Lectures on Supersymmetry

Matteo Bertolini
SISSA

March 12, 2023

1
 Foreword

This is a write-up of a course on Supersymmetry I have been giving for several

years to PhD students attending the curriculum in Theoretical Particle Physics at
SISSA, the International School for Advanced Studies of Trieste.
There are several excellent books on supersymmetry and many very good lecture
courses are available on the archive. The ambition of this set of notes is not to add
anything new in this respect, but to offer a set of hopefully complete and self-
consistent lectures, which start from the basics and arrive to some of the more
recent and advanced topics. The price to pay is that the material is pretty huge.
The advantage is to have all such material in a single, possibly coherent file, and
that no prior exposure to supersymmetry is required.
There are many topics I do not address and others I only briefly touch. In
particular, I discuss only rigid supersymmetry (mostly focusing on four space-time
dimensions), while no reference to supergravity is given. Moreover, this is a the-
oretical course and phenomenological aspects are only briefly sketched. One only
chapter is dedicated to present basic phenomenological ideas, including a bird eyes
view on models of gravity and gauge mediation and their properties, but a thorough
discussion of phenomenological implications of supersymmetry would require much
more.
There is no bibliography at the end of the file. However, each chapter contains its
own bibliography where the basic references I used to prepare the material (mainly
books and/or reviews available on-line) are reported – including explicit indication
of the corresponding pages and chapters, so to let the reader have access to the
original font (and to let me give proper credit to authors).

I hope this effort can be of some help to as many students as possible!

2
Contents

1 Supersymmetry: a bird eyes view 7

1.1 What is supersymmetry? . . . . . . . . . . . . . . . . . . . . . . . . . 9
1.2 What is supersymmetry useful for? . . . . . . . . . . . . . . . . . . . 10
1.3 Some useful references . . . . . . . . . . . . . . . . . . . . . . . . . . 20

2 The supersymmetry algebra 24

2.1 Lorentz and Poincaré groups . . . . . . . . . . . . . . . . . . . . . . . 24
2.2 Spinors and representations of the Lorentz group . . . . . . . . . . . 28
2.3 The supersymmetry algebra . . . . . . . . . . . . . . . . . . . . . . . 31
2.4 Exercises . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 39

3 Representations of the supersymmetry algebra 41

3.1 Massless supermultiplets . . . . . . . . . . . . . . . . . . . . . . . . . 43
3.2 Massive supermultiplets . . . . . . . . . . . . . . . . . . . . . . . . . 51
3.3 Representation on fields: a first try . . . . . . . . . . . . . . . . . . . 56
3.4 Exercises . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 60

4 Superspace and superfields 62

4.1 Superspace as a coset . . . . . . . . . . . . . . . . . . . . . . . . . . . 62
4.2 Superfields as fields in superspace . . . . . . . . . . . . . . . . . . . . 65
4.3 Supersymmetric invariant actions - general philosophy . . . . . . . . . 69
4.4 Chiral superfields . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 70
4.5 Real (aka vector) superfields . . . . . . . . . . . . . . . . . . . . . . . 73
4.6 (Super)Current superfields . . . . . . . . . . . . . . . . . . . . . . . . 75
4.6.1 Internal symmetry current superfields . . . . . . . . . . . . . . 76
4.6.2 Supercurrent superfields . . . . . . . . . . . . . . . . . . . . . 77
4.7 Exercises . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 80

5 Supersymmetric actions: minimal supersymmetry 81

5.1 N = 1 Matter actions . . . . . . . . . . . . . . . . . . . . . . . . . . 81
5.1.1 Non-linear sigma model I . . . . . . . . . . . . . . . . . . . . . 88

3
 5.2 N = 1 SuperYang-Mills . . . . . . . . . . . . . . . . . . . . . . . . . 92
5.3 N = 1 Gauge-matter actions . . . . . . . . . . . . . . . . . . . . . . . 96
5.3.1 Classical moduli space: examples . . . . . . . . . . . . . . . . 101
5.3.2 Non-linear sigma model II . . . . . . . . . . . . . . . . . . . . 108
5.4 Exercises . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 110

6 Supersymmetric actions: extended supersymmetry 113

6.1 N = 2 supersymmetric actions . . . . . . . . . . . . . . . . . . . . . . 113
6.1.1 Non-linear sigma model III . . . . . . . . . . . . . . . . . . . . 116
6.2 N = 4 supersymmetric actions . . . . . . . . . . . . . . . . . . . . . . 118
6.3 On non-renormalization theorems . . . . . . . . . . . . . . . . . . . . 120
6.4 Exercises . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 127

7 Supersymmetry breaking 129

7.1 Vacua in supersymmetric theories . . . . . . . . . . . . . . . . . . . . 129
7.2 Goldstone theorem and the goldstino . . . . . . . . . . . . . . . . . . 132
7.3 F-term breaking . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 133
7.4 Pseudomoduli space: quantum corrections . . . . . . . . . . . . . . . 145
7.5 D-term breaking . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 149
7.6 Indirect criteria for supersymmetry breaking . . . . . . . . . . . . . . 152
7.6.1 Supersymmetry breaking and global symmetries . . . . . . . . 152
7.6.2 Topological constraints: the Witten Index . . . . . . . . . . . 155
7.6.3 Genericity and metastability . . . . . . . . . . . . . . . . . . . 161
7.7 Exercises . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 162

8 Supersymmetry breaking and the Standard Model 164

8.1 Towards dynamical supersymmetry breaking . . . . . . . . . . . . . . 165
8.2 The Supertrace mass formula . . . . . . . . . . . . . . . . . . . . . . 166
8.3 Beyond the MSSM . . . . . . . . . . . . . . . . . . . . . . . . . . . . 169
8.4 Spurions, soft terms and the messenger paradigm . . . . . . . . . . . 170
8.5 Mediating the breaking . . . . . . . . . . . . . . . . . . . . . . . . . . 173
8.5.1 Gravity mediation . . . . . . . . . . . . . . . . . . . . . . . . 174

4
 8.5.2 Gauge mediation . . . . . . . . . . . . . . . . . . . . . . . . . 176
8.6 Exercises . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 183

9 Non-perturbative effects and holomorphy 184

9.1 Instantons in a nutshell . . . . . . . . . . . . . . . . . . . . . . . . . . 184
9.2 Anomalies in a nutshell . . . . . . . . . . . . . . . . . . . . . . . . . . 187
9.3 ’t Hooft anomaly matching condition . . . . . . . . . . . . . . . . . . 191
9.4 Holomorphy . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 192
9.5 Holomorphy and non-renormalization theorems . . . . . . . . . . . . 195
9.6 Holomorphic decoupling . . . . . . . . . . . . . . . . . . . . . . . . . 204
9.7 Exercises . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 207

10 Supersymmetric gauge dynamics: N = 1 209

10.1 Confinement and mass gap in QCD, YM and SYM . . . . . . . . . . 209
10.1.1 Intermezzo: Wilson loops as order parameters for confinement 219
10.2 Phases of gauge theories: examples . . . . . . . . . . . . . . . . . . . 222
10.2.1 Coulomb phase and free phase . . . . . . . . . . . . . . . . . . 222
10.2.2 Continuously connected phases . . . . . . . . . . . . . . . . . 224
10.3 N=1 SQCD: perturbative analysis . . . . . . . . . . . . . . . . . . . . 225
10.4 N=1 SQCD: non-perturbative dynamics . . . . . . . . . . . . . . . . 227
10.4.1 Pure SYM: gaugino condensation . . . . . . . . . . . . . . . . 228
10.4.2 F < N : the ADS superpotential . . . . . . . . . . . . . . . . . 229
10.4.3 Integrating in and out: the linearity principle . . . . . . . . . 235
10.4.4 F ≥ N : quantum moduli space . . . . . . . . . . . . . . . . . 240
10.4.5 Conformal window . . . . . . . . . . . . . . . . . . . . . . . . 247
10.4.6 Electric-magnetic duality (aka Seiberg duality) . . . . . . . . . 249
10.5 The phase diagram of N=1 SQCD . . . . . . . . . . . . . . . . . . . . 258
10.6 Exercises . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 260

11 Dynamical supersymmetry breaking 262

11.1 Calculable and non-calculable models: generalities . . . . . . . . . . . 262
11.2 The one GUT family SU(5) model . . . . . . . . . . . . . . . . . . . . 265

5
 11.3 The 3-2 model: instanton driven SUSY breaking . . . . . . . . . . . . 267
11.4 The 4-1 model: gaugino condensation driven SUSY breaking . . . . . 273
11.5 The ITIY model: SUSY breaking with classical flat directions . . . . 275
11.6 DSB into metastable vacua. A case study: massive SQCD . . . . . . 278
11.6.1 Summary of basic results . . . . . . . . . . . . . . . . . . . . . 278
11.6.2 Massive SQCD in the free magnetic phase: electric description 280
11.6.3 Massive SQCD in the free magnetic phase: magnetic description281
11.6.4 Summary of the physical picture . . . . . . . . . . . . . . . . . 289

12 Supersymmetric gauge dynamics: extended supersymmetry 292

12.1 Low energy effective actions: classical and quantum . . . . . . . . . . 292
12.1.1 N = 2 effective actions . . . . . . . . . . . . . . . . . . . . . . 294
12.1.2 N = 4 effective actions . . . . . . . . . . . . . . . . . . . . . . 301
12.2 Monopoles, dyons and electromagnetic duality: a recap . . . . . . . . 301
12.3 Seiberg-Witten theory . . . . . . . . . . . . . . . . . . . . . . . . . . 311
12.3.1 N = 2 SU (2) pure SYM . . . . . . . . . . . . . . . . . . . . . 312
12.3.2 Intermezzo: confinement by monopole condensation . . . . . . 324
12.4 Seiberg-Witten theory: generalizations . . . . . . . . . . . . . . . . . 327
12.4.1 A case study: N = 2 SU (2) SQCD with one flavor . . . . . . 330
12.5 N = 4: Montonen-Olive duality . . . . . . . . . . . . . . . . . . . . . 339
12.6 Exercises . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 342

6
1 Supersymmetry: a bird eyes view

Why should a theoretical particle physicist ever know about supersymmetry?

Why embark on reading a book as long as this? There are several reasons why it’s
worth it. Let me mention the ones I think are the most relevant.
The first such reasons, at least from a chronological point of view, is also the
more phenomenological in nature. Back in 2012 the missing building block of the
Standard Model, the Higgs particle, has been discovered at CERN’s Large Hadron
Collider (LHC). This has been an impressive achievement, one of the greatest suc-
cess of the way we think Nature works at short distances and of the tool we use
to describe it, i.e. Quantum Field Theory (QFT). On the other hand, there ex-
ist many reasons - some of which we will review in the following - which suggest
that this cannot be the end of the story: new physics should show-up at energy
scales higher than those we have been able to have access to, so far (but way lower
than, say, the Planck scale). It turns out that of all possible options, the most
compelling and motivated scenario for such beyond the Standard Model physics is
supersymmetry. So, when it comes to try and understand how particles behave at
high energy, equivalently at shorter and shorter distance, supersymmetry is a piece
of basic knowledge any particle physicist should have. It should be said that at
the time of writing no supersymmetric particles have been discovered yet, nor we
have any indirect evidence for their existence. While very few high energy physicists
doubt that supersymmetry is actually realized in Nature, this lack of experimental
signature is putting the very idea of low energy supersymmetry into question, sug-
gesting at least some twist in the way we think about it. Things might be slightly
more involved than we imagined, supersymmetric particles might not be around the
corner but actually a few more steps ahead, implying that the way supersymmetry
tackles the different phenomenological problems it is expected to solve, might be
more tricky than we thought. However, I do not think we are yet at a stage to
declare supersymmetry phenomenology dead, and I keep on thinking there is still
room for such a phenomenological motivation for supersymmetry.
An even more profound role supersymmetry is believed to play in the dynamics
and ultimate structure of space-time, in the way gravity behaves at very high energy,
as high as the Planck scale, via string theory. The latter is the more successful
framework to describe all interactions, including gravity, in a way consistent with
quantum mechanics. However, differently from an ordinary quantum field theory,
string theory is inherently supersymmetric. From this point of view, no matter the

7
scale at which it might show up, supersymmetry looks as a crucial ingredient in our
understanding of the ultimate laws of Nature.
Supersymmetry is also at the core of what is probably the more amazing and
far-reaching discovery in theoretical physics in the last few decades, the celebrated
AdS/CFT correspondence. In short, this correspondence predicts that a (non-
gravitational) QFT in d space-time dimensions can actually be dual to a theory
of quantum gravity in one dimension higher. This means that the two theories are
equivalent at the full quantum level and, upon using a proper dictionary, all ob-
servables agree. The best studied (and solid) examples of such remarkable duality
involve supersymmetric QFTs in d-dimensional Minkowski space and (super)string
theory in d + 1-dimensional anti-de Sitter space. This is why supersymmetric quan-
tum field theories have now also become a tool to study quantum gravity.
Supersymmetry turns out to be relevant also outside the realm of particle physics,
like in some condensed matter systems, and it has also be at the core of what is
probably the more amazing and far-reaching discovery in theoretical physics in the
last decades, namely the celebrated AdS/CFT correspondence.
One other thing we, theoretical physicists, want to understand is the behavior
of quantum field theories at strong coupling. This is a regime where usual pertur-
bative techniques fail and we lack analytical tools. However, many phenomena we
observe in Nature are described by the behavior of quantum field theories in such
a regime, the most notable example being the way phenomena like confinement,
dynamical mass generation and chiral symmetry breaking are realized in Nature.
One spectacular property of supersymmetry is that it makes these phenomena more
accessible: supersymmetric quantum field theories turn out to have a much more
constrained dynamics with respect to non-supersymmetric ones, so constrained that
it is often possible to understand their strong coupling regime analytically. In this
regard, supersymmetry is seen (and is being used) as a theoretical laboratory to
study quantum field theories at strong coupling and get some intuition on how phe-
nomena like those mentioned above are realized in non-supersymmetric field theories
(as QCD). Remarkably, several ideas that had been proposed to account for such
phenomena and which could only be conjectural as far as ordinary quantum field
theories, have been analytically proven in the supersymmetric context, notable ex-
amples being that confinement is due to monopole condensation, or that at strong
coupling fermion bilinears condense. From this point of view, even setting aside its
phenomenological or formal applications, supersymmetry is useful in that is a way in

8
which we can deepen our understanding of QFT in general, seeing all of its features
at work in a well-controlled setting.
I won’t be able to discuss all these aspects in detail. The aim of this course is
just to provide the minimum foundation you need to get into this fascinating subject
and to give you some taste of some advanced topics. What to do with it... will be
your choice.
In this first lecture I will give a brief overview on what is supersymmetry and why
it is interesting to study it. In the rest of the course I will try to provide (much) more
detailed answers to these two basic questions. I hope you will enjoy the journey!

1.1 What is supersymmetry?

Supersymmetry (SUSY) is a space-time symmetry mapping particles and fields of
integer spin (bosons) into particles and fields of half integer spin (fermions), and
viceversa. The generators Q act as

Q |F ermioni = |Bosoni and viceversa (1.1)

From its very definition, this operator has two obvious but far-reaching properties
that can be summarized as follows:

• It changes the spin of a particle (meaning that Q transforms as a spin-1/2

particle) and hence its space-time properties. This is why supersymmetry is
not an internal symmetry but a space-time symmetry.

• In a theory where supersymmetry is realized, each one-particle state has at

least a superpartner. Therefore, in a SUSY world, instead of single particle
states, one has to deal with (super)multiplets of particle states.

Supersymmetry generators have specific commutation properties with other gener-

ators. In particular:

• Q commutes with translations and internal quantum numbers (e.g. gauge and
global symmetries), but it does not commute with Lorentz generators

[Q, Pµ ] = 0 , [Q, G] = 0 , [Q, Mµν ] 6= 0 . (1.2)

This implies that particles belonging to the same supermultiplet have different
spin but same mass and same quantum numbers.

9
A supersymmetric field theory is a set of fields and a Lagrangian which exhibit such a
symmetry. As ordinary field theories, supersymmetric theories describe particles and
interactions between them: SUSY manifests itself in the specific particle spectrum
a theory enjoys and in the way particles interact between themselves.
A supersymmetric model which is covariant under general coordinate transfor-
mations is called supergravity (SUGRA) model. In this respect, a non-trivial fact,
which again comes from the algebra, in particular from the (anti)commutation re-
lation
{Q, Q} ∼ Pµ , (1.3)
is that having general coordinate transformations is equivalent to have local SUSY,
the gauge mediator being a spin 3/2 particle, the gravitino. Hence local supersym-
metry and General Relativity are intimately tied together.
One can have theories with different number of SUSY generators Q: QI I =
1, ..., N . The number of supersymmetry generators, however, cannot be arbitrarily
large. The reason is that any supermultiplet contains particles with spin at least as
large as 41 N . Therefore, N can be at most as large as 4 for theories with maximal
spin 1 (gauge theories) and as large as 8 for theories with maximal spin 2 (gravity).
Thus stated, this statement is true in four space-time dimensions. Equivalent state-
ments can be made in higher/lower dimensions, where the dimension of the spinor
representation of the Lorentz group is larger/smaller (for instance, in 10 dimen-
sions, which is the natural dimension where superstring theory lives, the maximum
allowed N is 2). What really matters is the number of single state supersymmetry
generators, which is a dimension-independent statement.
Finally, notice that since supersymmetric theories automatically accomodate
both bosons and fermions, SUSY looks like the most natural framework where to
formulate a theory able to describe matter and interactions in a unified way.

1.2 What is supersymmetry useful for?

Let us briefly outline a number of reasons why it might be meaningful (and useful)
to have such a bizarre and unconventional symmetry actually realized in Nature.

i. Theoretical reasons.

• What are the more general allowed symmetries of the S-matrix? In 1967 Cole-
man and Mandula proved a theorem which says that in a generic quantum

10
 field theory, under a number of (very reasonable and physical) assumptions,
like locality, causality, positivity of energy, finiteness of number of particles,
etc..., the only possible continuos symmetries of the S-matrix are those gener-
ated by Poincaré group generators, Pµ and Mµν , plus some internal symmetry
group generators G commuting with them

[G, Pµ ] = [G, Mµν ] = 0 , (1.4)

where the group G is a semi-simple group times abelian factors.

In other words, the most general symmetry group enjoyed by the S-matrix is

Poincaré × Internal Symmetries

The Coleman-Mandula theorem can be evaded by weakening one or more of

its assumptions. One such assumptions is that the symmetry algebra only in-
volves commutators, all generators being bosonic generators. This assumption
does not have any particular physical reason not to be relaxed. Allowing for
fermionic generators, which satisfy anti-commutation relations, it turns out
that the set of allowed symmetries can be enlarged. More specifically, in 1975
Haag, Lopuszanski and Sohnius showed that supersymmetry (which, as we will
see, is a very specific way to add fermionic generators to a symmetry algebra)
is the only possible such option. This makes the Poincaré group becoming Su-
perPoincaré. Therefore, the most general symmetry group the S-matrix can
enjoy turns out to be

SuperPoincaré × Internal Symmetries

From a purely theoretical view point, one could then well expect that Nature
might have realized all possible kind of allowed symmetries, given that we
already know this is indeed the case for all known symmetries, but supersym-
metry (i.e., the Standard Model.

• The history of our understanding of physical laws is an history of unifica-

tion. A famous example is Newton’s law of universal gravitation, which says
that one and the same equation describes the attraction a planet exert on
another planet and on... an apple! Maxwell equations unify electromagnetism
with special relativity. Quantumelectrodynamics unifies electrodynamics with
quantum mechanics. And so on and so forth, till the formulation of the Stan-
dard Model which describes in an unified way all known non-gravitational

11
 interactions. Supersymmetry (and its local version, supergravity), is the most
natural candidate to complete this long journey. It is a way not just to de-
scribe in a unified way all known interactions, but in fact to describe matter
and radiation all together. This sounds compelling, and from this view point
it sounds natural studying supersymmetry and its consequences.

• There is one more reason as to why one could expect that supersymmetry is out
there, after all. As already emphasized, as of today string theory stands up as
the most satisfactory theory where to describe quantum gravity in a consistent
way and, also, to describe all known interactions in a unified framework. So,
it might very well be that Nature, at high enough energy, is described by
string theory. Unlike a theory of fields, a theory of strings can only be made
consistent if it is supersymmetric. So, in this sense, supersymmetry is predicted
to be realized in Nature, if string theory is correct. Supersymmetry is in fact
one of the two more striking predictions of string theory (the other being the
existence of extra-dimensions).

Note: all above arguments suggest that supersymmetry maybe realized in Nature.
However, none of such arguments give any obvious indication on the energy scale at
which supersymmetry might show-up. In principle, this scale can be very high, as
high as the Planck scale. Below, we will present few arguments, more phenomeno-
logical in nature, which suggest that low energy supersymmetry (as low as TeV scale
or slightly higher) would be the preferred option.

ii. Phenomenological reasons.

• Naturalness and the hierarchy problem. Three out of four of the fundamental
interactions among elementary particles (strong, weak and electromagnetic)
are described by the Standard Model (SM). The typical scale of the SM, the
electroweak scale, is

Mew ∼ 250 GeV ⇐⇒ Lew ∼ 10−16 mm . (1.5)

The SM is very well tested up to such energies. This cannot be the end
of the story, though: for one thing, at high enough energies, as high as the
Planck scale Mpl , gravity becomes comparable with other forces and cannot
be neglected in elementary particle interactions. At some point, we need a
quantum theory of gravity. Actually, the fact that Mew /Mpl << 1 calls for

12
 new physics at a much lower scale. One way to see this, is as follows. The
Higgs potential reads

V (H) ∼ µ2 |H|2 + λ|H|4 where µ2 < 0 . (1.6)

p
Experimentally, the minimum of such potential, hHi = −µ2 /2λ, is at around
174GeV. This implies that the bare mass of the Higgs particle is roughly
around 100 GeV or so, m2H = −µ2 ∼ (100GeV)2 . What about radiative cor-
rections? Scalar masses are subject to quadratic divergences in perturbation
theory. The SM fermion coupling −λf Hf f induces a one-loop correction to
the Higgs mass as
∆m2H ∼ − 2 λ2f Λ2 (1.7)
due to diagrams as the one in Fig. 1.1. A natural physical UV cut-off Λ

f
<latexit sha1_base64="Ugy/fqiFLFxo8F7s7CQYM9xgh5c=">AAAB6HicdVDJSgNBEK2JW4xb1KOXxiB4GiYxkngLevGYgFkgGUJPpyZp07PQ3SOEIV/gxYMiXv0kb/6NnQ1U9EHB470qqup5seBKO86nlVlb39jcym7ndnb39g/yh0ctFSWSYZNFIpIdjyoUPMSm5lpgJ5ZIA09g2xvfzPz2A0rFo/BOT2J0AzoMuc8Z1UZq+P18wbFLV1XnokoWpFxZkRIp2s4cBVii3s9/9AYRSwIMNRNUqW7RibWbUqk5EzjN9RKFMWVjOsSuoSENULnp/NApOTPKgPiRNBVqMle/T6Q0UGoSeKYzoHqkfnsz8S+vm2i/6qY8jBONIVss8hNBdERmX5MBl8i0mBhCmeTmVsJGVFKmTTY5E8LqU/I/aZXs4qXtNMqF2vUyjiycwCmcQxEqUINbqEMTGCA8wjO8WPfWk/VqvS1aM9Zy5hh+wHr/AkeMjUM=</latexit>

H
<latexit sha1_base64="XkUC/lSHe3+EOmbSi0AEbToBXk8=">AAAB6HicdVDJSgNBEK1xjXGLevTSGARPw0yMJN6CXnJMwCyQDKGnU0na9Cx09whhyBd48aCIVz/Jm39jZwMVfVDweK+Kqnp+LLjSjvNpra1vbG5tZ3ayu3v7B4e5o+OmihLJsMEiEcm2TxUKHmJDcy2wHUukgS+w5Y9vZ37rAaXiUXinJzF6AR2GfMAZ1UaqV3u5vGMXrsvOZZksSLG0IgXi2s4ceVii1st9dPsRSwIMNRNUqY7rxNpLqdScCZxmu4nCmLIxHWLH0JAGqLx0fuiUnBulTwaRNBVqMle/T6Q0UGoS+KYzoHqkfnsz8S+vk+hB2Ut5GCcaQ7ZYNEgE0RGZfU36XCLTYmIIZZKbWwkbUUmZNtlkTQirT8n/pFmw3SvbqRfzlZtlHBk4hTO4ABdKUIEq1KABDBAe4RlerHvryXq13hata9Zy5gR+wHr/AhoUjSU=</latexit>

f¯
<latexit sha1_base64="XOg/xTqi2AIfGlPbEz/+yl2iY/g=">AAAB7XicdVDLSgMxFL3js9ZX1aWbYBFcDTO10rorunFZwT6gHUomzbSxmWRIMkIZ+g9uXCji1v9x59+YvkBFz+pwzr3cc0+YcKaN5306K6tr6xubua389s7u3n7h4LCpZaoIbRDJpWqHWFPOBG0YZjhtJ4riOOS0FY6up37rgSrNpLgz44QGMR4IFjGCjZWa3RArFPUKRc8tXVa98yqak3JlSUrId70ZirBAvVf46PYlSWMqDOFY647vJSbIsDKMcDrJd1NNE0xGeEA7lgocUx1ks7QTdGqVPoqkPSyFQTP1+0aGY63HcWgnY2yG+rc3Ff/yOqmJqkHGRJIaKsj8UJRyZCSavo76TFFi+NgSTBSzWREZYoWJsQXlbQnLT9H/pFly/QvXuy0Xa1eLOnJwDCdwBj5UoAY3UIcGELiHR3iGF0c6T86r8zYfXXEWO0fwA877F5YAjyY=</latexit>

Figure 1.1: One-loop radiative correction to the Higgs mass due to fermion couplings.

should then be at around the TeV scale in order to protect the Higgs mass,
and the SM should then be seen as an effective theory valid up to energies
E ≤ Meff ∼ TeV, well below the Planck scale.
What can be the new physics beyond such scale and how can such new physics
protect the otherwise perturbative divergent Higgs mass? New physics, if any,
may include new fermionic and bosonic fields, possibly coupling to the SM
Higgs. Each of these fields will give radiative contributions to the Higgs mass
of the kind above, hence, no matter what new physics will show-up at high
energy, the natural mass for the the Higgs field would always be of order the
UV cut-off of the theory, generically around ∼ Mpl . We would need a huge
fine-tuning to get it stabilized at ∼ 100GeV (we now know that the physical
Higgs mass is at 125 GeV, in fact)! This is known as the hierarchy problem: the
experimental value of the Higgs mass is unnaturally smaller than its natural
theoretical value.
In principle, there is a very simple way out of this. This resides in the fact that
(as you should know from your QFT course!) scalar couplings provide one-loop

13
 radiative contributions which are opposite in sign with respect to fermions.
Suppose there exist some new scalar, S, with Higgs coupling −λS |H|2 |S|2 .
Such coupling would also induce corrections to the Higgs mass via the one-
loop diagram in Figure 1.2.

S
<latexit sha1_base64="pkq/2YLEDOl5uzDzJ6k+SlFIZDU=">AAAB6HicdVDJSgNBEK2JW4xb1KOXxiB4GmZiJPEW9OIxQbNAMoSeTk3Spmehu0cIIV/gxYMiXv0kb/6NnUVQ0QcFj/eqqKrnJ4Ir7TgfVmZldW19I7uZ29re2d3L7x80VZxKhg0Wi1i2fapQ8AgbmmuB7UQiDX2BLX90NfNb9ygVj6NbPU7QC+kg4gFnVBupftPLFxy7eFFxzipkQUrlL1Ikru3MUYAlar38e7cfszTESDNBleq4TqK9CZWaM4HTXDdVmFA2ogPsGBrREJU3mR86JSdG6ZMglqYiTebq94kJDZUah77pDKkeqt/eTPzL66Q6qHgTHiWpxogtFgWpIDoms69Jn0tkWowNoUxycythQyop0yabnAnh61PyP2kWbffcduqlQvVyGUcWjuAYTsGFMlThGmrQAAYID/AEz9ad9Wi9WK+L1oy1nDmEH7DePgEqwI0w</latexit>

H
<latexit sha1_base64="XkUC/lSHe3+EOmbSi0AEbToBXk8=">AAAB6HicdVDJSgNBEK1xjXGLevTSGARPw0yMJN6CXnJMwCyQDKGnU0na9Cx09whhyBd48aCIVz/Jm39jZwMVfVDweK+Kqnp+LLjSjvNpra1vbG5tZ3ayu3v7B4e5o+OmihLJsMEiEcm2TxUKHmJDcy2wHUukgS+w5Y9vZ37rAaXiUXinJzF6AR2GfMAZ1UaqV3u5vGMXrsvOZZksSLG0IgXi2s4ceVii1st9dPsRSwIMNRNUqY7rxNpLqdScCZxmu4nCmLIxHWLH0JAGqLx0fuiUnBulTwaRNBVqMle/T6Q0UGoS+KYzoHqkfnsz8S+vk+hB2Ut5GCcaQ7ZYNEgE0RGZfU36XCLTYmIIZZKbWwkbUUmZNtlkTQirT8n/pFmw3SvbqRfzlZtlHBk4hTO4ABdKUIEq1KABDBAe4RlerHvryXq13hata9Zy5gR+wHr/AhoUjSU=</latexit>

Figure 1.2: One-loop radiative correction to the Higgs mass due to scalar couplings.

Such corrections would have opposite sign with respect to those coming from
fermion couplings, that is
∆m2H ∼ λS Λ2 . (1.8)
Therefore, if the new physics would be such that each quark and lepton of
the SM were accompanied by two complex scalars having the same Higgs
couplings of the quark and lepton, i.e. λS = |λf |2 , then all Λ2 contributions
would automatically cancel, and the Higgs mass would be stabilized at its tree
level value! Such conspiracy, however, would be quite ad hoc, and not really
solving the fine-tuning problem mentioned above; rather, just rephrasing it.
A natural thing to invoke to have such magic cancellations would be to have a
symmetry protecting mH , right in the same way as gauge symmetry protects
the masslessness of spin-1 particles. A symmetry imposing to the theory the
correct matter content (and couplings) for such cancellations to occur. This is
exactly what supersymmetry is: in a supersymmetric theory there are fermions
and bosons (and couplings) just in the right way to provide exact cancellation
between diagrams like the ones above. In summary, supersymmetry is a very
natural and economic way (though not the only possible one) to solve the
hierarchy problem.
Known fermions and bosons cannot be partners of each other. For one thing,
we do not observe any degeneracy in mass in elementary particles that we
know. Moreover, and this is possibly a stronger reason, quantum numbers do
not match: gauge bosons transform in the adjoint representations of the SM
gauge group while quarks and leptons in the fundamental or singlet representa-
tions. Hence, in a supersymmetric world, each SM particle should have a (yet

14
not observed!) supersymmetric partner, usually dubbed sparticle. Roughly,
the spectrum of such supersymmetric Standard Model (SSM) should be as
follows

SM particles SUSY partners

gauge bosons gauginos
quarks, leptons scalars
Higgs higgsino

Notice: the (down) Higgs has the same quantum numbers as the scalar part-
ner of neutrino and leptons, sneutrino and sleptons respectively, (Hd0 , Hd− ) ↔
(ν̃, ẽL ). Hence, one can imagine that the Higgs is in fact a sparticle. This
cannot be. In such scenario, there would be phenomenological problems, e.g.
lepton number violation and (at least one) neutrino mass in gross violation of
experimental bounds.
In summary, the world we already had direct experimental access to, is not su-
persymmetric. If at all realized, supersymmetry should be a (spontaneously)
broken symmetry in the vacuum state chosen by Nature. However, in order to
solve the hierarchy problem without too much fine-tuning this scale should be
not much higher than 1 TeV. Including lower bounds from present day exper-
iments, it turns out that the SUSY breaking scale should be in the following
energy range

102 GeV ≤ SUSY breaking scale ≤ 103 − 104 GeV .

Let us emphasize that these bounds are just a crude and rough estimate, as
they depend very much on the specific SSM one is actually considering. In
particular, the upper bound can be made higher by enriching the structure
of the SSM in various ways, while keeping naturalness as a guiding principle.
In any event, these bounds are the basic reason why it was believed SUSY to
show-up at the LHC.
It is worth stressing that, as of today, no signal of supersymmetry has been
found at LHC or elsewhere and this has made the above upper bounds more
and more in tension with experimental data, and in turn the very idea of
naturalness being reconsidered, at least in this context. There are ongoing
discussions on these aspects, including the idea that the resolution of the hi-
erarchy problem should not use naturalness as a guiding principle and that

15
 it should be explained by something different, as for instance anthropic argu-
ments or something we do not yet fully understand.

• Gauge coupling unification. There is another reason to believe in (low energy)

supersymmetry; possibly stronger, from a phenomenological point of view,
than that provided by the hierarchy problem. Forget about supersymmetry for
a while, and consider the SU (3)×SU (2)L ×U (1)Y SM as it stands. Interesting
enough, besides the EW scale, the SM contains in itself a new scale of order
1015 GeV. The three SM gauge couplings run according to RG equations like
4π bi µ
= ln i = 1, 2, 3 . (1.9)
gi2 (µ) 2π Λi

At the EW scale, µ = MZ , there is a hierarchy between them, g1 (MZ ) <

g2 (MZ ) < g3 (MZ ). But RG equations make this hierarchy changing with the
energy scale. In fact, supposing there are no particles other than the SM
ones, at a much higher scale, MGU T ∼ 1015 GeV, the three couplings tend to
meet! This naturally calls for a Grand Unified Theory (GUT), where the three
interactions are unified in a single one, two possible GUT gauge groups being
SU (5) and SO(10). The symmetry breaking pattern one should have in mind
would then be as follows

SU (5) → SU (3) × SU (2)L × U (1)Y → SU (3) × U (1)em

φ H

where φ is an heavy Higgs inducing spontaneous symmetry breaking at energies

MGU T ∼ 1015 GeV, and H the SM light Higgs, inducing EW spontaneous
symmetry breaking around the TeV scale. This idea makes a lot of sense but
poses several problems. First, there is a new hierarchy problem (generically,
the SM Higgs mass is expected to get corrections from the heavy Higgs φ).
Second, there is a proton decay problem: some of the additional gauge bosons
predicted by the GUT group mediate baryon number violating transitions,
allowing processes as p → e+ + π0 . This makes the proton not fully stable and
it turns out that its expected lifetime in such GUT framework is violated by
present experimental bounds. Finally, on a more theoretical side, if we do not
allow for new particles besides the SM ones to be there at some intermediate
scale, the three gauge couplings only approximately meet and it turns out
that this cannot be taken care of just by experimental uncertainties. The

16
 latter is an unpleasant feature: small numbers are unnatural from a theoretical
view point, unless there are specific reasons (as symmetries) justifying their
otherwise unnatural smallness.
Remarkably, making the GUT supersymmetric (SGUT) solves all of these
problems in a glance! As already emphasized, with supersymmetry, the Higgs
mass is automatically protected. Moreover, just allowing for the minimal su-
persymmetric extension of the SM spectrum, known as MSSM, the three gauge
couplings do meet (more precisely, they miss but now well within experimen-
tal uncertainties). Finally, the GUT scale is raised enough, up to around 1016
GeV, so to let proton decay rate being compatible with experimental bounds.
So, supersymmetry makes the very natural idea of gauge coupling unification
via a GUT free of any apparent drawbacks.

Couplings Couplings

1 1
<latexit sha1_base64="e44U31FHPaJbNupv9HZ8IeY4V34=">AAAB7nicdVDLSgNBEOz1GeMr6tHLYBC8uMwGg3oLevEYwTwgWcPsZDYZMju7zMwKYclHePGgiFe/x5t/42weoKIFDUVVN91dQSK4Nhh/OkvLK6tr64WN4ubW9s5uaW+/qeNUUdagsYhVOyCaCS5Zw3AjWDtRjESBYK1gdJ37rQemNI/lnRknzI/IQPKQU2Ks1PLwfXbqTXqlMna9yxwIuxVsUV2QCvJcPEUZ5qj3Sh/dfkzTiElDBdG64+HE+BlRhlPBJsVuqllC6IgMWMdSSSKm/Wx67gQdW6WPwljZkgZN1e8TGYm0HkeB7YyIGerfXi7+5XVSE174GZdJapiks0VhKpCJUf476nPFqBFjSwhV3N6K6JAoQo1NqGhDWHyK/ifNiutVXXx7Vq5dzeMowCEcwQl4cA41uIE6NIDCCB7hGV6cxHlyXp23WeuSM585gB9w3r8AkJiPEg==</latexit> <latexit sha1_base64="e44U31FHPaJbNupv9HZ8IeY4V34=">AAAB7nicdVDLSgNBEOz1GeMr6tHLYBC8uMwGg3oLevEYwTwgWcPsZDYZMju7zMwKYclHePGgiFe/x5t/42weoKIFDUVVN91dQSK4Nhh/OkvLK6tr64WN4ubW9s5uaW+/qeNUUdagsYhVOyCaCS5Zw3AjWDtRjESBYK1gdJ37rQemNI/lnRknzI/IQPKQU2Ks1PLwfXbqTXqlMna9yxwIuxVsUV2QCvJcPEUZ5qj3Sh/dfkzTiElDBdG64+HE+BlRhlPBJsVuqllC6IgMWMdSSSKm/Wx67gQdW6WPwljZkgZN1e8TGYm0HkeB7YyIGerfXi7+5XVSE174GZdJapiks0VhKpCJUf476nPFqBFjSwhV3N6K6JAoQo1NqGhDWHyK/ifNiutVXXx7Vq5dzeMowCEcwQl4cA41uIE6NIDCCB7hGV6cxHlyXp23WeuSM585gB9w3r8AkJiPEg==</latexit>

10 10
SU(3) SU(3)

SU(2) SU(2)
2 2
<latexit sha1_base64="g1qIhHsbVWF3F+CSmj4zhYQGMlw=">AAAB7nicdVDLSgNBEOz1GeMr6tHLYBC8uMwGg3oLevEYwTwgWcPsZDYZMju7zMwKYclHePGgiFe/x5t/42weoKIFDUVVN91dQSK4Nhh/OkvLK6tr64WN4ubW9s5uaW+/qeNUUdagsYhVOyCaCS5Zw3AjWDtRjESBYK1gdJ37rQemNI/lnRknzI/IQPKQU2Ks1PLwfXZamfRKZex6lzkQdivYorogFeS5eIoyzFHvlT66/ZimEZOGCqJ1x8OJ8TOiDKeCTYrdVLOE0BEZsI6lkkRM+9n03Ak6tkofhbGyJQ2aqt8nMhJpPY4C2xkRM9S/vVz8y+ukJrzwMy6T1DBJZ4vCVCATo/x31OeKUSPGlhCquL0V0SFRhBqbUNGGsPgU/U+aFderuvj2rFy7msdRgEM4ghPw4BxqcAN1aACFETzCM7w4ifPkvDpvs9YlZz5zAD/gvH8Bkh2PEw==</latexit> <latexit sha1_base64="g1qIhHsbVWF3F+CSmj4zhYQGMlw=">AAAB7nicdVDLSgNBEOz1GeMr6tHLYBC8uMwGg3oLevEYwTwgWcPsZDYZMju7zMwKYclHePGgiFe/x5t/42weoKIFDUVVN91dQSK4Nhh/OkvLK6tr64WN4ubW9s5uaW+/qeNUUdagsYhVOyCaCS5Zw3AjWDtRjESBYK1gdJ37rQemNI/lnRknzI/IQPKQU2Ks1PLwfXZamfRKZex6lzkQdivYorogFeS5eIoyzFHvlT66/ZimEZOGCqJ1x8OJ8TOiDKeCTYrdVLOE0BEZsI6lkkRM+9n03Ak6tkofhbGyJQ2aqt8nMhJpPY4C2xkRM9S/vVz8y+ukJrzwMy6T1DBJZ4vCVCATo/x31OeKUSPGlhCquL0V0SFRhBqbUNGGsPgU/U+aFderuvj2rFy7msdRgEM4ghPw4BxqcAN1aACFETzCM7w4ifPkvDpvs9YlZz5zAD/gvH8Bkh2PEw==</latexit>

10 10
U(1) U(1)

GeV GeV
1016 1016
<latexit sha1_base64="YPp+f2QT03iqJSaCZ4zmeCJmMRE=">AAAB7nicdVDJSgNBEK1xjXGLevTSGARPQ08wLregF48RzALJGHo6PUmTnp6hu0cIQz7CiwdFvPo93vwbe7KAij4oeLxXRVW9IBFcG4w/naXlldW19cJGcXNre2e3tLff1HGqKGvQWMSqHRDNBJesYbgRrJ0oRqJAsFYwus791gNTmsfyzowT5kdkIHnIKTFWann4PvPOJr1SGbveZQ6E3Qq2qC5IBXkunqIMc9R7pY9uP6ZpxKShgmjd8XBi/Iwow6lgk2I31SwhdEQGrGOpJBHTfjY9d4KOrdJHYaxsSYOm6veJjERaj6PAdkbEDPVvLxf/8jqpCS/8jMskNUzS2aIwFcjEKP8d9bli1IixJYQqbm9FdEgUocYmVLQhLD5F/5NmxfWqLr49Ldeu5nEU4BCO4AQ8OIca3EAdGkBhBI/wDC9O4jw5r87brHXJmc8cwA8471+eSY8b</latexit> <latexit sha1_base64="YPp+f2QT03iqJSaCZ4zmeCJmMRE=">AAAB7nicdVDJSgNBEK1xjXGLevTSGARPQ08wLregF48RzALJGHo6PUmTnp6hu0cIQz7CiwdFvPo93vwbe7KAij4oeLxXRVW9IBFcG4w/naXlldW19cJGcXNre2e3tLff1HGqKGvQWMSqHRDNBJesYbgRrJ0oRqJAsFYwus791gNTmsfyzowT5kdkIHnIKTFWann4PvPOJr1SGbveZQ6E3Qq2qC5IBXkunqIMc9R7pY9uP6ZpxKShgmjd8XBi/Iwow6lgk2I31SwhdEQGrGOpJBHTfjY9d4KOrdJHYaxsSYOm6veJjERaj6PAdkbEDPVvLxf/8jqpCS/8jMskNUzS2aIwFcjEKP8d9bli1IixJYQqbm9FdEgUocYmVLQhLD5F/5NmxfWqLr49Ldeu5nEU4BCO4AQ8OIca3EAdGkBhBI/wDC9O4jw5r87brHXJmc8cwA8471+eSY8b</latexit>

Standard Model …+ Supersymmetry

Figure 1.3: On the left a qualitative picture describing the running of the three SM
couplings, which approximately meet at a scale of order 1015 GeV. On the right, the
same picture in a minimal supersymmetric extension of the SM, where the couplings
exactly meet (within experimental uncertainties) at a scale of order 1016 GeV.

Disclaimer: the MSSM is not the only possible option for supersymmetry
beyond the SM, just the most economic one. In the MSSM one just adds a
superpartner to each SM particle, therefore introducing the higgsino, the wino,
the zino, together with all squarks and sleptons, and no more. [There is in
fact an exception. To have a meaningful model one has to double the Higgs
sector, and have two Higgs doublets. One reason for that is gauge anomaly

17
 cancellation: the higgsinos are fermions in the fundamental representation of
SU (2)L hence two of them are needed, with opposite hypercharge, not to spoil
the anomaly-free properties of the SM. A second reason is that in the SM
the field H gives mass to down quarks and charged leptons while its charge
conjugate, H c (∼ H) gives mass to up quarks. As we will see, in a SUSY model
H cannot enter in the potential, which is a function of H, only. Therefore,
in a supersymmetric scenario, to give mass to up quarks one needs a second,
independent Higgs doublet.] There exist many non-minimal supersymmetric
extensions of the Standard Model (which, in fact, are in better shape against
experimental constraints with respect to the MSSM). One can in principle
construct any SSM one likes. In doing so, however, several constraints are
to be taken into account. For example, it is not so easy to make such non-
minimal extensions keeping the nice exact gauge coupling unification enjoyed
by the MSSM.
It is worth stressing that gauge coupling unification and the hierarchy problem
are independent issues. Indeed, for the former to hold one does not need a
full supersymmetric spectrum at low energy. Only light fermionic partners are
needed. Scalar partners of SM fermions sit in full GUT families so they do
not contribute to gauge coupling unification; they just shift all couplings by
one and the same constant. At the price of forgetting about naturalness, this
observation opened-up the idea that the SUSY spectrum can be split - with
light fermions and heavy scalars - with supersymmetry being realized only at
high energy. This scenario goes under the name of Split Supersymmetry.

• Supersymmetry and dark matter. Another context where supersymmetry might

play an important role is cosmology. There are various evidences which indi-
cate that around 26% of the energy density in the Universe should be made
of dark matter, i.e. non-luminous and non-baryonic matter. The only SM
candidates for dark matter are neutrinos, but they are disfavored by available
experimental data (basically, neutrinos are too light to account for such an
enormous energy density). Supersymmetry provides instead a valuable and
very natural dark matter candidate: the neutralino. Neutralinos are mass
eigenstates of a linear superposition of the supersymmetric partners of the
neutral Higgs and of the SU(2) and U(1) neutral gauge bosons

e 0 + αi2 W
χi = αi1 B f 0 + αi3 H
e 0 + αi4 H
e0 . (1.10)
u d

18
 In most SUSY frameworks the neutralino is the lightest supersymmetric par-
ticle (LSP), and fully stable, as a dark matter candidate should be.

iii. Supersymmetry as a theoretical laboratory for strongly coupled gauge dynamics.

• What if supersymmetry will turn out not to be the correct theory to describe
beyond the Standard Model physics? Or, worse, what if supersymmetry will
turn out not to be realized at all, in Nature (something we could hardly ever
being able to prove, in fact)? Interestingly, there is yet another reason which
makes it worth studying supersymmetric theories, independently from the role
supersymmetry might or might not play as a theory describing high energy
physics.
Let us consider non-abelian gauge theories, which strong interactions are an
example of. Every time a non-abelian gauge group remains unbroken at low
energy, we have to deal with strong coupling. The typical questions one should
try and answer (in QCD or similar theories) are:

– The bare Lagrangian is described in terms of quark and gluons, which

are UV degrees of freedom. Which are the IR (light) degrees of freedom
of QCD? What is the effective Lagrangian in terms of such degrees of
freedom?
– Strong coupling physics is very rich. Typically, one has to deal with
phenomena like confinement, charge screening, the generation of a mass
gap, etc.... Is there any theoretical understanding of such phenomena?
– It is believed that the QCD vacuum is populated by vacuum conden-
sates of fermion bilinears, hΩ|ψψ|Ωi =
6 0, which induce chiral symmetry
breaking. What is the microscopic mechanism behind this phenomenon?

Most of the IR properties of QCD have eluded so far a clear understanding,

since we lack analytical tools to deal with strong coupling dynamics. Most
results come from lattice computations, but these do not furnish a first prin-
ciple understanding of the above phenomena. Moreover, they are formulated
in Euclidean space and are not suited to discuss, e.g. transport properties.
Because of their nice renormalization properties, supersymmetric theories are
more constrained than ordinary field theories and let one have a better control
on strong coupling regimes, sometime. Therefore, one might hope to use them

19
 as toy models where to study properties of more realistic theories, such as
QCD, in a more controlled way. Indeed, as we shall see, supersymmetric
theories do provide examples where some of the above strong coupling effects
can be studied exactly! This is possible due to powerful non-renormalizations
theorems supersymmetric theories enjoy, and because of a very special property
of supersymmetry, known as holomorphy, which in certain circumstances lets
one compute several non-perturbative contributions to the Lagrangian exactly.
We will spend a sizeable amount of time discussing these issues in the second
part of this course.

This is all we wanted to say in this introductory chapter, which should be re-
garded just as an invitation to supersymmetry and its fascinating world. Let us
end by just adding a curious historical remark. Supersymmetry did not first ap-
pear in ordinary four-dimensional quantum field theories but in string theory, at
the very beginning of the seventies. Only later it was shown to be possible to have
supersymmetry in ordinary quantum field theories.

1.3 Some useful references

The list of references in the literature is endless. Below I list some old and more
recent books plus some reviews which are available on the Archive dialing at

https://arxiv.org/multi?group=grp

Some of these references may be better than others, depending on the specific topic
one is interested in (and on personal taste). In preparing these lectures I have used
most of them, some more, some less. A collection of references is also given at
the end of each chapter, where I refer to some original papers, review articles or
textbooks that I found useful for preparing the material presented there. This will
help the reader to be guided if she/he wants to deepen any specific topics and have
access to the original font... and it also let me give proper credit to authors.

1. Historical references

• J. Wess and J. Bagger

Supersymmetry and supergravity
Princeton, USA: Univ. Pr. (1992) 259 p.

20
 • P. C. West
Introduction to supersymmetry and supergravity
Singapore: World Scientific (1990) 425 p.
• M. F. Sohnius
Introducing Supersymmetry
Phys. Rep. 128 (1985)

2. Some more recent books

• S. Weinberg
The quantum theory of fields. Vol. 3: Supersymmetry
Cambridge, UK: Univ. Pr. (2000) 419 p.
• J. Terning
Modern supersymmetry: Dynamics and duality
Oxford University Press (2006) 324 p.
• M. Dine
Supersymmetry and string theory: Beyond the standard model
Cambridge University Press (2007) 515 p.
• H.J. Müller-Kirsten and A. Wiedemann
Introduction to Supersymmetry
Singapore: World Scientific (2010) 439 p.
• S. Cecotti
Supersymmetric field theories
Cambridge University Press (2015) 424 p.

3. On-line reviews: bases

• J. D. Lykken
Introduction to Supersymmetry
TASI 96
arXiv:hep-th/9612114
• S. P. Martin
A Supersymmetry Primer
Adv. Ser. Direct. High Energy Phys. 21 (2010) [arXiv:hep-ph/9709356]
• A. Bilal
Introduction to supersymmetry
arXiv:hep-th/0101055

21
 • J. Figueroa-O’Farrill
BUSSTEPP Lectures on Supersymmetry
arXiv:hep-th/0109172
• M. J. Strassler
An Unorthodox Introduction to Supersymmetric Gauge Theory
TASI 2001
arXiv:hep-th/0309149
• R. Argurio, G. Ferretti and R. Heise
An introduction to supersymmetric gauge theories and matrix models
Int. J. Mod. Phys. A 19 (2004) 2015 [arXiv:hep-th/0311066]

4. On-line reviews: advanced topics

• K. A. Intriligator and N. Seiberg

Lectures on supersymmetric gauge theories and electric-magnetic duality
Nucl. Phys. Proc. Suppl. 45BC (1996) [arXiv:hep-th/9509066]
• A. Bilal
Duality in N=2 SUSY SU(2) Yang-Mills Theory: A pedagogical introduc-
tion to the work of Seiberg and Witten
arXiv:hep-th/9601007
• L. Alvarez-Gaume and S. F. Hassan
Introduction to S duality in N=2 supersymmetric gauge theories: A ped-
agogical review of the work of Seiberg and Witten
Fortsch. Phys. 45 (1997) [arXiv:hep-th/9701069]
• M. E. Peskin
Duality in Supersymmetric Yang-Mills Theory
TASI 96
arXiv:hep-th/9702094
• M. Shifman
Non-Perturbative Dynamics in Supersymmetric Gauge Theories
Prog. Part. Nucl. Phys. 39 (1997) [arXiv:hep-th/9704114]
• P. Di Vecchia
Duality in supersymmetric N = 2, 4 gauge theories
Les Houches Summer School 1997
arXiv:hep-th/9803026

22
 • E. D’Hoker and D. H. Phong
Lectures on supersymmetric Yang-Mills theory and integrable systems
CRM Summer School 1999
arXiv:hep-th/9912271
• P. Argyres
Lectures on Supersymmetry
available at http://www.physics.uc.edu/~argyres/661/index.html
• Y. Tachikawa
N=2 supersymmetric dynamics for pedestrians
Lect.Notes Phys. 890 (2014) [arXiv:1312.2684 [hep-th]]

5. On-line reviews: supersymmetry breaking

• G. F. Giudice and R. Rattazzi

Theories with Gauge-Mediated Supersymmetry Breaking
Phys. Rept. 322 (1999) [arXiv:hep-ph/9801271]
• E. Poppitz and S. P. Trivedi
Dynamical Supersymmetry Breaking
Ann. Rev. Nucl. Part. Sci. 48 (1998) [arXiv:hep-th/9803107]
• Y. Shadmi and Y. Shirman
Dynamical Supersymmetry Breaking
Rev. Mod. Phys. 72 (2000) [arXiv:hep-th/9907225]
• M. A. Luty
2004 TASI Lectures on Supersymmetry Breaking
arXiv:hep-th/0509029
• Y. Shadmi
Supersymmetry breaking
arXiv:hep-th/0601076
• K. A. Intriligator and N. Seiberg
Lectures on Supersymmetry Breaking
Class. Quant. Grav. 24 (2007) [arXiv:hep-ph/0702069]

23
2 The supersymmetry algebra

In this lecture we will introduce the supersymmetry algebra, which is the al-
gebra encoding the set of symmetries a supersymmetric theory should enjoy. The
supersymmetry algebra is an extension of the Poincaré algebra so in the first two
sections we will start recalling a few facts about Lorentz and Poincaré algebras, the
corresponding groups and their representations. In particular, we will emphasize
the relation of the Lorentz algebra with the SU (2) and SL(2, C) algebras and define
(two component) spinors as the basic representation of the Lorentz group. In the
last section we will introduce the concept of graded Lie algebra and, finally, the
supersymmetry algebra, which is a specific instance of a graded Lie algebra.

2.1 Lorentz and Poincaré groups

The Lorentz group SO(1, 3) is the subgroup of matrices Λ of GL(4, R) with unit
determinant, detΛ = 1, and which satisfy the following relation

ΛT η Λ = η , (2.1)

where η is the (mostly minus in our conventions) flat Minkowski metric

ηµν = diag(+, −, −, −) . (2.2)

The Lorentz group has six generators (associated to space rotations and boosts)
enjoying the following commutation relations

[Ji , Jj ] = iijk Jk , [Ji , Kj ] = iijk Kk , [Ki , Kj ] = −iijk Jk . (2.3)

Notice that while the Ji are hermitian, the boosts Ki are anti-hermitian, this being
related to the fact that the Lorentz group is non-compact (topologically, the Lorentz
group is R3 ×S3 /Z2 , the non-compact factor corresponding to boosts and the doubly
connected S3 /Z2 corresponding to rotations). In order to construct representations
of this algebra it is useful to introduce the following complex linear combinations of
the generators Ji and Ki
1
Ji± = (Ji ± iKi ) , (2.4)
2
where now the Ji± are hermitian. In terms of Ji± the algebra (2.3) becomes

[Ji± , Jj± ] = iijk Jk± , [Ji± , Jj∓ ] = 0 . (2.5)

24
This shows that the Lorentz algebra is equivalent to two SU (2) algebras. As we will
see later, this simplifies a lot the study of representations of the Lorentz group, which
can be organized into (couples of) SU (2) representations. This isomorphism comes
from the theory of Lie Algebra which says that at the level of complex algebras

SO(4) ' SU (2) × SU (2) . (2.6)

The Lorentz algebra is a specific real form of that of SO(4). This difference can be
seen from the defining commutation relations (2.3): for SO(4) one would have had
a plus sign on the right hand side of the third such commutation relations. This
difference has some consequence when it comes to study representations. In particu-
lar, while in Euclidean space all representations are real or pseudoreal, in Minkowski
space complex conjugation interchanges the two SU (2)’s. This can also be seen at
the level of the generators Ji± . In order for all rotation and boost parameters to be
real, one must take all the Ji and Ki to be imaginary and hence from eq. (2.4) one
sees that
(Ji± )∗ = −Ji∓ . (2.7)
In terms of algebras, all this discussion can be summarized noticing that for the
Lorentz algebra the isomorphism (2.6) changes into

SO(1, 3) ' SU (2) × SU (2)∗ . (2.8)

For later purpose let us introduce a four-vector notation for the Lorentz genera-
tors, in terms of an anti-symmetric tensor Mµν defined as

Mµν = −Mνµ with M0i = Ki and Mij = ijk Jk , (2.9)

where µ = 0, 1, 2, 3. In terms of such matrices, the Lorentz algebra reads

[Mµν , Mρσ ] = −iηµρ Mνσ − iηνσ Mµρ + iηµσ Mνρ + iηνρ Mµσ . (2.10)

Another useful relation one should bear in mind is the relation between the
Lorentz group and SL(2, C), the group of 2 × 2 complex matrices with unit determi-
nant. More precisely, there exists a homomorphism between SL(2, C) and SO(1, 3),
which means that for any matrix A ∈ SL(2, C) there exists an associated Lorentz
matrix Λ, and that
Λ(A) Λ(B) = Λ(AB) , (2.11)

25
where A and B are SL(2, C) matrices. This can be proved as follows. Lorentz
transformations act on four-vectors as

x0µ = Λµν xν , (2.12)

where the matrices Λ’s are a representation of the generators Mµν defined above.
Let us introduce 2 × 2 matrices σµ where σ0 is the identity matrix and σi are the
Pauli matrices defined as
! ! !
0 1 0 −i 1 0
σ1 = , σ2 = , σ3 = . (2.13)
1 0 i 0 0 −1

Let us also define the matrices with upper indices, σ µ , as

σ µ = (σ 0 , σ i ) = (σ0 , −σi ) . (2.14)

The matrices σµ are a complete set, in the sense that any 2 × 2 complex matrix can
be written as a linear combination of them. For every four-dimensional vector xµ
let us construct the 2 × 2 complex matrix

ρ : xµ → xµ σµ = X . (2.15)

The matrix X is hermitian, since the Pauli matrices are hermitian, and has deter-
minant equal to xµ xµ , which is a Lorentz invariant quantity. Therefore, ρ is a map
from Minkowski space to H, the space of 2 × 2 hermitian complex matrices

M4 −→ H . (2.16)
ρ

Let us now act on X with a SL(2, C) transformation A

A : X → AXA† = X 0 . (2.17)

This transformation preserves the determinant since detA = 1 and also preserves
the hermicity of X since

X 0† = (AXA† )† = AX † A† = AXA† = X 0 . (2.18)

Therefore A is a map between H and itself

H −→ H . (2.19)
A

26
We finally apply the inverse map ρ−1 to X 0 and get a four-vector x0µ . The inverse
map is defined as
1
ρ−1 = Tr[• σ̄ µ ] (2.20)
2
(where, as we will later see more rigorously, as a complex 2 × 2 matrix σ̄ µ is the
same as σµ ). Indeed

1 1 1 1
ρ−1 X = Tr[X σ̄ µ ] = Tr[xν σ ν σ̄ µ ] = Tr[σ ν σ̄ µ ]xν = 2 η µν xν = xµ . (2.21)
2 2 2 2
Assembling everything together we then get a map from Minkowski space into itself
via the following chain

M4 −→ H −→ H −→
−1
M4
ρ A ρ
(2.22)
xν −→ xν σ ν −→ Axν σ ν A† −→
−1
1
2
Tr[Axν σ ν A† σ̄ µ ] = x0µ
ρ A ρ

This is nothing but a Lorentz transformation obtained by the SL(2, C) transforma-

tion A as
1
Λµν (A) = Tr[σ̄ µ A σν A† ] . (2.23)
2
It is now a trivial exercise, provided eq. (2.23), to prove the homomorphism (2.11).
Notice that the relation (2.23) can in principle be inverted, in the sense that
for a given Λ one can find a corresponding A ∈ SL(2, C). However, the relation is
not an isomorphism, since it is double valued. The isomorphism holds between the
Lorentz group and SL(2, C)/Z2 (in other words SL(2, C) is a double cover of the
Lorentz group). This can be seen as follows. Consider the 2 × 2 matrix
!
e−iθ/2 0
M (θ) = (2.24)
0 eiθ/2

which corresponds to a Lorentz transformation producing a rotation by an angle θ

about the z-axis. Taking θ = 2π which corresponds to the identity in the Lorentz
group, one gets M = −1 which is a non-trivial element of SL(2, C). It then follows
that the elements of SL(2, C) are identified two-by-two under a Z2 transformation
in the Lorentz group. Note that this Z2 identification holds also in Euclidean space:
at the level of groups SU (2) × SU (2) = Spin(4), where Spin(4) is a double cover of
SO(4) as a group (it has an extra Z2 ).
The Poincaré group is the Lorentz group augmented by the space-time trans-
lation generators Pµ . In terms of the generators Pµ and Mµν the Poincaré algebra

27
reads

[Pµ , Pν ] = 0

[Mµν , Mρσ ] = −iηµρ Mνσ − iηνσ Mµρ + iηµσ Mνρ + iηνρ Mµσ (2.25)

[Mµν , Pρ ] = −iηρµ Pν + iηρν Pµ .

2.2 Spinors and representations of the Lorentz group

We are now ready to discuss representations of the Lorentz group. Thanks to the
isomorphism (2.8) they can be easily organized in terms of those of SU (2) which
can be labeled by the spins. In this respect, let us introduce two-component spinors
as the objects carrying the basic representations of SL(2, C). There exist two such
representations. A spinor transforming in the self-representation M is a two complex
component object !
ψ1
ψ= (2.26)
ψ2
where ψ1 and ψ2 are complex Grassmann numbers, which transform under a matrix
M ∈ SL(2, C) as
ψα → ψ α0 = Mαβ ψβ α, β = 1, 2 . (2.27)
The complex conjugate representation is defined from M∗ , where M∗ means com-
plex conjugation, as
0
ψ α̇ → ψ α̇ = M∗ α̇β̇ ψ β̇ α̇, β̇ = 1, 2 . (2.28)

These two representations are not equivalent, that is it does not exist a matrix C
such that M = CM∗ C −1 .
There are, however, other representations which are equivalent to the former.
Let us first introduce the invariant tensor of SU (2), αβ , and similarly for the other
SU (2), α̇β̇ , which one uses to raise and lower spinorial indices as well as to construct
scalars and higher spin representations by spinor contractions
! !
0 −1 0 1
αβ = α̇β̇ = αβ = α̇β̇ = . (2.29)
1 0 −1 0

We can then define

β̇ α̇
ψ α = αβ ψβ , ψα = αβ ψ β , ψ α̇ = α̇β̇ ψ , ψ = α̇β̇ ψ β̇ . (2.30)

28
The convention here is that adjacent indices are always contracted putting the ep-
silon tensor on the left.
Using above conventions one can easily prove that ψ 0α = (M−1T )αβ ψ β . Since
M−1T ' M (the matrix C being in fact the epsilon tensor αβ ), it follows that
the self-representation ψα and the representation ψ α are equivalent. A similar
α̇ 0 α̇
story holds for ψ which transforms in the representation M∗−1T , that is ψ =
β̇
(M∗−1T )α̇β̇ ψ , which is equivalent to the complex conjugate representation ψ α̇ (the
matrix C connecting M∗−1T and M∗ is now the epsilon tensor α̇β̇ ). From our con-
ventions one can easily see that the complex conjugate matrix (Mαβ )∗ (that is, the
matrix obtained from Mαβ by taking the complex conjugate of each entry), once
expressed in terms of dotted indices, is not M∗ α̇β̇ , but rather (M∗−1T )α̇β̇ . Finally,
lower undotted indices are row indices, while upper ones are column indices. Dot-
α̇
ted indices follow instead the opposite convention. This implies that (ψα )∗ = ψ ,
while under hermitian conjugation (which also includes transposition), we have, e.g.
ψ α̇ = (ψα )† , as operator identity.
Due to the homomorphism between SL(2, C) and SO(1, 3), it turns out that the
α̇
two spinor representations ψα and ψ are representations of the Lorentz group, and,
because of the isomorphism (2.8), they can be labeled in terms of SU (2) represen-
tations as
 
1
ψα ≡ ,0 (2.31)
2
 
α̇ 1
ψ ≡ 0, . (2.32)
2
P P
To understand the identifications above just note that i (Ji+ )2 and i (Ji− )2 are
Casimir of the two SU (2) algebras (2.5) with eigenvalues n(n + 1) and m(m + 1)
with n, m = 0, 12 , 1, 23 , . . . being the eigenvalues of J3+ and J3− , respectively. Hence
we can indeed label the representations of the Lorentz group by pairs (n, m) and
since J3 = J3+ + J3− we can identify the spin of the representation as n + m, its
dimension being (2n + 1)(2m + 1). The two spinor representations (2.31) and (2.32)
are just the basic such representations. Note, in passing, that the representations ψα
and ψ α are nothing but the fundamental and anti-fundamental representations of
SU (2). That they are equivalent is specific to SU (2) and does not hold for SU (N )
with N > 2.
Recalling that Grassmann variables anticommute (that is ψ1 χ2 = −χ2 ψ1 , ψ1 χ2̇ =

29
−χ2̇ ψ1 , etc...) we can now define a scalar product for spinors as

ψχ ≡ ψ α χα = αβ ψβ χα = −αβ ψα χβ = −ψα χα = χα ψα = χψ (2.33)

β̇ α̇ α̇ α̇
ψχ ≡ ψ α̇ χ α̇ = α̇β̇ ψ χ α̇ = −α̇β̇ ψ χ β̇ = −ψ χα̇ = χα̇ ψ = χψ . (2.34)

Under hermitian conjugation we have

α̇
(ψχ)† = (ψ α χα )† = χα† ψ α† = χα̇ ψ = χψ . (2.35)

In our conventions, undotted indices are contracted from upper left to lower right
while dotted indices from lower left to upper right (this rule does not apply when
raising or lowering indices with the epsilon tensor). Recalling eq. (2.17), namely
that under SL(2, C) the matrix X = xµ σµ transforms as AXA† and that the index
structure of A and A† is Aαβ and A∗β̇α̇ , respectively, we see that σµ naturally has a
dotted and an undotted index and can be contracted with an undotted and a dotted
spinor as
ψσ µ χ ≡ ψ α σαµα̇ χ α̇ . (2.36)
Similarly one can define σ̄ µ as

σ̄ µ α̇α
= α̇β̇ αβ σβµβ̇ = (σ0 , σi ) , (2.37)

and define the product of σ̄ µ with a dotted and an undotted spinor as

ψσ̄ µ χ ≡ ψ α̇ σ̄ µ α̇β
χβ . (2.38)

A number of useful identities one can prove are

1 1
ψ α ψ β = − αβ ψψ , (θφ) (θψ) = − (φψ) (θθ)
2 2
χσ µ ψ = −ψσ̄ µ χ , χσ µ σ̄ ν ψ = ψσ ν σ̄ µ χ

(χσ µ ψ)† = ψσ µ χ , (χσ µ σ̄ ν ψ)† = ψσ̄ ν σ µ χ

1


1


(θψ) θσ µ φ = − (θθ) ψσ µ φ , θψ θσ̄ µ φ = − θθ ψσ̄ µ φ
2 2
1
(φψ) · χα̇ = (φσ µ χ) (ψσµ )α̇ . (2.39)
2
As some people might be more familiar with four component spinor notation,
let us close this section by briefly mentioning the connection with Dirac spinors. In
the Weyl representation Dirac matrices read
! !
µ
0 σ 1 0
γµ = , γ5 = iγ0 γ1 γ2 γ3 = (2.40)
σ̄ µ 0 0 −1

30
and a Dirac spinor is
!    
ψα 1 1
ψ= implying r(ψ) = , 0 ⊕ 0, . (2.41)
χ α̇ 2 2

This shows that a Dirac spinor carries a reducible representation of the Lorentz
algebra. Using this four component spinor notation one sees that
! !
ψα 0
and (2.42)
0 χ α̇

are Weyl (chiral) spinors, with chirality +1 and −1, respectively. One can easily
show that a Majorana spinor (ψ C = ψ) is a Dirac spinor such that χα = ψα . To
prove this, just recall that in four component notation the conjugate Dirac spinor
T
is defined as ψ = ψ † γ0 and the charge conjugate is ψ C = Cψ with the charge
conjugate matrix in the Weyl representation being
!
−αβ 0
C= . (2.43)
0 −α̇β̇

Finally, Lorentz generators are

!
i 1 1 σ µ σ̄ ν − σ ν σ̄ µ 0
Σµν = γ µν , γ µν = (γ µ γ ν − γ ν γ µ ) = (2.44)
2 2 2 0 σ̄ σ − σ̄ ν σ µ
µ ν

while the 2-index Pauli matrices are defined as

1 µ ν γ̇β
1  µ α̇γ ν 
(σ µν )αβ = σαγ̇ (σ̄ ) − (µ ↔ ν) , (σ̄ µν )α̇β̇ = (σ̄ ) σγ β̇ − (µ ↔ ν) (2.45)
4 4
From the last equations one then sees that iσ µν acts as a Lorentz generator on ψα
α̇
while iσ̄ µν acts as a Lorentz generator on ψ .

2.3 The supersymmetry algebra

As we have already mentioned, a no-go theorem provided by Coleman and Mandula
implies that, under certain assumptions (locality, causality, positivity of energy,
finiteness of number of particles, etc...), the only possible symmetries of the S-matrix
are, besides C, P, T

• Poincaré symmetries, with generators Pµ , Mµν

31
 • An internal symmetry group G with generators Bl being Lorentz scalars and
with the structure of a compact semi-simple group times U (1) factors.

The full symmetry algebra hence reads

[Pµ , Pν ] = 0 (2.46)

[Mµν , Mρσ ] = −iηµρ Mνσ − iηνσ Mµρ + iηµσ Mνρ + iηνρ Mµσ (2.47)

[Mµν , Pρ ] = −iηρµ Pν + iηρν Pµ (2.48)

[Bl , Bm ] = iflmn Bn (2.49)

[Pµ , Bl ] = 0 (2.50)

[Mµν , Bl ] = 0 , (2.51)

where flmn are structure constants and the last two commutation relations simply
say that the algebra is the direct sum of the Poincaré algebra and the algebra G
spanned by the scalar bosonic generators Bl , that is

ISO(1, 3) × G , (2.52)

at the level of groups.

The Coleman-Mandula theorem can be evaded by weakening one (or more) of
its assumptions. The theorem assumes, in particular, that the symmetry algebra
involves only commutators but there are not any specific physical requirements for
this to be needed. Haag, Lopuszanski and Sohnius generalized the notion of Lie
algebra to include algebraic systems involving, in addition to commutators, also
anticommutators. This extended Lie algebra goes under the name of graded Lie
algebra. Allowing for a graded Lie algebra weakens the Coleman-Mandula theorem
enough to allow for supersymmetry, which is nothing but a specific graded Lie
algebra.
Let us first define what a graded Lie algebra is. Recall that a Lie algebra is a
vector space (over some field, say R or C) which enjoys an additional composition
rule, called product
[ , ]:L×L→L , (2.53)

32
with the following properties

[v1 , v2 ] ∈ L

[v1 , (v2 + v3 )] = [v1 , v2 ] + [v1 , v3 ]

[v1 , v2 ] = −[v2 , v1 ]

[v1 , [v2 , v3 ]] + [v2 , [v3 , v1 ]] + [v3 , [v1 , v2 ]] = 0 ,

where vi are elements of the algebra. A graded Lie algebra of grade n is a vector
space
i=n
L = ⊕i=0 Li (2.54)
where Li are all vector spaces, and the product

[ , }:L×L→L (2.55)

has the following properties

[Li , Lj } ∈ Li+j mod n + 1

[Li , Lj } = −(−1)ij [Lj , Li }

[Li , [Lj , Lk }}(−1)ik + [Lj , [Lk , Li }}(−1)ij + [Lk , [Li , Lj }}(−1)jk = 0 .

From the first such properties it follows that L0 is a Lie algebra while all other Li ’s
with i 6= 0 are not. The second property is called supersymmetrization while the
third one is nothing but the generalization to a graded algebra of the well known
Jacobi identity any algebra satisfies.
The supersymmetry algebra is a graded Lie algebra of grade one, namely

L = L0 ⊕ L1 , (2.56)
I
where L0 is the Poincaré algebra and L1 = (QIα , Qα̇ ) with I = 1, . . . , N , wherre
I
QIα , Qα̇ is a set of N + N = 2N anticommuting fermionic generators transforming
in the representations ( 12 , 0) and (0, 21 ) of the Lorentz group, respectively. Haag,
Lopuszanski and Sohnius proved that this is the only possible consistent extension
of the Poincaré algebra, given the other (very physical) assumptions one would not
like to relax of the Coleman-Mandula theorem. For instance, generators with spin
higher than one, like those transforming in representation ( 12 , 1), cannot be there.
The generators of L1 are spinors and hence they transform non-trivially under
the Lorentz group. Therefore, supersymmetry is not an internal symmetry. Rather

33
it is an extension of Poincaré space-time symmetries. Moreover, acting on bosons,
the supersymmetry generators transform them into fermions (and viceversa). Hence,
this symmetry naturally mixes radiation with matter.
The supersymmetry algebra, besides the commutators (2.46)-(2.51), contains the
following (anti)commutators
 
Pµ , QIα = 0 (2.57)
h I
i
Pµ , Qα̇ = 0 (2.58)
 
Mµν , QIα = i(σµν )αβ QIβ (2.59)
h I α̇
i I β̇
Mµν , Q = i(σ̄µν )α̇β̇ Q (2.60)
n J
o
QIα , Qβ̇ = 2σαµβ̇ Pµ δ IJ (2.61)
 I J
Qα , Qβ = αβ Z IJ , Z IJ = −Z JI (2.62)
n I Jo
Qα̇ , Qβ̇ = α̇β̇ (Z IJ )∗ (2.63)

Let us discuss a bit the above structure.

I
• Eqs. (2.59) and (2.60) follow from the fact that QIα and Qα̇ are spinors of
the Lorentz group, recall eq.(2.45). From these same equations, recalling that
M12 = J3 , one also sees that
  1   1
J3 , QI1 = QI1 , J3 , QI2 = − QI2 . (2.64)
2 2
Taking the hermitian conjugate of the above relations we get
h I
i 1 I h I
i 1 I
J3 , Q1̇ = − Q1̇ , J3 , Q2̇ = Q2̇ (2.65)
2 2
I
and so we see that QI1 and Q2̇ rise the z-component of the spin by half unit
I
while QI2 and Q1̇ lower it by half unit.

• Eq.(2.61) has a very important implication. First notice that given the trans-
J
formation properties of QIα and Qβ̇ under Lorentz transformations, their an-
ticommutator should be symmetric under I ↔ J and should transform as
     
1 1 1 1
, 0 ⊗ 0, = , . (2.66)
2 2 2 2

34
 The obvious such candidate is Pµ which is the only generator in the algebra
with such transformation properties (the δ IJ in eq. (2.61) is achieved by diag-
onalizing an arbitrary symmetric matrix and rescaling the Q’s and the Q’s).
Hence, the commutator of two supersymmetry transformations is a translation.
In theories with local supersymmetry (i.e. where the spinorial infinitesimal pa-
rameter of the supersymmetry transformation depends on xµ ), the commutator
is an infinitesimal translation whose parameter depends on xµ . This is noth-
ing but a theory invariant under general coordinate transformation, namely a
theory of gravity! The upshot is that theories with local supersymmetry auto-
matically incorporate gravity, the two things are tight together. Such theories
are called supergravity theories, SUGRA for short.

• Eqs.(2.57) and (2.58) are not at all obvious. Compatibility with Lorentz sym-
metry would imply the right hand side of eq. (2.57) to transform as
       
1 1 1 1 1
, ⊗ , 0 = 0, ⊕ 1, , (2.67)
2 2 2 2 2
and similarly for eq. (2.58). The second term on the right hand side cannot
be there, due to the theorem of Haag, Lopuszanski and Sohnius which says
that the only allowed fermionic generators in the algebra are supersymmetry
generators, which are spin 12 . In other words, there cannot be a consistent
extension of the Poincaré algebra including generators transforming in the
(1, 12 ) under the Lorentz group. Still, group theory arguments by themselves
do not justify eqs.(2.57) and (2.58) but rather something like
  J β̇
Pµ , QIα = C IJ σµ αβ̇ Q (2.68)
h I
i
Pµ , Qα̇ = (C IJ )∗ σ̄µ α̇β QJβ . (2.69)

where C IJ is an undetermined matrix. We want to prove that this matrix

vanishes. Let us first consider the generalized Jacobi identity which the super-
symmetry algebra should satisfy and let us apply it to the (Q, P, P ) system.
We get
 I       
Qα , Pµ , Pν + [Pµ , Pν ] , QIα + Pν , QIα , Pµ =
h J β̇ i h J β̇ i
−C IJ σµ αβ̇ Q , Pν + C IJ σν αβ̇ Q , Pµ =

C IJ C JK∗ σµ αβ̇ σ̄νβ̇γ QKγ − C IJ C JK∗ σν αβ̇ σ̄µβ̇γ QKγ =

4 (C C ∗ )IK (σµν )αγ QKγ = 0 .

35
 This implies that
C C∗ = 0 , (2.70)
as a matrix equation. Note that this is not enough to conclude, as we would,
that C = 0. For that, we also need to show, in addition, that C is symmetric.
To this aim we have to consider other equations, as discussed below.

• Let us now consider eqs. (2.62) and (2.63). As for the first, from Lorentz
representation theory we would expect
   
1 1
,0 ⊗ , 0 = (0, 0) ⊕ (1, 0) , (2.71)
2 2

which explicitly means


QIα , QJβ = αβ Z IJ + βγ (σ µν )αγ Mµν Y IJ . (2.72)

The Z IJ , being Lorentz scalars, should be some linear combination of the

internal symmetry generators Bl and, given the antisymmetric properties of
the epsilon tensor under α ↔ β, they should be anti-symmetric under I ↔ J.
On the contrary, given that βγ (σ µν )αγ is symmetric in α ↔ β, the matrix Y IJ
should be symmetric under I ↔ J.
Let us now consider the generalized Jacobi identity between (Q, Q, P ), which
can be written as

[{QIα , QJβ }, Pµ ] = {QIα , [QJβ , Pµ ]} + {QJβ , [QIα , Pµ ]} . (2.73)

If one multiplies it by αβ , only the anti-symmetric part under α ↔ β of the

left hand side survives, which, from eq. (2.72), can be seen to vanish since the
matrix ZIJ commutes with Pµ . So we get

0 = αβ {QIα , [QJβ , Pµ ]} + αβ {QJβ , [QIα , Pµ ]}

β̇ β̇
= αβ CI K σµ β β̇ {QJα , QK } − αβ CJK σµ β β̇ {QIα , QK } ∼ (CIJ − CJI ) σ̄µγ̇α σ ναγ̇ Pν

= 2 (CIJ − CJI )Pµ ,

which implies that the matrix C is symmetric. So the previously found equa-
tion C C ∗ = 0 can be promoted to CC † = 0, which in turn implies C = 0 and
hence eq. (2.57). A similar rationale leads to eq. (2.58).

36
 Let us now come back to eq. (2.62), which we have not yet proven. To do so,
we should plug eq. (2.57) into the Jacobi identity (2.73), getting

[{QIα , QJβ }, Pµ ] = 0 , (2.74)

which implies, by (2.72), that the matrix Y IJ vanishes because Pµ does not
commute with Mµν . This finally proves eq. (2.62). Similarly, one can prove
eq. (2.63), which is just the hermitian conjugate of (2.62).

What about the commutation relations between supersymmetry generators and in-
ternal symmetry generators, if any? In general, the Q’s will carry a representation
of the internal symmetry group G. So one expects something like
 I 
Qα , Bl = (bl )IJ QJα (2.75)
 
QI α̇ , Bl = −QJ α̇ (bl )JI . (2.76)

The second commutation relation comes from the first under hermitian conjugation,
recalling that the bl are hermitian, because so are the generators Bl . The largest
possible internal symmetry group which can act non trivially on the Q’s is thus
U (N ), and this is called the R-symmetry group (the relation between a Lie algebra
with generators S and the corresponding Lie group with elements U is U = eiS ;
hence hermitian generators, S † = S, correspond to unitary groups, U † = U −1 ).
In presence of non-vanishing central charges one can prove that the R-symmetry
group reduces to U Sp(N ), the compact version of the symplectic group Sp(N ),
U Sp(N ) ∼= U (N ) ∩ Sp(N ).
As already noticed, the operators ZIJ are Lorentz scalars and should then cor-
respond to some linear combination of the internal symmetry group generators Bl
of the compact Lie algebra G, say

Z IJ = al|IJ Bl . (2.77)

Using the above equation, the supersymmetry algebra (2.57)-(2.63) and eqs. (2.75)
and (2.76) one can actually prove that the Z’s are central charges, that is they
commute with the whole supersymmetry algebra, and within themselves. Contrary
to what one could naively think, this does not imply they are ineffective. Indeed,
central charges are not numbers but quantum operators and their value may vary
from state to state. For a supersymmetric vacuum state, which is annihilated by all
supersymmetry generators, they are trivially realized, recall eqs. (2.62) and (2.63).

37
However, they do not need to vanish in general. For instance, as we will see in the
subsequent chapter, massive representations are very different if Z IJ vanishes or if
it is non-trivially realized on the representation.
Let us end this section with a few more comments. First, if N = 1 there are only
two supersymmetry generators, which correspond to one Majorana spinor in four
component notation. In this case we speak of unextended or minimal supersym-
metry (and there are no non-trivial central charges). For N > 1 we have extended
supersymmetry (and there can now be a central extension of the supersymmetry
algebra). From an algebraic point of view there is no limit to N . However, as we
will see later, increasing N the theory must contain particles of increasing spin. In
particular we have

• N ≤ 4 for theories without gravity (spin ≤ 1)

• N ≤ 8 for theories with gravity (spin ≤ 2)

Therefore, to avoid theories with spin higher than two (that is focusing on local,
interacting theories) N = 8 is an upper bound. As discussed in the previous chap-
ter, thus stated this statement is true in four space-time dimensions. Equivalent
statements can be made in higher/lower dimensions, where the dimension of the
spinor representation of the Lorentz group is bigger/smaller. What matters is the
number of single state supersymmetry generators, which is a dimension-independent
statement (e.g., N = 8 corresponds to 32 supercharges).
For N = 1 the R-symmetry group is just U (1) (one can see it from the Jacobi
identity between (Q, B, B) which implies that the flmn are trivially realized on the
supersymmetry generators). In this case the hermitian matrices bl are just real
numbers and by rescaling the generators Bl one gets

[R, Qα ] = −Qα , [R, Qα̇ ] = +Qα̇ . (2.78)

This implies that in minimal supersymmetric models, supersymmetric partners

(which are related by the action of the Q’s) have different U (1) R-charge. In par-
ticular, given eqs. (2.78), if a particle has R = 0 then its superpartner has R = ±1.
An important physical consequence of this property is that in a theory where the
R-symmetry is preserved, the lightest supersymmetric particle (LSP) is stable.
Let us finally comment on the relation between two and four component spinor
notations, when it comes to supersymmetry. In four component notation the 2N

38
 I
supersymmetry generators QIα , Qα̇ constitute a set of N Majorana spinors
!
QI
I
 
α I
QI = I α̇ Q = Q Iα
Q α̇ (2.79)
Q

and the supersymmetry algebra reads

J
{QI , Q } = 2δ IJ γ µ Pµ − i I ImZ IJ − γ5 ReZ IJ
i
[QI , Pµ ] = 0 [QI , Mµν ] = γµν QI [QI , R] = iγ5 QI . (2.80)
2
Depending on what one needs to do, one notation can be more useful than the other.
In the following, we will stick to two component spinor notation, unless otherwise
stated.

2.4 Exercises
1. Prove the following spinor identities
1 1
ψ α ψ β = − αβ ψψ , (θφ) (θψ) = − (φψ) (θθ)
2 2
χσ µ ψ = −ψσ̄ µ χ , χσ µ σ̄ ν ψ = ψσ ν σ̄ µ χ

(χσ µ ψ)† = ψσ µ χ , (χσ µ σ̄ ν ψ)† = ψσ̄ ν σ µ χ

1


1


(θψ) θσ µ φ = − (θθ) ψσ µ φ , θψ θσ̄ µ φ = − θθ ψσ̄ µ φ
2 2
1
(φψ) · χα̇ = (φσ µ χ) (ψσµ )α̇ .
2

2. The operators Z IJ are linear combinations of the internal symmetries gener-

ators Bl , eq. (2.77). Hence, they commute with Pµ and Mµν . Prove that ZIJ
are in fact central charges of the supersymmetry algebra, namely that it also
holds that
 IJ      h K
i
Z , Bl = 0 , Z IJ , Z KL = 0 , Z IJ , QKα = 0 , Z IJ
, Qα̇ = 0 .

References

[1] A. Bilal, Introduction to supersymmetry, Section 2, arXiv:hep-th/0101055.

39
[2] H.J.W. Müller-Kirsten, A. Wiedemann, Introduction to Supersymmetry, Chap-
ter 1, Singapore: World Scientific (2010).

[3] L. Castellani, R. D’Auria and P. Fre, Supergravity And Superstrings: A Geo-

metric Perspective. Vol. 1: Mathematical Foundations, Chapter II.2, Singapore:
World Scientific (1991).

[4] J. D. Lykken, Introduction to Supersymmetry, Appendix, TASI 96, arXiv:hep-

th/9612114

[5] M.F. Sohnius, Introducing Supersymmetry, Section 3, Phys. Rep. 128 (1985).

[6] S. Weinberg, The quantum theory of fields. Vol. 3: Supersymmetry, Chapter

24.B, Cambridge University Press (2000).

40
3 Representations of the supersymmetry algebra

The goal of this chapter is to construct representations of the supersymmetry

algebra. Let us first recall how things go for the Poincaré algebra. The Poincaré
algebra (2.25) has two Casimir operators (i.e. two operators which commute with
all generators)
P 2 = Pµ P µ and W 2 = Wµ W µ , (3.1)
where W µ = 21 µνρσ Pν Mρσ is the so-called Pauli-Lubanski vector. Casimir operators
are useful to classify irreducible representations of a group. In the case of the
Poincaré group such representations are nothing but what we usually call particles.
Let us see how this is realized for massive and massless particles, respectively.
Let us first consider a massive particle with mass m and go to the rest frame,
Pµ = (m, 0, 0, 0). In this frame it is easy to see that the two Casimir reduce to
P 2 = m2 and W 2 = −m2 j(j + 1) where j is the spin. The second equality can be
proven by noticing that Wµ P µ = 0 which implies that in the rest frame W0 = 0.
Therefore, in the rest frame Wµ = (0, 21 i0jk m M jk ) from which one immediately
gets W 2 = −m2 J~2 . So we see that massive particles are distinguished by their mass
and their spin.
Let us now consider massless particles. In the rest frame Pµ = (E, 0, 0, E). In
this case we have that P 2 = 0 and W 2 = 0, and W µ = M12 P µ . In other words, the
two operators are proportional for a massless particle, the constant of proportionality
being M12 = ±j, the helicity. For these representations the spin is then fixed and
the different states are distinguished by their energy and by the sign of the helicity
(e.g. the photon is a massless particle with two helicity states, ±1).
Now, as a particle is an irreducible representation of the Poincaré algebra, we call
superparticle an irreducible representation of the supersymmetry algebra. Since the
Poincaré algebra is a subalgebra of the supersymmetry algebra, it follows that any
irreducible representation of the supersymmetry algebra is a representation of the
Poincaré algebra, which in general will be reducible. This means that a superparticle
corresponds to a collection of particles, the latter being related by the action of the
I
supersymmetry generators QIα and Qα̇ and having spins differing by units of half.
Being a multiplet of different particles, a superparticle is often called supermultiplet.
Before discussing in detail specific representations of the supersymmetry algebra,
let us list three generic properties which any of such representations enjoy, all of them
having very important physical implications.

41
1. As compared to the Poincaré algebra, in the supersymmetry algebra P 2 is
still a Casimir, but W 2 is not anymore (this follows from the fact that Mµν
does not commute with the supersymmetry generators). Therefore, particles
belonging to the same supermultiplet have the same mass and different spin,
since the latter is not a conserved quantum number of the representation. The
mass degeneracy between bosons and fermions is something we do not observe
in known particle spectra; this implies that supersymmetry, if at all realized,
must be broken in Nature.
Note: what’s above is true in Minkowski space. If one wants to discuss super-
symmetry in, e.g. anti-de Sitter space, things are different. In anti-de Sitter
space the generators Pµ do not commute with Lorentz generators, nor with
the supercharges. The consequence is that P 2 is not anymore a Casimir but
rather C = M 2 + αP 2 is, with α a dimension-full quantity proportional to the
anti-de Sitter radius squared. So, in anti-de Sitter space states belonging to
the same multiplet do have the same C-eigenvalue but different P 2 eigenvalue,
so different masses.

2. In a supersymmetric theory the energy of any state is always ≥ 0. Consider

an arbitrary state |φi. Using the supersymmetry algebra, we easily get
n I
o
hφ| QIα , Qα̇ |φi = 2σαµα̇ hφ|Pµ |φi δ II
 I 

Qα̇ = (QIα )† = hφ| QIα (QIα )† + (QIα )† QIα |φi

= ||(QIα )† |φi||2 + ||QIα |φi||2 ≥ 0 .

The last inequality follows from positivity of the Hilbert space. Summing over
α = α̇ = 1, 2 and recalling that Tr σ µ = 2δ µ0 we get

4 hφ|P0 |φi ≥ 0 , (3.2)

as anticipated.

3. A supermultiplet contains an equal number of bosonic and fermionic d.o.f.,

nB = nF . Define a fermion number operator
(
−1 fermionic state
(−1)NF = (3.3)
+1 bosonic state

42
 NF can be taken to be twice the spin, NF = 2s. Such an operator, when acting
on a bosonic, respectively a fermionic state, gives indeed

(−1)NF |Bi = |Bi , (−1)NF |F i = −|F i . (3.4)

We want to show that Tr (−1)NF = 0 if the trace is taken over a finite dimen-
sional representation of the supersymmetry algebra. First notice that

QIα , (−1)NF = 0 → QIα (−1)NF = −(−1)NF QIα . (3.5)

Using this property and the cyclicity of the trace one easily sees that
 J J

0 = Tr −QIα (−1)NF Qβ̇ + (−1)NF Qβ̇ QIα
 n J
o  
= Tr (−1) NF I
Qα , Qβ̇ = 2σαµβ̇ Tr (−1)NF Pµ δ IJ .

Summing on I, J and choosing any Pµ 6= 0 it follows that Tr (−1)NF = 0,

which implies that nB = nF .

In the following, we discuss (some) representations in detail. Since the mass is

a conserved quantity in a supermultiplet, it is meaningful distinguishing between
massless and massive representations. Let us start from the former.

3.1 Massless supermultiplets

Let us first assume that all central charges vanish, i.e. Z IJ = 0 (we will see later that
this is the only relevant case, for massless representations). Notice that in this case
it follows from eqs. (2.62) and (2.63) that all Q’s and all Q’s anticommute among
themselves. The steps to construct the irreps are as follows:

1. Go to the rest frame where Pµ = (E, 0, 0, E). In such frame we get

!
0 0
σ µ Pµ = (3.6)
0 2E

Plugging this into eq. (2.61) we get

!
n J
o 0 0 n J
o
QIα , Qβ̇ = δ IJ −→ QI1 , Q1̇ = 0 . (3.7)
0 4E
αβ̇

43
 I
Due to the positiveness of the Hilbert space, this implies that both QI1 and Q1̇
are trivially realized. Indeed, from the equation above we get
n I
o I
0 = hφ| QI1 , Q1̇ |φi = ||QI1 |φi||2 + ||Q1̇ |φi||2 , (3.8)

I I
whose only solution is QI1 = Q1̇ = 0. We are then left with just QI2 and Q2̇ ,
hence only half of the generators.

2. From the non-trivial generators we can define

1 1 I
aI ≡ √ QI2 , a†I ≡ √ Q2̇ . (3.9)
4E 4E
These operators satisfy the anticommutation relations of a set of N creation
and N annihilation operators
n o n o
† IJ † †
aI , aJ = δ , {aI , aJ } = 0 , aI , aJ = 0 . (3.10)

These are the basic tools we need in order to construct representations of the
supersymmetry algebra. Recall that in our conventions the operators QI2 and
I
Q2̇ (and hence aI and a†I ) lower respectively raise the helicity of half unit on
the state they act on.

3. To construct a representation, one can start by choosing a state annihilated

by all aI ’s (known as the Clifford vacuum): such state will carry some irrep
of the Poincaré algebra. Besides having m = 0, it will carry some helicity λ0 ,
and we call it |E, λ0 i (|λ0 i for short). For this state

aI |λ0 i = 0 . (3.11)

Note that this state can be either bosonic or fermionic, and should not be
confused with the actual vacuum of the theory, which is the state of minimal
energy: the Clifford vacuum is a state with quantum numbers (E, λ0 ) and
which satisfies eq. (3.11).

4. The full representation (aka supermultiplet) is obtained acting on |λ0 i with

the creation operators a†I as follows
1
|λ0 i , a†I |λ0 i ≡ |λ0 + iI , a†I a†J |λ0 i ≡ |λ0 + 1iIJ ,
2
N
. . . , a†1 a†2 . . . a†N |λ0 i ≡ |λ0 + i .
2
44
 Hence, starting from a Clifford vacuum with helicity λ0 , the state with highest
helicity in the representation has helicity λ = λ0 + N2 . Due to the antisymmetry
in I ↔ J, at helicity level λ = λ0 + k2 we have
!
k N
# of states with helicity λ0 + = , (3.12)
2 k

where k = 0, 1, . . . , N . The total number of states in the irrep will then be

!
XN
N


= 2N = 2N −1 B + 2N −1 F , (3.13)
k=0
k

half of them having integer helicity (bosons), half of them half-integer helicity
(fermions).

5. CPT flips the sign of the helicity. Therefore, unless the helicity is distributed
symmetrically around 0, which is not the case in general, a supermultiplet is
not CPT-invariant. This means that in order to have a CPT-invariant the-
ory one should in general double the supermultiplet we have just constructed
adding its CPT conjugate. This is not needed if the supermultiplet is self-
CPT conjugate, which can happen only if λ0 = − N4 (in this case the helicity
is indeed distributed symmetrically around 0).

Let us now apply the above procedure and construct several (physically inter-
esting) irreps of the supersymmetry algebra.

N = 1 supersymmetry

• Matter multiplet (aka chiral multiplet):

   
1 1
λ0 = 0 → 0, + ⊕ − ,0 . (3.14)
2 CP T 2

The degrees of freedom of this representation are those of one Weyl fermion
and one complex scalar (on shell; recall we are constructing supersymmetry
representations on states!). Note that since the two representations above
are exchanged by CPT, the two spin 0 states have opposite parity, so if one
corresponds to a scalar the other is a pseudoscalar. In a N = 1 supersymmetric
theory this is the representation where matter sits; this is why such multiplets
are called matter multiplets.

45
 • Gauge (or vector) multiplet:
   
1 1 1
λ0 = → + , +1 ⊕ −1, − . (3.15)
2 2 CP T 2
The degrees of freedom are those of one vector and one Weyl fermion. This
is the representation one needs to describe gauge fields in a supersymmetric
theory. Notice that since internal symmetries (but the R-symmetry) commute
with the supersymmetry algebra, the representation the Weyl fermion should
transform under gauge transformations should be the same as the vector field,
i.e. the adjoint. Hence, usual Standard Model matter (quarks and leptons)
cannot be accommodated in these multiplets.

Although in this course we will focus on rigid supersymmetry and hence not con-
sider supersymmetric theories with gravity, let us list for completeness (and future
reference) also representations containing states with higher helicity.

• Spin 3/2 multiplet:

   
3 3
λ0 = 1 → 1, + ⊕ − , −1 . (3.16)
2 CP T 2

The degrees of freedom are those of a spin 3/2 particle and one vector.

• Graviton multiplet:
   
3 3 3
λ0 = → + , +2 ⊕ −2, − . (3.17)
2 2 CP T 2
The degrees of freedom are those of a graviton, which has helicity 2, and a
particle of helicity 3/2, known as the gravitino (the supersymmetric partner
of the graviton).

Representations constructed from a Clifford vacuum with higher helicity will in-
evitably include states with helicity higher than 2. Hence, if one is interested in
interacting local field theories, the story stops here. Recall that in an interacting
local field theory massless particles with helicity higher than 12 should couple to con-
served quantities at low momentum. The latter are: conserved internal symmetry
generators for (soft) massless vectors, supersymmetry generators for (soft) gravitinos
and four-vector Pµ for (soft) gravitons. The supersymmetry algebra does not allow
for generators other than these ones. Hence, supermultiplets with helicity λ ≥ 25 are

46
ruled out: they may exist, but they cannot have couplings that survive in the low
energy limit.
The above discussion also implies that in a local interacting field theory a spin
3/2 particle is inevitably associated to local supersymmetry and hence, in turn,
with gravity. Therefore, there is no much meaning for a theory without the graviton
multiplet and a spin 3/2 multiplet, which would be a non-interacting one in fact. In
other words, the physical gravitino is the one sitting in the graviton multiplet.

N = 2 supersymmetry

• Matter multiplet (aka hypermultiplet):

   
1 1 1 1 1
λ0 = − → − , 0, 0, + ⊕ − , 0, 0, + . (3.18)
2 2 2 CP T 2 2

The degrees of freedom are those of two Weyl fermions and two complex
scalars. This is where matter sits in a N = 2 supersymmetric theory. In
N = 1 language this representation corresponds to two chiral multiplets with
opposite chirality (CPT flips the chirality).
Note: in principle this representation enjoys the necessary condition to be
CPT self-conjugate, λ0 = − N4 . However, a closer look shows that an hyper-
multiplet cannot be self-conjugate (that’s why we added the CPT conjugate
representation). The way the various states are constructed out of the Clifford
vacuum shows that under the compact part of the R-symmetry group, SU (2),
the helicity 0 states behave as a doublet while the fermionic states are sin-
glets. If the representation were CPT self-conjugate the two scalar degrees of
freedom would have been both real. Such states cannot form a SU (2) doublet
since a two-dimensional representation of SU (2) is pseudoreal, and hence the
doublet should be complex.

• Gauge (or vector) multiplet:

   
1 1 1 1
λ0 = 0 → 0, + , + , +1 ⊕ −1, − , − , 0 . (3.19)
2 2 CP T 2 2

The degrees of freedom are those of one vector, two Weyl fermions and one
complex scalar. In N = 1 language this is just a vector and a matter multiplet
(both transforming in the adjoint representation of the gauge group).

47
 • Spin 3/2 multiplet:
   
3 3 1 1 3
λ0 = − → − , −1, −1, − ⊕ + , +1, +1, + . (3.20)
2 2 2 CP T 2 2

The degrees of freedom are those of a spin 3/2 particle, two vectors and one
Weyl fermion.

• Graviton multiplet:
   
3 3 3 3
λ0 = −2 → −2, − , − , −1 ⊕ +1, + , + , +2 . (3.21)
2 2 CP T 2 2
The degrees of freedom are those of a graviton, two gravitini and a vector,
which is usually called graviphoton in the supergravity literature.

N = 4 supersymmetry

• Gauge (or vector) multiplet:

 
1 1
λ0 = −1 → −1, 4 × − , 6 × 0, 4 × + , +1 . (3.22)
2 2
The degrees of freedom are those of a vector, four Weyl fermions and three
complex scalars. In N = 1 language this corresponds to one vector multiplet
and three matter multiplets (all transforming in the adjoint). Notice that
this multiplet is CPT self-conjugate. This time there are no issues with R-
symmetry transformations. The vector is a singlet under SU (4), fermions
transform under the fundamental representation, and scalars under the two
times anti-symmetric representation, which is the fundamental of SO(6), and
is real. The fact that the representation under which scalars transform is real
also explains why for N = 4 supersymmetry, the R-symmetry group is not
U (4) but just SU (4).

For N = 4 it is not possible to have matter in the usual sense, since the number
of supersymmetry generators is too high to avoid helicity one states. Therefore,
N = 4 supersymmetry cannot accommodate fermions transforming in fundamental
representations. Besides the vector multiplet there are of course also representations
with higher helicity, but we refrain to report them here.
One might wonder why we did not discuss N = 3 representations. This is just
because as far as non-gravitational theories are concerned, N = 3 and N = 4 are

48
physically equivalent: when constructing N = 3 representations with maximal spin
one, once the CPT conjugate representation is added (in this case we cannot satisfy
the condition λ0 = − N4 ) one ends up with a multiplet which is the same as the
N = 4 vector multiplet. N = 4 and N = 3 differ only for representations including
states with spin higher than one.

N > 4 supersymmetry
In this case one can easily get convinced that it is not possible to avoid gravity
since there do not exist representations with helicity smaller than 32 when N > 4.
Hence, theories with N > 4 are all supergravity theories. It is interesting to note
that N = 8 supergravity allows only one possible representation with highest helic-
ity smaller than 52 and that for higher N one cannot avoid states with helicity 25 or
higher. Therefore, N = 8 is an upper bound on the number of supersymmetry gen-
erators, as far as interacting local field theories are concerned. Beware: as stated, all
these statements are valid in four space-time dimensions. The way to count super-
symmetries depends on the dimension of space-time, since spinorial representations
get bigger, the more the dimensions. Obviously, completely analogous statements
can be made in higher dimensions. For instance, in ten space-time dimensions the
maximum allowed supersymmetry to avoid states with helicity 52 or higher is N = 2.
A dimension-independent statement can be made counting the number of single
component supersymmetry generators. Using this language, the maximum allowed
number of supersymmetry generators for non-gravitational theories is 16 (which is
indeed N = 4 in four dimensions) and 32 for theories with gravity (which is N = 8
in four dimensions).
The table below summarizes all results we have discussed.
1 3
N λmax = 1 λmax = 2 λmax = 2 λmax = 2

 
8 none none (2), 8( 32 ), 28(1), 56( 12 ), 70(0) none
   3 
6 none none (2), 6( 32 ), 16(1), 26( 12 ), 30(0) ( 2 ), 6(1), 15( 12 ), 20(0)
   3 
5 none none (2), 5( 32 ), 10(1), 11( 12 ), 10(0) ( 2 ), 6(1), 15( 12 ), 20(0)
     3 
4 (1), 4( 21 ), 6(0) none (2), 4( 23 ), 6(1), 4( 21 ), 2(0) ( 2 ), 4(1), 7( 12 ), 8(0)
     3 
3 (1), 4( 21 ), 6(0) none (2), 3( 23 ), 3(1), ( 12 ) ( 2 ), 3(1), 3( 12 ), 2(0)
   1     3 
2 (1), 2( 21 ), 2(0) 2( 2 )4(0) (2), 2( 32 ), (1) ( 2 ), 2(1), ( 12 )
   1     3 
1 (1), ( 21 ) ( 2 )2(0) (2), ( 32 ) ( 2 ), (1)

49
The numbers in parenthesis represent the helicity, while others represent the multi-
plicity of states with given helicity. Notice that, as anticipated, any supermultiplet
contains particles with spin at least as large as 14 N . The N = 7 theory has not been
reported since that allows only the graviton multiplet which, once CPT invariance
is required, is identical to the N = 8 graviton multiplet. In other words, at the
interacting level, N = 7 supergravity is the same as N = 8 supergravity (this is the
same argument we used for not discussing N = 3 representations as far as maximal
spin one is concerned).

A final, very important comment regards chirality. The Standard Model is a

chiral theory, in the sense that there exist particles in the spectrum whose chiral
and anti-chiral components transform differently under the gauge group (weak in-
teractions are chiral). When it comes to supersymmetric extensions, it is easy to
see that only N = 1 theories allow for chiral matter. That N = 1 irreps can be
chiral is obvious: Wess-Zumino multiplets contain one single Weyl fermion. There-
fore, in N = 1 supersymmetric extensions of the Standard Model one can accom-
modate left and right components of leptons and quarks in different multiplets,
which can then transform differently under the SU (2) gauge group. What about
N > 1 supersymmetry? First notice that all helicity 21 states belonging to multi-
plets containing vector fields should transform in the adjoint representation of the
gauge group, which is real. Therefore, the only other representation which might
allow for helicity 21 states transforming in fundamental representations is the N = 2
hypermultiplet. However, as already noticed, a hypermultiplet contains two Wess-
Zumino multiplets with opposite chirality. Since for any internal symmetry group G,
we have that the corresponding algebra commutes with the SuperPoincaré algebra,
[G, SuperPoincaré] = 0, these two Wess-Zumino multiplets transform in the same
representation under G. Therefore, N = 2 is non-chiral: left and right components
of leptons and quarks would belong to the same matter multiplet and could not
transform differently under the SU (2) Standard Model gauge group. Summarizing,
if extended supersymmetry is realized in Nature, it should be broken at some high
enough energy scale to an effective N = 1 model. This is why at low energy people
typically focus just on N = 1 extensions of the Standard Model.

