The Higgs as a Composite Nambu-Goldstone Boson

R. Contino
arXiv:1005.4269v1 [hep-ph] 24 May 2010

Dipartimento di Fisica, Università La Sapienza,

and INFN, Roma, Italy
roberto.contino@roma1.infn.it

Abstract
This is an introduction to theories where the Higgs is a composite Nambu-
Goldstone boson of a new strongly-interacting dynamics not much above the
weak scale. A general discussion is presented based on the pattern of global
symmetries at low energy, and the analogy with the QCD pion is analyzed. The
last part of the lectures shows how a composite Higgs can emerge as the hologram
of a 5-dimensional gauge field.

Lectures presented at the 2009 TASI Summer School,

‘Physics of the Large and the Small’
University of Colorado, Boulder, June 1-29, 2009
Contents

1 Introduction 3

2 Two paradigms for Electroweak Symmetry Breaking 4

2.1 The Higgs model . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8
2.2 Technicolor models . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 10

3 The Higgs as a composite Nambu-Goldstone boson 18

3.1 An SO(5)/SO(4) example . . . . . . . . . . . . . . . . . . . . . . . . . 20
3.2 Comparing with QCD: the pion potential . . . . . . . . . . . . . . . . . 26
3.3 Electroweak precision tests and flavor constraints in composite Higgs
models . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 31
3.4 Higgs potential from the top quark . . . . . . . . . . . . . . . . . . . . 41

4 The holographic Higgs 46

4.1 A5 as a pseudo Nambu-Goldstone boson . . . . . . . . . . . . . . . . . 50
4.2 Effective Potential of the Holographic Higgs . . . . . . . . . . . . . . . 56

5 Epilogue 62

Appendix 63

2
1 Introduction
More than a century of experimental results and theoretical progress has led us to the
formulation of an extremely elegant and compact theory of the fundamental interac-
tions among particles. Its success in reproducing a huge amount of experimental data,
spanning several orders of magnitude in energy, is impressive. Despite their profoundly
different manifestations on macroscopic scales, the electromagnetic, weak and strong
forces are all described within the same mathematical framework of gauge theories.
The electromagnetic and weak interactions are associated to the same SU (2)L × U (1)Y
gauge invariance at short distances, although only electromagnetism is experienced as
a long-range force. The rest of the electroweak symmetry is hidden at large distances
or low energies, i.e. it is spontaneously broken by the vacuum. As a matter of fact,
despite the abundance of experimental information, we do not know much about the
dynamics responsible for such spontaneous breaking.
The Standard Model (SM) of particle physics gives an extremely economical for-
mulation of the electroweak symmetry breaking (EWSB) in terms of only one new
fundamental degree of freedom: the Higgs boson. It does not explain, however, the dy-
namical origin of the symmetry breaking, nor why the Higgs boson should be light, as
required to comply with the LEP data. An old and still attractive idea is that the Higgs
boson might be a bound state of a new strongly-interacting dynamics not much above
the weak scale. Its being composite would solve the SM hierarchy problem, as quan-
tum corrections to its mass are now saturated at the compositeness scale. Significant
theoretical progress on the construction of these theories has recently come from the
intriguing connection between gravity in higher-dimensional curved spacetimes and
strongly-coupled gauge theories. Fully realistic models have been built where some
longstanding problems of the original constructions are solved.
These lectures are aimed to give an introduction to composite Higgs theories. We
start in Section 2 by recalling why new dynamics is expected to emerge at the TeV scale
in the electroweak sector. We then present the two paradigms for such new dynamics:
the weakly-interacting Higgs model, and strongly-coupled Technicolor theories. The
idea of the Higgs as a composite pseudo Nambu-Goldstone (pNG) boson is introduced
in Section 3, as an interpolation between these two scenarios. We illustrate one explicit
example of symmetry breaking pattern, SO(5) → SO(4), and make use of symmetry
arguments to derive the expression of the Higgs potential in terms of form factors. The
same approach is then followed to describe the electromagnetic potential of the pion in
QCD, and the analogy with the composite Higgs is analyzed in detail. The constraints
from electroweak precision tests and from flavor-changing neutral currents (FCNC)
are then discussed at length, and the concept of partial compositeness introduced.
Section 4 shows how a composite pNG boson can emerge as the fifth component of
a gauge field living in a 5-dimensional spacetime. The basic features of this kind of
theories are illustrated by means of a simple abelian model in a flat extra dimension.
We discuss how the bulk of the fifth dimension gives a model of the 4-dimensional strong

3
dynamics which is perturbative and thus calculable. As an important application we
compute the form factors that parametrize the couplings of the composite Higgs and
obtain an analytic expression for its potential. We conclude with a few words on the
phenomenology of composite Higgs models.
In selecting the above topics I had necessarily to omit some other important ones,
as for example warped extra dimensional models and holography in curved spacetimes,
and Little Higgs theories. Fortunately excellent reviews exist on these subjects, such
as for example the Les Houches lectures by T. Gherghetta on holography [1] and the
review by M. Schmaltz and D. Tucker-Smith on Little Higgs models [2]. The lectures
by R. Sundrum [3] at TASI 2004 and the review [4] by M. Serone partly overlap
with Section 4 and contain interesting complementary topics and discussions. General
introductions to flat and warped extra dimensions are given for example in the parallel
TASI lectures by H. C. Cheng [5] and T. Gherghetta, the TASI lectures by C. Csaki [6],
G. Kribs [7], and the Cargese lectures by R. Rattazzi [8]. Extra dimensional models
as theories of electroweak symmetry breaking are for example discussed in the TASI
lectures by C. Csaki, J Hubisz and P. Meade [9]. More detailed references are given
throughout the text. They are meant to introduce the reader to the vast literature on
the subject and form a necessarily incomplete and partial list. I apologize in advance
for the omissions.

2 Two paradigms for Electroweak Symmetry Breaking

The vast amount of data collected so far in high-energy experiments can be explained
and compactly summarized by the Lagrangian

L = L0 + Lmass
3
1 a a µν 1 1 X
L0 = − Wµν W − Bµν B µν − Gµν Gµν + Ψ̄(j) i D
6 Ψ(j)
4 4 4 j=1

2 1 (1)
Lmass = MW Wµ+ W − µ + MZ2 Z µ Zµ
2
X  (i) 
ūL Miju uR + d¯L Mijd dR + ēL Mije eR + ν̄L Mijν νR
(j) (i) (j) (i) (j) (i) (j)
−
i,j

+ h.c. ,

where Ψ = {qLi , uiR , diR , lLi , eiR , νRi } is a collective index for the Standard Model fermions
and i, j are generation indices. A remarkable property of L is that while all the fun-
damental interactions among the particles (determined by L0 ) are symmetric under
local SU (2)L × U (1)Y transformations, the observed mass spectrum (determined by
Lmass ) is not. In other words, the electroweak symmetry is hidden, i.e. spontaneously
broken by the vacuum. Although successful at the energies explored so far, the above

4
 Im(a)

Re(a)

Figure 1: Unitary circle: at energies below (above) the inelastic threshold the amplitude
al is constrained to lie on (inside) the circle.

mathematical formulation leads to an inconsistency if extrapolated to arbitrarily high

energies: when used in a perturbative expansion, it predicts scattering amplitudes that
grow with the energy and violate the unitarity bound. The latter prescribes that the
elastic scattering amplitude al of each l-th partial wave must satisfy

Im(al ) = |al |2 + |ain 2

l | , (2)

where ainl is the inelastic scattering amplitude. This means that at energies below the
inelastic threshold al is constrained to lie on the unitary circle Re2 (al ) + (Im(al ) −
1/2)2 = 1/4, while at higher energies it is bounded to be inside it, see Fig. 1. Since at
tree level the amplitude is real and an imaginary part only arises at the 1-loop level,
perturbativity is lost when the imaginary and real part are of the same order, that is
when the scattering phase is large, δ ≈ π.
It turns out that the violation of perturbative unitarity occurs in processes that
involve longitudinally polarized vector bosons as external states. For example, at tree
level the amplitude for the elastic scattering of two longitudinally polarized W ’s grows
as E 2 at energies E  mW :

g2
A(WL+ WL− → WL+ WL− ) ' (s + t) . (3)
4m2W
Here s, t are the Mandelstam kinematic variables, and terms subleading in mW /E
have been dropped. Each longitudinal polarization brings one factor of E, since at

5
large energies µL (p) = pµ /mW + O(mW /E), so that each diagram naively grows as E 4 .
When all the diagrams are summed, however, the leading E 4 term cancels out, and
the amplitude grows as E 2 . We will see shortly that this cancellation can be easily
understood by performing the calculation in a renormalizable gauge. By projecting on
partial wave amplitudes,
Z +1
1
al = d cos θ A(s, θ)Pl (cos θ), (4)
32π −1

where Pl (x) are the Legendre polynomials (P0 (x) = 1, P1 (x) = x, P2 (x) = 3x2 /2 − 1/2,
etc.), one finds the following expression for the s-wave amplitude (l = 0):
1 s
a0 (WL+ WL− → WL+ WL− ) ' . (5)
32π v 2
1
The loss of perturbative unitarity in the s-wave scattering thus occurs for
√
π ≈ δ ' 2Re(a0 ) , i.e. for: s ≈ Λ = 4πv ' 3 TeV . (6)

The role of the longitudinally polarized vector bosons suggests that the inconsis-
tency of the Lagrangian (1) is in the sector that breaks spontaneously the electroweak
symmetry and gives mass to the vector bosons. The connection can be made explicit
by introducing, as propagating degrees of freedom, the Nambu-Goldstone bosons χa
that correspond to the longitudinal polarizations of the W and Z bosons:

σa a σ3
Σ(x) = exp(iσ a χa (x)/v), Dµ Σ = ∂µ Σ − ig Wµ Σ + ig 0 Σ Bµ . (7)
2 2
In terms of the chiral field Σ, the mass terms can be rewritten as follows: 2
u (j)
!
2 λ u
v h i v X 
(i) (i)

Lmass = Tr (Dµ Σ)† (Dµ Σ) − √ ūL dL Σ ij R
(j)
+ h.c. (8)
4 2 i,j λdij dR

The local SU (2)L × U (1)Y invariance is now manifest, since Σ transforms as

Σ → UL (x) Σ UY† (x) ,

(9)
UL (x) = exp iαLa (x)σ a /2 UY (x) = exp iαY (x)σ 3 /2 ,

1
√ √
A slightly stronger bound, s . 2 2πv = 2.2 TeV, is obtained by including the effect of the
channel W + W − → ZZ, see Ref. [10]. Notice that sometimes the bound Re(al ) ≤ 1/2 or |al | ≤ 1 is
imposed, instead of δ ≤ π. All are in fact acceptable as an estimates of the energy where perturbative
unitarity is lost. The difference in the values of the cutoff Λ thus obtained can be interpreted as the
theoretical uncertainty of the estimate.
2
For simplicity, from here on I will omit the lepton terms and concentrate on the quark sector.

6
although it is non-linearly realized on the χa fields:
v v
χa (x) → χa (x) + αLa (x) − δ a3 αY (x) . (10)
2 2
In the unitary gauge, hΣi = 1, the chiral Lagrangian (8) reproduces the mass term
of eq.(1) with
M2
ρ ≡ 2 W2 = 1. (11)
MZ cos θW
This relation is consistent with the experimentally measured value to quite good ac-
curacy. It follows as the consequence of a larger approximate invariance of (8) under
SU (2)L × SU (2)R global transformations,
Σ → UL Σ UR† , (12)
which is spontaneously broken to the diagonal subgroup SU (2)c by hΣi = 1, and
explicitly broken by g 0 6= 0 and λuij 6= λdij . In the limit of vanishing g 0 the fields χa
transform as a triplet under the “custodial” SU (2)c , so that MW = MZ . This equality
is replaced by Eq.(11) at tree level for arbitrary values of g 0 . Further corrections
proportional to g 0 and (λu − λd ) arise at the one-loop level and are small. In fact, the
success of the tree-level prediction ρ = 1 a posteriori justifies the omission in the chiral
Lagrangian (8) of the additional term
2
v 2 Tr Σ† Dµ Σ σ 3

(13)
that is invariant under the local SU (2)L × U (1)Y but explicitly breaks the global
SU (2)L × SU (2)R . In other words, the coefficient of such additional operator is exper-
imentally constrained to be very small.
The chiral Lagrangian (8) makes the origin of the violation of perturbative unitarity
most transparent. Let us work in a renormalizable ξ-gauge, with a gauge-fixing term
2
g0v 3

1  3 gv 3 2 1
LGF = − ∂µ W µ + ξ χ − ∂µ Bµ + ξ χ
2ξ 2 2ξ 2
0 2
(14)
1 g v
− ∂µ Wµ+ + ξ χ+ .
2ξ 2
The Equivalence Theorem [11,10] states that at large energies the amplitude for the
emission or absorption of a Goldstone field χ becomes equal to the amplitude for the
emission or absorption of a longitudinally-polarized vector boson:

µ
WL χ

!
m2W

= × 1+O .
E2

7
In particular, the amplitude for the scattering of two longitudinal W ’s becomes equal,
at energies E  mW , to the amplitude for the scattering of two Goldstone bosons.
For the latter process there is only one diagram which contributes at leading order in
E/mW :

1
A(χ+ χ− → χ+ χ− ) = (s + t) . (15)
v2

The growth of the amplitude with E 2 thus originates from the derivative interaction
among four Goldstones contained in the kinetic term of Σ in Eq.(8). Ultimately, the
violation of perturbative unitarity can be traced back to the non-renormalizability of
the Lagrangian (8). The merit of the chiral formulation is that of isolating the problem
to the sector of the Lagrangian which leads to the mass terms for the vector bosons
and the fermions.
There are thus two possibilities: i) either new particles associated to new dynamics
come in to restore unitarity before perturbativity is lost, or ii) the χχ scattering grows
strong until the interaction among four χ’s becomes non-perturbative. This latter pos-
sibility must also be seen as the emergence of new physics, as the description of the
theory changes, at the strong scale, in terms of new, more fundamental, degrees of
freedom. These two paradigms for the electroweak symmetry breaking are well exem-
plified by the two theories that we will discuss in the next sections: the Higgs model,
and Technicolor theories. Whatever mechanism Nature has chosen, it is generally true
that
There has to be some new symmetry-breaking dynamics acting as an ultraviolet
completion of the electroweak chiral Lagrangian (8).
As required by the experimental evidence, such new dynamics must be (approximately)
custodially symmetric, so as to prevent large corrections to the ρ parameter. The most
important question then is the following: is this dynamics weak or strong ?

2.1 The Higgs model

The most economical example of new custodially-invariant dynamics is that of just one
new scalar field h(x), singlet under SU (2)L × SU (2)R . Assuming that h is coupled to
the SM gauge fields and fermions only via weak gauging and (proto)-Yukawa couplings,
the most general EWSB Lagrangian has three free parameters a, b, c 3 at the quadratic
3
In general c can be a matrix in flavor space. We will assume it is proportional to unity, so that
no flavor-changing neutral current effects originate from the tree-level exchange of h.

8
order in h [12]:

v2 h i h2

1 2 † h
LH = (∂µ h) + V (h) + Tr (Dµ Σ) (Dµ Σ) 1 + 2a + b 2 + . . .
2 4 v v
(j)
! (16)
λuij uR
 
v X  (i) (i)  h
−√ ūL dL Σ 1 + c + · · · (j)
+ h.c.
2 i,j v λdij dR

Here V (h) denotes the potential, including a mass term, for h. Each of these parame-
ters controls the unitarization of a different sector of the theory. For a = 1 the exchange
of the scalar unitarizes the χχ → χχ scattering 4

s2
 
+ − 1
+ − 2
A(χ χ → χ χ ) = 2 s − a + (s ↔ t)
v s − m2h
 2
s+t 2

mh
= 2 1−a +O .
v E2

Since we have introduced a new particle in the theory, we have to check that also the
inelastic channels involving h are unitarized. The χχ → hh scattering (equivalent to
WL WL → hh at high energy), is perturbatively unitarized for b = a2 :

m2h
 
s
A(χ χ → hh) = 2 b − a2 + O
+ −


.
v E2

Finally, the χχ → ψ ψ̄ scattering (equivalent to WL WL → ψ ψ̄ at high energy) is unita-

rized for ac = 1

4
In the diagrams showed in present section, dashed and solid lines denote respectively the fields χ
and h, whereas solid lines with an arrow denote fermions.

9
 √  2
+ − mψ s mh
A(χ χ → ψ ψ̄) = 2
(1 − ac) + O .
v E2

Only for a = b = c = 1 the EWSB sector is weakly interacting (provided the scalar h
is light), as for example a 6= 1 implies √
a strong W√ W → W W scattering with violation
of perturbative unitarity at energies s ≈ 4πv/ 1 − a2 , and similarly for the other
channels.
The point a = b = c = 1 in fact defines what I will call the “Higgs model”: LH (with
vanishing higher-order terms in h) can be rewritten in terms of the SU (2)L doublet
 
1 iσa χa (x)/v 0
H(x) = √ e (17)
2 v + h(x)

and gets the usual form of the Standard Model Higgs Lagrangian. In other words, χa
and h together form a linear representation of SU (2)L × SU (2)R . The unitarity of the
model can be thus traced back to its renormalizability. In terms of the Higgs doublet
H, the custodial invariance of the Lagrangian appears like an accidental symmetry:
at the renormalizable level, all the (SU (2)L × U (1)Y )-invariant operators are functions
of H † H = i ωi2 , where ωi are the four real components parametrizing the complex
P
doublet H. This implies that the theory is invariant under an SO(4) ∼ SU (2)L ×
SU (2)R invariance, broken down to SO(3) ∼ SU (2)c in the vacuum hH † Hi = v 2 ,
under which the ωi components are rotated.
The weakly-interacting Higgs model has two main virtues: it is theoretically at-
tractive because of its calculability, and it is insofar phenomenologically successful, as
it satisfies the LEP and SLD electroweak precision tests [13]. Both calculability (which
stems from perturbativity) and the success in passing the precision tests follow from
the Higgs boson being light. It is however well known that an elementary light scalar,
such as h, is unstable under radiative corrections: its mass receives quadratically diver-
gent corrections, which makes a light Higgs scalar highly unnatural in absence of some
symmetry protection. In this sense, the Higgs model should perhaps be regarded as
a parametrization rather than a dynamical explanation of the electroweak symmetry
breaking.

