Download as pdf or txt
Download as pdf or txt
You are on page 1of 190

Università degli Studi di Padova

Dipartimento di Fisica e Astronomia ”Galileo Galilei”

STANDARD MODEL

Lecture Notes

1st 2nd 3rd generation Goldstone outside


standard matter unstable matter force carriers bosons standard model

2.3 MeV 2/3 1.28 GeV 2/3 173.2 GeV 2/3 charge strong nuclear force (color) 125.1 GeV
R

colors
u c
/G

/G

/G

t h
/B

/B

/B
(+6 anti-quarks)

mass
6 quarks

up 1/2 charm 1/2 top 1/2 spin Higgs


electromagnetic force (charge) 0

4.8 MeV −1/3 95 MeV −1/3 4.7 GeV −1/3

g
C
R

s
O
/G

/G

/G

d b
LO
/B

/B

/B

down 1/2 strange 1/2 bottom 1/2 gluon 1


weak nuclear force (weak isospin)

511 keV −1 105.7 MeV −1 1.777 GeV −1

gravitational force (mass)


e µ τ γ
(+6 anti-leptons)
6 leptons

electron 1/2 muon 1/2 tau 1/2 photon 1

< 2 eV < 190 keV < 18.2 MeV 80.4 GeV ±1 91.2 GeV

νe νµ ντ W± Z
e neutrino 1/2 µ neutrino 1/2 τ neutrino 1/2 1 1 graviton

12 fermions 5 bosons
(+12 anti-fermions) (+1 opposite charge W )
increasing mass →

Lectures by Written by
Paride Paradisi Alessandro Lenoci

Last Update: June 30, 2020


2
Contents

1 The Standard Model Lagrangian 7


1.1 The bosonic sector . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8
1.2 The fermionic sector . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 10
1.3 The CKM matrix . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 12

2 Discrete symmetries 15
2.1 C, P and T symmetries . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 15
2.2 C, P and T in the Standard Model . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 16
2.3 CPT theorem . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 19
2.4 Sakharov conditions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 20

3 Determination of the CKM matrix 23


3.1 Chiral symmetries of QCD . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 23
3.2 |Vud | from superallowed β-decays . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 26
3.3 |Vus | from semileptonic kaon decays . . . . . . . . . . . . . . . . . . . . . . . . . . . . 28
3.4 |Vcb | and |Vub | from semileptonic B meson decays . . . . . . . . . . . . . . . . . . . . . 28
3.5 |Vtd | and |Vts | from B-B oscillation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 29
3.6 Unitarity triangle and CP violation . . . . . . . . . . . . . . . . . . . . . . . . . . . . 29
3.7 Leptonic pion decay π → `ν` . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 33

4 Neutrino physics 37
4.1 Dirac and Majorana masses . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 37
4.2 The see-saw mechanism . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 39
4.3 The PMNS matrix . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 40
4.4 The Standard Model flavor group . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 41
4.5 Neutrino oscillation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 42
4.5.1 Oscillation in vacuum . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 42
4.5.2 CP violation in neutrino oscillation . . . . . . . . . . . . . . . . . . . . . . . . . 45
4.5.3 Neutrino oscillation and the uncertainty principle . . . . . . . . . . . . . . . . . 46
4.5.4 Oscillation in matter and the MSW effect . . . . . . . . . . . . . . . . . . . . . 47
4.5.5 Atmospheric neutrinos . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 55
4.5.6 Determination of Ue3 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 57
4.6 Kinematical Tests of Neutrino Masses . . . . . . . . . . . . . . . . . . . . . . . . . . . 59
4.6.1 Beta decay . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 60
4.6.2 Neutrinoless double beta decay . . . . . . . . . . . . . . . . . . . . . . . . . . . 62
4.6.3 Charged Lepton Flavor Violation . . . . . . . . . . . . . . . . . . . . . . . . . . 64

5 Effective Field Theories 67


5.1 The Schrodinger equation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 67
5.2 The Euler-Heisenberg Lagrangian . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 68
5.3 EFT of the blue sky . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 69
5.4 Decoupling of Heavy Particle . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 69
5.5 Fermi theory . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 71
5.6 The Weinberg operator . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 73

3
4 CONTENTS

5.7 Renormalization of EFTs . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 74


5.8 EFTs and unitarity . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 75
5.8.1 The optical theorem . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 75
5.8.2 Partial wave unitarity bounds . . . . . . . . . . . . . . . . . . . . . . . . . . . . 76

6 The weak neutral current discovery 79


6.1 νµ e− → νµ e− and ν µ e− → ν µ e− . . . . . . . . . . . . . . . . . . . . . . . . . . . . 80
6.2 νe e− → νe e− and ν e e− → ν e e− . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 82
6.3 The weak NC effects in e+ e− → µ+ µ− . . . . . . . . . . . . . . . . . . . . . . . . . 83

7 The Standard Model at LEP-I 87


7.1 The optical theorem and the Z boson propagator . . . . . . . . . . . . . . . . . . . . . 87
7.2 The process e+ e− → f + f − at LEP . . . . . . . . . . . . . . . . . . . . . . . . . . . . 88
7.2.1 Cross-sections, Partial Widths and Asymmetries . . . . . . . . . . . . . . . . . 91
7.2.2 Invisible Width and Number of Neutrinos . . . . . . . . . . . . . . . . . . . . . 93

8 The Standard Model at LEP-II 95


8.1 Standard Model Processes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 96
8.1.1 Photon-Pair Production . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 96
8.1.2 Fermion-Pair Production . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 96
8.1.3 W W and ZZ Production . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 96
8.1.4 Four-Fermion Production . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 97
8.2 Tests of the non abelian structure of the Standard Model . . . . . . . . . . . . . . . . 98
8.2.1 e+ e− →W + W − . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 98
8.3 The Goldstone boson equivalence theorem . . . . . . . . . . . . . . . . . . . . . . . . . 101
8.3.1 e+ e− → W + W − . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 102
8.3.2 t → W + b . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 103
8.3.3 W + W − → W + W − and the no-lose-Higgs theorem . . . . . . . . . . . . . . . 105

9 Renormalization 109
9.1 Dimensional regularization and subtraction schemes . . . . . . . . . . . . . . . . . . . 110
9.2 Vacuum polarization . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 116
9.3 Electron self-energy . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 121
9.4 Vertex correction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 122
9.5 The anomalous magnetic moment g − 2 . . . . . . . . . . . . . . . . . . . . . . . . . . 124
9.6 Ward identity of QED . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 127
9.7 The β functions of QED and QCD . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 128

10 Electro-Weak Precision Tests 131


10.1 Oblique corrections . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 133
10.2 Computation of EW vacuum polarization loops . . . . . . . . . . . . . . . . . . . . . . 138
10.3 Custodial SU (2)C symmetry . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 141
10.4 One loop corrections to ∆ρ . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 143
10.4.1 Yukawa contribution . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 144
10.4.2 Higgs-boson contribution . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 144
10.5 One loop corrections to Z → f¯f . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 146
10.6 The S, T and U parameters . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 150

11 Flavor Changing Neutral Currents 153


11.1 Meson-antimeson oscillation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 154
11.2 FCNC in the SM and the GIM mechanism . . . . . . . . . . . . . . . . . . . . . . . . . 157
11.3 The hierarchy problem . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 160
CONTENTS 5

12 Anomalies 161
12.1 Axial-vector current anomaly and π 0 → γγ . . . . . . . . . . . . . . . . . . . . . . . . 162
12.2 ABJ anomalies . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 163
12.3 Anomalies in the Standard Model . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 168

13 Higgs physics 171


13.1 Theoretical bounds on Mh : triviality and vacuum stability bounds . . . . . . . . . . . 171
13.2 Experimental bounds on Mh . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 175
13.3 Higgs production: gluon-gluon fusion and the parton model . . . . . . . . . . . . . . . 176
13.4 Higgs decay modes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 178
13.4.1 h → f f . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 179
13.4.2 h → W W ? and h → ZZ ? . . . . . . . . . . . . . . . . . . . . . . . . . . . . 180
13.4.3 h → gg and h → γγ . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 184
6 CONTENTS
Chapter 1

The Standard Model Lagrangian

Any model of particle physics has to be constructed specifying

1. the gauge group G of the theory;

2. the particle content of the theory;

3. the possibility of spontaneous symmetry breaking of the group G into a smaller group GSSB .

Moreover, it has to be renormalizable, i.e. physical quantities extracted from the theory have to be
finite: as proven by a theorem, this is the case for SSB theories. These properties need to be satisfied
not only by the Standard Model but also by theories beyond it. People have also shown that the SM is
a gauge anomaly-free theory. The gauge group of the Standard Model is

GSM = SU (3)c ⊗ SU (2)L ⊗ U (1)Y . (1.1)

Remembering that the number of generators of SU (N ) is N 2 − 1 and the number of charged (diagonal)
generators of SU (N ) is N − 1, we get 8 gluons from SU (3)c the W ± , Z, γ bosons from SU (2)L ⊗ U (1)Y .
The SM symmetry group breaks into

GSSB
SM = SU (3)c ⊗ U (1)em , (1.2)

which shows the physical fact that gluons and photons stay massless after SSB. The SM Lagrangian
can be written in a short-handed form as
1
LSM = − F µν Fµν + iψγ µ Dµ ψ + (Dµ ϕ)† (Dµ ϕ) − V (ϕ† ϕ) + ψyϕψ + h.c.
4 (1.3)
= Lg.b. f h h
kin + Lkin + Lkin + Lpot + LY .

In particular:

1. Lg.b.
kin contains all the kinetic terms of gauge bosons; what’s non trivial is the presence of non-abelian
type interactions between gauge bosons:

∼ gP µ ∼ g2 (1.4)

2. Lfkin contains the fermion kinetic terms and also the gauge boson interactions with fermions;

3. Lhkin is the kinetic term for the Higgs field which gives masses to the gauge bosons m2V ∼ hϕi2 g 2
through the SSB mechanism, once we pick up a vacuum state hϕi. In this term we also have the
interactions between the Higgs and the gauge bosons:

7
8 CHAPTER 1. THE STANDARD MODEL LAGRANGIAN

(1.5)

4. Lhpot is the Higgs potential. It gives the Higgs field mass and the Higgs self interactions

(1.6)

5. LY is the Yukawa sector of the SM. It gives rise to the fermion masses mψ ∼ hϕiy and also to
the Higgs interactions with fermions.
The fermions of the SM are grouped in three generations, i = 1, 2, 3, and classified by their behavior
under the gauge group:
 SU (2)L doublets: ! !
uiL νLi
QiL = , LiL = . (1.7)
diL eiL

 SU (2)L singlets charged under U (1)Y :

uiR , diR , eiR , (νR


i
), (1.8)

with the right neutrino not included in the SM.


We set the relation between charge, isospin and hypercharge as

Q = T3L + Y. (1.9)

Since ϕ is an SU (2) doublet, it contains 4 degrees of freedom. The transformation properties of ϕ


under SU (2)L ⊗ U (1)Y are
1
 
ϕ ∼ 2; . (1.10)
2

1.1 The bosonic sector


We can start looking at the bosonic sector of the SM Lagrangian, neglecting the gluons:
1 a µν,a 1
LSM
b = − Wµν W − Bµν B µν + (Dµ ϕ)† (Dµ ϕ) − V (ϕ† ϕ), (1.11)
4 4
where, explicitly:
a
Wµν =∂µ Wνa − ∂ν Wµa + gεabc Wµb Wνc ,
Bµν =∂µ Bν − ∂ν Bµ ,
~σ ~ (1.12)
0
Dµ ϕ =(∂µ − ig · W µ − ig Yϕ Bµ )ϕ,
2
V (ϕ† ϕ) =µ2 ϕ† ϕ + λ(ϕ† ϕ)2 .

We chose µ2 < 0 in order to realize EW SBB and λ > 0 in order to have the Higgs potential V bounded
from below at large field ϕ values. The SU (2)L gauge transformation is given by

(1.13)
~ ~
σ
U (θ(x)) = e−iθ(x)· 2 ,
1.1. THE BOSONIC SECTOR 9

and the U (1)Y by


U (β(x)) = e−iβ(x)Y , (1.14)
and are applied to the fields in the following way:

~ µ )U (θ)−1 + i U (θ)∂µ U (θ)−1 ,


~ µ0 =U (θ)(~σ · W
~σ · W
g
(1.15)
1
Bµ0 =Bµ + 0 ∂µ β(x).
g

The Higgs field can be written in the following exponential form

~
!
1 ξ(x) · ~σ
 
0
ϕ(x) = √ exp i (1.16)
2 v v + h(x)

where ξ(x)
~ stands for the three Goldstone bosons. Exploiting the unitary gauge i.e. applying
~
ξ(x)·~
σ
U = e−i v on every doublet in the SM Lagrangian we get for the Higgs field
!
1 0
ϕu.g. (x) = √ (1.17)
2 v + h(x)

where in our convention v = 246 GeV. In the unitary gauge, the bosonic sector reads:
1 a µν,a 1 1 1
LSM
b |u.g = − Wµν W − Bµν B µν + (∂µ h)(∂ µ h) − m2h h2 +
4 4 2 2
2 2 2 0 2 (1.18)
(v + h) 2 + −µ (v + h) (g + g ) λ
+ g Wµ W + Zµ Z µ − λvh3 − h4
4 2 4 4
where
Wµ1 ∓ iWµ2
Wµ± = √
2
3 (1.19)
Zµ = cW Wµ − sW Bµ
Aµ = cW Bµ + sW Wµ3

having set cW = cos θW , sW = sin θW , tan θW = g0


g so that

gg 0
e = gsW = g 0 cW = q . (1.20)
g2 + g02

The bosonic mass spectrum (before and after SSB) is summarized in the following table On the

µ2 < 0 µ2 > 0
d.o.f particle m2 d.o.f particle m2
1 h 2λv 2 4 ϕ 2µ2
3×2 Wµ± 1 2 2
4g v 2×3 Wµa 0
3×1 Zµ 1 2
4 (g + g 0 2 )v 2 2×1 Bµ 0
2 Aµ 0
12 12

experimental side, the boson masses have been measured with a remarkably high precision:

MZ = 91.1876 ± 0.00021 GeV LEP


MW = 80.385 ± 0.015 GeV Tevatron (1.21)
Mh = 125.10 ± 0.14 GeV LHC .
10 CHAPTER 1. THE STANDARD MODEL LAGRANGIAN

The Higgs interactions with bosons are described by the following Feynman rules

h
h h
λ
h = −iλv × 3! = −i × 4! (1.22)
4
h h
h
h Wµ± h Zµ
g2 g2 + g02
= i ηµν × 2! =i ηµν × 2! × 2! (1.23)
4 8
h Wν∓ h Zν

Wµ± Zµ
g2 g2 + g02
h = iv ηµν h = iv ηµν × 2! (1.24)
2 4

Wν∓ Zν

where the combinatorial factors n! (n = 2, 3, 4) stem from equivalent permutations of identical fields.

1.2 The fermionic sector


Now we can study the fermionic sector which is represented by the following Lagrangian

LSM / ψ − ψyϕψ + h.c.


f = ψiD (1.25)

where the fermions can be decomposed into left and right chiral components ψ = ψL + ψR where

1 ∓ γ5
ψL = PL ψ ψ R = PR ψ PL,R = (1.26)
2

with the projectors PL,R satisfying the properties PL,R


2 = PL,R , PL PR = PR PL = 0. The ψL spinor
describes a particle with helicity −1/2 and an antiparticle with helicity +1/2 while ψR describes a
particle with helicity +1/2 and an antiparticle with helicity −1/2. Fermions are classified in three
generations with charges under the SM gauge groups shown in the following table

! SU (3)c SU (2)L U (1)Y T3 ! Q! Y = Q −!T3


νL +1/2 0 −1/2
LL = 1 2 −1/2
eL −1/2 −1 −1/2
eR ! 1 1 −1 0 ! −1 ! −1 !
uL +1/2 +2/3 1/6
QL = 3 2 +1/6
dL −1/2 −1/3 1/6
uR 3 1 +2/3 0 +2/3 +2/3
dR ! 3 1 −1/3 0 ! −1/3! −1/3!
ϕ+ +1/2 +1 +1/2
ϕ= 1 2 +1/2
ϕ0 −1/2 0 +1/2
?
! ! ! !
ϕ0 +1/2 −1 −1/2
ϕ̃ = 1 2 −1/2
−ϕ− −1/2 0 −1/2
1.2. THE FERMIONIC SECTOR 11

Let us write the fermion Lagrangian in the unitary gauge


X X
LSM
f |u.g. = f L i∂/ fL + f R i∂/ fR Kinetic terms
f f
g
− √ (Wµ− J +µ + Wµ+ J −µ ) Weak charged current
2
g (1.27)
− Zµ JZµ Weak neutral current
cW
µ
−eAµ Jem Electromagnetic current
+LY , Yukawa sector

where the electromagnetic current is given by


X
Jµem = f γµ f Qf , (1.28)
f

while the weak charged-currents read

Jµ− = uL γµ dL + ν L γµ eL
(1.29)
Jµ+ = (Jµ− )† = dL γµ uL + eL γµ νL ,

and, finally, the weak neutral-current reads

JZµ = J3L
µ
X X
µ
− s2W Jem = f L γ µ T3 fL − f γ µ f Qf s2w
f f
X 
= f L gLf γ µ fL + f R gR
f µ
γ fR
(1.30)
f
X  (g f + g f ) (g f − gR
f
) µ

L R µ
= f γ f −f L γ γ5 f ,
f
2 2

where the chiral couplings gL(R)


f
and the vector and axial couplings gVf (A) are given by

g f + gR
f
T f − 2s2W Qf

f
gV = L = 3L
 
g f = T f − Qf s2


2 2 (1.31)
L 3L W
.
g f = −Q s2  g f − gR
f f
T3L
R f W g f = L


A =
2 2
Let’s consider now the Yukawa sector in the unitary gauge
v + h(x) i ij j
 
LY |u.g =− √ uL yu uR + diL ydij djR + eiL yeij ejR + h.c (1.32)
2

where yfij are flavor mixing matrices with i, j being flavor indices. Notice that the Yukawa sector is the
only one that flips chiralities. In the flavor space we organize matter in triplets:
     
u d e
uiL,R = c diL,R = s eiL,R = µ . (1.33)
     
t L,R b L,R τ L,R

The n × n matrix yf with f = u, d, e, being a complex matrix can depend in general on 2n2 parameters.
Specializing to the SM (n = 3), it has 2 × 32 = 18 parameters for each u, d and e, so in total we would
have have 54 parameters. We would like to diagonalize the Yukawa matrices. Let us rotate the fields

0

0
uL = Lu uL uR = Ru uR

 

d0 = L d
L d L Rd0 = R d
d R (1.34)
 
e0 = L e
 e0 = R e

L e L R e R
12 CHAPTER 1. THE STANDARD MODEL LAGRANGIAN

with
RR† = R† R = 1 LL† = L† L = 1 (1.35)
So the Yukawa sector yields

v+h 0
 
LY |u.g = − √ uL (Lu yu Ru† )u0R + d0L (Ld yd Rd† )d0R + e0L (Le ye Re† )e0R (1.36)
2

Now we will demonstrate that it is always possible to diagonalize generic matrices (as yu,d,e ) by means
of a biunitary transformation and therefore it turns out that Lf yf Rf† are diagonal matrices
 
ŷf 1
ŷf = Lf yf Rf† =  ŷf 2 . (1.37)
 
ŷf 3


Since yf yf† = (yf yf† ) , it is possible to diagonalize yf yf† by means of a unitary matrix and the resulting
eigenvalues are real. Let us call Mf2 the diagonal matrix, which will be identified afterwards with ŷf2 :

yf yf† = Uf Mf2 Uf† = Uf Mf Mf Uf† = (Uf Mf )(Mf Uf† ) (1.38)


−1
We want to write yf in such a way that yf = Uf ŷf Kf† . If we consider Kf† = Mf Uf† (yf† ) ,

Kf Kf† = yf−1 Uf Mf Mf Uf† (yf† )−1 = 1, (1.39)

Kf† Kf = Mf Uf† (yf† )−1 (yf )−1 Uf Mf = Mf U


† †−1  U f M f = 1,
(Mf2 )−1 Uf−1

(1.40)
f (U
f)

  
so Kf exists and is unitary. We get

yf = Uf Mf Kf† =⇒ Mf = Uf† yf Kf ∼ Lf yf Rf† . (1.41)

The complete fermionic sector including the three generations is therefore:

v + h(x) i ij j
 
Lf = − √ uL yu uR + diL ydij djR + eiR yeij ejR + h.c. −
2
g
 
− √ uiL W/ + diL + ν iL W/ + eiL + h.c.
2
g X i
  (1.42)
− f L γµ gLf fLi + f iR γµ gR
f i
fR Z µ −
cW f
X X 
−eAµ f γµ f Qf + f iL i∂/ fLi + f iR i∂/ fRi .
f f

1.3 The CKM matrix


The previous Lagrangian has been written in the unitary gauge. Passing to the fermion mass basis,
the Yukawa sector reads
v + h i ii i
 
LY |m.b. =− √ uL ŷu uR + diL ŷdii diR + eiL ŷeii eiR + h.c (1.43)
2
where the fermion masses are given by
v ii
mii
f = √ ŷf (1.44)
2
Now we consider the effect of the change of basis on the other fermionic sectors:

 Kinetic sector: it is not affected as the matrices Lf and Rf are unitary.


1.3. THE CKM MATRIX 13

 Electromagnetic interaction: it’s not affected too since γ µ affects the spinorial space, Lf and Rf
are in the flavor space, and the interaction couples fermions of the same type.

 Weak neutral interaction: it’s not affected as the electromagnetic interactions.

 Weak charged interaction: are indeed affected since the interaction is among up-type and
down-type fermions so we cannot exploit the unitarity of Lf

g
 
LCC = − √ u0L (Lu L†d ) W
/ + d0L + ν 0L (Lν L†e ) W
/ + e0L + h.c. (1.45)
2 | {z } | {z }
VCKM UPMNS

VCKM (Cabibbo-Kobayashi-Maskawa) and UPMNS (Pontecorvo-Maki-Sakata-Nakagawa) are uni-


tary matrices with flavor mixing properties
If we introduce a right-handed neutrino field νR , we could write an SU (2)L ⊗ U (1)Y invariant term

LνY = −yν LL ϕ̃νR + h.c. (1.46)

with LL ϕ̃ being a singlet under SU (2)L and of hypercharge YL + Yϕ̃ = 0 and so νR has to be also (11, 0),
i.e. completely sterile under SU (2)L ⊗ U (1)Y . If neutrinos were massless, that is if yν = 0, we can
always exploit whichever unitary matrix we want in order to represent the neutrinos in the mass basis:

νL0 = Lν νL . (1.47)

Therefore in the mν = 0 approximation we can choose Lν ≡ Le and then UPMNS = 1.


Let’s count now the number of angles and phases of the VCKM matrix. A generic unitary n × n
matrix U has n2 (instead of 2n2 ) independent parameters because the unitary condition U † U = U U † =
1 → U = eiH with H = H † implies that the initial 2n2 parameters are not all independent but we have
to subtract n elements on the diagonal and n(n − 1) elements out of the diagonal. As a result we get
2n2 − n − n(n − 1) = n2 independent parameters. In order to compute the number of angles in U we
look for the number of parameters of an othogonal matrix O, such that OT O = OOT = 1 → O = eA
with A = −AT . These conditions imply that from the initial n2 parameters in O we have to subtract
n terms from the diagonal and n(n − 1)/2 terms out of the diagonal. As a result we end up with
n2 − n − n(n − 1)/2 = n(n − 1)/2 angles. All the other parameters in U that are not angles are phases
and therefore we obtain
n(n − 1) n(n + 1)
angles = , phases = . (1.48)
2 2
Now we look for physical phases: we have the freedom to apply phases to the quark fields

q i → eiφiL q i
(1.49)
L L
q i → eiφiR q i
R R

Notice that the up and down components of the SU (2)L doublet are rephased in the same way, but
the phase can be different between left-handed and right-handed fields. Hence we can subtract this
freedom from the physical phases
2×Nflavors
n(n + 1)
(1.50)
n=3
z}|{
phases = − 2n = 0
2
Actually, if all the quarks are rotated by the same phase, the VCKM remains invariant, so we can
subtract only 2n − 1 phases due to the conservation of baryonic number (the baryonic number is indeed
a symmetry, associated to a coherent phase shift in all the quark fields)
n(n + 1) (n − 1)(n − 2) n=3
#phases = − (2n − 1) = = |{z}
1 (1.51)
2 2
CP

14 CHAPTER 1. THE STANDARD MODEL LAGRANGIAN

We will see that the remaining phase is related to CP violation.


To conclude we put together the different terms and write explicitly the SM Lagrangian invariant
under the SU (3)c ⊗ SU (2)L ⊗ U (1)Y for massless neutrinos
1 i 1 1
LSM = − Wµν Wiµν − B µν Bµν − Gaµν Gµν a
4 4 4
+ (Dµ ϕ)† (Dµ ϕ) − µ2 (ϕ† ϕ) − λ(ϕ† ϕ)2
(1.52)
X 
+ LλL D
/ LλL + QλL D
/ QλL + eλR D
/ eλR + uλR D
/ uλR + dλR D
/ dλR
λ
X 0 0 0 0 0 0

− LλL yeλλ ϕeλR + QλL ydλλ ϕdλR + QλL yuλλ (−iϕ† σ2 )T uλR + h.c.
λ,λ0

with i = 1, 2, 3 and a = 1, . . . , 8 gauge indices, λ = 1, 2, 3 generation index and


σj λa
Dµ = ∂µ + ig Wjµ + ig 0 Y Bµ + igs Aaµ
2 2
i
Wµν = (∂µ Wνi − ∂ν Wµi ) + gεijk Wµj Wνk (1.53)
Bµν = (∂µ Bν − ∂ν Bµ )
Gaµν = (∂µ Aaν − ∂ν Aaµ ) + gs εabc Abµ Acν

After SSB the Higgs field takes the vev (0, v/ 2) and the SM gauge group breaks into SU (3)c ⊗ U (1)em .
Finally, in the unitary gauge and passing to the fermion mass basis, we have
2
1 a µν 1
LSSB
SM = − Gµν Ga − ∂µ Aν − ∂ν Aµ − ie(Wµ− Wν+ − Wµ+ Wν− )
4 4
2
1
− ∂µ Wν+ − ∂ν Wµ+ − ie(Wµ+ Aν − Wν+ Aµ ) + ig 0 cW (Wµ+ Zν − Wν+ Zµ )
2
2
1
− ∂µ Zν − ∂ν Zµ + ig 0 cW (Wµ− Wν+ − Wµ+ Wν− )
4
1 1 λ
+ (∂µ h)(∂ µ h) − Mh2 h2 − λvh3 − h3
2 2 4
2 
h 1 2 µ
 
2 µ,+ −
− 1+ MW W W µ + M Z Z Z µ
v 2 (1.54)
X  λa a λ λa a λ

+ i eλ ∂/ eλ + ν λL ∂/ νLλ + uλ (∂/ + igs A/ )u + dλ (∂/ + igs A/ )d
λ
2 2
g
   
λ,λ0 λ0 0
X
− √ Wµ+ uλ γ µ (1 − γ5 )VCKM d + ν λ γ µ (1 − γ5 )eλ + h.c.
λ,λ0
2 2
f
XX g  T3L − 2s2W Qf Tf
    
− Zµ f λ γ µ − 3L γ5 f λ − e Aµ f λ γµ Qf f λ
f λ
cW 2 2
X X h

− 1+ mλf f λ f λ
f 6=ν λ
v

with f = e, ν, u, d and
gg 0
e = gsW = g 0 cW = q (1.55)
g2 + g02

v2g2 v 2 (g 2 + g 0 2 ) vyfλ
Mh2 = 2λv 2 2
MW = MZ2 = mλf = √ . (1.56)
4 4 2
Chapter 2

Discrete symmetries

2.1 C, P and T symmetries


The discrete symmetries P (parity), C (charge conjugation) and T (time reversal) act as follows
 
~
 x → −~x (
~
 x → ~x
xµ → xµ
 
P : t→t C: T : t → −t . (2.1)

q → q
 q → −q 
q → q

Let us find now a representation for C, P and T in the Dirac basis:

 Parity transformation: let us consider the Dirac equation coupled with the e.m. field

(2.2)
h     i
(iγµ Dµ − m) ψ(x) = iγ0 ∂ 0 + iqφ + iγi ∂ i + iqAi − m ψ(x) = 0

where Dµ = ∂ µ + iqAµ is the covariant derivative and Aµ is the e.m. field transforming under a
parity transformation as P A0 (t, ~x) = A00 (t, ~x0 ) = A0 (t, ~x) and P A
~ (t, ~x) = A
~ 0 (t, ~x0 ) = −A
~ (t, ~x)
where ~x = −~x. Under parity the above equation becomes
0

(2.3)
h     i
iγ0 ∂ 0 + iqφ − iγi ∂ i + iqAi − m ψP = 0

where we have assumed the exstence of the linear transformation P such that ψP (t, −~x) = P ψ(x).
Applying the operator P on the left of eq. (2.2) and writing ψ = P −1 ψP , which assumes that the
operator P is invertible, we find

(2.4)
h     i
P iγ0 ∂ 0 + iqφ + iγi ∂ i + iqAi − m P −1 ψP = 0 .

In order to make the Dirac equation covariant under P , we equate eqs. (2.3) and (2.4) finding

−1 [P, γ0 ] = 0
( 
P γ0 P = γ0

=⇒ {P, γ } = 0
i =⇒ P = eiϕP γ0 (2.5)
P γi P −1 = −γi 
{γ , γ } = 2η µν

µ ν

 Time reversal: as before we act with T on the Dirac equation and using the fact that
T A0 (t, ~x) = A00 (t0 , ~x) = A0 (t, ~x) and T A
~ (t, ~x) = A ~ (t, ~x) where t0 = −t, we find
~ 0 (t0 , ~x) = −A

(2.6)
h     i
iγ0 −∂ 0 + iqφ + iγi ∂ i − iqAi − m ψT (−t, ~x) = 0

Notice that eqs. (2.2) and eq. (2.6) are not formally equivalent as the relative signs of the terms in
the covariant derivatives are opposite. Therefore, we consider the complex conjugate of eq. (2.2):

(2.7)
h     i
iγ0 −∂ 0 + iqφ − iγi? ∂ i − iqAi − m ψ ? (x) = 0

15
16 CHAPTER 2. DISCRETE SYMMETRIES

and postulate the existence of a transformation T acting as ψT = T ψ ? = T ψ. Notice that T is


an antilinear operator as it contains the complex conjugation operation T iT −1 = −i. Applying
the operator T to eq. (2.7) and writing ψ ? = T −1 ψT , which assumes that T is invertible, we find

(2.8)
h     i
T iγ0 −∂ 0 + iqφ − iγi? ∂ i − iqAi − m T −1 ψT = 0 .

Finally, equating now eqs. (2.6) and eq. (2.8) we find



−1 [T, γ0 ] = 0
( 
T γ0 T = γ0

=⇒ [T, γ ] = 0
2 =⇒ T = eiϕT γ1 γ3 . (2.9)
T γi? T −1 = −γi 
{T, γ } = 0

1,3

 Charge conjugation: since in this case q → −q, xµ → xµ and Aµ → Aµ , eq. (2.2) becomes

(2.10)
h     i
iγ0 ∂ 0 − iqφ + iγi ∂ i − iqAi − m ψC = 0

Similarly to what done in the case of time reversal, the covariance of eqs. (2.2) and (2.10) can be
restored assuming that ψC = Cψ ? . Then, following the steps outlined above, we find

Cγ0 C −1 = −γ0
( (
[C, γ2 ] = 0
=⇒ =⇒ C = eiϕC γ2 . (2.11)
Cγi? C −1 = −γi {C, γ0,1,3 } = 0

So, under a CPT transformation, we get

ψCP T = CP T ψ = eiϕCP T γ0 γ1 γ2 γ3 ψ = eiϕCP T γ5 ψ . (2.12)

2.2 C, P and T in the Standard Model


Since we know from experiments that P and CP are violated by weak interactions, we are going now
to study the transformation properties of the SM Lagrangian under P and CP . The pieces of the SM
Lagrangian that are non-trivially transformed are the following fermion interaction terms
g  † −

LSM
f,int = − √ u V
L CKM W/ +
dL + d V
L CKM W
/ uL
2
g  
f L γµ gLf fL + f R γµ gRf
X
− fR Z µ
cW f
v + h(x) 
(2.13)
X 
− eAµ f γ µ f Qf − √ uL yu uR + dL yd dR + h.c.
f
2

where LSM
f,int is given in the mass basis and, for simplicity, we have neglected leptons in eq. (2.13).
These terms will be affected by transformations under discrete symmetries
(
~x → −~x
P : =⇒ ψP (−~x, t) = eiϕP γ0 ψ(~x, t) (2.14)
t→t
(
~x → ~x
T : =⇒ ψT (~x, −t) = eiϕT γ1 γ3 ψ ? (~x, t) (2.15)
t → −t
(
xµ → xµ
C: =⇒ ψC (~x, t) = eiϕC γ2 ψ ? (~x, t) (2.16)
q → −q

1
! ! !
0 φ χ
CP T : ψCP T = e iϕCP T
γ5 ψ = e iϕCP T
=e iϕCP T
. (2.17)
1 0 χ φ
2.2. C, P AND T IN THE STANDARD MODEL 17

Notice that the CP T transforms a particle into an antiparticle. Neglecting the phases, let us study
some bilinear transformation under parity:
P
ψψ −→ ψ † γ 0 γ 0 γ 0 ψ = ψψ scalar
P {γ5 ,γµ }=0
ψγ5 ψ −→ ψ † γ 0 γ 0 γ5 γ 0 ψ = −ψ † γ 0 γ5 ψ = −ψγ5 ψ pseudoscalar
( 0
P +ψγ ψ, µ = 0
ψγ µ ψ −→ ψ † γ 0 γ 0 γ µ γ 0 ψ = ψ † γ µ γ 0 ψ = i
= (−1)µ ψγ µ ψ vector
−ψγ ψ, µ = i
−ψγ 0 γ5 ψ, µ = 0
(
µ P † 0 0 µ † µ 0
0
ψγ γ5 ψ −→ ψ γ γ γ γ5 γ ψ = −ψ γ γ γ5 ψ = = −(−1)µ ψγ µ γ5 ψ axial vector
+ψγ i γ5 ψ, µ = i
(2.18)
where we exploited the convention
(
+1, µ = 0
(−1)µ = . (2.19)
−1, µ = 1, 2, 3

If we contract the vector bilinear with a generic vector field Aµ

(2.20)
P P
ψγ µ Aµ ψ −→ ψγ µ Aµ ψ if Aµ −→ (−1)µ Aµ

and therefore we have invariance under parity, like in electromagnetism. Let us study now the
transformation properties of the weak neutral-current under parity:
g Xh i
LNC = − f L γµ gLf fL + f R γµ gR
f
fR Z µ
cW f
P g Xh i (2.21)
=− f R γµ gLf fR + f L γµ gR
f
fL Z µ
cW f

since Z µ f L γµ fL = Z µ f γµ PL f −→ Z µ f γµ PR f = Z µ f R γµ fR , i.e. the vector part stays invariant while


P

the axial part changes sign. We see an explicit parity violation if gLf 6= gR f
. Let us assume, per assurdo
that gL = gR ≡ g . Then
f f f

g X f
 
− g f R γµ fR + f L γµ fL Z µ , (2.22)
cW f | {z }
f γµ f

and therefore we have conservation of P . This is what happens in QED. For the charged current terms
of the Lagrangian we have parity violation too. This is because the electroweak theory is a so-called
V − A theory that is violated under parity because V − A −→ V + A. Notice that the SM Yukawa
P

sector doesn’t violate P. Indeed,


(v + h) (v + h) (v + h)
   
LY = − √ yt f L fR + h.c. = − √ yt f PR f + yt† f PL f = − √ yt f f (2.23)
2 2 2

and, as we have already seen, scalar bilinears like f f are P invariant.


Now let us study the C transformation, choosing the phase convention ϕC = π/2

ψ → iγ2 ψ ?



(2.24)

C: ψ ? → −iγ2? ψ = iγ2 ψ

 ?T T

ψ=ψ γ0 → (iγ2 ψ) γ0

acting on a scalar bilinear (where i, j indices denote different particles)

C {γ2 ,γ0 }=0,γ22 =−1


ψ i ψj −→ (iγ2 ψi )T γ 0 iγ2 ψj? = − ψiT γ2T γ 0 γ2 ψJ? = −ψiT γ 0 ψj?
 . (2.25)
ψα ,ψβ? =0
= −ψiα γαβ
0
ψjβ? = ψjβ? γαβ
0
ψiα = ψj† γ0T ψi = ψ j ψi
18 CHAPTER 2. DISCRETE SYMMETRIES

As expected, we have found the important result that C transforms a particle in the antiparticle

(2.26)
C
ψ i ψj −→ ψ j ψi

Let us write also the transformation rules for the EW gauge bosons

(2.27)
P C
Zµ −→ (−1)µ Zµ −→ −(−1)µ Zµ

(2.28)
P C
Aµ −→ (−1)µ Aµ −→ −(−1)µ Aµ

(2.29)
P C
Wµ± −→ (−1)µ Wµ± −→ −(−1)µ Wµ∓
One could wonder why neutral fields (uncharged under electromagnetism) get a minus sign with C. This
is because the change of sign of the coupling e, the electric charge, cannot be accounted for by a unitary
transformation (e is a number that doesn’t transform under a unitary transformation). Therefore, we
need to postulate a trasformation rule for the neutral fields. Let us gather all the transformation rules
under discrete symmetries in a table. Now, exploiting the transformation properties

Sym. ψψ ψγ5 ψ ψγ µ ψ ψγ µ γ5 ψ ψσ µν ψ ∂µ
P +1 −1 (−1)µ −(−1)µ (−1)µ (−1)ν (−1)µ
T +1 −1 (−1)µ (−1)µ −(−1)µ (−1)ν −(−1)µ
C +1 +1 −1 +1 −1 +1
CP T +1 +1 −1 −1 +1 −1

Table 2.1: Transformation properties of fermionic bilinears and ordinary derivative under C, P, and T.

(2.30)
CP CP
/ ψj −→ ψ j A
ψiA / ψi / γ5 ψj −→ ψ j A
ψiA / γ5 ψi

we can see how the charged-current terms in the SM Lagrangian transform under CP
g h †
i
LCC = − √ uL VCKM W / + dL + dL VCKM W/ − uL
2
g (2.31)
 
CP †
T
= − √ dL VCKM W/ − uL + uL (VCKM )T W/ + dL
2 | {z } | {z }
† ?
VCKM
(VCKM )?

so we would have CP invariance only if VCKM = VCKM ? . However, the CKM matrix contains a physical
phase and therefore we conclude that CP is violated in the charged-current sector of the SM. It is
straightforward to verify that, in the mass basis, all the other sectors of the SM are CP invariant.
Now we may argue that it is only in the fermion mass basis that we are violating CP since only in
this basis we have the presence of VCKM . Obviously this is not true, since physics does not depend
on the basis we choose. So let us see what happens in the interaction basis, where there is no VCKM ,
considering the up-quark Yukawa sector

v + h(x) h i
LuY = − √ uL yu uR + uR yu† uL
2
♥ v + h(x) h i
=− √ u(yu + yu† )u + u(yu − yu† )γ5 u
2 2
(2.32)
CP v + h(x) h i
= − √ u(yu + yu† )T u − u(yu − yu† )T γ5 u
2 2
♦ v + h(x) h † i
=− √ u(yu + yu )? u + u(yu − yu† )? γ5 u
2 2
Comparing ♥ and ♦ we get CP invariance only if yu = yu? . But the yu matrix is completely general
and therefore yu 6= yu? so that CP is violated also in the interaction basis, as it must be.
2.3. CPT THEOREM 19

2.3 CPT theorem


Given a relativistic quantum field theory described by a density Lagrangian L(x) which is i) hermitian,
ii) local1 , iii) a bosonic operator2 , iv) invariant under proper Lorentz transformations, and v) a normal
product of fields, it must be that
θL(x)θ† = L(−x) (2.33)
where θ = CP T and therefore the action S is invariant under CP T
Z Z Z
θSθ =† 4
d x θL(x)θ = † 4
d x L(−x) = d4 (−x) L(−x) = S . (2.34)

The above result goes under the name of CPT theorem.


In order to demonstrate eq. (2.33), let us consider a L(x) which fulfills all the above conditions

(2.35)
X
L(x) = ci Oi (x)
i
Oi (x) = : ... Aµ (x) ... (ψ̄a Γψb )(x) ... ∂ ν ... : (2.36)

where ci are complex coefficients. The classical transformation of states ψCP T = θψ(x) = iγ5 ψ ∗ (−x)
becomes, at the quantum level, the following transformations of operators

θψ(x)θ† = iγ5 ψ † (−x)


(
(2.37)
θψ † (x)θ† = −iψ(−x)γ5

and therefore we find


h i
θ(ψ̄a Γψb )(x)θ† = θψa† (x)θ† θγ 0 Γθ† θψb (x)θ† = ψaα (−x) γ5 (γ 0 Γ)∗ γ5 ψb†β (−x) =
| {z } | {z } | {z } αβ
−iψa (−x)γ5 (γ 0 Γ)∗ iγ5 ψ † (−x)
b
h iT
− ψb†β (−x)[γ5 (γ 0 Γ)∗ γ5 ]αβ ψaα (−x) = −ψb† (−x) γ5 (γ 0 Γ)∗ γ5 ψa (−x) =

(2.38)
h i
− ψb† (−x)γ5 (γ 0 Γ)† γ5 ψa (−x) = + ψ̄b γ 0 γ5 Γ† γ5 γ 0 ψa (−x)

where we have used that fermionic operators anticommute, γ5 is hermitian and anticommuting with γ 0
and that θ is an antilinear operator. Moreover,

γ 0 (γ5 Γ† γ5 )γ 0 = γ 0 (γ5 Γγ5 )† γ 0 = (−1)N γ 0 Γ† γ 0 = (−1)N Γ (2.39)

where the factor (−1)N , being N the number of Lorentz indices of Γ, arises from the anticommutation
properties of γ5 with γ µ . As a result, we obtain

θ(ψ̄a Γψb )(x)θ† = (−1)N (ψ̄b Γψa )(−x) = (−1)N [(ψ̄a Γψb )(−x)]† . (2.40)

Therefore, L(x) transforms under CPT as

θL(x)θ† = c∗i (−1)Ntot : ... Aµ (−x) ... [(ψ̄a Γψb )(−x)]† ... ∂ ν ... : (2.41)
X

where Ntot is the total number of Lorentz indices of L(x) since also the ordinary derivative (N = 1),
vercor fields (N = 1) as well as field-strenght tensors (N = 2) transform under CPT proportionally
to a factor of (−1)N . Notice also that θci θ† = c∗i as the operator θ is antilinear. Moreover, within a
normal product, we can commute bosonic operators and therefore

θL(x)θ† = (−1)Ntot [ci Oi (−x)]† . (2.42)


X

i
1
L(x) is local if it is a function of fields and their derivatives evaluated at the same space-time point.
2
Fermionic fields can appear only with an even number so that L contains only fermionic bilinears.
20 CHAPTER 2. DISCRETE SYMMETRIES

Since L(x) is required to be Lorentz invariant, the Lorentz indeces must be summed over invariant
tensors. In a 4-dimensional space-time we have just the metric tensor gµν and the Levi-Civita tensor
µνρσ . Since a contraction with either gµν or µνρσ reduces the free indeces by an even number, it
follows that Ntot is even in order to make L(x) Lorentz invariant. Therefore, we finally find that

θL(x)θ† = [ci Oi (−x)]† = L† (−x) = L(−x) (2.43)


X

where the last equality follows from the fact that L(x) is hermitian.

2.4 Sakharov conditions


As a beautiful application of what we have learnt so far let us prove the famous Sakharov conditions
for the baryon-antibaryon asymmetry in the early universe. This can arise from a perfectly symmetric
universe through a dynamical mechanism leading to the matter excess over antimatter only if the
following conditions are fulfilled:
1. B : baryon number violation.

2. C and : charge conjugation and its combination with parity are violated symmetries
CP

3. Out of thermodynamical equilibrium.


Let us prove them under the hypotesis of a symmetric early Universe, that is that B(t0 ) = 0.
1. B : The baryon number is an operator B(t) hence it satisfies
dB(t)
i = [B, H] (2.44)
dt
Per assurdo, if [B, H] = 0 (B is conserved) we would have B(t) = B(t0 ) = 0. Hence B .

2. C and : The mean value of the baryon number at temperature T is computed as
CP

hB(t)iT = Tr(ρ(t)B(t)) (2.45)

with a suitable density matrix ρ(t). The baryon number is the conserved charge for the following
global accidental symmetry of the SM Lagrangian
ω
qL,R → ei 3 qL,R
(
B: (2.46)
`L,R → `L,R

where q stands for quark and ` for lepton fields. The associated conserved current is
∂LSM 1
JBµ = δq = qγ µ q (2.47)
∂∂µ q 3
∂ 1 ~ · J~ ⇐⇒ ∂ B(t) = 0
Z Z
0= ∂µ JBµ = ∂0 JB0 + ∂i JBi = d x q†q +
3
d3 x ∇ (2.48)
∂t V 3 ∂t
|RV {z }
~
dσ n̂·J=0
σV

If we now sum over quarks Z X1


B(t) = d3 x q†q (2.49)
V q 3
and we exploit the following transformation properties,
C



 q † q −→ −q † q
(2.50)

P
 q † q −→ +q † q

 † T


q q −→ +q q
2.4. SAKHAROV CONDITIONS 21

which are easy to demonstrate

(2.51)
C
q † q −→ (iγ2 q)T iγ2 q ? = −q T γ2T γ2 q ? = q T q ? = −q † q
(2.52)
P
q † q −→ (iγ0 q)† iγ0 q = q † q
(2.53)
T
q † q −→ (iγ1 γ3 q ? )† iγ1 γ3 q ? = q T q ? = q † q

we have
CBC −1 = −B (2.54)
(CP )B(CP )−1 = −B (2.55)
T BT −1 = B (2.56)
then
hB(t)iT = Tr(ρ(t)B(t)) = Tr CC −1 ρ(t)CC −1 B(t)
 

= Tr C −1 ρ(t)C C −1
 
| | {zBC} = −hB(t)iT
{z } (2.57)
ρ(t) −B

=⇒ hB(t)iT = 0
We exploited that ρ(t) = eiHt ρ(0)e−iHt therefore, imposing per assurdo [C, H] = 0,
[C,H]=0
C −1 ρ(t)C = C −1 eiHt ρ(0)e−iHt C = eiHt C −1 ρ(0)C e−iHt = ρ(t) (2.58)
| {z }
ρ(0)

where we have assumed a symmetric early universe so that C −1 ρ(0)C = ρ(0). Hence, if per
assurdo C is conserved, we obtain hB(t)iT = 0 i.e. a forever symmetric universe. The necessity
of  can be demonstrated by the same argument as of C .
CP

3. Out of equilibrium: per assurdo if we are at thermal equilibrium the density matrix would have
the form ρ(t) = e−H/T . Calling θ = CP T we have

hB(t)iT = Tr e−H/T B(t) = Tr θθ−1 e−H/T θθ−1 B(t)


   

= Tr θ| −1 e{z
−H/T
θ} θ−1 B(t)θ = −Tr e−H/T B(t) = −hB(t)iT
   
| {z } (2.59)
e−H/T −B(t)

=⇒ hB(t)iT = 0

where we exploited the CP T theorem and B −→ −B. Also in this case the thermal equilibrium
CP T

hypotesis leads to a symmetric universe.

We have seen how powerful symmetries are to demonstrate theorems in a QFT sound way.
22 CHAPTER 2. DISCRETE SYMMETRIES
Chapter 3

Determination of the CKM matrix

3.1 Chiral symmetries of QCD


We continue our study to determine the parameters of the SM. We have to stop our discussion to
address some fondamental properties of the QCD Lagrangian: the chiral symmetries of QCD. By
means of these symmetries we can determine the hadronic matrix elements that are crucial in order
to extract the VCKM elements. Since QCD is non perturbative at low energies (i.e. αS ∼ 1) the only
way to compute the form factors of the hadronic matrix elements is to exploit either lattice QCD, a
sophisticated approach based on first principles of quantum field theories, or symmetry arguments. We
will follow the second approach. So we are interested in
 
LfQCD (3.1)
X
= / qk − mk q k qk
q k iD
k=u,d,s,c,b,t

where we sum over all quarks. At the energy scales of√ interest the EW phase transition has been
occurred already so the quarks have masses mk = vyk / 2. Moreover

(3.2)
 
Dµ qk = ∂µ + igS Gaµ ta qk a = 1, . . . , 8

where ta are the 8 generators of the SU (3)c group. Now let us analyze the mass spectrum of this QCD
Lagrangian. Since the masses are running parameters whose values depend on the energy scale we
specify a energy scale of the mass of the quark itself.

mu md ms mc mb mt
2.2 MeV 4.7 MeV 94 MeV 1.3 GeV 4.2 GeV 173 GeV
 1 GeV > 1 GeV
light heavy

We see that there’s a kind of splitting in the QCD mass spectrum in two groups of different orders
of magnitude. In a first approximation we can write the QCD Lagrangian as

LfQCD = Lf,light f,heavy


QCD + LQCD (3.3)

where in Lf,light
QCD we neglect the light quark masses mu = md = ms = 0:

X  
Lf,light (3.4)
X
QCD = q k iD
/ qk = q kL iD
/ qkL + q kR iD
/ qkR .
k=u,d,s k=u,d,s

This Lagrangian enjoys a global symmetry group which is

U (3)L ⊗ U (3)R = U (1)V ⊗ U (1)A ⊗ SU (3)L ⊗ SU (3)R 1 . (3.5)


1
We exploited U (N ) ' SU (N ) ⊗ U (1) which is properly an isomoprhism between the relative algebras.

23
24 CHAPTER 3. DETERMINATION OF THE CKM MATRIX

Defining the quark triplet  


u
q ≡ d (3.6)
 
s
we can find the conserved currents associated with the subroups U (1)V,A and SU (3)L,R

 U (1)V

q 0 = e−iα q JVµ = qγ µ q (3.7)


The conserved charge is nothing but the total (light) baryon number
Z
B= d3 x q † q (3.8)

 U (1)A

q 0 = e−iαγ5 q JAµ = qγ µ γ5 q (3.9)


Obviously this symmetry is not there in the limit of massive quarks.

 SU (3)L,R

(3.10)
a
qL = UL qL = e−iαL ta qL a
JµL = q L γµ ta qL
qR = U R qR = e −iαa
R ta qR a
JµR = q R γµ ta qR (3.11)

However it is known that at low energies there are not free quarks, so a quark-antiquark condensate of
mass scale Λ3 has to form such that SSB occurs:

hqqi 6= 0 (3.12)

Indeed, lattice QCD studies observed this chiral symmetry phase transition generating the condensate.
The consequence of this SSB is that mesons and baryons are not massless. The initial global symmetry
group is broken into the coset
hqqi6=0
SU (3)L ⊗ SU (3)R −→ SU (3)V (3.13)
| {z } | {z }
dim G=8+8=16 dim H=8

In fact
(3.14)
G
0 6= hqqi = hq L qR + q R qL i −→ hq L UL† UR qR + q R UR† UL qL i
and the condensate hqqi is not invariant under G if αL a 6= αa : the group is explicitly broken by the
R
condensate. However, if αR = αL , our chiral symmetry breaks into a unique SU (3)V symmetry which
a a

does not distinguish chirality. Still from the number of broken generators (8) of the SU (3)L ⊗ SU (3)R
group we expect 8 massless Goldstone bosons. These are, being the lightest particles made of u, d, s
quarks, the following mesons:

π 0 = (uu + dd) K 0 = (sd) K + = (su)


π + = (du) K 0 = (sd) K − = (us)
π − = (ud) η = (ss)

mπ 0 mπ± mK 0 mK 0 mK ± mη
135 MeV 140 MeV 500 MeV 500 MeV 495 MeV 550 MeV

We see that while the pions have masses about 100 MeV the kaons and eta are about 5 times
heavier. So the SU (3)V symmetry is not an accurate approximation: we cannot consider the kaons
3.1. CHIRAL SYMMETRIES OF QCD 25

as being the massless Goldstone bosons. However in certain cases it can be useful to exploit this
large symmetry. Anyway the mass spectrum of pions and kaons suggests we can further restrict our
symmetry group to a smaller one. The kaons and eta are heavier since they are made of strange quarks
(i.e. they carry strangeness) which are more that one order of magnitude heavier than up and down
quarks. So we can just repeat the reasoning setting mu = md = 0 but ms 6= 0. In this limit we expect
the pions to be massless while the kaons and eta to be massive. Therefore, defining the SU (2) doublet
!
u
q= (3.15)
d

the transformation properties of q under the symmetry group SU (2)L ⊗ SU (2)R (the so-called Nf = 2
chiral symmetry) and the associate conserved global currents are
σa
(3.16)
a σa
SU(2)L : qL0 = UL qL = e−iαL 2 qL a
JµL = qLγ µ qL
2
σa
(3.17)
0 a σa
SU(2)R : qR = UR qR = e−iαR 2 qR a
JµR = q R γ µ qR
2
As before, we know from experiments and theory (lattice QCD) that the symmetry is broken by a
quark-antiquark condensate hqqi 6= 0

SU (2)L ⊗SU (2)R


0 6= hqqi = hq L qR + q R qL i −→ hq L UL† UR qR + q R UR† UL qL i 6= hqqi (3.18)

However, if UL = UR , i.e. αL
a = αa , the condensate is invariant and the residual symmetry group is
R

hqqi6=0
SU (2)L ⊗ SU (2)R −→ SU (2)V (3.19)

If the remnant group of symmetry is SU (2)V we have 3 broken generators and we expect 3 Goldstone
bosons, that are indeed the three pions π 0 and π ± . Clearly, the SU (2)V symmetry is a better symmetry
than the SU (3)V one, however, even SU (2)V has to be violated at low-energy scales since the up and
down quarks are massive. Let us find the conserved currents and the charges associated with SU (2)V

µ σa
Z
µ
Ja,V = qγ q Ia = d3 x JaV
0
(x) (3.20)
2
or explicitly
1
  
µ 1
Z
µ µ 

 J1V = uγ d + dγ u I1 = d3 x u† d + d† u
 
2

 








 2
i
 
i
Z
(3.21)
 
Jµ = dγ µ u − uγ µ d , d3 x d† u − u† d .
 
2V I2 =


 2 

 2Z
I3 = 1 d3 x u† u − d† d
 
1
     
 µ µ µ

J3V = uγ u − dγ d


2 2
It is straightforward to check that these charges form a closed algebra colled isospin algebra

[Ia , Ib ] = εabc Ic (3.22)

thus they form a representation of the SU (2) group. If we define the currents and charges
(
uγµ d (+) Z
±
JµV = 1
JµV ± 2
iJµV = I ±
= ±
d3 x J0V (3.23)
dγµ u (−)

the matrix elements of I ± between the states |I, I3 , pi and |I 0 , I30 , p0 i are found to be
q
hI 0 , I30 , p0 |I ± |I, I3 , pi = I(I + 1) − I3 (I3 ± 1) hI 0 , I30 , p0 |I, I3 , pi
q (3.24)
p − p~0 )
= I(I + 1) − I3 (I3 ± 1) (2π)3 δI,I 0 δI30 ,I3 ±1 2p0 δ 3 (~
26 CHAPTER 3. DETERMINATION OF THE CKM MATRIX

Finally, we show that electromagnetic interactions violate isospin, that is they are not invariant
under the isospin group SU (2)V . In fact, the conserved currents associated with the symmetry group
U (2)V = U (1) ⊗ SU (2)V are JUµ (1) = qγ µ q and JaV
µ
= qγ µ σ2a q with q = (u, d) and the electromagnetic
current can be written as a linear combination of them
2 1 σ3
µ
Jem = uγ µ u − dγ µ d = a JUµ (1) + b J3V
µ
= a qγ µ q + b qγ µ q (3.25)
3 3 2
where a = 1/6 and b = 1. Now, the piece of the electromagnetic current proportional to the coefficient
a is invariant under an SU (2)V isospin transformation while the one proportional to b is not. Therefore,
electromagnetic interactions violate isospin. The underlying physical reason of this result is that
electromagnetic interactions distinguish between the component of the isospin doublet q = (u, d) since
u and d quarks have different electric charges.
We proceed now to calculate hadronic matrix elements, which are relevant to extract information
on the VCKM matrix, using QCD chiral symmetries.

3.2 |Vud | from superallowed β-decays


Superallowed β−decays are processes like
14
O −→14 N e+ νe (3.26)
34
Cl −→34 S e+ νe (3.27)
26
Al −→26 Mg e+ νe (3.28)
that involve transitions (using the J P notation) 0+ → 0+ due to the process at quark level

u −→ de+ νe (3.29)

Let us focus on the first of these processes. Since the mass difference between the nuclei is ∆M ∼
5 MeV  MW we can use the effective Fermi theory described by the following Lagrangian

GF ? 
dγµ (1 − γ5 )u ν e γ µ (1 − γ5 )e (3.30)
 
LF = − √ Vud
2

and the following Feynman diagram:

14 O(p)
..
.. ..
.. 14 N(p0 )
u ? d
Vud
W−
νe (k)

e+ (k 0 )
The amplitude of the process is given by

iGF ? 14
M = − √ Vud h N(p0 )|dγµ (1 − γ5 )u|14 O(p)i u(k)γ µ (1 − γ5 )v(k 0 ) (3.31)
2 | {z }
hadronic matrix element

The hadronic matrix element is the difficult part to compute: we cannot exploit a perturbative approach
since at low energy αs ∼ 1 but we can exploit the chiral symmetries we have just studied. Since the
process is a 0+ → 0+ transition the axial current does not contribute because of parity conservation

h0+ |dγ µ (1 − γ5 )u|0+ i = h0+ |dγ µ u|0+ i . (3.32)


3.2. |VU D | FROM SUPERALLOWED β-DECAYS 27

Noting that the bra and ket states have mass dimension −1 and that [dγ µ u] = M 3 , we deduce that
the hadronic matrix element has a mass dimension +1. Moreover, it carries a Lorentz index so, since
we have the only dependence on the initial and final momenta p and p0 , it can be expressed as

h14 N(p0 )|dγ µ u|14 O(p)i = F (q 2 )(p + p0 )µ + G(q 2 )(p − p0 )µ (3.33)

where F and G are dimensionless functions called form factors and q = p − p0 . Notice that in
principle the form factors could depend on scalars as p2 , p0 2 , p · p0 , q 2 but p2 = m2O , p0 2 = m2N and
q 2 = m20 + m2N − 2p · p0 so there is only one independent scalar involved. Now we can exploit the
SU (2)V symmetry and the fact that JV−µ = dγ µ u is conserved
0
0 = ∂µ JV−µ ⇐⇒ 0 = ∂µ h14 N(p0 )|dγ µ u(x)|14 O(p)i = ∂µ h14 N(p0 )|dγ µ u(0)e−i(p−p )·x |14 O(p)i
⇐⇒ 0 = −i(p − p0 )µ h14 N(p0 )|dγ µ u(x)|14 O(p)i = −i (p2 − p02 ) F (q 2 ) + q 2 G(q 2 ) (3.34)
| {z }
SU (2)V
m2O −m2N → 0

where we exploited the SU (2)V limit in which mu = md = 0 and so the nuclei mass difference goes to
zero. Therefore, by symmetry arguments we can set one form factor to zero

∂µ JV−µ = 0 ⇐⇒ G(q 2 ) = 0 . (3.35)

Actually by definition we insert a 1/2 factor


1 14
h N(p0 )|dγ µ u|14 O(p)i = F (q 2 )(p + p0 )µ ≈ F (0) 2pµ (3.36)
2
since qR = p − p0 → 0 being q 2 ∼ few (MeV)2 . In order to compute F (0) we exploit the conserved charge
I − = d3 xJV−0 (x) taking µ = 0 and integrating in the spatial coordinates
1 14 1
Z
h N(p0 )|I − |14 O(p)i = d3 x h14 N(p0 )|dγ 0 u|14 O(p)i
2 2
1 14
Z
0
= h N(p )| d3 x (dγ 0 u)(0)e−i(p−p )·x |14 O(p)i
0
2 (3.37)
0 00 1
p − p~0 ) e| −i(p{z−p )t} h14 N(p0 )|dγ 0 u|14 O(p)i .
= (2π)3 δ 3 (~
0
e−iq t ≈1
|2 {z }
F (0)2p0

On the other hand, we can calculate the hadronic matrix element also by means of eq. (3.24) finding
1 14 1
h N(p0 ) |I − | 14 O(p)i = h1, 0, p0 |I − |1, 12 , pi
2 | {z } | {z } 2
7p7n 8p6n
(3.38)
1s 1√
= I(I + 1) − I3 (I3 − 1) h1, 0, p0 |1, 0, pi = p − p~0 ) .
2 2p0 (2π)3 δ 3 (~
2 | {z } | {z } 2
2 0

Equating the two results we get that


1
F (0) = √ (3.39)
2
that is valid in the SU (2)V limit, i.e mu = md = 0. Comparing the expected rate of the process, which
depends on |Vud |2 , with the experimental rate we get
δVud
|Vud | = 0.974147 ± 0.00021 ' 2 × 10−4 . (3.40)
Vud
We stress the high precision of the result, well below the permille level.
2
Remembering that q = (u, d) and T3 = (1/2, −1/2), the value of I3 for 14 O is 8p6n = 22u + 20d = 11 − 10 = 1 and
that of 14 N is 7p7n = 0. Instead the value of I is more difficult to establish. In general the ground state of a light nucleus
is characterized by the minimum value of I = |p − n|/2. In this case however 14 N is in an excited state forming with 14 O
and 14 C the so-called “isobaric triplet”.
28 CHAPTER 3. DETERMINATION OF THE CKM MATRIX

3.3 |Vus | from semileptonic kaon decays


In order to extract the matrix element Vus , we need to consider tree-level charged-current processes
with an underlying s → u flavor transition, such as

K 0 −→ π − e+ νe (3.41)

which is a 0− −→ 0− transition. At the quark level, the relevant transition is s → ue+ νe which is
described by the following Feynman diagram:

d d − 0
K 0 (p) π (p )
s ?
u
Vus
W−
νe (k)

e+ (k 0 )
In order to have an idea of where we stand, let’s estimate the rate of the process K 0 → π − e+ νe assuming
that QCD is perturbative at low energy. In NDA we find
G2F
Γ(s → ue+ νe ) ' m5 |Vus |2 (3.42)
192π 3 K
where we took the phase space of the three body muon decay. This result is “just” a factor of four off
from the correct result which is obtained by taking into account the relevant hadronic matrix element,
as we will see in the next sections. Comparing the theoretical prediction of Γ(K 0 → π − e+ νe ) with the
corresponding experimental result, one finds that
δVus
|Vus | = 0.2248 ± 0.0006 ∼ 3 × 10−3 . (3.43)
Vus
We observe that the accuracy of Vus is roughly one order of magnituce worse than that of Vud , due to
lower experimental resolutions as well as less precise theoretical predictions which are based on the
SU (3)V instead of the SU (2)V symmetry.