50
3.2 Massive supermultiplets
The logical steps one should follow for massive representations are similar to previous
ones. There is however one important difference. Let us consider a state with mass
m in its rest frame, Pµ = (m, 0, 0, 0). One can easily see that, differently from
the massless case, the number of non-trivial generators gets not diminished: there
remain the full set of 2N creation and 2N annihilation operators. Indeed, eq. (2.61)
is now n o
J
Qα , Qβ̇ = 2m δαβ̇ δ IJ
I
(3.23)
and no supersymmetric generators are trivially realized. This means that, gener-
ically, massive representations are longer than massless ones. Another important
difference is that we better speak of spin rather than helicity, now. A given Clifford
vacuum will be defined by mass m and spin j, with j(j + 1) being the eigenvalue
of J 2 . Hence, the Clifford vacuum will have itself degeneracy 2j + 1 since j3 takes
values from −j to +j.
N = 1 supersymmetry

The annihilation and creation operators, satisfying the usual oscillator algebra,
now read
1 1
a1,2 ≡ √ Q1,2 , a†1,2 ≡ √ Q1̇,2̇ . (3.24)
2m 2m
As anticipated these are twice those for the massless case. Notice that a†1 lowers the
spin by half unit while a†2 raises it. We can now define a Clifford vacuum as a state
with mass m and spin j0 which is annihilated by both a1 and a2 and act with the
creation operators to construct the corresponding massive representations.

• Matter multiplet:  
1 1
j=0 → − , 0, 00 , + . (3.25)
2 2
The number of degrees of freedom is the same as the massless case (but with
no need to add any CPT conjugate, now). It is worth noticing that also in the
present case the two spin 0 states, 00 and 0, as one can easily proof playing a bit
with the operator algebra. Summarizing, the multiplet is made of a massive
complex scalar and a massive Majorana fermion.

• Gauge (or vector) multiplet:

 
1 1 1
j= → −1, 2 × − , 2 × 0, 2 × + , 1 . (3.26)
2 2 2

51
 The degrees of freedom one ends-up with are those of one massive vector, one
massive Dirac fermion and one massive real scalar (recall the comment after
eq. (3.23), which in this case implies that the Clifford vacuum, j = 21 contains
two single particle states, with j3 component |1/2i and | − 1/2i, respectively).
The representation is longer than that of a massless vector supermultiplet, as
expected. Notice, though, that the number of these degrees of freedom is the
same as those of a massless vector multiplet plus one massless matter multiplet.
This is reassuring, since we do not like massive vectors to start with, and only
allow Higgs-like mechanisms to generate masses for vector fields in a unitary
and renormalizable theory. One can generate massive vector multiplets by a
generalization of the Higgs mechanism, in which a massless vector multiplet
eats-up a chiral multiplet while preserving supersymmetry.

Since we cannot really make sense of massive particles with spin higher than one
(and we are not much interested in supergravity theories in this course, anyway),
we stop here and move to extended supersymmetry representations.

Extended supersymmetry

Let us then consider N > 1 and allow also for non-trivial central charges. A
change of basis in the space of supersymmetry generators turns out to be useful for
the following analysis. Since the central charge matrix Z IJ is antisymmetric, with
a U (N ) rotation one can put it in the standard block-diagonal form
 
0 Z1
 
−Z1 0 
 
 0 Z2 
 
 −Z2 0 
 
Z IJ =  (3.27)
 ... ... 
 
 ... ... 
 
 
 0 ZN/2 
−ZN/2 0

52
(we have assumed for simplicity that N is even). One can now define
1

a1α = √ Q1α + αβ (Q2β )†
2
1

b1α = √ Q1α − αβ (Q2β )†
2
1

a2α = √ Q3α + αβ (Q4β )†
2
1

b2α = √ Q3α − αβ (Q4β )†
2
... = ...

... = ...
1 −1 †


aN/2
α = √ QN α + αβ (QN
β )
2
1 N −1 N †


bN/2
α = √ Qα − αβ (Qβ )
2
which satisfy the oscillator algebra
 r
aα , (asβ )† = (2m + Zr ) δrs δαβ
 r s †
bα , (bβ ) = (2m − Zr ) δrs δαβ
 r s † 
aα , (bβ ) = arα , asβ = · · · = 0 .

where r, s = 1, . . . , N/2. As anticipated, we have now 2N creation operators

(arα )† , (brα )† r = 1, . . . , N/2 , α = 1, 2 (3.28)

which we can use to construct massive representations starting from some given
Clifford vacuum. Notice that, from their very definition, it follows that creation
operators with spinorial index α = 1 lower the spin by half unit, while those with
spinorial index α = 2 raise it.
Several important comments are in order. Due to the positiveness of the Hilbert
space, from the oscillator algebra above one can show that
N
2m ≥ |Zr | , r = 1, . . . , . (3.29)
2
This means that the mass of a given representation is always larger (or equal) than
(half) the modulus of any central charge eigenvalue. The first important consequence

53
of the bound (3.29) is that for massless representations (for which the left hand side
is identically 0) the central charges are always trivially realized, i.e. Z IJ = 0. That’s
why we did not discuss massless multiplets with non vanishing central charges in
the previous section.
There is another important consequence of the bound (3.29). Suppose none of
the central charge eigenvalues saturate it, namely 2m > |Zr | , ∀r. Proceeding as
before, starting from a Clifford vacuum λ0 annihilated by all operators arα , brα and
acting on it with the creation operators (3.28) one creates 22N states, 22N −1 bosonic
and 22N −1 fermionic, with spin going from λ0 − N/2 to λ0 + N/2. Therefore, the
representation has states with spins spanning 2N + 1 half-integer values.
Suppose instead that some Zr saturate the bound (3.29), say k ≤ N/2 of them
do so. Looking at the oscillator algebra one immediately sees that k b-type operators
become trivial (we are supposing, without loss of generality, that all Zr are posi-
tive), and the dimension of the representation diminishes accordingly. The multiplet
contains only 22(N −k) states now. These are called short multiplets. The extreme
case is when all Zr saturate the bound (k = N/2). In this case half the creation
operators trivialize and we get a multiplet, known as ultrashort, whose dimension is
identical to that of a massless one: the number of states is indeed 2N , 2N −1 bosonic
and 2N −1 fermionic.
The upshot of the discussion above is that in theories with extended supersym-
metry one can have massive multiplets with different lengths:



long multiplets 22N = 22N −1 B + 22N −1 F

short multiplets 22N −2k = 22(N −k)−1 B
+ 22(N −k)−1 F

N


ultra-short multiplets 22N −2 2 = 2N = 2N −1 B
+ 2N −1 F
.
As it happens for massless representations, states belonging to some representa-
tion of supersymmetry also transform into given representations of the R-symmetry
group, since the supercharges do so. One should just remember that the R-symmetry
group is U (N ) in absence of central charges but reduces to U Sp(N ) if central charges
are present.

N = 2 supersymmetry

In this case we have one only central charge eigenvalue, Z, and we have four
oscillators aα and bα (we have dropped the now inessential upper index r). In

54
the following, the Clifford vacuum will be defined as a state annihilated by all
undaggered operators, unless otherwise stated.
Let us first consider the case of long multiplets, namely a situation in which the
bound (3.29) is not saturated. In this case we cannot have massive matter since
we have too many creation operators to avoid spins higher than 12 . So the only
possibility are (massive) vector multiplets.

• Gauge (or vector) multiplet:

 
1 1
j=0 → −1, 4 × − , 6 × 0, 4 × + , 1 . (3.30)
2 2
The degrees of freedom correspond to a massive vector, two Dirac fermions,
and five real scalars. Their number equals that of a massless N = 2 vector
multiplet and a massless N = 2 hypermultiplet. As before, such massive
vector multiplet should be thought of as obtained via some supersymmetric
Higgs-like mechanism.

Let us now consider shorter representations. Since in this case there is only one
central charge eigenvalue, Z, the only possible short representation is in fact the
ultrashort, whose length should equal that of the corresponding massless represen-
tation. The only non-trivial oscillators are now aα , since the bα are trivially realized
(we are assuming Z > 0)..

• Matter multiplet (short rep.):

 
1 1
j=0 → 2 × − , 4 × 0, 2 × + , (3.31)
2 2

(where the doubling of states arises for similar reasons as for the massless hy-
permultiplet). The degrees of freedom are those of one massive Dirac fermion
and two massive complex scalars. As expected the number of degrees of free-
dom equals those of a massless hypermultiplet.

• Vector multiplet (short rep.):

 
1 1 1
j= → −1, 2 × − , 2 × 0, 2 × + , +1 . (3.32)
2 2 2
The degrees of freedom are those of one massive vector, one massive Dirac
fermion and one massive real scalar. While rearranged differently in terms of

55
 fields, the number of bosonic and fermionic degrees of freedom equals that of
a massless vector multiplet. What’s interesting here is that a massive ultra-
short vector multiplet can arise dynamically, via some Higgs-like mechanism
involving only a massless vector multiplet, something peculiar to N = 2 su-
persymmetry and related to the fact that massless vector multiplets contain
scalars, and can then self-Higgs.

N = 4 supersymmetry

For N = 4 supersymmetry long multiplets are not allowed since the number
of states (actually 256!) would include at least spin 2 states; such a theory would
then include a massive spin 2 particle which is not believed to be possible in a local
quantum field theory. What are are possible are short multiplets, actually only
the ultrashort, whose field content amounts to rearrange the fields characterizing a
massless vector multiplet into massive states: one would get a massive vector, two
Dirac fermions and five real scalars. The construction is left to the reader.
Let us finally notice that all short multiplets are supersymmetry preserving,
meaning they are annihilated by the supersymmetry generators whose correspond-
ing central charge eigenvalue saturates the bound. In general one can then have
, , . . . , N/2
1 2
N N N
supersymmetry preserving multiplets (the numerator is nothing but
the integer k previously defined). For instance, ultrashort multiplets, for which
k = N/2, are 21 supersymmetry preserving states. These multiplets accommodate
states which have very important properties at the quantum level; most notably,
it turns out that they are more protected against quantum corrections with re-
spect to states belonging to long multiplets. Short multiplets are also called BPS,
since the bound (3.29) is very much reminiscent of the famous Bogomonlyi-Prasad-
Sommerfeld bound which is saturated by solitons, tipically. This is not just a mere
analogy, since the bound (3.29) is in fact not just an algebraic relation but it has a
very concrete physical meaning: it is nothing but a specific BPS-like bound. Indeed,
short multiplets often arise as solitons in supersymmetric field theories, and central
charges correspond to physical (topological) charges. We will see concrete examples
of BPS states later in this course.

3.3 Representation on fields: a first try

So far we have discussed supersymmetry representations on states. However, we
would like to discuss supersymmetric field theories, eventually. Therefore, we need

56
to construct supersymmetric representations in terms of multiplets of fields rather
than multiplets of states. In principle, following our previous strategy this can be
done quite easily.
Let us start focusing on N = 1 supersymmetry. To build a representation of the
supersymmetry algebra on fields, we start from some field φ(x) for which
 
Qα̇ , φ(x) = 0 . (3.33)

The field φ is the analogous of the Clifford vacuum |λ0 i we used previously, the
ground state of the representation. Similarly as before, acting on this ground state
φ(x) with the supersymmetry generator Qα , we can generate new fields out of it, all
belonging to the same supermultiplet.
For definiteness, we choose φ(x) to be a scalar field, but one can also have ground
states which are fields with some non-trivial tensor structure, as we will later see.
Not much of what we want to say here depends on this choice.
The first thing to notice is that the scalar field φ(x) is actually complex. Suppose
it were real. Then, taking the hermitian conjugate of eq. (3.33) one would have
obtained
[Qα , φ(x)] = 0 . (3.34)
One can now use the generalized Jacobi identity for (φ, Q, Q) and get
      
φ(x), Qα , Qα̇ + Qα , Qα̇ , φ(x) − Qα̇ , [φ(x), Qα ] = 0

−→ 2σ µ [φ(x), Pµ ] = 0 −→ [Pµ , φ(x)] ∼ ∂µ φ(x) = 0 , (3.35)

which should then imply that the field is actually a constant (not a field, really!).
So better φ(x) to be complex. In this case eq. (3.34) does not hold, but rather

[Qα , φ(x)] ≡ ψα (x) . (3.36)

This automatically defines a new field ψα belonging to the same representation (since
φ is a scalar, ψ is a Weyl spinor). The next step is to see whether acting with su-
persymmetry generators on ψα gives new fields or just derivatives (or combinations)
of fields already present in the representation. In principle we have

{Qα , ψβ (x)} = Fαβ (x) (3.37)


Qα̇ , ψβ (x) = Xα̇β (x) . (3.38)

57
Enforcing the generalized Jacobi identity on (φ, Q, Q), and using eq. (3.36), after
some trivial algebra one gets

Xα̇β = ψβ (x), Qα̇ = 2σβµα̇ [Pµ , φ] ∼ ∂µ φ , (3.39)

which implies that Xα̇β is not a new field but just the space-time derivative of the
scalar field φ. Let us now enforce the generalized Jacobi identity on (φ, Q, Q). Since
the Q’s anticommute (recall we are considering N = 1 supersymmetry and hence
there are no central charges) one simply gets

{Qα , [Qβ , φ]} − {Qβ , [φ, Qα ]} = 0 −→ Fαβ + Fβα = 0 . (3.40)

This says that the field Fαβ is antisymmetric under α ↔ β, which implies that

Fαβ (x) = αβ F (x) . (3.41)

No other constraints are imposed on F by other consistency conditions. So we find

here a new complex scalar field F . Again, we should now ask whether acting on it
with supersymmetry generators produces new fields. We get

[Qα , F ] = λα (3.42)
 
Qα̇ , F = χα̇ . (3.43)

Using the generalized Jacobi identities for (ψ, Q, Q) and (ψ, Q, Q), one can easily
prove that λα is actually vanishing and that χα̇ is proportional to the space-time
derivative of the field ψ. So no new fields in this case: after a certain number of
steps the representation closes. The multiplet of fields we have found is then

(φ, ψ, F ) . (3.44)

If φ is a scalar field, as we have supposed here, this multiplet is a matter multiplet

since it contains particles with spin 0 and 1/2 only. It is called chiral multiplet and
it is indeed the field theory counterpart of the chiral multiplet of states we have
constructed before. Notice that the equality of the number of fermionic and bosonic
states for a given representation still holds: we are now off-shell, and the spinor ψα
has four degrees of freedom; this is the same number of bosonic degrees of freedom,
two coming from the scalar field φ and two from the scalar field F

(Reφ, Imφ, ReF, ImF )B , (Reψ1 , Imψ1 , Reψ2 , Imψ2 )F . (3.45)

58
While we see the expected degeneracy between bosonic and fermonic degrees of free-
dom, they do not match those of the chiral multiplet of states we have constructed
before, which are just 2B + 2F . This is because we are off-shell, now. Going on-shell,
the 4 fermionic degrees of freedom reduce to just 2 propagating degrees of freedom,
due to Dirac equation. The same sort of reduction should occur for the bosonic
degrees of freedom, in order to match the 2B + 2F on-shell condition. But Klein-
Gordon equation does not diminish the number of independent degrees of freedom!
What happens is that F turns out to be a non-dynamical auxiliary field: as we will
see when constructing Lagrangians, the equation of motion for F simply tells that
this scalar field is not an independent field but rather a (specific) function of other
fields, F = F (φ, ψ) . This is not specific to the chiral multiplet we have constructed,
but it is in fact a completely general phenomenon. We will come back to this point
in next lectures.
The procedure we have followed to construct the multiplet (3.44) can be eas-
ily generalized. Modifying the condition (3.33) one can construct other kind of
multiplets, like linear multiplets, vector multiplets, etc... And/or construct chiral
multiplets with different field content, simply by defining a ground state carrying
some space-time index, letting φ being a spinor, a vector, etc...
Out of a set of multiplets with the desired field content, one can construct suitable
Lagrangian made out of these fields. In order for the theory to be supersymmetric,
this Lagrangian should (at most) transform as a total space-time derivative under
supersymmetry transformations. Indeed, in this case, the action constructed out of
it Z
S = d4 x L , (3.46)

will be supersymmetric invariant.

In principle, this is a well defined program. In practice, however, to see whether
a given action is invariant under supersymmetry is rather cumbersome: one should
take any single term in the Lagrangian, act on it with supersymmetry transforma-
tions and prove that the variations of all (possibly very many) terms sum-up to
a total space-time derivative. This turns out to be very involved, in general, but
theoretical physicists came up with a brilliant idea to circumvent this problem.
This difficulty is due to the fact that the formulation above is a formulation in
which supersymmetry is not manifest. Ordinary field theories are naturally defined
in Minkowski space and in such formulation it is easy to construct Lagrangians
respecting Poincaré symmetry. It turns out that supersymmetric field theories are

59
naturally defined on an extension of Minkoswki space, known as superspace, which,
essentially, takes into account the extra space-time symmetries associated to the
supersymmetry generators. In such extended space it is much easier to construct
supersymmetric Lagrangians, and indeed the superspace formalism is what is most
commonly used to discuss supersymmetric field theories. This is the formalism
we will use along this course, and next chapter will be devoted to a throughout
introduction of superspace.

3.4 Exercises
1. Prove that P 2 and W 2 are Casimir of the Poincaré algebra.

2. Prove that CPT flips the sign of the helicity.

3. Construct explicitly, in terms of creation operators acting on the Clifford vac-

uum, the massive N = 1 vector multiplet (3.26) and the massive N = 2 BPS
vector multiplet (3.32). For the latter, determine the SU (2)R representations
under which bosonic and fermionic states transform.

4. Construct explicitly the N = 4 1/2 BPS vector multiplet (Hint: the Clif-
ford vacuum has j = 0). Discuss its (massive) content and its relation with
the massless vector multiplet. Can one construct a 1/4 BPS N = 4 vector
multiplet?

5. Enforcing the generalized Jacobi identity on (ψ, Q, Q) and (ψ, Q, Q), using
eqs. (3.37), (3.38), (3.42) and (3.43), prove that λα = 0 and χβ̇ ∼ ∂µ (σ µ ψ)β̇ .

References

[1] A. Bilal, Introduction to supersymmetry, Section 3, arXiv:hep-th/0101055.

[2] L. Castellani, R. D’Auria and P. Fre, Supergravity And Superstrings: A Geo-

metric Perspective. Vol. 1: Mathematical Foundations, Chapter II.4 and II.5,
Singapore: World Scientific (1991).

[3] J.D. Lykken, Introduction to Supersymmetry, Section 2, TASI 96, arXiv:hep-

th/9612114.

60
[4] M.F. Sohnius, Introducing Supersymmetry, Sections 3 and 5.4, Phys. Rep. 128
(1985).

[5] S. Weinberg, The quantum theory of fields. Vol. 1: Foundations, Chapter 13.1,
Cambridge University Press (2000).

[6] M. T. Grisaru and H. N. Pendleton, Soft Spin 3/2 Fermions Require Gravity
and Supersymmetry, Phys. Lett. B 67 (1977), 323.

61
4 Superspace and superfields

The usual space-time Lagrangian formulation is not the most convenient one for
describing supersymmetric field theories. This is because in ordinary space-time
supersymmetry is not manifest. In fact, an extension of ordinary space-time, known
as superspace, happens to be the best and most natural framework in which to for-
mulate supersymmetric theories. The basic idea of (N = 1) superspace is to enlarge
the space-time labelled with coordinates xµ , associated to the generators Pµ , by
adding 2 + 2 anti-commuting Grassmann coordinates θα , θ̄α̇ , associated to the su-
persymmetry generators Qα , Qα̇ , and obtain a eight coordinate superspace labelled
by (xµ , θα , θ̄α̇ ). In such an apparently exotic space many mysterious (or hidden)
properties of supersymmetric field theories become manifest. As we will see, at the
price of learning a few mathematical new ingredients, the goal of constructing super-
symmetric field theories will be achieved much more easily, and within a framework
in which many classical and quantum properties of supersymmetry will be more
transparent.
In this lecture we will introduce superspace and superfields. In subsequent lec-
tures we will use this formalism to construct supersymmetric field theories and study
their dynamics.

4.1 Superspace as a coset

Let us start recalling the relation between ordinary Minkowski space and the Poincaré
group. Minkowski space is a four-dimensional coset space defined as
ISO(1, 3)
M1,3 = , (4.1)
SO(1, 3)
where ISO(1, 3) is the Poincaré group and SO(1, 3) the Lorentz group. The Poincaré
group ISO(1, 3) is nothing but the isometry group of this coset space, which means
that any point of M1,3 can be reached from the origin with a Poincaré transfor-
mation. This transformation, however, is defined up to Lorentz transformations.
Therefore, each coset class (≡ a point in space-time) has a unique representative
which is just a translation and can be parametrized by a coordinate xµ
µP
xµ ←→ e(x µ)
. (4.2)

Superspace can be defined along similar lines. The first thing we need to do is to
extend the Poincaré group to the so-called superPoincaré group. In order to do this,

62
given that a group is the exponent of the algebra, we have to rewrite the whole
supersymmetry algebra in terms of commutators, namely as a Lie algebra. This is
achieved by introducing a set of constant Grassmann numbers θα , θ̄α̇ , defined as to
anti-commute with everything fermionic and commute with everything bosonic

{θα , θβ } = 0 , {θ̄α̇ , θ̄β̇ } = 0 , {θα , θ̄β̇ } = 0 . (4.3)

This allows to transform anti-commutators of the supersymmetry algebra into com-

mutators, and get
   
θQ, θ̄Q = 2 θσ µ θ̄Pµ , [θQ, θQ] = θ̄Q, θ̄Q = 0 , (4.4)
α̇
where as usual θQ ≡ θα Qα , θ̄Q ≡ θ̄α̇ Q . This way, one can write the supersymmetry
algebra solely in terms of commutators. Exponentiating this algebra one gets the
superPoincaré group. A generic group element can then be written as
1
G(x, θ, θ̄, ω) = exp(ixP + iθQ + iθ̄Q + iωM ) , (4.5)
2
where xP is a shorthand notation for xµ Pµ and ωM a shorthand notation for
ω µν Mµν .
The superPoincaré group, mathematically, is Osp(4|1). Let us open a brief paren-
thesis and explain such a notation. Let us define the graded Lie algebra Osp(2p|N )
as the grade one Lie algebra L = L0 ⊕ L1 whose generic element can be written as
a matrix of complex dimension (2p + N ) × (2p + N )
!
A B
(4.6)
C D

where A is a (2p × 2p) matrix, B a (2p × N ) matrix, C a (N × 2p) matrix and D a

(N × N ) matrix. An element of L0 respectively L1 has entries
! !
A 0 0 B
respectively (4.7)
0 D C 0

where

AT Ω(2p) + Ω(2p) A = 0

DT Ω(N ) + Ω(N ) D = 0

C = Ω(N ) B T Ω(2p)

63
and
Ω2(2p) = −I , ΩT(N ) = Ω(N ) , ΩT(2p) = −Ω(2p) . (4.8)
This implies that the matrices A span a Sp(2p, C) algebra and the matrices D a
O(N, C) algebra. Therefore we have that

L0 = Sp(2p) ⊗ O(N ) , (4.9)

hence the name Osp(2p|N ) for the whole superalgebra. A generic element of the
superalgebra has the form
Q = q a ta + q l tl , (4.10)
where ta ∈ L0 and tl ∈ L1 are a basis of the corresponding vector spaces, and we
have introduced complex numbers q a for L0 and Grassman numbers q l for L1 (recall
why and how we introduced the fermionic parameters θα , θ̄α̇ before).
Taking now p = 2 we have the algebra Osp(4|N ). This is not yet what we are
after. The last step, which we do not describe in detail here, amounts to take the
so-called Inonu-Wigner contraction. Essentially, one has to rescale (almost) all gen-
erators by a constant 1/ē, rewrite the algebra in terms of the rescaled generators
and take the limit ē → 0. What one ends up with is the N -extended supersym-
metry algebra in Minkowski space we all know, dubbed Osp(4|N ), where in the
aforementioned limit one gets the identification

A → Pµ , Mµν D → Z IJ B, C → QI , Q̄I . (4.11)

Taking N = 1 one finally gets the unextended supersymmetry algebra Osp(4|1).

Given the generic group element of the superPoincaré group (4.5), N = 1 super-
space is defined as the (4+4 dimensional) group coset

Osp(4|1)
M4|1 = . (4.12)
SO(1, 3)
A point in superspace (point in a loose sense, of course, given the non-commutative
nature of the Grassman parameters θα , θ̄α̇ ) gets identified with the coset represen-
tative corresponding to a so-called super-translation through the one-to-one map


xµ , θα , θ̄α̇ ←→ e(x Pµ ) e(θQ+θ̄Q) .
µ
(4.13)

The 2 + 2 anti-commuting Grassmann numbers θα , θ̄α̇ can then be thought of as co-

ordinates in superspace (in four-component notation they correspond to a Marojana
spinor θ). For these Grassmann numbers all usual spinor identities hold.

64
 Thus far we have introduced what is known as N = 1 superspace. If discussing
extended supersymmetry one should introduce, in principle, a larger superspace.
There exist (two, at least) formulations of N = 2 superspace. However, these
formulations present some subtleties and problems whose discussion is beyond the
scope of this course. And no formulation is known of N = 4 superspace. In this
course we will use N = 1 superspace even when discussing extended supersymmetry,
as it is typically done in most of the literature.

4.2 Superfields as fields in superspace

Superfields are nothing but fields in superspace: functions of the superspace coordi-
nates (xµ , θα , θ̄α̇ ). Since θα and θ̄α̇ anticommute, any product involving more than
two θ’s or two θ̄’s vanishes: given that θα θβ = −θβ θα , we have that θα θβ = 0 for
α = β and therefore θα θβ θγ = 0, since at least two indices in this product are the
same. Hence, the most general superfield Y = Y (x, θ, θ̄) has the following simple
Taylor-like expansion

Y (x, θ, θ̄) = f (x) + θψ(x) + θ̄ χ(x) + θθ m(x) + θ̄θ̄ n(x) +

+ θσ µ θ̄ vµ (x) + θθ θ̄ λ(x) + θ̄θ̄ θρ(x) + θθ θ̄θ̄ d(x) . (4.14)

Each entry above is a field (possibly with some non-trivial tensor structure). In this
sense, a superfield is nothing but a finite collection (a multiplet) of ordinary fields.
We aim at constructing supersymmetric Lagrangians out of superfields. In such
Lagrangians superfields get multiplied one another, sometime we should act on them
with derivatives, etc... Moreover, integration in superspace will be needed, eventu-
ally. Therefore, it is necessary to pause a bit and recall how operations of this kind
work for Grassman variables.
Derivation in superspace is defined as follows
∂ ∂
∂α ≡ and ∂ α = −αβ ∂β , ∂¯α̇ ≡ α̇ and ∂¯α̇ = −α̇β̇ ∂¯β̇ , (4.15)
∂θα ∂ θ̄
where
∂α θβ = δαβ , ∂¯α̇ θ̄β̇ = δα̇β̇ , ∂α θ̄β̇ = 0 , ∂¯α̇ θβ = 0 . (4.16)

Let us consider a single Grassmann variable θ (either θ1 , θ2 , θ̄1̇ or θ̄2̇ in our case).
Integration in θ is defined as follows
Z Z
dθ = 0 dθ θ = 1 . (4.17)

65
This implies that for a generic function f (θ) = f0 + θf1 , the following results hold
Z Z Z
dθ f (θ) = f1 , dθ δ(θ)f (θ) = f0 −→ = ∂ , θ = δ(θ) . (4.18)

These relations can be easily generalized to N = 1 superspace, provided

1 1
d2 θ ≡ dθ1 dθ2 , d2 θ̄ ≡ dθ̄2̇ dθ̄1̇ . (4.19)
2 2
With these definitions one can prove the following useful identities
Z Z Z
2 2
d θ θθ = d θ̄ θ̄θ̄ = 1 , d2 θd2 θ̄ θθ θ̄θ̄ = 1
Z Z
1 1
d θ = αβ ∂α ∂β ,
2
d2 θ̄ = − α̇β̇ ∂¯α̇ ∂¯β̇ . (4.20)
4 4
Another crucial question we need to answer is: how does a superfield transform
under supersymmetry transformations? In order to answer this question we first
need to realize the supersymmetry generators Qα , Qα̇ as differential operators, in
the same way we do for the generators of Poincaré algebra (translations, rotations
and boosts).
Let us recall how the story goes in ordinary space-time and consider a translation
generated by Pµ with infinitesimal parameter aµ , on a field φ(x) (for notational
convenience, we will use momentarily calligraphic letters for the abstract operator
and latin ones for the representation of the same operator as a differential operator
in field space). This is defined as

φ(x + a) = e−iaP φ(x)eiaP = φ(x) − iaµ [Pµ , φ(x)] + . . . . (4.21)

On the other hand, Taylor expanding the left hand side we get

φ(x + a) = φ(x) + aµ ∂µ φ(x) + . . . (4.22)

Equating the right hand sides of the two equations above we then get

[φ(x), Pµ ] = −i∂µ φ(x) ≡ Pµ φ(x) , (4.23)

where Pµ is the generator of translations and Pµ is its representation as a differential

operator in field space (recall that ∂µ is an operator and from (∂µ )∗ = ∂µ one gets
that (∂µ )† = −∂µ ; hence Pµ is hermitian, as it should). So, a translation of a field
by parameter aµ induces a change on the field itself as

δa φ ≡ φ(x + a) − φ(x) = iaµ Pµ φ . (4.24)

66
Notice that here and below we are using right multiplication, when acting on fields.
We now want to apply the same procedure to a superfield. A translation in
superspace (i.e. a supersymmetry transformation) on a superfield Y (x, θ, θ̄) by a
quantity (α , ¯α̇ ), where α , ¯α̇ are spinorial parameters, is defined as

Y (x + δx, θ + δθ, θ̄ + δ θ̄) = e−i(Q+¯Q) Y (x, θ, θ̄)ei(Q+¯Q) , (4.25)

with
δ,¯ Y (x, θ, θ̄) ≡ Y (x + δx, θ + δθ, θ̄ + δ θ̄) − Y (x, θ, θ̄) (4.26)
the variation of the superfield under the supersymmetry transformation.
To find the representation of Q and Q as differential operators, there are two
questions we need to answer, first. What is the explicit expression for δx, δθ, δ θ̄?
Why are we supposing here δx 6= 0, given we are not acting with the generator of
space-time translations Pµ , but just with supersymmetry generators?
First notice that eq. (4.25) can be written as

Y (x + δx, θ + δθ, θ̄ + δ θ̄) = e−i(Q+¯Q) e−i(xP+θQ+θ̄Q) Y (0; 0, 0)ei(xP+θQ+θ̄Q) ei(Q+¯Q)

(4.27)
Let us now evaluate the last two exponentials, for which we need to recall the Baker-
Campbell-Hausdorff formula for non-commuting objects which says that
X∞
A B C 1
e e =e where C = Cn (A, B) (4.28)
n=1
n!

with
1 1
C1 = A + B , C2 = [A, B] , C3 = [A, [A, B]] − [B, [B, A]] . . . . (4.29)
2 2
We then have



exp{i xP + θQ + θ̄Q } exp{i Q + ¯Q } =
1  1 
= exp{ixµ Pµ + i( + θ)Q + i(¯ + θ̄)Q − θ̄Q, Q − θQ, ¯Q }
2 2
= exp{ixµ Pµ + i( + θ)Q + i(¯ + θ̄)Q + σ µ θ̄ Pµ − θσ µ ¯ Pµ }

= exp{i(xµ + iθσ µ ¯ − iσ µ θ̄)Pµ + i( + θ)Q + i(¯ + θ̄)Q} (4.30)

which means that 

 µ µ µ
δx = iθσ ¯ − iσ θ̄
δθα = α (4.31)

 α̇
δ θ̄ = ¯α̇

67
This answers the first question. Notice, further, the expression for δxµ , which is
non-vanishing. This is needed, in order to be consistent with the supersymmetry

algebra, Qα , Qα̇ ∼ Pµ : two subsequent supersymmetry transformations generate
a space-time translation. This answers the second question.
We can now find the representation of the supersymmetry generators Qα and
Qα̇ as differential operators. Let us take eq. (4.26) and, recalling eqs. (4.31), let us
Taylor expand the right hand side which becomes


δ,¯ Y (x, θ, θ̄) = Y (x, θ, θ̄) + i θσ µ ¯ − σ µ θ̄ ∂µ Y (x, θ, θ̄) +

+ α ∂α Y (x, θ, θ̄) + ¯α̇ ∂¯α̇ Y (x, θ, θ̄) + · · · − Y (x, θ, θ̄)



= α ∂α + ¯α̇ ∂¯α̇ + i θσ µ ¯ − σ µ θ̄ ∂µ + . . . Y (x, θ, θ̄) (4.32)

On the other hand, from eq. (4.25) we get

δ,¯ Y (x, θ, θ̄) = 1 − iQ − i¯Q + . . . Y (x, θ, θ̄) 1 + iQ + i¯Q + . . . − Y (x, θ, θ̄)
   
= −iα Qα , Y (x, θ, θ̄) + i¯α̇ Qα̇ , Y (x, θ, θ̄) + . . . , (4.33)
α̇
(recall that i¯Q ≡ i¯α̇ Q = −i¯α̇ Qα̇ ). Defining
 
[Y, Qα ] ≡ Qα Y , Y, Qα̇ ≡ Qα̇ Y , (4.34)

the previous result implies that the supersymmetry variation of a superfield by

parameters , ¯ is represented as


δ,¯ Y = iQ + i¯Q Y . (4.35)

Comparing with eq. (4.32) we get the following expression for the differential oper-
ators Qα , Qα̇ (
Qα = − i∂α − σαµβ̇ θ̄β̇ ∂µ
(4.36)
Qα̇ = + i∂¯α̇ + θβ σ µ ∂µ β α̇

Notice that, consistently, Q†α = Qα̇ (to prove this, recall that (σαµβ̇ )† = σβµα̇ ).
One can check the validity of the expressions (4.36) by showing that the two
differential operators close the supersymmetry algebra, namely that

{Qα , Qβ } = {Qα̇ , Qβ̇ } = 0 , {Qα , Qβ̇ } = 2σαµβ̇ Pµ . (4.37)

We can now give a more precise definition for what a superfield actually is: a super-
field is a field in superspace which transforms under a super-translation according
to eq. (4.25). This implies, in particular, that a product of superfields is still a
superfield.

68
4.3 Supersymmetric invariant actions - general philosophy
Having seen that a supersymmetry transformation is simply a translation in super-
space, it is now easy to construct supersymmetric invariant actions. In order for
an action to be invariant under superPoincaré transformations it is enough that the
Lagrangian is Poincaré invariant (actually, it should transform as a scalar density)
and that its supersymmetry variation is a total space-time derivative.
Here is where the formalism we have introduced starts to manifest its powerful-
ness. The basic point is that the integral in superspace of any arbitrary superfield
is a supersymmetric invariant quantity. In other words, the following integral
Z
d4 x d2 θ d2 θ̄ Y (x, θ, θ̄) (4.38)

is manifestly supersymmetric invariant, if Y is a superfield. This can be proven as

follows. The integration measure is translationally invariant by construction since
Z Z
dθθ = d(θ + ξ)(θ + ξ) = 1 (4.39)

This implies that

Z Z
δ,¯ d x d θ d θ̄ Y (x, θ, θ̄) = d4 x d2 θ d2 θ̄ δ,¯ Y (x, θ, θ̄) .
4 2 2
(4.40)

Now, using eqs. (4.35) and (4.36) we get



δ,¯ Y = α ∂α Y + ¯α̇ ∂¯α̇ Y + ∂µ −i σ θ̄ − θσ¯ Y . (4.41)

Integration in d2 θd2 θ̄ kills the first two terms since they do not have enough θ’s or
θ̄’s to compensate for the measure, and leaves only the last term, which is a total
derivative. In other words, under supersymmetry transformations the integrand in
eq. (4.40) transforms as a total space-time derivative plus terms which get killed by
integration in superspace. Hence the full integral is supersymmetric invariant
Z
δ,¯ d4 x d2 θ d2 θ̄ Y (x, θ, θ̄) = 0 . (4.42)

Supersymmetric invariant actions are constructed this way, i.e. by integrating in

superspace a suitably defined superfield. Such superfield, call it A, should not be
generic, of course. It should have the right structure to give rise, upon integration
on Grassman coordinates, to a Lagrangian density, which is a real, dimension-four

69
operator, transforming as a scalar density under Poincaré transformations. The end
result will be a supersymmetric invariant action S
Z Z
S = d x d θ d θ̄ A(x; θ, θ̄) = d4 x L (φ(x), ψ(x), Aµ (x), . . . ) .
4 2 2
(4.43)

Let us emphasize again: one does not need to prove S to be invariant under super-
symmetry transformations. If it comes from an integral of a superfield in superspace,
this is just automatic: by construction, the Lagrangian L on the r.h.s. of eq. (4.43),
an apparently innocent-looking function of ordinary fields, is guaranteed to be su-
persymmetric invariant, up to total space-time derivatives.
The superfield A will be in general a product of superfields (recall that a prod-
uct of superfields is still a superfield). However, the general superfield (4.14) cannot
be the basic object of this construction: it contains too many field components to
correspond to an irreducible representation of the supersymmetry algebra. We have
to put (supersymmetric invariant) constraints on Y and restrict its form to contain
only a subset of fields. Being the constraint supersymmetric invariant, this reduced
set of fields will still be a superfield, and hence it will carry a representation of the
supersymmetry algebra. In what follows, we will start discussing a few such con-
straints, the so-called chiral and real constraints. These will be the relevant ones for
our purposes, as they will lead to chiral and vector superfields, the right superfields
to accommodate matter and radiation, respectively, and to linear superfields, where
conserved currents sit.

4.4 Chiral superfields

One can construct covariant derivatives Dα , Dα̇ defined as
(
Dα = ∂α + i σαµβ̇ θ̄β̇ ∂µ
(4.44)
Dα̇ = ∂¯α̇ + i θβ σ µ ∂µβ α̇

and which anticommute with the supersymmetry generators Qα , Qα̇ . More precisely
we have

{Dα , Dβ̇ } = 2i σαµβ̇ ∂µ = −2σαµβ̇ Pµ , (4.45)

{Dα , Dβ or Qβ or Qβ̇ } = 0 (similarly for Dα̇ ) . (4.46)

This implies that

δ,¯ (Dα Y ) = Dα (δ,¯ Y ) , (4.47)

70
since Dα commutes both with Q and ¯Q. Therefore, if Y is a superfield, that is a
field in superspace transforming as dictated by eq. (4.25) under a supersymmetry
transformation, so is Dα Y . This means that Dα Y = 0 is a supersymmetric invariant
constraint we can impose on a superfield Y to reduce the number of its components,
while still having the field carrying a representation of the supersymmetry algebra
(the same holds for the constraint Dα̇ Y = 0).
Recall the generic expression (4.14) for Y and consider ∂¯α̇ Y : this has fewer
components with respect to Y itself, since, for instance, there is no θθθ̄θ̄ term.
However
 
∂¯α̇ , Q = β σβµα̇ ∂µ . (4.48)
This implies that a supersymmetry transformation on X α̇ = ∂¯α̇ Y would generate a
θθθ̄θ̄ term which X α̇ did not contain. Hence ∂¯α̇ Y is not a superfield, in the sense
of providing a basis for a representation of supersymmetry. On the other hand, the
covariant derivatives defined in (4.44) anticommute with Qα and Qα̇ . Hence, if Y is
a superfield, Dα Y, Dα̇ Y are also superfields (and so is ∂µ Y , since Pµ commutes with
Qα and Qα̇ ).
A chiral superfield Φ is a superfield such that

Dα̇ Φ = 0 . (4.49)

Seemingly, an anti-chiral superfield Ψ is a superfield such that

Dα Ψ = 0 . (4.50)

Notice that if Φ is chiral, its hermitian conjugate, Φ, is anti-chiral (unless otherwise

stated, here and in the following we will use the symbol − instead of † to mean
hermitian conjugation, to adapt to the two-component spinor notation for which
ψ † = ψ α̇ ). This implies that a chiral superfield cannot be real (i.e. Φ = Φ). Indeed,
in this case it is easy to show that it should be a constant. Taking the hermitian
conjugate of eq. (4.49) one would conclude that the field would also be anti-chiral.
Acting now on it with the anticommutator in eq.(4.45) one would get ∂µ Φ = 0. This
is the superfield analogue of what we have seen in the previous lecture, when we
constructed the chiral multiplet and we learned that the scalar field φ had to be
complex.
We would like to find the most general expression for a chiral superfield in terms
of ordinary fields, as we did for the general superfield (4.14). In other words, we have

71
to integrate the constraint (4.49). To this aim, it is useful to define new coordinates

y µ = xµ + iθσ µ θ̄ , ȳ µ = xµ − iθσ µ θ̄ . (4.51)

It easily follows that

Dα̇ θβ = Dα̇ y µ = 0 , Dα θ̄β̇ = Dα ȳ µ = 0 . (4.52)

Recalling the definition (4.49) this implies that Φ depends only on (y µ , θα ) explicitly,
but not on θ̄α̇ (the θ̄-dependence is hidden inside y µ ). In this (super)coordinate
system the chiral constraint is easily solved by
√
Φ(y, θ) = φ(y) + 2θψ(y) − θθF (y) . (4.53)

Taylor-expanding the above expression around x we get for the actual Φ(x, θ, θ̄)
√ i 1
Φ(x, θ, θ̄) = φ(x)+ 2θψ(x)+iθσ µ θ̄∂µ φ(x)−θθF (x)− √ θθ∂µ ψ(x)σ µ θ̄− θθθ̄θ̄φ(x) ,
2 4
(4.54)
iθσ µ ∂µ θ̄
which can also be conveniently recast as Φ(x, θ, θ̄) = e Φ(x, θ). We see that,
as expected, this superfield has less components than the general superfield Y , and
some of them are related to each other.
The chiral superfield (4.54) is worth its name, since it is a superfield which
encodes precisely the degrees of freedom of the chiral multiplet of fields we have
previously constructed. On-shell, it corresponds to a N = 1 multiplet of states,
hence carrying an irreducible representation of the N = 1 supersymmetry algebra.
A similar story holds for the anti-chiral superfield Φ for which we would get
√
Φ(x, θ, θ̄) = φ(ȳ) + 2θ̄ψ(ȳ) − θ̄θ̄F (ȳ) (4.55)
√ i 1
= φ(x) + 2θ̄ψ(x) − iθσ µ θ̄∂µ φ(x) − θ̄θ̄F (x) + √ θ̄θ̄θσ µ ∂µ ψ(x) − θθθ̄θ̄φ(x) .
2 4

Let us now try and see how does a chiral (or anti-chiral) superfield transform under
supersymmetry transformations. This amounts to compute


δ,¯ Φ(y; θ) = iQ + i¯Q Φ(y; θ) (4.56)

(and similarly for Φ). To compute eq. (4.56) it is convenient to write the differential
operators Qα , Qα̇ in the (y µ , θα , θ̄α̇ ) coordinate system. This amounts to trade the

72
partial derivatives taken with respect to (xµ , θα , θ̄α̇ ) for those taken with respect to
the new system (y µ , θα , θ̄α̇ ) and plug this into eqs. (4.36). The final result reads
(
Qnew
α = −i∂α
new (4.57)
Qα̇ = i∂¯α̇ + 2θα σαµα̇ ∂y∂µ

Plugging these expressions into eq. (4.56) one gets

 
α α µ β̇ ∂
δ,¯ Φ(y; θ) =  ∂α + 2iθ σαβ̇ ¯ Φ(y; θ)
∂y µ
 
√ µ ∂ √ ∂
= 2ψ − 2θF + 2iθσ ¯ φ + 2θ µ ψ (4.58)
∂y µ ∂y
   
√ √ √ √ µ ∂ √ µ ∂
= 2ψ + 2θ − 2F + 2iσ ¯ µ φ − θθ −i 2¯σ̄ ψ .
∂y ∂y µ

Therefore, the final expression for the supersymmetry variation of the different field
components of the chiral superfield Φ reads
 √

 δφ = 2ψ
√ √
δψα = 2i(σ µ ¯)α ∂µ φ − 2α F (4.59)

 √ µ
δF = i 2∂µ ψσ ¯

It is left to the reader to derive the corresponding expressions for an anti-chiral

superfield. In this case, one should write the generators Qα , Qα̇ in the (ȳ µ , θα , θ̄α̇ )
coordinate system.

4.5 Real (aka vector) superfields

In order to have gauge interactions we clearly need to find a supersymmetric invari-
ant projection which saves the vector field v µ in the general expression (4.14) and
makes it real (this was not the case for the constraint (4.49), for which the vector
component is ∼ ∂µ φ). The right thing to do is to impose a reality condition on the
general superfield Y . Indeed, under hermitian conjugation, Y → Y , one has that
v µ → v µ ; so imposing a reality condition, the vector component not only survives
as an independent degree of freedom, but becomes real.
A real (aka vector) superfield V is a superfield such that

V =V . (4.60)

73
Looking at the general expression (4.14) this leads to the following expansion for V
i
V (x, θ, θ̄) = C(x) + i θχ(x) − i θ̄χ(x) + θσ µ θ̄vµ + θθ (M (x) + iN (x))
2
 
i i µ
− θ̄θ̄ (M (x) − iN (x)) + i θθθ̄ λ(x) + σ̄ ∂µ χ(x) (4.61)
2 2
   
i µ 1 1 2
− i θ̄θ̄θ λ(x) + σ ∂µ χ(x) + θθθ̄θ̄ D(x) − ∂ C(x) .
2 2 2
Notice that, as such, this superfield has 8B + 8F degrees of freedom. The next step
is to introduce the supersymmetric version of gauge transformations. As we shall
see, after gauge fixing, this will reduce the number of off-shell degrees of freedom to
4B + 4F , which become 2B + 2F on-shell (for a massless representation), as it should
be the case for a massless vector multiplet of states, see eq. (3.15).
First notice that if Φ is a chiral superfield, then Φ + Φ is a (very special) vector
superfield. So, under the following transformation on V
V →V +Φ+Φ (4.62)
one gets a a superfield V which is still real. Under the shift (4.62) the vector
component vµ in (4.61) transforms as vµ → vµ −∂µ (2 Imφ). This is precisely how an
ordinary (abelian) gauge transformation acts on a vector field. Therefore, eq. (4.62)
is a natural definition for the supersymmetric version of a gauge transformation.
Under eq. (4.62) the component fields of V transform as


 C → C + 2 Reφ

 √

 χ → χ − i 2ψ





M → M − 2 ImF
N → N + 2 ReF (4.63)



 D → D



 λ → λ



 v µ → v µ − 2 ∂ Imφ
µ

where the components of Φ have been dubbed (φ, ψ, F ). From the transformations
above one sees that properly choosing Φ, namely choosing
C i N M
Reφ = − , ψ = − √ χ , ReF = − , ImF = . (4.64)
2 2 2 2
one can gauge away (namely put to zero) C, M, N, χ. The choice above is called
Wess-Zumino gauge. In this gauge a vector superfield can be written as
1
VW Z = θσ µ θ̄ vµ (x) + iθθ θ̄λ(x) − iθ̄θ̄ θλ(x) + θθ θ̄θ̄D(x) . (4.65)
2
74
Therefore, taking into account gauge invariance (that is, the redundancy of one
of the vector degrees of freedom, the one associated to the transformation vµ →
vµ − ∂µ (2 Imφ), which the WZ gauge does not fix), we end-up with 4B + 4F degrees
of freedom off-shell. As we shall see later, D will turn out to be an auxiliary field;
therefore, by imposing the equations of motion for D, the spinor λ and the vector
v µ , one will end up with 2B + 2F degrees of freedom on-shell. Since we like to
formulate gauge theories keeping gauge invariance manifest off-shell, the WZ gauge
is defined as a gauge where C = M = N = χ = 0, but no restrictions on v µ . This
way, while remaining in the WZ gauge, we still have the freedom to do ordinary
gauge transformations. In other words, once in the WZ gauge, we can still perform a
supersymmetric gauge transformation (4.62) with parameters φ = −φ, ψ = 0, F = 0.
Let us end this section with two important comments. First notice that in the
WZ gauge each term in the expansion of VW Z contains at least one θ. Therefore
1
VW2 Z = θθθ̄θ̄vµ v µ , VWn Z = 0 n ≥ 3 . (4.66)
2
These identities will simplify things a lot when it comes to construct supersymmetric
gauge actions. Second, it should be remarked that the WZ gauge does not commute
with supersymmetry. Acting with a supersymmetry transformation on a vector
superfield in the WZ gauge, one obtains a new superfield which is not in the WZ
gauge. Hence, when working in this gauge, after a supersymmetry transformation,
one has to do a compensating supersymmetric gauge transformation (4.62), with a
properly chosen Φ, to come back to the WZ gauge.

4.6 (Super)Current superfields

The two superfields described above are what we need to describe matter and ra-
diation in a supersymmetric theory, if we are not interested in gravitational inter-
actions. However, in a supersymmetric theory, also composite operators should sit
in superfields. There are at least two other types of superfields which accommodate
important composite operators, i.e. conserved currents and the supersymmetry cur-
rent (supercurrent for short), the latter being ubiquitous in a supersymmetric QFT,
as this is the current associated to the supersymmetry charge itself. Both these su-
perfields turn out to be real superfields, as the superfield described in the previous
section, but current conservation implies extra supersymmetric invariant conditions
they should satisfy which make them a particular class of real superfields. In what
follows, we will briefly describe both of them.

75
4.6.1 Internal symmetry current superfields

Because of Nöether theorem, in a local QFT any continuous symmetry is associated

to a conserved current jµ satisfying ∂ µ jµ = 0, and to the corresponding conserved
R
charge Q defined as Q = d3 x j 0 . Here we are referring to non-R symmetries.
R-symmetry will be discussed later.
As any other operator, in a supersymmetric theory a conserved current should
sit in a superfield. It turns out that this is a real scalar superfield J satisfying the
following extra constraint
2
D2 J = D J = 0 . (4.67)
A real superfield satisfying the constraint above is called linear superfield. Working
a little bit one can show that a real superfield subject to the conditions (4.67) has
the following component expression
1 1 1
J = J(x)+iθj(x)−iθ̄j̄(x)+θσ µ θ̄jµ (x)+ θ2 θ̄σ̄ µ ∂µ j(x)− θ̄2 θσ µ ∂µ j̄(x)+ θ2 θ̄2 J(x) ,
2 2 4
(4.68)
where J is a real scalar and jα a spinor. By imposing eq. (4.67) on the above
expression one easily sees that the current jµ satisfies ∂ µ jµ = 0, i.e. is a conserved
current. So the constraint (4.67) is indeed the correct supersymmetric generaliza-
tion of current conservation. Note that while the condition (4.67) is compatible with
2
supersymmetry (both D2 and D commute with supersymmetry transformations),
it stands on a slightly different footing with respect to the conditions (4.49), (4.50)
and (4.60). The latter constrain the dependence of a superfield as a function of
the fermionic coordinates (θα , θ̄α̇ ), but they do not say anything about space-time
dependence. On the contrary, eq. (4.67) constrains also the space-time dependence
of some of the fields imposing differential equations in x-space, one obvious exam-
ple being the conservation equation ∂ µ jµ = 0. In this sense, (4.67) is an on-shell
constraint.
A few comments are in order. First notice that, as compared to a general real
superfield (4.61), a linear superfield has less independent components. This is due
to the extra condition (4.67) a linear superfield has to satisfy. Another comment
regards the spin content of J . One condition that J should (and does) satisfy is
that it should not contain fields with spin higher than one. If this were the case,
one could not gauge the current j µ without introducing higher-spin gauge fields,
something which is expected not to be consistent in a local interacting QFT with
rigid supersymmetry (recall our discussion in the previous lecture). This implies

76
that J should be a real scalar superfield, namely its lowest component J should be
a scalar. Finally, it is worth notice that the detailed structure of J is not uniquely
fixed, but in fact defined up to Schwinger terms entering the current algebra.
This can be understood as follows. Because the conserved charge Q is a non-R
symmetry charge, it commutes with supersymmetry generators, [Qα , Q] = 0. This
implies that the operator on the right hand side of the current algebra

[Qα , jµ ] = Oαµ , (4.69)

should be an operator which vanishes when acting with ∂µ , because so is jµ , and it

should also be a total space-time derivative for µ = 0, say Oα0 = ∂ ν Aαν , so that it
integrates to zero, because so happens to the left hand side given that
Z Z
4
d x [Qα , j0 ] = dt [Qα , Q] = 0 . (4.70)

An operator of this kind is known as Schwinger term. Different Schwinger terms

provide different versions of the superfield J , which is hence not univocally defined.
The superfield defined in eq. (4.68) is one possible such completions, for which
Oαµ = −2i(σµν )αβ ∂ ν jβ . This can be easily checked using eqs. (4.34)-(4.35). This
said, the first terms in the expansion of J , namely J + iθj − iθ̄j̄ + θσ µ θ̄jµ (x), are
universal in the sense that they turn out not to depend on the specific Schwinger
term appearing in eq. (4.69).

4.6.2 Supercurrent superfields

While currents associated to internal symmetries might or might not be there, in

any supersymmetric theory there always exists, by definition, a conserved current,
the supersymmetry current Sαµ , associated to the conservation of the supercharge
R
Qα . In terms of the supercurrent, the supersymmetry charge is Qα = d3 x Sα0 .
Such supercurrent should sit in a superfield.
An equation analogous to eq. (4.69) is imposed by the supersymmetry algebra,
which reads

Qα̇ , Sαν = 2σαµα̇ Tµν + Oαα̇ν , (4.71)
where Tµν is the (conserved) energy-momentum tensor and Oαα̇ν is again a Schwinger
term. Note that, unlike eq. (4.69), the ν = 0 component of the left hand side of
R
eq. (4.71) does not integrate to zero now but is proportional to dt Pµ by the
R
supersymmetry algebra, namely to d4 x Tµ0 . This is why, on top of a Schwinger

77
term, the energy-momentum tensor appears on the right hand side of eq. (4.71).
This also shows that the supercurrent and the energy-momentum tensor sit in the
same superfield, Tµν being the highest spin field of the representation (otherwise,
it would be problematic coupling supersymmetry with gravity avoiding higher spin
currents). This is the current operators counterpart of the fact that the graviton
and the gravitino sit in the same multiplet. These properties are enough to fix the
universal structure of the supercurrent superfield which reads


Jµ = jµ + θ (Sµ + . . . ) + θ̄ S µ + . . . + θσ ν θ̄ (2Tµν + . . . ) + . . . . (4.72)

The arbitrariness of the Schwinger term gives rise to different possible completions
of Jµ . The most known such completions is due to Ferrara and Zumino. The FZ
multiplet can be described by a pair of superfields (Jµ , X) satisfying the relation
α̇
2 D σαµα̇ Jµ = Dα X , (4.73)

with Jµ being a real vector superfield, and X a chiral superfield, Dα̇ X = 0. The
same comment we made on the on-shell nature of the condition (4.67) holds also
in this case. From the defining equation above one can work out the component
expression of these two superfields. They read
   
1 1 i i
Jµ =jµ + θ Sµ − σµ S + θ̄ S µ + σ̄µ S + θ2 ∂µ x∗ − θ̄2 ∂µ x
3 3 2 2
  (4.74)
2 1
+ θσ ν θ̄ 2Tµν − ηµν T + εµνρσ ∂ ρ j σ + . . .
3 2
and  
2 2 2 µ
X = x + θS + θ T + i ∂ jµ + . . . , (4.75)
3 3
where . . . stand for the supersymmetric completion and we have defined the trace
α̇
operators T ≡ T µµ and Sα ≡ σαµα̇ S µ . All in all, the FZ superfield contains a (in
general non-conserved) R-current jµ , a symmetric and conserved Tµν , a conserved
Sαµ , and a complex scalar x. That jµ is a R-current follows from the fact that it sits in
the same superfield where Sαµ sits and so it does not commute with supersymmetry.
From the above expression one can also see that whenever X vanishes the current
jµ becomes conserved and all trace operators vanish. In this case the theory is
conformal and jµ becomes the always present (and conserved) superconformal R-
current.
We have seen that the FZ multiplet contains a non-conserved R-current. What if
a theory admits a U (1)R R-symmetry? For theories with an R-symmetry there exists

78
an alternative supermultiplet accommodating the energy-momentum tensor and the
supercurrent, the so-called R multiplet. It turns out this is again defined in terms
of a pair of superfields (Rµ , χα ) which now satisfy a different on-shell condition
α̇
2 D σαµα̇ Rµ = χα , (4.76)

where Rµ is a real vector superfield and χα a chiral superfield which, besides Dα̇ χα =
0, also satisfies the identity Dα̇ χα̇ − Dα χα = 0. This implies that ∂ µ Rµ = 0, from
which it follows that the lowest component of Rµ is now a conserved current, the R-
current jµR . The component expression of the superfields making-up the R multiplet
reads
 
R ν 1 ρ σ ρσ
Rµ = jµ + θSµ + θ̄ S µ + θσ θ̄ 2Tµν + εµνρσ (∂ j + C ) + . . . (4.77)
2

and

α̇
χα = −2Sα − 4δαβ T + 2i (σ ρ σ̄ τ )βα Cρτ θβ + 2i θ2 σαν α̇ ∂ν S + . . . (4.78)

where again . . . stand for the supersymmetric completion, and Cµν is a closed two-
form. That jµR is an R-current can be easily seen noticing that the current algebra
 
now reads Qα , jµR = Sαµ . Taking the time-component and integrating, this implies
R   R
that dt Qα , QR = dt Qα , which is what is expected for a R-symmetry, recall
eq. (2.78). Notice, finally, that when X = 0, the FZ multiplet (4.74) becomes
a (special instance of an) R-multiplet, one for which χα = 0. Indeed, its lowest
component jµ becomes now the conserved superconformal R-current.
The FZ and R multiplets are the more common supercurrent multiplets. How-
ever, there are instances in which a theory does not admit a R-symmetry (and hence
the R multiplet cannot be defined) and the FZ multiplet is not a well-defined oper-
ator, e.g. it is not gauge invariant. In these cases, one should consider yet another
multiplet where the supercurrent can sit, the so-called S multiplet, which is bigger
than the two above. We will not discuss the S multiplet here, and refer to the
references given at the end of this lecture. On the contrary, there exist theories in
which both the FZ and the R multiplets can be defined. In such cases it turns out
that the two are related by a so-called shift transformation defined as
1   1 2 3 2
Rµ = Jµ + σ̄µα̇α Dα , Dα̇ U , X=− D U , χα = D Dα U , (4.79)
4 2 2
where U is a real superfield associated to a non-conserved (and non-R) current.

79
4.7 Exercises
1. Prove identities (4.20).

2. Check that the differential operators Qα and Qα̇ (4.36) close the supersymme-
try algebra (4.37).
Hint: recall that all θ’s and θ̄’s anti-commute between themselves, and that
∂ ∂aj ∂
{ai , aj } = 0 −→ aj = − aj , (4.80)
∂ai ∂ai ∂ai
which implies that, e.g.
{∂α , θ̄γ̇ } = 0 , {∂α , θβ } = δαβ , {∂¯α̇ , θ̄β̇ } = δα̇β̇ . (4.81)

3. Check that the covariant derivatives Dα and Dα̇ (4.44) anticommute between
themselves and with the supercharge operators (4.36).

4. Compute how the field components of an anti-chiral superfield Ψ transform

under supersymmetry transformations. Show that if Ψ = Φ one gets the
hermitian conjugate of the transformations (4.59).

5. Compute the supersymmetric variation of a vector superfield in the WZ gauge,

and find the explicit form of the chiral superfield Φ which, via a compensating
supersymmetric gauge transformation, brings the vector superfield back to the
WZ gauge.

References

[1] A. Bilal, Introduction to supersymmetry, Section 4, arXiv:hep-th/0101055.

[2] L. Castellani, R. D’Auria and P. Frè, Supergravity And Superstrings: A Geomet-

ric Perspective. Vol. 1: Mathematical Foundations, Chapter II.2.5 and II.2.6,
Singapore: World Scientific (1991).

[3] J. D. Lykken, Introduction to Supersymmetry, Sections 2 and 3, TASI 96,

arXiv:hep-th/9612114.

[4] Z. Komargodski and N. Seiberg, Comments on Supercurrent Multiplets, Super-

symmetric Field Theories and Supergravity, Section 1, arXiv:1002.2228 [hep-th].

[5] T. T. Dumitrescu and Z. Komargodski, Aspects of supersymmetry and its break-

ing, Sections 2 and 4, Nucl. Phys. Proc. Suppl. 216 (2011) 44.

80
5 Supersymmetric actions: minimal supersymmetry

In the previous lecture we have introduced the basic superfields one needs to
construct N = 1 supersymmetric theories, if one is not interested in describing
gravitational interactions. We are now ready to look for supersymmetric actions
describing the dynamics of these superfields. We will first concentrate on matter
actions and construct the most general supersymmetric action describing the inter-
action of a set of chiral superfields. Then we will introduce SuperYang-Mills theory
which is nothing but the supersymmetric version of Yang-Mills. Finally, we will
couple the two sectors with the final goal of deriving the most general N = 1 super-
symmetric action describing the interaction of radiation with matter. In all these
cases, we will consider both renormalizable as well as non-renormalizable theories,
the latter being relevant to describe effective low energy theories.
Note: in what follows we will deal with gauge theories, and hence gauge groups,
like SU (N ) and alike. In order to avoid confusion, in the rest of these lectures we
will use calligraphic N when referring to the number of supersymmetry, N = 1, 2 or
4.

5.1 N = 1 Matter actions

Following the general strategy outlined in §4.3 we want to construct a supersym-
metric invariant action describing the interaction of a (set of) chiral superfield(s).
Let us first notice that a product of chiral superfields is still a chiral superfield and a
product of anti-chiral superfields is an anti-chiral superfield. Conversely, the product
of a chiral superfield with its hermitian conjugate (which is anti-chiral) is a (very
special, in fact) real superfield.
Let us start analyzing the theory of a single chiral superfield Φ. Consider the
following integral Z
d2 θ d2 θ̄ ΦΦ . (5.1)

This integral satisfies all necessary conditions to be a supersymmetric Lagrangian.

First, it is supersymmetric invariant (up to total space-time derivative) since it is the
integral in superspace of a superfield. Second, it is real and a scalar object. Indeed,
the first component of ΦΦ is φφ which is real and a scalar. Now, the θ2 θ̄2 component
of a superfield, which is the only term contributing to the above integral, has the
same tensorial structure as its first component since θ2 θ̄2 does not have any free

81
space-time indices and is real, that is (θ2 θ̄2 )† = θ2 θ̄2 . Finally, the above integral has
also the right physical dimensions for being a Lagrangian, i.e. [M ]4 . Indeed, from
the expansion of a chiral superfield, one can see that θ and θ̄ have both dimension
[M ]−1/2 (compare the first two components of a chiral superfield, φ(x) and θψ(x),
and recall that a spinor in four dimensions has physical dimension [M ]3/2 ). This
means that the θ2 θ̄2 component of a superfield Y has dimension [Y ] + 2 if [Y ] is
the dimension of the superfield (which is that of its lowest component). Since the
dimension of ΦΦ is 2, it follows that its θ2 θ̄2 component has dimension 4 (notice that,
R
consistently, d2 θd2 θ̄ θ2 θ̄2 is dimensionless since dθ (dθ̄) has opposite dimensions
with respect to θ (θ̄), given that the differential is equivalent to a derivative, for
Grassman variables). Summarizing, eq. (5.1) is an object of dimension 4, is real, and
transforms as a total space-time derivative under SuperPoincaré transformations.
To perform the integration in superspace one can start from the expression of
Φ and Φ in the y (resp. ȳ) coordinate system, take the product of Φ(ȳ, θ̄)Φ(y, θ),
expand the result in the (x, θ, θ̄) space, and finally pick up the θ2 θ̄2 component, only.
The end result is
Z
i

Lkin = d2 θ d2 θ̄ ΦΦ = ∂µ φ ∂ µ φ+ ∂µ ψ σ µ ψ − ψ σ µ ∂µ ψ +F F + total der . (5.2)
2
What we get is precisely the kinetic term describing the degrees of freedom of a
free chiral superfield! In doing so we also see that, as anticipated, the F field is
an auxiliary field, namely a non-propagating degree of freedom. Integrating it out
(which is trivial in this case since its equation of motion is simply F = 0) one gets a
(supersymmetric) Lagrangian describing physical degrees of freedom, only. Notice
that after integrating F out, supersymmetry is realized on-shell, only, namely upon
imposing the equation of motions on the propagating degrees of freedom.
The equations of motion for φ, ψ and F following from the Lagrangian (5.2) can
be easily derived using superfield formalism readily from the expression in super-
space. This might not look obvious at a first sight since varying the action (5.1)
with respect to Φ we would get Φ = 0, which does not provide the equations of
motion we would expect, as it can be easily inferred expanding it in components.
The point is that the integral in eq. (5.1) is a constrained one, since Φ is a chiral
superfield and hence subject to the constraint Dα̇ Φ = 0. One can rewrite the above
integral as an unconstrained one noticing that
Z Z
2 2 1
d θ d θ̄ ΦΦ = d2 θ̄ ΦD2 Φ . (5.3)
4

82
 R
In getting the right hand side we have used the fact that dθα = Dα , up to total
space-time derivative, and that Φ is a chiral superfield (hence Dα Φ = 0). Now,
varying with respect to Φ we get

D2 Φ = 0 , (5.4)

which, upon expansion in (x, θ, θ̄), does correspond to the equations of motion for
φ, ψ and F one would obtain from the Lagrangian (5.2).
That’s great. However, we want to describe interactions, not just a set of freely
propagating fields. Can we have a more general Lagrangian than just (5.1)? Let us
try to consider a more generic function of Φ and Φ, call it K(Φ, Φ), and consider
the integral Z
d2 θ d2 θ̄ K(Φ, Φ) . (5.5)

In order for the integral (5.5) to be a promising object to describe a supersymmetric

Lagrangian, the function K should satisfy a number of properties. First, it should
be a superfield. This ensures supersymmetric invariance. Second, it should be a real
and scalar function. As before, this is needed since a Lagrangian should have these
properties and the θ2 θ̄2 component of K, which is the only one contributing to the
above integral, is a real scalar object, if so is the superfield K. Third, K should
have mass dimension 2, since then its θ2 θ̄2 component will have dimension 4, as a
Lagrangian should have. Finally, K should be a function of Φ and Φ but not of
Dα Φ and Dα̇ Φ. The reason is that, as it can be easily checked, covariant derivatives
would provide θθθ̄θ̄-term contributions giving a higher derivative theory (third order
and higher), which cannot be accepted for a local field theory. It is not difficult to
get convinced that the most general expression for K which is compatible with all
these properties is
∞
X m
K(Φ, Φ) = cmn Φ Φn where cmn = c∗nm . (5.6)
m,n=1

where the reality condition on K is ensured by the relation cmn = c∗nm . Not that all
coefficients cmn with either m or n greater than one have negative mass dimension,
while c11 is dimensionless. This means that, in general, a contribution as that in
eq. (5.5) will describe a supersymmetric but non-renormalizable theory, typically
defined below some cut-off scale Λ. Indeed, the coefficients cmn will be of the form

cmn ∼ Λ2−(m+n) (5.7)

83
with the constant of proportionality being a pure number. The function K is called
Kähler potential. The reason for such fancy name will become clear later (see §5.1.1).
If renormalizability is an issue, the lowest component of K should not contain
operators of dimension bigger than 2, given that the θ2 θ̄2 component has dimension
[K] + 2. In this case all cmn but c11 should vanish and the Kähler potential would
just be equal to ΦΦ, the object we already considered before and which leads to the
renormalizable (but free) Lagrangian (5.2).
In passing, notice that the combination Φ + Φ respects all the physical require-
ments discussed above. However, a term like that would not give any contribution
since its θ2 θ̄2 component is a total derivative. This means that two Kähler potentials
K and K 0 related as
K(Φ, Φ)0 = K(Φ, Φ) + Φ + Φ , (5.8)
are different, but their integrals in full superspace, which is all what matters for us,
are the same
Z Z
0
d x d θ d θ̄ K(Φ, Φ) = d4 x d2 θ d2 θ̄ K(Φ, Φ) .
4 2 2
(5.9)

This is the reason why we did not consider m = 0 or n = 0 in the expansion (5.6)
(recall that products of (anti)chiral superfields is still a (anti)chiral superfield so
(5.8) holds with arbitrary powers of Φ and Φ)).
Thus far, we have not been able to describe any renormalizable interaction,
like non-derivative scalar interactions and Yukawa interactions. How to describe
them? As we have just seen, the simplest possible integral in superspace full-filling
the minimal and necessary physical requirements, ΦΦ, already gives two-derivative
contributions, see eq. (5.2). What can we do, then?
When dealing with chiral superfields, there is yet another possibility to construct
supersymmetric invariant superspace integrals. Let us consider a generic chiral
superfield Σ (which can be obtained from products of Φ’s, in our case). Integrating
it in full superspace would give
Z
d4 x d2 θ d2 θ̄ Σ = 0 , (5.10)

since its θ2 θ̄2 component is a total derivative. Consider instead integrating Σ in half
superspace Z
d4 x d2 θ Σ . (5.11)

84
Differently from the previous one, this integral does not vanish, since now it is
the θ2 component which contributes, and this is not a total derivative for a chiral
superfield. Note that in computing (5.11) one can work in the (y, θ, θ̄) coordinate
space, taking Σ = Σ(y, θ), and then evaluate the result at y µ = xµ . The terms one
is missing would just provide total space-time derivatives, which do not contribute
R
to d4 x. Another way to reach the same conclusion is to notice that in (xµ , θα , θ̄α̇ )
coordinate, the chiral superfield Σ reads Σ(x, θ, θ̄) = exp(iθσ µ θ̄∂µ )Σ(x, θ).
Besides being non-vanishing, (5.11) is also supersymmetric invariant, since the
2
θ component of a chiral superfield transforms as a total derivative under supersym-
metry transformations, as can be seen from eq. (4.59)
An integral like (5.11) is more general than an integral like (5.5). The reason
is the following. Any integral in full superspace can be re-written as an integral in
half superspace. Indeed, for any superfield Y
Z Z
4 2 2 1 2
d x d θ d θ̄ Y = d4 x d2 θ D Y , (5.12)
4
2
(in passing, let us notice that for any arbitrary Y , D Y is manifestly chiral, since
3
D = 0 identically). This is because when going from dθ̄ to D the difference is just
a total space-time derivative, which does not contribute to the above integral. On
the other hand, the converse is not true in general. Consider a term like
Z
d4 x d2 θ Φn , (5.13)

where Φ is a chiral superfield. This integral cannot be converted into an integral in

full superspace, essentially because there are no covariant derivatives to play with.
Integrals like (5.13), which cannot be converted into integral in full superspace, are
called F-terms. All others, like (5.12), are called D-terms.
Coming back to our problem, it is clear that since the simplest non-vanishing
integral in full superspace, eq. (5.1), already contains field derivatives, we must turn
to F-terms. First notice that any holomorphic function of Φ, namely a function
W (Φ) such that ∂W/∂Φ = 0, is a chiral superfield, if so is Φ. Indeed
∂W ∂W
Dα̇ W (Φ) = Dα̇ Φ + Dα̇ Φ = 0 . (5.14)
∂Φ ∂Φ
The proposed term for describing interactions in a theory of a chiral superfield is
Z Z
Lint = d θ W (Φ) + d2 θ̄ W (Φ) ,
2
(5.15)

85
where W is a holomorphic function of Φ and the hermitian conjugate has been
added to make the whole thing real. The function W is called superpotential. Which
properties should W satisfy? First, as already noticed, W should be a holomorphic
function of Φ. This ensures it to be a chiral superfield and hence (5.15) to be a
supersymmetric invariant quantity (modulo total space-time derivatives). Second,
W should not contain covariant derivatives since Dα Φ is not a chiral superfield, given
that Dα and Dα̇ do not (anti)commute. Finally, [W ] = 3, to make the expression
(5.15) have dimension 4. The upshot is that the superpotential should have an
expression like
X∞
W (Φ) = an Φn (5.16)
n=1

If renormalizability is an issue, the lowest component of W should not contain opera-

tors of dimensionality bigger than 3, given that the θ2 component has dimensionality
[W ] + 1. Since Φ has dimension one, it follows that to avoid non-renormalizable op-
erators the highest power in the expansion (5.16) should be n = 3, so that the θ2
component will have operators of dimension 4, at most. In other words, a renormal-
izable superpotential should be at most cubic.
The superpotential is also constrained by R-symmetry. Given a chiral superfield
Φ, if the U (1)R charge of its lowest component φ is r, then that of ψ is r − 1 and
that of F is r − 2. This follows from the commutation relations (2.78). Therefore,
we have
R[θ] = 1 , R[θ̄] = −1 , R[dθ] = −1 , R[dθ̄] = 1 (5.17)
(recall that dθ = ∂/∂θ, and similarly for θ̄). In theories with a R-symmetry, it the
follows that the superpotential should have R-charge equal to 2

R[W ] = 2 , (5.18)

in order for the Lagrangian (5.15) to have R-charge 0 and hence be R-symmetry
invariant. Does R-symmetry constraint also the Kähler potential? Let us first notice
that the integral measure in full superspace has R-charge 0, because of eqs. (5.17).
This implies that for theories with a R-symmetry, the Kähler potential should have
itself R-charge 0. This is trivially the case for a canonical Kähler potential, since
ΦΦ has R-charge 0. If one allows for non-canonical Kähler potential, then besides
the reality condition, one should also impose that cnm = 0 whenever n 6= m, see
eq. (5.6).

86
 The integration in superspace of the Lagrangian (5.15) is easily done recalling
the expansion of the superpotential in powers of θ. We have
 
√ ∂W ∂W 1 ∂ 2W
W (Φ) = W (φ) + 2 θψ − θθ F+ ψψ , (5.19)
∂φ ∂φ 2 ∂φ∂φ

where
∂ nW ∂ nW
≡ . (5.20)
∂φn ∂Φn Φ=φ

So, modulo total space-time derivatives, we get

Z Z
2 2 ∂W 1 ∂ 2W
Lint = d θ W (Φ) + d θ̄ W (Φ) = − F − ψψ + h.c. , (5.21)
∂φ 2 ∂φ∂φ
where, again, the rhs is already evaluated at xµ .
Summing up (5.5) and (5.15) we can now write down the most generic N = 1
supersymmetric Lagrangian describing the dynamics of a chiral superfield Φ, which
reads Z Z Z
2 2
L = d θ d θ̄ K(Φ, Φ) + d θ W (Φ) + d2 θ̄ W (Φ) .
2
(5.22)

Renormalizability restricts the structure of the Kähler potential and of the super-
potential as
X3
K(Φ, Φ) = ΦΦ , W (Φ) = an Φn . (5.23)
1

In this case, upon integrating in superspace we get

Z Z Z
L = 2 2
d θ d θ̄ ΦΦ + d θ W (Φ) + d2 θ̄ W (Φ)
2
(5.24)

i
∂W 1 ∂ 2W
= ∂µ φ ∂ µ φ + ∂µ ψ σ µ ψ − ψ σ µ ∂µ ψ + F F − F− ψψ + h.c. .
2 ∂φ 2 ∂φ∂φ

We can now integrate the auxiliary fields F and F out by substituting in the La-
grangian their equations of motion which read

∂W ∂W
F = , F = . (5.25)
∂φ ∂φ
Doing so, we get a Lagrangian where only physical fields enter, that is

i
∂W 2 1 ∂ 2W 1 ∂ 2W
Lon−shell = ∂µ φ ∂ µ φ + ∂µ ψ σ µ ψ − ψ σ µ ∂µ ψ − − ψψ − ψψ .
2 ∂φ 2 ∂φ∂φ 2 ∂φ∂φ
(5.26)

87
From this we can read the scalar potential which is
∂W 2
V (φ, φ) = = FF , (5.27)
∂φ

where the last equality holds on-shell, namely upon use of eqs. (5.25).
All what we said so far can be easily generalized to a set of chiral superfields Φi
where i = 1, 2, . . . , n. In this case the most general Lagrangian reads
Z Z Z
2 2 i
L = d θ d θ̄ K(Φ , Φi ) + d θ W (Φ ) + d2 θ̄ W (Φi ) .
2 i
(5.28)

For renormalizable theories we have

1 1
K(Φi , Φi ) = Φi Φi and W (Φi ) = ai Φi + mij Φi Φj + gijk Φi Φj Φk , (5.29)
2 3
where summation over dummy indices is understood (notice that a quadratic Kähler
potential can always be brought to such diagonal form by means of a GL(n, C)
transformation on the most general term Kji Φi Φj , where Kji is a constant hermitian
matrix). In this case the scalar potential reads

Xn
∂W 2
V (φi , φi ) = = F iF i , (5.30)
i=1
∂φi

where
∂W ∂W
Fi = , Fi = . (5.31)
∂φi ∂φi

5.1.1 Non-linear sigma model I

The possibility to deal with non-renormalizable supersymmetric field theories we al-

luded to previously, is not just academic. In fact, one often has to deal with effective
field theories at low energy. The Standard Model itself, though renormalizable, is
best thought of as an effective field theory, valid up to a scale of order the TeV scale
or slightly higher. Not to mention other effective field theories which are relevant
beyond the realm of particle physics. In this section we would like to say some-
thing more about the Lagrangian (5.28) once the most general Kähler potential and
superpotential are allowed, and show that what one ends-up with in this case is a
supersymmetric version of a non-linear σ-model. Though a bit heavy notation-wise,
the effort we are going to do here will be very instructive as it will show the deep
relation between supersymmetry and geometry.

88
 Since we do not care about renormalizability here, the superpotential is no more
restricted to be cubic and the Kähler potential is no more restricted to be quadratic
(though it must still be real and with no covariant derivatives acting on the chiral
superfields Φi ). For later purposes it is convenient to define the following quantities

∂ i ∂ j ∂2
Ki = K(φ, φ) , K = K(φ, φ) , K i = K(φ, φ)
∂φi ∂φi ∂φi ∂φj
∂ i ∂2
Wi = W (φ) , W = W i , W ij = W (φ) , W ij = W ij ,
∂φi ∂φi ∂φj
where in the above fromulæ both the Kähler potential and the superpotential are
meant as their restriction to the scalar component of the chiral superfields, while φ
stands for the full n-dimensional vector made out of the n scalar fields φi (similarly
for φ).
Extracting the F-term contribution in terms of the above quantities is pretty
simple. The superpotential can be written as
1
W (Φ) = W (φ) + Wi ∆i + Wij ∆i ∆j , (5.32)
2
where we have defined
√
∆i (y) = Φi − φi (y) = 2θψ i (y) − θθF i (y) , (5.33)

and we get for the F-term

Z Z  
2 i 2 i 1 i j
d θ W (Φ ) + d θ̄ W (Φi ) = −Wi F − Wij ψ ψ + h.c. , (5.34)
2

where, see the comment after eq. (5.11), all quantities on the rhs are evaluated in
xµ .
Extracting the D-term contribution is more tricky (but much more instructive).
Let us first define

∆i (x) = Φi − φi (x) , ∆i (x) = Φi − φi (x) (5.35)

which read
√ i 1
∆j (x) = 2θψ j (x) + iθσ µ θ̄∂µ φj (x) − θθF j (x) − √ θθ∂µ ψ j (x)σ µ θ̄ − θθ θ̄θ̄φj (x)
2 4
√ i 1
∆j (x) = 2θ̄ψ j (x) − iθσ µ θ̄∂µ φj (x) − θ̄θ̄F j (x) + √ θ̄θ̄θσ µ ∂µ ψ j (x) − θθ θ̄θ̄φj (x) .
2 4

89
Note that ∆i ∆j ∆k = ∆i ∆j ∆k = 0. With these definitions the Kähler potential can
be written as follows
1 1
K(Φ, Φ) = K(φ, φ) + Ki ∆i + K i ∆i + Kij ∆i ∆j + K ij ∆i ∆j + Kij ∆i ∆j +
2 2
1 k i j 1 1
+ Kij ∆ ∆ ∆k + Kkij ∆i ∆j ∆k + Kijkl ∆i ∆j ∆k ∆l . (5.36)
2 2 4
We can now compute the D-term contribution to the Lagrangian. We get
Z
1 1 1 1
d2 θ d2 θ̄ K(Φ, Φ) = − Ki φi − K i φi − Kij ∂µ φi ∂ µ φj − K ij ∂µ φi ∂ µ φj +
4 4 4 4
 
j i 1 i µ i i µ i i µ
+ Ki F F j + ∂µ φ ∂ φj − ψ σ ∂µ ψ j + ∂µ ψ σ ψ j
2 2 2
i k i µ
i
+ Kij ψ σ ψ k ∂µ φj + ψ j σ µ ψ k ∂µ φi − 2iψ i ψ j F k − Kkij (h.c.) +
4 4
1 kl i j
+ K ψ ψ ψk ψl , (5.37)
4 ij
up to total derivatives. Notice now that

K(φ, φ) = Ki φi + K i φi + 2Kij ∂µ φj ∂ µ φi + Kij ∂µ φi ∂ µ φj + K ij ∂µ φi ∂ µ φj . (5.38)

Using this identity we can eliminate Kij and K ij , and rewrite eq. (5.37) as
Z  
2 2 j i i µ i i µ i i µ
d θ d θ̄ K(Φ, Φ) = Ki F F j + ∂µ φ ∂ φj − ψ σ ∂µ ψ j + ∂µ ψ σ ψ j +
2 2
i k i µ
i
+ Kij ψ σ ψ k ∂µ φj + ψ j σ µ ψ k ∂µ φi − 2iψ i ψ j F k − Kkij (h.c.) +
4 4
1 kl i j
+ K ψ ψ ψk ψl , (5.39)
4 ij
again up to total derivatives.
A few important comments are in order. As just emphasized, independently
whether the fully holomorphic and fully anti-holomorphic Kähler potential compo-
nents, Kij and K ij respectively, are or are not vanishing, they do not enter the final
result (5.39). In other words, from a practical view point it is as if they are not there.
The only two-derivative contribution entering the effective Lagrangian is hence Kij .
This means that given any holomorphic function of φ, Λ(φ), the transformation

K(φ, φ) → K(φ, φ) + Λ(φ) + Λ(φ) , (5.40)

90
known as Kähler transformation, is a symmetry of the theory (in fact, such symmetry
applies to the full Kähler potential, as we have already observed). This is important
for our second comment.
The function Kij which normalizes the kinetic term of all fields in eq. (5.39), is
hermitian, i.e. Kij = Kji † , since K(φ, φ) is a real function. Moreover, it is positive
definite and non-singular, because of the correct sign for the kinetic terms of all
non-auxiliary fields. That is to say Kij has all necessary properties to be interpreted
as a metric of a manifold M of complex dimension n whose coordinates are the
scalar fields φi themselves. The metric Kij is in fact the second derivative of a (real)
scalar function K, since
∂2
Kij = K(φ, φ) . (5.41)
∂φi ∂φj
In this case we speak of a Kähler metric and the manifold M is what mathematically
is known as Kähler manifold. The scalar fields are maps from space-time to this
Riemanniann manifold, which supersymmetry dictates to be Kähler. This is the
(supersymmetric) σ-model. Actually, in order to prove that the Lagrangian is a
σ-model, with target space the Kähler manifold M, we should prove that not only
the kinetic term but any other term in the Lagrangian can be written in terms of
geometric quantities defined on M, e.g. the affine connection and the curvature
tensor. With some work one can compute both of them out of the Kähler metric
Kij and, using the auxiliary field equations of motion
1
F i = (K −1 )ik W k − (K −1 )ik Klm
k
ψl ψm (5.42)
2
(remark: the above equation shows that when the Kähler potential is non-canonical,
the auxiliary fields can depend also on fermion fields!), get for the Lagrangian
 
j i i i µ
L = Ki ∂µ φ ∂ φj + Dµ ψ σ ψ j − ψ σ Dµ ψ j − (K −1 )ij Wi W j
i µ i µ
(5.43)
2 2
1
1
1 kl i j
− Wij − Γkij Wk ψ i ψ j − W ij − Γij
kW
k
ψ i ψ j + Rij ψ ψ ψk ψl ,
2 2 4
where
V (φ, φ) = (K −1 )ij Wi W j (5.44)
is the scalar potential, and the covariant derivatives for the fermions are defined as

Dµ ψ i = ∂µ ψ i + Γijk ∂µ φi ψ k

Dµ ψ i = ∂µ ψ i + Γkj
i ∂µ φk ψ j .

91
With our conventions on indices, Γijk = (K −1 )im Kjk
m
while Γkj
i = (K
−1 m kj
)i Km and
kl kl m −1 n kl
Rij = Kij − Kij (K )m Kn .
As anticipated, a complicated component field Lagrangian is uniquely charac-
terized by the geometry of the target space. Once a Kähler potential is specified,
anything in the Lagrangian (masses and couplings) depends geometrically on this
potential (and on W ). This shows the strong connection between supersymmetry
and geometry. There are of course infinitely many Kähler metrics and therefore
infinitely many N = 1 supersymmetric σ-models. The normalizable case, Kij = δij ,
is just the simplest such instances.
The more supersymmetry the more constraints, hence one could imagine that
there should be more restrictions on the geometric structure of the σ-model for
theories with extended supersymmetry. This is indeed the case, as we will see
explicitly when discussing the N = 2 version of the supersymmetric σ-model. In
this case, the scalar manifold is further restricted to be a special class of Kähler
manifolds, known as special-Kähler manifolds. For N = 4 constraints are even
sharper. In fact, in this case the Lagrangian turns out to be unique, the only
possible scalar manifold being the trivial one, M = R6n , if n is the number of
N = 4 vector multiplets (recall that a N = 4 vector multiplet contains six scalars).
So, for N = 4 supersymmetry the only allowed Kähler potential is the canonical
one! As we will discuss later, this has drastic consequences on the quantum behavior
of theories with N = 4 supersymmetry.

5.2 N = 1 SuperYang-Mills
We would now like to find a supersymmetric invariant action describing the dynamics
of vector superfields. In other words, we want to write down the supersymmetric
version of Yang-Mills theory (SYM for short). Let us start considering an abelian
theory, with gauge group G = U (1). The basic object we should play with is
the vector superfield V , which is the supersymmetric extension of a spin one field.
Notice, however, that the vector v µ appears explicitly in V so the first thing to do is
to find a suitable supersymmetric generalization of the field strength, which is the
gauge invariant object which should enter the action. Let us define the following
superfield
1 1
Wα = − DDDα V , W α̇ = − DDDα̇ V , (5.45)
4 4

92
and see if this can do the job. First, Wα is obviously a superfield, since V is a
superfield and both Dα̇ and Dα commute with supersymmetry transformations. In
3
fact, Wα is a chiral superfield, since D = 0 identically. The chiral superfield Wα is
invariant under the gauge transformation (4.62). Indeed
1
1 β̇
Wα → Wα − DDDα Φ + Φ = Wα + D Dβ̇ Dα Φ
4 4
1 β̇ i β̇
= Wα + D {Dβ̇ , Dα }Φ = Wα + σαµβ̇ ∂µ D Φ = Wα . (5.46)
4 2
This also means that, as anticipated, as far as we deal with Wα , we can stick
to the WZ-gauge without bothering about compensating gauge transformations or
anything.
In order to find the component expression for Wα it is useful to use the (y, θ, θ̄)
coordinate system, momentarily. In the WZ gauge the vector superfield reads
1
VW Z = θσ µ θ̄ vµ (y) + iθθ θ̄λ(y) − iθ̄θ̄ θλ(y) + θθ θ̄θ̄ (D(y) − i∂µ v µ (y)) . (5.47)
2
It is a simple exercise we leave to the reader to prove that expanding in (x, θ, θ̄)
coordinate system, the above expression reduces to eq. (4.65). Acting with Dα
written in the (y, θ, θ̄) coordinate system, we get
β̇
Dα VW Z = σαµβ̇ θ̄β̇ vµ + 2iθα θ̄λ − iθ̄θ̄λα + θα θ̄θ̄D + 2i(σ µν )αβ θβ θ̄θ̄∂µ vν + θθθ̄θ̄σαµβ̇ ∂µ λ
(5.48)
and finally


Wα = −iλα + θα D + i (σ µν θ)α Fµν + θθ σ µ ∂µ λ α , (5.49)
where Fµν = ∂µ vν − ∂ν vµ is the usual gauge field strength and y-dependence of all
fields is understood. Since it contains the field strength Fµν it seems this is the right
superfield we were searching for! Wα is the so-called supersymmetric field strength
and it is an instance of a chiral superfield whose lowest component is not a scalar
field, as we have been used to, but in fact a Weyl fermion, λα , the gaugino. For this
reason, Wα is also called gaugino superfield.
Given that Wα is a chiral superfield, a putative supersymmetric Lagrangian could
be constructed out of the following integral in chiral superspace
Z
d2 θ W α W α , (5.50)

which, notice, has dimension 4. Plugging eq. (5.49) into the expression above and
computing the superspace integral one gets after some simple algebra
Z
1 i
d2 θ W α Wα = − Fµν F µν − 2iλσ µ ∂µ λ + D2 + µνρσ Fµν Fρσ . (5.51)
2 4

93
One can get a real object by adding the hermitian conjugate to (5.51), having finally
Z Z
α̇
Lgauge = d θ W Wα + d2 θ̄ W α̇ W = −Fµν F µν − 4iλσ µ ∂µ λ + 2D2 . (5.52)
2 α

This is the supersymmetric version of the abelian gauge Lagrangian (up to an overall
normalization to be fixed later). As anticipated, D is an auxiliary (real) field.
The Lagrangian (5.52) has been written as an integral over chiral superspace, so
one might be tempted to say it is a F-term. This is wrong since (5.52) is not a true
F-term. Indeed, it can be re-written as an integral in full superspace (while F-terms
cannot) Z Z
d2 θ W α W α = d2 θ d2 θ̄ Dα V · Wα , (5.53)

and so it is in fact a D-term. As we will see later, this fact has important conse-
quences at the quantum level, when discussing renormalization properties of super-
symmetric Lagrangians.
All what we said, so far, has to do with abelian interactions. What changes if
we consider a non-abelian gauge group G? First, we have to promote the vector
superfield to
V = Va T a a = 1, . . . , dim G , (5.54)
where T a are hermitian generators and Va are n = dim G vector superfields. Second,
it is useful to define the finite version of the gauge transformation (4.62) which can
be written as
eV → eiΛ eV e−iΛ . (5.55)
One can easily check that at leading order in Λ this indeed reduces to (4.62), upon
the identification Φ = −iΛ. Again, it is straightforward to set the WZ gauge for
which
1
eV = 1 + V + V 2 . (5.56)
2
In what follows this gauge choice is always understood. The gaugino superfield is
generalized as follows
1
1

Wα = − DD e−V Dα eV , W α̇ = − DD eV Dα̇ e−V (5.57)
4 4
which again reduces to the expression (5.45) to first order in V . Let us look at

94
eq. (5.57) more closely. Under the gauge transformation (5.55) Wα transforms as
1 h  i
Wα → − DD eiΛ e−V e−iΛ Dα eiΛ eV e−iΛ
4
1 


= − DD eiΛ e−V Dα eV e−iΛ + eV Dα e−iΛ
4
1

= − eiΛ DD e−V Dα eV e−iΛ = eiΛ Wα e−iΛ , (5.58)
4
where we used the fact that, given that Λ (and products thereof) is a chiral superfield,
Dα̇ e−iΛ = 0, Dα eiΛ = 0 and also DDDα e−iΛ = 0. The end result is that Wα
transforms covariantly under a finite gauge transformation, as it should whenever
one is dealing with non-abelian groups. Similarly, one can prove that

W α̇ → eiΛ W α̇ e−iΛ . (5.59)

Let us now expand Wα in component fields. We would expect the non-abelian

generalization of eq. (5.49). We have
   
1 1 2 1 2
Wα = − DD 1 − V + V Dα 1 + V + V
4 2 2
1 1 1
= − DDDα V − DDDα V 2 + DDV Dα V
4 8 4
1 1 1 1
= − DDDα V − DDV Dα V − DDDα V · V + DDV Dα V
4 8 8 4
1 1
= − DDDα V + DD [V, Dα V ] .
4 8
The first term is the same as the one we already computed in the abelian case. As
for the second term we get
 
1 1 µν i µ β̇
DD [V, Dα V ] = (σ θ)α [vµ , vν ] − θθσαβ̇ vµ , λ . (5.60)
8 2 2
Adding everything up simply amounts to turn ordinary derivatives into covariant
ones, finally obtaining


Wα = −iλα (y) + θα D(y) + i (σ µν θ)α Fµν + θθ σ µ Dµ λ(y) α (5.61)

with
i i
Fµν = ∂µ vν − ∂ν vµ − [vµ , vν ] , Dµ = ∂µ − [vµ , ] , (5.62)
2 2
which provide the correct non-abelian generalization for the field strength and the
(covariant) derivatives.

95
 In view of coupling the pure SYM Lagrangian with matter, it is convenient to
introduce the coupling constant g explicitly, making the redefinition

V → 2gV ⇔ vµ → 2gvµ , λ → 2gλ , D → 2gD , (5.63)

which implies the following changes in the final Lagrangian. First, we have now

Fµν = ∂µ vν − ∂ν vµ − ig [vµ , vν ] , Dµ = ∂µ − ig [vµ , ] . (5.64)

Moreover, the gaugino superfield (5.61) should be multiplied by 2g and the (non-
abelian version of the) Lagrangian (5.52) by 1/4g 2 . The end result for the SYM
Lagrangian is
 Z 
1 2 α
LSY M = Im τ d θ TrW Wα
32π
h 1 1 i θY M 2
= Tr − Fµν F µν − iλσ µ Dµ λ + D2 + g TrFµν Feµν , (5.65)
4 2 32π 2
where we have introduced the complexified gauge coupling
θY M 4πi
τ= + 2 (5.66)
2π g
and the dual field strength
1
Feµν = µνρσ Fρσ , (5.67)
2
while gauge group generators are normalized as Tr T a T b = δ ab .

5.3 N = 1 Gauge-matter actions

We want now to couple radiation with matter in a supersymmetric consistent way.
To this end, let us consider a chiral superfield Φ transforming in some representation
R of the gauge group G, T a → (TRa )ij where i, j = 1, 2, . . . , dimR. Under the gauge
transformation (5.55) we expect Φ to transform as

Φ → Φ0 = eiΛ Φ , Λ = Λa TRa . (5.68)

Note that since Λ is a chiral superfield, Φ0 is still a chiral superfield. This looks
promising but in this way it turns out that the chiral superfield kinetic action we
have derived previously would not be gauge invariant since

ΦΦ → Φe−iΛ eiΛ Φ 6= ΦΦ . (5.69)

96
This means that we have to change the kinetic action. As we are going to show in
the following, the correct expression for the kinetic term happens to be

ΦeV Φ , (5.70)

which can be easily shown to be a supersymmetric invariant quantity (modulo total

space-time derivatives) and also gauge invariant, when integrated in superspace.
With this modification, the complete Lagrangian for charged matter hence reads
Z Z Z
Lmatter = d θ d θ̄ Φe Φ + d θ W (Φ) + d2 θ̄ W (Φ) .
2 2 V 2
(5.71)

Obviously the superpotential should be compatible with the gauge symmetry, i.e.
it should be gauge invariant itself. This means that a term like

ai1 i2 ...in Φi1 Φi2 . . . Φin (5.72)

is allowed only if ai1 i2 ...in is an invariant tensor of the gauge group and if R×R×· · ·×R
n times contains the singlet representation of the gauge group G.
As an explicit example, take the gauge group of strong interactions, G = SU (3),
and consider quarks as matter field. In this case R is the fundamental representation,
R = 3. Since 3 × 3 × 3 = 1 + . . . and ijk is an invariant tensor of SU (3), while
1 6⊂ 3×3 it follows that a supersymmetric and gauge invariant cubic term is allowed,
but a mass term is not. In order to have mass terms for quarks, one needs R = 3 + 3̄
corresponding to a chiral superfield Φ in the 3 (quark) and a chiral superfield Φ e in
e is gauge invariant and does correspond to a mass
the 3̄ (anti-quark). In this case ΦΦ
term. This is consistent with the fact that a chiral superfield contains a Weyl fermion
only and quarks are described by Dirac fermions. The lesson is that to construct
supersymmetric actions with colour charged matter, one needs to introduce two sets
of chiral superfields which transform in conjugate representations of the gauge group.
This is just the supersymmetric version of what happens in ordinary QCD or in any
non-abelian gauge theory with fermions transforming in complex representations
(G = SU (2) is an exception because 2 ' 2̄) .
Let us now compute the D-term of the Lagrangian (5.71). We have (as usual we
work in the WZ gauge)
1
ΦeV Φ = ΦΦ + ΦV Φ + ΦV 2 Φ . (5.73)
2

97
The first term is the one we have already calculated, so let us focus on the D-term
contribution of the other two. After some algebra we get
i µ i 1 i i 1
ΦV Φ|θθ θ̄θ̄ = φv ∂µ φ − ∂µ φv µ φ − ψσ̄ µ vµ ψ + √ φλψ − √ ψλφ + φDφ
2 2 2 2 2 2
1 µ
ΦV 2 Φ|θθ θ̄θ̄ = φv vµ φ .
2
Putting everything together we finally get (up to total derivatives)

i i 1
ΦeV Φ|θθ θ̄θ̄ = Dµ φ Dµ φ − iψσ̄ µ Dµ ψ + F F + √ φλψ − √ ψλφ + φDφ , (5.74)
2 2 2

where Dµ = ∂µ − 2i vµa TRa .

Performing the rescaling V → 2gV and rewriting ψσ̄ µ Dµ ψ = ψσ µ Dµ ψ (recall
the spinor identity χσ µ ψ = −ψσ̄ µ χ) we get finally

√ √
Φe2gV Φ|θθ θ̄θ̄ = Dµ φ Dµ φ−iψσ µ Dµ ψ+F F +i 2gφλψ−i 2gψλφ+gφDφ , (5.75)

where now Dµ = ∂µ −igvµa TRa . The above result shows that the D-term in the La-
grangian (5.71) not only provides matter kinetic terms but also interaction terms
between matter fields φ, ψ and gauginos λ, where it is understood that

φλψ = φi (TRa )ij λa ψ j , (5.76)

and similarly for the other couplings.

To get the most general action there is one term still missing: the so called Fayet-
Iliopulos term. Suppose that the gauge group is not semi-simple, i.e. it contains
U (1) factors. Let V A be the vector superfields corresponding to the abelian factors,
A = 1, 2, . . . , m, where m is the number of abelian factors. The D-term of V A
transforms as a total derivative under supersymmetric gauge transformations, since

V A → V A − iΛ + iΛ : DA → DA + ∂µ ∂ µ (. . .) . (5.77)

Therefore a Lagrangian of this type

X Z 1X
LF I = ξA d2 θd2 θ̄ V A = ξA DA (5.78)
A
2 A

is supersymmetric invariant (since V A are superfields) and gauge invariant, modulo

total space-time derivatives.

98
 We can now assemble all ingredients and write down the most general N = 1
supersymmetric Lagrangian (with canonical Kähler potential, hence renormalizable,
if the superpotential is at most cubic) which reads

L = LSY M + Lmatter + LF I =
 Z  X Z
1 2 α
= Im τ d θ TrW Wα + 2g ξA d2 θd2 θ̄ V A +
32π A
Z Z Z
+ 2 2
d θd θ̄ Φe Φ + d θ W (Φ) + d2 θ̄ W (Φ)
2gV 2
(5.79)
h 1 1 2 i θY M 2
µν µ
= Tr − Fµν F − iλσ Dµ λ + D + g TrFµν Feµν
4 2 32π 2
X
√
+ g ξA DA + Dµ φ Dµ φ − iψσ µ Dµ ψ + F F + i 2gφλψ
A

√ ∂W i ∂W 1 ∂ 2W i j 1 ∂ 2W
− i 2gψλφ + gφDφ − F − F i − ψψ − ψψ .
∂φi ∂φi 2 ∂φi ∂φj 2 ∂φi ∂φj i j

Both Da and F i are auxiliary fields and can be integrated out. Their equations of
motion read
∂W
Fi = , Da = −gφT a φ − gξ a (ξ a = 0 if a 6= A) . (5.80)
∂φi
These can plugged back into (5.79) leading to the following on-shell Lagrangian
h 1 i θ
g 2 TrFµν Feµν + Dµ φDµ φ − iψσ µ Dµ ψ +
µν µ YM
L = Tr − Fµν F − iλσ Dµ λ + 2
4 32π
√ √ 1 ∂ W i j 1 ∂ 2W
2
+ i 2gφλψ − i 2gψλφ − ψψ − ψ ψ − V (φ, φ) , (5.81)
2 ∂φi ∂φj 2 ∂φi ∂φj i j

where the scalar potential V (φ, φ) is

∂W ∂W g2 X
V φ, φ = + |φi (T a )ij φj + ξ a |2 = (5.82)
∂φi ∂φi 2 a
1
= F F + D2 ≥0.
2 on the solution
where in the last step we have used eqs. (5.80). The above equation shows that the
potential is a semi-positive definite quantity in supersymmetric theories.
Expressing the potential in terms of auxiliary fields, as above, provides a very
direct way to understand whether a supersymmetric theory admits supersymmetric
vacua and also suggests how to parametrize such vacua.

99
 First recall that a vacuum is a Lorentz invariant state configuration. This means
that all field derivatives and all fields but scalar ones should vanish in a vacuum
state. Hence, the only non trivial thing of the Hamiltonian which can be different
from zero in the vacuum is the non-derivative scalar part, which, by definition, is
the scalar potential. Therefore, the vacua of a theory, which are the minimal energy
states, are in one-to-one correspondence with the (global or local) minima of the
scalar potential.
As we have already seen, in a supersymmetric theory the energy of any state is
semi-positive definite. This holds also for vacua. For a vacuum Ω we have
X

hΩ|P 0 |Ωi ∼ ||Qα |Ωi||2 + ||Q†α |Ωi||2 ≥ 0 . (5.83)
α

This means that the vacuum energy is 0 if and only if it is a supersymmetric state,
that is Qα |Ωi = 0, Qα̇ |Ωi = 0 with α, α̇ = 1, 2. Conversely, supersymmetry is broken
(in the perturbative theory based on this vacuum) if and only if the vacuum energy is
positive. This implies that supersymmetric vacua are in one-to-one correspondence
with the zero’s of the scalar potential. From eq. (5.82) we see that, if they exist,
they are described by the set of scalar field VEVs which solve simultaneously the
so-called D-term and F-term equations

F i (φ) = 0 , Da (φ, φ) = 0 . (5.84)

To find such zero’s, the most convenient thing to do is to look first for the space
of scalar field VEVs such that

Da (φ, φ) = 0 , (5.85)

which is called the space of D-flat directions. If a superpotential is present, one

should then consider the F-term equations, which may put further constraints on
the subset of scalar field VEVs already satisfying the D-term equations (5.85). The
subspace of the space of D-flat directions which is also F-flat, i.e. which also satisfies
the equations
i
F (φ) = 0 , (5.86)
is called (classical) moduli space and represents the space of (classical) supersym-
metric vacua, Mcl . Note that in solving for the D-term equations, one should mod
out by gauge transformations, since solutions which are related by gauge transfor-
mations are physically equivalent and describe the same vacuum state.

