicago. As it was much too large for trains, airplanes or trucks,
it was shipped via the Atlantic Ocean, the Mexican Bay and the Mississippi
river, see the home page at http://muon-g-2.fnal.gov/bigmove/. The ring is
124 Chapter 11

in operation already, we can expect the first results in the near future.

Exercise 11.1
Why is the polarization of the muon definite when the muon is emitted in
pion decay?

Exercise 11.2
Derive Eq. (11.2).

Exercise 11.3
Show how the muon storage ring facilitates direct measurement of the anoma-
lous magnetic moment of the muon.

Exercise 11.4
How could new physics phenomena beyond the standard model be discovered
via measuring (g − 2) of the muon?

Exercise 11.5
How do we measure the polarization of the muon?

Exercise 11.6
Why does the muon have deeper penetration in matter than both the much
lighter electron or the much heavier proton and neutron?
Chapter 12

Kaons and CP violation

MOTTO:
Alice laughed. “There’s no use trying,” she
said: “one can’t believe impossible things.”
“I daresay you haven’t had much practice,”
said the Queen. “When I was your age, I al-
ways did it for half-an-hour a day. Why, some-
times I’ve believed as many as six impossible
things before breakfast.”

(Lewis Carroll: Through the Looking Glass)

12.1 Kaons
As discussed in Chapter 1, CPT invariance is one of the most fundamen-
tal symmetries in physics. As for elementary particle interactions the time
reversal invariance is also quite basic and generally established by experi-
mental evidence, it was generally assumed that CP invariance also has to be
valid for all microscopic processes, even when parity symmetry has proved
to be completely violated by the weak interaction. Fig. 12.1 shows the four
possible CP-variations for pion decay: only two of them will be allowed,
both C and P have to be reflected in order to get from the case where the
neutrino is left-polarized to the right-polarized anti-neutrino.
125
126 Chapter 12

Figure 12.1: Pion decay: out of the four possible CP-variations only two
will be allowed, both C and P have to be reflected.

Strangeness I3 = + 12 I3 = − 12
S = +1 K+ = us K0 = ds
0
S = −1 K = ds K− = us

Figure 12.2: The lightest K-mesons. Left: kaon formation in pion-nucleon colli-
sions. Right: the strong eigenstates of kaons.

Kaons are the second lightest mesons after the pions (Fig. 12.2). They
contain s quarks which carry the strangeness S flavour quantum number.
Pseudo-scalar mesons, their spin-parity quantum number is J P = 0− . We
met them already in Chapter 11: Lee and Yang discovered parity violation
when noted that K+ , the positive kaon appeared as two different particles
in the experiment. Among the kaons, in addition to K± we have another
particle-antiparticle pair: K0 and K0 .

12.2 Neutral kaons

Neutral kaons are among the so-called V-particles: their decay appeared in
bubble chamber experiments as two tracks out of nothing (as neutral particles
do not leave tracks in chambers) curving into opposite directions in the
magnetic field. They were copiously produced and slowly decayed, so they
are created in strong and decay in weak processes. There is a slight difference
in masses between charged and neutral kaons: mK± = 493.7 MeV/c2 and
mK0 = mK0 = 497.6 MeV/c2 .
Kaons are formed when pions collide with nucleons at high enough
energies (Fig. 12.2): a light quark-antiquark pair annihilates into a gluon and
 Kaons and CP violation 127

makes an ss pair which then separate in two particles carrying strangeness.

In pion scattering on protons at certain pion energies (around a few GeV)
well regulated beams can be produced of given kinds of kaons. One can
produce neutral kaons at accelerators at 1 < T(π − ) < 6 GeV where T denotes
kinetic energy: π − p → ΛK0 with the subsequent decay of the Λ hyperon
(with the lifetime of τΛ = 0.26 ns) via:
*
pπ − (64%)
Λ→
nπ 0 (36%)

that helps to tag the production of the kaon.

Neutral anti-kaons can be made with positive pion scattering,

π + p → K + K0 p (T(π + ) > 1.5 GeV),

or in
π − p → Λ K0 n n (T(π − ) > 6 GeV).

The kaons are created via colour interaction in flavour eigenstates and decay
via weak interaction in weak eigenstates. Assuming that the weak interaction
conserves CP parity, one can look for weak decays in the forms of CP
0
eigenstates. K0 and K , however, are not those: CP |K0 >= |K0 > and
CP |K0 >= |K0 >. We can construct CP eigenstates:

K10 1 1
0
= √ (K0 ± K0 ) = √ (ds ± ds)
K2 2 2

As the pion is pseudo-scalar meson, its CP eigenvalue is κ(CP) = −1.

CP conservation demands that K01 with κ(CP) = +1 decays to two pions,
but the K02 with κ(CP) = −1 to three pions only, which increases its lifetime
by 3 orders of magnitude as zero orbital momentum dominates between the
emitted particles: K10 →ππ (τ1 ∼ 89 ps) whereas K20 →πππ (τ2 ∼ 52 ns).
Because of the slightly different virtual loop corrections, even their masses
are different, although the difference is extremely small.

12.3 Kaon oscillation

Thus a free K0 meson will decay via weak interaction in two different modes
with two different lifetimes. After a long enough flight only the longer-lived
component will survive. The amplitude of the free meson when it forms
128 Chapter 12

at t = 0 is K0 = √1 (K1 + K2 ), a mixture of two states with different time-

2
dependencies: a1 (t) = √1 exp{− Γ21 t +im1 t} and a2 (t) = √1 exp{− Γ22 t +im2 t}.
2 2
The time dependence of the total squared wave function is

1 −Γ1 t
IK0 (t) = 12 (a1 + a2 ) · (a1∗ + a2∗ ) = (e + e−Γ2 t + Re(a1 a2∗ ))
4
where
   
Γ1 t Γ2 t Γ1 t Γ2 t
Re(a1 a2∗ ) = exp − −im1 t − +im2 t +exp − +im1 t − −im2 t
2 2 2 2
 
Γ1 + Γ2
= exp − t [ei(m2 −m1 )t + e−i(m2 −m1 )t ] .
2

Using eix = cos(x) + i sin(x), the expression in the brackets can be written
as cos(Δmt),
   
1 −Γ1 t −Γ2 t Γ1 + Γ2
IK0 (t) = e +e + 2 exp − t cos Δmt ,
4 2

where Δm = |m1 − m2 |. This means a periodic changing mixture of the two

states with the frequency ω = Δm. If a particle is produced in the state K0 ,
the probability to detect it in the same state at time t is proportional to IK0 (t).
Thus we see a kaon oscillation between the two flavour states with a
frequency depending on the mass difference, Δm = 3.5 · 10−6 eV, implying
a relative difference Δm/m = 0.7 · 10−14 . Assuming also the exchange of
quarks, the measured mass difference allowed to make predictions for the
mass of a fourth quark, the c quark. The study of kaon oscillation with the
small mass difference between the weak eigenstates facilitated to achieve the
most precise experimental test of CPT invariance:

0
|m(K0 ) − m(K )|
< 10−18
m(mean)
Another interesting feature of kaon physics is the regeneration process
(Fig. 12.4). After a long flight the neutral kaon beam consists mostly of
long-lived K20 = √1 (ds − ds). As shown by Pais and Piccioni in 1955, if it
2
is injected into matter, the d quark is just scattered on nucleons, whereas its
antiparticle component can annihilate and so K0 is partially reconfigured.
 Kaons and CP violation 129

Figure 12.3: Kaon oscillation between two CP eigenstates. Left: exchange

of weak bosons, right: origin of the mass difference.

Figure 12.4: Kaon oscillation. Left: A kaon starting as K0 becomes in time

a mixture of K0 and K0 . Right: K0 regeneration: anti-quarks annihilate and
K0 is reconfigured from the long-lived mixture

12.4 CP violation
Around 1960 everybody believed in the conserved CP symmetry, but af-
ter the discovery of parity violation, CP invariance called for experimen-
tal testing as well. J.H. Christenson, James W. Cronin, Val L. Fitch and
R. Turlay published the observation of the CP-violating decay K02 →ππ in
1964, (Cronin and Fitch were awarded the Nobel Prize in 1980). Their
method was based on the different kinematics of 2- and 3-pion decay modes
of the neutral kaon: a 3-pion decay produces pions more-or-less uncorrelated
in angle, but for the 2-pion decay the kinematic conditions are determined.
The experiment fixed the angle and energy of the detected pions and gave
45 ± 9 π + π − pairs above background (note the 5σ excess!). This means
130 Chapter 12

Figure 12.5: Penguin diagram describing direct CP-violation. Its name is

the result of a bet: the loser, John R. Ellis, had to introduce the word penguin
in his next paper.

that the long-lived neutral kaon is not a clean CP-odd eigenstate, it has a
CP-even component as well. This mixing can be written as KS = K1 + K2
and K L = K2 + K1 where KS is the short-lived and K L is the long-lived
neutral kaon. Here | | = (2.28 ± 0.02) · 10−3 so — as opposed to parity
violation — CP invariance is just a little bit violated.
CP invariance would mean |e+ νe π − > = |e− ν e π + >, but CP violation
should violate that as well. We expect that in the decay of the long-lived
neutral kaon,
* + − 
e νe π
KL →
e− νe π +

a charge asymmetry can be present. This direct CP-violation (Fig. 12.5)

was observed at CERN by the NA48 experiment. With a very sophisticated
setup they simultaneously observed kaon decay in four channels with very
similar geometric conditions and detection efficiencies. They deduced the
following ratio:

N(K L →π 0 π 0 ) N(KS →π + π − )
R= .
N(KS →π 0 π 0 ) N(K L →π + π − )

This helped to eliminate quite a few systematic effects: calibration and

geometry.
Experiment NA48 at the SPS of CERN did not go without problems:
once the chambers around the beam line imploded and the whole setup had
 Kaons and CP violation 131

to be rebuilt. Finally, the result was the following:

N(K L →π 0 π 0 ) N(KS →π + π − )
= 0.99098 ± 0.00101 ± 0.00126
N(KS →π 0 π 0 ) N(K L →π + π − )
with the first uncertainty being statistical, the second one systematic. Clearly,
the experiment has demonstrated the direct CP-violation.

Exercise 12.1
In what sense can we speak of anti-kaons? How are antiparticles introduced
in particle physics?

Exercise 12.2
Show that K01 and K02 are CP eigenstates and that is why they can only decay
to 2 and 3 pions.

Exercise 12.3
How can energy and momentum conservation simultaneously fulfilled when
the kaon changes its mass in flight?

Exercise 12.4
What is the mechanism of kaon regeneration? How does the short-lived kaon
become long-lived again?

Exercise 12.5
What is the most precise test of CPT invariance and why?
Chapter 13

Neutrinos

MOTTO:
I have done a terrible thing, I have postulated
a particle that cannot be detected.

(Wolfgang Pauli)

13.1 Weak currents

The weak interaction has two kinds of currents, a charged current when
the interaction is mediated by W± bosons and a neutral one with the Z
boson. When an unstable particle decays to leptons it produces particle-
antiparticle pairs or a lepton with a neutrino. Typical weak decays are
shown in Fig. 13.1. We have determined the number of fermion families via
studying the decay of the Z boson to neutrino pairs by measuring its invisible
width (see Section 10.1.1) and obtained 3. Thus there are 3 kinds of light
neutrinos with charged leptons associated.
As neutrinos are produced overwhelmingly in the nuclear reactions in
stars (most of the energy of a supernova is carried away by neutrinos) space
is full of them, 60 billion neutrinos fly through our fingertips every second,
most of them from the Sun. Let us list the main neutrino sources, noting
the kind of neutrinos and their distance between production and eventual
detection on the surface of Earth.
133
134 Chapter 13

Figure 13.1: Weak interaction processes. From left to right: neutron beta-
decay at nucleon level, depicted in a crossing symmetric diagram, and at
quark level (charged currents); antineutrino scattering on an electron (neutral
current). The arrows follow the CPT convention

• Cosmic neutrinos: all kinds from cosmic distances.

• Sun neutrinos: nuclear fusion, 4 H → He + 2 e+ + 2 νe . They are
electron neutrinos coming from a distance of L 108 km.
• Atmospheric neutrinos: secondary particles of cosmic rays.
π ± →μ± νμ ; μ± →e± νμ νe ; L 30 km, νe, 2 νμ , ν representing both
neutrinos and antineutrinos.
• Neutrino beam from an accelerator: Analogous to the atmospheric
case, L 1–700 km.
• Geo-neutrinos: Geological antineutrinos from natural radioactivity:
β-decay of uranium and thorium: n→p + e + ν e .
• Reactor neutrinos: In nuclear reactors the energy is produced by nu-
clear fission resulting in neutron-rich daughter nuclei that subsequently
undergo β− decay releasing low-energy electron-antineutrinos.

Neutrinos were first observed by Clyde L. Cowan and Frederick Reines

in 1952-56 with antineutrinos from a nuclear reactor. They used the reaction
of neutrino scattering on the proton, ν e p→e+ n with double identification:
the two prompt photons from positron annihilation and the delayed photons
from the capture of the neutron in cadmium. F. Reines received the Nobel
Prize almost 40 years later, in 1995. Since then neutrinos have been in the
front line of particle research: at the moment there are dozens of working
neutrino experiments (some of them are noted in Fig. 17.26). Most of them
 Neutrinos 135

are large, built under water, in deep mines, or tunnels under earth, and one
is sunk in the ice of the Antarctica.
It was quite a breakthrough when in 1987 the neutrinos from supernova
SN1987A were detected by the existing large neutrino experiments, the
Irvine–Michigan–Brookhaven (IMB) detector in the USA, Kamiokande in
Japan and Baksan in Russia, although the whole neutrino burst lasted 13
seconds only.

13.2 Neutrino mass

In the original formulation of the standard model neutrinos were postu-
lated to have zero mass. Many experiments were done in the last almost
hundred years to measure their mass. Most of them rely on tritium decay
t→3 He + e− + ν e and study the high-energy end of the electron spectrum for
deviations, as the end point should be sensitive to neutrino mass. One of
the largest and newest such experiment is KATRIN, the KArlsruhe TRItium
Neutrino experiment (Fig. 17.27). It is so large that when built in Germany
it could not be transported on the highway. Similarly to the Brookhaven
muon storage ring, it was shipped down the Danube, along the Black Sea,
the Mediterranean, the Atlantic Ocean, and up the Rhine River, altogether a
travel of 8600 km instead of the 400 km as the crow flies. KATRIN hopes
to measure the mass of the electron neutrino with the precision of ∼ 1 eV
(Fig. 13.2). Cosmological observations suggest that the sum of the masses
of the three neutrino types is below 1 eV.

13.3 Early neutrino mysteries

The first mystery was connected to neutrinos from the Sun. The nuclear
fusion of the Sun makes helium from hydrogen in several steps resulting in

41 H→4 He + 2e+ + 2νe,

i.e. producing 2 νe with energies up to 18 MeV. Those neutrinos were first

observed by the Homestake Experiment of Raymond Davis and his group
in 1964-69 (he was awarded the Nobel Prize in 2002). The enormous
difficulty of detecting neutrinos is well demonstrated by Solar Neutrino Unit
of detection:
10−36 ν-interactions
1 SNU = ,
atom · sec
136 Chapter 13

Figure 13.2: The expected sensitivity of the KATRIN spectrometer for

determining the mass of the electron neutrino.

which practically means one detected neutrino interaction a day per 1030
atoms, i.e. per 10–100 tons of material. The Homestake experiment of
Davis collected and analysed radioactive 37 Ar atoms from the reaction
νe +37 Cl→37 Ar + e− .
The expected rate from the standard solar model was 8.2 ± 1.8 SNU, but they
found only 2.56 ± 0.23 SNU. As all subsequent measurements confirmed
both this result and the validity of the model, this huge discrepancy remained
mysterious for a long time.
A second mystery is connected to the disappearance of the atmospheric
muon neutrinos. As cosmic particles (mostly protons and helium nuclei) hit
the atmosphere they produce pions which decay to muons and the muons to
electrons :
π+ −→ μ+ νμ π− −→ μ− ν μ
| and |
−→ e+ ν μ νe −→ e− ν e νμ
 Neutrinos 137

Figure 13.3: Neutrino-less double beta decay with a Majorana neutrino in

nuclear decays and on parton level.

These processes should yield twice as many muon neutrinos than electron
ones, but the measurements gave much less than the expected factor of two.
As the neutrinos are neutral point-like fermions, they could be Majorana
particles which are their own antiparticles with the opposite polarization:
νL = ν R . In that case, however, the neutrino-less double beta-decay should
be possible as shown in Fig. 13.3. Many experiments were and are looking
for such a process, and so far upper limits were estimated only for its half-life,
typically about 1021 years (on a 90 % confidence level).

13.4 Neutrino oscillation

13.4.1 Phenomenology
Bruno Pontecorvo (Joint Institute for Nuclear Research, Dubna, Russia)
developed in 1963 the theory of neutrino oscillations to explain the lack of
detected atmospheric and Sun neutrinos, but he passed away before they were
experimentally observed. Let us assume two kinds of neutrinos, with flavour
eigenstates νe and νμ . Weak interaction will mix them into mass eigenstates
ν1 and ν2 and they can oscillate between the two states: νe ⇔νμ (similarly
0
to the K0 ⇔K case, see Chapter 12). Obviously, energy and momentum
cannot be conserved at the same time.
If the momentum is conserved, then the energy difference is (c = 1)
(  )
m22 m12
E2 − E1 = p + m2 − p + m1 ≈ p 1 + 2 − 1 + 2
2 2 2 2
2p 2p

m22 − m12 Δm2

= ≡ .
2p 2p
138 Chapter 13

In the case of energy conservation the momentum difference is

m22 − m12 Δm2
p2 − p1 = E 2 − m22 − E 2 − m12 = ≡ .
2E 2E
As the neutrino mass is very small, it is quite relativistic and Eν ≈ pν , so the
two approaches yield similar results.
A free neutrino can be treated as a propagating field of plane wave: e−ip ·x
where a difference in mass leads to a change in phase. This phase change
for a neutrino of energy E in time t and on a distance of L ≈ ct is
Δm2 t Δm2 L
Δ(p · x) = Δ(Et − px) ≈ ≈ .
2E 2E
Of course, for this to be valid one needs small enough violation of the
conservation laws to fit in the uncertainty relation of Heisenberg, or an
interaction with some outside field or medium. The Sun neutrinos do have
a thick dense medium to interact with, but in the case of the atmosphere
we need the uncertainty relation. There the momentum change is δpν ≈
10−12 eV, much lower than the pion or muon natural widths: Γπ ≈ 10−8 eV,
Γμ ≈ 10−10 eV.

13.4.2 Super-Kamiokande experiment

Super-Kamiokande (SKK) is a huge detector in the Kamioka mine in Japan
under 1 km of earth. It was originally built to detect proton decays (Kamio-
kande stands for KAMIOKA Nucleon Decay Experiment). Its heart is
a Cherenkov detector with 50000 tons of extra clean water watched by
11146 photo-multiplier tubes of 50 cm diameter (Fig. 17.28). It has 100%
efficiency of detecting muons (and thus converted muon neutrinos) with
pμ > 100 MeV/c. It committed “suicide” after the discovery of the oscillation
of the atmospheric neutrinos: in 2001 one of its phototubes imploded during
water refilling and the shock wave destroyed more than half of its phototubes.
By 2006 it was completely reconstructed.
SKK was looking for muon and electron events originating inside the
detector as the incoming neutrinos leave no other trace. It could identify
from the Cherenkov cones the type of particles (muon or electron) and their
energies. Even the direction of the detected neutrinos could be reconstructed
by measuring the shape and timing of the Cherenkov ellipse (Figs. 17.28 and
17.29).
SKK confirmed the lack of Sun neutrinos with a very high precision:
*
SKK data +0.014
= 0.406 ± 0.004
SSM MC −0.013
 Neutrinos 139

where the first uncertainty is statistical, the second one systematic. Thus it
was confirmed that only 40% of the electron neutrinos of the Sun predicted by
the standard solar model arrive to Earth. The Sun neutrinos were identified
by type and energy only and their origin reconstructed by the Cherenkov
cones nicely showed the shape of the Sun (Fig. 17.29).
Part of the atmospheric muon neutrinos also disappeared, SKK measured
the following yield:
(Nμ /Ne )data
= 0.688 ± 0.016 ± 0.050
(Nμ /Ne )MC
where, as usual, the first uncertainty is statistical, the second one systematic.
However, SKK could also separate muons and electrons, and sort them by
energy and direction, which allowed to make an amazing discovery: the
electron neutrinos of cosmic origin (i.e. of higher energies than those from
the Sun) are in order, whereas the higher energy muon neutrinos from the
atmosphere at the opposite side of Earth partially disappear (Fig. 17.30).
The ratio (flux up)/(flux down)
N(−1.0 < cos θ < −0.2)
= 0.54 ± 0.04
N(0.2 < cos θ < 1.0)
The conclusion from the SKK results was that (1) low-energy electron
neutrinos oscillate into other types on the Sun-Earth distance and (2) high-
energy muon neutrinos at the distance of Earth’s diameter oscillate into
τ neutrinos as they do not appear in the electron neutrino spectra. The
measured flux ratio was consistent with 1.3 × 10−3 eV2 ≤ ΔMatm 2 ≤ 3.0 ×
−3 2
10 eV . Note that these measurements cannot determine the neutrino
masses, just the differences in their mass squares.

13.4.3 SNO experiment (1999-2003)

The Sudbury Neutrino Observatory (SNOlab) is located in the Creighton
nickel mine of Sudbury, Canada at 2 km underground. The SNO detector
worked in 1999-2003, it contained 1000 t of 99.92% pure D2 O inside an
acrylic ball (Fig. 17.31) floating in 7500 t of H2 O. For the experiment SNOlab
borrowed the heavy water from the Canadian Atomic Energy Agency. SNO
used three different reactions to detect neutrinos:
• νe + d→p + p + e− : charged current, sensitive to electron neutrinos
only;
• νx + d→p + n + νx : neutral current , sensitive to all neutrinos;
140 Chapter 13

• νx + e− →νx + e− : electron scattering, sensitive to all neutrinos.

Thus SNO could measure the fluxes of electron neutrinos separately and
of all neutrinos simultaneously. Between 2001 and 2003 the heavy water had
2 tons of Na35 Cl dissolved to catch the neutrons released from deuterons in
the interaction with neutrinos. As neutron capture in 35 Cl produces several
gamma photons, adding Na35 Cl helped to distinguish between the neutral
current and electron scattering reactions.
The result of SNO has consolidated the proof of the oscillation of the
Sun neutrinos. It has shown that although some of the electron neutrinos
disappear, the total flux of Sun neutrinos agrees very well with the expecta-
tions of the standard solar model, so the missing electron neutrinos transit
into the other two kinds on the Sun-Earth distance.
For the observation of cosmic neutrinos with the Super-Kamiokande ex-
periment Masatoshi Koshiba received the Nobel Prize in 2002. For the dis-
covery of neutrino oscillations the 2015 Nobel Prize in physics was awarded
to Takaaki Kajita (SKK) and Arthur B. McDonald (SNOlab).

13.4.4 Neutrino mixing and masses

Thus at least two of the three neutrinos have masses (most likely all three) and
the mass eigenstates are different from the weak eigenstates, so they mix.
The mixing can be described by the Pontecorvo-Maki-Nakagawa-Sakata
mixing matrix U f i , see Section 22.8.
We have found two mass differences among the three neutrinos: the
atmospheric neutrinos oscillate with a mass-square-difference of
(ΔM 2 )atm ≈ 2.4 × 10−3 eV2
and the Sun neutrinos with
(ΔM 2 )Sun ≈ 7.6 × 10−5 eV2 .
This does not allow to determine the neutrino mass hierarchy, the order-
ing of neutrinos by mass eigenvalues. For simplicity let us assume m1 < m2
for the two masses closer to each other (from the oscillation of the Sun
neutrinos), then m3 will be more different from the first two and either
m1 < m2 < m3 or m3 < m1 < m2 .
Recent data by many experiments confirmed [Patrignani et al., 2016] that
the mixing angle Θ12 is associated mainly with the solar νe , and Θ23 with
the atmospheric νμ oscillations. That is why in the literature the 12 index is
often substituted by  and 23 by atm or A, like Θ12 →Θ  and Θ23 →Θatm . It
is also agreed upon that sin Θ13 must be quite small.
 Neutrinos 141

13.4.5 Neutrino experiments at nuclear reactors

As mentioned above, the first detection of neutrinos was made near a nuclear
power plant. The experimental proof of flavour oscillations of neutrinos
along distances from the Sun and through the diameter of Earth raised the
interest to study neutrino oscillations on various distances. We shall briefly
describe three examples of such experiments: the Daya Bay experiment made
at a few kilometres from a nuclear power plant, two long-distance neutrino
beam accelerator experiments and two short-distance ones at accelerators.
Reactor anti-neutrinos are commonly detected by the reaction

ν e + p→e+ + n

with energies between 1.8 and 8 MeV. Energy conservation prohibits the
reactions involving muons or tau-leptons, and the oscillation is detected by
the disappearance of the electron anti-neutrinos.
For three families the survival probability of ν e is [An et al., 2017]

Psur ≈ 1 − cos4 Θ13 sin2 (2Θ12 ) sin2 Δ21 − sin2 (2Θ13 ) sin2 Δee ,

where Θi j is the mixing angle between two mass states, and

L[m]
Δi j ≈ 1.267 Δmi2j [eV2 ]
Eν [MeV]
with the mass-squared difference measured in eV, the flight distance in meters
and the neutrino energy in MeV. The effective neutrino disappearance phase
Δee (not to be mixed with Δi j ) is an empirical quantity independent of the
neutrino mass hierarchy. The reactor experiments help to determine the
two quantities Θ13 and Δee . From the latter an empirical mass-squared
difference Δmee 2 can be deduced, characterizing the disappearance of reactor

anti-neutrinos.
The Daya Bay experiment (Fig. 17.32) is built near Hong Kong in China.
It consists of eight ν detectors in three groups in the vicinity of six nuclear
reactors. Each detector contains 20 tons of liquid scintillator. It is aimed
at determining the mixing of neutrinos of the first and third families. The
results published in 2017 [An et al., 2017] were the following:
• For the ν1 ↔ν3 mixing they measured a mixing angle of

sin2 (2Θ13 ) = (8.41 ± 0.27 stat ± 0.19 syst) × 10−2,

a very well established experimental result.

142 Chapter 13

• For the empirical ν neutrino mass-squared difference they got

|Δmee
2
| = 2.50 ± 0.06 stat ± 0.06 syst) × 10−3 eV2 .

• They have excluded the existence of an additional ν4 sterile neutrino

(that does not belong to a charged lepton) in the mass range of 10−3 <
Δm142 < 0.1 eV2 .

13.4.6 Long distance neutrino experiments

As neutrino oscillations were confirmed both at the Sun–Earth distance and
at the distance of Earth’s diameter, several experiments were set up to check
shorter distances. We shall consider later LSND at 30 m and MiniBooNE at
500 m. The CNGS (CERN Neutrinos to Gran Sasso) experiment creates a
neutrino beam with the 450 GeV proton beam of CERN’s Super Proton Syn-
chrotron and shoots neutrinos toward the Gran Sasso National Laboratory,
built in a tunnel 1 km underground south from Rome at a distance of 732 km
from CERN. The relativistic pions decay in vacuum, they produce muons
and muon neutrinos. The charged muons will slow down and decay or get
absorbed by nuclei, but the muon neutrinos follow the forward direction of
the pions and form a beam: at Gran Sasso the beam diameter is about 2.8 km.
A similar setup operates at Fermilab: there is a 736 km distance between the
injector synchrotron of the Tevatron and the Soudan mine where the MINOS
neutrino detector is located.
Gran Sasso has several neutrino experiments. The OPERA experiment,
e.g. managed to observe the oscillation of νμ into ντ on the CNGS beam.
They measured the neutrino flight time between CERN and Gran Sasso as
well. When — as a result of a faulty cable connection — they announced
that they saw neutrinos travelling faster than the speed of light in vacuum,
another experiment in their neighbourhood, ICARUS, immediately checked
it and found ordinary subluminal neutrinos only.

13.4.7 Sterile neutrinos?

One of the numerous mysteries of neutrino physics is connected to the Liquid
Scintillator Neutrino Detector (LSND) experiment. LSND collected data
from 1993 to 1998 at Los Alamos: the 800 MeV proton beam from LAMPF,
the Los Alamos Meson Physics Facility was injected in a thick water target
where hadrons, mostly positive and negative pions were produced. The
pions were stopped in a copper target that absorbed the negative pions and
 Neutrinos 143

Figure 13.4: The beam line of the MiniBooNE experiment at Fermilab.

The booster protons are injected into the target, produce pions of which
the negative ones get absorbed in nuclei and the positive pions decay into
muons and neutrinos. The result is 3 kinds of neutrinos, νμ , ν μ and νe . The
appearance of ν e is interpreted as ν μ →ν e oscillation.

muons (they get captured in atomic orbits and then absorbed by the nuclei),
whereas the positive particles decay: π + →μ+ + νμ , μ+ →e+ + ν μ + νe . These
reactions produce three kinds of neutrinos, but not electron antineutrinos; if
they appear, it is a sign of ν μ →ν e oscillation.
LSND saw a 4σ excess of ν e above expected background using the
ν e p→e+ n reaction. Considering that we do not see this oscillation with
the atmospheric neutrinos, the difference has to be connected to the different
energy of neutrinos (in Los Alamos it was up to 52.8 MeV) and their different
flight distances (30 m between target and detector). The conclusion of the
collaboration was that Mν > 0.4 eV.
The LSND result contradicts the other experiments. If it is valid then
a fourth, sterile neutrino must exist. As usual, no new observation is ac-
cepted in particle physics before another experiment confirms it (although the
LSND result did not reach the required 5σ level). A new collaboration was
formed and it built the MiniBooNE (Booster Neutrino Experiment) detector
at Fermilab (Fig. 13.4) in 2005 to confirm or refute the LSND result.
Fig. 13.4 presents the beam line of the MiniBooNE experiment at Fer-
milab. It reproduces the conditions of LSND at an order of magnitude
higher energy and distance. LSND was located at L1 = 30 m from the
target and received E1 = 30 MeV neutrinos on average. The MiniBooNE
detector is located at L2 = 500 m from the production target which pro-
vides E2 = 500 MeV neutrinos. Thus the ratios determining the oscillation
frequency were the same, E1 /L1 = E2 /L2 .
In 2007 the MiniBooNE Collaboration stated that they did not see the
antineutrinos of LSND. In 2009 they announced that they started seeing
something and in 2012 they published a 3.8σ excess, almost exactly the
144 Chapter 13

same as LSND. Again, this excess is not considered to be significant enough

to accept as the signal of new physics.

13.5 Present neutrino mysteries

One would think that the neutrino mysteries are solved, we found the missing
neutrinos: the atmospheric νμ transforms into ντ on the way through Earth,
and the Sun’s νe metamorphoses into νμ and ντ while travelling from the
Sun to Earth. The oscillations are well described by three neutrino flavours
and the PMNS mixing matrix.
Unfortunately (or fortunately!) the observation of neutrino oscillations
added more to the mysteries that were already known for the neutrinos.
Although we may add neutrino masses to the Lagrangian of the standard
model like we do with the charged leptons, we would not be able to explain
any phenomenology with such an addition. The standard model has to be
extended in order to accommodate non-zero neutrino masses.

• If the neutrinos have masses, there could exist right-polarized neutri-

nos, νR , and left-polarized antineutrinos, ν L , with zero hypercharge
Y = 0, i.e. with no charged lepton attached (sterile neutrinos), with no
interaction within the standard model. While this would be a natural
extension of the standard model, such particles were not observed in
experiment, except by the LSND experiment which was not confirmed.

• Maybe the neutrinos are Majorana particles, ν = ν: the neutrinoless

double-beta-decay experiments so far did not observe such phenom-
ena.

• Maybe there is an extended Higgs-sector with more symmetry-break-

ing potentials to create those masses? The LHC experiments found
just one Higgs-boson and its properties are close to those predicted by
the standard model.

• Why do neutrinos oscillate? They feel only the weak interaction, no

reason for them to have different eigenstates. Maybe there is another,
so far unobserved interaction mixing those states?

• Why the neutrino masses are so unreasonably small? The see-saw

mechanism is proposed to solve this problem. It assumes that the
neutrinos exist in pairs: light Dirac and heavy sterile neutrinos, but
such heavy ones were not observed yet.
 Neutrinos 145

• What is really the mass of the neutrinos? The oscillation gives the
difference of mass-squares only.

We can conclude that neutrinos present a real enigma: they are bright
signal posts showing the holes and deficiencies in our standard model in
spite of all its success in describing the experimental data, and also a way to
extend our knowledge.

Exercise 13.1
Why does the existence of neutrino oscillations point toward the existence
of new physics beyond the standard model whereas that of kaon oscillation
does not?
Exercise 13.2
How can energy and momentum conservation simultaneously fulfilled while
the neutrino changes its mass during flight?

Exercise 13.3
What is the difference between mass and flavour eigenstates for neutrinos?

Exercise 13.4
How was the number of light neutrinos measured?

Exercise 13.5
How does the existence of neutrino oscillations prove the non-zero mass of
neutrinos?
Exercise 13.6
What is the difference between left and right polarised neutrinos and an-
tineutrinos?
Chapter 14

Higgs boson

MOTTO:
It was in 1972 . . . that my life as a boson really
began.

(Peter Ware Higgs)

The standard model seems to describe all experimental data so well that
any observed deviation from it is looked upon with deep suspicion. For
almost 40 years, more and more precise new data have been acquired at the
particle accelerators and all seem to agree very well with the predictions of
the standard model. Hundreds of experiments are summarized in Fig. 17.33
according to the LEP Electroweak Working Group [Schael et al., 2013]. It
shows the 2012 situation of the analysis of electroweak data: all experimen-
tal data and theoretical estimates agree within the experimental uncertain-
ties. The only parameter which deviates at more than 2σ uncertainty is the
forward-backward asymmetry of the decay of the Z boson to two b quarks.
Since the observation of the top quark in 1995 the Higgs boson was
the only missing elementary particle of the standard model. Its observa-
tion in 2012 put the dot on i, proving the mechanism of mass creation via
spontaneous symmetry breaking.
147
148 Chapter 14

14.1 Search for the Higgs boson

The search for the Higgs boson is the best example to show how systematic
data handling and analysis are organized in high-energy physics. For 40
years larger and larger experiments tried to find it or exclude its existence,
until finally, in 2012 the ATLAS and CMS experiments at the Large Hadron
Collider managed to observe it.

14.1.1 Methodology
When searching for a new particle, what we usually try to observe is a
resonance. As described in Chapter 2, for a particle with lifetime τ = Γ−1
and decay rate Γ the event rate against the energy corresponding to the
invariant mass of the decay products is
1
| χ(E)| 2 = ,
(E − M)2 + Γ2 /4
i.e. a Lorentz curve. It shows a peak at the invariant mass M of the decaying
system with a full width Γ at half maximum. We claim the discovery of a
new particle if we see a resonance with a confidence level at least 5σ at the
same invariant mass of the particle in all expected decay channels, by all
related experiments.
The search involves several consecutive steps as outlined already in
Chapter 9.
• Compose a complete standard model background using Monte Carlo
simulation taking into account all types of possible events normalized
to their cross sections.
• Compose Higgs-boson signals, simulations of all possible production
and decay processes at all possible Higgs-boson masses.
• Put all these through the detector simulation to get events analogous
to the expected measured ones.
• Optimize the event selection via reducing the B background and en-
hancing the S signal via maximizing some figure of merit.
• Check the background, i.e. the description of data by the simulation for
the given luminosity: the simulation should reproduce the observed
background distributions in all details. For instance, you can check
the background of the decay of a neutral particle to charged leptons
by selecting lepton pairs of identical charges.
 Neutrinos 149

• Check the signal: does it agree with the expectation by the theoretical
model?

Once we are happy with the simulations and the event selection, we
must choose a test statistic. That could be any kind of probability variable
characteristic of the given phenomenon: probabilities for having background
only, signal or combinations. One of the favourite is the Q likelihood ratio of
signal + background over background: Q = L s+b /L b . What most frequently
plotted is

Nch ⎡
   ⎤
⎢ nk
sk (mH )Sk (x jk ; mH ) ⎥⎥
−2 ln Q(mH ) = 2 ⎢
⎢sk (mH ) − ln 1 +
bk Bk (x jk ) ⎥
k=1 ⎢⎣ j=1 ⎥
⎦
where the variables are the following:

• nk : events observed in channel k, k = 1 . . . Nch .

• sk (mH ) and bk : signal and background events in channel k for Higgs-

boson mass mH .

• Sk (x jk ; mH ) and Bk (x jk ): signal and background probability distribu-

tions for events for Higgs-boson mass mH at test point x jk .

• x jk : position of event j of channel k on the plane of its reconstructed

Higgs-boson mass and cumulative testing variable, the latter con-
structed of various special features of the event like b-tagging, signal
likelihood, neural network output etc.

Several other testing variables can be constructed on the same basis,

rather frequently used ones are probabilities of not having the expected
signal on the basis of the expected background and the collected data:

• CLb , the signal confidence level assuming background only, i.e. the
complete absence of the signal, or

• The so-called p-value: the probability of obtaining a test statistic at

least as extreme as the one that was actually observed, assuming that
the null hypothesis is true. Translated to our language that means
the probability that a random fluctuation of the measured background
could give the observed excess.
150 Chapter 14

14.2 Exclusion at LEP

Although the four large experiments at the Large Electron Positron (LEP)
collider saw no new physics, no deviation from the standard model, LEP
provided an incredible amount of very precise measurements, some of which
are presented in Fig. 17.33. In its last two years of working, LEP was mostly
devoted to the search for the Higgs boson, collecting more luminosity at
higher energies than in the previous 10 years together.
At LEP the dominant formation process is Higgs-strahlung e− e+ →ZH
(the name comes from the funny English word Bremsstrahlung1), the asso-
ciated production of Higgs and Z bosons. The dominant Higgs boson decay
at these low masses is to a bb̄-quark pair (Fig. 17.34). The various channels
are different only due to the various decay processes of the accompanying Z
boson.
Statistics, the master of bad jokes at particle physics experiments (see
the Foreword of [Patrignani et al., 2016]), have shown up at LEP as well:
one of the experiments, ALEPH, saw in one of the possible Higgs decay
channels a quite strong signal corresponding to a Higgs boson of a mass of
115 GeV, while the rest of LEP have not seen anything [Barate et al., 2003]
(Fig. 17.36). ALEPH saw the excess in 4-jet events only, in those events
where the Higgs boson decays to a pair of b quarks and the accompanying Z
boson also decays to a quark pair (Fig. 17.35). The b quark is identified by
its long lifetime leading to a secondary decay vertex in the event. Another
strange thing was that the Higgs signal seen by ALEPH by far exceeded the
expectations of the standard model. Also, the observed Higgs mass was
critical as it coincided with the average kinematic limit of LEP: in 2000
the average collision energy of LEP was about 206 GeV and the observed
resonance was found at 115 GeV, the difference was very close to the mass
of the Z boson, 91 GeV.
A quite interesting feature of data analysis was the plotting of spaghetti
diagrams. Those are signal weight distributions of each selected event
as a function of the assumed Higgs mass. Fig. 17.37 shows the weight
distributions of 17 selected Higgs-like candidate events observed by the four
LEP experiments [Barate et al., 2003]. The ALEPH events crowd around
115 GeV whereas for the other three experiments there are less of them with
a rather random mass distribution. This caused quite an excitement at LEP:
many physicists signed the petition to the Director General of CERN to
extend the life of LEP by another year, but that was refused: the simulated
1Other languages use the simple mirror translation of the German word: braking
radiation
 Neutrinos 151

projections were not very promising for a discovery of the SM Higgs boson
(the effect seen by ALEPH only was far too large, much higher than the
prediction of the standard model), and the contractors for building the LHC
were already prepared to start.

14.3 Search and observation at LHC

Figure 17.38 shows the cross sections of various production processes of
the SM Higgs boson in p-p collisions at the LHC. The dominant reaction is
gluon fusion, but vector boson fusion is also significant.
Just like LEP had, the Large Hadron Collider has also four interaction
points (Fig. 17.7) with a major experiment (and sometimes a smaller one as
well) in each. The two largest ones, ATLAS and CMS were designed with
the main aim of discovering the Higgs boson, ALICE is specialized on heavy
ion collisions and LHCb on studying rare processes involving b quarks. The
authors belong to CMS, so most of the results we mention are from CMS,
but all will be compared to those of ATLAS pointing out the similarities and
the (very few and not significant) differences.
The ATLAS and CMS collaborations are really huge. According to
the official statistics at the time of the Higgs discovery CMS had 3275
physicists (including 1535 students) and 790 engineers and technicians from
179 institutions of 41 countries (ATLAS was even somewhat larger). The
largest participant of CMS is the USA, then Italy, Germany and Russia.
It is quite remarkable how similar and different are ATLAS and CMS.
ATLAS uses a lot of new detector techniques while CMS consists of mostly
traditional parts. CMS is based on the largest superconducting solenoid on
Earth whereas ATLAS has a smaller solenoid encircled by 8 huge magnets
making a toroidal field. CMS is full of steel, it weighs 14000 tons, twice
the weight of ATLAS in an order of magnitude smaller volume. The event
triggering was also different, CMS developed a quite innovative, two-level
trigger scheme, while the ATLAS trigger had three levels. And in spite of
all these differences, the two collaborations got (and still get) very similar
results, and that makes all their observations most reliable.
The design of the LHC and its experiments started well before the actual
start of LEP, which means that the construction of the LHC detectors took
two decades of hard work before the actual data acquisition started. The
LHC devoted its first two years of operation to development rather than data
taking, which really started in 2011 only.
Even before the LHC started, the parameter fitting of the standard model
152 Chapter 14

had pointed toward a light Higgs boson with a mass around 100 GeV. As
LEP excluded the Higgs boson below 114 GeV the LHC experiments had
to be prepared for detecting the Higgs boson in the most complicated mass
region, around 120 GeV, with several competing decay channels (Fig. 17.34).
It is well seen that below 110 GeV the bb̄ decay, above 160 GeV the WW
dominates, that is why those mass regions could be excluded earlier. The-
oretical calculations have shown that the best channels to observe a light
Higgs boson at the LHC should be the two-photon, H→γγ and four-lepton,
H→ZZ∗ → + − + − channels. The hadronic decay channels are hampered
by the very high hadron background. However, these signal channels have
very low branching ratios, for H→γγ it is BR = 2.27 × 10−3 and the sig-
nal/background ratio is small, S/B 1, and for H → ZZ∗ → + − + − (as
usual, = e, μ, i.e. in the experiment only the electron and the muon are
considered leptons, because the taus can decay to hadrons as well) it is even
lower, BR = 1.24 × 10−4 , but with a much cleaner signal, S/B > 1. In 2012
the LHC luminosity was already so high that every bunch crossing (event in
every 50 ns) contained 10–20 p-p collisions leading to copious hadron pro-
duction. Both large experiments, CMS and ATLAS designed their tracking
systems and electromagnetic calorimeters with these two processes in mind.
The electromagnetic calorimeter of CMS consists of 75,848 PbWO4 single
crystal scintillators, whereas that of ATLAS is a sampling calorimeter based
on liquid argon shower detectors.
By the beginning of 2012, when all 2011 data were analysed, the possible
mass of the SM Higgs boson was already confined to the region of 114 <
MH < 127 GeV by CMS with very similar results from ATLAS. In that
region 2– 3 σ excesses were found at 125 GeV in the two main decay
channels, H →γγ and H → ZZ. It seemed more and more probable that the
Higgs boson would be observed at the LHC in 2012. It was even decided
by the CERN administration to extend the data taking scheduled for 2012
before the long shutdown for accelerator development (for increasing its p-p
collision energy from 8 to 13 TeV and the event rate from 20 to 40 MHz) if
necessary for the discovery.
On July 4th, at the beginning of the large annual high-energy physics
congress in Melbourne, the spokespersons of ATLAS and CMS gave talks
from CERN (in internet connection to the whole word, including, of course,
the main auditorium of the Australian conference) on Higgs search. They
announced that at the LHC collision energies 7 and 8 TeV, in the two most
significant decay channels H → γγ and H → ZZ → + − + − , at an in-
variant mass of m 125 GeV a new boson is seen by both experiments at
a convincing statistical significance of 5σ confidence level, with properties
 Neutrinos 153

corresponding to those of the standard model Higgs boson. The fact that
the new particle could decay to two photons or Z bosons confined its spin
to an even integer, i.e. a boson of S = 0 or S = 2. Of course, as the data
analysis was optimized to find the SM Higgs, it was very unlikely to find
something very different. Nevertheless, the two experiments emphasized
that it has to be studied, whether or not its spin is really zero with a +
parity (the pseudo-scalar mesons have spin 0 with negative parity), and that
its decay probabilities to various final states follow the predictions of the
standard model. After reanalysing their data the Tevatron experiments, CDF
and D0 also found an excess at this mass (after the LHC started the Tevatron
accelerator of Fermilab was stopped).

14.3.1 Reactions of the media

The saying that three people can keep something secret only if two of them are
dead is attributed to Benjamin Franklin. As any result of a collaboration has
to be approved by all members before it is made public, the more than 6000
participants of ATLAS and CMS knew well in advance the developing result.
Thus two days before the 4th July announcement Nature Online already re-
ported the discovery. Of course, the fact that the CERN management invited
to the seminar all leading scientists of the field, including the theoreticians
who developed spontaneous symmetry breaking for the standard model, also
helped people to guess that something dramatic would be announced.
CERN produced some figures concerning the media echo of the day:
55 media organizations were represented at the talks of 4 July. The talks
were broadcasted via close to half a million internet connections (many
of them being conference rooms in partner institutions, 1034 TV stations
devoted 5016 news broadcasts to the event for more than a billion (109 )
people. Many-many news articles and even more blogs and talks discussed
the conditions and importance of the discovery.