3.4 |Vcb | and |Vub | from semileptonic B meson decays


The CKM matrix elements |Vcb | and |Vub | are determined by the B-meson decays Bd0 −→ D− e+ νe and
Bd0 → π − e+ νe , respectively, where Bd0 = (bd). The relevant Feynman diagrams for these processes are

d d d d − 0
Bd0 (p) D− (p0 ) Bd0 (p) π (p )
b c b u
Vcb? ?
Vub
W− W−
νe (k) νe (k)

e+ (k 0 ) e+ (k 0 )
Since the b quark is involved in the above flavor transitions, the use of the SU (3)V symmetry is no
longer justified. Another effective theory called heavy quark effective field theory has to be used.
Matching the theoretical results with the experimental measurements, it has been found that

|Vcb | = (40.5 ± 1.5) × 10−3 , |Vub | = (4.09 ± 0.39) × 10−3 . (3.44)

where we observe that the relative uncertainties in |Vcb | and |Vub | are comparable.
3.5. |VT D | AND |VT S | FROM B-B OSCILLATION 29

3.5 |Vtd | and |Vts | from B-B oscillation


The CKM matrix elements Vui and Vci (i = d, s, b) are completely determined experimentally by
semileptonic decays of π, K, D, B mesons, that are charged-current tree-level decays. Instead, with
regards to Vti , since we don’t have a top factory, we cannot exploit tree-level semileptonic top decays.
Therefore, Vts and Vtd have to be extracted by loop-induced processes, like B 0q ↔ Bq0 oscillation
(q = d, s), described by the following Feynman diagram (where the crossed diagram is understood)
Vib Vid?
b u d
i
B 0d (p) W W Bd0 (p0 )

d b
?
Vkd uk V
kb
This is a ∆F = 2 violating amplitude (F = b, d) which can be estimated in NDA as follows

G2F m2i G2F 2


2 mt
 
(3.45)
X
? 2 ?
M∼ (Vid Vib ) f 2 ∼ (V td V tb ) 2 ,
i=u,c,t
16π 2 MW 16π 2 MW

where for simpliticy we took i = k and f is the loop function which can be Taylor expanded as
f (xi ) = f (0) + f 0 (0)xi + · · · with xi = m2i /MW
2 . Since the CKM matrix V is unitary, it turns out that

the flavour-independent term f (0) doesn’t contribute to the ampliture and, in turn, the top quark
contribution is the by far dominant one.

3.6 Unitarity triangle and CP violation


We have analyzed the charged current quark sector in the SM Lagrangian:

g
 
Lquarks
C.C. / + dL + dL V † W
= − √ uL V W / − uL ,
2

or, diagrammatically:

µ µ

W+ W−
ui dj dj ui (3.46)

−i √g2 γ µ (1−γ
2
5)
Vij −i √g2 γ µ (1−γ
2
5)
Vji?

We can write the VCKM matrix in the following schematic way:

p
 
π π

 n νe K νe B νe 

 
 W W W 
 
e− e− e− 
 

 
π
 
 K D 
(3.47)
 
VCKM =
 D νe D νe B νe 
 .
W W W
 
 
 

 e− e− −
e 
 
b
 
 
 
 0
 Bd W W Bd0 B 0s W W Bs0 t 

W
30 CHAPTER 3. DETERMINATION OF THE CKM MATRIX

We explicitly see that while the first two rows’ elements are found studying semileptonic B-decays of
mesons the first two elements of the third row have to be found considering loop-induced processes
(oscillation of B mesons). Regarding the Vtb element one could study the top quark decay into the
bottom quark and W boson but since we are experimentally not that much accurate it is better to
exploit the unitarity of the VCKM to set it.
We have seen that the VCKM matrix depends on 3 mixing angles and a phase responsible of CP
violation. Therefore we can decompose the VCKM matrix as the product of three rotation matrices one
of which contains the CP phase. Calling θij the mixing angle between the ith and the jth generations
and cij and sij its cosine and sine respectively,
   
1 0 0 c13 0 s13 eiδ c12 s12 0
VCKM = V (θ23 )V (θ13 )V (θ12 ) = 0 c23 s23   0 1  −s12 c13 0
   
0 −s23 c23 −s13 e−iδ 0 c13 0 0 1
  (3.48)
c13 c12 c13 s12 s13 e−iδ
= −c23 s12 − s23 c12 s13 e
 iδ c23 c12 − s23 s13 s12 e iδ c13 s23  .

s23 s13 − c23 c12 s13 eiδ −s23 c12 − c23 s12 s13 eiδ c13 c23

The choice of putting the phase in V (θ13 ) is arbitrary: through a field redefinition it is always possible
to move the phase. We will stress many times that physical quantities have to be invariant under
phase redefinitions. From experiments on semileptonic β−decays we know that s13 ∼ 10−3 and
s23 ∼ 10−2 so we can assume c13 ∼ c23 ∼ 1. This fact triggers the possibility of expanding the VCKM .
Following Wolfenstein we can replace the set of 4 parameters (θ13 , θ12 , θ23 , δ) that is called the standard
parametrization with the Wolfenstein parameters (λ, A, ρ, η) defined in the following way:

|Vcb | Vub
λ ≡ |Vus |, A≡ , ρ − iη ≡ . (3.49)
λ2 Aλ3
From the process K → πeν we have that λ ∼ 0.2 so an expansion in λ is possible. We can get conditions
on the VCKM matrix elements exploiting the unitarity:

VCKM VCKM = 1. (3.50)

So exploiting the fact that |Vub |  1 since it couples first to third generation quarks,

λ2
|Vud |2 + |Vus |2 + |Vub |2 = 1 ⇐⇒ |Vud |2 + λ2 ' 1 ⇐⇒ |Vud | = 1 − + O(λ4 ). (3.51)
| {z } 2
1

One can show that the Wolfenstein parametrization of VCKM gives


   2 
Vud Vus Vub 1 − λ2 λ Aλ3 (ρ − iη)
4
(3.52)
2
VCKM =  Vcd Vcs Vcb  =  −λ 1 − λ2 Aλ2  + O(λ ).
   
Vtd Vts Vtb Aλ3 (1 − ρ − iη) −Aλ2 1

Now 
? ? ?
Vub Vud + Vcb Vcd + Vtb Vtd = 0
 (♥)
(3.53)
  
† ?
VCKM VCKM = 0 ⇐⇒ Vus Vud + Vcs? Vcd + Vts? Vtd = 0 (♦)
i6=j 
V ? V + V ? V + V ? V = 0

(♣)
ub us cb cs tb ts

Substituting the values of the matrix element and we get


 ?
 Vub − λVcb? + Vtd = 0
|{z} (♥)

 | {z } |{z}
 ∼λ3 ∼λ3 ∼λ3



 V ? + V + V ?V = 0

(♦)
us
|{z}
cd
|{z} td
| ts{z } (3.54)


 ∼λ ∼λ ∼λ 5

? ?

λVub + Vcb Vcs + Vts = 0 (♣)





| {z } | {z } |{z}
∼λ4 ∼λ2 ∼λ2
3.6. UNITARITY TRIANGLE AND CP VIOLATION 31

We note that only in the (♥) equation all the three terms have the same order of magnitude in λ:
this tells us that this is indeed the equation to study in order to experimentally check unitarity of the
VCKM . In fact the three terms of the (♥) equation define a triangle, the so called unitarity triangle.
We can rewrite the (♥) equation dividing by Vcb? Vcd
?V
Vub Vtb? Vtd
(3.55)
ud
+ +1=0.
Vcb? Vcd Vcb? Vcd
Defining
?V
Vub q
Vtb? Vtd q
(3.56)
ud
Rb ≡ = ρ2 + η 2 , Rt ≡ = (1 − ρ)2 + η 2 .
Vcb? Vcd Vcb? Vcd
where the last equalities of 3.56 have been obtained exploiting the Wolfstein parametrization of
eq. (3.52). Notice that eq. (3.55) defines a triangle in the (ρ, η) complex plane:
η

(ρ, η)
Rb α Rt

γ β
(0, 0) (1, 0) ρ

If we did the same thing with the (♦) and (♣) equations we would have ended up with very flattened
triangles. We also find that
2η(1 − ρ) 2ρη
sin 2β = sin 2γ = . (3.57)
(1 − ρ)2 + η 2 ρ2 + η 2
Let us see how we can test the unitarity of the VCKM matrix. Since the area of the triangle is clearly
invariant under phase redefinitions of the VCKM , it can be seen as a measure of CP violation in the
SM. One can see this observing that, in the limit η → 0 (remember that where η is the source of CP
violation in the Wolfenstein parametrization), the area of the triangle vanishes.
We can define the Jarlskog invariant as

JCP = 2 × Area4 = Rt sin β. (3.58)

Then, any CP violating observable has to be proportional to JCP . In particular, observables which
survive in the limit η → 0 are not CP violating observables.
So far, we have found the unknown matrix elements of VCKM exploiting experimental results and
therefore we haven’t done any test of the SM. However, since the number of observables at our disposal
is much bigger than the numbers of unknown parameters (λ, A, ρ, η), we can perform a consistency
check of the unitarity triangle by fitting the (ρ, η) parameters and seeing if we obtain a consistent
result. Let us see how to proceed.
 |Vub |: we have that
q
|Vub |2
|Vub (B → πeν)| = Aλ3 ρ2 + η 2 =⇒ ρ2 + η 2 = = const. (3.59)
|Vcb |2 |Vus |2
that is the equation of a circumference centered in the origin of the (ρ, η) plane.

 ∆Md,s : it is an observable related to the B mesons oscillation. We have


 h i
∆Md ∝ |V ? Vtb |2 ∼ (1 − ρ)2 + η 2 λ6 =⇒ (1 − ρ)2 + η 2 = const.
(3.60)
td
∆M ∝ |V ? V |2 ∼ λ4
s ts tb
32 CHAPTER 3. DETERMINATION OF THE CKM MATRIX

Sψks (ρ, η)
0.4
0.2
0 0.6
-0.2 0.4
0.6 -0.4
0.4 0.2
0.2 0
1
sin 2β 0 −0.2
2
−0.2 −0.4
−0.4 −0.6
−0.6
4
3
2
−4 −3 1
0
−2 −1 −1 η
0 1 −2
ρ 2 3 −3
4 5 −4

Figure 3.1: Level sets of Sψks varying the value of β.

thus ∆Md and ∆Md /∆Ms define two circumferences centered in (1, 0) in the (ρ, η) plane. Actually,
the observable ∆Md /∆Ms is theoretically cleaner than ∆Md as many hadronic uncertainties
related to the hadronic matrix elements cancel in the ratio.

 Sψks : it is an observable related to CP violation in b → d transitions:


Im (Vtd? Vtb )2
 
η(1 − ρ) 1
Sψks ∝ ∼ 2 2
= sin 2β. (3.61)
∆Md (1 − ρ) + η 2
This is the equation of a conical centered in (1, 0), parametrized with β. In particular 12 sin 2β ∈
[− 12 , 12 ] and for β = π/4 we have simply a line η = 1 − ρ. In general, the level sets of the function
are lines converging in (1, 0) as shown in Fig.(3.1).

 εk : it is an observable related to CP violation in s → d transitions:


Im (Vtd? Vts )2
 
∼ Im (1 − ρ − iη)2 ∼ (1 − ρ)η. (3.62)
 
εk ∝
∆Mk
This is the equation of an hyperbola centered in (1, 0).
All these observables can be compared in a single plot. In Fig.(3.2) we show qualitatively all the
constraints and we see that there is a region where all the constraints overlap, i.e. the unitarity
condition is satisfied being the unitarity triangle well defined. Therefore, we conclude that very likely,
flavor violation and CP violation in flavor changing processes are dominated by the CKM mechanism.
Let us conclude this section reporting the best result obtained so far in the determination of the
magnitudes of the VCKM elements:
0.97427 ± 0.00015 0.22534 ± 0.00065 0.00351+0.00015
   
|Vud | |Vus | |Vub | −0.00014
 |Vcd | |Vcs | |Vcb |  = 0.22520 ± 0.00065 0.97344 ± 0.00016 0.0412+0.0011 (3.63)
   
−0.0005 
|Vtd | |Vts | |Vtb | 0.00867+0.00029
−0.00031 0.0404+0.0011
−0.0005 0.999146+0.000021
−0.000046

The Wolfenstein parameters had also been determined:


λ = 0.2257+0.0009
−0.0010 A = 0.814+0.021
−0.022 ρ = 0.135+0.031
−0.016 η = 0.349+0.015
−0.017 . (3.64)
3.7. LEPTONIC PION DECAY π → `ν` 33

η
∆Md
∆Ms
Sψks εk ∆Md

(1, 0)
ρ

|Vub |2

Figure 3.2: Unitarity triangle plot showing that, very likely, flavor violation and CP violation in flavor
changing processes are dominated by the CKM mechanism.

3.7 Leptonic pion decay π → `ν`


The pion decay into leptons π + → l+ νl where π + = (u, d) and l = e, µ is a very important
process as it is the main source of atmospheric neutrinos. Within the SM, π + → l+ νl is in-
duced already at tree-level by charged-current interactions as illustrated by its Feynman diagram
u l+ (k)

g g
π + (pπ )
W+

d νl (k 0 )
Since q 2 ∼ m2π  MW 2 , the Fermi theory provides an excellent approximation of the SM and therefore

we can effectively consider the following Feynman diagram


u l+ (k)

π + (pπ ) G
√F
2
d νl (k 0 )
where GF is universal, i.e. it is the same irrespective of the fermion generations. The lepton flavor
universality (LFU) in electroweak processes arises from the flavor blind nature of the gauge couplings
g and g 0 . The only breaking of LFU, i.e. the only dependence of physical observables on the flavor,
may come from lepton masses. Let us test this fact computing the rate of π + → l+ νl . We start from
the Fermi Lagrangian:
GF ? 
Lπ→lν dγµ (1 − γ5 )u ν l γ µ (1 − γ5 )l] (3.65)

F = − √ Vud
2
from which we get the following amplitude
iGF ?
M(π → lν) = − √ Vud h0|dγ µ (1 − γ5 )u|π(pπ )iu(k 0 )γ µ (1 − γ5 )v(k) (3.66)
2

Let us evaluate the hadronic matrix element. Consider the vector part of the current. Since [dγ µ u] = M 3 ,
[|πi] = M −1 , and [h0|] = M 0 we can write

h0|dγ µ u|π(pπ )i = const. × fπ pµπ (3.67)


34 CHAPTER 3. DETERMINATION OF THE CKM MATRIX

where [fπ ] = M 1 . Actually fπ is not a form factor but a constant, called pion decay constant as in two
body decays all the kinematical invariants p2π , k 2 , k 2 and k · k 0 are fixed by the masses. Let us prove
0

now that this matrix element is zero. The pion is a pseudoscalar, so under parity it transforms as

P |π(p0π , p~π )i = −|π(p0π , −~


pπ )i (3.68)

while the current transforms as a vector

P (dγ µ u)P −1 = ηνµ (dγ ν u) (3.69)

where ηνµ stands for (−1)µ defined in previous sections. Therefore, Eq. (3.67) can be rewritten as

h0|P −1 P (dγ µ u)P −1 P |π(p0π , p~π )i = −ηνµ h0|dγ ν u|π(p0π , −~


pπ )i = const. × fπ ηνµ pνπ (3.70)

which gives for µ = 0 and p~π = 0:

− h0|dγ 0 u|π(p0π , 0)i = const. × fπ p0π . (3.71)

On the other hand, from Eq. (3.67) we have

h0|dγ 0 u|π(p0π , 0)i = const. × fπ p0π =⇒ const. × fπ p0π = 0 . (3.72)

Since in general p0π 6= 0 and fπ 6= 0, it turns out that const. = 0 and therefore

h0|dγ µ u|π(pπ )i = 0 . (3.73)

Repeating the steps for the axial part of the current we see that it actually survives. We could have
guessed this result from the beginning noting that the hadronic matrix element between a pseudoscalar
state and the vacuum can be different from zero only for pseudoscalar or axial-vector currents. We set

h0|dγ µ γ5 u|π(pπ )i = i 2pµπ fπ (3.74)

and the resulting amplitude reads:


?
M = GF Vud fπ pµπ u(k 0 )γ µ (1 − γ5 )v(k) = GF Vud
?
fπ u(k 0 )(k/ + k/0 )(1 − γ5 )v(k)
?
= −GF Vud fπ ml u(k 0 )(1 + γ5 )v(k),

having exploited k/v(k) = −mv(k), u(k 0 )k/0 = mu(k 0 ) and {k/, γ5 } = 0. The modulus squared of the
unpolarized amplitude, being the pion a spin 0 particle, reads

|M|2 = G2F |Vud |2 fπ2 m2l Tr[u(k 0 )(1 + γ5 )v(k) (u(k 0 )(1 + γ5 )v(k))† ]
(3.75)
| {z }
v(k)(1−γ5 )u(k0 )
0 0
= G2F |Vud |2 fπ2 m2l Tr[u(k )u(k )(1 + γ5 )v(k)v(k)(1 − γ5 )].

Exploiting the energy projectors

u(k 0 , s)u(k 0 , s) = k/0 + mν = k/0 (3.76)


X X
v(k, s)v(k, s) = k/ − ml ,
s s

we compute the trace:

Tr[k/0 (1 + γ5 )(k/ − ml )(1 − γ5 )] =Tr[k/0 k/(1 − γ5 )2 ] = 2Tr[k/0 k/(1 − γ5 )]


(3.77)
= 8kµ0 kν η µν = 4(m2π − m2l )

and finally
|M|2 = 4G2F |Vud |2 fπ2 m2l (m2π − m2l ). (3.78)
The rate of the process is given by
1
Z
Γ= |M|2 dΦ, (3.79)
2mπ
3.7. LEPTONIC PION DECAY π → `ν` 35

where
d3 k d3 k 0
dΦ = × (2π)4 δ 4 (pπ − k − k 0 ). (3.80)
(2π)3 2k 0 (2π)3 2k 0 0
We compute the phase space integral

1 d3 k d3 k 0
Z
(3.81)
0
Φ= δ(mπ − k 0 − k 0 ) δ 3 (~
pπ − ~k − k~0 )
16π 2 k0 k0 0

in the pion rest frame (~


pπ = 0 and ~k = −~k 0 ) and after integrating over ~k 0 we obtain

4π ∞ |k|2 1 ∞ |k|
Z Z
00
(3.82)
0
Φ= 2
d|k| 0 0 0 δ(m π − k 0
− k ) = d|k| 0 0 δ(mπ − k 0 − k 0 )
16π 0 k k 4π 0 k
1 ∞ |k|
Z  q 
= d|k| q δ mπ − |k| − |k|2 + m2l (3.83)
4π 0 2
|k| + m 2
l

where we have exploited that k 0 = |k 0 | = |k| since mν = 0. Now we perform the change of variable
q x2 − m2l
x = |k| + |k|2 + m2l =⇒ |k| = (3.84)
2x
|k|d|k| d|k| dx
dx = d|k| + q =⇒ q = (3.85)
|k|2 + m2l |k|2 + m2l x

and then ! !
1 ∞ dx x2 − m2l 1 m2π − m2l
Z
Φ= δ(x − mπ ) = . (3.86)
4π 0 x 2x 8π m2π
Finally we can write the rate of the pion decay:
!2
G2 m2
Γ(π → lνl ) = F |Vud |2 fπ2 m2l mπ 1 − 2l . (3.87)
4π mπ

Note that, as we argued before, there is an explicit dependence on the mass of the lepton involved, the
so-called helicity suppression. Since the pion is a spin 0 particle we need a spin 0 final state. However,
if we work in the ml → 0 limit, both l+ and νl are left-handed. Given that chirality coincides with
helicity in the massless limit we would end up with a spin 1 final state since the antilepton and the
neutrino would travel in opposite direction in the center of mass frame. The only way to avoid this is
through an explicit dependence on the lepton mass so that we can flip chirality making the antilepton
right-handed so that the helicities of l+ and νl can point in opposite directions and cancel. Therefore,
the ml dependence of M(π → lν) is the result of the total angular momentum conservation.
Now we define the ratio
!2
Γ(π → eν) m2 1 − m2e /m2π
Rπe/µ ≡ = 2e ≈ 10−4 (3.88)
Γ(π → µν) mµ 1 − m2µ /m2π

showing that the electron decay channel is very much suppressed compared to muon chsannel because
of the helicity suppression. This is quite controintuitive since the decay into electrons is kinematically
favoured, as one can see from the phase space factor.
Rπ provides us with a tool to check the validity of LFU. Comparing the experimental result
e/µ

(Rπe/µ )exp = (1.218 ± 0.014) × 10−4 (3.89)

with the tree level prediction of eq. (3.87) (Rπ )tree ' 1.283 × 10−4 , we find a discrepancy at the level
e/µ

of 5 ÷ 6 σ. Are we missing something or LFU is violated? The right answer is the former: we are
neglecting loop effect which can depend on the lepton masses. Clearly, loop corrections like
36 CHAPTER 3. DETERMINATION OF THE CKM MATRIX

u νl

W+ W+

d l+
called oblique corrections, do not affect the ratio as the contribution is the same for electrons and
muons in the final state. Instead, flavor-dependent loop corrections which affect Rπ are the following
e/µ

u νl u νl u γ νl

g g g g g g
+ + .
W+ W+ W+
γ γ
d l+ d l+ d l+
By an explicit calculation of the one-loop QED corrections, one can find that

(Rπe/µ )1−loop = (Rπe/µ )tree (1 + δem


e/µ
) (3.90)

with
αem mµ
 
e/µ
δem = −3 × log ≈ −few × 10−2 . (3.91)
π me
Taking into account one-loop effects one obtains (Rπ )1−loop ' 1.235 × 10−4 that is indeed in good
e/µ

agreement with the experimental result. It is remarkable that LFU tests, such as Rπ , probe the
e/µ

validity of the SM at the quantum level.


Now let us consider a kind of “inverse process” of the pion decay, that is the tau decay τ − −→ π − ντ .
We consider the Fermi Lagrangian
GF ?  µ
LτF→πν = − √ Vud dγ (1 − γ5 )u ν τ γ µ (1 − γ5 )τ . (3.92)
 
2
The amplitude is, by analogy with the pion decay,
?
M = −GF Vud fπ mτ u(k 0 )(1 + γ5 )u(k) (3.93)

In order to compute the rate we have to be careful since the initial state has non-zero spin
1 1
dΓ = |M|2 dΦ (3.94)
2mτ 2sτ + 1
| {z }
1
2

Now by analogy with the rate of the pion decay in Eq. (3.87) we obtain
!2
1 G2F m2
Γ(τ → πντ ) = |Vud |2 fπ2 m2τ mτ 1 − π2 . (3.95)
2 4π mτ
Note that, in contrast to the π → `ν case, the helicity suppression is no longer at work now as the
required tau lepton mass coincides with the energy scale of the process. Moreover, τ → πντ offers the
possibility to test LFU also in the e/τ and µ/τ sectors through the ratios
! !2 !−2
Γ(π → lν) mπ m2l m2 m2
R l/τ
≡ =2 1 − 2l 1 − π2 (3.96)
Γ(τ → πν) m3τ mπ mτ
which have been computed at the loop level in the SM finding an agreement with the experimental
results at 1 ÷ 2σ.
Chapter 4

Neutrino physics

Neutrinos physics is interesting because neutrino experiments recently discovered something new, rather
than giving only more precise measurements of SM parameters, or stronger bounds on unseen new
physics. ‘Solar’ and ‘atmospheric’ data directly show that lepton flavour is not conserved. The next
step was to identify the new physics responsible of these anomalies. The observed flavour conversions
are perfectly explained assuming oscillations of massive neutrinos, provided that mixing angles among
the SM neutrinos are unexpectedly large. On the other hand, present data strongly disfavor alternative
exotic possibilities, such as neutrino decay or oscillations into extra ‘sterile’ neutrinos νs (i.e. light
fermions with no SM gauge interactions, as opposed to the three ‘active’ SM neutrinos) and show some
hints for the characteristic features of oscillations. Oscillations can be directly seen by precise reactor
and long-baseline beam experiments, that are nowadays respectively testing the solar and atmospheric
anomalies. Non-oscillation experiments should detect neutrino masses, and test if they violate lepton
number. Understanding neutrino propagation will allow to do astrophysics, cosmology, geology using
neutrinos.
Neutrino oscillations necessarely require that neutrinos are massive particles. The origin of the
extremely small neutrino mass scale, mν ∼ (0.001 − 0.05) eV is still an open question in particle
physics. Presumably neutrino masses are of Majorana type and are the first manifestation of a new
scale in nature, MR ∼ v 2 /mν ∼ 1014 GeV, which will be never tested experimentally. This could be
the mass of new particles, maybe 3 right-handed neutrinos with a 3 × 3 matrix of Yukawa couplings.
Leptogenesis, charged lepton flavor violating processes such as µ → eγ and related processes could be
other manifestations of the new physics behind neutrino masses.

4.1 Dirac and Majorana masses


On general grounds, fermion mass terms connect left-handed (LH) and right-handed (RH) fields.
A Dirac-type mass connects LH and RH components of the same field:

− Lm
D = D (ψ L ψR + ψ R ψL ) = D ψψ , (4.1)

where the mass eigenstate is


ψ = ψL + ψR . (4.2)
A Majorana-type mass connects LH and RH components of conjugate fields:

− Lm c c c c
M = A(ψ L ψL + ψ L ψL ) + B(ψ R ψR + ψ R ψR ) = A χχ + B ωω (4.3)

where the mass eigenstates are self-conjugate fields

χ = ψL + ψLc ≡ χc , (4.4)
ω = ψR + c
ψR c
≡ω . (4.5)

In the above expressions ψ c and ψ c are defined as

ψ c = iγ2 ψ ? = Cγ0 ψ ? , ψ c ≡ (ψ)c = (iγ2 ψ ? )† γ0 = −iψ T γ2† γ0 = ψ T C (4.6)

37
38 CHAPTER 4. NEUTRINO PHYSICS

where C = iγ2 γ0 . Therefore, the transformations properties of ψL,R under charge conjugation are

ψLc = (ψL )c = iγ2 (PL ψ)? = PR (iγ2 ψ ? ) = (ψ c )R =⇒ ψ cL = ψ c PL (4.7)


c
ψR c ?
= (ψR ) = iγ2 (PR ψ) = PL (iγ2 ψ ) = (ψ )L ? c
=⇒ ψ cR = ψ PR ,c
(4.8)

showing that ψLc is a RH field while ψR


c is LH. For later convenience let us show that ψ̄ c ψ c = ψ̄ψ

ψ̄ c ψ c = (iγ2 ψ ? )† γ0 (iγ2 ψ ? ) = ψ T γ2† γ0 γ2 ψ ? = −ψ T γ0 ψ ? = ψ ? γ0T ψ T = ψ̄ψ (4.9)

From the above relations we can also find the following important property
c
ψL,R = C ψ TL,R (4.10)

which can be easily demonstrated as follows

C ψ TL = iγ2 γ0 (ψ † γ0 PR )T = iγ2 γ0 PR γ0 ψ ? = PR (iγ2 ψ ? ) = (ψ c )R = ψLc . (4.11)

The most general fermion mass including both Dirac and Majorana mass terms can be written as

−Lm c c
DM = D ψ L ψR + Aψ L ψL + Bψ R ψR + h.c. (4.12)
! !
D D
(4.9) A χ
= (χω + ωχ + χγ5 ω − ωγ5 χ) + Aχχ + Bωω = (χ, ω) D 2 . (4.13)
2 | {z } | {z } 2 B ω
ψ̄ψ+ψ̄ c ψ c ψ̄ψ−ψ̄ c ψ c

A diagonalization of the above mass matrix provides us with the following eigenvalues
p
(A + B) ∓ (A − B)2 + D2
m1,2 = (4.14)
2
while the eigenvectors are given by the self-conjugate fields

η1 = cos θ χ − sin θ ω ≡ η1c (4.15)


η2 = cos θ ω + sin θ χ ≡ η2c (4.16)

where the angle θ is given by


D
tan 2θ = . (4.17)
A−B
As a result, the most general mass term for a four-component fermion field describes two Majorana
particles with distinctive masses.
Extending the previous discussion to three fermion families, we end up with 3 × 3 mass matrices in
the flavour space. In the Dirac case, we already know that it is always possible to diagonalize a generic
mass matrix by means of two unitary matrices. Instead, Majorana mass terms satisfy the property
(4.10)
−LM = ψ L M ψLc + h.c. = ψ L M Cψ TL + h.c. = (ψ L M Cψ TL )T + h.c.
(4.18)
C=−C T
= − ψ L M T C T ψ TL + h.c. = ψ L M T Cψ TL + h.c. = ψ L M T ψLc + h.c.

so that M is a symmetric complex matrix, i.e. M = M T , which can be always diagonalized by only
one unitary matrix U
M = MT ⇐⇒ M = U M̂ U T . (4.19)
Let us notice that a Majorana mass term violates any U (1) symmetry (both global and local) by two
units. Indeed, under a U (1) transformation we have that

U (1) iα
ψL,R −→

e ψL,R U (1)
=⇒ c
ψ L,R M ψL,R c
−→ e2iα ψ L,R M ψL,R . (4.20)
ψ c U (1)
−→ eiα ψ cL,R .

L,R
4.2. THE SEE-SAW MECHANISM 39

This means that Majorana mass terms would violate the electric charge conservation by two units
which is experimentally not allowed. For this reason, neutrinos are the only candidates within the
SM to be Majorana particles as they are chargeless. However, Majorana neutrinos would violate the
leptonic number by two units inducing processes such as neutrinoless double β decay or K → πµ± µ± .
If observed, these processes would provide a clear cut proof of the Majorana nature of neutrinos.
Actually, there is no fundamental reason for the conservation of the lepton number. In fact, the SM
is built upon the gauge group SU (3)c ⊗ SU (2)L ⊗ U (1)Y which breaks into SU (3)c ⊗ U (1)em after SSB,
and the conservation of the lepton and baryon numbers are just the result of accidental symmetries.

4.2 The see-saw mechanism


In the SM neutrino are massless because the simple Higgs structure of the theory leads to a global
symmetry corresponding to lepton-number conservation which forbids the Majorana mass term ν̄Lc νL
and there are no right-handed neutrinos νR that could combine with νL to form a Dirac mass term. In
practice, neutrino are massless because of the restricted particle content of the SM. In order to see this,
let us remind the transformation properties of the lepton fields and the Higgs scalar under the SM
gauge group SU (2)L ⊗ U (1)Y
! !
1 ϕ+ 1
   
νL
LL = ∼ 2, − , eR ∼ (1, −1) , ϕ= ∼ 2, (4.21)
eL 2 ϕ0 2
where the hypercharge Y is defined by the relation Q = T3 + Y . All possible lepton bilinears are
L̄L eR ∼ (2, 1/2) × (1, −1) = (2, −1/2) (4.22)
L̄cL LL ∼ (2, −1/2) × (2, −1/2) = (1, −1) + (3, −1) (4.23)
ēcR eR ∼ (1, −1) × (1, −1) = (1, −2) . (4.24)
As a result, only the Yukawa coupling ēR LL ϕ + h.c. is allowed in the SM and we have a global symmetry
corresponding to lepton number conservation.
Extensions of the SM able to give masses to neutrinos include extra scalar particles and/or leptons.
The first attempt is to generate a Dirac mass term by introducing a νR . In the mass basis we obtain
vyν
(4.25)
mass
− LlY = LL ϕ̃yν νR + h.c. = ν L mν νR + h.c. , mν = √
2
where ϕ̃ = iσ2 ϕ? and νR is sterile under SU (2)L ⊗ U (1)Y since the term LL ϕ̃ transforms as (11, 0). The
experimental bounds on the neutrino masses translates into a bound on the neutrino Yukawa coupling
mν . 0.1 eV =⇒ yν . 10−12 (4.26)
which is exceedingly small especially when compared to the top Yukawa coupling yt ' 1. This
naturalness problem calls for a dynamical mechanism accounting for such small neutrino masses.
The see-saw mechanism introduces a heavy right-handed Majorana neutrino NR which is totally
sterile under the SM gauge group GSM . The most general Lagrangian invariant under GSM is
−Lsee−saw = LL ϕ̃yν NR + MR N cR NR + h.c.
mass v
= √ ν L yν NR + MR N cR NR + h.c.
2 (4.27)
vy
√ν
! !
0 2 2 χ
= (χ, ω) vy √ν
.
2 2
MR ω

Since we assume the right-handed neutrino to lie at the GUT scale (MR ∼ 1015 GeV), we end up with
the following mass eigenstates and mass spectrum
(vyν )2
η1 ' χ = νL + νLc m1 ' (4.28)
8MR
η2 ' ω = NR + NRc m2 ' M R (4.29)
40 CHAPTER 4. NEUTRINO PHYSICS

where we assumed the limit vyν  MR . We see that the see-saw mechanism provides a way to
dynamically suppress the νL mass without assuming any innatural value of the Yukawa coupling yν
!
1014 GeV
mν ≡ m1 ∼ 0.1 eV yν2 (4.30)
MR

The suppressing factor v/MR in m1 is enough to give the right order of magnitude of the neutrino
masses even for yν ∼ yt ∼ 1, provided the right-handed neutrino mass MR is at the GUT scale.

4.3 The PMNS matrix


Irrespectively of the nature of neutrino masses, flavor mixing effects in the leptonic sector appear, in
the lepton mass basis, in charged-current interactions
g 
(4.31)

/ + eL + eL U † W
LlCC = − √ ν L U W / − νL
2
where, if neutrinos are Dirac particles, the matrix U contains three flavour mixing angles and one CP
violating phase, in complete analogy with the CKM matrix in the quark sector. The situation changes
if neutrinos are Majorana particles. In order to understand this point, let us rotate the lepton fields by
flavour-dependent phases ( i ν
νL −→ eiφi νLi
e . (4.32)
ejL −→ eiφj ejL
As a result, LlCC changes into

g X i −iφν ij + iφej j
 
LlCC = −√ ν Le i U W
/ e eL + h.c.
2 i,j
(4.33)
g
 
e −φν ) X ν ν e −φe ) j
= − √ e| i(φ{z
1 1 / + e| i(φ{z
ν iL e| −i(φ{zi −φ1}) U ij W j 1 e + h.c. .
} L
2 }
i,j
1 phase n−1 phases n−1 phases

Now let us consider two cases:


 Dirac ν: In this case we can rotate all 2n lepton fields but we can subtract only 2n − 1 phases
contained in U . Indeed, a common phase rotation of fields leaves the Lagrangian invariant as it
is associated with the conservation of the total lepton number L = Le + Lµ + Lτ . Hence

n(n + 1) (n − 1)(n − 2) n=3


nphases
Dirac = − 2n + 1 = = 1 (4.34)
2 2

 Majorana ν: In this case we can rotate only charged lepton fields and therefore we can remove
only n phases of U . The phases in the neutrino sector cannot be reabsorbed as neutrino mass
terms are not invariant under phase rotations of neutrino fields. As a result, we find
n(n + 1) n(n − 1) n=3
nphases
Majorana = −n= = 3 (4.35)
2 2

Concerning the parametrization of the U matrix, we can write


 
eiα 0 0
Majorana
UPMNS Dirac
= UPMNS (θ; φ, ψ, γ)  0 eiβ 0 (4.36)
 
0 0 1

containing 3 angles, 1 Dirac phases and 2 Majorana phases for 6 parameters in total. Unfortunately, the
neutrino oscillation phenomenon which ultimately proved the neutrinos to be massive is not sensitive
to Majorana phases but only to the Dirac phase. However, we can exploit other processes to discover
whether neutrinos are Majorana or Dirac particles, as we will discuss shortly.
4.4. THE STANDARD MODEL FLAVOR GROUP 41

4.4 The Standard Model flavor group


Let us formalize what we have just seen using symmetry arguments. The SM flavor group Gflavor
SM is a
global symmetry group acting on the 3 × 3 generations space. Setting the Yukawa couplings to zero

Gflavor 5
SM (yu,d,e,ν = 0) = U (3) = U (3)QL ⊗ U (3)dR ⊗ U (3)uR ⊗ U (3)LL ⊗ U (3)eR . (4.37)

The transformation properties of fermion fields under Gflavor


SM are


QL −→ UQL QL

dR −→ UdR dR




u −→ U
R uR R u (4.38)

LL −→ ULL LL






eR −→ UeR eR

and since all the fermionic bilinears which preserve chirality are invariant under Gflavor
SM

/fL −→ f L Uf†L D
f L D / UfL fL ≡ f L D
/fL

(4.39)
R /fR −→ f R UfR D
/ UfR fR ≡ f R D
f D /fR

where f = u, d, ν, e, it is clear that kinetic terms, charged and neutral current interactions are invariant
under Gflavor
SM . Instead the SM Yukawa sector, which involves both L and R fermions, transforms as

−LY = QL yu ϕ̃uR + QL yd ϕdR + LL ye ϕeR + · · · + h.c. (4.40)


Gflavor
SM
= †
QL UQ †
y U ϕ̃uR + QL UQ
L u uR
y U ϕdR + LL UL† L ye UeR ϕeR + · · · + h.c.
L d dR
(4.41)

and therefore is not invariant under Gflavor


SM . Thus, if LY 6= 0, the residual flavour group is

Gflavor
SM (yu,d,e 6= 0, yν = 0) = U (1)B ⊗ U (1)e ⊗ U (1)µ ⊗ U (1)τ (4.42)

which states that baryon number is conserved and, considering massless neutrinos, lepton family
numbers are also separately conserved. If neutrinos are massive we have two possible cases:
 Dirac ν, yν 6= 0:
Gflavor
SM = U (1)B ⊗ U (1)L . (4.43)
The quark and lepton sectors are symmetric and the baryon and total lepton numbers are
conserved. Instead, quark and lepton family numbers are broken by the Yukawa couplings.

 Majorana ν, yν 6= 0:
Gflavor
SM = U (1)B . (4.44)
The total lepton number is violated by two units ∆L = 2 by the neutrino mass terms. Therefore,
we are left only with the conservation of the baryon number.
Let us conclude this section counting the SM input parameters.
 Bosonic sector: in addition to the three coupling constants (g, g 0 , gs ) related to gauge groups
SU (3)c ⊗ SU (2)L ⊗ U (1)Y , we have two other parameters (λ, µ2 ) from the Higgs sector as well
as the so-called θQCD parameter associated with the renormalisable and gauge invariant operator
1
LθQCD = θQCD Gaµν G̃a,µν G̃a,µν = εµναβ Gαβ . (4.45)
2
Notice that LθQCD violates the CP symmetry and therefore generates CP violating observables
such as the neutron electric dipole moment dn which is tightly constrained experimentally as
dexp
n ≤ 10−26 e cm. As a result θQCD < 10−10 . The required smallness of θQCD which is not
supported by symmetry arguments poses another naturalness problem in the SM which goes
under the name of QCD θ problem or strong CP problem.
42 CHAPTER 4. NEUTRINO PHYSICS

 Fermionic sector: from the quark sector we have 6 quark masses and 4 parameters (3 angles
and 1 phase) of the CKM matrix. In the lepton sector there are 6 charged lepton and neutrino
masses and the parameters of the PMNS matrix which are 3 angles and 1 or 3 phases depending
on whether neutrinos are Dirac or Majorana particles, respectively.
To sum up, the total number of SM input parameters are
(
26 Dirac ν
# of SM parameters = (4.46)
28 Majorana ν
These parameters cannot be predicted by the theory but have to be set by experiments. The crucial
point is to have more observables than input parameters so that, once the latters have been determined,
we can make predictions for other observables which can be experimentally tested.

4.5 Neutrino oscillation


In this section we will study another phenomenon arising from the fact that neutrinos are massive:
neutrino oscillation. Neutrinos produced with different flavors (flavor eigenstates) mix into each other
in different mass eigenstates by means of the UPMNS matrix. We will discuss atmospheric neutrinos
which will imply the mixing νµ ↔ ντ and solar neutrinos in where νe ↔ νµ,τ mixings are involved. The
crucial ingredient to have a mixing is the fact that UPMNS 6= 1. We remember that the lepton flavor
numbers Le , Lµ , Lτ are conserved individually (classically1 ) in the SM.
Neutrinos are produced in weak charged current processes, described by the Lagrangian
g
 
LCC = − √ Wµ+ J −µ + Wµ− J +µ , (4.47)
2
with 
µ
ν L γ eL interaction basis


J −µ = †
γ µ eL (4.48)
ν L Lν Le
 | {z }
mass basis

UPMNS

If the neutrinos are massless we have always the freedom to choose Lν = Le because we don’t have any
Yukawa matrix in the neutrino sector to diagonalize: if so the mixing matrix can be set to unity and
there is no neutrino oscillation. Therefore we immediately understand that the ocillation phenomenon
is provided only if the neutrinos are massive, independently on their Dirac or Majorana nature.

4.5.1 Oscillation in vacuum


Let us see in practice what is the neutrino oscillation and how we can detect this phenomenon. Let us
consider a source of neutrinos, which are always produced in a flavor eigenstate να , such as the charged
current processes π + −→ µ+ νµ or µ+ −→ e+ νe ν µ . After traveling for a dinstance L neutrinos interact
with a detector where their flavor is probed by means of another charged current process involving a
charged lepton of defined flavor, see fig. (4.1).

να L νβ ↔ eβ
? 
Source Detector

Figure 4.1: Schematic picture of a neutrino oscillation experiment.

We can distinguish two different kind of neutrino oscillation experiments


1. α 6= β: appearance experiments. These experiments measure transitions between different
neutrino flavors. If the final flavor to be searched for in the detector is not present in the initial
beam, the background can be very small. In this case, an experiment can be sensitive to rather
small values of the mixing angle.
1
As we shall see they are broken at the quantum level through anomalies.
4.5. NEUTRINO OSCILLATION 43

2. α = β: disappearance experiments. These experiments measure the survival probability


of a neutrino flavor by counting the number of interactions in the detector and comparing it
with the expected one. Since, even in the absence of oscillations, the number of detected events
has statistical fluctuations, it is very difficult to reveal a small disappearance and therefore to
measure small values of the mixing angle.

Let νi be mass eigenstates and να flavor eigenstates. Neutrinos are produced through charged
current interactions in a given flavor eigenstate that is the superposition of different mass eigenstates

|να i = Uαi |νi i. (4.49)

We define oscillation probability the following quantity

Pαβ (t) = |hνα (0)|νβ (t)i|2 (4.50)

that is the probability of detecting at the time t a neutrino of flavor β if at the time t = 0 such a
neutrino was of flavor α. From now on, we will assume that neutrinos are ultrarelativistic (which
is always an excellent approximation in the considered scenarios) and we will identify the distance
between source and detector L with the time of flight t (we assume natural units where c = 1).

N = 2 analysis
Let us start studying neutrino oscillation in a framework with two flavors. At the source we have

|νe (0)i = Ue1 |ν1 i + Ue2 |ν2 i (4.51)

The state at the detector is found applying the quantum evolution in the mass basis

|νe (t)i = Ue1 e−iE1 t |ν1 i + Ue2 e−iE2 t |ν2 i (4.52)

where the energies associated to the mass eigenstates are, assuming a monochromatic beam2
s
q
m2i m2i m2i
Ei=1,2 = p2 + m2i = p 1 + ≈ p + ≈ E + , (4.53)
p2 2p 2E

where we have exploited E ≈ p. Thanks to this expression we can get rid of an overall phase e−iEt .
The survival probability is defined as
m2
 (m2 2  2
2 −i 1L 2 2 −i 2 −m1 )L
Pee (L) = |hνe (0)|νe (t)i| = e 2E |Ue1 | + |Ue2 | e 2E

" # (4.54)
2 2 2 (m22 − m21 )L
=1 − 4|Ue1 | |Ue2 | sin ,
4E

where in the N = 2 case the U matrix has the form


!
cos θ − sin θ
U= (4.55)
sin θ cos θ

and therefore we find " #


∆m2 L
2
Pee (L) = 1 − sin 2θ sin 2
(4.56)
4E

The conservation of probability (which is guaranteed by the unitarity of UPMNS ) implies that the
conversion probability is given by
Peµ = 1 − Pee . (4.57)
2
Of course this assumption is unphysical. However it can be proved that a proper description of the beam with a wave
packet gives the same result we are about to derive.
44 CHAPTER 4. NEUTRINO PHYSICS

The phenomenon of neutrino oscillation can be observed experimantally only provided ∆m2 L/4E ∼
O(1), see Eq. (4.56). A useful numerical approximation which shows the values of the parameters for
which the oscillation is at work is the following
!
∆m2 L ∆m2 1 GeV L

' 1.27 . (4.58)
4E 1 eV2 E 1 km

Now we show in Fig. (4.2) the behavior of the oscillation probability as a function of the distance
travelled by the neutrino. We see that 1 − sin2 2θ ≤ Pee ≤ 1. Moreover we distinguish two cases:

 ∆m2 L
4E  1. In this limit, i.e. very close to the source we have Pee ≈ 1.

 ∆m2 L
4E  1. Very far away from the source hsin2 (∆m2 L/4E)i = 1/2 and Pee = 1 − (sin2 2θ)/2.

Pee = 1 − sin2 2θ sin2 [1.27L]


1.2 1 − sin2 2θ
1 − 12 sin2 2θ
1
Probability νe → νe

0.8

0.6

0.4

0.2

0
0.1 1 10 100 1000
Neutrino Flight Distance L [km]

Figure 4.2: Survival probability Pee as a function of the flight distance L for ∆m2 = 1 eV2 , E = 1 GeV,
and θ = 33◦ . For large values of L we can take the average hsin2 (∆m2 L/4E)i = 1/2 and therefore
Pee = 1 − (sin2 2θ)/2, see the dot-dashed line.

Now let us see which are the requirements to probe the mass differences of neutrinos in physical cases.
Consider the following table:

ν sources L [km] E [MeV] ∆m2 [eV2 ]


Atmospheric ν e,µ , νe,µ 20 ÷ 104 103 ÷ 104 10−4 ÷ 10−3
Reactor 1 1 1 10−3
Solar νe 108 1 ÷ 10 10 ÷ 10−4
−5

Reactor 2 100 1 10−5

Notice that since the depth of the atmosphere and the energy range of the secondary cosmic rays
are given we are lucky that it is of the right amount to observe the oscillation phenomenon due to
mass splitting of 10−4 ÷ 10−3 eV2 . We can also build an experiment (reactor 1), with chosen distance
and energy of the neutrinos to probe with increasing precision the mass splitting. Also in the case of
solar neutrinos the distance is fixed and so is the energy range, but it is such that we can probe much
smaller mass splittings: accounting also for interactions of neutrino with matter (mostly inside the Sun)
through CC and NC processes we pass from 10−11 ÷ 10−10 to 10−5 ÷ 10−4 eV2 . Also people constructed
experiments (reactor 2) to probe the mass splitting highlighted by the solar neutrino oscillation.
4.5. NEUTRINO OSCILLATION 45

N = 3 analysis
We can generalize the previous analysis to the case of three generations of neutrinos. Consider two
flavors ` and `0 , then
3
e−iEi t U`i |νi (0)i, (4.59)
X
|ν` (t)i =
i=1
3
U`?0 i e−iEi t U`i . (4.60)
X
M(ν` → ν`0 ) = hν`0 (0)|ν` (t)i =
i=1

Therefore the oscillation probability is given by

3 2
∆m2ki L
 
U`?0 i e−iEi t U`i
X X
P(ν` → ν`0 ) = = U`?0 i U`0 k U`i U`k
?
exp −i
2E
i=1 i,k
(4.61)
∆m2ki L
X X  
= |U`0 i |2 |U`i |2 + 2 Re U`?0 i U`0 k U`i U`k
?
exp −i
i=k i>k
2E

where in the last equality we have decomposed the sum with i 6= k into two sums that are complex
conjugate of each other (one with i > k for which ∆m2ki > 0 and the other with i < k for which
∆m2ki < 0) and then we used the property z + z ? = 2Re z. Now we exploit the identities
X X X
U`?0 i U`0 k U`i U`k
?
= U`?0 i U`i ?
U`0 k U`k = δ`0 ` δ``0 = δ`0 `
i,k i k

(4.62)
X X
= |U`0 i | |U`i |2 + 2 Re
2
U`?0 i U`0 k U`i U`k
?

i=k i>k

where we used that ? = δ``0 . Substituting Eq. (4.62) into Eq. (4.61), we find
P
i U`0 i U`i
 
∆m2ki L
  
(4.63)
X
P(ν` → ν`0 ) = δ``0 − 2 Re  U`?0 i U`0 k U`i U`k
?
1 − exp −i 
i>k
2E
 
∆m2ki L 
X   
= δ``0 − 2 Re  U`?0 i U`0 k U`i U`k
?
1 − cos
i>k
2E
 
∆m2ki L 
 
(4.64)
X
+ 2 Im  U`?0 i U`0 k U`i U`k
?
sin
i>k
2E

where in the last equality we exploited the identity Re(ab) = Re a Re b − Im a Im b.

4.5.2 CP violation in neutrino oscillation


If we repeat the same calculations for the oscillation ν ` → ν 0` , which is related to the oscillation ν` → ν`0
by a CP transformation, we find

(4.65)
CP ?
P(ν` → ν`0 ) =⇒ P(ν ` → ν `0 ) = P(ν` → ν`0 )(Uαi → Uαi ).

and since in the substitution Uαi → Uαi


? the real part of the product U ? U U U ? stays unchanged while

the imaginary part changes sign we can define a CP violating asymmetry A`` CP as follows
0

∆m2ki L
 
(4.66)
0 X
A``
CP = P(ν` → ν`0 ) − P(ν ` → ν `0 ) = 4 Im(U`?0 i U`0 k U`i U`k
?
) sin .
i>k
2E

An important property satisfied by A``


CP is
0

(4.67)
0 0
A`` ``
CP = −ACP =⇒ A``
CP = 0 .
46 CHAPTER 4. NEUTRINO PHYSICS

In the standard parameterization of U , it is strightforward to obtain the explicit expression for Aeµ
CP

∆m212 L ∆m223 L ∆m213 L


     
Aeµ
CP = 16 J sin sin sin . (4.68)
4E 4E 4E

where J is the so-called Jarlskog invariant defined as

J = c12 c23 c213 s12 s23 s13 sin δ ∝ s12 s23 s13 sin δ . (4.69)

It is important to remark that the Jarlskog invariant J and therefore any CP violating effect in neutrino
oscillation are non-vanishing only provided there are three non-vanishing mixing angles as well as a
non-vanishing Dirac phase δ.
Let’s analyse more closely the transformations properties of the oscillation probabilities under the
discrete symmetries C, P, T . Looking at the leptonic CC Lagrangian
g g
LCC = − √ Wµ+ J −µ + Wµ− J +µ = − √ Wµ+ ν L U γ µ eL + Wµ− eL U † νL (4.70)
 
2 2
we see that U and U ? describe, in our convention, flavor transitions between antineutrinos and neutrinos,
respectively. For instance, the amplitude for transitions between electronic and muonic (anti)neutrinos
of energy E at distance L from the source reads
im2 L
X
− i ?
M(νe → νµ , E, L) = e 2E Uµi Uei
i
im2 L
(4.71)
i
e−
X
?
M(ν e → ν µ , E, L) = 2E Uµi Uei
i

and therefore P(ν e → ν µ , E, L) 6= P(νe → νµ , E, L) since U 6= U ? . However,

(4.72)
CP T
P(ν e → ν µ , E, L) −→ P(νe → νµ , E, −L) −→ P(νµ → νe , E, L) .

As a result, the neutrino oscillation probability in invariant under a CP T transformation

(4.73)
CP T
P(ν e → ν µ ) −→ P(νµ → νe ) ≡ P(ν e → ν µ )

Summarizing, the oscillation probability Pαβ (L) in presence of N families is such that

1. Pαβ (L) is sensitive only to mass differences.

2. Pαβ (L) is invariant under the rephasing Uαk → eiϕα Uαk eiϕk and therefore is sensitive to
0

 (
N (N − 1) 3 N =3
≡ mixing angles




2 1 N =2
(4.74)

(

 (N − 1)(N − 2) 1 N =3

 ≡ Dirac phases
2 0 N =2

3. Pαβ (L) is not sensitive to CP violating Majorana phases as they enter the UPMNS matrix through
a diagonal matrix, see eq. (4.36), and therefore they simplify in Pαβ (L) = | i Uβi Uαi |2 .
P ? −iEi t
e
Dirac phases affect conversion but not survival probabilities, see eqs. (4.66) and (4.67).

4.5.3 Neutrino oscillation and the uncertainty principle


The neutrino oscillation phenomenon is deeply related to the uncertainty principle of QM. We have
seen that neutrinos are produced in a well defined flavor state which is a coherent superposition of
states with different masses
|νµ i = Uµ1 |ν1 i + Uµ2 |ν2 i + Uµ3 |ν3 i . (4.75)
4.5. NEUTRINO OSCILLATION 47

The above relation assumes that small neutrino masses can be neglected in matrix elements of weak
processes in which flavor neutrinos are produced. However, this assumption is trustable only if neutrinos
with different masses are not distinguishable. We will demonstrate now that this is indeed the case.
Let us consider the pion decay π + → µ+ νµ . From standard kinematics of two body decays it is
straightforward to evaluate the neutrino energy in the pion rest frame
m2π − m2µ + m2i
Ei = , (4.76)
2mπ
and the corresponding modulus of the momentum
m2π − m2µ m2π + m2µ
!
q
m2 m2i
pi = Ei2 − m2i ≈ Ei − i ≈ − . (4.77)
2Ei 2mπ 2mπ m2π − m2µ

Therefore, the difference between momenta corresponding to different neutrino mass eigenstates is
m2π + m2µ
!
|∆m2ik |
|∆pik | = |pk − pi | ≈ , (4.78)
2mπ m2π − m2µ

so, if we want to distinguish among the various mass eigenstates, we need an uncertainty ∆p on
momenta smaller than ∆pik . However, for the uncertainty principle ∆x∆p & ~, so the size of the wave
packet of the decaying pions is indetermined by
!
~ 2.4×10−3 eV2
∆x & ∼ 6 km . (4.79)
∆p ∆m2

So, in order to have a resolution enabling us to distinguish among the three mass eigenstates we have
an indetermination over the position of the decaying pions that is many order of magnitudes greater
than the atomic size of the beam. Hence we understand that is the uncertainty principle itself that does
not allow us to know the mass eigenstates. This behavior is analogous to the case of the interference
phenomenon due to the particle-wave dualism. Then we need to sum up coherently over the states
we cannot determine. In other words if we could pin down precisely the mass eigenstates we would
destroy the oscillation phenomenon as the wave interference would be destroyed if we could determine
with infinite precision the momentum of a photon. The neutrino oscillation phenomenon is therefore a
purely quantum interference phenomenon.

4.5.4 Oscillation in matter and the MSW effect


So far we have developed the formalism necessary to study the neutrinos oscillation in vacuum, in
particular, we have made use of the dispersion relation of a free particle. However, an important case
in which this approximation breaks down is the case of neutrinos produced inside the Sun: the medium
through which the neutrinos travel provides a non negligible interaction probability that has to be
taken into account when writing the neutrino energy. The study of the oscillation of neutrinos in a
dense medium provides the solution to the famous solar neutrinos problem.
In the Sun nuclear reactions produce a huge amount of electronic neutrinos that travel to the Earth.
In 1967 in the Homestake Gold Mine in Lead, South Dakota, it was studied the process
37
Cl + νe −→ 37
Ar + e− (4.80)

where solar neutrinos produce a transition of the Chlorine isotope to Argon. In general, experiments
involving neutrinos study processes of the type

ν+A→B+C (4.81)

where the Mandelstam variable s = 2Eν mA + m2A ≥ (mB + mC )2 , therefore the neutrino energy
threshold for the process is given by
(mB + mC )2 mA
Ethr = − (4.82)
2mA 2
48 CHAPTER 4. NEUTRINO PHYSICS

which corresponds to an energy Ethr = 0.81 MeV in the case of the Homestake experiment. Solar
neutrinos have a wide energy spectrum depending on the nuclear reaction in which they are produced.
In Fig. (4.3), we show the solar neutrino fluxes (together with their uncertainties) vs. the neutrino
energy for the leading nuclear reactions in the Sun. The main regions of interest are those regarding
neutrinos produced in the pp chain reaction pp →2 H + e+ + νe (which is by far the dominant one at
low energies) and in the Boron decay 8 B →8 Be + e+ + νe , which is the main source of solar neutrinos
at work in the Homestake experiment. In the figure it is also shown the energy range of interest (above
the threshold) for experiments with Gallium, Chlorine or others. Experimentally, one computes the
quantity
observed flux
R= , (4.83)
expected flux
that was expected to be 1 in the absence of neutrino oscillation. The Homestake experiment obtained

RHomestake = 0.33 ± 0.05, (4.84)

hence an anomaly was found, namely a huge deficit of solar neutrinos.

1014 100 %
dΦν / dEν in cm-2 sec-1 MeV-1

Gallium
1012 pp 80 %
Chlorine

Survival probability
Water
1010 Be pep
60 %
8 CNO
10
night 40 %
106
B day
4 20 %
10
hep
102 0%
0.1 1 10
Energy of solar neutrinos in MeV

Figure 4.3: The predicted unoscillated flux spectrum dΦ/dEν of solar neutrinos, together with the
energy thresholds of the experiments performed so far and with the best-fit oscillation survival probability
Pee (Eν ) (dashed line).

Gallium Chlorine Water

pp Be
CNO
pp B
B
B
CNO Be

Figure 4.4: The predicted fractional contributions to the neutrino rates of present experiments, assuming
energy-independent oscillations.

The situation became even worse much later when the results of GALLEX or Gallium Experiment
(1991-1997) at the Gran Sasso came out. As we see in Fig. (4.3) the Gallium experiment allows for a
lower energy threshold allowing to detect many low energy neutrinos produced in the pp chain. The
reaction involved was
νe +71 Ga −→71 Ge + e− (4.85)
4.5. NEUTRINO OSCILLATION 49

with threshold of 233.2 keV. The quantity R was measured to be

RGallex = 0.60 ± 0.06 (4.86)

so we have still an anomaly but also a strong disagreement with the results of the Homestake experiment.
We will see that both data can be explained through the neutrino oscillation phenomenon once matter
effects are taken into account.
If neutrinos oscillate, the ratio R is given by

(1 − ξ)Φνe σ(νe e− ) + ξΦνe σ(νµ,τ , e− )


R= , (4.87)
Φνe σ(νe e− )

where Φνe is total incident flux of solar neutrinos, so that if there is no neutrino oscillation ξ = 0 and
R = 1. If neutrinos oscillate instead, some of the initial electronic neutrinos change flavor and decrease
the ratio R since σ(νµ,τ e− ) is smaller than σ(νe e− ) by a factor of about six. As we will see in the next
chapter, the cross-section of the process να e− → να e− , where α = e, µ, τ , is given by σ(να e− ) ∼ cα G2F s
where cα is a flavor dependent numerical factor since νµ,τ e− → νµ,τ e− receives only NC contributions
while νe e− → νe e− is induced by both NC and CC effects, as shown by the following Feynman diagrams

νµ,τ νµ,τ νe νe e− νe

Z Z + W
(4.88)

e− e− e− e− e− νe
| {z } | {z }
NC only NC+CC contribution

By the time the Gallex experiment was at work the scientific community was still skeptical about
the existence of the neutrino oscillation phenomenon. However, the clear cut proof came with the
experiment performed at the Sudbury Neutrino Observatory (SNO) in Ontario. The experiment was
sensitive to three reactions

1. A pure NC process
ν ν

ν + d −→ ν + p + n : Z (4.89)
n
d
p

2. A pure CC process

νe e−

ν + d −→ e− + p + p : W (4.90)
p
d
p
50 CHAPTER 4. NEUTRINO PHYSICS

3. A process involving both NC and CC contributions

νe,µ,τ νe,µ,τ e− νe

ν + e− −→ ν + e− : Z + W (4.91)

e− e− e− νe

Notice that the CC process selects only electronic neutrinos. So the SNO experiment was able to
disentangle and study separately the NC and CC contributions as well as their sum. There were
observed anomalies in the second and third reactions which involved CC processes but no anomalies in
the pure NC process. This was the definitive proof of the existence of neutrino oscillation. Indeed,
the pure NC process is sensitive to all neutrino flavors and their contributions sum to give the same
effect as in absence of any neutrino oscillation. This fact ruled out other hypotheses that suggested
the solar neutrinos disappeared or were lost somehow or that our standard solar model predicted a
wrong neutrino flux: the number of neutrinos was conserved in the journey from the Sun to the Earth.
The data of the second and third processes were showing anomalies but they are well explained by the
neutrino oscillation phenomenon that was eventually confirmed.
However, we are left with an unsolved puzzle. How do we explain the disagreement between the
Homestake and Gallex experiments? Clearly there is a mechanism which, in the neutrino oscillation
framework, makes a distinction between the two experiment. We already know that this mechanism
should be energy dependent since the only significant difference between the two experiments were
the energy thresholds of the reactions they considered. The solution of the puzzle is provided by the
Mikheyev-Smirnov-Wolfenstein (MSW) effect. The Sun is made of electrons, protons and neutrons. If
neutrinos are produced in nuclear reactions occurring in the core of the Sun, they interact with the
dense medium through NC and CC interactions as shown by the following diagrams

νe,µ,τ νe,µ,τ e− νe

Z + W (4.92)

e− p n e− p n e− νe
| {z } | {z }
NC: flavor blind CC: only electronic neutrinos

and their dispersion relation in matter is given, in the flavor basis, by


α
Em = E0α + ECC
int α int
δe + ENC . (4.93)
|{z} | {z }
vacuum matter

Since the oscillation probability depends only on differences of neutrino energies, NC contributions
cancel out in the oscillation probability as they are flavor universal. Therefore, hereafter we will safely
neglect ENC
int . Let us estimate instead E int in naive dimensional analysis. We know that weak CC
CC
effects must be iproportional to the Fermi constant GF . Moreover, since neutrino CC interactions
involve exclusively electrons, ECC
int has to be proportional to the number of electrons N that are in the
e
medium otherwise if Ne = 0 we would be in vacuum where ECC int = 0. As a result, we find that

Ne
int
ECC ∼ GF ≡ GF n e (4.94)
V
4.5. NEUTRINO OSCILLATION 51

where by dimensional arguments we divided the energy by the volume obtaining a dependence on the
number density of electrons ne ≡ Ne /V . Let us derive now this result more rigorously. Neutrino CC
interactions are described by the Fermi Lagrangian

GF
  
Hint = −Lint = √ eγµ (1 − γ5 )νe ν e γ µ (1 − γ5 )e
2
(4.95)
GF
  
= √ eγµ (1 − γ5 )e ν e γ µ (1 − γ5 )νe
2

where in the third equality we exploited a Fierz identity. With this trick the CC looks formally as a
NC. To compute the interaction energy we need to average over the matter in the Sun, namely over the
electrons. Being neutrinos ultrarelativistic, the electrons can be considered at rest (this is an excellent
approximation even at high temperatures) and therefore

2GF (1 − γ5 )
Hint = √ he(~p = 0)|eγµ (1 − γ5 )e|e(~
p = 0)i ν e γµ νe . (4.96)
2 | {z } 2 }
| {z
heγµ (1−γ5 )ei PL

However, the average of the axial vector current over parity invariant states vanishes

heγµ γ5 ei = 0 (4.97)

whereas, in the vector current case, we obtain

(4.98)
p
~=0
~ −→
heγi ei = hJi 0 heγ0 ei = he† ei = ne 6= 0

since the current density J~ vanishes (classically) if the electrons are at rest. Therefore the only non
vanishing contribution to Hint is provided by the number density of electrons. As a result, the total
Lagrangian density for electron neutrinos is given by3

(4.99)
h i
L = Lfree + Lint = ν e i∂/ − 2GF ne γ 0 νe

where the interaction piece modifies the time component of L according to



i∂0 −→ i∂0 − 2GF ne (4.100)

and therefore, in the flavor basis, we obtain the following expression for the electon neutrino energy

Emα
= E0α + 2GF ne δeα (4.101)

where Em α are the eigenvalues of the Hamiltonian in the flavor basis which account for matter effects.