100
 The space of flat directions is the space of fields the potential does not depend
on and is called moduli space because each flat direction has a massless particle
associated to it, a modulus. The moduli represent the lightest degrees of freedom
of the low energy effective theory (think about the supersymmetric σ-model we
discussed in §5.1.1). As one moves along the moduli space one spans physically
inequivalent (supersymmetric) vacua, since the mass spectrum changes from point
to point as, generically, particle masses will depend on scalar field VEVs.
Let us anticipate an important and far-reaching fact that we will derive later.
While in a non-supersymmetric theory (or in a supersymmetry breaking vacuum of
a supersymmetric theory) the space of classical flat directions, if any, is generically
lifted by radiative corrections (which can be computed at leading order by e.g. the
Coleman-Weinberg potential), in supersymmetric theories this does not happen. If
the ground state energy is zero at tree level, it remains so at all orders in pertur-
bation theory. This is because perturbations around a supersymmetric vacuum are
themselves described by a supersymmetric Lagrangian and quantum corrections are
protected by cancellations between fermionic and bosonic loops. This means that the
only way to lift a classical supersymmetric vacuum, namely to break supersymme-
try, if not at tree level by some cleverly chosen superpotential, are non-perturbative
corrections. We will have much more to say about this issue in later lectures.

5.3.1 Classical moduli space: examples

To make concrete the previous discussion on moduli space, in what follows we would
like to consider two examples explicitly. Before we do that, however, we want to
rephrase our definition of moduli space presenting an alternative (but equivalent)
way to describe it.
Suppose we are considering a theory without superpotential. For such a theory
the space of D-flat directions coincides with the moduli space. The space of D-flat
directions is defined as the set of scalar field VEVs satisfying the D-flat conditions

Mcl = {hφi i / Da = 0 ∀a}/gauge transformations . (5.87)

Generically it is not at all easy to solve the above constraints and find a simple
parametrization of Mcl . An equivalent, though less transparent definition of the
space of D-flat directions can help in this respect. It turns out that the same space
can be defined as the space spanned by all (single trace) gauge invariant operator

101
VEVs made out of scalar fields, modulo classical relations between them

Mcl = {hGauge invariant operators ≡ Xr (φ)i}/classical relations . (5.88)

The latter parametrization is very convenient since, up to classical relations, the

construction of the moduli space is unconstrained. In other words, the gauge in-
variant operators provide a direct parametrization of the space of scalar field VEVs
satisfying the D-flat conditions (5.85).
Notice that if a superpotential is present, this is not the end of the story: F-
equations will put extra constraints on the Xr (φ)’s and may lift part of (or even all)
the moduli space of supersymmetric vacua. In later lectures we will discuss some
such instances in detail. Here, in order clarify the equivalence between definitions
(5.87) and (5.88), we will instead consider two models with no superpotential term.

Massless SQED. The first example we want to consider is (massless) SQED, the su-
persymmetric version of quantum electrodynamics. This is a supersymmetric gauge
ei ) hav-
theory theory with gauge group U (1), F (couples of) chiral superfields (Qi , Q
ing opposite charge with respect to the gauge group (we will set for definiteness
the charges to be ±1) and no superpotential, W = 0. The vanishing of the super-
potential implies that for this system the space of D-flat directions coincides with
the moduli space of supersymmetric vacua. The Lagrangian is an instance of the
general one we derived before and reads
 Z  Z  
1 e† e−2V Q
ei (5.89)
LSQED = Im τ d θ W Wα + d2 θd2 θ̄ Q†i e2V Qi + Q
2 α
i
32π

(in order to ease the notation, we have come back to the most common notation to
indicate hermitian conjugation).
The (only one) D-equation reads

e† Q
D = Q†i Qi − Q ei = 0 (5.90)
i

ei appear.
where here and in the following a h i is understood whenever Qi or Q
What is the moduli space? Let us first use the definition (5.87). The number of
putative complex scalar fields parametrizing the moduli space is 2F . We have one
D-term equation only, which provides one real condition, plus gauge invariance

Qi → eiα Qi , ei → e−iα Q
Q ei , (5.91)

102
which provides another real condition. Therefore, the complex dimension of the
moduli space is
1 1
dimC Mcl = 2F − − = 2F − 1 . (5.92)
2 2
At a generic point of the moduli space the gauge group U (1) is broken. Indeed, the
−1 above corresponds to the complex scalar field which, together with its fermionic
superpartner, gets eaten by the vector superfield to give a massive vector superfield.
One component of the complex scalar field provides the third polarization to the
otherwise massless photon; the other real component provides the real physical
scalar field a massive vector superfield has. Finally, the Weyl fermion provides the
extra degrees of freedom to let the photino become massive. This is nothing but
the supersymmetric version of the Higgs mechanism. As anticipated, the vacua are
physically inequivalent, generically, since e.g. the mass of the photon depends on
the VEV of the scalar fields.
Let us now repeat the above analysis using the definition (5.88). The only gauge
invariants we can construct are

ej ,
M ij = Qi Q (5.93)

the so-called mesons. They look as F 2 degrees of freedom but there are classical
relations between them, that we now want to find. The matrix (5.93) is a symmetric
F × F complex matrix with rank one since so is the rank of Q and Q e (Q and Q e are
vectors of length F , since the gauge group is abelian). This implies that the meson
matrix has only one non-vanishing eigenvalue which means

det (M − λ1) = λF −1 (λ − λ0 )(−1)F . (5.94)

Recalling that for a matrix A

i1 i2 ...iF Aij11 Aij22 . . . AijFF = det A j1 j2 ...jF (5.95)

with i1 i2 ...iF the fully antisymmetric tensor with F indices, we have

i1 i2 ...iF (M ij11 − λδ ij11 ) . . . (M ijFF − λδ ijFF ) = λF −1 (λ − λ0 )(−1)F j1 j2 ...jF (5.96)

which means that from the left hand side only the coefficients of the terms λF and
λF −1 survive. The next contribution, proportional to λF −2 , should vanish, that is

i1 i2 ...iF M ij11 M ij22 j1 j2 ...jF = 0 . (5.97)

103
One can show that of the above complex equations, only (F − 1)2 give indepen-
dent conditions, while the contributions proportional to lower powers of λ, that is
λF −3 , λF −4 , . . . do not give new constraints. So we finally get that

dimC Mcl = F 2 − (F − 1)2 = 2F − 1 . (5.98)

which coincides with what we have found before, eq. (5.92)! A quick way to get
the same result looking at the meson matrix is to observe that being of rank one,
the matirx Mji is fully determined by the first row and the first column, namely by
2F − 1 elements.
The parametrization in terms of (single trace) gauge invariant operators is very
useful if one wants to find the low energy effective theory around the supersym-
metric vacua. Indeed, up to classical relations, these gauge invariant operators (in
fact, their fluctuations) directly parametrize the massless degrees of freedom of the
perturbation theory constructed upon these same vacua.
Using that Q, Qe and that the meson matrix M have rank one, one can show that
on the moduli space (5.90)
√
e† Q
Tr Q† Q = Tr Q e = Tr M †M . (5.99)

Therefore, the Kähler potential, which is canonical in terms of the microscopic UV

degrees of freedom Q and Q,e once projected on the moduli space reads
h i √
† e †e
K = Tr Q Q + Q Q = 2Tr M † M . (5.100)

The Kähler metric of the non-linear σ-model hence reads

1 1
ds2 = KM M † dM dM † = √ dM dM † (5.101)
2 M M †

which is manifestly non-canonical. Notice that the (scalar) kinetic term

Z
1 1
d4 x √ ∂µ M ∂ µ M † (5.102)
2 †
M M
is singular at the origin, since the Kähler metric diverges there. This has a clear
physical interpretation: at the origin the theory is unhiggsed, the photon becomes
massless and the correct low energy effective theory should include it in the de-
scription. This is a generic feature in this all business: singularities showing-up at
specific points of the moduli space are a signal of extra-massless degrees of freedom
that, for a reason or another, show up at those specific points

Singularities ←→ New massless d.o.f. . (5.103)

104
The correct low energy, singularity-free, effective description of the theory should
include them. The singular behavior of KM M † at the origin is simply telling us that.

Massless SQCD. Let us now consider the non-abelian version of the previous
theory. We have now a non-abelian gauge group which we take for definiteness to
be SU (N ), F flavors and again no superpotential, W = 0. The quarks superfields
Q and Q e are F × N complex matrices. Looking at the Lagrangian, which is the
obvious generalization of (5.89), we see there are two independent flavor symmetries,
one associated to Q and one to Q, e SU (F )L and SU (F )R respectively. To make the
different (flavor and gauge) symmetries manifest we split matter indexes as (i, a) with
i an index in the (anti)fundamental of SU (F )L,R and a in the (anti)fundamental of
SU (N ). This is summarized in the table below

SU (N ) SU (F )L SU (F )R
Qia N F 1 (5.104)
eb
Q N 1 F
j

where i, j = 1, 2, . . . , F and a, b = 1, 2, . . . , N . The convention for gauge indices is

that lower indices are for an object transforming in the fundamental representation
and upper indices for an object transforming in the anti-fundamental. The conven-
tion for flavor indices is chosen to be the opposite one. Given these conventions, the
D-term equations read
i i
b f† (T a )b Q
DA = Q† i (TNA )cb Qci + Q
b f† (T A )b Q
ec = Q† (T A )c Q i − Q ec
b N̄ c i i b c b c i
i
b f† = 0 ,
e b (T A )c Q
= Q† i (T A )cb Qci − Q (5.105)
i b c

where A = 1, 2, . . . , N 2 − 1 is an index in the adjoint of SU (N ), and we used the

fact that (TNA )ab = −(TN̄A )ab ≡ (T A )ab .
Let us first focus on the case F < N . Using the two SU (F ) flavor symmetries
and the (global part of the) gauge symmetry SU (N ), one can show that on the
moduli space (5.105) the matrices Q and Q e can be put, at most, in the following
form (recall that the maximal rank of Q and Qe is F in this case, since F < N )
 
v1 0 ... 0 ...
 0 v2 . . . 0 ...
  eT
Q= =Q (5.106)
. . . ... ... ... ... 
0 0 ... vF ...

105
This means that at a generic point of the moduli space the gauge group is broken
to SU (N − F ). So, the complex dimension of the classical moduli space is
  
dimC Mcl = 2F N − N 2 − 1 − (N − F )2 − 1 = F 2 . (5.107)
Let us now use the parametrization in terms of gauge invariant single trace operators
(5.88). In this case we have
ea
M ij = Qia Q (5.108)
j

(notice the contraction on the N gauge indices). The meson matrix has now maximal
rank, since F < N , so there are no classical constraints it has to satisfy: its F 2
entries are all independent. In terms of the meson matrix the classical moduli space
dimension is then (trivially) F 2 , in agreement with eq. (5.107). Again, playing with
global symmetries, the meson matrix can be diagonalized in terms of F complex
eigenvalues Vi , which, not suprprisingly, turn out to be the square of the ones in
(5.106), Vi = vi2 .
Also in this case one can write down the (classical) effective action. On the
a f† i . Using this identity we get
ea Q
moduli space we have Q† Qi = Q
i b i b

i f† i Q† a Qk Q f† i ea f† k eb
M †M j = Q a k
eb
b j = Q aQ k Q bQ j (5.109)
√
e† Q
which implies Q e = M † M as a matrix equation. So the Kähler potential is
√
K = 2Tr M † M . (5.110)
The Kähler metric is singular whenever the meson matrix M is not invertible. This
does not only happen at the origin of field space as for SQED, but actually on the
 
subspace where some of the N 2 − 1 − (N − F )2 − 1 = (2N − F )F massive gauge
bosons parametrizing the coset SU (N )/SU (N −F ) become massless (and they need
to be included in the low energy effective description).
Let us now consider the case F ≥ N . Following a similar procedure as the one
before, the matrices Q and Q e can be brought to the following form on the moduli
space
   
v1 0 ... 0 ṽ1 0 ... 0
   
 0 v2 . . . 0   0 ṽ2 . . . 0 
   
. . . . . . . . . . . .  . . . . . . . . . . . . 
   
  e T  
Q= 0 0 . . . vN  , Q =  0 0 . . . ṽN  (5.111)
   
 0 0 0 0    
  0 0 0 0 
. . . . . . . . . . . .  . . . . . . . . . . . . 
   
0 0 0 0 0 0 0 0

106
where |vi |2 − |ṽi |2 = a, with a a i-independent number. Since F ≥ N , at a generic
point on the moduli space the gauge group is now completely higgsed. Therefore,
the dimension of the classical moduli space is now


dimC Mcl = 2N F − N 2 − 1 . (5.112)

The parametrization in terms of gauge invariant operators is slightly more involved,

in this case. The mesons are still there, and defined as in eq. (5.108). However,
there are non-trivial classical constraints one should take into account, since the
rank of the meson matrix, which is at most N , is now smaller than its dimension, F ,
like in the SQED example. Moreover, besides the mesons, there are now new gauge
invariant single trace operators one can build, the baryons, which are operators
e with fully anti-symmetrized indices.
made out of N fields Q respectively N fields Q,
As an explicit example of this richer structure, let us apply the above rationale
to the case N = F . According to eq. (5.112), in this case dimC Mcl = F 2 + 1. The
gauge invariant operators are the meson matrix plus two baryons, B and B, e defined
as

B = a1 a2 ...aN Qa11 Qa22 . . . QaNN

e = a1 a2 ...a Q
B ea11 Q
ea22 . . . Q
eaN .
N N

Notice that since F = N the anti-symmetrization on the flavor indices is automat-

ically taken care of, once anti-symmetrization on the gauge indices is imposed. All
in all we have, naively, F 2 + 2 complex moduli space directions. There is however
one classical constraints between them which reads

e=0,
det M − B B (5.113)

as can be easily checked from the definition of the meson matrix (5.108) and that of
the baryons above. Hence, the actual dimension of the moduli space is F 2 + 2 − 1 =
F 2 + 1, as expected. As for the case F < N , there is a subspace in the moduli space,
which includes the origin, where some fields become massless and the low energy
effective analysis should be modified to include them.
All we said, so far, is true classically. As we will see when discussing the quantum
dynamics of SQCD, quantum corrections sensibly change this picture and the exact
structure of the moduli space differs in many respects from the classical one. For
instance, focusing again on F = N SQCD it turns out that the classical constraint
(5.113) is modified at the quantum level. This has the effect of excising the origin

107
of field space from the actual quantum moduli space removing all singular subspace
and the corresponding extra massless degrees of freedom, which are then just an
artifact of the classical analysis, in this case.
We will have much more to say about SQCD and its classical and quantum
properties at some later stage.

5.3.2 Non-linear sigma model II

In section 5.1.1 we discussed the supersymmetric non-linear σ-model for matter

fields, which is relevant to describe supersymmetric low energy effective theories.
Though it is not often the case, it may happen to face effective theories with some left
over propagating gauge degrees of freedom at low energy. Therefore, in this section
we will generalize the σ-model of section 5.1.1 to such a situation: a supersymmetric
but non-renormalizable effective theory coupled to gauge fields. Note that the choice
of the gauge group cannot be arbitrary here. In order to preserve the structure of
the non-linear σ-model one can gauge only a subgroup G of the isometry group of
the scalar manifold.
Following previous strategy, one gets easily convinced that the pure SYM part
changes simply by promoting the (complexified) gauge coupling τ to a holomorphic
function of the chiral superfields, getting
Z Z
2 α
τ d θ TrW Wα −→ d2 θ Fab (Φ) W αa Wαb , (5.114)

where the chiral superfield Fab (Φ) should transform in the Adj × Adj of the gauge
group G in order for the whole action to be G-invariant. Notice that for Fab =
τ TrTa Tb one gets back the usual result (recall that we have normalized the gauge
group generators as Tr Ta Tb = δab ). For this reason the function Fab (actually its
restriction to the scalar fields) is dubbed generalized complex gauge coupling.
As for the matter Lagrangian, given what we have already seen, namely that
whenever one has to deal with charged matter fields the gauge invariant combination
is (Φe2gV )i Φi , one should simply observe that the same holds for any real G-invariant
function of Φ and Φ. In other words, the σ-model Lagrangian for charged chiral
superfields is obtained from the one we derived in section 5.1.1 upon the substitution

K(Φi , Φi ) −→ K(Φi , (Φe2gV )i ) . (5.115)

108
The end result is then
Z 
1 2 αa b
L = Im d θ Fab (Φ) W Wα +
32π
Z Z Z
+ d θd θ̄K(Φ , (Φe )i ) + d θ W (Φ ) + d2 θ̄ W (Φi ) . (5.116)
2 2 i 2gV 2 i

By expanding and integrating in superspace one gets the final result. The derivation
is a bit lengthy and we omit it here. Let us just mention some important differences
with respect to our previous results. The gauge part has the imaginary part of
Fab multiplying the kinetic term (the generalized gauge coupling) and the real part
multiplying the instanton term (generalized θ-angle). Moreover, there are higher
order couplings between fields belonging to vector and scalar multiplets which are
proportional to derivatives of Fab with respect to the scalar fields, and which are
obviously absent for the renormalizable Lagrangian (5.79). As for the matter part,
one important difference with respect to the σ-model Lagrangian (5.43) is that all
derivatives are (also) gauge covariantized. More precisely we have

e µ ψ i = ∂µ ψ i − igv a T a ψ i + Γi ∂µ φi ψ k
D µ R jk

e µ ψ j = ∂µ ψ j − igv a T a ψ j + Γki ∂µ φk ψ i ,
D µ R j

which are covariant both with respect to the σ-model metric and the gauge con-
nection. As compared to the Lagrangian (5.79) the Yukawa-like couplings have the
Kähler metric inserted, that is

φλψ −→ Kji φi λψ j = Kji φi TRa λa ψ j = Kji (φi )M (TRa )M j N

N λa (ψ ) , (5.117)

where M, N are gauge indices. Moreover, the term gφDφ is also modified into
gφi DK i , where as usual K i = ∂K/∂φi .
All these changes are important to keep in mind. However, it is worth noticing
that in N = 1 supersymmetry vectors belong to different multiplets with respect to
those where scalars sit. Hence, any geometric operation on the scalar manifold M
will not have much effect on the vectors, and viceversa. In other words, the structure
of the N = 1 non-linear σ-model is essentially unchanged by gauging some of the
isometries of the scalar manifold. This is very different from what happens in models
with extended supersymmetry, as we will see in the next lecture.

109
 After solving for the auxiliary fields which read (with obvious notation)

1 g2 a b
F i = (K −1 )ij W j − Γijk ψ j ψ k − i (K −1 )ij (Fab,j )† λ λ (5.118)
2 16π
 
a 4π −1 b i 2 1  i c

D = − 2 (Im F)ab gφi T K + g √ (Fbc,i )ψ λ + h.c. (5.119)
g 8π 2
one finds for the potential

V (φ, φ) = (K −1 )ij Wi W j + 2π (Im F)−1 a i b j

ab (φi T K ) (φj T K ) , (5.120)

which is the σ-model version of the potential (5.82).

As far as the potential, we cannot resist making a comment which will actually
be relevant later, when we will discuss supersymmetry breaking. Whenever the
effective theory one is dealing with does not have any propagating gauge degrees of
freedom (due to Higgs mechanism, confinement or alike) the scalar potential (5.120)
gets contributions from the first term, only. In this case the zero’s of the potential,
which correspond to the supersymmetric vacua of the theory, are described just by

Wi = 0 , (5.121)

as in cases where the Kähler potential is canonical, since K is a positive definite

matrix (provided the integrating out procedure has been done correctly along the
whole moduli space). This means that it is possible to see whether supersymmetry
is broken/unbroken independently of any knowledge of the Kähler potential! This
is very different from non-supersymmetric σ models and it is related to what was
mentioned earlier, namely that in a supersymmetric theory if supersymmetry is
unbroken classically it cannot be broken at any order in perturbation theory but
only non-perturbatively. This said, other important features which characterize the
vacua, as field VEVs, the exact value of the vacuum energy (if not zero), the mass
and the interactions of the lightest excitations, etc... do depend on K. With an abuse
of notation, eqs. (5.121) are usually referred to as F-term equations, even though
for a theory with non-canonical Kähler potential the correct F-term equations are
eqs. (5.118).

5.4 Exercises
1. Derive eq. (5.2). Using eqs. (4.59), show that the resulting expression trans-
forms as a total space-time derivative under supersymmetry transformations.

110
 2. Compute D2 Φ = 0 and show that the different components provide the equa-
tions of motion for a free massless WZ multiplet.

3. Consider a theory of a chiral superfield Φ with canonical Kähler potential,

K = ΦΦ and superpotential W (Φ) = 12 mΦ2 + 13 gΦ3 . This is the renowned
Wess-Zumino model. Derive the equation of motion in superfield formalism
and the corresponding ones in component fields. Compute the off-shell and
on-shell space-time Lagrangians. Show that once auxiliary fields are integrated
out supersymmetry closes only on-shell, namely that the algebra closes only
upon use of (some of) the equations of motion.

4. Consider the theory of a single chiral superfield Φ and Kähler potential K =

Λ2 ln(1 + ΦΦ/Λ2 ), where Λ is a given mass scale. Compute the off-shell and
on-shell Lagrangians and study the geometry of the one-dimensional super-
symmetric non-linear σ model.

5. Using all possible available symmetries, show that in SU (N ) SQCD with F <
N flavors, the complex scalar field matrices parametrizing the moduli space
can be put in the form (5.106). Using the same procedure, show that the
structure (5.111) holds for F ≥ N .

6. Consider the following matter theories

1
1. K = QQ , W = mQ2
2
2. K = XX + Y Y , W = (X − m) Y 2

3. K = XX + Y Y + ZZ , W = gXY Z
p
4. K = Λ XX , W = λX .

Determine whether there are supersymmetric vacua and, if they exist, compute
the mass spectrum of the theory around them.

References

[1] A. Bilal, Introduction to supersymmetry, Sections 4, 5 and 7, arXiv:hep-

th/0101055.

[2] M. F. Sohnius, Introducing Supersymmetry, Sections 5, 7, 8, 9 and 10, Phys.

Rept. 128 (1985) 39.

111
[3] J. Terning, Modern supersymmetry: Dynamics and duality, Chapter 3.4, Oxford
University Press (2006).

[4] M. J. Strassler, An unorthodox introduction to supersymmetric gauge theory,

Sections 4.5 and 5.1, arXiv:hep-th/0309149.

[5] M. A. Luty and W. Taylor, Varieties of vacua in classical supersymmetric gauge

theories, Phys. Rev. D 53 (1996) 3399 [hep-th/9506098].

112
6 Supersymmetric actions: extended supersymmetry

Until now we have discussed theories with N = 1 supersymmetry. In this lecture

we will discuss the structure of theories with extended supersymmetry. This will
also let us emphasize the basic differences which arise at the quantum level between
theories with different number of supersymmetries.

6.1 N = 2 supersymmetric actions

In this section we would like to construct the most general N = 2 supersymmetric
action in four dimensions. We will follow the same logic of the previous lecture,
but we will not develop the corresponding N = 2 superspace approach, whose
formulation is beyond our scope. Rather, we will use the (by now familiar) N = 1
superspace formalism and see which specific properties does more supersymmetry
impose on an otherwise generic N = 1 Lagrangian.
We have two kinds of N = 2 multiplets we have to deal with, vector multiplets
and hypermutiplets. What we noticed at the level of representations of the super-
symmetry algebra on states, see. lecture 3, holds also at the field level. In particular,
in N = 1 language a N = 2 vector superfield can be seen as the direct sum of a
vector superfield V and a chiral superfield Φ (having, the same internal quantum
numbers, e.g. they both transform in the adjoint representation of the gauge group).
Similarly, in terms of degrees of freedom a hypermultiplet can be constructed out of
two N = 1 chiral superfields, H1 and H2 . Schematically, we have

[N = 2 vector multiplet] : V = (λα , vµ , D) ⊕ Φ = (φ, ψα , F )

[N = 2 hypermultiplet] : H1 = (H1 , ψ1 α , F1 ) ⊕ H 2 = H 2 , ψ 2α̇ , F 2

(notice that H1 and H 2 transform in the same representations of internal symmetries

since they belong to the same supersymmetry representation, while H2 transforms
in the complex conjugate representations).
Let us start considering pure SYM. In N = 1 language this is a Lagrangian of
the type (5.79) on which, however, two specific requirements should be imposed.
First, as already stressed, the chiral superfield Φ should transform in the adjoint
representation of the gauge group, as the vector superfield V does. Second, notice
that we have now a larger R-symmetry group, whose compact component, SU (2)R ,
should be a symmetry of the Lagrangian. All bosonic degrees of freedom vµ , D, F

113
and φ are singlets under SU (2)R , but (λα , ψα ) transform as a doublet. This is
because (Q1α , Q2α ) transform under the fundamental representation of SU (2)R , and
the same should hold for λα and ψα (recall from (3.19) that they are obtained acting
with the two supersymmetry generators on the Clifford vacuum |j = 0i).
The Lagrangian reads
 Z  Z
N =2 1
LSY M = Im τ d θ Tr W Wα + d2 θd2 θ̄ Tr Φe2gV Φ =
2 α
32π

1 θY M 2
= Tr − Fµν F µν − iλσ µ Dµ λ − iψσ µ Dµ ψ + Dµ φDµ φ + g Fµν Feµν +
4 32π 2

1 2 √ √  
+ D + F F + i 2gφ{λ, ψ} − i 2g{ψ, λ}φ + gD φ, φ (6.1)
2
where

φ = φa Ta , ψα = ψαa Ta , F = F a Ta ; λα = λaα Ta , vµ = vµa Ta , D = Da Ta ,

with a = 1, 2, . . . , dim G. The reason why commutators and anti-commutators ap-

pear in the Lagrangian (6.1) is just because all fields transform in the adjoint rep-
a
resentation of G. Indeed, given that (Tadj )bc = −ifabc , we have that, e.g. the term
φλψ is actually
b b b b
φ λa (Tadj
a
)bc ψ c = −iφ λa fabc ψ c = iφ λa fbac ψ c = φ λa ψ c Tr Tb [Ta , Tc ] = Tr φ{λ, ψ} ,
(6.2)
and similarly for all other contributions in eq. (6.1).
As compared to a N = 1 Lagrangian describing matter coupled SYM theory,
the above Lagrangian is special in many respects.
First and foremost, despite N = 2 supersymmetry is not manifest using N = 1
superspace language, the above Lagrangian is invariant under two independent su-
persymmetries, as it should. This can be seen using the SU (2)R symmetry the La-
grangian enjoys, actually just its center Z2 , under which the two generators Q1α , Q2α
are exchanged. Eq. (6.1) is written in terms of two N = 1 superfields and, cor-
respondingly, it is obviously N = 1 invariant. Acting now with a Z2 R-symmetry
rotation which acts as ψα → λα and λα → −ψα , while leaving the bosonic fields
invariant, one sees that the same Lagrangian shows an invariance under an inde-
pendent N = 1 supersymmetry acting on two different superfields with entries
(φ, λα , F ) and (vµ , −ψα , D) . So we conclude that the Lagrangian is indeed N = 2
supersymmetric invariant.

114
 The existence of a manifest SU (2)R symmetry has also other related conse-
quences on the structure of the Lagrangian (6.1). The kinetic terms for λ and ψ
have the same normalization. Moreover, and more importantly, the Lagrangian
has no superpotential, W = 0. Indeed, a superpotential would give ψ interactions
and/or mass terms, that are absent for λ. This is clearly forbidden by the SU (2)R
symmetry. While there is no superpotential, there is a potential, which comes from
D-terms. Indeed, the auxiliary fields equations of motion are in this case
 a
Fa = 0 , Da = −g φ, φ (6.3)

(the auxiliary fields F a appear only in the non-dynamical kinetic term F a F a and
therefore are trivial). The potential hence reads
1 1  2
V (φ, φ) = Da Da = g 2 Tr φ, φ . (6.4)
2 2
The above expression shows that pure N = 2 SYM enjoys a huge moduli space of
supersymmetric vacua. Indeed, the potential vanishes whenever the fields φ belong
to the Cartan subalgebra of the gauge group G. At a generic point of the moduli
space the scalar field matrix can be diagonalized and the gauge group is broken to
U (1)r , with r the rank of G. The low energy effective dynamics is that of r massless
vector multiplets and dimG − r massive vector multiplets whose masses depend on
the scalar fields VEVs. The theory is said to be in a Coulomb phase, since charged
external sources will feel a Coulomb-like potential. The (classical) moduli space
is a r-dimensional complex manifolds, parametrized by r massless complex scalars.
Singularities arise whenever some VEVs become degenerate and the theory gets
partially unhiggsed (in particular, at the origin of the moduli space one recovers the
full gauge symmetry G). This is classical analysis and, as for N = 1 theories, in later
lectures we will see if and how this description gets modified once (non-perturbative)
quantum corrections are taken into account.
Let us now consider the addition of hypermultiplets. In this case the scalar
fields, H1 and H 2 form a SU (2)R doublet (again, recall how they were constructed
from the ground state of the corresponding N = 2 supersymmetry representation).
Hypermultiplets cannot interact between themselves since no cubic SU (2) invariant
is possible. Therefore, for renormalizable theories a superpotential is not allowed
and interactions turn out to be all gauge interactions.
Let us now suppose that matter transforms under some non-trivial representation

115
of the gauge group. We get for the corresponding N = 2 hypermultiplet Lagrangian
Z Z √
N =2 2 2 2gVR −2gVR


LM atter = d θd θ̄ H 1 e H1 + H 2 e H2 + d2 θ 2gH1 ΦH2 + h.c. , (6.5)

where the suffix R on the vector superfield V refers to the representation of the gauge
group G carried by the hypermultiplets. The F-term coupling the hypermultiplets
with the chiral multiplet Φ belonging to the N = 2 vector multiplet is there because
of N = 2 supersymmetry (it is, say, the supersymmetric partner of the kinetic
terms which couple the hypermultiplets to V ). So we see that eventually a cubic
interaction does arise, but it is a gauge interaction, in the sense that it vanishes once
the gauge coupling g is switched off.
Eliminating the auxiliary fields F1 and F2 , the scalar potential for the hypermul-
tiplets can be recast as a D-term contribution only and reads
1 1

V (H1 , H2 ) = D2 = g 2 |H 1 TRa H1 −H 2 TRa H2 |2 , Da = g Tr H 1 TRa H1 − H 2 TRa H2 .
2 2
(6.6)
Notice finally that a mass term can be present and has the form

mH1 H2 . (6.7)

However, a term of this sort can be there only for BPS hypermultiplets (which, as
discussed in lecture 3, are short enough to close the algebra within maximal spin
1/2 particle states).

6.1.1 Non-linear sigma model III

In the previous section we have constructed the most general renormalizable N = 2

supersymmetric Lagrangian. Like for N = 1 supersymmetry, one can relax renor-
malizability and get a N = 2 σ-model.
Let us start with pure SYM. Differently from the N = 1 case, this is a meaningful
thing to do, since scalar fields are present in a N = 2 vector multiplet and a σ-
model can exist. To write it down it is sufficient to take the N = 1 σ-model
Lagrangian (5.116), set the superpotential to zero, and take into account that the
chiral superfield Φ transforms in the adjoint representation of the gauge group.
On general grounds, one would expect he Kähler potential K to be related, in a
N = 2 consistent way, to the generalized complexified gauge coupling Fab , since
the scalars spanning the manifold M sit in the same multiplets where the vectors

116
sit (in particular, one would expect that an isometry transformation on M should
have effects on the vectors, too). Equivalently, one can notice from (5.116) that the
imaginary part of the generalized complexified gauge coupling multiplies the gaugino
kinetic term while the Kähler metric that of the matter fermion fields. These should
transform as a doublet under SU (2)R and then one would expect Fab and Kab to be
exchanged under a Z2 R-symmetry rotation
b
Im Fab × λa σ µ Dµ λ ←→ Kab × ψ a σ µ Dµ ψ b . (6.8)

What one actually finds is that Fab and K can be written in terms of one and the
same holomorphic function F(Φ), dubbed prepotential and read
∂ 2 F(Φ)
Fab (Φ) = (6.9)
∂Φa ∂Φb
i ∂F(Φ) i i
K(Φ, Φ) = − Φa a
+ h.c. = − Φa F a (Φ) + Fa (Φ)Φa . (6.10)
32π ∂Φ 32π 32π
This is the very non-trivial statement that the σ-model action is uniquely determined
by a single holomorphic function, the prepotential F(Φ). The end result reads
Z Z
N =2 1 2 αa b 1
a
Lef f = d θFab (Φ)W Wα + d2 θd2 θ̄ Φe2gV Fa (Φ) + h.c.
64πi 32πi
Z Z 
1 2 αa b 2 2


2gV a
= Im d θFab (Φ)W Wα + 2 d θd θ̄ Φe Fa (Φ) . (6.11)
32π
Using eqs. (6.9)-(6.10) one can compute the Kähler metric and see its relation with
the complexified gauge coupling which is
!
2 2 2
∂ K(φ, φ) i ∂ F(φ) ∂ F(φ) 1
Kab (φ, φ) = = − a b
− = ImFab (φ) . (6.12)
∂φa ∂φb 32π ∂φ ∂φ ∂φa ∂φb 16π

Therefore, we finally get for the potential (recall that W = 0)

1  a  b
V (φ, φ) = − (Im Fab (φ))−1 φ, Fc (φ)T c φ, Fd (φ)T d . (6.13)
2π
A Kähler manifold where the Kähler potential can be written in terms of a holomor-
phic function as in eq. (6.10) is called special Kähler manifold. From a geometric
point of view this corresponds to a Kähler manifold endowed with a symplectic
structure (a 2nv symplectic bundle, where nv is the number of vector multiplets).
The renormalizable Lagrangian (6.1) is recovered just taking F(Φ) = 12 τ TrΦ2 .
This is not the end of the story, though. To the σ-model action we have con-
structed one can add hypermultiplets. This will be a (very) special version of the

117
Lagrangian (5.43). We refrain to present its precise structure here and just make
two comments. Hypermultiplets contain two complex scalars. What one finds is
that the corresponding σ-model is defined on a quaternionic manifold, known as
HyperKähler manifold, which is, essentially, the quaternionic extension of a Kähler
manifold (in particular, there are three rather than just one complex structures).
To sum-up, in N = 2 supersymmetry, due to the existence of two sets of scalars,
those belonging to matter multiplets and those belonging to gauge multiplets, the
most general scalar manifold is (classically) of the form

M = MV × M H , (6.14)

where MV is a special Kähler manifold and MH a HyperKähler manifold. Notice

that, once renormalizability is relaxed, quartic superpotential couplings (and higher,
if SU (2)R singlets) are possible. We will have much more to say about N = 2 σ-
models in later lectures.

6.2 N = 4 supersymmetric actions

Let us now discuss the structure of the N = 4 Lagrangian. In this case there is only
one kind of multiplet, the vector multiplet. So, from a N = 4 perspective, we speak
of pure SYM theories. The decomposition of the N = 4 vector superfield in terms
of N = 1 representations is as follows


[N = 4 vector multiplet] : V = (λα , vµ , D) ⊕ ΦA = φA , ψαA , F A A = 1, 2, 3 .

The propagating degrees of freedom are a vector field, six real scalars (two for each
complex scalar φA ) and four gauginos. The Lagrangian is a special instance of the
Lagrangian (5.79) and very much constrained by N = 4 supersymmetry. First, the
chiral superfields ΦA should transform in the adjoint representation of the gauge
group, since internal symmetries commute with supersymmetry. Moreover, we have
now a large R-symmetry group, SU (4)R . The four Weyl fermions transform in the
fundamental of SU (4)R , while the six real scalars in the two times anti-symmetric
representation, which is nothing but the fundamental representation of SO(6) (the
fact that the scalar fields transform under the fundamental representation of SO(6),
which is real, makes the R-symmetry group of the N = 4 theory being at most SU (4)
and not U (4)). Finally, the auxiliary fields are singlets under the R-symmetry group.

118
 Using N = 1 superfield formalism the Lagrangian reads
 Z  Z X
N =4 1
LSY M = Im τ d θ Tr W Wα + d2 θd2 θ̄ Tr
2 α
ΦA e2gV ΦA
32π A
Z √
− d2 θ 2g Tr Φ1 [Φ2 , Φ3 ] + h.c. , (6.15)

where, similarly to the N = 2 Lagrangian (6.1), the commutator in the third term
appears because the three chiral superfields ΦA transform in the adjoint represen-
tation of the gauge group. Notice that in writing the expression (6.15) we made
a choice of a specific N = 1 supersymmetry generator out of the four QI ’s. This
breaks the full SU (4)R R-symmetry to SU (3) × U (1)R , meaning that the full non-
abelian R-symmetry is not manifest using N = 1 superspace formalism. The three
complex scalars belonging to the chiral superfields ΦA are related to the six real
scalars of the N = 4 vector multiplet as
1 1 1
φ1 = √ (X1 + iX2 ) , φ2 = √ (X3 + iX4 ) , φ3 = √ (X5 + iX6 ) . (6.16)
2 2 2
They transform in the 3 of SU (3) and have R-charge R = 2/3 under the U (1)R .
In N = 1 language the four gauginos λI get split as one gaugino and three Weyl
fermions transforming in the adjoint representation of the gauge group

λI = (λ, ψ1 , ψ2 , ψ3 ) . (6.17)

The gaugino λ is a singlet of SU (3) and has R-charge R = 1 under the U (1)R , while
the three ψA transform as a triplet of SU (3) and have U (1)R R-charge R = −1/3.
Note that from the point of view of the N = 1 supersymmetry which is manifest in
the Lagrangian (6.15), the SU (3) symmetry acts as a flavor (non-R) symmetry.
It is an easy but tedious exercise to perform the integration in superspace and
get an explicit expression in terms of ordinary fields. Then, one can solve for the
auxiliary fields and get an expression where only propagating degrees of freedom
are present and where, in terms of the N = 4 basis λI , Xi , see eqs. (6.16)-(6.17),
SU (4)R invariance is manifest. We refrain to perform the calculation here and just
report the end result which reads
h 1 1 i θ
g 2 TrFµν Feµν +
YM
L = Tr − Fµν F µν − iλI σ µ Dµ λI + Dµ X i Dµ Xi +
4 2 32π 2
1 abc IJ i 1 2 X 6
+ gf Σi Xa λIb λJc − g Tr [Xi , Xj ]2 , (6.18)
2 2 i,j=1

119
where sum on dummy indices is understood, the scalar potential is

1 2 X 6
V = g Tr [Xi , Xj ]2 , (6.19)
2 i,j=1

and the six 4 × 4 matrices Σi are the structure constants of the SU (4)R R-symmetry
group.
From eq. (6.19) we see that N = 4 SYM enjoys a large moduli space of vacua.
At a generic point, very much like pure N = 2 SYM, the gauge group is broken as
G → U (1)r , where r is the rank of G, and the dynamics is that of r copies of N = 4
U (1) gauge theory. At the origin, and more generally whenever some VEVs become
degenerate the theory gets partially unhiggsed and non-abelian gauge factors may
survive at low energy.
One might ask whether a N = 4 version of non-linear σ-model exists. After all,
we are plenty of scalar fields, actually 3n complex scalars, if n is the dimension of
G. The answer is that there is only one possible σ-model compatible with N =
4 supersymmetry (the stringent constraint comes from the SU (4)R R-symmetry),
which is nothing but the trivial one, M = R6n . So, the Lagrangian (6.15) is actually
the only possible N = 4 Lagrangian one can build. This also implies that, unlike
pure N = 2 SYM, the moduli space of vacua has a trivial topology. As we will
see, this is related to the very much constrained dynamics N = 4 SYM enjoys at
quantum level.

6.3 On non-renormalization theorems

One of the advantages, in fact the advantage of supersymmetry is that it makes
quantum corrections much better behaved with respect to ordinary field theories.
Many relevant results about UV properties of supersymmetric field theories were
obtained back in the 1980’s and can be summarized in terms of powerful non-
renormalization theorems. At that time, a very efficient approach was developed
to deal with supersymmetric quantum field theories, a version of Feynman rules,
known as supergraph techniques, which let one work directly with superfields in su-
perspace with no need to expand into component fields. Most non-renormalization
theorems were proved using such techniques whose description, however, is beyond
the scope of these lectures. Here I just want to mention what is possibly the main
result thus obtained: in a N = 1 supersymmetric quantum field theory containing

120
chiral and vector superfields, the most general term that can be generated by loop
diagrams has only one Grassman integral over all superspace
Z
d4 x1 . . . d4 xn d2 θd2 θ̄ G(x1 , . . . , xn )F1 (x1 , θ, θ̄) . . . Fn (xn , θ, θ̄) , (6.20)

where G(x1 , . . . , xn ) is a translationally invariant function and the Fi ’s are products

of superfields and their covariant derivatives. Such term is a D-term and does not
contribute to superpotential terms, which are F-terms, implying that the superpo-
tential is tree-level exact, i.e. it is not renormalized at any order in perturbation
theory! The only possible corrections may arise at the non-perturbative level (and in
some cases, namely when only chiral superfields are present, the latter also vanish,
as we will see later). On the contrary, because of eq. (6.20), the Kähler potential is
renormalized and generically receives corrections at any order in perturbation theory
(and non-perturbatively).
Let us try and see what non-renormalization theorems imply for theories with
different number of supersymmetries.
Let us first focus on a renormalizable N = 1 action describing a chiral superfield
Φ (the generalization to many chiral superfields is straightforward and does not
present any relevant difference). There are three supersymmetric contributions to
the action. One, the kinetic term, is a D-term, and undergoes renormalizations. Two
are F-terms (the mass term and the cubic term) and are hence exact, perturbatively.
Concretely
Z Z
4 2 2
d xd θd θ̄ ΦΦ → ZΦ d4 xd2 θd2 θ̄ ΦΦ (6.21)
Z Z
4 2 2
m d xd θ Φ + h.c. → m d4 xd2 θ Φ2 + h.c. (6.22)
Z Z
4 2 3
λ d xd θ Φ + h.c → λ d4 xd2 θ Φ3 + h.c. . (6.23)

This means that m and λ do get renormalized but only logarithmically at one loop,
instead of quadratically and linearly, respectively. Indeed, from the above equations
one concludes that

Zm ZΦ = 1 which implies m → ZΦ−1 m

3/2 −3/2
Zλ ZΦ = 1 which implies λ → ZΦ λ.

Something similar happens for the renormalization of the factor e2gV . However,

121
things are more subtle, here. Notice that
Z
d2 θd2 θ̄ Φe2gV Φ (6.24)

is a D-term and therefore it do renormalize. However, what one can see using
for example the background field method is that gV does not renormalize. An
independent renormalization for g and V leading to a kinetic term of the form
1/2
Φ e2Zg gZV V
Φ (6.25)

would correspond to counterterms of the form

Φ V egV Φ (6.26)

which are not gauge invariant and cannot be generated by loop diagrams. This
implies that the integral (6.24) should renormalize as the kinetic term (6.21) (not
because of supersymmetry, but just due to gauge invariance!), meaning that g and V
compensate each other upon renormalization. In other words, gV is not renormal-
ized. Another way to see this is the following. Consider pure SYM. In this theory
the only possible counterterm would correspond to something proportional to the
action itself Z
d2 θ Tr W α Wα + h.c. , (6.27)

which would then correspond to a wave-function renormalization of the full La-

grangian (this is certainly there since the integral above is not a F-term, but rather
a D-term, as already noticed). This means that one should multiply by the same
function both the kinetic terms dV dV as well as the interaction terms gV V dV and
g 2 V 4 to keep gauge invariance (recall we are considering a non-abelian gauge group).
In order for this to be the case one needs that if
1/2
V → ZV V , (6.28)

then
−1/2
g → ZV g, (6.29)
−1/2
which implies that gV is not renormalized, as anticipated, namely that Zg = ZV .
The conclusion is that in renormalizable theories with N = 1 supersymmetry there
are only two independent renormalization, ZΦ and ZV , which are just logarithmically
divergent at one-loop, and correspond to wave-function renormalization of chiral and
vector superfields, respectively.

122
 Notice, in passing, that the fact that ZV 6= 1 means that the integral (6.27)
is renormalized. This does not contradict non-renormalization theorems, since, as
already observed, (6.27) is not a F-term, really, but actually a D-term. That ZV 6= 1
also implies that Zg 6= 1, meaning that in N = 1 SYM theories the gauge coupling
runs, and can get corrections at all loops, in general.
What about higher supersymmetry? All what we said, so far, still applies, since
any extended supersymmetry model is also N = 1. However, extended supersym-
metry imposes further constraints. In what follows, we stick to the notation we
have used in sections 6.1 and 6.2 when discussing N = 2 and N = 4 Lagrangians,
respectively.
Let us start from N = 2 supersymmetry. For one thing, since V and Φ belong
now to the same multiplet, we have that

ZV = ZΦ . (6.30)

As for the hypermultiplets, from the cubic interaction gH1 ΦH2 which appears in
the superpotential (and which is then tree level exact, being a F-term) we get the
following condition
Zg2 ZΦ ZH1 ZH2 = 1 . (6.31)
The first two contributions cancel since ZΦ = ZV and since gV is not renormalized,
Zg2 = ZV−1 . So we get in the end

ZH1 ZH2 = 1 . (6.32)

Hence the wave-functions of the two chiral superfields making up an hypermultiplet

are not independent. All in all, we have then only two independent renormalizations
in N = 2 supersymmetry, ZV and, say, ZH1 . In fact, for massless representations
there is the SU (2)R symmetry rotating the scalar components of H1 and H2 into
each other. Hence, they should have the same renormalization, which means, using
eq. (6.32), that ZH1 = ZH2 = 1. The same holds for massive (BPS) representations.
In this case the existence of a non-trivial central charge does break the R-symmetry
group to U Sp(2); however, the algebra of such group is the same as that of SU (2)
and one can again conclude that ZH1 = ZH2 = 1.
It turns out that because of the relation between ZΦ and Zg , not only N = 2
SYM has a unique renormalization but it is one-loop exact in perturbation theory.
In other words, the gauge coupling β-function gets only one-loop contributions, per-
turbatively. We will derive this important result in a later lecture, when discussing

123
the dynamics of supersymmetric gauge theory. There, we will use a very powerful
approach which is based on a crucial property of supersymmetry, known as holo-
morphy. For the time being let us just stress that this one-loop exactness of N = 2
SYM gauge coupling does not hold for N = 1 SYM, whose physical gauge coupling
receives corrections at all orders in perturbation theory.
Let us finally consider N = 4 supersymmetry. Here we have a single superfield,
the vector superfield which, in N = 1 language, can be seen as one vector superfield
V and three chiral superfields ΦA (all transforming in the adjoint representation of
the gauge group). The SU (3) symmetry rotating the three chiral superfields implies
that the latter should have all and the same wave-function renormalization

ZΦ1 = ZΦ2 = ZΦ3 = Z . (6.33)

The N = 4 Lagrangian can be thought as a N = 2 SYM Lagraingian coupled

to a hypermultiplet transforming in the adjoint representation of the gauge group,
with one of the three ΦA playing the role of Φ and belonging to the N = 2 vector
multiplet. Then, using eq. (6.30), one concludes that Z should equal ZV , the wave-
function of the vector superfield. Plugging this into eq. (6.31), which for the N = 4
Lagrangian (6.15) is
Zg2 ZΦ1 ZΦ2 ZΦ3 = Zg2 Z 3 = 1 , (6.34)
−1/2
and recalling that Zg = ZV it follows that

ZV = ZΦ1 = ZΦ1 = ZΦ1 = 1 , (6.35)

meaning that N = 4 SYM is perturbatively finite; in other words, the theory does
not exhibit ultraviolet divergences in the correlation functions of canonical fields!
Though we are not going to prove it here, it turns out that in fact N = 4 is finite
also once non-perturbative corrections are taken into account. More precisely, the
latter give finite contributions only, and therefore the theory is believed to be UV
finite.
There is yet another important property of N = 4 SYM we would like to mention:
the theory is superconformal invariant, and it is so at the full quantum level. Let
us see how this goes. A theory whose Lagrangian contains only dimension four
operators, like the N = 4 Lagrangian (and many others, in fact) is classically
scale invariant. For any relativistic field theory this implies a larger symmetry
algebra, the conformal Poincaré algebra which, besides Poincaré generators, includes
also dilations and special conformal transformations, the corresponding group being

124
 SO(2, 4) ' SU (2, 2). The generators associated to dilations and special conformal
transformations, D and K µ , respectively, act as follows

D : xµ −→ λ xµ
x µ + aµ x 2
K µ : xµ −→ ,
1 + 2xν aν + a2 x2
and have the following commutation relations between themselves and with the
generators of the Poincaré algebra

[Pµ , D] = iPµ , [D, Mµν ] = 0 , [Kµ , D] = −iKµ , [Kµ , Kν ] = 0

[Pµ , Kν ] = 2i (Mµν − ηµν D) , [Kµ , Mρσ ] = i (ηµρ Kσ − ηµσ Kρ ) . (6.36)

Supersymmetry enlarges further the symmetry group. A conformal invariant su-

persymmetric theory enjoys an even larger algebra, the superconformal algebra,
which includes, besides dilations and special conformal transformations, also con-
I
formal supersymmetry transformations SαI , S α̇ (which appear in the commutator of
the supersymmetry charges QIα with the generators of special conformal transfor-
mations K µ ), and the generators associated to R-symmetry transformations, T JI
(where I, J = 1, . . . , N ), which are now part of the algebra and do not act just as
external automorphisms (they appear in the anti-commutator of the supersymmetry
charges with the SαI ’s). The associated supergroup is SU (2, 2|N ). The non-vanishing
(anti)commutators involving the new generators are
  β̇I   i   i
Kµ , QIα = 2iσµαβ̇ S , {SαI , S β̇ J } = 2σαµβ̇ Kµ δ IJ , D, QIα = − QIα , D, SαI = SαI
2 2

1 µν
  β̇I
{QIα , SβJ } = αβ δ IJ D + T JI + δ IJ σαβ Mµν , Pµ , SαI = σµαβ̇ Q . (6.37)
2
The N = 4 SYM action is invariant under this larger symmetry algebra, SU (2, 2|4)
in this case, but it is certainly not the only theory having this property, at the
classical level. Classical superconformal invariance is shared by any supersymmet-
ric Lagrangian made solely by dimension four operators (in other words, with di-
mensionless, hence classically marginal couplings), well-known examples being the
massless WZ model, and in fact any SYM theory, like N = 1 SQCD discussed in
the previous lecture.
What makes N = 4 SYM special is that, as we have observed above, the La-
grangian does not renormalize (recall that essential to this proof was the use of
the SU (3) subgroup of the R-symmetry group rotating the three scalar superfields

125
ΦA ). In particular, as we have seen before, Zg = 1. In other words, the N = 4
β-function vanishes identically: the theory remains scale invariant at the quantum
level, and the superconformal symmetry SU (2, 2|4) is then an exact symmetry of
the theory. An equivalent conclusion can be reached by observing that N = 4 SYM
is a (very special) N = 2 theory. The N = 2 gauge coupling is one-loop exact
and since in N = 4 SYM this is the only coupling appearing in the Lagrangian, it
is enough to compute the one-loop β-function for g. One can easily see that such
one-loop coefficient vanishes, concluding that the theory is superconformal also at
the quantum level. In fact, the equivalence of these proofs lies in the fact that the
gauge coupling β-function and the R-symmetry are in the same supermultiplet, the
(N = 4) supercurrent multiplet.
This non-renormalization property is not shared by other theories, in general:
typically, the superconformal algebra is broken by quantum corrections and cou-
plings run. For instance, in the massless WZ model, the coupling, which is classically
marginal, becomes irrelevant quantum mechanically (i.e., it flows to zero and the
theory becomes free in the IR). On the contrary, UV-free supersymmetric gauge the-
ories, like pure N = 1, 2 SYM, enjoy dimensional transmutation and a dynamically
scale is generated at the quantum level.
What we said above about the finiteness of N = 4 does not mean that any
operator has protected dimension. The scaling dimension of canonical fields (gauge
fields, gauginos and adjoint scalars) is unaffected by quantum corrections, but this
does not happen, in general, to composite gauge invariant operators. Yet, in a su-
perconformal theory there are special operators whose dimension is protected. To
see how this comes, let us start considering the conformal algebra (6.36). In unitary
theories there is a lower bound for the scaling dimension ∆ of a field (e.g. ∆ ≥ 1
for a scalar field in four dimensions). Since Kµ lowers the scaling dimension of a
field, any representation of the conformal algebra should admit an operator with
minimal dimension ∆ which is annihilated by Kµ (at xµ = 0). Such states are called
conformal primary operators. Since the conformal algebra is a subalgebra of the su-
perconformal algebra, representations of the latter decompose into representations
of the former. By definition, a superconformal primary operator is an operator
I
which is annihilated (at xµ = 0) both by Kµ and SαI , S α̇ . From the commutator
 
Kµ , QIα in (6.37) it also follows that any operator which is obtained from a super-
conformal primary by the action of QIα , and hence sits in the same supermultiplet,
is a primary operator of the conformal algebra. Superconformal primary operators

126
which are annihilated by some of the supercharges are called chiral primaries and,
most importantly, their dimension is fixed by their R-symmetry representation, and
as such are protected against quantum corrections. This can be proven from the
next-to-last commutation relation in eqs. (6.37), which lets express the scaling di-
mension ∆ of a chiral primary operator (for which the left hand side is zero) in
terms of Lorentz and R-symmetry representations which are quantum numbers and
as such does not renormalize. By supersymmetry, this implies that in a supercon-
formal theory operators belonging to supersymmetry representations which include
a chiral primary operator do not renormalize. For instance, in N = 4 SYM, a class
of superconformal primaries are all operators made of symmetric traceless products
of the scalar fields Xi ’s, e.g. Tr (X {i X j} ) = Tr (X i X j ) − 61 δ ij Tr (X k X k ).
As a final comment, let us notice that in N = 4 SYM superconformal invariance
is/is not realized depending on the point of the moduli space one is sitting. The
phase where all scalar field VEVs hXi i vanish is called superconformal phase since at
the origin of the moduli space the gauge group remains unbroken and superconformal
invariance is preserved. In other words, physical states are not only gauge invariant,
but carry unitary representations of SU (2, 2|4). On the contrary, at any other
point of the moduli space, where gauge symmetry is broken, also superconformal
symmetry is broken since scalar VEVs hXi i set a dimension full scale in the theory.

6.4 Exercises
1. Derive the potential (6.6) starting from the matter Lagrangian (6.5).

2. Starting from the expression (6.15), compute the off-shell N = 4 Lagrangian

and derive, upon integrating out auxiliary fields, the on-shell one, eq. (6.18).
For scalars and fermion fields use bases Xi and λI , which carry faithful repre-
sentations of SU (4)R , defined in eqs. (6.16) and (6.17), respectively.

References

[1] A. Bilal, Introduction to supersymmetry, Section 8, arXiv:hep-th/0101055.

[2] M. F. Sohnius, Introducing Supersymmetry, Sections 12 and 13, Phys. Rept.

128 (1985) 39.

127
[3] P. Frè, Lectures on special Kahler geometry and electric - magnetic duality
rotations, Nucl. Phys. Proc. Suppl. 45BC (1996) 59 [hep-th/9512043].

[4] E. D’Hoker and D. Z. Freedman, Supersymmetric gauge theories and the

AdS/CFT correspondence, Section 3, arXiv:hep-th/0201253.

[5] M. J. Strassler, An unorthodox introduction to supersymmetric gauge theory,

Sections 2.3 and 3.1, TASI 2001, arXiv:hep-th/0309149.

[6] R. Argurio, Introduction on Supersymmetry, Chapter 6. You can find this at:
https://ptm.ulb.be/rargurio/susycourse.pdf

128
7 Supersymmetry breaking

If supersymmetry is at all realized in Nature, it must be broken at low enough

energy: we do not see any mass degeneracy in the elementary particle spectrum, at
least at energies of order 102 GeV or lower. The idea is then that supersymmetry
is broken at some scale Ms , such that at energies E > Ms the theory behaves as a
supersymmetric theory, while at energies E < Ms it does not. On general ground,
there are two ways supersymmetry can be broken, either spontaneously or explicitly.

• Spontaneous supersymmetry breaking: the theory is supersymmetric but has

a scalar potential admitting (stable, or metastable but sufficiently long-lived)
supersymmetry breaking vacua. In such vacua one or more scalar fields acquire
a VEV of order Ms , which then sets the scale of supersymmetry breaking.

• Explicit supersymmetry breaking: the Lagrangian contains terms which do not

preserve supersymmetry by themselves. In order for them not to ruin the nice
and welcome UV properties of supersymmetric theories, these terms should
have positive mass dimension, in other words they should be irrelevant in the
far UV. In this case we speak of soft supersymmetry breaking. In such scenario,
the scale Ms enters explicitly in the Lagrangian.

As we will show later, soft supersymmetry breaking models can (and typically do)
actually arise as low energy effective descriptions of models where supersymmetry is
broken spontaneously. Therefore, we will start focusing on spontaneous supersym-
metry breaking. Only after we will discuss supersymmetry breaking induced by soft
terms.

7.1 Vacua in supersymmetric theories

We have already seen that supersymmetric vacua are in one-to-one correspondence
with the zero’s of the scalar potential. In other words, the vacuum energy is zero
if and only if the vacuum preserves supersymmetry. Hence, non-supersymmetric
vacua correspond to minima of the potential which are not zero’s. In this case
supersymmetry is broken in the perturbative theory based on these positive energy
vacua.
Notice how different is spontaneous supersymmetry breaking with respect to
spontaneous breaking of ordinary internal symmetries (being them global or local).

129
There, what matters is the location of the minima of the potential in field space,
while here it is the absolute value of the potential at such minima. This implies that
in, e.g. a supersymmetric gauge theory there can be minima which preserve both
gauge symmetry and supersymmetry, others which break both, and others which
preserve gauge symmetry and break supersymmetry, or viceversa. A schematic
picture of these different situations is reported in figure 7.1.

V V
<latexit sha1_base64="A1XDWPnWvrr38Qg2Hbjj1OG2Oh0=">AAAB6HicdVDLSgNBEOz1GeMr6tHLYBA8LbPBoN6CXjwmYB6QLGF2MpuMmZ1dZmaFsOQLvHhQxKuf5M2/cTYPUNGChqKqm+6uIBFcG4w/nZXVtfWNzcJWcXtnd2+/dHDY0nGqKGvSWMSqExDNBJesabgRrJMoRqJAsHYwvsn99gNTmsfyzkwS5kdkKHnIKTFWarT6pTJ2vascCLsVbFFdkgryXDxDGRao90sfvUFM04hJQwXRuuvhxPgZUYZTwabFXqpZQuiYDFnXUkkipv1sdugUnVplgMJY2ZIGzdTvExmJtJ5Ege2MiBnp314u/uV1UxNe+hmXSWqYpPNFYSqQiVH+NRpwxagRE0sIVdzeiuiIKEKNzaZoQ1h+iv4nrYrrVV3cOC/XrhdxFOAYTuAMPLiAGtxCHZpAgcEjPMOLc+88Oa/O27x1xVnMHMEPOO9fB++NFw==</latexit> <latexit sha1_base64="A1XDWPnWvrr38Qg2Hbjj1OG2Oh0=">AAAB6HicdVDLSgNBEOz1GeMr6tHLYBA8LbPBoN6CXjwmYB6QLGF2MpuMmZ1dZmaFsOQLvHhQxKuf5M2/cTYPUNGChqKqm+6uIBFcG4w/nZXVtfWNzcJWcXtnd2+/dHDY0nGqKGvSWMSqExDNBJesabgRrJMoRqJAsHYwvsn99gNTmsfyzkwS5kdkKHnIKTFWarT6pTJ2vascCLsVbFFdkgryXDxDGRao90sfvUFM04hJQwXRuuvhxPgZUYZTwabFXqpZQuiYDFnXUkkipv1sdugUnVplgMJY2ZIGzdTvExmJtJ5Ege2MiBnp314u/uV1UxNe+hmXSWqYpPNFYSqQiVH+NRpwxagRE0sIVdzeiuiIKEKNzaZoQ1h+iv4nrYrrVV3cOC/XrhdxFOAYTuAMPLiAGtxCHZpAgcEjPMOLc+88Oa/O27x1xVnMHMEPOO9fB++NFw==</latexit>

<latexit sha1_base64="iK4LIai9Ml/FphQ8ddTj4CWppl4=">AAAB63icdVDLSsNAFL3xWeur6tLNYBFchUmxqLuiG5cV7APaUCbTSTN0JgkzE6GE/oIbF4q49Yfc+TdO+gAVPXDhcM693HtPkAquDcafzsrq2vrGZmmrvL2zu7dfOThs6yRTlLVoIhLVDYhmgsesZbgRrJsqRmQgWCcY3xR+54EpzZP43kxS5ksyinnIKTGF1E8jPqhUsetdFUDYrWGL+pLUkOfiGaqwQHNQ+egPE5pJFhsqiNY9D6fGz4kynAo2LfczzVJCx2TEepbGRDLt57Nbp+jUKkMUJspWbNBM/T6RE6n1RAa2UxIT6d9eIf7l9TITXvo5j9PMsJjOF4WZQCZBxeNoyBWjRkwsIVRxeyuiEVGEGhtP2Yaw/BT9T9o116u7+O682rhexFGCYziBM/DgAhpwC01oAYUIHuEZXhzpPDmvztu8dcVZzBzBDzjvX2i3jnw=</latexit> <latexit sha1_base64="iK4LIai9Ml/FphQ8ddTj4CWppl4=">AAAB63icdVDLSsNAFL3xWeur6tLNYBFchUmxqLuiG5cV7APaUCbTSTN0JgkzE6GE/oIbF4q49Yfc+TdO+gAVPXDhcM693HtPkAquDcafzsrq2vrGZmmrvL2zu7dfOThs6yRTlLVoIhLVDYhmgsesZbgRrJsqRmQgWCcY3xR+54EpzZP43kxS5ksyinnIKTGF1E8jPqhUsetdFUDYrWGL+pLUkOfiGaqwQHNQ+egPE5pJFhsqiNY9D6fGz4kynAo2LfczzVJCx2TEepbGRDLt57Nbp+jUKkMUJspWbNBM/T6RE6n1RAa2UxIT6d9eIf7l9TITXvo5j9PMsJjOF4WZQCZBxeNoyBWjRkwsIVRxeyuiEVGEGhtP2Yaw/BT9T9o116u7+O682rhexFGCYziBM/DgAhpwC01oAYUIHuEZXhzpPDmvztu8dcVZzBzBDzjvX2i3jnw=</latexit>

GAUGE SUSY GAUGE SUSY

V V
<latexit sha1_base64="A1XDWPnWvrr38Qg2Hbjj1OG2Oh0=">AAAB6HicdVDLSgNBEOz1GeMr6tHLYBA8LbPBoN6CXjwmYB6QLGF2MpuMmZ1dZmaFsOQLvHhQxKuf5M2/cTYPUNGChqKqm+6uIBFcG4w/nZXVtfWNzcJWcXtnd2+/dHDY0nGqKGvSWMSqExDNBJesabgRrJMoRqJAsHYwvsn99gNTmsfyzkwS5kdkKHnIKTFWarT6pTJ2vascCLsVbFFdkgryXDxDGRao90sfvUFM04hJQwXRuuvhxPgZUYZTwabFXqpZQuiYDFnXUkkipv1sdugUnVplgMJY2ZIGzdTvExmJtJ5Ege2MiBnp314u/uV1UxNe+hmXSWqYpPNFYSqQiVH+NRpwxagRE0sIVdzeiuiIKEKNzaZoQ1h+iv4nrYrrVV3cOC/XrhdxFOAYTuAMPLiAGtxCHZpAgcEjPMOLc+88Oa/O27x1xVnMHMEPOO9fB++NFw==</latexit> <latexit sha1_base64="A1XDWPnWvrr38Qg2Hbjj1OG2Oh0=">AAAB6HicdVDLSgNBEOz1GeMr6tHLYBA8LbPBoN6CXjwmYB6QLGF2MpuMmZ1dZmaFsOQLvHhQxKuf5M2/cTYPUNGChqKqm+6uIBFcG4w/nZXVtfWNzcJWcXtnd2+/dHDY0nGqKGvSWMSqExDNBJesabgRrJMoRqJAsHYwvsn99gNTmsfyzkwS5kdkKHnIKTFWarT6pTJ2vascCLsVbFFdkgryXDxDGRao90sfvUFM04hJQwXRuuvhxPgZUYZTwabFXqpZQuiYDFnXUkkipv1sdugUnVplgMJY2ZIGzdTvExmJtJ5Ege2MiBnp314u/uV1UxNe+hmXSWqYpPNFYSqQiVH+NRpwxagRE0sIVdzeiuiIKEKNzaZoQ1h+iv4nrYrrVV3cOC/XrhdxFOAYTuAMPLiAGtxCHZpAgcEjPMOLc+88Oa/O27x1xVnMHMEPOO9fB++NFw==</latexit>

<latexit sha1_base64="iK4LIai9Ml/FphQ8ddTj4CWppl4=">AAAB63icdVDLSsNAFL3xWeur6tLNYBFchUmxqLuiG5cV7APaUCbTSTN0JgkzE6GE/oIbF4q49Yfc+TdO+gAVPXDhcM693HtPkAquDcafzsrq2vrGZmmrvL2zu7dfOThs6yRTlLVoIhLVDYhmgsesZbgRrJsqRmQgWCcY3xR+54EpzZP43kxS5ksyinnIKTGF1E8jPqhUsetdFUDYrWGL+pLUkOfiGaqwQHNQ+egPE5pJFhsqiNY9D6fGz4kynAo2LfczzVJCx2TEepbGRDLt57Nbp+jUKkMUJspWbNBM/T6RE6n1RAa2UxIT6d9eIf7l9TITXvo5j9PMsJjOF4WZQCZBxeNoyBWjRkwsIVRxeyuiEVGEGhtP2Yaw/BT9T9o116u7+O682rhexFGCYziBM/DgAhpwC01oAYUIHuEZXhzpPDmvztu8dcVZzBzBDzjvX2i3jnw=</latexit>
<latexit sha1_base64="iK4LIai9Ml/FphQ8ddTj4CWppl4=">AAAB63icdVDLSsNAFL3xWeur6tLNYBFchUmxqLuiG5cV7APaUCbTSTN0JgkzE6GE/oIbF4q49Yfc+TdO+gAVPXDhcM693HtPkAquDcafzsrq2vrGZmmrvL2zu7dfOThs6yRTlLVoIhLVDYhmgsesZbgRrJsqRmQgWCcY3xR+54EpzZP43kxS5ksyinnIKTGF1E8jPqhUsetdFUDYrWGL+pLUkOfiGaqwQHNQ+egPE5pJFhsqiNY9D6fGz4kynAo2LfczzVJCx2TEepbGRDLt57Nbp+jUKkMUJspWbNBM/T6RE6n1RAa2UxIT6d9eIf7l9TITXvo5j9PMsJjOF4WZQCZBxeNoyBWjRkwsIVRxeyuiEVGEGhtP2Yaw/BT9T9o116u7+O682rhexFGCYziBM/DgAhpwC01oAYUIHuEZXhzpPDmvztu8dcVZzBzBDzjvX2i3jnw=</latexit>

GAUGE SUSY GAUGE SUSY

Figure 7.1: A schematic picture of possible patterns of spontaneous gauge symmetry

and supersymmetry breakings. The potential on the upper left does not admit any
symmetry breaking vacuum. The one on the lower right, instead, admits two vacua
breaking both gauge symmetry and supersymmetry. The other two represent mixed
situations where either supersymmetry or gauge symmetry are broken.

While non-supersymmetric vacua can be either global or local minima of the po-
tential (corresponding to stable or metastable vacua, respectively), supersymmetric
vacua, if present, are obviously global minima of the potential, since in a supersym-
metric theory the scalar potential is a semi-positive definite quantity.

130
 Recall the expression (5.82), that is
1
V (φ, φ) = F F + D2 , (7.1)
2
where
∂W a a i j a


Fi = , D = −g φi (T ) j φ + ξ (7.2)
∂φi
(we focus here on models with canonical Kähler potential; later we will also discuss
situations where the Kähler potential is not canonical).
Supersymmetric vacua are described by all possible set of scalar field VEVs
satisfying the D and F-term equations
i
F (φ) = 0 , Da (φ, φ) = 0 . (7.3)

If there exist more than one solution, it means there are more supersymmetric vacua,
generically a moduli space of vacua, if these are not isolated. If there does not exist
a set of scalar field VEVs for which eqs. (7.3) are satisfied, then supersymmetry is
broken and the minima of the potential are all necessarily positive, Vmin > 0.
Notice, on the contrary, that on any vacuum, supersymmetric or not, global or
local, the following equations always hold
∂V (φ, φ) ∂V (φ, φ)
=0 , =0, (7.4)
∂φi ∂φi
which simply say that vacua sit at extrema of the scalar potential.
An equivalent statement about supersymmetric vacua is that on supersymmetric
vacua the supersymmetric variations of fermion fields vanish. This can be seen as
follows. Due to Lorentz invariance, on a vacuum any field’s VEV or its derivative
should vanish, but scalar fields. Recalling how the different field components of
a chiral or vector superfield transform under supersymmetry transformations, it
follows that on a vacuum state we have

δhφi i = 0 , δhF i i = 0 , δhψαi i ∼ α hF i i

a
δhFµν i = 0 , δhDa i = 0 , δhλaα i ∼ α hDa i . (7.5)

Therefore, in a generic vacuum the supersymmetric variations of the fermions is not

zero: it is actually proportional to the vacuum expectation values of the auxiliary
fields. A supersymmetric vacuum state is by definition supersymmetric invariant
(!). Hence, from the above equations it follows that on a supersymmetric vacuum
also the supersymmetric variations of the fermions should be zero, the latter being
equivalent to the D and F-term equations (7.3), as anticipated.

131
7.2 Goldstone theorem and the goldstino
When a global symmetry is spontaneously broken, Goldstone theorem says that
there is a massless mode in the spectrum, the Goldstone field, whose quantum num-
bers should be related to the broken symmetry. We should expect this theorem to
work also for spontaneously broken supersymmetry. In fact, given that supersym-
metry is a fermionic symmetry, the Goldstone field should be in this case a Majorana
spin 1/2 fermion, the so-called goldstino.
Consider the most general supersymmetric Lagrangian with gauge and matter
fields, eq. (5.81), and suppose it admits some vacuum where supersymmetry is bro-
ken. In this vacuum eqs. (7.4) hold, while (some of) eqs. (7.3) do not. Recalling
eqs. (5.80) and (5.82) we have in this vacuum that

∂V (φ, φ) ∂ 2W
= F j
(φ) − gDa φj (T a )ji = 0 . (7.6)
∂φi ∂φi ∂φj
On the other hand, since the superpotential is gauge invariant, we have that
∂W a i
δaW = δ φ = F i (T a )ij φj = 0 . (7.7)
∂φi
Combining the former equation with the complex conjugate of the latter evaluated
in the vacuum, we easily get a matrix equation
! 2W
!
hF j i h ∂φ∂ i ∂φ j i −ghφl i(T a l
) i
M a
= 0 where M = b l
. (7.8)
hD i −ghφl i(T ) j 0

The above equation implies that the matrix M has an eigenvector with zero eigen-
value. Now, this matrix is nothing but the fermion mass matrix of the Lagrangian
(5.81)! This can be seen looking at the non-derivative fermion bilinears of (5.81),
which on the vacuum get contributions also from the cubic coupling between scalar
fields, their superpartners and gauginos, and which can be written as
!
1  i √ b ψj
··· − ψ , 2iλ M √ a + h.c. + . . . (7.9)
2 2iλ

Hence, on the supersymmetry breaking vacuum the spectrum necessarily admits a

massless fermion, the goldstino. It is easy to see that in terms of spin 1/2 particles
belonging to the different multiplets, the goldstino ψαG corresponds to the following
linear combination
ψαG ∼ hF i iψαi + hDa iλaα . (7.10)

132
 The proof we have provided of the goldstino theorem has been based on the
Lagrangian (5.81). In fact, one can provide a similar proof using properties of the
supercurrent and Ward identities, which does not rely on the existence of an explicit
classical Lagrangian. The supersymmetry Ward identity reads

h∂ µ Sµα (x)S ν β̇ (0)i = −δ 4 (x)hδα Sν β̇ i = −2 σαµβ̇ hTµν i δ 4 (x) , (7.11)

where the last equality follows from the current algebra, see eq. (4.71) (Schwinger
terms cannot have a non-vanishing VEV in a Lorentz invariant vacuum, while this
is possible for the energy-momentum tensor, Tµν ∼ ηµν ). Integrating eq. (7.11), one
gets for the two-point function of the supercurrent that
xρ
hSµα (x)S ν β̇ (0)i = · · · + (σµ σ̄ ρ σν )αβ̇ hT i , (7.12)
x4
where hT i = η µν hTµν i and the . . . are terms which are not relevant to the present
discussion. Upon Fourier transforming we finally get
hT i
hSµα (k)S ν β̇ (−k)i = · · · + (σµ σ̄ ρ σν )αβ̇ kρ , (7.13)
k2
which shows the presence of a massless pole (the goldstino), proportional to the
vacuum energy density, in the supercurrent two-point function. The above equation
shows that the goldstino is the lowest energy excitation of the supercurrent, and it is
so if and only if the vacuum energy is non-vanishing. This shows, as anticipated, that
the goldstino theorem holds universally, i.e. also for vacua where supersymmetry
is broken in a strongly coupled phase of the theory, where classical arguments may
not apply.

7.3 F-term breaking

From our discussion it is clear that given a generic Lagrangian, there are two a
priori independent ways we can break supersymmetry: either giving a non vanishing
expectation value to (some) F-terms or to (some) D-terms. We will consider both
options in turn.
In this section we will start considering F-term breaking and therefore we assume,
for the time being to deal with a theory with chiral superfields, only.
The most general renormalizable Lagrangian of this sort reads
Z Z Z
L = d θd θ̄ Φi Φ + d θ W (Φ ) + d2 θ̄ W (Φi ) ,
2 2 i 2 i
(7.14)

133
where
1 1
W (Φi ) = ai Φi + mij Φi Φj + gijk Φi Φj Φk . (7.15)
2 3
The equations of motions for the auxiliary fields read
∂W
F i (φ) = i
= ai + mij φj + gijk φj φk , (7.16)
∂φ
and the potential is
X
V (φ, φ) = |ai + mij φj + gijk φj φk |2 . (7.17)
i

Supersymmetry is broken if and only if there does not exist a set of scalar field VEVs
such that all F-terms vanish, hF i i = 0. This implies that in order for supersymmetry
to be broken, it is necessary some ai to be different from zero. If not, the trivial
solution hφi i = 0 solves all F-equations. So, any model of F-term supersymmetry
breaking needs a superpotential admitting linear terms
Notice that this conclusion applies also for a superpotential with higher non-
renormalizable couplings. In fact, it does also in presence of a non-canonical Kähler
potential! This can be seen recalling the expression of the scalar potential when a
non-canonical Kähler metric is present

i ∂W ∂W
V (φ, φ) = K −1 j ∂φi
. (7.18)
∂φj

From this expression it is clear that unless it were singular (something signalling,
as already discussed, an inconsistency of the effective theory analysis), a non-trivial
Kähler metric could not influence the existence/non existence of supersymmetric
vacua, which is still dictated by the possibility/impossibility to make the first deriva-
tives of the superpotential vanish. What gets modified by a non-trivial Kähler po-
tential, instead, is the value of the vacuum energy (for non-supersymmetric vacua,
only!) and the particle spectrum around a given vacuum (for both supersymmetric
and non-supersymmetric vacua).
In what follows we will consider several examples of F-term breaking.

Example 1 : The Polonyi model.

Let us consider the theory of a single chiral superfield with canonical Kähler
potential and a linear superpotential

K = XX , W = λX . (7.19)

134
This is the most minimal set-up one can imagine for a F-term supersymmetry break-
ing model. The potential reads
∂W 2
= |λ|2 , (7.20) V =
∂X
Supersymmetry is clearly broken for any |X|, and the latter is in fact a flat direction.
The supersymmetry breaking scale is set by the modulus of λ itself, |λ| = Ms2 . In
Figure 7.2 we report the (trivial) shape of the scalar potential.

V
<latexit sha1_base64="A1XDWPnWvrr38Qg2Hbjj1OG2Oh0=">AAAB6HicdVDLSgNBEOz1GeMr6tHLYBA8LbPBoN6CXjwmYB6QLGF2MpuMmZ1dZmaFsOQLvHhQxKuf5M2/cTYPUNGChqKqm+6uIBFcG4w/nZXVtfWNzcJWcXtnd2+/dHDY0nGqKGvSWMSqExDNBJesabgRrJMoRqJAsHYwvsn99gNTmsfyzkwS5kdkKHnIKTFWarT6pTJ2vascCLsVbFFdkgryXDxDGRao90sfvUFM04hJQwXRuuvhxPgZUYZTwabFXqpZQuiYDFnXUkkipv1sdugUnVplgMJY2ZIGzdTvExmJtJ5Ege2MiBnp314u/uV1UxNe+hmXSWqYpPNFYSqQiVH+NRpwxagRE0sIVdzeiuiIKEKNzaZoQ1h+iv4nrYrrVV3cOC/XrhdxFOAYTuAMPLiAGtxCHZpAgcEjPMOLc+88Oa/O27x1xVnMHMEPOO9fB++NFw==</latexit>

Ms4 = | |2
<latexit sha1_base64="2ZKv2nj/rPzNndtx5DhDZmvsYYI=">AAAB+3icdVDLSsNAFJ3UV62vWJduBovgKiShRV0IRTduhAr2AW0aJpNJO3TyYGYilrS/4saFIm79EXf+jZM+QEUPDBzOuYd753gJo0Ka5qdWWFldW98obpa2tnd29/T9ckvEKcekiWMW846HBGE0Ik1JJSOdhBMUeoy0vdFV7rfvCRc0ju7kOCFOiAYRDShGUkmuXr5xRb96ASc9pkI+mvRtV6+YhnWeA5qGbSrUlsSGlmHOUAELNFz9o+fHOA1JJDFDQnQtM5FOhrikmJFpqZcKkiA8QgPSVTRCIRFONrt9Co+V4sMg5upFEs7U74kMhUKMQ09NhkgOxW8vF//yuqkMzpyMRkkqSYTni4KUQRnDvAjoU06wZGNFEOZU3QrxEHGEpaqrpEpY/hT+T1q2YdUM87ZaqV8u6iiCQ3AEToAFTkEdXIMGaAIMHsAjeAYv2lR70l61t/loQVtkDsAPaO9fPMeT7w==</latexit>

|X|
<latexit sha1_base64="gxD/pfcb7DPht99+FiQI281IHDI=">AAAB6nicbVBNS8NAEJ34WetX1aOXxSJ4Kokoeix68VjRfkAbyma7aZduNmF3IpS0P8GLB0W8+ou8+W/ctjlo64OBx3szzMwLEikMuu63s7K6tr6xWdgqbu/s7u2XDg4bJk4143UWy1i3Amq4FIrXUaDkrURzGgWSN4Ph7dRvPnFtRKwecZRwP6J9JULBKFrpYdwad0tlt+LOQJaJl5My5Kh1S1+dXszSiCtkkhrT9twE/YxqFEzySbGTGp5QNqR93rZU0YgbP5udOiGnVumRMNa2FJKZ+nsio5ExoyiwnRHFgVn0puJ/XjvF8NrPhEpS5IrNF4WpJBiT6d+kJzRnKEeWUKaFvZWwAdWUoU2naEPwFl9eJo3zindZce8vytWbPI4CHMMJnIEHV1CFO6hBHRj04Rle4c2Rzovz7nzMW1ecfOYI/sD5/AF6I43s</latexit>

Figure 7.2: The potential of the Polonyi model.

A few comments are in order. First notice that the theory possesses an R-
symmetry, the R-charge of X being R(X) = 2. At a generic point of the moduli
space, then, both supersymmetry and R-symmetry are broken. Second, notice that
although supersymmetry is broken, the spectrum is degenerate in mass: |X|, its
phase α, and ψX are all massless. The fermion field has a good reason to be mass-
less: it is the goldstino predicted by Goldstone theorem. Seemingly, the phase of X
is expected to be massless: it is nothing but the goldstone boson associated to the
broken R-symmetry. Finally, the modulus of the scalar field |X| is massless since it
parametrizes the (non-supersymmetric) moduli space. However, there are no rea-
sons to expect this moduli space to be protected, in principle, against quantum
corrections, since supersymmetry is broken. Hence, generically, one would expect it
to be lifted at the quantum level and |X| to get a mass. This is not the case in this
simple theory, since it is a non-interacting theory, and there are no quantum correc-
tions whatsoever. In general, however, things are different: a non-supersymmetric
moduli space gets typically lifted at one or higher loops, and the putative moduli
get a mass. For this reason, non-supersymmetric moduli spaces are dubbed pseudo-
moduli spaces, and the moduli parametrizing them, pseudo-moduli. We will see
examples of this sort soon.

135
 Let us now consider the following innocent-looking modification of the model
above. Let’s add a mass term to X,
1
∆W = mX 2 . (7.21)
2
Things drastically change, since supersymmetry is now restored. Indeed, the F-term
equation now reads
F (X) = mX + λ = 0 , (7.22)
λ
which admits the solution hXiSU SY = − m . Hence, there is a choice of scalar field
VEV which makes the potential V = |λ + mX|2 vanish, as illustrated in figure 7.3.

V
<latexit sha1_base64="A1XDWPnWvrr38Qg2Hbjj1OG2Oh0=">AAAB6HicdVDLSgNBEOz1GeMr6tHLYBA8LbPBoN6CXjwmYB6QLGF2MpuMmZ1dZmaFsOQLvHhQxKuf5M2/cTYPUNGChqKqm+6uIBFcG4w/nZXVtfWNzcJWcXtnd2+/dHDY0nGqKGvSWMSqExDNBJesabgRrJMoRqJAsHYwvsn99gNTmsfyzkwS5kdkKHnIKTFWarT6pTJ2vascCLsVbFFdkgryXDxDGRao90sfvUFM04hJQwXRuuvhxPgZUYZTwabFXqpZQuiYDFnXUkkipv1sdugUnVplgMJY2ZIGzdTvExmJtJ5Ege2MiBnp314u/uV1UxNe+hmXSWqYpPNFYSqQiVH+NRpwxagRE0sIVdzeiuiIKEKNzaZoQ1h+iv4nrYrrVV3cOC/XrhdxFOAYTuAMPLiAGtxCHZpAgcEjPMOLc+88Oa/O27x1xVnMHMEPOO9fB++NFw==</latexit>

| |2
<latexit sha1_base64="FnnLb2ttjbBiExMfppXWQbSHpsE=">AAAB8nicbVDLSgMxFM3UV62vqks3wSK4KjNF0WXRjcsK9gHTsWQymTY0kwzJHaG0/Qw3LhRx69e4829M21lo64HA4Zxzyb0nTAU34LrfTmFtfWNzq7hd2tnd2z8oHx61jMo0ZU2qhNKdkBgmuGRN4CBYJ9WMJKFg7XB4O/PbT0wbruQDjFIWJKQvecwpASv5k66w2YhMHmu9csWtunPgVeLlpIJyNHrlr26kaJYwCVQQY3zPTSEYEw2cCjYtdTPDUkKHpM98SyVJmAnG85Wn+MwqEY6Vtk8Cnqu/J8YkMWaUhDaZEBiYZW8m/uf5GcTXwZjLNAMm6eKjOBMYFJ7djyOuGQUxsoRQze2umA6IJhRsSyVbgrd88ipp1areZdW9v6jUb/I6iugEnaJz5KErVEd3qIGaiCKFntErenPAeXHenY9FtODkM8foD5zPHzKVkTE=</latexit>

|X|
<latexit sha1_base64="gxD/pfcb7DPht99+FiQI281IHDI=">AAAB6nicbVBNS8NAEJ34WetX1aOXxSJ4Kokoeix68VjRfkAbyma7aZduNmF3IpS0P8GLB0W8+ou8+W/ctjlo64OBx3szzMwLEikMuu63s7K6tr6xWdgqbu/s7u2XDg4bJk4143UWy1i3Amq4FIrXUaDkrURzGgWSN4Ph7dRvPnFtRKwecZRwP6J9JULBKFrpYdwad0tlt+LOQJaJl5My5Kh1S1+dXszSiCtkkhrT9twE/YxqFEzySbGTGp5QNqR93rZU0YgbP5udOiGnVumRMNa2FJKZ+nsio5ExoyiwnRHFgVn0puJ/XjvF8NrPhEpS5IrNF4WpJBiT6d+kJzRnKEeWUKaFvZWwAdWUoU2naEPwFl9eJo3zindZce8vytWbPI4CHMMJnIEHV1CFO6hBHRj04Rle4c2Rzovz7nzMW1ecfOYI/sD5/AF6I43s</latexit>

Figure 7.3: The potential of the massive Polonyi model.

The spectrum is supersymmetric and massive: all fields have mass m. This agrees
with physical expectations: ψX is no more the goldstino, since this is not expected to
be there now; |X| is no more a (pseudo)modulus since the supersymmetric vacuum
is isolated (the VEV of |X| is not a flat direction); finally, α is not anymore the
goldstone boson associated to the broken R-symmetry since the superpotential term
∆W breaks the R-symmetry explicitly. More precisely, W = λX + 12 mX 2 does not
admit any R-charge assignment for X such that R(W ) = 2, meaning that the theory
does not admit a R-symmetry to start with.
Things might also change (both qualitatively and quantitatively) if one allows
the Kähler potential not being canonical. Suppose we keep W = λX but we let the
Kähler metric be non-trivial, that is

∂ 2K
V = (KXX )−1 |λ|2 with KXX = 6= 1 . (7.23)
∂X∂X

136
A non-trivial Kähler metric can deform sensibly the pseudo-moduli space of figure
7.2. A sample of possible different behaviors, which depend on the asymptotic
properties (or singularities) of the Kähler metric, is reported in Figure 7.4.

V V
<latexit sha1_base64="A1XDWPnWvrr38Qg2Hbjj1OG2Oh0=">AAAB6HicdVDLSgNBEOz1GeMr6tHLYBA8LbPBoN6CXjwmYB6QLGF2MpuMmZ1dZmaFsOQLvHhQxKuf5M2/cTYPUNGChqKqm+6uIBFcG4w/nZXVtfWNzcJWcXtnd2+/dHDY0nGqKGvSWMSqExDNBJesabgRrJMoRqJAsHYwvsn99gNTmsfyzkwS5kdkKHnIKTFWarT6pTJ2vascCLsVbFFdkgryXDxDGRao90sfvUFM04hJQwXRuuvhxPgZUYZTwabFXqpZQuiYDFnXUkkipv1sdugUnVplgMJY2ZIGzdTvExmJtJ5Ege2MiBnp314u/uV1UxNe+hmXSWqYpPNFYSqQiVH+NRpwxagRE0sIVdzeiuiIKEKNzaZoQ1h+iv4nrYrrVV3cOC/XrhdxFOAYTuAMPLiAGtxCHZpAgcEjPMOLc+88Oa/O27x1xVnMHMEPOO9fB++NFw==</latexit> <latexit sha1_base64="A1XDWPnWvrr38Qg2Hbjj1OG2Oh0=">AAAB6HicdVDLSgNBEOz1GeMr6tHLYBA8LbPBoN6CXjwmYB6QLGF2MpuMmZ1dZmaFsOQLvHhQxKuf5M2/cTYPUNGChqKqm+6uIBFcG4w/nZXVtfWNzcJWcXtnd2+/dHDY0nGqKGvSWMSqExDNBJesabgRrJMoRqJAsHYwvsn99gNTmsfyzkwS5kdkKHnIKTFWarT6pTJ2vascCLsVbFFdkgryXDxDGRao90sfvUFM04hJQwXRuuvhxPgZUYZTwabFXqpZQuiYDFnXUkkipv1sdugUnVplgMJY2ZIGzdTvExmJtJ5Ege2MiBnp314u/uV1UxNe+hmXSWqYpPNFYSqQiVH+NRpwxagRE0sIVdzeiuiIKEKNzaZoQ1h+iv4nrYrrVV3cOC/XrhdxFOAYTuAMPLiAGtxCHZpAgcEjPMOLc+88Oa/O27x1xVnMHMEPOO9fB++NFw==</latexit>

|X| |X|
<latexit sha1_base64="gxD/pfcb7DPht99+FiQI281IHDI=">AAAB6nicbVBNS8NAEJ34WetX1aOXxSJ4Kokoeix68VjRfkAbyma7aZduNmF3IpS0P8GLB0W8+ou8+W/ctjlo64OBx3szzMwLEikMuu63s7K6tr6xWdgqbu/s7u2XDg4bJk4143UWy1i3Amq4FIrXUaDkrURzGgWSN4Ph7dRvPnFtRKwecZRwP6J9JULBKFrpYdwad0tlt+LOQJaJl5My5Kh1S1+dXszSiCtkkhrT9twE/YxqFEzySbGTGp5QNqR93rZU0YgbP5udOiGnVumRMNa2FJKZ+nsio5ExoyiwnRHFgVn0puJ/XjvF8NrPhEpS5IrNF4WpJBiT6d+kJzRnKEeWUKaFvZWwAdWUoU2naEPwFl9eJo3zindZce8vytWbPI4CHMMJnIEHV1CFO6hBHRj04Rle4c2Rzovz7nzMW1ecfOYI/sD5/AF6I43s</latexit> <latexit sha1_base64="gxD/pfcb7DPht99+FiQI281IHDI=">AAAB6nicbVBNS8NAEJ34WetX1aOXxSJ4Kokoeix68VjRfkAbyma7aZduNmF3IpS0P8GLB0W8+ou8+W/ctjlo64OBx3szzMwLEikMuu63s7K6tr6xWdgqbu/s7u2XDg4bJk4143UWy1i3Amq4FIrXUaDkrURzGgWSN4Ph7dRvPnFtRKwecZRwP6J9JULBKFrpYdwad0tlt+LOQJaJl5My5Kh1S1+dXszSiCtkkhrT9twE/YxqFEzySbGTGp5QNqR93rZU0YgbP5udOiGnVumRMNa2FJKZ+nsio5ExoyiwnRHFgVn0puJ/XjvF8NrPhEpS5IrNF4WpJBiT6d+kJzRnKEeWUKaFvZWwAdWUoU2naEPwFl9eJo3zindZce8vytWbPI4CHMMJnIEHV1CFO6hBHRj04Rle4c2Rzovz7nzMW1ecfOYI/sD5/AF6I43s</latexit>

Stable vacuum, lim KXX = 0 No vacuum, lim KXX = 1

<latexit sha1_base64="E66zs5gHR8M8VNf4duvNmPH8ZKs=">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</latexit> <latexit sha1_base64="fAgFx+PUg71kVH+Ubsjbmori4Lo=">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</latexit>

|X|!1 |X|!1

V V
<latexit sha1_base64="A1XDWPnWvrr38Qg2Hbjj1OG2Oh0=">AAAB6HicdVDLSgNBEOz1GeMr6tHLYBA8LbPBoN6CXjwmYB6QLGF2MpuMmZ1dZmaFsOQLvHhQxKuf5M2/cTYPUNGChqKqm+6uIBFcG4w/nZXVtfWNzcJWcXtnd2+/dHDY0nGqKGvSWMSqExDNBJesabgRrJMoRqJAsHYwvsn99gNTmsfyzkwS5kdkKHnIKTFWarT6pTJ2vascCLsVbFFdkgryXDxDGRao90sfvUFM04hJQwXRuuvhxPgZUYZTwabFXqpZQuiYDFnXUkkipv1sdugUnVplgMJY2ZIGzdTvExmJtJ5Ege2MiBnp314u/uV1UxNe+hmXSWqYpPNFYSqQiVH+NRpwxagRE0sIVdzeiuiIKEKNzaZoQ1h+iv4nrYrrVV3cOC/XrhdxFOAYTuAMPLiAGtxCHZpAgcEjPMOLc+88Oa/O27x1xVnMHMEPOO9fB++NFw==</latexit> <latexit sha1_base64="A1XDWPnWvrr38Qg2Hbjj1OG2Oh0=">AAAB6HicdVDLSgNBEOz1GeMr6tHLYBA8LbPBoN6CXjwmYB6QLGF2MpuMmZ1dZmaFsOQLvHhQxKuf5M2/cTYPUNGChqKqm+6uIBFcG4w/nZXVtfWNzcJWcXtnd2+/dHDY0nGqKGvSWMSqExDNBJesabgRrJMoRqJAsHYwvsn99gNTmsfyzkwS5kdkKHnIKTFWarT6pTJ2vascCLsVbFFdkgryXDxDGRao90sfvUFM04hJQwXRuuvhxPgZUYZTwabFXqpZQuiYDFnXUkkipv1sdugUnVplgMJY2ZIGzdTvExmJtJ5Ege2MiBnp314u/uV1UxNe+hmXSWqYpPNFYSqQiVH+NRpwxagRE0sIVdzeiuiIKEKNzaZoQ1h+iv4nrYrrVV3cOC/XrhdxFOAYTuAMPLiAGtxCHZpAgcEjPMOLc+88Oa/O27x1xVnMHMEPOO9fB++NFw==</latexit>

|X| |X|
<latexit sha1_base64="gxD/pfcb7DPht99+FiQI281IHDI=">AAAB6nicbVBNS8NAEJ34WetX1aOXxSJ4Kokoeix68VjRfkAbyma7aZduNmF3IpS0P8GLB0W8+ou8+W/ctjlo64OBx3szzMwLEikMuu63s7K6tr6xWdgqbu/s7u2XDg4bJk4143UWy1i3Amq4FIrXUaDkrURzGgWSN4Ph7dRvPnFtRKwecZRwP6J9JULBKFrpYdwad0tlt+LOQJaJl5My5Kh1S1+dXszSiCtkkhrT9twE/YxqFEzySbGTGp5QNqR93rZU0YgbP5udOiGnVumRMNa2FJKZ+nsio5ExoyiwnRHFgVn0puJ/XjvF8NrPhEpS5IrNF4WpJBiT6d+kJzRnKEeWUKaFvZWwAdWUoU2naEPwFl9eJo3zindZce8vytWbPI4CHMMJnIEHV1CFO6hBHRj04Rle4c2Rzovz7nzMW1ecfOYI/sD5/AF6I43s</latexit> <latexit sha1_base64="gxD/pfcb7DPht99+FiQI281IHDI=">AAAB6nicbVBNS8NAEJ34WetX1aOXxSJ4Kokoeix68VjRfkAbyma7aZduNmF3IpS0P8GLB0W8+ou8+W/ctjlo64OBx3szzMwLEikMuu63s7K6tr6xWdgqbu/s7u2XDg4bJk4143UWy1i3Amq4FIrXUaDkrURzGgWSN4Ph7dRvPnFtRKwecZRwP6J9JULBKFrpYdwad0tlt+LOQJaJl5My5Kh1S1+dXszSiCtkkhrT9twE/YxqFEzySbGTGp5QNqR93rZU0YgbP5udOiGnVumRMNa2FJKZ+nsio5ExoyiwnRHFgVn0puJ/XjvF8NrPhEpS5IrNF4WpJBiT6d+kJzRnKEeWUKaFvZWwAdWUoU2naEPwFl9eJo3zindZce8vytWbPI4CHMMJnIEHV1CFO6hBHRj04Rle4c2Rzovz7nzMW1ecfOYI/sD5/AF6I43s</latexit>

K singular at finite distance in field Metastable vacuum, lim KXX = 1

<latexit sha1_base64="fAgFx+PUg71kVH+Ubsjbmori4Lo=">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</latexit>

|X|!1
space. Effective description breaks
down at |X| = |X0 | , lim KXX = 1
<latexit sha1_base64="huNTDdXKxAoD5rj3hVtMM4mZk28=">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</latexit>

<latexit sha1_base64="ekpeODJw84PR/WfNKCfoq7qXKNM=">AAAB8HicbVBNSwMxEJ2tX7V+VT16CRbBU9kVRS9C0YvHCrZdaZeSTbNtaJJdkqxQuv0VXjwo4tWf481/Y9ruQVsfDDzem2FmXphwpo3rfjuFldW19Y3iZmlre2d3r7x/0NRxqghtkJjHyg+xppxJ2jDMcOonimIRctoKh7dTv/VElWaxfDCjhAYC9yWLGMHGSo+Zn11nftfNuuWKW3VnQMvEy0kFctS75a9OLyapoNIQjrVue25igjFWhhFOJ6VOqmmCyRD3adtSiQXVwXh28ASdWKWHoljZkgbN1N8TYyy0HonQdgpsBnrRm4r/ee3URFfBmMkkNVSS+aIo5cjEaPo96jFFieEjSzBRzN6KyAArTIzNqGRD8BZfXibNs6p3UXXvzyu1mzyOIhzBMZyCB5dQgzuoQwMICHiGV3hzlPPivDsf89aCk88cwh84nz+XVJBE</latexit>

|X|!|X0 |

Figure 7.4: Qualitatively different potentials of non-canonical Polonyi-like models.

Physically, the different behaviors reported in figure 7.4 should be understood as

follows. In presence of a classical pseudomoduli space like the one in figure 7.2, the
lifting of the pseudomoduli at quantum level occurs because the massless particles
which bring from a vacuum to another get a mass at one loop. Sometime, such effect
can be mimicked by a non-canonical Kähler potential. In fact, these seemingly ad-
hoc theories can and sometime do arise at low energies as effective theories of more
complicated UV-renormalizable theories: the mass scale entering the non-canonical
Kähler potential is nothing but the UV cut-off of these low energy effective theories.
Let us try to make the above discussion more concrete by considering an explicit
example.

137
Example 2 : A Polonyi model with quartic Kähler potential.
Let us consider the following model
c
2
K = XX − 2 XX , W = λX , (7.24)
Λ
where c > 0. Notice that the R-symmetry is not broken by the non-canonical
Kähler potential (7.24), which is R-symmetry invariant. So the U (1)R symmetry is
a symmetry of the theory. The Kähler metric and the scalar potential now read
4c |λ|2
KXX = 1 − XX , V = K XX |λ|2 = . (7.25)
Λ2 1 − Λ4c2 XX
The Kähler potential is an instance of models like those in the upper-left diagram of
√
Figure 7.4: the Kähler metric KXX vanishes for large enough |X|, |X| → |Λ|/2 c,
which is order the natural cut-off of the theory. The potential, which is depicted
in figure 7.5, admits a (unique) mimimum at hXi = 0. Therefore, there is an

V
<latexit sha1_base64="A1XDWPnWvrr38Qg2Hbjj1OG2Oh0=">AAAB6HicdVDLSgNBEOz1GeMr6tHLYBA8LbPBoN6CXjwmYB6QLGF2MpuMmZ1dZmaFsOQLvHhQxKuf5M2/cTYPUNGChqKqm+6uIBFcG4w/nZXVtfWNzcJWcXtnd2+/dHDY0nGqKGvSWMSqExDNBJesabgRrJMoRqJAsHYwvsn99gNTmsfyzkwS5kdkKHnIKTFWarT6pTJ2vascCLsVbFFdkgryXDxDGRao90sfvUFM04hJQwXRuuvhxPgZUYZTwabFXqpZQuiYDFnXUkkipv1sdugUnVplgMJY2ZIGzdTvExmJtJ5Ege2MiBnp314u/uV1UxNe+hmXSWqYpPNFYSqQiVH+NRpwxagRE0sIVdzeiuiIKEKNzaZoQ1h+iv4nrYrrVV3cOC/XrhdxFOAYTuAMPLiAGtxCHZpAgcEjPMOLc+88Oa/O27x1xVnMHMEPOO9fB++NFw==</latexit>

| |2
<latexit sha1_base64="FnnLb2ttjbBiExMfppXWQbSHpsE=">AAAB8nicbVDLSgMxFM3UV62vqks3wSK4KjNF0WXRjcsK9gHTsWQymTY0kwzJHaG0/Qw3LhRx69e4829M21lo64HA4Zxzyb0nTAU34LrfTmFtfWNzq7hd2tnd2z8oHx61jMo0ZU2qhNKdkBgmuGRN4CBYJ9WMJKFg7XB4O/PbT0wbruQDjFIWJKQvecwpASv5k66w2YhMHmu9csWtunPgVeLlpIJyNHrlr26kaJYwCVQQY3zPTSEYEw2cCjYtdTPDUkKHpM98SyVJmAnG85Wn+MwqEY6Vtk8Cnqu/J8YkMWaUhDaZEBiYZW8m/uf5GcTXwZjLNAMm6eKjOBMYFJ7djyOuGQUxsoRQze2umA6IJhRsSyVbgrd88ipp1areZdW9v6jUb/I6iugEnaJz5KErVEd3qIGaiCKFntErenPAeXHenY9FtODkM8foD5zPHzKVkTE=</latexit>

⇤
<latexit sha1_base64="yFrnXXViWirehqLFkWd9XsZYx1w=">AAAB7nicbVBNS8NAFHypX7V+VT16WSyCp5KIoseiFw8eKlhbaEPZbDbt0s0m7L4IJfRHePGgiFd/jzf/jds2B20dWBhm5rHvTZBKYdB1v53Syura+kZ5s7K1vbO7V90/eDRJphlvsUQmuhNQw6VQvIUCJe+kmtM4kLwdjG6mfvuJayMS9YDjlPsxHSgRCUbRSu3enY2GtF+tuXV3BrJMvILUoECzX/3qhQnLYq6QSWpM13NT9HOqUTDJJ5VeZnhK2YgOeNdSRWNu/Hy27oScWCUkUaLtU0hm6u+JnMbGjOPAJmOKQ7PoTcX/vG6G0ZWfC5VmyBWbfxRlkmBCpreTUGjOUI4toUwLuythQ6opQ9tQxZbgLZ68TB7P6t5F3b0/rzWuizrKcATHcAoeXEIDbqEJLWAwgmd4hTcndV6cd+djHi05xcwh/IHz+QMM949h</latexit>

|X|
<latexit sha1_base64="gxD/pfcb7DPht99+FiQI281IHDI=">AAAB6nicbVBNS8NAEJ34WetX1aOXxSJ4Kokoeix68VjRfkAbyma7aZduNmF3IpS0P8GLB0W8+ou8+W/ctjlo64OBx3szzMwLEikMuu63s7K6tr6xWdgqbu/s7u2XDg4bJk4143UWy1i3Amq4FIrXUaDkrURzGgWSN4Ph7dRvPnFtRKwecZRwP6J9JULBKFrpYdwad0tlt+LOQJaJl5My5Kh1S1+dXszSiCtkkhrT9twE/YxqFEzySbGTGp5QNqR93rZU0YgbP5udOiGnVumRMNa2FJKZ+nsio5ExoyiwnRHFgVn0puJ/XjvF8NrPhEpS5IrNF4WpJBiT6d+kJzRnKEeWUKaFvZWwAdWUoU2naEPwFl9eJo3zindZce8vytWbPI4CHMMJnIEHV1CFO6hBHRj04Rle4c2Rzovz7nzMW1ecfOYI/sD5/AF6I43s</latexit>

Figure 7.5: The potential of a Polonyi model with quartic Kähler potential (7.24).

isolated vacuum now and it is a supersymmetry breaking one. One can compute the
spectrum around such vacuum and find that ψX is consistently massless (it is the
goldstino), while now the scalar field is massive, m2X ∼ c|λ|2 /Λ2 .

Example 3 : Supersymmetry restoration by new degrees of freedom.

Let us now deform the basic Polonyi model by adding a new superfield, Y , while
keeping the Kähler potential canonical. The superpotential reads
1
W = λX + hXY 2 . (7.26)
2
Notice that this model has an R-symmetry, with R-charge assignment R(X) =
2 , R(Y ) = 0. From the F-equations one can compute the potential which reads
1
V = |hXY |2 + | hY 2 + λ|2 , (7.27)
2
138
implying that there are two supersymmetric vacua at
r
λ
hXiSU SY = 0 , hY iSU SY = ± −2 . (7.28)
h
So we see that the additional degrees of freedom have restored supersymmetry.
Interestingly, there are other local minima of the potential, a pseudo-moduli space
in fact, where supersymmetry is broken

hXiSB = any , hY iSB = 0 where V = |λ|2 . (7.29)

The physical interpretation is as follows. For large hXi, the superfield Y gets a
large mass and affects the low energy theory lesser and lesser. The theory reduces
effectively to the original Polonyi model, which breaks supersymmetry and whose
vacuum energy is indeed V = |λ|2 . It is a simple but instructive exercise to compute
the mass spectrum around the non-supersymmetric minima. The chiral superfield
X is obviously massless while Y gets a mass. There is a first (obvious) contribution
to both the scalar and the fermion components of Y from hhXi, and a second
contribution which affects only the scalar component of Y coming from FX , which
is non-vanishing. The end result is

m2Y = |hhXi|2 ± |hλ| , mψY = hhXi . (7.30)

From the above expressions, we see that the supersymmetry breaking pseudomoduli
space has a tachyonic mode which develops (and destabilizes the vacuum) for

|X|2 < |λ/h| ≡ |Xc |2 . (7.31)

In such region the potential decreases along the hY i direction towards the super-
symmetry vacua. A qualitative picture of the potential is reported in Figure 7.6.