2.2 Technicolor models

The Higgs model is an extremely economical way to perturbatively unitarize the theory
and parametrize the symmetry breaking, but we know that it is not the solution that
Nature has chosen in another similar physical system: QCD. At low energy the con-
densation of the color force dynamically breaks the SU (2)L × SU (2)R chiral symmetry

10
to its vectorial subgroup SU (2)V , the three pions π a being the associated Nambu-
Goldstone bosons. Their dynamics is described by a non-linear sigma model analogous
to the chiral Lagrangian (8) for the χ fields

fπ2 h i
Lπ = Tr (∂µ Σ)† (∂ µ Σ) , Σ(x) = exp(iσ a π a (x)/fπ ) , (18)
4
where fπ = 92 MeV is the pion decay constant. Consequently, the pion-pion scattering
is affected by the same unitarity problems encountered in the W W scattering. In this
case however, we know from experiment that there is no light scalar resonance playing
the role of the Higgs boson h. Rather, a whole tower of heavier resonances is exchanged
in pion-pion scattering at high energies, which eventually enforces unitarity. Experi-
mentally, the most important contribution comes from the lightest vector resonance,
the ρ meson (J = 1, I = 1)

The new symmetry breaking dynamics is thus strongly interacting in this case, and
its dual description in terms of more fundamental degrees of freedom is the quark
Lagrangian. QCD can be considered as the prototype for strong symmetry breaking,
and the study of its properties can shed light on the UV completion of the electroweak
Lagrangian (8).
It is interesting, for example, to discuss what happens to QCD and to the pions
when one turns on the weak interactions. 5 In the limit of vanishing quark masses
(chiral limit), and before turning on the weak interactions, the pions are exact NG
bosons associated to the global symmetry breaking SU (2)L × SU (2)R × U (1)B →
SU (2)V × U (1)B , and are thus massless. The SU (2)L × U (1)Y interactions gauge only
part of the full global symmetry, and in this way they introduce an explicit breaking
of the symmetry, see Fig. 2. In other words, the QCD vacuum breaks the electroweak
invariance and the pions are eaten to give mass to the W and the Z. The surviving
unbroken group is exactly the electromagnetic U (1)em , with the electric charge given
by Q = TL3 + TR3 + B/2. To see how the weak bosons get mass, let us consider for
example the W propagator in the Landau gauge ξ = 0. As the result of the coupling of
the W to the conserved weak current J µ ± = q̄L γ µ T ± qL (T ± = T1 ±iT 2 ), its propagator
gets corrected from the QCD dynamics:
5
In the following we consider for simplicity QCD with two quark flavors, the generalization to the
six-flavor case is trivial.

11
 SU (2)L × U (1)Y

QCD

SU (2)L × SU (2)R × U (1)B

π
SU (2)V × U (1)B

Figure 2: Cartoon of QCD with part of its chiral symmetry gauged by the weak interactions.

= +

+ + ...

−i qµ qν
Gµν (q) = (PT )µν , (PT )µν ≡ ηµν − , (19)
q2 − g 2 Π(q 2 )/2 q2
where
Z
iΠµν (q) = − d4 x e−iq·x h0|T Jµ+ (x)Jν− (0) |0i

  (20)
qµ qν
Πµν (q) = ηµν − 2 Π(q 2 ) .
q
Then, a mass for the W arises if Πµν (q 2 ) has a pole at q 2 = 0. The pole in fact exists
as a result of the symmetry breaking, due to the exchange of the pion:
fπ
h0|Jµ+ |π − (p)i = i √ pµ (21)
2

fπ2
=⇒ Π(q 2 ) = .
2
This implies that the W acquires a mass
gfπ
mW = ' 29 MeV .
2
Although this number is far from the experimental value, the above discussion shows
that QCD is, at the qualitative level, a good example of electroweak symmetry breaking
sector. This is even more true considering that the unbroken SU (2)V isospin invariance
acts as a custodial symmetry so that ρ = 1 at tree level in the QCD vacuum.

12
 SU (2)L × U (1)Y

QCD TC
(SU (3)c) (SU (NT C ))

SU (2)L × SU (2)R SU (2)L × SU (2)R

π πT C
SU (2)V SU (2)V

Figure 3: Cartoon of a new Technicolor sector and QCD with part of their global symmetries
gauged by the weak interactions.

This suggests that the actual EWSB dynamics could be just a scaled-up version of
QCD, with
fπ −→ Fπ ' v = 246 GeV . (22)
In general, one can think of an SU (NT C ) “Technicolor” gauge group with a global
SU (2)L × SU (2)R invariance broken down to SU (2)V at low energy due to confine-
ment [14], see Fig. 3. A linear combination of the QCD pions and the new set of
‘technipions’ is thus eaten to form the longitudinal polarizations of the W and the Z
|VL i = sin α |πQCD i + cos α |πT C i
fπ (23)
v 2 = fπ2 + Fπ2 tan α = ,
Fπ
while the orthogonal combination can be identified with the physical pion. Since fπ 
Fπ ' v, the longitudinal polarizations of the W and the Z will mostly come from the
technipions, although it is true that they will have a small component of the QCD
pions. In fact, this is true in general for any EWSB sector, Technicolor being just a
specific example of symmetry breaking dynamics where the role of the χ fields in the
chiral Lagrangian (8) is played by the technipions.
In order to derive the properties of the Technicolor sector, the large-N formalism
developed in Refs. [15,16] is extremely useful. Here we just summarize the results that
we will use and refer to Ref. [16] for the proof. Let us consider an SU (N ) gauge theory,
with a large number N of ‘colors’. We know that QCD at N = 3 is a confining theory
and we will assume that this behavior persists for N  1. Under this assumption, the
large-N theory has the following properties:
1. At leading order in N the two-point function of a quark local bilinear operator
J(x), like the scalar q̄q or the current q̄γ µ q, is given by an infinite exchange of
one-meson states:
X f2
n
hJ(q)J(−q)i = 2 − m2
,
n
q n

13
 where mn is the mass of the n-th meson and fn = h0|J|ni the amplitude for J to
create it from the vacuum. This in turn implies that:

2. For large N the mesons are free, stable and non-interacting. Their number is
infinite and their masses have a smooth large-N limit. The mass of the lowest
lying modes is of the order
mρ ∼ gρ fπ , (24)
where gρ denotes the coupling among mesons.

3. Since the two-point function hJJi is of order N/16π 2 , it follows that fn scales
like √
N
fn ∼ . (25)
4π
From the behavior of the n-point Green functions of J it follows that a local
vertex with n mesons scales like ∼ gρn−2 , where

4π
gρ ∼ √ . (26)
N

Using the above results, the vectorial and axial conserved currents of the Technicolor
sector can be written, for large NT C , in terms of an infinite sum over vectorial (ρn ) and
axial (an ) resonances:

X fρ2n
hT {JVµ (q)JVν (−q)}i 2 µν
= q η −q q µ ν
(27)
n
q 2 − m2ρn
" #
X fa2n Fπ2
hT {JAµ (q)JAν (−q)}i = q 2 η µν − q µ q ν


+ . (28)
n
q 2 − m2an q2

Here we have defined the amplitude for JV to create a vectorial resonance with mo-
mentum q and polarization r to be h0|JVµ |ρ(q, r )i = µr mρ fρ , and similarly for the
axial current. Notice that the latter have the quantum numbers to create, in addition
to spin-1 axial mesons, also the technipion. As a consequence, the two-point function
hJA JA i has a pole at q 2 = 0. At √ large NT C the vector, axial and technipion decay
constants scale like fρ , fa , Fπ ∼ NT C /4π, while all the masses are constants. In par-
ticular, the mass of the lowest-lying vectorial resonance of the Technicolor sector, the
‘technirho’, is expected to be of order
r
3 Fπ
mρT C ∼ mρ , (29)
NT C fπ
where mρ = 770 MeV is the mass of the QCD rho meson, fπ = 92 MeV is the QCD
pion decay constant and mρT C , Fπ are the corresponding Technicolor quantities. For
example, for NT C = 4 one has mρT C ∼ 1.8 TeV.

14
 The most attractive feature of Technicolor theories, and in general of theories with
strong electroweak symmetry breaking, is that the hierarchy problem of the Higgs
model is solved by dimensional transmutation: the electroweak scale v is generated
dynamically as the scale at which the Technicolor coupling gT C grows strong in the
infrared (β0 < 0):

8π 2
 
d 1 β0
µ (µ) = − 2 =⇒ v = MP l exp − 2 . (30)
dµ gT2 C 8π gT C (MP l )(−β0 )

This is in complete analogy with the dynamical generation of the QCD scale from the
Planck scale MP l . On the other hand, the simplest Technicolor constructions, like the
naive scaled-up version of QCD, lead to predictions in conflict with the experimental
data. The two most serious problems are a parametrically too large correction to the
Peskin-Takeuchi S parameter, and too fast flavor-changing neutral-current processes.
Let us review both in turn.
The Peskin-Takeuchi S parameter is defined as [17]

∂
S ≡ −16π Π3B (q 2 ) q 2 =0
(31)
∂q 2
µ
where the vacuum polarization of a W3L and an hypercharge boson B µ , Π3B (q 2 ), is
defined according to eq.(20). The leading contribution to S from new heavy states can
be parametrized in terms of the dimension-6 operator (see Refs. [18, 19])
S
Tr T aL Wµν
aL
Σ B µν T 3R Σ†
 
(32)
16π
where T aL = σ a /2 = T aR are the generators of SU (2)L ×SU (2)R . Since the Technicolor
sector is strongly coupled, a perturbative calculation of the S parameter is not possible.
However, one can estimate its size using Naive Dimensional Analysis (NDA) [20]: it will
arise at the 1-loop level, thus carrying a factor NT C /16π 2 , and it will be proportional
to the number of technidoublets ND :
3L
Wµν Bµν

S NT C ND NT C ND
∼ 2
=⇒ S∼ . (33)
16π 16π π
A more sophisticated calculation that makes use of QCD data rescaled up to the EW
scale gives a similar result [17]. From the estimate above one deduces that even minimal
models (with NT C and ND small) tend to predict S ∼ 1. Such values are ruled out by
the LEP data, which bound (assuming an optimal contribution to the T parameter) [13]

S . 0.3 @ 99% CL . (34)

15
 √
Conversely, S < 0.3 requires mρT C & 3 TeV ND , which is difficult to accommodate in
Technicolor models given that the mass of the first vectorial resonance is tied to the
EW scale v.
The second important difficulty with simplest Technicolor models is the way in
which quark masses are generated. So far we have not discussed how the quark sector
feels the electroweak symmetry breaking. For this to occur, some interaction must
exist between quarks and techniquarks. A simple solution is to assume that both the
color group SU (3)c and the Technicolor SU (NT C ) are embedded in a larger Extended
Technicolor (ETC) group,

SU (NET C ) ⊃ SU (3)c × SU (NT C ) ,

which is assumed to be spontaneously broken at the scale ΛET C [21]. The exchange
of the broken ETC gauge bosons connects quarks with techniquarks and generates, at
the scale ΛET C , four-fermion operators with two SM quarks and two technifermions:
q ψT C

q̄ ψ̄T C

2
gET C


Lint = 2
(q̄q) ψ̄T C ψT C . (35)
ΛET C
At the lower scale ΛT C ≈ Fπ ' v the SU (NT C ) group condenses giving rise to the
quark masses 2
2

gET C ΛT C
mq = 2 hψ̄T C ψT C i ∼ ΛT C . (36)
ΛET C ΛET C
In order to explain the large hierarchies in the quark masses it is thus clear that the
generation of the four-fermion interactions (35) for different flavors cannot happen just
at one single scale ΛET C . Rather, one has to assume that different SM quark families
are embedded into a single ETC multiplet and that SU (NET C ) undergoes a cascade
of breakings, thus generating several different scales. The ETC breaking scale relevant
for any given quark flavor cannot be too large, otherwise the corresponding quark
mass that follows from eq.(36) becomes too small. For example, if ΛT C ' v one needs
ΛET C ≈ 10 TeV in order to reproduce the s quark mass.
The same exchange of ETC gauge fields that leads to the four-fermion operator
(35), however, also generates operators with four SM fermions, (q̄q)2 /Λ2ET C . Quite
generically, these operators violate flavor and CP, since different SM flavors have to be
embedded into the same ETC multiplet, and thus give rise to various FCNC processes.
The bounds from K K̄ and B B̄ mixing and rare meson decays, for example, put very

16
 β

gT∗ C gT C
E→0

Figure 4: Flow to an IR non-perturbative fixed point.

stringent limits on the scale at which such operators can be generated: ΛET C & 105 TeV
(103 −104 TeV) from CP-violating (-conserving) processes [22]. Thus, there is a tension
between generating large enough quark masses and avoiding too fast FCNC processes.
One mechanism that has been proposed to resolve this tension is that of Walking
Technicolor [23]. It is based on the following general observation: if a term ∆L =
λ O(x) is generated in the Lagrangian at the scale Λ with dimension [O(x)] = d, its
contribution to physical amplitudes at the low-energy scale E goes like ∼ λ (E/Λ)d−4 .
The energy factor is due to the classical running of the coupling λ for d 6= 4, so that the
higher the dimension of the operator O, the more suppressed its contribution at low
energy. In writing the formula (36) for the quark masses we have implicitly assumed
that the dimension of the four-fermion operator (35) is equal to its classical value
[(q̄q)(ψ̄T C ψT C )] = 6, although quantum corrections due to the Technicolor interactions
can change it. In general, the anomalous dimension γ is small if the SU (NT C ) theory
is asymptotically free above the scale ΛT C , so that the coupling gT C (µ) quickly runs
to small values for µ > ΛT C :

[(q̄q)(ψ̄T C ψT C )] = 6 + γ γ(µ) ∼ O(αT C ) . (37)

However, it is possible that starting from high energies, ΛET C & E  ΛT C , and flowing
down to lower scales, the SU (NT C ) dynamics reaches a non-perturbative infrared fixed
point, see Fig. 4. In that case the theory behaves like a conformal field theory down
to energies E ∼ ΛT C , at which it condenses and the conformal behavior is lost. In the
conformal regime, any operator O is characterized by its scaling dimension d∗ at the
fixed point:
1
[O] = d∗ , hO(x)O(0)i ∝ 2d∗ . (38)
x
Notice that d∗ can significantly differ from its perturbative (classical) value, since
the Technicolor coupling at the fixed point is large and its evolution above ΛT C is
slow: it ‘walks’ towards the fixed-point value gT∗ C . Once applied to the operator
O = (q̄q)(ψ̄T C ψT C ), the above argument shows that the formula for the quark masses

17
generalizes to  2+γ
ΛT C
m q ∼ ΛT C . (39)
ΛET C
Hence, if the anomalous dimension γ is sizable and negative, the suppression in the
quark masses can be reduced or, equivalently, ΛET C can be larger. This ameliorates the
FCNC problem, since no large anomalous dimension is generated by the SM color and
weak interactions, so that the suppressing factor in front of the flavor-violating four-
quark operators is still 1/Λ2ET C . However, naive arguments suggest that the smallest
consistent value for the anomalous dimension is γ = −1, so that the suppression in the
quark masses can be ameliorated but not completely avoided. For example, a simple
way to deduce the lower bound γ > −1 is the following [24]: Neglecting the contribution
coming from SM interactions, γ entirely arises from the anomalous dimension of the
quark bilinear H = (ψ̄T C ψT C ), which plays the role of the Higgs field in acquiring a
vacuum expectation value and giving mass to the SM quarks. The unitarity bound
on primary scalar operators of a conformal field theory requires γ to be larger than
−2, a value at which the dimension of H becomes equal to that of the corresponding
free field. In the limit of large NT C or γ → −2 the dimension of the SM scalar singlet
H † H is well approximated by twice the dimension of H: [H † H] ' 2[H] = 6 + 2γ.
Then, for γ < −1 the operator H † H becomes relevant and it will reintroduce the
problem of UV instability that plagues the Higgs model. In particular, for γ = −2
the operator H † H has dimension 2 and its radiative correction will be quadratically
divergent. The possibility is still open for a Walking Technicolor theory at small NT C
where [H] is not too much above 1 (in order not to suppress the quark masses), while
[H † H] stays close to 4 (so that no hierarchy problem is present) [24], although strong
constraints have been derived on this scenario based on general properties of conformal
field theories [25, 26].

3 The Higgs as a composite Nambu-Goldstone boson

There is an interesting variation of the strong symmetry breaking paradigm that in-
terpolates between simple Technicolor theories and the Higgs model: a light Higgs
boson could emerge as the bound state of a strongly interacting sector, rather than
being an elementary field. A composite Higgs would solve the hierarchy problem of
the Standard Model, as its mass is not sensitive to virtual effects above the compos-
iteness scale, in the same way as the mass of the QCD pion does not receive Planckian
corrections. Having a light Higgs in the spectrum, on the other hand, would allow the
theory to satisfy the LEP electroweak precision tests more easily than in the case of
simple Technicolor constructions.
As first pointed out by Georgi and Kaplan in the eighties in a series of seminal
papers, the composite Higgs boson can be naturally lighter than the other resonances
if it emerges as the pseudo Nambu-Goldstone boson of an enlarged global symmetry

18
 H0
EWSB
G
Sector

H
G H1 H0 H1
H

Figure 5: Cartoon of a strongly interacting

EWSB sector with global symmetry G broken
Figure 6: The pattern of symmetry
down to H1 at low energy. The subgroup H0 ⊂ G
breaking.
is gauged by external vector bosons.

of the strong dynamics [27–32], see also [33]. Consider for example the general case
in which the strongly interacting sector has a global symmetry G dynamically broken
to H1 at the scale f (the analog of the pion decay constant fπ ), and the subgroup
H0 ⊂ G is gauged by external vector bosons, see Fig. 5. The global symmetry breaking
G → H1 implies n = dim(G) − dim(H1 ) Goldstone bosons, n0 = dim(H0 ) − dim(H) of
which are eaten to give mass to as many vector bosons, so that H = H1 ∩ H0 is the
unbroken gauge group, see Fig. 6. The remaining n − n0 are pseudo Nambu-Goldstone
bosons. In this picture the SM fields, both gauge bosons and fermions, are assumed to
be external to the strong sector, and in this sense we will refer to them as ‘elementary’,
as opposed to the composite nature of the resonances of the strong dynamics. The
SM gauge fields, in particular, are among the vector bosons associated to gauge group
H0 . For simplicity, in the following we will identify H0 with the SM electroweak group,
H0 = GSM ≡ SU (2)L × U (1)Y , so that the SM vectors are the only elementary gauge
fields coupled to the strong sector.
In order to have a composite pNG Higgs boson one has to require two conditions:

1. The SM electroweak group GSM must be embeddable in the unbroken subgroup

H1 :
G → H1 ⊃ GSM

2. G/H1 contains at least one SU (2)L doublet, to be identified with the Higgs dou-
blet.

If the above two conditions are realized, at tree level GSM is unbroken
1 and the Higgs
doublet is one of the pNG bosons living on the coset G/H1 . Its potential vanishes at
tree level as a consequence of the non-linear Goldstone symmetry acting on it. On
the other hand, the global symmetry G is explicitly broken by the couplings of the SM
fields to the strong sector, as they will be invariant under GSM but not in general under

19
G. Loops of SM fermions and gauge bosons thus generate a Higgs potential, which in
turn can break the electroweak symmetry. In this context the electroweak scale v is
dynamically determined and can be smaller than the sigma-model scale f , differently
from Technicolor theories where no separation of scale exists. The ratio ξ = (v/f )2 is
determined by the orientation of GSM with respect to H in the true vacuum (degree
of misalignment), and sets the size of the parametric suppression in all corrections to
the precision observables. By naive dimensional analysis, indeed, the mass scale of the
resonances of the strong sector is mρ ∼ gρ f , with 1 . gρ . 4π. The Higgs instead gets
a much lighter mass at one-loop, mh ∼ gSM v where gSM . 1 is a generic SM coupling.
The limit f → ∞ (ξ → 0) with fixed v is thus a decoupling limit where the Higgs stays
light and all the other resonances become infinitely heavy.
Let us explain in detail all the above points by considering an explicit example.

