Universit`
a di Pisa
Dipartimento di Fisica “Enrico Fermi”
Tesi di Laurea Magistrale in Fisica
Axion from high-quality accidental
Peccei-Quinn symmetry
Candidato: Marco Ardu
Relatori : Prof. Alessandro Strumia, Dott. Daniele Teresi
Abstract
We propose two models where a U(1) Peccei-Quinn (PQ) global symmetry arises accidentally and is respected up to high-dimensional operators, so that the axion solution to the strong CP problem is successful even in the presence of Planck-suppressed operators. One model is SU(N ) gauge interactions with fermions in the fundamental and a scalar in the symmetric. The axion arises from spontaneous symmetry breaking to SO(N ), that confines at a lower energy scale. Axion quality in the model needs N & 10. SO bound states and possibly monopoles provide extra Dark Matter (DM) candidates beyond the axion. In the second model the scalar is in the anti-symmetric: SU(N ) broken to Sp(N ) needs even N & 20. The cosmological DM abundance, consisting of axions and/or super-heavy relics, can be reproduced if the PQ symmetry is broken before inflation (Boltzmann-suppressed production of super-heavy relics) or after (super-heavy relics in thermal equilibrium get partially diluted by dark glue-ball decays).
Ringraziamenti
Vorrei ringraziare i miei relatori, Prof. Alessandro Strumia e Dott. Daniele Teresi, per avermi guidato, con la loro disponibilit`a, conoscenza ed esperienza, nella stesura di questa tesi. Posso solo sperare di avvicinarmi alla vostra padronanza e comprensione della Fisica ad Alte Energie.
Ringrazio Luca di Luzio, Giacomo Landini e Jin-Wei Wang, con i quali ho condiviso la mia prima esperienza nella scrittura di un articolo scientifico, sempre disponibili nel chiarire i miei dubbi.
Un ringraziamento speciale alla mia famiglia, in particolare ai miei genitori, per l’immenso supporto psicologico ed economico di questi 5 anni. Ogni mio traguardo `e anche merito vostro.
Grazie ad amici e colleghi, per avermi accompagnato in questi anni. Poter condividere con voi le mete raggiunte `e ci`o che le rende veramente speciali.
Un grazie infinito ad Eleonora, per esser stata sempre al mio fianco. Non credo di meritare quanto mi fai sentire amato e il tuo supporto mi tiene sano nei momenti pi`u difficili.
Contents
1 Introduction 5
2 The U(1)A problem 7
2.1 Quantum Chromodynamics . . . 7
2.1.1 Non-Abelian SU(Nc) gauge theory . . . 7
2.1.2 Discrete symmetries . . . 10
2.1.3 Perturbative QCD and running coupling . . . 12
2.2 Continuous chiral symmetries . . . 13
2.2.1 Quantum Symmetries and Anomalies . . . 14
2.2.2 Spontaneous symmetry breaking and the effective chiral lagrangian . 19 2.2.3 The U(1)A problem . . . 25
3 Instantons and the strong CP Problem 26 3.1 Classical instantons . . . 26
3.1.1 The Cartan-Maurer integral . . . 27
3.1.2 Bogomol’ny inequality . . . 30
3.2 The Strong CP problem . . . 31
3.2.1 Topological term . . . 31
3.2.2 Neutron dipole moment . . . 33
3.2.3 Strong CP problem . . . 34
4 The Axion solution and quality problem 35 4.1 Axion models . . . 35
4.1.1 General EFT formalism . . . 35
4.1.2 The original PQ Model . . . 38
4.1.3 DFSZ axion models . . . 40
4.1.4 KSVZ axion models . . . 40
4.2 Axion quality problem . . . 41
5 Axion cosmology and Dark Matter 44 5.1 Standard Cosmology . . . 44
5.1.1 Forms of matter . . . 45
5.1.2 Thermal Equilibrium . . . 46
5.1.3 Boltzmann equations . . . 48
5.1.4 Freeze out of massive particles . . . 50
5.1.5 Big Bang Nucleosynthesis . . . 53
5.2 Axion cosmology . . . 54
5.2.1 Domain Walls (DWs) and Strings . . . 55
5.2.2 Vacuum Misalignment Mechanism . . . 56
Contents Chapter 0
6 Accidental axion with high-quality PQ symmetry 60
6.1 The model . . . 60
6.1.1 Anomalies . . . 60
6.2 Symmetric scalar that breaks SU(N ) → SO(N ) . . . . 61
6.2.1 Spontaneous symmetry breaking . . . 63
6.2.2 Tree level spectrum and interactions . . . 64
6.2.3 SM running couplings and Landau Poles (LP) . . . 66
6.2.4 Confinement and bound states . . . 67
6.2.5 Axion effective Lagrangian . . . 68
6.2.6 Higher dimensional operators . . . 68
6.2.7 Cosmology and Dark Matter . . . 72
6.3 Anti-symmetric scalar that breaks SU(N ) → Sp(N ) . . . . 78
6.3.1 Spontaneous symmetry breaking . . . 79
6.3.2 Tree level spectrum and interactions . . . 79
6.3.3 Confinement and bound states . . . 81
6.3.4 Axion effective lagrangian . . . 81
6.3.5 Higher dimensional operator . . . 81
6.3.6 Cosmology and Dark Matter . . . 82
7 Conclusion 85 Appendices 89 A Lie group algebras 90 A.1 SU(N ) . . . . 90
A.2 SO(N ) . . . . 90
Chapter 1
Introduction
A deeper understanding of the vacuum structure of Quantum Chromodynamics, closely related to the discovery of instanton field configurations [1], has furnished a solution to the so-called ‘ U(1)A problem’ [2], while introducing a potential CP-violating operator in
the QCD Lagrangian. The value of the θ angle, the relevant parameter that quantifies the (strong) CP violation, is strictly bounded by the non-observation of the neutron dipole electric moment dn< 1.8× 10−26 e cm [3], which impliesθ < 10−10 [4, 5].
Given that the possible values of θ belong to the interval [0, 2π], such a small value is regarded as highly unnatural. One of the most theoretically appealing way to explain the smallness of the θ parameter, which is still not ruled out experimentally, is known as the Peccei-Quinn (PQ) mechanism [6,7].
A global U(1)PQ symmetry is postulated, anomalous under SU(3)c and spontaneously
broken at some energy scale fa. The corresponding pNGB goldstone in the spectrum, the
axion [8, 9], acquires a VEV which actively cancels out the CP-violating term, solving dynamically the strong CP problem (also providing a Dark Matter (DM) candidate)
However, global symmetries are not regarded as fundamental features of Quantum Field Theory, but they rather arise accidentally in gauge theories, famous examples being the baryon and lepton numbers.
This conflicts with a global U(1)PQsymmetry that must be realized at very high precision
to relax the topological θ-term. Adopting an Effective Field Theory (EFT) perspective, new physics can generate higher-dimensional operators suppressed by some energy scale ΛUV,
generally breaking the accidental U(1)PQ symmetry of the renormalizable Lagrangian. The
strong CP problem solution requires the UV explicit breaking contribution to the axion potential to be 10−10 smaller than the QCD energy density
fa ΛUV d−4 f4 a . 10−10Λ 4 QCD (1.1)
estimated for a d dimensional breaking operator. Assuming new physics at the Planck’s scale ΛUV ' MPl = 1.2× 1019 GeV and taking into account the phenomenological bound
fa & 109 GeV [10], d & 9 is needed to satisfy eq. (1.1). This is known as the PQ quality
problem [11, 12,13, 14].
This encourages us to seek axion models in which the PQ symmetry is an accidental symmetry of the Lagrangian, protected by early explicit breaking by some fundamental principle, for instance, gauge invariance. In the present work, we propose a gauge theory based on the SM gauge group GSMplus a SU(N ) [15], in which an accidental chiral U(1)PQ
symmetry it’s protected by higher dimensional operator up to N dependent dimension, in principle arbitrary.
Chapter 1
The thesis is outlined as follows: in Chapter 2 a brief review of the U(1)A problem is
laid out, along with some aspect of chiral perturbation theory.
The discovery of the instanton solutions and how their inclusion in the path integral results in the strong CP problem is discussed in Chapter 3.
In Chapter 4 the general approach of an axion solution is presented, with the original Peccei-Quinn model as well as the more popular invisible axion models. Furthermore, a more detailed discussion on the PQ quality problem is given.
In Chapter5we give a summary of Early Universe cosmology, discuss thermal production of Dark Matter and axion cosmology.
Finally, in Chapter 6 we present two model with a high quality PQ symmetry, studying their phenomenological implication.
Chapter 2
The U(1)
A
problem
2.1
Quantum Chromodynamics
Quantum Chromodynamics (QCD) has been established, through numerous experimental results, theoretical arguments and lattice simulations, as the description for the strong inter-action. At low energies, the strong force is the strongest of all fundamental interactions and it’s responsible for the binding of quarks inside hadrons, and for the interaction between the latters, thanks to which protons and neutrons are held together to form the nuclei of atoms. QCD is a quantum field theory that features Nf = 6 fermion fields, the quarks, that
carry Nc = 3 colors. The interactions between the quarks, mediated by the 8 gluons, arise
by the requirement of invariance under a local (position dependent) SU(Nc) transformation
of the quark fields.