14.3.2 Observations
On 31 July the two experiments submitted papers of the discovery to Physics
Letters B, they were published 14 August. Both papers, [Aad et al., 2012]
and [Chatrchyan et al., 2012b], were 15 pages long followed by 16 pages
long lists of close to 3000 authors, and both were dedicated to the memory
of those participants who could not live to see the result of the more than
two decades of construction work. Their measured mass distributions, after
analysing about a quarter of the data to be collected in 2012, are shown in
154 Chapter 14

Figs. 17.41 and 17.42. The mass distributions of the di-photon and 4-lepton
decay channels were quite similar for the two experiments. They showed a
significant peak at the same invariant mass of about 125 GeV. In all cases the
signal strengths agreed within uncertainties to the predictions of the standard
model.
What was really convincing of the observation was the distribution of
the p-values of the events selected in the various decay channels of the
hypothetical Higgs boson. It was a joke of statistics that in July 2012 adding
together two decay channels, H → γγ and H → 4 gave the same 5σ
significance for both ATLAS and CMS, whereas adding to it the results for
other channels increased the significance to 6σ for ATLAS and left it at 5σ
for CMS (Fig. 17.42).

14.3.3 Is it really the Higgs boson?

Analysing most of the data collected in 2012 confirmed the existence of the
new boson and led to the conclusion that all observed properties of the newly
discovered particle were close to those predicted for the Higgs boson of the
standard model within statistics. All these results were further confirmed by
the LHC data collected in 2016-17 at higher energy (13 TeV p-p collisions)
and much higher luminosity. The corresponding 4-lepton and 2-photon
spectra are presented in Figs. 17.43 and 17.44.
The fact that the new boson decays to two photons points to its having
spin 0 or 2. The charged lepton spectra bore the features of its having S = 0+
as ascertained by both experiments. The measured signal strengths of the
new particle were fully compatible with that expected for the standard model
Higgs boson: for CMS it was ∼ 20% less while for ATLAS ∼ 40% more
than the SM prediction, but both deviations were within the experimental
uncertainties. As a theoretician remarked, whenever ATLAS had an excess
CMS came up for everybody’s annoyance with a deficit, bringing the average
close to the SM prediction.
The final analysis of the 2012 data resulted in a joint mass value by
ATLAS and CMS as [Aad et al., 2015]

mH = 125.09 ± 0.21(stat) ± 0.11(syst) GeV .

The LHC experiments studied the cross sections of the processes con-
nected to the new particle. Fig. 17.45 shows the signal strengths of production
and decay in various possible channels of the Higgs-like boson measured by
CMS [Chatrchyan et al., 2013] as compared to those predicted by the stan-
dard model for the Higgs boson with a mass of 125 GeV. The amplitudes of
 Neutrinos 155

all observed signals are in agreement with the expectations of the standard
model. ATLAS got similar results, of course.
Thus what we found is most likely the standard model Higgs boson. On
one hand this is a great success of particle physics. On the other hand this
is somewhat of a disappointment as the SM has theoretical shortcomings
which need new physics to resolve. Just to list a few of them: it cannot unite
the interactions at large energies, cannot include gravitation, cannot account
for the dark matter of the Universe or for the dominance of matter against
antimatter, and cannot explain neutrino masses and oscillations. There are
many extensions of the theory, that can resolve some of those problems, most
popular among them is supersymmetry, which among other things predicts 5
kinds of Higgs bosons, but none of its predicted phenomena could be found
yet experimentally. The observables of the Higgs boson should be sensitive
to some of the features of new physics and these studies will be the main job
of the LHC experiments in the future.
Right after the confirmation of the discovery, in 2013 François Englert
and Peter Higgs were awarded the Nobel Prize in physics.
It is very interesting that the 125 GeV mass of the Higgs boson seems to
be exciting for theoreticians. There was even a special workshop in Madrid
organized to discuss this mass in 2013. The reason is that mH = 125 GeV is
at the border line of the stability of electroweak vacuum on the plane of top
mass against Higgs mass.

14.4 Vacuum stability

Let us recall that in the standard model the potential of the symmetry-
breaking BEH field is V(φ) = μ2 |φ| 2 + λ|φ| 4 where the second term with
|φ| 4 is the self-interaction of the field with a strength depending on λ.
The observation of the Higgs boson at mass 125 GeV proved that the
vacuum of our world has a non-zero minimum potential value. The strengths
of all interactions depend on the energy of interaction, at LHC energies and at
mH 125 GeV λ 0.13. Well before the introduction of the BEH potential
it was stated that the coupling of a self-interacting scalar field (λ in this case)
can change sign at high energies, which is called vacuum instability. The
question is, of course, at what energy we can expect this sign change. This
energy dependence is affected by the interaction of the field with all particles.
As the t quark has the largest mass, hence highest interaction potential with
the BEH field, it has the largest effect on λ.
In certain conditions several BEH minima could exist [Lee and Wick, 1974].
156 Chapter 14

If the BEH vacuum has one minimum only, our vacuum is stable, if it has
more, then it can be metastable. If our vacuum is metastable and we are
not in the deepest minimum, then the universe exist in a false vacuum, and
the world can shift into the deeper one by a quantum mechanical tunnelling
effect. In some sense it could be the end of our world as we know it. At the
new BEH minimum all the particles would have different masses and also
the strength of the weak interaction would change. Of course, this assumes
that the standard model is valid up to extreme high energies. As Turner and
Wilczek stated in 1982 [Turner and Wilczek, 1982], even if our vacuum is
metastable, its expected lifetime could be much longer than the age of the
Universe, hence appear stable to us.
These estimations were refined in time and as the mass of the Higgs
boson became definite, it became clear that our world is located somewhere
at the edge of stability. Fig. 17.46 shows such an estimation, presently most
precise in the standard model. The location point of the measured masses of
the Higgs boson and of the t quark fall in a narrow metastable region.

14.5 BEH field and inflation

According to the generally accepted model of cosmology our world was
born in a Big Bang. The theory has many shortcomings and for their
solution Alan Guth proposed the model of inflation, according to which
the space of the Universe in the first 10−32 second expanded by 1020 –
1040 times. This needed a force field called inflaton which had to dis-
appear after the inflation. There are many speculations about its nature,
here we quote one connected to the Higgs boson. Bezrukov and Sha-
poshnikov [Bezrukov and Shaposhnikov, 2008] have shown that the BEH
field is a suitable candidate for that assuming that its ξ coupling to grav-
ity is strong enough. Then the Lagrangian of the standard model should
be completed by − 12 M 2 R − ξH † HR, where R is the gravitational space
curvature and H is the strength of the BEH field. M is a mass-like pa-
rameter, less than the Planck mass related to the gravitational constant,
0 < M < MP = 1/ 8π/G 2.4 · 1018 GeV. M = MP should result in
too high initial density fluctuations, and M = 0 in too large Higgs boson
masses. Subsequent estimation showed that the Higgs boson should be light,
mH < 135 GeV/c2 for the model to function. From the demand that the in-
stability of the BEH field coupled to gravity, started at the Planck scale only,
Shaposhnikov and Wetterich [Shaposhnikov and Wetterich, 2010] predicted
mH = 126 GeV/c2 . Of course, this assumes that the standard model is valid
 Neutrinos 157

up to the Planck scale. If below that energy some new physics starts, the
picture is invalid.
Thus we do not have to be afraid that our Universe shifts into another
vacuum. At the same time we see another excellent feature of the BEH
mechanism: it could explain the cosmic inflation without assuming another
force field. Apparently we get more and more proof for Lederman’s joke
that the Higgs boson and the BEH mechanism remind him of the deus-ex-
machina in the Greek-Roman dramas, that puts everything in order.

Exercise 14.1
Why do we need symmetry breaking in order to create masses? Why does
the existence of the Higgs boson prove the validity of the symmetry breaking
mechanism?
Exercise 14.2
Why do we need such overly sophisticated statistical methods like −2 ln Q or
the p-value in order to exclude or discover new physics?

Exercise 14.3
Why do hadron colliders have higher chance to discover new physics in spite
of the more difficult data analysis requirements than the electron-positron
colliders with their much cleaner events?

Exercise 14.4
What reactions helped to discover the Higgs boson at the LHC? Why those
decay channels were used which have the lowest yields?

Exercise 14.5
How can we tell that the Higgs-like signal observed at the LHC belongs to
the Higgs boson predicted by the standard model?
Chapter 15

Heavy ion physics

MOTTO:
“It seems very pretty, ... but it’s RATHER
hard to understand! ... Somehow it seems to
fill my head with ideas — only I don’t exactly
know what they are!”

(Lewis Carroll: Through the Looking Glass)

15.1 Quark gluon plasma

The basic idea of heavy ion physics is to reproduce and study the properties
of matter right after the Big Bang. That extremely hot matter is assumed to
have been a soup of de-confined quarks and gluons, the quark gluon plasma.
Figure 17.47 shows a simulated collision of heavy ions. As they are
accelerated to high energies, they are relativistic. In the direction of the
beam the ions are Lorentz-contracted, which also increases the density and
temperature of the plasma. The fire ball will expand and cool down, and the
quarks hadronize, form colourless hadrons that can be detected experimen-
tally. Most of the components and also the hadrons formed in the collision
will fly forward. If a new particle forms and decays in the soup, it will decay
in more-or-less its own centre-of-mass system and emit its decay products
in all directions. Thus what is really interesting should fly to some extent
159
160 Chapter 15

perpendicularly to the colliding beams. Let us recall the definitions of trans-

verse momentum, pT = p2x + p2y and pseudorapidity, η = − ln tg θ2 . The
latter is used instead of polar angle at hadron colliders. These variables have
important use in heavy ion physics.
Note that in heavy ion physics the reaction yields are determined by
the energy of the individual nucleon-nucleon collisions and not by the total
energy of the projectiles, so it expresses the collision energies in units of
GeV/nucleon pair (GeV/NN) or at LHC energies TeV/NN. Thus, when pro-
tons are accelerated to 3.5 TeV on each side, the total energy of p-p collisions
will be 7 TeV, whereas in the case of Pb-Pb collisions its characteristic energy
will be 2.76 TeV/NN.
Generally speaking, heavy ion physics is very much related to quan-
tum chromodynamics as it treats quasi-free, interacting quarks and gluons.
Citing the recent review [Busza et al., 2018]: “Heavy ion collisions are a
laboratory that is rich with unique ways to probe fundamental aspects of
QCD empirically, with some control over varying conditions.” According to
the theory of heavy ion collisions, the appearance of the quark-gluon plasma,
as a phase transition, should be characterized by the following experimental
signatures:

• High pressure in the medium, increasing pT of the released particles.

• Jet quenching: As the medium becomes coloured, the stopping power
for coloured particles increases, quarks and gluons have difficulties
getting out of the system, so the number of detected jets should be
reduced.
• Low-viscosity liquid: The quarks and gluons move quasi-free reflect-
ing to some extent the original nucleon-like lumpiness of the system,
which at higher viscosity should even out.
• Direct photons: With the appearance of quasi-free quarks the medium
contains fast decelerating charged particles that should radiate high-
energy direct photons, similarly to ordinary bremsstrahlung.
• Effective hadron masses: In the high-temperature plasma the chiral
symmetry is partially restored, heavier quarks become lighter, so the
identified heavy mesons should have lighter masses .
• Debye screening: When the colour content of the medium becomes
generally large because of release of quarks and gluons, that will
decrease the binding energy of individual hadrons. This will lead
 Heavy ion physics 161

Figure 15.1: Lead-lead collision event at LHC energy of 5.02 TeV per
nucleon pair recorded by the time projection chamber of the ALICE detector
in November 2015. In this collision 1582 positively-charged (darker tracks)
and 1579 negatively-charged (lighter tracks) particles are produced; about
80 % of them are pions

to such effects as the break-up of heavy hadrons, like the reduced

detection of J/ψ = (cc) mesons.

15.2 Hydrodynamics
As mentioned above, due to the low internal viscosity of the quark gluon
plasma its partons move almost free. If their interaction were stronger,
the system could not possibly remember its structure before the collisions,
but it does: hydrodynamics converts spatial anisotropies into momentum
anisotropy. Another characteristic feature is the elliptic flow. The central-
ity (related to the overlap) of the nuclei in nucleus–nucleus collisions (see
Fig. 15.2) is rarely perfect. If the newly formed quark matter were gas-like
162 Chapter 15

Figure 15.2: Centrality regions and charged particle multiplicities in

Pb-Pb collisions at 2.76 TeV/nucleon-pair energy measured by ALICE
[Aamodt et al., 2010]. Note that 0 % corresponds to the highest centrality.
Regions 60–70 % and 70–80 % are not labelled

then it should expand in all directions with the same momentum. How-
ever, it seems to behave as an almost perfect liquid and expands elliptically
(Fig. 17.47). This azimuthal anisotropy in particle production can be quanti-
tatively characterized by the elliptic flow, the coefficient v2 of the second term
in the Fourier transform of the angular distribution of measured (charged)
particles:
 
dN N  ∞
= 1+2 v n cos (n(φ − ψ n )) . (15.1)
dφ 2π n=1

The observables are the following: φ is the angle in the transverse plane,
ψn are the event plane angles (where the nth harmonic component has its
maximal track multiplicity) and N is the average number of particles per
event. They can be measured for different particle species as functions of
rapidity, centrality and transverse momentum. v2 is the elliptic flow reflect-
ing two-parton correlations in the quark gluon plasma. Higher correlation
coefficients can also be measured and calculated. Useful quantities are also
the event-by-event values of these data, not only the averages.
Elliptic flow is one of the most important measured quantities in heavy
 Heavy ion physics 163

ion physics as it can be predicted theoretically using various hydrodynamic

models.

15.3 Experiments
For heavy ion experiments the main problem is to follow and identify the
thousands of emitted particles (Fig. 15.1). The detectors of heavy ion physics
look very similar to those in high-energy particle physics: a large magnet, a
vertex detector closest to the beam pipe, an inner tracker, usually a chamber
encircled by electromagnetic and hadronic calorimeters, and muon detectors
outside. The experiment specialized on heavy ion physics at LHC, ALICE
(A Large Ion Collider Experiment) sacrificed data acquisition speed for
precision: it has the largest existing time projection chamber (TPC, see
Chapter 7) around its beam pipe, which helps to follow and identify thousands
of particles, but cannot handle the amount of luminosity ATLAS and CMS
accept with their semiconductor tracker systems. ALICE was placed in
and around the huge magnet of the former LEP experiment L3. Fig. 15.2
shows the correlation between centrality and particle multiplicity measured
by ALICE at 2.76 TeV/NN [Aamodt et al., 2010].
The measuring apparatus of the fixed-target NA49 experiment, now con-
tinued as NA61 (NA stands for North Area of SPS and the number is a
sequence number of accepted experimental proposals) also based on TPC’s.
It mapped the transition from ordinary nuclear matter to quark-gluon plasma
by comparing the following reactions from the point of view of particle
yields:

• Pb+Pb collisions at 158, 80, 40, 30, 20 GeV/NN;

• C+C, Si+Si collisions at 158 and 40 GeV/NN;

• p+Pb, p+C, p+p and n+p at 158 GeV/NN;

• π+p and π+Pb also at 158 GeV/NN.

They had a beam of 105 Pb-ion/sec in about 1 mm diameter at 158 GeV/NN

and typical event multiplicities of more than 1500 charged hadrons in every
collision.
Another place of heavy ion experiments is RHIC, the Relativistic Heavy
Ion Collider at Brookhaven National Laboratory, USA. RHIC accelerates
two beams of copper and gold ions and collides them at several interaction
points. Its two largest experiments are PHENIX and STAR. The RHIC
164 Chapter 15

experiments were the first to demonstrate the transition from nuclear matter
to quark matter by detecting jet quenching in central Au+Au collisions as
compared to proton-proton, proton-Au, deuteron-Au events and peripheral
Au+Au collisions. In the case of peripheral collisions most of the nucleons
will avoid collisions and just shoot along the beam pipe. In the case of
central collisions there is a large activity in the transverse directions, which
is easily measured by the cylindrically built detector systems. Every heavy
ion experiment has a zero degree calorimeter, a detector close to the beam
pipe, in order to check the centrality of the collisions by detecting the forward-
scattered particles and also for measuring the total number of collisions, the
luminosity.

15.4 Jet quenching

The observation of jet quenching was a real break-through in heavy ion

physics as it proved that at high enough energies the quarks are released
from the nucleons and the matter becomes all coloured in a volume much
larger than the size of the nucleon.
In Fig. 17.48 RHIC results are shown. The π 0 yield observed by the
PHENIX experiment (Fig. 17.48, upper) from central Au+Au collisions is
suppressed as compared to the d+Au collisions of minimum bias, i.e. at
minimal selection criteria [Adler et al., 2003]. In other spectra PHENIX has
shown how the suppression depends on the energy and collision centrality.
The suppression of neutral pions proves that this is not a Coulomb effect, but
the result of the coloured medium.
The STAR experiment made a very similar discovery (Fig. 17.48, lower).
They measured the azimuthal angles of tracks in the events taking φ = 0 at
the track with the largest pT for 2-jet events in d+Au and Au+Au collisions
[Adams et al., 2003] for minimum-bias events. In central Au+Au collisions
the track with the highest pT will more probably come from 2 → 2 collisions
near the surface of the fireball. Then the other particle produced in the
collision flies in the opposite direction. As it has to traverse thicker coloured
medium, it is more likely stopped inside the fireball. There is no such effect
in d+Au collisions. This was observed at the very beginning of RHIC and
published in 2003. All RHIC experiments saw similar effects both with
neutral and with charged hadrons, and have shown the formation of coloured
matter in high-energy heavy-ion collisions.
 Heavy ion physics 165

15.5 Heavy ions at LHC

At the end of every p+p run (usually lasting 6–8 months) the LHC collides
heavy ions for a few weeks. There were Pb+Pb runs at the maximal LHC-
energies, but there were also p+Pb collisions and p-p runs at lower energies
(corresponding to the per-nucleon-pair energies of heavy ions) to establish
reference measurements for the interpretation of the Pb+Pb data. There was
also a Xe–Xe run at the end of 2017.
At the low-luminosity start-up of LHC, ALICE could handle the p+p
collisions very well and produced the first paper with LHC results, a test of
QCD at high energy. It made many measurements with heavy ions later.
We saw that the RHIC experiments had observed the one-sided suppres-
sion of the tracks of two-jet events from central collisions of heavy ions
as compared to those from collisions with protons or deuterons. This was
interpreted as quenching of partons going through the bulk of the coloured
quark matter. At the LHC ATLAS reproduced this result (Fig. 17.49) with
the observation of direct event-by-event jet quenching [Aad et al., 2010].
The natural reference for determining the effect of the coloured medium
on the observables is the nuclear modification factor,
dN AA/dpT
R AA(pT ) =
< Ncoll > dNpp /dpT
where < Ncoll > is the average number of colliding nucleons, N AA and Npp
are the numbers of events selected for AA and pp collisions and pT is the
transverse momentum. The effect of the nuclear medium is demonstrated in
comparison of peripheral and central p-p and Pb-Pb collisions by the CMS
experiment [Chatrchyan et al., 2012a] (Fig. 17.50). CMS also observed
an interesting new effect on particle correlations: apparently the degree of
two-particle correlations increases with the event multiplicity, the number
of tracks in the event.

15.6 Big questions

In spite of the great wealth of data collected at SPS, RHIC and LHC there
are several open questions [Busza et al., 2018] in heavy ion physics. How
does a strongly coupled liquid emerge from an asymptotically free gauge
theory? How can the quark-gluon-plasma behave as a freely flowing liquid
with strong coupling among its constituents? How could such a liquid form
within 1 fm/c? What are the limits of the applicability of hydrodynamics in
heavy ion collisions?
166 Chapter 15

Exercise 15.1
Why pions dominate the hadron jets?

Exercise 15.2
What is the importance of the transverse momentum of jets and collision
centrality in heavy ion collisions?

Exercise 15.3
What are the characteristic features of the transition of nuclear matter into
quark-gluon plasma?

Exercise 15.4
What are the advantages and disadvantages of using time projection cham-
bers for studying heavy ion collisions?

Exercise 15.5
Deduce how one gets 2.76 GeV/NN collision energy at colliding 3.5 TeV Pb
beams. At what beam energies one gets 2.76 GeV/NN for Pb-Pb collisions
in the case of a fixed Pb target?
Chapter 16

Practical applications

MOTTO:
Prediction is very difficult, especially about
the future.

(Niels Bohr)

Particle physics is, of course, basic research, driven by the curiosity of

the researchers. However, as shown in this chapter, its spin-off has proven to
be extremely useful for the general society. In general one can state that any
kind of basic research can bring applications in the long run, even if we do
not know exactly how and when.

16.1 Informatics

The experiments at high-energy accelerators generate large amount of data.

At the LHC more than 10 PB data are recorded every year. Special techniques
were and are developed to handle them. Some of these techniques have found
already their ways into society.
167
168 Chapter 16

16.1.1 World Wide Web

The first and probably most important example is the World Wide Web. It
was invented and first developed by Tim Berners-Lee and his group at CERN
in 1989-90. As presented in the CERN museum, his boss in 1989 wrote on
the original proposal Vague but exciting. Tim Berners-Lee wrote in 1991:
“The WWW project was started to allow high energy physicists to share data,
news, and documentation. We are very interested in spreading the web to
other areas, and having gateway servers for other data.” First, of course, it
was used at HEP laboratories like Stanford, Fermilab and at the associated
universities, but in 1992 CERN announced that anybody can use the Web
free of royalty and the platform-independent Mosaic browser was opened
for the public: that began such an information explosion in the world that
seems to surpass even Gutenberg’s printing. Since Mosaic many browsers
have been developed and a lot of features added. Of course, with time the
Web became commercial as more and more companies use it for trading, but
it retains its completely free access established by CERN. Thus the free use
of the internet browsers is due to particle physics research. We mean free in
the monetary sense. Internet providers recognized quickly that information
is a new source of wealth and they collect information from their users

16.1.2 Grid computing

At the start of designing the LHC it was clear that it would become impossible
quickly to store and analyse all its data at CERN, so CERN established its
Worldwide LHC Computing Grid, WLCG. The experimental data at LHC
are analysed in several steps, in a distributed manner. The WLCG consists
of tiers. The centre, called Tier-0, is located at CERN and in the Wigner
data centre in Budapest. It receives the primary data from the LHC. After
a preliminary event reconstruction the data are distributed to the Tier-1
data storage centres, for instance these stations of CMS are located in Lyon,
Barcelona, Oxford, Bologna, Karlsruhe, Fermilab (near Chicago) and Taipei.
The actual data analysis is done at about a hundred Tier-2 computer centres
which get their data from the Tier-1 storage stations. This system was
developed by CERN for the LHC, but it helps other applications as well. For
a few years CERN helped to establish MammoGrid, a virtual organization
for diagnosing breast cancer.
 Practical applications 169

16.1.3 Computer simulation

For the detection and interpretation of new physics at high-energy particle
collisions need elaborate Monte Carlo techniques. The experience in pro-
gramming in general and in simulation methods in particular, that young
scientists learn by the time they obtain their degrees can be used in other
fields as well. Several researchers with PhD-s in high-energy physics went
to work in the insurance or banking sectors to make financial and indus-
trial forecasts using simulations. CERN has developed the GEANT general
simulation program package [Agostinelli et al., 2003] which is now widely
used, far outside high energy physics.

16.2 Radiation
Some of the discoveries in particle physics were immediately used in practice,
mostly in medicine. Wilhelm Röntgen discovered the X-rays in 1895 and
used it right away to photograph the bones in his wife’s hand. He refused to
patent the X-rays as he realized its enormous significance for the society. In
1901 Röntgen was awarded the very first Nobel Prize in Physics, which he
donated to his university.
Henri Becquerel discovered natural radioactivity in 1896, which was
followed the separation of radium by Pierre and Marie Curie in 1898, the
three of them were awarded the 1903 Nobel Prize in Physics. As early as in
1902 radioactive irradiation was already used in Stockholm for local tumour
therapy.

16.3 Accelerators
There are nearly 40 000 particle accelerators in operation worldwide, about
half of them employed for biomedical uses and the other half in industry
(for surface treatment and to produce microchips). These days less than 200
accelerators are used for particle and nuclear research.
The first cyclotron was created in Berkeley by Ernest Lawrence in 1930,
it was as small as his palm. This was followed by larger and larger machines
(Fig. 17.51) and Lawrence was awarded a Nobel Prize in physics in 1939.
His brother, John Lawrence, a physician, also worked at Berkeley and in
1936 together they produced phosphorus-32 and used it as the first artificial
radioactive isotope for treating leukaemia. This is an excellent example of
how an interdisciplinary environment helps both science and applications.
170 Chapter 16

The neutron was discovered by James Chadwick in 1932 (Nobel Prize,

1935) and since then it is used to form artificial radioactive isotopes. In 1934
an Italian group led by Enrico Fermi (Nobel Prize, 1938) created 50 new
radioactive isotopes (including those of iodine, very popular in medicine)
using slow neutrons.
The linear accelerator was invented by Wideroe in 1927, and the first
modern linac was built by Alvarez in 1946. Electron linacs are used in
hospitals worldwide, see Fig. 17.51.

16.4 Medical diagnostics

The particle detectors are more than eyes for particle physicists, they be-
came irreplaceable in medical diagnosis as well. The multiwire proportional
chamber invented by Georges Charpak in 1968 (Nobel Prize, 1992) not only
made a revolution in experimental physics, it also started very serious ap-
plications in everyday life. Its faster image registration means also lower
radiation doses for the studied tissues. When they are combined with to-
mography (Fig. 17.52, right) they can produce a 3-dimensional picture of
the inner organs.
Here we just list some of these methods in medicine.
• Magnetic resonance imaging (MRI) is based on the nuclear magnetic
resonance method and gives morphological information by measuring
the density of water (actually, of hydrogen in various molecules) in the
tissues. More than simple morphology it can give information even
on the operation of organs.
• Technetium-99m (“m” stands for metastable, its lifetime is τ = 6 h)
can be produced by irradiating molybdenum by slow neutrons:
98
Mo + n =99 Mo + γ
99
Mo(τ = 66h) =99m Tc(τ = 6h) + e− + ν e
and the 99 Tc isotope emits a 0.14 MeV photon. 85% of the nuclear
medical diagnoses use this isotope.
• Positron emission tomography (PET) uses proton-bombarded glucose
to produce Fluoro-Deoxy-D-Glucose (FDG): proton absorption makes
18 F of 18 O. 18 F emits positrons and their annihilation tells us the

location of the glucose. The glucose metabolism of the tumours is

much faster than that of healthy tissue, so the tumours will appear as
hot spots of the tomography picture (Fig. 17.52).
 Practical applications 171

16.5 Medical therapy with radiation

There are three kinds of radiotherapy used to cure various kinds of cancer:

• Teletherapy or external beam radiotherapy is the most common form

of radiotherapy. The patient sits or lies on a couch and an external
source of radiation is pointed at a particular part of the body. Most
frequently, X-rays of keV energies are used for treating skin cancer
and superficial structures and MeV X-rays are used to treat deep-
seated tumours. Electron beams are also used, but nowadays hadron
therapy (with protons or carbon ions) is spreading more.

• Brachytherapy is internal radiotherapy or sealed source radiotherapy.

It is the type of radiotherapy when radioactive sources are put straight
into the tumour to be destroyed. Fig. 17.53 shows the treatment of
prostate cancer with Brachytherapy. About 100 radioactive seeds, pins
with radioactive isotopes are placed into the tumour. The positions of
the seeds are continuously checked with an ultrasound probe.

• In radio-immunotherapy the radioactive isotope is carried by a molecule

aimed at the organ to be cured. In cancer therapy, an antibody spe-
cific for a tumour-associated antigen is used to deliver a lethal dose of
radiation to the tumour cells. For instance, α particles from Bismuth-
213 are used for leukaemia, while electrons from the beta decay of
Yttrium-90 for glioblastoma (a kind of brain tumour).

16.5.1 Teletherapy
The main problem of teletherapy is that the irradiation is not selective, it
destroys tissue not only in the tumour, but also that around it. In the case of
X-rays several beams are used from different directions with collators (In-
tensity Modulation Radiation Therapy, IMRT): using computer simulations
the crossing beams are shaped to minimize the dose in the healthy organs at
risk. Fig. 17.54 shows a simulation of a 3-dimensional conformal radiation
therapy, in which the profile of each radiation beam is shaped to fit the profile
of the target using a multi-leaf collimator and a variable number of beam
shots. Note that the radiation dose is most commonly measured in absorbed
radiation energy per unit mass: 1 Gy = 1 J/kg.
Another modern way of applying conformal radiation therapy is to-
motherapy: the accelerator rotates around the patient with changing colli-
mators. The gamma knife uses hundreds of 60 Co sources for local operation
172 Chapter 16

of brain tumours or arteries. The cyber knife is similar, but uses X-rays from
an electron linac (Fig. 17.54).
There is a way to prevent loosing cancerous tissue during operations,
which can cause metastasis, achieved by electron irradiation of the tumour
area during operation. Usually electrons of 3–9 MeV energy are applied for
a couple of minutes delivering a dose of about 10 Gy.

16.5.2 Hadron therapy

All these teletherapies hurt the tissue around the tumour, especially at en-
tering the body. This can be minimized using hadron therapy. Charged
particles lose most of their energy at the end of their flight path (see the
Bragg curve, see Fig. 17.55). This makes it advantageous to use protons and
carbon ions to irradiate tumours. With careful planning one hardly hurts the
healthy tissue and destroys the tumour.
As shown in Fig. 17.55 the apparatus for proton therapy is rather so-
phisticated and expensive. More than that, it needs a proton or heavy ion
synchrotron to provide the ion beam. Nevertheless, many developed coun-
tries have such facilities in hospitals. There are dozens in the USA, Europe
and Asia. Hadron therapy can be made much more cost-efficient with com-
pact proton synchrotrons, one for each gantry without sophisticated beam
lines. There exist designs for such tabletop accelerators.

16.5.3 Neutron therapy

Slow neutrons can easily be captured by nuclei and cause very high local
dose deposits. The depth can be modulated by irradiation with neutrons of
varying energy. As early as in 1936 (just four years after the discovery of
the neutron) G. L. Locher suggested to use neutrons for cancer treatment by
putting such nuclei in the cancer cells which undergo nuclear fission after
neutron absorption and release a lot of local energy. For that the 10 B boron
isotope seems to be ideal as it absorbs neutrons, its fragments are not tissue
unfriendly and its chemistry is well know. The problem is that it is hard to
bring it selectively into the tumour.

16.6 Conclusion
We see that particle physics offers useful methods to many other fields of
science, especially to informatics and medicine. For the latter physicists,
engineers and medical doctors should closely work together. Thus particle
 Coloured figures 173

physics brings not only new knowledge about Nature, but also useful new
techniques and technologies.

Exercise 16.1
Why was the World Wide Web developed for particle physics?

Exercise 16.2
What is the advantage of hadrons to photons (X- or γ-rays) for cancer
therapy?

Exercise 16.3
Why should we prefer to use protons and carbon ions for hadron therapy to
other nuclei?
Chapter 17

Coloured figures

MOTTO:
Colour in certain places has the great value
of making the outlines and structural planes
seem more energetic.

(Antoni Gaudí)

175
176 Chapter 17

Figure 17.1:
Graphic picture of the Lagrangian of quantum chromodynamics (QCD)

Figure 17.2: Resonance curves. Left: Breit-Wigner resonance curve: normalized

cross section against centre-of-mass energy. Right: The Z-resonance: hadronic decay
cross section against centre-of-mass energy as measured in high-energy electron-
positron collisions

Figure 17.3: Left: Bending of charged particles in a dipolar magnetic field. Right:
a dipole magnet, with two (blue) quadrupoles in the back
 Coloured figures 177

Figure 17.4: Left: Cross section of a dipole magnet of the Large Hadron Collider.
Right: Microwave resonator to accelerate electrons for the TESLA project at DESY,
Hamburg

Figure 17.5: The accelerator complex of CERN in the LEP era. LEAR: Low
Energy Antiproton Ring, EPA: Electron-Positron Accumulator, PS: Proton
Synchrotron, PSB: Proton Synchrotron Booster, AAC: Antiproton Accu-
mulator Collector, SPS: Super Proton Synchrotron, LEP: Large Electron-
Positron collider
178 Chapter 17

Figure 17.6: The Low Energy Ion Ring (LEIR) at CERN. The Pb ions are collected
and stored in the ring while continuously cooled. The pipes of the stochastic cooling
are above the ring. The electron cooling is set up on left: the two guns are for
introducing and removing the electron cloud. The bending magnets are painted
orange, the quadrupole doublets and triplets blue. LEIR uses the elements of the
former Low Energy Antiproton Ring (LEAR) functioning in 1982–1996

Figure 17.7: The accelerator complex of CERN in the LHC era, after 2008
 Coloured figures 179

Figure 17.8: Left: LHC dipoles waiting for installation. They are 15 m long,
weigh 35 tons, maintain up to B = 8 T magnetic field at T = 1.9 K. Right: The
radio-frequency cavities accelerating the particles in the LHC ring

Figure 17.9: Normal LHC operation at 13 TeV p-p collision energy . The upper
plot shows the energy (black line corresponding to the right scale) and the intensities
of the two beams (red and blue lines). On the bottom the background radiations of
the two lines around the four experiments. ATLAS and CMS needed the maximal
possible luminosities, whereas ALICE and LHCb much less
180 Chapter 17

Figure 17.10: The Antiproton Decelerator of CERN: its building and main
experimental area

Figure 17.11: 78 identified proton-proton collisions in the same CMS event.

Note that the vertices (collision points) are practically lined up. This event
was recorded, because it contains two energetic muons

Figure 17.12: The toroidal magnet of the ATLAS detector before the detection
elements and the solenoid magnet were installed in it
 Coloured figures 181

Figure 17.13: Left: Working principle of a time projection chamber. Right: The
time projection chamber of the ALICE detector at the LHC

CMS Detector
SILICON TRACKER
Pixels (100 x 150 μm2)
~1m2 ~66M channels
Microstrips (80-180μm)
Pixels ~200m2 ~9.6M channels
CRYSTAL ELECTROMAGNETIC
Tracker CALORIMETER (ECAL)
~76k scintillating PbWO4 crystals
ECAL
HCAL
Solenoid PRESHOWER
Steel Yoke Silicon strips
Muons ~16m2 ~137k channels

STEEL RETURN YOKE

~13000 tonnes

SUPERCONDUCTING
SOLENOID
Niobium-titanium coil
carrying ~18000 A FORWARD
CALORIMETER
Steel + quartz fibres
~2k channels
HADRON CALORIMETER (HCAL)
Total weight : 14000 tonnes Brass + plastic scintillator MUON CHAMBERS
Overall diameter : 15.0 m ~7k channels Barrel: 250 Drift Tube & 480 Resistive Plate Chambers
Overall length : 28.7 m Endcaps: 468 Cathode Strip & 432 Resistive Plate Chambers
Magnetic field : 3.8 T

Figure 17.14: Structure and main parts of the CMS detector at the LHC
182 Chapter 17

Figure 17.15: Preparation of the PbWO4 single crystal scintillators for CMS.
The electromagnetic calorimeter of CMS consists of 75,848 such crystals
read by avalanche photo-diodes

Figure 17.16: Scintillation tiles of CMS: the wavelength-shifting fibres collect and
carry the light to hybrid photo-diodes to be converted to electronic signals
 Coloured figures 183

Figure 17.17: The forward part of the hadron calorimeter of CMS: quartz
fibres in steel to collect Cherenkov light from secondary electrons

Figure 17.18: Inserting the central ring with the superconducting solenoid into the
CMS detector in the experimental cave of LHC underground
184 Chapter 17

Figure 17.19: The detection system of the OPAL experiment at LEP

Figure 17.20: Electron-photon events recorded by the OPAL experiment. The

yellow blocks show the energy deposits in the electromagnetic calorimeter. Left:
Bhabha scattering e+ e− →e+ e− ; middle: Bhabha scattering with photon emission
e+ e− →e+ e− γ; right: electron-positron annihilation e+ e− →γγ
 Coloured figures 185

Figure 17.21: Leptonic decays of Z bosons recorded by the OPAL experi-

ment. The yellow blocks show the energy deposits in the electromagnetic
calorimeter, the purple ones those in the hadron calorimeter. Left: Z decay
to neutrinos with photon emission from the initial state (i.e. by the electron or
the positron), e+ e− →Zγ→ννγ; middle: Z decay to muons, e+ e− →Z→µµ;
right: Z decay to a tau pair with their subsequent decay to electron, muon
and neutrinos: e+ e− →Z→τ + τ − →µe + 4ν

Figure 17.22: Hadron jets recorded by the OPAL experiment. The yellow
blocks show the energy deposits in the electromagnetic calorimeter, the purple ones
those in the hadron calorimeter. Left: two-jet event, Z decay to a quark pair,
e+ e− →Z→qq; middle: three-jet event, Z decay to a quark pair when one of them
emits a gluon: e+ e− →Z→qqg; right: WW production and decay to quark pairs,
e+ e− →W+ W− →qqqq where the quarks have different flavours to conserve charge
186 Chapter 17

Figure 17.23: Events recorded in pp collisions at the Tevatron. Left: Observation

of the t quark in the reaction pp → tt → bW+ bW− → `νjjjj. Right: Energy deposit
on the plane of pseudorapidity vs. azimuthal angle of the mysterious CDF event
pp→e+ e− γγ + Emiss

Figure 17.24: Search for charged Higgs-bosons at LEP. Left: Preselection cuts of
the OPAL search in the 4-jet channel [Horvath, 2003]. Right: Likelihood analysis in
the 4-jet channel for two assumed H± masses [Abbiendi et al., 2012]. Note how well
the data (black dots) are described by the background simulation. The signal curves
are filled yellow (left) and blue (right). The cut values are denoted by arrows
 Coloured figures 187

Figure 17.25: Multi-hadronic events detected by the OPAL experiment at LEP.

Left: e+ e− → W+ W− → qqqq → 75 charged particles in 4 hadron jets. Right: a 3-jet
event characteristic for gluon production.

Figure 17.26: Some of the major neutrino experiments. At the end of 2017 there
were 37 such detector systems under earth, water, ice, and near nuclear power plants
188 Chapter 17

Figure 17.27: KATRIN, the KArlsruhe TRItium Neutrino experiment. The spec-
trometer lets one electron of the 1010 tritium decays per second through into the
detector

Figure 17.28: Detecting neutrinos in water. Left: Cherenkov light detection in

water from neutrino reactions. Right: Filling the Super-Kamiokande detector with
water, three technicians in a boat check the phototubes
 Coloured figures 189

Figure 17.29: Left and middle: Neutrino events detected by Super-Kamiokande.

Right: The shape of the Sun depicted by the identified electron neutrinos of the Sun

Figure 17.30: The discovery of neutrino oscillations by Super-Kamiokande: neu-

trino yields vs. zenith angle for electron-like and muon-like events of sub-GeV and
multi-GeV energies. The dashed red line is simulation without, the thick green line
is with oscillation, the measurement is in black
190 Chapter 17

Figure 17.31: The SNO detector in the Sudbury Neutrino Observatory (1999-2003).
Left: a 3-dimensional drawing of the apparatus, right: the acrylic ball

Figure 17.32: The Daya Bay experiment near Hong Kong: eight ν detectors in
three halls in the vicinity of six nuclear reactors
 Coloured figures 191

Figure 17.33: Parameters of the standard model [Schael et al., 2013] as determined
by the experiment (2nd column) with uncertainties (3rd column), its prediction or fit
by the standard model (4th column) and a bar plot showing the difference between
theory and experiment divided by the experimental uncertainty. The agreement is
purely statistical as the difference is in only one case more than 2 uncertainties
192 Chapter 17

Figure 17.34: The various decay channels of the Higgs boson according to the
standard model. At lower masses bb, at higher ones W+ W dominates. Note how
small is the contribution of the important γγ channel

Figure 17.35: Higgs-like event detected by the ALEPH experiment at LEP: an

e+ e− collision produced four hadron jets containing b quarks recognized by their
secondary vertices due to longer lifetimes
 Coloured figures 193

Figure 17.36: Exclusion of the Higgs boson at LEP [Barate et al., 2003]. The
test statistic, −2 ln Q shows a significant signal (well above the standard model
expectation) for ALEPH and nothing for the other three LEP experiments at equivalent
statistical and experimental conditions

Figure 17.37: Exclusion of the Higgs boson at LEP [Barate et al., 2003]. Spaghetti
diagrams of 17 Higgs-like events detected by the four LEP experiments: signal
weights against the simulated Higgs mass. The ALEPH events crowd around
115 GeV, whereas for the other three experiments there are less of them with a
rather random mass distribution
194 Chapter 17

gluon fusion

vector boson
fusion

Figure 17.38: Production of the SM Higgs boson in p-p collisions at LHC. Gluon
fusion (upper right) has the highest cross section, vector boson fusion (lower right)
comes next

Figure 17.39: Higgs-boson-like event detected by the CMS experiment at LHC: a

boson is formed in a p-p collision and decays to two energetic gamma photons. The
block sizes at the end of the invisible photon trajectories correspond to the photon
energies deposited in the electromagnetic calorimeter
 Coloured figures 195

Figure 17.40: Higgs-boson-like event detected by the ATLAS experiment at LHC:

a p-p collision produces 4 electrons, one pair is from Z-decay as identified by the
invariant mass. The 4-electron mass corresponds to a decaying mass of 125 GeV.
Upper right: x − y view, lower right: lego plot of energy deposits in the calorimeters.
Middle right shows that there were 9 other identified vertices of p-p collisions in
the neighbourhood of the candidate Higgs decay belonging to the same LHC bunch
crossing.
196 Chapter 17

Figure 17.41: First observation of the Higgs-like boson in invariant mass distribu-
tions at around 125 GeV by ATLAS [Aad et al., 2012]. Left: H→γγ, the raw events
and also the sum of event weights according to their signal-likelihood. Top right:
H→` + ` − ` + ` − . Bottom right: the local p-values averaged for all channels. The total
p values have shown more excess events than the prediction of the standard model
(dotted line in the lower right plot)

Figure 17.42: First observation of the Higgs-like boson in invariant mass distri-
butions at around 125 GeV by CMS [Chatrchyan et al., 2012b]. Left: H→γγ, the
raw events and also the sum of event weights according to their signal-likelihood.
Middle: H→` + ` − ` + ` − . Right: the local p-values for the individual channels and
their averages. CMS saw somewhat less events than the prediction of the standard
model (dotted line in the right plot)
 Coloured figures 197

Figure 17.43: Invariant mass spectra obtained by ATLAS [Aaboud et al., 2017]
and CMS [Sirunyan et al., 2017] with 13 TeV p-p collisions for the H→` + ` − ` + ` −
reaction

Figure 17.44: Invariant mass spectra obtained by ATLAS [ATLAS, 2017] and CMS
[Sirunyan et al., 2017] with 13 TeV p-p collisions for the H → γγ reaction
198 Chapter 17

Figure 17.45: Relative signal strengths in the most significant decay channels
normalized to the predictions of the standard model as measured by ATLAS and
CMS [Aad et al., 2016] for MH = 125 GeV. Left: for various production channels
against the invariant mass of the new boson. Middle: for the various decay channels.
Right: 68% contours for various decay channels when produced in weak vs. strong
processes. All results agree within statistics with each other and with the expectations
of the standard model

Figure 17.46: Stable, unstable and metastable regions of the BEH vacuum on the
plane of the masses of the Higgs boson and the t quark according to the standard model
[Branchina et al., 2014]. The experimental point of the two masses at m H ' 125
and mt ' 173 GeV are encircled by 1σ, 2σ and 3σ contours

Figure 17.47: Simulated collision of heavy ions. In the central part the
temperature is high enough to de-confine coloured quarks and gluons
 Coloured figures 199

Figure 17.48: Jet quenching observed at RHIC. Left: The π 0 yield observed by
PHENIX from central Au+Au collisions is suppressed as compared to the d+Au
collisions of minimum bias. Right: The 2-jet events in STAR are suppressed on the
side opposite to the highest energy track for central Au+Au collisions, whereas they
are free to come out in d+Au collisions

Figure 17.49: Jet quenching in lead-lead collisions at the LHC observed by

√
the ATLAS experiment [Aad et al., 2010] at s N N = 2.76 TeV per nucleon
pair
200 Chapter 17

Figure 17.50: Jet–jet asymmetry in peripheral (left) and central (right) p-p and Pb-
Pb collisions at the LHC observed by the CMS experiment [Chatrchyan et al., 2012a]
√
at s N N = 2.76 TeV per nucleon pair. The simulation shows the expected asymmetry
with no nuclear medium

Figure 17.51: Left: Cyclotron used in medicine to produce radioactive isotopes.