3.1 An SO(5)/SO(4) example

Let us consider the case in which the strongly interacting sector has a global symmetry
G = SO(5) × U (1)X broken down to H1 = SO(4) × U (1)X [34, 35]. 6 In section 4
we will provide an explicit example of dynamics that leads to this pattern of global
symmetries. As shown in the Appendix, SO(4) is isomorphic to (that is: it has the
same Lie algebra of) SU (2)L × SU (2)R . The SM electroweak group SU (2)L × U (1)Y
can be thus embedded into SO(4) × U (1)X ∼ SU (2)L × SU (2)R × U (1)X , so that
hypercharge is realized as Y = T 3R + X. The coset SO(5)/SO(4) implies four real NG
bosons transforming as a fundamental of SO(4), or equivalently as a complex doublet
H of SU (2)L . The doublet H is the composite Higgs. Under an SU (2)R rotation it
mixes with its conjugate H c = iσ 2 H ∗ , so that (H, H c ) transforms as a bidoublet (2, 2)
representation of SU (2)L × SU (2)R .
Let us derive the effective action that describes the composite Higgs and the SM
elementary fields. As our final goal is to compute the Higgs potential generated at one-
loop by the virtual exchange of SM fields, we will integrate out the strong dynamics
encoding its effects into form factors and keep terms up to quadratic order in the SM
fields. The four NG bosons living on the coset SO(5)/SO(4) can be parametrized in
terms of the linear field Σ,

Σ0 = (0, 0, 0, 0, 1)
Σ(x) = Σ0 eΠ(x)/f √ (40)
Π(x) = −iT â hâ (x) 2 ,

where T â are the SO(5)/SO(4) generators. Using the basis of SO(5) generators given
in the Appendix, one can easily compute the explicit expression of Σ in terms of its
four real components hâ :
sin(h/f ) 1 2 3 4
p
Σ= h , h , h , h , h cot(h/f ) , h≡ (hâ )2 . (41)
h
6
For an analysis of the less minimal coset SO(6)/SO(5) see Ref.[36].

20
The most general effective action for the SM gauge fields in the background of Σ can
be derived just based on symmetries by using a trick: let us assume that the full
SO(5) × U (1)X global symmetry of the strong sector is gauged, so that the external
gauge fields form a complete adjoint representation of SO(5) × U (1)X . Then, at the
quadratic level and in momentum space, the most general (SO(5) × U (1)X )-invariant
action has the form:
1 2 2 2
L = (PT )µν ΠX t
 
0 (q ) Xµ Xν + Π0 (q ) Tr(Aµ Aν ) + Π1 (q ) ΣAµ Aν Σ . (42)
2
Here Xµ and Aµ = Aaµ T a + Aâµ T â are the U (1)X and SO(5) gauge bosons 7 , and
PT is the transverse projector defined by eq.(19). Since we want to derive only the
Higgs potential and not its derivative interactions, the field Σ has been treated as a
classical background, with vanishing momentum. The form factors ΠX 0 , Π0,1 encode
the dynamics of the strong sector, including the effect of the fluctuations around the
background Σ (i.e. the NG fields). A few useful properties of the form factors can be
derived as follows.
By expanding around the SO(4)-preserving vacuum Σ = Σ0 , the effective action
(42) can be rewritten as
1 2 2 2
L = (PT )µν ΠX a a â â
 
0 (q ) Xµ Xν + Πa (q ) Aµ Aν + Πâ (q ) Aµ Aν , (43)
2
where
Π1
Πa = Π0 , Πâ = Π0 + (44)
2
are the form factors associated respectively to the unbroken and broken generators.
In the limit of large number of ‘colors’ N of the strong sector, they can be written in
terms of an infinite sum of narrow resonances using the large-N results of section 2.2:

µν 2 2 µν

X fρ2n
(PT ) Πa (q ) = hJaµ Jaν i = q η −q qµ ν
(45)
n
q 2 − m2ρn
" #
X fa2n 1 f2
(PT )µν Πâ (q 2 ) = hJâµ Jâν i = q 2 η µν − q µ q ν


+ . (46)
n
q 2 − m2an q 2 2

We have used the fact that the current Jâµ has the correct quantum numbers to excite
the NG bosons hâ from the vacuum. Thus, we deduce that at zero momentum Π0 must
vanish (and similarly ΠX
0 ), while Π1 does not:

Π0 (0) = 0 = ΠX
0 (0) , Π1 (0) = f 2 . (47)
7
Here and in the following T a and T â denote respectively the unbroken (SO(4)) and broken
(SO(5)/SO(4)) generators. Among the SO(4) generators, those of SU (2)L (SU (2)R ) will be denoted
as T aL (T aR ).

21
 At this point we turn back to the original action (42) and switch off the unphysical
gauge fields keeping only those of SU (2)L × U (1)Y . By using eq.(41) we obtain:
sin2 (h/f )
 
1 µν X 2 2 2
L = (PT ) Π0 (q ) + Π0 (q ) + Π1 (q ) Bµ Bν
2 4
sin2 (h/f )
 
2 2
+ Π0 (q ) + Π1 (q ) AaµL AaνL (48)
4

2 2 † aL aL
+ 2 sin (h/f ) Π1 (q ) Ĥ T Y Ĥ Aµ Bν ,

where Bµ is the hypercharge field and we defined

1 h1 − ih2
 
1
Ĥ ≡ H = . (49)
h h h3 − ih4
This is the effective action for the SM gauge fields in the background of Σ that we were
looking for. By expanding the form factors at momenta small compared to the mass
scale of the strong resonances, q 2  m2ρ , one obtains an effective Lagrangian in terms of
local operators. Without loss of generality, one can always perform an SO(4) rotation
and align the Higgs vev along the h3 direction, so that (h1 , h2 , h3 , h4 ) = (0, 0, 1, 0) and
Ĥ t = (0, 1). Hence, at order q 2 one has
  2 2 
µν 1 f sin (hhi/f )
Bµ Bν + Wµ3 Wν3 − 2Wµ3 Bν


L = (PT )
2 4
 2 2 
f sin (hhi/f )
+ Wµ+ Wν− (50)
4
q2 h 0 i 
aL aL 0 X0


+ Π (0) Wµ Wν + Π0 (0) + Π0 (0) Bµ Bν + . . .
2 0
where Π0 denotes the first derivative of Π with respect to q 2 . From the above Lagrangian
we can thus identify
1 1
= −Π00 (0) , = − Π00 (0) + ΠX 0


0 (0) (51)
g2 g 02

and
hhi v2 hhi
v = f sin , so that ξ ≡ 2 = sin2 . (52)
f f f
Notice that the formulas in eq.(51) show the contribution to the low-energy gauge
couplings from the strong dynamics only. If one adds to the effective action (42) bare
kinetic terms for the external SU (2)L × U (1)X fields, the expressions for g and g 0 will
be modified to
1 1 1 1
= −Π00 (0) + 2 , = − Π00 (0) + ΠX 0


2 02 0 (0) + 02 . (53)
g g0 g g0

22
 Starting from eq.(48) it is simple to derive the couplings of the physical Higgs boson
to the gauge fields. By expanding around the vev hhi,
 
0
 0 
hâ = 
hhi + h ,
 (54)
0

one has

2 hhi hhi hhi h

  
2 2 h 2
f sin =f sin + 2 sin cos
f f f f f
  2
2 hhi
 
h (55)
+ 1 − 2 sin + ...
f f
p
=v 2 + 2v 1 − ξ h + (1 − 2ξ) h2 + . . .
√
where, with a slight abuse of notation, h stands for hâ hâ on the left hand side, while
it denotes the physical Higgs boson on the right hand side. Compared to their SM
prediction, the couplings of the composite Higgs to the gauge bosons V = W, Z are
thus modified as follows:
p
gV V h = gVSMVh 1−ξ, gV V hh = gVSMV hh (1 − 2ξ) . (56)

If one compares with the effective Lagrangian for a generic scalar eq.(16), one finds
that the SO(5)/SO(4) theory predicts
p
a= 1−ξ, b = 1 − 2ξ . (57)

Using the results of section 2.1 on the W W scattering, we deduce that both the W W →
2
W W and W W → hh scattering amplitudes grow as √ ∼ (E/v) ξ at large√energies,
violating perturbative unitarity at a scale Λ ≈ 4πv/ ξ. This is a factor ξ larger
than what we found for a theory with no Higgs.
We see that the composite Higgs only partly unitarizes the scattering amplitudes,
simply postponing the loss of perturbative unitarity to larger scales. In the limit ξ → 0
(with v fixed) one recovers the standard Higgs model: the resonances of the strong
sector become infinitely heavy and decouple, while the Higgs boson fully unitarizes the
theory. For ξ → 1, on the other hand, the Higgs contribution vanishes and unitarity
in W W → W W scattering is enforced solely by the strong resonances. Furthermore,
f = v and there is no gap of scales in theory: in this limit the strong dynamics behaves
quite similarly to a minimal Technicolor theory, although a light scalar exists in the
spectrum. In the general case, for ξ small enough the strong resonances can be made
sufficiently heavy and their correction to the electroweak observables sufficiently small
to pass the LEP precision tests. We will illustrate this point in detail later on, in

23
 + + + ···

Figure 7: 1-loop contribution of the SM gauge fields to the Higgs potential. A grey blob
represents the strong dynamics encoded by the form factor Π1 .

section 3.3, as we are now ready to derive the Coleman-Weinberg potential for the
composite Higgs.
We will concentrate on the contribution from the SU (2)L gauge fields, neglecting the
smaller correction from hypercharge. The contribution from fermions will be derived
in section 3.4. The 1-loop Coleman-Weinberg potential resums the class of diagrams
in Fig. 7. From the effective action (48), after the addition of the gauge-fixing term
1
2
LGF = − 2
∂ µ AaµL , (58)
2g ζ
it is easy to derive the Feynman rules for the gauge propagator and vertex:

i ig 2
Gµν = (P T )µν − ζ (PL )µν
Π0 (q 2 ) q2

iΠ1 (q 2 ) 2
iΓµν = sin (h/f )(PT )µν
4

where (PL )µν = qµ qν /q 2 is the longitudinal projector. Resumming the series of 1-loop
diagrams of Fig. 7 then leads to the potential:
Z 4
1 Π1 (Q2 )
 
9 dQ 2
V (h) = log 1 + sin (h/f ) (59)
2 (2π)4 4 Π0 (Q2 )

where Q2 = −q 2 is the Euclidean momentum. The factor 9 originates from the sum
over three Lorentz polarizations and three SU (2)L flavors.
Let us argue on the behavior of the form factors at large Euclidean momentum and
on the convergence of the integral. We have seen that Π0 is related to the product of
two SO(4) currents
hJµa (q)Jνa (−q)i = Π0 (q 2 )(PT )µν (60)
where, we recall, the notation hO1 O2 i denotes the vacuum expectation of the time-
ordered product of the operators O1 and O2 . The form factor Π1 , on the other hand,

24
is given by the difference (see eq.(44))
1
hJµa (q)Jνa (−q)i − hJµâ (q)Jνâ (−q)i = − Π1 (q 2 )(PT )µν . (61)
2
At energies much above the scale of symmetry breaking, the SO(5) invariance is re-
stored, and the difference of two-point functions along broken and unbroken directions
is expected to vanish. In this sense Π1 is an order parameter : it is sensitive to the
symmetry-breaking IR dynamics, and it vanishes at large momenta. If Π1 goes to zero
fast enough, the integral in eq.(59) will be convergent and the Higgs potential finite.
This agrees with the intuition that if the Higgs is a bound state of the strong dynam-
ics, then its mass cannot receive corrections larger than the compositeness scale. To
support this intuition with a more rigorous argument, let us consider the Operator
Product Expansion (OPE) of two currents.
Following Wilson, the time-ordered product of two operators A(x1 ), B(x2 ) can be
expressed as an infinite sum of local operators of increasing dimension multiplied by
coefficients that depend on the separation (x1 − x2 ):
X (n)
T {A(x)B(0)} = C12 (x)On (0) . (62)
n

The equality is at the level of operators, thus implying the equality of any Green
function made of them. The sum extends over all operators with the same global
symmetries of the product AB. In particular, the OPE of the product of two conserved
currents Jµ reads, in momentum space:
Z X
i d4 x eiq·x T {Jµ (x)Jν (0)} = (q 2 ηµν − qµ qν ) C (n) (q 2 )On (0) . (63)
n

By dimensional analysis, the larger is the dimension of the operator On , the more
suppressed is its coefficient at large Euclidean momenta Q2 = −q 2 :
1
C (n) (Q2 ) ∼ for Q large. (64)
Q[On ]
The convergence of the integral in the Higgs potential then requires that the first
operator to contribute to the difference of the product of SO(4) and SO(5)/SO(4)
currents must have dimension 5 or greater:

hJµa (q)Jνa (−q)i − hJµâ (q)Jνâ (−q)i = (q 2 ηµν − qµ qν ) C (5) (q 2 )hO5 i + . . .

 
(65)

so that
1
Π1 (Q2 ) ∼ n≥5 for Q2 → ∞ . (66)
Qn−2
This makes use of the fact that Π0 grows at least as Q2 at large momenta, see eq.(51).

25
 Clearly, without knowing the details of the strong dynamics we cannot say more
about the behavior of the form factors, nor can we prove that the condition (66)
is satisfied in general. There is however a similar physical situation where we have
enough experimental and theoretical information to reconstruct the OPE and deduce
the convergence of the integral: this is the case of the electromagnetic correction to
the pion mass.

3.2 Comparing with QCD: the pion potential

Let us consider QCD in the chiral limit, so that the pion is an exact NG boson at
tree level, and turn on the electromagnetic interaction. Differently from the cartoon of
Fig. 2, where the full SU (2)L × U (1)Y symmetry was gauged, in this case the external
U (1)em group can be embedded into the unbroken subgroup SU (2)V . This means
that the pion is not eaten to form a massive photon, but remains in the spectrum as a
pseudo NG boson. At the radiative level, diagrams with loops of the elementary photon
will generate a potential and a mass term for the charged pion, while the neutral pion
remains massless. This is in complete analogy to the composite Higgs theory considered
in the previous section, although in the case of QCD we dispose of much more detailed
information on the strong dynamics. Following the same steps as we did for the case
of the composite Higgs, we can derive the effective action for the pion and compute its
potential. 8
In order to write down the effective action that describes the photon and the pion
we use the same trick of section 3.1 and assume that the whole SU (2)L × SU (2)R chiral
invariance of QCD is gauged by external fields. Treating the pion field Σ (see eq.(18)) as
a constant classical background, the most general (SU (2)L × SU (2)R )-invariant action,
in momentum space and at the quadratic order in the gauge fields, is
1 h
L = (PT )µν ΠL (q 2 ) Tr {Lµ Lν } + ΠR (q 2 ) Tr {Rµ Rν }
2 i (67)
2
 †
− ΠLR (q ) Tr Σ Lµ ΣRν .

Here Lµ , Rµ are the external gauge fields associated respectively to SU (2)L and SU (2)R
transformations. Since in the vacuum hΣi = 1 the chiral SU (2)L × SU (2)R symmetry
is broken down to SU (2)V , it is useful to rewrite the left and right gauge fields in term
of vectorial and axial ones:
1 1
Vµ = √ (Rµ + Lµ ) , Aµ = √ (Rµ − Lµ ) . (68)
2 2
8
Much of the material of this section is reviewed, for example, in Ref.[37]. See also the original
papers [38] and [39].

26
In the hΣi = 1 vacuum the effective action thus reads
1 h
L = (PT )µν ΠV V (q 2 ) Tr {Vµ Vν } + ΠAA (q 2 ) Tr {Aµ Aν }
2 i (69)
+ ΠV A (q 2 ) Tr {Vµ Aν + Aµ Vν } .

where we have defined

1
ΠV V = (ΠL + ΠR − ΠLR )
2
1
ΠAA = (ΠL + ΠR + ΠLR ) (70)
2
1
ΠV A = (ΠR − ΠL ) .
2
We know that only the axial current has the right quantum numbers to excite a pion
from the vacuum, so we expect that only the hAµ Aν i correlator has a pole at q 2 = 0
from the pion exchange. This implies that at zero momentum the form factors ΠV V
and ΠV A vanish, whereas ΠAA (0) = fπ2 . Equivalently,

ΠLR (0) = 2 ΠL (0) = 2 ΠR (0) = fπ2 . (71)

At this point we turn back to the effective action (67) and switch off all the external
gauge fields but the photon: we set Lµ (x) = T 3 vµ (x) = Rµ (x). Using

Σ = exp (iσ a π a /fπ ) = 1 cos (π/fπ ) + i π̂ a σ a sin (π/fπ )

p πa (72)
π ≡ (π a )2 , π̂ a = ,
π
we obtain
 
1 µ ν 1 2 2

2
 † 3 3
L = (PT )µν v v ΠL (q ) + ΠR (q ) − ΠLR (q ) Tr Σ T ΣT . (73)
2 2

After a bit of algebra one finds (π + π − ≡ (π1 )2 + (π2 )2 )

1 sin2 (π/fπ ) + −
Tr Σ† T 3 ΣT 3 = −

(π π ) , (74)
2 π2
hence
2
 
1 µ ν 2 2 sin (π/fπ ) + −
L = (PT )µν v v ΠV V (q ) + ΠLR (q ) (π π ) . (75)
2 π2
As expected the neutral pion does not couple to the photon at the quartic level, al-
though there can be interactions involving both the neutral and the charged pion. The

27
1-loop diagrams associated to the Coleman-Weinberg potential are these same as those
in Fig. 7. Their resummation gives
Z ∞
1 ΠLR (Q2 ) sin2 (π/fπ ) + −
 
3 2 2
V (π) = dQ Q log 1 + (π π ) . (76)
16π 2 0 2 ΠV V (Q2 ) π2

The convergence of the integral thus depends on the behavior of the form factors
ΠLR (Q2 ) and ΠV V (Q2 ) at large Euclidean momenta Q2 . To infer such behavior we can
use the information that comes from the OPE of the product of two vector and axial
currents, see eq.(63). The color-singlet, scalar 9 operators of dimension 6 or less are:

1 (identity operator) (d=0)

Om = ψ̄mq ψ (d=4)
OG = Gaµν Ga µν (d=4)
Oσ = ψ̄σ µν ta mq ψGaµν (d=6)



OΓ = ψ̄Γ1 ψ ψ̄Γ2 ψ (d=6)
abc
Of = f Gaν µ Gbρ ν Gcµρ (d=6)
where a, b, c are color indices and Γ1,2 are matrices in flavor, color and Lorentz space.
Notice that the operators Om and Oσ break explicitly the chiral symmetry and must be
thus proportional to the quark mass matrix mq . As such they vanish in the chiral limit.
On the other hand OΓ is the only chiral-invariant operator among those listed above
whose vacuum expectation value can violate the chiral symmetry and thus distinguish
between the axial and vector currents. In other words, OΓ is the operator with lowest
dimension to contribute to the form factor ΠLR :
  
2 2 2 2 δ 1
ΠLR (Q ) = Q COΓ (Q )hOΓ i + · · · = Q 6
+O , (77)
Q Q8

where δ is a numerical coefficient. 10 Since the form factor ΠV V grows as Q2 at

large Euclidean momenta (the leading term in its expansion corresponds to the kinetic
term of the photon), we deduce that the integral in the pion potential is convergent. A
reasonable approximation to the full potential is obtained by setting ΠV V (Q2 ) ' Q2 /e2
and expanding the logarithm at first order:
Z ∞
3 sin2 (π/fπ ) + −
V (π) ' 2 αem (π π ) dQ2 ΠLR (Q2 ) . (78)
8π π2 0
9
Operators of spin 1/2 and higher do not contribute to the vacuum expectation value hJµ Jν i and
are thus irrelevant to the following argument.
10
The coefficient δ can be computed perturbatively expanding in powers of αs and 1/Nc .
In the large Nc limit,

the matrix element hOΓ i factorizes into (hψ̄ψi)2 , and one finds: δ =
8π αs /π + O(αs ) (hψ̄ψi)2 [38, 39].
2 2