2.1.1
Non-Abelian SU(
N
c) gauge theory
Let us consider the lagrangian for Nf free fermions withNc colors
L0 F = Nf X f =1 Nc X i=1 ¯ ψf i(x)(iγµ∂µ− mf)ψf i(x) (2.1)
where the ψ-s are grassmann variables. The theory is invariant under a global U(Nc)
trans-formation
ψf i(x)→ ψ0f i(x) = Uijψf j(x) (2.2)
with U∗
jiUjk =δik. To lighten the notation, let us define
Ψf ≡ ψf 1 .. . ψf Nc Ψ 0 f =U Ψf with U†U = U U† =1Nc×Nc (2.3)
If we focus on the SU(Nc) subgroup (i.e U ∈ U(Nc) and detU = 1), every element can be
expressed in terms of the Lie algebra generators
U (θ) = exp N2 c−1 X a=1 iθaTa Ta, Tb = ifabcTc Tr TaTb = 1 2δab (2.4)
2.1. Quantum Chromodynamics Chapter 2
where Ta are tracelessN
c× Nc hermitian matrices and fabc is completely anti-symmetric in
a, b, c. The Nc colors furnish the defining representation of SU(Nc).
Clearly, the local version of the group action is not a symmetry of the Lagrangian, because if we perform a position dependent transformation
Ψf → Ψ0f(x) = U (x)Ψf(x) (2.5)
the derivative ∂µΨ does not transform as Ψ. This is inherently linked to the definition of
the derivative, which involves the sum of fields defined in different space-time points (xµ and
xµ+dxµ), that do not transform covariantly.
Parallel Transport
Given two space-time points xµ and yµ, a path C
y←x:zµ(s) such that s∈ [0, 1], zµ(0) =xµ
and zµ(1) =yµ, we define the parallel transport W (C
y←x) as:
1. W (Cy←x)∈ SU(Nc)
2. W (∅) = 1, where ∅ represents any curve with zero measure 3. If Cy←x =Cy←w◦ Cw←x → W (Cy←x) = W (Cy←w)W (Cw←x)
4. If we defineC−1
x←y :zµ(1− s) → W (Cx−1←y) = W (Cy←x)−1
5. Under a gauge trasformation W (Cy←x)→ W (Cy←x)0 =U (y)W (Cy←x)U−1(x)
It follows from the property 5) that ˜Ψ(y)≡ W (Cy←x)Ψ(x) transforms under a gauge
trans-formation as Ψ(y).
Since W belongs to the gauge group, we can write it as the exponential of a matrix in the Lie algebra W (Cy←x) = exp(iA(x, y)). If we consider a straight infinitesimal path from
xµ toxµ+dxµ, we expect the parallel transport to be infinitesimally close to the identity
W (Cx+dx←x)' 1 + igsAµ(x)dxµ ' eigsAµ(x)dx
µ
(2.6)
where all equalities are true at first order in dx and the factorization of gs is a matter of
convention. Aµ is known as the connection and lives in the vector space spanned by the
generators, it can therefore be expanded as Aµ(x)≡ Aaµ(x)Ta. TheAaµ are the Nc2− 1 gluon
fields.
From the transformation rule of the parallel transport
W0(Cx+dx←x) =U (x + dx)W (Cx+dx←x)U−1(x), (2.7)
it is straightforward to obtain the gauge transformation of the connection, expanding eq. (2.7) at first order indx
A0µ(x) = U (x)Aµ(x)U−1(x)−
i gs
(∂µU (x))U−1(x). (2.8)
We can now define a covariant differential
DΨ(x) ≡ W (Cx−1←x+dx)Ψ(x + dx)− Ψ(x) (2.9)
2.1. Quantum Chromodynamics Chapter 2
and, consequently, a covariant derivative Dµ such that DΨ(x) = DµΨ(x)dxµ
Dµ=∂µ− igsAµ=∂µ− igsAaµT a
. (2.10)
By definition, DµΨ(x) transforms as Ψ(x), therefore
LF = Nf X f =1 ¯ Ψf(iγµDµ− mf)Ψf =LF0+ Nf X f =1 gsAaµΨ¯fTaγµΨf (2.11) is gauge invariant.
Pure gauge lagrangian
We were able to extend the global symmetry to a local one introducing N2
c − 1 new spin-1
fields that interacts with the quarks. We need to introduce the kinetic terms of the gluons, without spoiling the local symmetry. Actually, since we are using gauge invariance as our guiding principle to construct the theory, we should add every term allowed by the gauge symmetry, stopping at the renormalizable level if we want the theory to be predictive.
The following identity
[Dµ, Dν]Ψ = [∂µ− igsAµ, ∂ν − igsAν]Ψ = −igs(∂µAν − ∂νAµ− igs[Aµ, Aν])Ψ (2.12)
holds for every Ψ, such that
[Dµ, Dν] =−igs(∂µAν − ∂νAµ− igs[Aµ, Aν]) =−igsGµν. (2.13)
The anti-symmetric tensor Gµν is known as the curvature and its transformation under the
gauge group is easily found by observing that D0
µ=U DµU−1 G0µν = i gs D0 µ, D0ν = i gs U [Dµ, Dν]U−1 =U GµνU−1. (2.14)
The curvature transforms in the adjoint of the gauge group and it can also be expanded in terms of the generators as
Gµν =GaµνTa = (∂µAaν − ∂νAaµ+gsfabcAbµAcν)Ta. (2.15)
We can now construct a gauge invariant Lorentz scalar
− 12Tr(GµνGµν) =− 1 4G a µνG aµν (2.16)
where the coefficient in front has been chosen to canonically normalized the kinetic term. The gauge symmetry allows an analogous term, built using the field tensor and its dual
µνρσTr(GρσGµν) = 2 Tr ˜GµνGµν
. (2.17)
We can further expand it in terms of the connection
µνρσTr(GρσGµν) =4µνρσTr((∂µAν − igAµAν)(∂ρAσ − igAρAσ))
2.1. Quantum Chromodynamics Chapter 2
having used that the term containing four connections is zero, because any cyclic permutation of the 4 indices leaves the trace invariant but gets a − from the tensor.
From eq. (2.18) it’s easy to see that
µνρσTr(GρσGµν) =∂µKµ with Kµ= 4µνρσTr Aν∂ρAσ −2igs 3 A νAρAσ (2.19)
where Kµ is known as the Chern-Simons current. Naively, one can think that since this
term is a total derivative, its integral over all space-time vanishes. However, as it will be discussed in the next sections, there exist topologically non trivial field configurations, known as instantons, with a non vanishing R ∂µKµd4x and finite action [1].
In the original QCD Lagrangian, this CP odd term1 was not considered, so that P and CP were automatically conserved as was experimentally verified in strong processes, scoring a feature that was considered one of the major success of QCD. What will be explored in this and the next chapter is that simply disregarding this term is inconsistent with the existence of instanton solutions and with the heaviness of the η mesons, a puzzle dubbed the ‘ U(1)A
problem’ [2].
Explaining the absence (or a very tiny presence) of CP violation in the strong interaction is what ultimately motivates the introduction of the axion.
Finally, the full QCD Lagrangian density reads
LQCD =− 1 4G a µνG aµν+ Nf X f =1 ¯ Ψf(i /D− mf)Ψf +θ g2 s 32π2G˜ aµνGa µν (2.20)
where the choice of normalization of the last term foreshadows the upcoming discussions.
2.1.2
Discrete symmetries
In this section we discuss the properties of the QCD lagrangian density under discrete trans-formation: ˆC, ˆP and ˆT .
Parity
Parity corresponds to the change of sign of the spatial components of the 4-position
xµ = (x0, ~x)−→ xPˆ µ0
= (x0,−~x) xµ −→ gPˆ µµxµ. (2.21)
Summation convention is suspended in writing the second expression. In the Hilbert space of a quantum theory, parity acts as a unitary (and hermitian) operator U ( ˆP ). On the quark and gluon fields 2
U ( ˆP ) ˆΨ(x0, ~x)U†( ˆP ) = γ0Ψ(xˆ 0, −~x) U ( ˆP )Aaµ(x 0 , ~x)U†( ˆP ) = gµµAˆaµ(x 0 ,−~x). (2.22)
We remind that we are working in the Weyl representation for the γ matrices
1Discrete symmetries of the lagrangian will be discussed in the next section
2Fermion transformations under discrete transformation can be derived imposing the invariance of the
free Dirac action
2.1. Quantum Chromodynamics Chapter 2
Charge conjugation
The charge conjugation effectively switches a particle with its antiparticle. On fermions acts as
U ( ˆC) ˆΨ(x0, ~x)U†( ˆC) = iγ2Ψˆ†(x0, ~x). (2.23)
In the previous section we have derived the expression for the covariant derivative using the parallel transport, defined to act on fermion fields that transforms in the fundamental rep-resentation of SU(Nc). Clearly, the same procedure can be applied in general, the difference
being that the generators have to be considered in the desired representation.