Right: Linear accelerator used in a hospital

Figure 17.52: Left: computer tomography at work. Right: PET pictures of healthy
and cocaine-addict brains
 Coloured figures 201

Figure 17.53: Brachytherapy on prostate cancer: about 100 radioactive pins are
placed in the tumour, while checking the placement with an ultrasound probe

Figure 17.54: Left: conform dose distribution with IMRT: the tumour gets 70 Gy,
the metaphase region 50 Gy and the spinal cord less than 25 Gy dose. Right: Cyber
knife, electron linac on a robot arm with very precise, multiple orientations and depth
penetration
202 Chapter 17

Figure 17.55: Left: a charged particle when penetrating matter loses most of its
energy at the end of its flight path (Bragg curve). Right: Gantry for proton therapy
at a hospital

Figure 17.56: Experimental measurements of the R-ratio as a function of

the total centre-of mass energy (taken from Ref. [Beringer et al., 2012]).
 Coloured figures 203

Figure 17.57: Distribution of thrust as measured at LEP compared to per-

turbative QCD predictions obtained assuming vector and scalar gluon

0.6

ALEPH 0.3 ALEPH

x =1 x =1
2 2
s = (91.187 GeV) s = (91.187 GeV)
0.4
0.2
R3(ycut)

R4(ycut)

0.2 0.1
LO LO
NLO NLO
NLO+NLL NLO+NLL
NLO+NLL+K 0.0 NLO+NLL+K
0.0

0.2 NLO+NLL+K ALEPH & PYTHIA 0.2 NLO+NLL+K PYTHIA

3(ycut)

4(ycut)

0.1 0.5 < x < 2 ALEPH & HERWIG 0.1 0.5 < x < 2 HERWIG
0.0 0.0
-0.1 -0.1
-0.2 -0.2
-3.0 -2.5 -2.0 -1.5 -1.0 -3.0 -2.5 -2.0 -1.5 -1.0
log10(ycut) log10(ycut)

Figure 17.58: Comparison of perturbative QCD predictions to data for three-

and four-jet rates R3 and R4 at LEP [Nagy and Trocsanyi, 1999]. The data
points include corrections from hadrons to partons based on Monte Carlo
simulations. The lower panels show the differences between theoretical
√ pre-
dictions and experimental data, normalized by the latter. xµ = µ/ s, while
K refers to some well-defined terms (called cusp anomalous dimension) of
the NNLL corrections
204 Chapter 17

Figure 17.59: Jets in a proton-proton scattering event obtained with the kT

(left), anti-kT (right) clustering algorithms

Figure 17.60: QCD predictions for the thrust distribution (τ = 1 − T)

in electron-positron annihilation. Left: Predictions obtained by us-
ing the CoLoRFulNNLO method compared to data measured by the
ALEPH collaboration. The lower panels show the comparison to the
predictions of Ref. [Weinzierl, 2009] (denoted by SW) and to those of
Ref. [Gehrmann-De Ridder et al., 2007] (denoted by GGGH). Right: Same
as the left one supplemented with the prediction obtained by matching the
NNLO prediction to resummed NNNLL one
 Coloured figures 205

e−(k ")
e−(k) θ
q
p(P )
X

Figure 17.61: Deeply inelastic electron-proton scattering in the H1 detector

(left) and parton model interpretation of such an event (right). X denotes the
unobserved final state (hadrons)

Figure 17.62: Left: Combination of F2 structure functions on proton

and deuteron targets measured by the NMC collaboration and fitted with
CTEQ6D parton distribution functions. Right: CTEQ6D valence and sea
quark distributions
206 Chapter 17

Figure 17.63: Evolution of valence quark, sea quark and gluon distributions
 Coloured figures 207

Figure 17.64: Artistic view of a proton-proton scattering event at high energy

(courtesy of F. Krauss)
 Introduction to Particle Physics 209

Intermezzo

MOTTO:
What is it that breathes fire into the equations
and makes a universe for them to describe?
. . . Why does the universe go to all the bother
of existing?

(Stephen Hawking)

The first three parts of this book discussed the most important results
of experimental particle physics during the past fifty (or in some cases
more) years. As a consequence we know with certainty that the physical
manifestations of the three fundamental particle interactions–the weak, the
electromagnetic, and the strong forces–below the TeV energy scale can be
described by a model based on local quantum field theory that we call the
standard model of particle interactions. There are three almost identical
copies of particle families, containing two quarks and two leptons each (see
Table 3.5), the only difference being the mass of the particles. In the standard
model the neutral leptons are massless, but we already have experimental
evidence (the observation of neutrino oscillations) that they have masses even
210 Intermezzo

if we have not been able to measure them directly. While the experimental
evidence for the existence of only three families is strong, it remains a mystery
why there are exactly three families and how neutrinos get masses.
The mathematical model that describes the interaction is a gauge field
theory based on the SU(3)c ⊗ SU(2)L ⊗ U(1)Y symmetry where “c” refers
to colour degrees of freedom, Y stands for hypercharge and “L” abbreviates
left-handed. The gauge fields that emerge due to the gauge symmetries me-
diate the forces. These are the massless mediators of the electromagnetic and
strong forces, the photon and the gluon, furthermore the massive mediators
of the weak force, the W and Z bosons. In addition there is also a scalar
boson, the elementary excitation of the BEH field, called Higgs particle that
emerge in the spontaneous symmetry breaking of the BEH field together with
the longitudinal components of the W and Z bosons. There is continuously
increasing evidence that the scalar particle observed at the LHC has the exact
properties as predicted in the standard model. The interaction of the BEH
field with the fermions explains the masses of the fermions, although the
reason for the large hierarchy of the fermion masses is unknown. Accord-
ing to the standard model the interactions between the Higgs particle and
the fermions are proportional to the masses of the fermions, but complete
experimental verification will require new accelerators. The same is true for
the precise experimental exploration of the Higgs potential. We also lack
the reason for the particular choice for the gauge interactions by Nature and
we do not know if there is only one scalar particle.
We mentioned that apart of the continuous symmetries of gauge interac-
tions and space-time, there are the discrete symmetries of charge conjugation,
parity and time reversal, too. We discussed at length the experimental proof
for the chiral characteristics of the electroweak interactions, which means
that weak force violate the parity symmetry maximally. We also argued that
the combined symmetry of CPT is expected to be a fundamental symmetry
of Nature and mentioned the strong constraints on the experimental searches
for CPT violation. As a result, the naive expectation that particle interac-
tions respect time reversal symmetry would imply combined CP symmetry,
which is however weakly violated. While this violation can be explained
in the standard model, there is also another symmetry violation — called
baryon-anti-baryon asymmetry — that the standard model cannot explain.
There are two more big mysteries at the fundamental level in Nature.
The energy density in the Universe comes from several main sources. Some
of these are well known, like electromagnetic radiation (which turns out to
provide a small fraction, not reaching 1 % of the total energy density) and
baryonic matter that has also a quite small contribution of about 5 %. The
 Introduction to Particle Physics 211

rest of the energy density is in unknown forms, called dark matter and dark
energy. While these results emerge in cosmological observations, it might
be that the explanations will come from particle physics, but clearly beyond
the standard model, therefore these subjects are omitted from this book.
We devote the last part of our book to the introduction to the quantum
field theoretical basis of the standard model of elementary particle interac-
tions, needed to understand high energy collision events. We explore what
we can learn from studying the theory of the standard model, to what extent
the experimental observations mentioned above can be understood theoreti-
cally, and what calls for a new theory beyond the standard model. In order to
facilitate the comprehension of the mathematical details, most of the compu-
tations are performed step by step, or in some cases left to the exercises. As
prerequisite familiarity with quantum electrodynamics is assumed, although
its basics are summarized.
Our primary goal is to describe collisions at the LHC, i.e. proton-proton
scatterings at high momentum transfer. As the initial state contains parti-
cles composed of coloured constituents of the proton–quarks and gluons–,
quantum chromodynamics (QCD) is the key to the mathematical descrip-
tion. QCD is a predictive theory due to its two main features: (i) asymptotic
freedom that allows for a perturbative description and (ii) the factorization
theorem that separates the short and long-distance physics in a universal
(process independent) way. Thus we put significant emphasis to develop
these two concepts in detail and exhibit the theoretical uncertainties that
are inherent in this theoretical framework, which is crucial to estimate the
theoretical systematics of experimental results correctly. In closing the book
we discuss the basics of the theory of electroweak interactions.
 Part IV

Standard model of
elementary particles

213
Chapter 18

Gauge theories in the

standard model

MOTTO:
We need something new. We can’t predict
what that will be or when we will find it be-
cause if we knew that, we would have found it
already!

(Stephen Hawking)

18.1 Underlying gauge group

The standard model (SM) of elementary particle interactions in based upon
the principle of local gauge invariance, which means that the Lagrangian
remains invariant if the fields are transformed by an arbitrary element of a
continuous group. The underlying gauge group is

G = SU(3)c ⊗ SU(2)L ⊗ U(1)Y ,

where c stands for “colour”, L for “left-handed” and Y for “hypercharge”.

The meaning of these terms will be explained later. U(1) is an abelian
group, which means that its elements, eiϕ ∈ U(1) commute (ϕ ∈ R is a real
215
216 Chapter 18

number). Quantum electrodynamics (QED) is the prototype gauge theory

with a local U(1) symmetry. SU(N) are non abelian groups: their elements
eiH ∈ SU(N) do not commute. H is not a number any more, but an Hermitian
matrix H † = H with Tr H = 0. There are N 2 − 1 generators T a of SU(N)
that satisfy the SU(N) algebra:
+ a b, 
T , T = i f abc T c , with normalization (T a )i j (T b ) ji = TR δ ab
i, j
(18.1.1)
where TR ∈ R is the normalization factor. A general matrix H can be written
as a linear combination of the generators,

N
2 −1

H= a Ta ,  a ∈ R.
a=1

The constants f abc are the structure constants of the Lie algebra. The
structure constants are completely antisymmetric. For SU(2) the generators
satisfy the well-known angular momentum algebra
+ a b,
J , J = i abc J c .

In this case the structure constants f abc =  abc are the Levi-Civita symbols
(if any two of its indices are equal, the symbol is zero, otherwise  abc =
(−1) p  123 , with  123 = 1 where p is the number of interchanges of indices
necessary to unscramble a, b and c into the order 1, 2, 3). In the SM the
SU(3)c symmetry is unbroken, this means that SU(3) is a symmetry of the
Lagrangian and also of the vacuum, whereas the SU(2)L ⊗ U(1)Y symmetry
is spontaneously broken to U(1). First we shall discuss the unbroken part
and turn to the broken part in the last chapter.
The gauge theory based on local SU(3)c invariance is called quantum
chromodynamics (QCD). There are many similarities and some important
differences between QED and QCD. In order to see these clearly, we recall
the basics of QED.

18.2 Basics of quantum electrodynamics

First√of all a remark about the notation and the units: we use the notation
i = −1, e for the unit charge, natural units  = c = 1 and the convention

g μν = diag (1, −1, −1, −1)

 Gauge theories in the standard model 217

for the metric tensor (see (Eq. (4.1)).

The classical Dirac Lagrangian (see chapter 4) for n types of non-
interacting fermion fields is


n
- .
Lf = f¯j i∂/ − m j f j
j=1

(in QED n = 1, the electron) where

• f j is the spinor valued wave function describing a plane wave of

momentum p j ,

• f¯j = f j† γ 0 is the Dirac adjoint of f j ,

• a/ = γ μ aμ where the γ μ matrices (see chapter 4) satisfy the Clifford

algebra,
/ 0
{γ μ, γν } = 2g μν , γ μ, γ 5 = 0 . (18.2.2)

Let us perform a global (i.e., independent of the +space-time , coordinate)

phase transformation of the field, f j → f j = exp i e j θ f j where θ ∈ R
and e j expresses the charge of the fermion j in units of e (e.g. for the
electron e j = −1). The Lagrangian is invariant under such a global phase
shift, L( f j ) = L( f j), which implies the existence of a conserved charge, the
electric charge, according to Noether’s theorem.
One may wonder whether we can modify L f such that it is also invariant
under local changes of gauge,

f j (x) → f j(x) = exp[i e j θ(x)] f j (x) ≡ U j (x) f j (x) .

The physical meaning of this question is the following: can we choose the
phase of the fermion field at any x μ at will, without changing observable
quantities? The answer is yes! To make the Lagrangian invariant under
local phase transformations, called gauge transformations, we have to take
the following steps:

1. Introduce a vector-valued field Aμ (x) with transformation property

i
Aμ (x) → Aμ (x) = Aμ (x) + [∂μ U (x)]U −1 (x) ,
e
U(x) = exp [i θ (x)] .
218 Chapter 18

2. Replace the partial derivative ∂μ with Dμ [A] ≡ ∂μ + ie Q̂ Aμ ,


n
- .
Lf = /
f¯j i D(A) − m j fj
j=1

where Q̂ f j = e j f j is the electric charge operator. If

f j (x) → f j (x) = U j (x) f j (x),

then
Dμ [A] f j (x) → Dμ [A] f j(x) = U j (x)Dμ [A] f j (x) ,
so Dμ [A] f j (x) transforms the same way as the fermion field f j (x),
hence Dμ [A] is called covariant derivative.
3. The new field requires a gauge-invariant kinetic term:
1
L A = − Fμν [A] F μν [A] Fμν [A] = ∂μ Aν − ∂ν Aμ .
4
L A is a Lorentz scalar and therefore Lorentz-invariant. Since F μν [A]
is gauge invariant (the proof is left as exercise), L A is also gauge
invariant: L A = L A .
The Feynman rules can be read off the action
∫
S = i d4 x(L f + L A) ≡ S0 + SI ,

where ∫ ∫
S0 = i 4
d xL0 SI = i d4 xL I .

L0 contains all terms that are bilinear in the fields, L I contains the rest (called
interactions). The photon propagator Δγμν is the inverse of the bilinear term
in Aμ [Ryder, 1996]. In momentum space we have the condition
+ , ρ
Δγμν (p) · i p2 gνρ − pν pρ = δμ ,

i.e. equal to the unit matrix in Lorentz space. However,

+ 2 νρ ,
p g − pν pρ pρ = 0 ,
+ ,
which means that the inverse does not exist, the matrix p2 gνρ − pν pρ is
not invertible. We can exploit gauge invariance to rewrite L in a physically
 Gauge theories in the standard model 219

equivalent form (action remains the same) such that Δγμν exists, which
is called gauge fixing. For example, the covariant gauges are defined by
requiring ∂μ Aμ (x) = 0 for any space-time point x μ . Adding

1 - .2
LGF = − ∂μ Aμ , λ ∈ R,
2λ
to L, the action S remains the same. The term LGF can be seen as a
constraint, taken into account in the Lagrangian by a term with Lagrange
multiplier (like in classical mechanics). The bilinear term becomes in this
case    
2 νρ 1 ν ρ
i p g − 1− p p ,
λ
with inverse  
i pμ pν
Δγμν (p) = − 2 gμν − (1 − λ) 2 .
p p
Of course, physical results must be independent of λ. It is customary
to choose λ = 1 (called Feynman gauge). In covariant gauges unphysical
degrees of freedom (longitudinal and time-like polarizations) also propagate,
which can be avoided by choosing axial (physical) gauges, defined with an
arbitrary, but fixed vector nμ (different from pμ ):
1 - μ .2
LGF = − n Aμ ,
2λ
which leads to
 - 2 . 
i pμ nν + nμ pν n + λ p2 pμ pν
Δγμν (p, n) = − 2 gμν − + .
p p·n (p · n)2

Since p2 = m2 = 0, we have:

Δγμν (p, n) pμ = 0 , Δγμν (p, n) nμ = 0 .

Thus, only 2 degrees of freedom propagate (transverse ones in the nμ + pμ

rest frame). A usual choice is n2 = 0, λ = 0 (light cone gauge). The price we
pay by choosing this gauge instead of a covariant one is that the propagator
looks more complicated and the integral over pμ diverges when pμ becomes
parallel to nμ . In this gauge
i
Δγμν (p, n) = dμν (p, n)
p2
220 Chapter 18

with
pμ nν + nμ pν 
dμν (p, n) = −gμν + = μ(λ) (p)ν(λ) (p)∗ , (18.2.3)
p·n λ=1,2

where μ(λ) (p) is the polarization vector of the photon.

Feynman rules
In the Feynman gauge we have the following Feynman rules:

p
gμν
photon propagator: μ ν = Δγμν (p) = −i p2
p / +m j
p
electron propagator: = Δ j (p) = i p 2 −m2j
μ

μ
vertex: = Γγ f = −ie j eγ μ
j f¯j

• outgoing fermion: ū (p) • outgoing antifermion: v (p)

• incoming fermion: u (p) • incoming antifermion: v̄ (p)
• outgoing photon: μ(λ) (p)∗ • incoming photon: μ(λ) (p) .

Exercise 18.1
Show that in QED the covariant derivative transforms the same way as the
field itself, namely if f (x) → U(x) f (x), then Dμ f (x) → U(x)Dμ f (x) where
Dμ = ∂μ + i e Aμ .

Exercise 18.2
Show that in QED
+ ,
Dμ, Dν = i e Fμν
where Dμ = ∂μ + i e Aμ .

Exercise 18.3
Show that the generators of a special unitary group are traceless and her-
mitian.
 Gauge theories in the standard model 221

Exercise 18.4
The generators of the SU(2) group in the fundamental representation are the
Pauli matrices divided by two:
     
σa 0 1 0 −i 1 0
tf =
a
, σ1 = , σ2 = , σ3 =
2 1 0 i 0 0 −1
The adjoint representation of a group is defined as

(t Ab )ac = i f abc

• Compute the generators in the adjoint representation of SU(2)

We define the constant TR for a representation R by the condition

Tr[tRa tRb ] = δ ab TR

• compute this constant from the explicit form of the fundamental (TF )
and the adjoint (TA) representation.
The quadratic Casimir factor C2 (R) of a representation R is defined by

C2 (R)1 = tRa tRa
a

• compute the quadratic Casimir factor in the fundamental (CF ) and

the adjoint (C A) representation of SU(2) using the explicit form of the
representation matrices

18.3 Cross sections

The cross section of a given elementary scattering process is given by
∫
1 1
σ= 2
dφ n (p 1, . . . , p n ; Q) |M n | 2 , (18.3.4)
2ECM S spin

where ECM is the total incoming energy, Q μ is the total incoming four-
momentum (in the centre-of-mass system ECM = Q2 ), and

n 1n
dd p j  
 μ
dφn = (2π)d δ d Q μ − pj  δ+ p2
j − m 2
j (18.3.5)
j=1 j=1
(2π)d−1
222 Chapter 18

is the phase space for n final state particles in d = 4 − 2 dimensions (in

reality  → 0 such that d = 4, but we allow   0 for later purposes).
The index + of the δ-function means that we consider only positive-energy
solutions E = + m2 + p 2 , in other words
   
δ+ p2j − m2j = δ p2j − m2j Θ (E) ,

where Θ (E) is the Heaviside step-function (equals one if E is positive and

zero otherwise). The amplitude M n is obtained from all possible Feynman
diagrams G. The factor S −1 stands for the symmetry factor of identical
particles in the final state and factors needed for averaging over incoming

quantum (e.g. spin) states. One can obtain the integral of |M n | 2 over
the phase space directly from the so called “cut” diagrams following the
Cutkosky rules. Computing the amplitude for n final state partons from
Feynman diagrams,
1

iM n = G ,
G n

the cross section for the scattering of two incoming particles pa and pb into
n final-state particles obtained as the sum over all possible squared diagrams
and over all possible cuts of these diagrams:

1 1 
1
G G
σ(pa pb → p1, . . . , pn ) = .


2s S G,cuts n

The Feynman rules for the cut diagrams are the usual ones with the following
additional rules:
√
1. The sign of explicit factors of i = −1 and directions of fermion
arrows and those of all momenta are reversed in G̃ as compared to G.

2. We do not integrate over the loop momentum of initial-state momenta,

3. A cut line j in the initial state means a factor of

• /p j + m j if j is a fermion,
• /p j − m j if j is a anti-fermion,
• dμν (p j ) if j is a (massless) gauge boson.

In the final state the corresponding factors are

 Gauge theories in the standard model 223

- .  
• /p ± m j 2πδ+ p2j − m2j if j is a fermion/anti-fermion,
 
• dμν (p j ) 2π δ+ p2j if j is a (massless) gauge boson.

The δ+ distributions express the on mass-shell conditions. These convert an

integral over a loop momentum into the element of a one-particle phase-space
measure.

Example: e+ e− → μ+ μ−
We consider the reaction e+ e− → μ+ μ− as a very simple application of the
Cutkosky rules. At leading-order (LO) accuracy there is only one Feynman
diagram, shown in Fig. 18.1, that contributes to the amplitude at lowest order
to this reaction.
In order to describe the kinematics of the
reaction we use the Mandelstam variables e− pe− μ−
(see Section 5.3)
p μ−
γ∗
- .2
s = (pe− + pe+ )2 = pμ− + pμ+ ,
- .2 - .2 q
t = pe − − pμ − = pe + − pμ + , pe+ p μ+
- .2 - .2 e+ μ+
u = pe − − pμ + = pe + − pμ − .
(18.3.6) Figure 18.1: Feynman dia-
gram for e+ e− → μ+ μ− at LO
Furthermore we express the coupling as
e2 = 4π α. The cross section is then given
by
p e+
pμ−

+ − 1 1 + −
σ(e e → μ μ ) = .
2s S p e−
pμ+

Using the Feynman rules we obtain the squared matrix element from this
diagram as

 (4πα)2 2     3
|M2 | 2 = 2
Tr /pe+ − me γ α /pe− + me γ β
spin
s
2    3
×Tr /pμ− + mμ γα /pμ+ − mμ γβ
224 Chapter 18

Evaluating the traces1 we obtain for the squared matrix element, summed
over spins
( )
 (u − me2 − mμ2 )2 + (t − me2 − mμ2 )2 mμ2 + me2
|M2 | = 8 (4πα)
2 2
+2 .
spin
s2 s
(18.3.7)
We leave the integration over the phase space as an exercise. As mμ 200me ,
we usually neglect the mass of the electron.

Exercise 18.5
Derive the result in Eqn. (18.3.7) and integrate it over the two-particle phase
space dφ2 .

Exercise 18.6
Use Mathematica and the Package Tracer.m (or FORM) to compute the
following traces:
 
Tr /p2 γν ( /p1 − k/ 1 )γ μ /p1 γμ ( /p1 − k/ 1 )γν

Tr γ μ1 γ μ2 γ μ3 γ μ4 γ μ5 γ μ6 γ μ7 γ μ8 γ μ9 γ μ10

× γμ1 γμ2 γμ3 γμ4 γμ5 γμ6 γμ7 γμ8 γμ9 γμ10

18.4 Quantum chromodynamics

In quantum chromodynamics the matter fields are again fermions with an
additional (inner) degree of freedom: the colour, which can take Nc values.
(We shall see in the next chapter that in QCD Nc = 3, but it is useful to keep
it arbitrary and specify only when making comparison with experiment.)
These fields are called quark fields: q kf (with f = 1, . . . , n f , where n f is the
number of different flavours, and k = 1, . . . , Nc ). The precise matter content
is shown in Table 18.1. Thus, the Lagrangian becomes


nf 
Nc
- .
Lq = q̄ kf i∂/ − m f δkl qlf .
f =1 k,l=1

1The Mathematica package Tracer.m is useful for computing the traces. (see
http://library.wolfram.com/infocenter/MathSource/2987/)
 Gauge theories in the standard model 225

f 1 2 3 4 5 6
qf u d s c b t
mf 2.12 MeV 4.69 MeV 92.5 MeV 1.27 GeV 4.2 GeV 172.6 GeV

Table 18.1: The six quark flavours. Their baryon number is B = 1/3.
For light flavours the masses correspond to running MS quark masses (see
definition in section 19.4) at 2 GeV, for heavy flavours c and b the mass
values of Ref. [Bazavov et al., 2018] and for t quark it is the pole mass.


If we apply a transformation q k → q k = Ukl ql , with

⎧
⎪ c −1
2 ⎫
⎪
⎨ N
⎪ ⎬
⎪
Ukl = exp i t θ
a a
≡ exp {it · θ} kl
⎪
⎪ a=1 ⎪
⎪
⎩ ⎭ kl

where θ a ∈ R, then the Lagrangian again remains invariant under the global
SU(Nc ) transformation: Lq (q) = Lq (q ), therefore the colour charges are
conserved. The (t a )kl are Nc × Nc matrices, which constitute the fundamental
representation of the generators T a (called colour-charge operators). Both
satisfy Eq. (18.1.1). For SU(3) the matrices t a are the Gell-Mann matrices.
Can we make Lq (q) invariant under local SU(Nc ) transformations? The
answer is again yes, we can and the steps are similar as in the case of QED:

1. Introduce an Aμa coloured vector field with the following transforma-

tion property under SU(Nc ) transformations:

i
t · Aμ → t · Aμ = U (x) t · Aμ U −1 (x) + (∂μ U (x))U −1 (x) ,
gs

where U (x) = exp {it · θ(x)}.

- .
2. Replace ∂μ δkl with Dμ [A]kl = ∂μ δkl + igs t · Aμ kl , where gs is the
strong coupling. Dμ [A]kl ql (x) transforms covariantly, i.e. the same
way as the quark field qk (x).

3. Introduce a kinetic term

1 a
L A = − Fμν [A] F aμν [A] ,
4
226 Chapter 18

a given by
with the non-abelian field strength Fμν
a
Fμν [A] = ∂μ Aνa − ∂ν Aμa − gs f abc Aμb Aνc ,
789:
 a
≡ Aμ × Aν

so the Lagrangian contains cubic and quartic self-interactions.

Thus we find that the gauge boson field is now self-interacting. In fact, these
self-interactions are the sources of the main difference between QED and
QCD. We shall see that as a result QCD is a ‘perfect theory’ in the sense
that it is asymptotically free.2 Furthermore, among quantum field theories
in d = 4 dimensions only non-abelian gauge theories are asymptotically
free. It is also plausible that the self-interactions are the sources of colour
confinement, i.e. the colour neutrality of hadrons, but we do not have a proof
based on first principles of QCD dynamics.
Similarly to QED, the quantization requires a gauge fixing. For example,
the covariant gauge fixing is achieved by adding the term
1 - μ .2
LGF = − ∂ Aμ
2λ
to the Lagrangian (the boldface means vector in colour space and λ is a
Lagrange multiplicator). As opposed to QED, in this case the covariant
gauge requires the introduction of ghost fields (coloured complex scalars
with Fermi statistics!–hence the name) η, with the following kinetic term:

LG = ∂ μ η a † Dμ [A]ab η b .
a,b

The necessity and form of the ghost-field Lagrangian can most easily be
derived from the path integral formalism. For a physical motivation we refer
to [Ellis, 1988].
In summary, the complete Lagrangian of QCD is

LQCD = Lcl + LGF + LG ,

where ‘cl’ stand for the gauge invariant classical Lagrangian


6
Lcl = Lq (q f , m f ) + Lg (A) . (18.4.8)
f =1

2To be defined precisely later.

 Gauge theories in the standard model 227

In Eq. (18.4.8)


Nc
Lq (q f , m f ) = q̄ kf (iγμ Dμ [A] − m f )kl qlf ,
k,l=1

with
 Nc2 −1
 
(Dμ )[A]ab = ∂μ + igs Aμc T c 
 c=1 ab
and the kinetic term for the gluon field, the gauge field of QCD:

Nc −1 2
1 
Lg (A) = − F a [A]F a μν [A] .
4 a=1 μν

The ghost fields are absent if we use physical gauges, which will be our
choice (defined precisely later).
It is clear that there is an unprecedented large number of degrees of
freedom we have to sum over:
1. spin and space-time as in any field theory, not exhibited above,
2. flavour and colour, specific to QCD.
As a result, computations in QCD are rather cumbersome. During the last
25 years a lot of effort was invested to find simpler ways of computing QCD
cross sections and to automate the computations.

Feynman rules of QCD

Propagators (Feynman’s ‘+i’-prescription is assumed, but not shown):
p
a b (p) = δ a b Δ
gluon propagator: a, μ b, ν = Δg, μν γ, μν (p)
p ij
quark propagator: i j = Δq (p) = δi j Δq (p)
p
ghost propagator: a b = Δa b (p) = δ a b i
p2
(not needed
in physical gauges)

Vertices: a, μ

μ, a
quark-gluon: j i
= Γgq q̄ = −igs (t a )i j γ μ
228 Chapter 18

a, α
p

q
a, b, c
three-gluon:
b, β
r
c, γ
≡ Γα, β, γ (p, q, r) =

= −igs (F a )bc Vα, β, γ (p, q, r) , p + q + r = 0

with Vα, β, γ (p, q, r) = (p − q)γ gαβ + (q − r)α gβγ + (r − p)β gαγ
a, α b, β

a,b,c,d
four-gluon: ≡ Γα, β, γ, δ
c, γ d, δ

a, μ
μ, a
ghost-gluon: j
≡ Γgη η̄ = −igs (F a )ab pμ (not needed in
i
physical gauges)

The matrices F a constitute the representation of the generators of the colour

algebra in the adjoint representation and are defined below.
The four-gluon vertex, given by
+
Γαβγδ
abcd
= −igs2 + f xac f xbd (gαβ gγδ − gαδ gβγ )
+ f xad f xbc (gαβ gγδ − gαγ gβδ )
,
+ f xab f xcd (gαγ gβδ − gαδ gβγ ) ,
differs from the rest of the Feynman rules in the sense that it is not in a
factorized form of a colour factor and a tensor factor. This is an inconvenient
feature because it prevents the separate summation over colour and Lorentz
indices and complicates automation. We can circumvent this problem by
introducing a fake field with propagator
γ δ
a b = 2i δ ab (g αβ gγδ − g αδ g βγ )
α β
that couples only to the gluon with vertex
a, α
ξ √
x = i 2gs f xac g αξ gγζ .
ζ

c, γ
We can check that a single four-gluon vertex can be written as a sum of three
diagrams (see exercise). This way the summation over colour and Lorentz
indices factorize completely, which helps automation and makes possible
for us to concentrate on the colour algebra independently of the rest of the
Feynman rules.
 Gauge theories in the standard model 229

Exercise 18.7
Show that in SU(N) gauge theories
+ ,
Dμ, Dν = igFμν
a a
T where a
Fμν = ∂μ Aνa − ∂ν Aμa − g f abc Aμb Aνc .

Exercise 18.8
a transforms according to the adjoint representation of SU(N):
Show that Fμν

a
Fμν → Fμν
a
− f abc θ b Fμν
c
.

Exercise 18.9
Show that the four-gluon vertex can be written as a sum of three graphs with
the help of the fake field such that in each graph the colour and Lorentz
indices are factorized.

18.5 Basics of colour algebra

Examining the Feynman rules, we find that there are two essential changes
as compared to QED. One is that there is an additional degree of freedom:
colour. Another is that there are new kind of couplings: the self couplings of
the gauge field. We now explore the effect of the first, and leave the second
to the next chapter.
In order to see how to treat the colour degrees of freedom, we set to
unity all but the colour part of the Feynman rules and try first to develop
an efficient technique to compute the coefficients involving the colour struc-
ture. This is possible because the colour degrees of freedom factorize from
the other degrees of freedom completely. We use the following graphical
representation for the colour charges in the fundamental representation:
a

= (t a )i j . The normalization of these matrices is given by

j i

- .
Tr t a t b ≡ a b TR a b = TR δ
ab . The usual choice is

TR = 2 , but TR = 1 is also often used. We shall use both.

1
230 Chapter 18

In the adjoint representation the colour charge T a is represented by the

matrix (F a )bc that is related to the structure constants by
a

- .
(F a )bc = F b ca = (F c )ab = −i fabc =

b c
- . . -
where F a with a = 1, . . . , (Nc2 − 1) are Nc2 − 1 × Nc2 − 1 matrices that
again satisfy the commutation relation (18.1.1). The graphical notation in
the adjoint representation is not unique. For the matrix (F a )bc we assume
an arrow pointing from index c to b, opposite to which we read the indices of
(F a ). On the structure constants the indices are not distinguished, therefore
arrows do not appear. However, these are completely antisymmetric in their
indices, so the ordering matters. By convention, in the graphical represen-
tation the ordering of the indices is counterclockwise. The representation
matrices are invariant under SU(N) - transformations.
.

The sums a tiaj t jk a and Tr F a F b have two free indices in the fun-

damental and adjoint representations. These are invariant under SU(N)

transformations, so must be proportional to the unit matrix,
  
tiaj t jk
a
= CF δik , Tr F a F b = CA δ ab . (18.5.9)
j,a

Eq. (18.5.9) is depicted graphically as

j j CF CA a
k i k i a b b .

Here CF and CA are the eigenvalues of the quadratic Casimir operator in

the fundamental and adjoint representations. In the familiar case of angular
momentum operator algebra (SU(2)), the quadratic Casimir operator is the
square of the angular momentum with eigenvalues j( j +1). The fundamental
representation is two dimensional, realized by the ( 12 times the) Pauli matrices
acting on two-component spinors, when j = 1/2 and CF = 1/2(1/2 + 1) =
3/4. In the adjoint representation j = 1 and CA = 2. Below we derive the
corresponding values for general SU(N).
The commutation relation (18.1.1) can be represented graphically by

a b b a a b

.
 Gauge theories in the standard model 231

Let us first multiply this commutator with another colour charge operator.
Next, sum over the fermion index and then multiply with δik . As a result, we
obtain the resolution of the three-gluon vertex as products of colour charges:

TR ⇒ Tr(t a t b t c ) − Tr(t c t b t a ) = i TR f abc .


The expression a tiaj tkl a is invariant under SU(N) transformations, thus

has to be expressible as a linear combination of δil δk j and δi j δkl (the third

combination of Kronecker δ’s is not possible, the direction of arrows do not
match). The two coefficients can be obtained by making contractions with
δil δ jk and δi j δkl . Thus we obtain the Fierz identity,
  
1
tiaj tkl
a
= TR δil δk j − δi j δkl ,
a
Nc

or graphically

TR 1
Nc

We now show some examples of how one can compute the colour algebra
structure of a QCD amplitude, in particular we will also find explicit values
for CF and C A in the exercises. Taking the trace of the identity in the
fundamental and in the adjoint representation we obtain

= Nc , = Nc2 − 1 .

Then, using the expressions for the fermion and gluon propagator corrections,
we immediately find

- .
= CF Nc , = CA Nc2 − 1 .

The generators are traceless,

232 Chapter 18
= Tr (t a ) = 0 , = Tr (F a ) = 0 .

These graphical rules make the evaluation of colour algebra easy. Never-
theless, nowadays computer algebra codes make computation of colour sums
an automated procedure. For instance, you can type In[1]:= Import["http:
//www.feyncalc.org/install.m"] in a Mathematica session to explore
some code for that purpose.

Exercise 18.10
Consider the process q q̄ → ggg. Compute the colour structures that appear
in the squared matrix element.

Exercise 18.11
Insert a quark loop on a single gluon loop and try to find the value of CF .
Compute CA in a similar way. Hint: use the resolution of the three-gluon
vertex.

Exercise 18.12
Determine the colour factors A,B,C in

A
,

B ,

C .

18.6 Are we done?

We now have the Feynman rules with colour and the rest (momentum and
couplings) factorized and we gained some insight how to perform the colour
algebra. Thus it seems that we are in the position to compute the cross
section of any process up to the desired accuracy in perturbation theory, just
 Gauge theories in the standard model 233

as we can do in QED. So it may appear that conceptually we are done. Well,

we are going to see big surprises!
The first conceptual challenge is due to a phenomenological observation.
In QED perturbation theory is applicable because the elementary excitations
of the quantum fields, the electrons and photons can be observed as stable,
free particles. Thus asymptotic states are parts of the physical reality. On
the contrary, free quarks and gluons (usually called simply partons) have
never been observed in nature. This experimental fact can be reformulated
saying that the probability of observing a final state with any fixed number
of on-shell partons is zero. This negative result has been turned into the
principle of ‘quark confinement’. Thus it is questionable whether a quantum
field theory (QFT) of quark and gluon fields can describe the observed world
of particles where in addition to leptons only hadrons have been found. In
fact, a main research project at the Large Electron Positron collider (LEP)
was to find an answer to this question in a well controlled and quantitative
manner. It turned out that the answer is positive if we make an assumption
that we cannot prove from first principles:

The result of a low-order perturbative computation in QCD is an

approximation to sufficiently inclusive hadronic cross section if (i)
the total centre-of-mass energy Q of partons is much larger than the
mass of quarks, Q  mq , and (ii) Q is far from hadronic resonances
and thresholds.

We shall define precisely what ‘sufficiently inclusive’ means later. Pre-

dictions made on the basis of this assumption agree with measurements
(e.g. made at LEP) within the expected accuracy of the prediction, which we
are to define also later.
Based on this assumption, it makes sense to make predictions with quark
and gluon asymptotic states. However, in QCD the complexity of the Feyn-
man rules will make higher order computations prohibitive. Indeed, the
largest effort in QCD computations during the past 25 years went into de-
vising ever more efficient methods to decrease the algebraic complexity of
the computations. This research is driven by the observation that the QCD
Lagrangian is highly symmetric, which has to be reflected in the final results.
Thus the complications somehow appear mainly because with our rules we
artificially introduce complications at intermediate steps of the computations,
which cancels to large extent in the final formulae. In the next subsection we
shall see an example of simplifications. But learning about the symmetries
of QCD is interesting and useful not only for technical purposes, so let us
make an inventory of those.
234 Chapter 18

Figure 18.2: Illustration of the approximation of hadronic final states by

partonic events in electron-positron annihilation.

18.7 Symmetries of the classical Lagrangian

The symmetries can be grouped into two large categories: exact symmetries
and approximate ones. Space-time symmetries are exact. These consist
of invariance against continuous transformations: translations and Lorentz-
transformations (rotations and boosts). In addition QCD is invariant under
scale transformation:

x μ → λx μ Aμ (x) → λ−1 Aμ (λx) q(x) → λ−3/2 q(λx) (18.7.10)

and conformal transformations. The Lagrangian is also invariant under

parity (P) transformation, as well as simultaneous charge conjugation (C),
space and time reversal (PT)–recall Chapter 1–, in agreement with observed
 Gauge theories in the standard model 235

properties of strong interactions (C, P and T violating strong decays are not
observed).
We have already discussed exact symmetry in colour space: gauge in-
variance. In addition to the classical Lagrangian of Eq. (18.4.8), there exists
an additional gauge invariant dimension-four operator, the Θ-term:

Θgs  a a,μν 1 μναβ a

LΘ = F F̃ , with F̃ a,μν =  Fαβ ,
32π 2 a μν 2

that violates P and T. As experimentally Θ < 10−9 , we set Θ = 0 in pertur-

bative QCD.
Another interesting feature of Lcl is that it is almost supersymmetric.3
For one massless flavour
1  a a,μν
Lcl = − F F / ,
+ q̄i Dq
4 a μν

which is very similar to the Lagrangian of N = 1 supersymmetric gauge

theory,
1  a a,μν
LclSUSY = − F F / .
+ λ̄i Dλ
4 a μν

The only difference is that the quark q transforms under the fundamental,
while the gluino λ under the adjoint representation of the gauge group.
An important approximate symmetry of the classical Lagrangian is re-
lated to the quark mass-matrix. Let us introduce the quark flavour triplet

u q
   1 
ψ =  d  =  q2  ,
 s   q3 
with each component being a four-component Dirac spinor , and the combi-
nations
1- .
P± = 1 ± γ5 , γ5 = iγ0 γ1 γ2 γ3 . (18.7.11)
2
The latter are projections:

P+ P− = P− P+ = 0 , P±2 = P± , P+ + P− = 1 .

3We note this interesting feature without detailed explanation of supersymmetry,

which can be found in [Martin, 1997].
236 Chapter 18

It follows from Clifford algebra that γμ P± = P∓ γμ . We define ψ± = P± ψ.

Using γ52 = 1, we find that ψ± are eigenvectors of γ5 with ±1 eigenvalues:
γ5 ψ± = ±ψ± .
From the definition of the Dirac adjoint, ψ = ψ † γ0 , we obtain ψ± = ψP∓ .
Thus the quark sector of the Lagrangian can be rewritten in terms of the
chiral fields ψ± :
Lcl = ψ i γμ Dμ ψ = ψ(P+ + P− ) i γμ Dμ (P+ + P− )ψ =
= ψP+ i γμ Dμ P− ψ + ψP− i γμ Dμ P+ ψ =
= ψ − i γμ Dμ ψ− + ψ + i γμ Dμ ψ+ =
= L− + L+ ≡ LL + LR .
This decomposition would not work if the gluon field in the covariant deriva-
tive were not a Lorentz-vector. In this chiral form the left- and right-handed
fields decouple, so the Lagrangian is invariant under separate U(Nf ) trans-
formations for the left- and right-handed fields, i.e. under UL (Nf ) × UR (Nf ),
hence it is called chiral symmetry. Indeed, if (gL, gR ) ∈ UL (Nf ) × UR (Nf ),
then under the transformation
ψL → gL ψL , ψ L → ψ L gL† , gR = 1
LL remains invariant. This symmetry is exact if the quarks are massless. The
group elements can be parametrized using 2Nf2 real numbers {α, αa, β, βb }
(a, b = 1 , . . . Nf2 − 1),
    
 
(gL, gR ) = exp(iα) exp(iβ) exp i αa T exp i
a
βb T ,
b

a b
   
 
exp(iα) exp(−iβ) exp i αa T a exp −i βb T b
a b
∈ UV (1) ⊗ SUL (Nf ) ⊗ UA (1) ⊗ SUR (Nf )
where the matrices T a represent the generators of the group (Nf × Nf

matrices). The  transformations (exp (i a αa T a ) , exp (i a αa T a )), acting
as ψ → exp (i a αa T 1) ψ, form
a
-  a vector . subgroup
- SUV.(Nf ). How-
ever, the transformations (exp i β T b , exp −i  β T b ), acting as
-  . b b b b
ψ → exp i b βb T b γ5 ψ, do not form an axial-vector subgroup because
the algebra is not closed,

[T a γ5, T b γ5 ] = i f abc T c 1 (γ52 = 1  γ5 ).
c
 Gauge theories in the standard model 237

This chiral symmetry is not observed in the hadron spectrum. Therefore,

we assume that vacuum has a non-zero vacuum expectation value (VEV) of
the light-quark operator,
; <
0| q̄q|0 = 0| ūu + d̄d|0 (250 MeV)3 ,

a chiral condensate that connects left- and right-handed fields,

0| q̄q|0 = 0| q̄L qR + q̄R qL |0 .

The condensate breaks chiral symmetry spontaneously to SUV (Nf ) ⊗ UV (1).

This remaining symmetry explains the existence of good quantum numbers
of isospin and baryon number, as well as the appearance of Nf2 − 1 = 8
massless mesons, the Goldstone bosons. As non-zero quark masses violate
the chiral symmetry, which is broken spontaneously, the Goldstone bosons
are not exactly massless. Thus we have natural candidates for the Goldstone
bosons: we can identify those with the pseudo scalar meson octet. In
practice, we assume exact chiral symmetry and treat the quark masses as
perturbation. This procedure leads us to chiral perturbation theory ( χPT),
which is capable to predict the (ratios of) masses of light quarks, scattering
properties of pions and many more. Although, χPT is a non-renormalizable
QFT4, it can be made predictive order by order in perturbation theory if
the measured values of sufficiently many observables are used to fix the
couplings of interaction terms at the given order.
The QCD Lagrangian was written 45 years ago. Since then many at-
tempts were tried to solve it and mature fields emerged that aim at solving
the theory in a limited range of physical phenomena. For instance, χPT is a
perturbation theory that uses low-energy information (in the MeV range) to
explain the world of hadrons and the masses of light quarks. In the same en-
ergy range non-perturbative approaches, notably lattice QCD and sum rules,
have been developed for the same purpose. By now it is possible to explain
the spectrum of the light hadrons from first principles using lattice QCD
[Durr et al., 2008]. The main goal at colliders, our focus in this book, is dif-
ferent. We shall prove that perturbation theory can give reliable predictions
for scattering processes at high energies, that is the topic of jet physics.
We have seen that the classical QCD Lagrangian shows many interesting
symmetry properties that can be utilized for (i) easing computations, (ii)
checking results, (iii) hinting on solving QCD. We shall see that some of these
symmetries are violated by quantum corrections, which leads to important
physical consequences. In QCD an important example is scaling violations.
4We discuss renormalization in section 19.2.
238 Chapter 18

Another example is the axial anomaly that provides strong constraints on

possible QFT’s.

Exercise 18.13
Show that the classical action of QCD in Eq. (18.4.8) is invariant under scale
transformations (18.7.10). Is the classical Lagrangian also scale invariant?

Exercise 18.14
Show that the classical Lagrangian of QCD in Eq. (18.4.8) is invariant under
the charge conjugation defined by C(ψ) = −iγ2 ψ ∗ .

18.8 SU(N)-amplitudes at tree level

Let us see an explicit example how we can exploit symmetry properties to
make QCD computations more efficient. The number of algebraic terms that
contribute to QCD amplitudes as obtained from the Feynman rules increase
dramatically with increasing number of multiplicity (number of partons in the
final state). For instance, the number of diagrams for the reaction gg → ng
at tree level is (nd denotes the number of diagrams):

n 2 3 4 5 6 7 8 9
nd 4 25 220 2485 34300 559405 10525900 224449225

Furthermore, these diagrams contain complicated non-abelian vertices. The

number of terms belonging to a given amplitude increases factorially. We can
organize the calculation of |M n | 2 better if we divide the full amplitude into
sums of gauge invariant contributions. These can be found by expanding the
amplitude in colour space using an orthogonal basis. Then each coefficient
of a basis vector has to be gauge invariant.
A purely gluonic tree5 amplitude can be written as

... a n
M n a11... n (p1, . . . , p n ) = Tr (t a1 . . . t an ) m (p1, 1 ; . . . ; pn, n )
{1,2,...,n}

where the sum runs over (n − 1)! non-cyclic permutations of the indices,
denoted by the prime on {. . . }. The usual normalization used in this repre-
sentation is TR = 1. The different traces are orthogonal only at leading order
5Similar, but more complex decomposition can be given also for loop amplitudes.
 Gauge theories in the standard model 239

in Nc :

c −1
2
N  ∗    
1
Tr (t a1 . . . t an ) Tr t b1 . . . t bn = Ncn−2 Nc2 − 1 δ {a } {b } + O
ai =1
Nc2

where {b} is a permutation of {a} ≡ (a1, . . . , an ). However, gauge invari-

ance must hold order-by-order in the 1/Nc -expansion. So we find that the
colour sub-amplitudes m (1, 2, . . . , n) ≡ m (p1, 1 ; . . . ; pn n ) have the follow-
ing properties:
1. they are gauge invariant,
2. they are invariant under cyclic permutations of indices,
3. they obey the relation m (n, n − 1, . . . , 1) = (−1)n m (1, 2, . . . , n),
4. they obey the relation

m (1, 2, 3, . . . , n) + m (2, 1, 3, . . . , n) + m (2, 3, 1, . . . , n) + . . .