28
 The information of the OPE on the asymptotic behavior of ΠLR allows us to proceed
further and compute the integral explicitly provided we make two approximations: the
large Nc limit and vector meson dominance. At leading order in 1/Nc the product of
two vector or axial currents can be written in terms of an infinite sum of resonances
poles, so that
X fρ2n
ΠV V (Q2 ) =Q2
n
Q2 + m2ρn
" # (79)
X fa2n fπ2
ΠAA (Q2 ) =Q2 + .
n
Q2 + m2an Q2
Given that
ΠLR (q 2 ) = ΠAA (q 2 ) − ΠV V (q 2 ) , (80)
the large-Q behavior that follows from eq.(77),

  Qlim
 ΠLR (Q2 ) = 0
1 1 2 →∞
ΠLR (Q2 ) ∝ +O =⇒ (81)
Q4 Q6  lim Q2 ΠLR (Q2 ) = 0 ,

2Q →∞

implies two sum rules on the spectrum of masses and decay constants of the strong
resonances:
Xh i
2 2
fρn − fan = fπ2 (82)
n
Xh i
fρ2n m2ρn − fa2n m2an = 0 . (83)
n

These relations where first derived by Weinberg [40], and are known respectively as his
first and second sum rules.
The vector meson dominance approximation then consists in assuming that the
dominant contribution to these relations, as well as to other observables, comes from
the first vector and axial resonances (the ρ and the a1 ). By neglecting the higher
resonances and saturating the two Weinberg sum rules with the ρ and the a1 we then
obtain
m2a1
fρ2 = fπ2 (84)
m2a1 − m2ρ
m2ρ
fa21 = fπ2 , (85)
m2a1 − m2ρ
and the ΠLR form factor can be written as

2
m2a1 m2ρ
ΠLR (Q ) ' fπ2 . (86)
(Q2 + m2a1 )(Q2 + m2ρ )

29
Using the above expression of ΠLR , the integral appearing in the pion potential gives
m2ρ m2a1
Z ∞  2 
2 2 2 ma1
dQ ΠLR (Q ) = fπ 2 2
log . (87)
0 ma1 − mρ m2ρ
For any value of the masses, the above expression is always positive (reflecting the
positivity of ΠLR in eq.(86)). This means that the pion potential is minimized for
hπ 1 i = hπ 2 i = 0 . (88)
In other words, the radiative corrections align the vacuum along the U (1)-preserving
direction, and the photon remains massless. It turns out that the positivity of the
integral (87) and the above conclusion on the alignment of the vacuum are much more
general that our approximate result. Witten [41] has shown that in a generic vector-like
confining gauge theory one has
ΠLR (Q2 ) ≥ 0 for 0 ≤ Q2 ≤ ∞ , (89)
so that the radiative contribution from gauge fields always tends to align the vacuum
in the direction that preserves the gauge symmetry.
The effect of the one-loop potential (78) is that of lifting the degeneracy of vacua
and give a (positive) mass to the charged pion, while leaving the neutral one massless.
Notice indeed that the potential vanishes in the vacuum (88), so that there is still
a flat direction along π 0 . All the results derived above are valid in the chiral limit,
that is for vanishing quark masses. When the quark masses is turned on, both the
charged and neutral pion get a mass, as a consequence of the explicit breaking of the
chiral symmetry. The difference of the charged and neutral pion mass, however, is
still dominantly accounted for by the electromagnetic correction that we have derived.
Thus, we can compare our prediction with the experimentally measured value and
check the accuracy of our approximations. From eqs.(78) and (87) one gets
3 αem m2ρ m2a1
 2 
2 2 ma1
mπ± − mπ0 ' 2 2
log . (90)
4π ma1 − mρ m2ρ
This result was first derived in 1967 by Das et al. using current algebra techniques [42].
Inserting the experimental values mρ = 770 MeV and ma1 = 1260 MeV into eq.(90) one
obtains the theoretical prediction
(mπ± − mπ0 )|TH ' 5.8 MeV , (91)
to be compared with the experimentally measured value
(mπ± − mπ0 )|EXP ' 4.6 MeV . (92)
Considering that corrections to the large-Nc approximation are expected to be of or-
der ∼ 30%, we conclude that the agreement of our theoretical prediction with the
experimental value is fully satisfactory.

30
 As an exercise useful for the following, we also derive the prediction for the chiral
coefficient L10 under the same assumptions that led to eq.(90). L10 is defined in terms of
the difference of the derivative of the axial and vector form factors at zero momentum:
fρ2 fa21
− 4L10 ≡ Π0AA (0) − Π0V V (0) = Π0LR (0) ' 2− 2 . (93)
mρ ma1

Under the assumption of vector meson dominance, the first Weinberg sum rule, eq.(82),
requires fa1 < fρ , and we know experimentally that ma1 > mρ . This implies that the
sign of L10 is fixed to be negative. Using both the Weinberg sum rules one obtains

fπ2 m2ρ
 
− 4L10 ' 2 1 + 2 . (94)
mρ ma1

3.3 Electroweak precision tests and flavor constraints in composite Higgs

models
Having discussed the QCD example in detail, we now turn back to the case of the
composite Higgs. We will assume that the form factor Π1 (Q2 ) goes to zero fast enough
for Q2 → ∞, so that the integral in eq.(59) is convergent. As for the pion potential, we
can expand the logarithm at first order and approximate Π0 (Q2 ) ' Q2 /g 2 to obtain
Z ∞
9 g2 2
V (h) = 2
sin (h/f ) dQ2 Π1 (Q2 ) . (95)
8 16π 0

According to Witten’s argument on vector-like gauge theories [41], we expect that the
above integral is positive and that the 1-loop gauge contribution to the potential aligns
the vacuum in an SU (2)L -preserving direction: v/f = sinhhi/f = 0. This is indeed
verified in explicit models, see for example Refs.[34, 35].
In the context of the original composite Higgs models, various solutions have been
proposed to solve this problem. For example, in their first paper Georgi and Kaplan
make use of an additional elementary scalar that mixes with the composite Higgs [27],
while in the model of Ref. [28] the vacuum is misaligned by the explicit breaking of the
global symmetry mediated at a higher scale by the exchange of (extended ultracolor)
heavy vectors. A more attractive mechanism, which does not rely on the existence of
elementary scalars or any hard breaking of the global symmetry, has been proposed by
Banks [29] and subsequently implemented in Refs.[30–32]. The idea is that of enlarging
the external gauge group to include an additional axial U (1)A and designing the pattern
of global symmetry breaking G → H1 such that while the electroweak SU (2)L × U (1)Y
can be embedded in the unbroken subgroup H1 , the full SU (2)L × U (1)Y × U (1)A
cannot. The impossibility of preserving the full gauge group implies that the 1-loop
contribution from the U (1)A vector boson to the potential necessarily destabilizes the
(SU (2)L × U (1)Y )-symmetric vacuum, leading to (v/f )2 = ξ 6= 0. In such models, the

31
degree of vacuum misalignment ξ depends on the ratio of the U (1)A and SU (2)L gauge
couplings, gA /g.
More recently, it has been shown that the SM top quark contribution can also
misalign the vacuum and break the electroweak symmetry in a natural way [34]. This
will be discussed in detail in section 3.4, where we compute the contribution of the SM
fermions to the Higgs potential. Here we want to discuss how composite Higgs theories
face the electroweak precision tests of LEP and the constraints from FCNC processes.
We will thus assume that some other contribution to the Higgs potential exists, for
example (though not necessarily) coming from the SM top quark, which triggers the
EWSB and gives ξ 6= 0. As before we will analyze the SO(5)/SO(4) model, although
the results that we will derive are generic.
Let us consider first the correction to the Peskin-Takeuchi S parameter. According
to its definition (31), S is given by the following term in the expansion of the effective
action (48) at small momenta:
1 ξ
L ⊃ Π01 (0) sin2 (h/f ) Wµν B Ĥ † T aL Y Ĥ → − Π01 (0) Wµν
aL µν 3
B µν . (96)
2 8
In analogy with the previous section, in the limit in which the number of colors N of
the strong sector is large we can use the results of section 2.2 and write the form factor
Π1 in terms of a sum over resonances. This gives
X  fρ2 fa2n

0 n
S = 2πξ Π1 (0) = 4πξ − 2 , (97)
n
m2ρn man
where mρn and fρn (man and fan ) denote respectively the mass and decay constant
of the SO(4) (SO(5)/SO(4)) spin-1 resonances. Equation (97) represents the large-N
leading contribution to S from the strong dynamics, interpreted as due to the tree-level
exchange of spin-1 resonances. It is clear at this point the strict analogy with the chiral
coefficient L10 that we computed in the previous section for the case of QCD. Indeed,
by using the Weinberg sum rules and assuming vector meson dominance we obtain
f2 m2ρ
 
S = 4πξ 2 1 + 2 , (98)
mρ ma
which is in complete analogy with the expression of L10 in eq.(94). In particular,√the
sign of S is fixed to be positive. By means of the large-N relation f /mρ ∼ gρ ∼ 4π/ N
(where gρ is the coupling among three composite states, see eqs.(24)-(26)), it is also
easy to see that its size is parametrically suppressed by a factor ξ compared to the
Technicolor estimate of eq.(33). This was in fact expected, considering that ξ → 0
with fixed v is a limit in which all the resonances of the strong sector except the Higgs
become infinitely heavy and decouple. Hence, for ξ small enough the LEP constraints
can be satisfied. If we use eq.(52) and the relation mρ /ma ' 3/5 valid in the 5-
dimensional SO(5)/SO(4) models of Refs. [34, 35], we get
 2
v
S = 4π (1.36) , (99)
mρ

32
 χ1 χ2 (χ1)

W3 B χ1 (χ2) χ1 (χ2)

χ2 B

h h

W3 B χ3 χ3

χ3 B

Figure 8: Logarithmically divergent contributions to S (left diagrams) and T (right dia-

grams) from loops of would-be NG Goldstones χ’s (upper row) and of the Higgs boson (lower
row). In the SM the Higgs divergent contribution exactly matches that from the χ’s to give
a finite result. At scales below mh , the upper
 left diagram contributes to the running of the
coefficient of the operator Tr Σ† Wµν ΣB µν , see eq.(32). Similarly, the upper right diagram

2
contributes to the running of the coefficient of Tr T 3 Σ† Dµ Σ . See Ref. [18].


which leads to a constraint on the mass of the lightest spin-1 resonance mρ . 11

Concerning ∆ρ (or equivalently the Peskin-Takeuchi T parameter), the tree-level
correction due to the exchange of heavy spin-1 resonances identically vanishes in the
SO(5)/SO(4) model as a consequence of the custodial symmetry of the strong sector.
In fact, the absence of this otherwise large correction to ∆ρ is the main reason to
consider this symmetry breaking pattern instead of more minimal ones (like for example
SU (3) → SU (2) × U (1), see Ref.[43]), where no custodial symmetry is present. Non-
vanishing corrections to ∆ρ will follow in general from loops of heavy fermions and
vectors. We do not discuss these effects here, referring to the literature [44–46] for
more details.
There is another important correction to both the S and T parameters, calculable
within the low-energy effective theory, that follows from the modified couplings of the
composite Higgs to the SM gauge bosons, see eq.(56). In the Standard Model the
1-loop contribution of the Higgs boson to the vector self energy exactly cancels the
logarithmic divergence arising from loops of would-be NG bosons χa (see for example
Ref. [18]). The relevant diagrams are shown in Fig. 8. The cancellation follows from the
11
Notice that resumming the effect of the whole tower of resonances, without assuming vector
meson dominance, will in general make the bound stronger. For example, the calculation of S in the
5-dimensional models of Refs. [34, 35] leads to a formula analogous to eq.(99) where the coefficient
1.36 is replaced by 2.08, see [34].

33
fact that the Standard Model is a renormalizable theory, and there are no additional
divergences that cannot be reabsorbed by a renormalization of the gauge couplings
and the Higgs wave function. In other words, there is no counterterm that can cancel
a possible divergence in S and T (at the level of renormalizable operators). Thus, in
the full theory the contribution to S and T must be finite when expressed in terms of
the renormalized parameters. On the other hand, working in a renormalizable ξ gauge
one finds that loops of NG bosons χa give a logarithmically divergent contribution to
S and T . 12 This must be then exactly matched by the Higgs boson contribution at
1-loop to give a finite result.
In a non-renormalizable composite Higgs theory the above argument on the finite-
ness of the S and T parameters does not hold anymore. In particular, as noticed by
the authors of Ref. [45], the modified Higgs couplings to the SM gauge bosons imply
that the contribution of the composite Higgs to the self-energy does not exactly cancel
the infrared log divergence arising from the χ’s. This mismatch leads to a correction
to S and T given by (∆T = ∆ρ/α)
 2
1 2 Λ
∆S = + (1 − a ) log , (100)
12π m2h
 2
3 1 2 Λ
∆T = − (1 − a ) log , (101)
2
16π cos θW m2h

where a parametrizes the shift of the coupling of one Higgs boson to two W ’s, see
eq.(16), and Λ ≈ 4πf is the strong cutoff scale of the theory (i.e. the scale at which
unitarity is ultimately restored in W W scattering). The LEP precision tests thus imply
a constraint on the parameter a. For example, assuming that the only correction
p to S
and T comes from eqs.(100),(101) and setting mh = 120 GeV, Λ = 1.2 TeV/ |1 − a2 |,
one obtains 0.8 . a2 . 1.6 at 99% CL, see Fig. 9. Larger deviations of a from 1
can of course be accommodated if the 1-loop contribution to S and T from the heavy
resonances (partly) compensates the infrared correction of eqs.(100),(101).
The corrections to S and T of Eqs. (97), (100) and (101) together put a strong
bound on the value of the mass of the lightest vector resonances. The size of the
corrections is controlled by the value of ξ and gρ , which in turn determine a and mρ .
As an illustrative example we consider the constraint√that follows in the 5-dimensional
√
SO(5)/SO(4) models of Refs.[34, 35], where a = 1 − ξ, mρ = (3π/8)gρ v/ ξ and
the UV correction to the S parameter is ∆S = 4π(2.08)(v/mρ )2 (see Eq.(99) and
footnote 11). Figure 10 shows the region in the plane (ξ, gρ ) excluded at 99% CL
(blue area) and the isocurves of constant mρ . The mass of the Higgs has been set to
mh = 120 GeV and the argument of the IR logarithm of Eqs.(100,101) to (m2ρ /m2h ).
One can see that values of ξ up to ∼ 0.2 are allowed for large gρ , while a smaller ξ
12
Of course the same result is obtained with any choice of gauge fixing. The renormalizable ξ gauge
is convenient because it shows that the log divergences solely arise from the EWSB sector, i.e. from
the contribution of the NG fields χa , and not from the transverse gauge bosons.

34
 0.01
68%, 90%, 99% CL mh "120 GeV

0.008

0.006
Ε1

0.004

0.002

0.003 0.004 0.005 0.006 0.007 0.008

Ε3

Figure 9: χ2 fit to the parameters 1,3 of Altarelli and Barbieri [47, 48] obtained from
LEP data [49]. The relation to the S and T parameters is as follows: 1 = SM 1 + αT,
2
3 = SM
3 + α/(4 sin θ W ) S. The solid curves represent the 68%, 90% and 99% CL contours
obtained by setting 2 , b to their SM value with mh = 120 GeV and mt = 171.3 GeV. The
black fat dot shows the SM prediction for mh = 120 GeV (a = 1). The blue smaller dots p show
how 1 and 3 are modified by varying a2 from 0 to 2 in steps of 0.1 (for Λ = 1.2 TeV/ |1 − a2 |
and mh = 120 GeV). No additional correction to S and T has been included other than that
of eqs.(100),(101).

is required to have more weakly coupled vector resonances. In any case the mass of
these latter, mρ , must be larger than 3 TeV, which makes their detection at the LHC
challenging (see for example [50]). The bound on mρ can be relaxed if an additional
positive contribution to T is present: for example the dotted black curve of Fig. 10
shows how the excluded region is reduced by adding an extra ∆T = +2 × 10−3 .
We have seen that LEP data constrain the parameter a. It is worth stressing,
however, that no bound exists on the coupling of two Higgses to two vector bosons,
i.e. on the parameter b defined in Eq.(16). Although a, b and c are related in specific
composite Higgs models (see for example Eqs.(57) and (121) for their prediction in the
SO(5)/SO(4) model), this shows that a direct measurement of b from the experiment
would be highly desirable. Unfortunately this seems to be quite difficult at the LHC:
while a can be extracted from both the Higgs decay branching fractions and the analysis
of the W W → W W scattering, the parameter b can be extracted only from the W W →
hh scattering. The exploratory analysis of Ref.[12] shows that observing this process
will be quite challenging at the LHC, although it should be possible at its planned
luminosity upgrade.
From the above discussion we conclude that a mild gap between v and f , such
as for example (v/f )2 = ξ . 0.1, can make a composite Higgs compatible with the

35
 12 6 TeV
5 TeV
7 TeV
8 TeV
10
4 TeV

8
3 TeV
gρ
6
2 TeV

4
1 TeV
2

0 0.1 0.2 0.3 0.4 0.5

ξ

Figure 10: The blue area denotes the region in the plane (ξ, gρ ) excluded at 99% CL
by the LEP data after including the UV and IR corrections to S and T as predicted in
the SO(5)/SO(4) models of Refs.[34, 35], see text. Superimposed in red are the isocurves of
constant mρ (dashed lines). The dotted black curve shows how the excluded region is reduced
by adding an extra ∆T = +2 × 10−3 . The Higgs mass has been set to mh = 120 GeV.

electroweak precision data from LEP. The original models constructed by Georgi and
Kaplan, however, suffer from a much more severe bound on f from CP-violating and
FCNC processes. In those theories the breaking of the electroweak symmetry is trans-
mitted to the quark sector trough the same mechanism of Extended Technicolor the-
ories: at some high scale ΛU V the exchange of massive vectors generates four-fermion
operators made of two SM fermions and two technifermions:

∆L = λ q̄qO , O(x) = ψ̄T C (x)ψT C (x) . (102)

Below the scale Λ the strong dynamics condenses and the composite operator O inter-
polates a Higgs field, λ O(x) ≈ (Λ/ΛU V )[O]−1 H(x). The term of eq.(102) thus becomes
a Yukawa coupling between H and the SM quarks, and gives quark masses of the order

 [O]−1 qL
4π Λ H
mq ∼ v √
N ΛU V qR

Similarly to the case of Technicolor, one can assume that above Λ the strong dynamics
stays close to an IR fixed point where its coupling slowly walks, and the dimension of the
operator O can significantly differ from its classical value. As discussed in section 2.2,

36
a naive argument shows that if one does not want to reintroduce UV instabilities in
the theory, the dimension of the operator O cannot be smaller than 2, thus implying
at least a factor (Λ/ΛU V ) in the quark masses. 13 Besides the term in eq.(102), the
UV dynamics at the scale ΛU V will also generate operators made of four SM fermions.
These are suppressed by 1/Λ2U V and are expected to violate flavor and CP, thus leading
to a strong bound on ΛU V . While in Technicolor theories this in turn implies too
small quark masses as a consequence of the suppression factor (Λ/ΛU V ) ≈ (v/ΛU V ), in
composite Higgs models one can still obtain large enough quark masses by making Λ
large, Λ ≈ f  v. This can be achieved, however, only at the price of fine tuning the
vacuum alignment parameter to be very small, ξ = (v/f )2  1. Therefore, the FCNC
problem of Technicolor can be solved, but the resulting model is highly tuned. At the
same time, a simple explanation of the hierarchy among quark masses is still missing,
the only possible mechanism being a complicated cascade of symmetry breakings as in
Extended Technicolor theories.
There is however a different mechanism that can transmit the EWSB to the SM
fermions and leads to much milder experimental constraints on f . Suppose that some
UV physics at the scale ΛU V generates a linear coupling between a composite operator
O and one SM fermion,
∆L = λ q̄O + h.c. (103)
In this case O must be a fermionic composite operator (made for example, but not
necessarily, of three technifermions) with the same SU (3)c × SU (2)L × U (1)Y quantum
numbers of the SM fermion to which it couples. Hence, there must be at least one
composite operator for each SU (3)c × SU (2)L × U (1)Y quark multiplet. At low energy
the composite Higgs field is interpolated by pairs of fermionic operators OR OL , and
the naive estimate for the quark masses is as follows:
H

√
qL qR N
mq = v λL (Λ)λR (Λ)
4π

The value of the coupling λ at low energy is determined by the dimension of the
corresponding operator O. Particularly interesting is the case in which from ΛU V
down to the scale Λ the strong dynamics is almost conformal and the dimension of
O is constant. Then, at large N the RG evolution of λ is governed by the following
equation:
d N 3
µ λ = γλ + c λ + ... (104)
dµ 16π 2
13
Strongly coupled theories at small N can however evade this conclusion, see the discussion at the
end of section 2.2.