If under charge conjugation fermions transform as stated in eq. (2.23), the covariant derivative should now act on a field that transforms in the anti-fundamental
Dµ ˆ C
−→ Dµ∗. (2.24)
The generators of the fundamental of SU(Nc) are hermitian, tracelessNc×Ncmatrices. Some
are real and symmetricTa
real, while the others are purely imaginary and anti-symmetricT b imag.
Since in the anti-fundamental the generators are −Ta∗ 3, the conjugation effectively acts on the corresponding gluon fields as
U ( ˆC) ˆAaµreal(x0, ~x)U†( ˆC)) =− ˆAaµ real
U ( ˆC) ˆAaµimag(x0, ~x)U†( ˆC) = ˆAaµ
imag. (2.25)
Time Reversal
Time reversal corresponds to changing the sign of time
x0 Tˆ
−→ −x0 xµ Tˆ
−→ −(gµµ)xµ. (2.26)
In this case, on quantum states, time reversal acts as an anti-unitary operator. On quark and gluon fields
U ( ˆT ) ˆΨ(x0, ~x)U†( ˆT ) = γ1γ3Ψ(ˆ
−x0, ~x)
U ( ˆT ) ˆAa
µ(x0, ~x)U†( ˆT ) = gµµAˆaµ(−x0, ~x). (2.27)
QCD discrete symmetries
We can now determine the symmetry properties of the QCD lagrangian under discrete trans-formations. FaµνFa µν ˆ P , ˆT −−→ g2 µµg 2 ννF aµνFa µν =F aµνFa µν (2.28) µνρσFa µνF a ρσ ˆ P , ˆT −−→ µνρσg µµgννgρρgσσFµνa F a ρσ =− µνρσFa µνF a ρσ (2.29)
where space-time coordinates are understood to be properly transformed. In the last equality of eq. (2.29) has been used that the four indicesµ, ν, ρ, σ are all different due to the contrac-tion with the tensor. For every local, Lorentz invariant and with a hermitian hamiltonian QFT, ˆC ˆP ˆT is a symmetry of the theory [16, 17].
Consequently, the two terms in eq.s (2.28) and (2.29) are, respectively, CP (CT) even and CP odd.
2.1. Quantum Chromodynamics Chapter 2 Bilinear Cˆ Pˆ Tˆ ¯ ΨΨ + + + ¯ Ψγ5Ψ + - -¯ ΨγµΨ - g µµ −gµµ ¯ ΨTa realγ µΨ - g µµ −gµµ ¯ ΨTa imagγµΨ + gµµ −gµµ
Table 2.1: Transformation properties of fermion bilinear
For fermions bilinears the symmetries properties are summarized in Table2.1.
We conclude by observing that every term in eq. (2.20), with the exception of theθ term, is separately invariant under all discrete transformations. However, if the mass matrix is complex, the mass term contains a CP odd operator. Indeed, in such case
− Lmass= ¯ΨLMΨR+ ¯ΨRM†ΨL (2.30) that is equal to −Lmass = ¯Ψ(A + iγ5B)Ψ (2.31) A = A† = M + M † 2 (2.32) B = B†=−iM − M † 2 . (2.33)
The term involving B is CP odd and we will see that can be rotated away and absorbed in the physical ¯θ definition. In the next sections we ignore every source of ˆC ˆP violation, hence we work with the QCD lagrangian with θ = 0 and with a real, diagonal, mass matrix in flavour space.
2.1.3
Perturbative QCD and running coupling
As in any interacting quantum field theory, vacuum-to-vacuum amplitude of time ordered products of gluon and quark fields can be computed perturbatively through Feynman dia-grams, considering a formal expansion in powers of the coupling gs.
UV divergences arise in computing loop diagrams, and a regularization procedure is nec-essary to deal with the divergent integrals. A UV cut-off Λ or a renormalization scale µ are introduced in the usual regularization schemes, and the parameters of the theory are renormalized, which means redefined in such a way that all correlation functions are finite when the regulator is removed. Many different renormalization schemes are possible.
The Lagrangian parameters and field normalizations are infinite and known as bare gs → gsB Aµ→ A
µ
B Ψ→ ΨB mf → mf B. (2.34)
They can be related to the renormalized fields and parameters, that are finite and appear in physical observables, through infinite scheme dependent multiplicative factor
gsB =ZggsR AµB =Z 1/2 A A µ R ΨB =Z 1/2 ψ ΨR mf B =Zmmf R. (2.35) 3U (θ)∗ = exp(iθa(
−Ta∗)), by conjugating the commutation relations for the Ta we get−Ta∗,
−Tb∗ =
ifabc(
−Tc∗)
2.2. Continuous chiral symmetries Chapter 2
Clearly, the underlying theory should not depend by the arbitrary regulator introduced. This statement, applied to the bare correlation functions, results in the Renormalization Group Equation (RGE), that relates renormalized correlation functions at different energy scales. One important consequence of the RGE solution is that, effectively, the renormalized coupling of the theory depends on the energy scale through the differential equation
µdgR(µ)
dµ =β(gR(µ)). (2.36)
β(gR(µ)) is accessible through perturbative calculations and admits an expansion in powers
of the renormalized coupling
β(gsR) =−β0gsR3 − β1gsR5 − β2g7sR+O g 9
sR. (2.37)
The coefficients of the expansion are in general scheme depedent, but it can be shown that β0, β1 are not. A two loop calculation yields [18]
β0 = 1 (4π)2 11Nc− 2Nf 3 β1 = 1 (4π)4 34 3 N 2 c − 13 3 NcNf − Nf Nc . (2.38)
For Nc = 3 and Nf = 6, β0 > 0 which implies that the theory is asymptotically free. To
lowest order
µdgR(µ)
dµ =−β0g
3
sR. (2.39)
Integrating it, we find
µe−12β0g2sR =cost. = ΛQCD → g2 sR(µ) = 1 2β0ln µ ΛQCD . (2.40)
From eq. (2.40) asymptotic freedom is apparent and we identify a physical mass scale, ΛQCD,
at which the theory is no longer perturbative.
2.2
Continuous chiral symmetries
In order to explore the continuous global symmetry of QCD, let us rewrite the fermion sector of LQCD in terms of chiral eigenstates Ψ = ΨL+ ΨR, whereγ5ΨL =−ΨL and γ5ΨR= ΨR.
LF = ¯ΨRDΨ/ R+ ¯ΨLDΨ/ L− ¯ΨRMΨL− ¯ΨLMΨR. (2.41)
The mass matrix is Mf f0 = mfδf f0 and flavour indices are implied. In general, the only
classical continuous symmetry is
Nf
Y
f
U(1)f, under which
Ψf U(1)f
−−−→ eiαfΨ
f. (2.42)
The combination of all U(1)f such that all flavours have the same charges U(1)V, corresponds
2.2. Continuous chiral symmetries Chapter 2
Note that if M ∝ 1 (mf =m for every f ) or M = 0 , the global symmetry group G is
much larger G = U(Nf)L⊗ U(Nf)R. ForUL ∈ U(Nf) andUR∈ U(Nf), the action of G on
the quark fields is
ΨL → Ψ0L =ULΨL Ψ¯L → ¯Ψ0L = ¯ΨLUL† (2.43)
ΨR → Ψ0R =URΨR Ψ¯R → ¯Ψ0R = ¯ΨRUR† (2.44)
G can be decomposed as G = U(Nf)L⊗ U(Nf)R= SU(Nf)L⊗ SU(Nf)R⊗ U(1)L⊗ U(1)R=
SU(Nf)V ⊗ SU(Nf)A⊗ U(1)V ⊗ U(1)A, in which
SU(Nf)V : ( ΨL→ V ΨL ΨR→ V ΨR SU(Nf)A: ( ΨL→ AΨL ΨR→ A†ΨR (2.45) U(1)V : ( ΨL→ eiαΨL ΨR→ eiαV ΨR U(1)A : ( ΨL→ eiβΨL ΨR→ e−iβΨR (2.46)
where V, A ∈ SU(Nf). Writing V and A as exponential, we shorten the notation using the
full Dirac spinor
SU(Nf)V : Ψ→ eiα
aTa
Ψ SU(Nf)A : Ψ→ eiβ
aTaγ5
Ψ (2.47)
U(1)V : Ψ→ eiαΨ U(1)A: Ψ→ eiβγ
5
Ψ. (2.48)
Many experimental and theoretical results, some of which will be mentioned in the upcoming discussion, suggest that the quark masses are neither equal nor zero, soG is not a symmetry. However, the masses of the lighter flavours (up, down and strange4) can be considered small compare to the physical scale ΛQCD of the theory, allowing us to treat the chiral symmetry
as an approximate symmetry, that is explicitly broken by the quark masses.
Even if we consider their masses negligible (m → 0, know as the chiral limit), we have to discuss how the continuous symmetry is realized at the quantum level.