In order to determine Emα we proceed as follows. Let’s consider first the oscillation in vacuum in a two

generation scheme4 which is described by the Schroedinger equation in the mass basis
 
m21
! ! !
d ν1 (t) ν1 (t) E + 0  ν1 (t)
i = H0 = 2E
m22 . (4.102)
dt ν2 (t) ν2 (t) 0 E + 2E ν2 (t)

Let’s pass now to the flavor basis through the rotation |νi i = Uαi
−1
|να i and multiplying eq. (4.102) by U
! ! !
d νe (t) νe (t) cos θ sin θ
i = U H0 U −1
U= , (4.103)
dt νµ (t) | {z } νµ (t) − sin θ cos θ
H0flavor

3
Note that, since neutrinos are left-handed, it turns out that ν e γµ νe = ν e γµ (PL + PR )νe = ν e γµ PL νe
4
This is a good approximation for the study of solar neutrinos as the mixing angle θ13 ≈ 8◦ can be safely neglected
compared to θ12 ≈ 33◦ , unless we are interested in CP violating effects which require three generations to show up.
52 CHAPTER 4. NEUTRINO PHYSICS

where we have assumed that U is time independent and therefore we find


2
!
1
E+ 2 2
2E (cos θm1 + sin2 θm22 ) sin 2θ ∆m
H0flavor = ∆m2 1
4E . (4.104)
sin 2θ 4E E + 2E (sin2 θm21 + cos2 θm22 )

Now, since ECCint is defined in the flavor basis, we can add it to H flavor in oder to find the desired
0
expression, in the flavor basis, of the Hamiltonian Hmflavor which includes matter effects

!
m2 + m22 ∆m2 − cos 2θ + λ sin 2θ
 
flavor
Hm = E+ 1 1+ , (4.105)
4E 4E sin 2θ cos 2θ

with
√ 4E
λ= 2GF ne . (4.106)
∆m2
The eigenvalues of Hm
flavor reads

m21 + m22 ∆m2 


 
(4.107)
p 
±
Em = E+ + λ ± λ2 − 4λ cos 2θ + 4 .
4E 8E

while the normalized eigenvectors, which are orthogonal being Hm


flavor symmetric, are given by

! !
χ 1 −1 1
V+ = p 2 V− = p 2 , (4.108)
χ +1 1 χ +1 χ

with
λ
(4.109)
p
χ= − cos 2θ + λ2 − 4λ cos 2θ + 4 .
2
If U m (θm ) is the unitary matrix diagonalizing Hm flavor , i.e. the matrix which has the normalized

eigenvectors as columns, we can define an effective oscillation mixing angle in matter5


2χ sin 2θ
tan 2θm = = λ
. (4.110)
1 − χ2 cos 2θ − 2

Notice that if λ = 0 we recover θm = θ as it must be. Interestingly, for cos 2θ = λ/2 we approach a
regime of maximal mixing, i.e. θm = π/4, indipendently on the mixing angle in vacuum. This happens
when the electron density reaches a critical value

cos 2θ∆m2
n̄e = √ . (4.111)
2 2EGF
This is the resonant matter effect in neutrino oscillation or MSW effect.
Let us finally explain the disagreement between Homestake and Gallex. In a three flavors framework,
  Z t  2
(4.112)
X
m
P(να → νβ ) = Uβi (t) exp −i Eim (t)dt m?
Uαi (t0 ) ,
i t0

with U m (t) being the unitary mixing matrix in the matter. Notice that U m (t) is time-dependent as it
contains the number density of electrons ne (t) which varies from the core of the Sun to its surface. We
can write the probability splitting the sum in two as we did in Eq. (4.61)
X
m m
P(να → νβ ) = |Uβi (t)|2 |Uαi (t0 )|2
i=k | {z }| {z }
matter
(4.113)
vacuum
X   Z t  .
m m?
+ Uβi (t)Uβk (t) exp −i (Eim − Ekm )(t)dt m?
Uαi (t0 )Uαk (t0 )
i6=k t0

 
5 A C
One can more rapidly say that a 2 × 2 matrix is diagonalized by a rotation matrix of angle tan 2θ = 2C
A−B
.
C B
4.5. NEUTRINO OSCILLATION 53

Since we detect the neutrinos at the time t on the Earth where the density of electrons is much smaller
than that in the Sun, the mixing matrix U m (t) will be in practice identical to the one in vacuum.
Instead, at time t0 when the neutrinos are produced, the matter effect are non negligible. Moreover,
the second term in Eq. (4.113) averages to zero. In fact, the neutrinos are not produced in a point-like
source but in a sphere of radius about 6 × 105 km: averaging over this huge region is tantamount to
average in time over many oscillations and the net effect is zero (no contribution to the probability).
Therefore, effectively, the oscillation probability is given by

(4.114)
X
m m
P(να → νβ ) = |Uβi (t)|2 |Uαi (t0 )|2
i=k | {z }| {z }
vacuum matter

The survival probability within a N = 2 framework is therefore given by


X
m
P(νe → νe ) = |Uei (x)|2 |Uei (x0 )|2 = cos2 θ cos2 θm (x0 ) + sin2 θ sin2 θm (x0 )
i
(4.115)
1
= (1 + cos 2θ cos 2θm )
2
where
1 cos 2θ − λ2
cos 2θm = p = q . (4.116)
1 + tan2 2θm (cos 2θ − λ2 )2 + (sin 2θ)2
In the high energy limit, namely when λ  2 cos 2θ, matter effects are dominant and cos 2θm → −1.
Instead, at low energies λ  2 cos 2θ, matter effects are neglibigle and we recover the results obtained
in vacuum, i.e. θm = θ. Therefore

2
 sin θ
 λ  2 cos 2θ
P(νe → νe ) ≈ sin2 2θ (4.117)
 1−
 λ  2 cos 2θ
2
In particular, the first case (λ  2 cos 2θ) is appropriate to model the Homestake experiment where
neutrinos mostly come from the Borum decay with a flux peak at E ∼ 10 MeV: at such high
energies matter effects are important thus the appropriate formula for the survival probability is
P(νe → νe ) = sin2 θ. Instead, the second case (λ  2 cos 2θ) is suitable to model the Gallex experiment
in which are involved mainly low energy neutrinos (E . 0.4 MeV) from the pp chain: therefore vacuum
approximation is still valid and P(νe → νe ) ≈ 1 − 0.5 sin2 2θ.
Since R = P(νe → νe ), we can find the value of the mixing angle θ that explains the two experiments:

=⇒ θ ≈ 33◦

2
 sin θ ≈ 0.3
 Homestake
R = P(νe → νe ) ≈ sin2 2θ (4.118)
 1−
 ≈ 0.6 =⇒ θ ≈ 33◦ Gallex
2
To sum up both experiments are perfectly explained by the neutrino oscillation phenomenon supple-
mented by the MSW effect which takes into account the neutrino interactions with matter.
Finally, le us check a posteriori whether our assumption on the importance of matter effets in the
Homestake and Gallex experiments were correct. A suitable numerical approximation for λ is given by
!
√ 4E E
 
10−4 eV2
λ= 2GF ne ∼ 0.38 . (4.119)
∆m2 MeV ∆m2

For Homestake (E = 10 MeV) we have λ ' 3.8  2 cos 2θ = 0.81 and for Gallex (E = 0.4 MeV) we
have λ ∼ 0.16  2 cos 2θ = 0.81, justifying our approximations.
Solar neutrino data are mostly sensitive to the solar mixing angle θ and only rather poorly to the
mass difference ∆m2 , see eqs. (4.115) and (4.117). Yet, their sensitivity to matter effects, made it
possible to establish the neutrino mass hierarchy m2 > m1 , i.e. ∆m2 > 0. A precise measurement of
∆m2 became possible only with the advent of long baseline experiments like KamLAND which detected
ν̄e emitted by terrestrial nuclear reactors located at distances L ≈ 180 km using the ν̄e p → e+ n reaction.
54 CHAPTER 4. NEUTRINO PHYSICS

Survival Probability
0.8

0.6

0.4

0.2 3-ν best-fit oscillation Data - BG - Geo νe


2-ν best-fit oscillation
0
20 30 40 50 60 70 80 90 100 110
L0/Eν (km/MeV)
e

Figure 4.5: Survival probability of ν̄e as a function of L0 /E where L0 = 180 km is the effective baseline.

Since the neutrino energy is directly linked to Ee+ as Eν̄ ≈ Ee+ + mn − mp , KamLAND data allow
to test if the ν̄e survival probability depends on the neutrino energy as predicted by oscillations, see
fig. 4.5. The effect of oscillations is given by

∆m2sun L0
 
P (ν̄e → ν̄e ) = 1 − sin 2θsun sin 2 2
(4.120)
4Eν̄e

up to minor corrections due to earth matter effects and to the small θ13 atmospheric mixing angle. Since
P (ν̄e → ν̄e ) shows two minima at L0 /Eν̄e ∼ (20, 50) km/MeV corresponding to ∆m2sun L0 /4Eν̄e = ( π2 , 3π
2 )
respectively, we deduce that ∆msun ≈ 10 eV . The global fit of fig.4.6 confirms the above estimate
2 −4 2

and shows that ∆m2sun is dominantly fixed by KamLAND data, which precisely fixes

∆m2sun = (7.39 ± 0.21) 10−5 eV2 , tan2 θsun = 0.45 ± 0.04 (4.121)

with other local minima disfavored at more than 4σ. Statistical and systematic errors are comparable.

KamLAND
95% C.L.
99% C.L.
99.73% C.L.
(eV )

best fit
2

10-4
∆m2⊙

Solar
95% C.L.
99% C.L.
99.73% C.L.
best fit

10-1 1
tan2θ⊙

Figure 4.6: Allowed region for tan2 θsun and ∆m2sun from KamLAND and solar neutrino experiments.
4.5. NEUTRINO OSCILLATION 55

4.5.5 Atmospheric neutrinos


The νµ → ντ oscillation was first observed by the SuperKamiokande (SK) experiment, studying
atmospheric neutrinos. Then, many other experiments such as T2K, NOVA, MINOS, and IceCube
confirmed the effect. Atmospheric neutrinos are generated by collisions of primary cosmic rays mainly
composed by H and He nuclei, yielding respectively ∼ 82 % and ∼ 12 % of the nucleons. Heavier nuclei
constitute the remaining fraction. The process can be schematized in 3 steps:

1. Primary cosmic rays hit the nuclei of air in the upper part of the earth atmosphere, producing
mostly pions (and some kaon).

2. Charged pions decay promptly generating muons and muonic neutrinos:

π + → µ+ νµ π − → µ− ν̄µ

(the decay rate into electrons is suppressed by m2e /m2µ ). The total flux of νµ , ν̄µ neutrinos is about
0.1/cm2 s at Eν ∼ GeV with a ∼ 20% error. At higher energy the flux dΦ/d ln Eν approximately
decreases as Eν−2±0.05 . The few kaons decay like pions, except that K → πe+ νe decays are not
entirely negligible.

3. The muons produced by π decays (at a distance ≈ 20 Km from the Earth) travel a distance6


 
d ≈ c τµ γµ ≈ 20 km
3 GeV

where τµ ' 2.2 × 106 s is the muon life-time and γµ = Eµ /mµ is the relativistic dilatation factor.
If all muons could decay

µ− → e− ν̄e νµ µ+ → e+ νe ν̄µ

one would obtain a flux of νµ and νe in proportion 2 : 1, with comparable energy, larger than
∼ 100 MeV. However, muons with energy above few GeV typically collide with the earth before
decaying, so that at higher energy the νµ : νe ratio is larger than 2.

The fluxes predicted by detailed computations is shown in fig.4.7, at SK location, averaged over zenith
angle and ignoring oscillations.

p...
ΝΜ
E3Ν dFΝ / dEΝ in m-2 sec-1 sr-1 GeV2

ϑ
102
SK
_ _
Νe Νe ΝΜ
ν

10 earth
p...
π

ϑ
1
10-1 1 10 102 103 104
Neutrino energy EΝ in GeV

Figure 4.7: Flux of atmospheric neutrinos in absence of oscillations, as predicted by FLUKA. The
cartoon at the right shows that, since the earth is spheric, without oscillations the flux of atmospheric
neutrinos would be up/down symmetric.
6
More precisely, d ≡ xLAB = βµ c τLAB = βµ γµ c τµ = γµ2 − 1 c τµ ≈ c τµ γµ for γµ  1, i.e. for Eµ  mµ .
p
56 CHAPTER 4. NEUTRINO PHYSICS

SK detects atmospheric neutrinos through CC scattering on nucleons, ν` N → `N 0 . SK is a cylindrical


tank containing 50000 ton of light water surrounded by photomultipliers, located underground in the
Kamioka mine in Japan. A relativistic charged lepton ` traveling in water gives rise to a detectable
Cerenkov ring. SK can distinguish νµ from νe but cannot distinguish particles from anti-particles. The
atmospheric fluxes contain a roughly equal number of ν and ν̄, and ν have roughly a two times larger
cross section on nucleons than ν̄.
Measuring the Cerenkov light SK reconstructs the energy E` and the direction ϑ` of the scattered
charged lepton. At high energy E`  mN the scattered lepton roughly keeps the direction of the
neutrino, whose zenith-angle ϑν is related to the pathlength L by
q
L= 2 cos2 ϑ − r | cos ϑ | +
h2 + 2hrE + rE ν E ν 2rE | cos ϑν | (4.122)
| {z }
in the earth, if cos ϑν < 0
| {z }
in the atmosphere

where rE = 6371 km is the radius of the earth and h ∼ (15 ÷ 20) km is the height of the atmosphere.
Down-ward
√ going neutrinos (cos ϑν = 1) travel L ∼ h. Horizontal neutrinos (cos ϑν = 0) travel
L ∼ 2rE h ∼ 500 km. Up-ward neutrinos (cos ϑν = −1) travel L = 2rE . Measuring E` and ϑ` is not
sufficient to reconstruct the neutrino energy, Eν & E` , since it is not know from which direction one
atmospheric neutrino arrives. Therefore SK grouped their data into energy ‘bins’ defined according to
the topology of the events:
1. Fully-contained electron or muon events: the scattered lepton starts and ends inside the
detector, so that its energy can be measured. These events are conveniently divided into

 sub-GeV events with E` < 1.33 GeV produced by neutrinos with Eν ∼ 1 GeV. The average
opening angle between the incoming neutrino and the detected charged lepton is ϑ`ν ∼ 60◦
with a poor angular resolution: Eν and L can only be estimated.

 multi-GeV events with E` > 1.33 GeV are produced by neutrinos with Eν ∼ f ew× GeV
and ϑ`ν ∼ 15◦ (decreasing at higher energy), allowing a rather precise measurement of L.

2. Partially contained muons: the muon is scattered inside the detector, but escapes from the
detector, so that its energy cannot be measured. These events originate from neutrinos with a
typical energy only slightly higher than those giving rise to multi-GeV muons. Therefore these
two classes of events can be conveniently grouped together.

3. Up-going stopping muons: the scattered µ is produced by neutrinos with Eν ∼ 10 GeV in the
rock below the detector (so that its energy cannot be measured) and stops inside the detector.
This technique cannot be used for studying νe (because the scattered electrons shower before
reaching the detector) nor for down-going νµ (due to the background of cosmic ray muons).

4. Through-going up muons: the µ is scattered in the rock below the detector by neutrinos with
Eν between 10 GeV and 10 TeV and crosses the detector without stopping.
The SuperKamiokandeI data are shown in fig.4.8. The crucial point is that since the Earth is a good
sphere, in absence of oscillations the neutrino rate would be up/down symmetric, i.e. dependent only
on | cos ϑ|7 . While the zenith-angle distribution of µ events is clearly asymmetric, e-like events show
no asymmetry. The flux of up-ward going muons is about two times lower than the flux of down-ward
muons. Therefore the data can be interpreted assuming that nothing happens to νe and that νµ
oscillate into ντ . Neglecting earth matter corrections, we assume
∆m2atm L
 
P (νe → νe ) = 1 P (νe ↔ νµ ) = 0 2
P (νµ → νµ ) = 1 − sin 2θatm sin 2
(4.123)
4Eν
The main result can be approximately extracted from very simple considerations. Looking at the
zenith-angle dependence we notice that down-ward going neutrinos (cos θν > 0) are almost unaffected by
7
The peak at cos ϑν ∼ 0 in multi-GeV µ−like events of fig.4.8 is due to the fact that horizontal muons have more time
for freely decaying before hitting the earth, while vertical muons cross the atmosphere along the shortest path.
4.5. NEUTRINO OSCILLATION 57

Super-Kamiokande I-IV
306 kt y
Prediction

1000 1000 200

Number of Events νµ → ντ

Sub-GeV e-like Sub-GeV µ-like UpStop µ


9775 Events 10147 Events 1370 Events
0 0 0
-1 0 1 -1 0 1 -1 -0.5 0

1000
400 1000

200 500 500

Multi-GeV e-like Multi-GeV µ-like + PC UpThrough µ


2653 Events 5485 Events 5896 Events
0 0 0
-1 0 1 -1 0 1 -1 -0.5 0
cos zenith

Figure 4.8: The zenith angle distributions of Super-Kamiokande atmospheric neutrino events. Fully
contained 1-ring e-like and µ-like events with visible energy < 1.33 GeV (sub-GeV) and > 1.33 GeV
(multi-GeV), as well as up-going stopping and through-going up µ samples are shown. Partially
contained (PC) events are combined with multi-GeV µ-like events. The blue histograms show the non-
oscillated Monte Carlo events, and the red histograms show the best-fit expectations for µ-oscillations.

oscillations, while up-ward going neutrinos (cos θν < 0) feel almost averaged oscillations, and therefore
their flux is reduced by a factor 1 − 12 sin2 2θatm . This must be equal to the up/down ratio ∼ 0.5, so
that sin2 2θatm ∼ 1. Furthermore, multi-GeV neutrinos have energy Eν ∼ 3 GeV, and according to
fig.4.8 they begin to oscillate around the horizontal direction (cos ϑ ∼ 0) i.e. at a pathlength of about
L ∼ 1000 km. Therefore ∆m2atm ∼ Eν /L ∼ 3 10−3 eV2 . The result of νµ → ντ oscillation fits derived
from several experiments is shown in fig. 4.9. The zenith-angle dependence of the number of µ-like
events discussed above represents the main evidence of the atmospheric neutrino oscillation.

4.5.6 Determination of Ue3


The PMNS matrix element Ue3 , which accounts for the admixture of electron neutrino νe in the mass
eigenstate ν3 , has been measured quite recently. Before its discovery, the CHOOZ experiment in France
looked for the survival probability of electron antineutrinos ν̄e from nuclear reactor. Antineutrino
energies were in the several MeV range, and the distance from the reactor to the detector was about 1
km. In these conditions, Earth matter effects were negligible so as the effects induced by ∆m2sun .
Thus, the survival probability of electron antineutrino is well approximated by the expression

∆m213 L
 
P(ν̄e → ν̄e ) = 1 − 4|Ue3 |2 (1 − |Ue3 |2 ) sin2 . (4.124)
| {z } 4E
sin2 (2θ13 )

The baseline corresponded to the maximum probability of neutrino oscillations at energy 2 − 3 MeV
for the mass squared difference |∆m213 | ≈ ∆m2atm ≈ 2.5 · 10−3 eV2 . Were |Ue3 | large enough, the
disappearance of electron antineutrinos and, most importantly, the distortion of their energy spectrum
would have been observed. These effects were not found, and the experiment results yielded the bound

|Ue3 | . 0.03 . (4.125)

On the other hand, the MINOS and T2K studied the conversion probability of muonic neutrinos into
58 CHAPTER 4. NEUTRINO PHYSICS

0.0035 Normal Hierarchy, 90% C.L.


Super-K
T2K
NOvA
IceCube
MINOS

∆ m232 [eV2]
0.003

0.0025

0.002
0.2 0.4 0.6 0.8
sin2 θ23

Figure 4.9: The 90% confidence regions for the sin2 θ23 -∆m232 plane assuming normal mass ordering
derived from the T2K, NOVA, MINOS, Super-Kamiokande, and IceCube experiments.

electronic neutrinos in an experimental setup where again ∆m2sun was neglegible


2
2 2 2
?
P(νµ → νe ) = Ue1 Uµ1 ?
+ Ue2 Uµ2 +e−i∆m13 L/2E Ue3 Uµ3
? ?
= Ue3 Uµ3 (1 − e−i∆m13 L/2E )
| {z }
?
−Ue3 Uµ3

∆m213 L ∆m213 L
    
= 2|Ue3 | 2 ? 2
|Uµ3 | 1 − cos 2 2
= sin θ23 sin (2θ13 ) sin 2
(4.126)
2E 4E

finding the 3σ range value


8◦ . θ13 . 9◦ (4.127)
which is almost insensitive to matter effects and to the neutrino mass ordering. A non-vanishing θ13 is
crucial to unveil CP violating effects in neutrino oscillations which are proportional to the Jarlskog
invariant J ∝ s12 s23 s13 sin δ where s13 ≡ sin θ13 and Ue3 ≡ s13 e−iδ in the standard parametrization.
Finally, in Table 4.1 we report the three-flavour oscillation parameters by the ν-fit collaboration.

Normal Ordering (best fit) Inverted Ordering (∆χ2 = 9.3)


bfp ±1σ 3σ range bfp ±1σ 3σ range
sin2 θ12 0.310+0.013
−0.012 0.275 → 0.350 0.310+0.013
−0.012 0.275 → 0.350

θ12 / 33.82+0.78
−0.76 31.61 → 36.27 33.82+0.78
−0.75 31.62 → 36.27

sin2 θ23 0.582+0.015


−0.019 0.428 → 0.624 0.582+0.015
−0.018 0.433 → 0.623

θ23 / 49.7+0.9
−1.1 40.9 → 52.2 49.7+0.9
−1.0 41.2 → 52.1

sin2 θ13 0.02240+0.00065


−0.00066 0.02044 → 0.02437 0.02263+0.00065
−0.00066 0.02067 → 0.02461

θ13 / 8.61+0.12
−0.13 8.22 → 8.98 8.65+0.12
−0.13 8.27 → 9.03

δCP /◦ 217+40
−28 135 → 366 280+25
−28 196 → 351

∆m221
7.39+0.21
−0.20 6.79 → 8.01 7.39+0.21
−0.20 6.79 → 8.01
10−5 eV2
∆m23`
+2.525+0.033
−0.031 +2.431 → +2.622 −2.512+0.034
−0.031 −2.606 → −2.413
10−3 eV2

Table 4.1: Three-flavour oscillation parameters from the ν-fit collaboration fit to global data (with SK
atmospheric data). The numbers in the 1st (2nd) column are obtained assuming NO (IO), i.e., relative
to the respective local minimum. Note that ∆m23` ≡ ∆m231 > 0 for NO and ∆m23` ≡ ∆m232 < 0 for IO.
4.6. KINEMATICAL TESTS OF NEUTRINO MASSES 59

4.6 Kinematical Tests of Neutrino Masses


Neutrino oscillation experiments are sensitive only to neutrino mass differences. In particular, current
experimental data indicate that

∆m212 ≡ ∆m2sun ≈ 10−4 eV2 , |∆m213 | ≈ |∆m223 | ≡ ∆m2atm ≈ 3 · 10−3 eV2 , (4.128)

where ∆m2ij = m2j − m2i . Notice that while solar neutrino data fix the ordering m2 > m1 8 , atmospheric
neutrino data leave open the possibilities m3 > m1 and m3 < m1 . The former case is referred to as
normal hierarchy while the latter as inverted hierarchy, see fig. 4.10.

νµ ντ ν3 νe νµ ντ ν2

sun
νe ν1
atm

νe νµ ντ ν2

atm
sun

νe ν1 νµ ντ ν3

Figure 4.10: Possible neutrino spectra: (a) normal (b) inverted.

On the other hand, oscillation experiments are insensitive to the absolute neutrino mass scale
(parameterized by the mass of the lightest neutrino) and to the 2 Majorana phases α and β. Other
types of experiments can study some of these quantities and the nature of neutrino masses. They are:

 β-decay experiments, that to good approximation probe m2νe ≡ i |Uei |mi ;


2 2
P

 neutrino-less double-beta decay (0ν2β) experiments, that probe the absolute value of the ee entry
of the neutrino Majorana mass matrix, |mee | ≡ | i Uei
P 2
mi |;

 cosmological observations (Large Scale Structures and anisotropies in the Cosmic Microwave
Background), that probe the sum of neutrino masses, mcosmo ≡ m1 + m2 + m3 .

Only 0ν2β probes the Majorana nature of the mass. The values |mee |, mνe , mcosmo are unknown,
and can be partially inferred from oscillation data. Ordering these probes according to their present
sensitivities, the list is cosmology, 0ν2β and finally β decay. Ordering them according to reliability
would presumably result into the reverse list: cosmological results are based on plausible theoretical
assumptions, and 0ν2β suffers from severe uncertainties in the nuclear matrix elements.

p
p
n
n
e
p e
ν ν
∆L = 2
n ν mass
ν
e e
e
n
n
ν p
p

Figure 4.11: Feynman diagrams for β decay, double-β decay, and neutrino-less double-β decay.

8
Since in the case of solar neutrinos matter effects are relevant, the oscillation probability depends on the sign of ∆m212
and experimental data can be fitted only if ∆m212 > 0, i.e. if m2 > m1 .
60 CHAPTER 4. NEUTRINO PHYSICS

4.6.1 Beta decay


The experimental manifestations of neutrino masses have been searched for in several different types of
experiments. The first type involves kinematic searches. The idea is that, if neutrinos have masses, this
will affect the electron energy distribution in nuclear beta decay. The process at the nuclear level is
(A, Z) → (A, Z + 1) + e− + ν̄e (4.129)
with the best bound coming from
3
H→ 3
He + e− + ν̄e . (4.130)
At the hadronic level we have
n → p + e− + ν̄e (4.131)
corresponding, at quark level, to the weak process d −→ u + e− + ν e as illustrated in fig. (4.12)

g
d u

W−
νj
g

e−

Figure 4.12: Feynman diagram relevant for β decay at the quark level.

The amplitude of the process is given by


iGF ?
M = − √ Vud hp(`0 )|dγµ (1 − γ5 )u|n(`)i ue (p)γ µ (1 − γ5 )vνe (k) . (4.132)
2 | {z }
hadronic matrix element

As already discussed in a previous chapter, the process is a 0+ → 0+ transition and therefore the axial
current does not contribute because of parity conservation. Further exploiting the SU (2)V symmetry
and the fact that the current JV−µ = dγ µ u is conserved, the hadronic matrix element reads
hp(`0 )|dγ µ u|n(`)i = F (q 2 )(` + `0 )µ ≈ F (0) 2`µ (4.133)
where the form factor F depends on the isospin of 3 He and 3 H and F (q 2 ) ≈ F (0) as q = ` − `0 → 0
being q 2 ∼ few (MeV)2 . Furthermore, since the energy release is much smaller than the nucleus mass,
the final nucleus is non-relativistic and the only non-negligible contribution will come from γ0
iGF ?
M = − √ Vud F (0) 2mn ūe (p)γ0 (1 − γ5 )vνe (k) . (4.134)
2
Now we can calculate the decay rate. Integrating over the 3-momentum of the final proton, we get
2
G2F |Vud |2 d3 p d3 k
2πδ(E0 − Ee − ω) (4.135)
X
dΓ = 2m2n |F (0)|2 ūe (p)γ0 (1 − γ5 )vνe (k)
2mn 2mp spin
(2π)3 2Ee (2π)3 2ω

Here, E0 denotes the total energy available to the lepton pair which, to a good approximation, is given
by the mass difference between the initial and final nuclei. The spin sum gives
2
1
= tr[k/γ0 p/γ0 (1 − γ5 )] = 2[Ee ω + p~ · ~k] = 2Ee ω[1 + βe βν cos θ] (4.136)
X
ūe (p)γ0 (1 − γ5 )vνe (k)
spin
2

where βe , βν are the magnitudes of the velocities of the outgoing leptons and θ is the angle between
their directions. Further integrating over the angular variables other than θ and setting mn = mp , we
obtain
2G2F |Vud |2
dΓ = p| Ee |~k| ω δ(E0 − Ee − ω) [1 + βe βν cos θ] dEe dω d(cos θ) .
|F (0)|2 |~ (4.137)
(2π)3
4.6. KINEMATICAL TESTS OF NEUTRINO MASSES 61

Integrating over cos θ, the term proportional to βe βν drops out and the other term gives 2. Finally,
integrating over ω we obtain

dΓ G2 |Vud |2
(4.138)
q
= F 3 |F (0)|2 |~
p| Ee (E0 − Ee ) (E0 − Ee )2 − m2νe .
dEe 2π

Defining the quantity Q = E0 − me and the electron kinetic energy Te = Ee − me , the number of
electrons emitted with energy between Ee and Ee + dEe is given by

G2F |Vud |2
(4.139)
q
2
dn = |F (0)| F (Ee , Z) |~
p| E (Q − Te ) (Q − Te )2 − m2νe dEe ,
2π 3
where F (Ee , Z) is the Coulomb correction factor.
If there is lepton mixing and νe has components in several mass eigenstates, then the square root
P q
expression is replaced by i (Q − Te )2 − m2i |Uei |2 . If mν = 0, then

1/2
dn/dEe

K(Ee ) ≡ ∝ Q − Te . (4.140)
F (Ee , Z) |~
p| Ee

If we plot the quantity K(Ee ) vs Te , we would obtain a straight line. Such a plot is called the Kurie
plot. The maximum value of Te would be Q. The linearity of the Kurie plot is lost if the neutrino has
a mass. The effect becomes appreciable only near the end point, where Q − Te is comparable to mνe .
The maximum kinetic energy of the electron becomes Q − mνe . If several νi are involved, the different
mi produce kinks in the Kurie-plot where the size of the kinks is a measure for the corresponding
|Uei |2 . This is illustrated in fig.4.13, where we show the combined effect of a Heavier neutrino with
little e component and of a Lighter neutrino with sizable e component.
(dN/dEe )1/2

Q − mνH Q − mνL Q
Ee

Figure 4.13: β-decay spectrum close to the end-point for a massless (dotted) and massive (continuous
line) neutrinos. The Kurie function K(Ee ) ≡ (dN/dEe )1/2 is plotted vs. the electron energy Ee .

In practice, the energy resolution is limited, and only broad features can be seen. If it is not possible
to resolve the difference between neutrino masses, we can identify to a good approximation mνe in
eq. (4.139) with the effective mass

(4.141)
X
m2νe ≡ 2
|Uei |m2i = cos2 θ13 (m21 cos2 θ12 + m22 sin2 θ12 ) + m23 sin2 θ13 .
i

The last equality holds in the standard three-neutrino case. The most recent bound on mνe from the
KATRIN experiment
mνe < 1.1 eV (4.142)
at 90% CL improves the previous bound from the Mainz and Troitsk experiments which constrained
mνe < 2.2 eV at 95% CL. KATRIN continues running with an estimated sensitivity limit of mνe < 0.2
eV, thanks to an energy resolution of 1 eV.
62 CHAPTER 4. NEUTRINO PHYSICS

4.6.2 Neutrinoless double beta decay


The nature (Dirac or Majorana) of massive neutrinos is one of the most important open problems
of particle physics. The discovery of the Majorana nature of neutrinos would represent a strong
argument in support of the see-saw mechanism which is broadly viewd as the most natural explanation
of the smallness of neutrino masses. Unfortunately, the phenomenon of neutrino oscillations does not
discriminate between Dirac and Majorana neutrinos, therefore, it is necessary to study processes in
which the total lepton number L = Le + Lµ + Lτ is violated such as the 0ν2β decay.
Let us consider an even-even nucleus (A, Z) with mass MA,Z . The mass of odd-odd nucleus
(A, Z + 1) with the same atomic number is larger than MA,Z . Therefore, the (A, Z) nucleus cannot have
usual β-decay (A, Z) → (A, Z + 1) + e− ν̄e . However, if there exists an even-even nucleus (A, Z + 2) with
mass smaller than MA,Z , the nucleus (A, Z) can decay into (A, Z + 2) with emission of two electrons.
Two types of ββ decays are possible

1. Two-neutrino double β decay (2ν2β-decay)

(A, Z) → (A, Z + 2) + e− + e− + ν̄e + ν̄e (4.143)

2. Neutrinoless double β decay (0ν2β-decay)

(A, Z) → (A, Z + 2) + e− + e− (4.144)

Hereafter, we will study 0ν2β-decay within the SM with Majorana neutrinos. The effective Fermi
Hamiltonian describing the process is

GF X
H(x) = √ ē(x)γα (1 − γ5 )Uei νi (x)j α (x) + h.c. (4.145)
2 i

where j α (x) is the hadronic charged-current and the field νi satisfies the Majorana condition

νi (x) = νic (x) = C ν̄iT (x) (4.146)

In the 0ν2β decay, which occurs at the second order in GF , neutrinos are virtual, see fig. (4.14), and
the matrix element of the process is given by the following expression

(−i)2 GF 2
  Z
hf |S 2 |ii = √ 4Np1 Np2 d4 x1 d4 x2 hNf |T (j α (x1 )j β (x2 ))|Ni i eip1 x1 eip2 x2
2! 2
(4.147)
X
T
2
Uei ūeL (p1 )γα h0|T (νiL (x1 )νiL (x2 ))|0i γβT ūTeL (p2 ) − (p1 ↔ p2 )
i

where p1,2 are the electron momenta, |Ni,f i are the states of the initial and the final nuclei and
Np = 1p
3/2 0
is the standard normalization factor.
(2π) 2p
Let us focus on the neutrino propagator. From the Majorana condition (4.146) we find
T
h0|T (νiL (x1 )νiL (x2 ))|0i = −PL h0|T (νi (x1 )ν̄i (x2 ))|0iPL C (4.148)

where
1 /q + mi
Z
h0|T (νi (x1 )ν̄i (x2 ))|0i = d4 qe−iq(x1 −x2 ) (4.149)
(2π)4 q 2 − m2i
is the standard propagator of a Dirac fermion. Thus, the propagator of eq. (4.148) becomes

i mi
Z
T
h0|T (νiL (x1 )νiL (x2 ))|0i =− d4 q e−iq(x1 −x2 ) PL C . (4.150)
(2π)4 q2 − m2i

As expected, the neutrino propagator is proportional to mi which is the source of the ∆L = 2 violation
required by the 0ν2β decay.
4.6. KINEMATICAL TESTS OF NEUTRINO MASSES 63

g
d u

W−
e−
gUei

mi
∆L = 2 ν i −→ q 2 −m2i

gUei
e−
W−

d u
g

Figure 4.14: Feynman diagram for 0ν2β at the quark level. The clash between the antineutrino lines
corresponds to a ∆L = 2 violation requiring a Majorana mass term in the neutrino propagator.

Since the process 2ν2β has been already observed, it is interesting to compare its decay rate with
the one of 0ν2β. Since both decays arise at the second order in the Fermi interactions, we expect that
2
Γ0ν2β mi
. 10−12 . (4.151)
X
∝ Uei Uei
Γ2ν2β i
Q

where we introduced, in the naive dimensional analysis approach, the dependence on the energy scale
of the 0ν2β process E ∼ Q = MP − MD ∼ 1 MeV and imposed the experimental bound mν . 1 eV.
For illustrative purposes, in fig.4.15, we sketch the 2ν2β and 0ν2β spectra to emphasize the
suppression of the latter. Another interesting kinematical feature emerging by this plot is that 0ν2β
gives two electrons with total energy equal to Q (in a two body decay the spectrum is fixed), while 2ν2β
decay gives two electrons with a continuous spectrum (since we have a three body decay) that extends
up to Q. In 1930 Pauli postulated the neutrino in order to explain why β-decay gives a continuous
electron spectrum rather than a line. This same kinematical feature now allows to distinguish 0ν2β
from ordinary 2ν2β decay.
counts

2ν2β 0ν2β

0 Q
Total energy in electrons

Figure 4.15: 2ν2β and 0ν2β spectra.

Assuming three Majorana neutrinos, mee can be written in terms of the neutrino masses mi , mixing
angles θij and Majorana CP-violating phases α, β as

(4.152)
X
2
mee = Uei mi = cos2 θ13 (m1 e2iβ cos2 θ12 + m2 e2iα sin2 θ12 ) + m3 sin2 θ13 .
i

Interestingly, mee and therefore 0ν2β decay rates are sensitive to the neutrino mass hierarchy and also
to Majorana CP violating phases. This is better illustrated in fig. 4.16.
64 CHAPTER 4. NEUTRINO PHYSICS

1 1
3 Bound from MAINZ and TROITSK disfavoured by 0Ν2Β
disfavoured by cosmology

1
10-1
Sensitivity of KATRIN
0.3
0.3 Dm223 < 0
mcosmo in eV

È mee È in eV
mΝe in eV

disfavoured by cosmology
disfavoured by cosmology
0.1 10-2

Dm223 < 0 Dm223 < 0


0.1 Dm223 > 0
0.03
-3
10
Dm223 > 0
0.01
Dm223 > 0
99% CL H1 dofL 99% CL H1 dofL
99% CL H1 dofL
0.03 0.003 10-4
10-3 0.01 0.1 1 10-3 0.01 0.1 1 10-4 10-3 10-2 10-1 1
lightest neutrino mass in eV lightest neutrino mass in eV lightest neutrino mass in eV

Figure 4.16: 99% CL expected ranges as function of the lightest neutrino mass for the parameters:
mcosmo probed by cosmology (fig.4.16a), mνe probed by β-decay (fig.4.16b), |mee | probed by 0ν2β
(fig.4.16c). ∆m223 > 0 corresponds to normal hierarchy (mlightest = m1 ) and ∆m223 < 0 corresponds to
inverted hierarchy (mlightest = m3 ), see fig.4.10. The darker regions show how the ranges would shrink
if the present best-fit values of oscillation parameters were confirmed with negligible error.

Summarizing, to let the 0ν2β process happen, we need a source of ∆L = 2 violation which is
provided by a Majorana neutrino mass. Therefore, the 0ν2β decay is the golden channel to unveil the
Majorana nature of neutrinos.

4.6.3 Charged Lepton Flavor Violation


The phenomenon of charged Lepton Flavor Violation (cLFV) is intimately related to neutrino masses.
Indeed, as long as neutrinos are massless, it is always possible to choose a basis for charged leptons and
neutrinos where the PMNS matrix is the identity matrix. Therefore, flavour-changing neutral-current
(FCNC) processes such as µ → eγ or µ → 3e are forbidden in the SM with massless neutrinos. Instead,
if neutrino are massive, FCNC processes occur at one loop through charged-current interactions
mediated by the W ± vector bosons as shown in fig. (4.17) for the case of µ → eγ.

γ γ

W− W− W− W−

mµ νi νi me
µR µL U ? eL
Uie µL Uµi ? eL
Uie eR
µi

Figure 4.17: Feynman diagrams generating the µ → eγ process.

The most general expression of the µ → eγ amplitude invariant under Lorentz transformations is

T (µ → eγ) = λ he|Jλem |µi (4.153)

where λ (q) is the photon polarization and he|Jλem |µi reads

he|Jλem |µi = ue (p0 ) γλ (A + Bγ5 ) + (pλ + p0λ )(C + Dγ5 ) + (pλ − p0λ )(E + F γ5 ) uµ (p) . (4.154)
 

Since λ qλ = 0, where qλ = pλ − p0λ , the last term of eq. (4.154) doesn’t contribute to T (µ → eγ) and
4.6. KINEMATICAL TESTS OF NEUTRINO MASSES 65

can be safely neglected. Exploiting the Gordon identities9


" #
(p0 + p)λ iσλν (p0 − p)ν
0
ue (p ) γλ uµ (p) = ue (p ) 0
+ uµ (p) (4.155)
me + mµ me + mµ
" #
(p0 + p)λ γ5 iσλν γ5 (p0 − p)ν
0
ue (p ) γλ γ5 uµ (p) = ue (p ) 0
+ uµ (p) (4.156)
me − mµ me − mµ

we can rewrite eq. (4.154) as

(4.157)
h i
he|Jλem |µi = ue (p0 ) γλ (Ã + B̃γ5 ) + iσλν q ν (C̃ + D̃γ5 ) uµ (p) .

Moreover, the requirement of electromagnetic gauge invariance gives the condition

(4.158)
h i
∂ λ Jλem = 0 =⇒ ue (p0 ) /q (Ã + B̃γ5 ) uµ (p) = 0

which is satisfied only provided that

à = B̃ = 0 . (4.159)

As a result, the most general gauge and Lorentz invariant amplitude of µ → eγ can be written as

T (µ → eγ) = λ ue (p0 ) [iq ν σλν (AL PL + AR PR )] uµ (p) . (4.160)

Notice that the above dipole transition requires a chirality flip between the electron and muon and
therefore AL ∝ me and AR ∝ mµ 10 . As eq. (4.160) corresponds to a dimension-five operator ēF µν σµν µ,
the amplitudes AL,R have mass dimension −1. Since µ → eγ is a loop-induced process by weak
interactions, we can estimate AR(L) in naive dimensional analysis as
!
g 2 mµ(e) X ? m2i
AR(L) =e 2 Uei f 2 Uµi (4.161)
16π 2 MW i MW

where 1/16π 2 is the loop suppression factor and g 2 /MW


2 is proportional to the Fermi constant G .
F
Moreover, the loop function f must depend on the mass ratio m2i /MW2 . Now, since m  M
i W we can
expand f around mν = 0 so that
!
g 2 mµ(e) X ? m2i

AR(L) 'e 2 Uei f (0) + f 0
(0) 2 Uµi . (4.162)
16π 2 MW i
MW

One could think that f (0) gives the leading contribution to AR(L) however, since U is unitary, it
turns out that (U U † )i6=j = 0 and therefore f (0) i Uie Uµi = 0. By an explicit loop calculation of the
P ?

diagrams of fig. (4.17) one can find that f (0) = 1/8 and thus we find that
0

GF mµ(e) µe µe
(4.163)
X
?
AR(L) ' e √ δ , δGIM = Uei Uµi (m2i /MW
2
)
2 16π 2 GIM i

where, exploiting the unitarity condition ? = 0 , we can rewrite δGIM


µe
as
P
i Uie Uµi

∆m2sun ∆m2atm
µe
δGIM ?
= Ue2 Uµ2 2 ± U ?
e3 Uµ3 2 . (4.164)
MW MW
9
From the definition σ µν = 2i [γ µ , γ ν ] where {γ µ , γ ν } = 2η µν we can write iσ µν = η µν − γ µ γ ν = γ ν γ µ − η µν and thus
ūe (p0 )iσ µν (p0ν − pν )uµ (p) = ūe (p0 )[(γ ν γ µ − η µν )p0ν − (η µν − γ µ γ ν )pν ]uµ (p) = ūe (p0 )[ p/γ
0 µ
/ ]uµ (p). Now by
− (p0µ + pµ ) + γ µ p
using the Dirac equation p /uµ (p) = mµ uµ (p) and ūe (p )p/ = me ūe (p ) it follows eq. (4.155). The expression of eq. (4.156)
0 0 0

is obtained in a very similar way further using the property {γ µ , γ5 } = 0.


10
Such a chirality flip arises in the external fermion lines by means of the equation of motion. Since only left-handed
leptons interact with the W − boson, it turns out that AL ∝ me and AR ∝ mµ , as shown in Fig. 4.17.
66 CHAPTER 4. NEUTRINO PHYSICS

where the signs ± refer to the cases of normal or inverted hierarchy, respectively. The calculation of
the decay rate of µ → eγ is straightforward and it turns out that

m3µ 
(4.165)

Γ(µ → eγ) = |AL |2 + |AR |2 ,
16π
where the factor 1/16π arises from the phase space of a two body decay while the m3µ dependence can
be understood by dimensional analysis remembering that AR(L) and the width Γ have mass dimension
−1 and +1, respectively. The above decay rate leads to the following branching ratio

Γ(µ → eγ) Γ(µ → eγ)


BR(µ → eγ) = P −
' (4.166)
f Γ(µ → f ) Γ(µ− → e− νν)

where, to an excellent approximation, we have assumed that the total decay rate of the muon is equal
to the decay rate of µ− → e− νν. Remembering that Γ(µ → eνν) = G2F m5µ /192π 3 , we finally find that

3α µe 2
BR(µ → eγ) = |δ | ≈ 3 × 10−55 (4.167)
32π GIM
far below the current experimental upper bound BRexp (µ → eγ) . 10−13 . The huge suppression of
cLFV processes within the SM with massive neutrinos tells us that any potential observation of such
processes at running or future facilities should be interpreted as an unambiguous signal of new physics.
Chapter 5

Effective Field Theories

The pioneers of quantum field theory tended to regard field theory as a fundamental description of
Nature, complete in itself. However, they realized soon after that the field theories they were studying
might be “merely” the low energy manifestation of a deeper structure, a structure first identified as
a grand unified theory and later as a string theory. Thus was developed an outlook known as the
effective field theory (EFT) approach. The general idea is that we can use field theory to say something
about physics at low energies or equivalently long distances even if we don’t know anything about
the ultimate theory, be it a theory built on strings or some as yet undreamed of structure. An EFT
provides an effective description of physics up to a certain energy scale Λ, a threshold of ignorance
beyond which physics not included in the theory comes into play.
Since EFTs are nonrenormalizable theories, they suffer from not being totally predictive, i.e. they
are not predictive all the way up to the Planck scale. However, if one is interested in making predictions
that can be compared to experiment, it is often better to use a nonrenormalizable theory rather than
a renormalizable one. By isolating the relevant degrees of freedom for a physical problem, one can
construct an appropriate effective theory which has a limited range of applicability, but is much more
predictive in that range than the corresponding renormalizable theory on which it is based. These
effective theories are generally non-renormalizable, but they are still predictive at the quantum level.
The EFT approach is equally useful irrespectively of whether the underlying ultraviolet (UV)
physics is known or not known. One may ask the question: “Why bother with an effective theory when
we know the complete theory?”. The answer is that in general the full theory can be quite complicated
and going to an effective theory simplifies things greatly. In particular, going to an effective theory
can manifest approximate symmetries that are hidden in the full theory and, as we know, increased
symmetry means increased predictive power. Furthermore, when the full theory contains several
disparate scales (m, M ) perturbation theory can be poorly behaved as typically one generates terms
containing powers of απ ln Mm . When these terms become large (because the theory is strongly coupled
α ∼ 1 and/or the logs are large), they need to be resummed in a systematic fashion in order to keep
perturbation theory under control. Working within an EFT simplifies the summation of these logs.

5.1 The Schrodinger equation


The Schrodinger theory of quantum mechanics describes a non-relativistic, non-renormalizable EFT.
Indeed, the Schrodinger equation in presence of some external potential V (r)
1 ~ 2
 
i∂0 ψ = − ∇ + V (r) ψ (5.1)
2m
can be obtained via the Euler-Lagrange equations starting from the Lagrangian density
1 ~ ? ~
L = iψ ? ∂0 ψ − ∇ψ · ∇ψ − V (r) ψ ? ψ (5.2)
2m

where the kinetic operator has mass dimension 5 as [∇]~ = 1 and [ψ] = 3 and therefore its coefficient 1
2 m
has correctly mass dimension −1. It is then clear that we should generalize the Schrodinger Hamiltonian

67
68 CHAPTER 5. EFFECTIVE FIELD THEORIES

to include all terms consistent with symmetries


!
p~ 2 p~ 2 p~ 4
H= 1 + a1 2 + a2 4 + · · · + V (r) (5.3)
2m m m

where the ai are numbers and the factors of m are added by dimensional analysis. Actually, we know
that the above terms correspond to the non-relativistic expansion of the Klein–Gordon Hamiltonian
q
p~ 2 p~ 4 p~ 6
H= p~ 2 + m2 − m = − + + ··· (5.4)
2m 8m3 16m5
so that a1 = − 14 , a2 = 18 , etc. As you well know from quantum mechanics, the Schrodinger equation is
useful even if we do not know about Lorentz invariance or that a1 = − 14 . The reason is that in the
non-relativistic limit |~
p |  m, the higher-order terms generally have a very small effect. Nevertheless,
through specially designed experiments, these coefficients can in fact be measured. For example, a1
contributes to the fine structure of the hydrogen atom, and a2 contributes to the hyperfine structure.
So even if a1 and a2 were not known from the Dirac equation, they could be measured from the
hydrogen atom. Once measured, they could then be used to make predictions: the fine structure of
helium, for example, or lots of other things.
Thus, the Schrodinger equation, and its generalization in Eq. (5.4), describe a very predictive
quantum theory. This theory is predictive despite it being non-renormalizable and having an infinite
number of terms. It is also important to note that the Schrodinger equation is not predictive for
momenta |~ p | & m, since all of the higher-order terms are then important. Thus, the Schrodinger
equation is predictive at low energy, but also indicates the scale at which perturbation theory breaks
down. If we can find a theory that reduces to the Schrodinger equation at low energy, but for which
perturbation theory still works at high energy, it is called a UV completion of the Schrodinger equation.
Thus, the Dirac and Klein-Gordon equations are UV completion of the Schrodinger equation.

5.2 The Euler-Heisenberg Lagrangian


The Euler-Heisenberg effective Lagrangian is the effective Lagrangian for photon-photon scattering at
energies much lower than the electron mass me . The lowest non-trivial operators must contain four
factors of the field strength Fµν and hence must be of dimension eight,
α2
 2
(5.5)

µν 2 µν
L= c1 (Fµν F ) + c2 Fµν F̃ .
m4e
Notice that terms with only three field strengths are forbidden by charge conjugation symmetry. The
effective interaction (5.5) is generated from the box diagram of Fig. 5.1.

(a) (b)

Figure 5.1: Light by light scattering in (a) QED and (b) in the Euler-Heisenberg effective theory. The
solid dot represents the four-photon interaction from the effective Lagrangian (5.5)

The box diagram contains four factors of the electric charge e, and one factor of 1/16π 2 for the loop.
In addition, the only dimensionful parameter other than the external momenta is the electron mass
me . This allows us to write the Lagrangian in the form (5.5), where c1,2 are dimensionless constants.
An explicit computation gives
1 7
c1 = , c2 = . (5.6)
90 90
5.3. EFT OF THE BLUE SKY 69

The low energy cross-section for γγ → γγ is obtained from the graph in the effective theory, Fig. 5.1.
The scattering amplitude is A ∼ α2 ω 4 /m4e , since each gradient of the photon field in (5.5) produces
one factor of ω. This produces a cross-section of order
!2
α2 ω 4 1
σ∼ . (5.7)
m4e ω2

The phase space factor 1/ω 2 is obtained using dimensional analysis. The cross-section must have
dimensions of area, so the phase space must have dimension −2. The only dimensionful parameter in
the effective theory is the photon energy ω, so the phase space must be proportional to 1/ω 2 . Thus we
find σ ∼ α4 ω 6 /m8e , with an error of order ω 2 /m2e from neglected higher order interactions in (5.5).

5.3 EFT of the blue sky


As a beautiful application of the EFT approach, consider the scattering of electromagnetic waves on
an electrically neutral spinless particle described by a scalar field Φ. Since Φ is neutral, the relevant
gauge invariant Lagrangian up to dimension-6 operators is given by
1 †
L = ∂µ Φ† ∂ µ Φ + m2 Φ† Φ + Φ ΦFµν F µν (5.8)
M2
with M some mass scale to be identified. The two powers of derivative in Fµν F µν tell us immediately
that the amplitude for photon scattering on this neutral particle goes like A ∝ ω 2 , with ω the frequency
of the electromagnetic wave. Thus we conclude that the scattering cross section varies like σ(ω) ∝ ω 4 .
We have arrived at Rayleigh’s celebrated explanation of the color of the sky. In passing through the
atmosphere red light scatters less than blue light on air molecules and hence the sky is blue.
For application
√ to spinless atoms or molecules, we can pass to the nonrelativistic limit setting
Φ = (1/ 2m)e −imt ϕ so that the effective Lagrangian now reads
1 1
L = ϕ† i∂0 ϕ − ∂ i ϕ† ∂ i ϕ + ϕ† ϕ (E
~2 − B
~ 2) + · · · (5.9)
2m 2mM 2
In this case, since we understand the microscopic physics governing atoms and molecules, we know
perfectly well what the mass scale M represents. The coupling of a photon to an electrically neutral
system such as an atom or a molecule must vanish like the characteristic size d of the system, since as
d → 0 the positive and negative charges are on top of each other, giving a vanishing net coupling to
the photon. Rotational invariance implies that the coupling ∼ ~k · d. ~ The scattering amplitude then
goes like A ∝ (ωd) , since the coupling has to act twice, once for the incoming photon and once for the
2

outgoing photon. Thus, invoking dimensional analysis, we obtain the cross section σ(ω) ∼ d6 ω 4 .

5.4 Decoupling of Heavy Particle


In this section we will study the procedure to construct an EFT when its UV completion is known.
Let’s assume a theory containing light degrees of freedom ϕl of mass m and heavy degrees of freedom
ϕh of mass M such that m  M . Starting from the total Lagrangian which is valid at any energy scale

L = L(ϕl , ϕh ) (5.10)

we can construct an effective Lagrangian which only depends on the light degrees of freedom ϕl by
properly getting rid of the heavy fields ϕh or, in the EFT jargon, integrating them out

L(ϕl , ϕh ) =⇒ Leff (ϕl ) . (5.11)

This Lagrangian describes physics only up to an energy scale E  M . Notice that our effective
Lagrangian may be valid well beyond m, provided E  M . If we consider the SM Lagrangian

LSM = LSM (γ, g, ν, l, u, d, s, c, b; W, Z, h, t) (5.12)


| {z } | {z }
ϕl ϕh
70 CHAPTER 5. EFFECTIVE FIELD THEORIES

the splitting into light and heavy degrees of freedom is quite well justified by the SM mass spectrum

m . few GeV , M ∼ few 100 GeV . (5.13)

In this case, being m/M  1, we can perform an expansion of LSM in the mass ratios m/M .
Let us see how an EFT is constructed considering a toy model for a Dirac fermion ψ interacting
with a real scalar field φ
1 1
L = ψ(i∂/ − m)ψ + ∂µ φ ∂ µ φ − M 2 φ2 − gψφψ (5.14)
2 2
We assume working at energy scales E ≈ m  M . The Feynman rules for this theory are

k i i
E 2 ∼k2 M 2
= −→ − = −ig (5.15)
k2 − M2 M2

Consider the elastic scattering ψψ → ψψ

ψ ψ
k
M= + crossed diagrams
φ (5.16)
ψ ψ
i
 
= (−ig)2 ABCD + crossed terms
k2 − M2
where A, B, C, D are v, u spinors.
We can construct our effective theory description of this process in two ways, through:
 the effective action Seff . Studying the scattering in configuration space, we can write an effective
action from our complete lagrangian neglecting the kinetic term and mass term of the scalar field
g2
Z Z Z
Seff = d4 x ψ(i∂/ − m)ψ + d4 x d4 y ψ(x)ψ(x)∆(x − y)ψ(y)ψ(y), (5.17)
2
where ∆ is the propagator:

1 e−ik·(x−y) 1 4
Z
k2 M 2
∆(x − y) = − d4 k −→ δ (x − y). (5.18)
(2π)4 k2 − M 2 M2
In the low energy limit E 2 ∼ k 2  M 2 the fermion currents meet in the coordinate space, so we
replace an interaction mediated by the φ scalar with a contact 4-fermion interaction:

y k2 M 2
x =⇒ x (5.19)
∆(x − y)

Our effective action reads:


g2
Z   
Seff = 4
d x ψ(i∂/ − m)ψ + ψ(x)ψ(x)ψ(x)ψ(x) (5.20)
2M 2
so at E  M our theory is effectively described by the light degrees of freedom and the heavy
degrees of freedom have been integrated out. The only remnant of the heavy field φ is contained
in the coefficient of the non-renormalizable operator ψ̄ψ ψ̄ψ through its dependence on M −2 .
5.5. FERMI THEORY 71

 the equations of motion (EoM). The procedure for the construction of the EFT is the following:

1. Write the EoM for the heavy field φ in the limit E  M :


g

( |{z} +M 2 )φ = −gψψ =⇒ φ = − ψψ. (5.21)
M2
∝E 2 M 2

We neglect the kinetic term since it is proportional to the energy scale at which we are
considering physical processes E  M . In this way we can express the heavy field φ in
terms of the light ones, the coupling and the heavy mass scale.
2. Substitute the expression for the heavy field in the complete Lagrangian ψ, φ: this way
we integrate out the heavy field. Also we neglect the kinetic term for the heavy degree of
freedom, while we cannot neglect the mass term since it turns out to be of the same order
in E/M with respect to the other terms:
2 2
1  M g g
 
φ ∂µφ −
Leff (ψ) = ψ(i∂/ − m)ψ + ∂µ 4
ψψψψ − g − 2 ψψ ψψ
|2 {z
 2 M M
(5.22)
}
∝E 2 M 2
g2
= ψ(i∂/ − m)ψ + ψψψψ
2M 2
that is the same effective Lagrangian we obtained from the contact interaction approach.

These procedures can obviously work only if the condition E  M is satisfied: we cannot consider a
process and extrapolate the behavior of the physical observables up to energy scales comparable to M .
We saw that is quite easy to obtain an EFT from a UV complete theory. If instead we do not know
what the UV theory is, we can only be guided by symmetry arguments. This is actually what is often
done when trying to extend the SM: we don’t know what kind of new physics there might be beyond it
but we know that it has to respect the SM gauge symmetry at energies reached by current accelerators.
If we follow this line, we end up with an extended SM Lagrangian where the remnants of heavy fields
are entirely captured by the coefficients of non-renormalizable operators invariant under the SM gauge
group. A remarkable fact is that an EFT exhibits the very same symmetries as the full theory. That is
why the Fermi theory works so well in its range of applicability.