37
Additional terms (indicated by the dots) can be neglected as long as (λ2 /16π 2 )  1 and
N is large. The first term in (104) corresponds to the classical scaling of λ according
to the anomalous dimension γ = [O] − 5/2. The second term, instead, comes from
the wave-function renormalization that (103) induces on the SM fermion, c being a
numerical coefficient of order 1.
For γ > 0 (so that eq.(103) is an irrelevant term in the Lagrangian), the coupling
λ becomes smaller at lower energies, and the second term in the RG equation can be
neglected as long as (λ2 N/16π 2 )  1 at the scale ΛU V . At low-energy one has:
 γ
Λ
λ(Λ) = λ(ΛU V ) . (105)
ΛU V

For γ < 0 (which means that (103) is a relevant deformation of the Lagrangian), if the
coupling λ starts small at the UV scale ΛU V , its RG evolution will be initially driven
by the first term of (104), so that λ increases when evolving to lower energies. For
c negative, λ rapidly grows and becomes non-perturbative, driving the strong sector
away from the fixed point. On the other hand, if c is positive, then the second term
has opposite sign compared to the first, and the strong sector is driven to a new fixed
point at which r
−γ 4π
λ ' λ∗ = √ . (106)
c N
For N large the value of the coupling at the fixed point is perturbative and thus our
derivation (where we neglected the additional terms in the RG equation) can be trusted.
Finally, for γ = 0 the second term in the RG equation leads to a logarithmic evolution
of the coupling.
Depending on whether the anomalous dimensions of the operators OL,R are positive
or negative, the corresponding quark mass can be large or very much suppressed. For
example, if both OL,R have positive anomalous dimensions, γL,R > 0, the naive estimate
for the quark mass reads √  γL +γR
N Λ
mq ∼ v . (107)
4π ΛU V
Although this expression looks similar to that obtained in the case of a bilinear coupling,
here the difference is that one can have γL + γR close or equal to zero without reintro-
ducing any UV instability. The unitary bound on the dimension of a fermionic operator
is 3/2 (which corresponds to the dimension of the free field), implying γL,R ≥ −1. Fur-
thermore, for γ ≥ 0 the operator ŌO (singlet under the SM gauge group) is irrelevant
at large N , and radiative corrections do not reintroduce any UV divergence. This
means that the UV scale ΛU V can be arbitrarily large, possibly equal to the Planck
scale, without suppressing the quark masses. On the other hand, a large value for ΛU V
suppresses any flavor- and CP-violating operator with four SM fermions generated at
that scale, thus resolving the problem of Technicolor theories without any fine-tuning.
As a bonus, when (Λ/ΛU V )  1, differences of O(1) in the anomalous dimensions can

38
 qi qk
√
λi λj λk λl yi yj yk yl
∼ ∼
Λ2 Λ2
qj ql

Figure 11: Flavor-Changing four fermion operators generated by the exchange of composite
states at the scale Λ are suppressed by four powers of the couplings λi . For (λL )i ' (λR )i
this suppression corresponds to the square root of the product of the Yukawa couplings yi of
the external quarks.

generate large hierarchies in the light quark masses. In other words, assuming linear
couplings between the SM fermions and the strong sector, and a vast energy range
over which these can evolve, gives a natural explanation of the hierarchies in the quark
masses [51, 52] and can lead to a qualitative explanation of the pattern of observed
flavor mixings [53–55]. Furthermore, the case with negative anomalous dimensions can
be relevant for reproducing the top quark mass. For example, if both γL,R are negative,
and assuming that the strong sector flows to a new IR fixed-point, one has
4π √
mq ∼ v √ γL γR , (108)
N
which can easily reproduce the experimental top mass in the range −1 ≤ γL,R < 0 even
for moderately large N .
To summarize, linear couplings between the SM fermions and the strong sector
represent an extremely interesting mechanism to communicate the breaking of the
electroweak symmetry to the quark sector and generate the quark masses. The possi-
bility of coupling one SM fermion to three technifermions was first proposed by D.B.
Kaplan [56], as an alternative mechanism to the more standard Extended Technicolor
approach. Ref. [56] however assumed a QCD-like dynamics for the strong sector, and
did not exploit the natural generation of hierarchies in the quark masses that follows
from the RG evolution over a vast energy domain. The importance of linear couplings
was re-discovered only later in the context of extra-dimensional warped field theories 14 ,
when their relevance to explain the flavor structure was realized.
Although flavor-violating local interactions generated at the scale ΛU V can be safely
suppressed in the case of linear couplings, there are still important flavor-violating ef-
fects that can be mediated by the strong sector at the lower scale Λ. In particular,
four-SM fermion operators can be generated by the exchange of the composite reso-
nances of the strong sector, see Fig.11. An interesting possibility is that the strong
14
To my knowledge, the connection between the mechanism of Ref. [56] and extra-dimensional
warped theories was pointed out for the first time by Ref. [57].

39
sector is flavor anarchic, and the flavor structure of the SM Yukawa couplings entirely
arises from the RG evolution of the composite-elementary couplings λi . In this case,
four-fermion operators involving light external quarks will be suppressed by their small
couplings to the strong sector at low energy, providing a sort of GIM protection against
large FCNC [52–55, 58]. Important effect can still arise, however, from the sizable cou-
pling of the third generation quarks to the strong sector. In particular, it has been
shown that important constraints on the scale Λ arise from CP violation in the K K̄
system [59–62], b → sγ [55, 58, 61, 62] and lepton-violating processes such as µ → eγ
and µ → 3e [54,63,64]. Here we will not discuss these constraints, referring to the vast
literature on the subject for more details.
There is an extremely interesting phenomenological consequence of linear couplings
which was already noticed in Ref. [56]: similarly to QCD, where a current made of
quarks has the quantum numbers to excite a heavy spin-1 resonance from the vacuum,
at energies below the scale Λ, at which the strong dynamics is assumed to condense,
a composite operator O can excite a heavy fermionic resonance. More exactly, there
will be a full tower of composite fermions of increasing mass that can be excited by
the operator O. 15 The linear coupling (103) thus becomes a mass mixing term at low
energy between the elementary fermion ψ and the tower of composite fermions χn :
X

Lmix = ∆n ψ̄χn + h.c. , h0|O|χn i = ∆n . (109)
n

Similarly, and in complete analogy with QCD, a conserved current Jµ associated with
the global symmetry G of the strong sector will excite a tower of spin-1 resonances ρn
which will mix with the elementary gauge fields Aµ :
X
Lmix = mρn fρn Aµ ρµn , h0|Jµ |ρn (r )i = rµ mρn fρn . (110)
n

The corresponding phenomenon is known as ρ-photon mixing in the QCD literature.

As a consequence of the mass mixings (109) and (110), the physical fermion and
vector eigenstates (to be identified with the SM fields) will be admixtures of elementary
and composite states. In this case one speaks of partial compositeness of the SM
particles [56,65]. A qualitative and simple understanding of the phenomenology of such
scenarios can be obtained by considering the simplifying limit in which one includes
only the first resonance of each tower in the low-energy theory, and neglects the other
heavy states [65]. For example, the effective Lagrangian describing one elementary
chiral field ψL and one composite heavy fermion χ is
L = ψ̄L i 6 ∂ ψL + χ̄ (i 6 ∂ − m) χ + ∆L ψ̄L χR + h.c. (111)
Notice that, as a result of the RG evolution above Λ, the mass mixing parameter
∆L can be naturally much smaller than the mass m∗ of the composite fermion. The
15
We are assuming that the operator O is vector-like, so that the excited composite fermions are
massive Dirac states.

40
Lagrangian (111) can be easily diagonalized by rotating the left-handed fields:
    
ψL cos ϕL sin ϕL ψL ∆L
→ , tan ϕL = . (112)
χL − sin ϕL cos ϕL χL m∗

The mass eigenstate fields, a light left-handed fermion

p (to be identified with the SM
field), and a heavy Dirac fermion of mass m = m2∗ + ∆2L , are superpositions of
elementary and composite states:

|lighti = cos ϕL |ψi + sin ϕL |χi

(113)
|heavyi = − sin ϕL |ψi + cos ϕL |χi .

The angle ϕL thus parametrizes the degree of partial compositeness of the correspond-
ing SM field. Similar formulas can be derived in the case of the mixing of a right-handed
elementary field in terms of a right handed angle ϕR . Since the origin of the breaking
of the electroweak symmetry resides, by assumption, in the composite sector, the mass
acquired by a SM fermion ψ entirely stems from the composite components of ψL and
ψR ,
y = Y∗ sin ϕL sin ϕR , (114)
where Y∗ is a Yukawa coupling among composites. Thus, heavier SM fields must have
larger degree of compositeness. In particular, light quarks and leptons are almost
elementary fields. This explains why all the standard bounds on the compositeness of
these particles can be easily evaded in the present framework. Furthermore, the Higgs
boson and the longitudinal components of the W and the Z are full composites. The
transverse polarization of the SM gauge fields will be instead partly composites, the
degree of compositeness this time being fixed in term of the ratio of elementary and
composite gauge couplings.
Besides those sketched above, this theoretical framework has simple and important
consequences for the physics at present and future colliders, as well as on the pattern
of deviations expected in precision measurements. We do not have time here to review
all of them, but the interested reader can find more details in Ref. [65].

3.4 Higgs potential from the top quark

So far we have assumed that the Higgs potential at its minimum can induce the cor-
rect amount of electroweak symmetry breaking. Here we want to show that this can
naturally follow due to the contribution of the SM top quark. Let us assume that the
fermionic content of the elementary sector is that of the Standard Model, and that
each SU (3)c × SU (2)L × U (1)Y multiplet couples linearly to a corresponding compos-
ite operator. The composite operators transform as a complete representations of the
global symmetry G of the strong sector, while, in general, the external fermions will
not. This means that the linear couplings violate G explicitly, so that loops of elemen-
tary fermions will induce a Higgs potential. The dominant contribution will come from

41
the top and bottom quarks, since heavier fermions have larger couplings to the strong
sector.
The calculation can be done by closely following the strategy adopted for the
gauge contribution in section 3.1. As before, we will work out the specific case of
the SO(5)/SO(4) model, but our results are generic. As a first step, one needs to spec-
ify how the composite operators transform under SO(5). Following Ref.[34], we will
assume that for each quark generation, qL , uR , dR , there are three composite operators
transforming as spinorial representations of SO(5) with U (1)X charge X = 1/6, such
that (a sum over the three flavors is understood)
L = λq q̄L Oq + λu ūR Ou + λd d¯R Od + h.c. (115)
In general the couplings λq,u,d can be arbitrary matrices in flavor space, but here we
will assume for simplicity that they are diagonal. A spinorial representation of SO(5)
consists of two spinors of SO(4), in the same way as the smallest irreducible represen-
tation of the SO(4, 1) Lorentz group in 5 dimension, a Dirac fermion, is made of two
Weyl fermions of SO(3, 1). Hence, a spinor of SO(5) decomposes as 4 = (2, 1) + (1, 2)
under SU (2)L × SU (2)R .
Similarly to the case of the gauge fields, a useful trick to derive the effective action
for the elementary quarks in the Higgs background is that of uplifting qL , uR and dR
to complete SO(5) spinorial representations,
 u   d 
qR q
 R0 
 
qL
Ψq = , Ψu = , Ψd = , (116)
 
QL
 uR  uR
0
dR dR

where QL , qRu,d , u0R and d0R are non-dynamical spurions. Each of the fields Ψ contains
one doublet of SU (2)L (the two upper components of the multiplet) and one doublet
of SU (2)R (the two lower components). Specifically, qL is the SU (2)L doublet inside
Ψq , uR is the upper component of the SU (2)R doublet inside Ψu , while dR is the
lower component of the SU (2)R doublet inside Ψd . Then, according to the definition
Y = T 3R + X, the hypercharge of qL , uR , dR is correctly reproduced if all the fields
Ψq,u,d are assigned U (1)X charge 1/6. Once written in terms of Ψq,u,d , the couplings of
eq.(115) formally respect SO(5) × U (1)X . Hence, the most general (SO(5) × U (1)X )-
invariant effective action for the elementary quarks, at the quadratic order and in
momentum space, is:
X h i
r r i
Leff = Ψ̄r 6 p Π0 (p) + Π1 (p) Γ Σi Ψr
r=q,u,d
X (117)
Ψ̄q M0r (p) + M1r (p) Γi Σi Ψr .
 
+
r=u,d

As before, we have treated Σ as a constant background and encoded the effect of the
strong dynamics in the form factors Πr0,1 and M0,1
r
(r = q, u, d). The poles of these

42
latter give the spectrum of the fermionic resonances of the strong sector. Using the
expression of the gamma matrices Γi of SO(5) given in the Appendix one easily obtains:
σ̂ ≡ σ â hâ /h
 
i 1 cos(h/f ) σ̂ sin(h/f )
Γ Σi = , (118)
σ̂ † sin(h/f ) −1 cos(h/f ) σ â = {~σ , −i1} .
At this point we keep only the top quark multiplets qL = (tL , bL ) and tR as physical,
dynamical fields, and set to zero all the other fields. The effect of the other elementary
fermions in the Higgs potential is negligible due to their small couplings to the strong
dynamics at low energy. We thus obtain the effective action for qL and tR we were
looking for:
L =q̄L 6 p Πq0 (p) + Πq1 (p) cos(h/f ) qL

+ t̄R 6 p Πu0 (p) − Πu1 (p) cos(h/f ) tR

(119)
+ sin(h/f ) M1u (p) q̄L Ĥ c tR + h.c.
Here Ĥ c = iσ 2 Ĥ and Ĥ has been defined in eq.(49). In particular, the top quark mass
can be extracted from the Yukawa term between tL and tR by taking the low-energy
limit p ' 0:
v M1u (0)
mt ' p q . (120)
f (Π0 (0) + Πq1 (0))(Πu0 (0) − Πu1 (0))
By expanding eq.(119) around the Higgs vev one also immediately obtains the expres-
sion of the parameter c defined in eq.(16):
p
c= 1−ξ. (121)
From the effective action one easily derives the 1-loop Coleman-Weinberg potential:
Z 4 h
dp 
2 2
i
V (h) = −2Nc log Π bL + log p Π Π
tL tR − Π tL tR , (122)
(2π)4
where Nc = 3 and we have defined
ΠtL = ΠbL ≡ Πq0 + Πq1 cos(h/f )
ΠtR ≡ Πu0 − Πu1 cos(h/f ) (123)
ΠtL tR ≡ M1u sin(h/f ) .
The first term in the integral of eq.(122) is the contribution of bL , while the second is
due to the top quark (tL and tR ). The potential can also be conveniently rewritten (up
to terms that do not depend on the Higgs field) as
Πq1
Z 4 
Πu1
   
dp h h
V (h) = − 2Nc 2 log 1 + q cos + log 1 − u cos
(2π)4 Π0 f Π0 f
(124)
(M1u sin(h/f ))2
 
+ log 1 − 2 q ,
p (Π0 + Πq1 cos(h/f ))(Πu0 − Πu1 cos(h/f ))

43
 + + + ···

tL tL tR

+ + ···
tR tL
tR

Figure 12: 1-loop contribution of the SM top and bottom quark to the Higgs potential.
Upper row: diagrams where the same elementary field, either qL = (tL , bL ) or tR , circulates
in the loop with a propagator i/(6 p Π0 ). A grey blob denotes the form factor 6 pΠ1 . Lower row:
diagrams where both tL and tR circulate in the loop with a Higgs-dependent propagator (see
text). In this case a grey blob denotes the form factor M1u .

where this time the first two terms in the integral can be thought of as due to the
resummation of 1-loop diagrams where only qL or tR are exchanged, see Fig. 12 (upper
row). The last term, instead, comes from resumming the diagrams where both tL and
tR circulate in the loop with a Higgs-dependent propagator, respectively
i i
, and ,
6 p (Πq0 + Πq1 cos(h/f )) 6 p (Πu0 − Πu1 cos(h/f ))
see Fig. 12 (lower row). As for the case of the gauge fields, the finiteness of the integral
is guaranteed by the convergence of the form factors M1u and Πu,q 1 at large Euclidean
momenta. Provided these decrease fast enough, the potential can be reasonably well
approximated by expanding the logarithms at first order. This gives:
h h
V (h) ' α cos − β sin2 , (125)
f f
where the coefficients α and β are defined in terms of integrals of the form factors.
Including the contribution of the gauge potential (59) to β, one has:
Πq1
Z 4  u 
dp Π1
α = 2Nc −2 q
(2π)4 Πu0 Π0
Z 4  (126)
(M1u )2

dp 9 Π1
β= 2Nc − .
(2π)4 (−p2 ) (Πq0 + Πq1 )(Πu0 − Πu1 ) 8 Π0

44
 We see that even though the gauge contribution to β is negative, the EWSB can
still be triggered by the top contribution if α ≤ 2β. In this case the potential has a
minimum at  2
2 hhi α
ξ = sin =1− . (127)
f 2β
This shows immediately that small values of ξ require a fine tuning between α and
β. In fact, this is a general feature of composite Higgs models: the misalignment of
the vacuum comes from the interplay of different terms in the potential (specifically,
sin2 and cos in eq.(125)), each of which is a periodic function of θ = h/f . One thus
naturally expects large values of the angle θ at the minimum (ξ ∼ 1), or no symmetry
breaking at all (ξ = 0). Small values of θ are unnatural and can arise only through a
fine-tuned cancellation among different terms of the potential. 16 Therefore, the value
of ξ gives a rough estimate of the tuning of the theory. In particular, models where
ξ ∼ 0.1 is required to pass the LEP precision tests are tuned at the level of 10%.
As a final exercise, it is instructive to derive the expression of the physical Higgs
mass that follows from eq.(125). Taking the second derivative of the potential at its
minimum one has m2h = 2βξ/f 2 . It is convenient to define

(M1u )2
F (Q2 ) = ,, (128)
(Πq0 + Πq1 )(Πu0 − Πu1 )
so that (neglecting for simplicity the gauge contribution)

Nc F (Q2 ) Nc
Z
β = 2 F (0) dQ2 ≡ 2 F (0) m2∗ , (129)
8π F (0) 8π
where the last equality defines m∗ . This is the scale at which the top loop is cut off,
and is naturally expected to be of the order pof the lightest fermionic resonance of the
strong sector. Using the fact that mt = ξ F (0), see eq.(120), one finally obtains
(yt = mt /v):
2 yt2 2
mh = 2Nc 2 m∗ ξ . (130)
8π
This result could have been guessed simply by naive dimensional analysis: the Higgs
mass is one loop suppressed compared to the scale of the heavy resonances m∗ , and
the SM coupling responsible for the explicit breaking of the Goldstone symmetry is
the top Yukawa coupling in this case. A further suppressing factor ξ comes from the
16
An interesting exception is when one term in the potential starts at order h4 (for example a sin4
term [34]), thus contributing only to the quartic coupling and not to the Higgs mass term. If the
coefficient of such term is slightly larger than that of the other terms in the potential, then a small
value of θ naturally follows at the minimum. This way of getting naturally a large gap between f and
v is analogous to the mechanism at work in Little Higgs theories, where the large quartic follows from
collective breaking (see [2] and references therein). Unfortunately, no fully natural mechanism have
been found so far (other than collective breaking), to make the coefficient of the h4 term parametrically
larger than that of the remaining terms in the potential.