2.2.1
Quantum Symmetries and Anomalies
Classically, Noether’s theorem states that if the Lagrangian is invariant under a global trans-formation
Φ→ Ψ + faδaΦ. (2.49)
we can find a set of conserved currents
Jµ a =−
∂L ∂(∂µΦ)
δaΦ ∂µJbµ = 0 (2.50)
and define the corresponding constant charges Qb
Qb(x0) = Qb =
Z
d3xJ(x0, ~x). (2.51)
A quantum field theory is defined through the partition function
Z = Z [DΦ] exp(iS(Φ, ∂Φ)) = Z [DΦ] exp i Z d4xL (Φ, ∂Φ) (2.52)
4This assumption for the strange quark is not very accurate, indeed the SU(3) flavour symmetry is more
badly realized than the SU(2) of Isospin.
2.2. Continuous chiral symmetries Chapter 2
where Φ represents all fields. If we perform a change of variables
Φ(x)→ Φ(x) + fa(x)δ
aΦ(x) (2.53)
with fa(x) arbitrary infinitesimal functions, the Lagrangian is replaced by
L (Φ, ∂Φ) → L (Φ, ∂Φ) +δL δΦf aδ aΦ + δL δ(∂µΦ) ∂µ(faδaΦ). (2.54)
If we assume that fa(x) goes to zero sufficiently fast at infinity for every a, we can partial
integrate to find Z d4x δL δ(∂µΦ) ∂µ(faδaΦ) = Z d4x δL δ(∂µΦ) ∂µ(δaΦ)− ∂µ δL δ(∂µΦ) δaΦ fa. (2.55) Defining δaL ≡ δL δΦδaΦ + δL δ(∂µΦ) ∂µ(δaΦ) (2.56)
and recognizing the classical Noether’s current in eq. (2.55), we see that Z d4xL (Φ, ∂Φ) → Z d4x [L (Φ, ∂Φ) + (δ aL + ∂µJaµ)f a]. (2.57)
Under a local change of variables, the measure in the path integral can also transform non trivially, which can be parameterize introducing a functionAa(x) known as anomaly, defined
such that [DΦ]→ [DΦ] exp i Z d4x Aa(x)fa(x) . (2.58)
The partition function, which transforms as
Z → Z [DΦ] exp i Z d4x (δaL + ∂ µJaµ+Aa)fa exp(iS(Φ, ∂Φ)) (2.59)
has to be invariant under a change of dummy variables, so that expanding the exponential at first order we find
Z [DΦ]
Z
d4x (δaL + ∂
µJaµ+Aa)faexp(iS(Φ, ∂Φ)) = 0. (2.60)
This has to be true for every choice of the arbitrary functions fa, leading to the condition
Z [DΦ] (δaL + ∂µJaµ+Aa) exp(iS(Φ, ∂Φ)) = 0 D ∂µJˆaµ E =−DδaL + ˆˆ Aa E . (2.61)
Even if the classical theory is invariant under the corresponding global transformation with constant fa
Φ(x)→ Φ(x) + fa
δaΦ(x) (2.62)
which means thatδaL = 0, in the quantum theory the Noether’s currents are not necessarily
2.2. Continuous chiral symmetries Chapter 2
Chiral Anomaly in a gauge theory
Let us consider a transformation on fermionic fields that fill a representation R under a gauge group G
Ψm(x)→ Umn(x)Ψn(x) (2.63)
wherem, n are collective indices that runs over all relevant spaces, for example in QCD they represents flavour, color and spinor indices. The transformation is equivalently written as
Ψm(x)→
Z
d4y δ4(x− y)U
mn(y)Ψn(y)≡ Uxm,ynΨyn (2.64)
where in defining Uxm,yn we formally extended the summation convention to the space-time
coordinates. Correspondingly ¯
Ψxm → ¯Ψyn U¯yn,xm where U¯yn,xm ≡ (γ0U†(x)γ0)nmδ4(x− y). (2.65)
In changing variables in Berezin integration, we have to multiply for the inverse of the Jacobian
[D ¯ΨDΨ]→ (det ¯U det(U))−1[D ¯ΨDΨ]. (2.66)
Let us consider a unitary non-chiral transformation U , meaning
U (x) = exp(iα(x)t) (2.67)
whereα(x) is an arbitrary function and t is an hermitian matrix which is equal to the identity in Dirac space. In this case
¯
UU = 1 (2.68)
hence the change of variables amount to a trivial transformation of the Berezin measure. On the other hand, for a chiral unitary transformation
U (x) = exp iα(x)γ5t (2.69) we find that ¯ U = U (2.70) therefore [D ¯ΨDΨ]→ (det U)−2[D ¯ΨDΨ]. (2.71)
The Dirac delta represents the identity matrix for the space-time coordinatesδ4(x−y) ≡ 1 xy
and the following formal equality holds
exp(A)⊗ 1xy = exp(A⊗ 1xy). (2.72)
Using it in writing U and employing the identity det ≡ exp Tr ln , we can write detU = exp Tr ln exp iα(x)γ5t⊗ 1xy = exp iα(x) Tr tγ5δ4(x− y)
(2.73)
where the trace is extend over all indices. In eq. (2.73) we identify the anomaly as defined in the previous section
A(x) = −2 Tr tγ5δ4(x− x). (2.74)
It is not very surprising that, as a consequence of the hand-waving manipulations that we have performed, the final result is ill-defined in dealing with continuum space-time coor-dinates. We need to give meaning to eq. (2.74) introducing a regularization for the delta
2.2. Continuous chiral symmetries Chapter 2
function. There is hope that the final result will be finite when removing the regulator, indeed while it’s true that the delta function is infinite, the trace of γ5 vanishes. The delta
function diverges in the UV, so we introduce a cut off Λ and a smooth function f (s) that satisfies the following conditions
f (0) = 1 lim s→∞f (s) = 0 (2.75) lim s→∞sf 0(s) = lim s→0sf 0(s) = 0. (2.76)
We regulate the anomaly as
A(x) = −2hTrntγ5f ( /D2 x/Λ2)
o
δ4(x− y)i
y→x (2.77)
where /Dx is the covariant derivative in the representation of the fermion fields Ψ and we
sub-stitute y→ x only after the action of f(− /D2x/Λ2) on the right. This choice of regularization
is manifestly gauge invariant. In momentum space
A(x) = −2 Z
d4k
(2π)4Trtγ 5
f (−(/∂x− ig /A(x))2/Λ2) eik(x−y)
y→x (2.78) =−2 Z d4k (2π)4 e
ik(x−y)Trtγ5f ((i/k + /∂
x− ig /A(x))2/Λ2) y→x (2.79) =−2 Z d4k (2π)4 Trtγ 5f ((i/k + /D x)2/Λ2) . (2.80)
After performing the change of variables k = k0Λ, we find
A(x) = −2 Z d4k0 (2π)4 Λ 4Trtγ5f ((i/k0+ /D x/Λ)2) (2.81)
and for notation’s sake we relabel k0 → k. Observing that, for two generic operators T µ Tν0, the product (γµT µ)(γνTν0) can be written as (γµT µ)(γνTν0) = 1 4([Tµ, T 0 ν][γ µ, γν] + {Tµ, Tν0}{γ µ, γν }) = 1 4 [Tµ, T 0 ν][γ µ, γν] + 2Tµ, T0 µ (2.82) we work out the argument of the function f
i/k + D/x Λ 2 =−k2+ 2ik· Dx Λ + / D2x Λ2 (2.83) =−k2+ 2ik· Dx Λ + D2 x Λ2 − ig 4Λ2Gµν[γ µ , γν] (2.84) ≡ −k2 + ∆. (2.85)
Since the function f is smooth, it admits a Taylor expansion
f (−k2+ ∆) =f (
−k2) +f0(−k2)∆ +f00(−k 2)
2 ∆
2 +. . . (2.86)
Higher orders either give a trivial contribution when taking the limit Λ → ∞ or do not contain the minimum number (4) of gamma matrices to have a non zero trace with the γ5.
2.2. Continuous chiral symmetries Chapter 2
The last condition is never satisfied at first order, and we therefore disregard it. At the end, the only non zero contribution is given by the square of the last term in eq. (2.84)
A(x) = g 2 16TrD(γ 5[γµ, γν][γρ, γσ]) Tr{tF µνFρσ} Z d4k (2π)4f 00(−k2) (2.87)
where TrDrepresents the trace over Dirac space. The proper way to evaluate the integral is to
work from the beginning in the euclidean path integral and perform an analytic continuation when going back to Minkowski space. Effectively, we can consider the rotation k0 → ikE4
where kE4 runs from −∞ to +∞5, whileki = (~kE)i and k2E =kE42 +|~kE|2 =−k2
Z d4kf00(−k2) =i Z d4k Ef00(kE2) = i Z ∞ 0 2π2k3 Ef00(kE2)dkE =iπ2 Z ∞ 0 sf00(s)ds (2.88) =iπ2sf0(s) ∞ 0 − iπ2 Z ∞ 0 f0(s)ds (2.89) =iπ2 (2.90)
where we have employed the properties of f written in eq. (2.76). Considering that
TrD(γ5[γµ, γν][γρ, γσ]) = 16iµνρσ (2.91)
and that the fermion fields fill the representation R under the gauge group
Gµν =GaµνTaR (2.92) we finally find A(x) = − g 2 16π2 µνρσ GaµνG b ρσTrtTaRTbR . (2.93)
Let us now consider a gauge theory involving only chiral fermions, for instance left-handed, in a representation R of the gauge group
L ∼ ¯ψLi /DψL. (2.94)
The anomaly of a symmetry transformation on ψL can be computed introducing a dummy
right handed component that do not couple with the gauge fields
Ldum ∼ ¯ψLi /DψL+ ¯ψRi/∂ψR (2.95)
and considering as the path integral measure [D ¯ψDψ] where ψ = (ψR, ψL)t. This does not
affect the theory, since the correlation functions do not involve the fictitious and decou-pled right handed field, but allows us to compute the anomaly using the previous results, considering a transformation
U (x) = exp(iα(x)PLt) = exp(iα(x)t/2) exp −iα(x)tγ5/2. (2.96)
Only the chiral component effectively changes the measure and in making the substitution α(x)→ −α(x)/2 we find AL (x) = g 2 32π2 µνρσ GaµνG b ρσTrtTaRTbR . (2.97)
5In other words, the analytic continuation from the Euclidean is straightforward because we do not
encounter any singularity thanks to the analyticity of f .