5.5 Fermi theory


The Fermi theory is perhaps the most famous EFT ever of particle physics. Let us construct it following
the EoM procedure previously discussed. We write the EW Lagrangian as
1 1
LSM = − W i,µν Wµν
i
− B µν Bµν + iψγ µ Dµ ψ + (Dµ ϕ)† (Dµ ϕ) − V (ϕ† ϕ) + ψyϕψ + h.c., (5.23)
4 4
where
g
q
iψγ µ Dµ ψ = iψ∂/ ψ − √ (Wµ− J +µ + Wµ+ J −µ ) − µ
g 2 + g 0 2 Zµ (J3L µ
− s2W Jem µ
) − eAµ Jem . (5.24)
2
Let us consider energy scales E  M ∼ MW , MZ . We study separately the heavy fields

1. W ± . We write the terms of the EW Lagrangian containing the physical W bosons:


± 1 µν,i g
LW
SM = − W Wµν,i +MW2
Wµ+ W −µ − √ (Wµ− J +µ + Wµ+ J −µ )
4 2 . (5.25)
 
+ W W W + W W W W + hW W + hhW W

We are not writing explicitly the terms with more than two W bosons or with the Higgs
bosons, because these terms are suppressed by the fact that they contain many powers of M .
72 CHAPTER 5. EFFECTIVE FIELD THEORIES

More explicitly, amplitudes relative to these couplings are suppressed by 1/M 3,4 factors, with
M = MW , Mh . Exploiting the EoM of the heavy degrees of freedom, neglecting the kinetic terms:
∂L ∂L
 = 0 ⇐⇒ M 2 W − − √g J − = 0, (5.26)

+ − ∂µ + W µ µ
∂Wν  ∂∂µ Wν 2

therefore
g 1 ±
Wµ± = √ 2 Jµ . (5.27)
2 MW
As in the toy model, the mass term has to be kept since it is of the same order in MW
2 . Substituting

Eq. (5.27) in Eq. (5.23), we obtain

g2
(5.28)
± + −µ
LW
eff = − 2 Jµ J .
2MW

2. Z 0 . We can easily repeat the steps in the case of the Z boson sector. The EoM in our
approximation read:
∂L ∂L
q
(5.29)

− ∂µ = 0 ⇐⇒ MZ2 Zµ − g 2 + g 0 2 JZµ = 0,
∂Zν  ∂∂µ Zν

therefore
1
q
Zµ = g2 + g02 JZµ . (5.30)
MZ2
We obtain:
(g 2 + g 0 2 ) µ Z
LZ
eff = − JZ Jµ . (5.31)
2MZ2

Combining the two effective Lagrangians Eq. (5.28) and Eq. (5.31) we obtain:
X 1
Leff (γ, ψf,light ) = ψ f (i∂/ − mf )ψf − F µν Fµν
f
4
(5.32)
µ g2 + −µ (g 2 + g 0 2 ) µ Zµ
− eAµ Jem − J
2 µ J − JZ J
2MW 2MZ2

Since we are considering E  100 GeV, it is understood that the top quark has been already integrated
out. The charged currents are therefore built with the quarks but the top and leptons:

Jµ+ = dL γµ uL + eL γµ νL (5.33)

Jµ− = uL γµ dL + ν L γµ eL = (Jµ+ )† (5.34)


Let us compute the first term:

g2 g2 1 
Jµ+ J −µ = − eγµ (1 − γ5 )ν νγ µ (1 − γ5 )e , (5.35)
 
− 2 2
2MW 2MW 4

that is exactly the (V − A) ⊗ (V − A) Fermi theory. Therefore, a tree level matching of the coefficients
provides us with the expression of the Fermi constant GF as a function of the SM parameters

GF g2 1
√ = 2 = 2 (5.36)
2 8MW 2v

Now, we can extract the value of GF and thus of v matching the SM prediction of the muon decay rate

G2F m5µ
Γ(µ → eνν) = × (1 + . . . ) (5.37)
192π 3
5.6. THE WEINBERG OPERATOR 73

with the corresponding experimental result. We obtain

GF = 1.1663787(6) × 10−5 GeV−2 =⇒ v = 246.22 GeV . (5.38)

It is remarkable that the strengths of charged-current and neutral-current interactions are the same

g2 4GF (g 2 + g 0 2 ) MW 2
2 = √ = =⇒ = 1 = ρ. (5.39)
2MW 2 2MZ2 MZ2 cos2 θW

The value of parameter ρ = 1 is an important feature arising from an underlying global symmetry
of the Higgs sector (custodial symmetry) which is due to the fact that the Higgs field is a doublet
under SU (2)L . As we shall see, this symmetry is broken at the loop level by U (1)Y gauge interactions
(proportional to g 0 ) as well as by Yukawa interactions (since yu 6= yd ). As a result, deviations of ρ from
1 will be allowed at the 0.1% level. Finally we can write the effective Lagrangian of the Fermi theory:
X 1
Leff (γ, ψf,light ) = ψ f (i∂/ − mf )ψf − F µν Fµν
4
f
(5.40)
4GF 
µ
− √ Jµ+ J −µ + JZµ J Zµ

− eAµ Jem
2

which explicitly shows that neutral-current and charged-current interactions are weighted by the same
factor proportional to the Fermi constant GF .

5.6 The Weinberg operator


The see-saw mechanism provides another example of how an EFT works. In this model, we introduce a
completely sterile heavy right-handed neutrino NR with mass MN  v interacting with light neutrinos
through Yukawa interactions. Integrating out the heavy neutrino, we generate a non-renormalizable
operator, the Weinberg operator, which is remarkably the only dimension-5 operator made of SM fields
and respecting the SM gauge symmetry. The relevant Lagrangian of the see-saw mechanism is

1
− Lsee−saw = N R ϕ̃† yν LL + N R CM N TR +h.c. (5.41)
2| {z }
c
N R M NR

where M = M T since NR has a Majorana mass and the following properties hold

N c = CN T , CC † = C † C = 1, C † = C T = −C, C 2 = −1 (5.42)

We write the EoM of NR to integrate it out

∂Lsee−saw
0= = yν ϕ̃† LL +CM N TR =⇒ N TR = −C −1 M −1 yν χ = CM −1 yν χ (5.43)
∂N R | {z }
χ

Substituting N TR in Lsee−saw and exploiting the fact that C and M commute and C 2 = −1, we find

1
−Lsee−saw
eff = χT C T yνT M −1 yν χ + χT C T yνT M −1 CM CM −1 yν χ + h.c. (5.44)
2
1 T
= χ C(yνT M −1 yν )χ + h.c. (5.45)
2

Since χ = ϕ̃† LL it has mass dimension [χ] = M 2 . As a result, the effective operator of Eq. (5.45) is
5

the famous dimension-5 Weinberg operator.


74 CHAPTER 5. EFFECTIVE FIELD THEORIES

5.7 Renormalization of EFTs


The Fermi theory is non-renormalizable since GF has mass dimension −2. Thus, we must add to it all
terms consistent with its symmetries. For example, we may have terms such as
−L = GF ψ̄ψ ψ̄ψ + a1 G2F ψ̄ψψ̄ψ + · · · , (5.46)
where a1 is a number and the factor of G2F has been added by dimensional analysis. Derivatives
and γ-matrices can be inserted anywhere and we neglected flavor indices for simplicity. Despite
these additional terms with unknown coefficients, the 4-Fermi theory is very predictive. In fact, the
higher-order terms in L will affect physical observables by factors of (GF E 2 )n for n > 1, where E is
the typical energy scale of the process. As long as we are interested in low-energy processes, i.e. if
GF E 2  1, these higher-order terms can be safely neglected.
Besides tree-level predictions, one can also calculate loops in this non-renormalizable theory and
derive testable predictions. For simplicity, let us imagine that all the fermions in Eq. (5.46) are identical.
At tree-level, the Lagrangian (5.46) generates S-matrix elements for the process ψ̄ψ → ψ̄ψ of the form
Mtree (s) ∼ GF + a1 G2F s + · · · , (5.47)
where s does not necessarily represent s = (p1 + p2 but any kinematical Lorentz-invariant quantity of
)2
mass dimension 2, and we are ignoring the external spinors for simplicity. At low energies, sGF  1,
this scattering is dominated by the leading term, with subleading terms suppressed by powers of
sGF  1. At 1-loop, there is a contribution of the form
!
d4 k 1 1 Λ2
Z
Mloop (s) = ∼ G2F ∼ G2F 2
b0 Λ + b1 s + b2 s ln . (5.48)
(2π)4 k/ k/ s

On the right, we have parametrized the possible forms the result could take with three finite and
calculable constants b0 , b1 , and b2 and a cut-off scale Λ. Without any symmetry arguments, there is no
reason to expect that any of the constants bi should vanish. Thus,
Mtree + Mloop ∼ (GF + b0 Λ2 G2F ) + sG2F (a1 + b1 + b2 ln Λ2 ) − b2 G2F s ln s + · · · , (5.49)
where we have grouped terms by their momentum dependence. The key term in this expression is the
b2 s ln s term, which has no analog coming from the classical Lagrangian, Eq. (5.46).
Indeed, UV enhanced contributions do just renormalize the tree-level couplings GF and a1 . There-
fore, divergent effects can be reabsorbed by renormalization of coefficient of lower dimension operators
Λ2
ḠF = GF + b0 Λ2 G2F , ā1 = a1 + b1 + b2 ln (5.50)
s0
where ḠF and ā1 are renormalized (finite) couplings and s0 an arbitrary scale. This renormalization
removes almost the entire result of the loop but one term ∼ ln s. The renormalized matrix element is
s
 
2
M(s) = Mtree + Mloop ∼ ḠF + sḠF ā1 − b2 ln + ··· . (5.51)
s0
At the scale s = s0 this is identical to the tree-level prediction. If the s dependence of the distribution
at low energies is well-enough measured, ḠF , ā1 , b2 , etc. can be extracted from data. Although the
constants ā1 is not calculable, the constant b2 is. More precisely, one could plot
M(s1 ) − ḠF M(s2 ) − ḠF s2
2
− 2
∼ b2 ln + O(ḠF s1 ) , (5.52)
s1 ḠF s2 ḠF s1
and see whether the logarithmic scale dependence agrees with the theoretical calculation. Thus, b2 is a
genuine testable prediction from a loop calculation in a non-renormalizable theory. The reason this
works is because the ln s dependence can never come from a tree-level calculation. This is because
tree-level calculations come from local Lagrangians that have only integer powers of derivatives, never
terms such as ψ̄ψ ψ̄ ln ψ. This is a general result:
Non-analytic energy dependence is a testable quantum prediction of non-renormalizable as well as
renormalizable theories.
5.8. EFTS AND UNITARITY 75

5.8 EFTs and unitarity


Now we want to discuss from a general point of view the energy limits up to which an EFT is valid. To
do this we will exploit the unitarity of the S matrix which stems from the conservation of probability.
This discussion is very useful in trying to get a feeling of the energy scales at which, constructing an
EFT to extend the SM, we can expect new physics and, experimentally, new resonances. The Fermi
theory is a classic example. The cross-section of any weak process within the Fermi theory goes as
σ ∝ G2F E 2 where E is the energy scale of the process. However, this would mean that the cross-section
diverges at very high energy. This is an unphysical behavior which is just telling us that an EFT is
trustable only up to an energy scale to be determined. In the following, we will discuss the so-called
partial wave unitarity bound which is based on the optical theorem.

5.8.1 The optical theorem


One of the most important implications of the unitarity of the scattering matrix S is the so-called optical
theorem which establishes a relation between scattering amplitudes and cross sections. Decomposing
the S matrix in terms of the transfer matrix T

S = 1 + iT (5.53)

and exploiting the unitarity condition SS † = S † S = 1 = (1 + iT )(1 − iT † ) we obtain

i (T † − T ) = T † T . (5.54)

The amplitude for a transition between the states i → f is related to the transfer matrix by

hf |T |ii = (2π)4 δ 4 (pi − pf )M(i → f ) (5.55)

and therefore we can write

hf |i(T † − T )|ii = ihi|T |f i? − ihf |T |ii = i(2π)4 δ 4 (pi − pf ) (M? (f → i) − M(i → f )) . (5.56)

Using the completeness relation


XZ
dΠn |nihn| = 1, (5.57)
n

being dΠn the differential phases space for the multiparticle state |ni

Y d3 pj 1
dΠn = , (5.58)
j∈n
(2π)3 2Ej

we obtain
XZ

hf |T T |ii = dΠn hf |T † |nihn|T |ii
(5.59)
n
X Z
= (2π)4 δ 4 (pf − pn )(2π)4 δ 4 (pi − pn ) dΠn M(i → n)M? (f → n).
n

Equating Eq. (5.56) and Eq. (5.59) we get the generalized optical theorem

XZ
M(i → f ) − M? (f → i) = i dΠn (2π)4 δ 4 (pi − pn )M(i → n)M? (f → n). (5.60)
n
76 CHAPTER 5. EFFECTIVE FIELD THEORIES

5.8.2 Partial wave unitarity bounds


An important implication of the optical theorem is that scattering amplitudes cannot be arbitrarily
large. That unitarity bounds should exist follows from conservation of probability. Roughly speaking,
the optical theorem says that Im M ≤ |M|2 , which implies |M| < 1. In order to make this more
quantitative, we discuss here the so-called partial wave unitarity bound. Let us consider the cross-section
of a AB → AB elastic process
1
Z
σ(AB → AB) = d cos θ|M(s, t(θ))|2 . (5.61)
32πs
We can decompose the amplitude in partial waves, remarkably factorizing the dependences on the
energy s and on the angles θ, as

(5.62)
X
M(s, θ) = 16π (2j + 1)aj (s)Pj (cos θ)
j=0

where Pj are the Legendre polynomials satisfying Pj (1) ≡ 1 and the orthogonality condition
Z 1 2
d cos θPj (cos θ)Pk (cos θ) = δjk . (5.63)
−1 2j + 1

Exploiting this properties, we obtain



16π X
σ(AB → AB) = (2j + 1)|aj (s)|2 . (5.64)
s j=0

If X is a multiparticle state, in the center-of-mass frame, the cross section is given by


1
Z
σ(AB → X) = dΠX (2π)4 δ 4 (pA + pB − pX )|M(AB → X)|2 . (5.65)
4ECM |~
pi |

Particolarizing the generalized optical theorem of Eq. (7.4) to the case |ii = |f i = |ABi, we have
1X
Z
Im M(AB → AB; θ = 0) = dΠX (2π)4 δ 4 (pA + pB − pX )|M(AB → X)|2 . (5.66)
2 X

The requirement that θ = 0 is necessary to impose that the final and initial states coincide. Being a
particle state defined by the type of particle involved and their momenta, the initial and final states
are the same only if the single momenta do not change, hence θ = 0. Therefore
X
Im M(AB → AB; θ = 0) = 2ECM |~
pi | σ(AB → X)
X
≥ 2ECM | p~i | σ(AB → AB) (5.67)

being the elastic cross section less then the total one. Therefore, substituting Eqs. (5.62) and (5.64)
in Eq. (5.67), exploiting the fact that the Legendre polynomials are real, and knowing that in the
√ √ √
center-of-mass frame |~pi | = s/2 and ECM = s (in the limit s  mA , mB ), we obtain

Im aj ≥ |aj |2 (5.68)

which can be rewritten as 2  2


1 1

2
(Re aj ) + Im aj − ≤ , (5.69)
2 2
implying, in general, the following constraints:
1
|aj | ≤ 1, 0 ≤ Im aj ≤ 1, |Re aj | ≤ . (5.70)
2
5.8. EFTS AND UNITARITY 77

In the complex plane (Re aj , Im aj ) we have a circle of avaiable aj .


Im aj

0.5

Re aj

The constraints on the aj coefficients, which only depend on s, impose in turn a bound on the energy s.
We are ready now to establish the validity limit of an EFT such as the Fermi theory. Let’s consider,
as an example, the weak process
eνµ −→ µνe (5.71)
which is described in the s  MZ2 limit by the Fermi theory. One can easily show that the unpolarized
squared amplitude for this process is given by

|M|2 = 32G2F s2 (5.72)

and in particular it is independent on the θ angle. Inverting Eq. (5.62) we find

1
Z +1
aj (s) = d cos θPj (cos θ)M(s, t). (5.73)
32π −1

Being |M(s)|2 independent on θ, i.e on t, in our case the only surviving coefficient is the monopole

GF s
|a0 (s)| = √ . (5.74)
2 2π
Imposing the constraints |a0 (s)| ≤ 1 of Eq. (5.70) we find that
s √
√ 2 2π
s≤ ≈ 900 GeV . (5.75)
GF

We have obtained an upper bound on the energy scale at which our EFT breaks down: if the energy
exceeds this upper bound, the S matrix is no more unitary and the EFT becomes meaningless.
Computing the cross section in the full SM theory one finds
 2
G s 2
 F s  MW

G2 s MW2 
σ(eνµ → µνe ) = F π (5.76)
π s + MW2 → 2 2
.
 F MW
 G
s 2
MW
π
As shown by eq. (5.76), the cross section in the SM has a good ultraviolet behavior: the mass of the
charged weak boson MW acts as a cut off of the Fermi EFT. The crucial point is that even without
knowing the full theory we could have expected something (new physics) appearing before the unitarity
bound of 900 GeV: this is precisely the charged gauge boson. Therefore we can say that the Fermi
theory is unitarized by the presence of vector bosons at high energy, that gives us a physical threshold
at which we stop trusting our EFT.
78 CHAPTER 5. EFFECTIVE FIELD THEORIES
Chapter 6

The weak neutral current discovery

In the 1960s, a requirement for any gauge model of electroweak interactions was that it should contain
the well-established V − A weak charged currents in the low-energy limit. But the model introduced
by Glashow, Weinberg and Salam, which became soon the SM of particle physics, also predicted a set
of new low-energy phenomena associated with the neutral weak currents
g g
LNC+CC
SM = − √ (Wµ− J +µ + Wµ+ J −µ ) − Zµ JZµ (6.1)
2 c W

where cW ≡ cos θW and sW ≡ sin θW and the weak charged-currents Jµ± read
Jµ− = uL γµ dL + ν L γµ eL Jµ+ = dL γµ uL + eL γµ νL , (6.2)
while the weak neutral-current JZµ is given by
Xh i 1 Xh f i
JZµ = gLf (f L γ µ fL ) + gR
f
(f R γ µ fR ) = f
gV (f γ µ f ) − gA (f γ µ γ5 f ) , (6.3)
f
2 f

where the chiral couplings gL(R) and the vector and axial couplings gV(A) = gL ± gR are given by
 
g f = T f − Qf s2 g f = T f − 2s2 Qf
(6.4)
L 3L W V 3L W
.
g f = −Q s2 g f = T f
R f W A 3L

Thus in the SM prediction for neutral currents there is only one unknown parameter, the Weinberg
angle θW , which reflects the relative strenght of the SU (2)L and U (1)Y gauge coupling constants.
The neutral currents were first discovered by the Gargamelle collaboration at CERN through the
elastic scattering processes of muon neutrinos and antineutrinos on atomic electrons, νµ e− → νµ e−
and ν µ e− → ν µ e− . High-energy neutrino processes were studied at FermiLab, at the Tevatron, and
at CERN during the 1970s. The results from these experiments were of crucial importance for the
development of the theory of elementary particles. Indeed, from the neutrino data it was possible to
extract sin2 θW ' 0.24 which was then used to predict the masses of the intermediate vector bosons
s
πα MW
MW = √ ≈ 77.5 GeV MZ = ≈ 89.1 GeV . (6.5)
2GF s2W cW

Although the above determination of MZ was not very precise, it was crucial to indicate the energy
scale at which a dedicate experiment had to run in order to produce on-shell the Z boson. Such a task
was subsequently accomplished by the LEP experiment at CERN which performed, in its first phase
(LEP-I), electroweak precision measurements at energies close to Z boson pole mass.
In the following sections, we will study the cross-section for elastic scattering processes of neutrinos
and antineutrinos on atomic electrons measured through of high-energy neutrino beams, with energies
Eν from a few GeV to several hundred GeV. In all cases s = 2me Eν  MW,Z 2 and therefore the use of
the effective Lagrangian
4GF 
Leff = − √ Jµ+ J −µ + JZµ JZµ (6.6)

2
is justified. We proceed now to analyse the processes νµ e− → νµ e− and ν µ e− → ν µ e− .

79
80 CHAPTER 6. THE WEAK NEUTRAL CURRENT DISCOVERY

6.1 νµ e− → νµ e− and ν µ e− → ν µ e−
In the SM, the processes νµ e− → νµ e− and ν µ e− → ν µ e− are induced, at tree-level, only by NC
interactions with a t-channel exchange of the Z boson, as shown by the Feynman diagrams of fig. 6.1.
Setting gVν = gA
ν = 1/2, the SM amplitude of ν e− → ν e− then reads
µ µ

2
ig i(−gµν + qµ qν /MZ2 ) 1
   
M= u(k 0 )γ µ (1 − γ5 )u(k) u(p0 )γ ν (gVe − gA
e
γ5 )u(p) (6.7)
cW q 2 − MZ2 8

Since for GF s  1 the EFT approach is justified, we can evaluate the amplitude of νµ e− → νµ e− also

νµ (k) νµ (k 0 ) ν̄µ (k) ν̄µ (k 0 )

Z(q) Z(q)

e− (p) e− (p0 ) e− (p) e− (p0 )

Figure 6.1: Feynman diagrams for the processes νµ e− → νµ e− and ν̄µ e− → ν̄µ e− in the SM.

by means of Leff , see Eq. (6.6). From the relevant Feynman diagram in the EFT, see fig. 6.2, we find

4GF 1
  
M = −i √ × × 2 u(k 0 )γ µ (1 − γ5 )u(k) u(p0 )γµ (gVe − gA
e
γ5 )u(p) (6.8)
2 8

where the factor of 2 stems from the two identical currents. As expected, the result of Eqs. (6.7) and
(6.8) coincide in the low-energy limit q 2  MZ2 as G
√F =
2
g2
8M 2 c2
with MW
2 = M 2 c2 .
Z W
Z W

νµ (k) νµ (k 0 ) ν̄µ (k) ν̄µ (k 0 )

G
√F G
√F
e− (p) 2 e− (p0 ) e− (p) 2 e− (p0 )

Figure 6.2: Feynman diagrams for the processes νµ e− → νµ e− and ν̄µ e− → ν̄µ e− in the Fermi theory.

Now we compute the amplitude modulus squared averaging (summing) over the initial (final) spin
states. Working in the massless limit for both neutrinos and electrons and setting now on gVe = gV and
e = g , we obtain
gA A

1 1 X G2 X
× |M|2 = F [u(k 0 )γ µ (1 − γ5 )u(k)u(p0 )γµ (gV − gA γ5 )u(p) × (6.9)
1 2 spins 4 spins
u(p)γν (gV − gA γ5 )u(p0 )u(k)γ ν (1 − γ5 )u(k 0 )] (6.10)
G2F
= Tr[k/0 γ µ (1 − γ5 )k/γ ν (1 − γ5 )] Tr[p
/0 γµ (gV − gA γ5 )p
/γν (gV − gA γ5 )] (6.11)
4
Exploiting the relations

Tr[γ µ γ ν γ σ γ δ ] = 4(η µν η σδ − η µσ η νδ + η µδ η νσ ) Tr[γ µ γ ν γ σ γ δ γ5 ] = −4iεµνσδ (6.12)

as well as
εµναβ εµνγδ = −2(δγα δδβ − δδα δγβ ) (kµ kν0 + kν kµ0 )εµναβ = 0 (6.13)
6.1. νµ e− → νµ e− AND ν µ e− → ν µ e− 81

we find that
1 X
 
|M|2 = 16G2F (p · k)(p0 · k 0 )(gA + gV )2 + (p0 · k)(p · k 0 )(gA − gV )2 . (6.14)
2 spins | {z } | {z }
s2 /4 t2 /4

Since the Mandelstam variables in the massless limit and in the laboratory frame read

(6.15)
m→0 lab
s = (p + k)2 = (p0 + k 0 )2 = 2p · k = 2p0 · k 0 = 2me Eν

(6.16)
m→0 lab
t = (p − k 0 )2 = (p0 − k)2 = −2p0 · k = −2p · k 0 = −2me Eν0
we find the following differential cross-section

(2π)4 δ 4 (p + k − p0 − k 0 ) 1 X d3 p0 d3 k
 
dσ = q |M|2
(p · k)2 − m 2m 
2 2 spins (2π)3 2Ee0 (2π)3 2Eν
e ν
2 
. (6.17)
G2F me Eν 0
2 Eν
 
2 3 0 4 0 0 3 0
= (gV + g A ) + (gV − gA ) d p δ (p + k − p − k )d k
(2π)2 Eν0 Ee0 Eν

Integrating over δ 4 (p + k − p0 − k 0 )d3 k 0 we obtain the momentum conservation ~k = ~k 0 + p~0 . Then we


write d3 p0 = Ee0 2 dEe0 d cos θdϕ and integrate over dϕ (which gives a factor of 2π):
2 
G2F me Eν Ee0 Eν0
 
dσ = 0
(gV + gA )2 + (gV − gA )2 dEe0 d cos θδ(Eν + me − Eν0 − Ee0 ) (6.18)
(2π) Eν Eν

Now we exploit the momentum conservation in the massless limit:


q q
|~k 0 | = |~k|2 + |~
p 0 |2 − 2|~k||~
p 0 | cos θ =⇒ Eν0 = Eν2 + Ee0 2 − 2Eν Ee0 cos θ (6.19)

and the properties of the Dirac delta


−1
1 df Eν0
Z 
δ(f (cos θ))d cos θ = 0
= = (6.20)
|f (cos θ0 )| d cos θ f =0 Eν Ee0

and we obtain 2 
G2 Eν0
 
dσ = F me (gV + gA )2 + (gV − gA )2 dEe0 . (6.21)
(2π) Eν
Then we exploit the energy conservation in the laboratory frame in the massless limit Eν = Eν0 + Ee0
Ee0 Eν0
and setting y = Eν
we have Eν = 1 − y. From the 2 body kinematics of elastic scattering we have that
0 ≤ y ≤ 1. Performing the integration over y
Z 1 1
dy(1 − y)2 = (6.22)
0 3
we can find the total cross section for this elastic scattering

G2 1
 
σνelµ e = F s (gV + gA )2 + (gV − gA )2 (6.23)
4π 3

The cross-section of the process ν µ e− → ν µ e− can be obtained in a straightforward way from the
above result. One could naively think that the cross sections for these processes are the same, but this
is not actually the case as weak interactions violate parity and charge conjugation. From the Feynman
diagram on the right of fig. 6.2, we can write the amplitude for ν µ e− → ν µ e−

GF
  
Mν µ e = −i √ v(k)γ µ (1 − γ5 )v(k 0 ) u(p0 )γµ (gV − gA γ5 )u(p) , (6.24)
2
82 CHAPTER 6. THE WEAK NEUTRAL CURRENT DISCOVERY

to be compared with the amplitude for νµ e− → νµ e−

GF
  
Mνµ e = −i √ u(k 0 )γ µ (1 − γ5 )u(k) u(p0 )γµ (gV − gA γ5 )u(p) . (6.25)
2
Since in the massless limit uu = vv, the only difference between the two amplitudes is the exchange
P P

between k and k 0 that corresponds from the point of view of the Mandelstam variables to the exchange
of s and t in Eq. (6.14). Therefore
1 X 1 X
|Mν µ e |2 = |Mνµ e (s ↔ t)|2 (6.26)
2 spins 2 spins

and we obtain the cross section for ν µ e− → ν µ e− :

G2F 1
 
σνelµ e = s (gV + gA )2 + (gV − gA )2 (6.27)
4π 3

where the factor 13 moved from the t term to the s term. Obviously, the overall dependence on s
remains unchanged since it has to do only with the kinematics.

6.2 νe e− → νe e− and ν e e− → ν e e−
Another important neutrino scattering process is given by ν e e− −→ ν e e− . In this case, both NC and
CC contributions have to be considered, as shown in fig. 6.3

ν e (k) ν e (k 0 )
e− (p) ν e (k 0 )

W−
Z

ν e (k) e− (p0 )
e− (p) e− (p0 )

Figure 6.3: Feynman diagrams for the process ν̄e e− → ν̄e e− in the SM.

The amplitudes for the two diagrams read:


GF
  
MNC
νee = −i √ v(k)γ µ (1 − γ5 )v(k 0 ) u(p0 )γµ (gV − gA γ5 )u(p) , (6.28)
2
GF
  
MCC
νee= −i √ v(k)γ µ (1 − γ5 )u(p) u(p0 )γµ (1 − γ5 )v(k 0 ) . (6.29)
2
In order to sum the two amplitudes, it is convenient to exploit the Fierz identity

ψ 1L γ µ ψ2L ψ 3L γµ ψ4L = ψ 1L γ µ ψ4L ψ 3L γµ ψ2L (6.30)


     

which makes the form of the two amplitudes equal. Therefore, the total amplitude reads
GF
       
MNC+CC
νee = −i √ v(k)γ µ (1 − γ5 )v(k 0 ) u(p0 )γµ gV + 1 − gA + 1 γ5 u(p) . (6.31)
2
and the cross section of the process can be readily obtained from Eq. (6.27)

G2 1
 
σνele e = F s (gV + gA + 2)2 + (gV − gA )2 (6.32)
4π 3
6.3. THE WEAK NC EFFECTS IN e+ e− → µ+ µ− 83

Finally, using the tricks outlined before, we can immediately write the cross-section for νe e → νe e

G2 1
 
σνele e = F s (gV + gA + 2)2 + (gV − gA )2 (6.33)
4π 3

The cross sections σνelµ e , σνelµ e , σνele e , and σνele e , which were measured long ago, represent four ellipses
in the (gA , gV ) plane, see fig. 6.4. Even if their measurements are consistent with the SM predictions, i.e.
the ellipses overlap within the experimental uncertainties, two regions are allowed and we need another
observable to remove the residual degeneracy. Such an observable is provided by the forward-backward
asymmetry of the process e+ e− → µ+ µ− which we are going to discuss now.

2002
1.0
-0.032

1987
-0.035 mH
–ν -
µe
gVl

mt

-0.038
∆α
0.5 +−
ll
+ −
ee
+ −
µµ
+ −
ττ 68% CL
-0.041
-0.506 -0.503 -0.5 -0.497
gAl

gV 0.0 e+ e - →
µ+ µ -

-0.5 –ν e -
e νµ e-

–ν e - νe e -
e
-1.0
-1.0 -0.5 0.0 0.5 1.0
gA

Figure 6.4: The neutrino scattering and e+ e annihilation data available in 1987 constrained the values
of gV` and gA
` to lie within broad bands, whose intersections helped establish the validity of the SM and

were consistent with the hypothesis of lepton universality. The inset shows the results of the LEP/SLD
measurements at a scale expanded by a factor of 65. The flavour-specific measurements demonstrate
the universal nature of the lepton couplings unambiguously on a scale of approximately 0.001.

6.3 The weak NC effects in e+ e− → µ+ µ−


Processes involving neutrinos are ideal candidates to test weak interactions (both CC and NC ones)
since neutrinos are not sensitive (at leading order) to neither QED nor QCD effects. Still, NC weak
contributions can play an important role also in processes such as e+ e− → `+ `− (with ` = e, µ, τ )
which can be mediated by the tree-level exchange of a photon or a Z boson, as shown in fig. 6.5
Whether weak-interactions are relevant or not depends on the energy we consider. In fact, by simple
NDA arguments, we can estimate the size of QED and weak NC contributions to the e+ e− → µ+ µ−
84 CHAPTER 6. THE WEAK NEUTRAL CURRENT DISCOVERY

e− (p− ) µ+ (k+ )

γZ

P
e+ (p+ ) µ− (k− )

Figure 6.5: Feynman diagram for the process e+ e− → µ+ µ− in the SM

cross-section as
 √ 2
s
+ −
σweak (e e → µ µ ) ∼ + −
G2F s ∼ 10 −10
(6.34)
1 GeV
2
α2 1 GeV

σQED (e+ e− → µ+ µ− ) ∼ ∼ 10−4 √ . (6.35)
s s

Since σweak /σQED ∝ s2 , QED interactions are by far dominant at low-energy s . 10 GeV while, for
√ √
increasing values of s, weak effects strart being relevant and σweak /σQED ∼ 1 for s ∼ 30 GeV. This

simple argument represented the main motivation to conceive e+ e− colliders running at s & 30 GeV
such as PETRA, PEP, and TRISTAN in order to probe NC weak interactions.
Writing the total amplitude of e+ e− → µ+ µ− has the sum of electromagnetic and weak contributions

M = Mγ + MZ (6.36)

the total unpolarized amplitude modulud squared reads

1 1 1 X
 
(6.37)
X
|M|2 = |M|2 = |Mγ |2 + |MZ |2 +2 Re(Mγ? MZ ) .
2se+ + 1 2se− + 1 spins 4 spins | {z } | {z } | {z }
∼1 ∼(GF s)2 GF s

As anticipated, at low-energy, the leading terms are in order the pure electomagnetic one, the interference
term, which goes as ∼ GF s  1, and the pure weak term of order (GF s)2 . The amplitudes Mγ,Z read

−iηµν
 
Mγ = (ie)2 v(p+ )γ µ u(p− )u(k− )γ ν v(k+ ) (6.38)
P2

4GF 1
MZ = −i √ × × 2v(p+ )γµ (gV − gA γ5 )u(p− )u(k− )γ µ (gV − gA γ5 )v(k+ ) (6.39)
2 4
and therefore their interference is given by
X 2GF e2 X
M?γ MZ = − √ 2
v(k+ )γ µ u(k− )u(p− )γµ v(p+ )
spins 2 P spins

× v(p+ )γ ν (gV − gA γ5 )u(p− )u(k− )(gV − gA γ5 )v(k+ ) . (6.40)



2GF e2    
=− Tr k
/ + γ µ k
/ − γ ν (gV − gA γ 5 ) Tr p
/ + γ µ p
/ − γ ν (gV − gA γ 5 )
P2

As we are assuming energies s  me , mµ we can safely work in the lepton massless limit me = mµ = 0.
In this limit the Mandelstam variables sum to zero s + t + u = 0 and it turns out that
s
p+ p− =k+ k− =
2
t
p+ k+ =p− k− = − (6.41)
2
s+t
p+ k− =p− k+ = .
2
6.3. THE WEAK NC EFFECTS IN e+ e− → µ+ µ− 85

Computing the traces we obtain the following expression



8 2
 
(6.42)
X
?
GF e2 gV2 t2 + (s + t)2 − gA
2 2
t − (s + t)2 .
  
Mγ MZ = −
spins
s

It is convenient to work in the CoM frame where the process e+ e− → µ+ µ− is schematized as

µ+
k+
p+
θ
e+ e−
p−

k−
µ−

In the CoM frame s = (2ECM )2 ≡ 4E 2 and the 4-momenta of e± and µ± read


√ √
s s
p+ = (1, 1, 0, 0), k+ = (1, cos θ, sin θ, 0),
√2 √2 (6.43)
s s
p− = (1, −1, 0, 0), k− = (1, − cos θ, − sin θ, 0).
2 2
Therefore, since
s
t = −2p+ · k+ = − (1 − cos θ) = −2E 2 (1 − cos θ), (6.44)
2
s2
t2 + (s + t)2 = 4E 4 (1 − cos θ)2 + 4E 4 (1 + cos θ)2 = (1 + cos2 θ), (6.45)
2
t2 − (s + t)2 = −s2 cos θ, (6.46)
we can write the three contributions to the unpolarized amplitude modulus squared
1 X √
M?γ MZ = − 2 2GF e2 s gV2 (1 + cos2 θ) + 2gA
2
 
cos θ
2 spins
1 X
|Mγ |2 =e4 (1 + cos2 θ) (6.47)
4 spins
1 X
|MZ |2 =2G2F s2 (gV2 + gA
2 2
) (1 + cos2 θ) + 8gV2 gA
2
 
cos θ
4 spins

Notice that the interference term is not invariant under the exchange gV ↔ −gA providing therefore
the possibility to remove the degeneracy in the gV and gA left by neutrino scattering data, see fig.6.4.
Finally the differential cross section of e+ e− → µ+ µ− can be written as

dσ 1 2
π αem
 
= |M|2 = A(1 + cos2 θ) + B cos θ , (6.48)
d cos θ 32πs 2 s

where we have factored out the QED contribution and defined the quantities A and B as follows

4GF s 2(GF s)2 2


A =1 − √ 2 gV2 + (gV + gA2 2
) ,
2e e4
(6.49)
8GF s 2 2(GF s)2 2 2
B = − √ 2 gA + 8gV gA .
2e e4

Note that A and B appear clearly like an expansion in the adimensional parameter GF s.
86 CHAPTER 6. THE WEAK NEUTRAL CURRENT DISCOVERY

Let us define now the forward-backward asymmetry AFB :


σF − σB
AFB = , (6.50)
σF + σB
where σF and σB are, respectively, the forward and the backward cross section
Z 1 

 2
π αem 4

1

σF = d cos θ = A+ B , (6.51)
0 d cos θ 2 s 3 2
Z 0 dσ
  2
π αem 4

1

σB = d cos θ = A− B . (6.52)
−1 d cos θ 2 s 3 2
Therefore the total cross section and forward-backward asymmetry read

2
4π αem 3B
σ= A AFB = . (6.53)
3 s 8A

In the low-energy limit (GF s  1), we can safely neglect quadratic terms in GF s and we obtain
2
4π αem

4GF s

3GF s 2
σ' 1 − √ 2 gV2 AFB ' − √ 2 gA . (6.54)
3 s 2e 2e
As we can see, NC weak-interactions induce a twofold effect: 1) modification of the total cross-section
σ(e+ e− → µ+ µ− ) from that of QED. At low-energy, this is proportional to gV2 , 2) generation of a
forward-backward asymmetry (which is absent in QED) which at low-energy measures gA 2 . Already at

the PETRA energiew s ≈ 30 GeV, these effects could be measured to map out an allowed region in
the (gV , gA ) plane, see fig. 6.4. This singles out one of the remaining two solutions, that of the SM
with sin2 θW ≈ 0.24. In particular, AFB turned out to be the most sensitive probe of weak interactions
√ !2
2
s gA
AFB ≈ 6 × 10 −2
(6.55)
30 GeV 0.5

as |gA | = 0.5 while |gV | ' 0.04. Finally, let us remark that AFB is generated, at low-energy, by the
interference between the vectorial QED current and the axial-vector current of the weak interactions.
Therefore, it is a purely parity-violating observable.
Chapter 7

The Standard Model at LEP-I

With the observation of neutral current interactions in neutrino-nucleon scattering in 1973 and the
discovery of the W and Z bosons in pp̄ collisions ten years later, the key features of the SM were
well established experimentally. The LEP and SLC accelerators were then designed during the 1980s
to operate at centre-of-mass energies of approximately 91 GeV, close to the mass of the Z boson, to
produce copious numbers of Z bosons via e+ e− annihilation, allowing detailed studies of the properties
of the Z boson to be performed in a very clean environment. The data accumulated by LEP and SLC
in the 1990s are used to determine the Z boson parameters with high precision: its mass, its partial
and total widths, and its couplings to fermion pairs. These results are compared to the predictions of
the SM and found to be in agreement. From these measurements, the number of neutrino generations
is also determined. Moreover, for the first time, the experimental precision is sufficient to probe the
predictions of the SM at the loop level, demonstrating not only that it is a good model at low energies
but that as a quantum field theory it gives an adequate description of experimental observations up
to much higher scales. The significant constraints which the data impose on the size of higher order
electroweak radiative corrections allow the effects of particles not produced at LEP and SLC, most
notably the top quark and the Higgs boson, to be investigated.
The LEP accelerator operated from 1989 to 2000, and until 1995, the running was dedicated to
the Z boson region (“LEP-I” phase). From 1996 to 2000, the centre-of-mass energy was increased to
161 GeV and ultimately to 209 GeV (“LEP-II” phase) allowing the production of pairs of W bosons,
e+ e → W + W . The latter process allowed to test the trilinear gauge couplings, a crucial feature of
non-abelian theories such as the SM.

7.1 The optical theorem and the Z boson propagator


In this section, we will derive the general expression of the propagator of an instable particle, such as
the Z boson, which accounts for width effects. We start with the tree-level expression of the Z boson
propagator in the unitary gauge

µ P i Pµ Pν
!
ν = −ηµν + . (7.1)
Z P 2 − MZ2 MZ2

This expression has clearly a pathology at P 2 ∼ MZ2 and cross-sections computed with this propagator
would diverge. This is obviously unphysical. To remove this pathology we should account for quantum
corrections in the expression of the propagator. Let us write the two point correlation (Green) function

+ Z Z = i (P 2 − M02 + ReTZZ + i ImTZZ )


i(P 2 − M02 ) (7.2)
i TZZ
= i (P 2 − MZ2 + i ImTZZ ),

87
88 CHAPTER 7. THE STANDARD MODEL AT LEP-I

where M0 is the bare (tree-level) mass of the boson and we have defined MZ2 = M02 − ReTZZ . So, being
the propagator the inverse of the Green two-point function, we get
−i
(7.3)
P 2 − MZ2 + i ImTZZ
In order to identify TZZ , we exploit exploit the generalized optical theorem derived in the EFT chapter
XZ
M(i → f ) − M? (f → i) = i dΠn (2π)4 δ 4 (pi − pn )M(i → n)M? (f → n). (7.4)
n

where we remind that the amplitude for a transition between the states i → f is related to the transfer
matrix T by
hf |T |ii = (2π)4 δ 4 (pi − pf )M(i → f ), (7.5)
Since we are interested in the Z boson two-point function, it turns out that |ii = |f i = |Zi and we can
define TZZ = hZ|T |Zi so that eq. (7.4) becomes
1 1X
Z
ImTZZ = (T T † )ZZ = dΠn |TZn |2 (7.6)
2 2 n

where in the last equality we used the completeness relation. Now, we observe that the total decay
width of the Z boson ΓZ is given by
1 X
Z
ΓZ = dΠn (2π)4 δ 4 (pZ − pn )|M(Z → n)|2
2MZ n
2

(7.7)
1 X 1 X
Z Z
= dΠn (2π)4 δ 4 (pZ − pn ) Z ... n = dΠn |TZn |2 .
2MZ n 2MZ n

and therefore ΓZ can be related to ImTZZ as

ImTZZ = MZ ΓZ (7.8)

This result is the so-called optical theorem. We don’t need to know the particle content of the loop in
the correction for the Z propagator because the term we need is easily expressed in terms of measurable
quantities. The final expression for the propagator is then

µ P i Pµ Pν
!
ν = −ηµν + . (7.9)
2 2
P − MZ + iMZ ΓZ MZ2
Z

and in the limit P 2 → MZ2 the divergence has been cured. However since ΓZ  MZ in the P 2 → MZ2
limit the cross section will get a huge enhancement: this is the phenomenon of resonance.

7.2 The process e+ e− → f + f − at LEP


The process under study is e+ e− → f f¯ which proceeds in lowest order via photon and Z boson exchange,
see fig. 7.1. Here the fermion f is a quark, charged lepton or neutrino. All known fermions except the
top quark are light enough to be pair produced in Z decays. The amplitude of e+ e− → f + f − induced
by the Z boson exchange reads
2 !
ig Pµ Pν

µ
M= v(p+ )γ (gVe − e
gA γ5 )u(p− )i −ηµν +
2cW MZ2
(7.10)
1
× 2 u(k− )γ ν (gVf − gA
f
γ5 )v(k+ ).
P − MZ2 + iMZ ΓZ
7.2. THE PROCESS e+ e− → f + f − AT LEP 89

e− (p− ) f + (k+ )

γ/Z

P
e+ (p +) f − (k− )

Figure 7.1: Feynman diagram for the process e+ e− → f f¯ in the SM where f = q, `, ν.

Exploiting the fact that P = p+ + p− = k+ + k− and the massless limit me = mµ = 0 so that

p
/u(p) = 0, u(p)p
/ = 0, (7.11)

p
/v(p) = 0, v(p)p
/ = 0, (7.12)
the Pµ Pν /MZ2 term gives a zero contribution (i.e. proportional to me mµ /MZ2 → 0) and
√ MZ2
M = −i 2GF v(p+ )γ µ (gVe − gA
e
γ5 ) u(p− )u(k− )γµ (gVf − gA
f
γ5 )v(k+ ), (7.13)
−P 2 + MZ2 − iMZ ΓZ
Notice that this amplitude is precisely the one obtained in the Fermi theory apart from an extra factor
MZ2
K(s) = (7.14)
MZ2 − s − iMZ ΓZ

such that if s  MZ2 (i.e. GF s  1) and neglecting width effects (ΓZ  MZ ) we get K(s) → 1. Now
we can easily extend the low-energy results with the substitution

GF =⇒ GF K(s) (7.15)

and therefore the differential cross-sectin of e+ e− → f + f − reads


dσf f̄ 2
παem
 
= Ncf A(1 + cos2 θ) + B cos θ , (7.16)
d cos θ 2s

where the color factor Ncf = 3 for f = q and Ncf = 1 for f = `, ν and

2 2 e f 2 h
(7.17)
ih i
A = 1 − 2 gV gV s Re[GF K(s)] + 4 (gVe )2 + (gA e 2
) (gVf )2 + (gAf 2
) s2 |GF K(s)|2 ,
e e

4 2 2 16
B = − 2 gA s Re[GF K(s)] + 4 gVe gAe f f
gV gA |GF K(s)|2 , (7.18)
e e
and
g2 s(MZ2 − s)
sRe[GF K(s)] = √ 2 (7.19)
4 2cW (MZ − s)2 + MZ2 Γ2Z
2

g4 s2
s2 |GF K(s)|2 = . (7.20)
32c4W (MZ2 − s)2 + MZ2 Γ2Z

At the Z pole ( s ≈ MZ ) the by far dominant contribution to the above cross-section is the one from
the Z exchange as it is resonantly enhanced. However, experimentally, it is not possible to tune exactly

s = MZ so we define a parameter ε which indicates how far we are from the resonant region

s − MZ2
ε= . (7.21)
MZ2
Therefore the differential cross section near the Z pole is

dσf f̄ Ncf g 4 1
h 
(7.22)
ih i
e 2 e 2 f 2 f 2 2 e e f f
= (gV ) +(gA ) (gV ) +(g A ) (1 + cos θ) + 8g g g g
V A V A cos θ
d cos θ 512π c4W Γ2Z + ε2 MZ2
90 CHAPTER 7. THE STANDARD MODEL AT LEP-I

10 5

Cross-section (pb)
Z
10 4
e+e−→hadrons

10 3

+ -
10 2
CESR
DORIS
PEP
WW
PETRA
KEKB
PEP-II
TRISTAN
SLC
10 LEP I LEP II
0 20 40 60 80 100 120 140 160 180 200 220
Centre-of-mass energy (GeV)

Figure 7.2: The hadronic cross-section as a function of centre-of-mass energy. The solid line is the
prediction of the SM, and the points are the experimental measurements. Also indicated are the energy
ranges of various e+ e accelerators. The cross-sections have been corrected for the QED effects.

and, integrating over the angles we find the total cross section

√ G2 (gVe )2 + (gAe )2 (g f )2 + (g f )2
  
σf f̄ ( s ≈ MZ ) ' F MZ4 Ncf V A
(7.23)
6π Γ2Z + ε2 MZ2

Moreover, the forward-backward asymmetry AFB = 3B


8A near the Z pole is given by

√ 3 gVe gA
e gf gf
f
AFB ( s ≈ MZ ) '  V A (7.24)
[(gVe )2 + (gA
e )2 ] (g f )2 + (g f )2

V A

Notice that, since AfFB is a parity violating observable, it must arise from an interference between a
vector and an axial-vector current and this explains why AFB ∼ gV2 gA 2 . Moreover, at the Z pole, Aµ
FB
is strongly suppressed by the leptonic vector-coupling (gVf )2 ≈ 1.6 × 10−3 which is axcidentally small.
This has to be compared, for instance, with (gVb )2 ≈ 3.7 × 10−21
Figure 7.3 illustrates two prominent features of the hadronic cross-section as a function of the
centre-of-mass energy. The first is the 1/s fall-off, due to virtual photon exchange, which leads to the
peak at low energies. The second is the peak at 91 GeV, due to Z exchange and allows LEP and SLC
to function as “Z factories”.
In order to determine the various branching ratios BR(Z → f f ), we proceed as follows. We
canculate the decay rate of Z in a fermion pair2

1
 
Γ(Z → f f ) = √ Ncf GF MZ3 (gVf )2 + (gA
f 2
) (7.25)
6π 2

with Ncf = 1 for leptons and Ncf = 3 for quarks. Then, it’s straightforward to see that we can use
Eq. (7.25) to write the cross section of e+ e− → f f at the Z pole (ε → 0) as

12π Γ(Z → e+ e− ) Γ(Z → f f ) 12π


σffpeak = 2 = 2 BR(Z → e+ e− ) BR(Z → f f ) , (7.26)
MZ ΓZ ΓZ MZ

where we have defined the cross-section at the peak σffpeak ≡ σ(e+ e− → f f )|√s=MZ .
1
Remember that the vector- and axial-vector couplings are defined as gVf = T3L − 2Qf s2W and gA f
= T3L .
2
This result holds in the massless limit which is an excellent approximation for f = e, µ, u, d, s but not for for f = τ, c, b.
7.2. THE PROCESS e+ e− → f + f − AT LEP 91

0.4

σhad [nb]

AFB(µ)
0 AFB from fit ALEPH
σ DELPHI
L3
40 QED corrected
OPAL
average measurements
ALEPH
DELPHI 0.2
L3
OPAL
30
AFB0

ΓZ 0
20
measurements (error bars
increased by factor 10)
-0.2
10 σ from fit
QED corrected

MZ
MZ
-0.4
86 88 90 92 94 88 90 92 94
Ecm [GeV] Ecm [GeV]

Figure 7.3: Average over measurements of the hadronic cross-sections (left) and of the muon forward-
backward asymmetry (right) by the four experiments, as a function of centre-of-mass energy. The
full line represents the results of model-independent fits to the measurements. Correcting for QED
photonic effects yields the dashed curves.

7.2.1 Cross-sections, Partial Widths and Asymmetries


The decays of the Z to charged leptons and to quarks are distinguished experimentally relatively easily,
and in addition some specific quark flavours can be identified. Total cross-sections for a given process
are determined by counting selected events, Nsel , subtracting the expected background, Nbg , and
normalising by the selection efficiency (including acceptance), sel , and the luminosity, L:
Nsel − Nbg
σ= . (7.27)
sel L
The expected background and the selection efficiencies are determined using Monte Carlo event
generators. The cross-sections as a function of centre-of-mass energy around the Z pole yield the Z
mass, mZ , and total width, ΓZ , together with a pole cross-section, see fig. 7.3. The ratios of cross-
sections for different processes give the partial widths and information about the relative strengths
of the Z couplings to different final-state fermions. The Z couples with a mixture of vector and
axial-vector couplings. This results in measurable asymmetries in the angular distributions of the
final-state fermions, the dependence of Z production on the helicities of the colliding electrons and
positrons, and the polarisation of the produced particles. One of the simplest such asymmetries to
measure is the number of forward events, NF , minus the number of backward events, NB , divided by
the total number of produced events:
NF − NB
AFB = . (7.28)
NF + NB
where “forward” means that the produced fermion (as opposed to anti-fermion) is in the hemisphere
defined by the direction of the electron beam (polar scattering angle θ < π/2). The forward-backward
asymmetries are usually derived from fits to the differential distribution of events as a function of the
polar angle of the outgoing fermion with respect to the incoming electron beam. This is the usual type
of asymmetry measured at LEP. Further asymmetries, can be measured if information is available
about the helicities of the incoming or outgoing particles. In particular, the polarised electron beam at
the SLC allowed the measurement of the left-right asymmetry:
NL − NR
ALR = . (7.29)
NL + NR
where, irrespective of the final state, NL is the number of Z bosons produced for left-handed electron
bunches, NR is the corresponding number for right-handed bunches.
92 CHAPTER 7. THE STANDARD MODEL AT LEP-I

The total cross-section can be written in terms of the partial decay widths of the initial and final
states, Γee and Γf f̄ , where Γf f̄ ≡ Γ(Z → f f¯) with f = `, ν, q 6= t

Γ2Z
σfZf̄ = σfpeak . (7.30)
f̄ Γ2Z + ε2 MZ2

where σfpeak

is a measurable quantity given by, see eq. (7.26),

12π Γee Γf f̄
σfpeak = . (7.31)
f̄ MZ2 Γ2Z

The overall hadronic cross-section is parametrised in terms of the hadronic width given by the sum
over all quark final states,
(7.32)
X
Γhad = Γqq̄ .
q6=t

The invisible width, Γinv = Nν Γν ν̄ , where Nν is the number of light neutrino species, is determined
from the measurements of the decay widths to all visible final states and the total width,

ΓZ = Γee + Γµµ + Γτ τ + Γhad + Γinv . (7.33)

Because the measured cross-sections depend on products of the partial widths and also on the total
width, the widths constitute a highly correlated parameter set. In order to reduce correlations among
the fit parameters, an experimentally-motivated set of six parameters is used to describe the total
hadronic and leptonic cross-sections around the Z peak. These are

1. the mass of the Z, mZ ;

2. the Z total width, ΓZ ;

3. the “hadronic pole cross-section”,


12π Γee Γhad
peak
σhad = ; (7.34)
MZ2 Γ2Z

4. the three ratios


Γhad Γhad Γhad
Re0 ≡ , Rµ0 ≡ , Rτ0 ≡ . (7.35)
Γee Γµµ Γτ τ
If lepton universality is assumed, the three ratios reduce to a single parameter R`0 ≡ ΓΓhad
``
where
Γ`` is the partial width of the Z into one massless charged lepton flavour. However, even assuming
lepton universality, Γτ τ differs from Γ`` by about δτ = 0.23% due to the sizable tau mass.

For those hadronic final states where the primary quarks can be identified, we define the ratios:
Γqq̄ Γ
Rq0 ≡ , e.g. Rb0 ≡ bb̄ (7.36)
Γhad Γhad
Experimentally, these ratios have traditionally been treated independently of the above set. The
leading contribution from γ/Z interference is proportional to the product of the vector couplings of

the initial and final states and vanishes at s = MZ , but becomes noticeable at off-peak energies and
therefore affects the measurement of the Z mass. Because an experimental determination of all quark
couplings is not possible, the γ/Z interference term in the hadronic final state is fixed to its predicted
SM value. The six parameters describing the leptonic and total hadronic cross-sections around the
Z peak are determined exclusively from the measurements of the four LEP collaborations, due to
the large event statistics available and the precise determination of the LEP collision energy. In the
measurement of Rb0 and Rc0 , however, the greater purity and significantly higher efficiency which SLD
achieved in identifying heavy quarks offset the statistical advantage of LEP, and yield results with
comparable, and in some cases better, precision.
7.2. THE PROCESS e+ e− → f + f − AT LEP 93

7.2.2 Invisible Width and Number of Neutrinos


Another very important achievement of LEP-I was the determination of the number of neutrinos Nν .
Let us see how. The total decay rate of the Z boson can be written as the sum of the decay rates into
visible particle (i.e. charged particles) and into invisible particles (i.e. neutrinos)

ΓZ = 3Γ``¯ + 3Γdd̄ + 2Γuu + Nν Γνν (7.37)


| {z } | {z }
Γvis Γinv

where Γhad = 3Γdd̄ + 2Γuu and we have assumed lepton universality. The visible and total decay rates
are experimentally measured while Γinv can be predicted in the SM as
  2  2 
GF MZ3 1 1
Γinv = Nν ΓSM (Z → νν) = Nν √ + ≈ Nν × 166 MeV . (7.38)
6π 2 2 2

Therefore, given Γexp


Z ' 2500 MeV and Γvis ' 2000 MeV, Eq. (7.37) allows to determine Nν
exp

Γexp exp
Z − Γvis
Nν = ≈3 (7.39)
ΓSM (Z → νν)

If the Z had no invisible width, all partial widths could be determined without knowledge of the
absolute scale of the cross-sections. Not surprisingly, therefore, the measurement of Γinv is particularly
sensitive to the cross-section scale. To see this point more explicitly, let us define Rinv
0 = Γ /Γ ,
inv ``
divide eq. (7.37) by Γ`` and exploit eqs. (7.34) and (7.35). We obtain
v
u 12πR0
u
Γν ν̄
 
0
Rinv = t peak ` − R`0 − 3 = Nν (7.40)
σhad MZ2 Γ`` SM

where the dependence on the absolute cross-section scale is explicit. The last equality is valid assuming
that the only invisible Z decays are to neutrinos coupling according to SM expectations. Thus, Eq. (7.40)
allows to extract Nν by comparing the measured Rinv 0 with the SM prediction for Γ /Γ . The strong
ν ν̄ ``
dependence of the hadronic peak cross-section on Nν is illustrated in fig. 7.4. The precision ultimately
achieved in these measurements allows tight limits to be placed on the possible contribution of any
invisible Z decays originating from sources other than the three known light neutrino species.
σhad [nb]


30 ALEPH
DELPHI

L3
OPAL
20
average measurements,
error bars increased
by factor 10

10

0
86 88 90 92 94
Ecm [GeV]

Figure 7.4: Measurements of the hadron production cross-section around the Z resonance. The curves
indicate the predicted cross-section for two, three and four massless neutrino species with SM couplings.
94 CHAPTER 7. THE STANDARD MODEL AT LEP-I
Chapter 8

The Standard Model at LEP-II



The electron-positron collider LEP at CERN increased its collision centre-of-mass energy, s from the
Z pole (LEP-I) up to 209 GeV during its second running phase (LEP-II) from 1995 to 2000. The four
LEP experiments ALEPH, DELPHI, L3 and OPAL collected a combined total integrated luminosity of
about 3 fb−1 in the LEP-II centre-of-mass energy range above the Z pole, 130 GeV to 209 GeV. This
large dataset explores the new energy regime accessed by LEP-II with high precision, allowing new tests
of the electroweak SM sector, and searches for new physics effects at higher mass scales. Photon-pair,
fermion pair and four-fermion production processes are analysed and the results are presented in the
form of total and differential cross-sections. Pair-production of W bosons yields measurements of the
mass, total decay width and decay branching fractions of the W boson. Together with other reactions
such as single-W , single-Z, W W γ, Z-pair, Zγ and Zγγ production, the data sample allows stringent
tests of the non-Abelian structure of the electroweak gauge group, by measuring triple and quartic
electroweak gauge boson couplings.

4
L3
10
+ − + – −
e e →e e qq
+ − −
3 e e →qq(γ)
10 + − + −
e e →µ µ (γ)
Cross section (pb)

2
10

10

+ − + −
e e →W W
1 + −
e e →ZZ
+ − + −
e e →W W γ
+ −
-1 e e →γγ
+ −
10 e e →HZ
mH = 115 GeV

80 100 120 140 160 180 200



√s (GeV)

Figure 8.1: Cross-sections of electroweak SM processes. The dots with error bars show the measurements,
while the continuous curves show the theoretical predictions based on the SM.

95
96 CHAPTER 8. THE STANDARD MODEL AT LEP-II

8.1 Standard Model Processes


The various SM processes occurring at high centre-of-mass energies in electron-positron collisions and
their cross-sections are shown as a function of the centre-of-mass energy in Figure 8.1.

8.1.1 Photon-Pair Production


The photon-pair production process, e+ e− → γγ(γ), is dominated by QED interactions. The cor-
responding Feynman diagrams at Born level are shown in Figure 8.2. Higher-order QED effects
play a significant role but the weak interaction is negligible for the present data set. Therefore this
reaction is different from the other processes discussed in this report as it provides a clean test of QED,
independent of other parts of the SM.

e− γ e− γ

e+ γ e+ γ

Figure 8.2: Feynman diagrams for the process e+ e → γγ at the Born level.

8.1.2 Fermion-Pair Production


Pair-production of fermions proceeds mainly via s-channel exchange of a photon or a Z boson as shown
in Figure 8.3. For energies above the Z resonance, QED radiative corrections are very large, up to
several 100% of the Born cross-section. This is caused by hard initial-state radiation
√ of photons, which

lowers the centre-of-mass energy, s, of the hard interaction down to values s0 close to the Z mass,

called radiative return to the Z. In order to probe the hard interaction at the nominal energy scale s,
cuts are applied to remove the radiative return to the Z and only keep the high-Q2 events. Further
cuts remove non-resonant pair corrections arising from four-fermion production not included in the
signal definition.

+ − + −
e f e f

γ Z

− −
e f e f

Figure 8.3: Feynman diagrams for the process e+ e → f f¯ at the Born level. For e+ e final states
additional t-channel diagrams contribute.

8.1.3 W W and ZZ Production


One of the most important processes at LEP-II consists of pair production of on-shell W bosons as
shown in Figure 8.4. These events allow a determination of the mass and total decay width of the
W boson. The non-Abelian nature of the electroweak gauge theory, leading to triple and quartic
gauge-boson vertices such as those appearing in the two s-channel W W production diagrams, is studied
and the gauge couplings are measured. Each W boson decays to a quark-antiquark pair, hadronising
into jets, or to a lepton-neutrino pair, resulting in a four-fermion final state. The W W events are thus
classified into fully hadronic, semileptonic and purely leptonic events. At higher centre-of-mass energies,
four-fermion final states are also produced via Z-pair production, as shown in Figure 8.5. Final-state
corrections arising from the interaction between the two W decay systems, such as colour reconnection
8.1. STANDARD MODEL PROCESSES 97

and Bose-Einstein correlations, may lead to a cross-talk effect. Such an effect potentially spoils the
assignment of decay products to decaying weak bosons in terms of four-momentum, with consequences
in the measurement of the W -boson mass and width in the all-hadronic channel. Radiative corrections
to W-pair production are particularly interesting as they allow the study of quartic-gauge-boson vertices
as shown in Figure 8.6.

+ + + +
e W e W
+ +
e W

νe
− −
γ Z
e W
− − − −
e W e W

Figure 8.4: Feynman diagrams for the process e+ e → W + W at the Born level.

+ + Z
e Z e

e e
− −
e Z e
Z

Figure 8.5: Feynman diagrams for the process e+ e → ZZ at the Born level.

+ + + +
e W e W

γ Z
γ/Z γ/Z

− − − −
e W e W

Figure 8.6: Feynman diagrams for the process e+ e → W W γ and W W Z at the Born level involving
quartic electroweak-gauge-boson vertices.

8.1.4 Four-Fermion Production


Besides the double-resonant W W and ZZ processes, single-resonant boson production channels such as
those shown in Figure 8.7, as well as non-resonant diagrams also contribute to fourfermion production.
Selections are devised to separate the various four-fermion processes, in particular W W , ZZ, single-W
and single-Z production. Single-W production is sensitive to the electromagnetic gauge couplings of
the W boson, as the t-channel photon exchange diagram dominates over the t-channel Z exchange
diagram at LEP-II energies. Bremsstrahlung diagrams with radiation of an on-shell Z boson off an
initial- or final-state fermion leg in Bhabha scattering contribute to single-Z production in the form of
Zee final states.

+ − + + + −
e νe e e e νe
W γ/Z W
+ −
W W γ/Z
γ/Z W W
− − − −
e e e νe e νe

Figure 8.7: Vector-boson fusion diagrams for the single W/Z/γ process at the Born level.
98 CHAPTER 8. THE STANDARD MODEL AT LEP-II

8.2 Tests of the non abelian structure of the Standard Model


The goal of this section is to test the SM as a non-abelian gauge theory. The most direct way to do this
is by studying processes involving self-interactions among gauge bosons, namely trilinear W + W − Z
and quadrilinear vertices. In particular, we will focus on e+ e− → W + W − , see Figure 8.4. We will
see that neglecting any of the three contributions of Figure 8.4 leads to a divergent cross section at
high energy. Instead, summing up all the contributions we get a cancellation essential for the internal
consistency of the theory and to probe the non-abelianity of the electroweak sector.