45
tuning among different terms in the potential. One can also use the NDA estimate
m2∗ ≈ f 2 N/16π 2 , where N is the number of ‘colors’ of the strong dynamics, to rewrite
mh as follows:
4Nc 2
m2h ∼ mt . (131)
N
This shows that the Higgs mass is naturally expected to be . mt , and that it remains
constant in the limit ξ → 0 with v fixed.

4 The holographic Higgs

So far we have discussed the phenomenology and the predictions of composite Higgs
models assuming that some dynamics exists which forms the Higgs as a bound state
at low energy. Here I want to give one example of such dynamics, illustrating an ex-
tremely fascinating possibility: the composite pNG Higgs might be identified with the
fifth component of a gauge field living in a 5-dimensional (5D) spacetime. Theories of
this kind are not just beautiful because of their profound implications on our under-
standing of Nature, but they are also extremely predictive: we will show, although in
the context of a simplified abelian model, how the form factors of the SO(5)/SO(4)
example discussed previously can be computed analytically.
Let us start by considering a gauge theory on a flat 5-dimensional interval: 17 the
metric is that of 5D Minkowski spacetime, ηM N = (+, −, −, −, −) (M, N = µ, 5), and
the fifth spatial coordinate runs from 0 to L, where L is the dimension of the extra
dimension: x5 ∈ [0, L]. One can also start with x5 defined on a circle of radius R and
identify opposite points by means of a Z2 symmetry:

x5 ∼ 2πR − x5 , x5 ∈ [0, 2πR] . (132)

The spacetime obtained in this way is called orbifold, and denoted with S 1 /Z2 . As a
consequence of the identification (132), only half of the points are physically inequiv-
alent, for example those lying between 0 and πR. The orbifold is thus equivalent to a
segment of length L = πR, see Fig. 13. 18
The action describing a (non-abelian) gauge field AM and a fermion field Ψ living
on the 5-dimensional interval is:
Z L  
1
Z
4 5 M N RS M


S= dx dx − 2 FM R FN S η η + Ψ̄ iDM Γ − mΨ Ψ , (133)
0 g5
17
I will assume that the reader is familiar with the formalism of field theories in higher-dimensional
spacetimes. Excellent introductions to the subject are Refs.[5–8]. Some of the topics discussed in this
section are also introduced in the TASI lectures by R. Sundrum [3] and the review [4] by M. Serone,
to which I refer for further details and a list of references on models of gauge-Higgs unification.
18
Although trivial at the level of spacetime, the equivalence between field theories on the orbifold
and on the interval is valid also at the level of field configurations. See for example the discussion in
Ref. [66].

46
 2πR − x5

Z2 0 πR
2πR 0 πR
=
0 πR

x5

Figure 13: The orbifold construction: opposite points on the circle are identified by a Z2
symmetry. The resulting space is equivalent to a segment of length L = πR.

where ΓM are the 5-dimensional gamma matrices,

ΓM = γ µ , −iγ 5 ,
 M N
= 2 ηM N .

Γ ,Γ (134)

The smallest irreducible representation of the 5-dimensional Lorentz group SO(4, 1) is

a Dirac fermion, so that the bulk fermion Ψ has both a left-handed and a right-handed
component 19
ΨL (x, x5 )
 
5
Ψ(x, x ) = . (135)
ΨR (x, x5 )
A gauge-invariant mass mΨ for the fermion field is thus allowed in the bulk. Notice
that the 5D gauge coupling has dimension of mass−1/2 , [1/g52 ] = 1, and this is a sign
that the theory described by the action (133) is non-renormalizable. Indeed, it is valid
up to energies of the order ΛS ≈ 16π 2 /g52 , below which it can be considered as the
low-energy effective description of some more fundamental theory. In spite of the non-
renormalizability, there are important physical observables – we will see that the Higgs
potential is one of those – which are UV finite and thus calculable.
Since the spacetime has boundaries, the action (133) alone does not completely
define the theory: one has to specify the fields’ boundary conditions at x5 = 0 and
x5 = L. These must be chosen so that the variation of the action vanishes, upon
evaluation on the equations of motion, both in the bulk and on the boundaries. For
example, in the case of the fermion Ψ, the variation of the action reads
δS δS
δS = δΨ + δ Ψ̄
δΨ δ Ψ̄
Z L (136)
 1
Z Z
4 5
L
d4 x Ψ̄γ 5 δΨ − δ Ψ̄γ 5 Ψ 0 .
 
= d x dx δ Ψ̄ DΨ + DΨ δΨ +
0 2
The first term on the second line of the previous formula vanishes upon evaluation on
19
Here ‘left’ and ‘right’ refer to the chirality in 4 dimensions, that is: γ5 ΨR,L = ±ΨR,L .

47
the bulk equations of motion,
(
M

i 6 ∂ΨR = (∂5 + mΨ ) ΨL
DΨ ≡ i∂M Γ − mΨ Ψ = 0 −→ (137)
i 6 ∂ΨL = (−∂5 + mΨ ) ΨR .
The second term instead,
1
Z
L
d4 x Ψ̄L δΨR − Ψ̄R δΨL − δ Ψ̄L ΨR + δ Ψ̄R ΨL 0 ,

2
must vanish when the boundary conditions at x5 = 0, L are imposed. As implied by the
coupled system of equations of motion (137), the boundary conditions of ΨL and ΨR
are not truly independent: fixing one determines automatically also the other. Thus, at
each boundary x5i = 0, L there are two possible choices of boundary conditions: either
ΨL (x5i ) = 0 and thus ∂5 ΨR (x5i ) = mΨ ΨR (x5i ) , (138)

or

ΨR (x5i ) = 0 and thus ∂5 ΨL (x5i ) = −mΨ ΨL (x5i ) . (139)

In the particular case of vanishing bulk mass, mΨ = 0, the above conditions simplify
to
( (
ΨL (x5i ) = 0 (Dirichlet, −) ∂5 ΨL (x5i ) = 0 (Neumann, +)
or (140)
∂5 ΨR (x5i ) = 0 (Neumann, +) ΨR (x5i ) = 0 (Dirichlet, −)
Similar conditions (of Neumann or Dirichlet type) also apply for the gauge field AM .
Since the spacetime is compact, each 5D field Φ can be decomposed in Fourier
harmonics, X
Φ(x, x5 ) = φ(n) (x)ξn (x5 ) , (141)
n
5
where the ξn (x ) form a complete set of orthogonal functions on the interval. The
φ(n) (x) (i.e. the Fourier harmonics of Φ(x, x5 )) are called Kaluza-Klein (KK) modes
and behave like 4-dimensional massive fields with masses increasing with n. Consider
for example the case of a fermion field: each chiral component has a decomposition in
(n)
terms of Kaluza-Klein modes ψL,R (x):
X (n) X (n)
ΨL (x, x5 ) = ψL (x)ξnL (x5 ) , ΨR (x, x5 ) = ψR (x)ξnR (x5 ) . (142)
n n

For mΨ = 0, a complete set of orthogonal wave functions ξn is given by (n = 0, 1, 2, . . . )

2nπx5 (2n + 1)πx5
     
(++) (+−)
ξn (x5 ) = cos , ξn (x5 ) = cos
L L
(143)
5 5
     
2nπx (2n + 1)πx
ξn(−−) (x5 ) = sin , ξn(−+) (x5 ) = sin ,
L L

48
 (s ,s )
where ξn 0 L satisfies a condition of type s0 at x5 = 0 and type sL at x5 = L, and
si = + (si = −) means Neumann (Dirichlet). As implied by eq.(140), if ξ L has (s0 , sL )
boundary conditions, then ξ R will have (−s0 , −sL ) conditions. In the case of (±, ∓)
fields, the Kaluza-Klein modes form a tower of four-dimensional Dirac fermions with
mass mn = (2n + 1)π/L (n = 0, 1, 2, . . . ). In the case of (±, ±) fields, the massive
levels are at mn = 2nπ/L. In addition to those, there is also a ‘zero mode’ (n = 0)
corresponding to a massless chiral fermion. Chirality comes from the fact that for n = 0
only the ξ (+,+) wave function admits a non-trivial solution (with a constant profile),
see eq.(143). These considerations remain valid even for mΨ 6= 0, although the value
of the masses of the non-zero KK modes will change.
The possibility of obtaining a spectrum of chiral fermions at low energy is in fact
one of the motivations to consider the interval rather than other compact spaces, like
for example the circle. There is another reason however: the boundary conditions
imposed on the gauge field can lead to an elegant mechanism of symmetry reduction
at low energy. Let us see how. In general, a 5D gauge transformation on AM has the
form
AM → ΩAM Ω† − i Ω∂M Ω†
(144)
Ω(x, x5 ) = P exp iT A αA (x, x5 ) ,


where T A are the generators of the bulk gauge group G, and P represents the path-
ordering of the exponential. An infinitesimal transformation with M = µ, in particular,
transforms AA A A
µ → Aµ − ∂µ α , which implies that each of the gauge parameters α
A
A
must respect the same boundary conditions of Aµ . Since the gauge field is part of
a covariant derivative and has a geometrical meaning, it follows that AA 5 must have
opposite boundary conditions compared to AA µ . The most generic set of boundary
conditions that can be consistently applied on the various components of the gauge
field is then 20
Aaµ (+, +) , Aa5 (−, −) T a ∈ Alg {H = H1 ∩ H0 }

Aāµ (+, −) , Aā5 (−, +) T ā ∈ Alg {H0 /H}

(145)
Aȧµ (−, +) , Aȧ5 (+, −) T ȧ ∈ Alg {H1 /H}

Aâµ (−, −) , Aâ5 (+, +) T â ∈ Alg {G/H0 } ∩ Alg {G/H1 } ,

where H0,1 are subgroups of G. One can notice two important facts: First, the set
of generators corresponding to the gauge fields that do not vanish at x5 = 0 (Aaµ and
Aāµ ) form the subgroup H0 , while those associated to the fields that do not vanish at
x5 = L (Aȧµ and Aâµ ) form the subgroup H1 . In other words, the bulk gauge symmetry
G is reduced to H0 (H1 ) on the boundary x5 = 0 (x5 = L), see Fig. 14. Second, the
20
Here we neglect the effect of possible mass terms localized on the boundaries.

49
 H0 G H1

x5 = 0 x5 = L

Figure 14: The bulk gauge symmetry G is reduced to the subgroup H0 at the boundary
x5 = 0, and to H1 at x5 = L.

gauge invariance at low energy is H = H1 ∩ H0 . Indeed, the basis of wave functions

relevant for the Kaluza-Klein decomposition of the fields in (145) is that of eq.(143).
Since only (+, +) fields have (massless) zero-modes, this means that the low-energy
spectrum comprises a set of gauge fields Aaµ T a ∈ Alg {H}, and a set of 4D scalars Aâ5
living in Alg {G/H0 } ∩ Alg {G/H1 }
In addition to these massless fields, there is a tower of massive spin-1 fields trans-
(n)
forming under the adjoint representation of G. Level by level, each of the modes A5
(n)
is eaten in a Higgs mechanism to form massive vectors together with Aµ . In fact,
there is a gauge transformation that eliminates the x5 dependence of A5 from the very
beginning, and leaves only its zero mode (which has a constant wave function). It can
be constructed as follows: In a 5D Minkowski spacetime one can go to an axial gauge
where A5 = 0 by performing the following gauge transformation:
( Z 5 )
x
Ω(x, x5 ) = P exp i dy A5 (x, y) . (146)

This is not an allowed gauge transformation on the interval, since if A5 has a constant
profile, the gauge parameter does not satisfy the correct boundary conditions. However,
one can obtain a proper gauge transformation by simply subtracting the zero mode of
A5 in the exponent:
( Z 5 )
x
x5 (0)
 
5
Ω(x, x ) = P exp i dy A5 (x, y) exp −i √ A5 (x) . (147)
0 L
√
The factor 1/ L comes from the normalization of the zero-mode (constant) wave
function, Z L

2
1= dy ξ (0) (y) . (148)
0
The existence of an axial gauge where A5 does not depend on x5 thus shows that only
(0)
the zero mode A5 is physical, all the other modes can be gauged away.

4.1 A5 as a pseudo Nambu-Goldstone boson

The existence of a massless scalar field in the spectrum, the zero mode of A5 , should
have raised a crucial question from the reader: is its lightness just an accident, perhaps

50
only valid at tree level, or there is a more profound reason ? In this latter case A5
would be a natural candidate to play the role of the Higgs boson. There is in fact
a simple reason why A5 is massless: locality and the 5D gauge invariance forbid a
potential for A5 at tree-level. This is because the only gauge invariant, local operators
one can write in the 5D theory involve the antisymmetric tensor FM N , and since Fµ5 =
∂µ A5 − ∂5 Aµ − i[Aµ , A5 ], there is no way to form terms with only A5 and no four-
dimensional derivative. 21
At the 1-loop level, on the other hand, a potential for A5 can arise from non-local
operators. One can indeed construct a gauge covariant variable, the Wilson line W (x),
which is a non-local function of A5 :
 Z L 
5 5
W (x) ≡ exp i dx A5 (x, x ) ≡ exp {iθ(x)} . (149)
0

After canonically normalizing the kinetic term of A5 one has

√ (0)
θ(x) = (g5 L)A5 (x) , (150)

so that W (x) is just the exponent of the zero mode of A5 . According to its definition,
under a gauge transformation the Wilson line transforms as

W (x) → Ω(x, L) W (x) Ω† (x, 0) , (151)

which intuitively suggests that a non-vanishing potential for A5 can arise from 5-
dimensional loops that stretch from one boundary to the other. We will show in
the following that this intuition is indeed correct. Here we just want to notice that a
possible potential for A5 must be of the form:
1
V (θ) = f (θ) , (152)
L4
where f (θ) is a periodic function of θ. Indeed, if V arises at the 1-loop level as the effect
of non-local operators, it must be finite, since there are no local counterterms which
could cancel possible divergences. Furthermore, it can depend on A5 only though the
Wilson line W , which is a periodic function of θ. This implies the periodicity of f , and
the fact that the overall dimension of V is set by the length L of the extra dimension.
In other words, the potential for A5 is a finite-volume effect, very much similarly to
the Casimir effect.
Thus, A5 is massless at the tree level and acquires a finite (i.e. non-divergent)
mass radiatively. This should sound familiar to the reader as the usual situation for a
pseudo-NG boson. Indeed, A5 is a pseudo Nambu-Goldstone boson of the 5D theory.
21
Notice, on the other hand, that in six dimensions a quartic coupling for A5 and A6 arises from
the non-abelian structure of the kinetic term. For an example of a 6-dimensional theory exploiting
this tree-level potential see [67] and references therein.

51
The easiest way to show it is by adopting the point of view of a 4-dimensional observer
located on one of the two boundaries of the extra dimension, for example at x5 = 0.
From her/his local perspective, the values of the bulk fields at the x5 = 0 boundary,
Φ0 (x) = Φ(x, x5 = 0), as well as possible additional localized fields, act like a 4D
sector with local invariance H0 . The dynamics associated with the degrees of freedom
living in the bulk and at x5 = L, on the other hand, is interpreted as a 4D strongly
interacting sector with a global invariance G broken down to H1 . As we now want to
show, this breaking is spontaneous rather than explicit, and the associated NG bosons
are the zero modes of A5 . The operative definition of the strong sector is through
the 5D functional integral, which can be performed in two steps: In the first step one
integrates over the bulk fields Φ(x, x5 ) while keeping their value at x5 = 0 fixed:
Z
Z = dΦ eiS[Φ]+iS0 [Φ]
Z Z
iS0 [Φ0 ]
= dΦ0 e dΦ eiS[Φ] (153)
Φ0
Z
≡ dΦ0 eiS0 [Φ0 ]+iSef f [Φ0 ] .

This is equivalent to integrating out the degrees of freedom in the bulk and at the
x5 = L boundary, and defines a 4-dimensional effective action,
Z
iSef f [Φ0 ] ≡ log dΦ eiS[Φ] , (154)
Φ0

which encodes their dynamics. In eq.(153) we have singled out possible terms in the
action localized at x5 = 0, Z
S0 [Φ] = d5 x δ(x5 )L0 , (155)

which thus depend only on Φ0 . As a second step, one integrates over all values of Φ0 .22
This defines a correspondence, or rather a ‘holographic dictionary’, which allows
one to translate the 5D theory into a 4-dimensional one defined in terms of an ‘elemen-
tary’ weakly-interacting sector (the boundary degrees of freedom at x5 = 0) coupled
to a strongly-interacting one (the dynamics of the bulk and of the x5 = L boundary),
see Fig. 15. In this perspective, the Kaluza Klein modes of the 5D theory must be
interpreted as the mass eigenstates resulting from the admixture of the massive res-
onances of the strong sector with the fields of the elementary sector. This is in fact
in complete analogy with the discussion on partial compositeness of section 3.3. The
above holographic description is clearly inspired by the notorious AdS/CFT correspon-
dence [69,70–72], but it does not mean to be an exact duality. Rather, it is a way to
22
At this level one can make use of Lagrange multipliers to ensure that bulk fields with Dirichlet
boundary conditions at x5 = 0 vanish, see for example ref.[68].

52
 G
H0 H1

L
0
=

=
x5

x5
Strong Sector
Elementary
Sector

Ψ
G → H1
Aµ (H0 )

Figure 15: The holographic dictionary.

define the 4D strong dynamics. Still, such 4D description of the 5D theory turns to be
extremely useful in order to get a quick qualitative understanding of its physics and
to devise models. At the same time, it indicates a different procedure, alternative to a
Kaluza-Klein reduction, to extract the 4D low-energy theory starting from the full 5D
one (see for example [70–74, 68, 75] and the reviews [4,1]).
To prove that the strong dynamics, as defined above via the holographic dictionary,
has a global symmetry G spontaneously broken to H1 , one can argue as follows.23 Let
us consider first the case in which H0 is a subgroup of H1 , so that there is a zero mode
of A5 for each of the G/H1 generators, see eq.(145). As shown above, it is always
possible to choose an axial gauge where A5 does not depend on x5 , so that only its
zero modes are non vanishing. Then, saying that the strong dynamics has a global
symmetry G means that if we rotate the boundary fields Φ0 by a global transformation
g ∈ G, we expect Sef f [Φ0 ] to be invariant:

Φ0 → g Φ0 , Sef f [g Φ0 ] = Sef f [Φ0 ] , g ∈G. (156)

23
See for example Refs. [43,75,4]

53
In this sense, rotating the external fields Φ0 coupled to the strong sector is a way to
probe its global symmetries. Intuitively, we expect Sef f [Φ0 ] to be at least invariant
under global H1 transformations, since the degrees of freedom in the bulk and on the
x5 = L boundary certainly respect this symmetry. In fact, Sef f [Φ0 ] is invariant under
local H1 transformations,

Φ0 → h(x)Φ0 , Sef f [h(x)Φ0 ] = Sef f [Φ0 ] , h ∈ H1 , (157)

since even after choosing the axial gauge, the action in the bulk and at x5 = L is
invariant under 5D gauge transformations that do not depend on x5 . The boundary
conditions at x5 = L imply that such x5 -independent transformations must belong
to H1 .
Starting from the axial gauge, one can further perform a field redefinition of the
form
Φ → Φ0 = ΩΦ
(158)
AM → A0M = ΩAM Ω† − iΩ∂M Ω† ,

with
Ω(x, x5 ) = exp i(x5 − L)A5 (x) .