2.2. Continuous chiral symmetries Chapter 2
The difference in the case of a right-handed fermion would have been replacing PL ↔ PR,
amounting in a sign difference
AL(x) =−AR(x). (2.98)
This is important when considering gauge theories that involve fermions of both chiralities, such as the Standard Model (SM).
In order to have a consistent gauge theory, all anomalies that correspond to the action of the gauge group must vanish. In QCD, the SU(Nc) anomaly is evidently zero because the
gauge transformation is vector-like (non-chiral) on the Dirac fermion. Employing the symmetry ofµνρσFa
µνFρσb under the exchangea ↔ b, we write the anomaly
in eq. (2.93) as A(x) = − g 2 16π2 µνρσ GaµνG b ρσTr tTaRTbR = − g2 32π2 µνρσ GaµνG b ρσTr tTaR, TbR . (2.99)
It is understood that the calculations have been computed in a background of fixed gauge fields, in the sense that the Noether’s current operator satisfies
D ∂µJˆµ E A=− D δ ˆLE A+ D ˆAE A (2.100) where hOiA ≡ Z [D ¯ψDψ]OeiS[A, ¯ψ,ψ] Z [D ¯ψDψ]eiS[A, ¯ψ,ψ] (2.101) Chiral anomaly in QCD
The axial SU(L)A and U(1)A of the chiral group in QCD, which correspond respectively to
the generators t = αa(x)Ta and t =1 acting on flavour space, are potentially anomalous.
The SU(L)A anomaly vanishes because the αa(x)Ta is traceless, while
A(x)U(1)A =− g2 s 32π2 µνρσGa µνG b ρσTrc({Tac, T c b}) Trf(1) (2.102) =−L g 2 s 32π2 µνρσGa µνGaρσ (2.103) ≡ −2LQ(x) (2.104)
where we have used Trc({Tac, Tbc}) = δab and defined the topological charge density
Q(x) = g 2 s 64π2 µνρσGa µνG a ρσ. (2.105)
The topological charge density is proportional to the divergence of the Chern Simons current defined in eq. (2.19), hence the change of the QCD action under an axial U(1)A rotation is
different from zero only if we take into account the instanton solutions.
2.2.2
Spontaneous symmetry breaking and the effective chiral
la-grangian
2.2. Continuous chiral symmetries Chapter 2
1. The symmetry is not anomalous;
2. The transformation acts as the unitary operator ˆU = eiωbQˆb on the Hilbert space and
ˆ
Qb|Ωi = 0, which means that the vacuum |Ωi is invariant ˆU|Ωi = |Ωi.
From the second property we see that the conserved charges are also the generators of the infinitesimal transformation, meaning that for ωa 1
e−iωaQˆaΨeiωaQˆa
' (1 − iωaQˆa)Ψ(1 +iωaQˆa)
' Ψ − iωah ˆQa, Ψi = Ψ +ωaδaΨ. (2.106)
If 1. is satisfied but 2. is not for some ˆQ˙a, the vacuum is not invariant under the group
symmetry and Spontaneous Symmetry Breaking (SSB) occurs. For every broken generators ˆ
Q˙a a spin 0 mass-less particle appears in the spectrum, with the same quantum numbers of
ˆ
Q˙a|Ωi [19].
Isospin and Gell-Mann flavour symmetries, that corresponds to the SU(2)V and SU(3)V
symmetries that emerge in taking respectively the limit mu, md → 0 and mu, md, ms →
0, were known to be approximate symmetries for the strong interaction even before the establishment of QCD.
From now on, we will assume that SU(L)V is Weigner-Weyl realized in the chiral limit.
If SU(L)Ais an exact symmetry, for every hadron state|hi with mass Mh and intrisic parity
ph, we can find a partner with the same mass and baryon number, but opposite parity. In
the algebra of the axial charges h ˆQa A, ˆQ b A i =ifabcQˆj A a, b, c = 1, . . . , L 2 − 1 (2.107)
we can identify a SU(2) sub-algebra h ˆQi A, ˆQ j A i =iijkQˆj A (2.108)
and define the ladder operators ˆQ±A = ˆQ1
A± ˆQ2A. Considering the state
h0± ≡ ˆQ±A|hi, we find that
1. If the symmetry is realized h ˆQ±A, ˆHi= 0, so that h0± and |hi are degenerate
2. h ˆQa A, ˆQV
i
= 0, where ˆQV is the generator of the U(1)V transformation, that
cor-responds to the baryon number. This can be verified observing that the conserved current of the U(1)V symmetry, JVµ = ¯ΨγµΨ, is invariant under an axial SU(L)A
transformation. 3. The ˆQa
A are pseudo-scalars, therefore
h0± has intrisic parity equal to −ph.
These hadronic degenerate partners with opposite parity have never been observed, ruling out the possibility that SU(L)A is an exact symmetry. An analogous parity doubling would
be present if the U(1)A was exact. Furthermore, through lattice simulations, it has been
found that [20]
hΩ| ¯ΨΨ|Ωi 6= 0. (2.109)
The expectation value in eq. (2.109) is known as the chiral condensate. Note that
hΩ|h ˆQa A, ¯Ψγ 5TbΨi |Ωi = hΩ|h ˆQa A, ¯Ψ i γ5TbΨ |Ωi + hΩ| ¯Ψγ5Tbh ˆQa A, Ψ i |Ωi (2.110) 20
2.2. Continuous chiral symmetries Chapter 2 but since h ˆQa A, ¯Ψ i =i(δa AΨ) =¯ − ¯ΨTaγ5 and h ˆQaA, Ψ i =i(δa AΨ) =−Taγ5Ψ, we find hΩ|h ˆQa A, ¯Ψγ5TbΨ i |Ωi = − hΩ| ¯ΨTa, Tb Ψ|Ωi (2.111) =−1 Lδ ab hΩ| ¯ΨΨ|Ωi − dabchΩ| ¯ΨTcΨ |Ωi (2.112)
where in the last equality we have used that Ta, Tb is a L × L hermitian matrix and
admits an expansion in the basis{1, Ta}. The coefficient of the identity matrix is determined
imposing the trace normalization Tr Ta, Tb = δab. The term involving the symmetric
tensor dabc is zero if the SU(L)
V symmetry is Wigner-Weyl realized6
hΩ|h ˆQa A, ¯Ψγ5TbΨ i |Ωi = −1 Lδ abhΩ| ¯ΨΨ|Ωi . (2.113)
Clearly, if SU(L)A was exact, the Left-Hand-Side would have been zero for everya, b.
SU(L)A breaking and effective chiral lagrangian
We conclude that SU(L)A undergoes complete SSB and we fall in the assumptions of the
Nambu-Goldstone Theorem. In the following breaking pattern
G0 = SU(L)V × SU(L)A → H = SU(L)V. (2.114)
L2 − 1 pseudo-scalar Nambu-Goldstone bosons (NGB) emerge, associated with the broken
generators of SU(L)A. Because of the explicitH breaking, due to the non zero quark masses
and electromagnetic interaction, the NGBs acquire masses and we refer to them as pseudo-NGBs.
They fill the same representation under the unbrokenH; in the relevant cases L = 2, 3 we identify them respectively as the iso-triplet of the pions (π−, π0, π+) and the pseudoscalars
octet (π−, π0, π+, η, K+, K−, ¯K0, K0) of Gell-Mann SU(3). In this framework, their lightness
compared to the other hadrons is naturally accounted for.
We can explore the low energy behaviour of the pNGBs with an effective field theory, arising from topological arguments [21,22].
In general, let us consider a symmetry groupG0 and a vacuum manifold that is invariant
under a subgroup H ⊂ G. The πa(x) (a = 1, . . . , N and N = dim(G0)− dim(H)) NGBs will
transform under the symmetry group with some mapping
~π −→ ~πG0 0 = ~f (g, ~π) (2.115)
where ~π = (π1, . . . , πN). Being a representation of the group, if e is the unit element and
g1, g2 ∈ G0, the mapping satisfies
~
f (e, ~π) = ~π f (g~ 2, ~f (g1, ~π)) = ~f (g2g1, ~π). (2.116)
The NGBs fields corresponds to quantum excitations along the vacuum manifold, so if we choose a vacuum configuration ~0, we can write the NGBs fields
~π = ~f (˜g,~0) = ~f (˜gh,~0) (2.117)
6We define C
ij= hΩ| ¯ΨiΨj|Ωi with i and j flavour space indices. An exact symmetry is implemented by
a unitary operator U that leaves the vacuum invariant, so that Cji= hΩ|U†Ψ¯iU U†ΨjU|Ωi = hΩ| ¯Ψ0iΨ0j|Ωi.