8.2.1 e+ e− →W + W −
Consider the diagram of Figure 8.4 with the s-channel photon exchange: we expect that the square of
this diagram contains a contribution to the cross section of the form

dσ α2
(e+ e− → W + W − ) ∼ × |ε(k+ ) · ε(k− )|2 , (8.1)
d cos θ s
where ε(k+ ) and ε(k− ) are the polarization vectors of the outgoing vector bosons. For transversely
polarized W bosons, this term is well behaved, but for longitudinally polarized W s it leads to problems.
Let us see why. Writing k for both k+ and k− , in its rest frame, a massive vector boson has momentum
k µ = (MW , 0, 0, 0) and generic polarization vector εµ that is the linear combination of the three
orthogonal unit vectors: (0, 1, 0, 0), (0, 0, 1, 0), (0, 0, 0, 1). Performing a boost in the z direction the
4-momentum becomes k µ = (Ek , 0, 0, k), the first two polarization vector remain the same (transverse
polarization) while the third one (longitudinal polarization) becomes

k Ek
 
εL (k) = , 0, 0, (8.2)
MW MW

so that gauge invariance εµ k µ = 0 , ε2 = −1, and the completeness relation i εµi (k)ενi ? (k) =
P
2 are satisfied. In the high energy limit s  M 2 i.e. k  M
−η µν + k µ k ν /MW W W the longitudinal
polarization is the only one among the other components that is enhanced. It is given by
kµ MW
 
εµL (k) = +O , (8.3)
MW Ek
which implies that
α2 s
σ(e+ e− → W + W − ) ∼ 4 (8.4)
MW
and therefore perturbative unitarity is violated. Fortunately, this bad high-energy behaviour is
automatically solved once we sum the three diagrams of Figure 8.4. In the following, we will collect all
the Feynman rules we need. The couplings of electrons to W , Z and γ are

ieγ µ 1
 
γ Z
e− = − ieγ µ e− = − + s2W
L L cW sW 2

(8.5)

γ Z ieγ µ 2
e− = − ieγ µ e− = s
R R cW sW W

The two couplings with the photon are obviously the same since QED is a vectorial theory. The relative
strengths of the Z couplings are determined by the SU (2) ⊗ U (1) quantum numbers T3L and Qf of
the left-handed and right-handed components of the electron. The three-gauge boson couplings in the
8.2. TESTS OF THE NON ABELIAN STRUCTURE OF THE STANDARD MODEL 99

diagrams arise from the cubic terms in the kinetic EW sector. Since the U (1) field strength is linear
in the gauge fields (it is abelian), these come only from the kinetic term of the SU (2) gauge field, in
particular from the non-abelian part. To identify the specific pieces that we need, we must rewrite this
cubic term in the basis of mass eigenstates. Keeping only cubic terms:
1 a µν,a 1
− Wµν W −→ − (∂µ Wνa − ∂ν Wµa )gεabc W µ,b W ν,c
4 2
= −g(∂µ Wν1 − ∂ν Wµ1 )W µ2 W ν3 + g(∂µ Wν2 − ∂ν Wµ2 )W µ1 W ν3
− g(∂µ Wν3 − ∂ν Wµ3 )W µ1 W ν2
 (8.6)
= ig (∂µ Wν+ − ∂ν Wµ+ )W µ− W ν3 − (∂µ Wν− − ∂ν Wµ− )W µ+ W ν3

1

+ (∂µ Wν3 − ∂ν Wµ3 )(W µ+ W ν− − W µ− W ν+ )
2

Finally, inserting Wµ3 = cW Zµ + sW Aµ and going to the momentum space we obtain the following
Feynman rules for the W W Z and W W γ vertices:

Wµ+ k+ Wν−

= ie η µν (k− − k+ )λ + η νλ (−q − k− )µ + η λµ (q + k+ )ν
 
k−
q


(8.7)
Wµ+ k+ Wν−
ecW  µν
η (k− − k+ )λ + η νλ (−q − k− )µ + η λµ (q + k+ )ν

=i
k− sW
q

Now we can calculate the amplitudes. We split the computation in two parts, considering a left-handed
initial electron and a right-handed one.

Amplitude for e− + +
R eL → W W

Since the coupling of the electron to the W − is purely left-handed the t-channel diagram does not
contribute. Instead, the s-channel contributions with exchange of γ/Z led to

e+
L W − (k− )
−i iesW −i iecW
    
γ/Z
= v R γλ uR (−ie) 2 (ie) + 2 2 ×
q cW q − MZ sW
e−
R W + (k+ )
 
µν λ λν
× η (k− − k+ ) + η (−q − k− ) + η µ λµ
(k+ + q) ε?µ (k+ )ε?ν (k− ).
ν
(8.8)

We have neglected the q λ q µ /MZ2 contribution in the Z boson propagator since in the me = 0 limit,
which we assume, we have
 λ µ
q q me =0

v R γλ uR −→ 0 (8.9)
MZ2
The second line of Eq. (8.8) contains the enhancement for longitudinal polarization of W bosons
mentioned above. If we approximate the longitudinal polarization at high-energy with Eq. (8.3)
100 CHAPTER 8. THE STANDARD MODEL AT LEP-II

dropping terms that do not grow as s → ∞, we find that

1 1 s
 
iM(e− +
R eL → WL+ WL− ) 2
= [v R γλ uR ](−ie ) 2 − 2 2 2 (k+ − k− )
λ
q q − MZ 2MW
! (8.10)
2
sMZ 1 λ ie2
= [v R γλ uR ] (k+ − k− )
s 2c2W

Notice that the gauge bosons trilinear interactions provide the necessary terms to cancel the bad
high-energy behavior stemming from the longitudinal polarization of W bosons. Therefore, it is the
non-abelian structure of the theory itself that guarantees the finiteness of the result.

Amplitude for e− + +
L eR → W W

For the amplitude with an initial left-handed electron, the computation is somewhat more involved.
Now also the tν channel diagram contributes, and since this diagram has a different kinematic structure,
it will be less clear how the three diagrams combine together. In what follows, we will demonstrate
the cancellation of the unitarity-violating enhanced terms, and we will indicate how the terms one
order smaller in MW 2 /s assemble into the correct structure. However, we will not account rigorously for

all of these smaller terms. It is actually straightforward, but lengthy, to show that the terms we will
omit do not contribute to the high-energy limit of the cross section. Let us start with the s-channel
contributions:

e+
R W − (k− )  
γZ
 
−i ie − 12 + s2W  −i iecW
 
= v L γλ uL (−ie) 2 (ie) + 2 2 ×
q sW cW q − MZ sW
e−
L W + (k+ )
 
µν λ
× η (k− − k+ ) + η (−q − k− ) + η λν µ λµ
(k+ + q) ε?µ (k+ )ε?ν (k− ) .
ν
(8.11)

This term differs from Eq. (8.8) only in the coupling of the electron to the virtual Z boson. In the
high-energy limit, the longitudinal component WL is dominant and we can simplify the expression as

MZ2 1 s
  
iMs (e− + + − 2
L eR → WL WL ) = ie v L γλ uL 2 − 2 2
λ
2 (k+ − k− ) . (8.12)
s(s − MZ ) 2sW (s − MZ ) 2MW

The second term in brackets is a potentially divergent contribution which must be canceled: this occurs
when we consider the t-channel. If ` is the momentum of the incoming electron,

e−
L (`) W − (k− )
2
ig i(/̀ − k/− ) ν

νe = √ vLγ µ γ uL (`)ε?µ (k+ )ε?ν (k− ), (8.13)
2 (` − k− )2

e+
R W + (k+ )

that in the high-energy limit reads

−ig 2 k/ /̀ − k/− k/− ig 2 k/+


iMt (e− + + −
L eR → WL WL ) = vL + uL (`) = v L 2 uL (`)
2 MW (` − k− )2 MW 2 MW
1 1
= ie2 2 λ
2 [v L γλ uL (`)] (k+ − k− ) . (8.14)
2sW 2MW

where we have exploited the equality

(8.15)
m =0
e
(/̀ − k/− )(k/− + /̀ − /̀)uL (`) = −(/̀ − k/− )2 uL (`) .
8.3. THE GOLDSTONE BOSON EQUIVALENCE THEOREM 101

As we can see, Eq. (8.14) cancels the bad high-energy behavior of Eq. (8.12). Now we can combine
this result with the ones of the s-channels. So far we have exploited εµL ≈ k µ /MW : this approximation
is good enough to see the cancelation of bad behaving terms at high-energy. However, to get the
subleading contributions we have to refine the expansion; doing that, the s channel result remains
unchanged while the t-channel gets an extra term since the O(MW 2 /s) term provides an overall factor

(1 + 2MW /s). Therefore, summing the two amplitudes we get M = Ms + Mt where


2

1 1
 
iM(e− +
L eR → WL+ WL− ) 2
= ie [v L γλ uL ] (k+ − k− ) λ
. (8.16)
s 4cW s2W
2

Therefore at high energies the total amplitude is finite and so is the cross section. This is an incredible
proof of the fact that the SM must be a non-abelian gauge theory. As we can see in Fig. (8.8) the
measurements obtained at LEP (phase 2) confirmed the SM prediction: cross section evaluated with

only the neutrino exchange diagram contribution or without the ZW W vertex diverge for s → ∞.

30
σWW (pb)

σWW (pb)
20
LEP LEP
YFSWW and RacoonWW

20

18
10
17
10
16
YFSWW/RacoonWW
no ZWW vertex (Gentle)
only νe exchange (Gentle)

190 195 200 205


0 0
160 180 200 160 180 200
√s (GeV) √s (GeV)

Figure 8.8: Measurements of the W-pair production cross-section, compared to the predictions of
RacoonWW and YFSWW. The shaded area represents the uncertainty on the theoretical predictions,

estimated as ±2% for s < 170 GeV and ranging from 0.7 to 0.4% above 170 GeV. The uncertainty on
the W mass MW = 80.35 GeV is expected to give a significant contribution only at threshold energies.

8.3 The Goldstone boson equivalence theorem


In this section, we will study the Goldstone boson equivalence theorem (GBET) and some important
applications of it. We will see that, in the high-energy limit, the unphysical Goldstone boson that is
eaten by a massive gauge bosons still controls the amplitude for emission or absorption of the gauge
boson in its longitudinal polarization state. Studying the Higgs mechanism for vector boson mass
generation, we notice that it involves a certain conservation of degrees of freedom. A massless gauge
boson, which has two transverse polarization states, combines with a scalar Goldstone boson to produce
a massive vector particle, which has three polarization states. When the massive vector particle is at
rest, its three polarization states are completely equivalent, but when it is moving relativistically, there
is a clear distinction between the transverse and longitudinal polarization direction: as we pointed out
in the previous section, the longitudinal polarization component dominates among the others. This
suggests that a rapidly moving, longitudinally polarized massive gauge boson might betray its origin
102 CHAPTER 8. THE STANDARD MODEL AT LEP-II

as a Goldstone boson. The idea is expressed by the GBET, diagrammatically:

WL+ φ+

!!
2
MW
= × 1+O (8.17)
E2

The proof the GBET is based on the Ward identities of the spontaneously broken gauge theory.
However, it is possible to understand the idea behind it by examining concrete examples.

8.3.1 e+ e− → W + W −
We want to exploit the GBET to obtain the results of the previous section. In order to do that we
need the Feynman rules for the couplings φφγ and φφZ.

p µ p µ ie 1
 
γ
φ+ = ie(p + p ) 0 µ
φ+
Z = − s2W (p + p0 )µ (8.18)
cW sW 2
p p

To get these rules we have to consider the scalar QED lagrangian:


1
Lscalar † µ
QED = (Dµ φ) (D φ) − Fµν F
µν
− m2 (φ† φ), (8.19)
4
with Dµ = ∂µ + ieAµ . The Feynman rules are straightforwardly obtained from the terms like
(∂µ φ + ieAµ φ)† (∂ µ φ + ieAµ φ). The rule for the coupling with the Z boson is similar, but the total
covariant derivative in the mass basis (in terms of physical gauge bosons) is more complicated:
ig ig
Dµ = ∂µ − √ (Wµ+ T + + Wµ− T − ) − Zµ (T3 − s2W Q) − ieAµ Q, (8.20)
2 cW

with
σ1 ± iσ2
Q = T3 + Y, T± = (8.21)
2
and of course g = e/sW . Therefore the Feynman rules are obtained taking into account the fact that
φ+ is the up component of the SU (2) doublet with T3 = 1/2 and Q = +1.

ig 1 ig µ 1
      
(Dµ φ)† (Dµ φ) −→ ∂µ φ† + Zµ − s2W φ† ∂µφ − Z − s2W φ
cW 2 cW 2
(8.22)
g 1
 
∂µ →−ipµ − µ †
−→ − s2W (p+
µ − pµ )Z φ φ
cW 2

with p+
µ and pµ both incoming while in the diagram above p is outgoing (φ = φ and φ = φ ). Now
− 0 + − †

we exploit the GBET and we write

e− W − (k− ) e− φ− (k− )
e+ W − (k −) e+ φ+ (k+ )
2
sMZ
γZ γZ
+ νe ≈ + νe (8.23)

e− W + (k+ ) e− φ− (k− )
e+ W + (k +) e+ φ+ (k+ )
| {z }
=0
8.3. THE GOLDSTONE BOSON EQUIVALENCE THEOREM 103

The second diagram in the RHS vanishes in the massless limit me = 0 since the φeν vertex is
proportional to the Yukawa coupling. Remembering the rules in Eq. (8.5), we compute the remaining
diagram decomposing the amplitude in helicity states

e+ φ+ (k+ )
γZ
= iM(e− + + − − + + −
L eR → φ φ ) + iM(eR eL → φ φ )

e− φ− (k− )
" 2 ! !#
1−2s2W 1−2s2W

1 1 1 1
2
= ie v L γµ uL + +v R γµ uR − (k+ −k− )µ . (8.24)
q2 4s2W c2W q 2 −MZ2 q2 2c2W q 2 −MZ2

It is easy to verify that in the s  MZ2 limit (q 2 = s) the two contributions are the same already
obtained in the previous section:
i 02 (k+ − k− )µ
iM(e− e
L R
+
→ φ +
φ − ) = − (g + g 2
) [v L γ µ uL ] (8.25)
4 s
i (k − k )µ
(8.26)
2 + −
iM(e− + + 0
R eL → φ φ− ) = − g [v R γµ uR ]
2 s
Notice that in the high energy limit, where the SU (2) ⊗ U (1) gauge group is unbroken we recover the
couplings with the Bµ and Wµ3 gauge bosons in the left-handed electron amplitude and only Bµ in the
right-handed electron one. The fact that we obtained the same result of the previous section seems a
kind of magic. We clearly see that the GBET does not tell us that every amplitude is invariant under
the replacement of every longitudinal polarized vector boson in each diagram with the scalar, but only
that the total amplitude is the same under this substitution. In fact we can see that

e−
R W − (k− ) e−
R φ− (k− )
e+
L W − (k− ) e+
L φ+ (k +)
γZ γZ
+ νe = + νe (8.27)

e−
R W + (k+ ) e−
R φ− (k− )
e+
L W + (k+ ) e+
L φ+ (k+ )
| {z } | {z }
=0 no e−
R coupling
=0 no Yukawa coupling

e−
L W − (k− ) e−
L φ− (k− )
e+
R W − (k− ) e+
R φ+ (k +)
γZ γZ
+ νe = + νe (8.28)

e−
L W + (k+ ) e−
L φ− (k− )
e+
R W + (k+ ) e+
R φ+ (k+ )
| {z }
=0 no Yukawa coupling

Therefore what is interesting is the fact that in Eq. (8.28) the two channels precisely sum to the s
channel with Goldstone bosons instead of the W ’s. Using the GBET we compute two diagrams only
and we get easily the result in the high-energy limit.

8.3.2 t → W + b
The top decay is another interesting case where we see the GBET at work. Let us first try to guess the
magnitude of the top quark width in NDA. The squared matrix element will contain a factor of g 2
times the largest mass of the process which is the top mass. Therefore,
g2
Γ(t → W + b) ∼ mt |Vtb |2 , (8.29)

104 CHAPTER 8. THE STANDARD MODEL AT LEP-II

where 1/8π is the two body phase-space factor. The correct expression, however, turns out to be
enhanced by a factor of m2t /MW
2 . The amplitude of the process is

W + (k)
ig 1 − γ5
 
iM(t → W + b) = t(p) = √ u(q)γ µ u(p)ε?µ (k)Vtb . (8.30)
2 2

b(q)

The unpolarized square amplitude reads

1 X g2 X
|M|2 = = |Vtb |2 (q µ pν + q ν pµ − η µν q · p) ε?µ (k)εν (k)
2 spins 2
(8.31)
pol
!
g2 2(k · p)(k · q)
= |Vtb |2 q · p + 2
2 MW

In the limit mb = 0

(k + q)2 = p2 = m2t = MW
2
+ 2k · q =⇒ 2k · q = m2t − MW
2

(p − q)2 = k 2 = MW
2
= m2t − 2q · p =⇒ 2p · q = m2t − MW
2
(8.32)
(p − k)2 = q 2 = 0 = m2t + MW
2
− 2k · p =⇒ 2k · p = m2t + MW
2

therefore ! !
g 2 m4t M2 M2
|M| = 2
2 1− W 1+2 W |Vtb |2 . (8.33)
4 MW m2t m2t

After multiplying by phase space, we find


!2 !
g 2 m3t
+ M2 M2
Γ(t → W b) = 2 1− W 1+2 W |Vtb |2
64π MW m2t m2t
! (8.34)
m2t MW
2
g2 m2t
−→ |Vtb |2 mt 2 ,
64π MW

that is larger than our estimate by the factor m2t /MW


2 . The origin of this enhancement can be traced

back by means of the GBET. In the gauge theory of weak interactions, the top quark obtains its mass
from its coupling to the Higgs sector. The top quark can be heavy only if the Yukawa coupling yt is
large. But then the amplitude for the top quark decaying into a Goldstone boson will be enhanced by

yt2 m2t
= . (8.35)
g2 2m2W

Let us compute now the width of t → W + b exploiting the GBET. The Yukawa coupling tbφ reads

Ltb −
Y = yt Vtb bL φ tR + h.c, (8.36)

with
1
φ± = √ (φ1 ± iφ2 ), (8.37)
2
while the Higgs field is
!
1 φ1 + iφ2
ϕ= √ . (8.38)
2 φ3 + iφ4
8.3. THE GOLDSTONE BOSON EQUIVALENCE THEOREM 105

Therefore the amplitude of the decay t → W + b reads

φ+ (k)
1 + γ5
 
+
iM(t → φ b) = t(p) = iyt Vtb u(q) u(p) . (8.39)
2

b(q)

We easily find
|M|2 = yt2 |Vtb |2 (q · p) , (8.40)
therefore
yt2 g2 m2t
 
Γ(t → φ+ b) =|Vtb |2 mt = |Vtb |2 mt 2 , (8.41)
32π 64π MW

which is precisely Eq. (8.34) (we exploited MW 2 = g 2 v 2 /4 and m = y v/ 2). Hence, in the m  M
t t t W
high-energy limit, the top decays into a longitudinally polarized vector boson which can be approximated
with a Goldstone boson, in agreement with the GBET.

8.3.3 W + W − → W + W − and the no-lose-Higgs theorem


The W W scattering is a fundamental process of the EW theory as it was instrumental to prove
(indirectly) the existence of the Higgs boson. In particular, imposing that the EW theory is unitary
we were able to find an upper bound on its mass of ∼ 1 TeV. This result is often called no lose
Higgs theorem and it was one of the reason why LHC was built. Let us see, firstly qualitatively and
then quantitatively exploiting the GBET, why this is the case. Let us consider the quadrilinear W
interaction arising from the product of igεW W terms in the field strength tensor. Considering only
longitudinally polarized W dominating in the high-energy limit and exploiting ε(p)ε(q) ∼ E 2 /2MW2

WL (`ν ) WL (pρ ) !
E4
= εµl (k)ενl (`)g 2 (2ηµρ ηνσ − ηµν ηρσ − ηµσ ηνρ )ερl (p)εσl (q) ∼ g 2
4 + ...
4MW
WL (k µ ) WL (q σ )
(8.42)
which exhibits a strong divergence. Considering all the diagrams with physical gauge bosons we get

WL (k µ ) WL (pρ )
WL (`ν ) WL (pρ ) WL (`ν ) WL (pρ ) 2
E

γZ
MG = + + γZ ∼ g 2
+ . . . (8.43)
MW
WL (k µ ) WL (q σ ) WL (k µ ) WL (q σ )
WL (q σ ) WL (`ν )

Despite the fact that in each diagram we have 4 longitudinal polarization vectors ε(k)µ ∼ k µ /MW ,
summing the three diagrams, the non abelian structure of the SM protects us from the strong divergence
we had considering only the quadrilinear diagram. However the amplitude is still divergent. In order
to get rid of a divergent behavior of the amplitude two channels involving the Higgs boson exchange
must be introduced:

WL (k µ ) WL (pρ )
WL (`ν ) WL (pρ )
2
E

Mh = h + h ∼ − g2 + ... (8.44)
MW
WL (k µ ) WL (q σ )
WL (q σ ) WL (`ν )
106 CHAPTER 8. THE STANDARD MODEL AT LEP-II

This is quite impressive: the high energy diverging terms perfectly cancel so we have to consider lower
order terms to write the total amplitude and it turns out that
2
Mh

Mtot = MG + Mh ≈ g 2
, (8.45)
2MW

which is finite at high energy.


Let us prove this result exploiting the GBET. To do this we need the Feynman rules for the resulting
diagrams. Neglecting all the couplings of the Higgs fields with the gauge fields in the Higgs sector i.e.
the couplings φ+ φ− γ, Z, the scalar sector of the SM Lagrangian is given by

LSM
s = (∂µ ϕ)† (∂ µ ϕ) − µ2 ϕ† ϕ − λ(ϕ† ϕ)2 (8.46)

with µ2 < 0 in order to realize a SSB and λ > 0 to have a scalar potential bounded from below and
!
1 φ1 + iφ2 µ2
ϕ= √ , φ = (φ1 , φ2 , φ4 ), σ = φ3 = v + h, v2 = − , (8.47)
2 φ3 + iφ4 λ

where
1
φ± = √ (φ1 ± iφ2 ) ←→ longitudinal W
2 (8.48)
Z = φ4 ←→ longitudinal Z.
With these definitions the Lagrangian reads

1 µ2 2 λ
 
LSM
s = (∂µ φ)2 + (∂µ σ)2 − (φ + σ 2 ) − (φ2 + σ 2 )2
2 2 4
2 2
(8.49)
1 Mh 2 Mh M2
 
= (∂µ φ)2 + (∂µ σ)2 − h − h(φ2 + h2 ) − h2 (φ2 + h2 )2
2 2 2v 8v

with
Mh2 =2λv 2 physical Higgs boson mass
(8.50)
m2φ =0 Goldstone bosons
Notice that in our notation
φ2 = 2φ+ φ− + ZZ (8.51)
Now we can read from the Lagrangian the couplings that we need to compute the φ+ φ− → φ+ φ−
process, that in the high energy limit, for the GBET is equivalent to the W W scattering

φ− φ+ !2
h iM 2 i
iMa = = − h
v s − Mh2
φ+ φ−

φ− φ−

!2
h iM 2 i
iMb = = − h
v t − Mh2
φ+ φ+

φ+ φ−
Mh2
iMc = = −i × 4 × 2! × 2! (8.52)
8v 2
φ− φ+
8.3. THE GOLDSTONE BOSON EQUIVALENCE THEOREM 107

The total amplitude is given by


M2 s t
 
Mhtot = − 2h + . (8.53)
v s − Mh2 t − Mh2

Now we can also consider the couplings of Goldstones φ+ φ− with gauge bosons γ and Z, exploiting
the Feynman rules we found at the beginning of the section:

φ− (p− ) φ+ (k+ )
1 1
 
γZ 2 2
iMd = = i(ie) + βW (p+ − p− ) · (k+ − k− )
s s − MZ2
φ+ (p+ ) φ− (k− )

φ− (p− ) φ− (k− )

γ/Z 
1 1

iMe = = i(ie)2 2
+ βW (p+ + k+ ) · (−p− − k− ) (8.54)
t t − MZ2
φ+ (k+ ) φ+ (p+ )

having called
2
1 1

2
βW = − s2W (8.55)
c2W s2W 2
Substituting the Mandelstam variables s = 2p+ · p− = 2k+ k− , t = −2p− · k− = −2p+ · k+ and
u = −2p+ · k− = −2p− · k+ we get the final total amplitude for the W W → W W scattering in the
GBET approach
M2 s t
 
Mtot = − 2h +
v s − Mh2 t − Mh2
(8.56)
1 1 1 1
    
− e2 2
+ βW (u − t) + + β 2
W (s − u)
s s − MZ2 t t − MZ2
In order to get the bound on the Higgs mass we recover the expression of the coefficients of the partial
wave decomposition of a Feyman amplitude M(s, t), discussed in the EFT chapter
1
Z 1
aj (s) = d cos θPj (cos θ)M(s, t). (8.57)
32π −1

We might consider only the dominant coefficient of the expansion a0 , even if there is a non trivial
dependence of the amplitude on t = −s(1 − cos θ)/2 and u = −s(1 + cos θ)/2 (in the center-of-mass
frame). We will compute the integral over the Mhtot only since for the diagram involving the gauge
bosons γ, Z the s-channel gives a zero contribution to a0 while the t-channel leads to a divergence
(as usual for t-channels with massless particles in the final state) which however has nothing to do
with the high energy behavior, but only to the direction dependence of the amplitude. Performing the
integration we obtain
2Mh2 Mh2 M2 s sMh2 GF Mh2
  
a0 (s) = − 2
2+ 2 − h log 1 + 2 −→ − √ . (8.58)
32πv s − Mh s Mh 4π 2
Since we have seen |Re(a0 )| ≤ 1/2 we obtain an upper bound on the mass of the Higgs boson
s √
2π 2
Mh ≤ ≈ 1 TeV. (8.59)
GF
This result obtained from first principles (unitarity of S matrix and GBET) shows that the Higgs
boson must exist in order to ensure a safe high-energy behavior of the W W cross-section. More in
general, the existence of the Higgs boson unitarizes the SM EW sector.
108 CHAPTER 8. THE STANDARD MODEL AT LEP-II
Chapter 9

Renormalization

In the following chapters we will address some very important processes in which quantum effect are
relevant to compare the SM theoretical predictions with the experimental measurements. In particular,
prominent examples which will be analysed in detail are

1. ElectroWeak Precision Tests (EWPT). So far we have analyzed processes at the Z pole
accounting only for tree-level effects. Since the experimental resolutions at LEP are at the
O(10−3 ) level, it is worth to consider also loop effects. As an example, we will consider the
process Z → bb, which receives loop corrections from the exchange of a W and top-quarks:

b
t
Z W
t

b (9.1)

2. Flavor Changing Neutral Currents (FCNC). As we have seen, in the experimental deter-
mination of the VCKM matrix elements we encounter the ∆F = 2 transitions, described by the
following loop diagrams

K 0 , B 0s W W K 0 , Bs0

3. Anomalies. With this term we mean symmetries that are exact at tree level but are broken
by quantum effects. A famous example of anomalous process is the decay of the neutral pion
π 0 → γγ induced by the triangle diagram:

π0
γ

Other examples of anomalies at work are the baryon number and lepton number violation at
loop level, that has a tremendous importance because of the Sakharov condition for baryogenesis.

4. Higgs physics. The search for the Higgs boson last until its discovery in 2012 at LHC. The
measured mass is Mh ' 125 GeV. Loop processes are of high importance in the Higgs physics.
For instance, the main production mechanism is the gluon-gluon fusion:

109
110 CHAPTER 9. RENORMALIZATION

g
t
h (9.2)
g

The main contribution in the fermion loops comes from the top quark again because of its order
1 Yukawa coupling. The Higgs boson has been detected through its decays. One important
decay mode, whose branching ratio is actually ∼60 %, is the one in a bottom quark-antiquark
pair. It is essential to have a precise theoretical prediction for this decay mode in order to get a
proper detection. If one computes the decay rate at tree level, he will find out a disagreement
with the experiments of a factor of two: this is due to the fact that, being involved the bottom
quark, large QCD loop effect (that are proportional to αs which is quite big at those energies)
are present. The complete SM prediction has to take into account at least gluon bremsstrahlung
(radiative correction) and loop contributions:

b b g b b

h + h g + h + h

g
b b b b
(9.3)
b b b
g g
h + h + h
g g
b b b

We know that BR(h → bb) ∝ m2b . Once you include these corrections the mb pole mass (mb ' 4.6
GeV at E ∼ mb ) gets evaluated at the Higgs scale mb (Mh ) and we recover the factor of 2 we
missed from the tree level computation.

Therefore is essential to consider loop contributions. Let us start from basic concepts of renormalization
and loop calculations in QED which will be our playground for further developements.

9.1 Dimensional regularization and subtraction schemes

Let us start considering the degree of superficial (UV) divergence of a diagram1 . If fi is the number of
internal fermions and bi the number of internal bosons we have
(
≥0 UV divergent
K = |{z}
4 − fi − 2bi (9.4)
|{z} |{z} < 0 UV finite
∼ d4 k ∼ k1 ∼ 1
k2

where k is the momentum running in each line. The main source of UV divergence (i.e. for k → ∞) is
the fact that loops involve integration on the virtual momentum inside them, hence, in four dimensions,
four powers of k; fermion and boson propagators instead ease this divergence. The fact that K < 0
does not protect us from IR divergences (i.e. for k → 0): these can be healed considering separately
bremsstrahlung and radiative corrections. It can be shown that in QED the degree of superficial
divergence depends only on the number of external legs of a diagram. Moreover it can be shown that

1
This discussion is valid only at one-loop level. The extension to higher orders is unnecessary for our purposes.
9.1. DIMENSIONAL REGULARIZATION AND SUBTRACTION SCHEMES 111

the only three UV divergent diagrams in QED are the following

k
1
Z
vacuum polarization ∼ d4 k ∼ k2
k/k/

k
(9.5)
1
Z
4
electron self energy ∼ d k 2 ∼k
k k/
1
Z
vertex correction k ∼ d4 k ∼ log k
k 2 k/k/

Actually there are other two potentially divergent diagrams –the fermion triangle loop with three
external photons and the fermion box loop with four external photons– but the amplitude for those
processes turns out to be zero because of Furry’s theorem and Ward identities, respectively.

Renormalization program: two steps


The renormalization procedure is achieved in two steps

1. Regularization: divergences are identified and isolated by means of a regulator. Examples


of regularization schemes are the cut-off regularization which introduces an energy scales Λ
eventually going to infinite, or the dimensional regularization in which the dimension of the
space is lowered by a number ε eventually going to zero. It is important to stress that physical
observables must be independent on regulators.

2. Renormalization: UV divergences are removed or (as in QFT gergo) subtracted by means of a


subtraction scheme as the on-shell scheme or the minimal subtraction scheme. At the end of the
subtraction procedure we get rid of regulators.

We will focus on dimensional regularization and the (modified) minimal subtraction scheme, called
MS scheme. When performing renormalization we can either subtract divergent terms only or we can
also subtract some finite terms. This freedom is called the subtraction scheme dependence. However
there is no physical dependence on this procedure, therefore when computing observables, taking into
account all terms, any subtraction scheme has to lead to the same results.
Now let us study the dimensional regularization procedure. The underlying idea is to realize that
when computing amplitude involving diagrams containing loops, they may UV diverge because of the
fact that d4 k has energy dimension four. Therefore, if we lower the dimesion of our space, the degree
of divergence of the diagram will be reduced. We will work in a d = 4 − ε dimension space: we will
perform calculations in this unphysical space and at the end we will set ε → 0, recovering our four
dimensions space. Namely before doing that we will perform renormalization, subtracting divergences
that will show up as terms ∼ 1/ε → ∞.
ε→0

Let us apply these ideas to the QED case. The bare (i.e not renormalized) Lagrangian reads
1 0 0,µν
L0QED = − Fµν F + iψ 0 γ µ (∂µ − ie0 A0µ )ψ 0 − m0e ψ 0 ψ 0 . (9.6)
4
In d dimensions the QED action reads
Z
SQED = dd xLQED . (9.7)

We know that the action has mass dimension zero [S] = 0 therefore, from this condition we can fix the
dimension of the bare quantities ψ 0 , e0 , m0e , A0µ . In fact, being [dd x] = −d, we obtain
0
[Fµν F 0,µν ] = 2[∂µ ] + 2[A0µ ] = d =⇒ [A0µ ] = 1 − ε/2 (9.8)
112 CHAPTER 9. RENORMALIZATION

3
[ψ 0 ∂/ ψ 0 ] = 1 + 2[ψ 0 ] = d =⇒ [ψ 0 ] = − ε/2 (9.9)
2
[m0e ψ 0 ψ 0 ] = [m0e ] + 2[ψ 0 ] = d =⇒ [m0e ] = 1 (9.10)
The fact that the bare mass has mass dimension 1 is non trivial: couplings can have in general mass
dimension dependent on ε. In fact
[e0 A0µ ψ 0 ψ 0 ] = d =⇒ [e0 ] = ε/2. (9.11)
We find out that the bare electric charge is not dimensionless in a d dimension space. This result has
very important consequences, as we shall see. Now we introduce the renormalized fields and parameters
1 1 1 0
Aµ = √ A0µ ; ψ = p ψ0; me = m . (9.12)
ZA Zψ Zm e
For the electric charge we require the renormalized electric charge to be dimensionless: therefore we
need to introduce a parameter µ, called renormalization scale with mass dimension [µ] = 1 (in this
sense it is an energy scale) so that
1 −ε/2 0
e= µ e , (9.13)
Ze
and [e] = 0. Substituting bare quantities in the bare QED Lagrangian we obtain
1
L0QED = − ZA Fµν F µν + iZψ ψγ µ (∂µ − iµε/2 Ze ZA eAµ )ψ − Zm Zψ me ψψ. (9.14)
p
4
Now we rewrite this Lagrangian in the same form of the original bare one (apart from the µ that
cannot be readsorbed), plus other terms needed to do so
1
L0QED = − F µν Fµν + iψγ µ (∂µ − iµε/2 eAµ )ψ − mψψ
4
1
− (ZA − 1)Fµν F µν + i(Zψ − 1)ψ∂/ ψ + i(Zψ Ze ZA − 1)ψ(−iµε/2 eA
p
/ )ψ − me (Zm Zψ − 1)ψψ
4
=LRen
QED + LCT
(9.15)
As we can see the terms in the first line Eq. (9.15) make a ”renormalized QED Lagrangian” that is
written in the same form as the bare one, but with renormalized quantities; however the price to pay
is to add to each term a so called counterterm. Now the sense of the renormalization procedure is
clear: the subtraction of divergences is performed adding counterterms depending on renormalization
parameters Zi that have to be fixed in order to remove precisely the divergences in the three diagrams
in Eq. (9.5). Going to momentum space, these counterterms can be considered interaction terms with
the following Feynman rules.
 
= i (ZA − 1)(pµ pν − p2 ηµν )
 
= i (Zψ − 1)p
/ − me (Zψ Zm − 1)
(9.16)

= ie(Zψ Ze ZA − 1)µε/2 γ µ
p

In order for the subtraction to be efficient the divergences must have the same form of the counterterms:
this is precisely what happens, as we shall see. Now we have to set our convention to fix the
renormalization parameters Zi . We decide to exploit the (modified) minimal subtraction scheme: we
will subtract only divergences and no finite terms2 . In this way

Z` (e)
(9.17)
X
Z =1+ ,
`=1
ε`
2
Actually, as we shall see, we will subtract few finite terms, that’s why this scheme is called modified minimal
subtraction scheme.
9.1. DIMENSIONAL REGULARIZATION AND SUBTRACTION SCHEMES 113

with Z` (e) that does not depend on ε: all the dependence on the dimension is in the 1/ε` part. We
will stop at ` = 1 loop. With this scheme we will fix Z to subtract terms ∼ 1/ε only.
Now let us write the complete integral for the calculation of the vacuum polarization diagram:

p+k
k k Z
dd p Tr γµ i(k/ + p

/ + me )γν i(p

/ + me )
= (−1) (ieµ ε/2 2
) (9.18)
(2π)d
 
| {z } (k + p) − me p − m2e ]
2 2 2
fermion loop
p

We need some mathematical tools to perform this kind of integrals.

Clifford algebra in d dimensions. The generalized metric takes the form η µν = diag(1, −1, . . . , −1).
Moreover γ 0 , γ 1 , . . . , γ d−1 are f (d) × f (d) matrices where f (d) −→ 4 (e.g. f (d) = d) which satisfy
d→4

{γ µ , γ ν } = 2η µν . Contractions of γ matrices give


γµ γ µ = d1
γµ γ α γ µ = − (d − 2)γ α (9.19)
α β µ α β αβ
γµ γ γ γ = (d − 4)γ γ + 4η
Traces of γ matrices read
Tr[1] = f (d)
α1 α2n+1
Tr[γ ...γ ] =0
(9.20)
Tr[γ γ ] = f (d)η αβ
α β

Tr[γ α γ β γ γ γ δ ] = f (d)(η αβ η γδ − η αγ η βδ + η αδ η βγ )

Feynman integrals The most general form of a one-loop integral is a rank r tensor of the form
dd k k µ1 . . . k µr
Z
Tnµ1 ...µr = , (9.21)
(2π)d D1 . . . Dn
where
Di = (k + ri )2 − m2i + i (9.22)
and the momenta ri are related with the external momenta pj (all taken to be incoming) through
i n
(9.23)
X X
ri = pj , with rn = pj = 0.
j=1 j=1

In the denominator of Eq. (9.21) products of the denominators of the propagators of particles in the
loop appear. It is convenient to combine these products in just one common denominator, introducing
the so-called Feynman parameters.

Feynman parameters We start from the equality


1
Z 1 1
= dx . (9.24)
ab 0 [ax + b(1 − x)]2
This can be generalized to the case of a and b of arbitrary power. For example if we want to write
1/(a2 b) we exploit the equality
1 d 1
 
=− (9.25)
a2 b da ab
and therefore we take the derivative of the integrated function. The same holds for b, so, in general
1 Γ(m + n) 1 xm−1 (1 − x)n−1
Z
= dx (9.26)
am bn Γ(m)Γ(n) 0 [ax + b(1 − x)]n+m
114 CHAPTER 9. RENORMALIZATION

Euler Γ function In the previous expression the Euler Γ functions appear just like combinatorial
factor, due to the choices of n and m. This is because of the properties of the Γ function:
Z ∞
Γ(z) = dt tz−1 e−t , (9.27)
0

Γ(z + 1) = zΓ(z), Γ(n + 1) = n! ifn ∈ N. (9.28)


We will often encounter Γ functions depending on the dimension regulator ε. We can write a series
expansion
π 2 ε2
 
Γ(1 + ε) = 1 − γε + γ 2 + + O(ε3 ) (9.29)
6 2!
where γ = 0.577 is the Euler constant. Our results will often depend on Γ ε
which is a divergent

2
quantity since the Γ function has a pole in 0: in fact

ε 2 ε 2
   
(9.30)
ε→0
Γ = Γ 1+ −→ − γ + O(ε).
2 ε 2 ε

Scalar integrals in d dimensions In general, in a dimensional regularization procedure we may


encounter scalar integrals of the form, called master integrals
r
dd k (k 2 )
Z
Ir,m = . (9.31)
(2π)d [k 2 − C + i]m

All the tensor integrals can be obtained from scalar integral of this form, for instance

dd k kµ
Z
= 0,
(2π)d [k 2 − C + i]m
(9.32)
dd k kµ kν η µν dd k k2
Z Z
= = I1,m ,
(2π)d [k 2 − C + i]m d (2π)d [k 2 − C + i]m

and so on. In order to compute Ir,m we consider C > 0 (the case C < 0 can be obtained by analytical
continuation) and we perform the integral in the complex plane for k 0 :
r
dd−1 k (k 2 )
Z Z
Ir,m = dk 0 (9.33)
(2π)d [k02 − |~k|2 − C + i]m
q
The integrated function has two poles for k0 = ± |~k|2 + C − i . If we choose a proper integration


path with no poles inside, shown in figure,


Im k0

γ1
×

Re k0
×
γ2

we can say, by Cauchy residue theorem, that


Z +∞ Z −i∞ Z
dk0 + dk0 + dk0 = 0. (9.34)
−∞ +i∞ γ1 ,γ2
9.1. DIMENSIONAL REGULARIZATION AND SUBTRACTION SCHEMES 115

Since the contribution of dk0 vanishes considering the radius of the arcs going to infinity we can
R
γ1 ,γ2
write
Z +∞ Z +i∞ 0
k0 →ikE
Z +∞
dk0 = dk0 = i 0
dkE (9.35)
−∞ −i∞ −∞

If we introduce the euclidean vector kE = (kE k) so that


0 ,~

0 2
k 2 = k02 − |~k|2 = −(kE ) − |~k|2 ≡ −kE
2
(9.36)

the integral in Eq. (9.33) becomes


r
dd kE (kE 2)
Z
Ir,m = i(−1)r−m 2 + C]m , (9.37)
(2π)d [kE

where the denominator is well defined since there are no more poles (C > 0 and kE > 0). This powerful
trick we have exploited is called Wick rotation. Now, in order to find an explicit expression for Ir,m ,
the fact that we are in d dimensions forces us to introduce generalized polar coordinates
Z Z
dd kE = d−1
dkE kE dΩd−1 , (9.38)

q
where kE = (kE k|2 is the length of the eucledian vector kE , and the integration over the solid
0 )2 + |~

angle is defined as, d > 2


Z Z π Z π Z 2π
dΩd−1 ≡ d−2
dθd−1 sin θd−1 ··· dθ2 sin θ2 dθ1 . (9.39)
0 0 0

For example,
Z Z 2π
d=2 Ω1 = dΩ1 = dθ1 = 2π,
0
Z Z π Z 2π
d=3 Ω2 = dΩ2 = dθ2 sin θ2 dθ1 = 2 · 2π = 4π, (9.40)
0 0
Z Z π Z π Z 2π π
d=4 Ω3 = dΩ3 = dθ3 sin2 θ3 dθ2 sin θ2 dθ1 = · 2 · 2π = 2π 2 .
0 0 0 2

To evaluate the integral over the solid angle we can use the gaussian integral

+∞ 2 +∞ 1 d
Z Z Z  
π d/2
= dx e −x2
= dΩd−1 dr r d−1 −r2
e = Ωd−1 Γ , (9.41)
−∞ 0 2 2

and obtain
π d/2
Ωd−1 = 2   . (9.42)
Γ d2

Moreover, introducing the Euler β function

+∞ z a−1 Γ(a)Γ(b)
Z
β(a, b) ≡ dz a+b
= , (9.43)
0 (1 + z) Γ(a + b)

we finally obtain the desired expression for the master integrals


(−1)r−m 4π 2+r−m Γ(2 + r − 2ε ) Γ(m − r − 2 + 2ε )

2
Ir,m =i C . (9.44)
(4π)2 C Γ(2 − 2ε ) Γ(m)
116 CHAPTER 9. RENORMALIZATION

9.2 Vacuum polarization


We have seen that the renormalization at 1-loop of the photon field and electric charge implies the
replacements (at the level of Feynman rules)

−→ + +
(9.45)
2 2 2 2 2
i(kµ kν − k ηµν ) −→ i(kµ kν − k ηµν ) + ie Πµν (k ) + i(ZA − 1)(kµ kν − k ηµν )

where we have called ie2 Πµν (k 2 ) our to-be-computed loop integral. First of all let us exploit Lorentz
decomposition to understand the Lorentz structure of the diagram:

Πµν (k 2 ) = −ηµν A(k 2 ) + kµ kν B(k 2 ), (9.46)

with A(k 2 ), B(k 2 ) scalar functions of the incoming photon momentum k. We exploit the Ward identity
of QED according to which by gauge invariance εµ (k) → εµ (k) + kµ we need kµ Mµ ≡ 0:

0 = k µ k ν Πµν (k 2 ) = −k 2 A(k 2 ) + k 4 B(k 2 ) = 0 =⇒ A(k 2 ) = k 2 B(k 2 ) ≡ k 2 Π(k 2 ). (9.47)

Therefore the loop diagram takes the form

ie2 Πµν = ie2 (kµ kν − ηµν k 2 )Π(k). (9.48)

This is the same form we have obtained for the corresponding counterterm: this is a very important
result because the cancelation cannot occur if the two terms have not the same Lorentz structure.
Hence our renormalization procedure will require
 
+ = i(k µ k ν − η µν k 2 ) (ZA − 1) + e2 Π(k 2 ) (9.49)

to be finite. This requirement gives us the value of ZA , or better the UV expression of ZA , since the
finite terms in ZA cannot be set in a unique way (renormalization scheme dependence). Therefore we
need to compute Π(k 2 ).

p+k
k k
=ie2 Πµν (k 2 )
(9.50)
p
2 Z
dd p
 
Tr γµ (p
/ + m)γν (p
/ + k/ + m)

ε/2
= ieµ d
(−1)i2
(2π) D1 D2

where
D1 = p2 − m2 + i D2 = (p + k)2 − m2 + i (9.51)
The trace is easy to compute
 
2 2
(9.52)
 
Tr γµ (p / + k/ + m) ≡ 4Nµν (p, k) = 4 2pµ pν + pµ kν + pν kµ − ηµν (p + p · k − m )
/ + m)γν (p

Now we introduce the Feynman parametrization

D2 x + D1 (1 − x) = p2 x + k 2 x + 2p · kx − m2 x + p2 − m2 − p2 x + m2 x + (k 2 x2 − k 2 x2 )
(9.53)
= (p + kx)2 − C(x, m2 , k 2 ),
9.2. VACUUM POLARIZATION 117

with C = m2 − k 2 x(1 − x). We managed to get rid of linear terms in the internal momentum p and to
group it in a single quadratic term. Thus we have
1 dd p Nµν (p, k)
Z Z
iΠµν = −4µε dx d
(9.54)
0 (2π) [(p + kx)2 − C + i]2
We perform the shift pµ → pµ − k µ x and we get
1 dd p Nµν (p − kx, k)
Z Z
iΠµν = −4µ ε
dx . (9.55)
0 (2π)d [(p)2 − C + i]2
We compute the shifted numerator

Nµν (p − kx, k) = 2(pµ − kµ x)(pν − kν x)+(pµ − kµ x)kν + (pν − kν )kµ


(9.56)
−ηµν [(p − kx)2 + (p − kx) · k − m2 ]

Thanks to the first of Eqs. (9.32) we can get rid of all linear terms in the integrated momentum pµ3 .
Hence
0
(p − kx, k) = 2pµ pν + 2x(x − 1)kµ kν − ηµν p2 − m2 + k 2 x(x − 1) . (9.57)
 
Nµν (p − kx, k) → Nµν

Now we exploit the second of Eqs. (9.32) and we decompose our integral in master integrals.
  Z 1
2
iΠµν = iΠ(k ) kµ kν − ηµν k 2
= −4 dxµε Mµν , (9.58)
0

with
0 (p − kx, k)
dd p Nµν
Z
µε Mµν = µε
(2π)d [p2 − C + iε]2
2
= µε ηµν I1,2 + µε 2x(x − 1)kµ kν I0,2 − µε ηµν I1,2 − µε ηµν (k 2 x(x − 1) − m2 )I0,2 (9.59)
d
ε 2−d
   
2 2
=µ ηµν I1,2 + − 2x(1 − x)kµ kν + ηµν (x(1 − x)k + m ) I0,2
d
We expand the fraction
2−d ε−2 2 1 1
= = − 1 = − + ε + O(ε2 ). (9.60)
d 4−ε 4−ε 2 8
Now we have to write down the results for the masters integrals, simplifying the expressions expanding
around small ε !ε
ε i 4πµ2 2 Γ( 2ε )
µ I0,2 =
16π 2 C Γ(2)
!
i ε 4πµ2 2 ε
  
ε→0
≈ 2
1 + log Γ 1+ +O(ε2 ) (9.61)
16π 2 C ε 2
| {z }
∼1−γ 2ε +...

i C
 
= 2
∆ε − log 2 + O(ε2 )
16π µ
with
2
∆ε = − γ + log 4π. (9.62)
ε
We have exploited the expansion
!ε ! !
4πµ2 2
ε 4πµ2 ε 4πµ2

ε→0
= exp log ≈ 1 + log + O(ε2 ) (9.63)
C 2 C 2 C
3
It is a common mistake to get rid of linear terms in the virtual momentum before having performed the shift!
118 CHAPTER 9. RENORMALIZATION

Similarly for the other master integral,



i 4πµ2 2
Γ(3 − 2ε ) Γ(−1 + 2ε ) ε→0 i C
 
ε
µ I1,2 =− C ≈ C 1 + 2∆ε − 2 log 2 + O(ε2 ) (9.64)
16π 2 C Γ(2 − 2ε ) Γ(2) 16π 2 µ

Inserting the master integrals in Eq. (9.59) and the expansion of the d dependent fraction we obtain

2ix(1 − x) C
 
ε
µ Mµν = 2
log 2 − ∆ε (+kµ kν − ηµν k 2 ) + O(ε2 ) (9.65)
16π µ

Therefore our loop integral reads, getting rid of the tensor part,

1 1Z 
C

Π(k 2 ) = 2
dx x(1 − x) ∆ε − log 2
2π 0 µ
(m2 − k 2 x(1 − x))
Z 1
=
1
∆ ε −
1
dx x(1 − x) log (9.66)
12π 2 2π 2 0 µ2
1
= ∆ε + ΠR (k 2 )
12π 2
We have reached our regularization goal: divergences have been isolated in the term ∆ε (actually only
a part of it it is divergent) and the ΠR integral is finite. Now we have to subtract divergences and
we will do this following the modified minimal subtraction scheme: we subtract the whole ∆ε (which
cointains also finite terms). Therefore we impose
 
µ ν µν 2 2 2
+ =i(k k − η k ) (ZA − 1) + e Π(k )
(9.67)
1
 
=i(k µ k ν − η µν k 2 ) (ZA − 1) + e2 ∆ε + e2 ΠR (k 2 )
12π 2

to be finite, hence we choose ZA to compensate the ∆ε term:

e2
ZA = 1 − ∆ε (9.68)
12π 2

The renormalization of the photon field lead us to a very important result: the running of the gauge
coupling of QED. First consider the analytic structure of ΠR (k 2 ). This expression is a correction to
the propagator of the photon. For k 2 < 0, as in the case when the photon propagator is in the t or
u channel, ΠR (k 2 ) is manifestly real and analytic. But for an s-channel process k 2 will be positive.
The logarithm function, as seen as a complex function, has a branch cut when its argument becomes
negative, that is when
m2 − x(1 − x)k 2 < 0 (9.69)
The product x(1 − x) for x between 0 and 1 is at most 1/4, so ΠR (k 2 ) has a branch cut beginning
at k 2 = 4m2 , i.e. at the threshold for creation of a real electron-positron pair. Now, in order to
examine how Π(k 2 ) modifies the electromagnetic interaction, let us consider the renormalization of the
propagator4 (the two-point function), which is a consequence of the renormalization of the photon field.

−→ + +
(9.70)
−iηαβ 2 −iηαβ 2 −iηαµ 2 µ ν −iηνβ
 
e0 −→ e + 2 ie (k k − k 2 η µν ) e2 ΠR (k 2 ) 2
k 2 + i 2
k + i k + i k + i
4
Note the dots in the diagrams which distinguish these ones from the ones we drew in Eq. (9.45)
9.2. VACUUM POLARIZATION 119

Let us study the RHS of this transformation. Terms proportional to k µ k ν are not relevant because of a
consequence Ward identity: photon propagators always occur with (i.e. have to stick to) conserved
fermion currents jµ jν , therefore since ∂ µ jµ = 0 we have k µ jµ = 0. So the RHS becomes

−iηαβ 2 −iηαβ e2
   
e 1 − e 2
Π R (k 2
) ≈ , (9.71)
k 2 + i k 2 + i 1 + e2 ΠR (k 2 )
where we have exploited the smallness of the renormalized coupling e, according to the perturbation
theory we are working with5 . This means the following propagator renormalization

1
   
−→ 1 − e2 ΠR ≈ (9.72)
1 + e 2 ΠR
The term in brackets in Eq. (9.71) can be thought as an effective charge:

e2 (µ2 )
e2eff (k 2 , µ2 ) ≡ (9.73)
1 + e2 (µ2 )ΠR (k 2 )

where we highlighted the fact that the renormalized charge depends on the (unphysical, i.e. observables
should not depend on it) renormalization scale µ. We will see soon how to get rid of this dependence.
First let us compute ΠR (k 2 ) in the −k 2  m2 approximation to study the UV behavior of the coupling.
In this case C = m2 − k 2 x(1 − x) ≈ −k 2 x(1 − x) therefore

e2 1 −k 2
Z    
e2 ΠR (k 2 ) ≈ − 2
dx x(1 − x) log + log[x(1 − x)]
2π 0 µ2
(9.74)
e −k 2 α −k 2
    
= 5 − 3 log = − log ,
36π 2 µ2 3π e5/3 µ2

having introduced α = e2 /(4π). We have obtained a formula showing the dependence on energy of the
effective gauge coupling costant. Absorbing the e5/3 factor in µ2 ,

α(µ2 )
αeff (k 2 , µ2 ) = α(µ2 )
 2
 (9.75)
1− 3π log − µk 2

To get rid of the dependence of the coupling constant on the unphysical renormalization scale, we
choose it as the electron mass, at which we know, by experimental measurements, at low energy k ∼
eV the value of the coupling is
1
α(µ2 = m2 ) ≡ αem ' (9.76)
137
Therefore we obtain the famous formula of the running of the gauge coupling constant of QED
! −1
αem k2

2
αeff (k ) = αem 1− log − 2 (9.77)
3π m

Therefore, given the value of the coupling constant at low energy (atomic energies, eV let us say) that
is αem we obtain an expression relating the effective coupling to the energy scale set by experiments
(e.g. k 2 ∼ MZ2 for physics around the Z pole). We understand how physical measurements are
necessary to get rid of ambiguities in the strength of the coupling. Not surprisingly the remnant of the
renormalization procedure, i.e. the renormalization scale µ has to be replaced, being unphysical, by a
value set by an experimental measure whom we associate a value of the coupling.
A prominent outcome of this discussion in the qualitative behavior of the QED gauge coupling: it
grows with the energy scale because of the minus sign in front of the logarithm. There’s an intuitive
interpretation of this phenomenon, sketched in Fig. (9.1). The vacuum polarization generates virtual
pairs somehow screening the bare charge of the electron: at high energy i.e. probing smaller distances
5
Remember that our framework is a perturbative expansion in the gauge coupling, in which we compute effects at first
order in α, i.e first order in e2
120 CHAPTER 9. RENORMALIZATION

we begin to penetrate the polarization cloud and see the bare charge. The bare charge is infinitely larger
than the observed charge, but we don’t mind since what can be observed is only the k 2 dependence of
the effective electric charge, that grows with the energy or, equally, the resolution of the experiment.

The running of couplings is a very important feature of gauge theories and has a very deep meaning:
all the gauge couplings in the SM, namely αem for U (1)em , αW for SU (2)L , αs for SU (3)c , vary
depending on the energy scale at which we are probing the theory. It turns out that αs behaves in an
opposite way with respect to αem . In Fig. (9.2) we show the behavior of the various coupling constants:
we see that at very high energy, namely 1015 ÷ 1016 GeV the coupling constant ”meet”: at these scales,
as claimed by Grand Unification Theories, there might be only a unique gauge group unifying all the
interactions.

Figure 9.1: Sketch of the vacuum polarization effect: virtual e+ e− pairs are effectively dipoles of length
∼ 1/m which screen the bare charge of the electron.

60
1/αem (k)
α−1 inverse coupling strength

1/αW (k)
50 1/αs (k)

40

30

20

10

0
2 4 6 8 10 12 14 16
Energy log

k
GeV

Figure 9.2: Plot of the behavior of coupling constants of the SM fundamental interactions (strong,
electromagnetic, weak) as a function of the energy scale of physical processes. Note how the lines
intersecate at the GUT scales.
9.3. ELECTRON SELF-ENERGY 121

9.3 Electron self-energy


We proceed in our journey to renormalization of QED studying the electron self-energy. We will denote
the loop with the symbol Σ(p/). We compute the loop integral

≡ie2 Σ(p
/)
p p (9.78)
p+k
dd k γ µ (p
/ + k/ + m)γµ
Z
ε/2 2
=(ieµ ) d
(+i)(−i) ,
(2π) D1 D2
with
D1 = k 2 + i D2 = (p + k)2 − m2 + i. (9.79)
We proceed exactly like in the previous section: we obtain
dd k N (p, k)
Z
/) = iµ
Σ(p ε
, (9.80)
(2π)d D1 D2
with
N (k, p) = (2 − d)(p
/ + k/) + dm. (9.81)
Notice that the form of the numerator is precisely the one we expect from the counterterm matching:
a term proportional to p
/ and another proportional to m. Exploiting the Feynman parametrization,

D2 x + D1 (1 − x) + p2 x2 − p2 x2 = (k + px)2 − C (9.82)

with C ≡ −p2 x(1 − x) + m2 x. Therefore


1 dd k N (p, k)
Z Z
/) = iµ
Σ(p ε
dx ; (9.83)
0 (2π)d [(k + px)2 − C + i]2
performing the shift k → k − px we find
1 dd k N (p, k − px)
Z
/) = iµ
Σ(p ε
dx , (9.84)
0 (2π)d [k 2 − C + i]2
with
(9.85)

N (k − px, p) = (2 − d) k/ + p
/(1 − x) + md.
Since we have performed the shift we can drop linear terms in k in the numerator, hence

N (k − px, p) → N 0 (k − px, p) = (2 − d)p


/(1 − x) + md. (9.86)

Finally, we obtain an expression of the form

/) = a(p2 )m + b(p2 )p
Σ(p / (9.87)

with, remembering that


i C
 
(9.88)
ε→0
µε I0,2 −→ 2
∆ε − log 2 ,
16π µ
Z 1
a(p2 ) = − i(ε − 4) dx µε I0,2
0
i 1 2 m2 x − p2 x(1 − x)
Z  
= 4i dx ∆ε − − log
16π 2 4 µ2
(9.89)
0
Z 1
2 ε
b(p ) = i(ε − 2) dx (1 − x)µ I0,2
0
i 1 2 m2 x − p2 x(1 − x)
Z  
= − 2i dx ∆ε − − log
16π 2 0 2 µ2
122 CHAPTER 9. RENORMALIZATION

Notice that if p2 → 0 we have an infrared divergence (we have m2 x in the logarithm), as expected in
this diagram involving a photon propagator. If instead we consider p2  m2 we have C ' −p2 x(1 − x)

i 3 −p2
 
2
a(p ) =4i ∆ ε + − log
16π 2 2 m2
(9.90)
i 1 −p2
 
2
b(p ) = − 2i ∆ε + 1 − log 2
16π 2 2 m

We obtain
ie2 −p2
 
2
ie Σ(p
/) = /
(p − 4m) ∆ ε − log + ... (9.91)
16π 2 µ2
with dots denoting other finite terms we do not care about. Now we can perform the normalization
procedure at the propagator level. We add the proper counterterm, obtaining

−→ + +
(9.92)
i i i i
 
−→ + ie2 Σ(p
/) + i(Zψ − 1)p
/ − i(Zm Zψ − 1)m .
/ − m0
p /−m p
p /−m /−m
p

In order to get rid of ∆ε , according to our usual modified minimal subtraction scheme, we need

e2

Z ψ = 1 − ∆

2 ε

2
16π (9.93)
Z Z = 1 − 4e ∆


m ψ 2 ε 16π

Now, exploiting a perturbation expansion in terms of e2 , we can write

4e2 4e2 e2 3e2


    
Zm = 1− ∆ ε Z −1
ψ ≈ 1 − ∆ε 1+ ∆ε ≈ 1 − ∆ε + O(e4 ). (9.94)
16π 2 16π 2 16π 2 16π 2

Finally, we have found the renormalization constants for the matter field and the mass

e2 3e2
Zψ = 1 − ∆ε Zm = 1 − ∆ε . (9.95)
16π 2 16π 2

9.4 Vertex correction


In this section we will compute the vertex correction which has two important outcomes: the validity
of the Ward identity of QED at any order in perturbation theory hence the conservation of the electric
charge and the correction to the gyromagnetic factor of leptons g − 2 which is one of the most important
successes of QED. Let us start studying the Lorentz structure of the QED vertex by means of a Lorentz
decomposition:
p p0

(9.96)
µ
Vµ = = u(p0 ) a1 γ µ + a2 pµ + a3 p0 u(p)


q
µ

with q = p0 − p. The coefficients ai are called form factors. Notice that in QED we don’t have to add
other terms carring the Lorentz index µ (like γ5 γ µ ): they should be added in a general full theory
approach. Exploiting the Ward identity we can write

0 = qµ V µ = u(p0 ) a1 /q + a2 p · q + a3 p0 · q u(p). (9.97)


 
9.4. VERTEX CORRECTION 123

The term proportional to a1 vanishes because u(p0 )(p /)u(p) ∝ m − m = 0; moreover


/0 − p

q · p =(p0 − p) · p = p0 · p − m2 ,
(9.98)
q · p0 =(p0 − p) · p0 = m2 − p0 · p = −q · p,
therefore we discover a2 = a3 . Thus there are only two independent form factors. The vertex takes the
form
V µ = u(p0 ) a1 γ µ + a2 (p + p0 )µ u(p) (9.99)
 

Now we exploit the Gordon identity in order to get a decomposition into current and spin density part
for the evaluated bilinear. Since {γ µ , γ ν } = 2η µν and σ µν = 2i [γ µ , γ ν ],

u(p0 )(p + p0 )µ u(p) = u(p0 ) 2mγ µ − iσ µν qν u(p). (9.100)




Thus we can rewrite our vertex redefining the form factor as

F2 (q 2 ) µν
 
V µ = u(p0 ) F1 (q 2 )γ µ + iσ qν u(p), (9.101)
2m

The chosen normalization of the form factors is such that, at tree level, F1 (q 2 ) = 1 and F2 (q 2 ) = 0.
Now we can start the renormalization procedure of the QED vertex, as usual adding the one-loop
contribution and the relative counterterm:

p0 p0 p0
p p0
−→ + +
p p p
q q q q
µ µ µ µ
e2
 
0 µ 0 /2 µ µ
u(p )ie0 γ u(p) −→ u(p )ieµ γ + e f1 (q )γ + f2 (q 2 )σ µν qν + (ZV − 1)γ µ u(p)
2 2
2m
√ (9.102)
where ZV = Ze Zψ ZA . Notice that the counterterm is proportional to γ only: this implies that, if
µ

QED is renormalizable (as we know it is) the UV divergent behavior of the one-loop vertex correction
is contained in the f1 (q 2 ) part, while f2 (q 2 ) is finite: it will be the term giving a correction to the
gyromagnetic factor. Also notice that we have set
F1 (q 2 ) = 1 + e2 f1 (q 2 ) + (ZV − 1)
(
(9.103)
F2 (q 2 ) = e2 f2 (q 2 ).
Now we are ready to compute the loop integral

k − p1

p1 p2
= ie2 Λµ
k q+k
q
µ
(9.104)
d4 k −iηνα i(q/ + k/ + m) i(k/ + m)
Z
=(−ieµ/2 )3 4 2
u(p2 )γ ν γµ 2 γ α u(p1 )
(2π) [(k − p1 ) + i] [(q + k) − m + i] [k − m2 + i]
2 2

d4 k γ ν (q/ + k/ + m)γ µ (k/ + m)γν


Z
/2 3
=(eµ ) u(p2 ) u(p1 )
(2π)4 [(k − p1 )2 + i] [(q + k)2 − m2 + i] [k 2 − m2 + i]
| {z }| {z }| {z }
C B A

We have defined the denumerators in such a way that we can exploit the following parametrization
1
Z 1 1
=2 dx dy dz δ(x + y + z − 1) (9.105)
ABC 0 [Ax + By + Cz]3
124 CHAPTER 9. RENORMALIZATION

with

Ax + By + Cz = k 2 + 2k(yq − zp1 ) + yq 2 + zp21 − (x + y)m2 + i = (k + yq − zp1 )2 − ∆ + i (9.106)

where we completed the square with the remaining terms ∆ = −xyq 2 + (1 − z)2 m2 . We are ready to
perform the shift k µ → k µ − yq µ + zpµ1 so that the denominator becomes just (k 2 − ∆ + i)3 . The
numerator instead is originally given by

N µ =u(p2 )γ ν (q/ + k/ + m)γ µ (k/ + m)γν u(p1 )


(9.107)
= − 2u(p2 ) k/γ µ /q + k/γ µ k/ + m2 γ µ − 2m(2k µ + pµ ) u(p1 ),
 

where we have contracted the two gamma matrices. Shifting also the numerator and neglecting linear
terms in k we get

N µ = −2u(p2 ) (k/ − yq/ + zp


/1 )γ µ /q + (k/ − yq/ + p
/1 )γ µ (k/ − yq/ + p

/1 )
(9.108)
+m2 γ µ − 2m(2k µ − 2ypµ + 2zpµ1 + q µ ) u(p1 ).