Example 4 : Runaway behavior.

A minimal modification of the above theory gives a completely different dynam-
ics. Let us suppose that the cubic term of the superpotential (7.26) has the square
shifted from Y to X. In this case we would have for the superpotential the following
ewxpression
1
W = λX + hX 2 Y . (7.32)
2
The F–equations are
1
F X = λ + hXY , F Y = hX 2 , (7.33)
2
139
 V
<latexit sha1_base64="A1XDWPnWvrr38Qg2Hbjj1OG2Oh0=">AAAB6HicdVDLSgNBEOz1GeMr6tHLYBA8LbPBoN6CXjwmYB6QLGF2MpuMmZ1dZmaFsOQLvHhQxKuf5M2/cTYPUNGChqKqm+6uIBFcG4w/nZXVtfWNzcJWcXtnd2+/dHDY0nGqKGvSWMSqExDNBJesabgRrJMoRqJAsHYwvsn99gNTmsfyzkwS5kdkKHnIKTFWarT6pTJ2vascCLsVbFFdkgryXDxDGRao90sfvUFM04hJQwXRuuvhxPgZUYZTwabFXqpZQuiYDFnXUkkipv1sdugUnVplgMJY2ZIGzdTvExmJtJ5Ege2MiBnp314u/uV1UxNe+hmXSWqYpPNFYSqQiVH+NRpwxagRE0sIVdzeiuiIKEKNzaZoQ1h+iv4nrYrrVV3cOC/XrhdxFOAYTuAMPLiAGtxCHZpAgcEjPMOLc+88Oa/O27x1xVnMHMEPOO9fB++NFw==</latexit>

|Xc |
<latexit sha1_base64="auHG8QLA4FJSoC9o092bCxnNCrg=">AAAB7HicbVBNS8NAEJ3Ur1q/qh69LBbBU0lE0WPRi8cKpi20oWy2m3bpZhN2J0Jp+xu8eFDEqz/Im//GbZuDtj4YeLw3w8y8MJXCoOt+O4W19Y3NreJ2aWd3b/+gfHjUMEmmGfdZIhPdCqnhUijuo0DJW6nmNA4lb4bDu5nffOLaiEQ94ijlQUz7SkSCUbSSP2l12aRbrrhVdw6ySrycVCBHvVv+6vQSlsVcIZPUmLbnphiMqUbBJJ+WOpnhKWVD2udtSxWNuQnG82On5MwqPRIl2pZCMld/T4xpbMwoDm1nTHFglr2Z+J/XzjC6CcZCpRlyxRaLokwSTMjsc9ITmjOUI0so08LeStiAasrQ5lOyIXjLL6+SxkXVu6q6D5eV2m0eRxFO4BTOwYNrqME91MEHBgKe4RXeHOW8OO/Ox6K14OQzx/AHzucP7NWOwg==</latexit>

|X|
<latexit sha1_base64="gxD/pfcb7DPht99+FiQI281IHDI=">AAAB6nicbVBNS8NAEJ34WetX1aOXxSJ4Kokoeix68VjRfkAbyma7aZduNmF3IpS0P8GLB0W8+ou8+W/ctjlo64OBx3szzMwLEikMuu63s7K6tr6xWdgqbu/s7u2XDg4bJk4143UWy1i3Amq4FIrXUaDkrURzGgWSN4Ph7dRvPnFtRKwecZRwP6J9JULBKFrpYdwad0tlt+LOQJaJl5My5Kh1S1+dXszSiCtkkhrT9twE/YxqFEzySbGTGp5QNqR93rZU0YgbP5udOiGnVumRMNa2FJKZ+nsio5ExoyiwnRHFgVn0puJ/XjvF8NrPhEpS5IrNF4WpJBiT6d+kJzRnKEeWUKaFvZWwAdWUoU2naEPwFl9eJo3zindZce8vytWbPI4CHMMJnIEHV1CFO6hBHRj04Rle4c2Rzovz7nzMW1ecfOYI/sD5/AF6I43s</latexit>

|YSUSY |
<latexit sha1_base64="1ok9aiCmTZbSevi7WphQgFVeoQE=">AAACAHicbVC7TsMwFHV4lvIKMDCwWFRITFWCQDBWsDAWlfShJqoc12mtOk5kO4gozcKvsDCAECufwcbf4LYZoOVIVzo6517de48fMyqVZX0bS8srq2vrpY3y5tb2zq65t9+UUSIwcXDEItH2kSSMcuIoqhhpx4Kg0Gek5Y9uJn7rgQhJI36v0ph4IRpwGlCMlJZ65uG408vc0I8eM1dRnsKG0+jk+bhnVqyqNQVcJHZBKqBAvWd+uf0IJyHhCjMkZde2YuVlSCiKGcnLbiJJjPAIDUhXU45CIr1s+kAOT7TSh0EkdHEFp+rviQyFUqahrztDpIZy3puI/3ndRAVXXkZ5nCjC8WxRkDCoIjhJA/apIFixVBOEBdW3QjxEAmGlMyvrEOz5lxdJ86xqX1Stu/NK7bqIowSOwDE4BTa4BDVwC+rAARjk4Bm8gjfjyXgx3o2PWeuSUcwcgD8wPn8ATeCW2w==</latexit>

|Y |
<latexit sha1_base64="0R/vwjUbmaGK0FiJxANgXXdvOwk=">AAAB6nicbVDLTgJBEOzFF+IL9ehlIjHxRHaNRo9ELx4xysPAhswOvTBhdnYzM2tCgE/w4kFjvPpF3vwbB9iDgpV0UqnqTndXkAiujet+O7mV1bX1jfxmYWt7Z3evuH9Q13GqGNZYLGLVDKhGwSXWDDcCm4lCGgUCG8HgZuo3nlBpHssHM0zQj2hP8pAzaqx0P34cd4olt+zOQJaJl5ESZKh2il/tbszSCKVhgmrd8tzE+COqDGcCJ4V2qjGhbEB72LJU0gi1P5qdOiEnVumSMFa2pCEz9ffEiEZaD6PAdkbU9PWiNxX/81qpCa/8EZdJalCy+aIwFcTEZPo36XKFzIihJZQpbm8lrE8VZcamU7AheIsvL5P6Wdm7KLt356XKdRZHHo7gGE7Bg0uowC1UoQYMevAMr/DmCOfFeXc+5q05J5s5hD9wPn8Ae6iN7Q==</latexit>

Figure 7.6: The potential of the model of Example 3: supersymmetry restoration.

and the scalar potential

1
V = | hX 2 |2 + |hXY + λ|2 . (7.34)
2
Differently from previous example, it is not possible to satisfy both F-equations and
so there are no supersymmetric ground states now. Notice, in passing, that the
R-symmetry is preserved by the superpotential as in the previous example, with
charge assignment R(X) = 2 and R(Y ) = −2 this time.
Now the question is: where is the minimum of this supersymmetry breaking
potential? An analysis of V shows that the global minimum is reached for Y → ∞.
λ
A quick way to see it is to set X = − hY , which kills the second contribution to the
potential, the FX -equation. By plugging this back into V one gets

λ2 2
V = −→ 0 . (7.35)
2hY 2 Y →∞

In other words, there is no stable vacuum but actually a runaway behavior and
supersymmetry is restored at infinity in field space.
A more physical way to reach the same conclusion is as follows. For large |Y | the
amount of supersymmetry breaking gets smaller and smaller and X mass larger and
larger. Hence the theory can be described by a theory where X is integrated out
solving its equation of motion, which for large enough mass reduces to ∂W/∂X = 0.

140
This sets X = −λ/(hY ) and the Y -dependent only superpotential becomes

λ2
Weff = − , (7.36)
2hY
which gives the runaway behavior described by the potential (7.35).
Notice that within all models discussed so far, the only renormalizable one which
breaks supersymmetry in a stable vacuum is the original Polonyi model (in fact there
is an all pseudo-moduli space). This model is however rather uninteresting per sé,
since it describes a non-interacting theory. One might wonder whether there exist
reasonably simple models which are renormalizable, interacting and break super-
symmetry in stable vacua. The simplest such model is the re-known O’Raifeartaigh
model, which we now describe.

Example 5 : The O’Raifeartaigh model.

Let us consider the theory of three chiral superfields with canonical Kähler po-
tential and a superpotential given by
1
W = hXΦ21 + mΦ1 Φ2 − µ2 X . (7.37)
2
The superpotential respects the R-symmetry, the R-charge assignment being R(X) =
2, R(Φ1 ) = 0 and R(Φ2 ) = 2. The F-term equations read

1 2 2

 F X = 2 hφ1 − µ
F 1 = hXφ1 + mφ2 (7.38)


F 2 = mφ1

Clearly the first and the third equations cannot be solved simultaneously. Hence
supersymmetry is broken. Let us try to analyze the theory a bit further. There are
two dimensionful scales, µ and m. Let us choose in what follows |µ| < |m| (nothing
crucial of the following analysis would change choosing a different regime). In this
regime one can show that the minimum of the potential is at

φ1 = φ2 = 0 , X = any (7.39)

and the vacuum energy is V = |µ2 |2 . Again, we find a pseudo-moduli space of vacua
since X is not fixed by the minimal energy condition. In Figure 7.7 we depict the
potential as a function of the scalar fields.
Let us compute the (classical) spectrum around the supersymmetry breaking
vacua. The full chiral superfield X is massless, right in the same way as for the

141
 V
<latexit sha1_base64="A1XDWPnWvrr38Qg2Hbjj1OG2Oh0=">AAAB6HicdVDLSgNBEOz1GeMr6tHLYBA8LbPBoN6CXjwmYB6QLGF2MpuMmZ1dZmaFsOQLvHhQxKuf5M2/cTYPUNGChqKqm+6uIBFcG4w/nZXVtfWNzcJWcXtnd2+/dHDY0nGqKGvSWMSqExDNBJesabgRrJMoRqJAsHYwvsn99gNTmsfyzkwS5kdkKHnIKTFWarT6pTJ2vascCLsVbFFdkgryXDxDGRao90sfvUFM04hJQwXRuuvhxPgZUYZTwabFXqpZQuiYDFnXUkkipv1sdugUnVplgMJY2ZIGzdTvExmJtJ5Ege2MiBnp314u/uV1UxNe+hmXSWqYpPNFYSqQiVH+NRpwxagRE0sIVdzeiuiIKEKNzaZoQ1h+iv4nrYrrVV3cOC/XrhdxFOAYTuAMPLiAGtxCHZpAgcEjPMOLc+88Oa/O27x1xVnMHMEPOO9fB++NFw==</latexit>

| 2|
<latexit sha1_base64="v47UxyWxzNC7r7dLjh3nAxbpQvQ=">AAAB73icbVDLTgJBEOzFF+IL9ehlIjHxRHaJRo9ELx4xkUcCGzI7zMKE2dl1pteEAD/hxYPGePV3vPk3DrAHBSvppFLVne6uIJHCoOt+O7m19Y3Nrfx2YWd3b/+geHjUMHGqGa+zWMa6FVDDpVC8jgIlbyWa0yiQvBkMb2d+84lrI2L1gKOE+xHtKxEKRtFKrUknGYhuZdItltyyOwdZJV5GSpCh1i1+dXoxSyOukElqTNtzE/THVKNgkk8LndTwhLIh7fO2pYpG3Pjj+b1TcmaVHgljbUshmau/J8Y0MmYUBbYzojgwy95M/M9rpxhe+2OhkhS5YotFYSoJxmT2POkJzRnKkSWUaWFvJWxANWVoIyrYELzll1dJo1L2Lsvu/UWpepPFkYcTOIVz8OAKqnAHNagDAwnP8ApvzqPz4rw7H4vWnJPNHMMfOJ8/BYqP9A==</latexit>

| 1|
<latexit sha1_base64="vU9tQEZwDfKfbrA210zhZJKc/QA=">AAAB73icbVBNS8NAEJ34WetX1aOXxSJ4Kokoeix68VjBfkAbymY7aZduNnF3I5S0f8KLB0W8+ne8+W/ctjlo64OBx3szzMwLEsG1cd1vZ2V1bX1js7BV3N7Z3dsvHRw2dJwqhnUWi1i1AqpRcIl1w43AVqKQRoHAZjC8nfrNJ1Sax/LBjBL0I9qXPOSMGiu1xp1kwLveuFsquxV3BrJMvJyUIUetW/rq9GKWRigNE1Trtucmxs+oMpwJnBQ7qcaEsiHtY9tSSSPUfja7d0JOrdIjYaxsSUNm6u+JjEZaj6LAdkbUDPSiNxX/89qpCa/9jMskNSjZfFGYCmJiMn2e9LhCZsTIEsoUt7cSNqCKMmMjKtoQvMWXl0njvOJdVtz7i3L1Jo+jAMdwAmfgwRVU4Q5qUAcGAp7hFd6cR+fFeXc+5q0rTj5zBH/gfP4ABAWP8w==</latexit>

|X|
<latexit sha1_base64="gxD/pfcb7DPht99+FiQI281IHDI=">AAAB6nicbVBNS8NAEJ34WetX1aOXxSJ4Kokoeix68VjRfkAbyma7aZduNmF3IpS0P8GLB0W8+ou8+W/ctjlo64OBx3szzMwLEikMuu63s7K6tr6xWdgqbu/s7u2XDg4bJk4143UWy1i3Amq4FIrXUaDkrURzGgWSN4Ph7dRvPnFtRKwecZRwP6J9JULBKFrpYdwad0tlt+LOQJaJl5My5Kh1S1+dXszSiCtkkhrT9twE/YxqFEzySbGTGp5QNqR93rZU0YgbP5udOiGnVumRMNa2FJKZ+nsio5ExoyiwnRHFgVn0puJ/XjvF8NrPhEpS5IrNF4WpJBiT6d+kJzRnKEeWUKaFvZWwAdWUoU2naEPwFl9eJo3zindZce8vytWbPI4CHMMJnIEHV1CFO6hBHRj04Rle4c2Rzovz7nzMW1ecfOYI/sD5/AF6I43s</latexit>

Figure 7.7: The (classical) potential of the O’Raifeartaigh model.

Polonyi model (notice that for larger and larger |X| the model gets closer and closer
to the Polonyi model since all other fields get heavier and heavier). The massless
fermion mode ψX is nothing but the goldstino. The only non vanishing F-term
in the vacuum is FX , so that the goldstino gets contribution only from ψX agrees
with eq. (7.10). The phase α of the scalar field X = |X|eiα is the Goldstone boson
associated to R-symmetry, which is spontaneously broken in the vacuum. Note that
while Φ2 is charged under the R-symmetry, the phase of φ2 does not contribute to
the R-axion since the VEV of φ2 is vanishing in the supersymmetry breaking vacua
(7.39). Finally, |X| is massless since it is a modulus (at least at classical level).
One can easily compute the (|X|-dependent) mass spectrum of all other fields
and get
1 1
m20 (|X|) = |m|2 + η|hµ2 | + |hX|2
2 2
1 p
± |hµ2 |2 + 2η|hµ2 ||hX|2 + 4|m|2 |hX|2 + |hX|4
2
1 p 2
m21/2 (|X|) = |hX| ± |hX|2 + 4|m|2 . (7.40)
4
where η = ±1, giving different masses to the four real scalar modes belonging to Φ1
and Φ2 . As expected, the spectrum is manifestly non-supersymmetric. Notice that
m21/2 (|X|) = m20 (|X|)|µ2 =0 . Note that these infinitely many (degenerate in energy)
vacua are in fact physically inequivalent, since the mass spectrum depends on |X|.

Example 6 : A modified O’Raifeartaigh model.

142
 Let us end this overview by considering a modification of the previous model.
Let us add a (small) mass perturbation for Φ2
1
∆W =  mΦ22 with  << 1 . (7.41)
2
Notice that this term breaks the R-symmetry enjoyed by the original O’Raifeartaigh
model. The only F-equation which gets modified is the one for Φ2 which now reads

F 2 = mφ1 + mφ2 . (7.42)

The presence of the second term removes the conflict we had before between this
equation and the F-equation for X. Hence we can solve all F-term equations simul-
taneously and supersymmetry is not broken anymore. The (two) supersymmetric
vacua are at r r
m 2µ2 1 2µ2
X= , φ1 = ± , φ2 = ∓ . (7.43)
h h  h
For  << 1 these vacua are far away, in field space, from where the supersymmetry
breaking vacua of the O’Raifeartaigh model sit (the VEV of φ2 becomes larger and
larger and hence very far from φ2 = 0, the value of φ2 in O’Raifeartaigh model
supersymmetry breaking vacua). In fact, near the origin of field space the potential
of the present model is practically identical to the one of the original O’Raifeartaigh
model and one can show by direct computation that a classically marginal pseudo-
moduli space of supersymmetry breaking minima is present, there.
Computing the mass spectrum near the origin one gets now
1n

m20 (|X|) = |hX|2 + |m|2 2 + ||2 + η|hµ2 |
2
q o
2
± [|hX| + |m| (2 + || ) + η|hµ |] − 4|m| [|hX − m| + η|hµ | (1 + || )]
2 2 2 2 2 2 2 2

m21/2 (|X|) = m20 (|X|)|µ2 =0 . (7.44)

A close look to the above spectrum shows that in order for mass squared eigenvalues
being all positive, the following inequality should be satisfied
hX 2 hµ2
1− > (1 + ||2 ) . (7.45)
m m2
For small  and µ/m, the marginally stable region described by the above inequality
includes a large neighborhood around the origin, and the tachyonic mode develops
only for |X| (parametrically) larger than a critical value |Xc |

|X| < |Xc | . (7.46)

143
Notice that this is quite the opposite of what we got in Example 4, where the
marginally stable region was above a critical value; as we will see, this difference has
crucial consequences at the quantum level. For  → 0 one gets that Xc , XSU SY → ∞
and the supersymmetric vacua are pushed all the way to infinity. This is consistent
with the fact that for  = 0 one recovers the original O’Raifeartaigh model where
supersymmetric vacua are not present. A rough picture of the potential is given in
Figure 7.8.

V
<latexit sha1_base64="A1XDWPnWvrr38Qg2Hbjj1OG2Oh0=">AAAB6HicdVDLSgNBEOz1GeMr6tHLYBA8LbPBoN6CXjwmYB6QLGF2MpuMmZ1dZmaFsOQLvHhQxKuf5M2/cTYPUNGChqKqm+6uIBFcG4w/nZXVtfWNzcJWcXtnd2+/dHDY0nGqKGvSWMSqExDNBJesabgRrJMoRqJAsHYwvsn99gNTmsfyzkwS5kdkKHnIKTFWarT6pTJ2vascCLsVbFFdkgryXDxDGRao90sfvUFM04hJQwXRuuvhxPgZUYZTwabFXqpZQuiYDFnXUkkipv1sdugUnVplgMJY2ZIGzdTvExmJtJ5Ege2MiBnp314u/uV1UxNe+hmXSWqYpPNFYSqQiVH+NRpwxagRE0sIVdzeiuiIKEKNzaZoQ1h+iv4nrYrrVV3cOC/XrhdxFOAYTuAMPLiAGtxCHZpAgcEjPMOLc+88Oa/O27x1xVnMHMEPOO9fB++NFw==</latexit>

| 2|
<latexit sha1_base64="v47UxyWxzNC7r7dLjh3nAxbpQvQ=">AAAB73icbVDLTgJBEOzFF+IL9ehlIjHxRHaJRo9ELx4xkUcCGzI7zMKE2dl1pteEAD/hxYPGePV3vPk3DrAHBSvppFLVne6uIJHCoOt+O7m19Y3Nrfx2YWd3b/+geHjUMHGqGa+zWMa6FVDDpVC8jgIlbyWa0yiQvBkMb2d+84lrI2L1gKOE+xHtKxEKRtFKrUknGYhuZdItltyyOwdZJV5GSpCh1i1+dXoxSyOukElqTNtzE/THVKNgkk8LndTwhLIh7fO2pYpG3Pjj+b1TcmaVHgljbUshmau/J8Y0MmYUBbYzojgwy95M/M9rpxhe+2OhkhS5YotFYSoJxmT2POkJzRnKkSWUaWFvJWxANWVoIyrYELzll1dJo1L2Lsvu/UWpepPFkYcTOIVz8OAKqnAHNagDAwnP8ApvzqPz4rw7H4vWnJPNHMMfOJ8/BYqP9A==</latexit>

| 1|
<latexit sha1_base64="vU9tQEZwDfKfbrA210zhZJKc/QA=">AAAB73icbVBNS8NAEJ34WetX1aOXxSJ4Kokoeix68VjBfkAbymY7aZduNnF3I5S0f8KLB0W8+ne8+W/ctjlo64OBx3szzMwLEsG1cd1vZ2V1bX1js7BV3N7Z3dsvHRw2dJwqhnUWi1i1AqpRcIl1w43AVqKQRoHAZjC8nfrNJ1Sax/LBjBL0I9qXPOSMGiu1xp1kwLveuFsquxV3BrJMvJyUIUetW/rq9GKWRigNE1Trtucmxs+oMpwJnBQ7qcaEsiHtY9tSSSPUfja7d0JOrdIjYaxsSUNm6u+JjEZaj6LAdkbUDPSiNxX/89qpCa/9jMskNSjZfFGYCmJiMn2e9LhCZsTIEsoUt7cSNqCKMmMjKtoQvMWXl0njvOJdVtz7i3L1Jo+jAMdwAmfgwRVU4Q5qUAcGAp7hFd6cR+fFeXc+5q0rTj5zBH/gfP4ABAWP8w==</latexit>

|Xc |
<latexit sha1_base64="auHG8QLA4FJSoC9o092bCxnNCrg=">AAAB7HicbVBNS8NAEJ3Ur1q/qh69LBbBU0lE0WPRi8cKpi20oWy2m3bpZhN2J0Jp+xu8eFDEqz/Im//GbZuDtj4YeLw3w8y8MJXCoOt+O4W19Y3NreJ2aWd3b/+gfHjUMEmmGfdZIhPdCqnhUijuo0DJW6nmNA4lb4bDu5nffOLaiEQ94ijlQUz7SkSCUbSSP2l12aRbrrhVdw6ySrycVCBHvVv+6vQSlsVcIZPUmLbnphiMqUbBJJ+WOpnhKWVD2udtSxWNuQnG82On5MwqPRIl2pZCMld/T4xpbMwoDm1nTHFglr2Z+J/XzjC6CcZCpRlyxRaLokwSTMjsc9ITmjOUI0so08LeStiAasrQ5lOyIXjLL6+SxkXVu6q6D5eV2m0eRxFO4BTOwYNrqME91MEHBgKe4RXeHOW8OO/Ox6K14OQzx/AHzucP7NWOwg==</latexit>

|XSUSY |
<latexit sha1_base64="9RAU6+c002TkclZMcidHBTCPh7Q=">AAACAHicbVC7TsMwFHXKq5RXgIGBxaJCYqoSBIKxgoWxqKQtaqLIcZ3WquNEtoOI0iz8CgsDCLHyGWz8De5jgJYjXenonHt17z1BwqhUlvVtlJaWV1bXyuuVjc2t7R1zd68l41Rg4uCYxaITIEkY5cRRVDHSSQRBUcBIOxhej/32AxGSxvxOZQnxItTnNKQYKS355sGo4+duFMSPuasoz2DTad4Xxcg3q1bNmgAuEntGqmCGhm9+ub0YpxHhCjMkZde2EuXlSCiKGSkqbipJgvAQ9UlXU44iIr188kABj7XSg2EsdHEFJ+rviRxFUmZRoDsjpAZy3huL/3ndVIWXXk55kirC8XRRmDKoYjhOA/aoIFixTBOEBdW3QjxAAmGlM6voEOz5lxdJ67Rmn9es27Nq/WoWRxkcgiNwAmxwAergBjSAAzAowDN4BW/Gk/FivBsf09aSMZvZB39gfP4ATEeW2g==</latexit>

|X|
<latexit sha1_base64="gxD/pfcb7DPht99+FiQI281IHDI=">AAAB6nicbVBNS8NAEJ34WetX1aOXxSJ4Kokoeix68VjRfkAbyma7aZduNmF3IpS0P8GLB0W8+ou8+W/ctjlo64OBx3szzMwLEikMuu63s7K6tr6xWdgqbu/s7u2XDg4bJk4143UWy1i3Amq4FIrXUaDkrURzGgWSN4Ph7dRvPnFtRKwecZRwP6J9JULBKFrpYdwad0tlt+LOQJaJl5My5Kh1S1+dXszSiCtkkhrT9twE/YxqFEzySbGTGp5QNqR93rZU0YgbP5udOiGnVumRMNa2FJKZ+nsio5ExoyiwnRHFgVn0puJ/XjvF8NrPhEpS5IrNF4WpJBiT6d+kJzRnKEeWUKaFvZWwAdWUoU2naEPwFl9eJo3zindZce8vytWbPI4CHMMJnIEHV1CFO6hBHRj04Rle4c2Rzovz7nzMW1ecfOYI/sD5/AF6I43s</latexit>

Figure 7.8: The (classical) potential of the modified O’Raifeartaigh model.

For future reference, let me notice the following interesting fact. In all models
we have been considering so far, the existence of (stable) supersymmetry breaking
vacua was always accompaigned by the existence of an R-symmetry in the theory
(think of the original Polonyi model in Example 1, the model in Example 2, the
O’Raifeartaigh model of Example 5 and, to some extent, the model of Example 4).
Its presence, however, does not seem to be a sufficient condition for supersymmetry
breaking: think of the model of Example 3 which does possess an R-symmetry but
does not break supersymmetry. On the contrary, whenever superpotential terms
explicitly breaking the R-symmetry were introduced (the massive Polonyi model of
Example 1 or Example 6), supersymmetric vacua were found. Finally, every time we
found locally stable supersymmetry breaking vacua (again Example 6), in the vicin-
ity of such vacua an approximate R-symmetry, which the theory does not possess
as an exact symmetry, was recovered (essentially, in Examples 6 the superpotential
perturbation responsible for the explicit breaking of the R-symmetry becomes neg-
ligible near the marginally stable supersymmetry breaking vacua). All this suggests

144
some sort of relation between R-symmetry and supersymmetry breaking. We will
discuss this issue later in this lecture, and put such apparent connection on a firm
ground.

7.4 Pseudomoduli space: quantum corrections

In most supersymmetry breaking models we discussed, we found a pseudo-moduli
space of non supersymmetric vacua. Associated to this, we found a massless scalar
mode |X|. While the masslessness of the goldstino and of the Goldstone boson
associated to R-symmetry breaking are protected by symmetries, there are no sym-
metries protecting the pseudo-modulus from getting a mass There isn’t any sym-
metry relating the (degenerate in energy) non supersymmetric vacua. Therefore, by
computing quantum corrections, one might expect this field to get a mass, some-
how. Let us stress the difference with respect to a moduli space of supersymmetric
vacua. Think about the harmonic oscillator. When we quantize the bosonic har-
monic oscillator, the energy of the ground state gets a 12 ~ω contribution. On the
contrary, if the ground state is fermionic, the contribution is the same in modulus
but with opposite sign (fermions tend to push the energy down). In a supersymmet-
ric situation, the mass degeneracy between bosonic and fermionic degrees of freedom
provides equal but opposite contribution to the vacuum energy and the total energy
hence remains zero. In a non-supersymmetric vacuum the degeneracy is not there
anymore (think about the spectrum we computed in the O’Raifeartaigh model) so
one expects things to change. In what follows we will try to make this intuition con-
crete by computing the one-loop Coleman-Weinberg effective potential for both the
O’Raifeartaigh and the modified O’Raifertaigh models. In practice, what we have
to do is to compute corrections in the coupling h at one loop in the background
where the pseudo-modulus |X| has a non-vanishing VEV.
For a supersymmetric theory the Coleman-Weinberg potential reads
 
1 4 M2 1 4 m2B 4 m2F
Veff = STr M log 2 = Tr mB log 2 − Tr mF log 2 , (7.47)
64π 2 Λ 64π 2 Λ Λ

where M = M(|X|) is the full tree level mass matrix, mB and mF correspond to
boson and fermion masses respectively, and Λ is a UV cut-off.
There are a few terms missing in the expression (7.47) of the effective potential,
if compared to a generic non-supersymmetric theory. Let us consider them in turn.

145
First, we miss the cosmological constant term

∼ Λ4 . (7.48)

This term is missing since it only depends on the spectrum, and not on the masses
of the different modes. In a supersymmetric theory the spectrum admits an equal
number of bosonic and fermionic degrees of freedom, no matter whether one is
in a supersymmetric or non supersymmetric vacuum. Since bosons and fermions
contribute opposite to this term, this degeneracy ensures this term to be vanishing.
A second term which is missing is the one proportional to

∼ Λ2 STr M2 . (7.49)

This is not expected to vanish in our supersymmetry breaking vacuum since particles
have different masses. In other words, the mass spectrum is not supersymmetric
along the pseudo-moduli space, recall for instance eqs. (7.40) describing the non-
supersymmetric mass spectrum of the O’Raifeartaigh model. However, an explicit
computation shows that also this term is vanishing. This is not specific to this
model. As we will show later, every time supersymmetry is broken spontaneously
at tree level, provided the Kähler potential is canonical and in the absence of FI
terms, cancellations occur so to give STr M2 = 0.
The only divergent term present in the expression (7.47) is proportional to

∼ log Λ2 STr M4 . (7.50)

This term does not vanish in general but it does not depend on |X|. As such, as we
will see momentarily, it can be reabsorbed in the renormalization of the tree level
vacuum energy |µ2 |2 . The upshot is that the only non-trivial |X|-dependent term
in (7.47) is the finite term
∼ STr M4 log M2 . (7.51)

Let us start considering the O’Raifertaigh model. We should simply plug the tree
level masses (7.40) into formula (7.47). A lengthy but straightforward computation
shows that Veff is a monotonic increasing function of |X| and can hence be expanded
in a power series in |X|2 . For small |X| we get

Veff (|X|) = V0 + m2X |X|2 + O(|X|4 ) (7.52)

where    
2 2 |h|2 |m|2 3 4
V0 = |µ | 1 + log 2 + v(y) + + O(h ) , (7.53)
32π 2 Λ 2

146
with
hµ2 y2
y= < 1 and v(y) = − + O(y 4 ) , (7.54)
m2 12
and
1 h4 µ4 2
m2X = 2 2
z(y) where z(y) = + O(y 2 ) . (7.55)
32π m 3
The minimum of the potential is at |X| = 0, and besides the tree level contribu-
tion |µ2 |2 it gets a contribution proportional ∼ |h2 | which is just a constant, |X|-
independent shift. As anticipated, the UV cut-off dependence can be reabsorbed in
a renormalization of the vacuum energy. Indeed, we can define a running coupling
   
2 2 |h|2 E2 3 4
µ (E) ≡ µbare 1 + log 2 + + O(h ) (7.56)
64π 2 Λ 2
getting  
2 2 |h|2 4
V0 = |µ (E = m)| 1 + v(y) + O(h ) (7.57)
32π 2
and the Λ-dependence has disappeared from the potential.
The upshot of this analysis is that loop corrections have lifted the classical
pseudo-moduli space, leaving just one isolated non supersymmetric vacuum. In this
vacuum the scalar field X gets a (one-loop) mass while ψX , which is the goldstino,
remains massless (notice, in passing, that in the unique supersymmetry breaking
vacuum R-symmetry is preserved, as in Example 2). The shape of the potential in
the X-direction becomes at all similar to that of Figure 7.5.
Let us now see what quantum corrections say about the marginally stable super-
symmetry breaking vacua of the modified O’Raifeartaigh model, the one including
the superpotential perturbation (7.41). We will just briefly sketch the main results.
The interested reader could try to work out all data in detail. One should again eval-
uate the Coleman-Weinberg potential using the tree level spectrum computed near
the origin of field space, where the putative marginally stable vacua live, eqs. (7.44).
Plugging the latter into formula (7.47), what one finds is that, again, the vacuum
degeneracy is lifted and a (locally) stable non-supersymmetric vacuum survives at
|X| = |Xmin | where in our regime,  << 1, Xmin is near the origin and very far,
in field space, from the two supersymmetric vacua sitting at |X| = |XSU SY |. More
precisely we get

Veff (|X|) = V0 + m2X |X − Xmin |2 + O(2 , |X − Xmin |4 ) , (7.58)

where Xmin ∼ m h
f (y) + O(3 ). The spectrum in the supersymmetry breaking
vacuum enjoys a massless fermion, ψX , the goldstino, while the X-field gets a (-

147
independent) mass, as in the original O’Raifeartaigh model. The effective potential,
once projected into X-direction, looks roughly like that in Figure 7.9.

V
<latexit sha1_base64="A1XDWPnWvrr38Qg2Hbjj1OG2Oh0=">AAAB6HicdVDLSgNBEOz1GeMr6tHLYBA8LbPBoN6CXjwmYB6QLGF2MpuMmZ1dZmaFsOQLvHhQxKuf5M2/cTYPUNGChqKqm+6uIBFcG4w/nZXVtfWNzcJWcXtnd2+/dHDY0nGqKGvSWMSqExDNBJesabgRrJMoRqJAsHYwvsn99gNTmsfyzkwS5kdkKHnIKTFWarT6pTJ2vascCLsVbFFdkgryXDxDGRao90sfvUFM04hJQwXRuuvhxPgZUYZTwabFXqpZQuiYDFnXUkkipv1sdugUnVplgMJY2ZIGzdTvExmJtJ5Ege2MiBnp314u/uV1UxNe+hmXSWqYpPNFYSqQiVH+NRpwxagRE0sIVdzeiuiIKEKNzaZoQ1h+iv4nrYrrVV3cOC/XrhdxFOAYTuAMPLiAGtxCHZpAgcEjPMOLc+88Oa/O27x1xVnMHMEPOO9fB++NFw==</latexit>

| |2
<latexit sha1_base64="FnnLb2ttjbBiExMfppXWQbSHpsE=">AAAB8nicbVDLSgMxFM3UV62vqks3wSK4KjNF0WXRjcsK9gHTsWQymTY0kwzJHaG0/Qw3LhRx69e4829M21lo64HA4Zxzyb0nTAU34LrfTmFtfWNzq7hd2tnd2z8oHx61jMo0ZU2qhNKdkBgmuGRN4CBYJ9WMJKFg7XB4O/PbT0wbruQDjFIWJKQvecwpASv5k66w2YhMHmu9csWtunPgVeLlpIJyNHrlr26kaJYwCVQQY3zPTSEYEw2cCjYtdTPDUkKHpM98SyVJmAnG85Wn+MwqEY6Vtk8Cnqu/J8YkMWaUhDaZEBiYZW8m/uf5GcTXwZjLNAMm6eKjOBMYFJ7djyOuGQUxsoRQze2umA6IJhRsSyVbgrd88ipp1areZdW9v6jUb/I6iugEnaJz5KErVEd3qIGaiCKFntErenPAeXHenY9FtODkM8foD5zPHzKVkTE=</latexit>

|XMETA |
<latexit sha1_base64="8gbBUnDVbIazI3GP+hAECRj7/Ss=">AAACAHicbVDLSsNAFJ34rPUVdeHCzWARXJVEFF1WRXAjVOgLmhAm00k7dDIJMxMxpNn4K25cKOLWz3Dn3zhts9DWAxcO59zLvff4MaNSWda3sbC4tLyyWlorr29sbm2bO7stGSUCkyaOWCQ6PpKEUU6aiipGOrEgKPQZafvD67HffiBC0og3VBoTN0R9TgOKkdKSZ+6POl7mhH70mDmK8hTe3TQu83zkmRWrak0A54ldkAooUPfML6cX4SQkXGGGpOzaVqzcDAlFMSN52UkkiREeoj7paspRSKSbTR7I4ZFWejCIhC6u4ET9PZGhUMo09HVniNRAznpj8T+vm6jgws0ojxNFOJ4uChIGVQTHacAeFQQrlmqCsKD6VogHSCCsdGZlHYI9+/I8aZ1U7bOqdX9aqV0VcZTAATgEx8AG56AGbkEdNAEGOXgGr+DNeDJejHfjY9q6YBQze+APjM8fB1uWrQ==</latexit>

|X|
<latexit sha1_base64="gxD/pfcb7DPht99+FiQI281IHDI=">AAAB6nicbVBNS8NAEJ34WetX1aOXxSJ4Kokoeix68VjRfkAbyma7aZduNmF3IpS0P8GLB0W8+ou8+W/ctjlo64OBx3szzMwLEikMuu63s7K6tr6xWdgqbu/s7u2XDg4bJk4143UWy1i3Amq4FIrXUaDkrURzGgWSN4Ph7dRvPnFtRKwecZRwP6J9JULBKFrpYdwad0tlt+LOQJaJl5My5Kh1S1+dXszSiCtkkhrT9twE/YxqFEzySbGTGp5QNqR93rZU0YgbP5udOiGnVumRMNa2FJKZ+nsio5ExoyiwnRHFgVn0puJ/XjvF8NrPhEpS5IrNF4WpJBiT6d+kJzRnKEeWUKaFvZWwAdWUoU2naEPwFl9eJo3zindZce8vytWbPI4CHMMJnIEHV1CFO6hBHRj04Rle4c2Rzovz7nzMW1ecfOYI/sD5/AF6I43s</latexit>

|XSUSY | ⇤
<latexit sha1_base64="9RAU6+c002TkclZMcidHBTCPh7Q=">AAACAHicbVC7TsMwFHXKq5RXgIGBxaJCYqoSBIKxgoWxqKQtaqLIcZ3WquNEtoOI0iz8CgsDCLHyGWz8De5jgJYjXenonHt17z1BwqhUlvVtlJaWV1bXyuuVjc2t7R1zd68l41Rg4uCYxaITIEkY5cRRVDHSSQRBUcBIOxhej/32AxGSxvxOZQnxItTnNKQYKS355sGo4+duFMSPuasoz2DTad4Xxcg3q1bNmgAuEntGqmCGhm9+ub0YpxHhCjMkZde2EuXlSCiKGSkqbipJgvAQ9UlXU44iIr188kABj7XSg2EsdHEFJ+rviRxFUmZRoDsjpAZy3huL/3ndVIWXXk55kirC8XRRmDKoYjhOA/aoIFixTBOEBdW3QjxAAmGlM6voEOz5lxdJ67Rmn9es27Nq/WoWRxkcgiNwAmxwAergBjSAAzAowDN4BW/Gk/FivBsf09aSMZvZB39gfP4ATEeW2g==</latexit>

<latexit sha1_base64="yFrnXXViWirehqLFkWd9XsZYx1w=">AAAB7nicbVBNS8NAFHypX7V+VT16WSyCp5KIoseiFw8eKlhbaEPZbDbt0s0m7L4IJfRHePGgiFd/jzf/jds2B20dWBhm5rHvTZBKYdB1v53Syura+kZ5s7K1vbO7V90/eDRJphlvsUQmuhNQw6VQvIUCJe+kmtM4kLwdjG6mfvuJayMS9YDjlPsxHSgRCUbRSu3enY2GtF+tuXV3BrJMvILUoECzX/3qhQnLYq6QSWpM13NT9HOqUTDJJ5VeZnhK2YgOeNdSRWNu/Hy27oScWCUkUaLtU0hm6u+JnMbGjOPAJmOKQ7PoTcX/vG6G0ZWfC5VmyBWbfxRlkmBCpreTUGjOUI4toUwLuythQ6opQ9tQxZbgLZ68TB7P6t5F3b0/rzWuizrKcATHcAoeXEIDbqEJLWAwgmd4hTcndV6cd+djHi05xcwh/IHz+QMM949h</latexit>

Figure 7.9: The effective potential of the modified O’ Raifeartaigh model project
onto the |X| direction.

One might ask whether such locally stable supersymmetry breaking minimum is
of any physical relevance. An estimate of its lifetime τ can be given looking at the
decay rate
Γ ∼ e−SB (7.59)
(recall that τ ∼ 1/Γ) where SB is the so-called bunch action, the difference between
the Euclidean action of the tunneling configuration and that of remaining in the
metastable vacuum. Its exact form depends on the details of the potential, but
a simple estimate can be given in the so-called thin wall approximation, which is
justified when |XSU SY − XM ax |4 >> VM eta , and which is the case here. In this
approximation the bunch action reads
h∆Xi4
SB ∼ where h∆Xi = hXiSU SY −hXiM eta , ∆V = VM eta −VSU SY = VM eta .
∆V
(7.60)
−α
An explicit computation shows that SB ∼  where α > 0. This implies that
for  << 1 the bunch action can be made arbitrary large so, in this limit, the
locally stable vacuum can be made parametrically long-lived. The upshot of this
analysis is that at the quantum level the classically marginal pseudo-moduli space
of the modified O’Raifertaigh model is lifted but a local, parametrically long-lived
supersymmetry breaking vacuum survives. It is an instructive exercise, which is left
to the reader, to repeat this quantum analysis for the classically marginal pseudo-
moduli space of Example 3. In this case, the pseudo-moduli space gets completely
lifted, and no locally stable supersymmetric minimum survives quantum corrections.

148
 Let us close this section stressing that nothing we said (and computed) about
quantum corrections, pseudomoduli lifting, etc... affects the supersymmetry break-
ing mechanism itself. All models we have been discussing so far, if breaking super-
symmetry, were doing it at tree level. We have not encountered examples where
supersymmetry was unbroken at tree level and one-loop quantum corrections in-
duced supersymmetry breaking. Everything coming from the one-loop potential
can and does modify the classical supersymmetry breaking vacua (which are not
protected by supersymmetry), while it leaves completely unaffected the supersym-
metric ones, if any. This agrees with non-renormalization theorems and our claims
about the robustness of supersymmetric moduli spaces against (perturbative) quan-
tum corrections.
We will have more to say about F-term breaking in due time. Let us pause a
bit now, and consider the other possibility we have alluded to, i.e. spontaneous
supersymmetry breaking induced at tree level by D-terms.

7.5 D-term breaking

In a generic theory, where chiral and vector superfields are present, in absence of
FI terms it is F-term dynamics which governs supersymmetry breaking. This is
because it so happens that whenever one can set all F-terms to zero, using (global)
gauge invariance acting on the scalar fields one can set to zero all D-terms, too. So,
if one wants to consider genuine D-term breaking, one should consider FI terms,
hence abelian gauge factors. In what follows, we will review the most simple such
scenario, where two massive chiral superfields with opposite charge are coupled to
a single U (1) factor, and a FI term is present in the Lagrangian. The Lagrangian
reads
 Z  Z
1

L = Im τ d θ W Wα + d2 θ d2 θ̄ ξV + Φ+ e2eV Φ+ + Φ− e−2eV Φ−
2 α
32π
Z
+ m d2 θ Φ+ Φ− + h.c. , (7.61)

where under a gauge transformation the two chiral superfields transform as Φ± →

e±ieΛ Φ± . The equations of motion for the auxiliary fields read
(
F ± = mφ∓
(7.62)
D = − 21 [ξ + 2e (|φ+ |2 − |φ− |2 )]

149
It is clearly impossible to satisfy all auxiliary fields equations, due to the presence
of the FI parameter ξ. Hence supersymmetry is broken, as anticipated. The scalar
potential reads
1
2

V = ξ + 2e |φ+ |2 − |φ− |2 + m2 |φ+ |2 + |φ− |2
8
   
1 2 1 1
= ξ + m − e ξ |φ− | + m + e ξ |φ+ |2 +
2 2 2
8 2 2
1 2
2
+ e |φ+ |2 − |φ− |2 . (7.63)
2
The vacuum structure and the low energy dynamics depends on the sign of m2 − 12 e ξ.
There is a qualitative difference between the depending on such sign

• m2 > 12 e ξ. All terms in the potential are positive and the minimum of V is
at hφ± i = 0, where V = 18 ξ 2 . Supersymmetry is broken but gauge symmetry
is preserved. The only auxiliary field which gets a non-vanishing VEV is D,
so in this case one speaks of pure D-term breaking. We are in a situation like
the one depicted in the upper right diagram of Figure 7.1.
One can compute the mass spectrum and find agreement with expectations.
The two fermions belonging to the two chiral multiplets have (supersymmetric)
mass m and hence form a massive Dirac fermion. The two scalar fields φ+ and
p p
φ− have masses m2 + 1/2eξ and m2 − 1/2eξ, respectively. Finally, both
the photon Aµ and the photino λ remain massless. The former, because gauge
symmetry is preserved, the latter because supersymmetry is broken and a
massless fermionic mode, the goldstino, is expected (in this case the goldstino
gets contribution from the photino only, since the only non-vanishing auxiliary
field VEV is that of the D-field, ψαG ∼ hDiλα ).

• m2 < 21 e ξ. Now the sign of the mass term for φ− is negative. The minimum
q
ξ 2
of the potential is at hφ+ i = 0 , hφ− i = 2e − me2 ≡ h. Hence both supersym-
metry and gauge symmetry are broken. Both the D-field and F+ get a VEV:
in this case we have a so-called mixed D and F-term breaking. The value of
the potential at its minumum is V = 18 ξ 2 − 12 e2 h4 . We are in a situation of the
type depicted in the lower right diagram of Figure 7.1.
In order to compute the mass spectrum one should expand the potential
around hφ+ i = 0 and hφ− i = h. A lengthy but simple computation gives
√
the following answer. The complex scalar field φ+ has mass mφ+ = 2m. The

150
 √
real part of φ− gets a mass mφR− = 2eh, while the imaginary part disappears
from the spectrum (in fact, it is eaten by the photon, which becomes massive,
√
with mass mAµ = 2eh). The three fermions mix between themselves (there
is a mixing induced from Yukawa couplings). One eigenfunction is massless,
and is nothing but the goldstino ψαG ∼ hDiλα + hF + iψ+α . The other two, ψe± ,
√ p
get equal mass mψe± = 2e2 h2 + m2 = eξ − m2 .

Figure 7.10 gives a summary of the mass spectrum of the FI model as a function
of 21 eξ which, at fixed m, is the order parameter of the supersymmetry breaking
transition.

2m2
<latexit sha1_base64="efKUm+HjJM16/+LP4mEk06d1uDE=">AAAB7HicbVBNS8NAEJ3Ur1q/qh69LBbBU0mKoseiF48VTFtoY9lsN+3S3U3Y3Qgl9Dd48aCIV3+QN/+N2zQHbX0w8Hhvhpl5YcKZNq777ZTW1jc2t8rblZ3dvf2D6uFRW8epItQnMY9VN8Saciapb5jhtJsoikXIaSec3M79zhNVmsXywUwTGgg8kixiBBsr+Q0kHhuDas2tuznQKvEKUoMCrUH1qz+MSSqoNIRjrXuem5ggw8owwums0k81TTCZ4BHtWSqxoDrI8mNn6MwqQxTFypY0KFd/T2RYaD0Voe0U2Iz1sjcX//N6qYmug4zJJDVUksWiKOXIxGj+ORoyRYnhU0swUczeisgYK0yMzadiQ/CWX14l7Ubdu6y79xe15k0RRxlO4BTOwYMraMIdtMAHAgye4RXeHOm8OO/Ox6K15BQzx/AHzucPw4GN/w==</latexit>

<latexit sha1_base64="ap2Q3yc/8MgBeEgdTQCpa/FKTS4=">AAAB7XicbVDLSgNBEOyNrxhfUY9eBoMgCGFXFD0GvXiMYB6QLGF2MpuMmccyMyuEJf/gxYMiXv0fb/6Nk2QPmljQUFR1090VJZwZ6/vfXmFldW19o7hZ2tre2d0r7x80jUo1oQ2iuNLtCBvKmaQNyyyn7URTLCJOW9Hoduq3nqg2TMkHO05oKPBAspgRbJ3U7CZD1jvrlSt+1Z8BLZMgJxXIUe+Vv7p9RVJBpSUcG9MJ/MSGGdaWEU4npW5qaILJCA9ox1GJBTVhNrt2gk6c0kex0q6kRTP190SGhTFjEblOge3QLHpT8T+vk9r4OsyYTFJLJZkvilOOrELT11GfaUosHzuCiWbuVkSGWGNiXUAlF0Kw+PIyaZ5Xg8uqf39Rqd3kcRThCI7hFAK4ghrcQR0aQOARnuEV3jzlvXjv3se8teDlM4fwB97nDzOAjuE=</latexit>

+
e±
<latexit sha1_base64="IKVwYvZcTdy/kklMl7uKp2O2rhY=">AAAB/HicbVBNS8NAEN34WetXtEcvi0XwVBJR9Fj04rGC/YAmhM1m0i7dbMLuRimh/hUvHhTx6g/x5r9x2+agrQ8GHu/NMDMvzDhT2nG+rZXVtfWNzcpWdXtnd2/fPjjsqDSXFNo05anshUQBZwLammkOvUwCSUIO3XB0M/W7DyAVS8W9HmfgJ2QgWMwo0UYK7Jr3yCLQjEeAvUyxwMuSwK47DWcGvEzcktRRiVZgf3lRSvMEhKacKNV3nUz7BZGaUQ6TqpcryAgdkQH0DRUkAeUXs+Mn+MQoEY5TaUpoPFN/TxQkUWqchKYzIXqoFr2p+J/Xz3V85RdMZLkGQeeL4pxjneJpEjhiEqjmY0MIlczciumQSEK1yatqQnAXX14mnbOGe9Fw7s7rzesyjgo6QsfoFLnoEjXRLWqhNqJojJ7RK3qznqwX6936mLeuWOVMDf2B9fkD18qU5A==</latexit>

m2
<latexit sha1_base64="t8U5zgwtMmVwam7Xrz1zSHyF+OM=">AAAB6nicbVDLSgNBEOz1GeMr6tHLYBA8hd2g6DHoxWNE84BkDbOT2WTIPJaZWSEs+QQvHhTx6hd582+cJHvQxIKGoqqb7q4o4cxY3//2VlbX1jc2C1vF7Z3dvf3SwWHTqFQT2iCKK92OsKGcSdqwzHLaTjTFIuK0FY1upn7riWrDlHyw44SGAg8kixnB1kn34rHaK5X9ij8DWiZBTsqQo94rfXX7iqSCSks4NqYT+IkNM6wtI5xOit3U0ASTER7QjqMSC2rCbHbqBJ06pY9ipV1Ji2bq74kMC2PGInKdAtuhWfSm4n9eJ7XxVZgxmaSWSjJfFKccWYWmf6M+05RYPnYEE83crYgMscbEunSKLoRg8eVl0qxWgouKf3derl3ncRTgGE7gDAK4hBrcQh0aQGAAz/AKbx73Xrx372PeuuLlM0fwB97nD/wwjZk=</latexit>

R
<latexit sha1_base64="SfjvK2KScFv/mziuJBKbhBLwNcE=">AAAB73icbVBNSwMxEJ2tX7V+VT16CRbBU9kVRY9FLx4r2FroLiWbZtvQJBuTrFCW/gkvHhTx6t/x5r8xbfegrQ8GHu/NMDMvVpwZ6/vfXmlldW19o7xZ2dre2d2r7h+0TZppQlsk5anuxNhQziRtWWY57ShNsYg5fYhHN1P/4Ylqw1J5b8eKRgIPJEsYwdZJnVAZ1guV6FVrft2fAS2ToCA1KNDsVb/CfkoyQaUlHBvTDXxloxxrywink0qYGaowGeEB7ToqsaAmymf3TtCJU/ooSbUradFM/T2RY2HMWMSuU2A7NIveVPzP62Y2uYpyJlVmqSTzRUnGkU3R9HnUZ5oSy8eOYKKZuxWRIdaYWBdRxYUQLL68TNpn9eCi7t+d1xrXRRxlOIJjOIUALqEBt9CEFhDg8Ayv8OY9ei/eu/cxby15xcwh/IH3+QMst5AO</latexit>

<latexit sha1_base64="VNDVO77nSVesVQuIGkNoExd08cU=">AAACCHicbVC7SgNBFJ2Nrxhfq5YWDgbBQsOuKFpGbSyjmAdk4zI7mSRD5rHMzIph2dLGX7GxUMTWT7Dzb5w8Co0euHA4517uvSeKGdXG876c3Mzs3PxCfrGwtLyyuuaub9S0TBQmVSyZVI0IacKoIFVDDSONWBHEI0bqUf9i6NfviNJUihsziEmLo66gHYqRsVLobp+FAU/gPgziHg0PbtOAR/I+DQwVA3idZaFb9EreCPAv8SekCCaohO5n0JY44UQYzJDWTd+LTStFylDMSFYIEk1ihPuoS5qWCsSJbqWjRzK4a5U27EhlSxg4Un9OpIhrPeCR7eTI9PS0NxT/85qJ6Zy2UirixBCBx4s6CYNGwmEqsE0VwYYNLEFYUXsrxD2kEDY2u4INwZ9++S+pHZb845J3dVQsn0/iyIMtsAP2gA9OQBlcggqoAgwewBN4Aa/Oo/PsvDnv49acM5nZBL/gfHwD6H+ZSg==</latexit>

± Aµ ,

<latexit sha1_base64="DBU5+KiqsU8MdTaspDjv2ubsBRU=">AAAB7XicbVDLSgNBEOyNrxhfUY9eBoPgxbArih6DXjxGMA9IljA7mU3GzGOZmRXCkn/w4kERr/6PN//GSbIHTSxoKKq66e6KEs6M9f1vr7Cyura+UdwsbW3v7O6V9w+aRqWa0AZRXOl2hA3lTNKGZZbTdqIpFhGnrWh0O/VbT1QbpuSDHSc0FHggWcwItk5qdpMh6531yhW/6s+AlkmQkwrkqPfKX92+Iqmg0hKOjekEfmLDDGvLCKeTUjc1NMFkhAe046jEgpowm107QSdO6aNYaVfSopn6eyLDwpixiFynwHZoFr2p+J/XSW18HWZMJqmlkswXxSlHVqHp66jPNCWWjx3BRDN3KyJDrDGxLqCSCyFYfHmZNM+rwWXVv7+o1G7yOIpwBMdwCgFcQQ3uoA4NIPAIz/AKb57yXrx372PeWvDymUP4A+/zBzaIjuM=</latexit>

1
<latexit sha1_base64="NoDlZorBTpq3uyfeNhDxyCn+KZY=">AAAB+nicbVBNS8NAEJ34WetXqkcvi0XwVJKi6LHoxWMF+wFNKJvtpl262YTdjVpifooXD4p49Zd489+4bXPQ1gcDj/dmmJkXJJwp7Tjf1srq2vrGZmmrvL2zu7dvVw7aKk4loS0S81h2A6woZ4K2NNOcdhNJcRRw2gnG11O/c0+lYrG405OE+hEeChYygrWR+nbFCyUmmZtn9RxR5D2yvl11as4MaJm4BalCgWbf/vIGMUkjKjThWKme6yTaz7DUjHCal71U0QSTMR7SnqECR1T52ez0HJ0YZYDCWJoSGs3U3xMZjpSaRIHpjLAeqUVvKv7n9VIdXvoZE0mqqSDzRWHKkY7RNAc0YJISzSeGYCKZuRWRETZZaJNW2YTgLr68TNr1mntec27Pqo2rIo4SHMExnIILF9CAG2hCCwg8wDO8wpv1ZL1Y79bHvHXFKmYO4Q+szx+IvZOG</latexit>

2 e⇠
<latexit sha1_base64="t8U5zgwtMmVwam7Xrz1zSHyF+OM=">AAAB6nicbVDLSgNBEOz1GeMr6tHLYBA8hd2g6DHoxWNE84BkDbOT2WTIPJaZWSEs+QQvHhTx6hd582+cJHvQxIKGoqqb7q4o4cxY3//2VlbX1jc2C1vF7Z3dvf3SwWHTqFQT2iCKK92OsKGcSdqwzHLaTjTFIuK0FY1upn7riWrDlHyw44SGAg8kixnB1kn34rHaK5X9ij8DWiZBTsqQo94rfXX7iqSCSks4NqYT+IkNM6wtI5xOit3U0ASTER7QjqMSC2rCbHbqBJ06pY9ipV1Ji2bq74kMC2PGInKdAtuhWfSm4n9eJ7XxVZgxmaSWSjJfFKccWYWmf6M+05RYPnYEE83crYgMscbEunSKLoRg8eVl0qxWgouKf3derl3ncRTgGE7gDAK4hBrcQh0aQGAAz/AKbx73Xrx372PeuuLlM0fwB97nD/wwjZk=</latexit>
<latexit sha1_base64="wTY4pRAErXIPxvk22dWORW2Xuho=">AAAB/nicbVDLSsNAFJ34rPUVFVduBovgqiSi6LLoQpcV7AOaWCbTSTt0HmFmIoYQ8FfcuFDErd/hzr9x+lho64ELh3Pu5d57ooRRbTzv21lYXFpeWS2tldc3Nre23Z3dppapwqSBJZOqHSFNGBWkYahhpJ0ognjESCsaXo381gNRmkpxZ7KEhBz1BY0pRsZKXXc/SDS9zwMeycc8MFRk8Looum7Fq3pjwHniT0kFTFHvul9BT+KUE2EwQ1p3fC8xYY6UoZiRohykmiQID1GfdCwViBMd5uPzC3hklR6MpbIlDByrvydyxLXOeGQ7OTIDPeuNxP+8TmriizCnIkkNEXiyKE4ZNBKOsoA9qgg2LLMEYUXtrRAPkELY2MTKNgR/9uV50jyp+mdV7/a0UrucxlECB+AQHAMfnIMauAF10AAY5OAZvII358l5cd6dj0nrgjOd2QN/4Hz+AN2elhA=</latexit>

G
Aµ ,
<latexit sha1_base64="AKvvBporU/PW+meDpNg+OnyDHM8=">AAAB+HicbVDLSgMxFM3UV62PVl26CRbBhZQZUXRZdeOygn1AZxgymUwbmmSGPIQ69EvcuFDErZ/izr8xbWehrQcCh3Pu4d6cKGNUadf9dkorq2vrG+XNytb2zm61trffUamRmLRxylLZi5AijArS1lQz0sskQTxipBuNbqd+95FIRVPxoMcZCTgaCJpQjLSVwlr1OvS5gafQZzYUo7BWdxvuDHCZeAWpgwKtsPblxyk2nAiNGVKq77mZDnIkNcWMTCq+USRDeIQGpG+pQJyoIJ8dPoHHVolhkkr7hIYz9XciR1ypMY/sJEd6qBa9qfif1zc6uQpyKjKjicDzRYlhUKdw2gKMqSRYs7ElCEtqb4V4iCTC2nZVsSV4i19eJp2zhnfRcO/P682boo4yOARH4AR44BI0wR1ogTbAwIBn8ArenCfnxXl3PuajJafIHIA/cD5/AG9Jkkw=</latexit>

m 2

Figure 7.10: The mass spectrum of the Fayet-Iliopoulos model as a function of the
FI parameter ξ.

One of the most attractive features of supersymmetric fields theories is the stabil-
ity of masses under quantum corrections. In models where the FI mechanism plays
a role, the physical mass spectrum depends on ξ which is not protected, a priori,
since it appears in a D-term. This is different from models where F-terms are re-
sponsible for the supersymmetry breaking dynamics, since these are superpotential
terms and protected by non-renormalization theorems. Therefore, it is important to
investigate the circumstances under which the FI term does not get renormalized.
It can be shown that the contribution renormalizing the FI term is proportional to
the trace of the U (1) generator taken over all chiral superfields present in the model.
This trace is proportional to the gravitational anomaly. Therefore, we can conclude
that the FI term does not renormalize for theories free of gravitational anomalies.

151
7.6 Indirect criteria for supersymmetry breaking
We have already alluded to some possible relation between supersymmetry breaking
and R-symmetry. In what follows, we will try to make this intuition precise and,
more generally, present a few general criteria one can use to understand whether a
theory might or might not break supersymmetry, without having a precise knowl-
edge of the details of the theory itself. These criteria might be useful as guiding
principles when trying to construct models of supersymmetry breaking in a bottom-
up approach and, at the same time, they allow to have a handle on theories which
are more involved than the simple ones we analyzed in previous sections. Finally,
having some general criteria, possibly being valid also beyond the realm of pertur-
bative physics might also be useful when one has to deal with theories in strongly
coupled phases, where a perturbative, semi-classical approach is not possible, and
where the direct study of the zero’s of the potential is not easy or even not possible.

7.6.1 Supersymmetry breaking and global symmetries

Let us consider a supersymmetric theory which has a spontaneously broken global

symmetry and which does not admit (non compact) classical flat directions. This
theory, generically, breaks supersymmetry. This can be easily proven as follows.
Since there is a broken global symmetry, the theory admits a goldstone boson (a
massless particle with no potential). If supersymmetry were unbroken then one
should expect a scalar companion of this goldstone boson, which, being in the same
multiplet of the latter, would not admit a potential either. But then, the theory
would admit a flat direction, contrary to one of the hypotheses. This is a sufficient
condition for supersymmetry breaking.
In the above reasoning we have assumed that the second massless scalar corre-
sponds to a non-compact flat direction. This is typically the case since the Goldstone
boson is the phase of the order parameter, and its scalar companion corresponds to
a dilation of the order parameter, and therefore represents indeed a non-compact
flat direction.
Consider now a theory of F chiral superfields Φi with superpotential W . Super-
symmetry is unbroken if
∂W
Fi = =0 , ∀i = 1, 2, . . . , F . (7.64)
∂φi
These are F holomorphic conditions on F complex variables. Therefore, if the su-

152
perpotenital W is generic, one expects (typically distinct) solutions to exist. Hence,
supersymmetry is unbroken. By the superpotential being generic we mean the fol-
lowing. The superpotential is, in general, a function of the Φi ’s of degree, say, n.
It is generic if all possible polynomials of degree n or lower compatible with the
symmetries of the theory are present.
Suppose now that W preserves some global non-R symmetry. Hence, W is a
function of singlet combinations of the Φi ’s. It is easy to see that in terms of these
reduced number of variables, eqs. (7.64) impose an equal number of independent
conditions. Suppose, for definiteness, that the global symmetry is a U (1) symmetry
and call qi the corresponding charge of the i-th chiral superfield Φi . Hence, we can
rewrite the superpotential as e.g.
−qi /q1
W = W (Xi ) where Xi = Φi Φ1 , i = 2, 3, . . . , F . (7.65)

If we now consider eqs. (7.64) we have

( −q /q
j 6= 1 ∂W
∂φj
= φ1 j 1 ∂W∂X(Xj i ) = 0
∂W ∂W (Xi ) ∂Xk
(7.66)
j=1 ∂φ1
= ∂Xk ∂φ1
=0 , k = 2, . . . , F .

We see that the equation for F 1 is automatically satisfied if the others F − 1 are
satisfied. Hence, having a system of F − 1 holomorphic equations in F − 1 variables,
generically the system allows for solutions. The same reasoning holds for a generic
global symmetry. A global symmetry (under which the superpotential is a singlet)
diminishes the number of independent variables, but it diminishes also the number
of independent F-equations by the same amount. Hence, again, if W is generic,
eqs. (7.64) can be solved and supersymmetry is unbroken.
Suppose now that the global symmetry under consideration is a R-symmetry.
The crucial difference here is that the superpotential is charged under this symmetry,
R(W ) = 2. Let us call ri the R-charge of the i-th superfield Φi . We can now rewrite
the superpotential as
2/r1 −ri /r1
W = Φ1 f (Xi ) where Xi = Φi Φ1 , i = 2, 3, . . . , F . (7.67)

If we now compute eqs. (7.64) we get

 2−rj
j 6= 1 ∂W = φ r1 ∂f (Xi ) = 0
∂φj 1 ∂Xj
2 2 (7.68)
 −1
j = 1 ∂φ1 = r1 φ1 f (Xi ) + φ1r1 ∂f∂X
∂W 2 r1 (Xi ) ∂Xk
k ∂φ1
=0.

153
Once the first F − 1 equations are satisfied, the remaining one reduces to f (Xi ) = 0,
which is not at all trivial. So now we have F independent equations in F −1 variables
and hence, generically, solutions do not exist. So we conclude that supersymmetry
is broken, generically. The upshot is that the existence of an R-symmetry is a
necessary condition for supersymmetry breaking, if the potential is generic (and, if
it is then spontaneously broken, it is a sufficient condition if there are no classical
flat directions).
This is known as the Nelson-Seiberg criterium. The O’Raifeartaigh model meets
this criterium. It possesses an R-symmetry (which is then spontaneously broken
along the pseudo-moduli space), the superpotential is generic, and it breaks su-
persymmetry. The modified O’Raifeartaigh model instead admits supersymmetry
preserving vacua. Indeed, the R-symmetry is absent since the mass perturbation
∆W breaks it explicitly. So, one would expect the model not to break supersym-
metry. And in fact it doesn’t. We have noticed, though, that somewhere else in
the space of scalar field VEVs this model admits non-supersymmetric vacua which,
if the mass perturbation is small enough, we have proven to be long-lived. In this
region the R-breaking perturbation is negligible and an approximate R-symmetry
(the O’Raifeartaigh model’s original one) is recovered. This property is in fact not
specific to the modified O’Raifeartaigh model, but is a generic feature of supersym-
metry breaking metastable vacua.
Summarizing, a rough guideline in the quest for supersymmetry breaking theories
can be as follows:
No R-symmetry −→ SUSY unbroken
R-symmetry −→ SUSY (maybe) broken
Approximate R-symmetry −→ SUSY (maybe) broken locally,
restored elsewhere

Since necessary conditions are quite powerful tools, let me stress again one im-
portant point. The existence of an R-symmetry is a necessary condition for super-
symmetry breaking under the assumption that the superpotential is generic. If this
is not the case, supersymmetry can be broken even if the R-symmetry is absent. An-
other possibility, which typically occurs when gauge degrees of freedom are present
in the Lagrangian, is that R-symmetry is absent, but then it arises as an accidental
symmetry in the low energy effective theory. Also in this case supersymmetry can be
broken even if R-symmetry was absent in the UV Lagrangian. We will see examples
of this sort later in this course.

154
7.6.2 Topological constraints: the Witten Index

Another powerful criterium exists which helps when dealing with theories with com-
plicated vacuum structure and for which it is then difficult to determine directly
whether supersymmetry is broken, i.e. to find the zero’s of the potential. This cri-
terium, which provides a necessary condition for supersymmetry breaking, has to do
with the so-called Witten index, which, for a supersymmetric theory, is a topological
invariant quantity.
The Witten index, let us dub it IW , is an integer number which measures the
difference between the number of bosonic and fermionic states, for any given en-
ergy level. In a supersymmetric theory, for any positive energy level there is an
equal number of bosonic and fermionic states. This is obvious if supersymmetry is
unbroken, but it also holds if supersymmetry is broken: every state is degenerate
with the state obtained from it by adding a zero-momentum goldstino (which is
certainly there, if supersymmetry is broken). In other words, a state |Ωi is paired
with |Ω + {pµ = 0 goldstino}i. On the contrary, zero energy states can be un-
paired since, due to the supersymmetry algebra, such states are annihilated by the
supercharges. Therefore, in a supersymmetric theory the Witten index can get con-
tribution from the zero energy states only, regardless the vacuum one is considering
preserves supersymmetry or it does not.
Strictly speaking the above argument holds only if we put the theory in a finite
volume. In an infinite volume, when supersymmetry is broken one has to deal with
IR singularity issues. In particular, the (broken) supercharge diverges and acting
with it on a physical state gives a non-normalizable state. This can be seen as
follows. From the current algebra (4.71) one sees that
1 µ α̇α ν
E η µν = hT µν i = σ̄ h{Qα , S α̇ }i . (7.69)
4
This shows that if the vacuum energy is non vanishing, the vacuum is transformed
into one goldstino states if acted with Qα (recall that the supercurrent creates a gold-
stino when acting on the vacuum). In an infinite volume a non-vanishing vacuum
energy density corresponds to an infinite total energy, implying that the supercharge
diverges and that the zero-momentum goldstino state is not defined (the correspond-
ing state does not exist in the Hilbert space). Putting the theory in a finite volume
is a way to regularize (translational invariance can be maintained imposing periodic
boundary conditions on all fields). Therefore, in what follows we will start consid-
ering the theory in a finite volume V and only later take the infinite volume limit.

155
As we will see, what is relevant for the argument we want to convey is not affected
by these issues and holds true also when V → ∞.
A theory in a finite volume has a discrete energy spectrum, all states in the
Hilbert space are discrete and normalizable and can be counted unambiguously.
Our goal is to compute the Witten index in such finite volume theory. To this
end, we can restrict to the zero-momentum subspace of the Hilbert space. In a
supersymmetric theory the energy of any state is semi-positive definite hence, from
the relativistic equation m2 = E 2 − |~p|2 , it follows that the energy is larger or equal
than the momentum, so zero energy states have p~ = 0. Setting |~p| = 0 we are
excluding from the counting massless states with E > 0. These states, though, do
not contribute to the Witten index. Massive states, on the contrary, never contribute
to it since they necessarily have E > 0, regardless the value of |~p|. So only massless
states with |~p| = 0 contribute to the Witten index. The upshot is that restricting
the Hilbert space to the subspace |~p| = 0 does not hart and so this is what we will
do in what follows. Nicely, in such subspace the supersymmetry algebra simplifies.
In particular, using four-component spinor notation in which the supercharge Q is
a Majorana spinor, the supersymmetry algebra in the subspace |~p| = 0 is just

{Q, Q} = 2γ 0 P0 . (7.70)

This implies that Q21 = Q22 = Q23 = Q24 = H, where H is the Hamiltonian of the
system and Qi are the four components of Q.
Suppose to have a bosonic state |bi for which Q2 |bi = E|bi, where Q is one of
the Qi ’s. Then the fermionic state obtained from |bi as
1
|f i ≡ √ Q|bi , (7.71)
E
has also energy E. This does not apply to zero-energy states since they are annihi-
lated by Q, and hence are not paired. So, as anticipated, the Witten index receives
contributions only from zero energy states. In Figure 7.11 we report the general
form of the spectrum of a supersymmetric theory.
In order to appreciate its topological nature, let us define the index a bit more
rigorously. A supersymmetric theory is a unitary representation of the Poincaré
superalgebra on some Hilbert space H. Let us assume that

H = ⊕ HE . (7.72)
E≥0

156
 Energy

Figure 7.11: The spectrum of a supersymmetric theory in a finite volume. Circles

indicate bosons, squares indicate fermions. The zero energy level is the only one
where there can exist a different number of circles and squares.

The Witten index is defined as

IW (β) = STrH eβH ≡ TrH (−1)F eβH , β ∈ R+ . (7.73)

It follows that
X X
IW (β) = eβE TrHE (−1)F = eβE [nB (E) − nF (E)] =
E≥0 E≥0

= nB (0) − nF (0) = TrH0 (−1)F = IW (0) . (7.74)

We have been rewriting what we have already shown to hold. The point is that this
way it is clear that the index does not depend on β: its value does not vary if we
vary β. More generally, one can prove that the Witten index does not depend on
any parameter, like in particular coupling constants, and can then be computed in
appropriate corners of the parameter space (say at weak coupling) and the result
one gets is exact. In other words, the Witten index is a topological invariant.
Suppose one starts from a situation like the one depicted in Figure 7.11. Varying
the parameters of the theory, like masses, couplings, etc..., it may very well be that
some states move around in energy. The point is that they must do it in pairs, in a
supersymmetric theory. Hence, it can happen that a pair of non-zero energy states
moves down to zero energy; or, viceversa, that some zero energy states may acquire
non-zero energy. But again, this can only happen if an equal number of bosonic and
fermionic zero energy states moves towards a non-zero energy level. The upshot is
that the Witten index does not change. This is summarized pictorially in Figure
7.12.

157
 Energy Energy Energy

IW = 1 IW = 1
<latexit sha1_base64="x4IkQN+41DXabsLrJ0qllppOuM4=">AAAB7HicdVBNSwMxEJ31s9avqkcvwSJ4WrLFoh6Eohe9VXDbQruUbJptQ7PZJckKpfQ3ePGgiFd/kDf/jdl+gIo+GHi8N8PMvDAVXBuMP52l5ZXVtfXCRnFza3tnt7S339BJpijzaSIS1QqJZoJL5htuBGulipE4FKwZDq9zv/nAlOaJvDejlAUx6UsecUqMlfzbbvPS65bK2PUuciDsVrBFdUEqyHPxFGWYo94tfXR6Cc1iJg0VROu2h1MTjIkynAo2KXYyzVJCh6TP2pZKEjMdjKfHTtCxVXooSpQtadBU/T4xJrHWozi0nTExA/3by8W/vHZmovNgzGWaGSbpbFGUCWQSlH+OelwxasTIEkIVt7ciOiCKUGPzKdoQFp+i/0mj4npVF9+dlmtX8zgKcAhHcAIenEENbqAOPlDg8AjP8OJI58l5dd5mrUvOfOYAfsB5/wJFsY5W</latexit> <latexit sha1_base64="x4IkQN+41DXabsLrJ0qllppOuM4=">AAAB7HicdVBNSwMxEJ31s9avqkcvwSJ4WrLFoh6Eohe9VXDbQruUbJptQ7PZJckKpfQ3ePGgiFd/kDf/jdl+gIo+GHi8N8PMvDAVXBuMP52l5ZXVtfXCRnFza3tnt7S339BJpijzaSIS1QqJZoJL5htuBGulipE4FKwZDq9zv/nAlOaJvDejlAUx6UsecUqMlfzbbvPS65bK2PUuciDsVrBFdUEqyHPxFGWYo94tfXR6Cc1iJg0VROu2h1MTjIkynAo2KXYyzVJCh6TP2pZKEjMdjKfHTtCxVXooSpQtadBU/T4xJrHWozi0nTExA/3by8W/vHZmovNgzGWaGSbpbFGUCWQSlH+OelwxasTIEkIVt7ciOiCKUGPzKdoQFp+i/0mj4npVF9+dlmtX8zgKcAhHcAIenEENbqAOPlDg8AjP8OJI58l5dd5mrUvOfOYAfsB5/wJFsY5W</latexit>

Energy Energy Energy

IW = 1 IW = 1
<latexit sha1_base64="x4IkQN+41DXabsLrJ0qllppOuM4=">AAAB7HicdVBNSwMxEJ31s9avqkcvwSJ4WrLFoh6Eohe9VXDbQruUbJptQ7PZJckKpfQ3ePGgiFd/kDf/jdl+gIo+GHi8N8PMvDAVXBuMP52l5ZXVtfXCRnFza3tnt7S339BJpijzaSIS1QqJZoJL5htuBGulipE4FKwZDq9zv/nAlOaJvDejlAUx6UsecUqMlfzbbvPS65bK2PUuciDsVrBFdUEqyHPxFGWYo94tfXR6Cc1iJg0VROu2h1MTjIkynAo2KXYyzVJCh6TP2pZKEjMdjKfHTtCxVXooSpQtadBU/T4xJrHWozi0nTExA/3by8W/vHZmovNgzGWaGSbpbFGUCWQSlH+OelwxasTIEkIVt7ciOiCKUGPzKdoQFp+i/0mj4npVF9+dlmtX8zgKcAhHcAIenEENbqAOPlDg8AjP8OJI58l5dd5mrUvOfOYAfsB5/wJFsY5W</latexit> <latexit sha1_base64="x4IkQN+41DXabsLrJ0qllppOuM4=">AAAB7HicdVBNSwMxEJ31s9avqkcvwSJ4WrLFoh6Eohe9VXDbQruUbJptQ7PZJckKpfQ3ePGgiFd/kDf/jdl+gIo+GHi8N8PMvDAVXBuMP52l5ZXVtfXCRnFza3tnt7S339BJpijzaSIS1QqJZoJL5htuBGulipE4FKwZDq9zv/nAlOaJvDejlAUx6UsecUqMlfzbbvPS65bK2PUuciDsVrBFdUEqyHPxFGWYo94tfXR6Cc1iJg0VROu2h1MTjIkynAo2KXYyzVJCh6TP2pZKEjMdjKfHTtCxVXooSpQtadBU/T4xJrHWozi0nTExA/3by8W/vHZmovNgzGWaGSbpbFGUCWQSlH+OelwxasTIEkIVt7ciOiCKUGPzKdoQFp+i/0mj4npVF9+dlmtX8zgKcAhHcAIenEENbqAOPlDg8AjP8OJI58l5dd5mrUvOfOYAfsB5/wJFsY5W</latexit>

Figure 7.12: Supersymmetric theory dynamics. Upon modifications of parameters

of the theory, the number of zero energy states can change, but the Witten index
remains the same.

What is this useful for? The crucial point is that the Witten index measures
the difference between zero-energy states only. Suppose it is different from zero,
IW 6= 0. This means that there exists some zero-energy state, hence supersymmetry
is unbroken. But, because of the topological nature of IW , this conclusion holds
at any order in perturbation theory and even non-perturbatively! A theory with
non vanishing Witten index cannot break supersymmetry. Suppose instead that
IW = 0. Now one cannot conclude anything, just that the number of bosonic and
fermionic zero-energy state is the same; but one cannot tell whether this number is
zero (broken supersymmetry) or different from zero (unbroken supersymmetry).
So we conclude that having a non-vanishing Witten index is a sufficient condition
for the existence of supersymmetric vacua, and that having it vanishing is a necessary
condition for supersymmetry breaking. And a robust one, since IW is an exact
quantity.
Few comments are in order at this point.

158
 First, the fact that we have been working in a finite volume does not question
our main conclusions. If supersymmetry is unbroken in an arbitrary finite volume
it means the ground state energy E(V ) is zero for any V. Since the large-V limit
of zero is still zero, supersymmetry is unbroken also in the infinite volume limit. If
one can explicitly compute the Witten index at finite volume and find that it is not
vanishing, one can safely conclude that supersymmetry is not broken even in the
actual theory, i.e. at infinite volume. On the contrary, the converse is not necessarily
true. It might be that supersymmetry is broken at finite volume and restored in the
infinite volume limit. Suppose that IW = 0 and that one knows that supersymmetry
is broken, that is the minimal energy states have positive energy. The energy density
goes as E(V )/V and it may very well be that for V → ∞ the increase of E is not
enough to compensate for the larger and larger volume. So the energy density can
very well become zero in the infinite volume limit and supersymmetry restored. But
this does not hurt much, since all what the vanishing of the Witten index provides
is a necessary condition for supersymmetry breaking, not a sufficient one.
A second comment regards the relation between classical and quantum results.
Suppose one can explicitly check at tree level that a given theory has non-vanishing
Witten index, IW 6= 0. For what we said above, this implies that supersymmetry is
unbroken classically and that it cannot be broken whatsoever, neither perturbatively
nor non-perturbatively. On the contrary, if IW = 0 at tree level and we know that
supersymmetry is unbroken classically, it can very well be that (non-perturbative)
quantum effects may break it.
The theorem we have discussed may find very useful applications. For one thing,
it turns out that pure SYM theories have non-vanishing Witten index (for SYM with
gauge group G, the index equals the dual Coxeter number of G, which for SU and Sp
groups is just r + 1, where r is the rank of G). So pure SYM theories cannot break
supersymmetry. As a corollary, SYM theories with massive matter (like massive
SQCD) cannot break supersymmetry either. This is because for low enough energy
all massive fields can be integrated out and the theory flows to pure SYM, which
has non-vanishing index. More generally, non-chiral theories, for which a mass term
can be given to all matter fields, are not expected to break supersymmetry.
What about chiral theories, instead? Chiral theories behave differently. In this
case some chiral superfield cannot get a mass anyway, and so these theories cannot
be obtained from deformation of vector-like theories, as massive SQCD. Hence,
one cannot conclude that these theories cannot break supersymmetry. As we will

159
see when discussing models of dynamical supersymmetry breaking, most known
examples of theories breaking supersymmetry are, in fact, chiral theories.
There is a subtlety in all what we said so far, which is sort of hidden in some of
our claims. We said that the Witten index is robust against any continuous change
of parameters. But it turns out that a perturbation that changes the asymptotic
behavior of the potential may induce a change in IW . This is related to the topolog-
ical nature of the index, which makes it depending on boundary effects. Consider
the following simple potential for a scalar field φ,

2
V (φ) = mφ − gφ2 . (7.75)

For g = 0 low energy states correspond to φ ∼ 0. For g 6= 0 low energy states may
correspond to φ = 0 but also to φ ∼ m/g (no matter how small g is). So we see here
that IW (g = 0) 6= IW (g 6= 0). What is going on? The point is that switching on and
off g changes the asymptotic behavior of V for large φ (that is, at the boundary of
field space): in the large φ region, for g = 0 V ∼ φ2 while for g 6= 0 V ∼ φ4 . The
punchline is that the Witten index is invariant under any change in the parameters
of a theory in which, in the large field regime, the potential changes by terms no
bigger than the terms already present. If this is not the case, the Witten index
can change discontinuously. In other words, IW is independent of numerical values
of parameters as long as these are non-zero. When sending a set of parameters
to zero, or switching on some couplings which were absent, one should check that
the asymptotic behavior of the potential is unchanged, in order to avoid new states
coming in from (or going out to) infinity. Coming back to our SQCD example, we see
that for massive SQCD the potential in the large field regime is quadratic, while for
massless SQCD is flat: the two theories do not have a priori the same Witten index.
Therefore, while massive SQCD is expected to be in the same equivalence class of
pure SYM (as far as supersymmetry breaking is concerned), this is not guaranteed
for massless SQCD. In other words, no conclusions can be drawn for the massless
regime by the analysis in the massive regime. We will see explicit examples of this
phenomenon in later lectures.
Let us finally notice that Witten index argument limits a lot the landscape of
possible supersymmetry breaking theories; for instance, non-chiral gauge theories
most likely cannot break supersymmetry.

160
7.6.3 Genericity and metastability

Before concluding this section, there is yet another important conclusion we can
draw from all what we have learned. Both the Nelson-Seiberg criterium and Witten
index argument seem to favor, at least statistically, supersymmetry breaking into
metastable vacua.
For one thing, thinking about R-symmetry one might have the impression to
fall into a vicious circle. Having an R-symmetry (which is a necessary condition for
supersymmetry breaking, if the superpotential is generic) forbids a mass term for the
gaugino which, being a fermion in a real representation and having R-charge R(λ) =
1, would have a R-symmetry breaking mass term. But we do not see any massless
gauginos around, so gauginos should be massive. If we have an R-symmetry which is
spontaneously broken, we could generate a mass for gauginos but we should also have
an R-axion, which is not observed. This suggests that R-symmetry should be broken
explicitly. But then, generically, we cannot break supersymmetry! The conclusion is
that asking for stable vacua compatible with phenomenological observations implies
one should look for non-generic theories, which are obviously much less than generic
ones. If one accepts, instead, that supersymmetry might be broken in metastable
vacua, then R-symmetry would not be an exact symmetry but only an approximate
one. In this case gaugino mass and supersymmetry breaking would be compatible,
at least in the metastable vacuum (it is worth noticing that in concrete models there
is, not unexpectedly, some tension between the magnitude of gaugino mass and the
lifetime of such supersymmetry breaking vacua).
Regardless the R-axion problem mentioned above (which in fact can also get a
mass by gravitational effects), there exists another argument favoring metastability.
This is related with the computations we performed in section 7.4 when we studied
one-loop corrections of the O’Raifeartaigh model. We have seen that quantum
corrections lift the classical pseudomoduli space, leaving one unique supersymmetry
breaking vacuum. However, such vacuum is (the only) one where R-symmetry is in
fact not broken, so gauginos cannot get a mass! This is not a specific feature of the
O’Raifeartaigh model but applies to any model where the R-charges of superfields
are either 0 or 2. It has been proven that a necessary condition for having the
true vacuum to break the R-symmetry is to have fields in the Lagrangian having R-
charge different from 0 and 2. Models of this kind exist and have been constructed.
However, in all such models supersymmetry preserving vacua also exist. Hence,
supersymmetry breaking vacua where also R-symmetry is in the end broken, are

161
actually metastable.
Also Witten index argument favors metastability, statistically. If accept we leave
in a metastable vacuum, it means we allow for the existence of supersymmetry pre-
serving vacua elsewhere in field space. Hence, all theories with non-vanishing Witten
index would not be anymore excluded from the landscape of possible supersymmetry
breaking and phenomenologically sensible theories. For example, non-chiral theories
would be back in business. A notable such example will be presented in section 11.
The punchline is that, generically, it may be more likely we leave in a metastable
vacuum rather than in a fully stable one. Just... we need to ensure that its lifetime
is long enough to be safe!

7.7 Exercises
1. Consider a theory of n chiral superfields Φi with superpotential (7.15). Prove
that, for an interacting theory (that is, some gijk should be non-vanishing), in
order to have spontaneous supersymmetry breaking one needs at least three
chiral superfields. Derive the generic form of the corresponding three-superfield
superpotential.

2. Compute the one-loop effective potential on the classically marginal non-

supersymmetric vacua of the model in Example 3. What is the fate of these
vacua after quantum corrections are taken into account?

3. Compute the mass spectrum of the FI model both in the pure D-term as well
as in the mixed D and F-terms breaking phases, and check explicitly that the
spectrum satisfies the so-called supertrace mass formula, that is STr M2 = 0.

4. Consider a theory of three chiral superfields with canonical Kähler potential

and superpotential
1 1
W = h1 XΦ21 + h2 Φ2 Φ21 + f X
2 2
Show the existence of a classical moduli space of supersymmetry breaking
vacua. Compute the one-loop corrections to the tree-level result and show
that the moduli space is not lifted at one-loop. Can you find a simple reason
to explain such a behavior?

5. Consider all models of F-term breaking of section 7.3 and discuss whether and
how the Nelson-Seiberg criterium applies or not.

162
References

[1] A. Bilal, Introduction to supersymmetry, Section 6, arXiv:hep-th/0101055.

[2] M. F. Sohnius, Introducing Supersymmetry, Sections 6 and 11, Phys. Rept. 128
(1985) 39.

[3] K. A. Intriligator, N. Seiberg, D. Shih, Dynamical SUSY breaking in meta-stable

vacua, Appendix A, JHEP 0604 (2006) 021 [hep-th/0602239].

[4] K. A. Intriligator, N. Seiberg, Lectures on Supersymmetry Breaking, Section 2,

Class. Quant. Grav. 24 (2007) S741-S772 [hep-ph/0702069].

[5] K. A. Intriligator, N. Seiberg and D. Shih, Supersymmetry Breaking, R-

Symmetry Breaking and Metastable Vacua, Section 3, JHEP 0707 (2007) 017
[arXiv:hep-th/0703281].

[6] S. R. Coleman, The Fate Of The False Vacuum. 1. Semiclassical Theory, Phys.
Rev. D 15 (1977) 2929 [Erratum-ibid. D 16 (1977) 1248].

[7] M. J. Duncan and L. G. Jensen, Exact tunneling solutions in scalar field theory,
Phys. Lett. B291 (1992) 109-114.

[8] S. Weinberg, The quantum theory of fields. Vol. 3: Supersymmetry, Chapter

27.6, Cambridge University Press (2000).

[9] E. Poppitz and S. P. Trivedi, Dynamical supersymmetry breaking, Sections 3

and 4.4, Ann. Rev. Nucl. Part. Sci. 48 (1998) [arXiv:hep-th/9803107].

[10] Y. Shadmi and Y. Shirman, Dynamical supersymmetry breaking, Sections III.A

and III.B, Rev. Mod. Phys. 72 (2000) 25 [arXiv:hep-th/9907225].

[11] E. Witten, Constraints On Supersymmetry Breaking, Nucl. Phys. B 202 (1982)

253.

[12] D. Shih, Spontaneous R-symmetry breaking in O’Raifeartaigh models, JHEP

0802 (2008) 091 [hep-th/0703196].

[13] S. Weinberg, The quantum theory of fields. Vol. 3: Supersymmetry, Chapters

29.1 and 29.2, Cambridge University Press (2000).

163
8 Supersymmetry breaking and the Standard Model

In this chapter we will elaborate a little bit on how the machinery we have been
constructing can be used to describe physics beyond the Standard Model, under the
hypothesis that this is described by a supersymmetric theory.
The basic idea is that the Standard Model should be viewed as an effective theory,
valid only up to some scale, and that Nature, at energy above such scale, is described
by some suitable N = 1 supersymmetric extension of the Standard Model itself. The
most economic option we can think of would be a N = 1 Lagrangian which just
includes known particles (gauge bosons, Higgs fields, leptons and quarks) and their
superpartners. In fact, strictly speaking this does not apply to the Higgs sector,
which should be doubled, not to spoil the anomaly-free properties of the Standard
Model, since the Higgs fermionic partner, the higgsino, would introduce SU (2)L
gauge anomalies. Therefore, two Higgs multiplets are needed, having opposite U (1)Y
charge. Another way to realize the need of two Higgs doublets in supersymmetric
extensions of the Standard Model is to notice that the Higgs field H gives mass to
down quarks (and charged leptons) and H to up quarks. Due to holomorphy of the
superpotential, H cannot enter W and so one needs a second, independent doublet
to give up quarks a mass. These two chiral superfields are dubbed Hu and Hd and
the minimal extension of the Standard Model MSSM.
Within such minimal supersymmetric extension, one might then ask whether is
it possible to break supersymmetry spontaneously in this theory and be consistent
with phenomenological constraints and expectations. Addressing this question will
be our main concern, in this lecture.
So far we have discussed possible supersymmetry breaking scenarios at tree level:
the Lagrangian is supersymmetric but the classical potential is such that the vac-
uum state breaks supersymmetry (or at least there exist metastable but sufficiently
long-lived supersymmetry breaking vacua, besides supersymmetric ones). Can these
scenarios, i.e. either F or D-term supersymmetry breaking at tree level, occur in
such a minimal extension of the Standard Model? In what follows, we will claim the
answer is no: things cannot be as simple as that.

164
8.1 Towards dynamical supersymmetry breaking
From a purely theoretical view point there is at least one point of concern as far as
tree level supersymmetry breaking. As discussed in the first lecture, there are several
reasons to prefer sparticle masses around the TeV scale or so. This scale is not much
different from the electro-weak scale and as such much smaller than any natural UV
cut-off one can think of, like the Planck mass. If supersymmetry is broken at tree
level, the mass setting the scale of supersymmetry breaking would be some mass
parameter entering the bare Lagrangian. For instance, in the O’Raifeartaigh model
that we discussed in previous chapter this scale is µ, the coefficient of the linear term
in the superpotential. This way, we would have a scenario where an unnaturally
small mass scale has been introduced in a theory in order to solve the unnatural
hierarchy between the electro-weak scale and, say, the Planck scale, and avoid a
fine-tuning problem for the Higgs mass. What we gain introducing supersymmetry
would then just be that this small parameter, put by hand into the Lagrangian,
would be protected against quantum corrections. Is this a satisfactory solution of
the hierarchy problem?
It would be much more natural for this small mass parameter to be explained in
some dynamical way. This is possible, in fact. In order to understand how it comes,
we should first recall two facts.
First, recall that due to non-renormalization theorems, in a supersymmetric the-
ory the superpotential is tree-level exact in perturbation theory, meaning that its
full structure looks like
Weff = Wtree + Wnp , (8.1)
where the subscript np stands for non-perturbative. As we already observed, this
implies that if supersymmetry is unbroken at tree level, then it cannot be broken at
any order in perturbation theory, but only non-perturbatively.
The second piece of knowledge we need comes from a well-known property that
many gauge theories share, i.e. dimensional transmutation. Due to the running of
the gauge coupling, which becomes bigger and bigger towards the IR, any UV-free
gauge theory possesses an intrinsic scale, Λ, which governs the strong-coupling IR
dynamics of the theory, is RG-invariant and is exponentially suppressed with respect
to the scale MX at which the theory is weakly coupled. Its one-loop expression reads
#
−
Λ ∼ MX e g 2 (MX )
<< MX , (8.2)

165
where # is a number which depends on the details of the specific theory, but is
roughly of order 1.
Suppose now we have some complicated supersymmetric gauge theory which does
not break supersymmetry at tree level (so all F-terms coming from Wtree are zero),
but whose strong coupling dynamics generates a contribution to the superpotential
Wnp which does provide a non-vanishing F-term. This F-term will be order the
dynamical scale Λ, since it should vanish in the classical limit Λ → 0, and so will
the scale of supersymmetry breaking. This would imply

Ms ∼ Λ << MX , (8.3)

hence giving a natural hierarchy between Ms and the UV scale MX (which can be
the GUT scale or any other scale of the UV-free theory under consideration). This
idea is known as Dynamical Supersymmetry Breaking (DSB) and can be regarded as
the most natural way we can think of supersymmetry breaking in a fully satisfactory
way.
We will discuss several DSB models in a subsequent chapter. For now, let me just
anticipate that what we learned about tree-level supersymmetry breaking in previous
chapter will be of great help also as far as DSB is concerned. As we will see, in DSB
models the effective superpotential has typically a O’Raifeartaigh-like structure: at
low enough energy gauge degrees of freedom typically disappear from the low energy
spectrum (because of confinement, higgsing and alike) and the effective theory ends-
up being a theory of chiral superfields, only. The analysis will then follow the one of
the previous chapter, but with the great advantage that the mass parameter setting
the scale of supersymmetry breaking and sparticle masses has been dynamically
generated (and with the complication that in general the Kähler potential will be
non-canonical, of course).

8.2 The Supertrace mass formula

There is yet another reason making tree-level supersymmetry breaking not welcome
in the MSSM. This is more phenomenological in nature, and related to the so-called
supertrace mass formula.
Let us consider the most general N = 1 renormalizable Lagrangian (5.79) and
suppose that supersymmetry is spontaneously broken at tree level. Suppose we want
to compute the trace over all bosonic and fermionic fields of the mass matrix squared

166
in an arbitrary supersymmetry breaking vacuum. To this aim, let us suppose that,
generically, all F and D auxiliary fields have some non-vanishing vacuum expectation
value.

• Vectors
If F and D-fields are non vanishing, it means that some scalar fields φi
have acquired a non-vanishing VEV. If such fields are charged under the
gauge group, a mass for some vector fields will be induced since Dµ φi Dµ φi ⊃
g 2 φT a T b φ va,µ vbµ . Hence, we have for the mass matrix squared of vector bosons
 ab
(M1 )2 = 2g 2 hφT a ihT b φi = 2hDia ihDbi i , (8.4)

where the lower index on the mass matrix refers to the spin, which is one for
vectors and we have defined Dia = ∂Da /∂φi , Dai = ∂Da /∂φi .

• Fermions
The fermion mass matrix can be easily read from the Lagrangian (5.79) to be
√ !
hFij i 2ihDib i
M1/2 = √ , (8.5)
2ihDja i 0

where Fij = ∂ 2 W/∂φj ∂φi . The matrix squared reads

√ !
† hFil ihF lj i + 2hDib ihDbj i − 2ihFil ihDbl i
M1/2 M1/2 = √ (8.6)
2ihDla ihF lj i 2hDla ihDbl i

where, with obvious notation, F ij = ∂ 2 W /∂φj ∂φi .

• Scalars
The scalar mass matrix squared is instead
2 2 !
2
h ∂φ∂i ∂φ
V
i h ∂φ∂ i ∂φ
V
li
(M0 ) = 2
k
2V . (8.7)
h ∂φ∂ ∂φ
V
i h ∂φ∂ l i
j k j ∂φ

Recalling that V = F i Fi + 12 Da Da , one can write it as

!
2 hFip ihF kp i + hDak ihDia i + hDa iDiak hF p ihFilp i + hDia ihDla i
(M0 ) = ,
hFp ihF jkp i + hDaj ihDak i hFlp ihF jp i + hDaj ihDla i + hDa iDlaj
(8.8)

where Djai = −gTjai , Fijk = ∂ 3 W/∂φi ∂φj ∂φk and F ijk = ∂ 3 W /∂φi ∂φj ∂φk .

167
Taking the trace over gauge and flavor indexes of the three matrices (8.4), (8.6) and
(8.8) we get

Tr (M1 )2 = 2hDia ihDai i

Tr M1/2 M†1/2 = hFil ihF li i + 4hDia ihDai i

Tr (M0 )2 = 2hFil ihF il i + 2hDai ihDia i − 2ghDa iTr T a

and finally for the supertrace

STr M2 = 3Tr (M1 )2 + Tr (M0 )2 − 2Tr (M1/2 )2 = −2ghDa i Tr T a . (8.9)

This formula puts severe phenomenological constraints. First notice that, because
of the trace on gauge generators, the rhs is non vanishing only in presence of U (1)
factors. If this is the case, then one needs non trivial FI terms to let it being
non-vanishing, since we know that if ξ = 0 then also hDa i = 0. Now, suppose
supersymmetry were broken spontaneously, at tree level, in the MSSM. We have
only two U (1) factors we can play with, the hypercharge generator U (1)Y and,
eventually, U (1)em . The latter cannot be of any use since if the corresponding FI
parameter ξ were non-vanishing, some squarks or sleptons would get a VEV and
hence would Higgs U (1)em (comparing with the FI model, being all MSSM scalars
massless at tree level, we will be in the mixed F and D-term phase, and hence the
potential would have a minimum at non-vanishing value of some scalar field VEV).
As for the hypercharge, this again cannot work, since the trace of U (1)Y taken over
all chiral superfields vanishes in the Standard Model (this is just telling us that the
Standard Model is free of gravitational anomalies). The upshot is that within the
MSSM, formula (8.9) reduces to

STr M2 = 0 . (8.10)

It is easy to see that this formula is hardly compatible with observations. Since
supersymmetry commutes with internal quantum numbers, the vanishing of the
supertrace would imply that for any given Standard Model set of fields with equal
charge we should observe at least a real component of a sparticle with a mass smaller
than all particles with the same charge. Take for instance a charged SU (3) sector.
Gluons are massless, since SU (3) is unbroken. From (8.4) it then follows that
hDia i = hDbi i = 0, which, by (8.5), implies that the corresponding gluinos are also
massless. Then, in such charged sector, only quarks and squarks can contribute non-
trivially to eq. (8.10). Since they contribute with opposite sign, the squarks cannot

168
all be heavier than the heaviest quark, and some must be substantially lighter. Take,
for instance, the color-triplet sector with electric charge e = −1/3, to which down,
strange and bottom quarks belong. We get m2d + m2s + m2b ' (5GeV)2 . In order
P
to satisfy eq. (8.10) we need scalar partners to satisfy i m2φi ' 2(5GeV)2 , which
implies that a charged scalar with mass smaller than 7 GeV should exist. This is
clearly excluded experimentally.
The upshot of this discussion is that we should give up with the idea that the
whole story is as simple as just tree level supersymmetry breaking in the MSSM.

8.3 Beyond the MSSM

The supertrace mass formula derived above comes from at tree-level analysis and,
for one thing, we know that masses get modified by loop effects. So one might hope
that at quantum level things could sensibly change. However, within the MSSM such
modifications are small since the Standard Model is a weakly interacting theory at
the electro-weak scale, so this would not help much.
A way to avoid the supertrace mass formula severe constraints, while still keeping
the MSSM, would be to allow for supersymmetry breaking beyond tree-level, that is
dynamical supersymmetry breaking. If supersymmetry breaking were transmitted
to the MSSM by quantum corrections, there would be effective corrections to ki-
netic terms from wave-function renormalization which could violate the supertrace
mass formula by in principle large amounts, hence allowing for phenomenologically
meaningful sparticle spectra. We do have a dynamical scale we can play with in
the Standard Model, the SU (3) strong coupling scale ΛQCD . However, DSB driven
by QCD strong coupling dynamics could not work either. Looking at eq. (8.3) we
would expect in this case a supersymmetry breaking scale of order 300 MeV, which
is by far too low for accommodating any sensible phenomenology.
The punchline is that in order to describe beyond the Standard Model physics
we need something more than just the MSSM . We might need new particles and
fields and/or new strong interactions. The options one can play with are many, and
understanding the correct path of supersymmetry breaking beyond the Standard
Model has been, and still is, a matter of concern and great challenge for theoretical
physicists. There are, however, at least two basic properties a competitive model
should have. Supersymmetry should be broken dynamically, so to generate the low
scale we need (much lower than, say, the Planck scale) in a natural way. Second, in

169
order to avoid the unpleasant constraints coming from the supertrace mass formula,
we should better rely on non-renormalizable couplings, or loop effects, to transmit
this breaking to the MSSM. As we will see shortly, besides invalidating formula
(8.10), such an option would also have the free-bonus of providing an extra suppres-
sion between the natural scale of the underlying UV theory and the scale of MSSM
sparticle masses. Hence, the primordial supersymmetry breaking scale would not
need to be comparable with electro-weak scale. It could be sensibly higher.

8.4 Spurions, soft terms and the messenger paradigm

Let us deviate, for a while, from what we have been saying so far, and come back to
what we said at the very beginning of previous chapter about possible mechanisms
for supersymmetry breaking. We have a second option we have not yet considered:
explicit supersymmetry breaking by soft terms. Let us suppose we add explicit su-
persymmetry breaking terms to the MSSM Lagrangian. In order to save the nice UV
properties of supersymmetry, these terms should be UV irrelevant. For instance, if
we were to add non-supersymmetric dimensionless couplings, like Yukawa couplings
and scalar quartic couplings, we would certainly destroy the pattern of UV cancel-
lations which makes supersymmetry solving, e.g. the hierarchy problem. We can
instead add mass terms, and more generally, positive dimension couplings, like cubic
scalar couplings. These would simply tell us below which scale UV cancellations will
stop working. Such soft supersymmetry breaking Lagrangian will schematically be
of the form
LSOFT = mλ λλ − m2 φφ + b φφ + a φ3 + h.c. , (8.11)
where λ represents gauginos and φ any possible scalar of the MSSM. The first two
terms provide masses for gauginos (wino, zino, photino, gluino) and scalar particles
(squarks and sleptons, Higgs particles), respectively. The third term, known as B-
term, may arise in the Higgs sector and couples the up and down scalar Higgs Hu
and Hd . Finally the fourth, known as A-term, corresponds to cubic gauge and flavor
singlet combinations of MSSM scalars, e.g. Higgs and left and right squark com-
ponents. A-terms are in one-to-one correspondence with Yukawa couplings (which
belong to the supersymmetric part of the MSSM Lagrangian): each quark and lepton
is just substituted by its scalar partner.
All terms appearing in eq. (8.11) are UV irrelevant and renormalizable, and it

170
was indeed shown time ago that the full Lagrangian

L = LMSSM + LSOFT (8.12)

is free of quadratic divergences to all orders in perturbation theory. Notice that

such a Lagrangian would automatically solve the supertrace mass formula problem.
Indeed, a Lagrangian like the one above would violate eq. (8.10) precisely by terms
of order the sparticle masses, see the expression (8.11), which is, by construction,
compatible with observations.
There is a number of very important issues one should discuss regarding the
Lagrangian (8.12), including a number of potential problems some of the soft terms
could pose, like the so-called supersymmetry flavor, CP and fine-tuning problems,
to name a few. Discussing this is beyond the scope of this course. What we want to
do instead, is to reconnect to our previous discussion and show how such a rather ad
hoc Lagrangian as (8.12), where supersymmetry is broken explicitly, can actually be
generated by spontaneous supersymmetry breaking in a larger theory, which includes
fields and interactions beyond the MSSM ones.
First, let us recall the idea of spurion fields. In a supersymmetric theory any
constant, non-zero value for the lowest component of a superfield (a VEV) does
not break supersymmetry. Hence, in a supersymmetric Lagrangian each coupling
constant can be promoted to a background superfield, a spurion, with non-vanishing
such lowest component VEV. Let us take, for instance, the WZ model
Z Z  
2 2 2 1 2 1 3
L = d θd θ̄ ZΦΦ + d θ M Φ + λΦ + h.c. , (8.13)
2 6

and think of Z, M and λ as real and chiral background superfields, respectively. If

only their lowest components have a non-vanishing VEV, this is just the WZ model
itself.
Interestingly, one can include supersymmetry breaking terms in the above La-
grangian by allowing these superfields having higher (scalar) component VEVs. We
can set in general

hZi = 1 + θ2 B + h.c. + c θ2 θ̄2

hM i = µ − θ2 FM

hλi = λ − θ2 Fλ .