Since Ψ0
i = VijΨj, Cji = Vki†VjlhΩ| ¯ΨkΨl|Ωi = (V CV†)ji. Due to Schur’s Lemma Cji ∝ δij and
hΩ| ¯ΨTcΨ
2.2. Continuous chiral symmetries Chapter 2
where in the last equality we’ve used that, by definition, h∈ H leaves the vacuum invariant. The transformation needs to be non-linear to get a non trivial results. If the mapping of eq. (2.117) is invertible, the NGBs fields can be uniquely identified with the elements of the coset space G/H. The coset space is divided into equivalence classes
g0 ∼ g if g0 =gh for some h∈ H (2.118)
and for every field configuration ~π we choose a representative ˜g(~π). Let us focus on the specific case at hand, G0 = SU(L)
L× SU(L)R and H = SU(L)V.
We write ˜g = ( ˜VL, ˜VR) = ( ˜VLV˜R†,1)( ˜VR, ˜VR) = nh, h ∈ SU(L)V. We have chosen the
representative as a SU(L)L transformation.
We define ˜VR=V , U = ˜VLV˜R† and we choose for U the canonical parametrization
U (~π) = exp L2−1 X a=1 iπaτa α ! where τa = 2Ta. (2.119) Under a G0 transformation ˜ g0 =g˜g = (VL, VR)(U V, V ) = (VLU VR†,1)(VRV, VRV ). (2.120)
In eq. (2.120) we read the trasformation forU
U (~π)−→ U(~πG0 0) =VLU (~π)VR†. (2.121)
Low energy effective lagrangian
UsingU and its transformation under the chiral group, we can write a G0 invariant effective
low energy Lagrangian
L(0)
χPT=f0(U ) + f2(∂µU, ∂µU ) +O p4
(2.122)
where f0 and f2 are hermitian operator that involves respectively 0 or 2 derivatives. No
term with one derivative of U is allowed by Lorentz invariance, while f0(U ) can only be a
function of the constant Tr U†U. We define the hermitian operator
∆µ=−iU†∂µU (2.123)
and work out its transformation under the G0 action
∆µ G0
−→ VR∆µVR†. (2.124)
It is possible to rewritef2(∂µU, ∂µU ) as an appropriate G0 invariant combinations of (∆µ)ij×
(∆µ)
kl. The only two possibilities are
Tr(∆µ) Tr(∆µ) Tr(∆µ∆µ). (2.125)
U , being unitary, can be written as exp(M ) with M traceless and anti-hermitian. Using the following identity ∂µeM = Z 1 0 dxexM∂µM e(1−x)M (2.126) 22
2.2. Continuous chiral symmetries Chapter 2 we see that Tr(∆µ)∝ Tr e−M∂µeM = Tr e−M Z 1 0 dxexM∂ µM e(1−x)M (2.127) = Tr(∂µM ) = ∂µTr(M ) = 0 (2.128)
where the ciclycity of the trace has been used in writing the second equality. We conclude that the effective Lagrangian takes the form
L(0)
χPT=C Tr(∆µ∆µ) = C Tr ∂µU†∂µU. (2.129)
There are 2 free parameters that need to be fixed: C and α in the definition of U . They can be determined by imposing a canonical normalization for the kinetic terms and identifying the pion decay constant fπ ' 130.4 MeV in this framework. Expanding U in powers of the
fields U (~π)' 1 + iπ aτa α (2.130) C Tr∂µU†∂µU ' C α2 Tr τ aτb∂ µπa∂µπb = 2C α2∂µπ a∂µπa (2.131)
we obtain the condition
2C α2 =
1
2. (2.132)
Under an infinitesimal axial transformation
U −−−−→ AUA ' (1 + iωSU(L)A a
τa/2)U (1 + iωaτa/2)' U + iωa
/2{τa
, U}. (2.133) The corresponding Noether’s currents are
Aµ a =− δLef f δ(∂µU )ij δaUij − δLef f δ(∂µU†)ij δaUij† =−iC/2 Tr ∂µU†{τa, U} + h.c =−iC/2 Tr ∂µU†U τa+∂µU†τaU + h.c =−iC/2 Tr τa∂µU†, U + h.c. (2.134) Since ∂µ(U†U ) = 0, we find ∂µ U†, U = −∂µ U, U† (2.135) so that using it in eq. (2.134), the leading order of the axial currents reads
Aµ
a =−iC Tr τ
a∂µU†, U ' −iC Tr τa−i(∂µπb)τb,1 = −4C
α ∂
µπa. (2.136)
The matrix elements of Aa
µ between a one-pion particle state and the vacuum is related to
fπ Ω Aa µ πb(p) = 4C α iδ abp µ. (2.137)
Indeed, in the weak decay of the charged pion π− → µ− + ¯ν, we define the pion decay
constant as
Ω Aµ
2.2. Continuous chiral symmetries Chapter 2
where Aµ is the axial hadronic currents involved in the V-A weak interaction. Considering
the case L = 2, in which Ψ = (u, d)T
Aµ= ¯uγµγ5d = ¯Ψγµγ50 1 0 0 Ψ = ¯Ψγµγ5(τ1+iτ2) 2 Ψ≡ A µ 1 +iA µ 2. (2.139)
In terms of the neutral components πa, the negative pion state is written as
π− = |
π1i − i |π2i
√
2 (2.140)
and the matching of the amplitudes leads to 4C
α =
fπ
√
2 ≡ Fπ ' 92 MeV. (2.141)
Finally, from eqs. (2.132) and (2.141), the effective Lagrangian reads
L(0) χP T = F2 π 4 Tr ∂µU †∂µU + O p4 with U (~π) = exp i~π· ~τ Fπ . (2.142)
Chiral pertubation theory The Lagrangian mass term
− Lmass= ¯ΨLMΨR+ ¯ΨRM†ΨL (2.143)
explicitly breaks the chiral symmetry G0. If we replace the constant mass matrix with a
dynamical field (spurion) that under the G0 action transforms as M → M0 =V
LMVR† (2.144)
the symmetry is restored. We want the pertubation to the effective chiral Lagrangian, induced by the massive quarks, to share the spurionic property of the mass term, fixing the form of the explicit breaking to
LχPT=LχPT(0) +δLmass (2.145) with δLmass = F2 π 2 Tr BMU + B ∗M†U†. (2.146)
Since the pions are pseudoscalars, U transforms under parity as
U (x0, ~x) Pˆ
−→ U†(x0,
−~x) (2.147)
so that we take real B to avoid parity violation
δLmass =
F2 π
2 B Tr MU + M
†U†. (2.148)
In the basis with a real and diagonal mass matrix 7
hΩ| ¯ΨfΨf|Ωi = − hΩ|
δLQCD
δmf |Ωi .
(2.149)
7Summation convention is suspend in the next expressions.
2.2. Continuous chiral symmetries Chapter 2
The effective chiral Lagrangian must reproduce the correlation functions of the full theory, so we require hΩ|δLQCD δmf |Ωi ≡ hΩ| δLχPT δmf |Ωi . (2.150)
Taking into account that the vacuum expectation value of U (~π) is equal to the identity (because for the physical pion fields hΩ|πa|Ωi = 0), we are led to the following relation
between the value of B and the chiral condensate
hΩ| ¯ΨfΨf|Ωi = −Fπ2B, (2.151)
accessible through numerical simulations.
2.2.3
The U(1)
Aproblem
The absence of the hadronic parity doubling historically led to believe that the U(1)A
sym-metry underwent SSB, adding a pNGB in the spectrum. Moreover, following a similar approach to the one adopted in section 2.2.2, it can be shown that
hΩ|h ˆQA, ¯Ψγ5Ψ
i
|Ωi = −2 hΩ| ¯ΨΨ|Ωi (2.152)
so the presence of the chiral condensate undermines the quantum invariance under U(1)A
of the QCD vacuum.
This picture immediately ran into troubles. Considering the chiral limit for the up and down quarks (L = 2), no signs of a light meson, with a mass ∼ mπ, were present. The
lightest isospin singlet meson known is the η, but its mass is significantly larger than the pion mass.
The effective Lagrangian approach of the last section teaches us that η has to regarded as the isospin singlet in the pNGB octet of the broken SU(L)A, which is heavier than the
pions due to the presence of the heavier strange quark. So assuming the U(1)A SSB in the
L = 3 case, the obvious candidate to be the SU (3)V singlet is theη0 meson, that once again
is much heavier than the Kaons and η.