Now we use that, according to the master integral properties k µ k ν → 14 η µν k 2 6 , that x + y + z = 1 and
the Gordon identity again to obtain

N µ = k 2 −2(1 − x)(1 − y)q 2 − 2(1 − 4z + z 2 )m2 u(p2 )γ µ u(p1 )


 
(9.109)
−2imz(1 − z)qν u(p2 )σ µν u(q1 ) − 2m(z − 2)(x − y)q µ u(p2 )u(p1 ).

The last term on the second line gives a null contribution when integrated over the Feynman parameter:
1 d4 k qµ
Z Z
dx dy dz δ(x + y + z − 1)(−2m)(z − 2)(x − y) u(p2 )u(p1 ) = 0, (9.110)
0 (2π)4 [k 2 − ∆ + i]3

since ∆ and the integral measure are symmetric under x ↔ y while the integrand (x−y) is antisymmetric.
We can split the integral in two contributions, namely the two form factors
1 d4 k k 2 − 2(1 − x)(1 − y)q 2 − 2(1 − 4z + z 2 )m2
Z Z
e2 f1 (q 2 ) = − 2ie2 µ dx dy dz δ(x + y + z − 1)
0 (2π)4 [k 2 − ∆ + i]3
(9.111)
1 d4 k z(1 − z)
Z Z
2 2
e f2 (q ) =8ie m µ 2 2 
dx dy dz δ(x + y + z − 1) (9.112)
0 (2π)4 [k 2 − ∆ + i]3

The only UV divergent term is the one proportional to k 2 in the ef1 (q 2 ) form factor (as we expected): it
carries a log divergence. All the other terms are k independent, hence are proportional to the integral,

d4 k 1 i
Z
4 2 3
=− (9.113)
(2π) [k − ∆ + i] 32π 2 ∆

and are UV finite, as one can deduce by dimensional arguments.

9.5 The anomalous magnetic moment g − 2


Before computing the divergent term by means of dimensional regularization, we compute the second
form factor, which will give us the correction to the gyromagnetic factor for leptons. Having performed
the integration over k and setting µ → 1 since the integral is finite for  → 0, we obtain

α 2
Z 1 z(1 − z)
e2 f2 (q 2 ) = m dx dy dz δ(x + y + z − 1) (9.114)
π 0 [(1 − z)2 m2 − xyq 2 ]
6
When we will generalize to the d dimensional case, we will have of course kµ kν → d1 η µν k2 .
9.5. THE ANOMALOUS MAGNETIC MOMENT G − 2 125

Since the gyromagnetic factors are defined in the limit q 2 → 0 (elastic scattering of the lepton by an
external electromagnetic field, e.g. the one created by a proton), we compute the integral in this limit:

α
Z 1 Z 1 Z 1 z 0 if 1 − y − z = x ∈
/ [0, 1]
e2 f2 (0) = dx dy dz δ(x + y + z − 1) = α (9.115)
π 0 0 0 1 − z  n if − z < y < 1 − z
π
with
1
1 z 1−z 1 z2 1
Z Z Z
n= dz dy = z= = (9.116)
0 1−z 0 0 2 0 2
Therefore
α
F2 (0) = e2 f2 (0) = (9.117)

Now we have to relate this factor to the magnetic moment of the lepton in the Schroedinger equation
coupled to the electromagnetic field. Let us start from the Dirac equation coupled to the electromagnetic
field
/ − m)ψ = 0
(iD (9.118)
multiplying both sides by (iD
/ + m),

/ 2 + m2 )ψ = 0
/ − m)ψ = 0 =⇒ (D
/ + m)(iD
(iD (9.119)

However we have
1
/ 2 = γ µ γ ν Dµ Dν = {γ µ , γ ν } + [γ µ , γ ν ] Dµ Dν = η µν − iσ µν Dµ Dν = Dµ Dµ − iσ µν Dµ Dν (9.120)
 
D
2
and
iσ µν Dµ Dν f (x) =iσ µν (∂µ − ieAµ )(∂ν − ieAν )f (x)
=iσ µν ∂µ ∂ν f − ie(f ∂µ Aν + Aν ∂µ f ) − ieAµ ∂ν f − e2 Aµ Aν f (9.121)


e
= σ µν (∂[µ Aν] )f (x),
2
where the first two terms and the last one vanish because of the antisymmetry of σ µν . Therefore our
Klein-Gordon-like equation reads
e
Dµ Dµ + m2 + Fµν σ µν ψ = 0. (9.122)

2
Now we evaluate 2e Fµν σ µν in the spinor space. Defining σµ ≡ (1, σi ) and σ µ ≡ (1, −σi ) with σi the
Pauli matrices, being !
0 σµ
µ
γ = (9.123)
σµ 0

in the Weyl basis, we have


!
e i ie σµσν − σν σµ 0
Fµν [γ µ , γ ν ] = Fµν
2 2 4 0 σµσν − σν σµ
! !
ie σiσ0 − σ0σi 0 ie σiσj − σj σi 0
2 · Fi0
= |{z} i i + F ij
4 0 0 0
σ σ −σ σ 4 0 σ σ − σj σi
i j
i↔0
! !
ie 2σi 0 ie −2iεijk σk 0
= − Ei − εijl B l
2 0 −2σi 4 0 −2iεijk σk
!
~ + iE)
(B ~ · ~σ 0
=−e ~ − iE)
~ · ~σ
0 (B
(9.124)
126 CHAPTER 9. RENORMALIZATION

having exploited F0i = Ei , Fij = −εijk B k , [σi , σj ] = 2iεijk σ k and εijk εijl = 2δkl . We obtain the
equation ( !)
~ + iE)
(B ~ · ~σ 0
2 2
(∂µ − ieAµ ) + m − e ~ − iE)~ · ~σ ψ = 0. (9.125)
0 (B
Now one can take the relativistic limit of the Klein-Gordon-like equation. First we introduce the spin
operator for spin 1/2 particles
~ = ~σ .
S (9.126)
2
It can be shown that in this limit the equations from the two doublets making up the Dirac spinor
ψ = (ϕ, χ) decouple and the only dynamic equation left is the one for the up doublet ϕ i.e. the large
component. In particular the Dirac equation in the presence of an external magnetic field produces a
Hamiltonian
(~ ~ 2
p − eA) e ~ ~
(9.127)

H= + V (r) + B L + gS ~ ,
2m 2m
acting on electron doublets |ϕi. The coupling g represents the relative strength of its intrinsic magnetic
dipole moment to the strength of the spin-orbit coupling. From the point of view of the Schrödinger
equation, g is a free parameter and could be anything. However, the Dirac equation implies that
g = 2, which was a historically important postdiction in excellent agreement with data when Dirac
presented his equation in 1932. A natural question is then: is g = 2 exactly, or does g receive quantum
corrections? The answer should not be obvious. For example, the charge of the electron is exactly
opposite to the charge of the proton, receiving no radiative corrections (we will prove this in a while),
so perhaps the magnetic moment is exact as well. By the late 1940s there were experimental data
that could be partially explained by the electron having an anomalous magnetic moment, that is, one
different from 2. Now we would like a way to extract the quantum corrections to g without having to
take the non-relativistic limit (which would be the proper way ndr). Going to momentum space with
the substitution i∂µ → (H, p~) we obtain from Eq. (9.125) the formal equation for the large component
of the Dirac spinor
(H − eA0 )2

m (~ ~ 2
p − eA) e ~ ~ e ~ ~

ϕ= + −2 B·S+i E · S ϕ, (9.128)
2m 2 2m 2m m
which can be compared directly7 to Eq. (9.127) to read off the strength of the magnetic dipole
e ~ ~8
interaction g 2m B · S . We find that, as anticipated, g = 2. However if the term proportional to σ µν
had another form we would have found a different result. Thus, a general and relativistic way to extract
corrections to g is to look for loops that have the same effect as an additional term proportional to
σ µν . We have found that the vertex correction at one loop gives us precisely a term of this form. Since
we knew the result from start (or better we cast our equation in a good-looking form in order to get
easily the result in the tree level case), we already have normalized that contribution with 2m so our
theory at one loop gives us
g = 2 + 2F2 (0) (9.129)
hence the anomalous magnetic moment of the electron is
α
g =2+ = 2.00232 (9.130)
π
which is the calculation Schwinger, Feynman and Tomonaga did in 1948 and whose agreement
with data was a triumph of quantum field theory, namely QED. The current best measurement is
g = 2.0023193043617 ± (3 × 10−13 ), so it’s impressive the result obtained just at 1-loop. The theory
calculation has been performed up to 4-loop level. One cannot compare theory to experiment directly,
since the theory is expressed as a function of α, which cannot be measured more precisely any other
way. Therefore g − 2 is now used to define the renormalized value of the fine-structure constant, which
comes out to αem
−1 = 137.035999070 ± (9.8 × 10−10 ).

7
This procedure is not rigorous, since one should perform the non relativistic limit but it has the power to show in the
right way which is the strength of the magnetic dipole interaction with the external magnetic field.
8
The E ~ term is not an electric dipole moment since it has an imaginary coefficient. Instead, it is the Lorentz-invariant
~ ·S
completion of the magnetic moment.
9.6. WARD IDENTITY OF QED 127

9.6 Ward identity of QED


Let us compute now the divergent term of f1 (q 2 ), exploiting dimensional regularization. The integral
is both UV and IR divergent, so we will regulate the UV divergence with dimensional regularization
and the IR divergence with a photon mass mγ 6= 0. In this case the integral gets simply modified to

dd k
Z Z
e2 f1 (q 2 ) = −2ie2 µε dx dy dz δ(x + y + z − 1)
(2π)d

4
 (9.131)
2− d k 2 − 2(1 − x)(1 − y)q 2 − 2(1 − 4z + z 2 )m2
× ,
[k 2 − ∆ + i]3

where
∆ = (1 − z)2 m2 − xyq 2 + zm2γ (9.132)

The UV (log-)divergent term is the k 2 one which, from the master integral formula can be evaluated
with    
4 2 4
Z d
d k 2 − d k i 2 − d d

4−d

ε ε
µ =µ d d Γ
(2π)d [k 2 − ∆ + i]3 (4π) 2 ∆2− 2 4 2
!
i 2 4πµ2 ε ε (9.133)
   
ε→0
≈ + log −γ 1− 1−
16π 2 ε ∆ 4 4
!
i 2 4πµ2
 
= 2
+ log −γ−1 .
16π ε ∆

The remaining terms are UV finite but IR divergent so we can set d = 4 in them and use

d4 k −2(1 − x)(1 − y)q 2 − 2(1 − 4z + z 2 )m2


Z

(2π)4 [k 2 − ∆ + i]3
(9.134)
q 2 (1 − x)(1 − y) + m2 (1 − 4z + z 2 )
=i .
16π 2 ∆

Therefore

e2 1 1 + γ 1
 Z
2 2 ε→0
e f1 (q ) ≈ − + dx dy dz δ(x + y + z − 1)
8π 2 ε 2 0
!  (9.135)
q 2 (1 − x)(1 − y) + m2 (1 − 4z + z 2 ) 4πµ2

× + log .
∆ ∆

Hence, in the q 2 → 0 limit this simplifies to

e2 1 1 + γ 1 m2 (1 − 4z + z 2 ) 4πµ2
 Z  
2
e f1 (0) = 2 − + dz (1 − z) + log
8π ε 2 0 (1 − z)2 m2 + zm2γ (1 − z)2 m2 + zm2γ
(9.136)
e2 1 1 4πµ2
5+γ m2γ
 
= + log + + log .
8π 2 ε 2 m2 2 m2

Our renormalization procedure requires that F1 (0) = 1 + e2 f1 (0) + (ZV − 1) is finite, therefore whithin
the minimal subtraction scheme framework, 9

e2 1 e2 2
 
ZV = 1 − = 1 + − (9.137)
8π 2 ε 16π 2 ε
9
Notice that in the previous sections we exploited the modified minimal subtraction scheme. Now instead we will
remove the (UV) divergent terms only.
128 CHAPTER 9. RENORMALIZATION

We can recast in the minimal subtraction scheme framework the other renormalization constant we
obtained in the previous sections, also expliciting the loop factor

e2 6
  

Zm = 1 + −
16π 2 ε






e 2 
2

(9.138)

Z =1+
ψ −


 16π 2 ε
2

e 8

  
ZA = 1 + −


2
16π 3ε

Notice that it turns out that


ZV = Zψ (9.139)

i.e. the vertex renormalization depends only on the electron self energy renormalization. This fact, due
to the Ward identity , i.e. to gauge invariance, has profound effects on the properties of QED. Also we
can find out what is the renormalization for the electric charge, which turns out to be dependent only
on the photon self energy renormalization:

−1
(9.140)
p
ZV = Ze Zψ ZA = Zψ =⇒ Ze = ZA 2

Therefore
e2 4
 
Ze = 1 + 2
(9.141)
16π 3ε

The fundamental implication of the equality ZV = Zψ is that the charge current ψγ µ ψ is not
renormalized but it is conserved also at the quantum level at any order:

∂µ (ψγ µ ψ) = 0 exact at any order (9.142)

This means that the electric charge is always conserved i.e the gauge symmetry is unbroken at any
order in perturbation theory. In other words, the ratio of charges of leptons and quarks is the same
in the quantum theory as they would be classically. This is pretty remarkable. It explains why the
observed charge of the proton and the charge of the electron can be exactly opposite, even in the
presence of vastly different interactions of the two particles. A priori, we might have suspected that,
because of strong interactions and virtual mesons surrounding the proton, the types of radiative
corrections for the proton would be vastly more complicated than for the electron. But, as it turns
out, this does not happen – the renormalization of the photon field strength rescales the electric
charge, but the corrections to the relative charges of the proton and the electron cancel. The symmetry
associated to charge conservation U (1) is exact in the QED Lagrangian (no anomalies for electric
charge conservation). Another way of saying this is, diagrammatically

p
lim p p0 + p p0 + p0 =0 (9.143)
q→0
q q q
µ µ µ

which is another consequence of the Ward-Takahashi identity of QED.

9.7 The β functions of QED and QCD


Using what we have learned in the previous section we can write a relation between bare and renormalized
coupling in QED at first order in e2 . The relation can be derived from the renormalization group
equation (RGE) of which we will consider a particular case.
9.7. THE β FUNCTIONS OF QED AND QCD 129

QED
We have seen that in the MS renormalization scheme
4 e2
  
(9.144)
ε ε
0
e = µ eZe = µ e 1 +
2 2
3 16π 2 ε
Since the bare coupling does not depend on the renormalization scale µ,
de0 d ε 1 µ dZe
 
(9.145)
ε  ε
0=µ =µ µ 2 eZe = µ 2 eZe + β(e) +
dµ dµ 2 e Ze dµ
where we have defined the QED β function
de
β(e) ≡ µ (9.146)

We obtain the following differential equation
ε dδe
β(e) = − e − e(1 − δe )µ (9.147)
2 dµ
with δe = e2
12π 2 ε
as found in the previous section. At first order in e2 we get
ε e2 d e2 ε 2e2 de
 
β(e) = −e + eµ 1 − 2 2
= −e − 2
µ +O(e3 ) (9.148)
2 12π ε dµ 12π ε 2 12π ε dµ
| {z }
β(e)

Let us solve the equation perturbatively. At leading order in e we have β(e) = −eε/2 + O(e2 ), so,
substituting in the differential equation we have
ε 2e2 ε ε e3
 
β(e) = −e − −e + O(e3 ) = −e + (9.149)
2 12π 2 ε 2 2 12π 2
Thus in the ε → 0 limit we have
e3
β(e) = (9.150)
12π 2
This function is positive and this is crucial. The β function gives the dependence on the renormalization
scale of the renormalized coupling e. Physical quantities do not depend on µ. For instance consider
the cross section σ(e+ e− → anything): the implicit µ dependence in the coupling is canceled by an
explicit dependence in the Feynman diagrams. One can check this by computing the finite parts of the
three divergent diagrams we have considered in the previous section. From
de e3
µ = (9.151)
dµ 12π 2
we find
1 1 1 µ2
= − log (9.152)
e2 (µ) e2 (µ0 ) 12π 2 µ20
Since e2 = 4πα, we obtain the same expression we have found when treating the vacuum polarization,
with the difference that now the relation comes from a more general formalism
α(µ0 )
α(µ) = α(µ0 ) µ2
(9.153)
1− 3π log µ20

The sign of the β function gives this relation its characteristic behavior at high energy. We can
rewrite this formula introducing the reference value α(me ) = αem ' 1/137 and the QED scale
Λ2QED = m2e exp(3π/αem )  GUT scale of non perturbativity
−3π
α(µ) = (9.154)
log(µ2 /Λ2QED )
Therefore we can safely say that QED is always perturbative.
130 CHAPTER 9. RENORMALIZATION

QCD
The case of QCD is more complicated because of the non-abelianity of the theory. It is also richer in
divergent diagrams and the β functions shows a very different behavior from the QED case, making
the strong interactions very peculiar. Let us sketch out qualitatively differences and analogies. Let us
consider as the QCD theory the one obtained from the simple Lagrangian
1
LQCD = − Gaµν Ga,µν + q(iD / − mq )q + counterterms (9.155)
4
where q = (u, d, s)T , mq = diag(mu , md , ms ) and
Gaµν = ∂µ Aaν − ∂ν Aaµ − gf abc Abµ Acν (9.156)
Dµ = ∂µ + igs Aaµ T a [T a , T b ] = if abc T c (9.157)
The gluon self-energy two point function contains two more diagrams, due to the non-abelianity of the
theory and the peculiar properties of strong interaction depend heavily on this difference.

+ + + (9.158)

The calculation of the two new diagrams is a lot more involved and we will not do it here. The
two-point functions for quark self-energies

+ (9.159)
can be obtained straightforwardly from QED with the substitutions e2 → gs2 T a T a = gs2 43 1 so that
gs2 1
q
 Zq = 1 −


2
12π ε (9.160)
 gs2 1
m =1−
Z

2π 2 ε
In the case of the three-point (gqq vertex) function we have one more diagram with respect to QED

+ + (9.161)

Let us simply quote the final result one should obtain performing the calculation of the new diagrams
and treating the terms like we have done in the QED case:
gs3 2
 
β(gs ) = − 2
11 − Nq + O(g 5 ) (9.162)
16π 3
where Nq is the number of quark flavors. Since Nq ≤ 6 the β(gs ) function is negative. This means
that QCD have opposite behavior with respect to QED: it is perturbative at high energy (asymptotic
freedom) and non perturbative at low energy (confinement). From the β function we can compute the
running of the strong coupling constant, in particular we can write
12π
αs (µ) = (9.163)
(33 − 2Nq ) log(µ2 /Λ2QCD )
where
 6π 
(9.164)
33−2Nq
ΛQCD = MZ exp − ' 200 MeV
αs (MZ )
for Nq = 3 and with αs (MZ ) = 0.1185 ± 0.0006. Notice that when µ ≈ ΛQCD the logarithm gets close
to 0 and QCD is non-perturbative and quarks get confined in hadrons (color singlets): δij q i q j (mesons)
and εijk q i q j q k (baryons).
Chapter 10

Electro-Weak Precision Tests

We have seen that the SM has 26 input parameters (or 28 depending whether neutrinos are Dirac or
Majorana particles) to be set by experiments and not predictable by theory. While 26 parameters
might seem like a lot, there are an infinite number of measurements that could conceivably be done.
Since the SM is renormalizable, the result of any of these infinite number of measurements can, in
principle, be expressed as a function of these 26 parameters. Thus, the SM is an overconstrained
system and we can test it by making enough measurements with enough precision. In this section, we
discuss how quantum field theory at loop level is required to connect measurements to the parameters
of the SM, in particular we discuss precision electroweak physics, which is concerned (mainly) with
observables constructed out of leptons and electroweak gauge bosons. There are a few quantities that
are basically only sensitive to electroweak physics and have been measured extremely well. We will
focus on five of them:
1. The electron magnetic dipole moment g = 2.0023193043617 ± (3 × 10−13 ).
2. The decay rate of the muon Γµ = Γ(µ− → νµ e− ν e ) = 2.99598 × 10−19 GeV.
3. The Z boson pole mass MZ? = 91.1876 ± 0.0021 GeV.
4. The W boson pole mass: MW
? = 80.385 ± 0.015 GeV.

5. The polarization asymmetry in the Z boson production


σ(e− + − +
L eL → Z) − σ(eR eR → Z)
ALR = = 0.1515 ± 0.0019 (10.1)
σ(e− + − +
L eL → Z) + σ(eR eR → Z)

This asymmetry, which can be measured using polarized electron beams, would vanish in a non-chiral
theory. Another important observable is the decay rate of the Z boson into electrons Γ(Z → e+ e− ).
In the SM, at leading order in perturbation theory, each one of these five observables depends only
on three electroweak parameters i.e the Lagrangian couplings (g, g 0 , v). We fairly trade these three
parameters with (e, s, v) i.e. the strength of the QED coupling e, the Higgs VEV v and weak mixing
angle s ≡ sin θW . The tree-level dependences of the Z and W masses on our three parameters are
ev ev
MZ = , MW = (10.2)
2sc 2s

where c ≡ cos θW = 1 − s2 . The muon decay rate, computed at tree-level is given by
G2 m5µ
!
m2e
Γµ = F 3 f . (10.3)
192π m2µ
The polarization asymmetry ALR is non-zero because the Z boson has different couplings to left-handed
and right-handed fermions. It turns out that
 2
1
− s2 − s4
(10.4)
2
ALR =  2 .
1
2 − s2 + s4

131
132 CHAPTER 10. ELECTRO-WEAK PRECISION TESTS

Now we would like to know whether all the measured values for these observables are consistent with
the SM. To do that, let us establish a clean definition of our set of parameters (e, s, v) in terms of well
measured quantities, physical observables which are renormalization-independent . We will denote the
values of these couplings extracted from the first three measurements above with a hat and any other
quantity related to these three by tree-level algebraic relations will also be denoted by an hat. In doing
this we trade our set (e, s, v) for the set of numbers (α̂, ĜF , M̂Z ). Let us see how. We begin by defining
M̂Z = MZ? at the Z pole (10.5)
Since g − 2 is known extremely well we use it to define α̂: we worked out in the previous section that
g − 2 = α/π at 1-loop but actually the calculation is known to very high order, competing with the
experimental precision. This gives the value of the fine structure constant at long distances q 2 → 0
α̂(0) = (137.035999074 ± 0.000000044)−1 (10.6)
For precision electroweak physics, it is more useful to work instead with
α̂(MZ ) = (127.944 ± 0.014)−1 . (10.7)
Technically α̂(MZ ) and α̂(M̂Z ) are different. However the difference is a correction that begins at
2-loop and beyond, so which scale we choose is beyond the order we are working in this chapter. Next,
since Γµ is measured extremely well we use it to define
1
ĜF = √ 2
= 1.16393 × 10−5 GeV−2 , (10.8)
2v̂
from which s
1
v̂ = √ = 246.48 GeV (10.9)
2ĜF
Finally for sin2 θW there are many reasonable definitions: one could define it from MW and MZ or from
ALR , or also from g and g 0 . For precision test a logical choice is to base it on the next-best-measured
quantity i.e. M̂Z at the Z pole:
s
π α̂ 1 1 4π α̂
2 2
ŝ ĉ = √ 2
=⇒ ŝ2 ≡ − 1− √ = 0.234289 (10.10)
2ĜF M̂Z 2 2 2ĜF M̂Z2

Therefore we have expressed (e, s, v) as functions of our measured quantities (α̂, ĜF , M̂Z ) closing the
dependences of the EW sector on free parameters. Now we can use these values to test the SM at
high precision making predictions for other observables and see whether they are in agreement with
experimental results or not. In particular we can compute the tree-level prediction of the SM for the 4.
and 5. observables in our starting list:
êv̂
M̂W,tree = = 79.794 GeV, (10.11)
2ŝ
 2
1
− ŝ2 − ŝ4
(10.12)
2
Âtree
LR = 2 = 0.1252
1
2 − ŝ2 + ŝ4
As we see by comparing these values to the ones listed at the beginning of this section, these are
well outside experimental bounds – by nearly 40 standard deviations in the MW case, in which the
predicted mass at tree level obviously cannot be compared with the pole mass without taking into
account at least 1-loop effects. Therefore this does not mean we have a contradiction within the SM:
we cannot make such a conclusion until we include loop corrections and carefully renormalize. We
will see that quantum effects will combine in a particular way in order give rise to a finite correction
solving our problem. Another important observable we can predict is the decay rate of the Z boson
into a electron-positron pair, in the massless electron limit:
2
v̂ê3 1 1
 
+ −
Γ̂(Z → e e ) = − + 2ŝ2 + . (10.13)
96πŝ3 ĉ3 2 4
10.1. OBLIQUE CORRECTIONS 133

10.1 Oblique corrections

There are many radiative corrections that can contribute to the observables listed above. For instance,
for the tree-level scattering process µν µ → e− ν e , we have to consider the following loop-diagrams

+ + + ...

b
+
t

but the last two are the only one which do not affect external legs. Since the observables are given at
tree-level by gauge boson exchange, the largest contributions will come from loops affecting the gauge
boson propagators, i.e. the last two diagrams reported. For historical reasons, these are called oblique
corrections. An advantage of these observables is that, since the Standard Model would have the same
structure with any number of generations, the oblique corrections from each generation will be gauge
invariant and finite. We will therefore focus on the largest corrections, which come from loops of the
third generation quarks (t, b)1 . Notice that out of all corrections, the vacuum polarizations ones are
flavor blind and therefore they do not break Lepton Flavor Universality (LFU).
Now we will compute the vacuum polarization corrections to our predictions for the observable
quantities in terms of observables: the final results will be that our prediction will be function of
numbers (known with high precisions) and loop functions. These loop contributions diverge considered
individually, but, as we know, observables cannot show divergences. Therefore the various contributions
will combine in a way that divergences will cancel and a finite correction to our prediction will come
out. In order to test the electroweak sector, we will proceed in four steps

1. Express our fiducial quantities (α̂, Ĝ, M̂Z ) (plus other observables related to those) in terms of
the Lagrangian parameters (e, v, s) including loop corrections.

2. Solve for the Lagrangian parameters in terms of the fiducial quantities.

3. Express any other quantity we want to compute, as M̂W


? ,A
LR , seff in terms of Lagrangian
2

parameters.

4. Substitute in the measured fiducial quantities to get our predictions for the other observables

To be clear, in the notation we use for this chapter e and MZ mean the MS renormalized electric
charge and Z boson MS mass. Quantities with hat are related to the fiducial quantities by tree-level
relations we have found above; the subscript 1-loop flags the fact we have corrected our prediction.

1
Corrections stemming from the Higgs boson will be discussed later.
134 CHAPTER 10. ELECTRO-WEAK PRECISION TESTS

EW oblique corrections at one loop


First let us recall some Feyman rules for the EW sector, namely the couplings between fermions and
gauge bosons.
f
Aµ =ieQf γ µ

f
f
ie µ  f
γ (T3L − Qf s2 )PL − Qf s2 PR (10.14)

Zµ =
cs
f
f0
ie
Wµ± = √ γ µ PL
s 2
f
We can write in the most general way the EW 1-loop vacuum polarization in the following way

µν
Vν0 ≡ ΠV V 0 (q) = i ΠV V 0 (q 2 )ηµν − ∆V V 0 (q 2 )qµ qν (10.15)


where in principle we take into account the fact that we can have different gauge bosons at the edges of
the loop (i.e. both Aµ or Zµ ). When the gauge bosons are massive, the vacuum polarization amplitudes
need not be transverse by themselves, so Πµν V V 0 (q) need not vanish at q = 0. In QED we had the
2

Ward identity that helped us to relate the two form factors Π and ∆, now instead we don’t have this
information so, in full generality we decomposed the loop that way. All we can say is that the qµ qν
term is not relevant since in precision EW physics the gauge bosons propagators couple to fermionic
current fields of light generations2 . Therefore in an excellent approximation we can say

(10.16)
m→0
q µ Jµlight f = f q µ γµ f = mf f −→ 0

and this will not affect our precision estimates. Exploiting mγ = 0, in the limit q 2 → 0, we obtain some
exact results from gauge symmetry. In fact, remembering that in QED ie2 Πµν 2 2
γγ = ie Π(q )(−η q +
µν 2

q µ q ν ) with Π(q 2 ) regular in q 2 = 0, with our new definition we have Πγγ = e2 q 2 Π(q 2 ) and so

lim Aµ A0ν = 0 =⇒ Πγγ (0) = 0 (10.17)


q 2 →0

and, since
Z
lim Aµ Aν = 0 (10.18)
q 2 →0

we get 3

ΠγZ (0) = ΠZγ (0) = 0 (10.19)


In general our sign conventions are chosen so that a positive value of ΠV V 2
0 (q ) gives a positive mass
shift to the gauge boson mass. Let us also define
dΠγγ
Π0γγ (0) = . (10.20)
dq 2 q 2 =0

Now we will give predictions for the observables, correcting with loop effect each of the quantities
considered above.
2
This is done in order to get precise measurements, being the light fermions easy to detect
3
The diagram in Eq. (10.18) can be seen as the product of two diagrams which at the end contain two factorized loops
µ,γZ (q )Πνα,Zγ (q ) = 0 i.e. a contribution next order to Πγγ . Hence ΠγZ (0) = ΠZγ (0) = 0
limq2 →0 Πα 2 2
10.1. OBLIQUE CORRECTIONS 135

Correction of M̂Z and M̂W (fiducial quantity and observable) In general, treating the optical
theorem, we have seen that vacuum polarization effects produce an effective shift on the mass of gauge
bosons4
MV2 −→ MV2 + ΠV V (MV2 ) (10.21)
therefore we can write the first contribution to the theoretical prediction.

e2 v 2
(M̂Z2 )1−loop = + ΠZZ (MZ2 ),
4c2 s2 (10.22)
2 e2 v 2 2
(M̂W )1−loop = 2 + ΠW W (MW ).
4s

Correction of α̂ (fiducial quantity) We can consider the correction to the propagator

Aµ Aµ Aµ Aµ
−→ +

which implies
i4π α̂ i4π α̂ Πγγ (q 2 )
 
−→ 2 1 + (10.23)
q2 q 2 →0 q q2 q 2 →0

Notice that Πγγ (0) = 0 does not imply that Πγγ (q 2 )/q 2 = 0 for q 2 → 0. Therefore we have

e2
1 + Π0γγ (0) (10.24)

α̂(0)1−loop =

Instead for the coupling constant evaluated at the Z pole we have, for the running coupling

α̂(0)1−loop e2 Πγγ (MZ2 )


 
α̂ ≡ α̂(M̂Z )1−loop = = 1+ (10.25)
1 − ∆α(MZ ) 4π MZ2

with
∆α(MZ ) = ∆αem (MZ ) + ∆αtop (MZ ) + ∆αhadr (MZ ), (10.26)
so that the contributions to the running of the coupling constant come from different effects. Where
the first two terms are computable in perturbation theory, the last term instead is not predictable but
measurable in experiments.

Correction of ĜF (fiducial quantity) Consider the muon decay:

ν ν ν

µ− µ− µ−
ν −→ ν + W− ν
W− W−
W−

e− e− e−

thus we get a correction on the W boson propagator such that, considering the q 2 → 0 limit (which is
a good approximation, being mµ  MW ),

(ĜF )1−loop g2 −i 1 ΠW W (0)


    
√ = 2 1 + iΠW W (q 2 ) 2
= 2 1− 2 (10.27)
2 8MW 2
q − M̂W q 2 →0 2v MW
4
Actually one has to take the real part of the loop integral correction, as we worked out in discussing the optical
theorem. This will be taken as understood.
136 CHAPTER 10. ELECTRO-WEAK PRECISION TESTS

Correction for ŝ2 and ÂLR (observables) We can write

f
Zµ = iγµ (cL PL + cR PR ) (10.28)

with
e f e
T3L − Qf s2 , Qf s2 (10.29)

cL = cR = −
sc sc
The left-right asymmetry is therefore given by

σ(Z → fL− fL+ ) − σ(Z → fR− fR+ ) c2L − c2R


ALR = = (10.30)
σ(Z → fL− fL+ ) + σ(Z → fR− fR+ ) c2L + c2R

If we consider loop effects, cL and cR are corrected.

f f q f q f
Zµ −→ Zµ + Zµ + Zµ
Aµ Zµ
f f f f

In particular the ZA loop diagram discriminates between cL and cR because the photon is blind to
chirality and so it affects only the part proportional to Qf ; instead the ZZ diagram shifts cL and cR
in the same way with a multiplicative factor so that the effect cancels in ALR . Thus in the q 2 → MZ2
limit,
e f −i e ΠγZ (MZ2 )
    
T3L − Qf s2 + iΠγZ (MZ2 ) f
− Qf s2 − sc (10.31)

cL = 2 eQf = T3L
sc MZ sc MZ2

es2 −i eQf 2 ΠγZ (MZ2 )


   
cR = − Qf + iΠγZ (MZ2 ) eQf = − s − sc (10.32)
sc MZ2 sc MZ2
Comparing this results with the tree level ones, we get that

ΠγZ (MZ2 )
ŝ2eff = s2 − sc (10.33)
MZ2
and  2
1
− (ŝ2 )eff − (ŝ4 )eff
(10.34)
eff 2
(ÂLR ) = 2
1
2 − (ŝ2 )eff + (ŝ4 )eff

Correction of Γ̂(Z → e+ e− ) (observable) Similarly to what we did in the previous paragraph

e− e− e− e−
Zµ −→ Zµ + Zµ + Zµ
Aµ Zµ
e+ e+ e+ e+
But in this case all the diagrams contribute to correct our observable. The result is, in the me  MZ
limit (quite a good approximation)
2
ZZ e2 1 1
 
+ −
Γ̂(Z → e e )1−loop = MZ − + 2ŝ2eff + (10.35)
48π s2 c2 2 4

with ŝ2eff which includes ΠγZ . The effect of the third diagram is hidden inside the renormalization
constant of the Z propagator. Since we didn’t treat EW renormalization in detail so far, let us sketch
10.1. OBLIQUE CORRECTIONS 137

the procedure. The Z propagator is modified because the mass is shifted MZ2 → MZ2 + ΠZZ (q 2 ).
However, if we perform a Taylor expansion around the pole mass of the Z boson,

ΠZZ (q 2 ) = ΠZZ (M̂Z2 ) + Π0ZZ (M̂Z2 )(q 2 − M̂Z2 ) + . . . (10.36)

if, as we have already seen M̂Z2 = MZ2 + ΠZZ (M̂Z2 ) the propagator can be written as
−iη µν −iη µν −iη µν ZZ
= ≡ . (10.37)
q 2 − MZ2 − ΠZZ (q 2 ) (q 2 − M̂Z2 )(1 − Π0ZZ (M̂Z2 )) q 2 − M̂Z2

Hence if ZZ = 1 + δZ we have that δZ = Π0ZZ (M̂Z2 ). Moreover the following approximation is commonly
used
ΠZZ (MZ2 ) − ΠZZ (0)
Π0ZZ (M̂Z2 ) ≈ (10.38)
MZ2

Step 1: Fiducial quantities at 1-loop


Step 1 is to express the three fiducial measured quantities in terms of Lagrangian parameters (e, v, s).
Collecting the results obtained above,
e2 Πγγ (MZ2 )
 
α̂(MZ2 )1−loop = 1+ (10.39)
4π MZ2
1 ΠW W (0)
 
(ĜF )1−loop = √ 2
1− 2 (10.40)
2v MW
e2 v 2
(M̂Z2 )1−loop =
+ ΠZZ (MZ2 ) (10.41)
4c2 s2
The left-hand sides of these three equations are the measured values while the right-hand sides are
formal expressions in terms of renormalized MS Lagrangian parameters.

Step 2: Lagrangian parameters at 1-loop


Step 2 is to invert these equations to leading order in α̂ giving e, v, s in terms of the fiducial quantities.

Πγγ (M̂Z2 )
 
e2 = 4π α̂(M̂Z2 )1−loop 1 − , (10.42)
M̂Z2
1 ΠW W (0)
 
2
v =√ 1− (10.43)
2(ĜF )1−loop (1 − ŝ2 )M̂Z2
e2 v 2 ΠZZ (M̂Z2 )
 
2 2
s c = 1 + (10.44)
4(M̂Z2 )1−loop M̂Z2
and, for the last one, substituting the other Lagrangian parameters we have just found, we obtain

π α̂(M̂Z2 )1−loop ΠZZ (M̂Z2 ) ΠW W (0) Πγγ (M̂Z2 )


  
2 2
s c = √ 1+ − − (10.45)
2(ĜF )1−loop (M̂Z2 )1−loop M̂Z2 M̂W2 M̂Z2

Since these vacuum polarization graphs are already 1-loop, we can use either M̂Z and M̂W or MZ and
MW as the arguments of these vacuum polarization graphs. The difference is formally beyond the
order we are working. It is not hard to get an expression for s2 instead of s2 c2 using trigonometric
identities. In terms of the value of sin2 θW calculated from fiducial quantities, i.e. ŝ, we get
ĉ2
 
2
s = ŝ 2
1+ 2 δs (10.46)
ĉ − ŝ2
having defined
ΠZZ (M̂Z2 ) ΠW W (0) Πγγ (M̂Z2 )
δs ≡ + − − (10.47)
M̂Z2 M̂W2 M̂Z2
138 CHAPTER 10. ELECTRO-WEAK PRECISION TESTS

Step 3: observables at 1-loop


Step 3 is to express the other observables first in terms of Lagrangian parameters corrected at 1-loop.
We have already done this for M̂W , ŝ2eff and Γ̂(Z → `+ `− ). We will focus on the first two:

e2 v 2
2
M̂W = 2
+ ΠW W (MW ) (10.48)
4s2
ΠγZ (MZ2 )
ŝ2eff = s2 − sc (10.49)
M̂Z2

Step 4: observables at 1-loop as function of fiducial observables

Then, Step 4, we express M̂W and ŝ2eff in terms of the measured quantities, substituting the values
of the Lagrangian parameters with their expression as function of fiducial quantities at 1-loop, or
derivates (we will omit the subscript for ease of notation but up to now it’s clear that everything is at
1-loop). We are finally ready to test the SM writing our theoretical predictions:

π α̂(M̂Z2 ) Πγγ (M̂Z2 ) ĉ2 2 )


ΠW W (0) ΠW W (M̂W
   
2
M̂W =√ 1− − δ s − + (10.50)
2ĜF ŝ2 M̂Z2 ĉ2 − ŝ2 M̂W2 M̂W2

and
ŝ2 ĉ2 ΠZZ (M̂Z2 ) ΠW W (0) ĉ2 − ŝ2 ΠγZ (M̂Z2 ) Πγγ (M̂Z2 )
 
ŝ2eff = ŝ2 + − − − (10.51)
ĉ2 − ŝ2 M̂Z2 M̂W2 ŝĉ M̂Z2 M̂Z2

10.2 Computation of EW vacuum polarization loops


Next, we need to evaluate the various vacuum polarization loop integrals, which we can then plug in
the expressions above to get our experimental predictions. If one goes through this he will experience
the very important cancellation of divergences in the various quantum effect that is assured by the
fact that the SM is a renormalizable theory. Remarkably we will only consider the loops with third
generation fermions so the cancellation occurs with this sector alone, without the need of the other two
generation or quantum effects involving gauge boson loops (apart from, as we shall see, little effects
due to the fact that |Vtb |2 is not exactly one). In some sense our precision tests of the SM check the
renormalizability of the theory, a posteriori. However, let us stress the fact that we are considering
oblique corrections only, which are the dominant ones in this case but not in general: an example is
the decay of the Z boson in bb which is sensitive to both oblique and non universal corrections.
Let us consider the most general case of a vacuum polarization fermion loop diagram. To compute
it we need to perform loops with left- or right-handed insertions. We will do this for general masses
and couplings and then insert the appropriate masses and couplings for the appropriate self-energy
function. The fermion contributions to the vacuum polarization functions come from loops such as

p+k
p mj p
iΠ µν
≡ (10.52)
mi

k
with mi 6= mj in principle (as in the ΠW W case). The LL and RR contributions at 1-loop are

dd k Tr[(iγ µ )PL i(k/ + mi )(iγ ν )PL i(k/ + p / + mj )]


Z
iΠµν
LL = iΠµν
RR = (−1)e µ 2 ε
d 2 2

2 2

(2π) [k − mi ] (k + p) − mj

d
 (10.53)
e2
Z 1 Γ 2− 2
µν 4−d µ ν
= iη µ dx d [2xm2j + 2(1 − x)m2i 2
− 4x(1 − x)p ] + p p terms,
(4π)d/2 0 ∆2− 2
10.2. COMPUTATION OF EW VACUUM POLARIZATION LOOPS 139

where
∆(x, p, mi , mj ) = xm2j + (1 − x)m2i − x(1 − x)p2 . (10.54)
We have exploited the usual procedure to compute the loop integrals. We can neglect terms since pµ pν
they do not contribute to the observables: they are contracted with external fermionic currents that go
to zero in the limit mf → 0. Now using Πµν LL,RR = η ΠLL,RR , we write, expanding in d = 4 − ε,
µν

e2 2 1 1 1 4πµ2
  Z  
ΠLL = ΠRR = 2 m2i + m2j − p2 − dx[x(1 − x)p − ∆] log 2
−γ (10.55)
4π 3 2ε 2 0 ∆
which contains a UV divergent term ∝ 1/ε and finite pieces. Considering the LR and RL integrals,
they require a mass insertion to turn R ↔ L so it must be odd in the masses, hence we expect to be
proportional to mi mj . In fact

dd k Tr (iγ µ )PL i(k/ + mi )(iγ ν )PR i(k/ + p


 
/ + mj )
Z
iΠµν
LR =iΠµν
RL = (−1)e µ 2 ε
d 2 2

2 2

(2π) [k − mi ] (k + p) − m2

d
 (10.56)
e2
Z 1 Γ 2− 2
= − iη µν µ4−d dx 2mi mj + pµ pν terms,
(4π)d/2 2− d2
0 ∆
so that !
e2 1 1 1 4πµ2
 Z
ΠLR + ΠRL = − 2 mi mj + dx log −γ . (10.57)
4π ε 2 0 ∆
As a check, the above calculation with mi = mj = m should reproduce the QED result for vacuum
polarization. In general, the vector-vector contribution ΠV V is
ΠV V =ΠLL + ΠLR + ΠRL + ΠRR
 
d
e2
Z 1 Γ 2− 2
=− µ4−d p2 dx 8x(1 − x) (10.58)
(4π)d/2 0 [m2 − p2 x(1 − x)]2−d/2
! !
e2 2 1 1 4πµ2
 Z
=− p + dx x(1 − x) log −γ .
2π 2 3ε 0 m2 − p2 x(1 − x)

This is proportional to p2 and agrees with our previous result since Πµν = η µν ΠV V . Now we simply
put in the various couplings and work out the vacuum polarization for the various gauge bosons. ΠW W
is proportional to ΠLL since it only involves left-handed fields; Πγγ is proportional to ΠV V being QED
insensitive to chirality and for ΠZZ and ΠZγ we can use that the Z couples to T3L − s2 Q with strength
−e/(sc) to write everything in terms of vector and left-handed contributions We will focus on the
contributions from the top quark, which gives the largest effect and from the bottom quark, as required
by SU (2)L invariance. Hence we get, being ∆ij ≡ ∆(m2i , m2j ),
X
Πγγ (p2 ) =Nc Q2i ΠV V (∆ii ),
i=t,b
Nc X 1
 
2 i 2 2
ΠγZ (p ) = T Qi ΠV V (∆ii ) − s Qi ΠV V (∆ii ) ,
sc i=t,b 3L 2
(10.59)
2 Nc 1
ΠW W (p ) =|Vtb |2 2 ΠLL (∆tb ),
s 2
Nc X 2 i Qi
 
i 2
ΠZZ (p2 ) = (T3L ) Π LL (∆ii ) − 2s T3L ΠVV (∆ ii ) + s4 2
Q Π
i VV (∆ii ,
)
s2 c2 i=t,b 2

where the 1/2 in ΠW W comes from the 2 normalization of W ± , the 1/2ΠV V comes from the
T3L /hypercharge mixing Π3Y ∝ ΠLR + ΠRR = 1/2ΠV V . Now we can check whether the divergent part
cancels out in the combination giving us corrections to obervables. Therefore, considering only the
divergent terms,
e2 2 e2 2
 
ΠLL = 2 m2i + m2j − p2 + O(ε0 ), ΠV V = − p + O(ε0 ), (10.60)
8π ε 3 6π 2 ε
140 CHAPTER 10. ELECTRO-WEAK PRECISION TESTS

we get

e2 MZ2 2
Πγγ (MZ2 ) = − (Qt + Q2b ),
2π 2 ε
3e2 2 2
 
2 2 2 2
ΠW W (MW ) =|Vtb | m b + m t − MW ,
16π 2 s2 ε 3
(10.61)
2
e M 2
ΠγZ (MZ2 ) = 2 Z (Qb − Qt + 4s2 (Q2t + Q2b )),
8π scε
e2
ΠZZ (MZ2 ) = [3m2b + 3m2t − 2MZ2 (1 + 2(Qb − Qt )s2 + 4(Q2b + Q2t )s4 ].
16s2 c2 π 2 ε
Using Qt − Qb = 1 we find that (s2eff )th and (MW 2 )
th are finite if and only if |Vtb | = 1. However we
2

know that |Vtb | is not exactly 1 hence we could be worried. However, one should include all the other
quark loops to obtain a sure finiteness of the expression. For example, for the divergent part of ΠW W
we find

2 3e2 
|Vtb |2 + |Vts |2 + |Vtd |2 m2t + |Vtb |2 + |Vcb |2 + |Vub |2 m2b + . . .
  
ΠW W (MW )= 2 2
16π s ε |
(10.62)
{z } | {z }
1 1
3e2
= [m2t + m2b + m2c + m2s + m2u + m2d − 4MW
2
],
16π 2 s2 ε
where unitarity of the VCKM matrix has been used. Thus, the CKM matrix elements drop out of the
divergent parts of the vacuum polarization graphs. Since the finite parts of loops involving light quarks
are proportional to the light-quark masses, we can neglect their corrections: including all the loop
corrections is therefore equivalent to including just the m2t contribution from the top-bottom loop and
set Vtb = 1. Since the divergent contributions to our predictions cancel, we can evaluate the 1-loop
corrections and compare the results to the experimental numbers we provided at the beginning of this
section, using fiducial quantities and derived ones. As we have said oblique corrections are maximized
in case of virtual top and bottom quarks circulating in the loops. Sometimes it is helpful to have
approximate analytic formulae for the oblique corrections. For example for mb → 0 and MZ  mt (
which is not true, there’s only a factor of two), we obtain

3α̂(M̂Z2 ) m̂2t
  
2
(M̂W )th 2
≈ ĉ M̂Z2 1+ = 80.285 GeV (10.63)
16πŝ2 (ŝ2 − ĉ2 ) M̂Z2

with m̂t =163 GeV (it is the scale-dependent top mass not the pole mass, there’s a difference of 10
GeV since for the top 1-loop and 2-loop effects are very large) and

3α̂(M̂Z2 ) m̂2t
  
(ŝ2eff )th ≈ ŝ2 1 − = 0.2314 (10.64)
16πŝ2 (ŝ2 − ĉ2 ) M̂Z2

which leads to Aeff LR = 0.1480 in close agreement with the experiments. In addition to the top/bottom
contribution, the other reasonably sized correction is from the Higgs boson (Mh = 125 GeV). However
the observables depend only logarithmically on the ratio M̂h2 /M̂W 2 , against the quadratic dependence on

m̂2t /M̂Z2 , as we shall see in more detail. In the following table we compare the results of EW precision
tests: notice the improvement with the introduction of 1-loop effects. Actually this precision test are

Observable Exp. value Tree-level 1-loop (t, b) 1-loop (t, b, h)


MW pole [GeV] 80.399 ± 0.023 79.794 80.368 80.333
ALR 0.1514 ± 0.0019 0.1252 0.1491 0.1470

very important even to get prediction for new particles masses: by tuning the theoretical results (at the
best order one can compute, to have the best estimate) in order to fit the experimental ones one can
get good constraints on the values of the most heavy particles. This is precisely the way the theoretical
community predicted bounds for mt and Mh before their discovery. Actually, being the dependence on
10.3. CUSTODIAL SU (2)C SYMMETRY 141

mt strong on precision observables it was easy to get a fair bound while in the case of the Higgs the
weak logarithmic dependence led to a very loose bound. Anyway the method of fitting the EW sector
in search of new physics is very powerful.
Summarizing, it should be clear the procedure we followed. We have a theory with Lagrangian
parameters to be fixed by experiments. In the EW sector we can express all these parameters in terms
of the best measured fiducial quantities α̂, ĜF , M̂Z . Then we correct our theoretical estimates at the
quantum level and re-express all the parameters in terms of these loop corrections. Having done that
we compute theoretical predictions for other observables MW , ALR , s2eff etc., which can be measured
with fair precision in order to test the impact of loop corrections. We eventually found out that the
SM is consistent with experiments at very high precision only considering quantum effects.

10.3 Custodial SU (2)C symmetry


In the Standard Model, the W -boson and Z-boson masses have a ratio determined by the relative
strengths of the weak and electromagnetic gauge couplings. That is,
2
MW g2
2 = 2 = cos2 θW (10.65)
MZ g + g02

This is a consequence of the way the SU (2)L ⊗ U (1)Y gauge symmetry is spontaneously broken through
the Higgs mechanism with a single SU (2) doublet. If the Higgs sector were more complicated, there
would have been deviations from this, even at tree-level. It is therefore useful to define the ρ-parameter,

MW 2
ρ≡ (10.66)
MZ2 c2W

Denoting the tree-level value of ρ as ρ0 we have that in the SM ρ0 = 1. This prediction arise from a
global symmetry of the Higgs sector, the custodial symmetry. Indeed recall that our original Higgs
field ϕ transforms as a doublet under SU (2). Writing
!
1 h3 + ih2
ϕ= √ , (10.67)
2 h1 + ih4

we see that the Higgs potential


2
v2 λ 2

(10.68)
† † 2
V (ϕ ϕ) = λ ϕ ϕ − = h + h22 + h23 + h24 − v 2
2 4 1

is invariant under a global SO(4) symmetry under which the quadruplet (h1 , h2 , h3 , h4 ) transforms in
the fundamental representation. Notice that SO(4) has 6 generators and is isomorphic to SU (2)⊗SU (2).
When the Higgs field gets a VEV, i.e. after SSB

hh1 i = v, hh2 i = hh3 i = hh4 i = 0 (10.69)

so the SO(4) symmetry breaks down to SO(3). Thus there are actually three unbroken (global)
symmetry directions in the Higgs sector of the Standard Model. In other words, there is a residual
global SO(3) ' SU (2) symmetry after EW SSB. This is known as custodial isospin or custodial
SU (2)C . Such symmetry is manifest in the SM mass spectrum providing ρ0 = 1.

The Higgs sector and gauge boson masses


Let us study in more detail the EW gauge sector. The vector boson mass spectrum is given by

 2 g2v2
 MW =

g 2 v 2 µ+ − v 2

Lgmass = W Wµ +
2
gWµ3 − g 0 Bµ =⇒ 4 (10.70)
4 8 |  (g 2 + g 0 2 )v 2
MZ2 =
{z } 

(g 2 +g 0 2 ) Zµ Z µ
4
142 CHAPTER 10. ELECTRO-WEAK PRECISION TESTS

and v 2 = MW
2 /c2 . Now, considering the g 0 → 0 limit, where c
W W → 1, we get a degenerate mass
spectrum
M2 g 0 →0 1
Lgmass = MW
2
Wµ+ W −µ + 2W Zµ Z µ −→ MW 2
(Wµ+ W −µ + Zµ Z µ ) (10.71)
2cW 2
i.e. a degenerate triplet. The custodial SO(3) ' SU (2)C symmetry is explicitly broken by U (1)Y
interactions: if g 0 = 0 we would have ρ = 1 at any order in perturbation theory, not just ρ0 = 1, which
is like that a tree-level only by construction. We will see that also Yukawa interactions break SU (2)C
and we can therefore expect simply from symmetry arguments the ρ-parameter to be of the form

g02 yt2
ρ=1+ f1 + f2 , (10.72)
16π 2 16π 2
where f1,2 are loop functions depending on fermion and vector boson masses and yt is the Yukawa
coupling of the top quark which is the dominant contribution in the Yukawa sector. Therefore at the
quantum level the ρ-parameter differs from 1. Let us see how it is the case.

Chiral SM Lagrangian and SU (2)C breaking


The Higgs sector has a SU (2)L ⊗ U (1)Y gauge symmetry but also an accidental global symmetry, the
custodial symmetry. Let us write
?
! !
φ+ φ0
(10.73)
?
ϕ= ϕc = iσ2 ϕ? = with φ− = φ+
φ0 |{z} −φ−
ε

We define a Higgs matrix or bi-doublet Higgs


?
!
1 1 φ0 φ+
Σ ≡ √ (ϕc , ϕ) = √ − φ0 . (10.74)
2 2 −φ

We can rewrite the Higgs Lagrangian as

(10.75)
2
LΣ = Tr[(Dµ Σ)† (Dµ Σ)] + µ2 Tr[Σ† Σ] − λ Tr[Σ† Σ] ,

with
g g0
Dµ Σ = ∂µ Σ + i σ · Wµ Σ − i Bµ Σσ3 , (10.76)
2 2
!
1 0
with the σ3 = coming from the fact that ϕ and ϕc have opposite hypercharges Yϕ = −Yϕc = 12 .
0 −1
The SU (2)L ⊗ U (1)Y gauge symmetry acts on Σ as follows

SU (2)L : Σ → LΣ
−i
σ θ
(10.77)
U (1)Y : Σ → Σe 2 3

being L a unitary matrix. Notice that

Tr[(Dµ Σ)† (Dµ Σ)] → Tr[(Dµ Σ)† L† L(Dµ Σ)] = Tr[(Dµ Σ)† (Dµ Σ)] (10.78)

so the Higgs sector is invariant under a global SU (2)L . To make the global symmetry manifest we
take the limit g 0 → 0 in which Dµ Σ = ∂µ + ig2 σ · Wµ Σ. Hence in this limit LΣ has another global
symmetry SU (2)R
SU (2)R : Σ → ΣR† (10.79)
so that
Tr[(Dµ Σ)† (Dµ Σ)] → Tr[R(Dµ Σ)† (Dµ Σ)R† ] = Tr[(Dµ Σ)† (Dµ Σ)], (10.80)
thus for g 0 → 0 the Higgs sector has a large global symmetry

SU (2)L ⊗ SU (2)R : Σ → LΣR† , (10.81)


10.4. ONE LOOP CORRECTIONS TO ∆ρ 143

where SU (2)L is just the global version of the corresponding gauge symmetry, while U (1)Y is a subgroup
of SU (2)R . Indeed if we consider the U (1) subgroup of SU (2)R given by e 2 σ3 θ this is equivalent to
i

U (1)Y : Σ → Σe− 2 σ3 θ . After SSB the Higgs bi-doublet takes a VEV,


i

!
1 v 0
hΣi = √ (10.82)
2 0 v
breaks both SU (2)L and SU (2)R individually: LhΣi 6= hΣi and RhΣi 6= hΣi. However there is a
residual SU (2)V = SU (2)L+R symmetry in the case L = R: LhΣiL† = hΣi: this residual symmetry is
the custodial symmetry SU (2)C :

(10.83)
SSB
SU (2)L ⊗ SU (2)R −→ SU (2)L+R

To this global symmetry breaking and in particular to the three broken generators are associated
three Goldstone bosons which are then eaten up by the Higgs mechanism to provide masses to W ± , Z
bosons. This really analogous to what we encountered in studying the QCD chiral symmetry. In that
case, even without any Higgs sector the large SU (2)L ⊗ SU (2)R symmetry was broken by the QCD
hqqi condensate and the residual group was a SU (2)V . In both cases we call this formalism the chiral
formulation of the SM.
Also the Yukawa sector explicitly breaks the custodial symmetry SU (2)L+R . In fact, consider the
Lagrangian
!
 yu 0
LY = QL yd dR ϕ + QL yu uR ϕc + h.c. = QL ϕc , ϕ QR + h.c (10.84)
| √{z } 0 yd

!
uR
where QR = . If now we apply the custodial symmetry
dR

QR → RQR

(10.85)

SU (2)L+R : QL → LQL

 †

Σ → LΣR
we notice that the Lagrangian is invariant under this symmetry only provided that yu = yd which
however is not true, as we know from the quark mass spectrum. To sum up the Higgs sector has a
SU (2)L ⊗ SU (2)R symmetry which after SSB breaks into SU (2)L+R . However both gauge interactions
and Yukawa interaction break explicitly SU (2)L+R : the former because SU (2)R is not a symmetry of
the SM, but only the subgroup U (1)Y is (so only SU (2)L is left, not SU (2)L+R ); the latter because
yu 6= yd i.e. up-type quarks have not the same masses as down-type quarks. To sum up the custodial
symmetry is an accidental symmetry that is broken by the fact that g 0 6= 0 and yu 6= yd .

10.4 One loop corrections to ∆ρ


Let us see how loop effects contribute to the violation of SU (2)C . The responsibles are both gauge
U (1)Y and the Yukawa interactions. Despite the fact that we have introduced this symmetry as acting
on the Higgs field, it is not hard to see that it actually just acts on the Goldstone bosons. Thus, it
should be present in the low-energy theory. In fact, it is even present in the Fermi theory. The CC and
NC Fermi interactions, coming from W, Z exchange respectively are
e2 e2
LCC+NC = − (Jµ
1
+ iJµ
2
)(J 1µ
− iJ 2µ
) − (J 3 − s2 Jµem )(J 3µ − s2 Jem
µ
). (10.86)
s2 MW 2 s2 c2 MZ2 µ

We define GF = 2e2
2
8s2 MW
so

−8GF  1
|Jµ + iJµ2 |2 + ρ(Jµ3 − s2 J em )2 . (10.87)

LCC+NC = √
2
144 CHAPTER 10. ELECTRO-WEAK PRECISION TESTS

So the custodial symmetry forcing ρ0 = 1 is just the symmetry that relates the strength of the weak
part of the NC interactions to the strength of the CC interaction. As an example, consider electroweak
symmetry breaking by QCD. Recall that even if we did not have a Higgs sector at all, we would still
have SU (2) ⊗ U (1) → U (1) by the QCD hqqi condensate. If this were the only source of electroweak
symmetry breaking, we would still find ρ0 = 1 because QCD has indeed a custodial SU (2) symmetry.
As we know in the massless quark limit QCD has a full SU (2)L ⊗ SU (2)R symmetry. After SSB since
only SU (2)L has associated gauge bosons, the breaking is SU (2)L ⊗ SU (2)R → SU (2)V where SU (2)V
is the custodial symmetry that relates the gauge bosons masses to the gauge charges.
In particular, the ρ parameter is defined as the difference between the CC and NC interaction
strengths that can be measured from the muon decay and the neutrino-neutrino pure scattering,
respectively. By analogy with the correction to the CC strength the NC one is given by

GNC e2 Π (0)
 
√F = 2 2 1 − ZZ 2 + . . . (10.88)
2 8s MZ MZ

so that
GNC ΠW W (0) ΠZZ (0)
∆ρ ≡ F
−1= 2 − (10.89)
CC
GF MW MZ2

10.4.1 Yukawa contribution


We have just shown that the Yukawa couplings in SM violate SU (2)C . Thus, the by far dominant
contribution to ∆ρ = ρ − 1 in the SM is from top quark. Taking into account that ΠV V (p2 = 0) = 0
and ∆ij (0) = xm2j + (1 − x)m2i we find

Nc 1
 
(10.90)
X
2
∆ρ = 2 2 ΠLL (∆tb (0)) − ρ0 T3L,i ΠLL (∆ii (0))
s MW 2 i=t,b
Nc e2 1 1 4πe−γ µ2 4πe−γ µ2
  
= 2 2 (m2
+ m2
) + m 4
log − m 4
log (10.91)
s MW 4π 2 16 t b
8(m2t − m2b ) t
m2t b
m2b
1 4πe−γ µ2 4πe−γ µ2
 
− m2t log − m4
b log (10.92)
8 m2t m2b
Nc αem 2m2b m2t m2t
 
= m2
t + m2
b − log . (10.93)
16πs2 c2 MZ2 m2t − m2b m2b

Note, as a consistency check, that if mt = mb (i.e. yu = yd in the relevant case) ∆ρ = 0 i.e. the
custodial symmetry is at work preserving ρ = 1. Instead we have mt  mb (huge SU (2)C breaking) so
the expression simplifies to (as also one can get from NDA)

mt mb 3αem m2t


∆ρt ≈ ∼ GF m2t = 0.008 ∼ 1% (10.94)
16πs2 c2 MZ2

Again we find a strong dependence of the correction from the top mass, which allows us to fit the
experimental result with fair precision in order to get bounds on the top mass.