(159)
This sets A5 to zero everywhere in the extra dimension (A05 = 0) except at x5 =
0. Notice indeed that eq.(158) is almost a gauge transformation: it would be so if
its exponent vanished at x5 = 0, as required for the parameter of a genuine gauge
transformation along the G/H1 direction. This means that away from x5 = 0 the
redefinition (158) acts like a real gauge transformation and leaves the action in the
bulk and at x5 = L invariant. Its only additional effect is that of changing all the
boundary conditions at x5 = 0 by an A5 -dependent factor,

Φ0 (x) → Φ00 (x) = Ω(x, x5 = 0)Φ0 (x) = e−iθ(x) Φ0 (x) , (160)

where exp(iθ(x)) is the Wilson line from x5 = 0 to x5 = L defined by eqs.(149) and

(150). This shows that the theory with A5 and boundary conditions Φ0 is equivalent
to a theory with vanishing A5 and boundary conditions Φ00 = e−iθ(x) Φ0 .
The invariance under a global G rotation, eq.(156), then follows for exp(iθ(x))
transforming according to the usual transformation rule of a NG field [76],
0
eiθ(x) → g eiθ(x) h† (θ(x), g) ≡ eiθ (x) , (161)

where h(θ(x), g) is an element of H1 . Indeed, from eqs.(161) and (160) one has that
Φ00 → h(θ, g)Φ00 , and from the invariance of the effective action under local H1 trans-
formations, eq.(157), one deduces the invariance under global G transformations. This
proves that the global symmetry G of the strong sector is non-linearly realized, and
that the zero modes of A5 are the associated NG bosons. In the general case in which

54
the local invariance of the x5 = 0 boundary, H0 , is not a subgroup of H1 , part of the
NG bosons are eaten to give mass to the corresponding elementary gauge fields. The
uneaten NG bosons are exactly dim(G/H1 ) − dim(H0 /H), and correspond to the zero
modes of A5 , see eq.(145).
At this point it should be clear that the same pattern of global and local symmetries
realized in 4-dimensional composite Higgs models can be obtained in a 5D gauge theory
on the interval. For example, the SO(5)/SO(4) model discussed in the previous section
can be obtained from a 5D theory with bulk gauge symmetry G = SO(5) × U (1)X
reduced to H0 = SU (2)L × U (1)Y at x5 = 0 and to H1 = SO(4) × U (1)X at x5 = L:

SO(5) × U (1)X

SU (2)L × U (1)Y SO(4) × U (1)X

The spectrum of zero-modes consists of the gauge fields of SU (2)L ×U (1)Y , Aaµ , and four
massless scalars Aâ5 transforming as a 4 of SO(4) or, equivalently, as a complex doublet
of SU (2)L . In addition to these, there is a tower of massive spin-1 states transforming
as adjoints of SO(5). Comparing with the SO(5) composite Higgs theory discussed in
the previous section, it is clear that the Aaµ ’s play the role of the SM gauge fields, while
the massive KK vectors correspond to the spin-1 resonances of the strong sector. From
the viewpoint of an observer on the x5 = 0 boundary, the dynamics of the bulk and
of the boundary at x5 = L act like a strongly-interacting sector with global invariance
SO(5) × U (1)X → SO(4) × U (1)X . The associated NG bosons, our holographic Higgs,
are the zero modes of A5 [34, 43]. As a consequence, Sef f has the same structure and
the same global symmetry of the effective action (48) obtained in the previous section
by integrating out the strong dynamics. By perturbatively solving the 5D theory one
can thus compute the analog of the form factors which were previously introduced to
encode the 4D strong dynamics. The 5D theory, in other words, provides a calculable
model for the 4D strong dynamics.
It has to be remarked that the above distinction between a strongly coupled sector
(corresponding to the dynamics in the bulk and at x5 = L) and the ‘elementary’ sector
living at x5 = 0 truly makes sense only if the latter is weakly coupled (within itself
and to the strong sector). In the case of a 5D theory on a flat extra dimension this
can be ensured by introducing large kinetic terms for the fields at x5 = 0, as part of
S0 [74]. In this way the two boundaries are treated differently, and the equivalence
between observers at x5 = 0 and x5 = L is lost. Remarkably, 5D theories defined on
a warped extra dimension [77] automatically satisfy the above requirement, provided
the elementary sector is identified with the degrees of freedom living on the so-called
UV brane. Most importantly, in that case the theory can be extrapolated up to the
Planck scale, and the Planck-electroweak hierarchy is generated by the 5-dimensional
geometry.

55
 (0)
(0) A5
A5
(0)
A5

X X
ψ (n) + ψ (n) + ···
n n (0)
A5
(0) (0)
A5 A5

Figure 16: 1-loop contribution to the effective potential of A5 from the KK modes of a
bulk fermion.

If one is interested in observables saturated in the infrared, on the other hand, the
exact ultraviolet completion of the theory is not important. The effective potential
of the holographic Higgs, as we discussed above, is one of these calculable quantities
independent of the UV physics. For its calculation we can thus consider a flat extra
dimension with no loss of generality.

4.2 Effective Potential of the Holographic Higgs

A straightforward way to compute the potential for A5 , our holographic Higgs from
the 5D theory, is that of performing a Kaluza-Klein decomposition of the bulk fields
and resumming the series of 1-loop diagrams induced by the virtual exchange of the
KK modes. For example, Figure 16 shows the diagrams corresponding to the 1-loop
contribution of a bulk fermion. As a consequence of locality and 5D gauge invariance,
we expect the final result to be finite, although the contribution of single modes is
divergent. 24
There is another way to proceed, however, which is more natural from the holo-
graphic point of view: one can first derive the effective holographic action Sef f [Φ0 ] at
x5 = 0 in the background of A5 . Using that action one then computes the contribution
to the potential of A5 that comes from loops of the elementary fields Φ0 . This way to
proceed is equivalent to the calculation in the Kaluza-Klein basis, since we know that
the potential is a non-local finite-volume effect, which means that any 5D loop has to
stretch from one boundary to the other in order to give a non-vanishing contribution:
24
Obviously, for the cancellation to happen properly it is crucial to choose a regulator that respects
the 5D gauge invariance. See for example [78] and references therein.

56
This is again simply understood from the holographic viewpoint, where the boundary
at x5 = 0 is interpreted as the source of the explicit breaking of the global symmetry
G, while the symmetry reduction G → H1 at x5 = L corresponds to a spontaneous
breaking. The above argument thus shows that any loop contributing to the potential
of A5 must involve the virtual propagation of the boundary degrees of freedom Φ0 :

Φ0

This means that one can perform the calculation by first deriving the effective action
for the Φ0 ’s, and then use that to make 4-dimensional loops. In full analogy with
the 4D composite Higgs model of the previous section, the finiteness of the result is
ensured in this case by the momentum dependence of the form factors that describe
the interactions of the holographic Higgs with the boundary fields. In this sense, even
in this case we can speak of the Higgs as a composite particle.
A crucial simplification in the explicit computation of the boundary action comes
from performing the field redefinition (158), which moves the dependence upon A5
to the fields’ boundary conditions at x5 = 0. This means that we can derive the
holographic action at x5 = 0 simply by setting A5 = 0 when solving the equations of
motions in the bulk, and using the new boundary conditions Φ00 = e−iθ(x) Φ0 .
All the above considerations can be more concretely illustrated by means of an
explicit example. We will consider a simplified abelian theory with an SO(2) bulk gauge
invariance fully broken on the boundaries. We introduce one bulk fermion transforming
as a doublet of SO(2). The fields’ boundary conditions are thus as follows:
" #
Ψ1L (+, +) Ψ1R (−, −)
[Aµ (−, −) A5 (+, +)] , , (162)
Ψ2L (−, −) Ψ2R (+, +)
and the 5D action reads:
Z L  
1 2 1
Z
4 5 M


S= dx dx − 2 FM N + 2 Ψ̄ iDM Γ − m Ψ . (163)
0 4g5 gΨ
For later convenience, and in analogy with the gauge field, we have introduced a pa-
rameter gΨ with dimensions of mass−1/2 , so that the dimension of the fermion field Ψ is

57
3/2 as in 4 dimensions. The value of gΨ can be for example chosen so as to canonically
normalize the kinetic term of the holographic fermion. We will see that, in absence of
localized kinetic terms for Ψ, gΨ does not enter the expression of the potential.
(0)
As a first step to compute the potential of A5 we need to derive the boundary
action at x5 = 0. As already stressed, in the case of a fermion field the bulk equations
of motion connect its chiral components ΨL and ΨR , and one cannot fix simultaneously
the value of both at x5 = 0. We will thus fix that of ΨL ,

ΨL (x, x5 = 0) ≡ Ψ0L , (164)

and let ΨR be free to vary. At x5 = L we instead impose the boundary conditions

specified by eq.(162):
(1) (2)
ΨR (x, x5 = L) = 0 , ΨL (x, x5 = L) = 0 . (165)

A consequence of demanding a fixed (non-zero) boundary value Ψ0L is that the variation
of the action does not vanish anymore at x5 = 0 (see eq.(136)):
Z L
1
Z
4
dx5 δ(x5 ) Ψ̄0L δΨR + δ Ψ̄R Ψ0L .
 
δS = dx (166)
2 0

We can however solve this problem by introducing an extra boundary action of the
form Z L
1
Z
S0 = 2 d x dx5 δ(x5 ) Ψ̄L ΨR + Ψ̄R ΨL ,
4
 
(167)
2gΨ 0

so that the variation of the total action is zero: δS + δS0 = 0 [68].

At this point we are ready to solve the bulk equations of motions and derive the
boundary action for Ψ0L . Let us work in mixed momentum-coordinate space and look
for a solution of the system of equations
(
− 6 p ΨL + (∂5 − m) ΨR = 0
(168)
− 6 p ΨR + (−∂5 − m) ΨL = 0

of the form
ΨL,R (p, x5 ) = fL,R (p, x5 )ΨL,R (p) . (169)
We require (
(∂5 − m) fR (p, x5 ) = α p fL (p, x5 )
(170)
(−∂5 − m) fR (p, x5 ) = β p fR (p, x5 ) ,
where α and β are numerical coefficients. If these conditions are satisfied, the Dirac
equations become (
− 6 p ΨR (p) = β p ΨL (p)
(171)
− 6 p ΨL (p) = α p ΨR (p) ,

58
thus requiring α = 1/β for consistency. Without loss of generality we can choose
α = 1 = β, so that
6 p ΨR (p) = p ΨL (p) . (172)
Using ΨL,R (p) = Ψ0L,R (p)/fL,R (p, x5 = 0), which follows from eq.(169) upon defining
Ψ0L,R (p) ≡ ΨL,R (p, x5 = 0), we finally get

fR (p, x5 = 0) 0
6 p Ψ0R (p) = p Ψ (p) , (173)
fL (p, x5 = 0) L

where fL,R satisfy

−∂52 + m2 fL,R (p, x5 ) = p2 fL,R (p, x5 ) .


(174)
Depending on whether ΨL (hence fL ) satisfies Neumann or Dirichlet boundary condi-
tions at x5 = L, there are two possible solutions to the above equations (respectively
dubbed as L+ and L− ):

fR (p, x5 ) = sin ω(x5 − L)

 

L+ : 5 1  5
  5
 (175)
fL (p, x ) = ω cos ω(x − L) − m sin ω(x − L)
p

fL (p, x5 ) = sin ω(x5 − L)

 

L− : 1  (176)
fR (p, x5 ) = − ω cos ω(x5 − L) + m sin ω(x5 − L) ,
  
p
p
where we have defined ω ≡ p2 − m2 . Evaluating the 5D action on the above solutions
leads to the boundary action. Since the 5D part of the action vanishes once evaluated on
the equations of motions, the only contribution to Sef f (Ψ0L ) comes from the boundary
term S0 . It is useful to define
(+)
1 fR (p, x5 = 0) 1 1
Σ(+) (p) = = −
pL fL(−) (p, x5 = 0) L m + ω cot(ωL)
(177)
(−)
1 fR (p, x5 = 0) 1
Σ(−) (p) =


= ω cot(ωL) − m
pL fL(+) (p, x5 = 0) p2 L

where the superscripts (+), (−) refer on whether the corresponding left-handed field
satisfies Neumann or Dirichlet conditions at x5 = L. From eqs.(169,173,175,176) and
eq.(165) one then obtains the following boundary action:
!
0 (1)
L
Z   Ψ
2
d4 x Ψ̄0L(1) Ψ̄0L(2) 6 p K(p) L
0 (2) , (178)
gΨ ΨL

59
where
Σ(+) (p)
 
0
K(p) = . (179)
0 Σ(−) (p)
Until this point we have proceeded as if A5 was vanishing. To reintroduce its
dependence back into the action we just have to adopt the following new boundary
conditions at x5 = L:
0 (i) 0 (j)
ΨL → [exp(−iθ(x))]ij ΨL . (180)
According to its definition (150), the expression of the Wilson line, in the case of our
SO(2) model, is
   
−iθ(x) −iAh(x)/f cos(h/f ) sin(h/f ) 0 −1
e =e = , A= , (181)
− sin(h/f ) cos(h/f ) +1 0

where A is the generator of SO(2) rotations, and we have knowingly defined

(0) 1
h(x) ≡ A5 (x) , f≡ √ . (182)
g5 L

After performing the redefinition of eq.(180), one obtains a boundary action of the
same form of eq.(178), but where now (∆Σ(p) ≡ Σ(+) (p) − Σ(−) (p))
" #
Σ(+) − sin2 (h/f ) ∆Σ sin(h/f ) cos(h/f ) ∆Σ
K(p) = . (183)
sin(h/f ) cos(h/f ) ∆Σ Σ(−) + sin2 (h/f ) ∆Σ

0 (2)
The last step before obtaining the final expression of Sef f is setting ΨL = 0. This
(2)
is required to reproduce the Dirichlet condition of ΨL at x5 = 0 in eq.(162), since
5D fields which vanish on the holographic boundary do not have any corresponding
elementary field. In other words, the 4D elementary sector consists of one single left-
0 (1) 0 (2)
handed field, ΨL . Its effective action is given by eqs.(178) and (183) with ΨL set
to zero:
 
L
2h
Z
4 0 (1) (+) (+) (−) 0 (1)
Sef f = 2 d x Ψ̄L 6 p Σ (p) − Σ (p) − Σ (p) sin ΨL . (184)
gΨ f

It closely resembles the expression of the effective action of the composite Higgs,
eq.(119), and in fact it is has exactly the form that one would have obtained, fol-
lowing the procedure of section 3, in the case of a composite Higgs theory where the
strong dynamics has an SO(2) global symmetry fully broken down in the infrared, and
an elementary sector consisting of one left-handed fermion.
It is at this point clear that h (hence A5 ) can be fully considered as a composite
scalar from the point of view of a low-energy 4D observer. Compositeness, indeed, is
experienced (and can be thus defined) as a non-trivial dependence of the couplings on
the 4D momentum. In the case of the holographic scalar h, such momentum dependence

60
is a consequence of the 5-dimensional profile of the corresponding bulk field. Also, h
is truly a NG boson with decay constant f , as the periodic dependence of Sef f upon
h/f testifies. Its Coleman-Weinberg potential can be easily derived starting from the
effective action (184): the series of 1-loop diagrams to resum is of the form showed
in the upper row of Fig. 12. After rotating to Euclidean momenta, Q2 = −p2 , one
obtains: Z 4
Σ(+) (Q) − Σ(−) (Q)
 
dQ 2 h
V (h) = −2 log 1 − sin . (185)
(2π)4 Σ(+) (Q) f
Similarly to the case of 4D composite Higgs theories, the convergence of the integral is
related to the behavior of the form factor (Σ(+) − Σ(−) ) at large virtual momenta. In
the Euclidean one has
Σ(+) (Q) − Σ(−) (Q) 1
= 1 − 2 ωE2 coth2 (ωE L) − m2 ,


(+)
(186)
Σ (Q) Q
p
where ωE ≡ Q2 + m2 . At large momenta, QL, Q/m  1, one has

Σ(+) (Q) − Σ(−) (Q) m2 −2QL

 
= −4 1 + 2 e + ··· (187)
Σ(+) (Q) Q

This means that the 5D bulk dynamics leads to an exponential convergence of the
integral. According to the discussion at the end of section 3.1, this is equivalent to
saying, in the 4D language, that the operator responsible for the symmetry breaking in
the infrared (i.e. the order parameter of the spontaneous breaking) has infinite dimen-
sion. A good approximation of the potential thus comes by expanding the logarithm
at leading order, so that:
1 1 h
V (h) ' − 2 4
f (m) sin2 ,
4π L f
(188)
Z ∞
dyE yE3 coth2 (yE ) − 1 .


f (m) ≡
m

For example f (0) = 3ζ(3)/2 ' 1.8.

All the above results and considerations clarify the meaning and the utility of the
holographic description of the 5D theory. We have already said that, by virtue of
the holographic interpretation, the 5D theory gives a model for the strong dynamics
of composite Higgs theories. Here we want to stress that such model of the strong
dynamics is especially interesting because it admits a perturbative expansion, thus
allowing one to compute a large class of infrared-saturated quantities. Calculability in
the 5D theory requires that the 5D expansion parameter be small

g52 L−1
 1.
16π 2

61
The corresponding perturbative parameter in the 4D picture is the number of ‘colors’
of the strong sector, which can thus be defined as:

1 g 2 L−1
≡ 5 2 . (189)
N 16π
This is completely consistent with the NDA expectation

N f2
≈ ,
16π 2 m2ρ

where now f is defined by eq.(182), and mρ ∼ 1/L sets the scale of the lightest
resonances of the strong sector, whose spectrum is given by the poles of K(p) in
eq.(183), see for example [68]. As already mentioned, the KK modes are to be identified
with the mass eigenstates obtained from the mixing of elementary and composite states
(mKK ∼ mρ ∼ 1/L). They are thus partial composites in the sense of section 3.3.