This qualitative picture can be made more quantitative considering an effective La-grangian for the Goldstone modes in the presence of the breaking pattern
SU(L)V ⊗ SU(L)A⊗ U(1)A→ SU(L)V (2.153)
constructed following the very same steps of the previous section. As shown by Weinberg [2], this leads to an iso-singlet light meson |li with a mass
Ml<
√
3mπ (2.154)
that is not observed.
The resolution of this puzzle relies on the presence of the quantum anomaly in eq. (2.104)
A(x)U (1)A =
g2 sL
16π2∂µK
µ (2.155)
whose effect is non trivial because of the existence of the instanton field configurations. It is tempting to argue that since
∂µJ µ U(1)A+ g2 sL 16π2∂µK µ A =D∂µK˜µ E A= 0 (2.156)
we can apply the Goldstone theorem to the conserved current ˜Kµ. This is technically true,
but the resulting Goldstone mode will result in un-physical excitation, because the conserved current ˜Kµ is not gauge invariant (because Kµ is not).
Chapter 3
Instantons and the strong CP
Problem
3.1
Classical instantons
Let us consider the Euclidean version of the pure gauge action for a non-Abelian SU(Nc)
gauge theory SE[A] = 1 4 Z d4x EGaEµνGaEµν (3.1) where Ga
E0i = −iGa0i and GEija = Gaij, evaluated at the Euclidean space time point xEµ,
defined after perfoming a π/2 rotation in the complex plane of Minkowski time
xµ
→ ˜xµ= (˜x0, ˜xi) = (x0eiπ/2, xi)
≡ (−ixE4, xEi). (3.2)
We drop the E index for simplicity, but it’s understood that from now on we are working in Euclidean spacetime. A field configuration contributes to the path integral if the action is finite, therefore the field tensor Gµν must vanish in the limit |x| → ∞ faster than 1/|x|2.
This can certainly be accomplished if the fieldsAa
µ → 0 sufficiently fast, but it’s also satisfied
if the connection approaches a pure gauge configuration
igsAaµ(x)T
a |x|→∞
−−−−→ ∂µU (ˆx)U−1(ˆx) (3.3)
where the gauge group elements are direction dependent. Note that the asymptotic config-uration is invariant if we replace U (ˆx)→ U(ˆx)U0, where U0 is a fixed SU (Nc) element. We
can choose U0 such that in any desired direction ˆx0, U ( ˆx0) = 1. This defines a mapping
from points ˆx of the S3 sphere S3 ={x ∈ R4| |x| = 1} to the group manifold
S3 → SU(Nc) (3.4)
with the point ˆx0 mapped into the unit element of the group1. A 3-sphere can be mapped
to the interior of a 3-cube I3 = [0, 1]3, with the boundary identified as a single point,
conveniently chosen to be ˆx0. The mapping in eq. (3.4) is therefore equivalent to
I3 → SU(Nc) (3.5)
1For the logic of the discussion, the base point doesn’t need to be the identity, but it could be any fixed
3.1. Classical instantons Chapter 3
with the boundary of I3 mapped to the identity of SU(Nc). Two different mappings
U1(z1, z2, z3), U1(z1, z2, z3) are said to be homotopically equivalent if it exists a continuous
function f (t, z1, z2, z3), with t∈ [0, 1] such that
f (0, z1, z2, z3) =U1(z1, z2, z3) f (1, z1, z2, z3) = U2(z1, z2, z3) (3.6)
f (t, p) =1 for every p in the boundary. (3.7) This is an equivalence relation, that can be employed to divide the set of all mappings into equivalence classes. For every element of the quotient space, i.e an equivalence class c, we can choose a representative U (z1, z2, z3;c) and define the product c1× c2 as the equivalence
class that contains the mapping
U (z1, z2, z3;c1× c2) =
(
U (2z1, z2, z3;c1) z1 ≤ 1/2
U (2z1− 1, z2, z3;c2) z1 ≥ 1/2
. (3.8)
It can be shown that this product gives to the quotient space, known as third homotopy group Π3( SU(Nc)), a group structure.
The perk of identifying topological non-equivalent configurations is that if we find a local minimum of the action in an certain equivalence class, we know that a finite perturbation cannot change the topology type, hence we find a local minimum for configurations in all equivalence classes.
3.1.1
The Cartan-Maurer integral
It is often convenient to characterize topological non-equivalent configurations in terms of topologically invariant quantities that distinguish from one another. Let us consider a map-ping from ad = 2n+1 dimensional manifold M , with local coordinates θ1, . . . , θd, to elements
U (θ1, . . . , θd) in a representation of a Lie group G. We define the Cartan Maurer integral as
C[U] = i1...id Z ddθ Tr ∂U (θ) ∂θi1 U−1(θ) . . .∂U (θ) ∂θid U−1(θ) . (3.9)
This is independent of the choice of local coordinates2and it’s invariant under a small change of U → U + δU, indeed δC[U] = di1...id Z ddθ Tr ∂U (θ) ∂θi1 U−1(θ) . . . δ ∂U (θ) ∂θid U−1(θ) . (3.10)
having used that we can push every δ∂U (θ)∂θ
ij U
−1(θ) to the right using the cyclic property
of the trace and adjusting the indices as written in eq. (3.10) is an even permutation of the indices (since d− 1 is even).
δ ∂U (θ) ∂θid U−1(θ) = ∂(δU (θ)) ∂θid U−1(θ)−∂U (θ) ∂θid U−1(θ)δU (θ)U−1(θ) (3.11) =U (θ) ∂ ∂θid U−1(θ)δU (θ) U−1(θ) (3.12)
2Considering a change of chart θ0
i(θ) and using the chain rule for the derivatives, we find that the
con-traction with the tensor amount to multiply the trace for the Jacobian of the change of variables.
3.1. Classical instantons Chapter 3
When we partial integrate we only act on the remainingd− 1 U−1(θ) terms, because, thanks
to the anti-symmetry of the tensor, second order partial derivatives vanish. The resulting d− 1 = 2n terms come in pairs of opposite sign and their sum is hence zero. This shows that every element of an equivalence class c, that can be continuously deformed in one another, shares the same value of C[c]. From the definition of multiplication between elements of the homotopy group, it is straightforward to see that
C[c1 × c2] =C[c1] +C[c2]. (3.13)
This tells us that if we find a non zero Cartan-Maurer invariant C[c] for a given equivalence class c, we find inequivalent classes cn for every n integer with C[cn] =nC[c], which means
that Z ⊂ Π3(G).3
In our case of interest Π3(SU (Nc)) = Z, so this procedure completely characterizes the
homotopy group. Let us discuss some general features that allow us to compute the invariant more easily. If we consider two elements of the Lie group U (θ), U (ψ) their product is
U (θ)U (ψ) = U (θ0(θ, ψ)) (3.14)
given that U is a representation. Differentiating with respect of θ0
j, keeping ψ fixed ∂U (θ) ∂θi ∂θi ∂θ0 j U (ψ) = ∂U (θ 0) ∂θ0 j (3.15)
and multiplying on the right of both sides by the inverse of U (θ0), we get
∂θi ∂θ0 j ∂U (θ) ∂θi U−1(θ) = ∂U (θ0) ∂θ0 j U−1(θ0). (3.16)
Employing the previous identity in the definition of the Cartan-Maurer invariant
C = i1...id Z ddθ0Tr ∂U (θ0) ∂θ0 i1 U−1(θ0). . .∂U (θ0) ∂θ0 id U−1(θ0) (3.17) =i1...idTr ∂U (θ) ∂θi1 U−1(θ) . . .∂U (θ) ∂θid U−1(θ) Z ddθ0det ∂θ ∂θ0 (3.18)
where θ and θ0 are regarded as independent variables considering that the element ψ is arbitrary. We define the metric
γij(θ0) =− 1 2Tr ∂U (θ0) ∂θ0 i U−1(θ0)∂U (θ 0) ∂θ0 j U−1(θ0) (3.19)
and observe that using eq. (3.16), the following holds
γij(θ0) = ∂θk ∂θ0 i ∂θl ∂θ0 j γkl(θ). (3.20)
Taking the determinant of both side
det(γ(θ0)) = det ∂θ ∂θ0
2
det(γ(θ)) (3.21)
3Substituting U
3.1. Classical instantons Chapter 3 we find C = i1...idTr ∂U (θ) ∂θi1 U−1(θ) . . .∂U (θ) ∂θid U−1(θ) s 1 det(γ(θ)) Z ddθ0pdet(γ(θ0)). (3.22)
Let us employ what we’ve found to calculate the Cartan-Maurer integral for a mapping S3 → SU(2), with an element in the defining representation of SU(2).
Bott [23] has shown that every mapping S3 → SU(Nc) can be continuously deformed to
a mapping S3 → SU(2) where SU(2) ⊂ SU(Nc) is the subgroup that acts on the first two
components of the fundamental representation of SU(Nc). Since the Cartan-Maurer integral
only depends on the equivalence class, we need to compute it for topologically non-equivalent configurations of the SU(2) subgroup.