171
Plugging these expressions into the Lagrangian (8.13), after integrating out the
auxiliary field FΦ we find for the potential
   
2

2 1 1 3
V = VSUSY − c − |B| φφ + (FM + Bµ) φ + Fλ + Bλ φ + h.c. , (8.14)
3 2
2
where VSUSY = µφ + 21 λφ2 . We see that the non-supersymmetric contribution
to the potential exactly reproduces the second, third and fourth soft terms of the
Lagrangian (8.11), upon the trivial identifications

m2 = c − |B|2

b = FM + Bµ
1 1
a = Fλ + Bλ .
3 2
Following the same logic for the SYM action
Z
L = d2 θ τ Waα Wαa , (8.15)

one can seemingly reproduce gaugino masses by promoting the complexified gauge
coupling τ to a chiral superfield and provide a non-vanishing VEV for its F-term

hτ i = τ + θ2 mλ . (8.16)

Applying this logic to the Lagrangian (8.12), one can actually write all soft terms
by means of spurion couplings, and rewrite (8.12) using a pure supersymmetric
formalism. This turns out to be a very convenient thing to do when it comes to
compute the divergence structure of the theory and prove, e.g. the absence of
quadratic divergences.
Although phenomenologically viable and logically consistent, this picture is still
not completely satisfactory. The Lagrangian (8.12) has more than 100 free parame-
ters (masses, phases, mixing angles, etc...), meaning that there are few unambiguous
predictions one can really make. One might want to find some organizing principle,
where these many parameters could be naturally explained in terms of some simpler
underlying theory.
Here is where we can close the gap between soft term breaking and sponta-
neous supersymmetry breaking. It is enough to promote spurions to fully fledged
superfields with their own Lagrangian and kinetic terms. By some suitable and

172
for the time being unspecified mechanism, they acquire non-vanishing F and D-
terms spontaneously, and then generate soft terms by their interactions with the
MSSM fields via couplings of the kind (8.13). This is the basic idea of the so-called
messenger paradigm: one imagines a fully renormalizable theory where supersym-
metry is broken spontaneously in some hidden sector and then communicated to the
MSSM fields by non-renormalizable interactions and/or loop effects. After integrat-
ing out heavy fields, this will generate effective couplings precisely as those in the
Lagrangian (8.13), with non-vanishing F and D-components for some fields. These
F and D-terms will then give rise to soft terms through a procedure like the one
above. This way, all specific properties that MSSM supersymmetric breaking soft
terms should have, will be ultimately generated (and explained) by a larger theory
in which supersymmetry breaking occurs spontaneously.

Hidden Visible
(SUSY breaking) MSSM
sector sector

Interactions

Figure 8.1: The messenger paradigm: supersymmetry is broken in a hidden sector

and then communicated to the visible MSSM sector (or any viable supersymmetric
extension of the Standard Model) via interactions felt by the MSSM particles.

8.5 Mediating the breaking

What are the possible ways in which a scenario as the one outlined above can actually
be realized?
An obvious candidate as messenger of supersymmetry breaking is gravity, since
any sort of particle couples universally to it. Gravity is inherently non-renormalizable,
at least as it manifests itself at energies lower than the Planck scale. Hence, cou-
plings like those appearing in eqs. (8.13) and (8.15) are precisely what one expects,
in this scenario.

173
 Another possibility is that supersymmetry breaking is mediated by gauge inter-
actions. We can imagine that supersymmetry is broken in the hidden sector and that
some fields, known as messenger fields, feeling (or directly participating in, this is a
model-dependent property) supersymmetry breaking are also charged under Stan-
dard Model gauge interactions. Gauginos will directly couple to suich messenger
fields and get a mass at one-loop. Scalar sparticles, instead, would get mass at two
loops, interacting with messenger fields via intermediate MSSM vector superfields,
to which gauginos belong to. In this scenario, soft terms will be generated after inte-
grating out heavy fields, ending-up again with effective couplings of the kind (8.13)
and (8.15). Obviously, the main source of mediation can be gauge interactions only
in a regime where the always present gravity mediation is suppressed. Below we will
give an estimate of the regime where such a situation can occur.
In what follows, we are not going to discuss these two mediation mechanisms
in detail, nor any of their diverse phenomenological benchmarks, neither the many
variants of the basic models which have appeared in the literature, with their pros
and cons. We just want to give a rough idea on how these two mediation mecha-
nisms work and show, in particular, how they can naturally generate, at low energy,
spurion-like couplings with MSSM fields and, eventually, give rise to soft terms.

8.5.1 Gravity mediation

From a low energy point of view, one can parameterize the effect of unknown physics
at the Planck scale MPl by higher order operators, suppressed by MPl . Suppose that
some hidden sector field X gets a non-vanishing F -term, that is

hXi = 0 , hFX i =
6 0. (8.17)

The most general form of the Lagrangian describing the gravitational interaction
between X and the visible sector fields will be something like
Z  
2 2 c i b0 b
Lint = d θd θ̄ XXQi Q + 2 XXHu Hd + XHu Hd + h.c.
MPl2 MPl MPl
Z  
2 s α a a i ei + h.c.
+ dθ XWa Wα + XQ Hu Q (8.18)
MP l MPl
plus, possibly, higher order operators. In the above expression Qi ’s represent all
matter superfields as well as the two Higgs doublets, while Hu and Hd obviously refer
to the up and down Higgs only. For the sake of simplicity, we have taken all order
one dimensionless coefficients in each term to be the same, that is i-independent.

174
 Plugging the values (8.17) into the above Lagrangian one gets all possible MSSM
soft terms! The first term on the rhs of eq. (8.18) gives rise to non-supersymmetric
masses for all sfermions (squarks, sleptons and scalar Higgs particles), while the
second and third terms provide mass terms for the scalar Higgs only (more below).
The first term of the second line provides gaugino masses. Finally, the last term
generates all A-terms. We see that we get a rather simple pattern of soft terms. Up
to order one coefficients, they share one and the same mass scale
hFX i
mSOFT ∼ . (8.19)
MPl
Imposing mSOFT to be order the TeV scale we see that in a gravity mediated scenario
the primordial supersymmetry breaking scale, the so-called intermediate scale, is
order
p p
Ms = hFX i ∼ mSOFT MPl ∼ 1011 GeV , (8.20)
somewhat in between the electro-weak scale and the Planck scale.
Let us spend a few more words on Higgs mass terms. From the Lagrangian (8.18)
we see three contributions to scalar Higgs mass . The first gives rise to mass terms
for the up and down Higgs, respectively (they are proportional to Hu† Hu and Hd† Hd ).
The second term is the -term, which gives rise to a quadratic term mixing Hu and
Hd . Finally, as for the third term, notice that it can be re-written as
Z Z
2 2 b hF X i
d θd θ̄ XHu Hd = b d2 θHu Hd . (8.21)
MPl MPl
This contribution is a so-called µ term contribution and upon integration in chiral
superspace it gives a quadratic contribution similar in structure to the first term.
Notice that all these three couplings are needed in order to trigger electro-weak
symmetry breaking. The first such terms gives masses to scalar Higgs particle,
and it can actually give a negative mass square to some of them, something we
certainly need to trigger spontaneous symmetry breaking. The second one is also
necessary. One can show the B term to be proportional to sin 2β where tan β is
the ratio between the VEVs of the up and down Higgs, tan β = vu /vd . Clearly,
if B = 0, either the up or the down Higgs do not get a VEV, and therefore one
cannot provide masses to all Standard Model particles. Finally, the µ term is the
only possible contribution which can provide higgsino a mass, and therefore should
certainly be there.
The way we have re-written the µ-term makes it clear that it can also be (and
generically is) generated from a perfectly supersymmetric superpotential coupling

175
in the MSSM Lagrangian
W = µSUSY Hu Hd . (8.22)
There is no a priori reason why the above term, which comes from a supersym-
metric contribution and is then not related to the dynamics driving the breaking of
supersymmetry, should come to be the same scale of the soft terms, as it should.
In principle, it could be any scale between mSOFT and MPl . This is the famous µ
(or µ/Bµ) problem: how to avoid large µ-terms and at the same time have them
the same order of magnitude of B-terms. Gravity mediation provides an elegant
and simple way to solve this problem. First, one can impose some PQ-like discrete
symmetry on the MSSM Lagrangian which forbids a tree level µ term in the super-
potential, µSUSY = 0. This is achieved giving charge 1 under this symmetry to Hu
and Hd and charge -1/2 to all other chiral superfields (quarks and leptons). This
way, both the µ and the B terms are generated radiatively. The non-trivial thing is
to make them be the same order of magnitude. However, as we have seen above, this
is exactly what happens in a gravity mediation scenario: up to coefficients of order
unity, all soft terms, including B and µ terms, are of the same order, eq. (8.19)!
A typical problem of gravity mediation scenarios is, instead, the so-called su-
persymmetry flavor problem. In order not to spoil the excellent agreement between
flavor changing neutral current (FCNC) effects predicted by the Standard Model
and known experimental bounds, any sort of new physics should not induce any
sensible extra FCNC. In order for this to be the case the interactions mediating su-
persymmetry breaking better be flavor-blind. This is not the case for gravity whose
UV completion is not actually guaranteed to couple universally to flavor. There-
fore, in general, in gravity mediation scenarios one has to confront with the flavor
problem. We will not discuss this further here and refer to the references at then
end of the chapter. Let us just remark that there exist different proposals on how
to overcome this problem, the most compelling and natural one being the so-called
anomaly mediation scenario.

8.5.2 Gauge mediation

Any gauge mediation model is characterized by the assumption that there exist
messenger fields. The latter, by definition, are those hidden sector fields which are
charged under the Standard Model gauge group. The basic idea of gauge mediation
is as follows.

176
 Messengers couple (in a model-dependent way) to hidden sector supersymmetry
breaking dynamics and this affects their mass matrix which, besides a supersym-
metric contribution (which is supposed to be large enough not to make messengers
appear at energies of order the electro-weak scale), receives a non-supersymmetric
contribution. By coupling radiatively with MSSM fields, supersymmetry breaking
is communicated to MSSM fields and provides all desired soft terms, as we are going
to show next. For instance, gaugini get a mass at one-loop while squarks, slep-
tons and Higgs fields feel supersymmetry breaking at two loops through ordinary
SU (3) × SU (2) × U (1)Y gauge boson and gauginos interactions. One of the beau-
ties of gauge mediation as opposed to gravity mediation, is that gauge mediation
supersymmetry breaking can be understood entirely in terms of loop effects in a
renormalizable framework. Hence, it has a high level of reliability and calculability.
There are different schemes for gauge mediation, e.g. minimal, direct and semi-
direct gauge mediation, which differ, ultimately, by the way the messenger mass
matrix is affected by the hidden sector supersymmetry breaking dynamics. This
provides different patterns for the MSSM soft terms texture. As an exemplification,
in what follows we will briefly discuss minimal gauge mediation (MGM) which is
a simple, still rich enough scenario to let one get a feeling on how things work.
In MGM all complicated hidden sector dynamics is parameterized in terms of a
single chiral superfield X which couples to the messenger sector through a tree-level
superpotential coupling. The messenger sector is made of two set of chiral superfields
Φ and Φ e transforming in complex conjugate representation of the SM gauge group,
so not to generate gauge anomalies. The interaction term is as simple as

e .
W = X ΦΦ (8.23)

A rough scheme of MGM is depicted in Figure 8.2.

The spurion-like field X inherits non-vanishing F and lower component term
VEVs from the hidden sector,

hXi = M + θ2 hFX i . (8.24)

Once plugged into the messenger Lagrangian, this gives a splitted messenger mass
spectrum
m2φ,φ̃ = M 2 ± hFX i , mψ,ψ̃ = M . (8.25)
While fermions receive only the supersymmetric contribution, scalars receive both
supersymmetric and non-supersymmetric contributions. Recalling that messenger

177
 SUSY
Messengers MSSM
breaking
e, sector
<latexit sha1_base64="NmAM1lBBtn+7IarTSGrNNCClD+0=">AAAB/3icdVDLSgMxFM34rPVVFdy4CRbBhQwZW7Huim5cVrAPaEvJZO60oZkHSUYpYxf+ihsXirj1N9z5N6bTCip64HIP59xLbo4bC640IR/W3PzC4tJybiW/ura+sVnY2m6oKJEM6iwSkWy5VIHgIdQ11wJasQQauAKa7vBi4jdvQCoehdd6FEM3oP2Q+5xRbaReYbdzyz3QXHiAO7UBx0dZ6xWKxC6dGVQwscuEEIfMCClhxyYZimiGWq/w3vEilgQQaiaoUm2HxLqbUqk5EzDOdxIFMWVD2oe2oSENQHXT7P4xPjCKh/1Imgo1ztTvGykNlBoFrpkMqB6o395E/MtrJ9qvdFMexomGkE0f8hOBdYQnYWCPS2BajAyhTHJzK2YDKinTJrK8CeHrp/h/0ji2nRObXJWL1fNZHDm0h/bRIXLQKaqiS1RDdcTQHXpAT+jZurcerRfrdTo6Z812dtAPWG+f6c6VYw==</latexit>

sector

Superpotential SM gauge
coupling X e
<latexit sha1_base64="sc9GR2NHOI34D3phAxb8eoJOFj8=">AAAB/3icdVDLSgMxFM3UV62vUcGNm2ARXA0ZW7Huim5cVrAP6JSSydy2oZkHSUYptQt/xY0LRdz6G+78G9NpBRU9cC+Hc+4lN8dPBFeakA8rt7C4tLySXy2srW9sbtnbOw0Vp5JBncUili2fKhA8grrmWkArkUBDX0DTH15M/eYNSMXj6FqPEuiEtB/xHmdUG6lr77Wwd8sD0FwEgL3agGetaxeJUzozqGDilAkhLpkTUsKuQzIU0Ry1rv3uBTFLQ4g0E1SptksS3RlTqTkTMCl4qYKEsiHtQ9vQiIagOuPs/gk+NEqAe7E0FWmcqd83xjRUahT6ZjKkeqB+e1PxL6+d6l6lM+ZRkmqI2OyhXiqwjvE0DBxwCUyLkSGUSW5uxWxAJWXaRFYwIXz9FP9PGseOe+KQq3Kxej6PI4/20QE6Qi46RVV0iWqojhi6Qw/oCT1b99aj9WK9zkZz1nxnF/2A9fYJKUGVjw==</latexit>

interactions

Figure 8.2: Minimal gauge mediation. Messengers feel supersymmetry breaking via
a cubic coupling with a spurion-like chiral superfield X which has a non-vanishing
F-term VEV inherited from the hidden sector non-supersymmetric dynamics.

fields are charged under the SM gauge group we see there is a stability bound which
forces us to take M 2 > hFX i (if not, some messenger scalars would get a non-
vanishing VEV and would break part of the SM gauge group). If M is large enough
we can then integrate the messengers out and the effective low energy theory at
scale lower than M breaks supersymmetry.
The net low energy effect boils down to radiative corrections to gaugini propa-
gator, which get a mass at one loop, while gauge bosons remain massless since they
are protected by gauge invariance. Via intermediate Standard Model gauge coupling
interactions, also MSSM scalar fields get a non-supersymmetric mass contribution,
though at two-loop order. Feynman diagrams contributing to gaugino and scalar
masses are reported in Figures 8.3 and 8.4, respectively.

Figure 8.3: The one-loop diagram providing gaugino mass. Black lines are MSSM
fields (just gauginos in this case), green lines are messenger fields. Dashed lines
correspond to scalar fields and continuous lines to fermion fields.

The gaugino mass computation is rather easy, since only one type of diagram
contributes. In the limit of small hFX i/M 2 the end result can be organized in a

178
 sfermion-gluon loop

gluon loop

fermion-gluino loop sfermion loop

Figure 8.4: Two-loops diagrams providing sfermion masses. There are four different
class of diagrams, the first three originating from a specific MSSM fields one-loop
diagram by inserting messenger loop corrections, as indicated. The last is a two-
loop diagram which comes from D-terms and mixes MSSM and messenger scalars.
Conventions are as in Figure 8.3.

series expansion and reads, to leading order in hFX i/M 2

 
g 2 hFX i hFX i 2
mλ ∼ 1 + O( ) . (8.26)
16π 2 M M2

Summing-up all two-loop contributions renormalizing scalar masses is instead quite

laborious even if conceptually straightforward. However, the end result is surpris-
ingly simple and, again in the limit of small hFX i/M 2 , reads
 2  
g2 hFX i 2 hFX i 2
m2sf ∼ 1 + O( ) . (8.27)
16π 2 M M2

In principle, there is also a one-loop contribution to sfermion masses originating

from the quartic scalar coupling involving two sfermions and two scalar messengers,
the same vertex which gives raise to the two-loop contribution in the last diagram
of figure 8.4. Due to contraction on gauge indexes, this can be non-vanishing for
abelian factors only, like e.g. U (1)Y . This contribution would be proportional to

179
the hypercharge of the corresponding sfermions and therefore may induce tachyonic
mass contributions which would be problematic, phenomenologically. Therefore, a
symmetry in the messenger sector is usually imposed in order to avoid such danger-
ous one-loop contributions.
A-terms are also generated radiatively, via the insertion of the renormalized
gaugino propagator of Figure 8.3 inside a fermion-higgsino-gaugino loop to which a
Higgs field and two sfermions can be attached as external legs, as shown in figure
8.5. Overall, this is again a two-loop effect.
Note, finally, that B and µ terms cannot be generated by any of the diagrams in
Figure 8.4 and require a separate discussion, as we will see shortly.

Higgs

higgsino

sfermion sfermion
fermion

Figure 8.5: The two-loop diagram generating A-terms in gauge mediation.

In agreement with the general philosophy advocated in section 8.4, one can get
these same results working within the effective low energy theory valid at scales
smaller than M , which is obtained by integrating messenger fields out. At E < M
the effect of the messengers is taken care of in the wave function renormalization of
gauge and matter kinetic terms of the MSSM fields. Soft terms arise from derivatives
in the X-field of the renormalized gauge and matter kinetic functions, ZV (X, µ) and
ZQ (X, X, µ), which can be evaluated at a scale µ by solving the RG equations. For
example, soft masses read
∂ ln ZV (X, µ) hFX i
mλ ∼ (8.28)
∂ ln X X=M M

∂ 2 ln ZQ (X, X, µ) hFX i 2
m2sf ∼ , (8.29)
∂ ln X∂ ln X X=M M
and similar formulae hold for the A-terms. This powerful method, originally pro-
posed by Giudice and Rattazzi, is not specific to gauge mediation but works when-
ever supersymmetry breaking is communicated by renormalizable perturbative in-
teractions. We refer to the bibliography at the end of the chapter for more details.

180
 We see from eqs. (8.26)-(8.27) that in MGM all soft terms come naturally of the
same order of magnitude
g 2 hFX i
mSOFT ∼ . (8.30)
16π 2 M
Imposing again that soft masses are order the TeV scale and setting g 2 /16π 2 ∼ 10−2
one then gets
hFX i
∼ 105 GeV , (8.31)
M
which implies that in MGM the primordial supersymmetry breaking scale Ms is
bounded from below as
p p
Ms = hFX i ∼ 10 mSOFT M ≥ 105 GeV , (8.32)

where the lower bound is reached for M 2 ∼ hFX i.

As we have already observed, gravity mediation is an always present contribu-
tion to supersymmetry breaking mediation mechanisms (i.e. the field X would also
interact gravitationally with the visible sector via a Lagrangian like (8.18), in gen-
eral). Hence, it is only when its contribution is suppressed with respect to that of
gauge mediation that the latter can play a role. In order for gravity effects to be
negligible, say to contribute no more than 1/1000 to soft mass squared, one gets an
upper bound for the scale M
g 2 hFX i 3/2 hFX i g2
2
≥ 10 −→ M ≤ 2
10−3/2 MPl ∼ 1015 GeV . (8.33)
16π M MPl 16π
√
Using the relation Ms ∼ 10 mSOFT M this gives an upper bound for Ms of order
1010 GeV. Together with the lower bound (8.32) this implies that the supersymme-
try breaking scale Ms can range from 105 to up to 1010 GeV, in gauge mediation
scenarios.
Let us close this brief overview on gauge mediation saying a few words about
flavor and µ problems. We are in a sort of reversed situation with respect to gravity
mediation. Gauge interactions are intrinsically flavor-blind. Hence, gauge mediation
does not provide any further FCNC contribution to the Standard Model and the
flavor problem is then automatically solved in this framework. On the contrary,
the µ problem is much harder. One can again avoid a supersymmetric µ term by
means of some discrete symmetry to be imposed on the Higgs sector supersymmetric
Lagrangian. What is problematic, though, is to generate radiatively µ and B terms
of the same order of magnitude. The two-loop diagrams in Figure 8.4 do not provide
B and µ terms and one should then argue for a direct coupling between the Higgs

181
and the messenger sectors. The simplest possible model one can think of, does not
work. Indeed, allowing a cubic coupling between Hu , Hd and the field X

WH = λH XHu Hd , (8.34)

one could in principle generate both a µ and a B-term from supersymmetry breaking
dynamics but they do not come of the same order of magnitude. In order for the µ
term being of the order of other soft masses, as it should be, we need

µ = λH M ∼ 1 Tev . (8.35)

This implies that λH is order 10−2 or smaller. This enhances the B-term. Indeed,
recalling that hFX i ≤ M 2 , the non-supersymmetric to supersymmetric mass ratio
contribution coming from the superpotential coupling (8.34) is
B λH hFX i hFX i
2
∼ 2 2 ∼ ∼ 102 , (8.36)
µ λH M µM
where in the last step we used the fact that hFX i/M ∼ 105 GeV. This gives an
unacceptably large B-term. This problem is not specific to MGM nor to the actual
way we have generated µ and B-terms here. It is a problem which generically plagues
any gauge mediation scenario. Even though several proposals has been put forward
to solve the µ-problem in gauge mediation, it is fair to say that a fully satisfactory
and natural framework to solve this problem is not yet available.
It is finally worth stressing that the simple mass pattern (8.26)-(8.27) is not a
generic feature of gauge mediation but specific to MGM only. Indeed, in another
popular scheme, direct gauge mediation, the soft spectrum tends to be split, that is
gauginos are typically suppressed with respect to scalar particles.
A generic, model-independent prediction of gauge mediation scenarios, instead,
is that the gravitino is the lightest supersymmetric particle. Gravitinos interact
only gravitationally and get a mass due to higgsing of order m3/2 ∼ hFX i/MMp .
Therefore, while in a gravity mediation scenario gravitonos have a mass of the same
order of magnitude of all other soft terms, in gauge mediation they are suppressed,
since
hFX i hFX i M hFX i
m3/2 ∼ = << = mSOFT , (8.37)
MPl M MPl M
and they can as light as few eV.

Let me conclude this brief overview stressing again what is the main point of
this all business. What all these mediation models are about is to provide a theory

182
of the soft terms, a predictive pattern for these extra terms that one can (and has
to) add to the MSSM Lagrangian or any desired supersymmetric extension of the
Standard Model. We have been trying to give an idea on how things might work,
and reviewed few aspects of the most basic mediation mechanisms. A throughout
analysis of the phenomenology of these schemes and their variants is not our goal
here and we refer to the bibliography at the end of the chapter for a more detailed
analysis. In the remainder of these lectures we will instead focus on the hidden
sector dynamics, trying to deepen our understanding of supersymmetric dynamics
at strong coupling. Besides its intrinsic interest (and the far reaching consequences
in our understanding of strong coupling regimes of gauge theories in general), this
will also allow us to study concrete models of dynamical supersymmetry breaking.

8.6 Exercises
1. Derive the gaugino mass formula (8.26) from the Feynman diagram of Figure
8.3.

2. Compute the contribution of two diagrams arbitrarily chosen out of those

depicted in Figure 8.4 to the sfermion mass formula (8.27).

References

[1] A. Bilal, Introduction to supersymmetry, Section 6.4, arXiv:hep-th/0101055.

[2] S. P. Martin, A Supersymmetry Primer, Sections 4, 5.3, 6.4, 6.6, 6.7 and 6.8,
arXiv:hep-ph/9709356v5.

[3] M. A. Luty, 2004 TASI lectures on supersymmetry breaking, Sections 5, 7, 8, 9,

10 and 11, arXiv:hep-th/0509029.

[4] Y. Shirman, TASI 2008 Lectures: Introduction to Supersymmetry and Super-

symmetry Breaking, Sections 6 and 7, arXiv:0907.0039 [hep-ph].

[5] G. F. Giudice and R. Rattazzi, Theories with gauge-mediated supersymme-

try breaking, Sections 2, 3 and 6, Phys. Rept. 322 (1999) 419 [arXiv:hep-
ph/9801271].

[6] J. Terning, Modern supersymmetry: Dynamics and duality, Chapters 6.1, 6.2
and 16, Oxford University Press (2006).

183
9 Non-perturbative effects and holomorphy

In this lecture we will start looking at the non-perturbative regime of supersym-

metric field theories. The main point here will be to introduce holomorphy, or better
put holomorphy, which is an intrinsic property of supersymmetric theories, at work.
Before doing that, however, there are a few standard non-perturbative field theory
results we need to review.

9.1 Instantons in a nutshell

Gauge theories might contain a so-called θ-term, which is
Z
θYM 1
Sθ = 2
d4 x Tr Fµν Feµν where Feµν = µνρσ Fρσ . (9.1)
32π 2
This term is a total space-time derivative, since
Z Z  
1 2
d x Tr Fµν Fe = d x 
4 µν 4 µνρσ
∂µ Tr Aν ∂ρ Aσ + Aν Aρ Aσ . (9.2)
2 3

This implies that the θ-term does not have any effect on the classical equations of
motion. However, when quantizing a theory one has to average over all fluctuations,
not just those satisfying the classical equations of motions and therefore the θ-term
can in fact be relevant in some cases.
In quantum field theory all information about physical observable (the spectrum
and the S-matrix) can be obtained from correlation functions of operators, which are
defined by the Feynman path integral. The most convenient formulation is where
these quantities are analytically continued in Euclidean space. Feynman rules can be
derived from the path integral but the latter is believed to contain more information,
including effects which are non-perturbative in the coupling constant. For a generic
gauge theory, the generating functional in Euclidean space reads, schematically
Z  Z 
1 4
Z[J] = DΦ exp − 2 S[Φ] + d xJ Φ , (9.3)
g

where Φ represents a set of fields with source J, S[Φ] is the Euclidean action, which
is real and bounded from below, and g the dimensionless gauge coupling. The
basic idea of semi-classical approximation (which corresponds to the limit of weak
coupling, g 2 → 0) is that the path integral is dominated by configurations of lowest
Euclidean action and one should proceed expanding around these confugurations.

184
The simplest are perturbative vacua, namely minima of the classical potential, and
the expansion is just the loop expansion. However, there can exist other minima
with finite action and one should expand in fluctuations around them, too. Note that
for a configuration of finite action, if it exists, the leading semi-classical contribution
2
goes as e−S/g so it is highly suppressed at weak coupling (and fluctuations lead to
corrections which are further suppressed by further powers of g 2 ).
A class of configurations of this kind, to which the term (9.1) is sensitive to,
are instantons. Instantons are classical solutions of the Euclidean action that ap-
proach pure gauge for |x| → ∞. Under a gauge transformation by a group element
U (x) = exp[λa (x)Ta ] the gauge connection Aµ (x) = gAaµ (x)Ta and the field strength
Fµν (x) = ∂µ Aν (x) − ∂ν Aµ (x) + [Aµ (x), Aν (x)] transform as follows

Aµ → U Aµ U −1 + U ∂µ U −1 (9.4)

Fµν → U Fµν U −1 , (9.5)

where x-dependence is everywhere understood. Note that if Fµν vanishes then Aµ is

a gauge-transform of 0, meaning that for some U we have that Aµ = U ∂µ U −1 . Since
(9.2) is a total derivative, it can be expressed as an integral on S 3 at infinity. So, in
order to make the action finite Fµν should go to 0 at infinity faster than 1/r2 . For
this to happen Aµ does not need to go as O(1/r2 ) but is enough it is a pure gauge,
that is
Aµ → U ∂µ U −1 + O(1/r2 ) . (9.6)
For these configurations one can show that the integral (9.2) is in fact an integer,
the so-called instanton number of the configuration (also known as winding number)
Z
θYM
Sθ = d4 x Tr Fµν Feµν = n θYM where n ∈ Z , (9.7)
32π 2
where n ∈ Z.
The instanton number is a topological quantity, in the sense that it does not
change upon continuous deformations of the gauge field configuration. Moreover,
R
since the action enters the path integral as DφeiSθ , the θ-angle indeed behaves as
an angle, in the sense that the shift

θYM → θYM + 2π , (9.8)

is a symmetry of the theory, , that’s why the name.

185
 An instanton field configuration interpolates between different vacua of the gauge
theory. Both vacua are gauge equivalent to the usual vacuum with zero gauge poten-
tial but the corresponding gauge transformation cannot be deformed to the identity
(these are known as large gauge transformations). Because if this were the case it
would have been possible to let the field strength being vanishing in all space-time,
contradicting eq. (9.7). The fact that Fµν cannot vanish identically for configura-
tions with n 6= 0 implies that there is a field energy associated with the gauge field
configuration interpolating between the different vacua. An energy barrier and an
associated quantum mechanical tunneling amplitude proportional to e−SE where SE
is the Euclidean action of a field configuration which interpolates between the dif-
ferent vacua. Finite action solutions have the interpretation of mediating quantum
tunneling effects. Instantons are nothing but just such interpolating field config-
urations. Notice that configurations related by large gauge transformations are
weighted differently in the action, because of eq. (9.7), implying that they should
not be identified as physically equivalent configurations.
Instantons have an intrinsic non-perturbative nature. Recall that the RG-equation
for the gauge coupling g reads
∂g b1 3
µ =− 2
g + O(g 5 ) , (9.9)
∂µ 16π
where b1 is a numerical coefficient which depends on the theory. The solution of this
equation at one loop is
1 b1 Λ
2
= − 2 log . (9.10)
g (µ) 8π µ
where the scale Λ is defined as the scale where the one-loop coupling diverges. It
sets the scale where higher-loop and non-perturbative effects should be taken into
account. For any scale µ0 we have that
8π 2
−
Λ ≡ µ0 e b1 g 2 (µ0 )
. (9.11)

It is important to stress that Λ does not depend on the energy scale: it is a RG-
invariant quantity. Indeed
  
∂Λ − 8π
2
8π 2 2 b1 3 5 − 8π
2

= e b1 g 2 (µ0 )
+ µ0 − 3 g (µ 0 ) + O(g ) e b1 g 2 (µ0 )
∂µ0 b1 g (µ0 ) µ0 16π 2
8π 2 1 − b g8π2 (µ2 )
−
b1 g 2 (µ0 )
= e + µ0 (− )e 1 0 = 0 . (9.12)
µ0
up to higher-order corrections. This can be reiterated order by order in perturbation
theory, getting the same result, namely that ∂Λ/∂µ0 = 0.

186
 An important point regarding instantons is that there exists a lower bound on
their Euclidean action. Indeed, we have that
Z  2 Z h i
4 e
0 ≤ d x Tr Fµν ± Fµν = d4 x 2 Tr Fµν F µν ± 2 TrFµν Feµν (9.13)

which implies
Z Z
4
d x Tr Fµν F µν
≥ d4 x Tr Fµν Feµν = 32π 2 n , (9.14)

where the last equality holds for an instanton configuration with instanton number
n. This implies that there is a lower bound to an instanton action: instanton
contributions to amplitudes are suppressed at least by (just multiply above equation
by 1/4g 2 )
 2
n  n b1
−Sinst − 8π
2
Λ
e = e g (µ) = (9.15)
µ
where in the last step we have used eq. (9.10). This shows that instantons are
inherently non-perturbative effects, since they vanish for Λ → 0, and are very weak,
if not negligible, in the perturbative regime.

9.2 Anomalies in a nutshell

Anomalies are classical symmetries of the action which are broken by quantum
effects. In other words, we have
quantum
∂µ j µ = 0 −−−−−−−−→ ∂µ j µ 6= 0 , (9.16)
corrections
where j µ is the current associated to the anomalous symmetry.
In what follows we will focus on chiral anomalies, that is anomalies associated to
chiral currents. These arise in field theories in which fermions with chiral symmetries
(symmetries under which opposite chirality fermions transform in complex conjugate
representations) are coupled to gauge fields. Let us consider, e.g., a Weyl fermion
ψα coupled to a gauge field Aµ with action
Z
S = d4 x ψ σ µ (∂µ + iAµ )ψ . (9.17)

This action is invariant under a U (1) global symmetry which rotates ψα → eiλ ψα
and ψ α̇ → e−iλ ψ α̇ . The corresponding conserved current can be computed by, e.g.

187
Nöether method. Considering a x-dependent transformation, that is λ = λ(x), we
get for the action
Z Z
S −→ S − d x ∂µ λ(x) · (ψσ µ ψ) = S + d4 x λ(x) · ∂µ (ψσ µ ψ) ,
4
(9.18)

where in the second step we have integrated by parts. Since for λ being a constant
this is a symmetry of the theory, it follows that ∂ µ (ψσ µ ψ) = 0. Actually, the current
jµ = ψσ µ ψ is conserved classically but it is not quantum mechanically. This can be
seen by computing loop diagrams involving three external currents.
Let us consider a set of free Weyl fermions transforming in some representation
R of some global symmetry group G and call jAµ the associated currents. Let us
compute the three-point function of the currents jA at one loop, diagram a of Figure
9.1. This will give something like

hjAµ (x1 )jBν (x2 )jCρ (x3 )i = Tr (tA tB tC ) fµνρ (xi ) , (9.19)

where the trace comes from contraction of the group generators around the loop.
As we are going to see in the following, this correlator has important properties but
it does not provide by itself any anomaly: the corresponding classical conservation
law is not violated quantum mechanically.

jB
<latexit sha1_base64="/70xllfzWZcJSlAvanxcOTZxAKY=">AAAB+3icbVDLSsNAFJ34rPUV69LNYBFclUQUXZa6cVnBPqAJYTKdtGMnM2FmIg0hv+LGhSJu/RF3/o3TNgttPXDhcM693HtPmDCqtON8W2vrG5tb25Wd6u7e/sGhfVTrKpFKTDpYMCH7IVKEUU46mmpG+okkKA4Z6YWT25nfeyJSUcEfdJYQP0YjTiOKkTZSYNceg9yLQzHNPU15BltFEdh1p+HMAVeJW5I6KNEO7C9vKHAaE64xQ0oNXCfRfo6kppiRouqliiQIT9CIDAzlKCbKz+e3F/DMKEMYCWmKazhXf0/kKFYqi0PTGSM9VsveTPzPG6Q6uvFzypNUE44Xi6KUQS3gLAg4pJJgzTJDEJbU3ArxGEmEtYmrakJwl19eJd2LhnvVcO4v681WGUcFnIBTcA5ccA2a4A60QQdgMAXP4BW8WYX1Yr1bH4vWNaucOQZ/YH3+AG6ElLA=</latexit>

VB VB
<latexit sha1_base64="GcQ3q2TZDmm040aVB8ov9zw1hco=">AAAB+3icbVDLSsNAFJ3UV62vWJduBovgqiSi6LLUjcsK9gFNCJPppB06MwkzE2kI+RU3LhRx64+482+ctllo64ELh3Pu5d57woRRpR3n26psbG5t71R3a3v7B4dH9nG9p+JUYtLFMYvlIESKMCpIV1PNyCCRBPGQkX44vZv7/SciFY3Fo84S4nM0FjSiGGkjBXa9F+QeD+NZ7mkqMtguisBuOE1nAbhO3JI0QIlOYH95oxinnAiNGVJq6DqJ9nMkNcWMFDUvVSRBeIrGZGioQJwoP1/cXsBzo4xgFEtTQsOF+nsiR1ypjIemkyM9UaveXPzPG6Y6uvVzKpJUE4GXi6KUQR3DeRBwRCXBmmWGICypuRXiCZIIaxNXzYTgrr68TnqXTfe66TxcNVrtMo4qOAVn4AK44Aa0wD3ogC7AYAaewSt4swrrxXq3PpatFaucOQF/YH3+AE7glJw=</latexit> <latexit sha1_base64="GcQ3q2TZDmm040aVB8ov9zw1hco=">AAAB+3icbVDLSsNAFJ3UV62vWJduBovgqiSi6LLUjcsK9gFNCJPppB06MwkzE2kI+RU3LhRx64+482+ctllo64ELh3Pu5d57woRRpR3n26psbG5t71R3a3v7B4dH9nG9p+JUYtLFMYvlIESKMCpIV1PNyCCRBPGQkX44vZv7/SciFY3Fo84S4nM0FjSiGGkjBXa9F+QeD+NZ7mkqMtguisBuOE1nAbhO3JI0QIlOYH95oxinnAiNGVJq6DqJ9nMkNcWMFDUvVSRBeIrGZGioQJwoP1/cXsBzo4xgFEtTQsOF+nsiR1ypjIemkyM9UaveXPzPG6Y6uvVzKpJUE4GXi6KUQR3DeRBwRCXBmmWGICypuRXiCZIIaxNXzYTgrr68TnqXTfe66TxcNVrtMo4qOAVn4AK44Aa0wD3ogC7AYAaewSt4swrrxXq3PpatFaucOQF/YH3+AE7glJw=</latexit>

jA jA + jA
<latexit sha1_base64="V/9k7UnXR0C7IyUNe8mLjcu2ys8=">AAAB+3icbVDLSsNAFJ34rPUV69LNYBFclUQUXVbduKxgH9CEMJlO2rEzkzAzkYaQX3HjQhG3/og7/8Zpm4W2HrhwOOde7r0nTBhV2nG+rZXVtfWNzcpWdXtnd2/fPqh1VJxKTNo4ZrHshUgRRgVpa6oZ6SWSIB4y0g3Ht1O/+0SkorF40FlCfI6GgkYUI22kwK49BrnHw3iSe5qKDF4XRWDXnYYzA1wmbknqoEQrsL+8QYxTToTGDCnVd51E+zmSmmJGiqqXKpIgPEZD0jdUIE6Un89uL+CJUQYwiqUpoeFM/T2RI65UxkPTyZEeqUVvKv7n9VMdXfk5FUmqicDzRVHKoI7hNAg4oJJgzTJDEJbU3ArxCEmEtYmrakJwF19eJp2zhnvRcO7P682bMo4KOALH4BS44BI0wR1ogTbAYAKewSt4swrrxXq3PuatK1Y5cwj+wPr8AWz+lK8=</latexit> <latexit sha1_base64="V/9k7UnXR0C7IyUNe8mLjcu2ys8=">AAAB+3icbVDLSsNAFJ34rPUV69LNYBFclUQUXVbduKxgH9CEMJlO2rEzkzAzkYaQX3HjQhG3/og7/8Zpm4W2HrhwOOde7r0nTBhV2nG+rZXVtfWNzcpWdXtnd2/fPqh1VJxKTNo4ZrHshUgRRgVpa6oZ6SWSIB4y0g3Ht1O/+0SkorF40FlCfI6GgkYUI22kwK49BrnHw3iSe5qKDF4XRWDXnYYzA1wmbknqoEQrsL+8QYxTToTGDCnVd51E+zmSmmJGiqqXKpIgPEZD0jdUIE6Un89uL+CJUQYwiqUpoeFM/T2RI65UxkPTyZEeqUVvKv7n9VMdXfk5FUmqicDzRVHKoI7hNAg4oJJgzTJDEJbU3ArxCEmEtYmrakJwF19eJp2zhnvRcO7P682bMo4KOALH4BS44BI0wR1ogTbAYAKewSt4swrrxXq3PuatK1Y5cwj+wPr8AWz+lK8=</latexit> <latexit sha1_base64="V/9k7UnXR0C7IyUNe8mLjcu2ys8=">AAAB+3icbVDLSsNAFJ34rPUV69LNYBFclUQUXVbduKxgH9CEMJlO2rEzkzAzkYaQX3HjQhG3/og7/8Zpm4W2HrhwOOde7r0nTBhV2nG+rZXVtfWNzcpWdXtnd2/fPqh1VJxKTNo4ZrHshUgRRgVpa6oZ6SWSIB4y0g3Ht1O/+0SkorF40FlCfI6GgkYUI22kwK49BrnHw3iSe5qKDF4XRWDXnYYzA1wmbknqoEQrsL+8QYxTToTGDCnVd51E+zmSmmJGiqqXKpIgPEZD0jdUIE6Un89uL+CJUQYwiqUpoeFM/T2RI65UxkPTyZEeqUVvKv7n9VMdXfk5FUmqicDzRVHKoI7hNAg4oJJgzTJDEJbU3ArxCEmEtYmrakJwF19eJp2zhnvRcO7P682bMo4KOALH4BS44BI0wR1ogTbAYAKewSt4swrrxXq3PuatK1Y5cwj+wPr8AWz+lK8=</latexit>

VC
<latexit sha1_base64="fgruEw5f48Vf2uPDZK056k8g0Q8=">AAAB+3icbVDLSsNAFJ3UV62vWJduBovgqiSi6LLYjcsK9gFNCJPppB06MwkzE2kI+RU3LhRx64+482+ctllo64ELh3Pu5d57woRRpR3n26psbG5t71R3a3v7B4dH9nG9p+JUYtLFMYvlIESKMCpIV1PNyCCRBPGQkX44bc/9/hORisbiUWcJ8TkaCxpRjLSRArveC3KPh/Es9zQVGWwXRWA3nKazAFwnbkkaoEQnsL+8UYxTToTGDCk1dJ1E+zmSmmJGipqXKpIgPEVjMjRUIE6Uny9uL+C5UUYwiqUpoeFC/T2RI65UxkPTyZGeqFVvLv7nDVMd3fo5FUmqicDLRVHKoI7hPAg4opJgzTJDEJbU3ArxBEmEtYmrZkJwV19eJ73LpnvddB6uGq27Mo4qOAVn4AK44Aa0wD3ogC7AYAaewSt4swrrxXq3PpatFaucOQF/YH3+AFBmlJ0=</latexit>

jC VC
<latexit sha1_base64="fgruEw5f48Vf2uPDZK056k8g0Q8=">AAAB+3icbVDLSsNAFJ3UV62vWJduBovgqiSi6LLYjcsK9gFNCJPppB06MwkzE2kI+RU3LhRx64+482+ctllo64ELh3Pu5d57woRRpR3n26psbG5t71R3a3v7B4dH9nG9p+JUYtLFMYvlIESKMCpIV1PNyCCRBPGQkX44bc/9/hORisbiUWcJ8TkaCxpRjLSRArveC3KPh/Es9zQVGWwXRWA3nKazAFwnbkkaoEQnsL+8UYxTToTGDCk1dJ1E+zmSmmJGipqXKpIgPEVjMjRUIE6Uny9uL+C5UUYwiqUpoeFC/T2RI65UxkPTyZGeqFVvLv7nDVMd3fo5FUmqicDLRVHKoI7hPAg4opJgzTJDEJbU3ArxBEmEtYmrZkJwV19eJ73LpnvddB6uGq27Mo4qOAVn4AK44Aa0wD3ogC7AYAaewSt4swrrxXq3PpatFaucOQF/YH3+AFBmlJ0=</latexit>

<latexit sha1_base64="tUg2P/00fDUXi2V1ZGUt6DfStlE=">AAAB+3icbVDLSsNAFJ34rPUV69LNYBFclUQUXRa7cVnBPqAJYTKdtGMnM2FmIg0hv+LGhSJu/RF3/o3TNgttPXDhcM693HtPmDCqtON8W2vrG5tb25Wd6u7e/sGhfVTrKpFKTDpYMCH7IVKEUU46mmpG+okkKA4Z6YWT1szvPRGpqOAPOkuIH6MRpxHFSBspsGuPQe7FoZjmnqY8g62iCOy603DmgKvELUkdlGgH9pc3FDiNCdeYIaUGrpNoP0dSU8xIUfVSRRKEJ2hEBoZyFBPl5/PbC3hmlCGMhDTFNZyrvydyFCuVxaHpjJEeq2VvJv7nDVId3fg55UmqCceLRVHKoBZwFgQcUkmwZpkhCEtqboV4jCTC2sRVNSG4yy+vku5Fw71qOPeX9eZtGUcFnIBTcA5ccA2a4A60QQdgMAXP4BW8WYX1Yr1bH4vWNaucOQZ/YH3+AHAKlLE=</latexit>

a b

Figure 9.1: One-loop diagrams contributing to correlators of one global current

with two global or two local currents. Diagram a does not provide any anomaly.
Diagrams b, instead, contribute to the anomaly of the global current jA .

Suppose now to gauge some (or all) global currents, by coupling the original
Lagrangian to gauge fields as
X µ
L = Lfree + AB
µ jB , (9.20)
B

188
and let us compute the correlator hjA AB AC i. The one-loop diagrams contributing
to such correlator are now diagrams b of Figure 9.1. By differentiating the result
one gets
∂µ jAµ ∼ Tr (tA {tB , tC }) FBµν FeC,µν , (9.21)
which says that the current is not anymore conserved (notice, on the contrary,
that as anticipated the rhs would vanish if none of the symmetries were gauged).
Remarkably, it turns out that this result is not corrected at higher loops, so the
anomaly can be entirely evaluated by a one-loop computation!
An important fact about anomalies is that local currents cannot be anomalous,
since they would imply violation of unitarity of the theory (we know how to couple
spin-1 fields in a way respecting unitarity to conserved currents, only). Hence a
quantum field theory, in order to make sense, should not have any gauge anomaly.
On the contrary, global chiral currents can be anomalous. So the currents jAµ in
eq. (9.21) should be global currents, since if this were not the case we would have
had a violation of unitarity in the quantum theory.
One other important thing one can learn looking at eq. (9.21) is that the anomaly
coefficient vanishes for real and pseudoreal representations. Indeed, for real or pseu-
doreal representations we have that tA = −(tA )T and it then easily follows that

Trr (tA {tB , tC }) = − Trr (tA {tB , tC }) . (9.22)

Therefore, only massless chiral fermions can contribute to the anomaly coefficient
(massive fermions always transform in the R + R representation, in order for the
mass term to be invariant under the symmetry). This result, once applied to local
currents, which cannot be anomalous, provides severe restrictions on the massless
fermion content of a quantum field theory.
Finally, from the above computation it is also clear that only abelian symmetries
can be anomalous. Suppose to have a theory with gauge group G with generators
tA , a global symmetry group G e with generators t̃A , and a set of Weyl fermions ψ i ,
α
transforming in the representations (ri , r̃i ) of the gauge and global symmetry groups,
respectively. In this case, the anomaly computation gives
X
∂µ jAµ ∼ Trr̃i t̃A Trri (tB tC ) FBµν FeCµν . (9.23)
i

Since Trr̃i t̃A = 0 for any simple algebra, only abelian factors U (1) ⊂ G e can be
anomalous. On the other hand, Trri (tB tC ) = C(ri )δBC , where C(ri ) is the quadratic

189
invariant (Casimir) of the representation ri . Working everything out (and paying
attention to numerical coefficients!), one finally gets for an abelian group
A µν eB
∂µ j µ = − F F , (9.24)
16π 2 B µν
P
where A = i qi C(ri ) is the anomaly coefficient, qi being the U (1) global charges
of fermion fields ψi .
This result makes it manifest the connection between anomalies and instantons.
Integrating eq. (9.24) in space-time we get
Z Z Z
4 µ 3 0 i


d x ∂µ j = dt d x ∂0 j − ∂i j = dt∂0 Q = ∆Q
Z
A µν eB
d4 x F F = 2An ,
16π 2 B µν
where in the first relation we have used the definition of charge Q and the fact
R
that d3 x∂i j i = 0, while in the second relation we have used eq. (9.7). Putting
everything together we finally get

|∆Q| = 2A n , (9.25)

where n is the instanton number and ∆Q the amount of charge violation due to the
anomaly. So we see that anomalous symmetries are violated by a specific amount,
given by eq. (9.25), in an instanton background. This also shows that anomalies are
IR effects, since the violation is very mild at weak coupling.
From eq. (9.18) one sees that the effect of the anomalous U (1) symmetry corre-
sponds to a shift in the θ-angle as

ψαi → eiqi λ ψαi =⇒ θYM → θYM − 2λA . (9.26)

since a shift in θYM generates the rhs of the anomalous conservation equation (9.24).
So the anomalous breaking can be seen as an explicit breaking: a term in the action,
the θ-term in fact, is not invariant under the anomalous symmetry.
Notice that if we perform a U (1) transformation but promote θYM to a spurion
field and assign to it transformation properties as to compensate for the shift, then
the anomalous U (1) is promoted to an actual symmetry of a larger theory (where
the complexified gauge coupling is promoted to a dynamical field). This symmetry,
however, is spontaneously broken by the coupling constant VEV (θYM in this case).
As we will later see, this way of looking at anomalous symmetries can be efficiently
used to put constraints on the construction of low-energy effective Lagrangians.

190
9.3 ’t Hooft anomaly matching condition
The correlator (9.19) does not provide any anomaly if only global currents are in-
volved. However, it does contain very important information, as originally pointed
out by ’t Hooft. This is because, as we review below, correlators as (9.19) compute
scale independent information about a quantum field theory and as such provide a
powerful tool to understand some of its non-perturbative properties.
Let us consider a Lagrangian L defined at some scale µ, with some non-anomalous
global symmetry group G generated by currents jAµ . Compute the triangle diagram
for three global currents (which is not an anomaly) and call AUV the result. Now
weakly gauge the global symmetry group G by adding new gauge fields AA µ and
define a new Lagrangian
1
L0 = L − TrFµν F µν + jAµ AA
µ . (9.27)
4g 2
This theory is inconsistent since it has a gauge anomaly, AUV , because we have
gauged G. We can make the theory consistent by adding some spectator free massless
fermion fields ψs (spectator in the sense that they couple only through the G-gauge
coupling) transforming in representations of G so to exactly cancel the anomaly, i.e.
As = −AUV . The resulting theory
1 µ


L00 = L − 2 TrFµν F µν + ψ s 6 ∂ ψs + js,A + jAµ AA
µ , (9.28)
4g
where js,A are the currents associated to the spectator fermions ψs , is non-anomalous,
and it is so for any value of the gauge coupling. Consider this anomaly-free theory at
some scale µ0 < µ. Since the spectator fermion fields and gauge fields can be made
arbitrarily weakly coupled by taking g → 0, the IR dynamics of the enlarged theory
(9.28) is just the IR dynamics of the original theory plus the arbitrarily weakly
coupled spectator theory. Therefore, As should be the same and since the theory is
anomaly free, we should have that AIR + As = 0, which implies

AIR = AUV . (9.29)

Taking g → 0 spectators fields completely decouple and (9.29) should still hold.
The punchline is that in a quantum field theory anomaly coefficients associated to
global currents are scale independent quantities, and their UV and IR values should
match. This is known as ’t Hooft anomaly matching condition.
A simple equation such as (9.29) puts severe constraints on the IR dynamics
of a quantum field theory, in particular as far as its massless spectrum: it implies

191
that a theory with global conserved currents but with ’t Hooft anomaly (that is,
a non-vanishing triangular anomaly associated to these global currents), does not
have a mass gap.
There exist two possible scenarios. If the global symmetry is preserved, a non-
vanishing ’t Hooft anomaly implies the existence of massless (typically composite)
fermions in the effective IR theory, so to match the anomaly. But it might hap-
pen that there does not exist any choice of quantum numbers for composite states
to match this anomaly. This suggests that the global symmetry is spontaneously
broken. But then the theory is again gapless since by Goldstone theorem massless
scalars, the goldstone bosons, are expected to exist. In this case one can match the
anomaly (9.29) by coupling the theory to background gauge fields and see that a
term in the effective action can (and should) be added which reproduces the UV
anomaly. This is known as gauged Wess-Zumino-Witten term. It is worth noting,
in passing, that it is precisely this latter argument which originally suggested that
the SU (3)L ×SU (3)R global symmetry of QCD should be spontaneously broken, the
pions being the corresponding (pseudo) Goldstone bosons. In that case, rather than
a background gauge field, an abelian symmetry is weakly gauged, which is nothing
but the electromagnetic U (1) gauge symmetry, and a term proportional to π 0 F µν Feµν
can and should be added to the pion Lagrangian which matches the anomaly (this
is also the way in which the puzzle of the π 0 → 2γ decay was solved).
Note that, quite remarkably, all these results can be claimed without doing any
sort of non-perturbative computation; just a one-loop one!

9.4 Holomorphy
We now want to discuss a property of supersymmetric theories, known as holomor-
phy, which plays a crucial role when it comes to understand the quantum properties
of supersymmetric theories and to what extent they differ from non-supersymmetric
ones.
Let us first briefly recall the concept of Wilsonian effective action.
When dealing with effective theories we deal with effective actions. The transi-
tion from a fundamental (bare) Lagrangian down to an effective one, involves inte-
grating out high-momentum degrees of freedom. The effective action (aka Wilsonian

192
action) is defined from the bare action Sµ0 defined at some UV scale µ0 , as
Z
iSµ
e = Dφ eiSµ0 (9.30)
φ(p) , p>µ

where Sµ is the effective action, of which we review below few basic properties.
The Wilsonian action correctly describes a theory’s degrees of freedom at energies
below a given scale µ (the cut-off). It is local on length scales larger than 1/µ, and
describes in a unitary way physical processes involving energy-momentum transfers
less than µ. As far as processes are concerned:

• at energies E ∼ µ, the effective action couplings and masses are given by the
tree-level couplings in the effective action (effects of all higher energy degrees
of freedom have already been integrated out),

• at energies E << µ there will be quantum corrections due to fluctuations of

modes of the fields in Sµ with energies between E and µ.

The upshot is that the Wilsonian action Sµ is the action which describes the physics
at the scale µ by its classical couplings.
Supersymmetry puts severe restrictions on the structure of the Wilsonian action,
more specifically to the superpotential (i.e., the F-term part). One way to see this,
is as follows. Any parameter in a supersymmetric Lagrangian can be thought of
as a VEV of a superfield. This implies, in particular, that each coupling (masses,
Yukawa couplings, etc...) appearing in the classical superpotential can be thought
of as the lowest component VEV of a (very heavy) chiral superfield (in other words,
the theory one is considering can be viewed as an effective theory of a bigger theory
where these fields have been integrated out and they act as spurions at low energy).
This implies that the superpotential is not only holomorphic in the fields but
also in the couplings and so is the effective superpotential in the Wilsonian action
(which is also holomporphic). The couplings of the effective action will be functions
of the couplings of the UV theory, and these should be holomorphic functions of
such UV couplings.
This important result can also be proven by means of supersymmetric Ward
identities. More specifically, they imply that all coupling constants appearing in the
tree level superpotential must only appear holomorphically in quantum corrections
to the superpotential (which is basically equivalent to what’s above).

193
 This property is important since, as we will see, promoting coupling constants
to chiral superfields one can often extend symmetries of the superpotential and put
severe constraints on the form (and sometime the very existence) of quantum correc-
tions. Holomorphy makes the restrictions on possible quantum corrections allowed
by supersymmetry apparent. It provides a supersymmetric version of selection rules.
In order to make this discussion more explicit, let us consider a concrete example.
Suppose we have a given supersymmetric theory and, following the logic outlined
previously, let us consider an enlarged symmetry group which includes a spurious
U (1) symmetry, associated to a coupling constant λ, which breaks this symmetry
spontaneously (a spurion) and has unit charge with respect to it, Q(λ) = 1. This
simply means that we have in the tree-level superpotential a term like

Wtree ⊃ λ O−1 , (9.31)

where the operator O has charge -1 with respect to the spurious U (1) symmetry.
Suppose we are interested in the appearance of a given operator of charge −10
among quantum corrections. In general, we expect it to appear at tenth or higher
order as
2
∆W ∼ λ10 O−10 + λ11 λ O−10 + · · · + λ10 e−1/|λ| O−10 , (9.32)
where we have assumed that the classical limit, λ → 0 is well defined and so we do
not allow any negative powers of λ to appear. Holomorphy implies that only the first
term can be generated. All other terms cannot be there since are non-holomorphic
in the coupling (both λ and λ appear). A corollary of the above discussion is that
any operator with positive U (1) charge is also disallowed. Indeed, we cannot have
negative powers of λ because we are supposing the theory is well defined in the
classical limit, while any power of λ is forbidden by holomorphy. Notice that the
latter property is due to supersymmetry, and it is not shared by an ordinary field
theory.
The basic message we want to convey is that holomorphy in the coupling con-
stants, usual selection rules for symmetries under which coupling constants may
transform and the requirement of smoothness of physics in various weak-coupling
limits, provide severe constraints on the structure of the effective superpotential of
a supersymmetric quantum field theory.
Let us close this section by recalling that the Wilsonian effective action is not
what we usually call the effective action Γ. The latter is obtained by integrating
out all degrees of freedom down to µ = 0 and it is the generating functional of

194
1PI graphs and calculates the Green functions of the original UV theory. It is not
holomorphic in the coupling constants and suffers from holomorphic anomalies. It is
not the correct thing to look at in asymptotically free gauge theories since it is not
well defined. The two effective actions are the same only if there are no interacting
massless particles, which are those making the 1PI effective action Γ suffer from IR
divergences.

9.5 Holomorphy and non-renormalization theorems

Using holomorphy one can prove many known non-renormalization theorems (and
go beyond them, as we will see).

Example 1: the Wess-Zumino (WZ) model. The tree level superpotential of the WZ
model has the following structure
1 1
Wtree = mΦ2 + λΦ3 . (9.33)
2 3
The question one might ask is: what is the form of the effective superpotential Weff ,
once quantum corrections (both perturbative and non-perturbative) are taken into
account? Let us try to answer this question using holomorphy. First, promote m
and λ to spurion superfields. This makes the theory enlarging its symmetries by a
flavor U (1) symmetry and a R-symmetry, according to the table below

U (1)R U (1)
Φ 1 1
(9.34)
m 0 −2
λ −1 −3

The superpotential has (correctly) R-charge 2 and flavor U (1) charge 0. Notice that
both symmetries are spontaneously broken whenever the spurion superfields, m and
λ, have a non-vanishing lower component VEV.
Because of what we discussed in the previous section, the effective (that is, exact)
superpotential should be a holomorphic function of Φ, m and λ, with R-charge equal
to 2 and flavor charge equal to 0. Its most general form can be written as a function
of λΦ/m as follows
  ∞
X
2 λΦ
Weff = mΦ f = an λn m1−n Φn+2 , (9.35)
m n=−∞

195
where ffree = 12 + 13 λΦ/m, and an are arbitrary coefficients. Note that due to holo-
morphicity of the superpotential neither λ nor m appear in (9.35).
The form of f can be fixed as follows. First, in the classical limit, λ → 0, we
should recover the tree level result. This implies that there cannot appear negative
powers of λ; hence n ≥ 0 and, in order to agree with (9.33) at tree-level, a0 = 12 and
a1 = 31 . Taking also the massless limit at the same time, m → 0, restricts n further,
i.e. n ≤ 1. The upshot is that the effective superpotential should be nothing but the
tree level one: holomorphy (plus some obvious physical requirements, more below)
tells us that the superpotential of the WZ model is not renormalized at any order
in perturbation theory and non-perturbatively!
The requirement about finiteness in the massless limit requires a few more com-
ments. Taking the massless limit at finite λ does not lead to a weakly-coupled
theory, so one could not use smoothness arguments so naively. However, taking
both m, λ → 0 such that m/λ → 0 we do achieve the result above, since the theory
is free in this case. One may still wonder whether this conclusion is correct since in
this limit there is a massless particle and so the effective theory should have some
IR divergences. This is not the case since we do not run the RG-flow down to µ = 0:
there are no IR divergences in the Wilsonian effective action, as opposed to the 1PI
effective action.
Another, equivalent way to see the absence of negative powers of m in the effec-
tive superpotential is to observe that all terms with n ≥ 0 are generated by tree-level
diagrams only, in the UV theory (it is a matter of number of vertices and propaga-
tors), see Figure 9.2 below. All diagrams of the kind of the one depicted in Figure

<latexit sha1_base64="bDO1vpc5sTYFTAs4PYLQyVF3DFQ=">AAAB63icbVBNS8NAEJ3Ur1q/qh69LBbBU0lE0WPRi8cI9gPaUDbbTbN0dxN2N0IJ/QtePCji1T/kzX/jps1BWx8MPN6bYWZemHKmjet+O5W19Y3Nrep2bWd3b/+gfnjU0UmmCG2ThCeqF2JNOZO0bZjhtJcqikXIaTec3BV+94kqzRL5aKYpDQQeSxYxgk0hDfyYDesNt+nOgVaJV5IGlPCH9a/BKCGZoNIQjrXue25qghwrwwins9og0zTFZILHtG+pxILqIJ/fOkNnVhmhKFG2pEFz9fdEjoXWUxHaToFNrJe9QvzP62cmuglyJtPMUEkWi6KMI5Og4nE0YooSw6eWYKKYvRWRGCtMjI2nZkPwll9eJZ2LpnfVdB8uG63bMo4qnMApnIMH19CCe/ChDQRieIZXeHOE8+K8Ox+L1opTzhzDHzifP+QLjiM=</latexit> <latexit sha1_base64="bDO1vpc5sTYFTAs4PYLQyVF3DFQ=">AAAB63icbVBNS8NAEJ3Ur1q/qh69LBbBU0lE0WPRi8cI9gPaUDbbTbN0dxN2N0IJ/QtePCji1T/kzX/jps1BWx8MPN6bYWZemHKmjet+O5W19Y3Nrep2bWd3b/+gfnjU0UmmCG2ThCeqF2JNOZO0bZjhtJcqikXIaTec3BV+94kqzRL5aKYpDQQeSxYxgk0hDfyYDesNt+nOgVaJV5IGlPCH9a/BKCGZoNIQjrXue25qghwrwwins9og0zTFZILHtG+pxILqIJ/fOkNnVhmhKFG2pEFz9fdEjoXWUxHaToFNrJe9QvzP62cmuglyJtPMUEkWi6KMI5Og4nE0YooSw6eWYKKYvRWRGCtMjI2nZkPwll9eJZ2LpnfVdB8uG63bMo4qnMApnIMH19CCe/ChDQRieIZXeHOE8+K8Ox+L1opTzhzDHzifP+QLjiM=</latexit> <latexit sha1_base64="bDO1vpc5sTYFTAs4PYLQyVF3DFQ=">AAAB63icbVBNS8NAEJ3Ur1q/qh69LBbBU0lE0WPRi8cI9gPaUDbbTbN0dxN2N0IJ/QtePCji1T/kzX/jps1BWx8MPN6bYWZemHKmjet+O5W19Y3Nrep2bWd3b/+gfnjU0UmmCG2ThCeqF2JNOZO0bZjhtJcqikXIaTec3BV+94kqzRL5aKYpDQQeSxYxgk0hDfyYDesNt+nOgVaJV5IGlPCH9a/BKCGZoNIQjrXue25qghwrwwins9og0zTFZILHtG+pxILqIJ/fOkNnVhmhKFG2pEFz9fdEjoXWUxHaToFNrJe9QvzP62cmuglyJtPMUEkWi6KMI5Og4nE0YooSw6eWYKKYvRWRGCtMjI2nZkPwll9eJZ2LpnfVdB8uG63bMo4qnMApnIMH19CCe/ChDQRieIZXeHOE8+K8Ox+L1opTzhzDHzifP+QLjiM=</latexit> <latexit sha1_base64="bDO1vpc5sTYFTAs4PYLQyVF3DFQ=">AAAB63icbVBNS8NAEJ3Ur1q/qh69LBbBU0lE0WPRi8cI9gPaUDbbTbN0dxN2N0IJ/QtePCji1T/kzX/jps1BWx8MPN6bYWZemHKmjet+O5W19Y3Nrep2bWd3b/+gfnjU0UmmCG2ThCeqF2JNOZO0bZjhtJcqikXIaTec3BV+94kqzRL5aKYpDQQeSxYxgk0hDfyYDesNt+nOgVaJV5IGlPCH9a/BKCGZoNIQjrXue25qghwrwwins9og0zTFZILHtG+pxILqIJ/fOkNnVhmhKFG2pEFz9fdEjoXWUxHaToFNrJe9QvzP62cmuglyJtPMUEkWi6KMI5Og4nE0YooSw6eWYKKYvRWRGCtMjI2nZkPwll9eJZ2LpnfVdB8uG63bMo4qnMApnIMH19CCe/ChDQRieIZXeHOE8+K8Ox+L1opTzhzDHzifP+QLjiM=</latexit>

1 1
<latexit sha1_base64="/H9TIqYR8ku3U3dEofnm/sRI+NE=">AAAB7XicdVDLSgNBEOz1GeMr6tHLYBC8uOzGSOIt6MVjBPOAZA2zk0kyZmZ2mZkVwpJ/8OJBEa/+jzf/xskLVLSgoajqprsrjDnTxvM+naXlldW19cxGdnNre2c3t7df11GiCK2RiEeqGWJNOZO0ZpjhtBkrikXIaSMcXk38xgNVmkXy1oxiGgjcl6zHCDZWqou79NQfd3J5zy1clL2zMpqRYmlBCsh3vSnyMEe1k/todyOSCCoN4Vjrlu/FJkixMoxwOs62E01jTIa4T1uWSiyoDtLptWN0bJUu6kXKljRoqn6fSLHQeiRC2ymwGejf3kT8y2slplcOUibjxFBJZot6CUcmQpPXUZcpSgwfWYKJYvZWRAZYYWJsQFkbwuJT9D+pF1z/3PVuivnK5TyODBzCEZyADyWowDVUoQYE7uERnuHFiZwn59V5m7UuOfOZA/gB5/0LpUKPMA==</latexit> <latexit sha1_base64="/H9TIqYR8ku3U3dEofnm/sRI+NE=">AAAB7XicdVDLSgNBEOz1GeMr6tHLYBC8uOzGSOIt6MVjBPOAZA2zk0kyZmZ2mZkVwpJ/8OJBEa/+jzf/xskLVLSgoajqprsrjDnTxvM+naXlldW19cxGdnNre2c3t7df11GiCK2RiEeqGWJNOZO0ZpjhtBkrikXIaSMcXk38xgNVmkXy1oxiGgjcl6zHCDZWqou79NQfd3J5zy1clL2zMpqRYmlBCsh3vSnyMEe1k/todyOSCCoN4Vjrlu/FJkixMoxwOs62E01jTIa4T1uWSiyoDtLptWN0bJUu6kXKljRoqn6fSLHQeiRC2ymwGejf3kT8y2slplcOUibjxFBJZot6CUcmQpPXUZcpSgwfWYKJYvZWRAZYYWJsQFkbwuJT9D+pF1z/3PVuivnK5TyODBzCEZyADyWowDVUoQYE7uERnuHFiZwn59V5m7UuOfOZA/gB5/0LpUKPMA==</latexit>

<latexit sha1_base64="bDO1vpc5sTYFTAs4PYLQyVF3DFQ=">AAAB63icbVBNS8NAEJ3Ur1q/qh69LBbBU0lE0WPRi8cI9gPaUDbbTbN0dxN2N0IJ/QtePCji1T/kzX/jps1BWx8MPN6bYWZemHKmjet+O5W19Y3Nrep2bWd3b/+gfnjU0UmmCG2ThCeqF2JNOZO0bZjhtJcqikXIaTec3BV+94kqzRL5aKYpDQQeSxYxgk0hDfyYDesNt+nOgVaJV5IGlPCH9a/BKCGZoNIQjrXue25qghwrwwins9og0zTFZILHtG+pxILqIJ/fOkNnVhmhKFG2pEFz9fdEjoXWUxHaToFNrJe9QvzP62cmuglyJtPMUEkWi6KMI5Og4nE0YooSw6eWYKKYvRWRGCtMjI2nZkPwll9eJZ2LpnfVdB8uG63bMo4qnMApnIMH19CCe/ChDQRieIZXeHOE8+K8Ox+L1opTzhzDHzifP+QLjiM=</latexit>
m m <latexit sha1_base64="bDO1vpc5sTYFTAs4PYLQyVF3DFQ=">AAAB63icbVBNS8NAEJ3Ur1q/qh69LBbBU0lE0WPRi8cI9gPaUDbbTbN0dxN2N0IJ/QtePCji1T/kzX/jps1BWx8MPN6bYWZemHKmjet+O5W19Y3Nrep2bWd3b/+gfnjU0UmmCG2ThCeqF2JNOZO0bZjhtJcqikXIaTec3BV+94kqzRL5aKYpDQQeSxYxgk0hDfyYDesNt+nOgVaJV5IGlPCH9a/BKCGZoNIQjrXue25qghwrwwins9og0zTFZILHtG+pxILqIJ/fOkNnVhmhKFG2pEFz9fdEjoXWUxHaToFNrJe9QvzP62cmuglyJtPMUEkWi6KMI5Og4nE0YooSw6eWYKKYvRWRGCtMjI2nZkPwll9eJZ2LpnfVdB8uG63bMo4qnMApnIMH19CCe/ChDQRieIZXeHOE8+K8Ox+L1opTzhzDHzifP+QLjiM=</latexit>

<latexit sha1_base64="XP5+6ZtEJA9uow/G8pwibH5AnOo=">AAAB7nicdVDLSsNAFL2pr1pfUZduBovgKiS10rorunFZwT6gDWUymbRDJ5MwMxFK6Ee4caGIW7/HnX/j9AUqemDgcM65zL0nSDlT2nU/rcLa+sbmVnG7tLO7t39gHx61VZJJQlsk4YnsBlhRzgRtaaY57aaS4jjgtBOMb2Z+54FKxRJxrycp9WM8FCxiBGsjdfrcREM8sMuuU7mquxd1tCDV2opUkOe4c5RhiebA/uiHCcliKjThWKme56baz7HUjHA6LfUzRVNMxnhIe4YKHFPl5/N1p+jMKCGKEmme0Giufp/IcazUJA5MMsZ6pH57M/Evr5fpqO7nTKSZpoIsPooyjnSCZrejkElKNJ8YgolkZldERlhiok1DJVPC6lL0P2lXHO/Sce+q5cb1so4inMApnIMHNWjALTShBQTG8AjP8GKl1pP1ar0togVrOXMMP2C9fwG5UY/W</latexit> <latexit sha1_base64="XP5+6ZtEJA9uow/G8pwibH5AnOo=">AAAB7nicdVDLSsNAFL2pr1pfUZduBovgKiS10rorunFZwT6gDWUymbRDJ5MwMxFK6Ee4caGIW7/HnX/j9AUqemDgcM65zL0nSDlT2nU/rcLa+sbmVnG7tLO7t39gHx61VZJJQlsk4YnsBlhRzgRtaaY57aaS4jjgtBOMb2Z+54FKxRJxrycp9WM8FCxiBGsjdfrcREM8sMuuU7mquxd1tCDV2opUkOe4c5RhiebA/uiHCcliKjThWKme56baz7HUjHA6LfUzRVNMxnhIe4YKHFPl5/N1p+jMKCGKEmme0Giufp/IcazUJA5MMsZ6pH57M/Evr5fpqO7nTKSZpoIsPooyjnSCZrejkElKNJ8YgolkZldERlhiok1DJVPC6lL0P2lXHO/Sce+q5cb1so4inMApnIMHNWjALTShBQTG8AjP8GKl1pP1ar0togVrOXMMP2C9fwG5UY/W</latexit> <latexit sha1_base64="XP5+6ZtEJA9uow/G8pwibH5AnOo=">AAAB7nicdVDLSsNAFL2pr1pfUZduBovgKiS10rorunFZwT6gDWUymbRDJ5MwMxFK6Ee4caGIW7/HnX/j9AUqemDgcM65zL0nSDlT2nU/rcLa+sbmVnG7tLO7t39gHx61VZJJQlsk4YnsBlhRzgRtaaY57aaS4jjgtBOMb2Z+54FKxRJxrycp9WM8FCxiBGsjdfrcREM8sMuuU7mquxd1tCDV2opUkOe4c5RhiebA/uiHCcliKjThWKme56baz7HUjHA6LfUzRVNMxnhIe4YKHFPl5/N1p+jMKCGKEmme0Giufp/IcazUJA5MMsZ6pH57M/Evr5fpqO7nTKSZpoIsPooyjnSCZrejkElKNJ8YgolkZldERlhiok1DJVPC6lL0P2lXHO/Sce+q5cb1so4inMApnIMHNWjALTShBQTG8AjP8GKl1pP1ar0togVrOXMMP2C9fwG5UY/W</latexit> <latexit sha1_base64="XP5+6ZtEJA9uow/G8pwibH5AnOo=">AAAB7nicdVDLSsNAFL2pr1pfUZduBovgKiS10rorunFZwT6gDWUymbRDJ5MwMxFK6Ee4caGIW7/HnX/j9AUqemDgcM65zL0nSDlT2nU/rcLa+sbmVnG7tLO7t39gHx61VZJJQlsk4YnsBlhRzgRtaaY57aaS4jjgtBOMb2Z+54FKxRJxrycp9WM8FCxiBGsjdfrcREM8sMuuU7mquxd1tCDV2opUkOe4c5RhiebA/uiHCcliKjThWKme56baz7HUjHA6LfUzRVNMxnhIe4YKHFPl5/N1p+jMKCGKEmme0Giufp/IcazUJA5MMsZ6pH57M/Evr5fpqO7nTKSZpoIsPooyjnSCZrejkElKNJ8YgolkZldERlhiok1DJVPC6lL0P2lXHO/Sce+q5cb1so4inMApnIMHNWjALTShBQTG8AjP8GKl1pP1ar0togVrOXMMP2C9fwG5UY/W</latexit>

Figure 9.2: Tree level (super)graph producing terms of the series (9.35) for n > 1.

9.2 are not 1PI for n > 1; they cannot be produced from loops, and they should
not be included in the effective action for finite m. So the integer n in eq. (9.35)

196
is indeed restricted to be either 0 or 1. It is easy to get convinced that in a non-
supersymmetric context, where e.g. λ can enter the expansion (9.35), operators with
negative powers of λ can come form 1PI diagrams and hence appear in the effective
action.
What we have just proven, namely that the superpotential of the WZ model
is not renormalized at any order in perturbation theory and non-perturbatively, is
not specific to the WZ model. It actually applies to all models where only chiral
superfields are present: in these cases, that is in the absence of gauge interactions,
the tree-level superpotential is an exact quantity (regardless its specific form, which
might even contain classically irrelevant operators!).

Example 2: As a second example, we want to illustrate what holomorphy can tell

us about the running of gauge couplings in supersymmetric gauge theories. Let us
focus, for definiteness, on SQCD. Recall that this is a supersymmetric gauge theory
e
with gauge group SU (N ), F flavors described by F pairs of chiral superfields (Q, Q)
transforming in the fundamental respectively anti-fundamental representation of
the gauge group and no tree-level superpotential. At the classical level, the global
symmetries are as detailed below

SU (F )L SU (F )R U (1)B U (1)A U (1)R0

Qia F • 1 1 a
ebj
Q • F −1 1 a
λ • • 0 0 1

where the convention on indices is the same as in lecture 5, see the discussion
below eq.(5.104), and the R-charges of Q and Qe are the same since under charge
conjugation (which commutes with supersymmetry), Q ↔ Q. e For later convenience,
we have also written down the charges of the gaugino field. The axial current and
the R0 current are anomalous, the anomaly coefficients being
1 1
AA = (+1)F + (+1)F = F
2 2
1
AR0 = [(a − 1)F + (a − 1)F ] + N = N + (a − 1)F
2
These two anomalous symmetries admit an anomaly-free combination (which is
obviously an R-symmetry) with current

(1 − a)F − N A
jµR = jµR0 + jµ , (9.36)
F
197
under which the matter fields have the following charges R(Qia ) = R(Q eb ) = F −N
j F
(note that the anomaly-free R-charge of matter fields does not depend on a, while the
gaugino has always R-charge equal to 1). Therefore, the group of continuos global
symmetries at the quantum level is GF = SU (F )L × SU (F )R × U (1)B × U (1)R .
Notice that for F = 0, namely for pure SYM, there do not exist an axial current
and in turn the R-symmetry is inevitably anomalous. This difference will play a
crucial role later on.
What we are interested in is the gauge coupling running, namely the β function
∂g b1 3
β=µ =− 2
g + O(g 5 ) . (9.37)
∂µ 16π
The one-loop coefficient b1 can be easily computed from the field content of the
classical Lagrangian and reads b1 = 3N −F . The question we would like to answer is
whether holomorphy can tell us something about higher-loop (and non-perturbative)
corrections.
Let us consider pure SYM, first, whose action is
Z
1
L= d2 θ τ Tr W α Wα + h.c. , (9.38)
16πi

with τ the complexified gauge coupling, τ = θY2πM + 4πig2

. Notice that τ appears
holomorphically in the action above, but the gauge fields are not canonically nor-
malized (to go to a basis where gauge fields are canonically normalized one should
shift V → gV , as we did already when constructing matter-coupled actions).
One thing that will prove useful for what we want to do, is to trade the dynamical
generated scale Λ for a complex parameter. For G = SU (N ) the one-loop running
of the gauge coupling is
 
1 3N |Λ| − 8π
2

= − log , |Λ| = µ 0 e 3N g 2 (µ0 )

, (9.39)
g 2 (µ) 8π 2 µ

where 3N is the one-loop coefficient of pure SYM β-function and |Λ| what we pre-
viously called Λ. We can then define a holomorphic intrinsic scale Λ as
θYM 2πiτ
Λ = |Λ|ei 3N = µe 3N , (9.40)

in terms of which the one-loop complexified gauge coupling reads

3N Λ
τ1-loop = log . (9.41)
2πi µ

198
What about higher order corrections? Suppose we integrate down to a scale µ, then
τ (Λ; µ)
Weff = Tr W α Wα . (9.42)
16πi
Since the physics is periodic under θYM → θYM + 2π, the following rescaling
2πi
Λ → e 3N Λ (9.43)

is a symmetry of the theory, under which τ → τ + 1. The most general form for τ
transforming this way under (9.43) is
3N Λ
τ (Λ; µ) = log + f (Λ; µ) , (9.44)
2πi µ
with f a holomorphic function of Λ having the following properties:

• f should have a positive Taylor expansion in Λ in such a way that in the limit
Λ → 0, which is a classical limit, we get back the one-loop result.

• Plugging the transformation (9.43) into the expression (9.41) shows that τ1-loop
already accounts for the shift of θ-angle by 2π. Hence, the function f should
be invariant under (9.43).

These two properties imply that the effective coupling (9.44) should have the fol-
lowing form
∞  3N n
3N Λ X Λ
τ (Λ; µ) = log + an . (9.45)
2πi µ n=1 µ
Recalling that the instanton action is
 3N
−Sinst Λ
e = , (9.46)
µ
we conclude that the function f receives only non-perturbative corrections and these
corrections come from n-instantons contributions. The upshot is that τ is one-loop
exact, in perturbation theory.
The one-loop exactness of the SYM gauge coupling can be equivalently proven
as follows. The θ-term is a topological term so it does not get renormalized pertur-
batively. Therefore the β-function, β = β(τ ) can only involve Im τ . If β should be
a holomorphic function of τ this implies that it can only be a imaginary constant (a
holomorphic function f (z), which is independent of Re z, is an imaginary constant).
Therefore
d
β(τ ) ≡ µ τ = ia , (9.47)
dµ

199
which implies
d d a
µ θYM = 0 , µ g = − g 3 . (9.48)
dµ dµ 8π
So we see that, indeed, the gauge coupling does not receive corrections beyond
one-loop, in perturbation theory (for the theory at hand a = 3N/2π).
All what we said above applies identically to SQCD (again, working in the ba-
sis where gauge fields are not canonically normalized and the complexified gauge
coupling enters holomorphically in the action), the only difference being that the
one-loop coefficient of the β-function is now 3N − F , with F being the number of
flavors.
Remarkably, in some specific cases one can show that also non-perturbative cor-
rections are absent. One such instances is pure SYM, and the argument goes as
follows. As already noticed, the R-symmetry of pure SYM is anomalous
quantum 2N a eµν
∂µ jRµ = 0 −−−−−−−−→ ∂µ jRµ = − F F . (9.49)
corrections 32π 2 µν a
The U (1)R , however, is not fully broken. This can be seen as follows. A R-symmetry
transformation with parameter α, under which the gaugino transforms as

λ → eiα λ (9.50)

is equivalent to a shift of the θ-angle

θYM → θYM − 2N α , (9.51)

recall eqs. (9.24) and (9.26). The point is that the transformation θYM → θYM + 2πk
where k ∈ Z, is a symmetry of the theory. So, whenever the U (1)R parameter α
equals πk/N , the theory is unchanged also at the quantum level. This implies that
a discrete subgroup of the original continuos abelian symmetry is preserved,

U (1)R −→ Z2N . (9.52)

Treating the complexified gauge coupling τ as a spurion field and assigning to it

transformation properties so to compensate for the shift, we can define a spurious
symmetry acting as
Nα
λ → eiα λ , τ → τ + . (9.53)
π
This constrains the coefficients an in the expansion (9.45). Indeed, under the spuri-
2πiτ
ous symmetry the holomorphic scale Λ = µ e 3N transforms as
2N α
Λ → ei 3N Λ. (9.54)

200
Hence we have
∞  3N n
N α 3N Λ X Λ
τ (Λ; µ) → + log + an e2iN αn . (9.55)
π 2πi µ n=1 µ

Since whenever n 6= 0 then e2iN αn 6= 0, it follows that to match the spurious sym-
metry (9.53) we need to have

an = 0 ∀n > 0 . (9.56)

Hence in pure SYM also non-perturbative corrections to the gauge coupling are
absent! This does not hold in presence of matter, namely for SQCD, since there the
R-symmetry is not anomalous and running the above argument one would not get
any constraint on the coefficients an (the rhs of eq. (9.49) would be zero in this case
and θY M would be insensitive to R-symmetry transformations).
If we collect all what we have learned so far we might have the feeling that some-
thing wrong is going on. There are three apparently incompatible results regarding
the running of the SQCD gauge coupling.

• Due to holomorphy, the supersymmetric gauge coupling runs only at one-loop

in perturbation theory, and the full perturbative β-function hence reads
g3
β=− (3N − F ) , (9.57)
16π 2
• There exists a well known result in the literature which claims that the exact,
all-loops β-function of SQCD is
h PF i
g 3 3N − i=1 (1 − γi )
β=− 2 2 (9.58)
16π 1 − N g2
8π

where γi = d log Zi (µ)/d log µ are matter fields anomalous dimensions. This
result gets contribution at all loops and is in clear contradiction with the
previous result. Eq. (9.58) is sometime called the NSVZ β-function.

• Another piece of knowledge we have about the β-function of SQCD (and,

in general, of any gauge theory) is that its one and two-loop coefficients are
universal, in the sense that are renormalization scheme independent. This
can be easily proven as follows. Changing renormalization scheme amounts to
define a new coupling g 0 which is related to g as

g 0 = g + ag 3 + O(g 5 ) . (9.59)

201
 Suppose that the β-function for g is

βg = b1 g 3 + b2 g 5 + O(g 7 ) . (9.60)

We get for the β-function for g 0

∂g 0

βg0 = βg = βg 1 + 3ag 2 + O(g 4 ) = b1 g 3 + (b2 + 3ab1 )g 5 + O(g 7 ) . (9.61)
∂g
We can invert the relation between g and g 0 and get

g = g 0 − ag 03 + O(g 05 ) , (9.62)

and finally

βg0 = b1 g 03 − 3ab1 g 05 + (b2 + 3ab1 ) g 05 + O(g 07 ) = b1 g 03 + b2 g 05 + O(g 07 ) , (9.63)

which shows that the first two coefficients of the β-function are indeed univer-
sal. Given the universality of the β-function up to two loops, the discrepancy
between the two expressions for the SQCD β-function (9.57) and (9.58), which
only agree at one loop, cannot just be a matter of renormalization scheme.

How can we reconcile this apparent contradiction? The answer turns out to be
surprisingly simple. Let us first consider pure SYM whose action is
Z
1
L= d2 θ τ Tr W α Wα + h.c. . (9.64)
16πi
As we already noticed, if one integrates in superspace one gets a space-time action
where gauge fields are not canonically normalized
1
L=− Tr Fµν F µν + . . . . (9.65)
4g 2
Let us call the gauge coupling defined in this frame holomorphic gauge coupling
gh , defined via the complexified gauge coupling as τ = 4πi/gh2 . In order to get a
Lagrangian in terms of canonically normalized fields one should rescale the vector
superfield V as V → gV . In other words, we should perform the change of variables
Vh = gp Vp . In terms of this physical gauge coupling gp the Lagrangian reads
Z  
1 2 1 θYM
L= dθ − i 2 Tr W α (gp Vp )Wα (gp Vp ) + h.c. , (9.66)
4 gp2 8π

Notice that the Lagrangian above is not holomorphic in the physical coupling since gp
is real as gp Vp should also be real. The crucial point now is that the two Lagrangians

202
(9.64) and (9.66) are not equivalent under the change of variables Vh = gp Vp in the
path integral, since there is a rescaling anomaly (there is an anomalous Jacobian in
passing from Vh to Vp ), that is D(gp Vp ) 6= DVp . In particular, one can show that
 Z   
i 2 2T (Adj) α
D(gp Vp ) = DVp exp − dθ log gp Tr W (gp Vp )Wα (gp Vp ) + h.c. .
4 8π 2
(9.67)
Hence we get for the partition function
Z  Z 
i 2 1 α
Z = DVh exp d θ 2 Tr W (Vh )Wα (Vh ) + h.c. = [Vh = gp Vp ]
4 gh
Z  Z   
i 2 1 2T (Adj) α
= DVp exp dθ − log gp Tr W (gp Vp )Wα (gp Vp ) + h.c. ,
4 gh2 8π 2
(9.68)

which implies
   
1 1 2T (Adj) 1 2N
= Re − log gp = Re − log gp . (9.69)
gp2 gh2 8π 2 gh2 8π 2

where in the last equality we used the fact that for SU (N ) the Dynkin index for the
adjoint T (Adj) = N . Differentiating with respect to log µ, and using the expression
(9.57) for the holomorphic gauge coupling β-function, one gets for the physical gauge
coupling gp precisely the NSVZ β-function (9.58), where in both expressions we have
set F = 0.
One can repeat an identical reasoning for SQCD where the relation between the
physical and the holomorphic gauge couplings reads
  X T (ri )
1 1 2T (Adj)
= Re − log g p − log Zi . (9.70)
gp2 gh2 8π 2 i
8π 2

Differentiating with respect to log µ (using again T (Adj) = N and taking matter
to be in the fundamental, for which T (r) = 1/2), one gets for the physical gauge
coupling exactly the expression (9.58).
We now see why there is no contradiction with the two-loops universality of
the β-function. The point is simply that the relation between the holomorphic and
the physical gauge coupling is not analytic. In other words, one cannot be Taylor-
expanded in the other, because of the log-term (it is a singular change of renor-
malization scheme: the so-called holomorphic scheme is not related continuously to
any other physical renormalization scheme). Furthermore, we now also understand

203
where higher-loop contributions to the physical β-function come from. This is just
because of wave function renormalization (both of the vector superfield as well as of
matter superfields): the physical gauge coupling differs from the holomorphic gauge
coupling by effects coming from wave-function renormalization, which get contribu-
tion at all loops. Consistently, the physical β-function can be expressed exactly in
terms of anomalous dimension of fields, once the one-loop coefficient (which agrees
with that of the holomorphic β-function) has been calculated.
One can repeat the same kind of reasoning for gauge theories with extended
supersymmetry which, after all, are just (very) special cases of N = 1 theories.
In doing so one immediately gets the result we anticipated when discussing non-
renormalization theorems in section 6.3. Let us start from N = 2 pure SYM. Using
N = 1 language we have a vector and a chiral multiplet, the latter transforming in
the adjoint of the gauge group. As we have already seen, due to N = 2 supersym-
metry, the kinetic terms of V and Φ are both changed according to the holomorphic
gauge coupling. Hence, going to canonical normalization for all fields we must rescale
them the same way, Vh = gp Vp and Φh = gp Φp . The crucial point is that the Jacobian
for V cancels exactly that from Φ! In other words, D(gp Vp )D(gp Φp ) = DVp DΦp ,
implying that the holomorphic and physical gauge couplings coincide (basically, the
second and third terms in (9.70) cancel each other). Adding matter nothing changes
since, as we have already noticed, kinetic terms for hypermultiplets do not renor-
malize. Hence, N = 2 gauge theories, with and without matter, are (perturbatively)
one-loop finite. Applying this result to N = 4 SYM we conclude that the latter is
tree-level exact, since the N = 4 one-loop β-function vanishes. It’s that simple!

9.6 Holomorphic decoupling

Holomorphy helps also in getting effective superpotentials when one has to integrate
out some massive modes and study the theory at scales lower than the corresponding
mass scale.
Let us consider a model of two chiral superfields, L and Φ, interacting via the
following superpotential
1 λ
W = M Φ2 + L2 Φ , (9.71)
2 2
This is a Wess-Zumino-like model, so this superpotential does not suffer from quan-
tum corrections, neither perturbatively nor non-perturbatively. The spectrum of
the theory is that of a massless chiral superfield and a massive one. If we want to

204
study the system at energies µ < M , we have to integrate Φ out. In order to do
so we can use holomorphicity arguments, and proceed as we did when proving the
exactness of the WZ superpotential. Let us first promote the couplings to spurion
fields and, consequently, enlarge the global symmetries as follows

U (1)a U (1)b U (1)R

L 0 1 1
Φ 1 0 0 (9.72)
M −2 0 2
λ −1 −2 0

The low energy effective superpotential, in which Φ should not enter, should be
a dimension-three function of λ, M and L respecting the above symmetries. In
particular it should have U (1)a,b charge 0 and U (1)R t charge 2. The only possible
answer is
λ2 L4
Weff = a , (9.73)
M
where a is an undetermined constant of order one.
The same result can be obtained by using the ordinary integrating out procedure.
At scales well below M , the chiral superfield Φ is frozen at its VEV (we do not have
enough energy to make it fluctuate). Therefore, we can integrate the field out by
solving its equation of motion, which is just an algebraic one, involving only the F-
term, since the kinetic term (the D-term) is trivially zero. The equation of motion
is
∂W λ λ 2
= M Φ + L2 = 0 −→ Φ = − L . (9.74)
∂Φ 2 2M
Substituting back into the superpotential we get

1 λ2 L4
Weff = − , (9.75)
8 M
which is the same as (9.73) (with the undetermined coefficient being fixed). Notice
that the superpotential we have just obtained is the effective superpotential one
generates in perturbation theory, in the limit of small λ, see Figure 9.3.
Let us emphasize that in this case, differently from the WZ model discussed in the
previous section, we have allowed for negative powers of M . In other words, we have
not required any smoothness in the M → 0 limit. The reason is that the effective
theory we are considering is valid only at energies lower than M , which is a UV
cut-off for the theory. Hence, we can accept (and actually do expect) singularities

205
 L L
<latexit sha1_base64="3N15HF9oM7jv7AKL9DiB8jk0nE8=">AAAB6HicbVA9SwNBEJ3zM8avqKXNYhCswp0oWgZtLCwSMB+QHGFvM5es2ds7dveEcOQX2FgoYutPsvPfuEmu0MQHA4/3ZpiZFySCa+O6387K6tr6xmZhq7i9s7u3Xzo4bOo4VQwbLBaxagdUo+ASG4Ybge1EIY0Cga1gdDv1W0+oNI/lgxkn6Ed0IHnIGTVWqt/3SmW34s5AlomXkzLkqPVKX91+zNIIpWGCat3x3MT4GVWGM4GTYjfVmFA2ogPsWCpphNrPZodOyKlV+iSMlS1pyEz9PZHRSOtxFNjOiJqhXvSm4n9eJzXhtZ9xmaQGJZsvClNBTEymX5M+V8iMGFtCmeL2VsKGVFFmbDZFG4K3+PIyaZ5XvMuKW78oV2/yOApwDCdwBh5cQRXuoAYNYIDwDK/w5jw6L8678zFvXXHymSP4A+fzB6TbjNQ=</latexit> <latexit sha1_base64="3N15HF9oM7jv7AKL9DiB8jk0nE8=">AAAB6HicbVA9SwNBEJ3zM8avqKXNYhCswp0oWgZtLCwSMB+QHGFvM5es2ds7dveEcOQX2FgoYutPsvPfuEmu0MQHA4/3ZpiZFySCa+O6387K6tr6xmZhq7i9s7u3Xzo4bOo4VQwbLBaxagdUo+ASG4Ybge1EIY0Cga1gdDv1W0+oNI/lgxkn6Ed0IHnIGTVWqt/3SmW34s5AlomXkzLkqPVKX91+zNIIpWGCat3x3MT4GVWGM4GTYjfVmFA2ogPsWCpphNrPZodOyKlV+iSMlS1pyEz9PZHRSOtxFNjOiJqhXvSm4n9eJzXhtZ9xmaQGJZsvClNBTEymX5M+V8iMGFtCmeL2VsKGVFFmbDZFG4K3+PIyaZ5XvMuKW78oV2/yOApwDCdwBh5cQRXuoAYNYIDwDK/w5jw6L8678zFvXXHymSP4A+fzB6TbjNQ=</latexit>

1
<latexit sha1_base64="2mnTciJmdL87cHHJ2Rvs9DOW7kM=">AAAB7XicdVDLSgNBEOyNrxhfUY9eBoPgxWU3RhJvQS9ehAjmAckaZiezyZjZnWVmVghL/sGLB0W8+j/e/BsnL1DRgoaiqpvuLj/mTGnH+bQyS8srq2vZ9dzG5tb2Tn53r6FEIgmtE8GFbPlYUc4iWtdMc9qKJcWhz2nTH15O/OYDlYqJ6FaPYuqFuB+xgBGsjdS4vktP3HE3X3Ds4nnFOa2gGSmVF6SIXNuZogBz1Lr5j05PkCSkkSYcK9V2nVh7KZaaEU7HuU6iaIzJEPdp29AIh1R56fTaMToySg8FQpqKNJqq3ydSHCo1Cn3TGWI9UL+9ifiX1050UPFSFsWJphGZLQoSjrRAk9dRj0lKNB8Zgolk5lZEBlhiok1AORPC4lP0P2kUbffMdm5KherFPI4sHMAhHIMLZajCFdSgDgTu4RGe4cUS1pP1ar3NWjPWfGYffsB6/wJ0Io8Q</latexit>

M
<latexit sha1_base64="XP5+6ZtEJA9uow/G8pwibH5AnOo=">AAAB7nicdVDLSsNAFL2pr1pfUZduBovgKiS10rorunFZwT6gDWUymbRDJ5MwMxFK6Ee4caGIW7/HnX/j9AUqemDgcM65zL0nSDlT2nU/rcLa+sbmVnG7tLO7t39gHx61VZJJQlsk4YnsBlhRzgRtaaY57aaS4jjgtBOMb2Z+54FKxRJxrycp9WM8FCxiBGsjdfrcREM8sMuuU7mquxd1tCDV2opUkOe4c5RhiebA/uiHCcliKjThWKme56baz7HUjHA6LfUzRVNMxnhIe4YKHFPl5/N1p+jMKCGKEmme0Giufp/IcazUJA5MMsZ6pH57M/Evr5fpqO7nTKSZpoIsPooyjnSCZrejkElKNJ8YgolkZldERlhiok1DJVPC6lL0P2lXHO/Sce+q5cb1so4inMApnIMHNWjALTShBQTG8AjP8GKl1pP1ar0togVrOXMMP2C9fwG5UY/W</latexit> <latexit sha1_base64="XP5+6ZtEJA9uow/G8pwibH5AnOo=">AAAB7nicdVDLSsNAFL2pr1pfUZduBovgKiS10rorunFZwT6gDWUymbRDJ5MwMxFK6Ee4caGIW7/HnX/j9AUqemDgcM65zL0nSDlT2nU/rcLa+sbmVnG7tLO7t39gHx61VZJJQlsk4YnsBlhRzgRtaaY57aaS4jjgtBOMb2Z+54FKxRJxrycp9WM8FCxiBGsjdfrcREM8sMuuU7mquxd1tCDV2opUkOe4c5RhiebA/uiHCcliKjThWKme56baz7HUjHA6LfUzRVNMxnhIe4YKHFPl5/N1p+jMKCGKEmme0Giufp/IcazUJA5MMsZ6pH57M/Evr5fpqO7nTKSZpoIsPooyjnSCZrejkElKNJ8YgolkZldERlhiok1DJVPC6lL0P2lXHO/Sce+q5cb1so4inMApnIMHNWjALTShBQTG8AjP8GKl1pP1ar0togVrOXMMP2C9fwG5UY/W</latexit>

L L
<latexit sha1_base64="3N15HF9oM7jv7AKL9DiB8jk0nE8=">AAAB6HicbVA9SwNBEJ3zM8avqKXNYhCswp0oWgZtLCwSMB+QHGFvM5es2ds7dveEcOQX2FgoYutPsvPfuEmu0MQHA4/3ZpiZFySCa+O6387K6tr6xmZhq7i9s7u3Xzo4bOo4VQwbLBaxagdUo+ASG4Ybge1EIY0Cga1gdDv1W0+oNI/lgxkn6Ed0IHnIGTVWqt/3SmW34s5AlomXkzLkqPVKX91+zNIIpWGCat3x3MT4GVWGM4GTYjfVmFA2ogPsWCpphNrPZodOyKlV+iSMlS1pyEz9PZHRSOtxFNjOiJqhXvSm4n9eJzXhtZ9xmaQGJZsvClNBTEymX5M+V8iMGFtCmeL2VsKGVFFmbDZFG4K3+PIyaZ5XvMuKW78oV2/yOApwDCdwBh5cQRXuoAYNYIDwDK/w5jw6L8678zFvXXHymSP4A+fzB6TbjNQ=</latexit> <latexit sha1_base64="3N15HF9oM7jv7AKL9DiB8jk0nE8=">AAAB6HicbVA9SwNBEJ3zM8avqKXNYhCswp0oWgZtLCwSMB+QHGFvM5es2ds7dveEcOQX2FgoYutPsvPfuEmu0MQHA4/3ZpiZFySCa+O6387K6tr6xmZhq7i9s7u3Xzo4bOo4VQwbLBaxagdUo+ASG4Ybge1EIY0Cga1gdDv1W0+oNI/lgxkn6Ed0IHnIGTVWqt/3SmW34s5AlomXkzLkqPVKX91+zNIIpWGCat3x3MT4GVWGM4GTYjfVmFA2ogPsWCpphNrPZodOyKlV+iSMlS1pyEz9PZHRSOtxFNjOiJqhXvSm4n9eJzXhtZ9xmaQGJZsvClNBTEymX5M+V8iMGFtCmeL2VsKGVFFmbDZFG4K3+PIyaZ5XvMuKW78oV2/yOApwDCdwBh5cQRXuoAYNYIDwDK/w5jw6L8678zFvXXHymSP4A+fzB6TbjNQ=</latexit>

Figure 9.3: The tree level (super)graph which produces the effective superpotential
(9.73) in the weak coupling limit.

as we send M to zero: new massless degrees of freedom are expected to arise when
M → 0. They are those associated to Φ, the superfield we have integrated out.

As a final instructive example, let us consider a perturbation of the previous

model. The superpotential we would like to analyze is
1 λ 
W = M Φ2 + L2 Φ + Φ3 . (9.76)
2 2 6
Again, if we want to study the system at energies µ < M , we have to integrate the
massive field out. The equation of motion for Φ gives
r !
M λL2
Φ=− 1∓ 1− . (9.77)
 M2

We have now two possible solutions, hence two different vacua. Consistently, as we
send  to zero one of the two vacua approaches the one of the unperturbed model
while the second is pushed all the way to infinity. Indeed
r !   
M λL2 M λL2 2
Φ+ = − 1+ 1− =− 1+ 1− + O( ) −−→ ∞
 M2  2M 2 →0

r !   
M λL2 M λL2 2 λ 2
Φ− = − 1− 1− 2
=− 1− 1− 2
+ O( ) −−→ − L
 M  2M →0 2M

Substituting (9.77) into (9.76) we get for the effective superpotential

"  r #
M3 3 λL2 λL2 λL2
Weff = 2 1 − ∓ 1− 1− . (9.78)
3 2 M2 M2 M2

There are again singularities, both in parameter space as well as in field L space,
now. Comparing to the unperturbed case one can suspect that at these points extra

206
massless degrees of freedom show up. Indeed, computing the (effective) mass for the
field we have integrated out, we get
r
∂ 2W λL2
Meff = = M + Φ = (on the solution) = ±M 1 − . (9.79)
∂Φ2 M2
√
The field Φ becomes massless at hLi = ±M/ λ, precisely the two singularities of
the effective superpotential (9.78). In the limit  → 0, keeping M fixed, one recovers,
again, the result of the unperturbed theory.
As in previous example, the same result could have been obtained just using
holomorphicity arguments. Promoting also the coupling  to a spurion field with
charges
U (1)a U (1)b U (1)R
(9.80)
 −3 0 2
and repeating the same argument as in the previous example, one could conclude
that the effective superpotential should have the following structure
 
M3 λL2
Weff = 2 f , (9.81)
 M2

which has precisely the structure of the exact expression (9.78). Taking various
limits one can actually fix also the form of the function f and get the expression
(9.78) (modulo an overall numerical coefficient, as before).
This way of integrating out in supersymmetric theories, which preserves holomor-
phy, is called holomorphic decoupling. We will heavily use holomorphic decoupling
when studying the quantum properties of SQCD in our next lecture. For instance,
using this technique it is possible to get the effective superpotential for SQCD with
an arbitrary number of flavors once the exact expression for a given number of flavors
is known. Everything amounts to integrate flavors in and out (more later).

9.7 Exercises
1. Using holomorphy and (spurious) symmetries, show that the superpotential

W = µ1 Φ + µ2 Φ2 + · · · + µn Φn , (9.82)

is not renormalized at any order in perturbation theory and non-perturbatively.

207
References

[1] S. Coleman, Aspects of Symmetry: Selected Erice Lectures -The uses of Instan-
tons, Chapter 7, Cambridge University Press (1985).

[2] M. A. Shifman and A. I. Vainshtein, Instantons versus supersymmetry: Fifteen

years later, In *Shifman, M.A.: ITEP lectures on particle physics and field
theory, vol. 2* 485-647 [hep-th/9902018].

[3] J. Terning, Modern supersymmetry: Dynamics and duality, Chapters 7 and 8,

Oxford University Press (2006).

[4] P. Argyres 1, An Introduction to Global Supersymmetry, Sections 2.3.2 and

2.3.3, http://www.physics.uc.edu/%7Eargyres/661/susy2001.pdf
P. Argyres 2, Introduction to Supersymmetry, Sections 7 and 14,
http://www.physics.uc.edu/%7Eargyres/661/susy1996.pdf

[5] N. Dorey, T. J. Hollowood, V. V. Khoze and M. P. Mattis, The Calculus of many

instantons, Phys. Rept. 371 (2002), 231-459 [arXiv:hep-th/0206063 [hep-th]].

[6] N. Seiberg, Naturalness Versus Supersymmetric Non-renormalization Theo-

rems, Phys. Lett. B 318 (1993) 469 [arXiv:hep-ph/9309335].

[7] N. Seiberg, Exact Results On The Space Of Vacua Of Four-Dimensional Susy

Gauge Theories, Phys. Rev. D 49 (1994) 6857 [arXiv:hep-th/9402044].

[8] N. Seiberg, The Power of holomorphy: Exact results in 4-D SUSY field theories,
arXiv:hep-th/9408013.

[9] V. A. Novikov, M. A. Shifman, A. I. Vainshtein and V. I. Zakharov, Exact Gell-

Mann-Low Function of Supersymmetric Yang-Mills Theories from Instanton
Calculus, Nucl. Phys. B 229 (1983), 381-393

[10] N. Arkani-Hamed and H. Murayama, Holomorphy, rescaling anomalies and

exact beta functions in supersymmetric gauge theories, JHEP 0006 (2000) 030
[arXiv:hep-th/9707133].

208
10 Supersymmetric gauge dynamics: N = 1

Given a quantum field theory, the very first questions one should answer regard
the way its symmetries are realized in its vacua, and what the dynamics around
such vacua is.

• Given a QFT with gauge group G and global symmetry group GF , how are
these realized in the vacuum?

• Which phases may enjoy such a theory?

• Are there tools to give not only qualitative but also quantitative answers to
these questions?