10.4.2 Higgs-boson contribution


Custodial symmetry help us to understand the properties of the theory at the quantum level. If we
neglect the Yukawa interactions (i.e. we take the quarks massless), radiative corrections to ρ due to the
gauge and Higgs sectors must be proportional to g 0 2 . In particular, we don’t expect a sensitivity of ∆ρ
to the Higgs mass sqared Mh2 as Mh2 ∼ λv 2 and λ does not break the custodial symmetry. Hence the
dependence of ∆ρ on λ can only be of the type λ0 ∼ log λ. As a result, the correction to ∆ρ stemming
from the Higgs sector is expected to the of the form
!
Mh2
02
∆ρ ∝ g log . (10.95)
MZ2
10.4. ONE LOOP CORRECTIONS TO ∆ρ 145

This naive expectation will be confirmed shortly by the explicit calculation of the following diagrams

h
h

ZW ZW
ZW ZW ZW (10.96)
Defining V = W, Z, we have
p+k

(1)µν h
iΠV V =V V
p k p
(10.97)
2
2iMV2 dd k i kµ kν i
 Z  
= µε − η µν +
v (2π) k − MV2
d 2 MV (p + k)2 − Mh2
2
2
2MV2 1 dd k 1 kµ kν
 Z Z  
= µε x − η µν +
v 0
d
(2π) [(k + px)2 − C]2 MV2
where C = −xp2 (1 − x) + MV2 (1 − x) + xMh2 . After performing the momentum shift and neglecting
terms linear in k µ and pµ pν , we obtain
2
2MV2 1 dd k 1 1 ε 2
 Z Z    
(10.98)
(1)µν ε
iΠV V = µ x d 2 2
η µν − 1 + 2 1+ k
v 0 (2π) [k − C] 4MV 4
2 Z
2MV2 1 1 ε
    
= dxη µν
− I0,2 + 2 1+ I1,2 (10.99)
v 0 4MV 4
2
2MV2 iη µν 1 µ2 1 µ2
 Z   
= dx − ∆ε + log + C ∆ ε + 1 + log . (10.100)
v (4π)2 0 C 2MV2 C
Now we compute the second diagram 5 , rewriting the loop integral in a suitable form to be summed
with the previous one
h

1 2iMV2 µν dd k i
  Z
(2)µν
iΠV V = V V = η
2 v2 (2π) k − Mh2
d 2
p p
MV2 µν dd (p + k) 1 k 2 − MV2
Z
=−
v2
η
(2π)d (p + k)2 − Mh2 k 2 − MV2 . (10.101)
MV2 µν 1 dd k k 2 + (1 − x)2 q 2 − MV2
Z Z
=− η dx
v2 0 (2π)d [k 2 − C]2
i MV2 µν 1
Z 
dx 2C + (1 − x)2 p2 − MV2 ∆ε

=− η
(4π)2 v 2 0
µ2

+ 2C + (1 − x)2 q 2 − M V 2 log

+C
C
Summing the two diagrams, we obtain
MV2 1  1 µ2
 Z  
2
− 2MV2 + p2 ∆ε + 3MV2 2 2
(10.102)

ΠV V (p ) = dx C − + (1 − x) q log .
(4π)2 v 2 3 0 C
5
Notice the factor 1/2 in front of the amplitude: it has to be added in addition to the Feynman rule of V V hh since
the Higgs legs are connected. Hence the symmetry factor of the diagram is S = 1 to be multiplied by the Lagrangian
coupling iMV2 /v 2 which gives us a 1/2 factor missing from the Feynman rule.
146 CHAPTER 10. ELECTRO-WEAK PRECISION TESTS

For our purposes (getting corrections to observables), we can neglect divergent terms proportional to
∆ε as they must cancel out after considering the contributions from the following two diagrams:

ZW ZW
ZW ZW (10.103)

We know this since divergences in oblique corrections from fermion loops cancel independently,
hence, also divergences in the gauge sector must cancel on their own. We can perform now the
integration over Feynman parameters in the large Higgs mass limit Mh2  MV2 , p2 where C ' xMh2
and log(µ2 /C) ' − log(Mh2 /µ2 ). Fixing the renormalization scale to µ2 = MZ2 , we obtain

MW2 
1 2

1 2

Mh2

2 2
ΠW W (p ) ' M + 3M + p log ,
(4π)2 v 2 2 h W
3 MZ2
(10.104)
MZ2 1 2 1 2 Mh2
   
2 2
ΠZZ (p ) ' M + 3MZ + p log 2 .
(4π)2 v 2 2 h 3 MZ

and finally the correction to the ρ-parameter from the Higgs sector in the large Mh limit reads

ΠW W (0) ΠZZ (0) 3GF MZ2 Mh2 g0 →0


∆ρ = 2 − ' − √ sin 2
θ W log −→ 0 (10.105)
MW MZ2 8π 2 2 MZ2

in agreement with our dimensional argument. The weak dependence on the Higgs mass unfortunately
provides us only a loose bound on the Higgs boson mass from the EW fit, as we have already stated.
We can compute corrections to other EW observables coming from the Higgs sector. We obtain
again logarithmic dependences on the Higgs boson mass. For instance, remembering that for the Higgs
case Πγγ = 0 and ΠγZ = 0, α = MW 2 s2 /πv 2

ŝ2 ĉ2 ΠZZ (MZ ) ΠW W (0)


 
∆s2eff = −
ĉ2 − ŝ2 MZ MW
2 2 2 2
(10.106)
ŝ ĉ MZ (1 + 9ŝ ) Mh2 α(1 + 9ŝ2 ) Mh2
= 2 log = log
ĉ − ŝ2 48π 2 v 2 MZ2 48π(ĉ2 − ŝ2 ) MZ2

ŝ2 ΠZZ (MZ2 ) ΠW W (0) ΠZZ (MZ2 ) ΠW W (MW2 )


   
2 2
∆MW =ĉ MZ2
− 2 − − −
ĉ − ŝ2 MZ2 MW2 MZ2 MW 2

2 ŝ2 ĉ2 M 2
(10.107)
11MW Z Mh2 11αĉ2 MZ2 Mh2
=− log = − log
48π 2 v 2 (ĉ2 − ŝ2 ) MZ2 48π(ĉ2 − ŝ2 ) MZ2

10.5 One loop corrections to Z → f¯f


As already anticipated, a very precise test of EW theory comes from the measurements of the Z boson
properties at LEP e+ e− collider: they operated at the centre of mass energy s ∼ MZ2 , producing
millions of Z at rest and allowing for detailed studies of its coupling to light and heavy fermions. The
tree level prediction for the decay rate Γ(Z → f f ) can be generalized considering that at 1-loop we
have
MW 2
= 1 + δρf (10.108)
MZ2 cos2 θW
Taking this into account we resum all corrections into the δρf term and, analogously to what we have
previously done, in an effective seff , So we can write the 1-loop prediction as
v
m2f  2 m2f  m2f 
u
4GF 3t1

(10.109)
u
f 2 2 2
Γ(Z → f f ) = √ (1 + δρf )Nc MZ − 2 (gV + gA ) 1 − 2 + 3(gV − gA ) 2 .
3π 2 4 MZ MZ MZ
10.5. ONE LOOP CORRECTIONS TO Z → F̄ F 147

with
T3f − 2Qf s2eff Tf
gV = gA = 3 (10.110)
2 2
In principle ρeff = 1 + δρf and seff do not contain only oblique correction to the Z boson propagator,
that we have studied so far, but also self energies and vertex corrections of the complete EW theory.
This process is therefore an interesting playground to get an insight into the EW renormalization. The
diagrams contributing to the Z boson decay at 1-loop are the following

f f f
Ma = Zµ , Mb = Z µ , Mc = Z µ
γ Zµ
f f f
f f
f
Md = Z µ , Me = Z µ , Mf = Z µ

f
f f

The tree level and oblique contributions are in the first row, while the second row contains vertex
corrections and fermion self energies. Denoting the momentum of the Z boson by q and the momenta
of the outgoing fermion and antifermion by p1 and p2 respectively, we have for these diagrams the
following amplitudes (the factor T3f − Qf s2 in the couplings can be written as ± 12 (1 − 2|Qf |s2 ) with
the sign + (−) for the up (down) component of SU (2)L fermion doublet):

ie
u(p1 )γ λ (1 − 2|Qf |s2 )PL + (−2|Qf |s2 )PR v(p2 )ελ (q)
 
Ma = ∓
2sc
ie ΠγZ (MZ2 )
 
Mb = ∓ u(p1 )γ λ 2sc|Qf | v(p2 )ελ (q)
2sc MZ2
ie 1 1
 
Mc = ∓ u(p1 )γ λ (1 − 2|Qf |s2 ) Π0ZZ (MZ2 )PL + (−2|Qf |s2 ) Π0ZZ (MZ2 )PR v(p2 )ελ (q)
2sc 2 2
ie
u(p1 )γ λ ΛL PL + ΛR PR v(p2 )ελ (q)
 
Md = ∓
2sc
ie

u(p1 )γ λ (1 − 2|Qf |s2 )δL PL + (−2|Qf |s2 )δR PR v(p2 )ελ (q)

Me + Mf = ∓
2sc
(10.111)
Notice that we have always exploited a chiral decomposition of the various terms. The factors ΛL and
ΛR denote 1-loop vertex amplitudes (note that e/2sc has been factored out), Π0ZZ (MZ2 ) is the derivative
at q 2 = MZ2 of the Z self energy as in Eq. (10.38) and δL,R are the external fermion wave-function
renormalization factors (each external leg carries a 12 δL,R factor).
We want to write the sum of all contributions as M = Ma + ... + Me + Mf

ie
u(p1 )γ λ (1 − 2|Qf |s2eff )PL + (−2|Qf |s2eff )PR v(p2 )ελ (q) (10.112)
 
M = ∓(1 + A)
2sc
where A and s2eff are defined by the two conditions

2 1 0 ΠγZ (MZ2 )

2 2

 cL ≡(1 − 2|Q f |s ) + (1 − 2|Qf |s ) Π (M Z ) + 2sc|Q f | 2 + ΛL + (1 − 2|Qf |s2 )δL
2 MZ





 =(1 + A)(1 − 2|Q |s2 )

f eff
2 2 1 0 2 ΠγZ (MZ2 )
≡(−2|Q |s |s | + ΛR + (−2|Qf |s2 )δR


 cR f ) + (−2|Q f ) Π (M Z ) + 2sc|Q f 2
2 MZ





2
=(1 + A)(−2|Qf |seff )

(10.113)
148 CHAPTER 10. ELECTRO-WEAK PRECISION TESTS

so that
1
1 + A = cL − cR = 1 + Π0 (MZ2 ) + ΛL − ΛR + (1 − 2|Qf |s2 )δL − (−2|Qf |s2 )δR . (10.114)
2
It is convenient to define the renormalized (scale-independent) vertex corrections FL and FR

FL ≡ ΛL + (1 − 2|Qf |s2 )δL FR ≡ ΛR + (−2|Qf |s2 )δR . (10.115)

Instead for s2eff ,


1 cL cR 1
 
1− = =⇒ −2|Qf |s2eff = = cR 1 − Π0 (MZ2 ) + FL − FR (10.116)
2|Qf |s2eff cR 1+A 2
hence
c ΠγZ (MZ2 ) 1
   
s2eff 2
=s 1− 2 − FL + 1 − FR . (10.117)
s MZ 2|Qf |s2
Now we have to express the MS parameters e, s, c through the measurable quantities ê, ŝ, ĉ (the
replacement is straightforward if one of these multiply directly a loop correction, since we are keeping
only first order terms). Then we can write the 1-loop decay rate as in Eq. (10.109), with

1 + δρf = (1 + A)2 ∼ 1 + 2A = 1 − (δs + . . . ) + Π0ZZ (MZ2 ) + 2(FL − FR ) (10.118)

and
ĉ2 ĉ ΠγZ (MZ2 ) 1
    
s2eff = ŝ2 1 + (δs + . . . ) 1− − FL + 1 − FR . (10.119)
ĉ2 − ŝ2 ŝ MZ2 2|Qf |ŝ2
where the dots take into account other possible contributions due to box diagrams. This factor
originates from expressing with 1-loop accuracy the factor e2 /s2 c2 though measurable quantities. In
practice, the vertex corrections and the self-energy contributions (namely FL,R ) are very small due to
the small fermion masses, so the corrections δρf and s2eff are dominated by ∆ρ present in δs

ΠZZ (0) Πγγ (MZ2 )


δs = −∆ρ − − (10.120)
MZ2 MZ2

which contains a dominant term proportional to m2t /MW 2 . Since the effective coefficient of ∆ρ is

positive and the decay rate is proportional to 1 + ∆ρ, Γ(Z → f f ) increases with increasing the top
quark mass. This is actually true for all fermions except for the bottom quark. For the bb pair in the
final state there is an additional, and competitor, contribution of order mt /MW
2 to the parameters δρ
b
and seff . It comes from the renormalized vertex correction FL and the renormalization of the external
b

b-quark leg (the diagrams we denoted Md ,Me and Mf ) since the top quark can be involved in loops
without paying any suppression from the CKM matrix (being Vtb ≈ 1). The reason of this fact can be
very easily seen if we leave our confortable unitary gauge, in which the Goldstone bosons are eaten
up by longitudinal components of the physical massive gauge bosons. Before fixing the gauge we can
write in general the Higgs doublet as
√ + !
1 2G
ϕ= √ (10.121)
2 v + h + iG0

Clearly the Goldstone bosons will remain in the EW complete Lagrangian mixing with gauge bosons
and in particular to fermions, in the Yukawa sector. If one introduces gauge fixing terms in the
Lagrangian (dependent on the parameter ξ), then he finds that MG0 = ξMZ and MG+ = ξMW .
Important couplings of Goldstone bosons with W and Z bosons are

Wµ± p1 G±
ie
G∓ = −iesMZ ηµν Zµ =∓ (1 − 2s2 )(p1 + p2 )µ (10.122)
2sc
Zν p2 G±
10.5. ONE LOOP CORRECTIONS TO Z → F̄ F 149

Regarding the Yukawa sector, if one goes through all calculations he will find, for instance, the charged
Goldstone boson couplings to the t, b quark pair, which, quite interestingly, are proportional to mt /MW

b t
ie mt mb ie mt mb
   
G+ =√ PR − PL G+ =√ PL − PR
2s MW mt 2s MW mt
t b
(10.123)
In this gauge we can easily compute the leading term of the contribution of order m2t /MW 2 to the

parameters δρb and sbeff . The relevant vertex corrections are shown in the following diagrams

W b t b
M1 = Z t M2 = Z W
W b t b
W b G+ b
M3 = Z t M4 = Z t
G+ b W b
G+ b t b
M5 = Z t M6 = Z G+
G+ b t b
From the Feynman rules we have just written above, we get straightforwardly that only diagrams (5)
and (6) contain the factor m2t /MW 2 coming from the Yukawa coupling G+ tb of the Goldstone field

G . Diagrams (3) and (4) are finite and the loop integral goes as 1/mt , compensating the factors
+

mt /MW coming from the single G+ tb vertices. Finally diagrams (1) and (2) do not have any mt /MW
factors in their vertices and the potentially logarithmically divergent loop integrals can at most give
log(mt /MW ). The computation can be simplified in the heavy top quark limit mt  MW , MZ , i.e. for
vanishing external momenta. Also neglecting the bottom mass mb  MW  mt , setting Vtb = 1 and
choosing ξ = 1 we have,
ie (5) e3 1 − 2s2 m2t dd k 1 2k λ
Z
M5 = Λ = 2 P R P L
2sc L 2s 2sc MW 2 (2π)d k/ − mt (k 2 − MW 2 )2
(10.124)
ie3 1 − 2s2 m2t 3 m2t 1
  
λ
=− 2 3 2 − ∆ ε − + log 2 +O γ PL .
(4π) 8s c MW 2 MW mt
Similarly, for the diagram (6) we find

ie (6) e e2 m2t dd k 1 1
Z
M6 = ΛL = 2 d 2 PR γλ
2sc 2sc 2s2 MW 2
(2π) k − MW k/ − mt
4 4 1
  
× 1 − s2 PL − s2 PR PL
3 3 k/ − mt
e3 m2t 4 22 − d dd k 1
 Z
= 3 2 − s (10.125)
4s c MW 3 d (2π) (k − MW )(k 2 − m2t )
d 2 2

2 dd k m2t
 Z 
λ
+ 1 − s2 2 )(k 2 − m2 )2 γ PL
3 (2π)d (k 2 − MW t
ie e2 m2t 4 2 1 m2t 2
    
=− 2 3 s − ∆ε − + log + 2 1 − s2 γ λ PL .
(4π)2 8s3 c MW 2 µ2 3

Now we need to compute the full contribution of order m2t /MW2 to the renormalization factor δ ≈ Σ :
L L
this can be obtained by calculating the contribution of G± to the self energy of the bottom quark. In
150 CHAPTER 10. ELECTRO-WEAK PRECISION TESTS

general
M

m (10.126)
i(cL PL + cR PR ) i(c?L PL + c?R PR )

dd k i[(c?V − c?A γ5 )(k/ + p


/ + m)γ ν (cV − cA γ5 )] µ
Z
2 ε
−iΣ =(i) µ γ
(2π)d [(k + p)2 − m2 ][k 2 − M 2 ]
i 1 (10.127)
=− [a(M ) − a(m) + (p2 + m2 − M 2 )b0 (p2 , m, M )]p /(cL c?L PL + cR c?R PR )
(4π) 2π 2
2

+ m(cL c?R PL + cR c?L PR )b0 (p2 , m, M )


with
m2
 
a(m) = m2 − ∆ε − 1 + log (10.128)
µ2
and
1 x(x − 1)p2 + xm2 + (1 − x)M 2
Z
2
b0 (p , m, M ) = −∆ε + dx log (10.129)
0 µ2
In our case we are interested in
1
b0 (0, mt , MW ) = 2 [a(mt ) − a(MW )] (10.130)
m2t − MW

hence
1 e2 m2t 3 m2 1
   
(10.131)
±
ΣG
L (0) = 2 − ∆ε − + log 2t +O .
(4π)2 4s2 MW 2 µ mt
Finally, collecting all contributions together and replacing s, c with ŝ, ĉ, we get

2 α̂ m2t
   
(10.132)
±
2FL = 2 ΛL + 1 − ŝ2 ΣG
L + . . . ≈ − .
3 4πŝ2 ĉ2 MZ2

There is no similar contribution of order m2t /MW2 to F . Comparing with ∆ρ, we find that for bottom
R
quarks the new negative contribution of order m2t /MW 2 to the parameter δρ overcompensates the
b
positive universal contribution of the same order. As a result Γ(Z → bb) decreases as mt grows. The
same correction is also responsible for the slightly slower , compared to the effective Weinberg angles
for other fermions, decrease of s2eff with mt .

10.6 The S, T and U parameters


In the same way that looking for deviations of the ρ-parameter from 1 can tell us about custodial
SU (2) violating interactions, it is useful to have some additional ways to constrain and characterize
new physics (NP) theories. In this way EWPT can be able to test extensions of the SM. To this end,
the Peskin–Takeuchi parameters S, T and U are often used. These are defined as

1 Πnew
W W (0) Πnew (0) ρ−1
 
T ≡ 2 − ZZ 2 = ,
α MW MZ α
4c2 s2 Πnew 2 new 2
ZZ (MZ ) − ΠZZ (MZ ) c2 − s2 Πnew 2
Zγ (MZ ) Πnew 2
γγ (MZ )
 
S≡ − − (10.133)
α MZ2 cs MZ2 MZ2
4s2 Πnew 2 new 2
W W (Mw ) − ΠW W (MZ ) c Πnew 2
Zγ (MZ ) Πnew 2 
γγ (MZ )

U≡ − − −S,
α MZ2 s MZ2 MZ2

where α = α(MZ ). Here new means that S, T and U are normalized by subtracting off the SM
prediction, so that they account purely for new physics contributions:

Πnew tot SM
V V 0 = ΠV V 0 − ΠV V 0 . (10.134)
10.6. THE S, T AND U PARAMETERS 151

The point S = T = U = 0 is defined with mt = 173 GeV and Mh = 126 GeV. Current experimental
measurements give

S = 0.03 ± 0.10, T = 0.05 ± 0.12, U = 0.03 ± 0.10 . (10.135)

The actual allowed region is an ellipse, as shown in Fig. 10.1. Thus, if you propose a model of physics

0.5 0.5
T

T
68%, 95%, 99% CL fit contours, U free 68%, 95%, 99% CL fit contours, U=0
0.4 (SM : MH=126 GeV, m t =173 GeV) 0.4 (SM : MH=126 GeV, m t =173 GeV)
ref ref

0.3 0.3

0.2 0.2

0.1 0.1

0 0
SM Prediction SM Prediction
MH = 125.7 ± 0.4 GeV MH = 125.7 ± 0.4 GeV
-0.1 -0.1
mt = 173.18 ± 0.94 GeV mt = 173.18 ± 0.94 GeV
MH MH
-0.2 -0.2
SM Prediction SM Prediction
-0.3 with MH ∈ [100,1000] GeV -0.3 with MH ∈ [100,1000] GeV

-0.4 Sep 12
-0.4

Sep 12
B B
G fitter SM G fitter SM

-0.5 -0.5
-0.5 -0.4 -0.3 -0.2 -0.1 0 0.1 0.2 0.3 0.4 0.5 -0.5 -0.4 -0.3 -0.2 -0.1 0 0.1 0.2 0.3 0.4 0.5
S S

Figure 10.1: Contours of 68%, 95%, and 99% confidence level in the T S-plane. The reference point at
which all oblique parameters vanish is defined by MH =126 GeV and mt =173 GeV. The small black
line shows the SM prediction with the current precision on mt and MH , whereas the grey region shows
the SM prediction with no assumption on MH . In the plot on the left (right) U is left free (is set to 0)
in the fit.

beyond the Standard Model, you can calculate S and T as a shortcut to comparing with electroweak
precision data. In practice, S and T tend to give stronger constraints on BSM physics than U. T
measures custodial isospin violation, since it is equivalent to ρ. S would get a contribution, for example,
from a new generation of fermions, even if custodial isospin were preserved. It is often useful to think
about S and T as coming from higher-dimension operators. For example, suppose that the SM were
extended with the following SU (2)L ⊗ U (1)Y invariant 6-dimensional operators

OS = H † σ i HWµν
i
B µν , OT = |H † Dµ H|2 . (10.136)

At tree level we would get contributions to S and T proportional to the Wilson coefficients for these
operators. In practice, one can take one’s favorite model of NP, for example supersymmetry, integrate
out the new particle before breaking EW symmetry, and then look at the coefficients CS and CT of the
operators OS and OT that are generated by integrating out the new particle. Then
CS OS CT OT
LNP = 2
+ (10.137)
Λ Λ2
generates
4sc v 2 1 v2
C S T = S=
− CT . (10.138)
α Λ2 2α Λ2
It is often easier to use this shortcut than to compute the contributions of new physics to the vacuum
polarization graphs and electroweak precision observables directly.
152 CHAPTER 10. ELECTRO-WEAK PRECISION TESTS
Chapter 11

Flavor Changing Neutral Currents

The expression flavour–changing neutral current processes (FCNC) denotes semileptonic processes with
a change of flavour but without net charge transfer to the leptons, or non-leptonic processes which
should occur at higher order in the electroweak theory, as a consequence of the fact that the neutral
current does not produce a change of flavour to lowest order. Well known examples of FCNC processes
are the decay KL → µ+ µ− and the ∆S = 2 transition K 0 − K̄ 0 . In the electroweak theory, the FCNC
amplitudes due to exchange of the intermediate bosons, W and Z, should be finite, as a consequence
of the fact that it is not possible to add flavour–changing counterterms to the neutral current. The
convergence of the integration over internal momenta is ensured by the GIM mechanism, or by its
extension to the six quark theory. This has a consequence that the relevant internal momenta are
of order or larger than the mass of the charm quark. The process is therefore dominated by short
distance amplitudes and it is possible to calculate the weak contribution without taking account of
strong interaction corrections, as a result of asymptotic freedom of the colour interactions. If there are
no further contributions dominated by long distances, and thus affected by effectively non-calculable
corrections due to strong interactions, eventual discrepancies of experimental data with the amplitudes
calculated in the standard theory can give significant indications of new physics beyond the SM.
The FCNC processes dominated by short distances have in recent years become an effective
instrument to test the SM at energies not yet accessible to available accelerators. Despite its historic
role as an FCNC prototype, the KL → µ+ µ− decay does actually not lend itself to provide precision
information since the process can occur also via the KL → γγ → µ+ µ− channel, which is dominated
by long distance contributions. Experimentally:

KL → γγ ' 6 × 10−4 , KL → µ+ µ− ' 7 × 10−9 . (11.1)

Since the probability of γγ → µ+ µ− is of order (α/π)2 ' 4 × 10−6 , it can be seen that the process
through the intermediate γγ state can account for a significant part of the process amplitude. In this
chapter, we calculate in an illustrative way the weak amplitudes for transitions with a double change
of flavour, ∆F = 2, processes in which the weak amplitude is dominant and for which searches in
different laboratories are under way.
As we have already seen flavor changing currents are allowed at tree-level only in the CC sector of
the SM Lagrangian. An example of FCCC process is

c e−

W− (11.2)

b ν

FCNC are instead not allowed at tree-level: as we know in the NC sector the unitarity of the CKM
matrix leaves this part of the Lagrangian invariant under a rotation of all quarks of the same kind
(up-down). However, FCNC are allowed at the loop level thanks to CC interactions, as it is illustrated

153
154 CHAPTER 11. FLAVOR CHANGING NEUTRAL CURRENTS

by the following Feynman diagram


ui
b s

W W
ui
s b (11.3)

11.1 Meson-antimeson oscillation


In nature, it is observed that neutral psudoscalar mesons, M 0 , oscillate into their antiparticle, M 0 .
Such a meson-antimeson oscillation M 0 ↔ M 0 regards many physical systems:

mM 0 ≡mM 0 [MeV]
 
 K 0 = (sd) K 0 = (sd)
438
 
K0
 
 0  0
D = (uc) D = (uc)
 
1865
 
D0
 
M0 ≡ Bd0 = (bd) ←→ M 0 ≡ B 0d = (bd) (11.4)
  Bd0 5280
Bs0 = (bs) B 0s = (bs) 5367
 
Bs0

 


 

 
... ... ...
 

As we will see M 0 ↔ M 0 oscillation is induced by weak interactions at second order G2F . Notice that a
flavor violation by two units ∆F = 2 is implied. This phenomen is allowed only provided that M 0
and M 0 are not mass eigenstates, otherwise, precisely like in the neutrino case there is no oscillation.
Therefore also in our case there is a mixing matrix relating flavor eigenstates to mass eigenstates. A
difference between meson-antimeson oscillation and neutrino oscillation is that mesons can decay: we
have to take into account also the probability of the meson decay by a suitable suppression factor.
Thus the Hamiltonian H describing the meson-antimeson system is made of two pieces: a mixing
matrix M and and a decay matrix Γ , which was not there in the neutrino case:
i
H =M − Γ (11.5)
2
Both the mixing matrix and decay matrix are 2 × 2 hermitian matrices in the flavor basis
! !
1 0
0
|M i = 0
|M i = (11.6)
0 1

while, for construction the Hamiltonian is not


(
M = M†
=⇒ H 6= H † (11.7)
Γ = Γ†

Notice that this way of proceed is somehow related to the optical theorem in which we had a ”correction”
to the mass (the mixing matrix mixes mass eigenstates) from the decay width of the particle (related
to the decay matrix). Let us assume that at time t = 0 we have a state |ψ(0)i. The time evolution of
this state is given by the usual formula

(11.8)
1
|ψ(t)i = e−iH
Ht
|ψ(0)i = e−iM
M t − 2 Γt
e |ψ(0)i

where M describes the oscillation phenomenon while Γ is associated with the depletion of the survival
probability, P (t), of the initial state |ψ(0)i due to the decay

P (t) = |hψ(t)|ψ(0)i|2 = e−2Γt . (11.9)

We write the Hamiltonian in the matricial form


!
M11 − 2i Γ11 M12 − 2i Γ12
H = , (11.10)
M21 − 2i Γ21 M22 − 2i Γ22
11.1. MESON-ANTIMESON OSCILLATION 155

with
( Γ12 = Γ?21
M = M†
=⇒ M12 = M21 ? (11.11)
Γ = Γ† M11,22 , Γ11,22 ∈ R
Moreover, by CPT invariance in a non-relativistic framework, it can be proved that Γ11 = Γ22 = Γ and
M11 = M22 = M . If we decompose a generic state of the system at time t in the flavor eigenstates we
get
|ψ(t)i = a(t)|M 0 i + b(t)|M 0 i (11.12)
The evolution of the system is decribed by the Schroedinger equation
! ! ! ! !
d a(t) M − 2i Γ M12 − 2i Γ12 a(t) α β a(t)
i = i ? = (11.13)
dt b(t) ?
M12 − 2 Γ12 M − 2i Γ b(t) γ α b(t)

Let us diagonalize this matrix. It is easy to show that the matrix V diagonalizes our Hamiltonian
! ! √ !
α β µ 0 α + βγ 0
V −1H V = V −1 V = L
= √ , (11.14)
γ α 0 µH 0 α − βγ

with √ √ !
1 β γ
V =p √ √ (11.15)
|β| + |γ| β − γ
We will denote the normalized column vectors of the diagonalization matrix V as ML and MH where
L stands for ”light” and H for ”heavy”:
1 1
! ! ! ! ! !
M0 p q ML ML 1 M0
= =⇒ = p
1
p
(11.16)
M0 p −q MH MH 2 q − 1q M0

hence, straightforwardly

β

p = p|β| + |γ|



√ with |p|2 + |q|2 = 1. (11.17)
 γ
q = p


|β| + |γ|
We can study the evolution of the system in the basis in which the Hamiltonian is diagonal:
! ! !
d a(t) a(t) a(t)
i =H =VV −1
HV V −1
, (11.18)
dt b(t) b(t) b(t)

Defining ! ! ! !
c(t) a(t) a(t) c(t)
≡V −1
, =V , (11.19)
d(t) b(t) b(t) d(t)
we obtain ! ! ! ! !
d c(t) µL 0 c(t) c(t) e−iµL t c(0)
i = =⇒ = (11.20)
dt d(t) 0 µH d(t) d(t) e−iµH t d(0)
and
1
  
β e−iµL t c(0) + e−iµH t d(0) |M 0 i
p
|ψ(t)i = p
|β| + |γ|
(11.21)

  
−iµL t −iµH t 0
+ γ e c(0) − e d(0) |M i .

The analytic expression of the two eigenvalues is


s
i  i ? − i Γ? i
(11.22)
 
µL,H = M− Γ ± M12 − Γ12 M12 12 = mL,H − ΓL,H
2 2 2 2
156 CHAPTER 11. FLAVOR CHANGING NEUTRAL CURRENTS

and so
i
∆µ = µH − µL = ∆m + ∆Γ (11.23)
2
where (
∆m = mH − mL
(11.24)
∆Γ = ΓL − ΓH
are observables, i.e. measurable quantities. Note that ∆m is positive by definition while ∆Γ can have
either sign. Experimentally the sign of ∆Γ is only known for Kaons and the sign convention in Eq.
(11.24) corresponds to ∆ΓK positive. The SM prediction for ∆ΓBd and ∆ΓBs is also positive, while no
reliable prediction is possible for the sign of ∆ΓD .
Now the natural question arising concernes the phases of the Hamiltonian: do we have some phase
associated with CP violation? The answer turns out to be yes. Experimentally we find that in the
case of Bd,s ↔ B d,s we have |Γ12 |  |M12 |, which implies

mL,H ' M ± |M12 | + O |Γ12 |


    

M12
 
|M12 | φ = arg − , (11.25)
 Γ12
ΓL,H ' Γ ∓ |Γ12 | cos φ

therefore,
(
∆m ' 2|M12 |
(11.26)
∆Γ ' 2|Γ12 | cos φ
this equation highlights which are the three physical quantities to be considered in the next of our
discussion to probe the possible presence of CP violation. Up to now, an experimental evidence of
∆Γ 6= 0 or ∆m 6= 0 is not an evidence of CP violation: if the CP violating phase were zero we could
nevertheless obtain ∆Γ 6= 0 and ∆m 6= 0. Let us rewrite the Hamiltonian in the form
0
! ! !
H H12 M |M12 |e2iθ i Γ |Γ12 |e2iθ
H = = − 0 . (11.27)
H21 H |M12 |e−2iθ M 2 |Γ12 |e−2iθ Γ

If we apply a CP transformation to our flavor eigenstates we simply transform the meson in an


antimeson (and viceversa) plus a phase α

CP iα
|M 0 i −→
! ! ! !
e |M 0 i M0 0 eiα M0 M0
(11.28)
CP
=⇒ −→ =U
CP
|M 0 i −→ e−iα |M 0 i M0 e−iα 0 M0 M0

Moreover the Hamiltonian operator transforms as


!
H H21 e2iα
(11.29)
CP †
H −→ U H U = −2iα
H12 e H

If we impose the CP conservation so that H = U H U † we obtain the two conditions


(
H12 = H21 e2iα
(11.30)
H21 = H12 e−2iα

and using M12 = M21


? and Γ
12 = Γ21 ,
?

i i
(11.31)
0 0
|M12 |e2iθ − |Γ12 |e2iθ = |M12 |e−2i(θ−α) − |Γ12 |e−2i(θ −α) ,
2 2
hence (
e2iθ = e−2i(θ−α)
0 0 =⇒ α = 2θ = 2θ0 (mod π) CP conservation (11.32)
e2iθ = e−2i(θ −α)
11.2. FCNC IN THE SM AND THE GIM MECHANISM 157

Now let us consider the quantity


 2 0
q ? − i Γ?
M12 2 12 |M12 |e−2iθ − 2i |Γ12 |e−2iθ
= =
p M12 − 2i Γ12 |M12 |e2iθ − 2i |Γ12 |e2iθ0
0 (11.33)
|M12 | + 2i |Γ12 |e−2i(θ −θ)+iπ
 
= e−4iθ
|M12 | + 2i |Γ12 |e2i(θ0 −θ)−iπ
 

Now we recognise that the angle φ we defined above is

M12 |M12 |e2iθ


   
φ = arg − = arg − = 2(θ − θ0 ), (11.34)
Γ12 |Γ12 |e−2iθ0

hence, for the CP conservation condition derived above, we discover that φ 6= 0 (mod π) implies a CP
violation. Therefore φ is responsible for CP violation in our meson-antimeson system. We can expand
Eq. (11.33) for small φ being the CP violation small, as suggested experimentally1 .
 2
q |Γ12 |
 
= e−4iθ 1 − φ + O(φ2 ) (11.35)
p |M12 | + 2i |Γ12 |

To get rid of the unphysical phase factor in front of this equation we can consider the square modulus,
still in the φ  1
q 2 |M12 ||Γ12 |
=1− 2 1 2
φ + O(φ2 ) (11.36)
p |M12 | + 4 |Γ12 |
Notice that this quantity only depends on physical measurable quantities hence is the observable
related to CP violation we were looking for.

11.2 FCNC in the SM and the GIM mechanism


Now we want to understand the meson-antimeson oscillation in a QFT framework. The mass difference
∆m = 2|M12 | + . . . arises at second order in GF . Let us focus for now on the Bd0 ↔ B 0d case. We have
two diagrams contributing to this process

Vib Vid? Vib Vid?


b d b ui d
W−
ui uk W− W+ (11.37)
W+ uk
d ? b d ? b
Vkd Vkb Vkd Vkb
(a) (b)

The Hamiltonian describing this ∆F = 2 process is, exploiting the effective Fermi theory, justified by
the fact that mB  MW ,
  
H(∆F = 2) = −L(∆F = 2) = const × dL γ bL µ
dL γµ bL (11.38)

The overall constant, which must have mass dimension −2, can be estimated by NDA:
!
G2 X m2i
const ∼ F2 (Vib Vid? )2 f 2 , (11.39)
16π i=t,c,u MW

where f is a function of the masses of the particles involved in the loops with mass dimension
2. Now suppose m2i  MW 2 which indeed is false if the top quark is involved in the loop. If so,

1
Interestingly, CP violation in mixing is small for the K, Bd and Bs systems. For D ↔ D mixing this is most likely
also the case, but the experimental data are not accurate enough at present
158 CHAPTER 11. FLAVOR CHANGING NEUTRAL CURRENTS

2 ) ≈ f (0) ∼ M 2 since M 2 is the relevant mass scale of the loop. Therefore f (0) would go
f (m2i /MW W W
out from the sum, being independent on mi and
mi MW G2F
(11.40)
X
const −→ f (0) (Vib Vid? )2 = 0
16π 2 i=t,c,u
| {z }
=0, V V † =1

Hence the constant has to feel the quark masses, otherwise it will vanish. A consequence of this is that
the dominant contribution will be given by the largest mass scale, which is the top one; moreover the
CKM coupling is maximized in case of the top. We can safely estimate
G2F 2
const ∼ m ηB (11.41)
16π 2 t
with η ' 0.55 a QCD correction factor. Now let us study the degree of superficial divergence of our
loop diagrams
1 1 k→∞ 2 2
Z
M(Bd ↔ B 0d ) ∼ G2F (Vib Vid? )(Vkb Vkd
?
) d4 k −→ GF ΛB (11.42)
k/ k/
Being the integral UV divergent it is necessary to introduce a UV cut-off scale ΛB to make the result
finite and in agreement with experiments which measure ∆mB . Notice that in our case the divergence
is only due to the fact that we are considering the Fermi theory which is valid only if k 2  MW 2 : this

is manifest in the appeareance of GF instead of MW . Therefore ΛB will be a cut-off scale at which our
EFT ceases to be valid: in particular ΛB will be associated to the presence of new degrees of freedom,
i.e. new particles . We can then fit the experimental result to obtain the value of ΛB . It turns out that
M(Bd ↔ B 0d ) ∼ G2F Λ2B =⇒ ΛB ∼ 170 GeV ≈ mt (11.43)
Thus we have an indirect (by a loop induced process) prediction of the mass of the top quark, in order
to agree with the experiments of the meson-antimeson oscillation. The discussion can be repeated in
the case of the kaons K 0 ↔ K 0 oscillation.

Vis Vid? Vis Vid?


s d s d
W− ui
ui uk W− W+ (11.44)
W+ uk
d ?
s d ?
s
Vkd Vks Vkd Vks
(a) (b)

This process allowed Glashow, Iliopoulos and Maiani to predict the charm quark mass (GIM mechanism)
throught the same argument we have followed above, in a four quarks (u, d, s, c) framework:
M(K 0 ↔ K 0 ) ∼ G2F Λ2K =⇒ ΛK ∼ 1.5 GeV ≈ mc (11.45)
Now let us compute, in the full SM framework, the loop amplitudes setting all the external momenta
to zero. Let q be the internal momentum. The fermion line of the (b) diagram corresponds to v d Sµν (q)us
where
a b c
 
Sµν (q) = γµ (1 − γ5 ) + + γν (1 − γ5 ) (11.46)
(q/ − mu ) (q/ − mc ) (q/ − mt )
having defined
?
a ≡ Vud Vus ; ?
b ≡ Vcd Vcs ; c = Vtd? Vts (11.47)
which by unitarity of the CKM matrix provide a + b + c = 0. Setting mu = 0, in an excellent
approximation, and exploiting a = −b − c, we get
1 1 1 1
    
Sµν (q) =γµ (1 − γ5 ) b − +c − γν (1 − γ5 )
/q − mc /q /q − mt /q
(11.48)
m2c m2t
 
=2γµ /q b 2 2 +c 2 2 γν (1 − γ5 )
q (q − m2c ) q (q − m2t )
11.2. FCNC IN THE SM AND THE GIM MECHANISM 159

Notice that exploiting CKM unitarity we have increased the UV degree of convergence of the loop
integral by 2, since S passed from being S ∝ 1/q/ to S ∝ /q m2 /q 4 ∼ 1/q 3 . Now we can compute the
amplitude in the full SM theory
!2 Z
g2 d4 q 1 qµqρ qν qσ
  
M(a) = 4
Sµν (q) 2 2 )2 − η µρ + 2 − η νσ + 2 Sσρ (q)
8 (2π) (q − MW MW MW
!2 Z
g2 d4 q   
=4 4
v d γµ /q γν (1 − γ5 )us ud γσ /q γρ (1 − γ5 )vs
8 (2π) (11.49)
m4 m2c m2t m4t
 
× b2 4 2 c 2 2 + 2bc 4 2 + c2
q (q − mc ) q (q − m2c )(q 2 − m2t ) q 4 (q 2 − m2t )2
qµqρ qν qσ 1
  
× − η µρ + 2 − η νσ + 2 2 2 )2 .
MW MW (q − MW
Exploiting
1
Z Z
d4 qq α q β F (q 2 ) = η αβ d4 qq 2 F (q 2 ), (11.50)
4
and contracting2

γµ γα γν (1 − γ5 ) ⊗ γ ν γ α γ µ (1 − γ5 ) = 4γµ (1 − γ5 ) ⊗ γ µ (1 − γ5 ), (11.51)
terms proportional to η µρ η νσ give
!2
g2 d4 q 
Z
µ  
M(a) = ·4 v d γ (1 − γ 5 )us ud γ µ (1 − γ 5 )v s
8 (2π)4
m4 m2c m2t m4t (11.52)
 
× b 4 2 c 2 2 + 2bc 4 2
2
+ c2
q (q − mc ) q (q − m2c )(q 2 − m2t ) q 4 (q 2 − m2t )2
q2 1 q4 1
 
× 1−2 2 + 2 2 )2 .
MW 4 MW (q 2 − MW
Computing the integral with standard tecniques by making use of master integrals, we obtain
G2F MW2
Ci Cj E(xi , xj ) v d γLµ us ud γµL vs (11.53)
X   
M(a) = −i
8π 2 i,j=c,t

where xi ≡ m2i /MW


2 and C = V ? V and
i id is
xt
E(xt , xt ) ≡ F (xt ), E(xc , xc ) = xc , E(xc , xt ) = −xc log xc (11.54)
2
where
1 3xt 1 ln xt 3 xt 1 2 ln xt
   
F (xt ) = + − + − 1+ − (11.55)
2 (xt − 1) xt (xt − 1) 2 (xt − 1)2 xt (xt − 1)
with F (xt ) ≈ 1.1. Similarly, for the diagram (b), we obtain the amplitude
G2F MW 2
Ci Cj E(xi , xj ) v d γLµ vs ud γµL us (11.56)
X   
M(b) = −i 2
8π i,j=c,t

The amplitudes M(a,b) can be obtained as matrix elements to perturbative first order in the effective
Lagrangian Lef f

G2F MW2
¯ µ (1 − γ5 )s (11.57)
X
Ci Cj E(xi , xj ) dγ µ (1 − γ5 )s dγ
 
Lef f = − 2
16π i,j=c,t
2
The identities γ ν γ α γ µ = g να γ µ +g αµ γ ν −g νµ γ α −iναµρ γρ γ5 and γµ γα γν = gνα γµ +gαµ γν −gνµ γα −iµανρ γ ρ γ5 when
combined give γµ γα γν (1 − γ5 ) ⊗ γ ν γ α γ µ (1 − γ5 ) = 10γρ (1 − γ5 ) ⊗ γ ρ (1 − γ5 ) − iµανρ γµ γα γν (1 − γ5 ) ⊗ γρ (1 − γ5 ). Moreover,
iµανρ γµ γα γν (1−γ5 )⊗γρ (1−γ5 ) = −6γ σ (1−γ5 )⊗γσ (1−γ5 ) and therefore we finally get γµ γα γν (1−γ5 )⊗γ ν γ α γ µ (1−γ5 ) =
4γµ (1 − γ5 ) ⊗ γ µ (1 − γ5 ).
160 CHAPTER 11. FLAVOR CHANGING NEUTRAL CURRENTS

Now the symbols s and d¯ denote the quark fields. If we take the matrix element of eq. (11.57) between
the physical external states and apply the contraction rules of the operators in the Lagrangian with
the creation operators of the particles or antiparticles in the initial and final states, we find a factor
two relative to the choice of the s field with which to annihilate the initial strange quark. There are
two possibilities according to the choice of annihilating the anti-d of the initial state with the d¯ field in
the same covariant to which the s field belongs, or in the other. In the first case the amplitude M(a) is
found for the diagram (a). In the second the amplitude M(b) for the diagram (b).

11.3 The hierarchy problem


As we have seen, the removal of divergences in flavor physics lead to the spectacular prediction (before
their discovery!) of the charm and top quarks with masses of the right order. A similar situation seem
to occurs also in the Higgs sector of the SM. Indeed, loop corrections to the Higgs boson mass, Mh ,
exhibit quadratic divergences

k
11 Λ2
Z
δMh2 ∼ h h ∼ d4 k ∼ . (11.58)
k/ k/ 16π 2
k

Without any yet undiscovered new physics, the only energy scale at which we know our description
must break down is the gravitational scale MPl ∼ 1018 GeV. Thus, the Higgs mass is pushed up to
1018 GeV which is way heavier than the observed value of 125 GeV. Although these enormously large
loop corrections can be reabsorbed in principle through the renormalization procedure (by adding
an appropriate counter-term), the required huge cancellation between loop-effects and counter-terms
makes this possibility very unplausible or unnatural. This naturalness problem is a generic feature of
theories containing fundamental scalars and is referred to as the hierarchy problem.
The solution to it generally requires the SM to break down at the TeV scale where new particles are
supposed to show up. There are two popular directions. One is Supersymmetry (SUSY). In a SUSY
framework, the SM particles are doubled and the loop contributions of SM particles are cancelled by
those from their super-partners. Supersymmetry has deep connections to an extension of space-time
symmetry since it relates particles with different spin, however no evidence of SUSY has been found
yet. The other solution assume that the Higgs no longer behaves like a scalar particles at the TeV
scale, but rather as a bound state of two fermions. This is precisely what happens with the mesons:
even though the pion is a scalar, there is no pion ‘hierarchy problem’ because as you probe smaller
distances, you realize the pion is actually a bound state of two quarks and it starts behaving as such.
Chapter 12

Anomalies

Most of the time, a symmetry of a classical theory is also a symmetry of the quantum theory based on
the same Lagrangian. When it is not, the symmetry is said to be anomalous. Since symmetries are
extremely important for determining the structure of a theory, anomalies are also extremely important.
If a symmetry is anomalous then it is not actually a symmetry and the associated current will not
be conserved. Such a situation has dire consequences for theories in which the current couples to
a massless spin-1 particle, such as QED or Yang–Mills theory. If the current to which a massless
spin-1 particle couples is not conserved, the Ward identity will be violated, unphysical longitudinal
polarizations can be produced, and unitarity will be violated. Thus, in a unitary quantum theory,
gauged symmetries (those with associated massless spin-1 particles) must be anomaly free. It turns
out that this is a strong requirement for a consistent quantum theory.
Before analyzing this interesting subjects let us recall some important symmetries of the SM. By
Noether’s theorem, for a system described by a Lagrangian density , any continuous

L φ(x), ∂ µ φ(x)
global symmetry transformation which leaves the action S = d xL invariant implies the existence of
R 4

a conserved current and a conserved charge:


Z
µ
∂µ J (x) = 0 =⇒ Q(t) = d3 xJ0 (x) = constant (12.1)

This is true only at the classical level. We have two types of symmetries.
1. Global symmetries. Examples of global symmetries are the U (1) accidental symmetry associ-
ated to the conservation of the baryon number B and the lepton family numbers Li (i = e, µ, τ ).
These symmetries are examples of global anomalies. Actually global anomalies do not lead to
inconsistencies (in this sense the phrase anomaly free refers to the absence of gauge anomalies).
In fact there is no massless spin-1 particle in the SM that couples to the classically conserved
JBµ . Indeed, baryon number violation is a necessary condition to explain the preponderance
of matter over antimatter in the Universe as pointed out studying the Sakharov conditions.
However, remarkably, the difference between baryon number and total lepton number, B − L, is
non-anomalous, as we shall check.
The QCD chiral symmetries provide another example of global anomalies. Indeed, in the
mu = md = 0 limit, LQCD is invariant under U (2)L ⊗U (2)R ' U (1)V ⊗U (1)A ⊗SU (2)L ⊗SU (2)R :
while U (1)V is related to the baryon number symmetry that is conserved in QCD also at quantum
level1 , the U (1)A symmetry is anomalous. Global anomalies help us to explain the so-called
U (1)A problem of QCD, that is why the η 0 meson is so heavy, mη0 ≈ 900 MeV  mπ ≈ 140
MeV2 . In fact, the η 0 mass receives dominant contributions from anomalies and not from the
quark masses because U (1)A is anomalous and therefore the symmetry breaking is not associated
to a ”massless” Goldstone boson.
2. Gauge symmetries. Examples of local (gauge) symmetries are the general SM gauge group
G = SU (3)c ⊗ SU (2)L ⊗ U (1)Y or the simple U (1)em of QED. Gauge anomalies concern the
1
As we shall see, in the full SM theory including weak interactions, the baryon number is anomalous.
2
As we have already seen, the three pions π 0 , π ± are the ”Goldstone bosons” associated with the vectorial QCD
symmetry SU (2)V .

161
162 CHAPTER 12. ANOMALIES

breaking of a gauge symmetry at the quantum level. This is not acceptable as the theory would
become inconsistent. For example if we take the QED in the interacting case where a spin-1
massless particle couples to a fermion current,

Aµ = eJµem with ∂µ Aµ = 0 (gauge symmetry) (12.2)

Gauge symmetry implies ∂ µ Aµ = 0 but if at the quantum level there is an anomaly

e∂ µ Jµem 6= 0 (12.3)

there is an inconsistency: gauge symmetry is violated and, with it, also the Ward identity which
is crucial for the renormalizability of the gauge theory. That is why in a unitary quantum theory,
gauged symmetries have to be anomaly free. We will check this in the following.

12.1 Axial-vector current anomaly and π 0 → γγ


Out of the most important physical consequences of anomalies, we will consider now the decay of a
pion in two photons π 0 → γγ which stem from the U (1)A anomaly of QED. This process is described
by the following Feynman diagram

π0
γ
(12.4)

where in the fermion-loop both up- and down-quarks can circulate. The actual calculation of the decay
rate gives the following result
2 m3
αem Nc
Γth (π 0 → γγ) = π
· Nc · Q2u − Q2d = 7.77 eV × (12.5)
 
3 2
64π fπ 3

and therefore
Γth (π 0 → γγ) Nc
0
≈ =⇒ Nc = 3 . (12.6)
Γexp (π → γγ) 3
Therefore this process provides an important way to measure the colour number of quarks.

Axial and vector currents in QED


Let us start from the general QED Lagrangian
1
LQED = − Fµν F µν + iψγ µ (∂µ + ieAµ )ψ − mψψ (12.7)
4
As we know this Lagrangian is invariant under a U (1)em gauge symmetry

0 −iα(x)
ψ (x) = e
 ψ(x)
Gauge U (1)em : 1 =⇒ Jµem = ψγµ ψ with ∂ µ Jµem = 0 (12.8)
A0µ (x) = Aµ (x) + ∂µ α(x)

e
This symmetry is anomaly free, and this is crucial for the conservation of the electric charge, which is
a well experimentally verified fact.
Now instead consider a U (1)A global axial symmetry, acting on matter fields

ψ 0 (x) = e−iβγ5 ψ(x)


( (
∂β δψ = −iβγ5 ψ
Global U (1)A : 0 −iβγ5
with = 0 =⇒ (12.9)
ψ (x) = ψ(x)e ∂x δψ = ψ(−iβγ5 )
12.2. ABJ ANOMALIES 163

Hence we compute the variation of the relevant pieces of the QED Lagrangian

δ ψγ µ ψ = δψγ µ ψ + ψγ µ δψ = −iβψ γ5 γ µ + γ µ γ5 ψ = 0 (12.10)


  
| {z }
{γ µ ,γ5 }=0

(12.11)

δ ψψ = δψψ + ψδψ = −iβψγ5 ψ − iβψγ5 ψ = −2iβψγ5 ψ 6= 0
Therefore
δLQED = +2imβψγ5 ψ = 0 ⇐⇒ m = 0 (12.12)
the U (1)A is a global symmetry of the QED Lagrangian (at the classical level) only if we consider
massless fermions. In the latter case, the conserved axial-vector current associated to the symmetry is

JµA = ψγµ γ5 ψ with ∂ µ JµA = 0 (12.13)

If instead m 6= 0 we can perform the variation of the action and exploit the equations of motion finding

∂LQED
Z Z  
4 4
δSQED =δ d xLQED = d x∂µ δψ
∂∂µ ψ
Z Z Z (12.14)
4 4  4 µ 
= d xδLQED = d x 2imβψγ5 ψ = d x∂µ βψγ γ5 ψ .

Therefore one finds ∂ µ JµA = 2imψγ5 ψ. We will prove that, including quantum corrections we obtain

e2 µνρσ
∂ µ JµA = 2imψγ5 ψ + ε Fµν Fρσ (12.15)
16π 2
The second term is the anomalous quantum correction which survives also in the symmetric classical
limit m = 0. Moreover the quantum correction is parity violating as one can immediately see by

(12.16)
~ ·B P
~ −→ ~ ·B
~
εµνρσ Fµν Fρσ ∼ E −E

12.2 ABJ anomalies


As we shall see, the validity of the axial Ward identity is not automatic when there are fermions in the
theory. This is because certain 1-loop diagrams (such as the triangle diagrams) introduce anomalous
terms which prevent the Ward identities from reproducing themselves recursively at higher orders
in the perturbative expansion. Such anomalies were discovered by Adler, Bell and Jackiw in their
current-algebra studies. We shall present an elementary introduction to this subject of ABJ anomalies.
Consider the three-point functions in QED:
Z
d4 x1 d4 x2 h0|T Vµ (x1 )Vν (x2 )Aλ (0) |0ieik1 ·x1 +ik2 ·x2

Tµνλ (k1 , k2 , q) =i
Z (12.17)
4 4  ik1 ·x1 +ik2 ·x2
Tµν (k1 , k2 , q) =i d x1 d x2 h0|T Vµ (x1 )Vν (x2 )P (0) |0ie

where Vµ , Aµ and P are the vector, axial vector, and pseudoscalar currents, respectively

Vµ (x) = ψ(x)γµ ψ(x),


Aµ (x) = ψ(x)γµ γ5 ψ(x), (12.18)
P (x) = ψ(x)γ5 ψ(x),

and q = k1 + k2 by conservation of momentum. To obtain the Ward identities relating Tµνλ and Tµν ,
we need the divergence of Vµ and Aµ which are calculated from the equation of motion, i.e. at the
classical level
∂ µ Vµ (x) = 0
(12.19)
∂ µ Aµ (x) = 2imP (x)
164 CHAPTER 12. ANOMALIES

as we have proved above. In general for a current Jµ (x) and a local operator O(y) the following identity
holds
∂xµ T(Jµ (x)O(y)) = T(∂ µ Jµ (x)O(y)) + [J0 (x), O(y)]δ(x0 − y0 ). (12.20)


In our case the equal-time commutators vanish [V0 (x), A0 (y)]δ(x0 − y0 ) = 0 and we can formally derive
the following vector and axial-vector Ward identities:

k1µ Tµνλ = k2ν Tµνλ = 0


(12.21)
q λ Tµνλ = 2mTµν

The lowest order contributions to the three-point functions are already at 1-loop. We write

k1 k2
p µ p ν
q q
Tµνλ : λ p − k1 + λ p − k2

p−q ν p−q µ
k2 k1
k1 k2
p µ p ν
q q
Tµν : λ p − k1 + λ p − k2

p−q ν p−q µ
k2 k1

Notice that in the left diagrams we have taken into account the boson exchange in the final state
exchanging k1 ↔ k2 and µ ↔ ν and setting the relative sign between diagrams to +. We find that
!
d4 p i i i µ↔ν
Z   
Tµνλ = i (−1) Tr γλ γ5 γν γµ + (12.22)
1 ↔ k2
(2π) 4 /−m
p / − /q − m p
p / − k/1 − m k
!
d4 p i i i µ↔ν
Z   
Tµν = i (−1) Tr γ5 γν γµ + (12.23)
(2π) 4 /−m p
p / − /q − m p / − k/1 − m k1 ↔ k2

To check the Ward identities, in particular q λ Tµνλ = 2mTµν , we can use the relation

/ − /q − m) + (p
/q γ5 = γ5 (p / − m)γ5 + 2mγ5 , (12.24)

to find that
q λ Tµνλ = 2mTµν + ∆(1) (2)
µν + ∆µν , (12.25)
with
d4 p i i i i
Z  
∆(1)
µν = 4
Tr γ5 γν γµ − γ5 γν γµ , (12.26)
(2π) /−m
p / − k/1 ) − m
(p / − k/2 ) − m
(p / − /q ) − m
(p
!
µ↔ν
∆(2)
µν = ∆(1)
µν . (12.27)
k1 ↔ k2

If both integrals ∆µν and ∆µν vanish the Ward identity is satisfied. Naively this appears to be the
(1) (2)

case. The two terms of ∆µν cancel each other if we can shift the integration variable p → p + k2
(1)

in the second term. Similarly, the two terms of ∆µν would cancel with the shift p → p + k1 in the
(2)
12.2. ABJ ANOMALIES 165

second term. But the integrals are linearly divergent and a translation of integration variable produces
extra finite terms so that ∆µν 6= 0 and ∆µν 6= 0. This spoils the Ward identity. This is a very subtle
(1) (2)

point that confused many people for a long time. The most obvious way to make a divergent integral
well-defined is to introduce a regulator. Unfortunately, none of our favorite regulators will work. For
example, dimensional regularization has trouble with γ5 since chiral fermions are a feature of four
dimensions. One can use dimensional regularization, but it is very subtle. Pauli–Villars, which would
introduce a heavy fermion, will not work either, since the fermion mass explicitly breaks the symmetry
we are trying to verify (remember that the RHS of the Ward identity is proportional to the mass of the
fermion involved in the loop). Instead, we proceed by trying to make sense of the linearly divergent
integrals directly.
Consider the one-dimensional integral
Z +∞
∆(a) = dx [f (x + a) − f (x)], (12.28)
−∞

where the function f (x) goes to a contant at x = +∞ and a different constant at x = −∞. Then each
terms is linearly divergent, and we would like to know if the difference is finite or infinite. If we are
allowed to shift x → x − a on the first term, then ∆(a) vanishes at the level of the integrand. On the
other hand if we Taylor expand around a = 0, we find
+∞ a2
Z  
dx af (x) + f 00 (x) + . . . = a f (+∞) − f (−∞) ,
0
(12.29)
 
∆(a) =
−∞ 2
where the higher-derivative terms do not contribute since f (±∞) = const. Thus the difference between
a linearly divergent integral and its shifted value has a linear dependence on the shift a. Therefore the
shift x → x + a can be performed only if f (±∞) = f 0 (±∞) = · · · = 0 i.e. the surface terms are zero:
this is the case if the integral −∞ dx f (x) converges (or at most diverges logarithmically) but not if it
R∞

diverges linearly. To generalize to four dimensions, we can do the same thing. In this case, we will
need to evaluate integrals such as
d4 k
Z
∆α (a) = (f α (k + a) − f α (k)) . (12.30)
(2π)4
Wick rotating with r ≡ kE this is
d4 r
Z
∆α (a) = i (f α (r + a) − f α (r)) . (12.31)
(2π)4
Taylor expanding the integrand around a = 0, we get
d4 r 1
Z  
α
∆ (a) = i aµ ∂µ f α (r) + aµ aν ∂µ ∂ν f α (r) + . . . . (12.32)
(2π)4 2
These derivative terms can then be integrated using Gauss’s theorem . Since the integral is supposed
to be linearly divergent, at large r our function f (r) must scale as

lim f α (r) = A , (12.33)
r→∞ r4
with A a proportionality constant. Therefore, everything but the term with one derivative vanishes
too fast at infinity to contribute. To evaluate the one-derivative term, we write it as a surface integral:
d4 r d3 Sµ α
Z Z
α
∆ (a) = ia µ
∂µ f α (r) = iaµ f (r) (12.34)
(2π)4 (2π)4

The surface element d3 Sµ is a 4-vector normal to the surface of a 4-sphere at r2 = R = ∞. So it can
be written as d3 Sµ = R2 Rµ dΩ4 . Thus, using Rα Rµ = R2 δµα /4 and Ω4 = 2π 2 ,

dΩ4 2 R2 Rµ f α (R) i
Z
α
∆ (a) = ia µ
lim R R µ f α
(r) = 2iπ 2 µ
a lim = Aaα . (12.35)
R→∞ (2π)4 R→∞ (2π)4 32π 2
166 CHAPTER 12. ANOMALIES

This is a general result: linearly divergent integrals that would vanish if we could perform the shift are
finite, with the result proportional to the necessary shift.
The one loop amplitude Tµνλ is superficially linearly divergent; hence it is not uniquely defined.
We chose a particular routing of the loop momentum p: the fermion line between the vector and axial
vector vertices carries momentum p. We could have chosen to route it differently so that this fermion
line carries p + a(k1 , k2 ), with a some (arbitrary) linear combination of k1 and k2 :

a = αk1 + (α − β)k2 . (12.36)

The fact that the integral is linearly divergent implies that Tµνλ has an ambiguity in its definition by
an amount
!
d4 p µ↔ν
Z  
∆µνλ (a) =Tµνλ (a) − Tµνλ (0) = (−1) Cµνλ (a) − Cµνλ (0) +
(2π)4 k1 ↔ k2 (12.37)
(1) (2)
=∆µνλ + ∆µνλ .

with
1 1 1
 
Cµνλ (a) = Tr γλ γ5 γν γµ , (12.38)
(p /) − m
/+a / − /q ) − m (p
/+a
(p / − k/1 ) − m
/+a

1 1 1
 
Cµνλ (0) = Tr γλ γ5 γν γµ . (12.39)
/−m
p / − /q ) − m (p
(p / − k/1 ) − m
Applying the result of Eq. (12.35), we have

d4 p ρ ∂ 1 1 i
Z  
(1)
∆µνλ =(−1) 4
a ρ
Tr γλ γ5 γν γµ
(2π) ∂p /−m
p / − /q ) − m (p
(p / − k/1 ) − m
−i2π 2 aρ pα pβ pδ
= lim p 2
p ρ Tr[γ α γ λ γ 5 γ β γ ν γ δ γ µ ] (12.40)
(2π)4 p→∞ p6
2iπ 2 aσ pσ pρ 1
= lim 4iεµνλρ = 2 ερµνλ aρ .
(2π)4 p→∞ p2 8π

Since ∆µνλ is related to ∆µνλ by exhanges k1 ↔ k2 and µ ↔ ν, we obtain, exploiting Eq. (12.36),
(2) (1)

β
(12.41)
(1) (2)
∆µνλ = ∆µνλ + ∆µνλ = ερµνλ (k1 − k2 )ρ .
8π 2
Thus the definition of Tµνλ has an ambiguity signified by the arbitrary parameter β:

β
Tµνλ (a) = Tµνλ (0) − εµνλρ (k1 − k2 )ρ ≡ Tµνλ (β). (12.42)
8π 2
We now attempt to determine β by imposing the Ward identities. We shall see that no value of β
exists such that Tµνλ satisfies both the vector and axial-vector Ward identities. Let us first check the
axial Ward identity. Like those in Eq. (12.37) the surface terms in Eq. (12.25) can be evaluated using
Eq. (12.35):
k2ρ ∂ Tr[(p / − k/1 + m)γµ
/ + m)γ5 γν (p
Z  
∆(1)
µν = − d4
p
(2π)4 ∂pρ (p2 − m2 )[(p − k1 )2 − m2 ]
ρ (12.43)
k2 2 pρ α β 1 σ ρ
= 2iπ lim 2 Tr[γα γ5 γν γβ γµ ]p k1 = − 2 εµνσρ k1 k2 ,
(2π)4 p→∞ p 8π

and therefore, turning out that ∆µν = ∆µν , we have


(2) (1)

1−β
q λ Tµνλ (β) = 2mTµν (0) − εµνσρ k1σ k2ρ . (12.44)
4π 2
12.2. ABJ ANOMALIES 167

For the vector Ward identity we have

d4 p 1 1 1
Z   
k1µ Tµνλ (0) =(−1) Tr γλ γ5 γν k/
(2π) 4 /−m
p / − /q ) − m (p
(p / − k/1 ) − m 1
(12.45)
1 1 1
 
+Tr γλ γ5 k/ γν .
/−m
p / − /q ) − m 1 (p
(p / − k/2 ) − m

Using
/ − m) − [(p
k/1 = (p / − k/1 ) − m] = [(p
/ − k/2 ) − m] − [(p
/ − /q ) − m], (12.46)
we can rewrite

d4 p 1 1
Z  
k1µ Tµνλ (0) =(−1) 4
Tr γλ γ5 γν
(2π) / − /q ) − m (p
(p / − k/1 ) − m
1 1

− γλ γ5 γν
(p/ − k/2 ) − m p /−m
ρ Z (12.47)
k ∂ Tr[γλ γ5 (p / − k/2 + m)γν (p / + m)]

= 1 4 d4 p
(2π) ∂pρ [(p − k2 ) − m ](p − m2 )
2 2 2

kρ pρ 1
= 1 4 2iπ 2 lim 2 Tr[γ5 γλ γα γν γβ ]k2α pβ = − 2 ελσνρ k1ρ k2σ .
(2π) p→∞ p 8π

Therefore
1+β
k1µ Tµνλ (β) = ενλσρ k1σ k2ρ . (12.48)
8π 2
Therefore it is clear that it is not possible to cancel the anomalous terms from Eq. (12.44) and Eq.
(12.48) simultaneously with any choice of β. As it turns out that there is no anomalous term in the
Ward identities for h0|T(V V V )|0i3 while there are anomalous terms for h0|T(AAA)|0i, it is logical to
associate the anomaly with the axial-vector current. Thus we fix the momentum routing so that the
vector Ward identity is mantained, i.e. if β = −1, the axial Ward identity becomes

1
q λ Tµνλ = 2mTµν − εµνσρ k1σ k2ρ . (12.49)
2π 2
This corresponds to a modification of the axial-vector current divergence as

1 µνρσ
∂ λ Aλ (x) = 2imP (x) + ε Fµν (x)Fρσ (x) (12.50)
16π 2

This extra terms, called ABJ anomaly, is thus produced by the renormalization effect and has the
following properties:

1. The anomaly is independent of the fermion masses and should be present also in the massless
theory: this is highly non-trivial.