+m (2, 3, . . . , 1, n) = 0,

5. they factorize on multi-gluon poles (this feature will be discussed

later),
6. they sum independently in the squared matrix element to leading order
in Nc :
   2  3
|M n | 2 = Ncn−2 Nc2 − 1 |m (1, . . . , n)| 2 + O Nc−2
colour

18.9 Spinor helicity formalism

We compute amplitudes of fixed helicities of the external legs, which has the
following advantages:
• Helicity is conserved along massless fermion lines (vector interaction
conserves helicity).
• We can exploit gauge invariance and select an explicit representation
for the polarization vectors.
• Different helicity configurations do not interfere. Thus, in computing

helicity |M n | , we sum the helicity amplitudes incoherently.
2
240 Chapter 18

We introduce the following formalism from Ref. [Mangano and Parke, 1991]:
• ψ (p) is a massless Dirac spinor that satisfies the Dirac equation,
/pψ (p) = 0, and the on-shell condition, p = 0 .
2

• We define two chiral projections

1
ψ± (p) ≡ P± ψ (p) = (1 ± γ5 ) ψ (p) = ψ∓ (p)c ,
2
P+ + P− = 1 , P±2 = P± ,
where ψ (p)c = Cψ (p)∗ and Cγμ ∗ C −1 = γμ is the charge conjuga-
tion.
• Furthermore, we use the following notation:

|p ± ≡ ψ± (p) ,  p ±| ≡ ψ± (p)

where  p ±| is an outgoing fermion, while |p ± is an outgoing anti-

fermion of momentum p and helicity ± and

pq ≡  p − | q+ = ψ− (p) ψ+ (q)

[pq] ≡  p + | q− = ψ+ (p) ψ− (q)
; <
with normalization p ± |γμ |p± = 2pμ . In this formalism, the fol-
lowing properties hold:

pq = − qp , [pq] = − [qp] (hence pp = [pp] = 0)

 
pq ∗ = sign (p · q) [qp] , |pq| 2 = |2 p · q| ≡ s pq 
which imply s pq = pq [qp]. Furthermore,

1
2 |p± q±| = (1 ± γ5 ) γ μ q±| γμ |p±
2
1
⇒ |p± p±| = (1 ± γ5 ) /p and /p = |p+ p+| + |p− p−| .
2
Furthermore,

p±| γμ1 . . . γμ2n+1 |q± = q∓| γμ2n+1 . . . γμ1 |p∓ ,

p±| γμ1 . . . γμ2n |q∓ = − q±| γμ2n . . . γμ1 |p∓ ,

AB CD = AD CB+AC BD (the same holds with [ . . . ]) ,

 Gauge theories in the standard model 241

A+| γμ |B+ C−| γ μ |D− = 2 [AD] CB .

For the polarization vectors we make the choice

p±| γμ |k±
μ± (p, k) = ± √ (18.9.12)
2 k ∓ |p±

where k μ is the reference momentum needed to define the transverse

directions in addition to pμ . Thus
√
2
μ± (p, k) γ μ =± (|p∓ k∓| + |k± p±|) .
k ∓ |p±

Moreover,
 ∗
μ± (p, k) = μ∓ (p, k) , (18.9.13)

μ± (p, k) pμ = μ± (p, k) k μ = 0 , (18.9.14)

μ+ (p, k) ν− (p, k) + μ− (p, k) ν+ (p, k) = dμν (p, k) , (18.9.15)
- .
 h (pi, k) ·  h p j , k = 0 , with h = +, − , (18.9.16)
- .
 h (pi, k) ·  −h p j , pi = 0 , with h = +, − , (18.9.17)
- .  =
/h pi, p j phj = 0 , with h = +, − , (18.9.18)
>  - .

phj  /−h pi, p j = 0 , with h = +, − . (18.9.19)

18.9.1 Helicity Feynman rules for colour sub-amplitudes

(with massless fermions)
We introduce the following notation:

• external outgoing fermion of momentum p, helicity ±: p±|,

• external outgoing antifermion of momentum p, helicity ±: |p∓,

• external outgoing vector of momentum p, reference k, helicity ±:

p± |γ |k±
μ± (p, k) = ± √ μ .
2 k∓ |p±
242 Chapter 18

The fermion propagator of momentum pμ in the direction of the fermion

arrow is i p/2 . The vector propagator of momentum pμ is: − pi2 gμν .
p

μ gs √
Γgq q̄ = i √ γ μ the factor of 2 is due to TR = 1,
2
gs
Γαβγ (p, q, r) = i √ Vαβγ (p, q, r) all momenta incoming, (18.9.20)
2
g2 - .
Γαβγδ = i s 2gαγ gβδ − gαδ gβγ − gαβ gγδ .
2

18.9.2 A simple application of the helicity formalism

We now compute the leading order contribution to the process e+ e− → μ+ μ−
using the helicity formalism. Although it is not a physical choice, we consider
the following crossing invariant kinematic configuration, which is useful for
obtaining the squared matrix element in either the annihilation or in the
scattering channels in a single computation:
μ μ μ μ
0μ → p1 + p2 + p4 + p5 .
As for the labelling convention, we anticipate further use of the result, with
an additional particle that will be labelled by ‘3’, which is omitted here.
There is only one Feynman diagram:
e− p5 p1 μ−
γ∗

q
p4 p2
e+ μ+

A general amplitude is given by

   
A4 1h1 , 2h2 , 4h4 , 5h5 = e2 a4 1h1 , 2h2 , 4h4 , 5h5
where the numbers are just a short-hand notation for the momenta, j ≡ p j ,
which is a standard in the helicity formalism. Choosing a specific helicity
configuration and applying the Feynman rules, we find
- . (−ig μν )
a4 1+, 2−, 4+, 5− ≡ a4 (+, −, +, −) = 1+| iγμ |2+ 5−| iγν |4−
s12
2 [14] 52 45 2 [12] 25 52
=i = −i
12 [21] 45 [21] 12 45
2 25 2
= −i , (18.9.21)
12 45
 Gauge theories in the standard model 243

- .2
where si j = pi + p j and in the third line we used momentum conservation
(4/ = −1/ − 2/ − 5/) to rewrite [14] 45 as
;   < ;   < ;   <
[14] 45 = 1 + 4/ 5+ = 1 + −1/ − 2/ − 5/ 5+ = 1 + −2/ 5+
= − [12] 25 . (18.9.22)
The other helicity amplitudes can be found from Eq. (18.9.21), using discrete
symmetry transformations of parity and charge conjugation. Parity transfor-
mation reverses all helicities of the helicity amplitude. It is implemented by
the complex conjugation operation which substitutes i j ↔ [ ji]. Charge
conjugation changes anti-fermions into fermions and vice versa, which in
the present case amounts to interchanging indices 4 with 5 and/or 1 with 2.
Thus,
a4 (+, −, −, +) = a4 (+, −, +, −)| 4↔5 (from charge conjugation)
2 24 2
= −i (18.9.23)
12 54
a4 (−, +, −, +) = a4 (+, −, +, −)| i j ↔[ji] (from parity transformation)
2 [25]2
= −i (18.9.24)
[12] [45]
2 [24]2
a4 (−, +, +, −) = −i . (18.9.25)
[12] [54]
Computing the square of the amplitude, summed over helicities, we obtain
 
2 + s2
 4 s25 24 t 2 + u2
|A| 2 = e4 2 = 8 (4πα)2 ,
helicity
s12 s45 s2

where we adopted the usual notation of the Mandelstam variables, s, t and

u:
s12 = s45 = s , s24 = t , s25 = u .
μ μ μ
The advantage of considering a kinematic configuration 0 → p1 + p2 + p4 +
μ
p5 is that it is easy to compute the matrix element for analogous reactions that
can be obtained by crossing symmetry. Depending on the crossed momenta
we speak of s- or t-channel. Integrating the squared matrix element over the
phase space of the two final state particles, averaging over the initial spin
states and dividing with the flux factor, the cross section is
∫ 0
1 1 dt t 2 + u2 4πα2
σ= 8 (4πα)2 = . (18.9.26)
4 2s −s 8πs s2 3s
244 Chapter 18

Exercise 18.15
Consider highly relativistic fermions in QED. Show that the electromagnetic
interaction conserves helicity. Is helicity conserved also for axial currents?

Exercise 18.16
Show that for colour ordered amplitudes m(1, . . . , n) ≡ m(p1, 1 ; . . . ; pn, n )
the following statements hold:

• they are invariant under cyclic permutations and that

• m(n, n − 1, . . . , 1) = (−1)n m(1, 2, . . . , n).

Exercise 18.17
Show the following properties of spinor products:

•  i j [ j i ] = 2k i · k j

•  i| γ μ | j] = [ j | γ μ | i

• Fierz identity: [ i| γ μ | j  k | γμ | l] = 2 [ i l ]  k j

• Schouten identity:  i j k l =  i k j l +  i l k j

The notation introduced here is

| i ≡ | i +   i| ≡  i − |
| i] ≡ | i −  [ i| ≡  i + |

Exercise 18.18
Consider the matrix element μ∗ (k 1 )ν∗ (k 2 )M μν (e+ e− → γ(k1 )γ(k 2 )) for the
QED process e+ e− → γ(k 1 )γ(k2 ) in leading order of perturbation theory.
Verify the Ward identity
μ
k1 M μν = 0 .

Exercise 18.19
What changes in QCD? We consider the matrix element M (q q̄ → g(k1 )g(k 2 )).

• Which diagrams contribute?

 Gauge theories in the standard model 245

• Verify that

 μ (k1 )k2ν M μν (q q̄ → g(k1 )g(k 2 )) = 0 ,

μ
in case of a transversely polarized gluon g(k2 ), namely if k 2 μ (k2 ) = 0.
The requirement that the external vector bosons must be transversely
polarized was not needed in QED (see the previous exercise).

Exercise 18.20
Nevertheless, in QCD the Ward-identity holds thanks to additional (unphys-
ical) fields called ghosts. Square the above amplitude both in covariant and
in physical gauge. Do the results match? Why?

Exercise 18.21
Compute the cross section for e+ e− → q q̄ in leading order in perturbation
theory.

Exercise 18.22
Consider the matrix element M (q q̄ → g(k1 )g(k 2 )) for the QCD process
q1 q¯2 → g(k 1 )g(k2 ) in leading order in perturbation theory.
• Determine graphically the colour structures and the corresponding
colour subamplitudes m(q1, q2, g1, g2 ).
• Decompose the latter into contributions of fixed helicities.

Now determine which helicity amplitudes have to be calculated explicitly:

• Make use of helicity conservation (see also exercise 18.15).

• Use invariance under charge conjugation.

• Carefully choose two reference momenta q1 and q2 of the polarization
vector of the gluons. Note for example  + (ki, k j ) ·  − (k j , q) = 0,
 + (ki, q) ·  − (k j , q) = 0 etc. (see also Eqs. (18.9.14)–(18.9.19)).

Exercise 18.23
Compute the remaining colour-ordered subamplitudes from the previous
exercise. Use the spinor helicity formalism.
Chapter 19

Electron-positron
annihilation into hadrons

19.1 Electron-positron annihilation into hadrons

MOTTO:
“Well, in OUR country,” said Alice, still pant-
ing a little, “you’d generally get to somewhere
else — if you ran very fast for a long time, as
we’ve been doing.”
“A slow sort of country!” said the Queen.
“Now, HERE, you see, it takes all the running
YOU can do, to keep in the same place. If you
want to get somewhere else, you must run at
least twice as fast as that!”

(Lewis Carroll: Through the Looking Glass)

We now use the assumption that a low-order perturbative computation

in QCD is an approximation to sufficiently inclusive hadronic cross section,
and make predictions for the cross section of electron-positron annihilation
into hadrons.
The leading-order (LO) perturbative contribution to the cross section
247
248 Chapter 19

σ (e+ e− → hadrons) is e+ e− → q q̄. The calculation is like in the case of

e+ e− → μ+ μ− , supplemented with colour and fractional electric charge of
q j where the subscript is index of flavour. The colour graph is a loop in
the fundamental representation that corresponds to a factor Nc as we have
seen in the previous chapter. While the annihilation into μ+ μ− contains only
one flavour in the final state, quarks can have three, four or five flavours
depending on the centre-of-mass (CM) energy.1 We have to sum over all
possible flavours that can appear. The ratio of the two cross sections is thus
given by
 
σ (e+ e− → q q̄) 
R≡ = eq Nc,
2
(19.1.1)
σ (e+ e− → μ+ μ− ) q

where eu = ec = 23 and ed = es = eb = − 13 . If we consider only the u, d, s

and c quarks (the CM energy is above the threshold of producing a c-quark

pair, but below the threshold of a b-quark pair) q eq2 = 2 49 + 2 19 = 10 9 .
Considering also the contribution of the b quark above the b-quark pair

threshold, q eq2 = 119 . This step-wise increasing behaviour of the R-ratio
was observed (see figure 17.56), providing an experimental confirmation of
the existence of three families of quarks with fractional charges and of the
SU(Nc ) gauge symmetry of QCD with Nc = 3.

19.2 Ultraviolet renormalization of QCD

The strong coupling is rather large as compared to the other couplings in the
SM and as a result, the QCD radiative corrections are also large. Therefore,
it is always important to compute at least the next-to-leading order (NLO),
but if possible, even higher order corrections.2
The computation of QCD radiative corrections is technically quite in-
volved and a good organization of the calculations is very important. Thus,
first we introduce some notation. The tensor product of the ket vectors
|c1, . . . , cm  ⊗ |s1, . . . , sm  denotes a basis vector in colour and helicity
space, |A m (p1, . . . , pm ) is a state vector of n = m − 2 final-state parti-
cles in colour and helicity space. (The remaining two indices stand for the
incoming leptons.) The amplitude for producing n final-state particles of

1The sixth flavour, the t quark is so heavy that it cannot contribute at CM energies
attained in e+ e− experiments so far.
2There is even a more serious reason that we shall point out at the end of this
chapter.
 Electron-positron annihilation into hadrons 249

colour (c1, . . . , cn ), spin (s1, . . . , sn ), momentum (p1, . . . , pn ) is

c1 ... cm,s1 ... sm
Am (p1 . . . pm ) = c1 . . . cm | ⊗ s1 . . . sm |A m (p1, . . . , pm )

(m = n + 2 and appears in subscript as opposed to m denoting mass of the

particles, that never appears in subscript) so
   2

{c }, {s }
A m i i (p1, . . . , pm ) = A m (p1, . . . , pm ) |A m (p1, . . . , pm ) .
colour helicity

The loop expansion in terms of the bare coupling gs(0) ≡ 4παs(0) , which is
the coupling that appears in the classical Lagrangian, is as follows:
  q2 (   )
α(0)  = αs(0)  (1) =  
 (0) (0) 2
|A m  = s A m + A m + O (αs ) , (19.2.2)
4π 4π

where q is a non-negative integer. When we compute loop corrections, we

shall find divergent integrals in d = 4 dimensions that can be regularized
using dimensional regularization. In order to keep αs dimensionless in
d = 4 − 2 dimensions we have to introduce a mass scale μ, called the
=scale. Each factor of αs acquires a factor of μ
dimensional regularization 2

 (i)
that we include in the A m amplitudes, not shown explicitly in Eq. (19.2.2).
The exponent q2 in the prefactor takes account of the power of αs at leading-
order (LO) accuracy. For instance, q = 0 for e+ e− → q q̄, while q = 1 for
e+ e− → q q̄g. The one-loop correction to A4(0) (1, 2, 4, 5) is computed from
three diagrams:

and .

We first look at the fermion propagator correction Σμ in covariant Feynman

gauge,
≡ iΣ / (p, m) =
∫
dd 4παs μ2 + - . ,
= CF + , γμ /p − / + m1 γν (−g μν )
(2π) (p − ) − m2 2
d 2
∫ - .
dd (d − 2) /p − / − d m1
= CF δkl 4παs μ 2
+ , ,
(2π)d (p − )2 − m2 l 2
250 Chapter 19

where 1 is the 4 × 4 unit matrix, and we used −γμ a/ γ μ = (d − 2) a/ , valid for

any four-momentum aμ in arbitrary dimensions d. The momentum of the
external fermion of mass m is denoted by pμ . Σ / is diagonal in the colour
indices of the external quarks, therefore we have suppressed these indices.
The necessary integrals are
∫
dd 1
I2 = + ,
(2π)( − p) − m2 2
d 2

i   − ∫ 1  m2  −
= Γ () p2
dx x − x (1 − x) − iε (19.2.3)
(4π)2− 0 p2
∫ μ
μ dd
I2 = + ,
(2π)d ( − p)2 − m2 2
i   − ∫ 1  m2  −
μ
= Γ () p 2
p dx x 2 − x (1 − x) − iε x (19.2.4)
(4π)2− 0 p

where the gamma (or factorial) function Γ () has the property

 Γ () 1
Γ () = = Γ (1 + ) .
 
It is important to distinguish the deviation from four dimensions, represented
by , from the Feynman prescription in the propagator, represented by ε.
Finally we obtain
2   - .  3
iΣ/ (p, m) ≡ i m Σ1 p2, m + /p − m Σ2 p2, m

where
     
αs 1 4π μ2
Σ1 p , m =
2
Γ (1 + ) (−3C F ) + O 0 ,
4π  p2
     
αs 1 4π μ2
Σ2 p2, m = Γ (1 + ) (+CF ) + O 0 .
4π  p2

This result becomes meaningless if  = 0 (that is d = 4), which is cured by

ultraviolet renormalization. The philosophy behind ultraviolet renormaliza-
tion is the following. The (bare) parameters in the (classical) Lagrangian are
not physically observable. Ultraviolet renormalization is a reformulation of
the Lagrangian using physical quantities, which means that instead of the
bare quantities, one should write the Lagrangian using renormalized ones
 Electron-positron annihilation into hadrons 251

that are related to the bare ones by simple multiplication, hence the name
renormalization (R):
√
q(0) = Zq q(R) , A(0) = Z A A(R) , η(0) = Zη η(R) ,
m(0) = Zm m(R) , μ gs(0) = μR Zg gs(R) , λ(0) = Zλ λ(R) .

Any physical observable can be calculated as a function of the renormalized

parameters (in perturbation theory) and their values can be extracted from
measured values of the physical observables. In perturbation theory each Z
has a perturbative expansion of the form
∞ 
αs  n  Z (n, j)
n α   Z (1,1)
  
s (1,0)
Z= = 1+ Z + +O αs2 , (19.2.5)
n=0
4π j=0
 j 4π 

where we simplified slightly the notation, denoting αs(R) with αs . The Z (n, j)
coefficients ( j > 0) can be determined order by order in perturbation theory
from the requirement that the renormalized Lagrangian,
 
L (R) q(R), A(R), η(R), m(R), g (R), λ(R) =
 
= L q(0), A(0), η(0), m(0), g (0), λ(0)
 
− LCT q(R), A(R), η(R), m(R), g (R), λ(R)

generates finite Green functions. Thus L = L (R) + LCT where CT stands

for counter terms. LCT must have the same functional dependence as L on
its arguments. Renormalizability means that such Z constants exist (in any
order of perturbation theory). The counter-term Lagrangian can be written
as
- . - . 1 - .2
LCT = Zq − 1 q̄i∂/q − Zq Zm − 1 mq̄q − (Z A − 1) ∂[μ Aν]
 4 
- .- .† - .
+ Zη − 1 ∂μ η (∂ μ η) − Zg Zq Z A − 1 μR 4παs q̄ T · A / q
 
1 3
+ Zg Z A − 1 μR 4παs ∂[μ Aν] Aμ × Aν
2
2
  - . μ ν
R 4παs Aμ × Aν (A × A )
+ Zg2 Z A2 − 1 μ2
  - .
+ Zg Zη Z A − 1 μR 4παs ∂ μ η† iT · Aμ η
(19.2.6)
252 Chapter 19

(the counter-term for gauge fixing is not included). In principle, the renor-
malization factors of the various terms in the Lagrangian have to be computed
independently for each term. However, one can derive relations among those,
called Slavnov-Taylor identities, which reflect the symmetry of the original
Lagrangian. These relations have been already taken into account, when we
expressed the vertex renormalization factors as combinations of field renor-
malization factors and Zg . We have also omitted the superscript (R) for the
sake of brevity. The constants Z (n,0) are not determined by the requirement
of ultraviolet renormalizability. The simplest choice is Z (n,0) = 0, which de-
fines the minimal subtraction (MS) scheme. In QCD the modified minimal
subtraction (MS) scheme is used, which is defined by

Z (n,0) = Z (n,1) (ln 4π − γE ) (19.2.7)

at one loop, with γE being the Euler constant, emerging in the expansion
 
Γ (1 + ) = 1 +  γE − O  2 .

Eq. (19.2.7) holds at one-loop accuracy. To go beyond, one has to be a little

more careful. However for the purpose of this book the given definition is
enough. In this scheme
   
Zm = 1 + Σ1 p2, m | p2 =m2 = 1 − c 3CF + O αs2
∂    
Zq−1 = 1 + 2m2 2 Σ1 p2, m | p2 =m2 + Σ2 p2, m | p2 =m2
∂p
 
= 1 + c CF + O αs2 ,

where
αs 1
c = Γ (1 + ) (4π)
4π 
thus

Zq(1,1) = −CF λR , Zm(1,1)

= −3CF ,
   
(1,1) 13 λR 4 3 − λR
ZA = CA − − TR nf , Zη(1,1) = CA ,
6 2 3 4
β0 β1
Zg(1,1) = − , Zg(2,1) = − , (19.2.8)
2 4
where
11 4 34 2 20
β0 = CA − TR n f > 0 , β1 = C − CATR nf − 4CFTR nf .
3 3 3 A 3
 Electron-positron annihilation into hadrons 253

The results in Eq. (19.2.8) are valid for any value of the gauge parameter λR .
When computing scattering amplitudes in massless QCD at one-loop
accuracy, the renormalization amounts to the simple substitution
( - . )
  αs μ2R β0  
(0) 2 MS
αs μ S → αs μR μR 1 − 2 2
+ O αs 2
(19.2.9)
4π 

where SMS = (4π) exp(− γE ). The renormalized coupling αs depends on

μR . In applying the substitution (19.2.9) μ disappears and μR appears in the
amplitude. The meanings of scales μ and μR are different:
• μ is the dimensional regularization scale to keep αs(0) dimensionless,
• μR is the renormalization scale: when we extract αs from measure-
ments, we have to define μR . We will come back to this and clarify it
later.
Why does (19.2.9) work correctly? Each Feynman diagram consists of
vertices with propagators connecting those and external lines. Moreover,
• each vertex receives a factor Zg (or Zg2 for quartic vertex) and factors
√
of Zi, i = q, A for each field connected to the vertex,
• each propagator of field i receives a factor of Zi−1 ,
−1
• each external leg of field i receives a factor of Zi 2 .
Thus the renormalization field factors cancel from each diagram and only
the charge renormalization (Zg ) is needed in practice! This can be seen
as a consequence of the fact that in massless QCD the only free parameter
besides the gauge-fixing parameter λ is αs . The scattering amplitudes are
physical, and any physical quantity has to be independent of λ, so the only
remaining parameter, which the amplitudes may depend on, is the coupling.
The renormalization factor Zg is most easily computed in background-field
gauge [Abbott, 1981], defined by
1  μ a 2
LGF = − ∂ Aμ − g f abc A μb Aμ c ,
2λ a

where A μb is a background field and Aμc describes the quantum fluctuations

on this background. It can be shown that in this gauge
−1
Z A2 = Zg,
254 Chapter 19

and Z A can be computed from loop insertions into the propagator.

Exercise 19.1
Compute the contribution to the beta function from the fermion loop:

1. Write down carefully the amplitude and compute the trace assuming
n f number of massless quarks..

2. The following types of integrals occur:

∫ μ ∫ μ ν
μ dd μν dd
I2 = , I2 =
(2π)d 2 ( − p)2 (2π)d 2( − p)2
(19.2.10)
Express these as linear combination of
∫
dd 1
I2 (p) = . (19.2.11)
(2π)d 2( − p)2

3. Obtain I2 from
∫
dd l 1
I2 (p, m) = + ,
(2π) (l − p) − m2 l 2
d 2

i   − ∫ 1  m2  −
= Γ () p2
dx x − x (1 − x) − iε
(4π)2− 0 p2
(19.2.12)

and find the divergent pieces.

The contribution to β0 is the coefficient of the 1/ pole without the coupling
factor.
 Electron-positron annihilation into hadrons 255

19.3 Running coupling

We consider a dimensionless
- . quantity that depends on a single dimensionful
parameter, R = R Q2 . This could, for example, be the ratio of cross
sections of the production of hadrons and a μ+ μ− pair in e+ e− collisions,
that we already discussed before:
σ (e+ e− → hadrons)
R= ,
σ (e+ e− → μ+ μ− )
√
and Q = s = Ee+ + Ee− (the second equality is valid in the centre-of-mass
frame). If R depends on other dimensionful parameters such as masses,
we assume that Q is much bigger than all those, which can be set to zero.
Classically, dim R = 0 and, since Q is dimensionful, it follows that dQ
dR
= 0.
So limQ2 →∞ R =constant. In renormalized QFT, R depends also on μR , thus
it need not be a constant,
 
Q2  
R = R 2 , αs μ2R  constant.
μR
This is called scaling violation. Note that the first term in parenthesis is
only a dimensionless combination of Q and μR . However, μR is arbitrary.
If R depended on μR , then its value could not be predicted. From now on
we drop the subscript “R” from μR and write μ, meaning μR . As R cannot
depend on μ, we can write:
 2    
2 d Q 2 ∂ 2 ∂αs ∂
0=μ R 2 , αs μ 2
= μ +μ R. (19.3.13)
dμ2 μ ∂ μ2 ∂ μ2 ∂αs
Eq. (19.3.13) is called the renormalization group equation (RGE). Keeping
αs(0) fixed, setting t = ln Q
2

μ2
and defining

∂αs  ∂αs (μ2 )
β (αs ) = μ2  =− ,
∂ μ αs(0) fixed
2 ∂t
we obtain the partial differential equation
 
∂ ∂ - .
− + β (αs ) R et , αs = 0. (19.3.14)
∂t ∂αs
- .
We can solve this equation by defining αs Q2 , the running coupling as
∫ αs (Q2 )
dx
t= . (19.3.15)
αs (μ )
2 β (x)
256 Chapter 19

- .
This
- 2gives
. αs Q2 implicitly for a known, arbitrarily fixed number αs ≡
αs μ . The derivative with respect to the variable t of (19.3.15) gives
- .
1 ∂αs Q2
1 = - - 2. . ,
β αs Q ∂t
which implies - .
  
∂αs Q2
β αs Q =
2
.
∂t
The derivative of (19.3.15) with respect to αs gives
- .
1 ∂αs Q2 1 ∂αs
0 = - - 2. . − ,
β αs Q ∂αs β(αs ) ∂αs
from which it follows that
- . - - ..
∂αs Q2 β αs Q2
= .
∂αs β(αs )
- - ..
It is now easy to prove that R 1, αs Q2 solves (Eq. (19.3.14)):
- .
∂    ∂R ∂αs Q2    ∂R
− R 1, αs Q 2
=− - 2. = −β αs Q2 - .
∂t ∂αs Q ∂t ∂αs Q2
and
   - .
∂ ∂αs Q2 ∂R
β(αs ) R 1, αs Q2 = β(αs ) - .
∂αs ∂αs ∂αs Q2
   ∂R
= β αs Q2 - ..
∂ αs Q2
- .
Thus if we
- know
. αs Q2 , we automatically
- - 2. . know the Q2 -dependence of
R. To get αs Q , we need the β αs Q function:
2

∂αs 
Q2 = β(αs ) . (19.3.16)
∂ Q2 αs(0) fixed
Our discussion so far is valid generally, with the single condition that R
depends on Q2 only and our computations were non-perturbative.
We can obtain the β function in perturbation theory. In the MS scheme
μ2 αs(0) = μ2
R Zg αs . Taking the derivative μ R dμ2 on both sides we obtain
2 2 d
R
( )
d
0 =  αs + αs μ2R + β (αs, ) Zg2
dμ2R
 Electron-positron annihilation into hadrons 257

and thus
μ2R d Zg2
β (αs, ) = −αs − αs .
Zg2 dμ2R
μ2R d Zg2
Let us define f (αs ) = Zg2 dμR2
. By definition, in the MS scheme Zg depends
- 2.
on μR only through αs μR , so
∂ Zg2 ∂ Zg2 ∂ Zg2
Zg2 f (αs ) = β (αs, ) = −αs − αs f (αs ) .
∂αs ∂αs ∂αs
Since  is arbitrary, the equality must hold at each power of . The terms
independent of  are

∞  α  n−1 2 Z (n,1)
s g
1 · f (αs ) = − αs n
n=1
4π 4π

(‘1’ refers to the first term in the expansion of Zg2 ). Thus,


∞  α  n−1 2 Z (n,1)
s g
β(αs ) = β (αs,  = 0) = −αs f (αs ) = αs2 n
n=1
4π 4π

∞  α  n+1
s
= −αs βn , (19.3.17)
n=0
4π

where we used the convention of setting Zg(n,1) = − β2n

n−1
. Another often used
convention is ( )
 ∞
β(αs ) = −b0 αs 1 +
2
bn αs ,
n
(19.3.18)
n=1
β0 β1 β1
where b0 = and b0 b1 = thus b1 = 4πβ
, . The first two coefficients
4π (4π)2 0
in the expansion of the β function are independent of the renormalization
scheme. - .
If αs Q2 is small we can truncate the series. The solution at leading
order (LO) is
∂αs ∂αs
Q2 = = −b0 αs2 ,
∂Q 2 ∂t
 
1 1
⇒ − - . − - . = −b0 t ,
αs Q2 αs μ2
  - .
αs μ2
⇒ αs Q =2
- ., (19.3.19)
1 + b0 t αs μ2
258 Chapter 19

- 2. - 2.
- 2 .gives αs Q as a function of αs μ if both are small. The coupling
which
αs μ is a number to be extracted from data. We observe that the running
coupling tends to zero at asymptotically large energies,
  Q2 → ∞ 1
αs Q2 −→ → 0. (19.3.20)
b0 t
This behaviour is called asymptotic freedom. The sign of b0 plays a crucial
role in establishing whether or not a theory is asymptotically free. If it is
positive, so the β function is negative, then the use of perturbation theory
is justified: the higher Q2 , the smaller the coupling. The coefficient b0
is easiest to compute in background-field gauge where only two diagrams
contribute:

The contribution of the quark loop is negative − 43 TR nf , while that of the

gluon loop is positive 11 3 CA . The net result is that the β function is negative
up to nf < 16 in QCD. In 2004 David J. Gross, Hugh D. Politzer and Frank
Wilczek received the Nobel prize for their discovery of asymptotic freedom
in QCD.
Clearly, it is the gluon self-interaction that makes QCD perfect in pertur-
bation theory. In QED b0 < 0, hence the coupling increases with energy, but
remains perturbative up to the Planck scale where we expect that any known
physics breaks down.
The analysis is also simple at next-to-leading order (NLO):
+ , −1 ∂αs
αs2 (1 + b1 αs ) = −b0 .
∂t
- .
αs Q2 is then given implicitly by the equation
- . - .
1 1 αs Q2 1 + b1 αs Q2
- . − - . + b1 ln - . − b1 ln - . = bt ,
αs Q2 αs μ2 αs μ2 1 + b1 αs μ2

which can be solved numerically.

Measuring αs (μ2 )
- . - .
We know αs Q2 if αs μ2 is known. Therefore, we have to measure αs
at some scale μ. The perturbative solution of the renormalization group
 Electron-positron annihilation into hadrons 259

equation (RGE,- Eq.. (19.3.13)) is never unique. The difference

- . between two
solutions at O αsn is suppressed by αs , i.e. at O αsn+1 . Nevertheless,
- .
this difference can lead to a significant numerical difference in αs Q2 if μ2
and Q2 are far from each other, which is important in present day precision
measurements. Therefore, the scale μ is chosen to be μ = MZ because
MZ 91.2 GeV is not far from the scales where αs (Q2 ) is used in current
experimental analyses.

Another approach to solving the RGE is to introduce a reference scale Λ

by
∫ ∞
Q2 dx
ln = - . .
Λ2 αs Q2 β(x)
The scale Λ indicates where the coupling becomes strong. The following
exercise is to explore the characteristics of this choice.

Exercise 19.2
The running of the strong coupling is given by Eq. (19.3.16). The pertur-
bative expansion of the QCD beta function is given by Eq. (19.3.18) with
b0, b1 ≥ 0. Determine (i) the expression for the coupling in leading order
(b0  0, b1 = 0) and the corresponding scale Λ0 (see below) (ii) the ex-
pression for the coupling in next-to-leading order (b0  0, b1  0) and the
corresponding scale Λ1 (see below).
Suggested steps:

1. Solve the differential equation for αs (μ); you’ll get an integration

constant.

2. Express your result in the form

1
αs (μ) =
K ln( Λμ 2 )
2

where K is a constant.

3. Solve the differential equation using b1  0

∫
1 b0 + b1 αs log(αs ) − b1 αs log(b0 + b1 αs )
dαs = +K
−b0 αs2 − b1 αs3 b20 αs
260 Chapter 19

4. This time the solution cannot be solved for αs analytically. One can
nevertheless find an approximate solution by expanding αs in log Λμ 2 .
2

1
The constant K is not equal to the one in the first part of this exercise.

5. Cast your equation for αs into the form

1 1
αs = ln(c2 +b0 αs )
μ2
K ln 1 + c1 μ2
Λ21 ln
Λ2
1

with a suitable choice of Λ1 .

6. Expand the right hand side of your equation in t = 1

μ2
and keep only
ln
Λ2
1
the first order term. Use the expansion
 
1 1
= 1 − t C1 ln + O(t) .
1 + C1 t ln(C3 1t + C2 ) t

19.4 Quark masses and massless QCD

Quark masses mq are parameters of LQCD like the gauge coupling, which
need to be renormalized. In QED the electron mass is measured in the
laboratories at μ2R = 0 (classical limit). We cannot similarly isolate a quark
at μ2R = 0 (at low scale quarks are confined). Instead, we can perform a
similar RGE analysis as with αs . For simplicity we assume one quark flavour
with mass m, which is yet another dimensionful parameter, so the RGE
becomes:
   2 
2 ∂ ∂ ∂ Q m
μ + β(αs ) − γm (αs )m R 2 , αs, = 0 , (19.4.21)
∂ μ2 ∂αs ∂m μ Q

where γm is called the mass anomalous dimension and the minus sign before
γm is a convention. In perturbation theory we can write the mass anomalous
dimension as   
γm (αs ) = c0 αs 1 + c1 αs + O αs2 , (19.4.22)
 Electron-positron annihilation into hadrons 261

with known coefficients up to c3 . At NLO accuracy we need only c0 = 1

π and
303−10 n f
c1 = 72π . As R is dimensionless, the dependence on the dimensionful
parameters has to cancel
   
∂ ∂ ∂ Q2 m
Q2 + μ2 2 + m2 R , αs, = 0. (19.4.23)
∂Q 2 ∂μ ∂ m2 μ2 Q

The difference of Eqs. (19.4.21) and (19.4.23) gives the dependence of R on

Q:
     2 
∂ ∂ 1 ∂ Q m
Q2 − β (αs ) + + γ (αs ) m R , α s, = 0.
∂ Q2 ∂αs 2 ∂m μ2 Q
(19.4.24)
This equation is solved
- .by introducing the running mass (in addition to the
running coupling) m Q2 obeying

∂m  
Q2 = −γ m (αs ) m Q2 , (19.4.25)
∂ Q2

from which - . ∫ Q2 2   
m Q2 dq
ln - 2. = − γ m αs q2 .
m μ μ 2 q 2

Exponentiating and using the definition of the β function, we obtain

( ∫ )
    αs (Q 2 )
γm (αs ) Q2 →∞
m Q = m μ exp −
2 2
dαs −→ 0, (19.4.26)
αs (μ2 ) β (αs )

which means that asymptotically free QCD is a massless theory at asymptot-

ically high energies. Thus we can use the massless approximation of QCD
for scattering processes at large energies. At LO in perturbation theory the
solution of (19.4.26) is given by

γm (αs ) c0   2   3 bc0
− = ⇒ m Q2 = m αs Q2 0 ,
β (αs ) b0 αs
- . + - . , − c0
where we introduced the abbreviation m = m μ2 αs μ2 b0 . At NLO
the solution becomes
  2   3 bc0  c0       
m Q = m αs Q
2 2 0
1+ (c1 − b1 ) αs Q − αs μ
2 2
+ O αs2 .
b0
262 Chapter 19

 2)

In terms of the running coupling and mass R 1, αs (Q2 ), m(Q Q is a solu-
- .
tion of Eq. (19.4.24), proven similarly as R 1, αs (Q2 ) being the solution of
Eq. (19.3.13). Expanding around m(Q2 ) = 0, we obtain
   2   ∞  n  2 
m(Q2 ) Q 1 m(Q2 ) Q
R 1, αs (Q2 ), = R 2 , αs, 0 + R(n) 2 , αs, 0 .
Q μ n=1
n! Q μ
(19.4.27)
We see from Eq. (19.4.27) that derivative terms are suppressed by factors of
m( Q 2 )
1/Q n at large Q2 . From the dependence of R on Q we can conclude
that the effect of mass is suppressed at high Q2 by its physical dimension
and also by its anomalous dimension, which justifies the assumption about
negligible quark masses for Q2  m2 , i.e. for the first three, four or five
flavours depending on the characteristic energy scale of the collision. The
condition Q2  mt2 is not generally fulfilled for the t quark in colliders at
present.

Exercise 19.3
Find the definition of γm (αS ) in (19.4.21). How does one determine γm
using Zm (μ)? Compute the value of c0 in the perturbative expansion of the
mass anomalous dimension (in Eq. (19.4.22)).

19.5 Consequences of renormalization and renor-

malization group equation
We now summarize several important consequences we could read off the
analysis carried out in the previous sections.

1. QCD is an asymptotically free theory, the gauge coupling and quark

masses decrease with increasing Q2 .

m( Q 2 )
2. If we want to avoid large logarithms of Q , we should consider
physical observables (that is physically measurable quantities) that
 Electron-positron annihilation into hadrons 263

have a finite zero-mass limit (called infrared-safe observable quantity):

 
Q2   m -Q 2 .  2
Q     p
m
R 2 , αs Q , 2
= R 2 , αs Q2 , 0 +O ,
μ Q μ Q
789:
infrared-safe observable

(p > 0). As a result, light quarks (u, s, d, always, c usually, and

often b too) are considered massless. For massless quarks helicity
is conserved in the interaction with a vector boson, which can be
exploited to simplify computations significantly (see the spinor helicity
method).
3. For scattering amplitudes, renormalization is achieved by the substi-
tution
( - . )
  αs μ2R β0
(0) 2 MS
αs μ S → αs μR μR 1 −
2 2
+ . . . , (19.5.28)
4π 

with
SMS = (4π) exp(−γE ) ,
which amounts to a simple shift in the amplitude:
 =  =
 (0)  (0)
M m = C(μR, μ, q; ) A m q∈N (19.5.29)
 =
 (1)
M m = C(μR, μ, q; )
- . ( 2    −1  = q β  =
)
αs μ2R μR  (1) 0  (0)
× SMS A m − A m
4π μ2 2 
(19.5.30)
 - . q
2
 μ2  q
2
  − q2
where C(μR, μ, q; ) = 4παs μ2R μ2
R
SMS . The ampli-
(i)
tudes A m are the terms in the expansion of the unrenormalized ampli-
tude, whereas the formal  perturbative
=  =expansion of the renormalized
 (0)  (1)
amplitude is |M m  = M m + M m + . . .. The factors μ2R /μ2 re-
move the dependence on the dimensional regularization scale μ and
introduce the dependence on the renormalization scale μR .
 =
 (1)
4. The renormalized theory is ultraviolet finite. We shall see that M m
is divergent also in the infrared. We can use dimensional regularization
264 Chapter 19

to regulate the amplitudes in the infrared by continuing- into. −d > 4

( < 0). The integrals that are scaleless in d = 4 have q2 mass
dimension in d = 4 − 2. Therefore, in the massless limit all integrals
can depend only on momentum invariants raised to a positive fractional
power ( < 0!). We conclude that when all external invariants vanish,
the continued integral must also vanish (“scaleless integrals vanish in
dimensional regularization”).

5. For infrared-safe observables the radiative corrections are finite in

d = 4, we can thus take the limit  → 0, or for the sake of simpler book-
q
keeping, we can set C(μR, μ, q; ) → (4παs ) 2 from the beginning of
the computations.

19.6 Re+ e− at NLO

We have seen that the loop insertion on a fermion propagator
- . −
∝ p2 where pμ is the four-momentum leaving the
loop and for the UV-renormalized theory  < 0 to regulate the IR diver-
gences. On a massless external leg, which is on-shell, p2 = 0, so
= 0. Therefore, only the vertex correction gives a
contribution, given by the graph

- .
≡ A4(1) 1h1 , 2h2 , 4h4 , 5h5 .

Applying the helicity Feynman rules, we find

αs (1)  h1 h2 h4 h5  4παs μ2

A4 1 , 2 , 4 , 5 = CF e2 eq
4π 2
∫
dd ; h1  μ α    <
d
1 γ /γ / − /p1 − /p2 γν 2−h2
(2π)
- .- .
−gμν −gαβ ; h  β  −h <
× 5 5  γ 4 4 (19.6.31)
2( −p ) ( −p −p ) s
2 2
1 1 2 12

4παs μ2
= CF e2 eq Iρσ (p1, p2 )
2s12
; h  μ ρ α σ  −h < ; h   −h <
× 1 1  γ γ γ γ γμ 2 2 5 5  γα 4 4 ,
 Electron-positron annihilation into hadrons 265

where

ρσ ρ
I ρσ (p1, p2 ) = I3 (p1, p2 ) − (p1 + p2 )σ I3 (p1, p2 )
∫ ρ σ
ρσ dd
I3 (p1, p2 ) = =
(2π)d 2( − p1 )2 ( − p1 − p2 )2
ρ ρ - ρ ρ.
a0 g ρσ s12 + a11 p1 p1σ + a22 p2 p2σ + a12 p1 p2σ + p1σ p2
∫ ρ
ρ dd ρ ρ
I3 (p1, p2 ) = = a1 p1 + a2 p2 .
(2π) d 2( − p1 ) ( − p1 − p2 )2
2

(19.6.32)

We now use algebraic relations to express these tensor integrals as a linear

combination of two scalar integrals (procedure called reduction to master
integrals),

∫
dd 1
I2 (p) =
(2π) d 2( − p)2
i   −
= −p2 Γ () B (1 − , 1 − ) (19.6.33)
(4π)2−
i 1 Γ2 (1 − ) Γ (1 + )  2  −
= −p ,
(4π)2−  Γ (2 − 2)

∫
dd 1
I3 (p1, p2 ) =
(2π) d 2(
− p1 ) ( − p1 − p2 )2
2

i1 Γ2 (1 − ) Γ (1 + )
=− (−s12 )−1− (19.6.34)
(4π)2−  2 Γ (1 − 2)
1 1 − 2
= I2 (p1 + p2 ) ,
s12 

which we can obtain using Feynman parametrization. The algebraic relations

can be derived by contracting the integrals with the external momenta and
266 Chapter 19

the metric tensor.

μ
2 p1 Iμ(3) (p1, p2 ) = a2 s12 , (19.6.35)
μ
2 p2 Iμ(3) (p1, p2 ) = a1 s12 , (19.6.36)
(3)
g μν Iμν (p1, p2 ) = (d a0 + a12 ) s12 , (19.6.37)
μ ν (3)
4 p1 p2 Iμν (p1, p2 ) = 2
s12 (2a0 + a12 ) , (19.6.38)
μ (3)
4 p1 pν1 Iμν (p1, p2 ) = 2
s12 a22 , (19.6.39)
μ (3)
4 p2 pν2 Iμν (p1, p2 ) = 2
s12 a11 . (19.6.40)
We can substitute the numerators in the integrands on the left hand sides
of Eqs. (19.6.35)–(19.6.40) with denominator factors using the algebraic
relations
 
2 · p1 = 2 − ( − p1 )2 p21 = 0 ,
 
2 · p2 = ( − p1 )2 − ( − p1 − p2 )2 + s12 p21 = p22 = 0 .

We do not need all coefficients in our calculation. We now perform the

spinor algebra.
γ μ γ ρ γ α γ σ γμ = (4 − d) γ ρ γ α γ σ − 2γ σ γ α γ ρ .
The gamma matrices γ ρ and γ σ will be contracted with either
(a) gρσ , giving γ μ γ ρ γ α γρ γμ = (d − 2)2 γ α , or
(b) p1 and p2 .
• Case (a):
; h μ ρ α  <;   <
1 1  γ γ γ γρ γμ 2−h2 5h5  γα 4−h4 =
;   <;   <
= (d − 2)2 1h1  γ α 2−h2 5h5  γα 4−h4
789:
−i s12 a4 (h1,h2,h4,h5 )

where a4 (h1, h2, h4, h5 ) was computed in Eqs. (18.9.21)–(18.9.25).