It is very difficult to fully or even partially answer these questions, in general. How-
ever, as we will discuss in the remainder of this course, for supersymmetric theories
this is possible, sometime. Before discussing any specific supersymmetric field theory
in detail, though, there are a number of general remarks we want to make regarding
the low energy behavior of field theories, more specifically asymptotically free gauge
theories. This will help to better appreciate what comes next.

10.1 Confinement and mass gap in QCD, YM and SYM

It is often said that asymptotically free gauge theories (e.g., QCD) enjoy many in-
teresting and fascinating phenomena at low energy, like confinement, the generation
of mass gap, chiral symmetry breaking etc... This is certainly true, but it turns out
that such phenomena may be realized very differently, for different theories. Below
we are going to consider three specific theories, namely QCD, YM and N = 1 SYM,
which are all UV-free and all said to be confining, and show how different the IR
dynamics of these theories actually is.

• QCD, the theory of strong interactions.

At high energy QCD is a weakly coupled theory, a SU (3) gauge theory of weakly
interacting quarks and gluons. It grows, through renormalization effects, to become
strong in low energy processes. So strong so to bind quarks into nucleons. The
strong coupling scale of QCD is

ΛQCD ∼ 300 MeV . (10.1)

209
Note that as compared to protons and neutrons (whose mass is order 1 GeV),
constituent quarks are relatively light (the u and d quarks are order of a few MeV;
the s quark is order 100 MeV). Most of the mass of nucleons comes from quark
kinetic energy and the interactions binding quarks together.
The reason why we cannot see free quarks, we usually say, is confinement: quarks
are bound into nucleons and cannot escape. In fact, this statement is not completely
correct: if we send an electron deep into a proton, we can make the quark escape!
If the electron is energetic enough, a large amount of energy, in the form of
chromoelectric field, appears in the region between the escaping quark and the rest
of the proton. When the field becomes strong enough, of order Λ4QCD ∼ (300 MeV)4 ,
flux lines can break and produce q − q pairs (this is a familiar phenomenon also in
electromagnetism: electric fields beyond a certain magnitude cannot survive; strong
fields with energy density bigger that m4e ∼ (1 Mev)4 decay by producing e+ − e−
pairs). The q quark binds to the escaping quark while the q quark binds to the
other two quarks in the proton. Therefore, the original quark does escape, the force
between it and the remaining proton constituent drops to zero. Just, the escaping
quark is not alone, it is bound into a meson. This should be better called charge
screening, rather than confinement.

strong chromoelectric field q - q̄ pair production

<latexit sha1_base64="sMjUtFiWns1026JYw2rVGf+3OPY=">AAACDXicbVC7SgNBFJ2Nrxhfq5Y2i1GwCruiKKQJ2lhGMA/IhjA7uZsMmccyMyuGJT9g46/YWChia2/n3zh5FJp4YOBwzr3cOSdKGNXG97+d3NLyyupafr2wsbm1vePu7tW1TBWBGpFMqmaENTAqoGaoYdBMFGAeMWhEg+ux37gHpakUd2aYQJvjnqAxJdhYqeMehTySD5k2SopeWCZ9JbkEBsQoSsJyTIF1Rx236Jf8CbxFEsxIEc1Q7bhfYVeSlIMwhGGtW4GfmHaGlaGEwagQphoSTAa4By1LBeag29kkzcg7tkrXi6WyTxhvov7eyDDXesgjO8mx6et5byz+57VSE1+2MyqS1IAg00NxyjwjvXE1Xpcqm5sNLcFEUftXj/SxwsTYAgu2hGA+8iKpn5aC85J/e1asXM3qyKMDdIhOUIAuUAXdoCqqIYIe0TN6RW/Ok/PivDsf09GcM9vZR3/gfP4A/cacww==</latexit>

<latexit sha1_base64="lIxnisOsLMoBP6HSXogPEL4DOcg=">AAACFXicbVDNS8MwHE39nPOr6tFLcAge5mhFUfAy9OJxgvuAtYw0TbewtKlJKo7Sf8KL/4oXD4p4Fbz535huFXTzQcLjvfcj+T0vZlQqy/oy5uYXFpeWSyvl1bX1jU1za7sleSIwaWLOuOh4SBJGI9JUVDHSiQVBocdI2xte5n77jghJeXSjRjFxQ9SPaEAxUlrqmdVbJ/T4fepUD51q5nhIwB/lPEZU6FtwP8F5OuuZFatmjQFniV2QCijQ6Jmfjs9xEpJIYYak7NpWrNwUCUUxI1nZSSSJER6iPulqGqGQSDcdb5XBfa34MOBCn0jBsfp7IkWhlKPQ08kQqYGc9nLxP6+bqODMTWkUJ4pEePJQkDCoOMwrgj4VBCs20gRhQfVfIR4ggbDSRZZ1Cfb0yrOkdVSzT2rW9XGlflHUUQK7YA8cABucgjq4Ag3QBBg8gCfwAl6NR+PZeDPeJ9E5o5jZAX9gfHwDFuGfbA==</latexit>

Nucleon Nucleon Meson

<latexit sha1_base64="NpIQc0CLaksfHLuUFiL6IN71l4w=">AAAB9XicbVDLSgMxFL1TX7W+qi7dBIvgqsyIosuiG1dSwT6gHUsmzbSheQxJRi1D/8ONC0Xc+i/u/BvTx0JbD1w4nHNvcu+JEs6M9f1vL7e0vLK6ll8vbGxube8Ud/fqRqWa0BpRXOlmhA3lTNKaZZbTZqIpFhGnjWhwNfYbD1QbpuSdHSY0FLgnWcwItk66b4tIPWU3KeFUyVGnWPLL/gRokQQzUoIZqp3iV7urSCqotIRjY1qBn9gww9oy9+Ko0E4NTTAZ4B5tOSqxoCbMJluP0JFTuihW2pW0aKL+nsiwMGYoItcpsO2beW8s/ue1UhtfhBmTSWqpJNOP4pQjq9A4AtRlmhLLh45gopnbFZE+1phYF1TBhRDMn7xI6ifl4Kzs356WKpezOPJwAIdwDAGcQwWuoQo1IKDhGV7hzXv0Xrx372PamvNmM/vwB97nDxxWkug=</latexit> <latexit sha1_base64="NpIQc0CLaksfHLuUFiL6IN71l4w=">AAAB9XicbVDLSgMxFL1TX7W+qi7dBIvgqsyIosuiG1dSwT6gHUsmzbSheQxJRi1D/8ONC0Xc+i/u/BvTx0JbD1w4nHNvcu+JEs6M9f1vL7e0vLK6ll8vbGxube8Ud/fqRqWa0BpRXOlmhA3lTNKaZZbTZqIpFhGnjWhwNfYbD1QbpuSdHSY0FLgnWcwItk66b4tIPWU3KeFUyVGnWPLL/gRokQQzUoIZqp3iV7urSCqotIRjY1qBn9gww9oy9+Ko0E4NTTAZ4B5tOSqxoCbMJluP0JFTuihW2pW0aKL+nsiwMGYoItcpsO2beW8s/ue1UhtfhBmTSWqpJNOP4pQjq9A4AtRlmhLLh45gopnbFZE+1phYF1TBhRDMn7xI6ifl4Kzs356WKpezOPJwAIdwDAGcQwWuoQo1IKDhGV7hzXv0Xrx372PamvNmM/vwB97nDxxWkug=</latexit> <latexit sha1_base64="1d2eigHdrm1X/96cpKz//XPUixk=">AAAB83icbVDLSgNBEOyNrxhfUY9eBoPgKeyKosegFy9CBPOA7BJmJ5NkyDyWmVkxLPkNLx4U8erPePNvnCR70MSChqKqm+6uOOHMWN//9gorq2vrG8XN0tb2zu5eef+gaVSqCW0QxZVux9hQziRtWGY5bSeaYhFz2opHN1O/9Ui1YUo+2HFCI4EHkvUZwdZJYShi9ZTdUaPkpFuu+FV/BrRMgpxUIEe9W/4Ke4qkgkpLODamE/iJjTKsLSOcTkphamiCyQgPaMdRiQU1UTa7eYJOnNJDfaVdSYtm6u+JDAtjxiJ2nQLboVn0puJ/Xie1/asoYzJJLZVkvqifcmQVmgaAekxTYvnYEUw0c7ciMsQaE+tiKrkQgsWXl0nzrBpcVP3780rtOo+jCEdwDKcQwCXU4Bbq0AACCTzDK7x5qffivXsf89aCl88cwh94nz+JoJIC</latexit>

a b c d

Figure 10.1: Charge screening: the way QCD confines.

Can we have confinement in a more strict sense? Suppose that quarks were
much more massive, say mq ∼ 1 TeV. Now proton mass would be order the TeV.
The dynamics drastically changes, now. Repeating the previous experiment, when
the chromoelectric field becomes order Λ4QCD , there is not enough energy now to
produce q − q pairs. The force between the escaping quark and the proton goes to
a constant: a tube of chromoelectric flux of thickness ∼ Λ−1 QCD and tension (energy
2
per unit length) of order ΛQCD connects the two. Not only the quark is confined,
it is the flux itself which is confined. This is certainly a more precise definition of
confinement: it holds regardless of quarks, in the sense that it holds also in the

210
limit mq → ∞, namely when the quarks disappear from the spectrum (they become
chromoelectric static sources and play no role in the dynamics). It is a property of
the pure glue. Strict confinement would be a property of QCD only if the quarks
were very massive, more precisely in the limit F/N << 1, where F is the number
of light quarks and N the number of colors. Real-life quarks are light enough to
let the chromoelectric flux tube break. Hence, actual QCD does not confine in the
strict sense.
Let us discuss the structure of QCD vacua in more detail, looking at how the
global symmetry group of QCD is realized in the vacuum. In what follows, we
consider only the three light quarks, u, d and s and forget the other ones (which are
dynamically less important). So we have F = 3 flavors. Moreover, we will first put
ourselves in the limit where the light quarks are massless, which is approximately
true for u, d and s quarks constituting ordinary matter (protons and neutrons). Only
later we will consider the effect of the small quark masses. In this massless limit the
QCD Lagrangian reads
1 X X
L = − Tr Fµν F µν + q iL i Dq
/ Li + q iR i Dq
/ Ri , i = 1, 2, 3 . (10.2)
4 i i

Quark quantum numbers under the global symmetry group are

SU (3)L SU (3)R U (1)A U (1)B
qL 3 1 1 1 (10.3)
qR 1 3̄ 1 −1
As well known there is an axial anomaly, in the sense that the U (1)A symmetry is
broken to Z2F at the quantum level. Therefore, the continuos global symmetries at
the quantum level are just

GF = SU (3)L × SU (3)R × U (1)B . (10.4)

As we have already discussed, the theory undergoes confinement (or better charge
screening) and quarks and gluons are bounded into color singlet states. But what
about GF ? Experimental and theoretical considerations plus several numerical sim-
ulations on the lattice lead to a definite picture of the realization of the global
symmetry at low energy. It is believed that at low energy only a subgroup survives

SU (3)D × U (1)B , (10.5)

under which hadrons are classified: SU (3)D gives the flavor quantum numbers and
U (1)B is the baryon charge. The remaining generators must be broken, somehow.

211
The intuitive picture is as follows. Due to confinement, at strong coupling quarks
and anti-quarks are bound into pairs, and the vacuum is filled by a condensate of
these color singlet quark bilinears

hqLi qRj i = ∆ δ ij , (10.6)

where ∆ ∼ Λ3QCD . This condensate is invariant under a diagonal SU (3) subgroup

of the original SU (3)R × SU (3)L group and is then responsible for the spontaneous
breaking of the chiral symmetry of the original symmetry group GF

SU (3)L × SU (3)R × U (1)B → SU (3)D × U (1)B . (10.7)

Eight global symmetries are broken by the quark condensates and hence we would
expect eight Goldstone bosons. The latter are indeed observed experimentally, and
0,+
correspond to the eight pseudoscalar mesons, the pions, π 0,± , K 0,1 , K , η

π + = ud , π − = du , π 0 = dd − uu , η = uu + dd − 2ss
0 + (10.8)
K 0 = sd , K − = us , K = sd , K = su

Let us first notice that if U (1)A were not anomalous, we would have had a ninth
meson, the so-called η 0 meson, which would have corresponded to a shift in the phase
of the condensate (10.6). The condensate breaks spontaneously the Z2F symmetry
down to Z2 , but this does not give rise to any massless particle. The η 0 has a periodic
potential with F minima, each of them being Z2 invariant, and related one another
by ZF rotations. These minima are not isolated, though, since they are connected
via SU (F )L × SU (F )R rotations. This means that there is a moduli space of vacua.
In Figure 10.2 a qualitative picture of QCD vacuum structure is reported.
Via a SU (3) rotation acting separately on qL and qR , the condensate (10.6) can
be put in the form
hqLi qRj i = ∆ U ij (10.9)
where U ij is a SU (3) matrix on which a SU (3)L × SU (3)R rotation acts as

U → A†L U AR , (10.10)

which shows there exists a SU (3)D rotation (AL = AR ) under which the matrix U
is invariant. So the moduli space of vacua is a SU (3) manifold.
The quantum fluctuations of the entries of this matrix represent the massless ex-
citations around the vacua of massless QCD, the pions. An effective Lagrangian for

212
 Z3
<latexit sha1_base64="GJcmt3eIuIoxq72/sriizggct4o=">AAAB83icbVDLSgMxFL1TX7W+qi7dBIvgqsz4QJdFNy4r2Ad2hpJJM21oJhmSjFCG/oYbF4q49Wfc+Tdm2llo64HA4Zx7uScnTDjTxnW/ndLK6tr6RnmzsrW9s7tX3T9oa5kqQltEcqm6IdaUM0FbhhlOu4miOA457YTj29zvPFGlmRQPZpLQIMZDwSJGsLGS78fYjMIwe5z2z/vVmlt3Z0DLxCtIDQo0+9UvfyBJGlNhCMda9zw3MUGGlWGE02nFTzVNMBnjIe1ZKnBMdZDNMk/RiVUGKJLKPmHQTP29keFY60kc2sk8o170cvE/r5ea6DrImEhSQwWZH4pSjoxEeQFowBQlhk8swUQxmxWREVaYGFtTxZbgLX55mbTP6t5l3b2/qDVuijrKcATHcAoeXEED7qAJLSCQwDO8wpuTOi/Ou/MxHy05xc4h/IHz+QP6tZGk</latexit>

(SU (3)L ⇥ SU (3)R )/SU (3)D

<latexit sha1_base64="UOv4rSp6iKXEG8xa56ec+8rXxhE=">AAACCXicdVDLSgMxFM3UV62vUZdugkVoN2PGVqy7oi5cuKiPPqAtQyZN29DMgyQjlKFbN/6KGxeKuPUP3Pk3pp0RVPRA4Nxz7uXmHjfkTCqEPozM3PzC4lJ2Obeyura+YW5uNWQQCULrJOCBaLlYUs58WldMcdoKBcWey2nTHZ1O/eYtFZIF/o0ah7Tr4YHP+oxgpSXHhIXreqFUdC5gRzGPSpiUV8X9hJw5Zh5ZpWONCkRWGSFko5SgErQtNEMepKg55nunF5DIo74iHEvZtlGoujEWihFOJ7lOJGmIyQgPaFtTH+ut3Xh2yQTuaaUH+4HQz1dwpn6fiLEn5dhzdaeH1VD+9qbiX147Uv1KN2Z+GCnqk2RRP+JQBXAaC+wxQYniY00wEUz/FZIhFpgoHV5Oh/B1KfyfNA4s+9BCl+V89SSNIwt2wC4oABscgSo4BzVQBwTcgQfwBJ6Ne+PReDFek9aMkc5sgx8w3j4BTVGW6w==</latexit>

Figure 10.2: The vacuum structure of 3-flavor massless QCD. The η 0 particle is
massive. The three vacua associated to the periodic potential along the η 0 direction
are rotated one another by the broken Z3 generators, but are not isolated since
there are flat directions connecting them, associated to eight massless excitations
(the pions) which parametrize the coset (SU (3)L × SU (F )R )/SU (3)D .

such excitations can be written in terms of U (x) and its derivatives. This Lagrangian
should be invariant under the full global symmetry group GF , hence non-derivative
terms are not allowed (the only GF -invariant function would be U † U = 1), which is
simply saying that the pions are massless in the massless QCD limit we are consid-
ering. The structure of the effective Lagrangian hence reads


Leff = fπ2 ∂µ U † ∂ µ U + κ ∂µ U † ∂ µ U ∂ν U † ∂ ν U + . . . , (10.11)

where κ ∼ 1/M 2 , with M being some intrinsic mass scale of the theory, and traces
on flavor indices are understood. At low momenta only the first term contributes
and we then get a definite prediction for pion scattering amplitudes, in terms of a
single parameter fπ .
In fact, quarks are not exactly massless and therefore the above picture is only
approximate. In reality, the SU (3)L × SU (3)R symmetry is only approximate since
quark masses correspond to (weak) GF -breaking terms. This has the effect to make
the pions be only pseudo-Goldstone bosons. Hence, one would expect them to be
massive, though pretty light, and this is indeed what we observe in Nature.
In principle, one should have gotten the chiral Lagrangian (10.11) from the UV
Lagrangian (10.2) by integrating out high momentum modes. This is difficult (next-
to impossible, in fact). However, we know in advance the expression (10.11) to
be right, since that is the most general Lagrangian one can write describing pion

213
dynamics and respecting the original symmetries of the problem. Combining the
expression (10.11) with weak GF -breaking terms induced by actual quark masses,
one gets a Lagrangian which, experimentally, does a good job.
Summarizing, we see that combining symmetry arguments, lattice simulations,
experimental observations and some physical reasoning, we can reach a rather rea-
sonable understanding of the low energy dynamics of QCD. This is all very nice but
one would like to gain, possibly, a theoretical (i.e. more microscopic) understanding
of QCD phenomena. As of today, this is still an open question for QCD. And, more
generally, it is so for any generic gauge theory. As we will later see, supersymmetry
lets one have more analytical tools to answer this kind of questions, having some-
time the possibility to derive strong coupling phenomena like confinement, chiral
symmetry breaking and the generation of a mass gap, from first principles.

• Yang-Mills (YM) theory, gauge interactions without matter fields.

Let us focus, for definiteness, on YM theory with gauge group SU (N ). This is
again a UV-free theory, the one-loop β-function being
 
g3 11
βg = − N . (10.12)
16π 2 3

There are two claims about this theory, coming from a combination of experimental
and theoretical reasoning, analytic and lattice calculations.

1. The theory has a mass gap, i.e. there are no massless fields in the spectrum.
Rather, there is a discrete set of states with masses of order Λ, the scale where
the one-loop gauge coupling diverges (higher loop and non-perturbative effects
do not change the actual value of Λ in any sensible way)
8π 2 11
−
g 2 (µ)b1
Λ = µe where b1 = N . (10.13)
3
The low energy spectrum consists of glueballs. These are sort of gluons bound
states which however do not consist of a fixed number of gluons (gluon number
is not a conserved quantum number in strong interactions), but rather of a
shifting mass of chromoelectric flux lines. Unlike gluons, for which a mass
term is forbidden (because they have only two polarizations), glueballs include
scalars and vectors with three polarizations (as well as higher spin particle
states), for which a mass term is allowed. Such mass should clearly be of order
of the dynamical scale, m ∼ Λ, so not to contradict perturbation theory.

214
 The low energy spectrum is very different from QCD. In QCD there is a mass
gap just because quarks are massive. If u, d and s quarks were massless, we
would not have had a mass gap in QCD since pions would have been exact
Goldstone bosons and hence massless. Here instead there is a genuine mass
gap.

2. The theory undergoes confinement (now in the strict sense). The chromoelec-
tric flux is confined, it cannot spread out in space over regions larger than
about Λ−1 in radius. How can we see confinement, namely the presence of
strings which contain the chromoelectric flux? Let us add some heavy quarks
to the theory and let us see whether these quarks are confined, as it was the
case for very massive QCD. The Lagrangian would read
1
L = − Tr F µν Fµν + iψDψ
/ − M ψψ , M >> Λ . (10.14)
4
In the limit M → ∞ the test particles become chromoelectrostatic sources,
and play no role in the dynamics.
If confinement occurs, we would expect a linear potential between the two
quarks. Indeed, in an unconfined theory, the electric flux is uniformly dis-
tributed over a sphere surrounding a charge, and falls-off as 1/r2 . In a con-
fining theory with flux tubes, the flux tube has a fixed cross-sectional area
∼ Λ−2 . Thus, for any sphere of radius r >> R ≡ Λ−1 , the flux is zero except
in a region of area Λ−2 , see Figure 10.3.

Area = ⇡R2
<latexit sha1_base64="zTzjV2iwz3QLPn/OoPZ4QP7W9Qo=">AAAB/nicbVDLSgMxFM3UV62vUXHlJlgEV2WmKLoRqm5cVrEP6Iwlk2ba0DyGJCOWoeCvuHGhiFu/w51/Y/pYaOuBwOGce7g3J0oY1cbzvp3cwuLS8kp+tbC2vrG55W7v1LVMFSY1LJlUzQhpwqggNUMNI81EEcQjRhpR/2rkNx6I0lSKOzNISMhRV9CYYmSs1Hb3Ah7Jx+zCZobwHAYJhbf35bZb9EreGHCe+FNSBFNU2+5X0JE45UQYzJDWLd9LTJghZShmZFgIUk0ShPuoS1qWCsSJDrPx+UN4aJUOjKWyTxg4Vn8nMsS1HvDITnJkenrWG4n/ea3UxGdhRkWSGiLwZFGcMmgkHHUBO1QRbNjAEoQVtbdC3EMKYWMbK9gS/Nkvz5N6ueSflLyb42LlclpHHuyDA3AEfHAKKuAaVEENYJCBZ/AK3pwn58V5dz4mozlnmtkFf+B8/gDMOpS4</latexit>

R⇠⇤ 1
<latexit sha1_base64="xLGfFiNaHbLbRxBUGbvm2mRtYII=">AAAB/HicbVC7TsMwFHXKq5RXoCOLRYXEQpUgEIwVLAwMBdGH1ITKcZzWqu1EtoMUReVXWBhAiJUPYeNvcNsM0HIkS0fn3KN7fYKEUaUd59sqLS2vrK6V1ysbm1vbO/buXlvFqcSkhWMWy26AFGFUkJammpFuIgniASOdYHQ18TuPRCoai3udJcTnaCBoRDHSRurb1TvoKcqhd2MyIXrIj91x3645dWcKuEjcgtRAgWbf/vLCGKecCI0ZUqrnOon2cyQ1xYyMK16qSILwCA1Iz1CBOFF+Pj1+DA+NEsIoluYJDafq70SOuFIZD8wkR3qo5r2J+J/XS3V04edUJKkmAs8WRSmDOoaTJmBIJcGaZYYgLKm5FeIhkghr01fFlODOf3mRtE/q7lnduT2tNS6LOspgHxyAI+CCc9AA16AJWgCDDDyDV/BmPVkv1rv1MRstWUWmCv7A+vwBYz2T9Q==</latexit>

Figure 10.3: Gauss law for unconfined (left) and confined (right) flux.

Hence, the electric field in that region has a magnitude which is r-independent,
which implies that the force it generates on a test charge is also r-independent,

215
 and so the potential V between charges would grow linearly in r. The force
goes to a constant, it never drops to zero, see Figure 10.4.

r behavior at
large distance
V(r)

1 r
<latexit sha1_base64="0pXKXSnQ+iAG/Z1RCbmkZrm/4QE=">AAAB83icbVDLSgMxFL3js9ZX1aWbYBHcWGZE0WXRjQsXFewDOmPJZDJtaJIZkoxQhv6GGxeKuPVn3Pk3pu0stPVA4HDOudybE6acaeO6387S8srq2nppo7y5tb2zW9nbb+kkU4Q2ScIT1QmxppxJ2jTMcNpJFcUi5LQdDm8mfvuJKs0S+WBGKQ0E7ksWM4KNlXz/zkYj/JifeuNeperW3CnQIvEKUoUCjV7ly48SkgkqDeFY667npibIsTKMcDou+5mmKSZD3KddSyUWVAf59OYxOrZKhOJE2ScNmqq/J3IstB6J0CYFNgM9703E/7xuZuKrIGcyzQyVZLYozjgyCZoUgCKmKDF8ZAkmitlbERlghYmxNZVtCd78lxdJ66zmXdTc+/Nq/bqoowSHcAQn4MEl1OEWGtAEAik8wyu8OZnz4rw7H7PoklPMHMAfOJ8/az2RRw==</latexit>

1/r behavior at
short distance

Figure 10.4: The potential between two test charges in YM theory.

Let us take a closer look to the potential, which generically reads

V (r) = TR r . (10.15)

The proportionality coefficient has dimension of an energy per unit length, and it is
the so-called string tension. On general ground, one would expect the string tension
to depend in some way on the gauge group representation the test charges transform.
This is pretty obvious since, e.g., for the singlet representation TR is clearly zero,
while for actual quarks, which transform in the fundamental representation, it is not.
In fact, as we are going to show below, the string tension does not depend on the
representation itself, but actually on what is called the N-ality of the representation.
Let us consider a (gauge) group G. Its center, CG is defined as the part of G
which commutes with all generators. For G = SU (N ), we have that
n o
CG = Uβ̄α = e2πik/N δβ̄α ; k = 0, 1, . . . , N − 1 mod N , (10.16)

where α an index in the fundamental and β̄ in the anti-fundamental of the gauge

group. Hence, in this case, CG = ZN . Let us now consider some representation R.
An element ρ of this representation is labeled by n upper indices αi and n̄ lower
indices β̄i , each upper index transforming in the fundamental and each lower index

216
transforming in the anti-fundamental representation. If one acts with the center of
the group on ρ one gets
CG : ρ → e2πik(n−n̄)/N ρ . (10.17)
The coefficient n − n̄ is called the N-ality of the representation ρ. If ρ has N-
ality p, then the complex conjugate representation ρC has N-ality N − p (from
eq. (10.17) it follows that the N-ality is defined modulo N ). For instance, while the
adjoint representation and the trivial representation have p = 0, the fundamental
representation has p = 1 and the anti-fundamental has p = N − 1.
Clearly, representations break into equivalence classes under the center of the
gauge group. It turns out that the string tension TR is not a function of the repre-
sentation but actually of the N-ality (the basic reason for this is that gluon number
is not a conserved quantity in YM theories, while N-ality is). Let us consider heavy
test particles transforming either in the anti-symmetric representation or in the sym-
metric representation of the gauge group, ψS and ψA , respectively. Each of them
will have its own string tension, TS and TA (but same N-ality, p = 2).
Suppose that TS > TA . Since gluon number is not a conserved quantity, we
can add a gluon Aµ coming from the chromoelectric flux tube next to ψS . The
charge of the bound state ψS Aµ is Symmetric ⊗ Adj = ⊕ Representations, where all
representations entering the sum have the same N-ality (the same as the symmetric
representation, in fact, since the N-ality is an additive quantity, and that of the
adjoint representation is zero). For G = SU (3) we have

6 ⊗ 8 = 3 + 6 + 15 + 24 , (10.18)

where the first representation on the r.h.s. is the anti-symmetric representation.

Since we have assumed that TS > TA it is energetically favored to pop a gluon out
of the vacuum and put it near to ψS (and another one near to ψ S ) since this has
an energy cost (of order Λ) which is lower than the energy gain, proportional to
(TS − TA )r, which for sufficiently large r always wins. In other words, in YM theory
the representation of a chromoelectric source is not a conserved quantum number;
only the N-ality is. Therefore, for all representations with the same N-ality, there
is only one stable configuration of strings, the one with lowest tension, as shown
in Figure 10.5. In summary, the tension of stable flux tubes are labeled by p, the
N-ality, not by R, the representation, as anticipated.
Notice that charge conjugation symmetry ensures that Tp = TN −p . Therefore,
there are order N/2 stable flux tube configurations for SU (N ). For SU (3) there is

217
 TA TS
<latexit sha1_base64="qmJ49ghcTzy+d5loGENNDRo1bGA=">AAAB6nicbVBNS8NAEJ3Ur1q/qh69LBbBU0lE0WPRi8dKW1toQ9lsJ+3SzSbsboQS+hO8eFDEq7/Im//GbZuDtj4YeLw3w8y8IBFcG9f9dgpr6xubW8Xt0s7u3v5B+fDoUcepYthisYhVJ6AaBZfYMtwI7CQKaRQIbAfju5nffkKleSybZpKgH9Gh5CFn1Fip0ew3+uWKW3XnIKvEy0kFctT75a/eIGZphNIwQbXuem5i/Iwqw5nAaamXakwoG9Mhdi2VNELtZ/NTp+TMKgMSxsqWNGSu/p7IaKT1JApsZ0TNSC97M/E/r5ua8MbPuExSg5ItFoWpICYms7/JgCtkRkwsoUxxeythI6ooMzadkg3BW355lTxeVL2rqvtwWand5nEU4QRO4Rw8uIYa3EMdWsBgCM/wCm+OcF6cd+dj0Vpw8plj+APn8wcJso2i</latexit>

<latexit sha1_base64="GjiDmyzpV8uLRjHhIVgZY9SLwXM=">AAAB6nicbVBNS8NAEJ3Ur1q/qh69LBbBU0lE0WPVi8eK/YI2lM120i7dbMLuRiihP8GLB0W8+ou8+W/ctjlo9cHA470ZZuYFieDauO6XU1hZXVvfKG6WtrZ3dvfK+wctHaeKYZPFIladgGoUXGLTcCOwkyikUSCwHYxvZ377EZXmsWyYSYJ+RIeSh5xRY6WHRv+6X664VXcO8pd4OalAjnq//NkbxCyNUBomqNZdz02Mn1FlOBM4LfVSjQllYzrErqWSRqj9bH7qlJxYZUDCWNmShszVnxMZjbSeRIHtjKgZ6WVvJv7ndVMTXvkZl0lqULLFojAVxMRk9jcZcIXMiIkllClubyVsRBVlxqZTsiF4yy//Ja2zqndRde/PK7WbPI4iHMExnIIHl1CDO6hDExgM4Qle4NURzrPz5rwvWgtOPnMIv+B8fAPuW42Q</latexit>

<latexit sha1_base64="srofXyAxpBvMicOjR9pxHy/tq/w=">AAAB+nicbVDLSsNAFL2pr1pfqS7dDBbBVUlE0WXVjcsK9gFNCJPptB06yYSZiVJiP8WNC0Xc+iXu/BsnbRbaemDgcM493DsnTDhT2nG+rdLK6tr6RnmzsrW9s7tnV/fbSqSS0BYRXMhuiBXlLKYtzTSn3URSHIWcdsLxTe53HqhUTMT3epJQP8LDmA0YwdpIgV31hLHzdOYlik2Dq8CuOXVnBrRM3ILUoEAzsL+8viBpRGNNOFaq5zqJ9jMsNSOcTiteqmiCyRgPac/QGEdU+dns9Ck6NkofDYQ0L9Zopv5OZDhSahKFZjLCeqQWvVz8z+ulenDpZyxOUk1jMl80SDnSAuU9oD6TlGg+MQQTycytiIywxESbtiqmBHfxy8ukfVp3z+vO3VmtcV3UUYZDOIITcOECGnALTWgBgUd4hld4s56sF+vd+piPlqwicwB/YH3+AM6YlFk=</latexit>

<latexit sha1_base64="sCyWbWSgemidgzf/7vJJSOo4uYM=">AAAB7XicbVBNSwMxEJ3Ur1q/qh69BIvgqeyKoseqF48V7Ae0S8mm2TY2myxJVihL/4MXD4p49f9489+YtnvQ1gcDj/dmmJkXJoIb63nfqLCyura+UdwsbW3v7O6V9w+aRqWasgZVQul2SAwTXLKG5VawdqIZiUPBWuHoduq3npg2XMkHO05YEJOB5BGnxDqp2U0M7133yhWv6s2Al4mfkwrkqPfKX92+omnMpKWCGNPxvcQGGdGWU8EmpW5qWELoiAxYx1FJYmaCbHbtBJ84pY8jpV1Ji2fq74mMxMaM49B1xsQOzaI3Ff/zOqmNroKMyyS1TNL5oigV2Co8fR33uWbUirEjhGrubsV0SDSh1gVUciH4iy8vk+ZZ1b+oevfnldpNHkcRjuAYTsGHS6jBHdShARQe4Rle4Q0p9ILe0ce8tYDymUP4A/T5A2WljwI=</latexit>

<latexit sha1_base64="ucM4TvPCPkux/N+semqbm45qq8c=">AAAB+nicbVDLSsNAFL2pr1pfqS7dDBbBVUlE0WXRjcuK9gFNCJPptB06mYSZiVJiP8WNC0Xc+iXu/BsnbRbaemDgcM493DsnTDhT2nG+rdLK6tr6RnmzsrW9s7tnV/fbKk4loS0S81h2Q6woZ4K2NNOcdhNJcRRy2gnH17nfeaBSsVjc60lC/QgPBRswgrWRArvqxcbO05mXKDYN7gK75tSdGdAycQtSgwLNwP7y+jFJIyo04Vipnusk2s+w1IxwOq14qaIJJmM8pD1DBY6o8rPZ6VN0bJQ+GsTSPKHRTP2dyHCk1CQKzWSE9Ugtern4n9dL9eDSz5hIUk0FmS8apBzpGOU9oD6TlGg+MQQTycytiIywxESbtiqmBHfxy8ukfVp3z+vO7VmtcVXUUYZDOIITcOECGnADTWgBgUd4hld4s56sF+vd+piPlqwicwB/YH3+AOnglGs=</latexit>

<latexit sha1_base64="cBkRlZe/IzENFLVao6u63rDUlaU=">AAAB7XicbVBNSwMxEJ3Ur1q/qh69BIvgqeyKoseiF48V7Qe0S8mm2TY2myxJVihL/4MXD4p49f9489+YtnvQ1gcDj/dmmJkXJoIb63nfqLCyura+UdwsbW3v7O6V9w+aRqWasgZVQul2SAwTXLKG5VawdqIZiUPBWuHoZuq3npg2XMkHO05YEJOB5BGnxDqp2U0M7933yhWv6s2Al4mfkwrkqPfKX92+omnMpKWCGNPxvcQGGdGWU8EmpW5qWELoiAxYx1FJYmaCbHbtBJ84pY8jpV1Ji2fq74mMxMaM49B1xsQOzaI3Ff/zOqmNroKMyyS1TNL5oigV2Co8fR33uWbUirEjhGrubsV0SDSh1gVUciH4iy8vk+ZZ1b+oenfnldp1HkcRjuAYTsGHS6jBLdShARQe4Rle4Q0p9ILe0ce8tYDymUP4A/T5A4DtjxQ=</latexit>

A A S S
decay
<latexit sha1_base64="a0PhYDgUi/yTTVwQOi4SQhTXH7U=">AAAB83icbVBNS8NAEN3Ur1q/qh69LBbBU0lE0WPRi8cKthaaUDabSbt0sxt2N2II/RtePCji1T/jzX/jts1BWx8MPN6bYWZemHKmjet+O5WV1bX1jepmbWt7Z3evvn/Q1TJTFDpUcql6IdHAmYCOYYZDL1VAkpDDQzi+mfoPj6A0k+Le5CkECRkKFjNKjJV8PwnlUxEBJflkUG+4TXcGvEy8kjRQifag/uVHkmYJCEM50brvuakJCqIMoxwmNT/TkBI6JkPoWypIAjooZjdP8IlVIhxLZUsYPFN/TxQk0TpPQtuZEDPSi95U/M/rZya+Cgom0syAoPNFccaxkXgaAI6YAmp4bgmhitlbMR0RRaixMdVsCN7iy8uke9b0Lpru3XmjdV3GUUVH6BidIg9doha6RW3UQRSl6Bm9ojcnc16cd+dj3lpxyplD9AfO5w+P4pIG</latexit>

TA
<latexit sha1_base64="GjiDmyzpV8uLRjHhIVgZY9SLwXM=">AAAB6nicbVBNS8NAEJ3Ur1q/qh69LBbBU0lE0WPVi8eK/YI2lM120i7dbMLuRiihP8GLB0W8+ou8+W/ctjlo9cHA470ZZuYFieDauO6XU1hZXVvfKG6WtrZ3dvfK+wctHaeKYZPFIladgGoUXGLTcCOwkyikUSCwHYxvZ377EZXmsWyYSYJ+RIeSh5xRY6WHRv+6X664VXcO8pd4OalAjnq//NkbxCyNUBomqNZdz02Mn1FlOBM4LfVSjQllYzrErqWSRqj9bH7qlJxYZUDCWNmShszVnxMZjbSeRIHtjKgZ6WVvJv7ndVMTXvkZl0lqULLFojAVxMRk9jcZcIXMiIkllClubyVsRBVlxqZTsiF4yy//Ja2zqndRde/PK7WbPI4iHMExnIIHl1CDO6hDExgM4Qle4NURzrPz5rwvWgtOPnMIv+B8fAPuW42Q</latexit>

T A < TS
<latexit sha1_base64="ub1sDf0KnEcIKk4TzIbNqkxHUYo=">AAAB8HicbVA9SwNBEJ2LXzF+RS1tFoNgFe5E0cIiamMZMV+SHMfeZi9Zsrt37O4J4civsLFQxNafY+e/cZNcodEHA4/3ZpiZFyacaeO6X05haXllda24XtrY3NreKe/utXScKkKbJOax6oRYU84kbRpmOO0kimIRctoORzdTv/1IlWaxbJhxQn2BB5JFjGBjpYdGcIUuUSO4D8oVt+rOgP4SLycVyFEPyp+9fkxSQaUhHGvd9dzE+BlWhhFOJ6VeqmmCyQgPaNdSiQXVfjY7eIKOrNJHUaxsSYNm6s+JDAutxyK0nQKboV70puJ/Xjc10YWfMZmkhkoyXxSlHJkYTb9HfaYoMXxsCSaK2VsRGWKFibEZlWwI3uLLf0nrpOqdVd2700rtOo+jCAdwCMfgwTnU4Bbq0AQCAp7gBV4d5Tw7b877vLXg5DP78AvOxzcd3I9O</latexit>

<latexit sha1_base64="ucM4TvPCPkux/N+semqbm45qq8c=">AAAB+nicbVDLSsNAFL2pr1pfqS7dDBbBVUlE0WXRjcuK9gFNCJPptB06mYSZiVJiP8WNC0Xc+iXu/BsnbRbaemDgcM493DsnTDhT2nG+rdLK6tr6RnmzsrW9s7tnV/fbKk4loS0S81h2Q6woZ4K2NNOcdhNJcRRy2gnH17nfeaBSsVjc60lC/QgPBRswgrWRArvqxcbO05mXKDYN7gK75tSdGdAycQtSgwLNwP7y+jFJIyo04Vipnusk2s+w1IxwOq14qaIJJmM8pD1DBY6o8rPZ6VN0bJQ+GsTSPKHRTP2dyHCk1CQKzWSE9Ugtern4n9dL9eDSz5hIUk0FmS8apBzpGOU9oD6TlGg+MQQTycytiIywxESbtiqmBHfxy8ukfVp3z+vO7VmtcVXUUYZDOIITcOECGnADTWgBgUd4hld4s56sF+vd+piPlqwicwB/YH3+AOnglGs=</latexit>

Sg g
<latexit sha1_base64="cBkRlZe/IzENFLVao6u63rDUlaU=">AAAB7XicbVBNSwMxEJ3Ur1q/qh69BIvgqeyKoseiF48V7Qe0S8mm2TY2myxJVihL/4MXD4p49f9489+YtnvQ1gcDj/dmmJkXJoIb63nfqLCyura+UdwsbW3v7O6V9w+aRqWasgZVQul2SAwTXLKG5VawdqIZiUPBWuHoZuq3npg2XMkHO05YEJOB5BGnxDqp2U0M7933yhWv6s2Al4mfkwrkqPfKX92+omnMpKWCGNPxvcQGGdGWU8EmpW5qWELoiAxYx1FJYmaCbHbtBJ84pY8jpV1Ji2fq74mMxMaM49B1xsQOzaI3Ff/zOqmNroKMyyS1TNL5oigV2Co8fR33uWbUirEjhGrubsV0SDSh1gVUciH4iy8vk+ZZ1b+oenfnldp1HkcRjuAYTsGHS6jBLdShARQe4Rle4Q0p9ILe0ce8tYDymUP4A/T5A4DtjxQ=</latexit>

<latexit sha1_base64="qmqxvw54qUUoUaz9MnxwYmJpOwM=">AAAB6HicbVBNS8NAEJ3Ur1q/qh69LBbBU0lE0WPRi8cW7Ae0oWy2k3btZhN2N0IJ/QVePCji1Z/kzX/jts1BWx8MPN6bYWZekAiujet+O4W19Y3NreJ2aWd3b/+gfHjU0nGqGDZZLGLVCahGwSU2DTcCO4lCGgUC28H4bua3n1BpHssHM0nQj+hQ8pAzaqzUGPbLFbfqzkFWiZeTCuSo98tfvUHM0gilYYJq3fXcxPgZVYYzgdNSL9WYUDamQ+xaKmmE2s/mh07JmVUGJIyVLWnIXP09kdFI60kU2M6ImpFe9mbif143NeGNn3GZpAYlWywKU0FMTGZfkwFXyIyYWEKZ4vZWwkZUUWZsNiUbgrf88ippXVS9q6rbuKzUbvM4inACp3AOHlxDDe6hDk1ggPAMr/DmPDovzrvzsWgtOPnMMfyB8/kDzceM7w==</latexit> <latexit sha1_base64="qmqxvw54qUUoUaz9MnxwYmJpOwM=">AAAB6HicbVBNS8NAEJ3Ur1q/qh69LBbBU0lE0WPRi8cW7Ae0oWy2k3btZhN2N0IJ/QVePCji1Z/kzX/jts1BWx8MPN6bYWZekAiujet+O4W19Y3NreJ2aWd3b/+gfHjU0nGqGDZZLGLVCahGwSU2DTcCO4lCGgUC28H4bua3n1BpHssHM0nQj+hQ8pAzaqzUGPbLFbfqzkFWiZeTCuSo98tfvUHM0gilYYJq3fXcxPgZVYYzgdNSL9WYUDamQ+xaKmmE2s/mh07JmVUGJIyVLWnIXP09kdFI60kU2M6ImpFe9mbif143NeGNn3GZpAYlWywKU0FMTGZfkwFXyIyYWEKZ4vZWwkZUUWZsNiUbgrf88ippXVS9q6rbuKzUbvM4inACp3AOHlxDDe6hDk1ggPAMr/DmPDovzrvzsWgtOPnMMfyB8/kDzceM7w==</latexit>

Figure 10.5: The string tensions corresponding to the antisymmetric and symmetric
representations. All strings have the same N-ality, p = 2. The flux tube in the sym-
metric representation, which is less energetic, decays into that of the anti-symmetric
one by popping-up a gluon out of the vacuum.

only one single confining string, that with N-ality p = 1, since T0 = 0 and T2 = T1 .
Multiple flux tubes only arise for larger gauge groups (for instance, for G = SU (4)
we have two string tensions, with N-ality p = 1 and p = 2, respectively).
All what we said let us also understand how to classify gauge singlets bound
states. While gluons are not confined by flux tubes, since TAdj = T0 = 0, any heavy
quark ψ with N-ality 6= 0 will experience a linear potential and a constant force
which will confine it to an antiquark ψ (these are the mesons) or, more generally,
to some combination of quarks and anti-quarks with opposite N-ality. For instance,
such combination can be made of N − 1 quarks and the bound state is called a
baryon.

• SYM, supersymmetric gauge interactions without matter fields.

We will study this theory in detail later. Here, we would just like to emphasize
similarities and differences with respect to pure YM theories and QCD.
Similarly to YM, SYM enjoys strict confinement, a mass gap and no pions.
Similarly to QCD, it has a sort of chiral symmetry breaking and an anomaly, which
makes the corresponding η 0 -like particle being massive. Finally, it differs from both
since it has multiple isolated vacua.
We stick again, for definiteness, to a SU (N ) gauge group, but qualitatively
identical statements can be done for more other choices. The structure of the (on-
shell) SYM Lagrangian
 
1 µν
LSYM = − Tr Fµν F + iλDλ / , (10.19)
4

218
where the trace on gauge indices is in the Adjoint representation of SU (N ). First
notice that gauginos do not break flux tubes since they transform in the Adjoint,
which is in the same N-ality class of the singlet representation. So gauginos behave
very differently from QCD quarks, in this respect. Basically, the presence of these
fermion fields does not change the confining behavior of pure YM glue, since gauginos
cannot break flux tubes. That is why SYM enjoys strict confinement, differently
from QCD.
On the other hand, the U (1)R symmetry resembles the axial symmetry of QCD,
since it is anomalous and it is broken to Z2N at the quantum level (recall gauginos
have R-charge equal to one). Finally, also SYM enjoys chiral symmetry breaking, in
the sense that gaugino bilinears acquire a non-vanishing VEV in the vacuum. More
precisely, we have

hλλi ∼ Λ3 e2πik/N , k = 0, 1, . . . , N − 1 , (10.20)

which breaks Z2N → Z2 . Hence there are N isolated vacua, each of them Z2 sym-
metric, related by ZN rotations, as shown in Figure 10.6.

ZN
<latexit sha1_base64="qU+n3abrzmk514ybJI5f3uyPwNk=">AAAB83icbVDLSgMxFL1TX7W+qi7dBIvgqsyIosuiG1dSwT6wM5RMmmlDM8mQZIQy9DfcuFDErT/jzr8x085CWw8EDufcyz05YcKZNq777ZRWVtfWN8qbla3tnd296v5BW8tUEdoikkvVDbGmnAnaMsxw2k0UxXHIaScc3+R+54kqzaR4MJOEBjEeChYxgo2VfD/GZhSG2eO0f9ev1ty6OwNaJl5BalCg2a9++QNJ0pgKQzjWuue5iQkyrAwjnE4rfqppgskYD2nPUoFjqoNslnmKTqwyQJFU9gmDZurvjQzHWk/i0E7mGfWil4v/eb3URFdBxkSSGirI/FCUcmQkygtAA6YoMXxiCSaK2ayIjLDCxNiaKrYEb/HLy6R9Vvcu6u79ea1xXdRRhiM4hlPw4BIacAtNaAGBBJ7hFd6c1Hlx3p2P+WjJKXYO4Q+czx8jsJG/</latexit>

Figure 10.6: The vacuum structure of pure N = 1 SYM. The N vacua are isolated,
and related by ZN rotations (compare with Figure 10.2).

The η 0 is the phase of the condensate (10.20) (similarly to QCD), but the vacua
are isolated, so there is a dynamical mass gap (unlike QCD and like YM).

10.1.1 Intermezzo: Wilson loops as order parameters for confinement

We have discussed the different ways in which confinement is realized in YM, QCD
and SYM by probing the theories with external heavy sources. In particular, we have
seen that pure YM enjoys strict confinement and the potential between static sources
grows linearly at large distance (and the same happens for SYM). QCD, instead,

219
enjoys a milder version of confinement, namely what is called charge screening. In
this case, the potential between static sources goes to a constant asymptotically
since the force drops to zero at large distance, as the charge is screened.
Interestingly, similar conclusions can be reached without resorting to external
sources but rather computing expectation values of specific line operators, known
as Wilson loops, as we now review.
In the same vein as in General Relativity the Levi-Civita connection tell us how to
parallel transport vectors around a manifold, in a gauge theory the gauge connection
Aµ tells how the internal degrees of freedom of a charged particle in representation
R of the gauge group (i.e. the vector ωi , with i = 1, . . . , dimR, describing its color
degrees of freedom) change as the particle moves along some given path γ. The way
the vector ωi rotates along γ depends on the initial and final points of γ, as well as
on γ itself.
If we take the path to be closed, this tells how the vector differs from its starting
value (hence measuring the holonomy). In this case one can form a gauge invariant
object, the Wilson loop, defined as
 Z 
µ
WR (γ) = TrR P exp i Aµ dx , (10.21)
γ

where P is the path-ordering operator (meaning that when expanding the exponen-
tial the matrices Aµ = Aaµ Ta are ordered so that those at earlier times are placed to
the left). The Wilson loop is a gauge invariant but non-local operator.
What does this have to do with confinement? Well, it so happens that the VEV
of the Wilson loop operator is an order parameter for confinement! Let us see how
this comes. The VEV of the Wilson loop reads
Z  Z 
hWR (γ)i = DA TrR P exp i Aµ dx e−S µ
(10.22)
γ

where S is the action. Let us now consider the closed path in figure 10.7 and let us
compute the VEV of the corresponding Wilson loop in the fundamental representa-
tion.
Let us take the time T to be very large, eventually T → ∞. For long enough
times the path integral projects the system onto lowest energy states. Before q and
q appear and after they disappear this energy is zero, while in between this is V (L).
Hence in this limit we get
(
e−σLT = e−σA[γ] Confinement
lim hWf (γ)i ∼ e−V (L)T = V0 (10.23)
T →∞ e−V0 T = e− 2 P [γ] Charge screening

220
 T
<latexit sha1_base64="jxv6pvKFSb2KAthx53LYkzvZtgo=">AAAB6HicbVBNS8NAEJ34WetX1aOXxSJ4Kokoeix68dhCv6ANZbOdtGs3m7C7EUroL/DiQRGv/iRv/hu3bQ7a+mDg8d4MM/OCRHBtXPfbWVvf2NzaLuwUd/f2Dw5LR8ctHaeKYZPFIladgGoUXGLTcCOwkyikUSCwHYzvZ377CZXmsWyYSYJ+RIeSh5xRY6V6o18quxV3DrJKvJyUIUetX/rqDWKWRigNE1Trrucmxs+oMpwJnBZ7qcaEsjEdYtdSSSPUfjY/dErOrTIgYaxsSUPm6u+JjEZaT6LAdkbUjPSyNxP/87qpCW/9jMskNSjZYlGYCmJiMvuaDLhCZsTEEsoUt7cSNqKKMmOzKdoQvOWXV0nrsuJdV9z6Vbl6l8dRgFM4gwvw4Aaq8AA1aAIDhGd4hTfn0Xlx3p2PReuak8+cwB84nz+w+4zc</latexit>

>

>
q q̄
<latexit sha1_base64="lwCGAz/uUq77l1nbzdeXenTD488=">AAAB7XicbVDLSgNBEOyNrxhfUY9eBoPgKeyKosegF48RzAOSJcxOZpMx81hnZoWw5B+8eFDEq//jzb9xkuxBEwsaiqpuuruihDNjff/bK6ysrq1vFDdLW9s7u3vl/YOmUakmtEEUV7odYUM5k7RhmeW0nWiKRcRpKxrdTP3WE9WGKXlvxwkNBR5IFjOCrZOa3Qhr9NgrV/yqPwNaJkFOKpCj3it/dfuKpIJKSzg2phP4iQ0zrC0jnE5K3dTQBJMRHtCOoxILasJsdu0EnTilj2KlXUmLZurviQwLY8Yicp0C26FZ9Kbif14ntfFVmDGZpJZKMl8UpxxZhaavoz7TlFg+dgQTzdytiAyxxsS6gEouhGDx5WXSPKsGF1X/7rxSu87jKMIRHMMpBHAJNbiFOjSAwAM8wyu8ecp78d69j3lrwctnDuEPvM8fK3KO3A==</latexit>

L
<latexit sha1_base64="lNpVo8UFt7FFDCUHLIHVTUfKtA0=">AAAB6HicbVBNS8NAEJ34WetX1aOXxSJ4Kokoeix68diC/YA2lM120q7dbOLuRiihv8CLB0W8+pO8+W/ctjlo64OBx3szzMwLEsG1cd1vZ2V1bX1js7BV3N7Z3dsvHRw2dZwqhg0Wi1i1A6pRcIkNw43AdqKQRoHAVjC6nfqtJ1Sax/LejBP0IzqQPOSMGivVH3ulsltxZyDLxMtJGXLUeqWvbj9maYTSMEG17nhuYvyMKsOZwEmxm2pMKBvRAXYslTRC7WezQyfk1Cp9EsbKljRkpv6eyGik9TgKbGdEzVAvelPxP6+TmvDaz7hMUoOSzReFqSAmJtOvSZ8rZEaMLaFMcXsrYUOqKDM2m6INwVt8eZk0zyveZcWtX5SrN3kcBTiGEzgDD66gCndQgwYwQHiGV3hzHpwX5935mLeuOPnMEfyB8/kD3O+M+Q==</latexit>

L
<latexit sha1_base64="3N15HF9oM7jv7AKL9DiB8jk0nE8=">AAAB6HicbVA9SwNBEJ3zM8avqKXNYhCswp0oWgZtLCwSMB+QHGFvM5es2ds7dveEcOQX2FgoYutPsvPfuEmu0MQHA4/3ZpiZFySCa+O6387K6tr6xmZhq7i9s7u3Xzo4bOo4VQwbLBaxagdUo+ASG4Ybge1EIY0Cga1gdDv1W0+oNI/lgxkn6Ed0IHnIGTVWqt/3SmW34s5AlomXkzLkqPVKX91+zNIIpWGCat3x3MT4GVWGM4GTYjfVmFA2ogPsWCpphNrPZodOyKlV+iSMlS1pyEz9PZHRSOtxFNjOiJqhXvSm4n9eJzXhtZ9xmaQGJZsvClNBTEymX5M+V8iMGFtCmeL2VsKGVFFmbDZFG4K3+PIyaZ5XvMuKW78oV2/yOApwDCdwBh5cQRXuoAYNYIDwDK/w5jw6L8678zFvXXHymSP4A+fzB6TbjNQ=</latexit>

<latexit sha1_base64="3N15HF9oM7jv7AKL9DiB8jk0nE8=">AAAB6HicbVA9SwNBEJ3zM8avqKXNYhCswp0oWgZtLCwSMB+QHGFvM5es2ds7dveEcOQX2FgoYutPsvPfuEmu0MQHA4/3ZpiZFySCa+O6387K6tr6xmZhq7i9s7u3Xzo4bOo4VQwbLBaxagdUo+ASG4Ybge1EIY0Cga1gdDv1W0+oNI/lgxkn6Ed0IHnIGTVWqt/3SmW34s5AlomXkzLkqPVKX91+zNIIpWGCat3x3MT4GVWGM4GTYjfVmFA2ogPsWCpphNrPZodOyKlV+iSMlS1pyEz9PZHRSOtxFNjOiJqhXvSm4n9eJzXhtZ9xmaQGJZsvClNBTEymX5M+V8iMGFtCmeL2VsKGVFFmbDZFG4K3+PIyaZ5XvMuKW78oV2/yOApwDCdwBh5cQRXuoAYNYIDwDK/w5jw6L8678zFvXXHymSP4A+fzB6TbjNQ=</latexit>

Figure 10.7: A rectangular Wilson loop in the fundamental representation of the

gauge group G. The physical interpretation of such a path can be thought of as
having put a quark q and an anti-quark q at some distance L and then compute the
interaction potential as a function of time. Equivalently, one can think of creating
a q/q pair out of the vacuum at some time in the past. These then propagate for a
time T before they annihilate back in the vacuum.

where A is the area enclosed by γ and P its perimeter. The second equality in the
second equation holds because in the limit T → ∞ we get P = 2T + 2L ' 2T .
So we learn that in a confining theory, for representations with non-vanishing N-
ality, as the fundamental, the Wilson loop follows the area law, while for unconfined
theories (including those enjoying charge screening, as QCD!) it follows the perimeter
law, no matter the representation.
Actually, in the limit in which one takes the loop infinitely large, one could sus-
pect, naively, that, regardless the phase the theory enjoys both VEVs in eqs. (10.23)
vanish, since both the area and the perimeter become infinite. Actually, this is not
the case. Infinities can be re-absorbed by counterterms. These however are local
so, for a line operator as the Wilson loop, one can regulate the perimeter but not
the area. The upshot is then that for γ → ∞, in a confining theory the VEV of the
Wilson loop operator vanish, while for a theory with charge screening it goes to a
positive constant
hWf (γ)i = 0 YM, SYM
(10.24)
hWf (γ)i =
6 0 QCD
In this sense, the VEV of the Wilson loop operator can be regarded as an order
parameter for (strict) confinement.

221
10.2 Phases of gauge theories: examples
After this detour on ”the meaning of confinement” and the different ways in which
confinement is realized in a sample of UV-free gauge theories, we would like to
consider, in more general terms, which kind of phases a generic gauge theory can
enjoy. Roughly, there are basically three different such phases

• Higgs phase: the gauge group G is spontaneously broken, all vector bosons
obtain a mass.

• Coulomb phase: vector bosons remain massless and mediate 1/r long range
interactions. This phase can be either interacting (this is sometime what is
referred to as actual Coulomb phase) or free, meaning that asymptotic states
do not interact, at low enough energy.

• Wilson or confining phase: color sources, like quarks, gluons, etc..., are bound
into color singlets. As discussed before, confinement can be realized as charge
screening or strict confinement, depending on the details of the theory under
consideration.

It is worth notice that the Coulomb phase is not specific to abelian gauge theories,
as QED. For example, a non-abelian gauge theory with enough matter content
may become IR-free, giving a long-range potential between color charges V (r) ∼
a(r) × 1/r, with a(r) a coefficient decreasing logarithmically with r. And, as we will
see in the following, also interacting non-abelian Coulomb phases can exist.
There can of course be intermediate situations, where for instance the original
gauge group is Higgsed down to a subgroup H, which then confines, or is in a
Coulomb phase (this is what happens in the SM of electroweak interactions), etc...
In these cases the phase of the gauge theory is defined by what happens to H in the
vacua, regardless of the fate of the original gauge group G.
Below we consider two examples which will hopefully clarify the meaning of some
of above statements, but also point out some subtleties one could encounter when
dealing with concrete models.

10.2.1 Coulomb phase and free phase

Let us consider SQED. The scalar potential reads

1
2
V = m2 |φ− |2 + m2 |φ+ |2 + e2 |φ+ |2 − |φ− |2 , (10.25)
2
222
where φ− and φ+ are the scalar fields belonging to the two chiral superfields Φ−
and Φ+ with electric charge ±1 respectively, and a superpotential mass term W =
mΦ− Φ+ has been also added. Let us consider massive and massless cases separately.

• m 6= 0. In this case the vacuum is at hφ− i = hφ+ i = 0. Heavy static probe

charges would experience a potential

α(r) 1
V ∼ , α(r) ∼ . (10.26)
r log r

However, the logarithmic fall-off is frozen at distance r = m−1 : for larger

distances α stops running. Hence, the asymptotic potential reads
α∗
V (r) ∼ , α∗ = α(r = m−1 ) , (10.27)
r
which simply says that massive SQED is in a Coulomb phase.

• m = 0. In this case the potential gets contributions from D-terms only. Now
there are more vacua, actually a moduli space of vacua. Besides the origin
of field space, also any hφ− i = hφ+ i 6= 0 satisfies the D-equations. One
can parameterize the supersymmetric vacua in terms of the gauge invariant
combination u = hφ− φ+ i. We have then two options. When u 6= 0 we are in a
Higgs phase, the gauge group U (1) is broken and the photon becomes massive
(the theory is described by a massive vector multiplet and a massless chiral
multiplet). When instead u = 0 the gauge group remains unbroken. Still, we
are in a different phase with respect to the massive case. The basic difference
is that the coupling α(r) does not stop running, now, since m = 0, and hence
it ends-up vanishing at large enough distances. In other words, the potential
again reads
α(r) 1
V (r) ∼ , α(r) ∼ (10.28)
r log r
but now α = 0 for r → ∞. This is not really a Coulomb phase but actually
what is called a free phase. At low energy (large enough distances) the theory
is a theory of free massless particles.

Let us emphasize again that both the interacting and the free Coulomb phases
are not specific to abelian gauge theories, and can be enjoyed also by non-abelian
theories. We will see examples of this phenomenon soon.

223
10.2.2 Continuously connected phases

Sometime there is no gauge-invariant distinction between different phases. Let us

show this non-trivial fact with a simple example.
Let us consider a SU (2) gauge theory with a SU (2) scalar doublet φ (a Higgs
field), a SU (2) singlet eR and a SU (2) doublet L = (νL , eL ), with interaction La-
grangian
Lint = LφeR + h.c. . (10.29)
This is nothing but a one-family EW theory model.
As it happens in standard EW theory, this theory can be realized in the Higgs
phase, where the field φ gets a non-vanishing VEV
!
1 0
hφi = √ . (10.30)
2 v

In this phase all three gauge bosons get a mass, the neutrino νL remains massless
√
while the electron gets a mass me = v/ 2.
Suppose instead that the theory were realized in a different phase, a confinement
phase. In such phase one would not observe massless gauge bosons (as above) while
fermions and Higgs bosons would bind into SU (2) singlet combinations

EL = φ† L , NL = ab φa Lb , eR (10.31)

in terms of which the interaction Lagrangian becomes

Lint = mE L eR + h.c. , (10.32)

where m ∼ Λ. So we see that EL and eR pair-up and become massive while NL

remains massless. The spectrum in this phase is the same as that of the Higgs phase:
there is no gauge-invariant distinction between the two phases! Consistently, also a
Wilson loop computation cannot tell in which phase the theory is. In a Higgs phase
gauge bosons are massive and cannot mediate long range forces, so the potential
between static test charges goes to a constant asymptotically. As we have seen,
already, confinement in presence of massless (charged) matter is realized as charge
screening and also in this case the force drops to zero asymptotically. Hence, in both
case the VEV of a Wilson loop follows a perimeter law and it cannot distinguish
between the two phases. This said, there are quantitative differences between the
two phases: for instance, EL is a composite field and its pair production would be

224
suppressed by a form factor which is not observed. Still, we need experiments to
discern between the two phases and understand which one is actually realized in
Nature.
The general lesson we want to convey here is that by adjusting some parameter
of a gauge theory, sometime one can move continuously from one type of phase
to another. In this non-abelian example, there is no invariant distinction between
Higgs and Wilson phase. Theories of this kind are said to enjoy complementarity.
We will encounter many such situations in a while.

10.3 N=1 SQCD: perturbative analysis

In what follows we will consider SQCD and its quantum behavior (including non-
perturbative effects) and try to answer the basic questions about its dynamical
properties in the most analytical possible way. Let us first summarize what we have
already learned about SQCD, about its classical and quantum - though perturbative
only - behavior.
SQCD is a renormalizable supersymmetric gauge theory with gauge group SU (N ),
e and no superpotential. Interaction terms are present and come
F flavors (Q, Q)
from D-terms. The group of continuous global symmetries at the quantum level is
GF = SU (F )L × SU (F )R × U (1)B × U (1)R with charge assignment

SU (F )L SU (F )R U (1)B U (1)R
F −N
Qia F • 1 F
eb
Q • F −1 F −N
j F

As already emphasized, for pure SYM the R-symmetry is anomalous, and only a
Z2N subgroup of U (1)R survives at the quantum level.
What do we know about the quantum properties of SQCD? We know there is a
huge moduli space of supersymmetric vacua, described by the D-term equations
i
b f† = 0 ,
e b (T A )c Q
DA = Q† i (T A )cb Qci − Q (10.33)
i b c

where A = 1, 2, . . . , N 2 − 1 is an index in the adjoint representation of SU (N ).

Up to flavor and global gauge rotations, a solution of the above equations can
be found for both F < N and F ≥ N . For F < N one can show that on the moduli

225
space (10.33) the matrices Q and Q e can be put, at most, in the following form
 
v1 0 ... 0 0 ... 0
 0 v ... 0 0 . . . 0
 2  eT
Q= =Q (10.34)
. . . . . . . . . . . . 0 . . . 0
0 0 ... vF 0 ... 0
Hence, at a generic point of the moduli space the gauge group is broken to SU (N −
F ). The (classical) moduli space can be parameterized in terms of mesons fields
eaj
M ij = Qia Q (10.35)

without any classical constraint between them, since the meson matrix has maximal
rank.
For F ≥ N the matrices Q and Q e can also be brought to a diagonal form on the
moduli space
   
v1 0 ... 0 ṽ1 0 ... 0
   
 0 v2 . . . 0   0 ṽ2 . . . 0 
   
. . . ... ... ... . . . ... ... ...
   
  eT =  
Q= 0 0 ... vN  , Q  0 0 ... ṽN  (10.36)
   
 0 0 0 0   0 0 0 0 
   
. . . ... ... ... . . . ... ... ...
   
0 0 0 0 0 0 0 0
where |vi |2 − |ṽi |2 = a, with a a i-independent number. At a generic point of
the moduli space the gauge group is now completely broken. The moduli space
is efficiently described in terms of mesons and baryons but there exist classical
constraints between them, now. The mesons are again defined as in eq. (10.35)
but the meson matrix does not have maximal rank anymore. Baryons are gauge
e with
invariant single trace operators made out of N fields Q respectively N fields Q,
fully anti-symmetrized indices and read

Bi1 ...iF −N = i1 i2 ...iF −N j1 ...jN a1 a2 ...aN Qja11 Qja22 . . . QjaNN
e i1 ...iF −N = i1 i2 ...iF −N j1 ...jN a1 a2 ...a Q
B ea1 Q
ea2 eaN
N j1 j2 . . . QjN , (10.37)

where ai are gauge indices and il , jl are flavor indices.

As far as quantum correction are concerned, we know the exact (perturbative)
expression for the gauge coupling which, in the holomorphic scheme, reads
θYM 4π b1 Λ
τ= +i 2 = log , b1 = 3N − F and Λ = µ e2πiτ /b1 . (10.38)
2π g (µ) 2πi µ

226
10.4 N=1 SQCD: non-perturbative dynamics
Our goal is to understand how the above picture is modified once non-perturbative
corrections are taken into account and get, if possible, quantum exact description
of the vacua of the theory and of the low energy dynamics around them. Given the
expression of the one-loop coefficient of the β-function (10.38), the harder region of
the parameter space where to gain such an understanding is obviously F < 3N .
A generic prediction for a UV-free theory with a classical moduli space, as SQCD
is for F < 3N , is that quantum corrections are expected to modify the perturbative
analysis only near the origin of field space. Indeed, for large value of scalar field
VEVs, the gauge group gets broken (and the gauge coupling hence stops running)
for small values of the gauge coupling constant
 3N −F
8π 2 Λ
− +iθYM
e g 2 (hQi) = −→ 0 for hQi → ∞ . (10.39)
hQi

This implies that for large field VEVs the gauge coupling freezes at a value g∗
where (semi)-classical analysis works properly. The smaller the field VEV the more
important are quantum corrections. Hence, generically, we expect non-perturbative
dynamics to modify the perturbative answer mostly near the origin of field space.

g 2 (µ)
<latexit sha1_base64="Ms0FlQmUgUvNDbuyPqzA67mULbQ=">AAAB73icbVBNSwMxEJ2tX7V+VT16CRahXspuqeix6MVjBfsB7VqyabYNTbJrkhXK0j/hxYMiXv073vw3pu0etPXBwOO9GWbmBTFn2rjut5NbW9/Y3MpvF3Z29/YPiodHLR0litAmiXikOgHWlDNJm4YZTjuxolgEnLaD8c3Mbz9RpVkk780kpr7AQ8lCRrCxUmf4UC33RHLeL5bcijsHWiVeRkqQodEvfvUGEUkElYZwrHXXc2Pjp1gZRjidFnqJpjEmYzykXUslFlT76fzeKTqzygCFkbIlDZqrvydSLLSeiMB2CmxGetmbif953cSEV37KZJwYKsliUZhwZCI0ex4NmKLE8IklmChmb0VkhBUmxkZUsCF4yy+vkla14l1U3LtaqX6dxZGHEziFMnhwCXW4hQY0gQCHZ3iFN+fReXHenY9Fa87JZo7hD5zPHxB2j1Q=</latexit>

g⇤2 (hQi)
<latexit sha1_base64="k7eEwFIt43W2HEiTWDsCokjE2UE=">AAACAXicbZDLSsNAFIYn9VbrLepGcDNYhOqiJEXRZdGNyxbsBZoYJtNJOnQyCTMToYS68VXcuFDErW/hzrdxmmah1R8GPv5zDmfO7yeMSmVZX0ZpaXllda28XtnY3NreMXf3ujJOBSYdHLNY9H0kCaOcdBRVjPQTQVDkM9Lzx9ezeu+eCEljfqsmCXEjFHIaUIyUtjzzILxreKc1hyEeMgLb0BE5nXhm1apbueBfsAuogkItz/x0hjFOI8IVZkjKgW0lys2QUBQzMq04qSQJwmMUkoFGjiIi3Sy/YAqPtTOEQSz04wrm7s+JDEVSTiJfd0ZIjeRibWb+VxukKrh0M8qTVBGO54uClEEVw1kccEgFwYpNNCAsqP4rxCMkEFY6tIoOwV48+S90G3X7vG61z6rNqyKOMjgER6AGbHABmuAGtEAHYPAAnsALeDUejWfjzXift5aMYmYf/JLx8Q2BR5Wl</latexit>

⇤
<latexit sha1_base64="yFrnXXViWirehqLFkWd9XsZYx1w=">AAAB7nicbVBNS8NAFHypX7V+VT16WSyCp5KIoseiFw8eKlhbaEPZbDbt0s0m7L4IJfRHePGgiFd/jzf/jds2B20dWBhm5rHvTZBKYdB1v53Syura+kZ5s7K1vbO7V90/eDRJphlvsUQmuhNQw6VQvIUCJe+kmtM4kLwdjG6mfvuJayMS9YDjlPsxHSgRCUbRSu3enY2GtF+tuXV3BrJMvILUoECzX/3qhQnLYq6QSWpM13NT9HOqUTDJJ5VeZnhK2YgOeNdSRWNu/Hy27oScWCUkUaLtU0hm6u+JnMbGjOPAJmOKQ7PoTcX/vG6G0ZWfC5VmyBWbfxRlkmBCpreTUGjOUI4toUwLuythQ6opQ9tQxZbgLZ68TB7P6t5F3b0/rzWuizrKcATHcAoeXEIDbqEJLWAwgmd4hTcndV6cd+djHi05xcwh/IHz+QMM949h</latexit>

hQi
<latexit sha1_base64="hQzsxjAMR3xfcJJQlIjBAoc5Pq0=">AAAB+nicbZDLSsNAFIYn9VbrLdWlm8EiuCqJKLosunHZgr1AE8pketIOnUzCzEQpsY/ixoUibn0Sd76N0zQLbf1h4OM/53DO/EHCmdKO822V1tY3NrfK25Wd3b39A7t62FFxKim0acxj2QuIAs4EtDXTHHqJBBIFHLrB5HZe7z6AVCwW93qagB+RkWAho0Qba2BXPU7EiANuYU/mNLBrTt3JhVfBLaCGCjUH9pc3jGkagdCUE6X6rpNoPyNSM8phVvFSBQmhEzKCvkFBIlB+lp8+w6fGGeIwluYJjXP390RGIqWmUWA6I6LHark2N/+r9VMdXvsZE0mqQdDFojDlWMd4ngMeMglU86kBQiUzt2I6JpJQbdKqmBDc5S+vQue87l7WndZFrXFTxFFGx+gEnSEXXaEGukNN1EYUPaJn9IrerCfrxXq3PhatJauYOUJ/ZH3+AJMqk44=</latexit>

µ
<latexit sha1_base64="lhdCbeeq104/O61Yae5JMm53kmA=">AAAB6nicbVA9SwNBEJ2LXzF+RS1tFoNgFe5E0TJoYxnRmEByhL3NXrJkd+/YnRNCyE+wsVDE1l9k579xk1yhiQ8GHu/NMDMvSqWw6PvfXmFldW19o7hZ2tre2d0r7x882iQzjDdYIhPTiqjlUmjeQIGSt1LDqYokb0bDm6nffOLGikQ/4CjloaJ9LWLBKDrpvqOybrniV/0ZyDIJclKBHPVu+avTS1imuEYmqbXtwE8xHFODgkk+KXUyy1PKhrTP245qqrgNx7NTJ+TEKT0SJ8aVRjJTf0+MqbJ2pCLXqSgO7KI3Ff/z2hnGV+FY6DRDrtl8UZxJggmZ/k16wnCGcuQIZUa4WwkbUEMZunRKLoRg8eVl8nhWDS6q/t15pXadx1GEIziGUwjgEmpwC3VoAIM+PMMrvHnSe/HevY95a8HLZw7hD7zPH16wjdo=</latexit>

Figure 10.8: The gauge coupling running of a UV-free theory. The large hQi region
is a weakly coupled region where classical analysis is correct, since the value at which
the gauge coupling stops running, g = g∗ , is small.

Not surprisingly, for any fixed value of N , several non-perturbative dynamical

properties change with the number of flavors, F . Hence, in what follows, we will
consider different cases separately.

227
10.4.1 Pure SYM: gaugino condensation

We have already discussed this case, at a qualitative level. Let us first recall that this
is the only case in which there does not exist an anomaly-free R-symmetry. At the
quantum level, only a discrete Z2N R-symmetry survives. Promoting τ to a spurion
field and using holomorphy arguments, it is easy to see what the structure of the
non-perturbative generated superpotential should be. Let us first notice that the
operator e2πiτ /N has R-charge R = 2. Indeed, due to the transformation properties
of θYM under R-symmetry transformations, θYM → θYM + 2N α, we have

e2πiτ /N → e2iα e2πiτ /N . (10.40)

Because of confinement, assuming a mass gap, the effective Lagrangian should de-
pend only on τ , and hence Weff , if any, should also depend only on τ . Imposing
R-symmetry, by dimensional analysis the only possible term reads

Weff = c µ3 e2πiτ /N = c Λ3 . (10.41)

where c is an undetermined coefficient (which in principle could also be zero, of

course). This innocent-looking constant superpotential contribution contains one
crucial physical information. Given the presence of a massless strong interacting
fermion field (the gaugino), one could wonder whether in SYM theory gauginos
undergo pair condensation, as it is believed to happen in QCD, where quark bilinears
condense. Looking at the SYM Lagrangian
Z 
1 2 α
L= Im d θ τ Tr W Wα , (10.42)
32π
we see that λα λα is the scalar component of W α Wα and (minus) Fτ acts as a source
√
for it (recall we are thinking of τ as a spurion superfield, τ = τ + 2θψτ − θθFτ ).
Therefore, in order to compute the gaugino condensate one should just differentiate
R R
the logarithm of the partition function Z = DV ei L with respect to Fτ . In fact,
under the assumption of a mass gap, the low energy effective action depends only
on τ , since gauge fields have been integrated out, and it coincides with the effective
superpotential (10.41), giving for the gaugino condensate
Z
∂ ∂ ∂
hλλi = −16π log Z = −16πi d2 θWeff (τ ) = 16πi Weff (τ ) (10.43)
∂Fτ ∂Fτ ∂τ
where in doing the second step we have used the fact that
 
√ ∂Weff ∂Weff 1 ∂ 2 Weff 2
Weff = weff (τ ) + 2 θψτ − θθ Fτ + ψ .
∂τ ∂τ 2 ∂τ 2 τ

228
Plugging (10.41) in eq. (10.43) we get

32π 2 3 2πiτ /N
hλλi = − cµ e ≡ a Λ3 , (10.44)
N
which means that if c 6= 0 gauginos do condense in SYM. Since gauginos have R = 1,
this implies that in the vacuum the Z2N symmetry is broken to Z2 and that there
are in fact N distinct (and isolated) vacua. All these vacua appear explicitly in the
above formula since the transformation

θYM → θYM + 2πk , (10.45)

which is a symmetry of the theory, sweeps out N distinct values of the gaugino
condensate
hλλi → e2iα hλλi , θYM → θYM + 2N α ' θYM + 2πk (10.46)
where k = 0, 1, . . . , 2N − 1, and k = i and k = i + N give the same value of the
gaugino condensate. In other words, we can label the N vacua with N distinct
phases of the gaugino condensate (0, 2π N1 , 2π N2 , . . . , 2π NN−1 ), recall Figure 10.6.
This ends our discussion of pure SYM. It should be stressed that to have a
definitive picture we should find independent ways to compute the constant c in
eq. (10.41), since if it were zero, then all our conclusions would have been wrong (in
particular, there would not be any gaugino condensate, and hence we would have
had a unique vacuum preserving the full Z2N symmetry). We will come back to this
important point later.

10.4.2 F < N : the ADS superpotential

For F < N classical analysis tells that there is a moduli space of complex dimen-
sion F 2 , parameterized by meson field VEVs. The question, again, is whether an
effective superpotential is generated due to strong coupling dynamics. Let use again
holomorphy, and the trick of promoting coupling constants to spurion superfields.
The effective superpotential could depend on meson fields and on the complexified
gauge coupling, through Λ. The quantum numbers of (well educated functions of)
these two basic objects are

U (1)B U (1)A U (1)R

det M 0 2F 2(F − N )
Λ3N −F 0 2F 0

229
Both above objects are invariant under the non-abelian part of the global symmetry
group (notice that det M is the only SU (F )L × SU (F )R invariant one can make out
of M ). From the table above it follows that the only superpotential term which can
be generated should have the following form
  N −F
1
Λ3N −F
Weff = cN,F , (10.47)
det M

where, again, the overall constant, which generically will be some function of N and
F , is undetermined.
That (10.47) is the only possible term can be understood as follows. The effective
superpotential should have R-charge two, U (1)A and U (1)B charges equal to zero,
should be invariant under the non-abelian global symmetry, SU (F )L ×SU (F )R , and
should have dimension three. The chiral superfields that might contribute to it are
Λ, W α and the meson matrix M . This implies that Weff should be made by (a sum
of) terms like
Weff ∼ Λ(3N −F )n (W α Wα )m (det M )p , (10.48)
where n, m and p are integer numbers. The invariance under the non-abelian global
symmetry and the baryonic symmetry is guaranteed by any such term. As for the
other two abelian symmetries we get the following constraints
( (
U (1)A : 0 = 2nF + 2pF n = −p
−→ (10.49)
U (1)R : 2 = 2m + 2p(F − N ) p = (m − 1)/(N − F )

Since 3N − F > 0, in order to have a meaningful weak couping limit, we should

have n ≥ 0, which implies that p ≤ 0 and m ≤ 1. On the other hand, we should
have m ≥ 0 in order for the Wilsonian action to be local (it needs to have a sensible
derivative expansion), which finally implies that m = 0, 1. The contribution m = 1
and hence p = n = 0 is the tree level result (the gauge kinetic term, in fact). The
contribution m = 0 which implies p = −1/(N − F ) and n = 1/(N − F ) is precisely
(10.47), the so-called Affleck-Dine-Seiberg (ADS) superpotential.
In what follows we would like to analyze several properties of the ADS super-
potential, trying to understand where it may come from, physically, and eventually
determine the coefficient cN,F .
Let us consider again the classical moduli space. At a generic point of the moduli
space the SU (N ) gauge group is broken to SU (N − F ). Suppose for simplicity that
all scalar field VEVs are equal, vi = v, recall expression (10.34). Clearly the theory

230
behaves differently at energies higher or lower than v. At energies higher than v the
gauge coupling running is that of SQCD with gauge group SU (N ) and F massless
flavors. At energies lower than v all matter fields become massive (and should be
integrated out) while the gauge group is broken down to SU (N − F ). Hence the
theory runs differently and, accordingly, the dynamical generated scale, ΛL is also
different. More precisely we have
4π 3N − F µ
E>v = log
g 2 (µ) 2π Λ
4π 3(N − F ) µ
E<v = log . (10.50)
gL2 (µ) 2π ΛL

If supersymmetry is preserved the two above equations should match at E = v. This

is known as scale matching (that there are no threshold factors reflects a choice of
subtraction scheme, on which threshold factors depend; this is the correct matching
in, e.g., the DR scheme). Hence we get
  N −F
1
3(N −F ) 3N −F 1 Λ3N −F Λ3N −F
ΛL =Λ = −→ Λ3L = . (10.51)
v 2F det M det M

This implies that

  N −F
1
Λ3N −F
Weff = cN,F = cN,F Λ3L , (10.52)
det M
which means that
cN,F = cN −F,0 . (10.53)
Besides getting a relation between c’s for different theories (recall these are (N, F )-
dependent constants, in general), we also get from the above analysis some physical
intuition for how the ADS superpotential is generated. One can think of Weff being
generated by gaugino condensation of the left over SU (N − F ) gauge group (recall
that gaugino condensation is in one-to-one correspondence with the very existence
of an effective superpotential for pure SYM theory: the two are fully equivalent
statements).
Let us now start from SQCD with a given number of flavors and suppose to give
a mass m to the F -th flavor. At high enough energy this does not matter much.
But below the scale m the theory behaves as SQCD with F − 1 flavors, as far as the

231
gauge coupling running is concerned. More precisely, we have
4π 3N − F µ
E>m = log
g 2 (µ) 2π ΛF
4π 3N − (F − 1) µ
E<m = log . (10.54)
gL2 (µ) 2π ΛL,F −1
Matching the scale at E = m we obtain the following relation between non-perturbative
scales
−F +1 −F
Λ3N
L,F −1 = m Λ3NF . (10.55)
Let us now use holomorphic decoupling, to connect the theories above and below the
scale m. The superpotential of SQCD with F − 1 massless flavors and one massive
one reads
 3N −F  N −F
1
ΛF eF .
Weff = cN,F + mQF Q (10.56)
det M
At low enough energy we can trade the equations of motion of mesons involving the
massive flavor by their F-term equations. The F-term equation for M Fi for i 6= F
implies M Fi = 0, and similarly for M iF . So the meson matrix can be put into the
form !
f 0
M
M= , t ≡ M FF , (10.57)
0 t
where M f is the meson matrix made out of F − 1 flavors. The F-term equation for
t gives
 3N −F  N −F
1  1+ N −F
1
cN,F ΛF 1
0=− +m (10.58)
N − F det M f t
which implies
" 1 # N −F +1
 3N −F  F −N F −N

N −F ΛF
t= m . (10.59)
cN,F f
det M
Plugging this back into eq. (10.56) one gets
  NN−F−F+1  −F  N −F +1
1
cN,F mΛ3N
F
Weff = (N − F + 1) . (10.60)
N −F f
det M
We can now use eq. (10.55) and get for the effective superpotential of SQCD with
F − 1 flavor, which is what the theory reduces for at low enough energies,
  NN−F−F+1 ! N −F1 +1 ! N −F1 +1
−F +1 3N −F +1
cN,F Λ3N
L,F −1 Λ L,F −1
Weff = (N − F + 1) = cN,F −1 .
N −F det Mf det M f
(10.61)

232
giving finally the following relation
  NN−F−F+1
cN,F
cN,F −1 = (N − F + 1) . (10.62)
N −F
Combining this result with the relation we found before, eq. (10.53), one concludes
that all coefficients are related one another as
1
cN,F = (N − F ) c N −F , (10.63)

with a unique common coefficient c to be determined. This result tells that if the
ADS superpotential can be computed exactly for a given value of F (hence fixing
c), then we know its expression for any other value!
Let us consider the case F = N − 1, which is the extreme case in the window
F < N . In this case
cN,N −1 = c . (10.64)
Interestingly, for F = N − 1 the gauge group is fully broken, so there is no left-over
strong IR dynamics. In other words, any term appearing in the effective action
should be visible in a weak-coupling analysis. Even more interesting, the ADS
superpotential for F = N −1 is proportional to Λ2N +1 which is nothing but how one-
instanton effects contribute to gauge theory amplitudes (recall that for F = N − 1
b1 = 2N + 1, and e−Sinst ∼ Λb1 ), suggesting that in this case the ADS superpotential
is generated by instantons. At weak coupling, a reliable one-instanton calculation
can indeed be done and gives c = 1. Via eq. (10.63) this result hence fixes uniquely
cN,F for arbitrary values of N and F as

cN,F = N − F , (10.65)

giving finally for the ADS superpotential the following exact expression
  N −F
1
Λ3N −F
WADS = (N − F ) . (10.66)
det M
Notice that this also fixes the coefficient of the effective superpotential of pure SYM
theory which is
WSYM = N Λ3 , (10.67)
implying, via eq. (10.43), that gauginos do condense!
Let us finally see how does the ADS superpotential affect the moduli space of
vacua. From the expression (10.66) we can compute the potential, which is expected

233
not to be flat anymore, since the effective superpotential WADS depends on scalar
fields (through the meson matrix). The potential
X ∂WADS 2 ∂WADS 2
VADS = + (10.68)
∂Qi ei
∂Q
i

e → ∞, where it reaches zero,

is minimized at infinity in field space, namely for Q = Q
see Figure 10.9. This can be easily seen noticing that, qualitatively, det M ∼ M F ,
2N
which implies that VADS ∼ |M |− N −F , which is indeed minimized at infinity. This

V
<latexit sha1_base64="Y6Q3IX+jnEX5vHSW+OoCVbeDy6w=">AAAB6HicbVBNS8NAEJ3Ur1q/qh69LBbBU0lE0WPRi8cW7Ae0oWy2k3btZhN2N0IJ/QVePCji1Z/kzX/jts1BWx8MPN6bYWZekAiujet+O4W19Y3NreJ2aWd3b/+gfHjU0nGqGDZZLGLVCahGwSU2DTcCO4lCGgUC28H4bua3n1BpHssHM0nQj+hQ8pAzaqzUaPXLFbfqzkFWiZeTCuSo98tfvUHM0gilYYJq3fXcxPgZVYYzgdNSL9WYUDamQ+xaKmmE2s/mh07JmVUGJIyVLWnIXP09kdFI60kU2M6ImpFe9mbif143NeGNn3GZpAYlWywKU0FMTGZfkwFXyIyYWEKZ4vZWwkZUUWZsNiUbgrf88ippXVS9q6rbuKzUbvM4inACp3AOHlxDDe6hDk1ggPAMr/DmPDovzrvzsWgtOPnMMfyB8/kDtAOM3g==</latexit>

No stable vacuum
<latexit sha1_base64="7yA1/Lg2Ri/d6nX6hMuK54Qb7dw=">AAACAnicbVC7SgNBFJ2Nrxhfq1ZisxgEq7ArikKaoI2VRDAPyC5hdjJJhsxjmUcwLMHGX7GxUMTWr7Dzb5wkW2jigYHDOfdw5544oURp3/92ckvLK6tr+fXCxubW9o67u1dXwkiEa0hQIZsxVJgSjmuaaIqbicSQxRQ34sH1xG8MsVRE8Hs9SnDEYI+TLkFQW6ntHoQsFg/prQjLSkObCstDiIxh47Zb9Ev+FN4iCTJSBBmqbfcr7AhkGOYaUahUK/ATHaVQaoIoHhdCo3AC0QD2cMtSDhlWUTo9YewdW6XjdYW0j2tvqv5OpJApNWKxnWRQ99W8NxH/81pGdy+jlPDEaMzRbFHXUE8Lb9KH1yESI01HlkAkif2rh/pQQqRtawVbQjB/8iKpn5aC85J/d1asXGV15MEhOAInIAAXoAJuQBXUAAKP4Bm8gjfnyXlx3p2P2WjOyTL74A+czx/DX5eq</latexit>

e
<latexit sha1_base64="w+i1B+t+KG5rsATvpJEXatKFPcg=">AAAB+HicbVDLSgNBEJz1GeMjqx69DAbBg4RdUfQY9OIxAfOAZAmzs51kyOyDmV4lLvkSLx4U8eqnePNvnCR70MSChqKqm+4uP5FCo+N8Wyura+sbm4Wt4vbO7l7J3j9o6jhVHBo8lrFq+0yDFBE0UKCEdqKAhb6Elj+6nfqtB1BaxNE9jhPwQjaIRF9whkbq2aX6Ge0+igBQyABovWeXnYozA10mbk7KJEetZ391g5inIUTIJdO64zoJehlTKLiESbGbakgYH7EBdAyNWAjay2aHT+iJUQLaj5WpCOlM/T2RsVDrceibzpDhUC96U/E/r5Ni/9rLRJSkCBGfL+qnkmJMpynQQCjgKMeGMK6EuZXyIVOMo8mqaEJwF19eJs3zintZceoX5epNHkeBHJFjckpcckWq5I7USINwkpJn8krerCfrxXq3PuatK1Y+c0j+wPr8AZ4bkmo=</latexit>

Q, Q

Figure 10.9: The runaway behavior of the quantum corrected potential of SU (N )

SQCD with F < N .

means that the theory does not admit any stable vacuum at finite distance in field
space: the (huge) classical moduli space is completely lifted at the quantum level!
This apparently strange behavior makes sense, in fact, if one thinks about it for a
while. For large field VEVs, eventually for v → ∞, we recover pure SYM which has
indeed supersymmetric vacua (that is, zero energy states). This is part of the space
of D-term solutions of SQCD; any other configuration would have higher energy
and would hence be driven to the supersymmetric one. Let us suppose this picture
were wrong and that SQCD had a similar behavior as QCD: confinement and chiral
symmetry breaking. Then we would have expected a quark condensate to develop
hψQi ψQej i =
6 0. Such condensate, differently from a gaugino condensate (which we
certainly have), would break supersymmetry, since it is nothing but an F-term for
the meson matrix Mij . Hence this configuration would have E > 0 and thus any
configuration with E = 0 would be preferred. The latter are all configurations like
(10.34) which, by sending vi all the way to infinity, reduce to SYM, which admits
supersymmetry preserving vacua. The ADS superpotential simply shows this.

234
 There is a caveat in all this discussion. In our analysis we have not included
wave-function renormalization effects. The latter could give rise, in general, to non-
canonical Kähler potential terms, which could produce wiggles or even local minima
in the potential. However, at most this could give rise to metastable vacua (which our

V V
<latexit sha1_base64="Y6Q3IX+jnEX5vHSW+OoCVbeDy6w=">AAAB6HicbVBNS8NAEJ3Ur1q/qh69LBbBU0lE0WPRi8cW7Ae0oWy2k3btZhN2N0IJ/QVePCji1Z/kzX/jts1BWx8MPN6bYWZekAiujet+O4W19Y3NreJ2aWd3b/+gfHjU0nGqGDZZLGLVCahGwSU2DTcCO4lCGgUC28H4bua3n1BpHssHM0nQj+hQ8pAzaqzUaPXLFbfqzkFWiZeTCuSo98tfvUHM0gilYYJq3fXcxPgZVYYzgdNSL9WYUDamQ+xaKmmE2s/mh07JmVUGJIyVLWnIXP09kdFI60kU2M6ImpFe9mbif143NeGNn3GZpAYlWywKU0FMTGZfkwFXyIyYWEKZ4vZWwkZUUWZsNiUbgrf88ippXVS9q6rbuKzUbvM4inACp3AOHlxDDe6hDk1ggPAMr/DmPDovzrvzsWgtOPnMMfyB8/kDtAOM3g==</latexit> <latexit sha1_base64="Y6Q3IX+jnEX5vHSW+OoCVbeDy6w=">AAAB6HicbVBNS8NAEJ3Ur1q/qh69LBbBU0lE0WPRi8cW7Ae0oWy2k3btZhN2N0IJ/QVePCji1Z/kzX/jts1BWx8MPN6bYWZekAiujet+O4W19Y3NreJ2aWd3b/+gfHjU0nGqGDZZLGLVCahGwSU2DTcCO4lCGgUC28H4bua3n1BpHssHM0nQj+hQ8pAzaqzUaPXLFbfqzkFWiZeTCuSo98tfvUHM0gilYYJq3fXcxPgZVYYzgdNSL9WYUDamQ+xaKmmE2s/mh07JmVUGJIyVLWnIXP09kdFI60kU2M6ImpFe9mbif143NeGNn3GZpAYlWywKU0FMTGZfkwFXyIyYWEKZ4vZWwkZUUWZsNiUbgrf88ippXVS9q6rbuKzUbvM4inACp3AOHlxDDe6hDk1ggPAMr/DmPDovzrvzsWgtOPnMMfyB8/kDtAOM3g==</latexit>

<latexit sha1_base64="YfpKXZJLRTdadz7nLh4lX3SK6Lg=">AAACA3icbVDLSgMxFM3UV62vUXe6CRahqzIjikI3RTeupIJ9QGcomTTThiaTIcmIZSi48VfcuFDErT/hzr8x085CWw8EDufcw809Qcyo0o7zbRWWlldW14rrpY3Nre0de3evpUQiMWliwYTsBEgRRiPS1FQz0oklQTxgpB2MrjK/fU+koiK60+OY+BwNIhpSjLSRevaBxwPxkN4I7dVioRQ1Qa9WoeGk1LPLTtWZAi4SNydlkKPRs7+8vsAJJ5HGDCnVdZ1Y+ymSmmJGJiUvUSRGeIQGpGtohDhRfjq9YQKPjdKHoZDmRRpO1d+JFHGlxjwwkxzpoZr3MvE/r5vo8MJPaRQnmkR4tihMGNQCZoXAPpUEazY2BGFJzV8hHiKJsDa1ZSW48ycvktZJ1T2rOren5fplXkcRHIIjUAEuOAd1cA0aoAkweATP4BW8WU/Wi/VufcxGC1ae2Qd/YH3+AKHVl34=</latexit>

Not possible (if

<latexit sha1_base64="UiIE1mWsee0ruoRhFlFiILLYLwM=">AAACBHicbVBLSwMxGMz6rPW16rGXYBH0UnZFUeil6MVjBfuA7lKyadqG5rEkWbEsPXjxr3jxoIhXf4Q3/43Zdg/aOhAYZr4vyUwUM6qN5307S8srq2vrhY3i5tb2zq67t9/UMlGYNLBkUrUjpAmjgjQMNYy0Y0UQjxhpRaPrzG/dE6WpFHdmHJOQo4GgfYqRsVLXLQU8kg8pR1oH1QGKgyqWShFsTibFrlv2Kt4UcJH4OSmDHPWu+xX0JE44EQYze2PH92ITpkgZihmZFINEkxjhERqQjqUCcaLDdBpiAo+s0oN9qewRBk7V3xsp4lqPeWQnOTJDPe9l4n9eJzH9yzClIk4MEXj2UD9h0EiYNQJ7NMvLxpYgrKj9K8RDpBA2tresBH8+8iJpnlb884p3e1auXeV1FEAJHIJj4IMLUAM3oA4aAINH8AxewZvz5Lw4787HbHTJyXcOwB84nz+RBpgG</latexit>

mass gap correct)

Possible
<latexit sha1_base64="R8ngeWD9huNDDOvNgPzjy9S6dLU=">AAAB+HicbVDLSgMxFL3js9ZHR126CRbBVZkRRZdFNy4r2Ae0Q8mkaRuax5BkxDr0S9y4UMStn+LOvzFtZ6GtBwKHc+5Jbk6ccGZsEHx7K6tr6xubha3i9s7uXsnfP2gYlWpC60RxpVsxNpQzSeuWWU5biaZYxJw249HN1G8+UG2Ykvd2nNBI4IFkfUawdVLXL3VErB6zmjKGucyk65eDSjADWiZhTsqQo9b1vzo9RVJBpSUcG9MOg8RGGdaWEXdfsZMammAywgPadlRiQU2UzRafoBOn9FBfaXekRTP1dyLDwpixiN2kwHZoFr2p+J/XTm3/KsqYTFJLJZk/1E85sgpNW0A9pimxfOwIJpq5XREZYo2JdV0VXQnh4peXSeOsEl5UgrvzcvU6r6MAR3AMpxDCJVThFmpQBwIpPMMrvHlP3ov37n3MR1e8PHMIf+B9/gBkYJOQ</latexit>

e e
<latexit sha1_base64="w+i1B+t+KG5rsATvpJEXatKFPcg=">AAAB+HicbVDLSgNBEJz1GeMjqx69DAbBg4RdUfQY9OIxAfOAZAmzs51kyOyDmV4lLvkSLx4U8eqnePNvnCR70MSChqKqm+4uP5FCo+N8Wyura+sbm4Wt4vbO7l7J3j9o6jhVHBo8lrFq+0yDFBE0UKCEdqKAhb6Elj+6nfqtB1BaxNE9jhPwQjaIRF9whkbq2aX6Ge0+igBQyABovWeXnYozA10mbk7KJEetZ391g5inIUTIJdO64zoJehlTKLiESbGbakgYH7EBdAyNWAjay2aHT+iJUQLaj5WpCOlM/T2RsVDrceibzpDhUC96U/E/r5Ni/9rLRJSkCBGfL+qnkmJMpynQQCjgKMeGMK6EuZXyIVOMo8mqaEJwF19eJs3zintZceoX5epNHkeBHJFjckpcckWq5I7USINwkpJn8krerCfrxXq3PuatK1Y+c0j+wPr8AZ4bkmo=</latexit> <latexit sha1_base64="w+i1B+t+KG5rsATvpJEXatKFPcg=">AAAB+HicbVDLSgNBEJz1GeMjqx69DAbBg4RdUfQY9OIxAfOAZAmzs51kyOyDmV4lLvkSLx4U8eqnePNvnCR70MSChqKqm+4uP5FCo+N8Wyura+sbm4Wt4vbO7l7J3j9o6jhVHBo8lrFq+0yDFBE0UKCEdqKAhb6Elj+6nfqtB1BaxNE9jhPwQjaIRF9whkbq2aX6Ge0+igBQyABovWeXnYozA10mbk7KJEetZ391g5inIUTIJdO64zoJehlTKLiESbGbakgYH7EBdAyNWAjay2aHT+iJUQLaj5WpCOlM/T2RsVDrceibzpDhUC96U/E/r5Ni/9rLRJSkCBGfL+qnkmJMpynQQCjgKMeGMK6EuZXyIVOMo8mqaEJwF19eJs3zintZceoX5epNHkeBHJFjckpcckWq5I7USINwkpJn8krerCfrxXq3PuatK1Y+c0j+wPr8AZ4bkmo=</latexit>

Q, Q Q, Q

Figure 10.10: The effect of a non-canonical Kähler potential on the ADS potential.
The picture on the right cannot hold if the assumption of mass gap for pure SYM
is correct.

holomorphic analysis cannot see), but it would not lift the absolute supersymmetric
minima at infinity, a region where the Kähler potential is nearly canonical in the
UV-variables Q and Q. e On the other hand, no supersymmetric minima can arise at
finite distance in field space. These would correspond to singularities of the Kähler
metric, implying that at those specific points in field space extra massless degrees
of freedom show-up. This cannot be, if the assumption of mass gap for pure SYM
(to which the theory reduces at low enough energy, at generic points in the classical
moduli space) is correct.

10.4.3 Integrating in and out: the linearity principle

The superpotential of pure SYM is sometime written as

 
S
WVY = N S 1 − log 3 (10.69)
Λ
1 α
where S = − 32π 2 Tr W Wα is the so-called glueball superfield and the subscript VY

stands for Veneziano-Yankielowicz. Let us first notice that integrating S out (recall
we are supposing pure SYM has a mass gap) we get
   
∂WVY S 1
= N 1 − log 3 + N S − =0, (10.70)
∂S Λ S

235
which implies
hSi = Λ3 . (10.71)
Plugging this back into the VY superpotential gives

WVY = N Λ3 , (10.72)

which is nothing but the effective superpotential of pure SYM we have previously
derived, eq. (10.67). From this view point the two descriptions seem to be equivalent,
at least as far as low enough energies are concerned: the effective superpotential
(10.67) can be obtained from the VY superpotential by integrating S out.
Analogously, the ADS superpotential is sometime written as
  N −F 
1 S det M
WTVY = (N − F ) S 1 − log , (10.73)
N −F Λ3N −F
where TVY stands for Taylor-Veneziano-Yankielowicz. Integrating S out one now
recovers the ADS superpotential. And, consistently, adding a mass term for all mat-
ter fields, ∼ Tr mM , and integrating M out, one gets from the TVY superpotential
the VY superpotential.
The fact is that one can also revert the procedure, and obtain the VY and TVY
superpotentials starting from the expressions (10.67) and (10.66) and integrating
the glueball superfield S in. Doing so one would imagine to loose some information.
After all, the TVY or VY superpotentials include one more dynamical field with
respect to (10.66)-(10.67), the glueball superfield S, so we expect them to contain
some dynamical information more. As we are going to discuss below, this intuition
is not correct: the two descriptions are completely equivalent.
Let us try to be as general as possible and consider a supersymmetric gauge the-
ory admitting also a tree-level superpotential Wtree . Given a set of chiral superfields
Φi , the generic form of such superpotential is
X
Wtree = λr Xr (Φi ) , (10.74)
r

where λr are coupling constants and Xr gauge invariant combinations of the chiral
superfields Φi . In general, one would expect the non-perturbative generated su-
perpotential Wnon-pert to be a (holomorphic) function of the couplings λr , the gauge
invariant operators Xr , and of the dynamical generated scales Λs (we are supposing,
to be as most general as possible, the gauge group not to be simple, hence we allow
for several dynamical scales). In fact, as shown by Intriligator, Leigh and Seiberg,

236
Wnon-pert does not depend on the couplings λr . This fact implies that the full ef-
fective superpotential (which includes both the tree level and the non-perturbative
contributions) is linear in the couplings, and hence this is sometime referred to as
linearity principle. The upshot is that, in general, we have
X
Weff = λr Xr + Wnon-pert (Xr , Λs ) . (10.75)
r

Let us focus on the dependence on, say, λ1 . At low enough energy (where the
superpotential piece dominates - let us suppose for now that X1 is massive) we
can integrate out the field X1 by solving its F-term equation, which, because of
eq. (10.75), reads
∂
λ1 = − Wnon-pert . (10.76)
∂X1
The above equation is the same as a Legendre transform. In other words, the
coupling λr and the gauge invariant operator Xr behave as Legendre dual variables.
Solving for X1 in terms of λ1 and all other variables, and substituting in eq. (10.75),
one obtains an effective superpotential with a complicated dependence on λ1 but
where X1 has been integrated out. Repeating the same reasoning for all Xr one can
integrate out all fields and end-up with an effective superpotential written in terms
of couplings only
" #
X
Weff (λr , Λs ) = λr Xr + Wnon-pert (Xr , Λs ) . (10.77)
r Xr (λ,Λ)

The point is that the Legendre transform is invertible. Therefore, as we can integrate
out a field, we can also integrate it back in, by reversing the procedure
∂
hXr i = Weff (λr , Λs ) . (10.78)
∂λr
The reason why the two descriptions, one in terms of the fields, one in terms of the
dual couplings, are equivalent is because we have not considered D-terms. D-terms
contain the dynamics (e.g. the kinetic term). Hence, if we ignore D-terms, namely
if we only focus on holomorphic terms as we are doing here, integrating out or in a
field is an operation which does not make us loose or gain information. As far as the
holomorphic part of the effective action is concerned, a field and its dual coupling
are fully equivalent.
What about the dynamical scales Λs ? Can one introduce canonical pairs for
them, too? The answer is yes, and this is where the physical equivalence between

237
ADS and TVY superpotentials we claimed about becomes explicit. Let us consider
pure SYM, for definiteness. One can write the gauge kinetic term as a contribution
to the tree level superpotential in the sense of eq. (10.74)
 
τ (µ) α Λ
Wtree = Tr W Wα = 3N log S, (10.79)
16πi µ

where S is a X-like field and 3N log (Λ/µ) the dual coupling. In other words, one
can think of S and log Λ as Legendre dual variables. From this view point, the SYM
superpotential (10.67) is an expression of the type (10.77), where the field S has
been integrated out and the dependence on the dual coupling is hence non-linear.
Indeed (10.67) can be re-written as
1 Λ
WSYM = N Λ3 = N µ3 e N 3N log µ , (10.80)

where the coupling appears non-linearly. Using now eq. (10.78) applied to this dual
pair, one gets
1 ∂
hSi = Λ Weff = Λ3 . (10.81)
3N ∂Λ
Therefore
 
Λ S
Wnon-pert (S) = Weff − Wtree = N S − 3N log S = N S − N S log 3 , (10.82)
µ µ
which is correctly expressed, according to the linearity principle, in terms of S only,
and not the coupling, log Λ. We can now add the two contributions, the one above
and (10.79) and get for the effective superpotential an expression in the form (10.75)
 
S
Weff = Wnon-pert + Wtree = N S 1 − log 3 (10.83)
Λ
which is nothing but the VY superpotential! The same reasoning can be applied to
a theory with flavor and/or with multiple dynamical scales. The upshot is one and
the same: integrating in (TVY) or out (ADS) fields holomorphically, are operations
which one can do at no cost. The two descriptions are physically equivalent.
In the table below we summarize the relation between couplings and dual field
variables for the most generic situation

Couplings | b11 log Λµ1 b21 log Λµ1 . . . λ1 λ2 . . .