2. Adler and Bardeen showed that the coefficient in the anomaly term is not affected by higher
order radiative corrections, i.e. triangle diagrams with more than one loop do not contribute to
the anomaly term. This can be understood heuristically by noting that the superfificial degrees
of divergence of the higher-order triangle diagrams are less than one and the momentum-routing
ambiguity does not exist for such diagrams.

3. As our presentation has been in terms of momentum routing and conventional cut-off regularization
we may inquire how this anomaly problem rears its head in the dimensional regularization scheme.
There the problem shows up as the difficulty of giving a proper definition to the Dirac γ5 matrix
in space-time dimensions other than four.
3
Actually in QED h0|T(V V V )|0i = 0 thanks to Furry’s theorem.
168 CHAPTER 12. ANOMALIES

4. It was pointed out by Fujikawa that the ABJ anomalous Ward identity could be formulated
rather directly in the path-integral formalism. He showed that the path integral measure for
gauge invariant fermion theory is not invariant under the γ5 transformation. The extra Jacobian
factor gives rise to the ABJ anomaly.

5. To connect to the π 0 → γγ decay we can interpret our result as as saying that the composite
operator to which the pion couples, namely ψγ5 ψ, has a non-zero value in the presence of a
background electromagnetic field.

12.3 Anomalies in the Standard Model


In non abelian theories, Green’s functions with odd number of axial vector couplings up to five-point
functions contribute with anomalous terms to the divergence of axial-vector current, as shown by
Bardeen. However the triangle anomaly may be regarded as the basic one since it is simplest and its
absence implies the absence of all other anomalous diagrams. In the following we shall continue to
restrict our discussion to the triangle anomaly. Consider the three-point functions
Z
abc
d4 x1 d4 x2 h0|T Vµa (x1 )Vνb (x2 )Acλ (0) |0ieik1 ·x1 +ik2 ·x2 , (12.51)

Tµνλ (k1 , k2 ; q) = i

Z
abc
d4 x1 d4 x2 h0|T Vµa (x1 )Vνb (x2 )P c (0) |0ieik1 ·x1 +ik2 ·x2 (12.52)

Tµν (k1 , k2 ; q) = i

where
Vµa (x) = ψ(x)T a γµ ψ(x),
Acλ (x) = ψ(x)T c γλ γ5 ψ(x), (12.53)
c c
P (x) = ψ(x)T γ5 ψ(x)
and the T a s are the internal symmetry matrices. The result for the anomaly in the axial-vector Ward
identity is
1
q λ Tµνλ
abc abc
= 2mTµν − 2 εµνρσ k1ρ k2σ Dabc + commutator terms. (12.54)

where
1 
Dabc = Tr {T a , T b }T c . (12.55)

2
Anomaly free gauge theories are such that Dabc = 0. Notice that the fact that the SM is an anomaly
free gauge theory is highly non-trivial. We have to check all possible triangle diagrams and we will see
that they combine in such a way to get rid of anomalies.

Gauge anomalies in the SM


Now we will check that the currents associated with the SU (3)c ⊗ SU (2)L ⊗ U (1)Y gauge symmetries
of the Standard Model are non-anomalous. If we write these three currents as Jµc , JµL and JµY ,
then we have to show that, exploiting a short hand notation for the three point functions defined
above, ∂ µ hJµi Jνj Jρk i for i, j, k any of the three forces. We will exploit the condition Dabc = 0 for every
combination of the three currents. First of all recall the charges of the SM independent degrees of
freedom, namely the matter content (particle and antiparticles). The generators for SU (3)c are T a
with a = 1, . . . 8, for SU (2)L are T a with a = 1, 2, 3 and for U (1) simply Y . We shall check for the sake
of simplicity the combination regarding SU (2)L and U (1)Y only; the check for combinations involving
SU (3)c can be done the same way.

1. SU (2)3L case: all SU (2)L generators.

1  1  1  1
Dabc = Tr T a {T b , T c } = Tr σ a {σ b , σ c } = Tr σ a δ bc 1 = δ bc Tr σ a = 0 (12.56)
  
2 16 8 8
12.3. ANOMALIES IN THE STANDARD MODEL 169

SU (3)c SU (2)L U (1)Y T3 Q Y = Q − T3


! ! ! !
ν +1/2 0 −1/2
L= 1 2 −1/2
e −1/2 −1 −1/2
! ! ! !
u +1/2 +2/3 +1/6
Q= 3 2 +1/6
d −1/2 −1/3 +1/6
ec 1 1 +1 0 +1 +1
uc 3 1 −2/3 0 −2/3 −2/3
dc 3 1 +1/3 0 +1/3 +1/3

2. U (1)Y SU (2)2L case: one U (1)Y generator and two SU (2)L generators

1  1  2δ bc 1  δ bc  
Dabc = Tr Y {T b , T c } = Tr Ydb (12.57)

= Tr Ydb
2 2 4 4

where Ydb denotes the charges of doublets only since only SU (2)L doublets couple with SU (2)2L .
Being Ydb = Qdb − T3db and Tr T3db = 0,


1  δ bc  2 1 δ bc
 
Tr Y {T b , T c } = = 0 ⇔ Nc = 3 (12.58)
 
Tr Qdb = 3 × −1 +Nc ·
0 |{z} −Nc ·
2 4 |{z}
νL | {z 3} | {z 3} 4
eL
uL dL

The anomaly cancelation occurs if and only if Nc = 3. This is an outstanding confirmation of


the colour number.

3. SU (2)L U (1)2Y case: one SU (2)L generator and two U (1)Y generators.

 σa 
Dabc = Tr T a Y 2 ∝ Tr (12.59)
 
=0
2

because Y is the same for both members of the same SU (2) doublet.

4. U (1)3Y case: three U (1)Y generators.

Dabc =Tr Y 3 ] = Tr Ydb


3
] + Tr Ysg3 ] = Tr (Qdb − T3db )3 + Tr (Qsg )3
      

Q3db
 − 3Q2 T3db + 3Qdb T 2 − T 3  3sg
 
=Tr  db 3db 3db + 
 TrQ

= − 3 Tr Q2db T3db +3Tr Qdb T3db


2 
  
| (12.60)
 {z }
2Tr 2
Qdb T3db

2  3  3 
= − Tr Qdb (σ3 )2 = − Tr Qdb = 0
  
= − 3Tr Qdb T3db
4 4

Also all possible anomalies involving SU (3)c generators vanish. Therefore the Standard Model is
(gauge) anomaly free as any consistent theory should be. The anomaly cancellation conditions reveal a
strict quark-lepton connection which may find a natural explanation in the context of GUTs.

Global anomalies in the SM


As we have already stressed, global anomalies do not lead to incosistencies in our theory. Nevertheless
they are important to be studied. The SM has four accidental symmetries: the baryon number B and,
for massless neutrinos, the three lepton numbers Li .
170 CHAPTER 12. ANOMALIES

1. U (1)B baryon number. We have to compute Tr B{T b , T c } . We have two cases4


 

 2 2  2 
1 1 2 1 Nc 3
 
2
(12.63)

Tr BY =3× Nc · 2 · −Nc − −Nc =− = − 6= 0,
3
|{z} 6 3 3 2 2
| {z! }| {z }| {z }
B uc dc
u
d

and
σb σc 1 2δ bc 1
    
Tr B , = 3 × Nc Tr = δ bc Nc 6= 0 (12.64)
2 2 3 4
since the doublets baryon number Bdb = 1/3. Therefore the baryon number is anomalous.

2. U (1)Li lepton number. We have to compute Tr Li {T b , T c } . We have two cases5


 

2
1 1
  
Tr Li Y 2 = 2 · − − (+1)2 = − = (12.65)
 
6 0, i = e, µ, τ each
2 | {z } 2
| {z! } ec
ν
e

σb σc δ bc
  
Tr Li , = Tr[Lidb ] = δ bc 6= 0 (12.66)
2 2 2
Therefore also the family lepton number is anomalous. However from the similar structure of
Eq. (12.64) and Eq. (12.66) we see that NBc − Li is not anomalous in contrast e.g. with B + Li .
Since Nc = 3 = number of generations it turns out that defining L = Le + Lµ + Lµ , B − L is
non-anomalous. There are theories (GUT theories) where actually B − L is promoted to a gauge
symmetry and this is only possible thanks to the fact that it is non anomalous.

4
Actually we would have to check two more cases, but they are quite easy:

Tr[B 2 T a ] = 0 (12.61)

since B is the same for quarks in the same SU (2)L doublet. Moreover
 
1 1 1 2 1 1
Tr[Y B 2 ] = Nc · ×2− · + · =0 (12.62)
9 6 9 3 9 3

5
Same here as above.
Chapter 13

Higgs physics

In this chapter we shall study the Higgs boson physics. Until its discovery at LHC on 2012 at CERN
it has been the last missing piece in the Standard Model. The importance of the Higgs sector in the
Standard Model is underlined by the fact that before its discovery many efforts were put in order to
find out the best bounds on its mass, both theoretically and experimentally, mainly at LEP and at
Tevatron. We shall review these bounds and also analyze consequences of the Higgs discovery. Finally
we will briefly outline some Higgs production mechanisms which are at work at colliders and also the
Higgs decay modes thanks to which the Higgs boson was discovered.

19.7 fb-1 (8 TeV) + 5.1 fb-1 (7 TeV)


×1043
S/(S+B) weighted events / GeV

CMS S/(S+B) weighted sum


3.5 H → γγ
Data
3
S+B fits (weighted sum)
B component
2.5
± 1σ
±2 σ
2

1.5

1
µ = 1.14 +− 0.26
0.23
0.5 mH = 124.70 ± 0.34 GeV

0
200
B component subtracted

100

-100
110 115 120 125 130 135 140 145 150
mγ γ (GeV)
Figure 13.1: Plot of the Higgs resonant peak at Mh ≈ 125 GeV in the gg → γγ channel at CMS (LHC)
at CERN. Figure from https://arxiv.org/abs/1510.01924.

13.1 Theoretical bounds on Mh : triviality and vacuum stability bounds


Theoretical bounds on the Higgs mass come from three main analysis:
1. Perturbative unitarity bound.

2. Triviality bound.

3. Vacuum instability bound.

171
172 CHAPTER 13. HIGGS PHYSICS

We have already studied the perturbative unitarity bound on the Higgs mass coming from the analysis
of the W W → W W scattering in the high-energy limit thanks to the Goldstone Boson Equivalence
Theorem and the partial wave expansion of the cross-section computed in that regime. Therefore,
simply imposing the unitarity of the S matrix describing the process we were able to give an upper
bound on the Higgs mass Mh ≤ 1 Tev which is indeed a non trivial result.
Now let us study the other two theoretical bounds. The Higgs potential, after spontaneous symmetry
breaking takes the form

SSB 1 λ
V (H) = M 2 (H † H) + λ(H † H)2 −→ Mh2 h2 + h4 + · · · , (13.1)
2 4
with the dots including all other interaction terms and Mh2 = 2λv 2 . We have already seen while
treating renormalization in QED that the QED coupling runs, because of the fact that in dimensional
regularization the bare Lagrangian coupling e0 has a non vanishing mass dimension: in that case we
had to introduce a dependence of the renormalized coupling on a unphysical renormalization scale µ of
mass dimension 1, therefore introducing a dependence on the energy scale at which the coupling is
probed. In the Higgs sector also λ is a running coupling. Studying the mass dimensions of the Higgs
sector Lagrangian in dimensional regularization,
d−2
[(∂µ H)2 ] = d = 2 + 2[H] =⇒ [H] = , (13.2)
2
d−2
 
[λH 4 ] = d = [λ] + 4 =⇒ [λ] = 4 − d = ε, (13.3)
2
therefore, being a dimensional coupling, λ acquires a dependence on the renormalization scale µ. The
evolution of the coupling λ as a function of energy is governed by the beta function β(λ), that is the
analogue of the one we showed in QED for the running of the electric charge. The equation describing
the running of λ is then

µ = β(λ). (13.4)

In the following, we show some relevant Feynman diagrams for the beta function β(λ)

t
t
+ + t + t t + ···
(13.5)
t

λ λ2 λyt2 yt4

The beta function is computed considering the tree level contribution and all the 1PI loop diagrams
giving corrections to the 4-point function. The final result reads:

1 9 3 4 3
 
(13.6)
2 2
β(λ) = 2
24λ2 + λ(4Nc yt2 − 9g 2 − 3g 0 ) − 2Nc yt4 + g 4 + g 0 + g 2 g 0
16π 8 8 4

Triviality bound
We can obtain an upper bound on the Higgs mass if we consider it to be large and so we study the
latter equation in the λ  1 limit. We can approximate
dλ 1 
24λ2 (13.7)

= 2
d log µ 16π
which, once solved, gives
λ(µ0 )
λ(µ) = 2 . (13.8)
1 − λ(µ0 ) 4π3 2 log µµ2
0
13.1. THEORETICAL BOUNDS ON MH : TRIVIALITY AND VACUUM STABILITY BOUNDS173

Now we fix the reference renormalization scale to be the EW scale (the value of the Higgs VEV, which
is the right scale to define the Higgs boson mass Mh2 = 2λv 2 ) µ0 = v and we vary the energy scale at
which we probe the coupling µ = Λ. We obtain
λ(v)
λ(Λ) = 2 . (13.9)
1 − λ(v) 4π3 2 log Λv2
Let us study the result in the two limits

 Λ  v. We have that λ(Λ) −→ 0: we obtain a trivial theory i.e. a non interacting theory, since
Λv

the coupling vanishes.

 Λ  v. In this case the logarithmic term grows and sooner or later it reaches the value of 1.
therefore the coupling λ(Λ) reaches non perturbativity values as the function approaches the
so-called Landau pole. The only possibility to avoid this issue is to assume that λ = 0.
We realize that in both case we are led to consider a vanishing coupling i.e. a non interacting, trivial
theory (whence the name triviality bound). This can be avoided if we impose perturbativity1 of the
theory up to a given energy scale Λ. This is tantamount to impose the denominator of Eq. (13.9)
always positive.
3 Λ2
1 − λ(v) 2 log 2 > 0 (13.10)
4π v
which, introducing the Higgs mass λ = Mh2 /2v 2 gives the upper bound
(
8π 2 v 2 Mh < 600 GeV Λ ≈ 2.5 TeV
Mh2 < Λ2
=⇒ (13.11)
3 log v2
Mh < 160 GeV Λ ≈ 2.5 × 1016 GeV

Thus, if we require perturbativity up to the GUT scale we imply a very low upper bound for the Higgs
mass. Nevertheless, if we are less demanding on the range of perturbativity, we end up with an upper
bound of 600 GeV which is low enough to be probed by experiments. This is why LHC, with its TeV
scale center of mass energy, was thought to be a Higgs discovery machine.

Vacuum stability bound


Let us now analyze the case in which the supposed Higgs mass is small, hence λ  1. In this limit we
can neglect all terms proportional to λ or λ2 in Eq. (13.6) and focus on the term proportional to yt4
which is by far the dominant one:
dλ 1 
= − 2Nc yt
4
(13.12)
d log µ 16π 2
which gives us, with the same choices for µ and µ0 , and assuming yt independent on the energy scale,
3 4 Λ2
λ(Λ) = λ(v) − y t log (13.13)
16π 2 v2
Now we impose λ(Λ) > 0: this is the vacuum stability condition. The Higgs potential contains a ∼ λϕ4
coupling hence λ > 0 is a fundamental condition for the stability of the vacuum state of the Higgs field.
Hence we obtain a lower bound on the Higgs boson mass
3v 2 4 Λ2
Mh2 > y log (13.14)
8π 2 t v2
Finally the theoretical bounds on the Higgs mass are given by the expression

3v 2 4 Λ2 8π 2 v 2
2
yt log 2 < Mh2 < 2 (13.15)
8π v 3 log Λv2
1
In other words we require the Landau pole to be above an arbitrary energy scale Λ.
174 CHAPTER 13. HIGGS PHYSICS

350

MH [GeV]
Perturbativity bound
Stability bound
300 Finite-T metastability bound
λ = 2π
Zero-T metastability bound
λ =π Shown are 1σ error bands, w/o theoretical errors

250

200

Tevatron exclusion at >95% CL

150
LEP exclusion
at >95% CL

100
4 6 8 10 12 14 16 18
log (Λ / GeV)
10

Figure 13.2: Perturbativity (triviality) bounds for two arbitrary values of λ(v) and vacuum stability
bound on the Higgs mass. The regions probed by LEP and Tevatron in which no Higgs boson was
found are also shown. The allowed region for the Higgs mass is below the λ curves and above the
stability bound.
180
107 108
200 Instability Instability 109
178
1010
Top mass Mt in GeV

ility 1011
150 stab
Non-perturbativity

Top pole mass Mt in GeV

ta- 176 1012


Me 1013
1016
100 Stability 174
1,2,3 Σ
Meta-stability
172 1019
50
170
1018
0 1014 Stability

0 50 100 150 200 168


120 122 124 126 128 130 132
Higgs mass Mh in GeV Higgs pole mass Mh in GeV

Figure 13.3: Left – regions of absolute stability, meta-stability and instability of the SM vacuum in the
Mh − mt (pole masses) plane. Right – Zoom in the region of the preferred experimental range of Mh
and mt ; the gray areas denote the allowed regions at 1, 2 and 3σ. The three boundary lines correspond
to αs (MZ ) = 0.1184 ± 0.0007, and the grading of the colors indicates the size of the theoretical error.
The dotted countour-lines show the instability scale Λ in GeV assuming αs (MZ ) = 0.1184.

which depends only on Λ. We obtain a bound of 60 GeV < Mh < 600 GeV for Λ ≈ 2.5 TeV. In Fig.
(13.2) we show the theoretical bounds together with some experimentally rejected regions. Considering
that λ(v) ∼ 0.1 (which is much less than π, as shown in figure) we see that there was no much room
for the Higgs boson. In Fig. (13.3) we show the phase plot with regions of stability, meta-stability,
instability and non-perturbativity (triviality) in the mt − Mh plane (which is equivalent to the yt − λ
plane). The stability region is such that the Higgs vacuum is stable up to Λ = MPl . The interesting
fact is that plotting the experimental values of mt and Mh we seem to live in a meta-stable region
in which the ”lifetime of the vacuum” is longer than the age of the universe (luckly enough!). More
detailed computation at 2-loop confirmed this result, showing that the observed Higgs vacuum is indeed
very close to the boundary between the stable and the meta-stable region.
13.2. EXPERIMENTAL BOUNDS ON MH 175

13.2 Experimental bounds on Mh


The experimental search of the Higgs boson was carried out over many years by the LEP collider and
after that by the Tevatron and LHC.

LEP: e+ e− collider machine

The main Higgs production channel in a e+ e− collider is the Higgs-strahlung described by the following
Feynman diagram
e+ h

Z? (13.16)

e− Z

Unfortunately the energy necessary to produce the final state is well above (more than twice) the

Z-pole. Since LEP was running at smax ' 210 GeV the missed discovery implies a lower bound on
the Higgs boson mass
Mh ≥ 114 GeV (13.17)

which is almost equivalent to the kinematical bound Mh ≥ smax − MZ ≈ 119 GeV.

LHC: pp collider machine

In a similar way to Tevatron, which is a pp collider, the dominant channel to produce the Higgs boson
at the LHC is the gluon-gluon fusion process, in which gluons fuse in a quark loop (dominated by the
top-quark) which at the end gives rise to a Higgs:

t h (13.18)

Moreover, there are also other processes called glu-glu fusion plus two jets (gg, gq, qq) where the gluon
fusion happens in three different cases depending on the partons (constituents of the colliding protons)
involved: (a) when the two gluons are radiated by two gluons (gg), (b) when a quark is involved (gq)
and (c) when both gluons are radiated by quarks (qq)

g g g g q q
g g g

t h t h t h (13.19)

g g g
g g q q q q
(a) (b) (c)

The gluons and quarks of high energy soon hadronize in jets producing hadronic showers in the
detectors. Another process which however is subdominant with respect to the ones above is the vector
176 CHAPTER 13. HIGGS PHYSICS

boson fusion, allowed at tree level by the fact that the vector bosons directly couple to the Higgs.
q q
W Z

h (13.20)
W Z
q q
(d)

It is important, in producing possible new particles, to monitor the number of possible events associated.
The expected number of events per unit time is estimated with the formula
dNh L  σ(gg → h) BR(h → final)
       
= 2s−1 · · · , (13.21)
dt 1034 cm−2 s−1 1 2 × 10−35 cm2 1
where we rescaled the quantities with appropriate values for the LHC luminosity and the σ(gg → h) ≈ 20
pb production cross section. For the efficiency  we assumed a value of order 1. Therefore at LHC it
was expected a large number of events to detect the Higgs boson.

Indirect bounds on Mh from EWPT


As we pointed out already at the end of the section regarding the EWPT the complete electroweak fit
only gives a loose bound on Mh due to the weak logarithmic dependence of quantum corrections to
Mh . For example, the ρ parameter receives the following contribution from the Higgs boson

h
h
g02 Mh2
 
+ ∼ log 2 . (13.22)
ZW ZW ZW ZW ZW 16π 2 MZ,W

The EW complete fit by the group GfitterSM is shown in Fig. (13.4– left): this gives indeed the loose
interval Mh = 94+23
−24 GeV, which is better understood as

Mh < 171 GeV at 95% CL (13.23)


The best value for Mh is found from the fit and extracted as the minimum ∆χ2 (ideally when ∆χ2 = 0,
when the measured and predicted value coincide). In Fig. (13.4 –right) the fit is repeated taking into
account also the LEP results, mainly the Mh ≥ 114 GeV bound. We already know the answer i.e. that
the EW fit result is quite below the experimental LHC value of the Higgs boson mass.

13.3 Higgs production: gluon-gluon fusion and the parton model


The gluon-gluon fusion process in a pp collider is represented schematically in the hadronic center of
mass by the following Feynman-like diagram:

P1
p g(x1 , µ)
pg1

g
t h (13.24)
g
pg2
p g(x2 , µ)
P2
13.3. HIGGS PRODUCTION: GLUON-GLUON FUSION AND THE PARTON MODEL 177

10 12
∆ χ2

∆ χ2
Tevatron 95% CL
LEP 95% CL

LEP exclusion at 95% CL


Mar 09
G fitter SM
9 3σ
10

Mar 09
8 G fitter SM

7
8

Tevatron exclusion at 95% CL


6

5 6

4 2σ
Theory uncertainty 4 2σ
3 Theory uncertainty
Fit including theory errors
2 Fit including theory errors
Fit excluding theory errors 2 Fit excluding theory errors
1 1σ 1σ
0 0
50 100 150 200 250 300 100 150 200 250 300
MH [GeV] MH [GeV]

Figure 13.4: Dependence on Higgs boson mass Mh of the ∆χ2 function obtained from the global fit of
the SM parameters to precision electroweak data, excluding (left) or including (right) the results from
direct searches at LEP and the Tevatron. Figure from https://arxiv.org/abs/0906.0954.

In order to give a proper theoretical description of the process we have to consider the hadrons as
composed by partons, point-like particles that are quarks or gluons. For example, as we all know, a
proton, at low energies appears to be made out of two up quarks and one down quark which we shall
call the valence quarks. However, deep inelastic scattering experiments show that at higher energies,
besides the valence quarks, we can find other quarks, antiquarks and gluons inside the proton. This
can be easily interpreted; at high enough energies any virtual gluon can become a quark-antiquark
pair. This gives rise to the parton distribution functions which are scale dependent. In particular
we consider the parton distribution functions for gluons g(x1,2 , µ) i.e. the probability densities of
finding a gluon of momentum pg1,2 = x1,2 P1,2 in the proton, where P1,2 are the proton momenta and
0 ≤ x1,2 ≤ 1 the momentum fractions carried out by gluons. In Fig. (13.5) we show the NNPDF3.1
NNLO PDFs evaluated at different scale of energy µ. The computation of those parton distribution
functions (PDFs), answering the simple question ”what is a proton made of?” is highly non-trivial.
Let’s relate now the parton center of mass frame (PCM) with the hadronic center of mass (HCM).
We are going to plot the cross section for very high HCM energies (> 103 mp ) so we can consider the
proton as massless. Therefore in the HCM frame we can write:

P1µ = (p, −~p) S = (P1 + P2 )2 = 2P1 · P2 = 4p2


µ =⇒ (13.25)
P2 = (p, p~) s = (pg1 + pg2 )2 = (x1 P1 + x2 P2 )2 = x1 x2 S ≡ τ S

We also define τ0 = Mh2 /S. The differential cross section in the PCM frame dσ̂, assuming gg in the
initial state, is, exploiting standard formula

1 1 d3 ph 1
dσ̂ = |M|2 (2π)4 δ(Eg1 + Eg2 − Eh )δ (3) (~
pg1 + p~g2 − p~h ) (13.26)
(2Eg )2 |c − (−c)| (2π)3 2Eh

In the PCM pµg1 = (Eg , p~g ), pµg2 = (Eg , −p~g ) and pµh = (Mh , 0) so the three dimensional delta gives us

p~h = 0. Introducing s = 2Eg = Mh , we obtain

2π 1 2 √
dσ̂ = |M| δ( s − Mh )
2Mh2 2Mh
(13.27)
π δ(s − Mh2 ) π
= 3 |M|2 1 = 2 |M|2 δ(s − Mh2 )
2Mh √ | 2 M
2 s s=M h h

and accounting for the initial state polarizations (2 spins and 8 gluons) and the fact that

|M(gg → h)|2 = |M(h → gg)|2


178 CHAPTER 13. HIGGS PHYSICS

11 11
NNPDF3.1
NNPDF3.1 (NNLO)
(NNLO) g/10
g/10
0.9
0.9 0.9
0.9
xf(x,µµ =10
xf(x, 22
=10 GeV
GeV )) 22
xf(x,µµ22=10
xf(x, =1044 GeV
GeV22))
ss
0.8
0.8 0.8
0.8
g/10
g/10
0.7
0.7 0.7
0.7
dd
0.6
0.6 0.6
0.6
uuvv
cc
0.5
0.5 0.5
0.5
uuvv
0.4
0.4 0.4
0.4 uu
ddvv
0.3
0.3 ss 0.3
0.3 bb ddvv

0.2
0.2 0.2
0.2
dd
uu
0.1
0.1 0.1
0.1
cc
00 00
−−33 −−22 −−11 −−33
10
10 10
10 10
10 11 10
10 10−−22
10 10−−11
10 11
xx xx

Figure 13.5: The NNPDF3.1 NNLO PDFs, evaluated at µ2 = 10 GeV2 (left) and µ2 = 104 GeV2
(right) obtained by the NNPDF collaboration from high-precision collider data. Figure from https:
//arxiv.org/abs/1706.00428 .

we obtain
1 1 π|M(h → gg)|2
σ̂(gg → h) = × δ(s − Mh2 ) . (13.28)
2·8 2·8 Mh2
Exploiting the properties of the Dirac delta
M2 1 τ0
    
δ(s − Mh2 ) = δ(x1 x2 S − Mh2 ) = δ x1 S x2 − h = δ x2 − (13.29)
x1 S x1 S x1
Now we have to account for the parton distribution functions. The total cross section is given by
Z Z
σ(gg → h) = dx1 dx2 g(x1 , µ)g(x2 , µ)σ̂(gg → h)
(13.30)
π|M(h → gg)|2 τ0 τ0
Z Z  
= dx1 dx2 g(x1 , µ)g(x2 , µ) δ x2 −
256Mh4 x1 x1
and finally
π|M(h → gg)|2 1 dx1
Z
σ(gg → h) = τ0 g(x1 , µ)g(τ0 /x1 , µ) (13.31)
256Mh4 τ0 x1
where the amplitude |M(h → gg)|2 will be evaluated at the end of this chapter. This cross section has
to be integrated numerically. In Fig. √
(13.6) it is shown the dependence of σ(gg → h) on the Higgs
mass at three
√ different HCM energies, S = 1.96 Tev, 7 TeV and 14 TeV. As expected, the LHC cross
section at S = 14 TeV is the dominating one. It reaches σ ≈ 30 pb for a 100 GeV Higgs mass. The
cross section at 7 TeV is approximately three times lower and finally the one corresponding to the
Tevatron energy is very low, more than one order of magnitude smaller. The bump in the cross section
near 350 GeV is due to the production of tt pairs in resonance when Mh = 2mt .

13.4 Higgs decay modes


Now let us study the dominant decay modes through which we can detect the Higgs boson.
13.4. HIGGS DECAY MODES 179


Figure
√ 13.6: Gluon-gluon fusion cross section at S = 7 and 14 TeV for pp collisions at the LHC and
at S = 1.96 TeV for pp collisions at Tevatron.

13.4.1 h → f f
The decay rate of the process h → f f can be estimated in a strightforward way in NDA. Starting from
the relevant Feynman rule for Yukawa interactions

f (p)

mf /v mf
M(h → f f ) = h(k) = u(p)v(q) (13.32)
v

f (q)

the tree level prediction for the decay rate should be

m2f f 1

4m2f α
Γtree (h → f f ) ∝ N c × M h × 1 − (13.33)
2
|v {z }
|{z} 8π Mh2
[Γ]=1 | {z }
from F. rules adim. phase space factor

where the phase space factor (1 − 4m2f /Mh2 )α (with α not predictable in NDA) can be safely neglected
for mf  Mh . Notice that, since Γtree (h → f f ) ∝ m2f , the largest decay rates are for the heaviest
fermions, that are the τ lepton and the bottom quark.
Let us compute now the decay rate in full detail. The unpolarized squared amplitude, including a
possible color factor Ncf , is

X m2f
|M|2 =Ncf · 1 · |M|2 = Ncf / + mf )(q/ − mf )]
Tr[(p
v2
spins . (13.34)
m2f 4m2f 
=Ncf 2 pα qβ 4η αβ − 4mf 2
= Ncf 2 p · q − m2f
 
v v
In the Higgs boson rest frame we have

(13.35)
q q
k µ = (Mh , 0, 0, 0) pµ = m2f + p2? , p? , 0, 0 qµ = m2f + p2? , −p? , 0, 0 ,
 

with v
u 2
t1 − 4 mf .
Mh u
(13.36)
q
Mh = 2 2 2
mf + p? =⇒ p? = 2 2 Mh
180 CHAPTER 13. HIGGS PHYSICS

Moreover, since p · q = m2f + 2p2? , the decay rate reads

2  32
p? GF mf
Z  
+ −
Γ(h → f f ) = |M| dΩ = √ Ncf m2f Mh 1 − 4
2
. (13.37)
32π 2 Mh2 4π 2 Mh

This result, obtained at tree-level, is not enough accurate in the case of quarks. Indeed, considering
the dominant decay channel h → bb, one should include the following QCD corrections

(tree level) h

b
b g b b

(1 − loop corrections) + h g + h + h (13.38)

g
b b b
b b
g
(bremsstrahlung) + h + h = IR and UV finite
g
b b

which include both loop corrections as well as gluon bremsstrahlung contributions. Both contributions
are of order αs as loop effects interfere with the tree level amplitude while the bremsstrahlung process
does not. As a result, we can write the QCD corrected decay ate of h → bb as

ΓQCD QCD
tot = Γtree (1 + δ1−loop ) + ΓB (h → bbg) . (13.39)

Only the sum of loop corrections and bremsstrahlung contributions is free of UV and IR divergences.
The inclusion of all QCD corrections are accounted for in a good approximation by replacing the
bottom quark pole mass entering at tree level with the running quark mass in the MS scheme at the
Higgs mass scale
m2b (Mh )
ΓQCD
tot ≈ Γtree . (13.40)
m2b (mb )

The masses of fermions run with the energy in a similar way to what happens to the coupling constant:
since gs decreases (asymptotic freedom), also the quark masses are smaller at higher energy.

13.4.2 h → W W ? and h → ZZ ?

The decays of the Higgs boson into on-shell vector bosons are not allowed. However, if one of the
vector bosons is virtual, we can have the following three body decays which contribute by a substantial
fraction to the Higgs width
h →W W ? → W + fu fd
h →W W ? → W − fd fu (13.41)
?
h →ZZ → Zf u,d fu,d
13.4. HIGGS DECAY MODES 181

where fu = u, c, νe , νµ , ντ and fd = d, s, b, e, µ, τ , taking into account that the decay in the top is not
kinematically allowed. The Feynman amplitude for the first process is given by

W + (p2 )
2iMW2

v
h(p1 ) q fd (p3 )
M(h → W + fu fd ) =
W?
(13.42)
g
√ V f γ β (1
2 2 ud
− γ5 )
f u (p4 )
2
2iMW g qα qβ −i
 
f ?α β
= √ Vud ε (p2 ) − ηαβ + 2 2 2 u(p3 )γ (1 − γ5 )v(p4 ).
v 2 2 MW q − MW
with q = p3 +p4 = p1 −p2 . We can neglect the qα qβ term in the amplitude because u(p3 )(p
/3 +p
/4 )v(p4 ) = 0.
The unpolarized squared amplitude yields, introducing also the possible color factor:
g2 pα2 pµ2 1
 
f 2 f αµ
|M|2 = 4M 4
|V
W ud | Nc − η + /3 γ ν (1 − γ5 )p
2 (−ηµν )(−ηαβ ) [q 2 − M 2 ]2 Tr[p /4 γ β (1 − γ5 )].
8v 2 MW W
(13.43)
The trace reads

T βν = Tr[p /4 γ β (2 − 2γ5 )] = 8 (pβ4 pν3 + pβ3 pν4 ) − η βν (p3 · p4 ) + 


/3 γ ν p iεβνρσ (13.44)
  
p 3ρ p4σ

where the Levi-Civita tensor term can be neglected since it will be contracted with a symmetric tensor
in the βν indices. Indeed
p2β p2ν 2(p3 · p2 )(p4 · p2 )
   
T βν − ηβν + 2 = 8 (p3 · p4 ) + 2 (13.45)
MW MW
and therefore we can write the unpolarized amplitude modulus squared as
1 2(p2 · p3 )(p2 · p4 )
 
|M|2 = NW 2 2
p3 · p4 + 2 (13.46)
[2(p3 · p4 ) − MW ] MW
where
4 |V f |2 N f
4g 2 MW ud
NW = c
. (13.47)
v2
Now we have to integrate over the three-body invariant phase space in order to get the rate
|M|2 d3 p3 d3 p4 d3 p2 4
dΓ = δ (p1 − p2 − p3 − p4 )
(2π)5 16Eh Efu Efd EW
(13.48)
N 1 1 (p2 · p3 )(p2 · p4 ) 3
 
= 2 ]2 p3 · p 4 + 2 d p2 d3 p3 d3 p4
(2π)5 16Eh E3 E4 EW [2(p3 · p4 − MW MW
First we will integrate in p3 and p4 . We define
pµ3 pν4 1
Z
I µν
(q) ≡ 3 3
d p3 d p4 4
2 ]2 δ (q − p3 − p4 ) (13.49)
E3 E4 [2(p3 · p4 ) − MW
and we can decompose as usual I µν as

I µν = Aη µν + Bq µ q ν , (13.50)

so that 
1 I2
 
A = 3 I 1 − q 2

I1 = ηµν I µν = 4A + Bq 2
( 

(13.51)
I2 = qµ qν I µν = q 2 A + Bq 4 1 4I2
 

A = 2 − I1


3q 2
q
182 CHAPTER 13. HIGGS PHYSICS

We compute I1 and I2 , putting ourselves in the center-of-mass frame of the two fermions: pµ3 = (|~
p|, p~)
µ 2 2
p|, −~
p4 = (|~ p), ~q = 0, p3 · p4 = (p3 + p4 ) /2 = 2|~
p|

p3 · p4 1
Z
I1 = d3 p3 d3 p4
2
4
2 ]2 δ (q − p3 − p4 )
|~
p| [2(p3 · p4 ) − MW
1
Z
= d3 p3 2 2 2 ]2 δ(q0 − 2|~ p|)
p | − MW
[4|~

Z
2

q0
 (13.52)
= d|~ p|2
p||~ 2 ]2 δ − |~
p |
2 [4|~p|2 − MW 2
q02 2 πq02
=2π
4 [q02 − MW
2 ]2 =
[q02 − MW 2 ]2 .

(q · p3 )(q · p4 ) 1
Z
I2 = d3 p3 d3 p4 2
4
2 ]2 δ (q − p3 − p4 )
|~
p| [2(p3 · p4 ) − MW
q02 |~
p|2 1
Z
= d3 p3 2 2 2 ]2 δ(q0 − 2|~
p|) (13.53)
|~p| [4|~
p| − MW
4π q02 q0 πq04 1
Z  
2
= p||~
d|~ p| 2 2 2
δ − |~
p| = 2 2 ]2 .
2 p| − MW ]
[4|~ 2 2 [q0 − MW

Being I1 and I2 scalars they do not depend on the frame we choose, so we can safely replace q02 → q 2 .
We get, straighforwardly,

1 πq 2 π 1
A= 2 )2 , B= 2 )2 , (13.54)
6 (q 2 − MW 2
3 (q − MW

and
1 π 2 µν π µ ν
 
I µν
= 2 2 2
q η + q q (13.55)
(q − MW ) 6 3
Therefore the differential decay rate, to be integrated in p2 , reads

NW 1 2p2µ p2ν µν 3
 
dΓ = 5
I1 + 2 I d p2
(2π) 16Eh EW MW
NW 1 πq 2 2 π 2 2 π
  
= 2 )2 + M 2 (q 2 − M 2 )2 6 q MW + 3 (p2 · q)
2
d3 p2 (13.56)
(2π)5 16Eh EW (q 2 − MW W W
NW 1 1 2π

(p · q) 2
2 2
= 2 )2 3 2q + M 2 .
(2π)5 16Eh EW (q 2 − MW W

Now we put ourselves in the rest frame of the Higgs boson, where pµ1 = (Mh , ~0), pµ2 = (EW , p~2 ) and

q 2 =(p1 − p2 )2 = Mh2 + MW
2
− 2p1 · p2
2
(13.57)
p2 · q =p2 · (p1 − p2 ) = p2 · p1 − MW ,

(p2 · q)2 2 )2
(Mh EW − MW
2q 2 + 2 =2M 2
h + 2M 2
W − 4M E
h W + 2
MW MW
(13.58)
M2 2
= 2h EW − 6Mh EW + 2Mh2 + 3MW
2
.
MW
We can write the differential decay rate as

Mh2 2 − 6M E 2 2
EW h W + 2Mh + 3MW
NW 1 2π (13.59)
2
MW
dΓ = · p2 |2 d|~
4π|~ p2 |.
(2π)5 16EW Mh 3 (Mh2 − 2EW Mh )2
13.4. HIGGS DECAY MODES 183

Now we define xW ≡ MW /Mh and y ≡ EW /Mh and change integration variable to EW i.e to y.
2 −2 2 (y 2 − x2W )
NW 1 2 y xW − 6y + 3xW + 2 3 q
dΓ = M 2 · 2πM y dy
(2π)4 48Mh6 y h (1 − 2y)2 h
y 2 − x2 W (13.60)
NW 1 y 2 x−2
W − 6y + 3x2W + 2q
= y 2 − x2W dy.
(2π)3 24Mh (1 − 2y)2
Before integrating in y we need to set the integration boundaries which, in a three body decay, read
M 2 + MW
2 1 1 MW
 
ymax = h = 1 + x2 ymin = = x. (13.61)
2Mh Mh 2 Mh
Therefore, the decay rate of our process is given by
NW 1
Γ(h → W + `− ν ` ) = 3
J (xW ) (13.62)
(2π) 24Mh
where 
1
Z 1+x2W y 2 x−2 − 6y + 3x2 + 2 q 2
(13.63)
2
J (xW ) = y − x2 dy .
x (1 − 2y)2
Setting Mh ≈ 125 GeV and MW = 80.4 GeV one can find numerically J (x? ) = 0.0273. This gives us
the decay rate for the process h → W + `− ν ` , exploiting g 2 v −2 = 8MW
2 G2
F

1 2 5 ?
Γ(h → W + `− ν ` ) = G M x J (x?W ) = 4 × 10−5 GeV (13.64)
6π 3 F W W
which corresponds to a branching ratio of about BR(h → W + `− ν ` ) = 0.01. Now we can compute the
inclusive decay rate h → W W ? i.e. with the virtual W boson that can decay in any up-down fermion
pair. To do that we need to evaluate the factor

(13.65)
X
Ncf |Vud |2 = 3 × 2 + 3 = 9
fu fd

which accounts for three lepton doublets and two active quark doublets, each in three colours. Finally,
multiplying by a factor of 2 corresponding to the cases in which the physical W boson is positively or
negatively charged we get
3 2 5 ?
Γ(h → W W ? ) = G M x J (x?W ) = 7.7 × 10−4 GeV, (13.66)
π3 F W W
which corresponds to BR(h → W W ? ) = 0.19, which is subdominant only to the decay channel h → bb.
Now let us discuss the decays of the Higgs into the ZZ ? pair. The Feynman diagram describing
the process h → ZZ ? → Zf u,d fu,d is the following

Z(p2 )
2
2iMZ
v
h(p1 ) q f (p3 )
?
M(h → ZZ → Zf u,d fu,d ) = (13.67)
Z?
g
cW γ β (gVf − gA
f
γ5 )
f (p4 )

We won’t repeat all the calculation we have already done, but instead we will derive the final result
by exploiting substitutions, mainly regarding the NW factor defined in Eq. (13.47). First of all we have
to substitute MW → MZ in NW 2 because of the different hZZ vertex; secondly, we have to multiply
2
This substitution has to be done in Eq. (13.47) only, since all the other masses will be taken into account by defining
a xZ . Also notice that introducing GF we used g 2 v −2 = 8MW 2
G2F in which the substitution shall not be performed.
184 CHAPTER 13. HIGGS PHYSICS


NW by 8/c2W to take into account the substitution 2 2 → cW ≡ cos θW in the Zf f vertex; thirdly we
have to multiply NW by (gV2 + gA 2 )/2 to take into account the different structure of the Zf f vertex

w.r.t the W f f one (of course gV ≡ gVf and gA ≡ gA f


for the ease of notation) 3 ; fourthly we need to
divide the final decay rate by 2 because, differently to the Ws, the Zs are indistinguishable. All of this
can be taken into account replacing
(gV2 + gA
2) 1
NW → NZ = 32MZ4 MW
2
Ncf G2F × 4 2 × . (13.68)
cW 2
Finally we have to substitute xW → xZ ≡ MZ /Mh . Now we are ready to write the decay rate,
exploiting MZ2 = MW
2 /c2
W

G2F 5 f 2
Γ(h → Zf u,d fu,d ) = 2
M N (g + gA )x?Z J (x?Z ), (13.69)
3π 3 Z c V
where x?Z = 0.7296 and J (x?Z ) = 0.00606. The inclusive decay rate Γ(h → ZZ ? ) is obtained summing
over Ncf (gV2 + gA
2 ). On can find that

Γ(h → ZZ ? ) ≈ 7.4 × 10−5 GeV, (13.70)

which corresponds to a branching ratio of 0.0182, suppressed by an order of magnitude with respect
to the W W ? case. All the results can be compared to experimental values, reported in http:
//pdg.lbl.gov/2018/reviews/rpp2018-rev-higgs-boson.pdf, with fair agreement (we are not
considering loop corrections to the propagators).

13.4.3 h → gg and h → γγ
We show in Fig. (13.7) the main channels of Higgs decays together with their branching ratios
computed at NLO, as functions of the Higgs boson mass. The direct comparison between LO and

1 _
bb
WW
BR(H)
ZZ
-1
10 + −
ττ
_
cc
tt-
gg
-2
10

γγ Zγ

-3
10
50 100 200 500 1000
MH [GeV]

Figure 13.7: Higgs decays branching ratios as functions of the Higgs boson mass, including NLO
corrections. Figure from https://arxiv.org/abs/hep-ph/9712334

NNLO computations is shown in the following table.


3
To see this, notice that in computing the trace T βν we obtained a factor of 2 from the multiplication (1 − γ5 )(1 − γ5 ):
instead now we shall have (gV2 + gA2
).
13.4. HIGGS DECAY MODES 185

Higgs decay mode ΓLO [MeV] ΓNNLO [MeV] Branching ratios [%]
h → bb 4.92 2.36 56.6
h → τ +τ − 0.26 0.26 6.2
h → cc 0.50 0.12 2.3
h → W ?W 0.28 0.34 22.5
h → Z ?Z 0.08 0.12 2.9
h → gg 0.22 0.35 8.5
h → γγ 0.0031 0.0036 0.23
Total 6.77 4.17 99.8

As we have already calculated the decays in fermion pairs as well as the inclusive decay rates in
boson pairs, we can focus now on the loop-induced processes h → gg and h → γγ. These are crucial
processes, even if they are loop-induced, since h → gg is related (M(h → gg) = M(gg → h)) to the
main production process at hadronic colliders such LHC and h → γγ is a very clean decay channel
through which the Higgs boson has been observed for the first time in 2012 at LHC. The two process
share some similarities. Let us start from the h → gg one. We have to compute the two triangle
diagrams (notice the inverted fermion arrows and the + sign) in which a virtual top quark is exchanged

k µ a g k µ a g

h t + h t (13.71)

ν a g ν a g

Firstly we can compute the degree of superficial divergence of the diagrams. A naive power counting
would suggest a linear divergence
1
Z
(13.72)
k→∞
Mhgg ∼ d4 k ∼ Λ
k/k/k/

However this is not the case. In particular let us notice that chirality arguments tell us that we need
to pick up at least one fermion mass in each propagator: Higgs couplings with fermions (Yukawa type
∼ yt htR tL ) flip chirality while gluon couplings conserve chirality (they do not care about it). Therefore,
since we impose chirality is conserved at gluon vertices and flipped at the Higgs vertex, we can place
either 1 insertion or 3 insertions:

tR tL tL tR
tL tL tL tR
tR tL
tR tR tR tL
tR tR tL tL (13.73)
| {z } | {z }

mt m3t 1
Z Z
4
d k 2 ∼ log Λ d4 k ∼ 2
k k/k/ k6 Λ
So we should have a logarithmic divergence from the first three diagrams with only one insertion, while
the last diagram is finite. However, also this ”improved” naive expectation is not accomplished since
the diagrams are convergent, i.e. finite.

Full calculation of the diagrams


Let us compute the first diagram in Eq. (13.71)

dd k Tr γ µ (k/ + p /3 + mt )γ ν (k/ + mt )
 
mt λa λb /2 + mt )(k/ − p
Z
M1 = (−i)gs2 εaµ (p2 )εbν (p3 ) (13.74)
(2π)d (k 2 − m2t ) (k + p2 )2 − m2t (k − p3 )2 − m2t
  
v |2{z2}
δab
2
186 CHAPTER 13. HIGGS PHYSICS

the trace reads

Tr[. . . ] = 4mt pµ3 pν2 + 4k µ k ν − 2k µ pν3 + 2pµ2 k ν − pµ2 pν3 + η µν (m2t − p2 · p3 ) − η µν k 2 ≡ 4mt N µν (13.75)


We exploit
1
Z 1 Z 1 Z 1 2
= dx dy dzδ(x + y + z − 1) (13.76)
ABC 0 0 0 [Ax + Bx + Cz]3
with
A = k 2 − m2t B = (k + p2 )2 − m2t C = (k − p3 )2 − m2t (13.77)
So
D = [Ax + By + Cz] =(k 2 − m2t )x + (k 2 + p22 − m2t + 2k · p2 )y + (k 2 + p23 − m2t − 2k · p3 )
p22 =0,p23 =0
= (k 2 − m2t )(x + y + z) + 2(k · p2 )y − 2(k · p3 )z (13.78)
2
=(k + p2 y − p3 z) + 2(p2 · p3 )yz − m2t 2
= (k + p2 y − p3 z) − a 2
| {z }
−a2

So we have to evaluate
dd k 1 1−y 8mt N µν
Z Z Z
µν
I ≡ dy dz  3
(2π)d 0 0 (k + p2 y − p3 z)2 − a2
(13.79)
dd k 1 1−y 8mt N µν (k → k − p2 y + p3 z)
Z Z Z
k→k−p2 y+p3 z
−→ dy dz
(2π)d 0 0 [k 2 − a2 ]3

Neglecting all terms linear in k which would vanish we obtain

N µν (k → k − p2 y + p3 z) =4k µ k ν − η µν k 2 + pµ3 pν2 (1 − 4yz) + pµ2 pν3 (−1 − 4yz + 2y + 2z)


+ pµ3 pν3 (4z 2 − 2z) + pµ2 pν2 (4y 2 − 2y) + η µν (m2t − p2 · p3 + 2p2 · p3 yz)
(13.80)
Note that the first two terms are the potential source of log divergence advertised before. We compute

dd k (4k µ k ν − η µν k 2 ) 4 i 0d d
Z    
= − 1 η µν (a2 ) Γ 2 − (13.81)
(2π)d (k 2 − a2 )3 d (4π)d/2 4 2

4 d d εd ε ε 2 1
     
−1 Γ 2− = Γ = · = (13.82)
d 4 2 d4 2 4  2
Therefore we can write
dd k (4k µ k ν − η µν k 2 ) i µν dd k 1 i 1
Z Z
= η =− (13.83)
(2π)d (k 2 − a2 )3 32π 2 (2π)d (k 2 − a2 )3 32π 2 a2

For on-shell gluons we have that ε2 · p2 = ε3 · p3 = 0, so the relevant I µν terms are

8imt 1 Z 1−y (pµ3 pν2 − η µν p2 · p3 )(1 − 4yz)


Z
I µν = dydz (13.84)
32π 2 2p2 · p3 yz − m2t
 
0 0

Therefore we find for M1 = M2


mt a
M = −igs2 ε (p2 )εbν (p3 )δab I µν (13.85)
2v µ
Then we use
(13.86)
X X
δab δab = 8 εa? a b? b
ρ (p2 )εµ (p2 )εσ (p3 )εν (p3 ) = ηµρ ησν
a,b spins

8m2t µν ? 4
4 mt (p2 · p3 )
2 1 2
(13.87)
X
|M1 + M2 |2 = gs4

I I µν = g s I(x t )
colors,spins
v2 v2 π4 4(p2 · p3 )2
13.4. HIGGS DECAY MODES 187

where xt = m2t /Mh2 and


Z 1 Z 1−y (1 − 4yz)
I(xt ) = dydz  m2t 
(13.88)
0 0 yz − 2p2 ·p 3

In the Higgs rest frame



µ ~
p1 = (Mh , 0)
 
 M = 2p Z
1
(13.89)
h
pµ2 = (p, p~) =⇒ Mh2 dQ2 = phase space
 p2 · p3 = 2p2 = 8π
pµ = (p, −~

p) 2
3

The exact result for the decay rate is therefore

α2 1 m4t 2 1
Γ(h → gg) = 2 2
I (xt ) · (13.90)
π 4π Mh v 2

where the 1/2 factor is because we have 2 identical particles in the final state. If we assume the
(unrealistic) case where xt = m2t /Mh2  1 (actually xt ' 2) we have
1 Z 1−y Mh2 1 Mh2
Z  
(13.91)
x 1
t
I(xt ) −→ dydz(1 − 4yz) · − =−
0 0 m2t 3 m2t

Hence in the heavy top limit

mt M αs2 Mh2 Mh
Γ(h → gg) −→ h ' 0.2 MeV (13.92)
π 2 v 2 72π
For the same input parameters the exact result provides 0.22 MeV, in excellent agreement with the
heavy top limit.

EFT approach to h → γγ, gg


To compute in an easy way the rates of the two processes and get the main physics out of it, we can
follow a EFT approach in the heavy top limit and heavy gauge boson limit MW , mt  Mh . Let us
focus on the h → γγ case. We have two additional diagrams with respect to the h → gg case:

t µ γ W µ γ W µ γ

h t + h W + h
t ν γ W ν γ W ν γ

which have to be taken into account all together to provide a reliable result. Let us focus, first of all
on the diagram containing the fermion loop. To compute the decay rate in our limit we can write the
effective Lagrangian containing a coupling betweeen the Higgs boson and the photon kinetic free term
1
Lm
eff
t ,MW Mh
= − Ah Fµν F µν (13.93)
4
which involves a 5-dimensional operator. Since [F µν ] = 2, [h] = 1 we have [A] = −1, therefore A has
to be proportional to the inverse of a mass scale. Since the cut-off of our theory is mt , one would be
tempted to assume that A ∼ m−1 t but indeed this is not the case. To understand this point we can
draw the Feynman diagram

tL tL yt mt 1
mt =⇒ A∼ 2 ∼ (13.94)
mt v
tR tR
188 CHAPTER 13. HIGGS PHYSICS

The fact that A ∼ yt mt is unavoidable since Yukawa interactions flip the fermion chirality while
electromagnetic interactions preserve it (same holds for the strong interactions in the gluon-gluon
case). The m2t in the denominator arises from dimensional reasons and the fact that the cut-off scale
is nevertheless mt . The fact that h → γγ and h → gg are not sensitive to yt , in constrast to the
naive expectation, tells us that they cannot be used to determine indirectly mt (we remember that
EWPT like ∆ρ receive loop contributions proportional to m2t ). Notice that this reasoning works only
in the limit Mh  2mt (and later for the limit Mh  2MW ) since in this case the loop momentum is
dominated by the top mass scale so indeed we can approximate yt (k/ + mt )/(k 2 − m2t ) ≈ yt /mt being
k  mt . Instead in the opposite limit the circulating momentum which is of the order (on average) of
Higgs boson we would have yt mt /Mh  1 so A gets suppressed.
Let us evaluate A in the mt  Mh limit. Notice that the diagram without the Higgs line is a
vacuum polarization-like diagram for the two final state (physical photons). Therefore we start from
the vacuum polarization diagram generalized from QED to the case of quarks and we will recover the
coupling to the Higgs external line. In QED the bare Lagrangian
1
L = − F µν Fµν (13.95)
4
receives 1-loop corrections from vacuum polarization

p+k
p p
2 2
ie Πµν (k ) = µ 2 2 2
ν = ie (pµ pν − k ηµν )Π(p ) (13.96)

k
We get from the QED result,
Π(p2 ) 1 2 1 m2 − p2 x(1 − x)
  Z  
2 = − γ + log 4π − 2 dx x(1 − x) log (13.97)
Nc Qf 12π 2 ε 2π µ2
| {z }
∆ε

and, in the p2 → 0
Π(0) ∆ε 1 m
2 = 2
− 2 log . (13.98)
Nc Qf 12π 6π µ
Now we have to remember that in the SM v is always accompaigned with the higgs boson: in particular
(v + h)
LY = −yf √ f L fR + h.c.. (13.99)
2

and m = yf v/ 2. We can expand Π[v](0) ≡ Π(0) at first order in h, seen as an expansion from the
vacuum expectation value
∂Π[v](0)
Π[v + h](0) ≈ Π[v](0) + h (13.100)
∂v
or diagrammatically

v+h ≈ + h (13.101)

So the term multiplying h is precisely the coupling A we are looking for


Π[v + h](0) ∆ε 1 yt (v + h) 1 ∂Π[v](0)
   
= − log √ = Π[v](0) + h + ... .
Nc Q2f 12π 2 6π 2 µ 2 Nc Q2f ∂v
(13.102)
∆ε 1 m 1
 
= − log − 2 h + ...
12π 2 6π 2 µ 6π v
13.4. HIGGS DECAY MODES 189

Notice that in the derivative both the divergent term ∼ ∆ε and the renormalization scale µ drop out.
We are left with a term proprtional to h, therefore we identify

∂Π[v](0)
A = e2 (13.103)
∂v

Considering h → γγ we get
Nc Q2t e2 2e2 1
Atγγ = − = − . (13.104)
6π 2 v 9π 2 v
This is indeed in agreement wih our NDA.
In order to compute the decay rate for the h → γγ process we need to evaluate the other two
diagrams contributing. In the limit mh  2MW , we obtain

∂ΠW [v](0) 7e2


AW
γγ = e
2
= 2 (13.105)
∂v 8π v

Notice that AWγγ has opposite sign and actually it is larger than Aγγ roughly by a factor of four. The
t

total contribution is
47e2
Aγγ = Atγγ + AW γγ = (13.106)
72π 2 v
To find the decay rate we have to write the amplitude of our process and to do this we need a Feynman
rule. We can get if from the Lagrangian

p2

h = −i(At + AW )(pµ1 pµ2 − p1 · p2 η µν ) (13.107)

p1

Then the unpolarized squared amplitude reads

1 X
|M|2 = |M|2 = |AW + At |2 ηµν ηαβ (pµ1 pα2 − (p1 · p2 )η µα )(pν1 pβ2 − (p1 · p2 )η νβ )
1 spins
  (13.108)
W t 2 W t 2
=|A +A | p21 p22 2 2
+ 4(p1 · p2 ) − (p1 · p2 ) − (p1 · p2 ) 2
= 2|A 2
+ A | (p1 · p2 )

having exploited the fact that on-shell photons are massless. Now we put ourselves in the h rest frame
where pµ1 = (Mh /2, ~q), pµ1 = (Mh /2, −~q) with |q| = Mh /2 and then
2
Mh2 Mh2 Mh4

|M|2 = 2|AW + At |2 + = |AW + At |2 (13.109)
4 4 2

Accounting for the phase space with two identical particles, the decay rate in the MW , mt  Mh limit
is given by
1 Mh 1
ΓMW ,mt Mh (h → γγ) = × |M|2 × 4π
2 2 32π 2 Mh2
(13.110)
|Atγγ + AWγγ |
2
α2 M 3 2 7 2
= Mh3 = em3 2h − + .
64π 4π v 9 8
Actually the Mh  2MW limit is not an excellent approximation. Corrections to this result originate
from the non-vanishing value of the Higgs mass, and can be expressed as a function of the ratio
ξ = Mh2 /m2 where m is either the top quark mass or the W mass. Notice that ξt ∼ 0.13 while ξW ∼ 0.6
for Mh = 125 GeV. It is clear then that the main correction comes from the W contribution, and can
only be estimated through a full calculation of the diagrams.
190 CHAPTER 13. HIGGS PHYSICS

The result for h → gg is obtained from the previous one considering only the first diagram with
the top quark loop, simply replacing Nc Q2t e2 → gs2 Tija Tjib = gs2 Tr[T a T b ] = gs2 TF δ ab with TF = 1/2 to
take into account the color structure of the amplitude. Therefore

gs2 1
Atgg = − δab (13.111)
12π 2 v
Finally the rate gives

|Atgg |2 3 X αs2 Mh3


Γmt Mh (h → gg) = Mh δab δab = 3 v2
, (13.112)
64π a,b
72π
| {z }
8

which agrees with the heavy top limit results obtained from the exact calculations.

You might also like