• Case (b): the only non-zero contributions are
; h  ρ α σ  −h < ;   <
1 1  γ γ γ 2 2 p1σ p2ρ = 1h1  2/ γ α 1/ 2−h2
;  <;   <;  <
= 1h1  2−h1 2−h1  γ α 1h2 1h2  2−h2
(19.6.41)
 Electron-positron annihilation into hadrons 267

;   < ;  <;   <;  <

1h1  γ σ γ α γ ρ 2−h2 p1ρ p2σ = 1h1  2−h1 2−h1  γ α 1h2 1h2  2−h2
;   <
= −s12 1h1  γ α 2−h2
(19.6.42)

where we used h1 = −h2 . Therefore, we need to compute a0, a1, a2 and a12
only. From Eqs. (19.6.35) and (19.6.36) we find
∫
1 dd ( − p1 )2 − ( − p1 − p2 )2 + s12
a1 =
s12 (2π) d 2(
− p1 )2 ( − p1 − p2 )2
⎡ ⎤
1 ⎢⎢ ⎥
⎥
= ⎢ I2 (p1 + p2 ) − I2 (p1 ) +s12 I3 (p1, p2 )⎥
s12 ⎢ 789: ⎥
⎢ ⎥
⎣ =0 ⎦
 
1 1
= − 1 I2 (p1 + p2 ) , (19.6.43)
s12 
∫ 2 − ( − p )2
1 dd 1
a2 =
s12 (2π)d 2 ( − p1 )2 ( − p1 − p2 )2
⎡ ⎤
1 ⎢⎢ ⎥
⎥
= ⎢ I2 (p2 ) −I2 (p1 + p2 )⎥
s12 ⎢789: ⎥
⎢ ⎥
⎣ =0 ⎦
1
= − I2 (p1 + p2 ) . (19.6.44)
s12
From Eqs. (19.6.37) and (19.6.38) we find

(d a0 + a12 ) s12 = I2 (p2 ) = 0

∫
dd
(2a0 + a12 ) s12 =
2
(2π)d
 + ,2
2 ( − p )2 − 2( − p1 − p2 )2 + 2s − ( − p 1 )2
1 12
×
2( − p1 )2 ( − p1 − p2 )2

( − p1 )2 ( − p1 − p2 )2 − ( − p1 )2 s12
+
2( − p1 )2 ( − p1 − p2 )2
μ
= s12 I2 (p2 ) + 2p1μ I2 (p1 + p2 ) − s12 I2 (p1 + p2 )
s12
=− I2 (p1 + p2 ) .
2
268 Chapter 19

In the last step we used

∫ μ
μ dd
I2 (p) = = a (p) pμ ,
(2π) d 2( − p) 2

where contraction with pμ yields an equation for the unknown coefficient

a(p),
∫
1 dd l l 2 − (l − p)2 + p2 p2 (2)
a (p) p2 = = I (p) .
2 (2π) d
l 2 (l − p)2 2
Therefore,
1
a0 = s−1 I2 (p1 + p2 )
2 (d − 2) 12
and
d
a12 = − s−1 I2 (p1 + p2 ) .
2 (d − 2) 12
From Eqs. (19.6.39) and (19.6.40) we can also compute the coefficients that
are not needed in our calculation:
d
a11 = − s−1 I2 (p1 + p2 ) , (19.6.45)
2 (d − 4) 12
1 −1
a22 = − s12 I2 (p1 + p2 ) . (19.6.46)
2
We can now simply collect all contributions to the vertex correction. We
obtain
αs (1)  h1 h2 h4 h5  4παs μ2 2  3
A4 1 , 2 , 4 , 5 = CF e2 eq −is12 a4 1h1 , 2h2 , 4h4 , 5h5
4π 2s12
?
× a0 s12 (d − 2)2 + a12 [(4 − d) (−s12 ) − 2 (−s12 )]

− [(4 − d) a2 (−s12 ) − 2a1 (−s12 )]
 
αs 4π μ2 1 Γ2 (1 − ) Γ (1 + )
= CF e eq 2
−
4π s12  Γ (1 − 2)
 
 1 3  
× a4 1h1 , 2h2 , 4h4 , 5h5 − − − 4 + O  2 .
 2
N 2 −1
Remember that in the helicity formalism TR = 1 and CF = Nc c . To get the
result for TR = 12 one can multiply the expression in [ . . . ] with 2 and use
Nc2 −1
CF = 2Nc , which of course, does not change the amplitude itself.
 Electron-positron annihilation into hadrons 269

The above method of computing the vertex correction is general, but it

becomes rather involved for more legs. At one loop the problem is considered
being solved although the algebraic complexity grows so fast that the general
method to compute the one-loop amplitudes is numerical, which becomes
computationally expensive and =may become unstable for many external legs.

For the second term M n(1) in the loop-expansion of the renormalized
amplitude a general form is known (with TR = 1
2 normalization):
 =   α(R)  =   =
 (1) s  (0)  (1) fin 2
M n = I n , μ2, {pi } M n ({pi }) + M n μ , {pi } + O ()
4π
where
 =   =   =
 (1)  (0)  (1) fin 2
M n = I n , μ2, {pi } M n ({pi }) + M n μ , {pi } + O ()

and the upper index “fin” indicates

- a term
. in the amplitude that remains finite
as  → 0. The factor I n , μ2 ; {pi } in massless QCD is given by3
 n n  2  
  (4π)  μ 1 1
n
I n , μ ; {pi } =
2
Ti · Tk − γi ,
Γ (1 − ) i k i −sik  2  i

where the sums run over all coloured external legs. Note that
 2   2   
μ 1 μ 1 1 π2
= + iπ − Θ (sik ) + O()
−sik  2 |sik |  2  2

where the second term on the right hand side is related to the +i prescription
in the Feynman propagator. The flavour constant γi are γq = 32 CF and
γg = 12 β0 . T i denotes the colour charge associated with the emission of a
gluon from parton i,

T i = Tic |c ⇒
 c1, . . . , ci, . . . cn, c |T i | b1, . . . , bi, . . . , bn  = δc1 b1 . . . Tcci bi . . . δcn bn
a = F a and T a = t a are the representation matrices in the adjoint
where Tcb cb kl kl
and fundamental representations. The colour algebra is as follows:

Ti · Tk = Tk · Ti if i  k and T 2i = Ci (CA or CF ) .
3The I n operator is known also with finite quark masses and also for the two-loop
amplitude.
270 Chapter 19

The amplitude |M m  is gauge invariant, i.e., it is a singlet in colour

space,
@ 1  A
 
a 
c1, . . . , cm |M m  = c1, . . . , cm  e i ε a T j M m

j
>   =
 
= c1, . . . , cm |M m  + i εa c1, . . . , cm  Tja M m + O(εa2 ) .
j

The infinitesimal parameters εa are arbitrary, so we find

>   =
 
c1, . . . , cm  Tja M m = 0
j

n
for any a and consequently i=1 T i |M n  = 0. We denote the general colour
connected squared matrix element as
 2
M n | T i · T k |M n  ≡ M n (i,k)  .

Γ2 (1− )Γ(1+ )
In our example q = 0 , n = 2 , s12 = s > 0 , Γ(1−2 ) = 1
Γ(1− ) +
- 3.
O  so

    2  
(4π) μ 1 3
I , μ ; p1, p2
2
= −2CF − CF . (19.6.47)
Γ (1 − ) −s12  2 

The comparison to our explicit computations gives

  =  =  
 (1) fin 2  Nc2 − 1
M4 μ ; p1, p2 = CF e2 eq (−8) M4(0) remember: CF = .
2Nc
(19.6.48)
The cross section σ is computed from
>  = >  =
(0)  (0) (0)  (1)
M m |M m  = M m M m + 2Re M m M m + . . . . (19.6.49)

In our case the virtual corrections

B 
are C
; (0)  (1) < 

Re M4 M4 = Re 


>   = >  =
  
= M4(0) I 2 ()M4(0) − 8CF M4(0) M4(0) .
 Electron-positron annihilation into hadrons 271

19.6.1 Real corrections in d = 4 dimensions

Another contribution to the cross section at the same order is given by the
real corrections
 
α (0) 2 q
μ 2>  =  3
 (0)
|M m+1 | 2 = s 4παs(0) μ2 A m+1
(0) (0)
A m+1 + O αs .
4π

 =  2
 (0) ; (0)  (0) <  
 .
Two diagrams contribute to M5 : M5 M5 =  
We compute them first using the helicity formalism and considering the
μ μ μ μ μ
crossing symmetric channel 0 → p1 + p2 + p3 + p4 + p5 (the new label ‘3’
refers to the gluon). Both diagrams contribute to the only colour structure
tic13i2 and thus a single helicity amplitude has the form
 
A5 1h1 , 2h2 , 3h3 , 4h4 , 5h5 = e2 eq tic13i2 a5 (h1, h2, h3, h4, h5 ) ,

where the colour amplitude’s are given by


i 1 1 ; h1  μ    <
a5 ({hi }) = − √ 1 γ /p1 + /p3 γν 2−h2
2 s45 s13

1 ; h1  ν    <
+ 1 γ − /p2 − /p3 γ μ 2−h2
s23
;   <
× 5h5  γν 4−h4 μh3 (p3, k3 )
>   =
h3   h3
h3 3 γμ k 3 ;   <
= −i >  = 5h5  γν 4−h4
2 s k −h3 3h3
45 3
 
1 ;   < ;   <
× 1h1  γ μ 1h1 1h1  γν 2−h2
s13
;   < ;   <
+ 1h1  γ μ 3h1 3h1  γν 2−h2
1  ; h1  ν  h1 < ; h1  μ  −h2 <
− 1 γ 2 2 γ 2
s23
;   < ;   < 3
+ 1h1  γν 3h1 3h1  γ μ 2−h2 .

For helicity configuration h3 = h1 = −h2 we choose reference momentum

for the gluon polarization (see Eq. (18.9.12)) k 3 = p2 , whereas for h3 = h2 =
272 Chapter 19

−h1 we choose k3 = p1 . The only terms which do not vanish are

;   < ;   < ;  <;  <
3h3  γμ 2h3 1h1  γ μ 1h1 = 2 3h3 1−h3 1−h3 2h3 ,
; h   h < ; h  μ h < ;  <;  <
3 3  γμ 2 3 1 1  γ 3 1 = δh1 h3 2 3h3 1−h3 3−h3 2h3 ,
; h   h < ; h  μ  −h < ;  <;  < (19.6.50)
3 3  γμ 1 3 2 1  γ 2 2 = 2 3h3 2−h3 2−h3 1h3 ,
; h   h < ; h  μ  −h < ;  <;  <
3 3  γμ 1 3 3 1  γ 2 2 = δh2 h3 2 3h3 2−h3 3−h3 1h3 .

In the case h3 = h1 = −h2 the amplitudes are

i 2  
a5 (+, −, +, −, +) = − [31] 12 [15] 42 + [31] 32 [35] 42
s45 23 s13
24 2  
=i 12 [15] + 32 [35]
23 31 45 [45]
789:
24[45]

24 2
= 2i
23 31 45
(19.6.51)

and from charge conjugation on the lepton line

25 2
a5 (+, −, +, +, −) = −2i .
23 31 45

In the case h3 = h2 = −h1 the amplitudes are

 
i 2
a5 (+, −, −, +, −) = − −
s45 [13] s23
 
× 32 [21] [14] 52 + 32 [31] [14] 53
[14] 2  
=i [21] 52 + [31] 53 (19.6.52)
[23] [31] [45] 54
789:
[14]54

[14] 2
= −2i
[23] [31] [45]

and
[15]2
a5 (+, −, −, −, +) = 2i . (19.6.53)
[23] [31] [45]
 Electron-positron annihilation into hadrons 273

The amplitudes in the other four cases (for h1 = − and h2 = +) are obtained
from parity inversion

a5 (h1, h2, h3, h4, h5 ) = a5 (−h1, −h2, −h3, −h4, −h5 ) | i j ↔[ji]

and give a factor of 2 in the squared matrix element, which is then given by
  2 2 + s2 + s2 + s2
s14
(0)  24 15 25
A5  = e4 eq2 Nc CF 8 .
hel.& col.
s s s
23 13 45

There is a mismatch between these calculations of real and virtual correc-

tions:

• in the loop calculation the gluon was treated in d = 4 − 2 dimensions

(had d − 2 polarization states),

• in the real correction the gluon had 2 polarization states.

In order to be consistent, we have to perform the calculation of real correc-

tions in d = 4 − 2, for which the helicity formalism is inconvenient. (At
the end of this chapter we shall discuss, how we can yet use the d = 4 result,
obtained using the helicity formalism.)

19.6.2 Real corrections in arbitrary d dimensions

If we are not interested in the angular correlations between initial and final
states, we can simplify the calculation (the full calculation in d = 4 − 2 is
rather lengthy). We can parametrize the total cross section as
- . 1
σtot e+ e− −→ hadrons = L μν (k1, k2 ) 2 Hμν (Q) ,
s
where s = Q2 and Q μ = (k1 + k2 )μ (formerly (p4 + p5 )). Hμν (Q) describes
the hadronic (or partonic) final state that depends only on the photon mo-
mentum (after summing over all final states and integrating over the phase
space). It also includes the flux factor 1/(2s) of the cross section. The only
possible Lorentz structure it can be decomposed into is

Q μ Qν
H μν = −H1 (s) g μν + H2 (s) .
s
L μν describes the leptonic part and is equal to (we take into account the
negative sign due to the fermion loop on the left hand side)
274 Chapter 19

1 e2
−L μν = μ ν= Tr (γ μ k/ 1 γν k/ 2 )
4 4
- μ μ .
= e2 k 1 k 2ν + k1ν k 2 − k1 · k2 g μν .

The lepton tensor is gauge invariant

+ ,
qμ L μν = e2 k1 · k 2 k 2ν + k1 · k 2 k 1ν − k1 · k 2 (k1 + k2 )ν = 0 .

As a result, contracting Hμν with L μν , only H1 (s) contributes to σtot . From

gμν L μν = e2 (d − 2) k 1 · k2 = e2 (1 − ) s = 4πα s (1 − ) ,

it follows by dimensional analysis that

1 1
dim σtot = dim = dim 2 L μν ,
s s
so that dim H1 (s) = 0 and thus H1 (s) must be a constant. Hμν is also U(1)
(of QED) gauge invariant and so

g μν Hμν
H1 (s) = H2 (s) ≡ H(s) = − .
d−1
There is one cut diagram that contributes to the hadron tensor Hμν (s) at LO
in perturbative QCD. It is formally the same as for the lepton tensor, only
the fermion lines refer to quarks of momenta p1 and p2 . Thus,
∫
1  - μ μ .
−H μν = dφ2 (p1, p2 ; Q) e2 eq Nc 4 p1 pν2 + pν1 p2 − p1 · p2 g μν ,
2s q
(19.6.54)
where dφ2 is the two-particle phase-space measure (see Eq. (18.3.5) for
n = 2) and summation
√ extends over quark flavours whose masses are much
smaller than s. From Eq. (19.6.54) we find
 
e2 q eq2 Nc ∫
(0)
−H (s) = − 2s (2 − d) dφ2
2 (d − 1) s
  ∫
 d−2
= Nc e2 eq2 dφ2 (19.6.55)
q
d−1
 
α 
= Nc eq (1 + O ()) .
2
3 q
 Electron-positron annihilation into hadrons 275

Then 
1   4πα2 
σtot = lim L μν 2 − gμν H (0) (s) = Nc eq2 ,
 →0 s 3s q

which we knew already, cf. with Eq. (19.1.1).

At NLO accuracy we apply all possible cuts on the three possible dia-
grams:

The virtual correction emerges from cuts that lead to the loop×tree diagrams:

We already know that only the vertex correction is non-zero, which gives
(cf. Eqs. (19.6.47) and (19.6.48))
   
CF αs 1 4π μ2 2 3
HV (s) = − 2 − − 8 + π 2 + O () H (0) (s) .
2π Γ (1 − ) s  

The real correction comes from cuts that lead to the tree×tree diagrams:

We write the real correction symbolically as:

∫ 2
1 μν
−HR (s) = g dφ3 d1 (1, 2, 3) + d2 (1, 2, 3)
d−1
3
+ d1 (2, 1, 3) + d2 (2, 1, 3) ,
μν

where
∫ ∫
g μν dφ3 d1 (1, 2, 3)μν = dφ3 μ μ

 ∫ 
1 
= − 4παs μ e2 2
Nc CF dφ3 (p1, p2, p3 ; q)
eq2
q
2s
1 2      3
× 2 Tr γμ − /p2 γ μ /p1 + /p3 γα /p1 γ α /p1 + /p3
s13
276 Chapter 19

(the negative sign is due to the fermion loop). We can evaluate the trace
using
 
γμ − /p2 γ μ = (d − 2) /p2 , γα /p1 γ α = − (d − 2) /p1,

which leads to
+ ,
Tr [. . .] = − (d − 2)2 (p1 + p3 )α (p1 + p3 )β Tr /p2 γα /p1 γβ
- .
= − (d − 2)2 (p1 + p3 )α (p1 + p3 )β 4 p1α p2β + p1β p2α − p1 · p2 gαβ
= − (d − 2)2 [2s13 (s12 + s23 ) − 2s13 s12 ] = −8 (1 − )2 s13 s23
and thus
∫  
 1
dφ3 g μν d1 (1, 2, 3)μν = 4παs μ2 e2 eq2 Nc CF
q
2s
789:
the other diagrams have the same prefactor
∫
s23
× dφ3 (p1, p2, p3 ; q) 8 (1 − )2 .
s13
Similarly ∫ ∫
g μν dφ3 d2 (1, 2, 3)μν = dφ3 μ μ
∫
−1
∝ dφ3 (p1, p2, p3 ; q)
s13 s23
⎡ ⎤
⎢ ⎥
⎢       ⎥⎥
⎢
× Tr ⎢⎢ γμ − /p2 γν − /p2 − /p3 γ μ /p1 γν /p1 + /p3 ⎥⎥ .
⎢ 789: ⎥
⎢ ⎥
⎢(4−d)[ p/ γν ( p/ + p/ )]−2( p/ + p/ )γν p/ ⎥
⎣ 2 2 3 2 3 2 ⎦
ν
Using the equality γν γα γβ γ = 4 gαβ − (4 − d) γα γβ , we find for the trace
* 2  3
Tr [. . .] = (4 − d) 2 (s12 + s13 ) Tr /p2 /p1 + /p3
2    3 
− (4 − d) Tr /p2 /p2 + /p3 /p1 /p1 + /p3
* 2  3
− 2 2s12 Tr /p2 + /p3 /p1 + /p3
2   3 
− (4 − d) Tr /p2 + /p3 /p2 /p1 /p1 + /p3
 Electron-positron annihilation into hadrons 277
2
= 2 (4 − d) (s12 + s13 ) 2 (s12 + s23 ) + s23 s13
3
− (s12 + s13 ) (s12 + s23 ) + (s12 + s13 + s23 ) s12
2 3
− (4 − d)2 s23 s13 − s12 (s12 + s13 + s23 ) + (s12 + s23 ) (s12 + s13 )
− 4s12 2 (s12 + s13 + s23 )
= −8s12 s + 4 [2s12 s + 2s13 s23 ] − 4 2 2s13 s23
= −8 (1 − ) [s s12 −  s13 s23 ] ,
where s12 + s13 + s23 = s. Collecting the contributions from the four cut
diagrams, we finally have
∫ 2
1
−HR (s) = g μν d1 (1, 2, 3) + d1 (2, 1, 3)
d−1
3
+ d2 (1, 2, 3) + d2 (2, 1, 3)
μν
 ∫
1
= 4παs μ2 e2 eq2 Nc CF dφ3 (p1, p2, p3 ; q)
q
2s
   
d−2 s23 s13 s s12
×4 (1 − ) + +2 − 2
d−1 s13 s23 s13 s23
where the three-body phase space can be rewritten as
  d−3
dφ3 (p1, p2, p3 ; q) = (2π)3−2d 2−1−d q2 dd−2 Ω dd−3 Ω
d−4
× (y12 y13 y23 ) 2 dy12 dy13 dy23
× δ (1 − y12 − y13 − y23 ) [dy123 δ (1 − y123 )] ,
s
with scaled invariants yi j = si j and y123 = y12 + y13 + y23 = 1. dd−1 Ω is the
measure of the hyper-surface element in d dimensions
∫ d
2π 2
dd−1 Ω = Ωd =   .
Γ d2

Ωd is the surface of the hyper-sphere of unit√radius in d dimensions, for

instance Ω2 = 2π and Ω3 = 4π (Γ(3/2) = π/2). Since the integrand
depends only on yi j , we can use the two-body phase space
  d−4
2 dd−2 Ω
dφ2 (p1, p2 ; q) = q2 dy12 δ (1 − y12 )
(2π)d−2 2d−1
278 Chapter 19





 














 
Figure 19.1: Region of integration for the real correction

to write the three-body phase space as

  1− 1 dd−3 Ω
dφ3 (p1, p2, p3 ; q) = q2 dφ2 (q) (y12 y13 y23 )−
(4π) (2π)
2 d−3
789:
(4π)
Γ(1− )

×δ (1 − y12 − y13 − y23 ) dy12 dy13 dy23 . (19.6.56)

Then
 
CF αs 1 4π μ2
HR (s) = H (0) (s)
2π Γ (1 − ) s
∫
× dy12 dy13 dy23 δ (1 − y12 − y13 − y23 ) (y12 y13 y23 )− (19.6.57)
    
y23 y13 y12
× (1 − ) + +2 − .
y13 y23 y13 y23
The region of integration is shown in Fig. 19.1. The integrand is singular
at y13 = 0 or y23 = 0. The singularities are regularized by the (y13 y23 )−
factor in the phase space measure.
The phase-space integral in HR (s) (equation (19.6.57)) can be computed
easily (see exercise) and it is equal to
2 3 19
+ + − π 2 + O () . (19.6.58)
2  2
We find that the sum of the real and virtual corrections is finite,
HR (s) + HV (s) CF αs 3 αs
= + O () = + O () . (19.6.59)
(0)
H (s) 2π 2 π
 Electron-positron annihilation into hadrons 279

Thus we see that σtot , a totally inclusive quantity is infrared finite (i.e.
infrared safe) at one loop.
The natural question at this point is whether there exist other (less inclu-
sive) infrared-safe quantities?

Exercise 19.4
In the lectures we have seen that the 3-point Green’s function I (3) (p1, p2 )
can be expressed–up to a "fairly simple" rational polynomial–in terms of
the two-point function I (2) (p1 + p2 ). Why was this relation so simple?
In this exercise we will derive this relation I (3) (p1, p2 ) = P(d)I (2) (p1, p2 )
(P(d) is a rational polynomial in d) from so called integration-by-parts
(IBP) identities. We do this without computing any integrals explicitly. In
dimensional regularization the following identities hold (generalization of
Gauss’ law):
∫
∂ μ
dd k
k I (k, p, m, . . .) = 0 ,
(2π) ∂k μ d
∫ d
d k ∂ μ
p I (k, p, m, . . .) = 0 .
(2π)d ∂k μ

By computing the derivatives and replacing the new scalar

 products in the

numerator by inverse propagators, such as k · p = 1/2 k 2 + p2 − (k − p)2 ,
we obtain a (homogeneous, linear) system of equations by means of which
more complicated integrals can be expressed in terms of simpler ones4

• Evaluate the equations

∫
∂ μ
dd k 1
μ
k = 0,
(2π) ∂k d
k 2 (k − p1 2 (k − p12 )2
)
∫ d
d k ∂ μ 1
d ∂k μ
p =0
(2π) k (k − p1 )2 (k − p12 )2
2

μ
(p12 = p1 + p2, p21 = p22 = 0, gμ = d) in order to reduce the three-
point function to two-point functions; the propagators of the two-point
functions might acquire higher exponents.
4This method of computing integrals has the advantage that it is easily automated.
The systems of equations can however become large.
280 Chapter 19

• Find two equations that reduce the (two) two-point functions with one
squared propagator each down to the so-called "master integral"
∫ d
(2) d k 1
I (p1 + p2 ) = .
(2π) k (k − p12 )2
d 2

• Check the relation using the integrals evaluated explicitly in the main
text.
Exercise 19.5
Show that the d-dimensional three-particle phase space for q → p1 + p2 + p3
can be expressed in terms of the Lorentz-invariants si j = (pi + p j )2 as

dd−1 p1 dd−1 p2 dd−1 p3

dφ3 = (2π)d δ d (q − p1 − p2 − p3 )
2E1 (2π)d−1 2E2 (2π)d−1 2E3 (2π)d−1
= (2π)3−2d 2−1−d (q2 )
2−d
2 dd−2 Ωdd−3 Ω
d−4
 
× (s12 s13 s23 ) 2 ds12 ds13 ds23 δ q2 − s12 − s13 − s23 .

Hints:
• The d-dimensional volume measure in spherical coordinates is given
by
dd+1 p = E dE d d pE = E d E E d−1 dd Ω
dd Ω = (sin θ 1 )d−1 dθ 1 dd−1 Ω .

• Show that
1 s12 s13 s23
sin2 θ 1 = ,
4 q2 E12 E22
where θ 1 is the angle between p1 and p2 .

Exercise 19.6
s
Let yi j = qi2j . Using the previous exercise, compute the real correction to
the process e+ e− → q q̄:
 
CF αS 1 4π μ2
HR (s) = H (0) (s)
2π Γ(1 − ) s
∫
× dy12 dy13 dy23 δ (1 − y12 − y13 − y23 ) (y12 y13 y23 )−
    
y23 y13 y12
× (1 − ) + +2 − .
y13 y23 y13 y23
 Electron-positron annihilation into hadrons 281

Hint: Transform the triangular integration region into the unit square
and evaluate the B (Euler β) functions.

19.7 Origin of the singular behaviour

A natural question is if we can construct the approximate cross section uni-
versally, i.e. independently of the process and observable. Our presentation
above suggests the affirmative answer. To understand how, we have to study
the origin of the singular behaviour in the squared matrix element (SME).
This singularity arises from propagator factors that diverge
ps , μ
1 1 1 1
Mm ∝ = =
pi (pi + ps ) 2 2 pi · p s 2Ei Es (1 − cos θ) Ei Es θ 2
In the collinear limit of two external momenta, characterized by θ → 0,
the matrix element behaves as M m+1 M m /θ + less singular terms (a
factor of θ appears in the numerator factors). In the soft limit Es → 0 the
leading singular behaviour is M m+1 M m /Es + less singular terms. The
phase space of the radiated gluon in these limits is approximated as

d3 p s 1 1
= Es dEs dcos θ dφ Es dEs dθ 2 dφ .
2Es 2 4
Therefore in the cross section we find logarithmic singularities both in the
dθ 2
soft and the collinear limits: dE s
Es or θ 2 . These are called the infrared
singular limits. In dimensional regularization the logarithmic singularities
appear as poles: ∫
−1− 1
dy13 y13 =− .

Thus, the singular behaviour arises at kinematically degenerate phase space
configurations, which at the NLO accuracy means that one cannot distinguish
the following configurations: (i) a single hard parton, (ii) the single parton
splitting into two nearly collinear partons, (iii) the single parton emitting a
soft gluon (on-shell gluon with very small energy). Then an answer to the
question posed at the end of the previous section is given by the Kinoshita-
Lee-Nauenberg (KLN) theorem:
In massless, renormalized field theory in four dimensions, transition rates
282 Chapter 19

are infrared safe if summation over kinematically degenerate initial and final
states is carried out.
For the e+ e− → hadrons process, the initial state is free of infrared
singularities that can emerge only off coloured particles (partons). Typical
infrared-safe quantities are (i) event shape variables and (ii) jet cross sections.

19.7.1 Event shapes

Thrust, thrust major/minor, C- and D-parameters, oblateness, sphericity,
aplanarity, jet masses, jet-broadening, energy-energy correlation, differential
jet rates are examples of event shape variables. The value of an event shape
does not change if a final-state particle further splits into two collinear
particles, or emits a soft gluon, hence it is (qualitatively) infrared safe. As
as example we consider the thrust T, which is defined by
m
i=1 | pi · n|
T = max  , (19.7.60)
i=1 | pi |
m
n

where n is a three-vector (the direction of the thrust axis) such that T is

maximal. The particle three-momenta pi are defined in the e+ e− centre-of-
mass frame. T is infrared safe because neither pi → 0, nor replacing pi
with zpi + (1 − z)pi change the value of T. At LO accuracy it is possible to
perform the phase space integrations and we find the differential distribution
in T:
( - .   )
1 dσ αs 2 3T 2 − 3T + 2 2T − 1 2−T
= CF ln − 3(3T − 2) .
σ dT 2π T(1 − T) 1−T 1−T
(19.7.61)
We see that the perturbative prediction for the thrust distribution is singu-
lar at T = 1. In addition to the linear divergence in 1 − T there is logarithmic
divergence, as well. The latter is characteristic to event shape distribu-
tions. In perturbation theory at nth order logarithms of 1 − T in the form
αsn lnm (1/(1−T)) , m ≤ 2n−1 appear in the differential distributions of event
shapes, like (19.7.61). These spoil the convergence of the perturbation series
and call for resummation if we want to make reliable predictions near the
edge of the phase space, for large values of T, where most of the data occur
and the best experimental statistics are available. Resummations of leading
(LL, m = 2n − 1) and next-to-leading (NLL, m = 2n − 2), next-to-next-to-
leading logarithms5 are available for many event shape variables (for some,
5The nomenclature in the literature is different as the resummation is performed
for a different formula than the differential distribution.
 Electron-positron annihilation into hadrons 283

like thrust even NNNLL is known), but the discussion of this technique is
beyond the scope of these lectures.

Exercise 19.7
The thrust variable T is used to characterize final states in e+ e− →hadrons.
For m particles in the final state it is defined by Eq. (19.7.60). The momenta
are taken in the centre of mass system.

• Verify that T is infrared and collinear safe.

• What is the range of values that T can take if
(a) there are only two particles in the final state, or
(b) there are three particles in the final state, or
(c) m → ∞ and all pi are distributed spherically.

Exercise 19.8
Compute the thrust distribution at LO accuracy. Hints:
• Write the squared matrix elements for three
√ partons in the final state
using scaled energy variables xi = 2Ei / s = 1 − y jk (i, j, k = 1, 2, 3,
with i  j  k).
• Convince yourself that the value of thrust is T = max{x1, x2, x3 }.
• Find the region of integration in terms of x1 and x2 .
• Find the region where T = x1 , x2 and x3 and integrate the fully
differential cross section with the δ(T − T(x1, x2 )) function.

19.7.2 Jet algorithms

Jets are sprays of energetic, on-shell, nearly collinear hadrons. The number
of jets does not change if a final-state particle further splits into two collinear
particles, or emits a soft gluon, hence it is again qualitatively infrared safe.
In order to quantify the jet-like structure of the final states jet algorithms
have been invented. These have a long history with rather slow convergence.
The reason is that the experimental and theoretical requirements posed to a
jet algorithm are rather different. Experimentally we need cones that include
almost all hadron tracks at cheap computational price. Theoretically the
284 Chapter 19

important requirements are infrared safety, so that perturbation theory can

be employed to make predictions and resummability, so that we can make
predictions in those region of the phase space where most of the data appear.6
For many years experimenters at hadron colliders preferred cone jet
algorithms (according to the ‘Snowmass accord’) [Berger, 1992]. These start
from a cone seed (centre of the cone) in pseudorapidity (η) and azimuthal-
angle (φ) plane: (ηc, φc ). We define a distance of a hadron track i from
the seed by dic = (ηi − η c )2 + (φi − φc )2 . A track belongs to the cone
if dic < R, with a predefined value for R (usually between 0.5 and 0.8).
However, it turned out that (i) this is an infrared unsafe definition and (ii)
there is a problem how to treat overlapping cones, so the cone jet definition
has been abandoned.
Theoreticians prefer iterative jet algorithms, consisting of the following
steps. (i) First we define a distance di j between two momenta (of partons or
hadron tracks) and a rule to combine two momenta, pi and p j into p(i j) . (ii)
Then we select a value for resolution, dcut and consider all pairs of momenta.
If the minimum of {di j } is smaller than dcut , then we combine the momenta pi
and p j and start the algorithm again. If the minimum is larger than dcut , then
the remaining momenta (after the combinations) are considered the momenta
of the jets, and the algorithm stops. The drawback of this algorithm is that it
becomes very expensive computationally for many particles in the final state.
This is not an issue in perturbative QCD computations because according
to our basic assumption there are only few partons, but a major problem for
the final states in the detectors where hundreds of hadrons may appear in a
single event.
At the experiments of LEP theory won and the Durham (or kT )-algorithm
was used. It was invented so that resummation of large logarithmic contri-
butions could be achieved [Catani et al., 1991]. The distance measure is
min(E 2,E 2 )
di j = 2 s
i j
Rir , where Ri j = 1 − cos θ i j and the recombination scheme
μ μ μ
is simple addition of the four momenta, p(i j) = pi + p j . The resolution
parameter ycut = dcut /s can take values in (0, 1]. The perturbative QCD pre-
diction contains logarithmically enhanced terms of the form αsn lnm (1/ycut ),
at any order. Such terms have to be resummed if we want to use small value
of ycut where we find the bulk of the data (see Figure 17.58). Predictions are
available with leading (m = 2n) and next-to-leading (2n − 1) logarithms (LL
and NLL) summed up to all orders [Catani et al., 1991].
Figure 17.58 shows the fixed order LO and NLO predictions, as well as

6Recently jet identification is considered pattern recognition and deep-learning

techniques are being developed to identify jets, which is however beyond our scope.
 Electron-positron annihilation into hadrons 285

predictions where NLO and NLL are matched. The curve at NLO accuracy
gives a good description of the data measured by the ALEPH collaboration
[Barate et al., 1998], but only for ycut > 0.01. As αs ln2 (100) = 2.5, for
smaller values of ycut resummation is indispensable. However, the resummed
prediction is not expected to give a good description at large ycut because
in the resummation formula only the collinear approximation of the matrix
element is used. Matching the two predictions gives a remarkably good
description of the data over the whole phase space.
At hadron colliders the k T -algorithm needs modifications. First, instead
of energy, the boost-invariant measure of hardness, transverse momentum
R2
is used to define the distance between tracks, di j = min(p2T i, p2T j ) Ri2j where
Ri2j = (yi − y j )2 + (φi − φ j )2 (distance in rapidity–azimuthal-angle plane),
R is a small positive real number, and we need a distance from the beam
diB = p2T i , too. Also, the algorithm needs modification because either
di j or diB can be the smallest distance. If a di j is the smallest value,
then i and j are merged, while if the smallest is a diB , then momentum pi
becomes a jet momentum and is removed from the tracks to be clustered. We
then call jet candidates with transverse momentum larger than a predefined
value, pT i > ER resolved jets. The merging rule changes as well. In the
usual merging we add transverse momenta, pT (i j) = pT i + pT j and we add
rapidities y and azimuthal angles φ weighted, y(i j) = (wi yi + w j y j )/(wi + w j )
and φ(i j) = (wi φi + w j φ j )/(wi + w j ). The weight can be wi = pT i , p2T i , ET i ,
or ET2 i . Such a merging is boost invariant along the direction of the beam.
The parameter R plays a similar role as dcut in electron-positron annihilation
or the cone radius R in the cone algorithms: the smaller R, the narrower the
jet. Popular values in the experimental analyses are R = 0.4 − 0.7.
The iterative k T -algorithm is infrared safe and resummation of large log-
arithmic contributions of the form αsn ln2n and αsn ln2n−1 is possible, which
is a clear advantage from the theoretical point of view. The logarithms are
those of 1/R and/or Q/ER , Q being the hard process scale. By taking ER
sufficiently large in hadron-hadron collisions, we avoid such leading con-
tributions from initial-state showering and the underlying event (collisions
not belonging to the hard parton-parton scattering), so these terms are deter-
mined by the time-like showering of final-state partons (when the virtuality
of the parent parton is always positive). Particles within angular separation
R tend to combine and particles separated by larger distance than R from all
other particles become jets. The algorithm assigns a clustering sequence to
particles within jets, so we can look at jet substructure.
Nevertheless, at the TEVATRON experiments the kT -algorithm did not
286 Chapter 19

become a standard for several reasons. The jets have irregular, often weird
shapes as seen on Fig. 17.59 (left) because soft particles tend to cluster
first (even arbitrary soft particles can form jets). As a result there is a
non-linear dependence on soft particles, energy calibration and estimating
acceptance corrections are more difficult. The underlying event correction
depends on the area of the jet (in η − φ plane). It was also very expen-
sive computationally, so experimenters had a clear preference of cone algo-
rithms. The breakthrough occurred with Refs. [Cacciari and Salam, 2006,
Cacciari et al., 2008] where variants of the kT -algorithm and an improved,
fast implementation was introduced. The distance formula was modified to
R2
di j = min(p2nT i, pT j ) R 2 (n = −1, 0, 1). Infrared safety is independent of n,
2n ij

as well as NLL resummation of large logarithms. It was found that with

n = −1 (called anti-k T -algorithm) particles close in angle cluster first, which
results in regular cone-like shapes as seen on Fig. 17.59 (right) without using
stable cones. As a result it became the standard jet algorithm at the LHC
experiments. Yet, one should keep in mind that there is no ‘perfect’ jet
algorithm. For instance, the anti-kT one does not provide useful information
on jet substructure. It is important to remember that in perturbative QCD
theoretical prediction can be made only with infrared-safe jet functions, but
among those the goal of the study may help decide which algorithms to use.
Analytically, measures of “jettyness”
/ μ 0 n or event shapes are represented by
“jet functions”, J ({pi }), where pi i=1 is the set of momenta of final state
particles. Thrust is an example of the jet function Jm . In perturbation theory
the requirement of infrared safety can be formulated as conditions for the jet
functions. For instance, at NLO accuracy, the jet function Jm of an m-jet
measure (m is the number of jets) must obey:
(m)
Jm (p1 . . . pm ) → 0 if any pi · p j → 0
(m+1)
Jm (p1 . . . pm+1 ) → 0 if pi · p j and pk · pl → 0 (i  k)
(19.7.62)
(for an m-jet measure we need m well separated momenta). Furthermore,
(m+1) - . (m)
Jm p1, . . . , p j = λ q, . . . , pm+1 → Jm (p1, . . . , pm+1 ) (19.7.63)
if λ → 0 , and q μ is a fixed vector ,
(m+1) - . (m)
Jm p1, . . . , pi, . . . , p j , . . . , pm+1 → Jm (p1, . . . , p, . . . , pm+1 )
(19.7.64)
μ μ
if pi → zpμ, p j → (1 − z) pμ . In the first case in the new jet function
(m) μ
Jm (p1, . . . , pm+1 ) we have omitted the soft momentum p j , while in the
 Electron-positron annihilation into hadrons 287

μ μ
second case the two collinear momenta pi and p j are replaced by a single
momentum pμ denoting the collinear direction. The cross section is the phase
space integral of the squared matrix element weighted by the jet function,
for instance ∫
1 (n)
σ= dφn |M n | 2 Jm (p1, . . . pn ) .
2s
In our case of electron-positron annihilation into partons, for an arbitrary jet
function J (n) we have the leading-order term
  ∫

(0) 2 d−2
H (s, J) = Nc e 2
eq dφ2 (p1, p2 ; q) J (2) (p1, p2 ) ,
q
d − 1

the virtual correction

   
CF αs 1 4π μ2 2 3
HV (s, J) = − − − 8 + π 2 + O ()
2π Γ (1 − ) s 2 
 
× H (0) s, J (2) ,

and the real correction

 
CF αs 1 4π μ2
HR (s, J) = H (0) (s)
2π Γ (1 − ) s
∫
× dy12 dy13 dy23 δ (1 − y12 − y13 − y23 ) (y12 y13 y23 )−
    
y23 y13 y12
× (1 − ) + +2 −  J (3) (p1, p2, p3 ) .
y13 y23 y13 y23

For the total cross section J (n) = 1 and cancellation of infrared poles takes
place as in Eq. (19.6.59). For thrust T (2) (p1, p2 ) = 1 (by momentum conser-
vation the two outgoing quarks are back-to-back, this direction defines the
thrust axis event-by event), and

T (3) (p1, p2, p3 ) ≤ 1 .

In general, we cannot integrate HR (s, J) analytically because of the presence

of the jet function J (3) . Therefore, we do not know how to combine HV and
HR . To find the solution, we have to understand the structure of infrared
singularities in |M n | 2 .
288 Chapter 19

Exercise 19.9
Generalize the requirements for the jet function for a computation at NLO
accuracy as given in Eqs. (19.7.62)–(19.7.64) to NNLO accuracy.

19.8 Factorization of |Mn | 2 in the soft limit

μ
The soft limit is defined by ps = λ q μ , with λ ∈ R+ (positive real number)
and λ → 0 for q μ fixed. In this limit the emission of the soft gluon from
(internal) propagators is infrared finite. If we consider the emission of a soft
gluon off an external quark (assumed to be massless here) we find
/p + /ps
ps , μ

Mm ∝ M mTis gs ū(pi, si )γ μ i
pi sis
μ
ps → 0 1 pi
Tis gs ū (pi, si ) γ μ /pi = Tis gs ū (pi, si ) ,
sis pi · ps
where Tis is the colour charge operator of the soft gluon s, gs is the strong
coupling, ū(pi, si ) is the Dirac adjoint spinor for the outgoing quark labelled
with i and sis = (pi + ps )2 = 2pi · ps . In taking the limit, we used the
μ
anti-commutation relation (18.2.2) to write γ μ /pi = − /pi γ μ + 2pi and the
μ
p
Dirac equation of the massless bi-spinor, ū(pi ) /pi = 0. The factor pi ·i ps is
the “square root” of the eikonal factor Sik (s) = si2ss si kk s .
The emission of a soft gluon off an external gluon (in light-cone gauge)
is given by
μ, s
ps 
Mm
λ, b
ν, a
∝ M m ε μ (ps, n) 1
si s d λλ (pi + ps, n)
pi

ν
× Γνμλ (−pi , −ps, pi + ps ) ε (pi , n)
asb

where in the three-gluon vertex in the limit becomes

Vνμλ (−pi, −ps, pi + ps ) =

= − (pi + 2ps )ν gμλ + (2pi + ps )μ gνλ − (pi − ps )λ gμν
+ ,
= 2piμ gλν + − (pi + ps )λ gμν − piν gμλ
+ ,
+ psμ gνλ + 2ps λ gμν − 2psν gμλ
ps → 0 + ,
2piμ gνλ − (pi + ps )λ gμν + piν gμλ .
 Electron-positron annihilation into hadrons 289


We use d λλ (pi + ps, n) (pi + ps )λ = 0 and εν (pi, n) pi ν = 0, thus

1 λλ ν ps → 0
d (pi + ps, n) Γνμλ
asb
 (−pi , −ps, pi + ps ) ε (pi , n)
sis
piμ  λλ  piμ
Tbs gs d (pi, n) gλν εν (pi, n) = Tbs gs (−ε λ (pi, n)).
pi · ps pi · p s

These two results can be unified and formalized by

Ŝs  cs | M m+1 (ps, . . .) = gs  μ (ps ) J μ (s) |M m (. . .)

where cs is the colour index of the soft gluon s, Ŝs is an operator which takes
the soft limit and keeps the leading singular term in 1/λ, and J μ (s) is given
by
m
pk μ
J μ (s) = T ks .
k=1
pk · p s

The soft gluon can be emitted from any of the external legs, therefore the sum
in the previous formula runs over all external partons. A soft quark leads
to an integrable singularity because the fermion propagator is less singular
than that of the gluon. Colour conservation implies that the current J μ (s) is
conserved,
m
μ
ps J μ (s) |M m  = T ks |M m  = + + = 0.
k=1
>  =
(0)  (0)
Then the soft limit of M m  M m is as follows:

Ŝs |M m+1 (ps, . . .)| 2 = 4παs μ2

m  μ
m
pi pνk
× μ (s) ν∗ (s) M m | T i · T k |M m 
p · p s pk · p s
i=1 k=1 789: i
dμν (ps ,n)
m  
1  (0) 2
= −8παs μ2 Sik (s) M m (i,k) 
+ gauge terms
i,k=1
2

∝ + + ... .
(19.8.65)
The gauge terms give zero contribution on on-shell matrix elements due to
gauge invariance.
290 Chapter 19

19.9 Factorization in the collinear limit

The collinear limit of momenta pi and pr is defined by the Sudakov parametriza-
tion:

μ μ
2
k iT nμ μ μ
2
krT nμ
pi = zi pμ + kiT − , pr = zr pμ + krT −
zi 2 p · n zr 2 p · n
μ μ
where k iT + krT = 0 and zi + zr = 1. The momentum pμ is the collinear
direction and

p2 = p2i = pr2 = n2 = 0 , k iT · p = krT · n = 0 ,

μ μ k2
In the collinear limit k iT, krT → 0 and sir = − zirzTr . We now state the
following theorem
In a physical gauge, the leading collinear singularities are due to the
collinear splitting of an external parton.
fi
This means that we need to compute M fir M in the collinear limit.
fr
There are three cases:

fir → fi + fr
q → q + g
g → q + q̄
g → g + g
We compute explicitly the first case and leave the second and third ones as
exercise. For the case of a quark splitting into a quark and a gluon we have
pr
/pi + /pr μ ν /p + /pr
= CF gs2 μ2 γ /pi γ dμν (pr , n) i
pi sir sir
 
/ /r
p + p / / n/ + n/ /pi /pr /pi + /pr
p p
= CF 4παs μ2 i −γ μ /pi γμ + r i .
sir pr · n sir
Using (19.9.66)

−γ μ /pi γμ = (d − 2) /pi , /pi /pi = pi 1 , /pi /pr /pi = sir /pi − pi /pr = sir /pi ,
2 2

we find
   
/pi + /pr −γ μ /pi γμ /pi + /pr = (d − 2) sir /pr ,
/pr /pi n/ = − /pi /pr n/ + sir n/ = /pi n/ /pr − 2 /pi pr ·n + sir n/ =
= −/n /pi /pr + 2pi ·n /pr − 2 /pi pr ·n + sir n/ .
 Electron-positron annihilation into hadrons 291

Then
   
/pi + /pr /pr /pi n/ + n/ /pi /pr /pi + /pr =
   
= 2 /pi + /pr pi ·n /pr − pr ·n /pi + pi · pr n/ /pi + /pr
2
= 2 pi ·n sir /pi − pr ·n sir /pr
   3
+pi · pr 2 (pi + pr )·n /pi + /pr − (pi + pr )2 n/
+ ,
= sir 4pi ·n /pi + 2pi ·n /pr + 2pr ·n /pi − sir n/ .
Substituting these results and then the Sudakov parametrization of the mo-
menta into Eq. (19.9.66) we obtain
( )
z 2
p i ! pr 1 i
CF 4παs μ2 2 (1 − ) zr + 4 + 2zi + 2zi + O (k T ) /p
sir zr
 
1 zi
= CF 8παs μ2 2 + (1 − ) zr /p
sir zr
789:
1+z 2
i
1−z i − (1−zi )

Similarly to the soft case we can define an operator Ĉir which takes the
collinear limit and keeps the leading singular (O(1/kT2 )) terms:
  1 > (0)   =
 (0) 2  (0)  (0)
Ĉir M m+1  = 8παs μ2 M m (p, . . .) P̂qg (zi, zr , k T ; ) M m (p, . . .) .
sir
(19.9.67)
The kernel P̂qg , called Altarelli-Parisi splitting kernel for the process q →
q + g. It is diagonal in the spin-state of the parent (splitting) parton:
 
zi
 s| P̂qg |s  = CF 2 + (1 − ) zr δss . (19.9.68)
zr
Similar calculations give the splitting kernels for the gluon splitting pro-
cesses, which display azimuthal correlations of the parent parton
( μ
)
(0) μν
k T kTν
μ| P̂q q̄ (zi, zr , kT ; ) |ν = TR −g + 4zi zr 2 , (19.9.69)
kT
(0)
μ| P̂gg (zi, zr , kT ; ) |ν =
(   μ
)
μν zi zr k T k Tν
= 2CA −g + − 2 (1 − ) zi zr 2 . (19.9.70)
zr zi kT
292 Chapter 19

The soft and collinear limits overlap when the soft gluon is also collinear to
its parent parton:

 2  2z j  2
 (0)  (0) 
Ĉ jr Ŝr M m+1 (pr , . . .) = −8παs μ2 M m(j,k) (. . .)
s
kj jr r
z
2 z j  (0) 2
= 8παs μ2 T 2j M m  . (19.9.71)
s jr zr

The notation for the splitting kernels in these lectures is different from the
usual notation in the literature. Usually, P̂i(0) j (z, k T ; ) denotes the splitting
kernel for the process fi (p) → f j (zp) + fk ((1 − z)p), which does not lead
to confusion for 1 → 2 splittings because the momentum fraction of parton
j determines that of parton k as their sum has to be one, z j + zk = 1.
For splittings involving more partons, it is more appropriate to introduce
as many momentum  fractions zi as the number of offspring partons, with
the constraint i zi = 1, and use the flavour indices to denote the offspring
partons in the order of the momentum fractions in the argument. For 1 → 2
(0)
splittings this means the use of P̂ir (zi, zr , kT ; ) for the splitting process
fk (p) → fi (zi p) + fr (zr p). The flavour of the parent parton fk is determined
uniquely by the flavour summation rules, q + g = q, q + q̄ = g + g = g.
These flavour summation rules are unique also for 1 → 3 splittings.