Fields | S1 S2 ... X 1 X2 . . .

Suppose that the mass spectrum of above (composite) fields is as follows

238
 m
<latexit sha1_base64="AV9aJWW4ksQKVrrKAQbweHyzc1s=">AAAB6HicbVBNS8NAEJ3Ur1q/qh69LBbBU0lE0WPRi8cW7Ae0oWy2k3btbhJ2N0IJ/QVePCji1Z/kzX/jts1BWx8MPN6bYWZekAiujet+O4W19Y3NreJ2aWd3b/+gfHjU0nGqGDZZLGLVCahGwSNsGm4EdhKFVAYC28H4bua3n1BpHkcPZpKgL+kw4iFn1FipIfvlilt15yCrxMtJBXLU++Wv3iBmqcTIMEG17npuYvyMKsOZwGmpl2pMKBvTIXYtjahE7WfzQ6fkzCoDEsbKVmTIXP09kVGp9UQGtlNSM9LL3kz8z+umJrzxMx4lqcGILRaFqSAmJrOvyYArZEZMLKFMcXsrYSOqKDM2m5INwVt+eZW0LqreVdVtXFZqt3kcRTiBUzgHD66hBvdQhyYwQHiGV3hzHp0X5935WLQWnHzmGP7A+fwB1t+M9Q==</latexit>

S, log ⇤
<latexit sha1_base64="ExxDh6pkSwmqTOkGyiHGYmEavTk=">AAAB+HicdVDLSgMxFM34rPXRUZdugkVwIUPGVqy7ohsXLiraB3SGkslk2tDMZEgyQh36JW5cKOLWT3Hn35g+BBU9EDiccw/35gQpZ0oj9GEtLC4tr6wW1orrG5tbJXt7p6VEJgltEsGF7ARYUc4S2tRMc9pJJcVxwGk7GF5M/PYdlYqJ5FaPUurHuJ+wiBGsjdSzSzdH0OOiD70rEwpxzy4jp3JmUIPIqSKEXDQnqAJdB01RBnM0eva7FwqSxTTRhGOlui5KtZ9jqRnhdFz0MkVTTIa4T7uGJjimys+nh4/hgVFCGAlpXqLhVP2eyHGs1CgOzGSM9UD99ibiX14301HNz1mSZpomZLYoyjjUAk5agCGTlGg+MgQTycytkAywxESbroqmhK+fwv9J69hxTxx0XS3Xz+d1FMAe2AeHwAWnoA4uQQM0AQEZeABP4Nm6tx6tF+t1NrpgzTO74Aest0+w3pJ6</latexit>

massive fields
<latexit sha1_base64="FaieaP/itsaPsuA4/Lr8hl2cTqY=">AAAB/3icbVDLSsNAFJ3UV62vqODGTbAIrkoiioKbohuXFewDmlAm05t26EwmzEyKJXbhr7hxoYhbf8Odf+O0zUJbD1w4nHPvzL0nTBhV2nW/rcLS8srqWnG9tLG5tb1j7+41lEglgToRTMhWiBUwGkNdU82glUjAPGTQDAc3E785BKmoiO/1KIGA415MI0qwNlLHPvB5KB4yjpWiQ/CvIgqsq8Ydu+xW3CmcReLlpIxy1Dr2l98VJOUQa8LMY23PTXSQYakpYTAu+amCBJMB7kHb0BhzUEE23X/sHBul60RCmoq1M1V/T2SYKzXioenkWPfVvDcR//PaqY4ug4zGSaohJrOPopQ5WjiTMJwulUA0GxmCiaRmV4f0scREm8hKJgRv/uRF0jiteOcV9+6sXL3O4yiiQ3SETpCHLlAV3aIaqiOCHtEzekVv1pP1Yr1bH7PWgpXP7KM/sD5/AM81lps=</latexit>

X 0, 0
<latexit sha1_base64="YhD0h3KRd7GX3y0WS8vqCm8R4c4=">AAAB8nicdVDLSgMxFM34rPVVdekmWKQuZMjYinVXdOOygn3AdCiZTKYNzSRDkhFK6We4caGIW7/GnX9j2o6gogcCh3POJfeeMOVMG4Q+nKXlldW19cJGcXNre2e3tLff1jJThLaI5FJ1Q6wpZ4K2DDOcdlNFcRJy2glH1zO/c0+VZlLcmXFKgwQPBIsZwcZKfrdy2uM2HeFKv1RGbvXSog6RW0MIeSgnqAo9F81RBjma/dJ7L5IkS6gwhGOtfQ+lJphgZRjhdFrsZZqmmIzwgPqWCpxQHUzmK0/hsVUiGEtlnzBwrn6fmOBE63ES2mSCzVD/9mbiX56fmbgeTJhIM0MFWXwUZxwaCWf3w4gpSgwfW4KJYnZXSIZYYWJsS0Vbwtel8H/SPnO9cxfd1sqNq7yOAjgER+AEeOACNMANaIIWIECCB/AEnh3jPDovzusiuuTkMwfgB5y3T2//kLY=</latexit>

X, massless fields
<latexit sha1_base64="837Cf+aNMrbXGGcTRrWAQB6PslE=">AAACAHicbVDLSsNAFJ3UV62vqAsXboJFcFUSURTcFN24rGAf0IQymd60Q2cmYWYilpCNv+LGhSJu/Qx3/o3TNgttPXDhcM69M/eeMGFUadf9tkpLyyura+X1ysbm1vaOvbvXUnEqCTRJzGLZCbECRgU0NdUMOokEzEMG7XB0M/HbDyAVjcW9HicQcDwQNKIEayP17AOfh/FjxrFSDJTyryIKrK/ynl11a+4UziLxClJFBRo9+8vvxyTlIDRh5rWu5yY6yLDUlDDIK36qIMFkhAfQNVRgDirIpgfkzrFR+k4US1NCO1P190SGuVJjHppOjvVQzXsT8T+vm+roMsioSFINgsw+ilLm6NiZpOH0qQSi2dgQTCQ1uzpkiCUm2mRWMSF48ycvktZpzTuvuXdn1fp1EUcZHaIjdII8dIHq6BY1UBMRlKNn9IrerCfrxXq3PmatJauY2Ud/YH3+AKxClxg=</latexit>

<latexit sha1_base64="OBKAS1OQCPXqYJXxl9+JeYWGZuo=">AAAB8HicdVDLSgMxFM3UV62vqks3wSK4kCFjK9Zd0Y3LCvYh7VAymUwbmmSGJCOUoV/hxoUibv0cd/6NaTuCih4IHM45l9x7goQzbRD6cApLyyura8X10sbm1vZOeXevreNUEdoiMY9VN8CaciZpyzDDaTdRFIuA004wvpr5nXuqNIvlrZkk1Bd4KFnECDZWuuue9LkNh3hQriC3emFRh8itIYQ8lBNUhZ6L5qiAHM1B+b0fxiQVVBrCsdY9DyXGz7AyjHA6LfVTTRNMxnhIe5ZKLKj2s/nCU3hklRBGsbJPGjhXv09kWGg9EYFNCmxG+rc3E//yeqmJ6n7GZJIaKsnioyjl0MRwdj0MmaLE8IklmChmd4VkhBUmxnZUsiV8XQr/J+1T1ztz0U2t0rjM6yiCA3AIjoEHzkEDXIMmaAECBHgAT+DZUc6j8+K8LqIFJ5/ZBz/gvH0CqtaQVA==</latexit>

where the X’s are a set of massless (possibly composite) fields, the X 0 ’s are massive
ones, and the Λ’s are dynamical scales (glueball superfields are all massive because
of mass gap of the pure glue theory, i.e. all Λ0 s 6= 0). The most Wilsonian thing
to do would be to describe the effective superpotential in terms of fields X, and
couplings λ0 and Λ
Weff = Weff (X, λ0 , Λ) . (10.84)
In this sense, the ADS superpotential is more Wilsonian than the TVY. Seemingly,
for pure SYM the most Wilsonian thing to do is to express the effective superpo-
tential as a function of the coupling only (since the glueball superfield is massive),
namely as Weff = N Λ3 . However, since (as far as the holomorphic part of the ef-
fective action is concerned) integrating in and out fields are equivalent operations,
one can very well choose to write down the effective superpotential by integrating X
fields out and X 0 and S fields in (or anything in between these two extreme cases)

Weff = Weff (λ, X 0 , S) , (10.85)

getting an equivalent way of describing the low energy effective theory superpoten-
tial. This said, one should bare in mind that as far as the massless fields X, there
is no actual energy range for which integrating them out makes real physical sense,
and this would be indicated by the Kähler potential of the effective theory being
ill-defined (in other words, there is no energy range in which the kinetic term of such
massless fields is negligible, since the energy is always bigger or equal than the field
mass, which is vanishing). On the contrary, in presence of a mass gap, that is in the
absence of X-like fields, the two descriptions, one in terms of couplings the other in
terms of fields, are equivalent, since now no singularities are expected in the Kähler
potential. And this is a more and more exact equivalence the lower the energy.

239
10.4.4 F ≥ N : quantum moduli space

Let us now go back to our analysis of the IR dynamics of SQCD with gauge group
SU (N ) and F flavors. What about the case F ≥ N ? As we are going to see, things
change drastically. For one thing, a properly defined effective superpotential cannot
be generated. There is no way of constructing an object respecting all symmetries,
with the correct dimension, and being vanishing in the classical limit, using couplings
and fields we have (mesons, baryons and the dynamical scale Λ). This has the effect
that for F ≥ N the classical moduli space is not lifted. This does not mean nothing
interesting happens. For instance, the moduli space can be deformed by strong
dynamics effects. Moreover, the perturbative analysis does not tell us what the low
energy effective theory looks like; as we will see instead (mainly using holomorphy
arguments), in some cases we will be able to make very non-trivial statements about
the way light degrees of freedom interact, and in turn about the phase the theory
enjoys.
In what follows, we will consider qualitatively different cases separately. Let us
start analyzing the case F = N . It is easy to see that in this case all gauge invariant
operators have R-charge R = 0, so one cannot construct an effective superpotential
with R = 2. However, as we are going to show, something does happen due to
strong dynamics.
Besides the mesons, there are now two baryons

B = a1 a2 ...aN Q1a1 Q2a2 . . . QN

aN

e = a1 a2 ...a Q
B ea11 Q
ea22 . . . Q
eaN .
N N

The classical moduli space is parameterized by VEVs of mesons and baryons. There
is, however, a classical constraint between them
e=0
det M − B B (10.86)

(this comes because for N = F we have that det Q = B and det Q e = Be and
the determinant of the product is the product of the determinants). One can ask
whether this classical constraint is modified at the quantum level. In general, one
could expect the quantum version of the above classical constraint to be
e = a Λ2N ,
det M − B B (10.87)

where a is a (undetermined for now) dimensionless and charge-less constant. One

can get easily convinced that this is the only possible modification compatible with

240
all physical requirements. First it has the correct physical dimension, namely the
same as the l.h.s, ∆ = 2N . Second, it correctly vanishes in the classical limit, Λ → 0.
Third, it has vanishing R-charge and U (1)A charge 2N , as the l.h.s has. Finally, it
is suggestive that the power Λ enters in eq. (10.87), 2N , is the one-loop coefficient of
the β-function and is exactly that associated with a one instanton correction, since
for N = F we have for the instanton action
2
− 8π2 +iθYM
e−Sinst ∼ e g ∼ Λ2N . (10.88)

There are no symmetry reasons not to allow for it (modulo the constant a which
can very well be vanishing, after all). So, given that in principle a modification like
(10.87) is allowed, everything boils down to determine whether the constant a is
vanishing or has a finite value.
The constraint (10.87) can be implemented, formally, by means of a Lagrange
multiplier, allowing a superpotential
 
e − aΛ2N
W = A det M − B B (10.89)

where A is the Lagrange multiplier, whose equation of motion is by construction the

constraint (10.87). The interesting thing is that one can use holomorphic decoupling
to fix the constant a. Adding a mass term for the N -th flavor, W = mM NN , the
low energy theory reduces to SQCD with F = N − 1. Imposing that after having
integrated out the N -th flavor one obtains an effective superpotential which matches
the ADS superpotential for F = N − 1, fixes a = 1, that is

e = Λ2N .
det M − B B (10.90)

So the quantum constraint is there, after all. Actually, it is necessary for it to

be there in order to be consistent with what we already know about the quantum
properties of SQCD with F < N !
Several comments are in order at this point.
This is the first case where a moduli space of supersymmetric vacua persists at
the quantum level. Still, the quantum moduli space is different from the classical
one. The moduli space (10.86) is singular. It has a singular submanifold reflecting
the fact that on this submanifold additional massless degrees of freedom arise. This
is the submanifold where not only (10.86) is satisfied, but also d(det M − B B) e =
0, which makes the tangent space singular and therefore good local coordinates
not being well-defined. This happens whenever baryon VEVs vanish, B = B e =

241
0, and the meson matrix has rank k ≤ N − 2. On this subspace a SU (N − k)
gauge group remains unbroken, and corresponding gluons (as well as some otherwise
massive matter fields) remain massless. The quantum moduli space (10.90) is instead
smooth. Basically, when B = B e = 0 the rank of the meson matrix is not diminished
since its determinant does not vanish, now: everywhere on the quantum moduli
space the gauge group is fully broken.
Classically, the origin is part of the space of vacua. Hence, chiral symmetry can
be unbroken. At the quantum level, instead, the origin is excised so in any allowed
vacuum chiral symmetry is broken (like in QCD). Moreover, being the moduli space
non-singular, means there are no massless degrees of freedom other than mesons and
baryons. But the latter are indeed massless, since are moduli. Hence in SQCD with
N = F there is no mass gap (as for massless QCD). By supersymmetry, there are
also massless composite fermions.
Obviously, the chiral symmetry breaking pattern is not unique. Different points
on the moduli space display different patterns. At a generic point, where all gauge
invariant operators get a VEV, all global symmetries are broken. But there are
submanifolds of enhanced global symmetry. For instance, along the mesonic branch,
defined as
M ij = Λ2 δji , B = Be=0, (10.91)
we have that

SU (F )L × SU (F )R × U (1)B × U (1)R −→ SU (F )D × U (1)B × U (1)R , (10.92)

a chiral symmetry breaking pattern very much similar to QCD. Along the baryonic
branch, which is defined as

M ij = 0 , e = ΛN ,
B = −B (10.93)

we have instead

SU (F )L × SU (F )R × U (1)B × U (1)R −→ SU (F )L × SU (F )R × U (1)R , (10.94)

which is very different from QCD (the full non-abelian chiral symmetry is preserved).
Which phases does the theory enjoy? The point where all field VEVs vanish,
i.e. the origin, is excised. Therefore, the gauge group is always broken and the
theory is hence in a Higgs phase. Still, near the origin the theory can be better
thought to be in a confined phase, since the effective theory is smooth in terms of

242
mesons and baryons, and, moreover, we are in the strongly coupled region of field
space, where an inherently perturbative Higgs description is not fully appropriate.
In fact, there is no order parameter which can distinguish between the two phases;
there is no phase transition between them (this is similar to the prototype example
of one-family EW theory we discussed already). In this respect, notice that the
Wilson loop is not a useful order parameter here since it follows the perimeter law,
no matter where one sits on the moduli space: we do not have strict confinement
but just charge screening, as in QCD, since we have (light) matter transforming in
the fundamental representation of the gauge group, and therefore flux lines can (and
do) break. The qualitative difference between classical and quantum moduli spaces,
and their interpretation is depicted in Figure 10.11.

Higgs

Confinement
Gluons
(charge screening)

Classical Quantum
moduli space moduli space

Figure 10.11: Classical picture (left): at the origin gauge symmetry is recovered,
and chiral symmetry is not broken. Quantum picture (right): the (singular) origin
has been replaced by a circle of theories where chiral symmetry is broken (rather
than Higgs phase, this resembles more closely the physics of a confining vacuum).

A non-trivial consistency check of this picture comes from computing ’t Hooft

anomalies in the UV and in the IR. Let us consider, for instance, the mesonic branch.
The charges under the unbroken global symmetries, SU (F )D × U (1)B × U (1)R of

243
the UV (fundamental) and IR (composite) degrees of freedom are as follows

SU (F )D U (1)B U (1)R
ψQ F 1 −1
ψQe F −1 −1
λ • 0 1

ψM Adj 0 −1
ψB • F −1
ψBe • −F −1

where we have used the constraint (10.90) to eliminate the fermionic partner of
Tr M , so that ψM transforms in the Adjoint of SU (F )D . We can now compute
diverse triangular anomalies and see whether computations done in terms of UV
and IR degrees of freedom agree. We get

UV IR
1
SU (F )2D U (1)R 2N 2
= −N
(−1) F (−1) = −F
(10.95)
U (1)2B U (1)R −2N F −2F 2
U (1)3R −2N F + N 2 − 1 −(F 2 − 1) − 1 − 1 = −F 2 − 1

Since (crucially!) F = N we see that ’t Hooft anomaly matching holds. A similar

computation can be done for the baryonic branch finding again perfect agreement
between the UV and IR ’t Hooft anomalies. This rather non-trivial agreement
ensures that our low energy effective description in terms of mesons and baryons,
subject to the constraint (10.90), is most likely correct.

Let us move on and consider the next case, F = N + 1. The moduli space is
again described by mesons and baryons. We have N + 1 baryons of type B and
e now
N + 1 baryons of type B

Bi = ij1 ...jN a1 a2 ...aN Qja11 Qja22 . . . QjaNN

e i = ij1 ...jN a1 a2 ...a Q
B ea2
ea1 Q eaN
N j1 j2 . . . QjN .

As we are going to show, differently from the previous case, the classical moduli
space not only is unlifted, but is quantum exact, also. In other words, there are no
quantum modifications to it.
This result can be proved using holomorphic decoupling. The rationale goes as
follows. As proposed by Seiberg, this system can be described, formally, by the

244
following superpotential
a  
Weff = ej
det M − Bi M ij B , (10.96)
Λ2N −1
where i = 1, 2, . . . , N +1 is a flavor index, 2N −1 is the one-loop β-function coefficient
and a, as usual, is for now an undetermined coefficient. The above superpotential
has all correct symmetry properties, including the R-charge, which is indeed equal to
2. Notice, though, that since the rank of the meson matrix k ≤ N , then det M = 0,
classically. So the above equation should be really thought of as a quantum equation,
valid off-shell, so to say.
Let us now add a mass m to the F -th flavor. This gives
a  
e j − mM N+1 .
Weff = 2N −1 det M − Bi M ij B N+1 (10.97)
Λ
Integrating out massive modes, which tantamounts to impose the F-flatness condi-
tions for M N+1i , M iN+1 , Bi and Be i for i < N + 1, reduces the meson matrix and the
baryons to ! ! !
M̂ij 0 0i 0i
M= , B= , B e= (10.98)
0 t B̂ ê
B
where now i, j = 1, . . . , N , and t = MN+1,N+1 . The F-flatness condition for t reads
a  ê

det M̂ − B̂ B −m=0 (10.99)
Λ2N −1
which implies
ê 1 1
det M̂ − B̂ B = m Λ2N −1 = Λ2N , (10.100)
a a L
where in the last step we have used the relation (10.55). This shows that the
ansatz (10.96) is correct, since upon holomorphic decoupling we get exactly the
quantum constraint of F = N SQCD (and a gets fixed to one). From eq. (10.96),
by differentiating with respect to M , Bi and B e i we get the moduli space equations
(i.e. the classical, still quantum exact, constraints between baryons and mesons)
(
M ·B e =B·M =0
(10.101)
j
det M · (M −1 )i − Bi Bej = 0

j
where detM · (M −1 )i ≡ minor {M }ji = (−1)i+j × det of the matrix obtained from
M by omitting the i-th row and the j-th column (recall that above equations are
on-shell, and on-shell detM itself vanishes).

245
 As a non-trivial check of this whole picture one can verify, choosing any preferred
point in the space of vacua, that ’t Hooft anomalies match (and hence that our
effective description holds).
Now that we know eq. (10.96) is correct, let us try to understand what does it
tell us about the vacuum structure of SQCD with F = N + 1. First, unlike F = N
SQCD, the origin of field space, M = B = B e = 0, is part of the moduli space. In
such vacuum chiral symmetry is unbroken. This is an instance of a theory displaying
confinement (actually charge screening) without chiral symmetry breaking. Theories
with such a property, like F = N + 1 SQCD at the origin of field space, are said to
be s-confining.
Classically, the singularities at the origin are interpreted as extra massless gluons
(and matter fields), since the theory gets unhiggsed for vanishing values of matter
field VEVs. At the quantum level, the physical interpretation is different, since
because the theory is UV-free, the region around the origin is the more quantum
one. Singularities are more naturally associated with additional massless mesons
and baryons which pop-up since eqs. (10.101) are trivially realized at the origin,
and do not provide any actual constraint between meson and baryon components.
In other words, at the origin the number of mesonic and baryonic massless degrees
of freedom is larger than the dimension of the moduli space. This can be checked,
again, by ’t Hooft anomaly matching.
Note, finally, that as for F = N SQCD, also this theory exhibits complementarity,
in the sense that one can move smoothly from a confining phase (near the origin)
to a Higgs phase (at large field VEVs) without any order parameter being able to
distinguish between them (i.e. the Wilson loop follows the perimeter law in both
phases).

One could try to go further, and apply the same logic to F = N + 2 (and on).
On general ground one would expect M , B ij , B eij (baryons have now two free flavor
indices) to be the dynamical degrees of freedom in the IR, and could then try to
construct an effective (off-shell) superpotential of the kind of (10.96). This, however,
does not work. Looking at the charges of the various gauge invariant operators and
dynamical scale Λ one can easily see that an effective superpotential with R-charge
equal to 2, correct physical dimensions and symmetries, cannot be constructed.
Indeed, the only SU (F )L × SU (F )R invariant superpotential one could construct

246
should be the obvious generalization of (10.96), that is

l e jm
Weff ∼ det M − Bil Mji Mm B , (10.102)

which, to start with, does not have R = 2 but actually R = 4 (things get worse
the larger the number of flavors). Even ’t Hooft anomaly matching condition can
be proven not to work. For instance, choosing for simplicity the origin of field
space where meson and baryons are unconstrained, one can easily see that ’t Hooft
anomalies do not match. More generally, it turns out that increasing F , ’t Hooft
anomaly coefficients computed using (unconstrained) IR degrees of freedom increase
much faster than those computed using UV degrees of freedom, and only for F =
N + 1 they match.
In fact, things turn out to be rather different. As we will show, the correct degrees
of freedom to describe the dynamics around SQCD vacua for F = N + 2 are those
of an IR-free theory (!) described by SU (2) SYM coupled to F chiral superfields
q transforming in the fundamental of SU (2), F chiral superfields q̃ transforming
in the anti-fundamental and F 2 singlet chiral superfields Φ, plus a cubic tree level
superpotential coupling q, q̃ and Φ. What’s that?
Two pieces of information are needed in order to understand this apparently
weird result and, more generally, to understand what is going on for F ≥ N + 2.
Both are due to Seiberg. In the following we will review them in turn.

10.4.5 Conformal window

A first proposal is that SQCD in the range 23 N < F < 3N flows to an interacting IR
fixed point (meaning it does not confine!). In other words, even if the theory is UV-
free and hence the gauge coupling g increases through the IR, at low energy g reaches
a constant RG-fixed value. Let us try to see how such claim comes about. The
SQCD β-function for the physical gauge coupling (which hence takes into account
wave-function renormalization effects) is

g 3 3N − F [1 − γ(g 2 )]
β(g) = − , (10.103)
16π 2 1 − N g 2 /8π 2

where γ is the anomalous dimension of matter fields and can be computed in per-
turbation theory to be

g2 N 2 − 1
γ(g 2 ) = − + O(g 4 ) . (10.104)
8π 2 N
247
Expanding formula (10.103) in powers of g 2 we get
   2 
g3 2 F g 4
β(g) = − 3N − F + 3N − 2F N + + O(g ) . (10.105)
16π 2 N 8π 2
From the above expression it is clear that there can exist values of F and N such
that the one-loop contribution is negative but the two-loops contribution is positive.
This suggests that in principle there could be a non-trivial fixed point, a value of
the gauge coupling g = g∗ , for which β(g∗ ) = 0.
Let us consider F slightly smaller than 3N . Defining
F
=3− << 1 (10.106)
N
we can re-write the β-function as
 
g3  2
 g2 4
β(g) = − N − 3(N − 1) + O() + O(g ) . (10.107)
16π 2 8π 2
The first term inside the parenthesis is positive while the second is negative and
hence we see we have a solution β(g) = 0 at

8π 2 N
g∗2 = , (10.108)
3 N2 − 1
up to O(2 ) corrections. This is called Banks-Zaks (BZ) fixed point. Seiberg argued
that an IR fixed point like the one above exists not only for F so near to 3N but
actually for any F in the range 23 N < F < 3N , the so-called conformal window.
According to this proposal, the IR dynamics of SQCD in the conformal window
is described by an interacting superconformal theory: quarks and gluons are not
confined but appear as interacting massless particles, the Coulomb-like potential
being
g2
V (r) ∼ ∗ . (10.109)
r
Hence, according to this proposal, SQCD in the conformal window enjoys a non-
abelian Coulomb phase.
Let us try to understand why the conformal window is bounded from below and
from above. The possibility of making exact computations in a SCFT shows that
for F < 23 N the theory should be in a different phase. In a SCFT the dimension of
a field satisfies the following relation
3
∆ ≥ |R| , (10.110)
2
248
where R is the field R-charge (recall that in a SCFT the generator of the R-symmetry
enters the algebra, and hence an R-symmetry is always present). The equality holds
for chiral (or anti-chiral) operators. This implies that
3 3 e = 3 F − N ≡ 2 + γ∗ ,
∆(M ) = R(M ) = R(QQ) (10.111)
2 2 F
given that M is a chiral operator. This means that the anomalous dimension of the
meson matrix at the IR fixed point is γ∗ = 1 − 3N/F .
The dimension of a scalar field must satisfy

∆≥1. (10.112)

Indeed, when ∆ < 1 the operator, which is in a unitary representation of the su-
perconformal algebra, would include a negative norm state which cannot exist in a
unitary theory. This implies that F = 32 N is a lower bound since there ∆(M ) = 1
and lower values of F make no sense (recall that the lowest component of the su-
perfield M is a scalar field): for F < 32 N the theory should be in a different phase.
A clue to what such phase could be is that at F = 32 N the field M becomes free.
Indeed, for F = 23 N we get that ∆(M ) = 1 which is possible only for free, non-
interacting scalar fields. Perhaps it is the whole theory of mesons and baryons which
becomes free, somehow. We will make this intuition more precise later.
As for the upper bound, let us notice that for F ≥ 3N SQCD is not asymp-
totically free anymore, since the β-function changes sign (actually, for F = 3N the
one-loop β function is 0 but one can show that the two-loop contribution is positive).
The spectrum at large distance consists of elementary quarks and gluons interacting
through a potential
g2 1
V ∼ with g 2 ∼ , (10.113)
r log(rΛ)
which implies that SQCD is in a non-abelian free phase. It is interesting to notice
that for F = 3N the anomalous dimension of M is actually zero, consistent with
the fact that from that value on, the IR dimension of gauge invariant operators is
not renormalized since the theory becomes IR-free. A summary of the IR behavior
of SQCD for F > 23 N is reported in Figure 10.12.

10.4.6 Electric-magnetic duality (aka Seiberg duality)

The second proposal put forward by Seiberg regards the existence of a electro-
magnetic-like duality. The IR physics of SQCD for F > N + 1 has an equivalent

249
 <latexit sha1_base64="PnMK3IfvzJcmT3kptw6oCsBznCg=">AAAB/3icbVDLSgNBEOyNrxhfq4IXL4NB8BR2E0UPHoKCeAoRzAOSJcxOZpMhsw9mZoWw7sFf8eJBEa/+hjf/xkmyB00saCiquunuciPOpLKsbyO3tLyyupZfL2xsbm3vmLt7TRnGgtAGCXko2i6WlLOANhRTnLYjQbHvctpyR9cTv/VAhWRhcK/GEXV8PAiYxwhWWuqZB11PYJJU0qScohq6RDe6KrWeWbRK1hRokdgZKUKGes/86vZDEvs0UIRjKTu2FSknwUIxwmla6MaSRpiM8IB2NA2wT6WTTO9P0bFW+sgLha5Aoan6eyLBvpRj39WdPlZDOe9NxP+8Tqy8CydhQRQrGpDZIi/mSIVoEgbqM0GJ4mNNMBFM34rIEOtAlI6soEOw519eJM1yyT4rWXenxepVFkceDuEITsCGc6jCLdShAQQe4Rle4c14Ml6Md+Nj1pozspl9+APj8wfTe5QF</latexit>

3 IR UV
N < F < 3N >
>tixetal/<==gxMO6j/8JOfEMHPPpirn3jd6dc5F953UIenR4AAgWNPAnqKAO8zEIOjBYUHvpe5Wk+cbVXiz78Jd5lFzLEanysjJjikajE3eLFKLB2EbmQ5Svm9LjYipAmRxX2oUDJLzY/vmQq6M//Vcjo1zA1GZHwj+oWGxYivruZDlEbsG2JS3qTCR3O24H1jkkUHDWHRILhGn6uUwZCcKDqPbsJn746UNRDoYkljZv66fVrHFClcyLTyawdVsrKXZa/SFLqOPUyDI6PBhDzfwSelKkefn6ujWAbIQRaKcCbIwNMYfZBaA6OVhYsghtYpe8UhFs093b3d2exbtgN3Y91WZld9f9GcFBI8y8w3wLeY4jtDuZbG//mJ/zqEDFe8ug+SQobsbSza3+Jsl9Qo9A2t84FiKpeokoK4JSxXOa1XteW43JEAN8SNBVbciH6BAAA>"=cz7jlbmXG9IjiEZfZyjht9/+L35"=46esab_1ahs tixetal<

2 g = g⇤ g=0
<latexit sha1_base64="rucJaNUN+Tpw2WGYXz8raRzDFdc=">AAAB6nicbVBNS8NAEJ3Ur1q/qh69LBbBU0lE0YtQ9OKxov2ANpTNdtMu3WzC7kQooT/BiwdFvPqLvPlv3LY5aOuDgcd7M8zMCxIpDLrut1NYWV1b3yhulra2d3b3yvsHTROnmvEGi2Ws2wE1XArFGyhQ8naiOY0CyVvB6Hbqt564NiJWjzhOuB/RgRKhYBSt9DC4dnvlilt1ZyDLxMtJBXLUe+Wvbj9macQVMkmN6Xhugn5GNQom+aTUTQ1PKBvRAe9YqmjEjZ/NTp2QE6v0SRhrWwrJTP09kdHImHEU2M6I4tAselPxP6+TYnjlZ0IlKXLF5ovCVBKMyfRv0heaM5RjSyjTwt5K2JBqytCmU7IheIsvL5PmWdW7qLr355XaTR5HEY7gGE7Bg0uowR3UoQEMBvAMr/DmSOfFeXc+5q0FJ585hD9wPn8Avd+NcA==</latexit>

<latexit sha1_base64="FL7ODKgRf2VVOY2hEJnLZUyPz+0=">AAAB7HicbVBNS8NAEJ2tX7V+VT16WSyCeCiJKHoRil48VjBtoQ1ls92kSzebsLsRSuhv8OJBEa/+IG/+G7dtDtr6YODx3gwz84JUcG0c5xuVVlbX1jfKm5Wt7Z3dver+QUsnmaLMo4lIVCcgmgkumWe4EayTKkbiQLB2MLqb+u0npjRP5KMZp8yPSSR5yCkxVvKim6h/1q/WnLozA14mbkFqUKDZr371BgnNYiYNFUTrruukxs+JMpwKNqn0Ms1SQkckYl1LJYmZ9vPZsRN8YpUBDhNlSxo8U39P5CTWehwHtjMmZqgXvan4n9fNTHjt51ymmWGSzheFmcAmwdPP8YArRo0YW0Ko4vZWTIdEEWpsPhUbgrv48jJpndfdy7rzcFFr3BZxlOEIjuEUXLiCBtxDEzygwOEZXuENSfSC3tHHvLWEiplD+AP0+QMtRI5E</latexit>

IR UV
F 3N >
<latexit sha1_base64="jlgzOMtRqaw0XxV7tVot6phkCic=">AAAB8HicbVDJSgNBEK2JW4xb1KOXxiB4CjMu6DEoiCeJYBZJhtDTqUmadM+M3T1CGPIVXjwo4tXP8ebf2FkOGn1Q8Hiviqp6QSK4Nq775eQWFpeWV/KrhbX1jc2t4vZOXcepYlhjsYhVM6AaBY+wZrgR2EwUUhkIbASDy7HfeESleRzdmWGCvqS9iIecUWOl+yvS7uEDOb7pFEtu2Z2A/CXejJRghmqn+NnuxiyVGBkmqNYtz02Mn1FlOBM4KrRTjQllA9rDlqURlaj9bHLwiBxYpUvCWNmKDJmoPycyKrUeysB2Smr6et4bi/95rdSE537GoyQ1GLHpojAVxMRk/D3pcoXMiKEllClubyWsTxVlxmZUsCF48y//JfWjsndadm9PSpWLWRx52IN9OAQPzqAC11CFGjCQ8AQv8Ooo59l5c96nrTlnNrMLv+B8fANeY494</latexit>

>tixetal/<==gxMO6j/8JOfEMHPPpirn3jd6dc5F953UIenR4AAgWNPAnqKAO8zEIOjBYUHvpe5Wk+cbVXiz78Jd5lFzLEanysjJjikajE3eLFKLB2EbmQ5Svm9LjYipAmRxX2oUDJLzY/vmQq6M//Vcjo1zA1GZHwj+oWGxYivruZDlEbsG2JS3qTCR3O24H1jkkUHDWHRILhGn6uUwZCcKDqPbsJn746UNRDoYkljZv66fVrHFClcyLTyawdVsrKXZa/SFLqOPUyDI6PBhDzfwSelKkefn6ujWAbIQRaKcCbIwNMYfZBaA6OVhYsghtYpe8UhFs093b3d2exbtgN3Y91WZld9f9GcFBI8y8w3wLeY4jtDuZbG//mJ/zqEDFe8ug+SQobsbSza3+Jsl9Qo9A2t84FiKpeokoK4JSxXOa1XteW43JEAN8SNBVbciH6BAAA>"=cz7jlbmXG9IjiEZfZyjht9/+L35"=46esab_1ahs tixetal<

g=0 any g
<latexit sha1_base64="rucJaNUN+Tpw2WGYXz8raRzDFdc=">AAAB6nicbVBNS8NAEJ3Ur1q/qh69LBbBU0lE0YtQ9OKxov2ANpTNdtMu3WzC7kQooT/BiwdFvPqLvPlv3LY5aOuDgcd7M8zMCxIpDLrut1NYWV1b3yhulra2d3b3yvsHTROnmvEGi2Ws2wE1XArFGyhQ8naiOY0CyVvB6Hbqt564NiJWjzhOuB/RgRKhYBSt9DC4dnvlilt1ZyDLxMtJBXLUe+Wvbj9macQVMkmN6Xhugn5GNQom+aTUTQ1PKBvRAe9YqmjEjZ/NTp2QE6v0SRhrWwrJTP09kdHImHEU2M6I4tAselPxP6+TYnjlZ0IlKXLF5ovCVBKMyfRv0heaM5RjSyjTwt5K2JBqytCmU7IheIsvL5PmWdW7qLr355XaTR5HEY7gGE7Bg0uowR3UoQEMBvAMr/DmSOfFeXc+5q0FJ585hD9wPn8Avd+NcA==</latexit>

<latexit sha1_base64="zbn6FBbKrUjGUMbIFyqxee1DWSI=">AAAB9HicbVDLSgNBEOyNrxhfUY9eFoPgKeyKouAl6MVjBPOA7BJmJ7PJkHmsM7PBsOQ7vHhQxKsf482/cZLsQRMLGoqqbrq7ooRRbTzv2ymsrK6tbxQ3S1vbO7t75f2DppapwqSBJZOqHSFNGBWkYahhpJ0ognjESCsa3k791ogoTaV4MOOEhBz1BY0pRsZKYcAj+ZQhMQ6uJ/1uueJVvRncZeLnpAI56t3yV9CTOOVEGMyQ1h3fS0yYIWUoZmRSClJNEoSHqE86lgrEiQ6z2dET98QqPTeWypYw7kz9PZEhrvWYR7aTIzPQi95U/M/rpCa+CjMqktQQgeeL4pS5RrrTBNweVQQbNrYEYUXtrS4eIIWwsTmVbAj+4svLpHlW9S+q3v15pXaTx1GEIziGU/DhEmpwB3VoAIZHeIZXeHNGzovz7nzMWwtOPnMIf+B8/gAa45JQ</latexit>

Figure 10.12: The IR behavior of SQCD in the window 32 N < F < 3N , where the
theory flows to an IR fixed point with g = g∗ , and for F ≥ 3N , where g∗ = 0 and
the theory is in a non-abelian IR-free phase.

description in terms of another supersymmetric gauge theory, known as the magnetic

dual theory. Such dual theory is IR-free for N + 1 < F ≤ 23 N , and UV-free for
F > 23 N . In the conformal window defined before, that is for 32 N < F < 3N , it
has a IR fixed point (the same as the original SQCD theory!), while for F ≥ 3N ,
where SQCD becomes IR-free, it enters into a confining phase. So these two theories
are very different: as such, the equivalence is an IR equivalence. SQCD, sometime
called electric theory in this context, and its magnetic dual are not equivalent in the
UV neither along the RG-flow. They just provide two equivalent ways to describe
the dynamics around the space of vacua (in fact, perturbing SQCD by suitable
operators, e.g., by quartic operators, one can sometime promote this IR duality to a
full duality, valid along the whole RG; however, discussing such instances is beyond
our present scope).
In order to understand this claim (and its implications), and define such dual
theory more precisely, we first need to do a step back. In trying to extend to higher
values of F the reasoning about SQCD with F = N + 1, one should consider the
following gauge invariant operators
e i1 i2 ...iF −N .
M ij , Bi1 i2 ...iF −N , B (10.114)

The baryons have N e = F − N free indices so, at least group-theoretically, one could
think about them as if they were bound states of N e components, some new quark-
like fields q and q̃ of some supersymmetric gauge theory theory with gauge group
SU (N e ) = SU (F − N ) for which q and q̃ transform in the Ne and Ne representations,
respectively. Then the SQCD baryons would have a dual description as
a
Bi1 i2 ...iNe ∼ a1 a2 ...aNe qia11 qia22 . . . qi eNe (10.115)
N

250
 e Recall that in terms of the original matter fields Q and Q,
and similarly for B. e the
baryons are composite fields made out of N components.
Seiberg made this naive idea concrete (and physical), putting forward the fol-
lowing proposal: SQCD with gauge group SU (N ) and F > N + 1 flavors can be
equivalently described, in the IR, by a different SQCD-like theory with gauge group
SU (F − N ) and F flavors plus an additional chiral superfield Φ which is a gauge
singlet and which transforms in the fundamental representation of SU (F )L and in
the anti-fundamental representation of SU (F )R , and which interacts with q and q̃
via a cubic superpotential
W = h qi Φij q̃ j . (10.116)
As bizarre this proposal may look like, let us try to understand it better. Let us first
consider the Seiberg dual theory (which from now on we dub mSQCD, where ’m’
stands for magnetic) without superpotential term, and let us focus on the SQCD
conformal window, 23 N < F < 3N , first. For W = 0 the field Φ is completely
decoupled and mSQCD is just SQCD with gauge group SU (F − N ) and F flavors.
Interestingly, as one can easily check, the SQCD conformal window is a conformal
window also for mSQCD! Hence mSQCD (without the singlet Φ) flows to an IR fixed
point for 32 N < F < 3N . At such fixed point the superpotential coupling, that we
now switch-on, is relevant, since ∆(W ) = ∆(Φ)+∆(q)+∆(q̃) = 1+ 32 N/F + 32 N/F <
3. The claim is that the perturbation (10.116) drives the theory to some new fixed
point (where the β-functions of the dual gauge coupling and of the coupling h both
vanish) which is actually the same fixed point of SQCD.
How does mSQCD look like for F ≤ 23 N ? The one-loop β-function coefficient of
mSQCD is b1 = 2F − 3N . Hence, for F = 23 N the β-function vanishes and for lower
values of F it changes its sign and mSQCD becomes IR-free. Hence, the bound
F = 23 N has the same role that the bound F = 3N has for SQCD (not surprisingly,
one can apply the BZ-fixed point argument to mSQCD for F slightly larger than 32 N
and find the existence of a perturbative fixed point). This explains why, if Seiberg
duality is correct, the IR dynamics of SQCD in the range N + 1 < F ≤ 32 N differs
from the behavior in the conformal window - something we had some indications
of, when studying the lower bound in F of the SQCD conformal window. Indeed,
we can now make our former intuition precise: using a clever set of variables (i.e.
the magnetic dual variables), one concludes that for N + 1 < F ≤ 23 N SQCD IR
dynamics is described by a theory of freely interacting (combinations of) meson and
baryon fields. These can be described in terms of free dual quarks interacting with

251
a Coulomb-like potential
gm2 1
Vm ∼ with gm2 ∼ , (10.117)
r log(rΛm )
where gm is the mSQCD gauge coupling and Λm the mSQCD strong coupling scale
(which in this regime of parameters is a UV cut-off, since the theory is IR-free).
This phase of SQCD is dubbed free magnetic phase, a theory of freely interacting
(dual) quarks. The fact that the IR dynamics of SQCD for N + 1 < F ≤ 23 N , where
the theory is confining, can be described this way is a rather powerful statement:
since mSQCD is IR-free, in terms of magnetic dual variables the Kähler potential
is canonical (up to subleading 1/Λ2m corrections), meaning that we know the full
effective IR Lagrangian of SQCD for N + 1 < F ≤ 23 N , at low enough energies!
As for the conformal window, which variables to use depends on F . The larger
F , the nearer to IR-freedom SQCD is, and the more UV-free mSQCD is. In other
words, the conformal window IR-fixed point is at smaller and smaller value of the
electric gauge coupling the nearer F is to 3N , and eventually becomes 0 for F = 3N .
For mSQCD things are reversed. The IR-fixed point arises at weaker coupling the
nearer F is to 32 N , and for F = 32 N we have that g∗m = 0. Therefore, the magnetic
description is the simplest to describe SQCD non-abelian Coulomb phase for F near
to 23 N ; the electric description is instead the most appropriate one when F is near
to 3N .
For F ≥ 3N the magnetic theory does not reach anymore an IR interacting fixed
point. The value F = 3N plays for mSQCD the same role the value F = 32 N plays
for SQCD. Indeed, the mSQCD meson matrix, U = q q̃ has ∆ = 1 for F = 3N , and
becomes a free field, while for larger values of F it would get a dimension lower than
one, which is not acceptable. For F ≥ 3N the theory should enter in a new phase.
This is something we know already: in this region we are in the SQCD IR-free phase.
Can we provide some consistency checks for the validity of this proposed duality?
Let us first note that two basic necessary requirements for its validity are met:
the two theories have the same global symmetry group as well as the same number
of IR degrees of freedom. In order to see this, let us first make the duality map
precise. The mapping between chiral operators of SQCD and mSQCD (at the IR
fixed point) is
1 i
M ←→ Φ : Φij = M (10.118)
µ j
B ←→ b : bj1 j2 ...jN = c i1 i2 ...iF −N j1 j2 ...jN Bi1 i2 ...iF −N (10.119)

252
 e The scale µ relating SQCD mesons with the mSQCD
and similarly for b̃ and B.
gauge singlet Φ appears for the following reason. In SQCD mesons are composite
fields and their dimension in the UV, where SQCD is free, is ∆ = 2. On the other
hand, Φ is an elementary field in mSQCD and its dimension in the UV is ∆ = 1.
Hence the scale µ needs to be introduced to match Φ to M in the UV. Clearly, upon
RG-flow both fields acquire an anomalous dimension and should flow to one and
the same operator in the IR, if the duality is correct. Applying formula (10.110),
which for chiral operators is an equality, one easily sees that this is indeed what
happens, since R(M ) = R(Φ). Finally, the scale c appearing in eq. (10.119) is there
for similar reasons as µ and has mass dimension F − 2N . Its precise value, which is
a function of µ in fact, will be fixed later.
From its very definition, it follows that the magnetic theory has a global sym-
metry group which is nothing but the one of SQCD, GF = SU (F )L × SU (F )R ×
U (1)B × U (1)R and, using the map (10.118), one can read-off the following charges
for the elementary fields
SU (F )L SU (F )R U (1)B U (1)R
N N
qia F • F −N F
q̃bj • F N
− F −N N
F
F −N
Φ F F 0 2 F
λ̃ • • 0 1
while the superpotential (10.116) has R = 2.
We can now use global symmetries to see that SQCD and mSQCD have the same
number of IR degrees of freedom. Basically, there is a one-to-one map between
gauge invariant operators, and these operators have the same global symmetries
(which counts physically distinct degrees of freedom). Indeed, the meson matrix M
enjoys the same symmetries as the mSQCD singlet Φ, and the SQCD baryons B, B e
the same as the mSQCD baryons b, b̃ (the latter being gauge invariant operators
constructed in terms of F − N dual quarks q, q̃). One might feel uncomfortable
since the mesons of the magnetic dual theory, Uij = qi q̃ j , seem not to match with
anything in the electric theory. This is where the superpotential (10.116) comes
into play. Recall the supposed equivalence between SQCD and mSQCD is just a
IR equivalence. The F-equations for Φ fix the dual meson to vanish on the moduli
space: FΦ = q q̃ = U = 0. Hence, in the IR the two theories do have the same
number of degrees of freedom!

Given what is above, let us now present several non-trivial checks for the validity

253
of Seiberg’s proposal.

• A very non-trivial check comes from ’t Hooft anomaly matching. The compu-
tation of ’t Hooft anomalies gives

SQCD
1
SU (F )2L U (1)B 2
N (+1) = 12 N
U (1)2B U (1)R 2N F (+1)(− N ) = −2N 2

3
F
4
U (1)3R −N
F
2N F + N 2 − 1 = −2 N F2
+ N2 − 1

mSQCD
1 N
SU (F )2L U (1)B − N ) F −N
(F = 12 N
2
N

2 N −F

U (1)2B U (1)R 2(F − N )F F −N = −2N 2
N −F

3
3
F
4
U (1)3R F
2(F − N )F + F 2 F −2N
F
+ (F − N )2 − 1 = −2 N
F2
+ N2 − 1

which shows there is indeed matching between SQCD and its IR-equivalent
mSQCD description.
Note that for the matching to work it turns out that the presence of dual
gauginos is crucial (as well as that of the magnetic superpotential term). This
explicitly shows that the description of SQCD baryons in terms of some sort
of dual quarks is not just a mere group representation theory accident. There
is a truly dynamical dual gauge group, under which dual quarks are charged,
and dual vector superfields (which include dual gauginos) which interact with
them.

• The duality relation is a duality, which means that acting twice with the
duality map one recovers the original theory (as far as IR physics!). Let us
start from SQCD with N colors and F flavors and act with the duality map
twice

SQCD : SU (N ) , F , W = 0

↓ duality
1
mSQCD : SU (F − N ) , F , W = qi M ij q̃ j = qi Φij q̃ j
µ
↓ duality
1 1
mmSQCD : SU (N ) , F , W = qi M ij q̃ j + di Ui j d˜j = qi Φij q̃ j + di Ψij d˜j
µ µ̃

254
 where Ui j = qi q̃ j is the meson matrix of mSQCD, Ψij is the gauge singlet chiral
superfield dual to U and belonging to the magnetic dual of mSQCD and, for
the ease of notation, we have put h = 1 in eq. (10.116). Choosing µ̃ = −µ, we
can rewrite the superpotential of mmSQCD as
1 h i
W = Tr U M − dU d˜ . (10.120)
µ

The fields U and M are hence massive and can be integrated out (recall we
claim the IR equivalence of Seiberg-dual theories, not the equivalence at all
scales). This implies

∂W ∂W
= 0 → M ij = di d˜j , =0→U =0 (10.121)
∂U ∂M
showing that the dual of the dual quarks are nothing but the original quark
e and that U = 0 (hence W = 0) in the IR . Summarizing,
superfields Q and Q,
after integrating out heavy fields, we are left with SQCD with gauge group
SU (N ), F flavors and no superpotential, exactly the theory we have started
with! In passing, let us note that in order to make the duality working we
have to set µ̃ = −µ, a mass scale which is not fixed by the duality itself.

• The duality is preserved under mass perturbations, namely upon holomorphic

decoupling. Let us again consider SQCD with gauge group SU (N ) and F
flavors and let us add a mass term to the F -th flavor, W = mM FF . This
corresponds to SU (N ) SQCD with F − 1 massless flavors and one massive
one. In the dual magnetic theory this gives a superpotential
1
W = qi M ij q̃ j + m M FF . (10.122)
µ
The F-flatness conditions for M FF and qF and q̃ F are

qFa q̃aF + µ m = 0 , (M · q̃a )F = (q a · M )F = 0 , (10.123)

where a is a SU (F − N ) gauge index. The first equation induces a VEV for

the dual quarks with flavor index F , which breaks the gauge group down to
SU (F − N − 1). The other two equations imply that the F -th row and column
of the SQCD meson matrix M vanish. We hence end-up with SU (F − N − 1)
SQCD with F − 1 flavors, a gauge singlet M which is a (F − 1) × (F − 1)
matrix, while the superpotential (10.122) reduces to eq. (10.116) where now

255
 i, j run from 1 to F − 1 only. This is the correct Seiberg dual mSQCD theory
at low energy.
This analysis shows that a mass term in the electric theory corresponds to
higgsing in the magnetic dual theory, according to the table below.

SQCD mSQCD
SU (N ), F ←→ SU (F − N ), F
DU AL
↓ mass ↓ higgsing
SU (N ), F − 1 ←→ SU (F − N − 1), F − 1
DU AL

The converse is also true (though slightly harder to prove): a mass term in
mSQCD corresponds to higgsing in SQCD.

• Let us use holomorphic decoupling to go from the last value of F where we

have the duality, F = N + 2, to F = N + 1. If Seiberg duality is correct, we
should recover the description of SQCD with F = N + 1 flavors we discussed
previously. Let us consider mSQCD when F = N + 2. The magnetic gauge
group is SU (2). Upon holomorphic decoupling, an analysis identical to the
one we did above produces a cubic superpotential at low energy as

W ∼ qi Mji q̃ j i, j = 1, . . . , N + 1 (10.124)

where qi are the baryons B i of SQCD with F = N + 1 and q̃j the baryons
Bej . At the same time, the VEVs for qN+2 and q̃ N+2 break the SU (2) gauge
symmetry completely. From mSQCD view point this is a situation similar to
SQCD with F = N − 1 where the full breaking of gauge symmetry group
allowed an exact instanton calculation providing the ∼ detM contribution to
the effective superpotential. The same happens here and the final answer one
gets for the low energy effective superpotential is


Weff ∼ qi Mji q̃ j − detM , (10.125)

which is precisely the effective superpotential of SQCD with F = N + 1!

This also shows that by holomorphic decoupling we can actually connect the
description of the IR dynamics of SQCD for any number of flavors, from F = 0
to any larger values of F , at fixed N .
Let us finally notice, in passing, that even for F = N + 1 we can sort of speak
of a magnetic dual theory. Just it is trivial, since there is no magnetic dual
gauge group.

256
• There is yet an important relation between the three a priori different mass
scales entering the duality: the electric dynamical scale Λel , the magnetic scale
Λm , and the matching scale µ. This reads

−F 3(F −N )−F
Λ3N
el Λm = (−1)F −N µF . (10.126)

That this relation is there, can be seen in different ways. First, one can check
that the relation is duality invariant, as it should. Indeed, applying the duality
map (recall that µ̃ = −µ, while Λel and Λm get interchanged by the duality)
one gets

−N )−F 3N −F
Λ3(F
m Λel = (−1)N µ̃F = (−1)N −F µF , (10.127)

which is identical to (10.126).

One can also verify the consistency of the relation (10.126) upon higgsing
and/or holomorphic decoupling. Finally, by matching F = N + 2 SCQD to
F = N + 1 SQCD via holomorphic decoupling, one can also fix the value of c
in eq. (10.119) to be
q p
−F
c = 1/ −(−µ)N −F Λ3N
el = 1/ −µN Λ3N m
−2F . (10.128)

Eq. (10.126) shows that as the electric theory becomes stronger (i.e. Λel in-
creases), the magnetic theory becomes weaker (i.e. Λm decreases). By using
the relation between dynamical scales and gauge couplings, this can be trans-
lated into a relation between gauge coupling constants, and gives an inverse
relation between them
gel2 ∼ gm−2 , (10.129)
showing that large values of the electric gauge coupling gel correspond to small
values of the magnetic one, and viceversa. This is why Seiberg duality is an
electric-magnetic duality.
Depending on where in the (F, N ) space one sits, the meaning of the dynamical
scales changes. In the conformal window both SQCD and mSQCD are UV-free.
Both theories have a non-trivial RG-flow and, upon non-perturbative effects,
driven by Λel and Λm , reach an IR fixed point (which is one and the same, in
fact). In the free-magnetic phase, mSQCD is IR-free and SQCD is UV-free.
Therefore, in this regime Λm should be better thought of as a UV-scale for
the magnetic theory, which is an effective theory. In this regime SQCD can

257
 be thought of as the (or better, a possible) UV-completion of mSQCD (the
electric free phase can be thought of in a similar way, with the role of SQCD
and mSQCD reversed). Within this interpretation it is natural to tune the
free parameter µ to make the two theories have one single non-perturbative
scale, the scale at which non-perturbative SQCD effects come into play and
the scale below which the magnetic effective description takes over. From the
relation (10.126) one sees that this is obtained by equating, up to an overall
phase, the matching scale µ with Λel and Λm
µ = Λel = Λm (≡ Λ) . (10.130)
Using the above relation for F = N + 2 and adding a mass term for the F -th
flavor, upon holomorphic decoupling one gets the expression (10.125) including
the correct power of Λ, that is
1  i ej

Weff = 2N −1 Bi Mj B − detM . (10.131)
Λ
where we have already used the fact that the dual quarks are nothing but the
baryon themselves, in this case.

Figure 10.13 contains a qualitative description of the three different regimes we

have just discussed.

SQCD SQCD mSQCD mSQCD

UV UV UV UV

⇤ ⇤
<latexit sha1_base64="yFrnXXViWirehqLFkWd9XsZYx1w=">AAAB7nicbVBNS8NAFHypX7V+VT16WSyCp5KIoseiFw8eKlhbaEPZbDbt0s0m7L4IJfRHePGgiFd/jzf/jds2B20dWBhm5rHvTZBKYdB1v53Syura+kZ5s7K1vbO7V90/eDRJphlvsUQmuhNQw6VQvIUCJe+kmtM4kLwdjG6mfvuJayMS9YDjlPsxHSgRCUbRSu3enY2GtF+tuXV3BrJMvILUoECzX/3qhQnLYq6QSWpM13NT9HOqUTDJJ5VeZnhK2YgOeNdSRWNu/Hy27oScWCUkUaLtU0hm6u+JnMbGjOPAJmOKQ7PoTcX/vG6G0ZWfC5VmyBWbfxRlkmBCpreTUGjOUI4toUwLuythQ6opQ9tQxZbgLZ68TB7P6t5F3b0/rzWuizrKcATHcAoeXEIDbqEJLWAwgmd4hTcndV6cd+djHi05xcwh/IHz+QMM949h</latexit> <latexit sha1_base64="yFrnXXViWirehqLFkWd9XsZYx1w=">AAAB7nicbVBNS8NAFHypX7V+VT16WSyCp5KIoseiFw8eKlhbaEPZbDbt0s0m7L4IJfRHePGgiFd/jzf/jds2B20dWBhm5rHvTZBKYdB1v53Syura+kZ5s7K1vbO7V90/eDRJphlvsUQmuhNQw6VQvIUCJe+kmtM4kLwdjG6mfvuJayMS9YDjlPsxHSgRCUbRSu3enY2GtF+tuXV3BrJMvILUoECzX/3qhQnLYq6QSWpM13NT9HOqUTDJJ5VeZnhK2YgOeNdSRWNu/Hy27oScWCUkUaLtU0hm6u+JnMbGjOPAJmOKQ7PoTcX/vG6G0ZWfC5VmyBWbfxRlkmBCpreTUGjOUI4toUwLuythQ6opQ9tQxZbgLZ68TB7P6t5F3b0/rzWuizrKcATHcAoeXEIDbqEJLWAwgmd4hTcndV6cd+djHi05xcwh/IHz+QMM949h</latexit>

⇤m
<latexit sha1_base64="e23eEMUlZ3i9nVMYj7WZhIOAigM=">AAACAXicbVDLSsNAFJ34rPUVdSO4GSyCq5KIosuiGxcuKtgHNCFMJpN26MwkzEzEEuLGX3HjQhG3/oU7/8Zpm4W2Hhg4nHMud+4JU0aVdpxva2FxaXlltbJWXd/Y3Nq2d3bbKskkJi2csER2Q6QIo4K0NNWMdFNJEA8Z6YTDq7HfuSdS0UTc6VFKfI76gsYUI22kwN73bkw4QkHu8TB5yD1NxQjyogjsmlN3JoDzxC1JDZRoBvaXFyU440RozJBSPddJtZ8jqSlmpKh6mSIpwkPUJz1DBeJE+fnkggIeGSWCcSLNExpO1N8TOeJKjXhokhzpgZr1xuJ/Xi/T8YWfU5Fmmgg8XRRnDOoEjuuAEZUEazYyBGFJzV8hHiCJsDalVU0J7uzJ86R9UnfP6s7taa1xWdZRAQfgEBwDF5yDBrgGTdACGDyCZ/AK3qwn68V6tz6m0QWrnNkDf2B9/gARk5dK</latexit>

⇤el
<latexit sha1_base64="P1QyY8L6+/xMUjZFQp5Ll1L5GKM=">AAACAnicbVDLSsNAFJ34rPUVdSVuBovgqiSi6LLoxoWLCvYBTQiTyaQdOpkJMxOxhODGX3HjQhG3foU7/8Zpm4W2Hhg4nHMud+4JU0aVdpxva2FxaXlltbJWXd/Y3Nq2d3bbSmQSkxYWTMhuiBRhlJOWppqRbioJSkJGOuHwaux37olUVPA7PUqJn6A+pzHFSBspsPe9GxOOUJB7SSgeck9TPoKEFUVg15y6MwGcJ25JaqBEM7C/vEjgLCFcY4aU6rlOqv0cSU0xI0XVyxRJER6iPukZylFClJ9PTijgkVEiGAtpHtdwov6eyFGi1CgJTTJBeqBmvbH4n9fLdHzh55SnmSYcTxfFGYNawHEfMKKSYM1GhiAsqfkrxAMkEdamtaopwZ09eZ60T+ruWd25Pa01Lss6KuAAHIJj4IJz0ADXoAlaAINH8AxewZv1ZL1Y79bHNLpglTN74A+szx/XxZe4</latexit>

g⇤el = 0
<latexit sha1_base64="aTHrm4qrfaxOEzQZCb1g4npfzOI=">AAACAnicbVDLSsNAFJ34rPUVdSVuBosgLkoiim6EohuXFewDmhgm00k7dGYSZiZiCcGNv+LGhSJu/Qp3/o3TNgttPXDhcM693HtPmDCqtON8W3PzC4tLy6WV8ura+samvbXdVHEqMWngmMWyHSJFGBWkoalmpJ1IgnjISCscXI381j2RisbiVg8T4nPUEzSiGGkjBfZu7y7zeBg/ZJ6mYggJy/PgCF5AJ7ArTtUZA84StyAVUKAe2F9eN8YpJ0JjhpTquE6i/QxJTTEjedlLFUkQHqAe6RgqECfKz8Yv5PDAKF0YxdKU0HCs/p7IEFdqyEPTyZHuq2lvJP7ndVIdnfsZFUmqicCTRVHKoI7hKA/YpZJgzYaGICypuRXiPpIIa5Na2YTgTr88S5rHVfe06tycVGqXRRwlsAf2wSFwwRmogWtQBw2AwSN4Bq/gzXqyXqx362PSOmcVMzvgD6zPH1VLlrc=</latexit>

g⇤m = 0
<latexit sha1_base64="Z/wVtH2KxTTZJRowKDYjx0Mowr4=">AAACAXicbVDLSsNAFJ34rPUVdSO4GSyCuCiJKLoRim5cVrAPaGKYTCft0JlJmJmIJcSNv+LGhSJu/Qt3/o3TNgttPXDhcM693HtPmDCqtON8W3PzC4tLy6WV8ura+samvbXdVHEqMWngmMWyHSJFGBWkoalmpJ1IgnjISCscXI381j2RisbiVg8T4nPUEzSiGGkjBfZu7y7zeBg/ZJ6mYgh5ngdH8AI6gV1xqs4YcJa4BamAAvXA/vK6MU45ERozpFTHdRLtZ0hqihnJy16qSILwAPVIx1CBOFF+Nv4ghwdG6cIolqaEhmP190SGuFJDHppOjnRfTXsj8T+vk+ro3M+oSFJNBJ4silIGdQxHccAulQRrNjQEYUnNrRD3kURYm9DKJgR3+uVZ0jyuuqdV5+akUrss4iiBPbAPDoELzkANXIM6aAAMHsEzeAVv1pP1Yr1bH5PWOauY2QF/YH3+AI7plkk=</latexit>

<latexit sha1_base64="pV3XG67YB1Bs2sPUOeEPBvglGpQ=">AAACH3icbZDLSgMxFIYz9VbrrerSTbAIIlJmxNtGKLpxWcHWQqeWTJq2obkMyRmxDH0TN76KGxeKiLu+jellodYfAj/fOYeT80ex4BZ8f+hl5uYXFpeyy7mV1bX1jfzmVtXqxFBWoVpoU4uIZYIrVgEOgtViw4iMBLuLelej+t0DM5ZrdQv9mDUk6Sje5pSAQ838aad5cJ+GMtKPaQhc9TETgwE+xDNcOhwqDfjCx7iZL/hFfyw8a4KpKaCpys38V9jSNJFMARXE2nrgx9BIiQFOBRvkwsSymNAe6bC6s4pIZhvp+L4B3nOkhdvauKcAj+nPiZRIa/sycp2SQNf+rY3gf7V6Au3zRspVnABTdLKonQgMGo/Cwi1uGAXRd4ZQw91fMe0SQyi4SHMuhODvybOmelQMTor+zXGhdDmNI4t20C7aRwE6QyV0jcqogih6Qi/oDb17z96r9+F9Tloz3nRmG/2SN/wG24eiLw==</latexit>

g⇤el , g⇤m 6= 0
IR mSQCD IR IR SQCD
3 3
<latexit sha1_base64="2Lawo4F3Cye8MZNRCTiuBRi4Jp0=">AAACBHicbVDLSgNBEOyNrxhfqx5zGQyCIITdqOjBQ1AQTyGCSYTsEmYns8mQ2Yczs0JY9uDFX/HiQRGvfoQ3/8bJ46CJBQ1FVTfdXV7MmVSW9W3kFhaXllfyq4W19Y3NLXN7pymjRBDaIBGPxJ2HJeUspA3FFKd3saA48DhteYPLkd96oEKyKLxVw5i6Ae6FzGcEKy11zGINHdroHF0hh9N75PgCk/QoSysZqnXMklW2xkDzxJ6SEkxR75hfTjciSUBDRTiWsm1bsXJTLBQjnGYFJ5E0xmSAe7StaYgDKt10/ESG9rXSRX4kdIUKjdXfEykOpBwGnu4MsOrLWW8k/ue1E+WfuSkL40TRkEwW+QlHKkKjRFCXCUoUH2qCiWD6VkT6WOegdG4FHYI9+/I8aVbK9knZujkuVS+mceShCHtwADacQhWuoQ4NIPAIz/AKb8aT8WK8Gx+T1pwxndmFPzA+fwAjipXi</latexit> <latexit sha1_base64="PnMK3IfvzJcmT3kptw6oCsBznCg=">AAAB/3icbVDLSgNBEOyNrxhfq4IXL4NB8BR2E0UPHoKCeAoRzAOSJcxOZpMhsw9mZoWw7sFf8eJBEa/+hjf/xkmyB00saCiquunuciPOpLKsbyO3tLyyupZfL2xsbm3vmLt7TRnGgtAGCXko2i6WlLOANhRTnLYjQbHvctpyR9cTv/VAhWRhcK/GEXV8PAiYxwhWWuqZB11PYJJU0qScohq6RDe6KrWeWbRK1hRokdgZKUKGes/86vZDEvs0UIRjKTu2FSknwUIxwmla6MaSRpiM8IB2NA2wT6WTTO9P0bFW+sgLha5Aoan6eyLBvpRj39WdPlZDOe9NxP+8Tqy8CydhQRQrGpDZIi/mSIVoEgbqM0GJ4mNNMBFM34rIEOtAlI6soEOw519eJM1yyT4rWXenxepVFkceDuEITsCGc6jCLdShAQQe4Rle4c14Ml6Md+Nj1pozspl9+APj8wfTe5QF</latexit>

N +1<F  N N < F < 3N F 3N

<latexit sha1_base64="jlgzOMtRqaw0XxV7tVot6phkCic=">AAAB8HicbVDJSgNBEK2JW4xb1KOXxiB4CjMu6DEoiCeJYBZJhtDTqUmadM+M3T1CGPIVXjwo4tXP8ebf2FkOGn1Q8Hiviqp6QSK4Nq775eQWFpeWV/KrhbX1jc2t4vZOXcepYlhjsYhVM6AaBY+wZrgR2EwUUhkIbASDy7HfeESleRzdmWGCvqS9iIecUWOl+yvS7uEDOb7pFEtu2Z2A/CXejJRghmqn+NnuxiyVGBkmqNYtz02Mn1FlOBM4KrRTjQllA9rDlqURlaj9bHLwiBxYpUvCWNmKDJmoPycyKrUeysB2Smr6et4bi/95rdSE537GoyQ1GLHpojAVxMRk/D3pcoXMiKEllClubyWsTxVlxmZUsCF48y//JfWjsndadm9PSpWLWRx52IN9OAQPzqAC11CFGjCQ8AQv8Ooo59l5c96nrTlnNrMLv+B8fANeY494</latexit>

2 2

Figure 10.13: The three qualitative different phases of SQCD with F > N +1. In the
magnetic-free and electric-free phases we have chosen (for convenience) the arbitrary
matching scale µ in such a way to make Λel and Λm being identified, eq. (10.130).

10.5 The phase diagram of N=1 SQCD

After this long tour on quantum properties of SQCD, it is time to wrap-up and
summarize its phase diagram.

258
 For F = 0 SQCD (pure SYM in this case) enjoys strict confinement, displays N
isolated supersymmetric vacua and a mass gap. For 0 < F < N the theory doesn’t
exist by its own. The classical moduli space is completely lifted and a runaway
potential, with no absolute minima at finite distance in field space, is generated.
For F = N, N + 1 a moduli space persists at the quantum level and SQCD
enjoys confinement with charge screening (the asymptotic states are gauge singlets
but flux lines can break) and no mass gap. Asymptotic states are mesons and
baryons. The theory displays complementarity, as any theory where there are scalars
transforming in the fundamental representation of the gauge group: there is no
invariant distinction between Higgs phase, which is the more appropriate description
for large field VEVs, and confinement phase, which takes over near the origin of field
space. The potential between static test charges goes to a constant asymptotically
since in the Higgs phase gauge bosons are massive and there are no long-range forces.
As already observed, this holds also in the confining description, since we actually
have charge screening and Wilson loops do not follow the area law in this case.
For N + 2 ≤ F ≤ 23 N we are still in a confinement phase, but the theory is in the
so-called free magnetic phase and can be described at large enough distance in terms
of freely interacting dual quarks and gluons. What is amusing here is that while
asymptotic massless states are composite of elementary electric degrees of freedom
(i.e. mesons and baryons), they are charged with respect to a magnetic gauge group
whose dynamics is not visible in the electric description and which is generated,
non-perturbatively, by the theory itself.
For F > 32 N SQCD does not confine anymore, not even in the weak sense:
asymptotic states are quarks and gluons (and their superpartners). The potential
between asymptotic states, though, differs if F ≥ 3N or F < 3N . In the former case
the theory is IR-free and it is described by freely interacting particles. Hence the
potential vanishes, at large enough distance. For 23 N < F < 3N , instead, the theory
(which is still UV-free) is in a non-abelian Coulomb phase. Charged particles are
not confined but actually belong to a SCFT, and interact by a 1/r potential with
coupling g = g∗ .
A diagram summarizing the gross features of the quantum dynamics of SQCD
is reported below.

259
 UV-free IR-free

0 N 3/2 N 3N

runaway free magnetic

<latexit sha1_base64="iOXbfpVC/54viPchKfXfekDWMM8=">AAAB/nicbVBNSwMxEM3Wr1q/VsWTl2ARPJVdUfRY9OKxgq2FdinZdLYNTTZLkhXLUvCvePGgiFd/hzf/jdl2D9r6YODx3kwy88KEM20879spLS2vrK6V1ysbm1vbO+7uXkvLVFFoUsmlaodEA2cxNA0zHNqJAiJCDvfh6Dr37x9AaSbjOzNOIBBkELOIUWKs1HMPuiKUj1mkAHBugWF0gntu1at5U+BF4hekigo0eu5Xty9pKiA2lBOtO76XmCAjyj7HYVLpphoSQkdkAB1LYyJAB9l0/Qk+tkofR1LZig2eqr8nMiK0HovQdgpihnrey8X/vE5qossgY3GSGojp7KMo5dhInGeB+0wBNXxsCaGK2V0xHRJFqLGJVWwI/vzJi6R1WvPPa97tWbV+VcRRRofoCJ0gH12gOrpBDdREFGXoGb2iN+fJeXHenY9Za8kpZvbRHzifP1GflbU=</latexit>

interacting Coulomb
<latexit sha1_base64="SYUoT87QYUMVx7sX+L9zGAgUkMg=">AAACA3icbVDLSgMxFM3UV62vUXe6CZaCqzIjii6L3bisYB/QDiWTZtrQPIYkI5ah4MZfceNCEbf+hDv/xkw7C209EDick8u954Qxo9p43rdTWFldW98obpa2tnd299z9g5aWicKkiSWTqhMiTRgVpGmoYaQTK4J4yEg7HNczv31PlKZS3JlJTAKOhoJGFCNjpb571OOhfEipMEQhbKgYwrpMmOThtO+Wvao3A1wmfk7KIEej7371BhInnAiDGdK663uxCVKkDMWMTEu9RJMY4TEakq6lAnGig3SWYQorVhnASCr7hIEz9fdEirjWE3sWrHBkRnrRy8T/vG5ioqvABowTQwSeL4oSBo2EWSFwQBXBhk0sQVhReyvEI5SVYWsr2RL8xcjLpHVW9S+q3u15uXad11EEx+AEnAIfXIIauAEN0AQYPIJn8ArenCfnxXl3PuZfC04+cwj+wPn8AdSmmEY=</latexit>

F
free electric
<latexit sha1_base64="QEPj1AenVklrUEeDihjTJh3p7oE=">AAAB/nicbVDLSgNBEJyNrxhfUfHkZTAInsKuKHoMevEYwcRAsoTZSW8yZB7LzKwYloC/4sWDIl79Dm/+jZNkD5pY0FBUddPdFSWcGev7315haXllda24XtrY3NreKe/uNY1KNYUGVVzpVkQMcCahYZnl0Eo0EBFxuI+G1xP//gG0YUre2VECoSB9yWJGiXVSt3zQEZF6zGINgIEDtZrRMe6WK37VnwIvkiAnFZSj3i1/dXqKpgKkpZwY0w78xIYZ0ZZRDuNSJzWQEDokfWg7KokAE2bT88f42Ck9HCvtSlo8VX9PZEQYMxKR6xTEDsy8NxH/89qpjS/DjMkktSDpbFGccmwVnmSBe0y7j/nIEUI1c7diOiCaUOsSK7kQgvmXF0nztBqcV/3bs0rtKo+jiA7RETpBAbpANXSD6qiBKMrQM3pFb96T9+K9ex+z1oKXz+yjP/A+fwBWLpW4</latexit>

<latexit sha1_base64="5UZrc1ubMN5y66gXsZc5EaD5R9I=">AAAB+HicbVDJSgNBEO2JW4xLRj16aQyCpzAjih6DXjxGMAskQ+jp9CRNehl6UeOQL/HiQRGvfoo3/8ZOMgdNfFDweK+Kqnpxyqg2QfDtFVZW19Y3ipulre2d3bK/t9/U0ipMGlgyqdox0oRRQRqGGkbaqSKIx4y04tH11G/dE6WpFHdmnJKIo4GgCcXIOKnnl7s8lo+ZsgI9oPEE9vxKUA1mgMskzEkF5Kj3/K9uX2LLiTCYIa07YZCaKEPKUMzIpNS1mqQIj9CAdBwViBMdZbPDJ/DYKX2YSOVKGDhTf09kiGs95rHr5MgM9aI3Ff/zOtYkl1FGRWoNEXi+KLEMGgmnKcA+VQQbNnYEYUXdrRAPkULYuKxKLoRw8eVl0jythufV4PasUrvK4yiCQ3AETkAILkAN3IA6aAAMLHgGr+DNe/JevHfvY95a8PKZA/AH3ucPPiGTdg==</latexit>

phase: g m ! 0 phase: g el,m ! g⇤el,m phase: g el ! 0

<latexit sha1_base64="PtJeOEg/kHgGo7PdWkHj0YEyp/Y=">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</latexit> <latexit sha1_base64="TPAKHQz7F/DSlNsF8Dyh8dzDG1M=">AAACGnicbVDLSgMxFM34rPU16tJNsAgupMyIorgqunFZwT6gU0smTdvQTDIkd9QyzHe48VfcuFDEnbjxb0wfC209cOFwzr3JvSeMBTfged/O3PzC4tJybiW/ura+selubVeNSjRlFaqE0vWQGCa4ZBXgIFg91oxEoWC1sH859Gt3TBuu5A0MYtaMSFfyDqcErNRy/SAK1UMa9+wb51lwiLu36UgKgMtBykSW4UDzbg+I1uoee7jlFryiNwKeJf6EFNAE5Zb7GbQVTSImgQpiTMP3YmimRAOngmX5IDEsJrRPuqxhqSQRM810dFqG963Sxh2lbUnAI/X3REoiYwZRaDsjAj0z7Q3F/7xGAp2zZsplnACTdPxRJxEYFB7mhNtcMwpiYAmhmttdMe0RTSjYNPM2BH/65FlSPSr6J0Xv+rhQupjEkUO7aA8dIB+dohK6QmVUQRQ9omf0it6cJ+fFeXc+xq1zzmRmB/2B8/UDHvGhgw==</latexit>

<latexit sha1_base64="9pjbBWjPwwFETOTg12OxHQgI9XQ=">AAAB/HicbVDLSsNAFJ3UV62vaJduBotQNyURRZdFNy4r2Ae0oUymk3boPMLMpBhC/RU3LhRx64e482+ctllo64ELh3Pu5d57wphRbTzv2ymsrW9sbhW3Szu7e/sH7uFRS8tEYdLEkknVCZEmjArSNNQw0okVQTxkpB2Ob2d+e0KUplI8mDQmAUdDQSOKkbFS3y33eCgfs6qQcIJwkvCzKey7Fa/mzQFXiZ+TCsjR6LtfvYHECSfCYIa07vpebIIMKUMxI9NSL9EkRniMhqRrqUCc6CCbHz+Fp1YZwEgqW8LAufp7IkNc65SHtpMjM9LL3kz8z+smJroOMirixBCBF4uihEEj4SwJOKCKYMNSSxBW1N4K8QgphI3Nq2RD8JdfXiWt85p/WfPuLyr1mzyOIjgGJ6AKfHAF6uAONEATYJCCZ/AK3pwn58V5dz4WrQUnnymDP3A+fwAvnZR2</latexit>

(no vacuum)
<latexit sha1_base64="qRckHVNmgGrQRJ/t7SaIXntbBa8=">AAACGXicbVC7SgNBFJ31bXytWtoMBsFCwq4oipVoY6lgVMiuYXYySYbMY5m5q4Zlf8PGX7GxUMRSK//GSdxCowcuHM65d+bek6SCWwiCT29sfGJyanpmtjI3v7C45C+vXFidGcrqVAttrhJimeCK1YGDYFepYUQmgl0mveOBf3nDjOVanUM/ZbEkHcXbnBJwUtMPIpnouzztujcOimgLd67zoRQBV/1cFgWODO90gRijb3GAm341qAVD4L8kLEkVlTht+u9RS9NMMgVUEGsbYZBCnBMDnApWVKLMspTQHumwhqOKSGbjfHhZgTec0sJtbVwpwEP150ROpLV9mbhOSaBrR72B+J/XyKC9H+dcpRkwRb8/amcCg8aDmHCLG0ZB9B0h1HC3K6ZdYggFF2bFhRCOnvyXXGzXwt1acLZTPTwq45hBa2gdbaIQ7aFDdIJOUR1RdI8e0TN68R68J+/Ve/tuHfPKmVX0C97HF07+oRU=</latexit>

confining phase non-confining phase

10.6 Exercises
1. Consider SQCD with F = 1 flavor and show that giving a mass m to the only
flavor, upon holomorphic decoupling one gets the pure SYM superpotential,
eq. (10.67), at low energy.

2. Consider SQCD with F = N with superpotential

 
e
W = A det M − B B − aΛ 2N eN
+ mQN Q (10.132)

By integrating out the massive flavor, show that one recovers the ADS super-
potential for F = N − 1 SQCD if and only if a = 1.

3. Check ’t Hooft anomaly matching for SQCD with F = N along the baryonic
branch, M = 0, B = −Be = ΛN .

4. Check ’t Hooft anomaly matching for SQCD with F = N + 1 at the origin of

the moduli space.

5. Consider mSQCD for F = 3(F − N ) − (F − N ) with  << 1, and find the BZ
perturbative fixed point, i.e. the values of the dual gauge coupling gm and of
the cubic superpotential coupling h, eq. (10.116), for which the corresponding
β-functions vanish.

6. Show that the addition of a mass term in mSQCD corresponds to higgsing in

SQCD (note: this is the inverse of what we have shown in the main text).

260
References

[1] M. J. Strassler, On confinement and duality, Chapter 1, ICTP Spring School

on Superstrings and Related Matters, ICTP Lect. Notes Ser. 7 (2002) 105.

[2] D. Tong, Lectures on Gauge Theory, Chapter 2,

https://www.damtp.cam.ac.uk/user/tong/gaugetheory/gt.pdf

[3] M. E. Peskin, Duality in supersymmetric Yang-Mills theory, Chapters 2, 3, 4

and 7, TASI 96, arXiv:hep-th/9702094.

[4] N. Dorey, T. J. Hollowood, V. V. Khoze and M. P. Mattis, The calculus of many

instantons, Chapter VII, Phys. Rept. 371 (2002) 231 [arXiv:hep-th/0206063].

[5] R. Argurio, G. Ferretti and R. Heise, An introduction to supersymmetric gauge

theories and matrix models, Chapter 3, Int. J. Mod. Phys. A 19, 2015 (2004)
[arXiv:hep-th/0311066].

[6] K. A. Intriligator, R. G. Leigh and N. Seiberg, Exact superpotentials in four-

dimensions, Phys. Rev. D 50 (1994) 1092 [hep-th/9403198].

[7] K. A. Intriligator, Integrating in’ and exact superpotentials in 4-d, Phys. Lett.
B 336 (1994) 409 [hep-th/9407106].

[8] K. A. Intriligator and N. Seiberg, Lectures on supersymmetric gauge theo-

ries and electric-magnetic duality, Nucl. Phys. Proc. Suppl. 45BC (1996) 1
[arXiv:hep-th/9509066].

[9] J. Terning, Modern supersymmetry: Dynamics and duality, Chapters 9 and 10,
Oxford University Press (2006).

[10] M. J. Strassler, The Duality cascade, Chapter 1, TASI 2003 [hep-th/0505153].

[11] M. A. Shifman, Nonperturbative dynamics in supersymmetric gauge theories,

Chapters 1, 2 and 3, Prog. Part. Nucl. Phys. 39 (1997) 1 [hep-th/9704114].

261
11 Dynamical supersymmetry breaking

After this long detour on quantum properties of supersymmetric gauge theories,

we can now go back to supersymmetry breaking, and finally discuss models where
supersymmetry is broken by strong coupling effects, aka dynamical supersymmetry
breaking (DSB). As we already emphasized, these models can in principle be used as
consistent (and natural) hidden sectors within gravity or gauge mediation scenarios
(or any of their possible variants).
We will first focus on models where supersymmetry is broken dynamically in
stable vacua, that is at absolute minima of the potential. The rough general picture
for models of this kind (with exceptions, as we will see) is as follows:

• The supersymmetric theory at hand is a gauge theory. This is because gauge

degrees of freedom are the only ones having some chance of generating non-
perturbative contributions to the superpotential. As we already discussed, in
models of chiral superfields only, the superpotential is tree-level exact.

• The tree level superpotential Wtree does not break supersymmetry but lifts
all flat directions. Since the superpotential is polynomial in the fields, this
typically gives a potential V which vanishes at the origin and grows for large
field VEVs. Since the superpotential is classically exact in perturbation theory,
supersymmetry is preserved at all orders, perturbatively.

• Strong coupling effects generate a non-perturbative superpotential which pro-

vides a contribution to the potential which is strong at the origin of field space
but decreases for large field VEVs (recall that the large field VEVs region
corresponds to the classical region, where quantum corrections are negligible).
An instance of such a potential is the effective potential of F < N SQCD.

DSB arises because of the interplay between the classical contribution and the non-
perturbatively generated one, as shown in Figure 11.1. Generically, supersymmetry
will be broken and the exact potential will display a stable non-supersymmetric
minimum at finite distance in field space.

11.1 Calculable and non-calculable models: generalities

Once a stable non-supersymmetric minimum is found, one would like to study quan-
tum fluctuations around it. The spectrum around such vacuum is non supersymmet-

262
 V

full potential Veff

classical potential

4
Ms
non−pert.
potential

< φ* > <φ>

Figure 11.1: A schematic view of dynamical supersymmetry breaking conspiracy.

The classical and non-perturbative contributions to the effective potential sum-up
and give a stable supersymmetry-breaking minimum at hφi = hφ∗ i.

ric and hence quantum corrections will not be protected by supersymmetry. More-
over, besides the superpotential, the knowledge of the Kähler potential will also be
important if one wants to make any sort of quantitative statement. A knowledge
of the Kähler potential is needed to know the exact point in field space where the
supersymmetry breaking vacuum sits, the values of the vacuum energy, i.e. the su-
persymmetry breaking scale Ms ∼ (Vmin )1/4 , and the masses and interactions of light
fields; in other words, one should know the structure of the effective Lagrangian.
It is in general a difficult task to control the form of the Kähler potential, since K
is corrected at all orders in perturbation theory (and non-perturbatively). There is
then a problem of calculability around a non-perturbatively generated supersymme-
try breaking vacuum, in general.
Looking at Figure 11.1, calling λ the generic tree level coupling(s), it should be
clear that if we decrease λ the tree-level potential Vtree becomes less and less steep,
and the supersymmetry breaking minimum is pushed more and more towards the
large VEVs region, where the theory is weakly coupled, see Figure 11.2.
There are three basic reasons why making the tree-level superpotential couplings
smaller, calculability is increased.
The smaller λ the smaller Ms , too. Eventually, it might become even smaller
than Λ (e.g., curve 4a in Figure 11.2). This is useful since at energies lower than Λ
gauge degrees of freedom can be safely integrated out giving rise to simpler models
(of O’R-like type, so to say). Hence, the analysis of the low energy effective theory

263
 1a 2a
V 3a

4a

4
Ms

Λ <φ>

Figure 11.2: Decreasing the perturbative coupling λ the DSB minimum moves to-
wards large field VEVs. For the curve 4a the supersymmetry breaking scale Ms is
smaller than the dynamical scale Λ.

around supersymmetry breaking vacua might be simpler, the potential having only
F-terms contributions
h 00 i−1 ∂W 2
V (φ, φ† ) = K (φ, φ† ) . (11.1)
∂φ
Second, as we have discussed at length in the previous lecture, most progresses
in understanding supersymmetric theories at the non-perturbative level regard the
deep IR, E < Λ (structure of vacua, lowest lying state excitations around them,
etc...). Hence, having Ms < Λ is a welcome feature.
One more reason why having λ small increases calculability has to do with the
very possibility of computing the Kähler potential. While the effective superpoten-
tial Weff can often be determined exactly, the Kähler potential is in general more
difficult to calculate, since it is not protected by holomorphy. Having supersymme-
try breaking vacua at large VEVs has the advantage that the theory is more and
more classical (i.e. weakly coupled) there. Therefore, one can in principle determine
the Kähler potential of the low energy effective fields just by projecting the UV-fields
canonical Kähler potential on such operators (which are typically some gauge and
flavor invariant combinations of UV fields), getting a correct result up to corrections
which, in such semi-classical region, are weak.
According to this general picture, DSB models can be roughly divided into three
classes, with increasing level of calculability.

• The worst case scenario is a situation where one cannot get any information

264
 on the full potential due to incapacity of computing both the effective super-
potential and the Kähler potential. That supersymmetry is broken can be
concluded based on indirect arguments, as those we discussed in Lecture 7
(like R-symmetry and/or global symmetry arguments). In these cases one can
reasonably say that supersymmetry is broken and that Ms ∼ Λ, but nothing
can be said about the massless excitations around the supersymmetry breaking
vacua nor on the effective Lagrangian describing their dynamics.

• A better situation occurs when one can compute the effective superpotential
and explicitly see that the latter generates some non-vanishing F-terms which
were vanishing at tree-level. In these cases one can safely say that supersym-
metry is broken and possibly tell which are the low energy degrees of freedom
around the supersymmetry breaking vacua. Still, the Kähler potential cannot
be determined. Hence, one cannot calculate any property of the ground states
nor determine the dynamics around them. DSB models belonging to this class
are known as non-calculable models.

• Finally, there can exist models where the scenario summarized in Figure 11.2
can be fully realized. There exists a region in parameter space where the theory
is weakly coupled and one can also compute the Kähler potential, then. In
these situations one can get also quantitative information about the low energy
effective theory, like the precise value of supersymmetry breaking scale, the
structure of the light spectrum and interactions. Possibly at an arbitrary high
level of accuracy, if supersymmetry breaking vacua can be made parametrically
far from the origin of field space. Models of this kind are known as calculable
models.

In what follows, we will present some concrete examples for each of above three
classes.

11.2 The one GUT family SU(5) model

Let us consider a supersymmetric gauge theory with gauge group SU (5), a chiral
superfield T transforming in the 10 (i.e. the antisymmetric representation), and
e transforming in the anti-fundamental representation, 5.
another chiral superfield Q
This theory is UV-free, the one-loop β function coefficient being b1 = 13.

265
 This theory does not have any classical flat direction since it is impossible to
construct holomorphic gauge invariant operators out of T and Q. e For the same
reason a superpotential cannot be added. Therefore, at the classical level there
exist one supersymmetric vacuum, sitting at the origin of field space, where the
gauge group is unbroken. Given that the theory is UV-free and so expected to go to
strong coupling in the IR and the not so large matter content, one can reasonably
argue that the theory confines and so that there are no leftover gauge degrees of
freedom in the IR.
At the origin the theory is strongly coupled and it is difficult to perform any
reliable computation. However, one can use indirect arguments to conclude that
non-perturbative corrections break supersymmetry. First, one can easily check that
there exist two non-anomalous global symmetries, GF = U (1) × U (1)R , under which
the fields have charges T ' (−1, 1) and Q e ' (3, −9), where the charges are fixed by
anomaly cancellations. We now use ’t Hooft anomaly matching to argue that GF is
spontaneously broken. We do not know what the low energy SU (5) invariant degrees
of freedom are, but if GF is unbroken, they should reproduce ’t Hooft anomalies for
U (1)3 , U (1)2 U (1)R , etc... of the original theory. One can be as general as possible
and allow for a set of putative low energy fields Xi with charges ' (qi , ri ) under
U (1) × U (1)R . One gets four equations for the qi ’s and ri ’s. Allowing charges not
larger than ∼ 50, one needs a least five fields (with rather bizarre charges) to obtain
a solution. This sounds quite unnatural. It is therefore quite possible the system
not to admit solutions, and the global symmetry group be spontaneously broken.
But then, since the theory does not have classical flat directions, according to the
indirect criteria we have discussed at the end of Lecture 7, supersymmetry is broken,
too.
An independent way to see that supersymmetry most likely is broken is to add
one pair of chiral superfields in the 5 and 5̄ representation. There are now classical
flat directions and by adding a mass term for the fundamentals one can show that
supersymmetry is broken, in fact. In the limit m → ∞ this theory reduces to the
original one, without extra matter. If there are no phase transitions in the limit of
large mass, then also the original theory breaks supersymmetry.
This is an instance of the first class of supersymmetry breaking models we dis-
cussed before. We do not have direct access to the effective superpotential nor
to the Kähler potential, so no quantitative statements can be made. However,
symmetry arguments indicate that supersymmetry is most likely broken at the non-

266
perturbative level. In principle, the hidden sector can be a model of this kind.
There exist generalizations of this model which break supersymmetry in a similar
manner. They are based on a gauge group SU (N ), with N odd, N − 4 chiral
superfields transforming in the antifundamental of SU (N ), one chiral superfield
transforming in the antisymmetric representation of SU (N ), and a superpotential
which lifts all otherwise present classical flat directions. One can show that at low
energy the dynamics of all these models essentially reduces to the one of the SU (5)
model described above, and, as the latter, they are therefore expected to break
supersymmetry.

11.3 The 3-2 model: instanton driven SUSY breaking

In what follows, we are going to describe an instance of a calculable model.
Let us consider a supersymmetric theory with gauge group G = SU (3) × SU (2)
and the following matter content

SU (3) SU (2) U (1)Y U (1)R

Qαi 3 2 1/3 1
ei
U 3̄ • −4/3 −8 (11.2)
Dei 3̄ • 2/3 4
Lα • 2 −1 −3

where i is a SU (3) index and α a SU (2) index, and there are two abelian anomaly-
free global symmetries, U (1)Y and U (1)R .
The above global symmetry charge assignment for matter fields comes from the
computation of triangle diagrams with global and gauge currents as detailed in
Figure 11.3. Anomaly-free global symmetries require (letters follow diagrams in
Figure 11.3)

a : 3Y (Q) + Y (L) = 0
e =0
b : 2Y (Q) + Y (U ) + Y (D)
1
c: [3(R(Q) − 1) + R(L) − 1] + 2 = 0
2
1h e ) − 1 + R(D)
i
e −1 +3=0 .
d: 2(R(Q) − 1) + R(U
2
Up to an overall (inessential) normalization, this system of equations admits the

267
 SU(2) SU(3)

U(1) Y anomaly: U(1) Y ; U(1) Y

SU(2) SU(3)
a b

SU(2) SU(3)

U(1) R anomaly: U(1) R ; U(1) R

SU(2) SU(3)
c d

Figure 11.3: One-loop diagrams contributing to anomalies of the abelian global

symmetries of the 3-2 model.

solution

R(Q) = 1 e ) = −8
, R(U e =4
, R(D) , R(L) = −3
e ) = −4/3 , Y (D)
Y (Q) = 1/3 , Y (U e = 2/3 , Y (L) = −1 ,

in agreement with table 11.2. Finally, the theory has a tree-level superpotential

e ,
Wtree = λ QDL (11.3)

which, given the above charge assignment, respects both R and non-R symmetries.
Let us start analyzing this theory at the classical level. The space of D-flat di-
rections has (real) dimension six. This can be seen using the usual parameterization
in terms of single trace gauge invariant operators. These read
   
XA = Q Q eA L = Qαi Q
ei Lβ αβ , Y = det QQ e = αβ AB Qαi Q ei Qβj Qej , (11.4)
A A B

where A = 1, 2, Qei1 ≡ Ue i, Q
ei2 ≡ D
e i and αβ is the invariant tensor of SU (2). That
these are the correct degrees of freedom to describe the space of D-flat directions
can be seen as follows. One can start constructing SU (3) invariants. The only ones
are Qα QeA and Lα , which are both SU (2) doublets. Using them to make (single
trace) operators which are also SU (2) invariant, operators XA and Y follow.
We should now ask whether the superpotential (11.3) affects this space of super-
symmetry preserving vacua, looking for the subspace of D-flat directions where all

268
F-terms also vanish. The F-equation for Lα reads
∂Wtree e2 = 0 .
= λQα Q (11.5)
∂Lα
Contracting with Lα itself this implies that on the moduli space X2 = 0. Similarly,
e1 so to construct the Y invariant, one can show that also Y = 0
multiplying by Qα Q
on the moduli space. Finally, the F-equation for Qe2 is
∂Wtree
= λQi L = 0 . (11.6)
ei2
∂Q
Contracting with Q ei one can show that also X1 = 0 on supersymmetric vacua.
1
The conclusion is that because of the presence of the superpotential (11.3) there do
not exist classical flat directions but rather one single supersymmetric vacuum at
the origin of field space. This implies that if we can prove that some of the global
symmetries are spontaneously broken, then we know supersymmetry is broken, too.
Let us start asking whether a dynamical superpotential is generated. In principle,
we would expect contributions from SU (3) and/or SU (2) gauge dynamics. Let us
choose for now a regime where Λ3 >> Λ2 and λ << 1. In this regime, at scales
lower than Λ3 and bigger than Λ2 , the SU (2) gauge group is weakly coupled while
SU (3) confines. Hence, up to subleading corrections, we can consider the SU (3)
gauge group as dynamical and the SU (2) gauge group acting as a global symmetry
group. Looking at the matter content of the model, we see that from the SU (3)
gauge theory view point this is nothing but SQCD with F = N − 1, where N = 3.
Hence a non-perturbative superpotential is generated and reads
Λ73
Wnon-pert = . (11.7)
Y
This is enough to conclude that supersymmetry is dynamically broken! Due to
(11.7) the minimum of the potential is certainly at some non-zero VEV for Y . Since
R(Y ) = −2 the R-symmetry is then spontaneously broken and since there are no
classical flat directions, supersymmetry is broken, too.
Summing up the tree-level and non-perturbative superpotential contributions,
we get for the full effective superpotential
Λ73
Weff = λX2 + . (11.8)
Y
From the above expression one can easily see that supersymmetry is broken because,
in terms of such low energy fields, we have
∂Weff
= λ 6= 0 . (11.9)
∂X2

269
In this derivation we have implicitly assumed that X1 , X2 and Y are the correct low
energy degrees of freedom, and that no other massless fields show up at any point
of field space. If this were the case, one could have met singularities in the Kähler
metric at such points, and the vacuum energy would have gone to zero there, hence
recovering supersymmetry preserving vacua. For small enough λ, we are safe on this
side. First notice that Wnon-pert brings the theory away from the origin. For λ << 1
the minimum of the potential is certainly in the large Q, Q e region. Since Q and Q e
are charged under both gauge groups, in the supersymmetry breaking vacuum the
gauge symmetry is completely broken, and (heavy) gauge bosons can be integrated
out. This suggests that X1 , X2 and Y are indeed the correct low energy degrees
of freedom and therefore we do not expect singularities (which correspond to extra
massless states) in the Kähler potential.
All what we said so far shows that the 3-2 model belongs, at least, to the sec-
ond class of supersymmetry breaking models we discussed at the beginning of this
section, the so-called non-calculable models. In fact, we can do more.
In the regime we chose, λ << 1, Λ3 >> Λ2 , the ground states are in a weakly
coupled region, and then we are in a situation similar to curve 4a of Figure 11.2.
Therefore, the Kähler potential can be safely taken to be canonical in terms of
UV-fields
K = Q† Q + Q e† Q
e + L† L . (11.10)
We can project this potential onto D-flat directions and get
A + Bx
K = 24 (11.11)
x2
where
 
1 †  1√ † √ 1 A
A= X1 X1 + X2† X2 , B = Y Y , x = 4 B cos arccos 3/2 .
2 3 3 B
(11.12)
We can now plug the above expression and that for the effective superpotential,
eq. (11.8), into eq. (11.1) and, upon minimization with respect to all scalar fields,
find the minima, and hence the vacuum energy E ∼ Ms .
The computation is doable but rather lengthy, so let us first try to get an estimate
of the different scales. The minima will be around a region of field space where the
classical and the non-perturbative contributions to the potential are the same order
(see Figure 11.1), which is the same to ask the two contributions to the effective
superpotential in eq. (11.8) being roughly comparable. In what follows, we think in

270
terms of fundamental UV fields (an acceptable thing to do, given we are in a weakly
coupled region). Calling v the generic VEV of (fundamental) scalar fields at the
non-supersymmetric minima, we get that
Λ37 Λ3
λv 3 ∼ that is v ∼ . (11.13)
v4 λ1/7
This implies that W ∼ Λ33 λ4/7 and hence ∂W/∂φ ∼ Λ32 λ5/7 . Therefore, since
the potential is proportional to the derivative of the superpotential squared (recall
that the Kähler potential is canonical in the UV fields around the supersymmetry
breaking vacua), we finally get

Ms4 ∼ Λ34 λ10/7 that is Ms ∼ Λ3 λ5/14 . (11.14)

Note that this is a leading order estimate. The Kähler potential receives perturbative
and non-perturbative corrections in inverse powers of v. However, in the regime we
are considering these are very small, in the sense that v is much larger than any
other scale in the theory. Indeed

v ∼ λ−1/7 Λ3 >> Λ3 >> Λ2 . (11.15)

λ<<1

From eq.(11.14) we see that Ms << Λ3 , as well as Ms << Λ2 , if λ is small enough.

This gives an a posteriori justification of our claim that X1 , X2 and Y were the
correct low energy degrees of freedom. Supersymmetry breaking occurs at an energy
scale below the confining scale of both non-abelian gauge groups. Therefore, the low
energy effective dynamics is certainly not including light gauge degrees of freedom.
The effective Lagrangian should be (and actually is - recall our conclusions below
eq (11.12)) of O’R-like type, with the only complication of a non-canonical Kähler
potential (in IR field variables); in fact, a supersymmetric σ-model.
As already stressed, our rough estimates do not prevent to compute everything
analytically, by means of eq. (11.1). The answer one gets this way is that the min-
imum of the potential is at X1 = 0, which means that the U (1)Y symmetry is
unbroken (X2 and Y are uncharged under this symmetry). On the other hand we
know that R-symmetry is broken, since the vacua are at finite value of Y , which is
charged under the R-symmetry. This suggests, and confirmed by explicit computa-
tions, the massless spectrum to be composed by a goldstino, an R-axion, associated
to the breaking of the R-symmetry, and finally a fermionic field with hypercharge
Y = −1, whose existence can be proved using t’Hooft anomaly matching condition
for the unbroken U (1)Y symmetry. All other fields have masses of order ∼ λv.

271
 What changes in our analysis if choosing a different regime, namely Λ2 >> Λ3 ?
One can derive an effective superpotential also in this case (which is that of SQCD
with F = N , now) and show that supersymmetry is still broken (though at a
different scale with respect to previous regime). However, generically the theory is
strongly coupled and hence the Kähler potential is unknown, regardless how small
the superpotential coupling λ is. Basically, this is because for F = N the effective
superpotential is not of runaway type and does not push the vacua towards large
field VEVs, where a semi-classical analysis can be done. Therefore, in this regime
the model is non-calculable.
Finally, one can be as general as possible, and consider the two dynamical scales
being the same order, leading to a superpotential of the following form

Λ73

Weff = λX2 + + A Z − Λ24 (11.16)
Y
where Z = ijk Qiα Qjβ αβ Qkγ Lδ γδ . The latter is nothing but just the gauge invariant
expression det M − B B e for the SU (2) theory, which classically is zero, Z = 0. This
shows why in the regime where the SU (2) gauge group is classical, the superpotential
reduces to the expression (11.8) we used before. Notice that since Z is classically
zero, the Z † Z term in the Kahler potential is suppressed by some function of Λ2 /v.
Restoring canonical normalization for Z kinetic term implies that the mass of Z is
enhanced by the inverse of this function. Therefore at low energy, in the regime
where the SU (2) group is nearly classical, one can safely integrate Z out and use
just XA and Y as low energy fields, as we did before. Obviously, the analysis in the
regime where both SU (3) and SU (2) have a quantum behavior is more complicated
but one can again conclude that supersymmetry is broken.
The 3-2 model is the prototype of calculable DSB models, and many interesting
generalizations are available, like the so-called SU (N )×SU (2) and SU (N )×SU (N −
1) models, plus several others.
A final comment is in order. The 3-2 model is a beautiful instance of a DSB
model, and provides a natural way to generate a (small) supersymmetry breaking
scale dynamically, without the need of having dimension-full parameters put by
hand in the theory, as it was the case for the supersymmetry breaking models we
discussed in Lecture 7. This holds at any point in the parameter space. Calculability,
though, does not. As we have seen, the model is fully calculable in the region of the
parameter space where the tree-level dimensionless coupling is parametrically small,
something not at all generic, from a naturalness point of view.

272
11.4 The 4-1 model: gaugino condensation driven SUSY
breaking
Let us now consider a model which to some extent is similar to the previous one,
but differs in that at low energy the theory is not fully higgsed but reduces to a
non-abelian SYM theory. In this case supersymmetry breaking will be driven by
gaugino condensation, and not by instanton effects as for the 3-2 model. Let us
consider a supersymmetric theory with gauge group G = SU (4) × U (1) and the
following matter content

SU (4) U (1) U (1)Y U (1)R

Qi 4 −3 1 1
ei
Q 4̄ −1 1 1 (11.17)
Aij 6 2 −1 −3
S • 4 −2 0
where i, j are (anti)fundamental SU (4) indices and U (1)Y and U (1)R are two abelian
non-anomalous global symmetries, Finally, there is the following tree-level superpo-
tential
e ,
Wtree = λS QQ (11.18)
which respects all symmetries.
As usual, let us start analyzing this theory at the classical level. We first consider
the SU (4) dynamics, only, and ignore the U (1) dynamics as well as the superpoten-
tial (11.18). The SU (4) D-flat directions can be described by the following gauge
invariant operators
e , PfA = ijkl Aij Akl /8 , S .
M = QQ (11.19)

For later purposes let us notice that M has U (1) charge equal to -4, while PfA and
S have U (1) charge equal to 4.
Along a generic flat direction a SU (2) ⊂ SU (4) gauge invariance survives under
which no matter is charged. At scales below the dynamical scale Λ2 of the effec-
tive SU (2) SYM, the theory confines and glueballs and their superpartners can be
integrated out: one is only left with M , PfA and S as low energy degrees of free-
dom. Gaugino condensation of pure SYM leads to the following non-perturbative
generated superpotential
Λ45
Wnon-pert ∼ Λ23 = √ (11.20)
M PfA

273
where the second equality comes from the usual scale-matching condition.
If we now switch-on the U (1) gauge interactions, we have to project the SU (4) D-
flat space onto the subspace which is also U (1) D-flat. The latter is parameterized by
two moduli, M PfA and SM , which hence parametrize the classical D-flat directions.
Note that the tree-level superpotential (11.18), exactly as in the 3-2 model, lifts them
all and leaves only the origin of field space as a supersymmetric vacuum (this can
be seen from the F-term equation for S which sets to zero M and hence both M PfA
and SM ).
Both the U (1) gauge coupling and the superpotential coupling are IR-free, so they
would not affect the above IR analysis leading to (11.20). Therefore, we can now
consider the full superpotential simply adding up the tree-level and non-perturbative
contributions and get
Λ5
Weff = √ 4 + λSM . (11.21)
M PfA
This superpotential is essentially the same as that of the 3-2 model, eq. (11.8).
Therefore, from this point on the analysis is the same as the one we performed in
the previous section. Supersymmetry is broken because of the interplay between
the dynamically generated runaway superpotential term (11.20) and the tree-level
contribution (11.18). Differently from the 3-2 model, though, the fact that, for small
enough λ, SM and M PfA are the correct low energy degrees of freedom, does not
follow from complete Higgsing of the gauge group, since on the moduli space there is
a surviving SU (2) SYM theory. Still, at energies below Λ2 the gauge group confines
and glueballs and their superpartners can be integrated out. Hence, at low enough
energy, the effective theory is indeed given in terms of SM and M PfA only.
Similarly to the 3-2 model, one can argue that for small values of the coupling
λ the model is calculable. This might look strange, given the left-over non-abelian
SU (2) gauge dynamics which is strongly coupled. How can that be? One expects
non-perturbative strong coupling dynamics associated to SU (2) to give rise to cor-
rections to the Kähler potential in terms of some function of ∼ Λ2 /v, where v, as
before, is taken to be the typical scale of a fundamental field VEV. Balancing the
two terms in eq. (11.21), recalling the expression (11.20) one finds

Λ2 Λ2
v∼ which implies << 1 for λ << 1 . (11.22)
λ1/3 v
Hence, quantum corrections to the Kähler potential are suppressed in this regime
and the model is calculable.