An element of SU(2) in the defining representation can be written as
U (θ) = θ0× 1 + 2i~t · ~θ (3.23)
where ~t = ~σ/2 and θ2
0 = 1− |~θ|2. Employing the algebraic properties of the pauli matrices
titj =δij/4 + iijktk/2, it can be shown that the following identities hold
Tr(titj) =δij/2 Tr(titjtk) = i 4ijk (3.24) Tr(titjtktl) = 1 8(δijδkl+δilδjk − δikδjl) (3.25) γij(θ) =− 1 2Tr θi θ0 + 2iti (θ0− 2itkθk) θj θ0 − 2it j (θ0+ 2itkθk) (3.26) =δij + θiθj θ2 0 =δij + θiθj 1− |~θ|2. (3.27)
The determinant of the metric is hence equal to
p
det(γ(θ)) = q 1 1− |~θ|2
. (3.28)
Substituting in eq. (3.22) and taking the limit θ → 0, the invariant reads C = −8iijkTr(t itjtk) Z d3θ0 q 1− |~θ0|2 . (3.29)
The integral is easily computed in polar coordinates, multiplying by a factor 2 to account for the two possible values of θ0
0 for fixed |~θ0|2
C[c] = 24π2. (3.30)
This holds for every mapping homotopic to U (θ) in the equivalence class c and we know that topologically non-equivalent configurations for every integer n C[cn] = 24nπ2 exist.4
4A configuration with a negative Cartan Maurer integral
−24π2is found considering the anti-fundamental
of SU(2).
3.1. Classical instantons Chapter 3
3.1.2
Bogomol’ny inequality
We rewrite the pure gauge action as
S[A] = 1 2 Z d4x Tr(G µνGµν) = 1 4 Z d4x Trh(G µν∓ ˜Gµν)2 i ±12 Z d4x TrG µνG˜µν (3.31)
where ˜Gµν = 1/2µνρσGρσ is the dual tensor. We’ve used that, in space-time with the metric
δµν,
µνρσµναβ = 2(δραδσβ− δρβδσα) (3.32)
to obtain that ˜GµνG˜µν =GµνGµν. From eq. (3.31) we identify a lower bound for the action
S[A]≥ 1 2 Z d4x TrG µνG˜µν (3.33)
which is reached when the curvature is dual G = ˜G or anti-dual G =− ˜G. Observe that the RHS of the above inequality is proportional to the divergence of the Chern Simons current, or more precisely the Euclidean version of it:
1 2Tr GµνG˜µν = ∂µKµ 4 whereKµ = 4µνρσTr Aν∂ρAσ− 2igs 3 AνAρAσ . (3.34)
Exploiting the antisymmetry of , the Chern-Simmons current is re-written as
Kµ= 4µνρσTr Aν∂ρAσ− 2igs 3 A ν AρAσ = 4µνρσTr AνGρσ 2 − igs 3 AνAρAσ . (3.35)
and the integral is evaluated using Gauss-Green theorem 1 2 Z d4x TrGµνG˜µν = lim r→∞ Z S3r dσµµνρσT r(AνGρσ/2− igs 3 AνAρAσ) (3.36) =−igs 3 rlim→∞ Z Sr 3 dσµµνρσT r(AνAρAσ) (3.37)
whereσµis a 4-vector normal to the hypersurface and with modulus equal to the infinitesimal
surface element on the 3-sphere. In the second row we have neglected theAνGρσterm because
the curvature must vanish in the limit r → ∞. This can be achieved, as we have previously discussed, if the connection approaches a pure gauge configuration at infinity
igsAµ(x) |x|→∞
−−−−→ ∂µU (ˆx)U−1(ˆx). (3.38)
We substitute the asymptotic behaviour inside the integral 1 2 Z d4x TrGµνG˜µν = 1 3g2 s lim r→∞ Z Sr 3 dσµµνρσT r((∂νU )U−1(∂ρU )U−1(∂σU )U−1). (3.39)
The infinitesimal surface element on the sphere in cartesian coordinates is d4xδ(|x| − r) and
the normal direction is given by the 4-versor ˆxµ. Definingθ
4 ≡ |x|, the latter is expressed as
ˆ xµ =∂
µθ4. If θi, with i = 1, 2, 3, identifies a point of S3
1 2 Z d4x TrGµνG˜µν = 1 3g2 s lim r→∞ Z S3r d4xδ(θ4− r)µνρσ ∂θ4 ∂xµ ∂θi ∂xν ∂θj ∂xρ ∂θk ∂xσ× Tr ∂U (θ) ∂θi U−1(θ)∂U (θ) ∂θj U−1(θ)∂U (θ) ∂θk U−1(θ) (3.40)
3.2. The Strong CP problem Chapter 3
where we’ve used that U is direction dependent. Noting that
µνρσ ∂θ4 ∂xµ ∂θi ∂xν ∂θj ∂xρ ∂θk ∂xσ =4ijkdet ∂θ ∂x (3.41)
we recognize the Jacobian of the change of variables{xµ} → {θ4, θi}, finding that the integral
is proportional to the Cartan-Maurer invariant of the asymptotic configuration 1 2 Z d4x TrGµνG˜µν = = 1 3g2 s ijk Z S3 d3θ Tr ∂U (θ) ∂θi U−1(θ)∂U (θ) ∂θj U−1(θ)∂U (θ) ∂θk U−1(θ) = 1 3g2 s C[U]. (3.42) Using the results of the previous section C = 24π2ν, where ν is an integer known as the Chern-Pontryagrin index or winding number, we conclude that
Z d4x TrGµνG˜µν = 16π 2ν g2 s (3.43)
which yields the following lower bound for the action
S[A]≥ 8π
2|ν|
g2 s
. (3.44)
Belavin et al. showed in [1] that the configurations
igsAµ =
(r− r0)2
(r− r0)2+ρ2
∂µU1(ˆx)U1−1(ˆx) (3.45)
have dual curvature (G = ˜G), hence are solutions for the classical equation of motion. U1
belongs to the SU(2) subgroup of SU(Nc) that acts on the first two components of the
defining representation U1(ˆx) = x4+i2~t· ~x r (3.46)
for which we computed theν = 1 winding number. The parameters r0 andρ are respectively
known as the instanton center and size.
Solutions with anyν ∈ Z exist, and can be explicitly constructed considering |ν| configu-rations as in eq. (3.45) (or with U1−1 if ν is negative) with centers r0i sufficiently distant that
non linearities of the field equation become negligible.
3.2
The Strong CP problem
3.2.1
Topological term
We can discuss how the instanton solutions enter in the path integral considering a generic weight factor f (ν) for configurations with winding number ν. If we have a local operatorO, its vacuum expectation value is in general
hOi = X ν f (ν) Z [DΦν]O exp − Z d4x L [Φ] X ν f (ν) Z [DΦν] exp − Z d4x L [Φ] (3.47) 31
3.2. The Strong CP problem Chapter 3
where Φ represents all fields of the theory and the measure [DΦν] is restricted to gauge fields
configurations with winding number ν. Suppose that we can decompose R4 = Ω
1 ∪ Ω2 in
two non overlapping large volumes and that the operator O has support contained in Ω1.
Approximately, we can divide the path integral in configurations with winding numberν1 in
Ω1 and ν2 in Ω2, such thatν = ν1+ν2.5
hOi = X ν1,ν2 f (ν1+ν2) Z [DΦν1]O exp − Z Ω1 d4x L [Φ] Z [DΦν2] exp − Z Ω2 d4x L [Φ] X ν1,ν2 f (ν1+ν2) Z [DΦν1] exp − Z Ω1 d4x L [Φ] Z [DΦν2] exp − Z Ω2 d4x L [Φ] . (3.48) The cluster decomposition principle (Chap. 4 of [24], Chap. 23.6 of [25]) states that the expectation value should be the same if we omit the volume Ω2, but this is true only if the
weight factor satisfies
f (ν1+ν2) =f (ν1)f (ν2). (3.49)
This fixes the form of f (ν) to
f (ν) = exp(iνθ). (3.50)
If the Poyntragin index is expressed in terms of gauge fields, the inclusion of the instanton in the path integral effectively amount to adding to the Euclidean Lagrangian density
LE → LE − iθ g2 s 32π2G a µνG˜ a µν. (3.51)
Going back to Minkowski space-time, substituting d4x
E → id4x and GE34 → −iG30, we get
LQCD → LQCD +θ g2 s 32π2G a µνG˜ aµν. (3.52)
Starting with the QCD Lagrangian with no topological term, the solution of the U(1)A
prob-lem furnished by existence of instantons requires us to include the θ term in the Lagrangian density.
ForNf flavours QCD, the partition function then reads
ZQCD = Z [DA][D ¯ΨDΨ] exp i Z d4xL0 QCD[A, ¯Ψ, Ψ] +LM+θQ(x) (3.53)
where LQCD0 is the QCD Lagrangian density with massless quarks and no topological term, while
− LM = ¯ΨLMΨR+ ¯ΨRM†ΨL. (3.54)
Let us perform the change of variables in flavour space
ΨL→ ULΨL ΨR→ URΨR (3.55)
with UL, UR∈ U(Nf). Defining the determinants
detUL,R=eiϕL,R (3.56)
5This is approximately true if the volumes are large compared to the instanton size, since the integral of
the Poyntragin density g2 s/32π
2Ga µνG˜
a