Exercise 19.10
Compute the Altarelli-Parisi-splitting kernel P̂qg (z) for the process q → qg
from the collinear limit of the matrix element for the process e+ e− → q q̄g:

   
M3 (e+ e− → q q̄g)2 = 8παs μ2 M2 (e+ e− → q q̄)2
    
1 y23 y13 y12
× (1 − ) + +2 − .
s y13 y23 y13 y23

Exercise 19.11
The Altarelli-Parisi splitting kernel P̂q q̄ (z) for the process g → qi q̄r is
 Electron-positron annihilation into hadrons 293

defined by the following collinear limit:

 (0) p i ! pr
(0)
M n+1 (pi, pr , . . .) M n+1 (pi, pr , . . .)

1  
8παs μ2 M n(0) (p, . . .) P̂q(0)q̄ (z, k T )M n(0) (p, . . .)
sir
   
1  
= 8παs μ2 M n(0) (p, . . .)  μμ P̂q(0)q̄ (z, kT ) νν M n(0) (p, . . .)
sir
 
 
Compute μ P̂q(0)q̄ (z, kT ) ν in leading order in k T .

Exercise 19.12
Derive the flavour summation rules for 1 → 3 splittings.

Exercise 19.13
Compute the soft limit of Eq. (19.9.67) and the collinear limit of Eq. (19.8.65).

19.9.1 Regularization of real corrections by subtraction

The cross section at NLO accuracy is a sum of two terms, the LO prediction
and the corrections at one order higher in the strong coupling,

σNLO = σ LO + σ NLO ,

where σ LO is the integral of the fully differential Born cross section over the
available phase space defined by the jet function, while σ NLO is the sum of
the real and virtual corrections:
∫
σ LO = dσ B Jm ({p}m ) ,
m
∫ ∫
σ NLO = dσ R Jm ({p}m+1 ) + dσV Jm ({p}m ) .
m+1 m
Both contributions to σ NLO are divergent in four dimensions, but their sum
is finite for infrared-safe jet functions.
The factorization of the squared matrix elements in the soft and collinear
limits allows for a method independent of the processes and observables
to regularize the real corrections in their singular limits. The essence of
294 Chapter 19

the method is to devise an approximate cross section dσ A that matches

the singular behaviour of the real cross section dσ R in all kinematically
degenerate regions of the phase space when one parton becomes soft or two
partons become collinear. Then we subtract this approximate cross section
from the real one and the difference can be integrated in four dimensions,
similarly to
∫ 1  
f (x) f (0)
dx 1− − 1− < ∞ (19.9.72)
0 x x
for arbitrary  (in particular when  = 0) provided the limit limx→0 f (x) < ∞.
Next, we integrate dσ A over the phase space of the unresolved parton and
we add it to dσV . In the example of Eq. (19.9.72) this step corresponds to
computing the integral
∫ 1
f (0) 1
dx 1− = f (0) .
0 x 
The integrated subtraction term cancels the explicit poles in the virtual
correction and the sum can also be integrated in four dimensions.
The key for this procedure is a proper mapping of the (m+1)-parton phase
space to the m-parton one which respects the limits, thus the approximate
cross section is defined with the m-parton jet function. After such a mapping
the phase space of the (m + 1) partons in the final state can be written in the
form
φm+1 (({p}m+1 )) → φm ({ p̃}m )φ1 (19.9.73)
where φ1 is related to the phase space measure of the unresolved parton. As
a result we integrate dσ A with the jet function Jm of m final state momenta.
This way we can rewrite the NLO correction as a sum of two finite terms,
∫
+ R ,
σ NLO = dσ Jm ({p}m+1 ) − dσ A Jm ({ p̃}m )  =0
m+1
∫  ∫  (19.9.74)
+ dσ + dσ
V A
Jm ({p}m ) .
m 1  =0

The definition of the approximate cross section is not unique and the best
choice may depend on further requirements, which we do not discuss here.
We also skip the precise definition of the momenta p̃μ which are obtained
by mapping the (m + 1)-particle phase space onto an m-particle phase space
times a one-particle phase space. A widely used general subtraction scheme
that can be used also for processes including massive partons with smooth
massless limits is presented in Ref. [Catani et al., 2002] where these defini-
tions are given explicitly. This method uses the factorization of the squared
 Electron-positron annihilation into hadrons 295

matrix element in the soft and collinear limits. The challenge posed by
the overlapping singularity in the soft-collinear limit is solved by a smooth
interpolation between these singular regions.
The factorization properties of Eqs. (19.8.65) and (19.9.67) play other
very important roles in pQCD. The numerical implementation of the squared
matrix element is a process prone to errors. Testing the factorization in the
kinematically degenerate phase space regions serves a good check of the
implementation. The computation is even more difficult for the virtual
corrections. Similar factorization holds for those, which more recently has
been utilized to find the virtual corrections (“collinear bootstrap”). The
factorized form of the squared matrix element can be used in resumming
logarithmically enhanced terms at all orders, or in devising a parton shower
algorithm for modelling events (see Sect. 21.3). The splitting kernels that
appear in the collinear factorization have a role in the evolution equations of
the parton distribution functions (see Sect. 20.7).
The state of the art in making precision predictions assaults on the one
hand the full automation of computations at NLO, and on the other the realm
of next-to-next-to-leading order (NNLO) corrections. The automation of
computing jet cross sections at NLO accuracy has been accomplished and
several programs are available with the aim to facilitate automated solutions
for computing jet cross sections at NLO accuracy:

• aMC@NLO (http://amcatnlo.web.cern.ch),
• BlackHat/Sherpa (https://blackhat.hepforge.org),
• FeynArts/FormCalc/LoopTools (http://www.feynarts.de),
• GoSam (https://gosam.hepforge.org),
• HELAC-NLO (http://helac-phegas.web.cern.ch),
• MadGolem
(http://www.thphys.uni-heidelberg.de/˜ lopez/madgolem-corner.html).

There is also the MadGraph5_aMC@NLO computer program that provides

full automation of computing QCD jet cross sections at NLO accuracy
(https://launchpad.net/mg5amcnlo).
In the NNLO case the infrared singularity structure is much more in-
volved than in the case of NLO computations due to complicated overlapping
single and double unresolved configurations. Several subtraction methods
have been proposed for the regularization of the infrared divergences. There
is intense research to find a general one that can be automated. To pro-
vide an impression about the importance of NNLO corrections, we present
QCD predictions at various accuracies (LO, NLO and NNLO) for the thrust
296 Chapter 19

distribution computed with αs = 0.118 and at a centre-of-mass energy of

√
s = 91.2 GeV in Fig. 17.60. For all calculations the uncertainty band re-
flects the uncertainty due
√ to the variation of the renormalization scale around
the default scale μ = s by factors of two in both directions. We find that
even the prediction at NNLO accuracy differs significantly from the (hadron-
level) data measured by the ALEPH collaboration at LEP. In Figure 17.60
(right) we show the effect of NNNLL resummation matched to the NNLO
prediction, which gives nice agreement between data and theory except for
small values of τ = 1 − T. The difference is most likely due to hadronization
corrections that we do not discuss in this book.
Chapter 20

Deeply inelastic
lepton-proton scattering

MOTTO:
If you want to learn about nature, to appre-
ciate nature, it is necessary to understand the
language that she speaks in.

(Richard P. Feynman)

20.1 Kinematics
Perturbative QCD stems from the parton model that was developed to un-
derstand deeply inelastic lepton-hadron scattering (DIS). The purpose of
those experiments was to study the structure of the proton by measuring
the kinematics of the scattered lepton. In Fig. 17.61 (left) we show a real
event in the H1 experiment at the HERA collider. The value of Q2 , which
is the modulus squared of the momentum transfer between the lepton and
the proton is 21475 GeV2 >> 1 GeV2 , signifying that the scattering is well
in the deeply inelastic region. The parton model interpretation of the event
is shown in Fig. 17.61 (right): the lepton is scattered by an angle θ due to
the exchange of a virtual photon with one of the constituents of the proton
297
298 Chapter 20

(a parton).
The measurement is inclusive from the point of view of hadrons (X
means any number of hadrons that are not observed separately), thus the
process can be described in pQCD.
The DIS kinematics is described by the following variables (using the
notation of Fig. 17.61 (right)):

centre-of-mass energy2 = s = (P + k)2 ,

momentum transfer = q μ = k μ − k μ ,
|momentum transfer| 2 = Q2 = −q2 = 2M E x y ,
scaling variable = x = Q2 /(2P · q) ,
energy loss = ν = (P · q)/M = E − E  ,
relative energy loss = y = (P · q)/(P · k) = 1 − E /E ,
1−x 2
recoil mass2 = W 2 = (P + q)2 = M 2 + Q ,
x

As q2 < 0, it is customary to introduce Q2 , the magnitude of q2 . E and E 

are the energies of the incoming and outgoing lepton.

20.2 Parametrization of the target structure

The cross section for e(k) + p(P) → e(k ) + X reads
 ∫
1 1
dσ = dφ |M | 2 . (20.2.1)
X
4M E 4 spin

The summation runs over all possible (unobserved) hadronic final states. We
factorize the phase space and the squared matrix element into two parts, one
for the lepton (denoted by L) and one for the hadrons (denoted H):

d3 k  1 e4
dφ = dφ L dφ H , dφ L = , |M | 2 = 4 L μν Hμν .
(2π)3 2E  4 spin Q

Then the hadron part of the cross

∫ section is the dimensionless tensor of
1 
Lorentz indices Wμν = 8π X dφ H
1
Hμν (the factor of 8π is included here
μ μ
by convention). As it depends on two momenta P and q , the most general
 Deeply inelastic lepton-proton scattering 299

gauge invariant expression for the Lorentz tensor can be written as

 
qμ qν
Wμν (P, q) = −gμν + 2 W1 (x, Q2 )
q
  
P·q P · q W2 (x, Q2 )
+ Pμ − qμ 2 Pν − qν 2
q q P·q

where the structure functions Wi (x, Q2 ) are dimensionless functions of the

scaling variable and the momentum transfer.
For the lepton part we express the kinematic relations E  = (1 − y)E,
xy M
cos θ = 1 − (1−y)E to change variables to scaling variable and relative energy
loss:
d3 k  dϕ E  dϕ yM E

= dE  d cos θ = dy dx
(2π) 2E
3 2π 8π 2 2π 8π 2

and compute the trace L μν = 12 Tr[/k γ μ k/  γν ] = k μ k ν + k ν k μ − g μν k · k  .

 

Then the differential cross section as a function of x and y is obtained from

Eq. (20.2.1) as
   
d2 σ 4πα2 2 1−y M2
= y W 1 (x, Q 2
) + − x y W2 (x, Q 2
) ,
dx dy y Q2 x Q2

which we rewrite in the scaling limit, defined by Q2 → ∞ with x fixed, as

 
d2 σ 4πα2 - 2. 1−y- .
= 1 + (1 − y) F1 + F2 − 2xF1 . (20.2.2)
dx dy y Q2 x

The dimensionless functions F1 and F2 were first measured by the SLAC-

MIT experiment [Miller et al., 1972]. The result of that measurement sup-
plemented by some later ones is shown in Fig. 20.1. The interesting feature
is that in the scaling limit F2 becomes independent of Q2 , F2 (x, Q2 ) → F2 (x)
(in fact, the independence starts at quite low values of Q2 ).

20.3 DIS in the parton model

Let us now describe the same scattering process by assuming the proton
is a bunch of free flying quarks and the lepton exchanges a hard virtual
photon with one of those quarks as shown in Fig. 17.61 (right). The struck
quark carries a momentum pμ , which is a fraction of the proton momentum,
300 Chapter 20

Figure 20.1: Measured value of the F2 structure function at several different

values of Q2

pμ = ξP μ , so we consider the process (k) + q(p) → (k ) + q(p). The

corresponding cross section is

∫
1 1
dσ̂ = dφ2 |M | 2 ,
2ŝ 4 spin

with ŝ = (p + k)2 . The hat refers to partonic quantities (cross section,

c.m. energy squared) here and below. The squared matrix element is pro-
portional to the product of the lepton tensor L μν and a similar quark tensor
 
Q μν = 12 Tr[ /pγ μ /p γν ] = pμ p ν + pν p μ − g μν p · p, i.e. L μν Q μν = 2(ŝ2 + û2 )
where û = (p − k )2 = −2p · k . As y = P · q/P · k = 2p · q/2p · k = (ŝ + û)/ŝ,
momentum conservation pμ = pμ + qμ implies for the on-shell condition of

the scattered quark 0 = p 2 = (p + q)2 = 2p · q + q2 = ŝ + û − Q2 . We have
y = Q2 /ŝ and û = (y − 1)ŝ so

1 eq2 e4 ŝ2 - .
|M | 2 = 4 L μν Q μν = 2eq2 e4 4 1 + (1 − y)2 .
4 spin Q Q
 Deeply inelastic lepton-proton scattering 301


Also Q2 = 2p · q = 2ξP · q, so p 2 = Q2 (ξ/x − 1). Then the two-particle
phase space is

d3 k  d4 p -  .
dφ2 = 2πδ+ p 2 (2π)4 δ4 (k + p − k  − p)
(2π) 2Ek  (2π)
3 4
(20.3.3)
dϕ E   x
= dE d cos θ 2 δ(ξ − x) ,
2π 4π Q
√
2yx
or using E  = ŝ (1 − y) and cos θ = 1 − , we obtain dφ2 =
2 ξ(1 − y)
dϕ y ŝ
dy dx δ(ξ − x) . The differential cross section in x and y
(4π)2 Q2

d2 σ̂ 4πα2 + ,1
= 2
1 + (1 − y)2 eq2 δ(ξ − x) . (20.3.4)
dx dy Q 2
Comparing Eqs. (20.2.2) and (20.3.4), we find the parton model predic-
tions

F1 (x) ∝ eq2 δ(ξ − x) , F2 − 2xF1 = 0 , called Callan-Gross relation.

(20.3.5)
Thus F2 probes the constituent of the proton with momentum fraction ξ = x.
However, this prediction for F2 cannot be correct because F2 (x) is not a δ
function as seen from Fig. 20.1, which makes us formulate the naïve parton
model in the following way:
the virtual photon scatters incoherently off the constituents (partons)
of the proton;
the probability that a quark q carries a momentum fraction of the
proton between ξ and ξ + δξ is fq (ξ)dξ.

Exercise 20.1
Compute the contribution to the DIS cross section in Eq. (20.2.2) with the
exchange of a transversely polarized photon. Hint: Use Eq. (18.2.3) for the
numerator in the propagator of the transversely polarized photon and the
Callan-Gross relation in Eq. (20.3.5). Can you identify the result with any
of the terms in Eq. (20.2.2)? What is the source of the remainder?
302 Chapter 20

20.4 Measuring the proton structure

With the assumptions of the naïve parton model the Callan-Gross relation
predicts
∫ 1 
F2 (x) = 2xF1 (x) = dξ fq (ξ) x eq2 δ(x − ξ) = x eq2 fq (x) .
q 0 q
(20.4.6)
For example, taking into account four flavours and simplifying the notation
by using fq (x) ≡ q(x), we obtain a prediction for the structure function mea-
sured in scattering of charged-lepton off proton (neutral current interaction):
 
4- . 1- .
F2 (x) = x
em
u(x) + ū(x) + c(x) + c̄(x) + d(x) + d(x) + s(x) + s̄(x) .
¯
9 9
Similarly, in charged current interactions the prediction is
+ ,
F2ν̄ (x) = 2x u(x) + d(x)
¯ + c(x) + s̄(x) (with W − ) ,
+ ,
F2ν (x) = 2x d(x) + ū(x) + s(x) + c̄(x) (with W + ) .
Further information can be obtained if we use different targets. Assuming
two flavours and isospin symmetry, the proton structure (with uud valence
quarks) is
 
p 4- . 1- .
F2 (x) = x u p (x) + ū p (x) + d p (x) + d¯p (x) , (20.4.7)
9 9
and that of the neutron (with udd valence quarks) is
 
4- . 1- .
F2 (x) = x
n
un (x) + ūn (x) + dn (x) + dn (x)
¯
9 9
  (20.4.8)
1- . 4- .
=x u p (x) + ū p (x) + d p (x) + d p (x) .
¯
9 9
The measurements are supplemented by sum rules. For instance, as the
proton consists of uud valence quarks, we have
∫ 1 ∫ 1
- . - .
dx u p (x) − ū p (x) = 2 , dx d p (x) − d¯p (x) = 1 ,
0 0
∫ 1 - .
dx s p (x) − s̄ p (x) = 0 .
0
 Deeply inelastic lepton-proton scattering 303

The combination of the measurements and sum rules gives separate infor-
mation on the quark distributions in the proton fq (x). The result of such
measurements performed by the NMC collaboration [Arneodo et al., 1997]
is shown in Fig. 17.62 (left) together with a fit to the data by the CTEQ
collaboration [Pumplin et al., 2002]. The parton distributions deduced from
the fit are shown in Fig. 17.62 (right).
We can infer the proton momentum from the measurements. The surpris-
ing result is that quarks give only about half of the momentum of the proton,
 ∫1
q 0 dx x fq (x) 0.5. By now we know that the other half is carried by
gluons, but clearly the naïve parton model is not sufficient to interpret the
gluon distribution in the proton. With our experience in perturbative QCD
we try to compute radiative corrections to the quark process to see if that
helps to find the role of the gluon distribution.

Exercise 20.2
It is not feasible to use a neutron target experimentally. Instead deuteron
is used which is the bound state of a proton and a neutron. The corre-
p p
sponding structure function is F2d (x) = 12 (F2 (x) + F2n (x)), with F2 and F2n
given in Eqs. (20.4.7) and (20.4.8), respectively. Which combination of the
structure function on proton and deuteron targets gives the u- and d-quark
distributions?

20.5 Improved parton model: perturbative QCD

 
Using the relations dy = dQ2 /ŝ and δ(ξ − x) = ξ1 δ 1 − ξx , we rewrite the
differential cross section (20.3.4) in a more usual notation,
∫  
d2 σ 1
dξ  d2 σ̂ x 2
= fi (ξ) , Q , (20.5.9)
dx dQ2 0 ξ i dx dQ2 ξ

which gives the cross section as a convolution of a long-distance component

(the parton distribution function, PDF) and a short-distance component (the
hard scattering cross section). This form of the cross section is the main
content of the factorization theorem, which we derived heuristically, but
304 Chapter 20

there exists a rigorous though not completely process independent proof,

based on QFT.
The factorization formula (20.5.9) raises some questions. Knowing that
the quarks do not give the total momentum of the proton, it is natural to
include the contribution of gluons in Eq. (20.5.9). However, we do not yet
know the corresponding hard scattering cross section. We also do not know
how we can apply perturbation theory. Furthermore, the scaling was exact
in the parton model. Is it so in QCD? There is a common answer to these
questions: DIS in perturbative QCD.
To develop perturbative QCD for DIS, let us revisit the infrared singu-
larities once more. Let us denote the hard scattering cross section for some
final state by σh . Then the cross section in the collinear approximation for
the same final state with an extra gluon of relative transverse momentum
kT = Eθ, carrying momentum fraction (1 − z) is
(1 − z)p

αs dE dθ αs dz dk T2
∝ σh+g σh 2CF = σh CF .
p
σh θ

zp π E θ π 1 − z k T2

Integrating over z up to one and over k T we find soft and collinear divergences.
In studying perturbative QCD we found that these infrared singularities in
the final state cancel against infrared divergences in the virtual correction for
infrared safe quantities:
p αs dz dk T2
σh
∝ σh+V −σh CF .
π 1 − z kT2

If there is a coloured parton in the initial state, then the splitting may
occur before the hard scattering and the momentum of the parton that enters
the hard process is reduced to zpμ , so
p zp σh αs dE dθ
θ ∝ σh+g (p) σh (zp)2CF
(1 − z)p
π E θ
αs dz dk T2
= σh (zp) CF .
π 1 − z kT2

Integrating over z up to one and over k T we again find soft and collinear
divergences. The corresponding  poles multiply σh (zp), while in the vir-
tual correction the poles multiply σh (p), irrespective whether the infrared
divergence is in the initial or final state:
p p αs dz dk T2
σh
∝ σh+V −σh (p) CF .
π 1 − z k T2
 Deeply inelastic lepton-proton scattering 305

The sum of the real and virtual corrections then contains an uncancelled
singularity,
∫ Q2 ∫
αs dkT2 1
dz
σh+g + σh+V CF [σh (zp) − σh (p)] ,
π m2g k T20 1−z
789: 789:
infinite if mg =0 finite

where we used a finite gluon mass to regulate the collinear divergence

(instead of dimensional regularization) to make manifest that the collinear
singularity remains, while the soft one (at z → 1) vanishes in the sum.
This uncancelled collinear singularity in the initial state is a general
feature of perturbative QCD computations with incoming coloured partons
and its form is universal, so we can find its precise form studying the structure
function at NLO accuracy. We know that in the parton model (QCD at LO)
the prediction for hard scattering cross section F̂2 is finite:
  
d2 σ̂  d2 σ̂ 
F̂2,q (x) = = eq2 x δ(1 − x) , F̂2,g (x) = = eq2 x · 0
dx dQ2 F2 2
dx dQ F2
q

i.e. it is zero in the gluon channel because the virtual photon does not interact
with the gluon directly. At one order higher in αs we find (the diagrams below
show the contributions to each term: one diagram at LO, one loop diagram
and two real radiation diagrams at NLO)
   
d2 σ̂  αs Q2 q
F̂2,q (x) = = eq x δ(1 − x) +
2
P (x) ln 2 + C2 (x) ,
qq
dx dQ2 F2 4π mg
(20.5.10)

and (only one real radiation diagram with gluon splitting in the initial state
contributes)
   
d2 σ̂   αs Q2 g
F̂2,g (x) = = eq x 0 +
2
P (x) ln 2 + C2 (x) ,
gq
dx dQ2 F2 q
4π mq
(20.5.11)

where Pi j (x) is the Altarelli-Parisi splitting function for i → j + k (regular-

ized at x = 1), while C2 (x) is the remaining finite term, called coefficient
306 Chapter 20

function. The space-like splitting function P ji (x) is obtained from the time-
like splitting kernel P̂ jk in four dimensions through the following steps: (i)
crossing the kernel for time-like splitting into space-like splitting:
 
1
P̂(0), i k (x, kT ) = −(−1)2s(i)+2s(j) x P̂(0)
¯
jk x T
, k ;  = 0 (20.5.12)

where i¯ is the antiparticle of parton i, s(i) denotes the spin of parton i;

(ii) averaging over the spin states of parton i to obtain P̂(0), i k (x, kT ); (iii)
¯

relabelling superscripts P̂(0), ik (x, kT ) → P j i (x) and (iv) adding the con-
¯ ¯ ¯

tribution from the one-loop diagram (see exercise). The flavours satisfy the
flavour summation rule fi = f j + fk . We see that at NLO the prediction
for F̂2 does not contain ultraviolet divergences. The final state infrared di-
vergences also cancel, but an un-cancelled singularity remains in the initial
state, regularized with a small mass (mg or mq ) here.
The hard scattering function is not measurable, only the structure function
is physical:
(

F2,q (x, Q ) = x
2
eqi fq(0)
2
i
(x)
i
∫      )
αs 1
dξ (0) qq x Q2 q x
+ f (ξ) P ln 2 + C2 .
2π 0 ξ qi ξ mg ξ

However, this function appears divergent if the regulator is removed, mg → 0.

While C2 (x) depends on the process under investigation, the divergence does
not because it is multiplied with universal splitting functions.

Exercise 20.3
g
Compute the coefficient C2 (x) in Eq. (20.5.11).

20.6 Factorization in DIS

If the remaining divergences are universal (and they are because do not
depend on the hard scattering), we can absorb the singularity into the parton
 Deeply inelastic lepton-proton scattering 307

distribution functions. For instance, defining

 ∫ (    ) 
αs
dξ (0) 1
qq x
μ2F qq x
fq (x, μF ) = fq(0) (x) + fq (ξ) P ln 2 + K ,
0 ξ
2π ξ mg ξ
(20.6.13)
the structure function becomes (recall that the strong coupling depends on
another scale, the renormalization scale μR )
(

F2,q (x, Q ) = x
2
eq2 i fi (x, μF )
i
∫      )
αs (μR ) 1
dξ x Q 2 - q . x
+ fi (ξ, μF ) P qq ln 2 + C2 − K qq
2π 0 ξ ξ μF ξ
(20.6.14)

where the summation runs over the light quark flavours.1 The long distance
physics is factored into the parton distribution functions that depend on
the factorization scale μF . The short distance physics is factored into the
hard scattering cross section that depends on both the factorization and the
renormalization scales. Both scales are arbitrary, unphysical scales. The
term K i j defines the factorization scheme. It is not unique, finite terms can
be shifted between the short and long distance parts, but it is important that it
must be chosen the same in all computations (the MS scheme is the standard
where K i j = 0).
Defining the convolution in x-space,
∫ 1 ∫ 1
( f ⊗ g)(x) ≡ dy dz f (y) g(z)δ(x − yz) ,
0 0

we see that the structure function is ‘factorized’ in the form of a convolution,

   Q2
F2,q (x, Q2 ) = x eq2 i fi (μF ) ⊗ F̂2,i (μR, t) (x) , t = ln .
i
μ2F

1Depending on what we consider light quark in the colliding proton, we speak

about four (u, d, c and s quarks) or five flavour number scheme (the b quark also
considered in addition to the lighter ones).
308 Chapter 20

Exercise 20.4
The regularization of the splitting function at x = 1 is achieved by the
+-prescription defined by
∫ 1 ∫ 1
- .
dx[g(x)]+ f (x) = dxg(x) f (x) − f (1)
0 0

for any smooth test function f (x). The contribution of the loop corrections
has the same kinematics as the LO one, so it has to be proportional to
δ(1 − x). Thus the complete regularized splitting function has the form
 
1 + x2
P (x) = CF
qq
+ Kδ(1 − x) .
(1 − x)+

We can obtain the parton distribution for a ‘quark in a quark’ from Eq. (20.6.13)
by the substitution fq(0) (x) → δ(1 − x)

αs qq μ2
fq (x, μF ) = δ(1 − x) + P (x) ln F2 .
2π mg

Integration over x gives the number of quarks in a quark that has to be

∫1
one, independently of μF . Thus we have the condition 0 dxP qq (x) = 0.
Compute the regularized splitting function.

20.7 DGLAP equations

We can compute the short-distance component of the factorized structure
function of Eq. (20.6.14) in QCD perturbation theory. It depends on the
renormalization scale, but recall that it has to satisfy the renormalization
group equation.
We cannot compute the parton distribution functions in perturbation
theory, so it seems that this is the end of the story: perturbative QCD appears
non-predictive for processes with hadrons in the initial state. However, the
arguments that lead to the renormalization group equation come to the rescue.
While the right hand side of Eq. (20.6.14) depends on both renormalization
and factorization scales, the measurable quantity F2 does not, which can be
expressed by the renormalization group equation. Of course, this statement
has to be understood perturbatively, namely at any order in PT, the right
 Deeply inelastic lepton-proton scattering 309

hand side of the renormalization group equation is not exactly zero, but may
contain terms that are higher order in perturbation theory. Only infinite order
is expected to give exact independence of the scales. The renormalization
group equation gives the missing piece of information needed to make the
theory predictive.
To write the renormalization group equation, we introduce Mellin trans-
∫1
forms defined by f (N) ≡ 0 dx x N −1 f (x), which turns a convolution into a
real product:
∫ 1 ∫ 1 ∫ 1
N −1
dx x dy dz δ(x − yz) f (y) g(z) =
0 0 0
∫ 1 ∫ 1
= dy dz (yz) N −1 f (y) g(z) = f (N)g(N) .
0 0

So F2,q (N, Q2 ) = x i eq2 i fi (N, μF )F̂2,i (N, μR, t) is independent of μF , ex-
pressed as  
dF2 - . - .
μF = 0 = O αsn+1 in PT at O αsn .
dμF
Let us explore the consequences of this renormalization group equa-
tion. For simplicity, we assume the existence of only one quark flavour,
F2,q (N, Q2 ) = x eq2 i fq (N, μF ) F̂2,i (N, μR, t). Then the renormalization group
equation reads

d fq dF̂2,q
F̂2,q (N, t) (N, μF ) + fq (N, μF ) (N, t) = 0 .
dμF dμF

Dividing with fq F̂2,q , it turns into

d ln fq d ln F̂2,q
μF (N, μF ) = −μF (N, t) ≡ −γqq (N) , (20.7.15)
dμF dμF
where γqq (N) is called the anomalous dimension because it acts as a factor
−γ (N )
μF q q in the dimensionless function fq (N, μF ). Taking the Mellin moment
of Eq. (20.6.13) and then its derivative with respect to μF , we obtain that the
anomalous dimension is
d ln fq αs (μR ) qq - .
γqq (N) = −μF (N, μF ) = − P (N) + O αs2 , (20.7.16)
dμF π
thus it is the Mellin transform of the splitting function, which can be com-
puted in PT. Equation (20.7.15) implies that the scale dependence of the
310 Chapter 20

parton distribution functions can be predicted in PT. This together with the
universality of parton distribution functions makes perturbative QCD pre-
dictive: we can measure the parton distribution functions in one process at
a certain scale and then use it in another process at another scale to make
predictions.
How shall we choose the renormalization and factorization scales? If we
want to avoid large logarithms that spoil the convergence of the perturbative
series, the scales should be chosen near the characteristic physical scale of
the process Q, e.g. μ2R = μ2F = Q2 . Then the renormalization group equation
becomes
 
2 d ln fq 1 - 2.
Q (N, Q ) = − γqq N, αs Q ,
2
(20.7.17)
dQ2 2
which is the Mellin transform of
- .
2 d fq αs Q2  qq - 2. 
Q (x, Q 2
) = (P ⊗ fq Q (x) . (20.7.18)
dQ2 2π
Our discussion was highly simplified by considering only one quark flavour
and neglecting the mixing of partons. If we make the full computation we
obtain the celebrated formula
- .
2 d fi αs Q2   i j - 2. 
Q (x, Q 2
) = P ⊗ f j Q (x) , (20.7.19)
dQ2 2π j

called DGLAP equation, named after Yuri Dokshitzer [Dokshitzer, 1977],

Vladimir Gribov, Lev Lipatov [Gribov and Lipatov, 1972] and Guido Altarelli,
Giorgio Parisi [Altarelli and Parisi, 1977].
Let us now solve the (simplified) DGLAP equation in Mellin space,
Eq. (20.7.17). It is a simple first order differential equation whose solution
is  ∫ t  
- .
fq (N, Q2 ) = fq (N, Q20 ) exp − dt γqq N, αs Λ2 et .
t0
- . 2
Let us recall the one-loop formula in Eq. (19.3.20), αs Q2 = 1
b0 t , t = ln Q
Λ2
γ (N )
and introduce the abbreviation dqq (N) = − q2πb q
0
≤ 0. Then
 ∫ t 
dt
fq (N, Q2 ) = fq (N, Q20 ) exp dqq (N) , (20.7.20)
t0 t
or   dq q (N )
t
fq (N, Q ) =
2
fq (N, Q20 ) , (20.7.21)
t0
 Deeply inelastic lepton-proton scattering 311

called scaling violation.

As γqq (1) = 0, the number of valence q-quark in the proton, given by
∫1
the integral 0 dx fq (x, Q2 ), is independent of Q2 . Higher moments vanish
more rapidly, therefore the average x decreases as Q2 increases. Thus we
predict that fq (x, Q2 ) increases at small x and decreases at large x. This
prediction is seen to be valid from the measurements shown in Fig. 20.2.

Exercise 20.5
Compute the anomalous dimension γqq (x) using Eq. (20.7.16).
312 Chapter 20

Figure 20.2: Measurement of the F2 structure function at different Q2 as a

function of x,
Chapter 21

Hadron collisions

MOTTO:
I cannot define the real problem, therefore I
suspect there’s no real problem, but I’m not
sure there’s no real problem.

(Richard P. Feynman)

21.1 Factorization theorem

While electron-positron annihilation and DIS played very important roles
in establishing perturbative QCD for understanding high-energy scattering
experiments, presently and in the mid-term future the experiments at the
energy frontier can be found at the LHC. Thus we are most interested in the
theoretical tools needed to understand high-energy proton-proton collisions.
Fortunately, the tools we have developed so far can be generalized
straightforwardly to hadron collisions. The most general form of the factor-
ization theorem includes convolution with two parton distribution functions,
one for each colliding parton, the hard scattering cross section, and possi-
bly a convolution with a fragmentation function (FF) of a parton into an
identified hadron in the final state. Thus, the differential cross section for a
hypothetical inclusive process pp → Z + π + X, in which we identify a Z
313
314 Chapter 21

boson and a pion only, has the form

dσpp→Z+π+X (s, x, αs, μR, μF ) =

∫ 1 ∫ 1
= dx1 fi/p (x1, αs, μF ) dx2 f j/p (x2, αs, μF )
i, j,k 0 0
∫ 1   (21.1.1)
dz x
× dσ̂i j→Z+k+X (ŝ, z, αs (μR ), μR, μF )Dπ/k , ŝ
x z z
 p
Λ
+O .
Q

In the factorization formula Eq. (21.1.1) s denotes the total centre-of-mass

energy squared, x/z is the longitudinal momentum fraction of the pion in
the parton k, μR and μF are the renormalization and factorization scales,
fi/p (x) is the parton distribution function for parton i in the proton with
momentum fraction x, dσ̂i j→Z+k+X (ŝ) is the hard scattering cross section for
the partonic process, Dπ/k (x) is the fragmentation function for the process
parton k → π. The last term shows that power-suppressed contributions at
high Q2 are neglected (p > 1). Substituting the parton distribution functions
and fragmentation functions with δ functions (in momentum and flavour) we
obtain the cross section formulae in DIS and electron-positron annihilation.
The parton distribution functions and fragmentation functions constitute
the long-distance, non-perturbative components of the cross section that can-
not be computed in perturbative QCD, only extracted from measurements.
Thus, it is a natural question whether or not the factorization theorem is
predictive. The answer is a clear yes for the following reasons.
We can compute the hard scattering cross section in perturbation theory,
which involves (i) renormalization of ultraviolet divergences (order by order
in perturbation theory), (ii) cancellation of infrared ones for infrared safe
observables using a subtraction method, (iii) absorbing initial state collinear
divergences into renormalization of parton distribution functions (and pos-
sibly uncancelled final state ones into that of the fragmentation function).
The non-perturbative components are universal, so can be measured in one
process and used to make predictions in another one. Furthermore, the
evolution of these with Q2 can be predicted in perturbation theory by the
DGLAP equations as shown in Fig. 17.63.
In summary, we are prepared to make predictions for any high-energy
scattering process. The theoretical framework for such predictions relies
on QCD perturbation theory and the factorization theorem. In perturbation
theory we can compute the hard scattering cross section and the evolution
 Hadron collisions 315

of the parton distribution functions. There are universal elements, such as

the parton distribution functions and fragmentation functions, as well as the
subtraction method for computing radiative corrections. These principles are
well understood and reasonably simple, though the actual implementations
are cumbersome and their improvements are subjects of active research.

21.2 Are we happy?

At this point theorists can make precision predictions for distributions of in-
frared safe observables. The main bottleneck to make such predictions is the
algebraic complexity of computing amplitudes and the analytic complexity
of evaluating loop integrals. The state of the art considers the computation
of NLO corrections a solved problem with automated implementations for
processes up to about five partons in the final state (at tree level). The exact
number depends on the process being considered because the numerical inte-
grations become too expensive eventually. Nevertheless, all processes listed
in the ‘Les Houches wishlist (2011)’ are known by now. Furthermore, there
is also a computer code to compute seven-jet production in electron-positron
annihilation [Becker et al., 2012].
For experimenters the situation is less satisfactory. While predictions
in perturbative QCD are based on a solid theoretical ground, those lack
important features. While perturbative QCD gives predictions for final states
with few partons, the experiments detect hadrons. A tool that can simulate
real events with hadrons at correct rates would be much more handy. To
finish these lectures we look into modelling events in a qualitative way. A
more detailed description can be found for example in Ref. [Skands, 2012].

21.3 Modelling events

Figure 17.64 shows our view of a proton-proton scattering event at high
energy. The three parallel lines ending in discs from both sides represent the
two incoming protons. At high energies these protons consist of (almost)
free-flying partons, two of which (one from each) collide at high centre-of-
mass energy and produce the hard scattering, with perturbatively computable
cross section. This is where signatures of new physics may appear. The hard
scattering cross section is process dependent. We have discussed how it can
be computed from first principles, from Feynman diagrams and rules. The
precision of the predictions can be improved systematically in perturbation
theory by computing the effect of radiative quantum corrections.
316 Chapter 21

Before collision the colliding partons may emit other partons nearly
collinear with the beam. These collinear emissions in the initial state give
rise to divergences that can be factored into the renormalized parton dis-
tribution functions. After collision few energetic partons appear that may
emit less energetic partons and each develops showers of partons. Emissions
into almost the same direction as the original parton occur with enhanced
probability (due to the collinear divergence) as well as emissions of soft glu-
ons. This is represented in Figure 17.64 by red quark and gluon lines. Both
factorization and parton showering can be described from first principles
based upon known physics of QCD and are universal, meaning that these are
independent of the process and observable. We have seen how factorization
works, but have not discussed how parton showers are modelled with shower
Monte Carlo (SMC) programs [Sjostrand et al., 2006, Corcella et al., 2001].
We mentioned marginally how the large logarithms emerging in the final
state splittings can be resummed. Such resummations give improved pre-
diction for the cross section (as seen in Fig. 17.60), but does not simulate
events.
Parton showers still only give a description of events in terms of quarks
and gluons, whereas detectors detect only hadrons. We do not know how
to compute hadronization, the transition from quarks and gluons to hadrons,
from first principles. Yet the idea of local parton-hadron duality (see,
e.g. Ref. [Dokshitzer et al., 1991]) provides some sort of theoretical un-
derstanding. It states that
after accounting for all gluon and quark production down to scales
ΛQCD , the transition from partons to hadrons is essentially local in
phase space, i.e. there is no rearrangement of energy and momentum.
Thus the directions and momenta of hadrons will be closely related to those
of the partons and the hadron multiplicity will reflect the parton multiplicity,
too. This is illustrated by the green lines with dots.
In addition to the energetic partons in the initial state, there are also low-
energy ones that may collide, which is energy and process dependent. This
low-energy physics is described in models of underlying event, which are
also part of modern shower Monte Carlo programs. The underlying event
produces low-energy partons. Also at the end of the shower low-energy
partons emerge. As QCD confines partons, these partons turn into hadrons
before detection, a process called hadronization. We do not have a theory
of hadronization based on first principles. Instead, shower Monte Carlo pro-
grams include models that describe hadronization in a process independent
way. These models contain parameters that are fixed experimentally.
 Hadron collisions 317

21.4 Conclusions
In the last three chapters we discussed the theoretical basis of interpreting
the results of high-energy collider experiments. We discussed how pertur-
bative QCD can be made predictive and also the main uncertainties in the
predictions. We used the following key ingredients in this tour:
(i) gauge invariance that allows us to write down the Lagrangian and
which predicts many important features of the theory;
(ii) renormalization that cancels ultraviolet divergences systematically
order-by-order in perturbation theory and introduces a dimension-
ful scale into even the scaleless Lagrangian of massless QCD, leading
to scaling violations of one-scale observables that would be scale
independent in the classical theory;
(iii) asymptotic freedom at high energies emerging from the quantum struc-
ture of the theory and the non-abelian nature of the gauge group;
(iv) need for infrared safety, emerging from asymptotic freedom, to en-
sure that the infrared divergences, associated with unresolved parton
emission, cancel between real and virtual contributions, allowing the
perturbative calculation of jet cross sections, without a detailed under-
standing of the mechanism by which partons become jets;
(v) factorization that makes possible to use perturbative QCD to calculate
the interactions of hadrons, since all the non-perturbative physics gets
factorized, into parton distribution functions;
(vi) evolution and universality of parton distribution functions that allows
us to extract those measuring cross sections in one process, like DIS,
and then used to predict the cross sections for any other process. Again,
this factorization introduces a scale dependence into the parton model
so that the structure functions of DIS, and other one-scale observables
become scale dependent.
These features make perturbative QCD predictive, without forcing us to
solve the theory at all possible scales: unknown or not calculated high and
low-energy effects can be renormalized, factorized and cancelled away.
In the last chapter we study the sector of the standard model with bro-
ken symmetry. The photon appears here, which is massless like the gluon.
However, the photon does not carry colour charge and does not have self in-
teraction, it only mediates the electromagnetic interaction between fermions.
318 Chapter 21

As the coupling of electromagnetism is much weaker than the strong cou-

pling, the processes with photon radiation can be approximated quite well
already at the lowest order in perturbation theory. Therefore in our introduc-
tory textbook we are satisfied with predictions at Born level, although the
precision of experiments require the inclusion of at least the first radiative
corrections.
Chapter 22

Electroweak sector of the

standard model

MOTTO:
And as they drifted up, their minds sang with
the ecstatic knowledge that either what they
were doing was completely and utterly and
totally impossible or that physics had a lot of
catching up to do.

(Douglas Adams: So Long, and Thanks for

All the Fish)

22.1 Weinberg mixing

The underlying gauge group of the broken part of the standard model is

G = SU(2)L ⊗ U(1)Y .

Here L stands for left (and later R for right) polarized particle currents and
Y is the hypercharge. The field content in one family (we consider the first
319
320 Chapter 22

one here) is
 
u
ψ1 = ψ2 = uR , ψ3 = dR quarks
d L
 
νe
ψ1 = ψ2 = 0 , ψ3 = eR− leptons
e− L

where for each flavour f = u, d, ν, e and

1
fL/R ≡ f∓ = (1 ∓ γ5 ) f .
2
One could imagine ψ2 = νR , but such leptons have not been observed and
thus not considered members of the standard model particles. There are three
families of each field ψi . For a matrix U ∈ G the following transformation
rules hold
1
Uψ1 (x) = eiT · α (x) ei y1 β(x) ψ1 (x) where T = (τ1, τ2, τ3 ) and
2
Uψ j (x) = ei y j β(x) ψ j (x) , j = 2, 3 .