5 Epilogue
In these lectures I tried to give an overview of the basic mechanisms behind the idea of
composite Higgs and of the central qualitative features a model have to incorporate to
be compatible with the present experimental data. Mastering the general mechanisms
should make the reader well equipped to go through the vast literature on the subject
and build her/his own model. Several realistic constructions have been proposed so far
whose phenomenological implications will soon be tested at the LHC. These include
SO(5) ‘minimal’ models in 5D warped [34, 35,44, 79, 80,81] and flat [82] spacetimes, as
well as modern 4D composite Higgs models [83]. Although I cannot discuss the detailed
predictions of each of these models, I would like to conclude by spending a few words
on some general aspects of the phenomenology of a composite Higgs at present and
future colliders. (Ref.[84] is an excellent place to start to learn more on this subject,
see also [85,12].)
If a light Higgs boson is discovered at the LHC or at Tevatron, the most important
questions to address will be: what is its role in the mechanism of electroweak symmetry
breaking ? Is it an elementary or a composite scalar ? Crucial evidence will come from
a precise measurement of the parameters a, b, c in the effective Lagrangian (16): any
deviation from the unitary point a = b = c = 1 will be the sign of a departure from
the simple SM description and will give hints on the nature of the symmetry breaking
dynamics. A first determination of a and c will come from the measurement of the
couplings of the Higgs to the SM fermions and vectors. This requires disentangling
possible modifications of both the Higgs production cross sections and decay rates.
Preliminary studies have shown that the LHC should be eventually able to extract
the individual Higgs couplings with a ∼ 10 − 20% precision [86], though much will
depend on the value of its mass. This would imply a sensitivity on (1 − a2 ) up to

62
0.1 − 0.2 [84]. While the determination of the Higgs couplings will give a first hint on
its nature, a more direct probe of the symmetry-breaking dynamics will come only from
a precise study of the scattering processes that the exchange of the Higgs is assumed
to unitarize. A smoking gun of the compositeness of the light Higgs would be finding
an excess of events in VL VL → VL VL at the LHC compared to the SM expectation.
Another important though difficult process to monitor is VL VL → hh [12].
Determining a, b, c gives direct information on the symmetry breaking structure
associated to the light composite Higgs. Indeed, while these three parameters are
independent for a generic composite scalar, we have seen that they are related to each
other in specific models where the Higgs is a pseudo √ Nambu-Goldstone boson. For
instance, the SO(5)/SO(4) coset implies a = 1 − ξ, b = 1 − 2ξ, while the value
of c depends on how the composite operators coupled to the SM fermions transform
under SO(5).
√ We have seen that in the case of spinorial representations of SO(5) one
has c = 1 − ξ. Different curves in the (a, b) plane are thus associated to different
symmetry-breaking cosets. It has been also shown by the authors of Ref. [84] that if
the light composite Higgs belongs to an SU (2)L doublet, regardless of whether it has a
NG interpretation, the parameters a and b follow a universal trajectory in the vicinity
of the unitary point (i.e. for small ξ = v 2 /f 2 ): a ' 1 − ξ/2, b ' 1 − 2ξ. Any deviation
from this curve would be the signal of a different origin for the light scalar h. It is
possible, for example, that a light dilaton arises from the spontaneous breaking of the
scale invariance of the strong sector [87]. In that case conformal invariance requires
a2 = b = c2 , with the Lagrangian (16) exactly truncated at quadratic order in h. For
this choice one can define the dilaton decay constant by v/a ≡ fD , and the dilaton
field as exp(φ(x)/fD ) = 1 + h(x)/fD .
Along with the study of the phenomenology of a light composite Higgs, crucial
information on the symmetry breaking dynamics will come from the production of res-
onances of the strong sector, in particular the fermionic resonances coupled to the top
quark. Extracting their masses and couplings by measuring their production cross sec-
tion and decay fractions will give the unique opportunity to understand the mechanism
by which the Higgs is light and identify the global symmetries of the strong dynamics.

Acknowledgments
I would like to thank Csaba Csaki, Scott Dodelson and K.T. Mahanthappa for inviting
me to TASI 2009, “Physics of the Large and the Small”, to give these lectures. I thank
all the students for the great atmosphere and for their many stimulating questions.
I am also grateful to Andrea Wulzer for discussions and to David Marzocca, Marco
Matassa and Natascia Vignaroli for reading the manuscript and pointing out several
errors.

63
Appendix
We collect here some useful group theory results and formulas.
The isomorphism between SO(4) and SU (2)L ×SU (2)R mentioned at the beginning
of section 3.1 can be shown by associating to any 4-dimensional vector v â a matrix
V ≡ σ â v â (σ â ≡ (~σ , −i1), with â = 1, 2, 3, 4). The group SO(4) acts on the vector v
as a rotation, preserving its norm:
SO(4) : v â → S âb̂ v b̂ , |v| = constant . (190)
The action of SU (2)L × SU (2)R can then be defined on the matrix V as the left
multiplication by L ∈ SU (2)L and right multiplication by R ∈ SU (2)R , so that the
determinant of V is unchanged:
SU (2)L × SU (2)R : V → L V R† , det(V ) = −|v|2 = constant . (191)
Then, for each SO(4) matrix S there are two SU (2)L × SU (2)R transformations that
act in the same way on V ,
S → { (L, R) , (−L, −R) } , (192)
and that differ by a sign. At the level of group elements such correspondence implies
the following exact equivalence:
SU (2)L × SU (2)R
SO(4) = . (193)
Z2
A suitable basis of SO(5) generators for the fundamental representation is the
following
 
aL,R i 1 abc b c b c

a 4 a 4


Tij = −  δi δj − δj δi ± δi δj − δj δi
2 2
(194)
â i â 5 â 5


Tij = − √ δi δj − δj δi ,
2
where i, j = 1, . . . , 5 and T aL,R (aL,R = 1, 2, 3) are the generators of SO(4) ∼ SU (2)L ×
SU (2)R . The spinorial representation of SO(5) can be defined in terms of the Gamma
matrices
   
â 0 σ â 5 1 0
Γ = â † , Γ = , σ â = {~σ , −i1} , (195)
σ 0 0 −1
as  
i 1 abc b c 4 i
T aL,R
=− √ a
 [Γ , Γ ] ± [Γ , Γ ] , T â = − √ [Γâ , Γ5 ] , (196)
2 2 2 4 2
so that
     
aL 1 σa 0 aR 1 0 0 âi 0 σ â
T = , T = , T = √ â † . (197)
2 0 0 2 0 σa 2 2 −σ 0

64
References
[1] T. Gherghetta, arXiv:hep-ph/0601213.

[2] M. Schmaltz and D. Tucker-Smith, Ann. Rev. Nucl. Part. Sci. 55, 229 (2005)
[arXiv:hep-ph/0502182].

[3] R. Sundrum, arXiv:hep-th/0508134.

[4] M. Serone, arXiv:0909.5619 [hep-ph].

[5] H. C. Cheng, arXiv:1003.1162 [hep-ph].

[6] C. Csaki, arXiv:hep-ph/0404096.

[7] G. D. Kribs, arXiv:hep-ph/0605325.

[8] R. Rattazzi, arXiv:hep-ph/0607055.

[9] C. Csaki, J. Hubisz and P. Meade, arXiv:hep-ph/0510275.

[10] B. W. Lee, C. Quigg and H. B. Thacker, Phys. Rev. D 16, 1519 (1977).

[11] J. M. Cornwall, D. N. Levin and G. Tiktopoulos, Phys. Rev. D 10, 1145 (1974)
[Erratum-ibid. D 11, 972 (1975)]; C. E. Vayonakis, Lett. Nuovo Cim. 17, 383
(1976); M. S. Chanowitz and M. K. Gaillard, Nucl. Phys. B 261, 379 (1985).

[12] R. Contino, C. Grojean, M. Moretti, F. Piccinini and R. Rattazzi, arXiv:1002.1011

[hep-ph].

[13] A. Collaboration et al., arXiv:0911.2604 [hep-ex].

[14] S. Weinberg, Phys. Rev. D 13, 974 (1976); Phys. Rev. D 19, 1277 (1979);
L. Susskind, Phys. Rev. D 20, 2619 (1979).
For introductory reviews on Technicolor, see for example:
C. T. Hill and E. H. Simmons, Phys. Rept. 381, 235 (2003) [Erratum-ibid. 390,
553 (2004)] [arXiv:hep-ph/0203079];
K. Lane, Lectures at Ecole de GIF at LAPP, Annecy-le-Vieux, France, 2001,
arXiv:hep-ph/0202255;
R. S. Chivukula, Lectures at TASI 2000, arXiv:hep-ph/0011264.

[15] G. ’t Hooft, Nucl. Phys. B 72, 461 (1974); B 75, 461 (1974).

[16] E. Witten, Nucl. Phys. B 160, 57 (1979).

[17] M. E. Peskin and T. Takeuchi, Phys. Rev. D 46, 381 (1992).

[18] H. Georgi, Nucl. Phys. B 363, 301 (1991).

65
[19] R. Barbieri, A. Pomarol, R. Rattazzi and A. Strumia, Nucl. Phys. B 703, 127
(2004) [arXiv:hep-ph/0405040].

[20] A. Manohar and H. Georgi, Nucl. Phys. B 234, 189 (1984); H. Georgi and L. Ran-
dall, Nucl. Phys. B 276, 241 (1986).

[21] E. Eichten and K. D. Lane, Phys. Lett. B 90, 125 (1980); S. Dimopoulos and
L. Susskind, Nucl. Phys. B 155, 237 (1979).

[22] See, for example, M. Bona et al. [UTfit Collaboration], JHEP 0803, 049 (2008)
[arXiv:0707.0636 [hep-ph]];

[23] B. Holdom, Phys. Rev. D 24, 1441 (1981); Phys. Lett. B 150, 301 (1985); K. Ya-
mawaki, M. Bando and K. i. Matumoto, Phys. Rev. Lett. 56, 1335 (1986); T. Ak-
iba and T. Yanagida, Phys. Lett. B 169, 432 (1986); T. W. Appelquist, D. Kara-
bali and L. C. R. Wijewardhana, Phys. Rev. Lett. 57, 957 (1986); T. Appelquist
and L. C. R. Wijewardhana, Phys. Rev. D 35, 774 (1987); Phys. Rev. D 36, 568
(1987).

[24] M. A. Luty and T. Okui, JHEP 0609, 070 (2006) [arXiv:hep-ph/0409274].

[25] R. Rattazzi, V. S. Rychkov, E. Tonni and A. Vichi, JHEP 0812, 031 (2008)
[arXiv:0807.0004 [hep-th]].

[26] V. S. Rychkov and A. Vichi, Phys. Rev. D 80, 045006 (2009) [arXiv:0905.2211
[hep-th]].

[27] D. B. Kaplan and H. Georgi, Phys. Lett. B 136, 183 (1984).

[28] D. B. Kaplan, H. Georgi and S. Dimopoulos, Phys. Lett. B 136, 187 (1984).

[29] T. Banks, Nucl. Phys. B 243, 125 (1984).

[30] H. Georgi, D. B. Kaplan and P. Galison, Phys. Lett. B 143, 152 (1984).

[31] H. Georgi and D. B. Kaplan, Phys. Lett. B 145, 216 (1984).

[32] M. J. Dugan, H. Georgi and D. B. Kaplan, Nucl. Phys. B 254, 299 (1985).

[33] S. Dimopoulos and J. Preskill, Nucl. Phys. B 199, 206 (1982).

[34] K. Agashe, R. Contino and A. Pomarol, Nucl. Phys. B 719, 165 (2005) [arXiv:hep-
ph/0412089].

[35] R. Contino, L. Da Rold and A. Pomarol, Phys. Rev. D 75, 055014 (2007)
[arXiv:hep-ph/0612048].

66
[36] B. Gripaios, A. Pomarol, F. Riva and J. Serra, JHEP 0904, 070 (2009)
[arXiv:0902.1483 [hep-ph]].

[37] E. de Rafael, arXiv:hep-ph/9802448.

[38] M. A. Shifman, A. I. Vainshtein and V. I. Zakharov, Nucl. Phys. B 147, 385

(1979); Nucl. Phys. B 147, 448 (1979).

[39] M. Knecht and E. de Rafael, Phys. Lett. B 424, 335 (1998) [arXiv:hep-
ph/9712457].

[40] S. Weinberg, Phys. Rev. Lett. 18, 507 (1967).

[41] E. Witten, Phys. Rev. Lett. 51, 2351 (1983).

[42] T. Das, G. S. Guralnik, V. S. Mathur, F. E. Low and J. E. Young, Phys. Rev.

Lett. 18, 759 (1967).

[43] R. Contino, Y. Nomura and A. Pomarol, Nucl. Phys. B 671, 148 (2003) [arXiv:hep-
ph/0306259].

[44] M. S. Carena, E. Ponton, J. Santiago and C. E. M. Wagner, Nucl. Phys. B 759,

202 (2006) [arXiv:hep-ph/0607106]; Phys. Rev. D 76, 035006 (2007) [arXiv:hep-
ph/0701055].

[45] R. Barbieri, B. Bellazzini, V. S. Rychkov and A. Varagnolo, Phys. Rev. D 76,

115008 (2007) [arXiv:0706.0432 [hep-ph]].

[46] P. Lodone, JHEP 0812, 029 (2008) [arXiv:0806.1472 [hep-ph]]; M. Gillioz, Phys.
Rev. D 80, 055003 (2009) [arXiv:0806.3450 [hep-ph]]; C. Anastasiou, E. Furlan
and J. Santiago, Phys. Rev. D 79, 075003 (2009) [arXiv:0901.2117 [hep-ph]]. See
also: R. Barbieri, G. Isidori, V. S. Rychkov and E. Trincherini, Phys. Rev. D 78,
036012 (2008) [arXiv:0806.1624 [hep-ph]].

[47] G. Altarelli and R. Barbieri, Phys. Lett. B 253, 161 (1991).

[48] G. Altarelli, R. Barbieri and S. Jadach, Nucl. Phys. B 369, 3 (1992) [Erratum-ibid.
B 376, 444 (1992)].

[49] Alessandro Strumia, private communication. The data used in the fit are those of
LEP1 (see Table 2 of Ref. [19]), and those from Atomic Parity Violation (APV)
(see Ref. [19], Table 3). The NuTeV data have not been included, and in any case
their inclusion would not significantly change the results of the fit.

[50] K. Agashe, S. Gopalakrishna, T. Han, G. Y. Huang and A. Soni, Phys. Rev. D

80, 075007 (2009) [arXiv:0810.1497 [hep-ph]]; K. Agashe et al., Phys. Rev. D 76,
115015 (2007) [arXiv:0709.0007 [hep-ph]].

67
[51] Y. Grossman and M. Neubert, Phys. Lett. B 474, 361 (2000) [arXiv:hep-
ph/9912408].

[52] T. Gherghetta and A. Pomarol, Nucl. Phys. B 586, 141 (2000) [arXiv:hep-
ph/0003129].

[53] S. J. Huber and Q. Shafi, Phys. Lett. B 498, 256 (2001) [arXiv:hep-ph/0010195].

[54] S. J. Huber, Nucl. Phys. B 666, 269 (2003) [arXiv:hep-ph/0303183].

[55] Phys. Rev. D 71, 016002 (2005) [arXiv:hep-ph/0408134].

[56] D. B. Kaplan, Nucl. Phys. B 365, 259 (1991).

[57] K. Agashe, A. Delgado and R. Sundrum, Annals Phys. 304, 145 (2003) [arXiv:hep-
ph/0212028].

[58] K. Agashe, G. Perez and A. Soni, Phys. Rev. Lett. 93, 201804 (2004) [arXiv:hep-
ph/0406101].

[59] C. Csaki, A. Falkowski and A. Weiler, JHEP 0809, 008 (2008) [arXiv:0804.1954
[hep-ph]].

[60] M. Blanke, A. J. Buras, B. Duling, S. Gori and A. Weiler, JHEP 0903, 001 (2009)
[arXiv:0809.1073 [hep-ph]].

[61] K. Agashe, A. Azatov and L. Zhu, Phys. Rev. D 79, 056006 (2009)
[arXiv:0810.1016 [hep-ph]].

[62] O. Gedalia, G. Isidori and G. Perez, Phys. Lett. B 682, 200 (2009)
[arXiv:0905.3264 [hep-ph]].

[63] K. Agashe, A. E. Blechman and F. Petriello, Phys. Rev. D 74, 053011 (2006)
[arXiv:hep-ph/0606021].

[64] K. Agashe, Phys. Rev. D 80, 115020 (2009) [arXiv:0902.2400 [hep-ph]].

[65] R. Contino, T. Kramer, M. Son and R. Sundrum, JHEP 0705, 074 (2007)
[arXiv:hep-ph/0612180].

[66] R. Barbieri, R. Contino, P. Creminelli, R. Rattazzi and C. A. Scrucca, Phys. Rev.

D 66, 024025 (2002) [arXiv:hep-th/0203039].

[67] C. Csaki, C. Grojean and H. Murayama, Phys. Rev. D 67, 085012 (2003)
[arXiv:hep-ph/0210133].

[68] R. Contino and A. Pomarol, JHEP 0411, 058 (2004) [arXiv:hep-th/0406257].

68
[69] J. M. Maldacena, Adv. Theor. Math. Phys. 2, 231 (1998) [Int. J. Theor. Phys.
38, 1113 (1999)] [arXiv:hep-th/9711200]; S. S. Gubser, I. R. Klebanov and
A. M. Polyakov, Phys. Lett. B 428, 105 (1998) [arXiv:hep-th/9802109]; E. Witten,
Adv. Theor. Math. Phys. 2, 253 (1998) [arXiv:hep-th/9802150].

[70] N. Arkani-Hamed, M. Porrati and L. Randall, JHEP 0108, 017 (2001) [arXiv:hep-
th/0012148].

[71] R. Rattazzi and A. Zaffaroni, JHEP 0104, 021 (2001) [arXiv:hep-th/0012248].

[72] M. Perez-Victoria, JHEP 0105, 064 (2001) [arXiv:hep-th/0105048].

[73] M. A. Luty, M. Porrati and R. Rattazzi, JHEP 0309, 029 (2003) [arXiv:hep-
th/0303116].

[74] R. Barbieri, A. Pomarol and R. Rattazzi, Phys. Lett. B 591, 141 (2004) [arXiv:hep-
ph/0310285].

[75] G. Panico and A. Wulzer, JHEP 0705, 060 (2007) [arXiv:hep-th/0703287].

[76] S. R. Coleman, J. Wess and B. Zumino, Phys. Rev. 177, 2239 (1969). C. G. Callan,
S. R. Coleman, J. Wess and B. Zumino, Phys. Rev. 177, 2247 (1969).

[77] L. Randall and R. Sundrum, Phys. Rev. Lett. 83, 3370 (1999) [arXiv:hep-
ph/9905221]; Phys. Rev. Lett. 83, 4690 (1999) [arXiv:hep-th/9906064].

[78] R. Contino and A. Gambassi, J. Math. Phys. 44, 570 (2003) [arXiv:hep-
th/0112161].

[79] A. D. Medina, N. R. Shah and C. E. M. Wagner, Phys. Rev. D 76, 095010 (2007)
[arXiv:0706.1281 [hep-ph]].

[80] M. Carena, A. D. Medina, N. R. Shah and C. E. M. Wagner, Phys. Rev. D 79,

096010 (2009) [arXiv:0901.0609 [hep-ph]].

[81] G. Panico, E. Ponton, J. Santiago and M. Serone, Phys. Rev. D 77, 115012 (2008)
[arXiv:0801.1645 [hep-ph]].

[82] G. Panico, M. Serone and A. Wulzer, Nucl. Phys. B 739, 186 (2006) [arXiv:hep-
ph/0510373]. G. Panico, M. Serone and A. Wulzer, Nucl. Phys. B 762, 189 (2007)
[arXiv:hep-ph/0605292].

[83] J. A. Evans, J. Galloway, M. A. Luty and R. A. Tacchi, arXiv:1001.1361 [hep-ph].

[84] G. F. Giudice, C. Grojean, A. Pomarol and R. Rattazzi, JHEP 0706, 045 (2007)
[arXiv:hep-ph/0703164].

[85] J. R. Espinosa, C. Grojean and M. Muhlleitner, arXiv:1003.3251 [hep-ph].

69
[86] See for example: M. Duhrssen, S. Heinemeyer, H. Logan, D. Rainwater, G. Wei-
glein and D. Zeppenfeld, Phys. Rev. D 70, 113009 (2004) [arXiv:hep-ph/0406323].

[87] W. D. Goldberger, B. Grinstein and W. Skiba, Phys. Rev. Lett. 100, 111802 (2008)
[arXiv:0708.1463 [hep-ph]]; J. Fan, W. D. Goldberger, A. Ross and W. Skiba, Phys.
Rev. D 79, 035017 (2009) [arXiv:0803.2040 [hep-ph]]; L. Vecchi, arXiv:1002.1721
[hep-ph].

70