Here τa represent the Pauli matrices, while α = (α1, α2, α3 ) and β are real
numbers. The number y j denotes the eigenvalue of the U(1) generator Y /2,
called weak hypercharge, for field ψ j (the factor 12 is included to maintain
the traditional Y = B + S, baryon number + strangeness definition for the
first three quarks). The Lagrangian for one family (family replication is
implicitly understood) is


3
L= / (j) ψ j (x),
iψ̄ j (x) D (22.1.1)
j=1

(j)
with Dμ = ∂μ + δ j1 ig T · W μ + ig  y j Bμ . The Lagrangian in Eq. (22.1.1) is
invariant under the gauge transformation of the ψ j fields, provided the four
gauge fields introduced in the covariant derivative transform according to
the rules
G 1
Bμ → Bμ (x) = Bμ (x) − ∂μ β(x) (22.1.2)
g
G i + ,
T · W μ (x) → T · W μ (x) = U(x) T · W μ (x) U † (x) + ∂μ U(x) U † (x)
g
(22.1.3)
 Electroweak sector of the standard model 321

where U(x) = exp [iT · α (x)]. The gauge invariant kinetic term for these
vector fields is
1 1
L B,W = − Bμν Bμν − W μν · W μν,
4 4
with Bμν = ∂μ Bν − ∂ν Bμ ≡ ∂[μ Bν] and W μν = ∂[μ W ν] − g W μ × W ν . Bμν
is invariant under G transformations, while T · W μν transforms covariantly:
G
T · W μν → U(x) T · W μν U † (x).
In quantum field theory the mass of a particle is the position of the pole
in the inverse two-point function in momentum space.1 Gauge symmetry
forbids such mass terms for gauge bosons. In case of the Dirac Lagrangian
the fermion propagator Δ j (q) = i q2 /+iε depends only on the momentum q,
q

i.e. does not contain mass. Thus, the mass of the fermion is zero. An extra
term proportional to iψ̄ψ would lead to a pole-position different from zero.
However, fermion masses must also be absent because

m ψ̄ψ = m ψ̄L ψR + m ψ̄R ψL ,

but the ψL , ψR fields transform differently under G. Thus the G-invariant

Lagrangian describes massless fields in contradiction to observation (for
instance mW 80 GeV, me 511 keV in natural units ). We return to the
problem of masses in the following sections.
The gauge invariant Lψ contains interactions of the fermion fields with
gauge bosons: 
/ ψ1 − g 
−g ψ̄1 T · W / ψj
y j ψ̄ j B
j

where  
1 Wμ3 Wμ1 − iWμ2
T · Wμ = .
2 Wμ + iWμ2
1 −Wμ3
The off-diagonal terms lead to charged-current interactions
g  
LCC = − √ Wμ† [ūγ μ (1 − γ5 ) d + ν̄γ μ (1 − γ5 ) e] + h.c. ,
2 2
with the stepping operators

1   1  
Wμ = √ Wμ1 + iWμ2 , Wμ† = √ Wμ1 − iWμ2 ,
2 2
1This is the pole mass definition that differs from the renormalized mass discussed
earlier. The two definitions are related in perturbation theory.
322 Chapter 22

and the letters u, d, e, ν stand for the spinor of the relevant particle. The
diagonal terms lead to neutral-current interactions
  
LNC = − / 3 + g yj B
ψ̄ j g T 3W / ψj . (22.1.4)
j

We would like to identify the fields Wμ3 and Bμ with the observed fields Zμ
and Aμ (the latter denoting the electromagnetic field), but both Wμ3 and Bμ
are massless, therefore an arbitrary combination of them is possible:
 3     
Wμ cos θ W sin θ W Zμ
= .
Bμ − sin θ W cos θ W Aμ

The angle θ W is called Weinberg (or weak mixing) angle. Equation (22.1.4)
becomes then
   
LNC = − / g T 3 sin θ W + g  y j cos θ W
ψ̄ j A
j
 
+ Z/ gT 3 cos θ W − g  y j sin θ W ψ j
(Z)
≡ LQED + LNC .

The term with Aμ should give QED and for this we require

Y
g sin θ W = g  cos θ W = e and Q= + T3 (22.1.5)
2
where Q is the electromagnetic charge operator, with
 
Qu/ν 0
Q1 = Q2 = Qu/ν Q3 = Qd/e .
0 Qd/e

In order to have the correct electromagnetic charge, the hypercharge of the

different species has to be
1 1
y1 = Qu/ν − = Qd/e + y2 = Qu/ν y3 = Qd/e . (22.1.6)
2 2
A possible right handed neutrino νR would have both QνR = 0 and yνR = 0
and thus it would not couple to anything (sterile neutrino). The existence
of such neutrino cannot be excluded, but since it does not interact with any
other particle of the model, we have no possibility to detect it in collider
 Electroweak sector of the standard model 323

experiments and for this reason we exclude it from the standard model. We
finally obtain
μ (Z) −e μ
LQED = −e Aμ JE M , LNC = Z μ JZ ,
sin θ W cos θ W

μ
JE M = Q j ψ̄ j γ μ ψ j ,
j
 
μ μ μ
JZ = T3 − sin2 θ W Q j ψ̄ j γ μ ψ j ≡ J3 − sin2 θ W JE M .
j

In terms of the usual fermion fields

−e  - .
(Z)
LNC = Zμ f¯γ μ v f − a f γ5 f , f = u, d or ν, e,
sin θ W cos θ W f
f f
where the axial coupling is a f = T3 and the vector coupling is v f = T3 −
2Q f sin2 θ W . We leave it as an exercise to compute the cubic and quartic
self-interactions among gauge bosons.

Exercise 22.1
The part of the standard model Lagrangian that determines the interaction
between fermions and electroweak gauge bosons is given by
   
1 - .
L= ψ̄L γ μ i∂μ + g τ · Wμ + g  yBμ ψL + ψ̄R γ μ i∂μ + g  yBμ ψR .
ψ
2 ψ
L R

The fields Wμ and Bμ denote the gauge fields of the weak isospin and the
hypercharges. The photon field Aμ and the Z boson field Zμ can be written
as linear combinations of these fields:
Aμ = Bμ cos θ W + Wμ3 sin θ W
Zμ = −Bμ sin θ W + Wμ3 cos θ W .
The relations of the various couplings are given by Eq. (22.1.5)
e = g sin θ W = g  cos θ W .
Show that one obtains the QED Lagrangian by substituting the photon and
the Z boson fields for the W 3 and the B fields in the Lagrangian density
above.
324 Chapter 22

22.2 U(1) Brout-Englert-Higgs mechanism

Our gauge symmetric Lagrangian is renormalizable, but the symmetry for-
bids masses, which contradicts reality. There is a way out: it is possible to
get non-symmetric result from an invariant Lagrangian. Let us consider a
Lagrangian L that

(a) is invariant under transformations of a continuous symmetry group G,

(b) has a degenerate set of states with minimal energy, which transform
under G as the members of a given multiplet.

The system arbitrarily selects one of these states as the ground state of
itself, we say in these cases that the symmetry is spontaneously broken. The
simplest example in field theory is the G = U(1) Brout-Englert-Higgs (BEH)
model. We consider the spontaneous symmetry breaking (SSB) of such a
model and move later on to the case of standard model.
Let us assume a Lagrangian L of a complex
scalar field φ which is symmetric under global U(1)
transformations

L = ∂μ φ∗ ∂ μ φ − V (φ)

where
V (φ) = μ2 φ∗ φ + λ (φ∗ φ)2 Figure 22.1: The Mex-
is the potential. L is invariant under global ican hat potential

φ → φ  = e−i eθ φ

phase transformations. The potential is bounded from below if λ > 0, which

leaves two options

(1) If μ2 > 0, then V has a single minimum at |φ| = 0 (a quadratic

function, rotationally symmetric around the V axis over the φ1 –φ2
plane with minimum at zero).

(2) If μ2 < 0, V is still rotationally symmetric, but has an infinite number

of degenerate states with minimum energy at |φ0 | = −μ
2
√v
2λ ≡ 2 (often
called Mexican hat potential, see Fig. 22.1). The value of the potential
at φ0 = √v eiθ is V (φ0 ) = − λ4 v4 < V(0) = 0.
2
 Electroweak sector of the standard model 325

If we choose for example θ = 0 as the ground state for the spontaneously

broken theory, and we parameterize the field around it as
1
φ (x) = √ [v + φ1 (x) + iφ2 (x)] , with φ1, φ2 real functions,
2
then the potential contains the following terms:
1 2 1 1
φ∗ φ = v + vφ1 + φ21 + φ22
2 2 2
1 1 2   1 2
(φ∗ φ)2 = v4 + v3 φ1 + v2 φ21 + v + 2vφ1 φ21 + φ22 + φ21 + φ22
4 2 4
1 4 3 2 2 1 2 2 1 2 2
= v + v φ1 + v φ1 + v φ1 + v φ2 + v φ1 φ2 +
3 3 2
φ1 + φ22 .
4 2 2 4
Consequently
  1 2 
V (φ) = V (φ0 ) + v μ2 + λ v2 φ1 + μ + 3λ v2 φ21
2
789:
=0
1 2    λ 2
+ μ + λ v2 φ22 + λ vφ1 φ21 + φ22 + φ21 + φ22 .
2 4
789:
=0

From the last equation we can read off that the mass of φ2 vanishes, mφ2 2 = 0,
which is a realization of Goldstone’s general theorem:
If a Lagrangian L is invariant under a continuous symmetry group
G but the vacuum is only invariant under a subgroup H ⊂ G, then
there exist as many massless spin-0 particles (Goldstone bosons) as
broken generators.
In the U(1) case there is one broken generator. The corresponding Goldstone
boson is φ2 that describes excitations around the flat direction in the potential.
We are more interested in the case when the symmetry is local. Then the
gauge-invariant Lagrangian is
 ∗   1
L = Dμ φ Dμ φ − V(φ) − Fμν F μν,
4
with Dμ [A] = ∂μ − i eAμ . We parameterize φ(x) around the ground state as
1 ξ (x)
φ (x) = √ [v + h(x)] ei v .
2
326 Chapter 22

The exponential factor is a phase that can be rotated (gauged) away by

exploiting the gauge invariance of L. In this gauge we use the transformed
fields
ξ (x) 1
φ (x) = e−i φ (x) = √ [v + h (x)]
v

2
 1
Aμ (x) = Aμ (x) − ∂μ ξ(x).
ev
In terms of the transformed fields (we drop the prime  to ease the notation)
the Lagrangian becomes
1- . 1 1 1
L= ∂μ h (∂ μ h) + e2 v2 A2 + e2 v h (x) A2 + e2 h2 A2 − Fμν F μν
2   2 2 4
v λ
− V √ − λ v2 h2 (x) − λ v h3 (x) − h4 (x) ,
2 4
from which we conclude
√ the following spectrum: a scalar Higgs particle h
with mass mh = 2λ v2 , cubic and quartic self interactions, and a massive
U(1) field Aμ with mass m A = e v. The massive vector field has three

components Ai (i = 1, 2, 3) and as A2 = A20 − 3i=1 A2i and A0 is unphysical
due to gauge fixing, the mass terms for the Ai field components have the
correct (negative) sign in the Lagrangian. The excitations of this gauge
field interact with the Higgs particle (third and fourth terms). We could
identify the U(1) symmetry with that of QED, but then the vacuum would
be electrically charged, which contradicts to our observations.

22.3 Brout-Englert-Higgs mechanism in the stan-

dard model
In the electroweak part of the standard model the covariant derivative acting
on the BEH field is given by
(φ)
Dμ = ∂μ + ig T · W μ + ig  yφ Bμ .
We introduce a new SU(2)-doublet
 +   
φ 1 φ1 + iφ2
φ= = √ .
φ0 2 φ3 + iφ4
Choosing yφ = 12 , the upper component has electric charge +1 and the lower
component is electrically neutral. After SSB of SU(2)L × U(1)Y into U(1)Q
 Electroweak sector of the standard model 327

the field becomes

 
1
φ = √ eiT · ξ (x)/v
0
.
2 v + h(x)

This ground state is electrically neutral, and so it does not couple to the
photon, which consequently remains massless as shown below. Eliminating
the unitary phase (in other words “choosing unitary gauge”), the scalar
kinetic term becomes
   2 
1 - . μ - . 0 
∂μ h (∂ h) +  ig T · W μ + ig  yφ Bμ 
2 v+h 

(the terms containing a single derivative of the h field cancel). Furthermore,

using yφ = 12 ,
  
1  - . 0 2 1  v √  2
3 
T · W + 
y  =  † 
, g − g
ig ig B 2g W B W μ 
2 v 
μ φ μ μ μ
2 2

v2
= g 2 Wμ†W μ
4

1  2
+ [g (A cos θW − Z sin θW ) − g (A sin θW + Z cos θW )] .
2 789:
g2
=[−Z(g sin θW +g cos θW )]2 =Zμ Z μ
cos2 θW

We rewrite the quadratic terms in the braces as

1 2
2
MW Wμ†W μ + M Zμ Z μ
2 Z
with
vg vg MW
MW = and MZ = = . (22.3.7)
2 2 cos θW cos θW
Thus we find that SSB predicts the masses of the vector bosons (note that the
Aμ field of QED is massless), with a non-trivial relation between MW and
MZ :
MW g
= cos θW = .
MZ g + g 2
2

The vector boson propagator is that of a massive vector field (with three
components propagating),
328 Chapter 22

p α pβ
g αβ − p
β M2
α = −i p2 −M 2 +iM V
, where V = W ±, Z.
MV V V ΓV

Low energy experimental facts (for example decay width of μ− →

e− ν¯e νμ , n → p e− ν̄e or p → n e+ νe , Λ→ p e− ν̄e , helicity suppression
of π − → e− ν̄e ) can be nicely described by
the four-fermion interaction term

GF
HI = √ J μ Jμ, (22.3.8)
2

introduced by Enrico Fermi, where GF = 1.16637 · 10−5 GeV−2 is Fermi’s

constant and

Jα = ūγα (1 − γ5 ) [d cos θ C + s sin θ C ] + ν̄e γα (1 − γ5 ) e + ν̄μ γα (1 − γ5 ) μ ,

with sin θ C 0.22 (θ C is called Cabibbo’s angle as this mixing of the quark
fields was introduced by Nicola Cabibbo).
In the low energy limit, the propagator of the massive gauge bosons
becomes
p
g αβ
α β +i 2 ,
MV |p|  MV MV
and from the electroweak theory we can recover the V − A model of weak
interactions (recall Table 4.1) with

GF g2 1
√ = = . (22.3.9)
2 8 M W
2 2 v2

As a result we deduce the vacuum expectation value of the BEH field in

terms of the Fermi constant, known precisely from low-energy experiments,
√  − 12
v= 2 GF 246 GeV.
We can also predict the value of MW as a function of sin θ W . Rearranging
Eq. (22.3.9) and substituting e = g sin θ W , we obtain
D
1 πα
MW = √
sin θ W 2 GF

where the fine structure α is not to be confused with the index α above. Using
the Feynman rules of the electroweak theory, from the tree-level diagrams
of Fig. 22.2,
 Electroweak sector of the standard model 329

ν ν
Z W
q q q

Figure 22.2: Neutral and charged current neutrino scattering at tree level

we can deduce the following predictions for neutral and charged current
neutrino scattering:

8π g 4  
σ (ν q → ν q) ≡ σNC (s) ∝ s a 2
q + a v
q q + v q ,
2
3 MZ4 cos4 θ W

8π g 4
σ (ν q → l q ) ≡ σCC (s) ∝ 4
s.
4 MW
(22.3.10)

(aq and vq denote the axial-vector and vector couplings of quark q). The
ratio of the two cross sections depends only on sin θ W (at LO accuracy):

 4  
σNC MW 4 2
= aq + aq vq + vq2 .
σCC MZ cos θ W 3
789:
f (sin θW )

Fitting this prediction to the measured value of the ratio one finds sin2 θ W =
0.231. Then with α(MZ ) 1/128, obtained from the measured value
α( 0) 1/137 by running the coupling, we can predict the values of the
gauge boson masses (at lowest order in perturbation theory),

SM prediction at LO measured
MW c2
GeV 80.23 80.385

MZ c2
GeV 91.49 91.188

We close this section with recording the complete Lagrangian in the

330 Chapter 22

Higgs sector:

1 1 M2 M2
Lh = ∂μ h ∂ μ h − Mh2 h2 − h h3 − h2 h4
2 2 2v 8v
  
+ −μ 1 2 μ h (x) h2 (x)
+ MW Wμ W
2
+ MZ Z μ Z 1+2 + .
2 v v2
√
where Mh = −2μ2 = 2λ v. The Higgs boson was observed at the
LHC by the ATLAS and CMS experiments. Its mass was measured to be
approximately 125.1 GeV/c2 , so λ 0.13. The Lagrangian also contains
cubic and quartic Higgs self interactions as well as vector boson-Higgs
interactions, proportional to MV2 . In order to prove that the standard model
Higgs boson was found the experiments have measured these couplings and
showed that these are consistent with the standard model predictions.

Exercise 22.2
In addition to explaining the mass of vector bosons, there is another puzzle
that the existence of the Higgs boson can explain. In the standard model
scattering of the electroweak vector bosons W + and W − is possible via
exchange of a photon or a Z boson both in the s and t-channels, as well as
through four-point interaction. Compute the cross section for longitudinally
polarized vector bosons assuming these processes at LO accuracy using the
Feynman rules in Sect. 22.7. What can you observe√ as the total centre-of-
mass energy of the colliding vector bosons grow, s → ∞?
With SSB, the colliding vector bosons can also interact through an off-
shell Higgs boson. Compute the contribution of these diagrams to the cross
section. How does the cross section now behave at large s?

Exercise 22.3
Compute the decay width of muon decay, μ− → e− ν¯e νμ (i) in the four-
fermion interaction theory of Fermi, (see Eq. (22.3.8)), (ii) in the standard
model (the necessary Feynman rules can be found in Sect. 22.7).

Exercise 22.4
Compute the cross sections in Eq. (22.3.10)
 Electroweak sector of the standard model 331

22.4 GIM (Glashow, Iliopoulos, Maiani) mecha-

nism
From the success of the low-energy interaction term in Eq. (22.3.8) we
conclude that the SU(2)L partner of the u quark should be

dc = d cos θ C + s sin θ C

instead of d quark only as seen in Eq. (22.1.1). However, if we use the field
(Z)
dc in LNC instead of d, we have

(Z) + μ ,
LNC ∝ cos θ C sin θ C Zμ dγ¯ (vd − ad γ5 ) s + s̄γ μ (vd − ad γ5 ) d
¯ μ (vd − ad γ5 ) d + · · · ,
+ cos2 θ C Zμ dγ

which gives flavour-changing neutral currents (FCNC) in the first two terms
with similar strength as Z d d¯ itself (third term), in contradiction to experi-
mental facts. For instance, K L = [ds̄] decays via neutral current into μ μ̄,
while K + = [us̄] decays via charged current into νμ μ̄ according to the fol-
lowing diagrams:
d μ− u νμ

Z0 W+
and
s̄ μ+ s̄ μ+

The ratio of their decay width is

Γ (K L → μ+ μ− )
- . = 2.8 · 10−9,
Γ K + → μ+ νμ

meaning that the neutral-current process is much suppressed. The solution

of Sheldon Lee Glashow, John Iliopoulos and Luciano Maiani (GIM) to the
problem is the following: in addition
 to u, d,
 s quarks
 there should exist a c
u c
quark, and two quark doublets and where
dc sc
       
dc cos θ C sin θ C d d
= ≡V .
sc − sin θ C cos θ C s s

Obviously, V † V = V V † = 1 from which

d̄c dc + s̄c sc = d̄d + s̄s ,

332 Chapter 22

which means that writing the Lagrangian in terms of the rotated fields is
equivalent to writing it using the original ones. In the latter case flavour
changing neutral currents do not appear, so these must be absent also when
we write the Lagrangian in terms of the rotated field. This prediction of
the charm, based on the absence of FCNC (in 1970) was confirmed by the
discovery of the particle J/ψ, a cc̄ bound state (in 1974). The problem
comes back at one loop if the c quark is too heavy, mc  1 GeV. The two
diagrams
μ− d μ−
d
W− W−

c νμ
u W+
νμ W+

μ+ μ+
s̄ s̄

cancel because both are proportional to cos θ C sin θ C , but with different
signs. This cancellation is incomplete if the second diagram is suppressed by
2 , which happens when mc  1 GeV, suggesting
1
mc in the propagator p2 −m
c
that the mass of the c quark should be around 1 GeV.

22.5 Fermion masses

We already discussed that explicit mass terms in the fermion sector would
break SU(2)×U(1) invariance. However, we can introduce a gauge-invariant
fermion-scalar couplings
 (+)   (0) ∗ 
- . φ - . φ
LY = cd ū, d L ¯ (0) dR + cu ū, d L
¯ uR
φ −φ(+) ∗
789:
≡ψ̄L ·φ (22.5.11)
 (+) 
φ
+ ce (ν̄e, ē)L eR + h.c. ,
φ(0)
where the hermitian conjugate terms are abbreviated. For the leptons it
  ν 
is of the form ēR φ(−), φ̄(0) e
and these are similar for quarks (see
e L
exercise). The terms in LY are called Yukawa terms and the couplings
cd, cu, ce are called Yukawa couplings. An infinitesimal transformation is
given by:
 
i
δψ̄L = ψ̄L (−i T ·  − i yL ) , δφ = i T ·  +  φ ,
2
 Electroweak sector of the standard model 333

so
 
- . i - .
δ ψ̄L · φ = −i T ·  + i T ·  − i yL  +  ψ̄L · φ
2
 
1 - .
= −i yL −  ψ̄L · φ
2
where yL = y1 = y3 + 12 (from Eq. (22.1.6)). Thus,
- . - . 
δ ψ̄L · φ = −i y3  ψ̄L · φ +- . ,
⇒ δ ψ̄L · φ dR = 0.
δ dR = i y3  dR
It is left as an exercise to check the gauge invariance of the second term,
proportional to cu . In a unitary gauge
 
1 0
φ (x) = √ ,
2 v + h (x)
from which
1 + ,
LY = √ (v + h(x)) cd d¯L dR + cu ūL uR + ce ēL eR + h.c.
2
We see that there are mass terms with mi = − c√i v where i = d, u, e:
2
 
h(x) + ,
LY = − 1 + ¯ + mu ūu + me ēe .
md dd
v
The standard model predicts that the coupling to the Higgs boson of the
fermions is proportional to the fermion masses. This prediction awaits
experimental confirmation. The first direct measurements of Higgs-fermion
couplings with large uncertainties are appearing at the time of writing this
book.

Exercise 22.5
Let us define a conjugate field φ̃ to the Higgs field φ by φ̃ = iτ2 φ∗ , where τ2
 (0) ∗ 
φ
is the second Pauli matrix, so φ̃ = . Check that the infinitesimal
−φ(+) ∗
- .
transformation of φ̃ under SU(2) L × U(1)Y is δ φ̃ = i T ·  − 2i  φ̃. Prove
that the
+- second , in Eq. (22.5.11) is invariant under SU(2) L × U(1)Y ,
. term
i.e. δ ψ̄L · φ̃ u R = 0.
334 Chapter 22

22.6 Flavour mixing

We have seen how the BEH mechanism, in addition to the gauge boson
masses, generates masses for the fermions if there is one family. However,
in the standard model there are three families. Let us explore if there are
additional consequences of the BEH mechanism in this case. The most
general Yukawa terms are

     (+) 
φ
 (0) ∗ 
φ

LY = ū j , d¯j c(d) d 
+ c (u)
u 
L jk φ(0) kR jk −φ(+) ∗ kR
j,k
   (+)  
 ¯ (l) φ 
+ ν̄ j , l c lkR + h.c. (22.6.12)
L jk φ(0)

where the fermion doublets are weak eigenstates and the couplings c(i) jk
are
arbitrary elements of 3 × 3 matrices (in flavour space). After spontaneous
symmetry breaking the Lagrangian becomes
 
h (x) ? ¯   
LY = − 1 + dL M d dR + ūL M u uR + l¯L M l lR + h.c. ,
v
where d , u  and l  are vectors in flavour space. The mass matrices are given
by
- . v - . v - . v
Md i j = −ci(d)
j √ , Mu i j = −ci(u)
j √ , Ml i j = −ci(l)j √ .
2 2 2
Their diagonalization determines the mass eigenstates. If none of the fermion
masses are zero (actually they are positive), M  is invertible. Furthermore,
the flavour space is finite dimensional. Then M can be decomposed as
M = H · U
where U U † = 1, H = H † and H is positive definite. (A general n × n
complex matrix has 2n2 independent real components, while both a unitary
and a hermitian matrix has n2 independent real components.) The hermitian
matrix H can thus be diagonalized (its diagonal form is denoted by M):
M  = S† M S U ,
with S S † = 1 and M has positive diagonal entries:
M d = diag (md, ms, mb ) ,
M u = diag (mu, mc, mt ) ,
- .
M l = diag me, mμ, mτ .
 Electroweak sector of the standard model 335

The mass eigenstates are

dL = S d dL , uL = S u uL , lL = S l lL ,

dR = S d U d dR , uR = S u U u uR , lR = S l U l lR

and the Lagrangian becomes

 
h (x)  ¯ 
LY = − 1 + d M d d + ūM u u + l¯M l l .
v

The matrices S and S U are unitary. Thus

ψ̄L · ψL = ψ̄L · ψL ψ̄R · ψR = ψ̄R · ψR

where ψ = d, u or l. Therefore, the absence of flavour changing neutral cur-

rents is preserved (this goes under the name of generalized GIM-mechanism).
For charged currents

ūL dL = ūL S u S †d dL ≡ ūLV dL ,

where V = S u S †d is the Cabibbo-Kobayashi-Maskawa (CKM) matrix2 that

couples u-type quarks with d-type quarks. Although, V V † = 1, V is different
from the identity matrix because S u  S d . The CKM matrix appears in
LCC when written in terms of mass eigenstates instead of weak eigenstates:
( )
g  
+ μ μ
LCC = √ Wμ ūi γ (1 − γ5 ) V i j d j + ν̄ γ (1 − γ5 ) + h.c. .
2 2 ij 

The presence of the CKM matrix in the charged current interactions is a

non-trivial prediction of the standard model even if the values of the matrix
elements are unknown parameters.
In the standard model mν = 0, thus we can rotate away V in the lepton
sector by redefining the phases of neutrino flavours such that Sν = S  . This is
ruled out by the observed masses for neutrinos seen via neutrino oscillations,
which we discuss in the next section.
The CKM matrix is a unitary 3 × 3 matrix. A general unitary matrix
has 9 independent parameters: three real rotational angles (as for a general
orthogonal matrix) and six complex phases. We can choose the phases of
the six quarks freely, thus can absorb all but one (five) complex phases of
2Named after Nicola Cabibbo, Makoto Kobayashi, and Toshihide Maskawa.
336 Chapter 22

the CKM matrix into the definition of the quark wave functions, which is
reflected by the usual parametrization of the CKM matrix:

V Vus Vub
 ud 
V =  Vcd Vcs Vcb 
 Vt d Vts Vtb 
1 0 0 c 0 s13 eiδ c s12 0
   13   12 
=  0 c23 s23   0 1 0   −s12 c12 0 
 0 −s23 c23   −s13 e
iδ 0 c13  0 0 1 
(22.6.13)

where ci j = cos θ i j , si j = sin θ i j and δ  0 is the remaining phase. At the

time of writing this book the experimental status of the CKM matrix is the
following:

|V | =
0.97446 ± 0.00010 0.22452 ± 0.00044 (3.65 ± 0.12) · 10−3
 +0.00010 
 0.22438 ± 0.00044 0.97359−0.00011 (42.14 ± 0.76) · 10−3  .
+0.24 −3 (41.33 ± 0.74) · 10−3
 (8.96−0.33 ) · 10 0.999105 ± 0.000032 
(22.6.14)

The one remaining phase is the signal of CP-violation, thus CP-violation

is predicted in the standard model with three families. The observed CP-
violation in the decay of kaons and B-mesons is in agreement with the
prediction of the standard model, thus the inventors of the three-family
CKM matrix, Makoto Kobayashi and Toshihide Maskawa were awarded the
Nobel prize in physics in 2008. (The two-family version was invented by
Nicola Cabibbo, hence the latter ‘C’, but that matrix is real, thus does not
predict CP-violation.)

Exercise 22.6
Prove that the CKM-matrix is real if there are only two families.
 Electroweak sector of the standard model 337

22.7 Parameters and Feynman rules of the

standard model
The predictive power of the standard model is strong because we can compute
the probability of many elementary particle processes with high precision.
However, in these computations we need the values of the input parameters of
the standard model that have to be extracted from measured data. Therefore,
it might be useful to collect the parameters of the standard model, even if
they are partially correlated.
In the gauge and scalar sector there are five parameters: the three gauge
couplings, gs , g and g , the vacuum expectation value v of the BEH field and
the mass of the Higgs particle MH . All other parameters can be expressed
as functions of these five:

1 1 2 2
MH g
MW = gv , MZ = g + g 2 v , λ= , tan θ W = ,
2 2 2v 2 g

and g sin θ W = g  cos θ W = e. In practice we had better choose parameters

that can be measured most precisely. These are Fermi’s constant GF , the
electromagnetic coupling αem and the mass of the Z0 boson, MZ . The
relations
1 g 2 g 2
GF = √ , and αem =
2v 2 4π(g 2 + g 2 )
imply that

1
v2 = √ ,
2GF
  
√ 4παem
g = 2 2 MZ G F 1 + 1 − √
2 2
,
2MZ2 GF
  
2
√ 4παem
g = 2 2 MZ G F 1 − 1 − √
2
2MZ2 GF

in the range tan θ W < 1.

In the fermion sector the input parameters are the Higgs-Yukawa cou-
plings. It is not simple to count the independent ones among them. There-
fore, assuming n fermion families, we consider the 3n fermion masses3 and
3Two quarks and one charged lepton are massive in each family – the neutrinos
are massless in the standard model.
338 Chapter 22

the (n − 1)2 independent elements of the CKM-matrix as input parameters.

Thus for n = 3 the number of input parameters is 5 + 3 · 3 + 22 = 18. We
have neglected the Θ coupling that characterizes the QCD vacuum as the
experimental upper bound is very small.

Feynman rules
We list the Feynman rules of electroweak interactions in unitary gauge, using
the notation of Sect. 18.4:4
• Cubic gauge field interactions V1,α V2,β V3,γ (with all-incoming kine-
matics, pμ + q μ + r μ = 0): Γα, β, γ (p, q, r) = ieCVα, β, γ (p, q, r)
where Vα, β, γ (p, q, r) = (p − q)γ gαβ + (q − r)α gβγ + (r − p)β gαγ is
the same function as in QCD, while C depends on the type of the
gauge boson participating in the interaction:

V1V2V3 C
γW +W − 1
cos θ W
ZW +W −
sin θ W

• Quartic gauge field + interactions V1,α V2,β V3,γ V4,δ, :

Γα, β, γ, δ = ie2 C 2gαβ gγδ − gαγ gβδ − gαδ gβγ , C again depends on
the type of the gauge boson participating in the interaction:

V1V2V3V4 C
1
W +W +W −W −
(sin
 θW )  2
2

cos θ W
W +W − Z Z −
sin θ W
cos θ W
W +W − γZ −
sin θ W
W +W − γγ −1
4We focus on the interaction vertices only as the propagators of the various fields
were already presented earlier, those of the photon and charged leptons in the section
on QED, those of the massless scalar and quarks among the QCD Feynman rules
and those of the massive gauge bosons in Sect. 22.2. Finally, the denominator in the
propagator of a massive scalar is the same as for the massive gauge boson, while its
numerator coincides with that for the massless scalar. Some authors use opposite
sign convention for the massive gauge bosons as we do (e.g. Ref. [Denner, 1993]).
As a result, their vertices have opposite signs if odd number of massive gauge bosons
are involved, which does not influence values of observable quantities.
 Electroweak sector of the standard model 339

• Cubic Higgs interaction: ieC where

3MH 2
C=− .
2 sin θ W MW
• Quartic Higgs interactions: ie2 C where
2
3MH
C=− .
4(sin θ W )2 MW
2

• Cubic gauge field-Higgs interactions V1,α V2,β H: iegαβ C where C

depends on the type of the gauge boson participating in the interaction:
V1V2 H C
MW
W +W − H
sin θ W
MW
ZZH
sin θ W (cos θ W )2

• Quartic gauge field-Higgs interactions Vα Vβ HH: ie2 gαβ C where C

depends on the type of the gauge boson participating in the interaction:
V1V2 HH C
1
W +W + HH
2(sin θ W )2
1
Z Z HH
2(cos θ W sin θ W )2

• Gauge field-fermion interactions Vα f¯i f j : −ieγα (C − P− +C + P+ ) where

C ± depend on the type of the gauge boson participating in the interac-
tion, the flavour f of fermions and family number i and j:
V f¯i f j C+ C−
γ f¯i f j e f δi j e f δi j
Z f¯i f j g +f δi j g −f δi j
1
W + ūi d j 0 √ Vi j
2 sin θ W
1
W − d¯j ui 0 √ Vi†j
2 sin θ W
1
W + ν̄i j 0 √ δi j
2 sin θ W
1
W − ¯j νi 0 √ δi j
2 sin θ W
340 Chapter 22

where
sin θ W T f3 − sin θ W
2 e
f
g +f = − ef , g −f = .
cos θ W sin θ W cos θ W

• Higgs-fermion interaction H f¯i f j : ieC where

1 m f ,i
C = −δi j .
2 sin θ W MW

The relations between the vector, axial-vector couplings of the Z boson and
the g ±f couplings read as

vf 1 − +
 T f3 − 2(sin θ W )2 e f
= gf + gf = ,
2 sin θ W cos θ W 2 2 sin θ W cos θ W
af 1  T f3
= g −f − g +f = .
2 sin θ W cos θ W 2 2 sin θ W cos θ W

22.8 Neutrino mixing and oscillation

If neutrinos also have masses, mν  0, then a mixing matrix , similar to the
CKM matrix, appears also in the charged currents of leptons. Such a matrix
is called Pontecorvo-Maki-Nakagawa-Sakata (PMNS) matrix.5 According
to the standard notation, instead of Sν used in conjunction of the BEH field,
we write
 
|νm  = U f m |ν f  , |ν f  = U ∗f m |νm  (22.8.15)
f m

where |νm  denotes the mass eigenstates (m = 1, 2, 3), while |ν f  corresponds

to the flavour eigenstates ( f = e, μ, τ). The kets are vectors in the space of
free neutrinos. For anti-neutrinos the mixing reads as
 
| ν̄m  = U ∗f m | ν̄ f  , | ν̄ f  = U f m | ν̄m  . (22.8.16)
f m

The PMNS matrix is a unitary (3 × 3) matrix. In the general case it has

nine independent parameters: three real rotations (just like in a (3 × 3)
orthogonal matrix) and six complex phases. If the neutrinos are Dirac
5Named after pioneering researchers Bruno Pontecorvo, Ziro Maki, Masami
Nakagawa and Soichi Sakata.
 Electroweak sector of the standard model 341

fermions, then five phases can be absorbed into the state vectors of charged
leptons, as for quarks, so only one phase remains. In such cases the usual
parametrization of the PMNS matrix is the same as for the CKM matrix in
(Eq. (22.6.13)), U = V, with c f m = cos θ f m , s f m = sin θ f m and δ  0, is
α21 α31
 neutrinos (ν ≡ ν̄), U = V P
the CP-violating phase. In case of Majorana
α n−1
where P = diag 1, ei 2 , ei 2 , . . . , ei 2 , (n is the number of Majorana
neutrinos that can be different from 3). We do not know the physical
origin of the PMNS matrix. According to experimental observations charged
leptons do not mix, but neutrinos may mix. If they do so, then the matrix
Sν = U PMNS is an analogue of the CKM matrix. In charged currents flavour
eigenstates participate (those are created and annihilated). In high-energy
scattering experiments of particles the masses of neutrinos are negligible.
Our measurements are by far not sufficiently precise to measure any effect
of neutrino masses.
Neutrinos are stable particles due to their small masses and weak inter-
actions with other particles. Thus, they exist in asymptotically free states.
Moreover, according to observations and depending on the question we raise,
we can use a macroscopic, point-like particle approach in describing their
motion. For instance, we can measure the speed of electrons directly using
time-of-flight measurements and thus their masses if we know their ener-
gies. Therefore, it depends on the question which description is the most
convenient: (i) classical, (ii) quantum mechanical, (iii) or quantum field
theoretical. As a matter of fact, neutrinos show classical, quantum and
relativistic features, similarly to electrons.
A characteristic quantum phenomenon is the double-slit interference
with particles. For example, even single photons or electrons produce the
characteristic interference lines if we let sufficiently many pass (one by one)
through the two slits. These “characteristic” interference lines are those
that are also produced by a plane wave of light. Such light is the classical
manifestation of an ensemble of plane wave photons.
Similar quantum mechanical phenomenon is the oscillation of neutrinos,
which is interference in flavour space. If neutrinos have different masses,
then their phases change differently as they travel the distance L between
the source and the detector. As a result flavour interference occurs, called
neutrino oscillation. We do not measure L, the time T needed to travel
the distance L, the energy E and momentum p of each individual neutrino,
but we measure their average values in a neutrino beam. Thus, we use
the plane wave approximation as for classical (beam of) light, |νm ( x, t) =
342 Chapter 22

e−iφm (x,t) |νm (0, 0) where the phase is φm ( x, t) = (Em t − pm · x)/.6 We
assume that the neutrinos move along the x-axis inertially and we use the
abbreviation | pm | ≡ pm . According to the dispersion relation Em 2 = p2 c 2 +
m
mm c . If m1  m2 , then the neutrinos travelling in time T a distance L will
2 4

arrive with different phase,

Δφ12 = (E1T − p1 L) − (E2T − p2 L) = ΔET − ΔpL

  (22.8.17)
ΔE 2 Δp2 L ΔE 2 T Δp2
= T− L= − .
E1 + E2 p1 + p2 2 E L p

The speed of the neutrinos in the laboratory frame V is v = L/T and

p = γmv = E v/c2 (γ(v) = (1 − v2 /c2 )−1/2 ). Thus E/T = pc2 /L. This ratio
is Lorentz-invariant because in a frame V  moving with speed u in V,

E  = γ(u)(E − up) = γ(u)E(1 − uv/c2 ) ,

T  = γ(u)(T − uL/c2 ) = γ(u)T(1 − uv/c2 ) ,

which means that E/T = E /T . As a result, in all inertial reference frames
T/L = E/(pc2 ).
In real experiments it is easier to measure L (instead of T) so we write
the phase difference as
   
L ΔE 2 E Δp2 L Δ(E 2 − pc2 ) Δm12
2 c4
Δφ12 = − = = L.
2 E pc2  p 2c2 p 2c2 p
(22.8.18)
Such a phase difference can lead to a change of neutrino flavour. The
probability of the change ν f → ν f  is
 2
 
 ∗ −iφ 
P(ν f → ν f  ) = |ν f  |ν f (t)| =  U f  mU f n e n νm |νn  
2
 m,n 
 2 (22.8.19)
 
 
=  U f  mU ∗f m e−iφm  .
m 

In order to compute the sum, let us first consider the simplified case of two

6In this section–as opposed to the rest of the book–we use SI instead of natural
units so  and c appears explicitly.
 Electroweak sector of the standard model 343

summands (Δφ = φ1 − φ2 ):

(u1 eiφ1 + u2 eiφ2 )(u1∗ e−iφ1 + u2∗ e−iφ2 ) = |u1 | 2 + |u2 | 2 + u1 u2∗ eiΔφ + u1∗ u2 e−iΔφ
= |u1 | 2 + |u2 | 2 + 2Re(u1 u2∗ ) cos(Δφ) − 2Im(u1 u2∗ ) sin(Δφ)
 
∗ Δφ
= |u1 + u2 | − 4Re(u1 u2 ) sin −
2 2
+ 2Im(u1 u2∗ ) sin(−Δφ) .
2
(22.8.20)

Generalization to n ≥ 2 leads to the usual form of neutrino oscillation

probability,
  
Δφmn
P(ν f → ν f  ) = δ f  f − 4 Re(U ∗f mU f  mU f n U ∗f  n ) sin2
n m>n
2

+2 Im(U ∗f mU f  mU f n U ∗f  n ) sin (Δφmn ) .
n m>n
(22.8.21)

For Dirac neutrinos, if the complex phase in the PMNS matrix vanishes,
δ = 0 (no CP-violation in the neutrino sector), then the PMNS matrix is real
and the last term vanishes in Eq. (22.8.21). In this case the only source of
neutrino oscillation is the second term, in which the square of the sinus is
periodic in π. Therefore, it is useful to rewrite its argument as

Δφmn L
= π (r)
2 Lmn

where
(r) 2hc2 p 2hcE pc
Lmn = = 2 c4 E
Δmmn c
2 4 Δmmn
is the characteristic length of oscillation. In oscillation experiments one can
usually measure the energy of the neutrinos and they travel with the speed
of light in vacuum. (Their energy is more than 1 MeV, while their masses
are smaller than 1 eV.) So it is more convenient to use the second form, in
which pc/E 1 so it is natural to drop it.
(r)
We obtain another form if we substitute Lmn expressed as a function of
the neutrino energy,
Δφmn π Δmmn2 c4
= L.
2 2 hcE
344 Chapter 22

Figure 22.3: Results of the KamLAND experiment compared to the theory

of neutrino oscillation (solid line).

As hc 1.24 · 10−9 eV km, and π/2.48 1.27, then

Δφmn π Δmmn
2 c4 Δmmn
2 c4
GeV L
= L 1.27 ,
2 2 hcE eV2 E km
while the oscillation length is

E or pc eV2 E or pc eV2

L (r) = 2.48 km = 2.48 m .
GeV Δm c 2 4 MeV Δm2 c4
Finally, using natural units, we obtain

Δφmn Δmmn
2
= L.
2 4E or p
Neutrinos can show interference in flavour space because they remain
coherent over long distances due to their weak interactions. Thus, the plane-
wave approximation can be applied. The experiments are performed with
neutrino beams. Oscillation experiments confirm our formula for neutrino
oscillation (see Fig. 22.3) within the uncertainty of the data. Neutrino oscil-
lation was observed in the following experiments:7
7The 2015 Nobel prize in physics was awarded to Profs. Takaaki Kajita and
Arthur B. McDonald for the discovery of neutrino oscillations.
 Electroweak sector of the standard model 345

• the difference between the observed and expected rates of atmospheric

neutrinos in the high-statistics data of the Super-Kamiokande experi-
ment;

• reduced event numbers as compared to expectations in radio-chemical

neutrino experiments (Chlorine, Galley/GNO, SAGE) and number of
observed events depending on time and energy at Super-Kamiokande,
SNO and Borexino;

• decrease of muon neutrinos and increase of electron neutrinos in far-

detector experiments (MINOS, T2K: Tokai to Kamioka, NOνA);

• disappearance of electron anti-neutrinos in not too far and close de-

tector experiments (Daya Bay, Chooz, RENO)

• change of spectra of electron anti-neutrinos in far detector experiment

(KamLAND).

All these experimental observations prove that neutrinos have masses and the
lepton flavour number is not conserved, which calls clearly for explanation
beyond the standard model.
In the near future the most important tasks in the neutrino sector from
the experimental side is to measure the mass of the neutrinos and the pa-
rameters of the PMNS mixing matrix. At present we have (sometimes
contradictory) experimental information only for the sum of the neutrino
masses. From cosmological observations the current conservative upper

bound is 3i=1 mi ≤ 0.23 eV. From oscillation experiments we know two
squared mass differences, δm2 and Δm2 , with the latter being about thirty
times larger than the former. We conclude that there are two mass eigenstates
of approximately equal masses (one of which can even vanish), while the
third one has much different mass. However, we do not know if the very
different mass is much bigger than the other two, called normal hierarchy
(NH), or much smaller than the other two, called inverse hierarchy (IH).
The further parameters of neutrino oscillation are the mixing angles θ i j
and the complex phase δ of the PMNS matrix. In the oscillation formula
sin2 θ i j appears, therefore those are measured directly.
The PMNS matrix, obtained from a global fit of the parameters to data,
(at 99 % confidence level) [Esteban et al., 2017] is the following:

0.800 − 0.844 0.515 − 0.581 0.139 − 0.155

 
|U| =  0.229 − 0.516 0.438 − 0.699 0.614 − 0.790  .
0.249 − 0.528 0.462 − 0.713 0.595 − 0.776
346 Chapter 22

Compare these numbers with the precision of the parameters of the CKM
matrix in (Eq. (22.6.14)). From the experimental point of view the main
difficulty is to improve the precision of the parameters. From the theoretical
point of view the challenge is to understand the origin of the PMNS matrix
(or in popular terms “to understand the origin of neutrino masses”) and why
it is so different from the CKM matrix. We do not know the answer to these
theoretical questions, but we can state with certainty that in order to answer
them, we have to go beyond the standard model of particle interactions.

22.9 Anomaly cancellation

Our final topic is a key feature of the SM, which is generally expected
from any QFT model of particle physics beyond the standard model. We
do not present all details of the computations, but those that are needed to
understand the concept of anomaly cancellation. The rest is left for exercise.
Let us consider a theory with both vector and axial-vector interactions,
such as the standard model. W μ is a gauge field. The current Jμ ∝ ψ̄γμ ψ
is a classically conserved vector current, ∂ μ Jμ = 0. The amplitude for the
process 
p p

Γμ

q
W

is proportional to Wμ Γμ , where Γμ =  p | J μ |p. The matrix element of the

divergence of the current is

 p | ∂ μ Jμ |p = (p − p)μ  p | Jμ |p =  p | q μ Jμ |p = 0.

Thus q μ Jμ = 0, which is the form of the vector Ward identity.

Jμ5 = ψ̄γμ γ5 ψ is a classical axial current for which

iγμ ∂ μ ψ = mψ ⇒ −γμ ∂ μ ψ = i mψ

−i∂μ ψ̄γ μ = mψ̄ ⇒ ∂ μ ψ̄γμ = i mψ̄ .

We multiply the first equation with ψ̄γ 5 from left and the second one with
γ 5 ψ from right and add the two equations. Using the anti-commutation
relation {γ 5, γ μ } = 0, we obtain

∂ μ Jμ5 = 2i mψ̄γ5 ψ ≡ 2m J5 ,
 Electroweak sector of the standard model 347

which vanishes if m = 0, implying conserved axial current at tree level. The

Feynman diagrams that contribute to Γμ and Γμ5 at higher orders are

Γμ
O(e5 ) ,

Γ5μ

· · · + O(e7 ) .

The last diagram8 which contributes to Γμ5 violates the conservation of the
axial current. We define

Tκλμ (p1, p2 ) = Sκλμ (p1, p2 ) + Sλκμ (p2, p1 )

where p1 p2

γκ γλ
Sκλμ (p1, p2 ) = − p1 + p2
q = p1 + p2
γμ γ5

and we expect the following Ward identities

axial: q μ Tκλμ = 0 ,
vector: p1κ Tκλμ = 0 ,
vector: p2κ Tκλμ = 0.

Sκλμ is given by (omitting coupling factors)

Sκλμ (p1, p2 ) ∝ − (−i)3

∫  
d4 
i i i 
× Tr γ μ γ5 γ κ γ λ 
(2π)4 / − /p1 − m / − m / + /p2 − m 
m=0
2    3
∫ / / /
d4 Tr γμ γ5 − /p1 γκ γλ + /p2
=− .
(2π)4 ( − p1 )2 ( + p2 )2 2
8This diagram represents two diagrams: itself and another with one of the fermion
arrows reversed.
348 Chapter 22

Using
2  3    
q μ γμ γ5 = / + /p2 − / − /p1 γ5 = / + /p2 γ5 + γ5 / − /p1

and /p /p = p2 1, we find

∫ ⎡    ⎤
⎢ / − /p1 γκ /γλ γκ /γλ / + /p2 ⎥
μ d4 ⎢ ⎥
q Sκλμ (p1, p2 ) = − Tr ⎢γ5 + γ5 ⎥.
(2π) 4 ⎢ ( − p1 )2 2
( + p2 )2 2 ⎥
⎢ ⎥
⎣ ⎦
We now shift by p1 : → + p1 in the first term and use γλ γ5 = −γ5 γλ
to obtain
⎡    ⎤
∫ ⎢ γλ /γκ / + /p1 γκ /γλ / + /p2 ⎥
d4 ⎢ ⎥
q μ Sκλμ (p1, p2 ) = Tr ⎢γ5 − γ5 ⎥
(2π)4 ⎢⎢ 2 ( + p )2
1
2 ( + p )2 ⎥
2 ⎥
⎣ ⎦
= −q μ Sλκμ .
Thus
q μ Tκλμ (p1, p2 ) = 0
provided the shift is allowed. However, the integral is divergent, so the shift
may be forbidden. For instance, with a cut-off regularization the momentum
shift leads to a finite term which is different from zero. Use of dimensional
regularization looks better as then the momentum shift is allowed. Neverthe-
less, the ambiguity remains also in this case, but disguised in a different form:
in d  4 dimensions γ5 is ambiguous (γ5 is intrinsically 4-dimensional). The
’t Hooft-Veltman prescription says for γ5 = iγ0 γ1 γ2 γ3 in d  4 dimensions
/ 0
γ μ , γ5 = 0 for μ = 0, 1, 2, 3
+ ,
γ μ , γ5 = 0 for other values of μ.

We now recompute q μ Sκλμ (p1, p2 ) with this prescription. Let μ = ˜μ + ˆμ

where ˜μ has μ = 0, 1, 2, 3 components and ˆμ has components in the other
μ μ
d − 4 dimensions. With pi = p̃i (external momenta kept in d = 4) and
       
q μ γμ γ5 = / + /p2 γ5 +γ5 / − /p1 −2γ5 /̂ = / + /p1 γ5 +γ5 / − /p2 −2γ5 /̂ ,

we find
∫ ( )
μ dd / − /p1 / / + /p2
q Tκλμ (p1 + p2 ) ∝ 2 Tr 2γ5 /̂ γκ γλ ,
(2π)d ( − p1 )2 2
( + p2 )2
 Electroweak sector of the standard model 349

which has 5 terms. The only one that does not vanish, after integrations
gives
∫ ˆ2
β dd
4 (−4i) αβκλ p1α p2 =
(2π) d 2(
− p1 )2 ( + p2 )2
β d − 4 (2)
= −16i αβκλ p1α p2 I (p1 + p2 )
d
1
= 2 αβκλ p1κ pλ2 + O().
2π
Dimensional regularization respects the vector Ward identities, but the axial-
vector Ward identity is violated at one loop. There is no regularization
which respects both vector and axial-vector Ward identities simultaneously.
We prefer to keep the vector Ward identities, which is an expression of
insisting on gauge invariance. The axial-vector Ward identity is then violated,
which
- . axial anomaly. It is supported by the observed decay rate
is called
Γ π 0 → γγ = (7.7 ± 0.6) eV. The chiral symmetry forbids π 0 → γγ at tree
level, but the anomaly triangle diagrams predict
   N  2 α 2 m2
c π
Γ π 0 → γγ = = 7.73 eV,
3 64π 3 fπ

hence the anomaly is physical and can be measured. According to the

Adler-Bardeen theorem

axial anomaly receives only 1-loop contributions.

A gauge anomaly destroys the renormalizability of a quantum field theory.

G = SU(2)L × U(1)Y is chiral, so axial anomalies are present and danger-
ous; they should exactly cancel. In SU(2)L × U(1)Y the triangle diagrams
also contain the generators of the gauge group and the total anomaly is
proportional to the trace of

T a1 T a2 T c + T a2 T a1 T c = {T a1 , T a2 } T c

for a given fermion in the loop (T c is the generator at the vertex with γμ γ5 ).
The generators are T a = τ2a or ya with anti-commutation relations
/ 0 / 0
τi, τj = 2δi j , yi, y j = 2yi y j .

Then
Tr ({T a1 , T a2 } τc ) ∝ Tr (τc ) = 0
350 Chapter 22

τa
for both T a = 2 and ya , while

Tr ({T a1 , T a2 } yc ) ∝ δ a1 a2 Q c,

or
Tr ({y a1 , y a2 } yc ) ∝ Q c (Tr (yc ) = Tr (Q c ) ∝ Q c )
Based on these considerations the total contribution
 in a family is propor-
tional to Q e + Qν + Nc (Qu + Qd ) = −1 + 0 + 3 23 − 13 = 0 for left-handed
members. For right-handed members Qν = 0 is absent from the sum. Con-
sequently, the axial anomaly cancels in the standard model. Any theory
beyond the standard model should also have this cancellation if we want a
renormalizable model.
Chapter 23

Outlook

In the theoretical part of this book we emphasized that QCD radiative cor-
rections play an essential role in understanding the structure of the theory,
interpreting data and finding signatures of new physics that cannot be ex-
plained by the standard model. The theory of radiative corrections in the
electroweak part of the standard model is equally important though the ef-
fects are usually less pronounced. As a result, electroweak corrections have
also been computed to many processes that play important role in particle
phenomenology. These computations are technically challenging and the re-
sulting corrections are usually much smaller than those of QCD, simply due
to the large difference in the size of the couplings. Nevertheless, in certain
kinematic regions as well as for interpreting the non-hadronic final states at
the LEP experiments, taking such corrections into account is crucial.
As the LHC collides hadrons, the strong interaction has a role in every
single interaction. As a thumb rule, we can state that the effect of first
order electroweak corrections is similar to that of the second order QCD
corrections. Thus in a computation at NNLO QCD accuracy (O(αs2 ) relative
correction) the first order electroweak corrections (O(α) relative correction)
should also be taken into account. Such computations are hampered by the
mixed QCD and electroweak corrections at O(αs α) relative accuracy.
We put more emphasis on the computation of QCD corrections because
their role of quantum corrections is qualitatively different in QCD than in
the electroweak theory. In general QCD predictions at LO accuracy provide
only an order of magnitude estimate due to the strong dependence on the
unphysical scales. The NLO corrections are crucial in order to decrease the
dependence on these scales. In order to estimate the accuracy of the scale
351
352 Chapter 23

dependence one needs to know the NNLO corrections. As opposed to this,

the electroweak theory gives a fairly good description of data already at LO
accuracy (see for instance the predictions for the masses of vector bosons in
Sect. 22.2). This description is improved by usually very little if we take into
account the radiative corrections.
The relative importance of quantum corrections changes in collisions at
energies larger than that of the LHC. Asymptotic freedom makes QCD cor-
rections relatively smaller. The electroweak corrections may even increase
due to increasing logarithmic contributions log MV /s if s increases, which
has the same numerical effect as decreasing MV . In future experiments it
will be interesting to see if any sign of restoration of the electroweak symme-
try can be observed. In the symmetric phase the electroweak vector bosons
are massless, just like the gluon. Then the electroweak theory will become
similar to QCD, but with an important difference. In QCD ΛQCD is a cut-off,
below which perturbation theory cannot be applied (we use hadronization
models). For the electroweak theory at large energies the perturbation theory
works both below and above the electroweak scale v, but the techniques may
change. At present there is very little known about what we should expect
at future colliders in this respect.
There are many exciting developments and several big unanswered ques-
tions in the current research of particle physics. Among the big questions
we mentioned the origin of neutrino masses. It is also an established exper-
imental fact that most of the energy density in the universe is in a form that
is different from the baryonic matter and electromagnetic radiation, whose
physics we can test in the laboratory. The research to uncover the nature of
dark matter and dark energy is flourishing at the time of writing this book.
Another big question is our very existence, namely that the standard model
predicts equal amount a matter and anti-matter, which is clearly not the case
in the universe. The mechanism of breaking this symmetry is unknown. At
present we do not even know whether the answer to these questions may come
from colliders, or some other fields like cosmology or neutrino experiments.
There are also long standing as well as recently emerging anomalies in
particle physics. The significance of these anomalies have not reached the
discovery level of “five sigma”, yet they are quite pronounced with devi-
ations from standard model predictions in the 3–4 sigma range. Among
the prominent ones we mentioned the long standing difference between the
measured and computed value of the anomalous muon magnetic moment.
Another interesting deviation concerns also the muon, which being the mas-
sive brother of the electron, can build muonic atoms if it is captured by a
proton. Due to the much larger mass the average distance of the muon to
 Outlook 353

the proton is much smaller than that of the electron in the ordinary hydrogen
atom. Thus it is more sensitive to the electric charge distribution inside the
proton. The radius of the proton measured in muonic atoms differs from the
radius measured in ordinary hydrogen.
Between writing and publishing this book several anomalies have been
observed by the LHCb experiment at the LHC. These concern a fundamental
assumption of the standard model that is known as flavour universality in
the lepton sector. The second and third particle families assumed to be exact
copies of the first one with only one exception, namely the particle masses.
Therefore the standard model predicts that processes with identical initial
state and with only difference in the final state in the type of leptons, such
as B0 → K ∗0 μ+ μ− and B0 → K ∗0 e+ e− ) should occur with approximately
equal probability. Quantitatively it means that the standard model predicts

B(B0 → K ∗0 μ+ μ− )
RK ∗0 = 1 (23.0.1)
B(B0 → K ∗0 e+ e− )
for the ratio of the branching fractions (the exact value depends on the value of
the invariant mass squared q2 of the charged lepton pair). Modern particle
detectors can identify the leptons in the final state with high accuracy so
such ratios can be measured fairly well. The results published by the LHCb
collaboration [Aaij et al., 2017]
⎧ +0.11
⎨ 0.66−0.07 (stat) ± 0.03(syst) for 0.045 < q < 1.1 GeV /c
2 4
⎪
⎪
2

RK ∗0 =
⎪
⎪ 0.69+0.11 (stat) ± 0.05(syst) for 1.1 < q2 < 6.0 GeV2 /c4
⎩ −0.07
(23.0.2)
seem to suggest violation of the principle of flavour universality. The ob-
servations currently show deviations from the standard model predictions at
about three sigma and await confirmation by other experiments. If deviations
with more than five sigma significance are confirmed, these will constitute
discoveries that will require non-trivial extension of the standard model,
similarly to the impact of the discovery of parity violation on the extension
of the Fermi theory.
In recent years particle physics has started to employ new techniques
inspired by developments in information technology. The particle detector
prepares a three-dimensional image of the final state measuring the direction,
energy and momentum of particle flow and the bend of particle tracks by
magnetic field. The purpose of data analysis is to find out the hard scattering
process from this information. This is a typical pattern recognition problem.
Engineering of pattern recognition has undergone tremendous development
354 Chapter 23

in this millennium, facilitated by the immense increase of computing power,

appearance of algorithms using neural networks and deep learning. Presently
it is a field of constant change and emerging new ideas, and thus beyond the
scope of this book.
While the items mentioned in this section are very exciting developments
in particle physics, we have deliberately not discussed those in detail, as they
are neither properly confirmed, nor fully established with scientific methods
yet. So instead of going into the details, let us close our book with several
quotes.

There is a theory which states that if ever anyone discovers exactly what the
Universe is for and why it is here, it will instantly disappear and be replaced by
something even more bizarre and inexplicable.
There is another theory which states that this has already happened.

(Douglas Adams)

Why God Particle? Two reasons. One, the publisher wouldn’t let us call it the
Goddamn Particle, though that might be a more appropriate title, given its villainous
nature and the expense it is causing. The title ended up offending two groups: (1)
those who believe in god and (2) those who do not.

(Leon Lederman)

We need something new. We can’t predict what that will be or when we will find
it because if we knew that, we would have found it already!

(Stephen Hawking)

“Would you tell me, please, which way I ought to go from here?”
“That depends a good deal on where you want to get to,” said the Cat.
“I don’t much care where–” said Alice.
“Then it doesn’t matter which way you go,” said the Cat
“–so long as I get SOMEWHERE,” Alice added as an explanation.
“Oh, you’re sure to do that,” said the Cat, “if you only walk long enough.”

(Lewis Carroll)
Bibliography

[Aaboud et al., 2017] Aaboud, M. et al. (2017). Measurement of the Higgs boson
√
coupling properties in the H → Z Z ∗ → 4 decay channel at s = 13 TeV with
the ATLAS detector.
[Aad et al., 2010] Aad, G. et al. (2010). Observation of a centrality-dependent dijet
√
asymmetry in lead-lead collisions at s N N = 2.77 TeV with the ATLAS detector
at the LHC. Phys. Rev. Lett., 105:252303.
[Aad et al., 2012] Aad, G. et al. (2012). Observation of a new particle in the search
for the Standard Model Higgs boson with the ATLAS detector at the LHC.
Phys.Lett., B716:1–29.
[Aad et al., 2015] Aad, G. et al. (2015). Combined measurement of the Higgs boson
√
mass in pp collisions at s = 7 and 8 TeV with the ATLAS and CMS experiments.
Phys. Rev. Lett., 114:191803.
[Aad et al., 2016] Aad, G. et al. (2016). Measurements of the Higgs boson produc-
tion and decay rates and constraints on its couplings from a combined ATLAS and
√
CMS analysis of the LHC pp collision data at s = 7 and 8 TeV. JHEP, 08:045.
[Aaij et al., 2017] Aaij, R. et al. (2017). Test of lepton universality with B0 →
K ∗0 + − decays. JHEP, 08:055.
[Aamodt et al., 2010] Aamodt, K. et al. (2010). Elliptic flow of charged particles in
Pb-Pb collisions at 2.76 TeV. Phys. Rev. Lett., 105:252302.
[Abbiendi et al., 2012] Abbiendi, G. et al. (2012). Search for charged Higgs bosons
√
in e+ e− collisions at s = 189 − 209 GeV. Eur. Phys. J., C72:2076.
[Abbott, 1981] Abbott, L. (1981). The background field method beyond one loop.
Nucl.Phys., B185:189.
[Adams et al., 2003] Adams, J. et al. (2003). Evidence from d + Au measurements
for final state suppression of high p(T) hadrons in Au+Au collisions at RHIC.
Phys. Rev. Lett., 91:072304.
[Adler et al., 2003] Adler, S. S. et al. (2003). Suppressed π 0 production at large
transverse momentum in central Au+ Au collisions at S(NN)**1/2 = 200 GeV.
Phys. Rev. Lett., 91:072301.
355
356 Bibliography

[Agostinelli et al., 2003] Agostinelli, S. et al. (2003). GEANT4: A simulation

toolkit. Nucl. Instrum. Meth., A506:250–303.
[Altarelli and Parisi, 1977] Altarelli, G. and Parisi, G. (1977). Asymptotic freedom
in parton language. Nucl.Phys., B126:298.
[An et al., 2017] An, F. P. et al. (2017). Measurement of electron antineutrino
oscillation based on 1230 days of operation of the Daya Bay experiment. Phys.
Rev., D95(7):072006.
[Arneodo et al., 1997] Arneodo, M. et al. (1997). Measurement of the proton and
deuteron structure functions, F2(p) and F2(d), and of the ratio sigma-L / sigma-T.
Nucl.Phys., B483:3–43.
[ATLAS, 2017] ATLAS, c. (2017). Measurements of Higgs boson properties in
the diphoton decay channel with 36.1 fb−1 pp collision data at the center-of-mass
energy of 13 TeV with the ATLAS detector.
[Aubert et al., 1974] Aubert, J. et al. (1974). Experimental observation of a heavy
particle J. Phys.Rev.Lett., 33:1404–1406.
[Augustin et al., 1974] Augustin, J. et al. (1974). Discovery of a narrow resonance
in e+ e- annihilation. Phys.Rev.Lett., 33:1406–1408.
[Barate et al., 1998] Barate, R. et al. (1998). Studies of quantum chromodynamics
with the ALEPH detector. Phys.Rept., 294:1–165.
[Barate et al., 2003] Barate, R. et al. (2003). Search for the standard model Higgs
boson at LEP. Phys.Lett., B565:61–75.
[Bazavov et al., 2018] Bazavov, A. et al. (2018). Up-, down-, strange-, charm-, and
bottom-quark masses from four-flavor lattice QCD.
[Becker et al., 2012] Becker, S., Goetz, D., Reuschle, C., Schwan, C., and Weinzierl,
S. (2012). NLO results for five, six and seven jets in electron-positron annihilation.
Phys.Rev.Lett., 108:032005.
[Berger, 1992] Berger, E. L. (1992). Research directions for the decade. Proceedings,
1990, Summer Study on High-Energy Physics, Snowmass, USA, June 25 - July
13, 1990.
[Beringer et al., 2012] Beringer, J. et al. (2012). Review of particle physics (RPP).
Phys.Rev., D86:010001.
[Bezrukov and Shaposhnikov, 2008] Bezrukov, F. L. and Shaposhnikov, M. (2008).
The Standard Model Higgs boson as the inflaton. Phys. Lett., B659:703–706.
[Bityukov and Krasnikov, 2000] Bityukov, S. and Krasnikov, N. (2000). On the
observability of a signal above background. Nucl.Instrum.Meth., A452:518–524.
[Branchina et al., 2014] Branchina, V., Messina, E., and Platania, A. (2014). Top
mass determination, Higgs inflation, and vacuum stability. JHEP, 09:182.
[Busza et al., 2018] Busza, W., Rajagopal, K., and van der Schee, W. (2018). Heavy
ion collisions: The big picture and the big questions.
 Bibliography 357

[Cacciari and Salam, 2006] Cacciari, M. and Salam, G. P. (2006). Dispelling the
N 3 myth for the k t jet-finder. Phys.Lett., B641:57–61.
[Cacciari et al., 2008] Cacciari, M., Salam, G. P., and Soyez, G. (2008). The anti-k(t)
jet clustering algorithm. JHEP, 0804:063.
[Catani et al., 2002] Catani, S., Dittmaier, S., Seymour, M. H., and Trocsanyi, Z.
(2002). The dipole formalism for next-to-leading order QCD calculations with
massive partons. Nucl.Phys., B627:189–265.
[Catani et al., 1991] Catani, S., Dokshitzer, Y. L., Olsson, M., Turnock, G., and
Webber, B. (1991). New clustering algorithm for multi - jet cross-sections in e+
e- annihilation. Phys.Lett., B269:432–438.
[Chatrchyan et al., 2012a] Chatrchyan, S. et al. (2012a). Jet momentum dependence
√
of jet quenching in PbPb collisions at s N N = 2.76 TeV. Phys. Lett., B712:176–
197.
[Chatrchyan et al., 2012b] Chatrchyan, S. et al. (2012b). Observation of a new boson
at a mass of 125 GeV with the CMS experiment at the LHC. Phys.Lett., B716:30–
61.
[Chatrchyan et al., 2013] Chatrchyan, S. et al. (2013). Observation of a new boson
√
with mass near 125 GeV in pp collisions at s = 7 and 8 TeV. JHEP, 1306:081.
[Cohen et al., 1998] Cohen, A. G., De Rujula, A., and Glashow, S. (1998). A matter
– antimatter universe? Astrophys.J., 495:539–549.
[Collins et al., 1989] Collins, P., Martin, A. D., and Squires, E. (1989). Particle
physics and cosmology.
[Corcella et al., 2001] Corcella, G., Knowles, I., Marchesini, G., Moretti, S., Oda-
giri, K., et al. (2001). HERWIG 6: An event generator for hadron emission
reactions with interfering gluons (including supersymmetric processes). JHEP,
0101:010.
[Cousins and Highland, 1992] Cousins, R. D. and Highland, V. L. (1992). In-
corporating systematic uncertainties into an upper limit. Nucl.Instrum.Meth.,
A320:331–335.
[Cowan et al., 2011] Cowan, G., Cranmer, K., Gross, E., and Vitells, O. (2011).
Asymptotic formulae for likelihood-based tests of new physics. Eur. Phys. J.,
C71:1554. [Erratum: Eur. Phys. J.C73,2501(2013)].
[Denner, 1993] Denner, A. (1993). Techniques for calculation of electroweak ra-
diative corrections at the one loop level and results for W physics at LEP-200.
Fortsch. Phys., 41:307–420.
[Dirac, 1931] Dirac, P. A. (1931). Quantized singularities in the electromagnetic
field. Proc.Roy.Soc.Lond., A133:60–72.
[Dokshitzer, 1977] Dokshitzer, Y. L. (1977). Calculation of the structure functions
for deep inelastic scattering and e+ e- annihilation by perturbation theory in
quantum chromodynamics. Sov.Phys.JETP, 46:641–653.
358 Bibliography

[Dokshitzer et al., 1991] Dokshitzer, Y. L., Khoze, V. A., Mueller, A. H., and Troyan,
S. I. (1991). Basics of perturbative QCD.
[Durr et al., 2008] Durr, S., Fodor, Z., Frison, J., Hoelbling, C., Hoffmann, R., et al.
(2008). Ab-Initio determination of light hadron masses. Science, 322:1224–1227.
[Ellis, 1988] Ellis, R. K. (1988). An introduction to the QCD parton model.
[Esteban et al., 2017] Esteban, I., Gonzalez-Garcia, M. C., Maltoni, M., Martinez-
Soler, I., and Schwetz, T. (2017). Updated fit to three neutrino mixing: Exploring
the accelerator-reactor complementarity. JHEP, 01:087.
[Gehrmann-De Ridder et al., 2007] Gehrmann-De Ridder, A., Gehrmann, T.,
Glover, E., and Heinrich, G. (2007). NNLO corrections to event shapes in e+
e- annihilation. JHEP, 0712:094.
[Glashow et al., 1970] Glashow, S., Iliopoulos, J., and Maiani, L. (1970). Weak
interactions with lepton-hadron symmetry. Phys.Rev., D2:1285–1292.
[Gribov and Lipatov, 1972] Gribov, V. and Lipatov, L. (1972). Deep inelastic e p
scattering in perturbation theory. Sov.J.Nucl.Phys., 15:438–450.
[Halzen and Martin, 1984] Halzen, F. and Martin, A. D. (1984). Quarks and leptons:
An introductory course in modern particle physics.
[Horvath, 2003] Horvath, D. (2003). Search for charged Higgs bosons with the
OPAL detector at LEP. Nucl. Phys., A721:C453–C456.
[James and Roos, 1975] James, F. and Roos, M. (1975). Minuit: A system for func-
tion minimization and analysis of the parameter errors and correlations. Com-
put.Phys.Commun., 10:343–367.
[Klein and Roodman, 2005] Klein, J. and Roodman, A. (2005). Blind analysis in
nuclear and particle physics. Ann.Rev.Nucl.Part.Sci., 55:141–163.
[Lee and Wick, 1974] Lee, T. D. and Wick, G. C. (1974). Vacuum stability and
vacuum excitation in a spin 0 field theory. Phys. Rev., D9:2291–2316.
[Leo, 1987] Leo, W. (1987). Techniques for nuclear and particle physics experi-
ments: a how to approach.
[Mangano and Parke, 1991] Mangano, M. L. and Parke, S. J. (1991). Multiparton
amplitudes in gauge theories. Phys.Rept., 200:301–367.
[Martin, 1997] Martin, S. P. (1997). A supersymmetry primer. pages 1–98. [Adv.
Ser. Direct. High Energy Phys.18,1(1998)].
[Miller et al., 1972] Miller, G., Bloom, E. D., Buschhorn, G., Coward, D., DeStae-
bler, H., et al. (1972). Inelastic electron-proton scattering at large momentum
transfers. Phys.Rev., D5:528.
[Mohr et al., 2016] Mohr, P. J., Newell, D. B., and Taylor, B. N. (2016). CODATA
recommended values of the fundamental physical constants: 2014. Rev. Mod.
Phys., 88(3):035009.
 Bibliography 359

[Nagy and Trocsanyi, 1999] Nagy, Z. and Trocsanyi, Z. (1999). Multijet rates in
e+ e- annihilation: Perturbation theory versus LEP data. Nucl.Phys.Proc.Suppl.,
74:44–48.
[Patrignani et al., 2016] Patrignani, C. et al. (2016). Review of particle physics.
Chin. Phys., C40(10):100001.
[Perkins, 1982] Perkins, D. (1982). Introduction to high energy physics.
[Prosper and Lyons, 2011] Prosper, H. B. and Lyons, L. (2011). Proceedings of
the PHYSTAT 2011 Workshop on statistical issues related to discovery claims in
search experiments and unfolding, CERN, Geneva, Switzerland, 17-20 January
2011.
[Pumplin et al., 2002] Pumplin, J., Stump, D., Huston, J., Lai, H., Nadolsky, P. M.,
et al. (2002). New generation of parton distributions with uncertainties from
global QCD analysis. JHEP, 0207:012.
[Ryder, 1996] Ryder, L. (1996). Quantum field theory.
[Schael et al., 2013] Schael, S. et al. (2013). Electroweak measurements in electron-
positron collisions at W-boson-pair energies at LEP. Phys.Rept., 532:119–244.
[Shaposhnikov and Wetterich, 2010] Shaposhnikov, M. and Wetterich, C. (2010).
Asymptotic safety of gravity and the Higgs boson mass. Phys. Lett., B683:196–
200.
[Sirunyan et al., 2017] Sirunyan, A. M. et al. (2017). Measurements of properties
of the Higgs boson decaying into the four-lepton final state in pp collisions at
√
s = 13 TeV. JHEP, 11:047.
[Sjostrand et al., 2006] Sjostrand, T., Mrenna, S., and Skands, P. Z. (2006). PYTHIA
6.4 physics and manual. JHEP, 0605:026.
[Skands, 2012] Skands, P. (2012). Introduction to QCD. arXiv.org, 1207.2389:1–85.
[Turner and Wilczek, 1982] Turner, M. S. and Wilczek, F. (1982). Might our vacuum
be metastable? Nature, 298:633–634.
[Weinzierl, 2009] Weinzierl, S. (2009). Event shapes and jet rates in electron-
positron annihilation at NNLO. JHEP, 0906:041.
[Woithe et al., 2017] Woithe, J., Wiener, G. J., and Van der Veken, F. F. (2017).
Let’s have a coffee with the Standard Model of particle physics! Phys. Educ.,
52(3):034001.
Index

Bold-face page numbers indicate sections dedicated to the given entry, whereas
page numbers in normal type point to textual references. In order to maintain an
alphabetic order, Greek symbols and numbers in entries are spelled out in Latin (e.g.
“τ-lepton” is listed as “Tau-lepton” or “3-jet events” as “Three-jet events”).

Abelian group, 215 Anti-unitary operator, 10

Adjoint representation, 221, 228, Antihydrogen, 11, 70
230, 231, 235, 269 Antimatter, 7, 11, 70, 155
Adjoint spinor, 33, 41 Antineutrino, 134, 143, 144
ALEPH experiment, 150, 192, 193, Antiparticle, 7–10, 22, 32, 70, 109,
204, 285, 296 111, 118, 126, 128, 133,
ALICE experiment, 69, 79, 137, 306
161–163, 165, 179, 181 Antiproton, 66, 70, 89
Altarelli-Parisi splitting function, Antiproton annihilation, see
305 Annihilation
Altarelli-Parisi splitting kernel, 291 Antiproton Decelerator (AD), 11,
70, 71, 180
Angular momentum, 4, 9, 22, 216
Associated production, 150
operator, 21, 230
Asymmetry, 8, 130, 147, 200, 210
spin, 6
Asymptotic freedom, 49, 165, 211,
Annihilation, 242, 248 226, 258, 261, 317
electron–positron, 7, 10, 45, ATLAS experiment, 74, 77, 81, 151,
46, 86, 134, 170, 184, 165, 179, 180, 195–197,
204, 234, 247, 285, 287, 330
314 Atmospheric neutrinos, 134,
proton–antiproton, 7 136–138, 140, 143, 345
Anomalous dimension, 203, 260, Axial current, 346, 347
262, 309, 311 Axial-vector coupling, 329, 340, 346
Anomalous magnetic moment,
120–122, 352 b quark, 50, 89, 111, 150, 151, 192,
Anomaly, 346 248
361
362 Index

Baryon, 13, 20, 25 heavy ion, 67

decuplet, 24, 26 proton-antiproton, 63, 66
number, 13, 225, 237, 320 Colour
octet, 24, 26 charge, see Charge, colour
Beam cooling, 62 confinement, 226
BEH field, 51, 52, 155, 156, 210, factor, 228
328, 337, 340 SU(3), 20
BEH mechanism, 52, 54, 324 Compton scattering, 45
Beta decay, 118, 134, 137, 171 Confinement of quarks, 20, 49, 233
Bethe-Bloch formula, 75, 76 Conservation law, 4, 138
Bhabha scattering, 44, 46, 184 Conserved current, 35, 41, 289, 345,
Big Bang theory, 59, 156, 159 346
Bilinear covariant, 33, 218, 219 Continuity equation, 35
Boson, 7 Contravariant vector, 31, 32
Breit-Wigner function, 17, 99, 176 Correlation matrix, 94
Bremsstrahlung, 77, 150, 160 Coulomb potential, 9, 42, 48
Bubble chamber, 79, 126 Coulomb scattering, 63, 74, 120
Coupling,
c quark, 24, 128, 248, 331, 332 electromagnetic, 52, 318, 337
Cabibbo angle, 328 running, 48, 255, 258, 260–262
Callan-Gross relation, 301, 302 strong, 49, 248, 288, 293, 307,
Casimir operator, 230 318
Charge, Covariance matrix, 94, 98
colour, 19–21, 45, 47, 87, 111, Covariant derivative, 41, 42, 218,
113, 225, 229, 230, 248, 320, 326
269, 288, 317 Covariant gauge, 219, 226, 249
conservation, 4 Covariant vector, 31
electric, xiv, 4, 26, 40, 110, 218 CP eigenstates, 127, 129
renormalization, 253 CP invariance, 11, 125, 128–130,
screening, 21 210
Charged current, 33, 52, 133, 302, CP violation, 129, 131, 210, 336
321, 329, 331, 340 CPT invariance, 10, 11, 210
Charm quark, 24, 332 Cross section, 15, 36, 45, 62, 86,
Chiral symmetry, 160, 236, 240, 349 110, 114, 151, 154, 176,
CKM matrix, 335, 338, 341 194, 221, 233, 243, 247,
Clifford algebra, 217, 236 270, 271, 273, 281, 287,
CMS experiment, 74, 80, 151, 293, 294, 298, 299, 301,
179–183, 194, 196, 197, 303, 313, 314
200, 330 Crossing, 44, 134, 242, 243, 271,
CNGS experiment, 70, 142 306
Collider, 62 Current conservation, see Conserved
electron-positron, 63, 177 current
hadron, 67 Current-current interaction, 42
hadron-electron, 63, 67 Cut diagram, 222, 274, 275, 277
 Index 363

Cutkosky rules, 222 Factorization theorem, 313

Cyclotron, 61, 169, 200 Fermi constant, 328, 337
Cyclotron frequency, 121, 122 Fermion, 7, 20, 33
charge (number), 4, 6, 35, 40
Daya Bay experiment, 141, 190 family, 7, 26, 109, 110, 334,
Decay rate, 17, 148, 349 337
Decay width, 17 propagator, 242, 264
Deeply inelastic scattering (DIS), Feynman diagram, 43, 44, 222, 242,
205, 297, 313, 317 253, 315, 347
DGLAP equations, 308, 310, 314 Feynman gauge, 219, 249
Dipole magnet, 61, 176, 177, 179 Feynman rule, 218, 220, 222, 227,
Dirac equation, 6, 9, 31, 41, 48, 120, 229, 232, 238, 241, 264,
240, 288 315, 337, 338
Dirac spinor, 32, 235, 240, 288 Fierz identity, 231, 244
Divergences, 264, 281, 281, 282, Fine structure constant, xiv, 45
293, 304–306, 314, 316, Flavour (of fermions), 13, 22, 24, 37,
317 110, 126, 127, 137, 185,
Drift chamber, 79, 86 224, 225, 235, 248, 262,
269, 292, 306, 307, 320,
Electromagnetic 331, 334, 340, 341, 353
calorimeter, 74, 79, 82, 86, Four-jet event, 87, 103, 150
152, 182, 184, 185, 194 Four-momentum, 221, 250, 264
current, 43, 52 Four-vector, 33
field, 42, 60, 322 Fragmentation, 21, 86
interaction, 6, 16, 39, 40, 42, function, 313, 314
112, 244, 317 Free fermion, 34
potential, 41, 52 Fundamental representation, 225,
Electron 229, 230, 235, 248, 269
cooling, 62, 178
propagator, 220 g factor, 120
radiation loss, 76 Gamma function, 250
Electron-photon shower, 77 Gamma matrices, 32
Electron-positron annihilation, see Gauge
Annihilation anomaly, 349
Electron-positron collider, see boson, 49–51, 53, 222, 226,
Collider, 321, 328, 329, 334, 338,
electron-positron 339
Electroweak interaction, 50, 338 field, 41, 48, 52, 210, 227, 320,
Electroweak Lagrangian, 54 326, 338
Euler constant, 252 fixing, 219, 226, 252, 253, 326
Euler-Lagrange equation, 33 invariance, 3, 40, 48, 215, 235,
Event registration, 85 239, 274, 289, 317, 326,
Event selection, 92 349
Exclusion principle, 7, 19, 20 theory, 165, 216
364 Index

transformation, 6, 217, 320 observation, 53, 151, 152, 153,

Gauge symmetry, see also Gauge 196
invariance production, 151, 194
Gauge symmetry, search for, 64, 148
generator, 47 Higgs field, see BEH field
global, 41 Higgs mechanism, see BEH
local, 41, 216 mechanism
violation, 41, 51, 53 Higgs potential, see BEH potential
Gell-Mann matrix, 47, 225 Higher order, 48, 233, 248, 293, 347
Ghost field, 226 Hypercharge, 13, 22, 52, 144, 210,
GIM mechanism, 26, 331, 331 215, 319, 320
Gluon, 40, 50, 304, 316
distributions (in proton), 206 ICARUS experiment, 142
field, 227, 236 Inflaton, 156
fusion, 50, 68, 151, 194 Infrared divergences, 263, 281, 287,
jet, 50, 87, 111, 185 295, 304, 306
production, 187 Infrared safe, 279, 282–286
propagator, 227 Intrinsic angular momentum, see
soft, 281–283, 288 Spin
splitting, 291, 305 Isospin, 11, 23, 27, 35, 237, 302
vertex, 228, 231, 288 of nucleons, 35
Goldstone boson, 237, 325 weak, 26–28, 52
Gyromagnetic ratio, see g factor
Jets, see also Hadron jet
Hadron, 12, 21, 49, 159, 233, 298, algorithm, 103, 283
313 gluon, see Gluon jet
calorimeter, 74, 80, 82, 86, in electron-positron collisions,
183, 185 49, 315
collision, 50, 87, 285, 313, 313 quark, see Quark jet
jet, 49, 50, 109, 111, 185, 187,
192 KamLAND experiment, 344, 345
mass, 160 Kaon, 125, 126, 131, 336
shower, see Hadron jet formation, 126
therapy, 172 lifetime, 127
Helicity, 245 neutral, 126, 129
conservation, 239, 244, 263 oscillation, 127, 129
formalism, 239, 242, 268, 271 regeneration, 128
HERA collider, 67, 297 KATRIN experiment, 135, 188
Higgs boson, 147, 333 Klein-Gordon equation, 34, 42
decay, 192 Kobayashi-Maskawa matrix, see
exclusion, 150, 193 CKM matrix
interaction, 326, 339–340
mass, 149, 152, 156, 198, 326, Lagrangian, 33, 34
330 of Dirac equation, 33, 217, 321
 Index 365

of Klein-Gordon equation, 42 Massive gauge boson, 327–329, 338

of Maxwell’s equations, 41 Meson nonet, 23
of QED, 41, 42 Mesons
of standard model, 54, 320–323 lightest, 7, 22, 112, 126
Lambda matrices for SU(3), see masses, 112, 160
Gell-Mann matrix pseudo-scalar, 24, 237
Larmor precession, 121 vector, 23, 24, 112, 112
Left-handed coordinates, 8 Metric tensor, 32, 43, 217, 266
Left-handed fermions, 52, 118, 125, MiniBooNE experiment, 143
210, 215, 236, 319 Missing energy (momentum), 73,
LEP (Large Electron Positron) 86, 88
collider, 63, 64, 177, 193, Mixing
233 angle, 23, 52, 110, 140, 141
LHC (Large Hadron Collider), 59, matrix, 140, 340
61, 67, 177–179 of kaons, 130
LHCb experiment, 69, 81, 113, 179, of neutrinos, 140, 340
353 weak (Weinberg), see Weak
Lifetime, 40, 89, 120–123, 170 mixing (angle)
Lifetime and line width, 17, 148 Momentum transfer, 43, 211, 297,
Linear accelerator (linac), 61, 170, 299
200, 201 Multiplet, 324
Local gauge symmetry, see Gauge baryon, 26
symmetry, local Multiplicity, 162, 163, 165, 238, 316
Lorentz curve, 17, 148 Multiwire proportional chamber
Lorentz invariance, 44, 88, 218, 342 (MWPC), 79, 122, 170
LSND experiment, 142–144 Muon
g − 2, see Anomalous magnetic
Møller scattering, 44, 46 moment
Magic momentum (of muon), 122 decay, 136
Magnetic detection, 138, 139
field, see Electromagnetic field polarization, 120
horn, 70 precession, 120
Magnetic moment, 37 production, 45, 112, 136, 142
anomalous, see Anomalous spin rotation, 119
magnetic moment storage ring, 121, 123
of muons, 120
Magnetic rigidity, 60 NA48 experiment, 130
Majorana particle, 137, 144, 341 Natural units, xiii, 17, 216, 321, 342,
Mandelstam variables, 44, 223, 243 344
Mass creation, 147 Neutral current, 52, 53, 110, 133,
Mass eigenstates, 23, 28, 137, 140, 134, 139, 302, 322, 329,
334, 335, 340 331, 332, 335
Mass term, 41, 321, 326, 332 Neutral kaon
Massive fermion, 41, 332, 337 CP violation, see CP violation
366 Index

K-long, K-short, 130 Parton model, 205, 297–307

oscillation, see Kaon Pauli matrices, 32, 47, 320
oscillation Pauli principle, see Exclusion
regeneration, see Kaon principle
regeneration Perturbation theory, 233, 237, 251,
Neutrino, 70, 133 256, 258, 260, 261, 282,
beam, 70, 134, 142, 341, 344 284, 286, 304, 308, 314,
flavour, 27, 335, 342 317, 321, 329, 352
mass, 135, 138–141, 144, 209, Perturbative QCD, 203, 247, 297,
335, 340, 341, 352 303–306, 310
mysteries, 135, 144 Phase invariance, 4, 217
oscillation, 70, 137, 139–141, Phase space, 222, 274, 277, 281,
189, 209, 335, 340, 341, 294, 301, 316
343 PHENIX experiment, 67, 164, 199
sterile, 142, 322 Photo-multiplier, 79, 81, 138
Neutron, 12, 170, 302 Photon, 39, 43
decay, 35, 50, 134 propagator, 43, 218, 220
magnetic moment, 37 virtual, 11, 120, 297, 299, 301,
therapy, 172 305
Noether’s theorem, 4, 35, 217 Pion, 12, 13, 22
Non-abelian group, 216, 226, 238, decay, 10, 40, 118, 119, 125,
317 126, 136, 142
Nuclear forces, 47, 47
scattering, 112, 127
Nuclear mass, 24, 36
Planck mass, 156
Nucleon, 12
PMNS mixing matrix, 140, 144,
quark composition, 37
340–346
Polarization, 119, 120, 137, 219, 271
OPAL experiment, 81, 86, 103,
state, 273
184–187
OPERA experiment, 142 vector, 220, 239
Positron annihilation, see
Pair annihilation, see Annihilation Annihilation
Pair production, 45, 64, 111 Positronium, 49
Parity, 8 Propagator
conservation, 9, 50, 117, 119 electron, see Electron
intrinsic, 9 propagator
operator, 8 gluon, see Gluon propagator
transformation, 234, 243 photon, see Photon propagator
violation, 117, 118, 119, 125, Proton, 11, 12, 37–38
353 acceleration, 65, 68
Parton, 111, 161, 233, 281, 289–293, radius, 47, 353
313 scattering, 204, 207, 211, 315
Parton distribution function, 303, structure, 67, 297, 302
314 therapy, see Hadron therapy
 Index 367

Pseudo-scalar mesons, see Mesons, group equation, 255, 262,

pseudo-scalar 308–310
Pseudorapidity, 87, 88, 160, 186, 284 scale, 253, 296, 307, 308
Representation, 6, 47
Quadrupole magnet, 60, 176, 178 adjoint, see Adjoint
Quantum chromodynamics (QCD), representation
224 fundamental, see Fundamental
corrections, 248, 293–295, representation
351, 352, see also Virtual Resonance, 17, 114, 148, 176, 233,
corrections see also Breit-Wigner
Feynman rules, 227, 338 function
vacuum, 338 acceleration, 61
Quantum electrodynamics (QED), Right-handed coordinates, 8
42, 216 Right-handed fields, 236
gauge invariance, 274 Running coupling, see Coupling,
running
Quantum field theory, 11, 233, 321
Running mass, 260–262
Quark
couplings, 329
Scalar particle, 34, 51, 210
gluon plasma, 159
Scaling
masses, 225, 237, 260, 262,
variable, 298, 299
269
violation, 237, 255, 311, 317
mixing, 328, 334, see also
Schouten identity, 244
CKM matrix
Scintillation
model, xii, 12, 19, 28
counter, 79
structure functions, see
tiles, 182
Structure functions
Scintillator, 80, 141, 182
Sea quark, 205
Radiation energy loss, 65, 76 Secondary particles, 80, 81, 88, 134,
Radiation length, 76, 77 183
Radiative corrections, 66, 264, 303, Secondary vertex, 89, 111, 150, 192
315, 351, see also Solar Neutrino Unit (SNU), 135
Quantum Spaghetti diagrams, 150, 193
chromodynamics (QCD) Spinor
corrections two-component, 230
Radio-frequency cavity, 61, 179 Spinor, adjoint, see Adjoint spinor
Rapidity, 88, 285, see also Spontaneous symmetry breaking,
Pseudorapidity 51, 52, 147, 210, 237,
Reactor neutrinos, 134, 141, 141 324, 334
Regularization, 249, 253, 263, 264, Standard solar model, 136, 139
281, 293, 305, 348, 349 STAR experiment, 67, 164, 199
Renormalization, 248, 253, 254, Stochastic cooling, 62, 178
314, 317 Storage ring, 61–63, 65, 71, 121
factors, 252, 253 Strange quark, 13
368 Index

Strangeness, 11, 13, 320 expectation value, 51, 237,

Strong coupling, see Coupling, 328, 337
strong polarization, 21
Structure constants, 216, 230, see QCD, see Quantum
also Fine structure chromodynamics (QCD)
constant vacuum
Structure functions, 205, 299–302, stability, 155, 198
305–308, 312, 317 Valence quark, 37, 38, 205, 302, 311
SU(2) symmetry, 6, 32, 35, 37, 41, Vector field, 34, 41, 217, 225, 321,
51, 216 326
SU(3) symmetry, 21, 23, 47, 48, 53, Vector mesons, see Mesons, vector
216, see also Colour Vector potential, 42, 60, see also
SU(3) Electromagnetic potential
Sudbury Neutrino Observatory Vertex
(SNO) experiment, 139, correction, 264, 268, 269, 275
190 detector, 86, 163
Super-Kamiokande (SKK) factor, 252, 253
experiment, 138, 138, gluon, see Gluon vertex
139, 189 secondary, see Secondary
Supersymmetry, 155, 235 vertex
Synchrotron, 61, 172, 177 Virtual corrections, 127, 270, 273,
Synchrotron radiation, 65, 67 275, 287, 293–295, 304,
305
Tau lepton, 26, 86, 141, 185
Virtual particles, 11, 39
Tau neutrino, 139
Tau-theta paradox, 117
Ward identity, 244, 346, 347, 349
Three-jet event, 50, 87, 111, 185,
Weak bosons (W, Z), 39, 51, 53, 129
187
Weak current, 133
Thrust, 203, 282, 283, 286, 287, 295
charged, see Charged current
Time reversal, 10, 11, 125, 210, 234
neutral, see Neutral current
Top (t) quark, 63, 89, 155, 186, 198,
Weak decay, 40, 118, 119, 126, 127,
225, 248
see also Beta decay
Transition rate, 15, 42, 282
Weak hypercharge, see Hypercharge
Transverse momentum, 63, 87, 88,
Weak isospin, see Isospin, weak
160, 165, 285, 304
Weak mixing (angle), 52, 319, 322
Two-jet event, 86, 164, 165, 185
U(1) symmetry, 6, 35, 40, 215, 324 Yukawa potential, 47
Uncertainty principle, 11, 43, 120 Yukawa term, 332, 334
Unitary gauge, 327, 333, 338
Universality (flavour), 110, 353 Z boson, 63
decay, 86, 109, 110, 147, 150,
V-A interaction, 33, 328 185
Vacuum, xiv, 216, 325, 326 production, 109
electroweak (BEH), 51, 155, width, 109
198