274
 Let us stress again, though, how different the dynamics is with respect to the 3-2
model. There, the smallness of λ ensures that both gauge groups are fully broken
at very high energy, and therefore quantum corrections due to gauge dynamics
suppressed. Here, instead, a fully unbroken gauge groups survives at low energy.
The computation of low energy spectrum and interactions goes along similar lines
as the 3-2 model, and we do not repeat it here (for instance, also for this model the
tree-level superpotential has an R-symmetry which is spontaneously broken in the
vacua; hence we expect, as for the 3-2 model, an R-axion in the massless spectrum).
Let us summarize, instead, the physical picture one should bare in mind. The theory
in the UV is a SU (4)×U (1) gauge theory. At a scale v this is broken down to SU (2).
This left-over non abelian gauge theory confines at a scale Λ2 << v, below which
we have a low energy effective theory with chiral superfields, only. Gaugino con-
densation gives rise to a superpotential contribution which induces supersymmetry
breaking at a scale Ms . Note that in the limit λ << 1 the supersymmetry breaking
scale is parametrically smaller than Λ2 (using the same rationale we used for the
3-2 model, one easily sees that Ms ∼ Λ2 λ1/6 which is well below Λ2 , if λ is small).
Hence, at the supersymmetry breaking scale all gauge degrees of freedom are heavy
and do not contribute to the effective action, which justifies the description in terms
of SM and M PfA, only.
The 4-1 model has several generalizations. The most straightforward ones are
theories with gauge group SU (2l) × U (1) and matter consisting of a chiral su-
perfield transforming in the anti-symmetric representation of SU (2l), 2l − 3 anti-
fundamentals Q,e one fundamental Q, and 2l − 3 singlets Si . Supersymmetry break-
ing is again driven by gaugino condensation of a IR left-over SU (2) gauge group,
provided a suitable tree-level superpotential is added which lifts all classical flat
directions. The 4-1 model corresponds to l = 2.

11.5 The ITIY model: SUSY breaking with classical flat

directions
Let us now consider an instance of a non-calculable model. Its interest lies in the
fact that supersymmetry is broken even though the theory is non-chiral and admits
classical flat directions (the latter get lifted by non-perturbative effects not leading
to runaway behavior).
Let us consider a gauge theory with group G = SU (2), four fundamental fields

275
Qi (which correspond to two flavors, since for SU (2) the fundamental and anti-
fundamental representations are equivalent) plus six singlets Sij and a superpotential

Wtree = λSij Qi Qj (11.23)

(notice that the product Qi Qj is antisymmetric since what it really means is Qiα Qjβ αβ
where α, β are SU (2) gauge indices). This theory admits a SU (4) flavor symme-
try group (this enhancement of the global non R-symmetry group from SU (F )L ×
SU (F )R × U (1)B to SU (2F ) is always there whenever the gauge group is SU (2)),
under which the Qi ’s transform in the fundamental and the singlets in the anti-
symmetric representations, respectively. Hence, the tree-level superpotential (11.23)
respects the flavor symmetry. As usual, let us start studying the classical behavior
of the theory. The SU (2) D-flat directions can be parameterized by six meson-like
operators M ij ∼ Qi Qj , which transform in the 6 of SU (4) and satisfy the classical
constraint of SQCD with N = F = 2

PfM = ijkl M ij M kl = 0 , (11.24)

where indices i, j should be seen as SO(4) indices (recall that SO(4) ' SU (2) ×
SU (2) and notice that for any nonzero value of M the global symmetry is broken
to SU (2) × SU (2)). The F-flatness condition for Sij sets all Qi ’s to zero hence all
flat directions are lifted but the singlets.
At the quantum level the classical constraint (11.24) is modified and the full
effective superpotential reads


Weff = λSij M ij + A ijkl M ij M kl − Λ4 , (11.25)

where A is a Lagrange multiplier. The F-equation for Sij still gives M ij = 0 but
now this is in conflict with the quantum constraint, i.e. the F-equation for the La-
grange multiplier A. Therefore, supersymmetry is broken. More precisely, working
out the potential from the expression (11.25) one can show that, up to symmetry
transformations, the minimum is at Mij = Λ22 , S13 = S14 = S23 = S24 = 0 and
S12 = S34 ≡ S. Therefore, there is a pseudoflat direction parametrized by S.
This model is instructive in many respects, which we consider in turn.
Having a flat direction, parametrized by S, one could be worried about where,
in field space, the supersymmetry breaking vacua lie, once quantum corrections in
the coupling λ are taken into account. In principle, there can also be a runaway.
A careful analysis, which we refrain to do here, shows that this is not the case: for

276
small enough λ and large λhSi the Kähler potential for S can be shown to grow
logarithmically for large S, hence ensuring that the actual minimum is stabilized at
a finite distance in field space.
Notice also that this model is non-chiral. Therefore, one could add a mass
term for all fields, lifting all classical flat directions. At low energy one could then
integrate all chiral fields out and end-up with pure SU (2) SYM, which does not
break supersymmetry (it has two vacua and Witten index equal to 2). How that
can be? The answer comes from a careful analysis of the massless limit.
Let us add a mass perturbation to the superpotential (11.25)
1

Weff = λSij M ij + mij M ij + m̃Pf S + A ijkl M ij M kl − Λ4 . (11.26)
2
ij
The F-equations for M and Sij set
 4 1/2
ij ijkl Λ
hM i ∼  mkl
Pf m
 4 1/2
mij Λ
hSij i ∼ .
m̃ Pf m
where the square root gets two values, corresponding to the two vacua of pure
SU (2) SYM. Take now the limit m̃, mij → 0 with the ratio fixed. This way, hM ij i
has a finite limit, but hSij i is pushed all the way to infinity. This implies that the
supersymmetry preserving vacua are also pushed to infinity and disappear from the
spectrum, recovering our previous result.
This is an instance of discountinuos change of the Witten index, which moves
from 2 to 0 in the limit of vanishing masses. This is because the mass terms change
the behavior of the Hamiltonian in the large field region. As the limit m̃ → 0 is
taken, the asymptotic behavior of the potential changes since now there are classical
flat directions (and the Witten index can, and does, change).
The ITIY model admits many generalizations. An interesting class is based on
SQCD with gauge group U Sp(2N ) and F = N + 1 flavors. This theory has a
SU (2F ) = SU (2N + 2) flavor symmetry, and enjoys a quantum deformed moduli
space, very much like SU (N ) SQCD with F = N flavors. Coupling the quark su-
perfields to a set of gauge singlets transforming in the antisymmetric representation
of the flavor symmetry group via a superpotential like (11.23), one can show super-
symmetry is broken in a way identical to that of the original ITIY model (in fact,
recalling that SU (2) ' U Sp(2), one sees that the ITIY model corresponds to the
case N = 1 of the above class).

277
11.6 DSB into metastable vacua. A case study: massive
SQCD
As a final project, we want to discuss the possibility that supersymmetry is broken
dynamically into metastable vacua.
A model of DSB into metastable vacua share some basic properties with or-
dinary DSB models. The theory should be a gauge theory and should not break
supersymmetry at tree level. Only non-perturbative corrections should. The differ-
ence is that the non-perturbative dynamics does not lift classical supersymmetric
vacua but just ensure that local minima of the potential whose nature is intrinsically
non-perturbative, arise.
On general ground, due to Witten index arguments, R-symmetry arguments,
etc... the landscape of theories admitting metastable DSB vacua is much larger
than those admitting fully stable DSB vacua. This has been known for a long time,
but only more recently it was made concrete. In 2006 Intriligator, Seiberg and Shih
(ISS) proved the existence of DSB metastable vacua in the most innocent-looking
supersymmetric gauge theory one can imagine: massive SQCD. Note that this is
a non-chiral theory, with supersymmetric vacua (a full moduli space, in fact, in
the massless limit), non-vanishing Witten index and no R-symmetry (quarks mass
terms explicitly break the non-anomalous R-symmetry of massless SQCD). Even
more strikingly, the model is calculable, in the sense that around these metastable
vacua one can compute both the superpotential and the Kähler potential, and hence
the effective Lagrangian describing the dynamics of light fields.
These results have been extended into several directions, and many interesting
applications have been found. In what follows, we will just review the basic model,
which represents the core of all these developments.

11.6.1 Summary of basic results

Since the derivation is rather lengthy, let us anticipate the upshot of the analysis
we are going to perform. This is as follows: SU (N ) SQCD with (light) massive
flavors in the free magnetic phase (that is for N + 1 ≤ F ≤ 32 N ) admits metastable
supersymmetry breaking vacua which, for m << Λ, where m is the scale of quark
masses and Λ the dynamical scale of the theory, can be made parametrically long
lived. More precisely, the theory admits:

278
 • N supersymmetric vacua along the mesonic branch, at

1/N
hM iSUSY = mF −N Λ3N −F e i1 i2 ...iF −N i = 0 . (11.27)
, hBi1 i2 ...iF −N i = 0 , hB

• A compact space of metastable supersymmetry breaking vacua along the bary-

onic branch

e i1 i2 ...iF −N i =
hBi1 i2 ...iF −N i, hB 6 0 , hM iMETA = 0 , (11.28)

with vacuum energy VMETA ∼ N |mΛ|2 .

One can also compute the life-time of the metastable vacua and find that
r
SB −4(3N −2F )/N m
τ ∼e where SB ∼  and  = , (11.29)
Λ
with SB the Coleman bounce action. This implies, as anticipated, that for small
masses, i.e.  << 1, the metastable vacua can be made arbitrarily long-lived, and
hence potentially viable, phenomenologically.

2
VPEAK ~ F |m Λ|

2
VMETA~ N |m Λ|

<M < M
susy

~
B,B

Figure 11.4: The scalar potential of massive SQCD in the free magnetic phase.
On the mesonic branch there are supersymmetric vacua. On the baryonic branch
there are supersymmetry breaking vacua, which are metastable and parametrically
long-lived.

279
11.6.2 Massive SQCD in the free magnetic phase: electric description

Consider SQCD with gauge group SU (N ) in the free magnetic phase, namely for
N + 1 ≤ F ≤ 23 N . This theory has many supersymmetric vacua, actually a full
moduli space. Let us add a mass term for all matter fields

e ≡ Tr mM .
Wm = Tr mQQ (11.30)

where the trace is taken on gauge and flavor indices. Notice that (11.30) breaks
the SQCD R-symmetry explicitly, while the flavor symmetry group is broken to a
subgroup H, whose structure depends on the specific form of the matrix m (more
later).
This theory has two mass scales, the quarks mass, which with a slight abuse of
language we call again m, and Λ, the dynamical scale of the theory. Let us consider
the two obvious possible regimes in turn.

a. m > Λ
The theory at low energy flows to pure SYM with gauge group SU (N ) and has
N (isolated) supersymmetric vacua. By scale matching, we obtain

Λ3N
L = det m Λ3N −F (11.31)

which implies

N1
Weff = N Λ3L = N det m Λ3N −F , (11.32)
an effective superpotential displaying, correctly, the N vacua of pure SU (N ) SYM.
What’s this, really? The mass matrix m and the meson matrix M are Legendre
dual variables. The effective superpotential above is nothing but the effective super-
potential once the mesons have been integrated out. Hence, using formula (10.78),
we get the matrix equation

N1 1
hM iSUSY = det m Λ3N −F (11.33)
m
which tells where in the moduli space the N supersymmetric vacua sit: they corre-
spond to the N roots of the above equation.

b. m < Λ
In this case, which is actually the one we will be interested in, eventually, it is
not completely correct to proceed as before since strong coupling dynamics, driven
by Λ, enters before being allowed to integrate the massive quarks out. The more

280
correct thing to do, in this case, is to notice that m and M are Legendre dual
variables, and integrate M in starting from eq. (11.32). In practice, one should take
the determinant of eq. (11.33), solve for det m and follow the procedure outlined in
section 10.4.3, getting finally
 3N −F  N −F
1
Λ
Weff = (N − F ) + Tr mM . (11.34)
det M
Then we can find eq. (11.33) simply solving the F-equations for M . Recall, however,
that strictly speaking det M = 0 for F ≥ N + 1, so one has to go a bit off-shell in
performing the computation. The final result, eq. (11.33), is of course a perfectly
meaningful on-shell result.
The upshot is that, no matter the value of m, there exist N supersymmetric vacua
on the mesonic branch. That baryon VEVs are vanishing can be easily argued as
follows. As m and M are Legendre dual variables, one can think of b and b̃ as
sources for the baryons B and B,e deforming the theory by ∆W = b B + b̃ B e (flavor
indexes are suppressed, for the ease of notation). At low energy the theory reduces
to pure SYM with gauge group SU (N ) and effective superpotential (11.32). Using
now eq. (10.78) applied to the Legendre dual variables b and B (respectively b̃ and
e one concludes that hBi = hBi
B) e = 0. Hence, indeed, the supersymmetric vacua
(11.33) have zero baryon number.
In general, m is a matrix transforming under the anti-fundamental of SU (F )L
and the fundamental of SU (F )R . This matrix can always be diagonalized via a
bi-unitary transformation and, from here on, we choose for simplicity all entries to
be equal, mi = m. The superpotential term hence reads

Wm = mTrM , (11.35)

where now m is just a number. With this choice, the SU (F )L × SU (F )R flavor

symmetry group is broken to SU (F )D . Similarly, eq. (11.33) now reads

N1 1
1 F −N
hM iSUSY = mF Λ3N −F = mF −N Λ3N −F N = 2 N Λ2 , (11.36)
m
p
where  ≡ m/Λ.

11.6.3 Massive SQCD in the free magnetic phase: magnetic description

So far, we have derived the first part of ISS statement, the easy one. We have ob-
tained, via holomorphic decoupling, the N supersymmetric vacua of massive SQCD,

281
and found they lie along the mesonic branch. In order to find something more inter-
esting, we have to turn to the Seiberg dual description of the theory, i.e. mSQCD.
Since we are in the magnetic-free phase we choose, in what follows, Λel = Λm =
µ ≡ Λ. The magnetic dual superpotential, including the mass deformation (11.35)
reads
Wm = hTr qΦq̃ − mΛ hTr Φ , (11.37)
where
1
Φ= M . (11.38)
hΛ
Let us start by recovering, using magnetic variables, the N supersymmetric vacua
we have found before. To this aim, let us suppose we give some non-vanishing VEV
to the gauge singlet Φ. This provides a mass to dual quarks, q and q̃, which can then
be integrated out. The theory reduces to pure SU (F − N ) SYM and the effective
superpotential one obtains, upon holomorphic decoupling, reads

Weff = −mΛhTr Φ + (F − N ) Λ3L . (11.39)

By matching at dual quarks mass scale we find the relation

3(F −N )
F −N
1
ΛL = hF det Φ Λ3(F −N )−F that is Λ3L = hF det Φ Λ2F −3N . (11.40)

We can substitute the above relation into the superpotential (11.39) and get

F −N
1
Weff = −mΛhTr Φ + (F − N ) hF det ΦΛ2F −3N . (11.41)

The F-equation for Φ gives

√ 3N −2F √
hhΦiSUSY = mΛ − N >> mΛ
F −N
= Λ 2 N << Λ , (11.42)

where the inequalities hold if  is small. The expression in the first line says that
the supersymmetric vacua are at a parametrically large distance from the origin of
√
field space in units of mΛ, while the second one ensures that the above analysis
is meaningful in mSQCD. Indeed, these vacua, though located in a very quantum
region from mSQCD point of view, are well below the Landau pole and hence reach-
able within the magnetic description. Self-consistently, using the map (11.38) one
can easily see that the vacua (11.42) are nothing but the vacua (11.36).
Also using magnetic variables one can easily conclude that in the supersymmetric
vacua baryon VEVs are vanishing. As already observed, on such vacua the dual

282
quarks are massive and can be integrated out, hence their VEVs vanish. The VEVs
of the magnetic baryons can be easily computed from that of dual quarks. Indeed,
the magnetic theory is IR free and in the supersymmetric vacua the VEV of the
product of qs is the same as the product of the VEVs of each q. Therefore

hbj1 j2 ...jN i = 0 (11.43)

and similarly for b̃j1 j2 ...jN . From the map (10.118) it then follows that hBi1 i2 ...iF −N i =
e i1 i2 ...iF −N i = 0, as anticipated.
hB
Notice that while the  parameter defined here and in the electric description is
one and the same, the  → 0 limit should be understood differently. In the electric
description Λ is a dynamical RG-invariant scale and the limit of small  is obtained
sending m → 0 keeping Λ fixed. In the magnetic description, Λ is a cut-off scale,
above which the theory is not defined. The limit should now be understood as
√
Λ → ∞ keeping mΛ, the mass scale entering the superpotential (11.37), fixed
p √
(notice that  = m/Λ = mΛ/Λ). This apparently pedantic observation will be
relevant later.
Let us now come back to the expression (11.37) and analyze the properties of
deformed mSQCD more closely. We will do this in steps and forget, for a while, that
the magnetic group SU (F − N ) is gauged. If gauge degrees of freedom are frozen,
the vacua of the theory are obtained solving F-equations only. From eq. (11.37)
these read 
i a i
F Φji = q̃a qj − mΛδj

F qi = hΦij q̃ j (11.44)


F q̃j = hqi Φij
where a are SU (F − N ) indices. We see that the first set of equations cannot be
solved. The rank of q̃ai qja is at most F −N while that of δji is clearly F . Hence we can
set to zero at most (F −N ) terms of FΦ -equations: we are left with F −(F −N ) = N
non-vanishing F-terms. On the other hand, the F-equations for q and q̃’s are easily
satisfied. We conclude that supersymmetry is broken, and is so by a rank condition.
The potential energy gets contribution from the N F-equations that cannot be set
to zero and hence reads
VMETA ∼ N |mΛ|2 . (11.45)
The supersymmetry breaking vacua are at
! ! !
0 0 q0 q̃0
hΦi = , hqi = , hq̃ T i = (11.46)
0 Φ0 0 0

283
where q0 q̃0 = mΛ1F −N , with q0 and q̃0 being F − N × F − N matrices, and Φ0 an
arbitrary N ×N matrix. Therefore, we find a pseudomoduli space of supersymmetry
breaking vacua parameterized by Φ0 , q0 and q̃0 . If this analysis were correct, the
picture we would obtain is what is summarized, schematically, in Figure 11.5.
V

?
VMETA~ N |m Λ|
{
2

Λ Φ

q,q~

Figure 11.5: Linearly deformed mSQCD classical potential.

The question we should now try to answer is to what extent the above results
are solid in the full quantum theory. So far, our analysis was classical, both because
we have been ignoring local SU (F − N ) gauge dynamics, and because, even within
the ungauged model, we have not taken into account quantum corrections coming
from the coupling h. Let us start considering quantum effects due to h. Later, we
will consider the role of gauge degrees of freedom and interactions.
Let us first notice that the supersymmetry breaking vacua lie relatively near to
the origin, which is the more classical region for mSQCD, which is a IR-free theory.
√
Indeed, as already observed, the scale mΛ is set to be the mass scale entering
the mSQCD Lagrangian by the superpotential (11.37), the natural mass unit to
measure dimensionfull quantities in the magnetic theory. Looking at eqs. (11.45)
and (11.46), we see that the energy density of the supersymmetry breaking minima
√
is order one in units of mΛ, and so are the values of q0 and q̃0 on such minima
(the Φ0 flat direction does not play any role here since, as we will see momentarily,
quantum corrections lift this degeneracy and set Φ0 = 0). On the contrary, looking
√
at eq. (11.42) we see instead that hΦiSUSY is parametrically large in units of mΛ.
Since mSQCD is IR-free, we can then safely take the Kähler potential to be canonical
in the region where the supersymmetry breaking vacua sit, that is


K = Tr Φ† Φ + q † q + q̃ † q̃ . (11.47)

284
A second comment regards global symmetries. In the limit where the magnetic
gauge group SU (F − N ) is taken to be ungauged, mSQCD has a global symmetry
group SU (F − N ) × SU (F )L × SU (F )R × U (1)B × U (1)R0 which is broken by the
second term in (11.37) to G = SU (F − N ) × SU (F )D × U (1)B × U (1)R0 , where
under the non-anomalous R-symmetry U (1)R0 the dual quarks are chargeless and Φ
has R-charge R0 = 2, as dictated by the tree-level superpotential (11.37).
On the supersymmetry breaking vacua (11.46) the group G is spontaneously
broken. The vacua with maximal unbroken global symmetry sit at (up to unbroken
flavor rotations)
√
Φ0 = 0 , q0 = q̃0 = mΛ 1F −N , (11.48)
which preserve H = SU (F − N )D × SU (N ) × U (1)B 0 × U (1)R0 (notice in particular
that the R-symmetry is not broken).
In order to study quantum corrections around the supersymmetry breaking
vacua, we can proceed as we did for the O’Raifeartaigh model, and compute the
masses of the fluctuations of Φ, q and q̃ as functions of the pseudomoduli Φ0 , q0 and
q̃0 . It is reasonable to expect that the actual vacuum will sit at a point of maxi-
mal symmetry, so as a working hypothesis let us expand around (11.48). We can
parametrize the fluctuations as (for ease of comparison we use the same notation of
ISS)
! √ ! √ !
δY δZ † mΛ + √12 (δχ+ + δχ− ) mΛ + √1 (δχ+ − δχ− )
Φ= ,q= , q̃ T = 2
δ Ze δ Φ̂ √1 (δρ+ + δρ− )
2
√1 (δρ+ − δρ− )
2
(11.49)
where we have expanded around the maximally symmetry preserving vacua (11.48).
What one finds is that the model looks as N copies of a O’Raifeartaigh-like model
and after computing the one-loop effective potential the spectrum is as follows:
√
• Some fields have (tree-level) mass ∼ |h mΛ| from the classical superpotential
(11.37).
√
• Pseudomoduli are all lifted and get non-tachyonic masses at one-loop ∼ |h2 mΛ|
from their coupling to massive fields (this shows in retrospective that our ed-
ucated guess was right, after all).

• Some fields remain exactly massless. These are: the Goldstone bosons associ-
ated to the coset G/H, a goldstino, as well as several fermionic partners of Φ0
pseudomoduli.

285
So, after taking into account quantum corrections in the tree-level coupling h we are
left with a compact moduli space of stable non-supersymmetric vacua. This moduli
space is robust against quantum corrections, because it is protected by symmetries.
What does it change of the above analysis if we now switch-on gauge interactions,
namely we let SU (F − N ) group being gauged? Interestingly, not much happens
around the supersymmetry breaking vacua (11.48).
First, besides F-equations (11.44) we have now to impose D-equations on the
supersymmetry breaking vacua (11.48). In fact, these are trivially satisfied, as one
can verify plugging VEVs (11.48) into
X

Tr q † TA q − q̃TA q̃ † = 0 . (11.50)
A

Hence, the compact space parameterized by (11.48) remains a minimum of the

potential (D-terms identically vanish and therefore do not contribute to the vacuum
energy).
Second, the SU (F − N ) gauge group is completely higgsed in the vacua (11.48),
√
since hqo i = hq̃0 i =
6 0. Gauge bosons acquire a mass ∼ g mΛ, eating some of the
previously massless Goldstone bosons of the ungauged model. The only change,
then, is that the compact moduli space is smaller since global symmetries in the
gauged model are less, to start with. In particular we have now that G = SU (F )D ×
U (1)B and H = SU (F − N ) × SU (N ) × U (1)B 0 . Notice that the R-symmetry of the
ungauged model U (1)R0 is now anomalous, while, as we already observed, the non-
anomalous R-symmetry which mSQCD shares with SQCD is explicitly broken by
the mass term in the superpotential (i.e. the linear term in Φ, in mSQCD language).
Finally, the gauging does not affect the computation of the one-loop effective
potential, either, since the tree level spectrum of massive SU (F − N ) fields is super-
symmetric and gives no contribution to StrM 2 . This happens because, as already
observed, D-terms vanish on the vacua (11.48), and the non-zero expectation values
of q and q̃ which provide masses to SU (F − N ) gauge fields do not couple directly
to any non-vanishing F-term.
So we conclude that, up to a restriction of the compact moduli space, the super-
symmetry breaking vacua we found classically in mSQCD survive quantum correc-
tions and are hence supersymmetry breaking vacua of our original theory!
Gauging the SU (F −N ) group does have (drastic) consequences on other regions
of field space, though. We already know, and we have proved it using both electric

286
and magnetic variables, that the theory has supersymmetric vacua on the mesonic
branch. Besides other things, this makes the supersymmetry breaking vacua (11.48)
being not absolute minima of the potential. Using magnetic language the effect of
gauging is the generation of a non-perturbative superpotential contribution
1 F
Wnp ∼ (detΦ) F −N ∼ Φ F −N . (11.51)

This contribution is irrelevant near the origin, where supersymmetry breaking vacua
sit, and becomes more and more important the farer we move along the mesonic
branch. This operator plays the same role that a mass term for the chiral super-
field Φ2 played in the modified O’Raifeartaigh model: it brings in supersymmetry
preserving vacua. The difference is that everything happens dynamically, here. So
we conclude that mSQCD has metastable supersymmetry breaking vacua semiclas-
sically, and non-perturbative restoration of supersymmetry by a a dynamical gener-
ated superpotential. On the other hand, in terms of the original SQCD theory, the
supersymmetry breaking vacua are highly quantum mechanical, since they sit in a
region which is, say, highly quantum, from SQCD view point.
The final picture we obtain is represented in Figure 11.6.
V

1 F
2
V PEAK
~ F |m Λ| Wnp ~ (det Φ) F−N~ Φ F−N

2
VMETA ~ N |m Λ|

<Φ
<
~ mΛ ε
−# Φ
susy
~ ~ mΛ
<q>=<q>

q,q~

Figure 11.6: The ISS potential in magnetic dual variables.

A final comment regards the R-symmetry breaking pattern. Here again, we find
a consistent picture with what we learned in previous lectures. The SQCD original
R-symmetry is explicitly broken by the mass term (11.30). Hence, the theory does
not satisfy the necessary condition for supersymmetry breaking, and indeed it has
N supersymmetric vacua, and non-vanishing Witten index. On the other hand, the

287
anomalous U (1)R0 R-symmetry is restored, approximately, near the origin. This is
more transparent using magnetic variables. The superpotential contribution (11.51)
breaks R0 explicitly, but this operator is irrelevant near the origin and this is why
this symmetry arises as an approximate R-symmetry around the vacua (11.48).
Therefore, by the (extended) NS criterium, one would expect metastable vacua to
arise there, and this is exactly what happens.
It is amusing to notice that in the ISS vacuum is plenty of massless fields, so
there is no mass gap: a theory with tree-level masses for all matter fields and strict
confinement, admits vacua without a mass gap!
In all our discussion there is one point that we have overlooked. The magnetic
theory has a UV cut-off, Λ. Do our results depend on the physics at scale Λ? Luckily,
not in the limit we are interested in, namely
r (
m SQCD : m → 0 , Λ fixed
= << 1 −→ √ (11.52)
Λ mSQCD : Λ → ∞ , mΛ fixed

First, the analysis within the macroscopic theory (i.e. ungauged mSQCD) is valid,
√
since this was done at scales of order mΛ = Λ, which are well below the UV
cut-off Λ, if  is small. Second, also the supersymmetry preserving vacua can be
seen in the magnetic theory: as we have already observed, second line of eq. (11.42),
for small  they are also well below the scale Λ
F −N
hΦiSUSY = Λ 2 N << Λ , (11.53)

and hence are very weakly affected by any Λ-physics effects. Finally, the one-loop ef-
√ √
fective potential gives pseudomoduli mass squares of order |mΛ|, that is mΛ mΛ,
which is not an holomorphic expression. On the other hand, corrections from Λ-
physics are holomorphic in mΛ and provide mass contributions of the form

mΛ mΛ
· = |mΛ|2 /|Λ|2 = |mΛ|2 << |mΛ| , (11.54)
Λ Λ
which are again subleading for  << 1. A direct way to see this is to note that
corrections in Λ would make the Kähler potential (11.47) not being canonical. In
particular, to leading order, we would get a contribution as δK = c/|Λ|2 (ΦΦ† )2 ,
with c a number of order one. This is reminiscent of the Polonyi model with quartic
Kähler potential we discussed previously. A similar computation as the one there
gives a contribution to the pseudomoduli mass as in eq. (11.54), δm2 ∼ |mΛ|2 /|Λ|2 .

288
 The last important check we have to do regards the life-time of the supersym-
metry breaking vacua. The life-time can be computed using the Coleman bounce
action. Intuitively, the more the two vacua are far in field space in units of the
energy difference between them, the more one might expect the life-time to be long.
This expectation is confirmed by an explicit computation. It turns out that in the
present case we are in a situation in which the so-called thin-wall approximation is
valid. In such a situation, up to inessential numerical factors, the bounce action is
proportional to the ratio between the fourth power of the distance, in field space,
between the supersymmetry breaking and the supersymmetry preserving vacua, and
the value of the energy difference between them. Using previous formulas SB hence
reads
(∆Φ)4 3N −2F
SB ∼ ∼ −4 N , (11.55)
VMETA
which is indeed large for  << 1. This ensures that the ISS vacua are parametrically
long lived, since τ ∼ eSB . Notice that the largeness of the bounce action is due to
different effects depending whether one is working in electric or magnetic variables.
√
From mSQCD view point it is large since ∆Φ is parametrically large in units of mΛ.
1/4
From SQCD view point, the bounce action is large because VMETA is parametrically
small in units of Λ.

11.6.4 Summary of the physical picture

Let us summarize the physical picture which emerges from our analysis of massive
SQCD in the free magnetic phase.
Since the theory is UV-free, at high energies, larger than the dynamical scale,
E > Λ, the theory is weakly coupled, it can be described in terms of electric variables
and the gauge coupling gel increases along the flow. The scale Λ is an IR cut-off
for SQCD and a UV cut-off for the IR-free dual magnetic theory. Hence, at scales
E ∼ Λ, in order to describe the dynamics of the theory one should better change to
magnetic variables. Below Λ but above hΦi the theory renormalizes as for an IR-free
theory, in the sense that the magnetic gauge copling gm decreases. This goes on until
one meets the scale hΦi. What happens next depends on the value of such scale.
If hΦi 6= 0 at E ∼ hΦi the dual quarks get massive and the theory reduces to pure
SYM and leads to N supersymmetric vacua. If instead hΦi = 0 the magnetic theory
becomes completely free, gauge degrees of freedom get frozen and one is driven to
√
the supersymmetry breaking vacua at E ∼ mΛ.

289
 Do exist metastable vacua in massive SQCD for different values of F ? The
existence of metastable vacua in massive SQCD has been proven only in the free
magnetic phase. For F ≥ 3N SQCD the dynamics is trivial and there do not exist
ISS-like metastable vacua whatsoever. In the conformal window, 32 N < F < 3N ,
the analysis is not easy since mSQCD is not IR-free. Moreover, one can show that
hΦiSUSY turns out to be very near to the origin of field space, hence making metasta-
bility difficult to achieve anyway. Finally, the non-perturbative generated super-
potential Wnp is relevant in the IR, indicating the difficulty in treating separately
classical and quantum effects. For F < N the runaway is too strong and there are
simply no tools to say whether local minima develop along the moduli space. Finally,
for F = N the existence of ISS vacua cannot actually be proven using the magnetic
dual theory, which does not exists for F = N , but can only be inferred using holo-
morphic decoupling starting from F = N + 1. Even though there are convincing
arguments in favor of ISS vacua also in this case, given the state we are speaking
about is not supersymmetric, the procedure requires some assumptions which are
not fully under control; hence, the case F = N is not completely understood, in fact.
A natural question is therefore whether is it possible to find ISS-like vacua in theo-
ries with a quantum deformed moduli space, as SU (N ) SQCD with F = N is. The
answer is for the affermative. It has been shown that suitable deformations of the
U Sp(2N ) ITIY model we discussed in section 11.5 allow for dynamically generated
metastable vacua, in a theory with a quantum deformed moduli space, as the ITIY
model and any of its generalizations actually are. Basically, giving supersymmetric
masses to some of the singlets Sij , one can show that supersymmetric vacua come
in from infinity (because, integrating out massive singlet(s), mesonic flat directions
develop) but dynamically generated non-supersymmetric local vacua survive. More-
over, such vacua can be made parametrically long lived in a region of the parameter
space which, interestingly enough, coincides with the region where Kähler potential
corrections are fully under control.
The ISS model admits many generalizations (including those above). In particu-
lar, at the price of some complications and subtleties which we cannot discuss here,
one can generalize the model in order to let the emergent IR R-symmetry to be
spontaneously broken in the supersymmetry breaking vacua. This is a feature that
the original ISS model does not have, since, as we have seen, quantum corrections
stabilize U (1)R0 charged moduli at the origin, Φ0 = 0. And, if one thinks of the
ISS model as a hidden sector in gravity or gauge mediation scenarios, having broken
R-symmetry is a necessary condition to let gauginos getting (Majorana) mass.

290
References

[1] E. Poppitz and S. P. Trivedi, Dynamical supersymmetry breaking, Section 5,

Ann. Rev. Nucl. Part. Sci. 48 (1998) 307 [arXiv:hep-th/9803107].

[2] Y. Shadmi, Supersymmetry breaking, Section 3, arXiv:hep-th/0601076.

[3] Y. Shadmi and Y. Shirman, Dynamical supersymmetry breaking, Sections IV

and VI.A, Rev. Mod. Phys. 72 (2000) 25 [arXiv:hep-th/9907225].

[4] K. A. Intriligator and N. Seiberg, Lectures on Supersymmetry Breaking, Sec-

tions 2.7, 2.8, 3.3, 3.7 and 4.3, Class. Quant. Grav. 24 (2007) S741 [arXiv:hep-
ph/0702069].

[5] K. A. Intriligator, N. Seiberg and D. Shih, Dynamical SUSY breaking in meta-

stable vacua, JHEP 0604 (2006) 021 [arXiv:hep-th/0602239].

[6] S. R. Coleman, The Fate Of The False Vacuum. 1. Semiclassical Theory, Phys.
Rev. D 15 (1977) 2929 [Erratum-ibid. D 16 (1977) 1248].

[7] M. J. Duncan and L. G. Jensen, Exact tunneling solutions in scalar field theory,
Phys. Lett. B291 (1992) 109-114.

[8] M. Bertolini, L. Di Pietro and F. Porri, Dynamical completions of generalized

O’Raifeartaigh models, JHEP 1201 (2012) 158 [arXiv:1111.2307 [hep-th]].

291
12 Supersymmetric gauge dynamics: extended supersym-
metry

In this lecture we will focus on asymptotically free gauge theories with extended
supersymmetry and try to understand their quantum dynamics, in analogy with
what we did for N = 1 supersymmetric theories in chapter 10.
Asymptotically free gauge theories can enjoy different phases at low energy. In
the case of N = 1 supersymmetry, thanks to powerful non-renormalization theorems
and more generally holomorphy, we were able to understand a great deal about the
possible phases such field theories can enjoy. This can obviously be done also for
theories with extended supersymmetry. In fact, the beauty of theories with extended
supersymmetry is that, quite often, one cannot only determine the phase in which
the theory is, but also derive the exact expression of the low energy effective action.
The main purpose of this lecture is to show when and how this is possible.

12.1 Low energy effective actions: classical and quantum

Let us start making some general comments, independent from supersymmetry.
Suppose to start from some matter coupled, asymptotically free gauge theory. At
low energy, its dynamics will be described by some (non-renormalizable) effective
action whose degrees of freedom will be in general very different from UV ones.
What is the structure one would expect for such an action?
Let us assume that in the vacuum we want to expand the theory about, the po-
tential vanishes, V = 0. This is not a restriction, since the minimum of the potential
can always be chosen to vanish via a constant shift in the Lagrangian. Moreover, in
the context of supersymmetric theories, which is what we are eventually interested
in, this is not even a choice but a necessary condition, as far as supersymmetric
vacua.
The leading dynamics around these vacua is governed by light fields, eventually
only massless ones, as we take the cutoff energy characterizing the low energy effec-
tive action to be lower than any scale in the theory. In this limit, the physics is, by
definition, scale invariant. However, the nature of the corresponding IR fixed point
is not unique. If no massless fields are present (like in the case of strict confinement)
there are no propagating degrees of freedom in the limit E → 0, so the IR theory is
empty and the IR fixed point a trivial one. The theory is gapped. If massless fields

292
are present, instead, the theory can be in a free or an interacting phase.
A necessary condition for having an interacting conformal field theory at low
energy is that massless non-abelian gauge fields are present in the effective action
(at least if one assumes that a local Lagrangian description can exist). Indeed, by the
Coleman-Gross theorem, in four space-time dimensions any theory of scalars, spinors
and abelian gauge fields flows in the IR to a free (or trivial, if everything get a mass)
theory. In presence of massless non-abelian gauge fields, one can either end up with,
say, confinement, hence a trivial IR fixed point, or, indeed, an interacting conformal
field theory (an example being N = 1 SQCD with 3/2 N < F < 3N ). There are no
general tools to describe strongly coupled, interacting conformal field theories, for
which, regardless of supersymmetry, typically one cannot easily derive an effective
Lagrangian (recall we are assuming that the UV theory is asymptotically free so we
are excluding the case in which there are enough massless charged matter fields to
make the β function being IR-free to start with, as e.g. N = 1 SQCD with F ≥ 3N ).
Sometime duality can help, like for N = 1 SQCD with N + 1 < F < 3/2 N , whose
dynamics can be described by a dual, IR-free, magnetic theory. But this is clearly
non generic.
Since our aim is to discuss low energy effective actions, in what follows we will
focus on effective theories where scalars, spinors and abelian gauge fields enter,
only. Remarkably, as far as what we want to discuss in this chapter, this is not
such a restriction. Indeed, as we will discuss momentarily, for both N = 2 and
N = 4 supersymmetric theories, a large fraction of the moduli space happens to
enjoy precisely such an abelian, IR-free phase.
A few more comments are in order.
In absence of supersymmetry, one expects the minima of the potential to be
isolated, and hence the space of vacua to be a set of isolated points in the space of
scalar field VEVs (there might exist a classical pseudo-moduli space which, however,
is typically lifted once quantum corrections are taken into account). If this is the
case, while scalar fields VEVs do parametrize the space of vacua, no scalar fields can
be actually massless. Hence, having truly massless scalar fields in the low energy
spectrum, implies the existence of a moduli space of vacua on which the potential
vanishes identically, V = 0 (this includes also the case of spontaneously broken
global symmetries, a scenario which can occur also in non-supersymmetric setups,
albeit in this case the moduli space, parametrized by goldstone bosons, is compact).
Supersymmetric theories typically admit moduli space of supersymmetric vacua.

293
Hence, in the following, we will assume we are in such a situation, and hence we
allow massless scalar fields to be present in the low energy effective action. These
scalar fields, or better their VEVs, parametrize the moduli space.
In writing down the most general form of a IR-free effective action, an important
simplification occurs. Suppose a charged massless field is present in the theory. Due
to one-loop running, the abelian gauge coupling τ under which the massless field is
charged vanishes in the far IR, that is Imτ → ∞. Hence, the abelian gauge field
associated to it is decoupled and does not partecipate to the low energy effective
dynamics. Notice, further, that a charged massless scalar field cannot parametrize
the moduli space. Indeed, a non vanishing VEV would Higgs the U (1) and thereby
give the field a mass, as the gauge field itself. They would both disappear from the
low energy effective action. If, on the contrary, all charged fields are massive, they
do not appear in the low energy effective action to start with. Therefore, in the
limit E → 0 the low energy effective action just contains massless neutral fields and
abelian gauge fields (plus fermions).
To sum up, the low energy effective action would be something like
1 I
L = gij (φ)∂µ φi ∂ µ φj + Im[τIJ (φ)Fµν F J µν ] + fermions , (12.1)
2
where i, j run on (neutral and massless) scalar fields, and I, J on (abelian) gauge
fields. The complexified gauge coupling matrix τIJ and field strength are defined,
respectively, as
θIJ 4πi I I i
τIJ = + 2 , Fµν = Fµν + µνρσ F Iρσ . (12.2)
2π gIJ 2
The σ-model metric gij = gij (φ) is the metric on the moduli space M, whose
coordinates are the massless scalar fields VEVs. Solving the theory boils down to
compute the exact expression of the metric gij and the gauge coupling matrix τIJ
(and of the coefficient functions of fermion kinetic terms which also depend on the
scalar fields φi ).
So far, we have been rather qualitative. In what follows we will show, focusing on
theories with extended supersymmetry, that one can be quantitative and understand
a great deal about actions as (12.1) and their quantum dynamics.

12.1.1 N = 2 effective actions

Let us focus on theories with N = 2 supersymmetry and suppose to start from some
asymptotically free N = 2 renormalizable action. Such an action is fully specified

294
by the gauge group G and matter content, see § 6.1.
If supersymmetry is preserved and a moduli space exists, the lightest excitations
are massless. Hence, for low enough energy, lower then any scale in the theory,
the dynamics on the moduli space is described by an effective action including
these massless fields, only. This action should preserve N = 2 supersymmetry.
Hence, it should be nothing but a special instance of the N = 2 non-linear σ-model
discussed in § 6.1.1. As such, it would be fully determined by knowing the exact
expression of the prepotential F(Φ), which gives both the special Kähler metric and
the generalized complexified gauge coupling, eqs. (6.9) and (6.10), and by knowing
the HyperKähler metric describing the hypermultiplets σ-model.
Scalar fields parametrize a complex manifold which, as discussed in §6.1.1, has
locally the following product structure

M = MV × M H . (12.3)

MV is a special Kähler manifold, whose coordinates are the massless scalar φI

VEVs belonging to vector multiplets, the subspace of M where hyperscalar field
VEVs are held fixed. MH is a HyperKähler manifold, whose coordinates are the
massless scalar H i VEVs belonging to hypermultiplets, the subspace of M where
vector multiplet scalar field VEVs are held fixed (we refer collectively to H1i and H2i
defined in §6.1 as H i in here).
Let us look at the structure of this (classical by now) moduli space more closely.
The first thing is that, in writing (12.3), we have assumed that the σ-model
metric is diagonal, i.e. that there are no kinetic terms mixing φI and H i . That this
is the case comes from a N = 2 selection rule. If a cross term where there in the
Lagrangian, its supersymmetry variation should be canceled (up to total space-time
derivatives) by the supersymmetry variation of some other term. Looking at the
supersymmetry variations of vector- and hypermultiplet component fields, one can
easily see that such a term cannot be constructed. Hence, metric cross terms are
zero.
The subspace MV where only the complex scalars φI get a VEV is called
Coulomb branch. This is because the scalars belonging to N = 2 vector multi-
plets transform in the adjoint of the gauge group G and, as such, can at most
Higgs G down to U (1)n , where n = rank G. More precisely, the scalar potential is
V ∼ Tr [φ, φ† ] and its supersymmetric minima are described by the adjoint scalars

295
VEVs being in the Cartan subalgebra of G
X
n
hφi = aI hI where hI ∈ CSA of G . (12.4)
I=1

For generic aI the gauge group is broken as G → U (1)n . To parametrize the moduli
space one should bare in mind that a set of aI fixes the gauge invariance only up
to the action of the Weyl group WG , the group of residual gauge transformations
which, while acting on φ, do not take it out from the Cartan subalgebra. So,
locally, MV = Cn /WG , meaning that the coordinates of the moduli space should be
invariant under Weyl transformations.
Taking, for example, G = SU (N ), the Weyl group is SN , the group of permu-
tations of N elements. A natural set of U (1)N −1 × SN invariant coordinates on the
(N − 1)-dimensional moduli space CN −1 /SN can be shown to be
X X
u2 = ai aj , u 3 = ai aj ak , . . . , uN = a1 a2 · · · aN , i, j, k = 1, . . . , N
i<j i<j<k
(12.5)
where, in this case, eq. (12.4) is
 1 
a
 .  XN
 
hφi =   with ak = 0 , (12.6)
 .  i=1
aN

since Tr φ = 0. At low energy, the effective Lagrangian describes n = N − 1 massless

N = 2 abelian vector superfields V I . The scalar fields φI belonging to these massless
abelian vector superfields are neutral and the gauge couplings τIJ are frozen at the
value corresponding to the (lightest) massive particles (whose masses are in fact
proportional to aI ). Hence the theory is in a IR-free abelian Coulomb phase. The
(qualitative) behavior of gauge coupling evolution is shown in figure 12.1, where we
refer to aI collectively as a.
The subspace MH , where only the scalars H i get a VEV is called Higgs branch.
This is because, for generic gauge group representations (hyper)scalar field VEVs
break the gauge group completely, now. So, on the Higgs branch, one does not
expect propagating massless gauge degrees of freedom. Here again, the scalars
parametrizing the moduli space MH are not only massless but also neutral (if this
were not the case, they would acquire a mass by Higgs mechanism and should be
integrated out for low enough energy, disappearing from the effective action).

296
 g2

U(1)0 s
<latexit sha1_base64="fvTU85QZ15z/n8b874OJfCw1nJk=">AAAB+HicdVDNSgMxGMz6W+tPVz16CRaxXpasrVhvRS8eK7htoV1KNs22odnskmSFuvRJvHhQxKuP4s23MW1XUNGBwDDzfXyTCRLOlEbow1paXlldWy9sFDe3tndK9u5eS8WpJNQjMY9lJ8CKciaop5nmtJNIiqOA03Ywvpr57TsqFYvFrZ4k1I/wULCQEayN1LdLvQjrURBmXsU9OVbTvl1GTvXCoA6RU0MIuSgnqApdB81RBjmaffu9N4hJGlGhCcdKdV2UaD/DUjPC6bTYSxVNMBnjIe0aKnBElZ/Ng0/hkVEGMIyleULDufp9I8ORUpMoMJOzmOq3NxP/8rqpDut+xkSSairI4lCYcqhjOGsBDpikRPOJIZhIZrJCMsISE226KpoSvn4K/yetU8c9c9BNrdy4zOsogANwCCrABeegAa5BE3iAgBQ8gCfwbN1bj9aL9boYXbLynX3wA9bbJxRHkrc=</latexit>

G
<latexit sha1_base64="2JA3HA+tdmq1Em/nDD+1RDNbnQI=">AAAB8XicdVDLSgMxFL1TX7W+qi7dBIvgasjYinVXdKHLCvaBbSmZNNOGZjJDkhHK0L9w40IRt/6NO//GtB1BRQ8EDufcS849fiy4Nhh/OLml5ZXVtfx6YWNza3unuLvX1FGiKGvQSESq7RPNBJesYbgRrB0rRkJfsJY/vpz5rXumNI/krZnErBeSoeQBp8RY6a4bEjPyg/Rq2i+WsFs+t6gi7FYwxh7OCC4jz8VzlCBDvV987w4imoRMGiqI1h0Px6aXEmU4FWxa6CaaxYSOyZB1LJUkZLqXzhNP0ZFVBiiIlH3SoLn6fSMlodaT0LeTs4T6tzcT//I6iQmqvZTLODFM0sVHQSKQidDsfDTgilEjJpYQqrjNiuiIKEKNLalgS/i6FP1Pmieud+rim0qpdpHVkYcDOIRj8OAManANdWgABQkP8ATPjnYenRfndTGac7KdffgB5+0TCu+RKg==</latexit>

⇤ a µ

Figure 12.1: Schematic behavior of gauge coupling running of N = 2 SYM with

gauge group SU (N ), higgsed to N − 1 U (1)’s at scale a > Λ on the Coulomb
branch. The gauge coupling stops running at a scale µ ' a. The low energy
effective theory is described by abelian vector superfields, containing photons and
neutral fields (scalars and fermions) and enjoys an IR-free Coulomb phase.

Branches where both φI and H i have non-vanishing VEVs are called mixed
branches.
So, all in all, we have to deal with a set of massless, neutral scalar fields and
n = rank G abelian gauge fields (plus fermionic superpartners). So, we are exactly
in a situation as the one advocated in the previous general discussion, see eq. (12.1).
As we have already seen discussing N = 1 theories, the moduli space needs
not to be smooth. There can exist singularities where submanifolds of different
dimensions meet. For example, classically, at the origin of field space, where the
Coulomb and the Higgs branch meet, the theory is fully un-higgsed and the metric
of the moduli space is expected to be singular: extra massless degrees of freedom
appear and they should be included in the low energy effective description. Fig. 12.2
provides a qualitative picture of the N = 2 classical moduli space.
All what we said above is the classical part of the story. How do quantum
corrections change it? Answering this question will be the basic goal of this chapter.
However, already at this stage, a few important facts can be anticipated.
First, the selection rule dictating a direct product for the moduli space M,
eq. (12.3), holds also at the quantum level, since it comes from the supersymmetry

297
 Higgs
Branch

Mixed
Branch

a I = hi = 0
<latexit sha1_base64="0wgChvb8BFPQeJwhcPpQx9YZkOw=">AAAB8HicbVBNSwMxEJ2tX7V+VT16CRbBU9kVRS+Fohe9VbAf0m5LNs22oUl2SbJCWforvHhQxKs/x5v/xrTdg7Y+GHi8N8PMvCDmTBvX/XZyK6tr6xv5zcLW9s7uXnH/oKGjRBFaJxGPVCvAmnImad0ww2krVhSLgNNmMLqZ+s0nqjSL5IMZx9QXeCBZyAg2VnrE3bvKsMsqbq9YcsvuDGiZeBkpQYZar/jV6UckEVQawrHWbc+NjZ9iZRjhdFLoJJrGmIzwgLYtlVhQ7aezgyfoxCp9FEbKljRopv6eSLHQeiwC2ymwGepFbyr+57UTE175KZNxYqgk80VhwpGJ0PR71GeKEsPHlmCimL0VkSFWmBibUcGG4C2+vEwaZ2Xvouzen5eq11kceTiCYzgFDy6hCrdQgzoQEPAMr/DmKOfFeXc+5q05J5s5hD9wPn8AwpyPuQ==</latexit>

Coulomb Branch

Figure 12.2: N = 2 classical moduli space. The mixed branch intersects the Higgs
branch on a Higgs submanifold and the Coulomb branch along a Coulomb subman-
ifold. The more singular point is the origin where the maximal number of degrees
of freedom become massless.

algebra.
Second, N = 2 supersymmetry implies that the special Kähler metric on MV
and the (imaginary part of the) generalized complexified gauge coupling, τIJ , are
related, see §6.1.1. The former is a function of the scalar fields φI only, and so is
the gauge coupling matrix τIJ . The latter undergoes renormalization, at one loop
and non-perturbatively in N = 2, its quantum corrected expression being some
(unknown for the time being) function of the strong coupling scale Λ. Since Λ
appears in τIJ , it appears in the Lagrangian in the same way as a VEV of a scalar
belonging to a vector multiplet (one can think of Λ as a spurion). Since the metric on
MH does not depend on vector multiplet scalars, it does not depend on Λ, either.
Taking the classical limit, Λ → 0, one then concludes that the metric on MH is
classically exact! The upshot is that the metric on the Coulomb branch can receive
quantum corrections, while that on the Higgs branch is classically exact. Therefore,
the exact low energy effective action will be described by a quantum corrected
Coulomb branch and a classically exact Higgs branch. Solving the quantum theory
boils down to determine the geometry on the Coulomb branch. Hence, in what
follows, we will mostly focus on Coulomb branches.
One more property which makes N = 2 special is that, unlike for N = 1 (an ex-

298
ample being SQCD with F < N ), in N = 2 theories a moduli space always survives
at quantum level. In other words, the classical moduli space can be modified, but
never completely lifted. As for MH , this is obvious. The HyperKähler manifold is
classically exact so, if it exists classically, it persists quantum mechanically. A way
to see that this holds also on the Coulomb branch is as follows. For large field VEVs,
aI >> Λ, we can use classical intuition where, by ordinary Higgs mechanism, the
gauge theory is higgsed to U (1)n at weak coupling, see eq. (12.4). The corresponding
n flat directions can be lifted at the quantum level, if given a mass. However, this
cannot occur since in such semi-classical region this can happen only by higgsing,
and abelian vector multiplets are neutral and so are the scalar fields φI , which can-
not then Higgs the theory further. Therefore, we conclude that at large fields VEVs
the moduli space persists even at quantum level. But then, by analytic continuation,
a moduli space persists also in the strongly coupled region, where aI ∼ Λ (complex
manifolds can become singular only on complex submanifolds, whose dimension is
then at least 2 real dimensions smaller, so there is no obstructions against analytic
continuation into a region of strong coupling where classical intuition would fail).
The classical moduli space has singularities of enhanced gauge symmetry and
one could wonder if such singularities survive at the quantum level. One of the
basic results we will show in the following is that the quantum moduli space does
admit singularities where massive particles become massless, but none of them are
gauge fields. So, there are no points of enhanced gauge symmetry, and the theory
is always in an abelian Coulomb phase. What can exist, instead, are other type of
singularities, known as Argyres-Douglas points, where mutually non-local particles,
as monopoles and dyons, become simultaneously massless. At these singularities the
low energy effective dynamics is described by an interacting conformal field theory
(which, however, does not admit a Lagrangian description). We will have more to
say about this later.
To sum up, apart from special points/curves where a Lagrangian description is
not available, the structure of the N = 2 low energy effective action is
I 1
L = KIJ ∂µ φ ∂ µ φJ + I
Im[τIJ (φ)Fµν F Jµν ] + Kij (h, h) ∂µ hi ∂ µ hj + fermions , (12.7)
2
where KJJ is the (special Kähler) metric on the Coulomb branch and Kji the (Hy-
perKähler) metric on the Higgs branch. The complexified gauge coupling is related
to the prepotential as
∂ 2F
τIJ = . (12.8)
∂φI ∂φJ

299
We have slightly changed normalizations with respect to previous chapters. With
present normalizations, eq. (6.12), which relates the special Kähler metric to the
complexified gauge coupling matrix, reads KIJ = Im τIJ . Defining

∂F
φDJ ≡ , (12.9)
∂φJ

we can re-write (12.7) as

  
I µ J 1 I Jµν
L = Im τIJ ∂µ φ ∂ φ + Fµν F + Kij (h, h) ∂µ hi ∂ µ hj + fermions
2
 I
 1
= Im ∂µ φ ∂ µ φDI + Im[τIJ (φ)Fµν I
F Jµν ] + Kij (h, h) ∂µ hi ∂ µ hj + fermions .
2
(12.10)

Solving the theory boils down to determine the quantum exact expression of the
prepotential F and, via eqs. (12.8), the Lagrangian (12.10).
A cartoon of the quantum corrected moduli space is depicted in fig. 12.3.

Higgs
Branch

Mixed
Branch

⇠⇤

Coulomb Branch

Figure 12.3: N = 2 quantum corrected moduli space. The (local) product structure
remains the same as in the classical limit and the Higgs branch is also unmodified.
The same happens for the Higgs directions of the mixed branch, even though they
may be deformed in the Coulomb directions. The Coulomb branch is modified at the
quantum level, instead (but never completely lifted!). Generically, this may excise
(part or all) classical singular submanifolds.

300
12.1.2 N = 4 effective actions

Let us now consider N = 4 supersymmetry. The story here is much simpler. First,
there exist only one class of scalar fields, all transforming in the adjoint representa-
tion of the gauge group G. So, at a generic point of the moduli space, the low energy
dynamics is that of a free U (1)n N = 4 theory, with n = rank G, and the moduli
space M is parametrized by 6n neutral real scalars. Hence, at a generic point on
the moduli space, we are in an (IR-free) abelian Coulomb phase. Moreover, as we
already discussed in § 6.2, the gauge coupling does not run, neither perturbatively
nor non-perturbatively, and then M = R6n . This is, though, not boring at all. As
we will discuss later, N = 4 non-renormalization theorems, which are the strongest
possible, let one get very interesting exact results.

12.2 Monopoles, dyons and electromagnetic duality: a re-

cap
Before proceeding, we need to recall a few properties that some gauge theories enjoy
and discuss how these are realized in supersymmetric contexts.
Let us start from a U (1) gauge theory without matter, namely electro-magnetism.
Maxwell equations in the vacuum, which in differential form notation can be written
as
d ∗ F = 0 , dF = 0 (12.11)
are invariant under the transformation F → ∗F , ∗ F → −F , which corresponds
to the exchange of electric and magnetic fields. This transformation is called S-
duality transformation. In presence of electric sources Maxwell equations can still
be invariant under S duality if one postulates the existence of magnetic sources (aka
monopoles) and the associated current jm , with the following action of S

F → ∗F , ∗ F → −F and je → jm , jm → −je . (12.12)

where now Maxwell equations read

d ∗ F = je , dF = −jm . (12.13)

The exchange of electric and magnetic currents implies, in particular, that under a
S duality transformation electric and magnetic charges are also exchanged.
One crucial consequence of the presence of magnetic monopoles is that the elec-
tric charge is quantized. More precisely, as shown by Dirac, it turns out that a theory

301
with both electric and magnetic charges, q and p respectively, can be consistently
quantized only if the following condition holds

q p = 2πn with n ∈ Z . (12.14)

This is the renown Dirac quantization condition, which implies that any electric
charge is an integer multiple of an elementary charge e ≡ (2π/p) n0 , for some integer
number n0 . Another important consequence of eq. (12.14) is that regimes where the
electric charge is small correspond to regimes where the magnetic charge is large
and viceversa. Therefore, S duality is a strong-weak coupling duality.
Maxwell equations are not affected if adding to the action a θ-term

θ e2
Fµν Feµν . (12.15)
32π 2
However, in presence of magnetic monopoles, a θ-term does have an interesting
physical effect. As shown by Witten, in this case the magnetic charge of a particle
contributes to its electric charge, too. Specifically, a particle with magnetic charge
p = 4π
e
and U (1) electric charge ne e has the following physical charges

4π θe2 θe
p= , q = ne e − 2 p = ne e − . (12.16)
e 8π 2π
In other words, if a θ-term is present a magnetic monopole always carries an electric
charge (even if ne = 0) and such electric charge is not a multiple of some basic unit.
This is known as Witten effect. In the following, with some abuse of language, we
will refer to the U (1) charge e as the electric charge.
Dirac quantization condition is generalized in presence of dyons, which are states
carrying both electric and magnetic charges, as

q1 p2 − q2 p1 = 2πn . (12.17)

This is known as Dirac-Schwinger-Zwanziger quantization condition. An aspect

regarding eq. (12.17) and that will play a relevant rôle later is that only if the right
hand side vanishes the corresponding states are local with respect to each other.
So, for instance, two electrically charged states are local with respect to each other
while an electrically charged state and a magnetic monopole (or a dyon) are not. As
such, they cannot be described within one and the same Lagrangian. Therefore, an
effective low energy theory where such mutually non-local objects are both present,
is believed not to admit a Lagrangian description.

302
 Let us emphasize that the duality transformation (12.12) is not a symmetry of the
theory, since it acts on the couplings. Rather, it maps a description of the theory to
another description of the same theory. There exists another transformation, known
as T -duality transformation, which does not act on the electro-magnetic field but
shifts the θ angle by 2π and, as such, is a symmetry of the theory. These two
transformations, S and T , generate a full group, SL(2, Z) ' Sp(2, Z), the duality
group of electro-magnetism.
As SL(2, Z) 2 × 2 matrices, S and T are
! !
0 −1 1 1
S= , T = (12.18)
1 0 0 1

The way S and T act on τ = θ/2π + 4πi/e2 , τ → −1/τ and τ → τ + 1, respectively,

shows that the group they generate acts on the complexified gauge coupling τ as a
fractional linear transformation
!
aτ + b a b
τ→ where ∈ Sp(2, Z) . (12.19)
cτ + d c d

In Maxwell theory, magnetic monopoles (or dyons) are introduced by hand as

extra degrees of freedom, they are pointlike and carry infinite energy. However,
monopole (and dyon)-like sources arise as solitons, i.e. localized, finite energy, non-
singular solutions of the equations of motion, in the context of spontaneously broken
gauge theories. The first and most famous example is the Giorgi-Glashow model, a
SU (2) gauge theory coupled to a scalar φ transforming in the adjoint of SU (2) and
quartic potential
λ
2
V = Tr φ2 − a2 , (12.20)
4
with a some real number. This theory undergoes a Higgs mechanism which breaks
SU (2) → U (1), and admits soliton solutions carrying monopole and/or dyonic
charges under the low energy effective U (1). For example, there exists a magneti-
cally charged soliton, the ’t Hooft-Polyakov soliton with charges ±(1, 0), as well as
a dyon, found by Julia and Zee, with charges ±(1, −1) (the plus and minus sign
solutions should both be there by anomaly cancellation of the low-energy effective
U (1) theory). For generic values of the parameters (charge g, scalar field VEV a
and quartic coupling λ) these solutions are not known analytically. However, there
exists a limit in which the equations of motion can be solved exactly. This is the
so-called BPS limit, which corresponds to take λ → 0 with g and a fixed while

303
retaining the boundary conditions on the Higgs field, that should tend towards a at
spatial infinity. In this limit, the minimal energy configurations satisfy the following
relation
√ a
M = 2 | (q + ip)| (12.21)
g
where M is the mass of the soliton and q = ne g and p = nm 4π g
its electric and
magnetic charges (the reason for the 4π in place of the 2π for the magnetic charge p
is just because in this theory we could add fields in the fundamental representation
of SU (2), which would carry electric charge ±g/2 and, in terms of such minimal
charge, one would get the usual Dirac quantization condition). In fact, in the BPS
limit all particles in the spectrum, including fundamental degrees of freedom (gauge
bosons and Higgs field), satisfy the mass formula (12.21) and so belong to the BPS
spectrum.
In presence of a θ-term, the analysis that lead to eq. (12.21) can be repeated
almost unchanged, the BPS mass formula becoming now
√ θ 4πi
M= 2|a (ne + τ nm )| where τ = + 2 . (12.22)
2π g
Due to Witten effect, acting with the transformation θ → θ + 2πn on the monopole
and dyon solutions, one can get a full tower of solutions with charges ±(1, −n)
and ±(1, −n − 1), n ∈ Z. Since this transformation is a symmetry of the theory,
these solutions are all physically equivalent. Note that the T -duality transformation
θ → θ + 2π which acts on the charge vector (nm , ne ) as

T : (nm , ne ) → (nm , −nm + ne ) , (12.23)

acts on the complexified gauge coupling as τ → τ + 1. Plugging this into the BPS
mass formula (12.22), we see that the BPS mass formula is left invariant. This is
consistent with the fact that since masses are physical observables, they should be
insensitive to symmetry transformations.
Looking at eq. (12.12), we see that a S transformation, which sends τ → −1/τ ,
should instead act on a charge vector as

S : (nm , ne ) → (−ne , nm ) . (12.24)

If we demand eq. (12.22) to be invariant under S duality, this should then also be
accompanied by the shift a → aτ .

304
 More generally, a matrix A ∈ Sp(2, Z) transforming (a τ, a)T as A · (a τ, a)T ,
should correspond to a change of the vector of electric and magnetic charges as
(nm , ne ) · A−1 . We will re-derive this important result later.
All above analysis is (semi) classical. In particular, the derivation of the BPS
bound and the construction of the monopole and dyon solutions. One might wonder
to what extent this still holds at the full quantum level. This is something difficult
to check in the Giorgi-Glashow model since an analytical handling of the quan-
tum/strong coupling regime is not possible in such a non-supersymmetric setup.
But, as usual, supersymmetry helps.
Let us consider N = 2 pure SYM with gauge group SU (2). Since we are going to
use slightly difference normalizations with respect to previous lectures, let us write
down the on-shell Lagrangian explicitly
1 h 1 θ 2  
eµν + Dµ φDµ φ − 1 φ, φ 2
L = 2 Tr − Fµν F µν + g F µν F
g 4 32π 2 2
√ √ i
−iλσ µ Dµ λ − iψσ µ Dµ ψ + i 2[φ, ψ]λ + i 2[φ, λ]ψ , (12.25)

where, in present normalizations, Dµ = ∂µ −iAaµ Ta . This theory has the same bosonic
content of the Giorgi-Glashow model (including a scalar potential which Higgses the
theory down to U (1) in the vacuum) and shares with it its basic dynamics. As such,
it also admits magnetically charged solitons, as monopoles and dyons.
There are, however, some important differences with respect to the Giorgi-
Glashow model.
First, in the Giorgi-Glashow model the BPS limit is a rather special limit, since
it consists in ignoring the quartic Higgs field potential, just retaining the boundary
conditions on the Higgs field at spatial infinity. This is automatically the set-up
 2
we have in our N = 2 example, since the potential, V ∼ φ, φ , is identically 0
on the moduli space (and it is so whenever φ is in the Cartan subalgebra, that is
φ = (0, 0, a), for any values of a). So here the BPS limit is built in, in a sense.
A more important difference regards the BPS mass formula (12.22). This formula
is reminiscent of the bound that massive states in the N = 2 spectrum should satisfy
and which 1/2 supersymmetry preserving states (short representations) saturate.
This suggests that, in presence of charged solitons in the spectrum, the central
charge may be related to their electric and magnetic charges. Witten and Olive
showed that this is indeed the case. Let us see how this comes about. In present

305
normalizations the N = 2 algebra and the corresponding bound read
√ √
{Q1α , Q2β } = 2 2αβ Z , M ≥ 2 |Z| . (12.26)

Starting from the Lagrangian of pure N = 2 SYM one can compute the correspond-
1 2
R
ing supercurrents Sαµ and Sαµ by Nöether theorem. Recalling that QIα = d3 x SαI 0
one finds (after dropping contributions which trivially vanish at spatial infinity) that
√ Z h  i
2 2 √
1 2
{Qα , Qβ } = αβ d 3
x ∂i F a0i
− i e
F a0i
φa ≡ 2 2αβ Z
g2
√ Z h  i
1 2 2 2 √
{Qα̇ , Qβ̇ } =  d 3
x ∂ i
−F a0i + i e
F a0i φ
a
≡ 2 2α̇β̇ Z ∗ . (12.27)
2 α̇β̇
g
Since the electric and magnetic charges of the U (1) low energy effective theory are
Z Z   4π
1
1 a0i a
q=− 3 0i a
d x ∂i Fa φ = gne , p = − 3 e
d x ∂i F φ = nm (12.28)
ag ag g
one finally finds (after taking into account the effect of a non-trivial θ-term) that

Re Z = ane , Im Z = aτ nm (12.29)

and hence eq. (12.22). So we learn that in presence of monopoles (and dyons) the
supersymmetry algebra must be modified with the addition of a non-trivial central
charge (which measures the electric and magnetic charges of soliton solutions). From
a geometric viewpoint this should not come as a surprise. Indeed, supersymmetry
charges are space integrals. In calculating their anticommutators one has to deal
with surface terms, which one usually neglects. However, as shown in eqs. (12.28),
in presence of electric and magnetic charges these surface terms are non-zero and
give rise to a non-vanishing Z.
Note that, unlike the Giorgi-Glashow model, here the relation between masses
and charges of BPS states, eq. (12.22), does not come from a (semi) classical analysis
but is dictated by the supersymmetry algebra. Hence, it cannot be spoiled quantum
mechanically and should remain valid even when perturbative and non-perturbative
corrections are taken into account. Indeed, BPS states are in short representations
and quantum corrections cannot generate the extra degrees of freedom needed to
convert a short multiplet in a long one. So, BPS saturated states remain so also at
the quantum level.
That eq. (12.22) persists quantum mechanically, does not mean that the quanti-
ties therein do not undergo renormalization. For one thing, in N = 2 we know that

306
the gauge coupling τ runs (at one loop and non perturbatively). Therefore, upon
taking into account renormalization effects, while by its very definition the central
charge Z is still a linear combination of conserved (electric and magnetic) abelian
charges, the coefficients multiplying nm and ne will be replaced by some (holomor-
phic) functions, which we dub a and aD , of the strong coupling scale Λ and field
VEVs
Z = a ne + aD nm . (12.30)
In the classical limit a is the VEV of the scalar field φ and aD = aτ . But one expects
this not to be true at the full quantum level. In particular, the expression of aD
in terms of a could be different. Seiberg and Witten proposed the following exact
relation between a and aD
∂F daD
aD ≡ that is τ = , (12.31)
∂a da
with F the prepotential. We will provide evidence for the proposal (12.31) later.
Here, just notice that this way, in the semi-classical limit, where Fcl = 21 τ a2 ,
eq. (12.30) correctly reduces to (12.29). But, unlike (12.29), it is by construction
renormalization group invariant.
One of our main goals in the following will be to check the proposal (12.31) and
compute the exact expression of a and aD in terms of the scalar field VEVs and
Λ. Given this information, the masses of all BPS states (fundamental fields as well
as magnetic monopoles and dyons) will be known exactly in terms of the moduli
parameters. More importantly, finding the exact expressions of a and aD amounts
to find the exact expression for τ and hence, by (12.10), the full effective action!
This discussion can be repeated for N = 4 SYM, which is also expected to
admit charged solitons in its spectrum. There, however, the relation aD = τ a is not
renormalized, since in this case τ is classically exact, as so is the moduli space. This
has important consequences which we will come back to, when discussing quantum
properties of N = 4 SYM.
So far, we have been considering pure N = 2 SU (2) SYM. One may want to
add matter fields, i.e. hypermultiplets. This amounts to add to the Lagrangian the
superpotential term
F 
X 
√
2H1i ΦH2i + mi H1i H2i + h.c. . (12.32)
i=1

For equal masses, the theory has a SU (F ) flavor symmetry, which is broken to
U (1)F for generic values of mi . One can repeat previous computations, calculate

307
the contribution of H1 and H2 to the supercurrent and, in turn, to the central charge
mass formula (12.30). The end result is
XF
1
Z = a ne + aD nm + √ mi Si , (12.33)
i=1
2

where Si are global conserved U (1) charges under which H1i and H2i have charges
+1 and −1, respectively.
There exist generalizations of this story. The Giorgi-Glashow model can be gen-
eralized to a gauge theory with gauge group G spontaneously broken to a subgroup
H by some Higgs-like field transforming in the adjoint representation of G. Thanks
to the topological nature of soliton solutions, it turns out that an analysis on their
existence can be carried out in the context of homotopy theory. In particular, in-
equivalent solutions are classified by the homotopy group Π2 (G/H). This is isomor-
phic to Π1 (H)G , the subgroup of closed paths in Π1 (H) which can be contracted to
a point when H is embedded in G. If G is simply connected, Π1 (H)G is isomorphic
to Π1 (H) and non-trivial soliton solutions are hence classified by Π1 (H). For exam-
ple, in the original Giorgi-Glashow model we have Π2 (G/H) = Π2 (SU (2)/U (1)) =
Π1 (U (1)) = Z, and one family of magnetic monopoles with integer charge is indeed
present. The same happens in GUT theories. Taking, e.g., GGU T = SU (5) one has
Π2 (G/H) = Π2 (SU (5)/SU (3) × SU (2) × U (1)) = Π1 (SU (3) × SU (2) × U (1)) =
Π1 (U (1)) = Z, so again magnetic monopoles are expected to exist. This is not the
case in the Standard Model, where the gauge group is not simple, G = SU (2)×U (1)Y
and, more importantly, Π1 (H)G = 0. Indeed, the generator of the unbroken elec-
tromagnetic U (1) gets contribution both from the generator of U (1)Y and of the
Cartan of SU (2). This implies that any closed path in U (1) may be deformed to
lie completely in U (1)Y , which, unless this path is trivial, cannot be deformed to
a point in G. This means that Π1 (H)G = 0 and hence magnetic monopoles do
not exist. Interestingly - recall the discussion around eq. (12.14) - that the electric
charge happens to be quantized can be seen as an evidence in favor of the existence
of monopoles and, in turn, of GUT theories.
Exactly as for the original Giorgi-Glashow model, this more general story finds
a natural embedding in supersymmetric contexts. One such situation is nothing
but the low energy effective theories describing the N = 2 (and N = 4) Coulomb
branch we are actually interested in. There, we have a gauge theory with gauge
group G broken to its Cartan subalgebra H = U (1)n , where n = rank G. From
previous general analysis, it follows that magnetically charged solitons are present

308
in the spectrum, in general. Most of what we said above holds unchanged. In par-
ticular, the IR-free effective theory is form-invariant under electro-magnetic duality
transformations which are the natural generalization to n > 1 of eq. (12.19), and
act on the couplings as



J −1
τIJ → ALI τLM + BIM C JN τN M + DM (12.34)
 
where now M ≡ CA DB ∈ Sp(2n, Z). The vector of electric and magnetic charges
is now a 2n-component row vector (nIm , neJ ). The corresponding BPS mass formula
which generalizes (12.30) is

Z = aI neI + aD,I nIm = a · ne + aD · nm , (12.35)

where, in the second step, matrix multiplication is understood and (12.31) is now
aDI ≡ ∂F/∂aI . Finally, the addition of (massive) flavors changes the central charge
formula in a way similar to eq. (12.33).
Let us conclude this section with a comment which will be relevant later. As
already noticed, electro-magnetic duality transformations are not symmetries of the
theory. They just express the equivalence of abelian theories coupled to massive
sources under a Sp(2n, Z) redefinitions of electric and magnetic charges. It is a
redundancy of the effective Lagrangian description. The point, though, is that this
redundancy can capture important features of the theory, when a moduli space is
present (which is the case we will actually be concerned with). That this is the case,
can be seen as follows.
Suppose there is a moduli space of vacua and that the effective dynamics on
this moduli space is described by n abelian gauge fields and a bunch of massless,
neutral scalars, collectively dubbed φ, which parametrize M. Upon traversing a
closed loop in M the physics must be the same at the beginning and at the end
of the loop. However, the Lagrangian does not need to: it can just be invariant
modulo a electro-magnetic duality transformation. Geometrically, this corresponds
to say that the matrix of couplings τIJ is a section of a Sp(2n, Z) bundle. In matrix
notation, this means that upon making a circle in the φ moduli space, the matrix τ
should transform as

τ (e2πi φ) = (A · τ (φ) + B) (C · τ (φ) + D)−1 . (12.36)

 
The element M = CA DB ∈ Sp(2n, Z) is called monodromy around the loop. If the
closed loop does not encircle any singularity, the monodromy is the identity element
of Sp(2n, Z). If it does, the monodromy is a non-trivial element of Sp(2n, Z), instead.

309
 As we discussed at length, singularities on the moduli space are associated to
massive particles becoming massless, there. The monodromy matrix tells about the
nature of such particle. Since masses are physical observables, the BPS mass formula
(12.35) should be invariant under monodromies. Hence, as already emphasized for
the case n = 1, a A ∈ Sp(2n, Z) transforming (aD , a)T as A · (aD , a)T , should
correspond to a change of the vector of electric and magnetic charges as (nm , ne )·A−1 .
This means that, in general, the action of the monodromy changes the quantum
numbers of charged states.
Now, the state of vanishing mass at a given singularity on the moduli space,
should be invariant under the action of the monodromy associated to the singularity
itself (it is the properties of such massless state which determine the monodromy
matrix). That is to say, it should be a left eigenvector of the matrix A with unit
eigenvalue. It is easy to see that the Sp(2n, Z) matrix for which the vector (nm , ne )
is an eigenvector with unit eigenvalue is
!
1 + 2nm ne 2n2e
A(nm , ne ) = . (12.37)
−2n2m 1 − 2nm ne

This means that at a singularity with monodromy matrix of the form above, a state
with charges ±(nm , ne ) becomes massless.
Actually, any state with charges l(nm , ne ) with l ∈ Z is left invariant by the
action of (12.37). However, stable dyons require lnm and lne to be relatively prime,
which is the case only for l = ±1. One way to see it is the following. Suppose
√
to start from a BPS saturated state with charges (Nm , Ne ) and mass M = 2|Z|.
Such state can decay into states whose sum of masses should be less or equal M .
√
For each of these states we have Zi = a · nie + aD · nim and Mi ≥ 2|Zi |. Since
P
charge conservation implies that Z = Zi , it follows from triangle inequality that
P
|Z| ≤ i |Zi | which in turn implies that
X
M≤ Mi . (12.38)
i

In order for the decay to occur the above bound should be saturated, which implies
P
|Z| = i |Zi | (so also the states with charges (nim , nie ) should be BPS). This can
happen if and only if the vectors (Nm , Ne ) and (nim , nie ) are proportional, that is if
Nm and Ne are not relatively prime, (Nm , Ne ) = l(nm , ne ). If they are, instead, the
decay cannot occur.

310
12.3 Seiberg-Witten theory
Let us now come back to our original problem. We would like to look at asymp-
totically free N = 2 gauge theories and try to see what can we say about their low
energy effective dynamics. As advertised, we will focus on the Coulomb branch,
which is the only part of the moduli space which is modified at the quantum level.
What this boils down to is to determine the exact expression of the prepotential
F, more specifically of the generalized complexified gauge coupling matrix, whose
imaginary part is the metric on the Coulomb branch.
Our starting point is some UV-free N = 2 matter-coupled Lagrangian. This
means that if, say, the gauge group is SU (N ) and matter multiplets transform in
the (anti)fundamental representation of SU (N ), we must require that F < 2N , since
the one-loop coefficient (which captures the full perturbative expression for the β
function) is proportional to 2N − F , in this case.
One thing which will play an important role later is the R-symmetry breaking
pattern. Let us first focus on pure SYM. Besides a compact component, SU (2)R ,
under which all bosons in the N = 2 vector multiplet are singlets and the two
gaugini transform as a doublet, there is also a U (1)R symmetry, under which (both)
gaugini have R(λ, ψ) = 1. This symmetry is anomalous and, following the same
discussion we had for N = 1 SYM, one can see that it gets broken as

U (1)R −→ Z4 T (Adj) , (12.39)

since now he anomaly coefficient is now A = 2 T (Adj).

The fact that R(ψ) = 1 implies that the adjoint scalars φ have R(φ) = 2,
meaning that on the Coulomb branch the residual symmetry gets further broken.
For example, we will see later that for G = SU (2) Coulomb branch vacua preserve
a Z4 subgroup of the full Z8 and, therefore, each point on the Coulomb branch is
paired with its mirror under the residual Z2 , which acts non-trivially on M. For
G = SU (3) a Z2 subgroup survives, only, while for higher ranks the U (1)R is fully
broken.
Interestingly, unlike N = 1, the addition of matter does not restore an anomaly-
free U (1)R symmetry, in general. Indeed, given that R(φ) = 2, from the cubic
superpotential term ∼ H1 ΦH2 one sees that the hyperscalars are neutral. Hence,
their fermionic partners ψ1 and ψ2 have R(ψ1 , ψ2 ) = −1. So, if the hypermultiplets
transform in the representation r of the gauge group G, the U (1)R is broken at the

311
quantum level as
U (1)R −→ Z4T (Adj)−4T (r) , (12.40)
if adding one hypermultiplet. For example, taking G = SU (N ) and F hypers in the
fundamental representation, one gets U (1)R → Z4N −2F . Note that for F = 2N the
R-symmetry is not anomalous, in agreement with the vanishing of the β function
and the supposedly conserved R-charge in superconformal field theories.

Let us start considering pure SYM and take, for definiteness, the gauge group to
be G = SU (N ). Following our general discussion in section § 12.1.1 the low energy
Coulomb branch effective (bosonic) Lagrangian looks like
I 1
L = Im(∂µ φ ∂ µ φDI ) + I
Im[τIJ (φ)Fµν F Jµν ] (12.41)
2
where I = 1, 2, . . . , N − 1. Solving the theory amounts to find the exact expression
for the prepotential F or, which is the same, for the effective abelian gauge coupling
matrix τIJ , as a function of Λ and of (gauge-invariant combination of) scalar field
VEVs. Recall that τIJ gets one-loop and non-perturbative corrections, only, and
reads (we refer collectively to a as the common VEV of all scalar fields φI )
∞  b1 n
2N Λ X Λ
τIJ (a, Λ) = CIJ log + dIJ,n , (12.42)
2πi a n=1 a

where 2N is the one-loop β-function coefficient, CIJ is the Cartan matrix of G and
dIJ,n ’s weight n-instanton corrections. Since the model is Higgsed at a scale a, which
can be taken arbitrarily large, these instanton effects can be made arbitrarily small
and are calculable. So, in principle, one could compute τIJ , and hence solve the
low energy effective theory exactly, by evaluating all instanton contributions. In
practice, this is hard. Seiberg and Witten came-up with a more physical approach
to determine τIJ , which is the one we will follow. We start analyzing the simplest
case, N = 2 SYM with gauge group SU (2).

12.3.1 N = 2 SU (2) pure SYM

N = 2 SYM with G = SU (2) admits a one-dimensional moduli space in which the

gauge group is broken to U (1). The gauge invariant coordinate on the (classical)
moduli space can be chosen to be
1
u = htr φ2 i = a2 , (12.43)
2
312
where hφi = aσ3 is the (adjoint) scalar field VEV. The u here corresponds to what
we called u2 in eq. (12.5).
The above formula is valid classically. Quantum corrections may change the
relation between u and a. In what follows, we will keep on calling u the coordinate
on the quantum moduli space but the above equation will be modified as
1
u = htr φ2 i = a2 + quantum corrections . (12.44)
2
√
While classically a = u, quantum mechanically one could expect a more general
√
relation, a = a(u), which only in the classical limit reduces to a = u.
The abelian low energy effective Lagrangian is
  
1 µν µ
L = Im τ (φ) Fµν F + ∂µ φ ∂ φ , (12.45)
2
and it is univocally determined knowing the exact expression of τ and so of the
prepotential F, since τ = ∂ 2 F/∂φ∂φ. Our goal, in the following, will then be to
find the exact expression for F and hence of the analogous of (12.42) which in this
case becomes  4 X ∞  4 n
1 Λ Λ
τ (a, Λ) = log 4
+ dn . (12.46)
2πi a n=1
a4
The first thing one should readily notice is that F cannot be a holomorphic function
of a all along M; it should be multivalued. Indeed, if this were not the case,
Im τ (a) = Im ∂ 2 F(a)/∂φ∂φ would be a harmonic function. As such, it could not
be positive definite everywhere (unless it were a constant). Hence, there would
necessary be regions in the moduli space where Im τ (a) would be negative, making
the effective gauge coupling squared g 2 being negative, too. This would correspond
to propagation of negative norm states, that cannot be. We need Im τ > 0. The way
out is to allow for different local descriptions which requires F(a) to be defined only
locally, say in a neighborhood of the classical region u → ∞. In regions where τ (a)
approaches zero, we need a different (but equivalent) description of the theory. Note
that this corresponds to regions where the gauge coupling is very large, eventually
infinite, so strongly coupled regimes. Geometrically, the moduli space should admit
singularities and in the vicinity of such singularities we expect a different coordinate
patch with respect to a (in other words, a is not a ”good coordinate” on the whole
moduli space M).
In order to understand how different local descriptions can emerge, we have
to understand how electric-magnetic duality is realized in the low energy effective

313
theory. The action (12.45) can be re-written as
!† ! !
1 1 φD φD 0 i
L = Im [τ (φ)Fµν F µν ] + ∂µ J ∂µ where J = , (12.47)
2 2 φ φ −i 0

where φD ≡ ∂F/∂φ. The scalar kinetic term is invariant under Sp(2, R) transfor-
mations acting on φD and φ as
! !
φD φD
→ M where M † JM = J . (12.48)
φ φ

This is the continuum version of the duality group of electro-magnetism previously

defined and it is generated by
! !
0 −1 1 b
S= , Tb = where b ∈ R . (12.49)
1 0 0 1

To see how the duality group acts on the Maxwell term we should first write the
Lagrangian introducing a Lagrange multiplier field ADµ , so to have both the equation
of motion and the Bianchi identity emerging as field equations (this is needed, since
S duality is a non-local transformation in the electromagnetic fields). Recalling that
Fµν = Fµν + iFeµν , we can write the Maxwell term action as
Z   2  Z
1
S = Im τ (φ) Fµν + iFeµν + ADµ ∂ν Feµν . (12.50)
2

where ADµ should be treated as an independent field, with field strength FDµν =
∂µ ADν − ∂ν ADµ .
In doing the path integral one should now integrate over F and AD . Path
integrating over AD we get the equation ∂ν Feµν = 0, namely dF = 0, which implies
F = dA and we are left with the original path integral. But one can path integrate
over F , first. Integrating by parts the second term in eq. (12.50) becomes 2FDµν Feµν
and completing the square one gets
"  #
Z
1  1  2 1 1  2
S = Im τ (φ) Fµν + iFeµν + FDµν + iFeDµν − FDµν + iFeDµν ,
2 τ (φ) τ (φ) 2
   (12.51)
where we used the identity Fµν + iFeµν FD + iFeD = 2i FDµν Feµν , which holds
µν µν

because Fµν FDµν = Feµν FeDµν . Performing the Gaussian path integral over F one

314
gets back the original action (up to an overall normalization) but in terms of the
dual gauge field ADµ as
Z  2 
1 1 e
S = Im − FDµν + iFDµν . (12.52)
τ (φ) 2
So we see that, as expected, the effect of a S-duality transformation which transforms
the gauge coupling as τ → τD = −1/τ , is to replace a gauge field, to which electric
sources couple locally, by a dual gauge field, to which magnetic sources couple
locally. The other generator of Sp(2, R), Tb , does not act on the gauge field but on
the coupling, only, shifting the θ angle. In order for it not to change the physics,
one should take b ∈ Z, hence obtaining the actual electromagnetic duality group,
which is Sp(2, Z). We will henceforth call T the generator T1 , to be consistent with
conventions in § 12.2.
In our previous discussion, we argued that whenever Im τ (a) approaches 0, a dif-
ferent description of the (same) physics should hold. The above discussion suggests
what that can be: an S-dual description in terms of a magnetic dual gauge field
ADµ , with τ → τD = −1/τ and φ and φD exchanged, see eq. (12.48).
Note, in passing, that the way the duality group acts on (φD , φ), and in turn
on (aD , a), eq. (12.48), provides further evidence for the BPS mass formula (12.30)
and the proposed relation (12.31). To see this, let us couple the low energy effective
theory to a charged hypermultiplet with charge ne . Its coupling to the adjoint chiral
superfield Φ is fixed by N = 2 supersymmetry to be
√
2 ne H1 ΦH2 . (12.53)

On the moduli space this induces a mass for the (BPS!) hypermultiplet whose cor-
responding central charge would then be Z = ne a. By a S-duality transforma-
tion, which acts on the adjoint scalar as eq. (12.48), it is clear that for a magnetic
monopole with magnetic charge nm we would have Z = nm aD (with aD = ∂F/∂a)
and, for dyons, the more general formula (12.30).

Singularities and monodromies

We now have all ingredients to understand the singularity structure of the moduli
space and the physical meaning of such singularities.
Let us start looking at the (semi)classical region, namely u → ∞. There one
can safely use the classical relation u = a2 and the one-loop expression for the

315
prepotential
i 2 a2
F1−loop = a log 2 . (12.54)
2π Λ
From this expression we can compute aD which is
 
∂F i a2
aD = = a log 2 + 1 . (12.55)
∂a π Λ
Let us take a counterclockwise contour in the u plane, say u → e2πi u, with very large
|u|. Since in such semiclassical region u = a2 we see that a transforms as a → −a.
For aD , instead, using (12.55), we get
 
i e2πi a2
aD → (−a) log + 1 = −aD + 2a . (12.56)
π Λ2
So, there is a non-trivial monodromy, which acts on the vector (aD , a) as
! ! !
aD aD −1 2
→ M∞ where M∞ = (12.57)
a a 0 −1

Note that, consistently with previous general discussion, M∞ ∈ Sp(2, Z). More
specifically, M∞ = −T −2 , with T the generator previously defined.
The log term in aD and the non-trivial monodromy show that a and aD are
multivalued functions: there is a branch cut extending from infinity, due to the
log term in the one-loop running. Given the singularity at u = ∞, there must be
singularities also somewhere else on the u plane, with their associated monodromies
Mi .
Since, as shown in Figure 12.4, a contour circling around infinity can be deformed
(it is topologically equivalent) to a contour circling around all other singularities,
say we have k of them, the following consistency relation should hold, in general

M∞ = M1 M2 ... Mk . (12.58)

Now, how many singularities, besides that at u = ∞, do we have on the u plane?

The R-symmetry breaking pattern helps, here. As already discussed, the U (1) R-
symmetry of the original theory is anomalous and broken to Z8 at the quantum
level. Since φ has R-charge 2, on the moduli space, parametrized by u ∼ hTr φ2 i,
this is further broken to Z4 . The residual Z2 symmetry acting on the moduli space
changes u as u → −u. Therefore, singularities should come in pairs on the moduli
space, but at the fixed points of the Z2 action u = ∞, 0. We conclude that if we
had one only more singularity beside the one at infinity, this should be at u = 0.

316
 u 1
P
<latexit sha1_base64="MyQmLRAee2ZdjfbUPnoaE6lqxdY=">AAAB73icbVDLSgNBEOyNrxhfUY9eBoPgKeyKosegF48RzAOSJcxOZpMh81hnZsWw5Ce8eFDEq7/jzb9xkuxBEwsaiqpuuruihDNjff/bK6ysrq1vFDdLW9s7u3vl/YOmUakmtEEUV7odYUM5k7RhmeW0nWiKRcRpKxrdTP3WI9WGKXlvxwkNBR5IFjOCrZPaXRGpp6w+6ZUrftWfAS2TICcVyFHvlb+6fUVSQaUlHBvTCfzEhhnWlhFOJ6VuamiCyQgPaMdRiQU1YTa7d4JOnNJHsdKupEUz9fdEhoUxYxG5ToHt0Cx6U/E/r5Pa+CrMmExSSyWZL4pTjqxC0+dRn2lKLB87golm7lZEhlhjYl1EJRdCsPjyMmmeVYOLqn93Xqld53EU4QiO4RQCuIQa3EIdGkCAwzO8wpv34L14797HvLXg5TOH8Afe5w9UOZAo</latexit>

<latexit sha1_base64="4n2vLQ4sEng6tQms8k1skrUvtdA=">AAAB9HicbVBNS8NAEN3Ur1q/qh69LBbBU0lE0WPRi8cK9gOaUCbbTbt0dxN3N4UQ+ju8eFDEqz/Gm//GbZuDtj4YeLw3w8y8MOFMG9f9dkpr6xubW+Xtys7u3v5B9fCoreNUEdoiMY9VNwRNOZO0ZZjhtJsoCiLktBOO72Z+Z0KVZrF8NFlCAwFDySJGwFgp8IcgBPR9JiOT9as1t+7OgVeJV5AaKtDsV7/8QUxSQaUhHLTueW5ighyUYYTTacVPNU2AjGFIe5ZKEFQH+fzoKT6zygBHsbIlDZ6rvydyEFpnIrSdAsxIL3sz8T+vl5roJsiZTFJDJVksilKOTYxnCeABU5QYnlkCRDF7KyYjUECMzaliQ/CWX14l7Yu6d1V3Hy5rjdsijjI6QafoHHnoGjXQPWqiFiLoCT2jV/TmTJwX5935WLSWnGLmGP2B8/kDBySSRA==</latexit>

> 1
>

>
…..
<latexit sha1_base64="Xqr3sWiiQnj+BsUDW2xkRvet8js=">AAAB73icbVA9SwNBEJ2LXzF+RS1tFoNgFe5E0TJoYxnBxEByhLnNXrJkd+/c3RNCyJ+wsVDE1r9j579xk1yhiQ8GHu/NMDMvSgU31ve/vcLK6tr6RnGztLW9s7tX3j9omiTTlDVoIhLditAwwRVrWG4Fa6WaoYwEe4iGN1P/4YlpwxN1b0cpCyX2FY85ReukVqePUmJ32C1X/Ko/A1kmQU4qkKPeLX91egnNJFOWCjSmHfipDceoLaeCTUqdzLAU6RD7rO2oQslMOJ7dOyEnTumRONGulCUz9ffEGKUxIxm5Tol2YBa9qfif185sfBWOuUozyxSdL4ozQWxCps+THteMWjFyBKnm7lZCB6iRWhdRyYUQLL68TJpn1eCi6t+dV2rXeRxFOIJjOIUALqEGt1CHBlAQ8Ayv8OY9ei/eu/cxby14+cwh/IH3+QMJfo/3</latexit>

>
1

Figure 12.4: The equivalence of a countour around infinity and one circling all
singularities on the u-plane. P is a base point for loops γ∞ and γi , i = 1, 2, . . . , k.

But this cannot be. If there were just one singularity at u = 0, because of
(12.58) we would have M0 = M∞ . But then, since a2 is left invariant by M∞ ,
u = a2 would be a good global coordinate on the full moduli space, not just in the
classical region. Then, F(a) would be a holomorphic function of a and so Im τ (a)
a harmonic function. But then, the latter could not be positive definite (unless it
were a constant, which we know it is not).
So, we conclude that there must be at least two singularities besides that at
infinity, located at, say u = ±u0 . If this is this case, u = 0, which is a singular
point on the classical moduli space, will not be a singular point anymore in the
quantum theory. In order to have a singularity at u = 0 one should have at least
three singularities on M. As we will see, this cannot be either. Having two (and
only two) singularities on M, located at u = ±u0 , seems to be the only consistent
possibility.
A natural question to ask is what the nature of the particles becoming massless
at u = ±u0 is. Interestingly, unlike to what happens classically, singularities in the
quantum moduli space are not associated to extended gauge symmetry, namely to
extra massless gauge bosons. This can be understood as follows. If an interacting
non-abelian Coulomb phase were there in the IR, a conserved R-symmetry should be
present (the superconformal R-symmetry). As we have already discussed, singulari-
ties on M occur at u 6= 0. Hence, if conformal invariance should be preserved, then
the dimension of u at the singularity should be zero. In a SCFT the dimension of an
operator is proportional to its R-charge, which for the operator u is R(u) = 4 since
R(φ) = 2. Therefore, at the singularity the operator u would have non-vanishing

317
scaling dimension and a VEV would break conformal invariance. This suggests that
a non-abelian Coulomb phase cannot emerge in the IR. The extra massless degrees
of freedom cannot be gauge bosons.
The analysis of the previous section, suggests what the other possibility could be.
The only other states in the spectrum (at least that we know ) are monopoles and
dyons. For example, magnetic monopoles are very heavy at weak coupling, because
of the BPS mass formula (12.30) and tend to become light at strong coupling. So it
might very well be that these are the states becoming massless at the strong coupling
singularities. Note that if this is the case, by the reasoning in the previous paragraph,
we conclude that they cannot sit in vector multiplets (which would include spin 1
particles). So they should correspond to hypermultiplets. Indeed, in the N = 2
version of the Giorgi-Glashow model, this was explicitly shown to be the case!
As a corollary, one would expect that the singularity at a = 0 of the classi-
cal moduli space, where extra massless gauge bosons did become massless, should
disappear at quantum level. From the exact expression we will eventually get for
a = a(u), we will see that this is indeed the case: the point a = 0 does not belong
to the moduli space, at quantum level (similarly to what happens for N = 1 SQCD
with F = N ).
Let us then focus on the strong coupling singularities at u = ±u0 . Note that
u0 should be proportional to Λ2 , since in the classical limit, Λ → 0, one should
recover the (only one) singularity at u = 0. Hence, from now on, without loss of
generality, we will take u0 = Λ2 . The structure of the moduli space, with punctures
and associated monodromies, is depicted in figure 12.5.
To find the structure of the monodromy matrices MΛ2 and M−Λ2 notice that they
should have a form like (12.37) in terms of the (integer) electric and magnetic charges
(nm , ne ), (n0m , n0e ) of the corresponding massless states. Imposing the consistency
relation M∞ = MΛ2 M−Λ2 and using (12.57) one finds that the unique solution
(modulo physically equivalent solutions, see the comment after eq. (12.22)) is

(nm , ne ) = ±(1, 0) , (n0m , n0e ) = ±(1, −1) (12.59)

corresponding to monodromy matrices

! !
1 0 −1 2
MΛ2 = , M−Λ2 = . (12.60)
−2 1 −2 3
So, we finally see what the nature of the singularities is: at u = Λ2 a monopole with
charge ±(1, 0) becomes massless and at u = −Λ2 a dyon with charge ±(1, −1) does.

318
 1
u >
3

2 > > 1
2 2
⇤x x ⇤

Figure 12.5: The u plane with the three singularities at ∞, Λ2 , −Λ2 . The mon-
odromies associated to the three cycles γi must satisfy the consistency relation
M∞ = MΛ2 M−Λ2 .

Note, in passing, that MΛ2 = ST 2 S −1 and M−Λ2 = T ST 2 S −1 T −1 = T MΛ2 T −1 ,

which nicely agrees with the fact that the residual Z2 symmetry connecting u = Λ2
and u = −Λ2 shift the θ angle by 2π.
One might wonder if there can be more than two singularities on M. For this
to be the case, one should be able to solve an equation like (12.58) with k > 2,
with M∞ given by (12.57) and the Mk ’s having a structure as (12.37) with integers
(nim , nie ). While a general proof is not available, one can explicitly show, for not too
large values of k, that there are no solutions for k > 2. Further evidence suggesting
that k = 2 is the correct answer will be provided shortly.

Seiberg-Witten curve

Given the knowledge of the singularity structure of the moduli space and its mon-
odromies, we want now to construct holomorphic functions a = a(u) and aD = aD (u)
satisfying the monodromies (12.57) and (12.60), and from them obtain the exact
expression for the complexified gauge coupling τ (u). A holomorphic function is uni-
vocally determined by its singularities. Therefore, if we are able to find a function
with the correct monodromies around u = ∞, Λ2 , −Λ2 , we can be sure we get the
correct answer. In principle, this can be done, following the so-called differential
equation approach, but we will follow a different, more geometric pattern. This
was the approach originally pursued by Seiberg and Witten and also the one which
makes it easier and more natural to understand generalizations to richer theories

319
(more general gauge groups and matter content).
The crucial observation comes from the property that we have learned τ should
have: a complex quantity with positive definite imaginary part and on which the
group Sp(2, Z) acts as a fractional linear transformation, eq. (12.19). Such quantities
are fundamental in the theory of Riemann surfaces, where they describe their moduli,
the positivity condition ensuring regularity of the surface. In the case at hand,
the relevant Riemann surface is just a torus, or equivalently, using the language of
algebraic geometry, an elliptic curve. This curve can be written as a complex surface

y 2 = (x − Λ2 )(x + Λ2 )(x − u) , (12.61)

where u parametrizes the modulus τ of the torus and x and y are complex coordi-
nates. Varying u we vary τ and hence eq. (12.61) describes a family of tori. If we
associate to any point on the Coulomb branch parametrized by the complex quan-
tity u a holomorphic varying torus, its modulus will have the same properties we
expect for the complexified gauge coupling τ : a holomorphic section of a Sp(2, Z)
bundle with positive imaginary part, Im τ > 0.
A way to understand that (12.61) describes a torus is as follows. From eq. (12.61)
we see that y is the square root of a polynomial in x so we can look at the x-plane
consisting into two sheets with branch points at Λ2 , −Λ2 , u and ∞, gluing along
the branch cuts (the two sheets corresponding to the ±y branches). One can take
one branch cut between −Λ2 and Λ2 and the second one between u and ∞. The
two sheets can be thought as spheres and the branch cuts as tubes connecting them.
Topologically, this is a torus, see figure 12.6.
On a torus the are two independent, non-trivial homology one-cycles, the A and
the B cycles, which we can take as in the figure. Degenerate tori (that is tori where
some cycles shrink to zero size) occur when any two zero’s of eq. (12.61) coincide.
In other words, when one of the branch cuts disappears. In particular, for u = Λ2
the B cycle shrinks to zero size, for u = ∞ the A cycle shrinks to zero size and for
u = −Λ2 a linear combination of the two, A + B, does.
The basis of one-cycles is not unique, but defined up to Sp(2, Z) transformations
which act as ! !
B B
→M where M ∈ Sp(2, Z) . (12.62)
A A
The modulus of the torus, τ (u), corresponds to the ratio of the periods ω and ωD
ωD
τ (u) = , (12.63)
ω
320
 x B
<latexit sha1_base64="w98e/zX9utWXK541Q7Vnj9kXmJU=">AAAB+XicdVDNS8MwHE39nPOr6tFLcAheLKmbOG9jXjxOcB+wlZFm6RaWpiVJB6XsP/HiQRGv/ife/G9Mtwoq+iDweO/3S16eH3OmNEIf1srq2vrGZmmrvL2zu7dvHxx2VJRIQtsk4pHs+VhRzgRta6Y57cWS4tDntOtPb3K/O6NSsUjc6zSmXojHggWMYG2koW0PQqwnfpA1z0lKOJ0P7QpyqtcGdYicGkLIRQVBVeg6aIEKKNAa2u+DUUSSkApNOFaq76JYexmWmuX3lQeJojEmUzymfUMFDqnyskXyOTw1yggGkTRHaLhQv29kOFQqDX0zmedUv71c/MvrJzqoexkTcaKpIMuHgoRDHcG8BjhikhLNU0MwkcxkhWSCJSbalFU2JXz9FP5POheOe+mgu1ql0SzqKIFjcALOgAuuQAPcghZoAwJm4AE8gWcrsx6tF+t1ObpiFTtH4Aest0/aRZPP</latexit>

cycle
x

x
x x A

>
>
B
x
1
⇤2 ⇤2 ux
x x

A cycle
<latexit sha1_base64="1pzvLQw9UppZh7ZzrTsT+RR+pTs=">AAAB+XicdVDNS8MwHE3n15xfVY9egkPwYmn3weZt6sXjBPcBWxlplm5haVqSdFDK/hMvHhTx6n/izf/GdKugog8Cj/d+v+TleRGjUtn2h1FYW9/Y3Cpul3Z29/YPzMOjrgxjgUkHhywUfQ9JwignHUUVI/1IEBR4jPS82U3m9+ZESBrye5VExA3QhFOfYqS0NDLNYYDU1PPTqwucYEYWI7NsW7XLpl1twBWpV3JSq0LHspcogxztkfk+HIc4DghXmCEpB44dKTdFQtHsvtIwliRCeIYmZKApRwGRbrpMvoBnWhlDPxT6cAWX6veNFAVSJoGnJ7Oc8reXiX95g1j5TTelPIoV4Xj1kB8zqEKY1QDHVBCsWKIJwoLqrBBPkUBY6bJKuoSvn8L/SbdiOXXLvquVW9d5HUVwAk7BOXBAA7TALWiDDsBgDh7AE3g2UuPReDFeV6MFI985Bj9gvH0CACaT6g==</latexit>

Figure 12.6: On the left the elleptic curve in the (two sheeted) x plane. A and B
are the two one-cycles of the torus. On the right the corresponding torus.

the integrals over the A and B cycles of the unique holomorphic (closed) one-form
on the torus, Ω = dx/y
I I
dx dx dx dx
ω= , ωD = with =p . (12.64)
A y B y y (x − Λ2 )(x + Λ2 )(x − u)

Note that the periods ω and ωD inherit from the A and B cycles the transformation
properties under Sp(2, Z) and so also the same monodromies at the three singular
points u = ∞, ±Λ2 , where two branch points collide.
The identification between the SU (2) gauge theory and the above family of
tori parametrized by u holds via identifying the modulus of the torus with the
complexified gauge coupling, and the periods with the u derivative of a and aD ,
that is
ωD ∂aD ∂aD /∂u
τ= ≡ τ (a) = = , (12.65)
ω ∂a ∂a/∂u
with the identification
I I
∂a dx ∂aD dx
=ω= , = ωD = , (12.66)
∂u A y ∂u B y
and an overall normalization that we will fix momentarily. Note, in passing, that
since u is globally defined, a and aD have the same monodromies of the periods ω
and ωD .
Integrating in u on both sides one obtains
I I
a= dλ , aD = dλ , (12.67)
A B

321
where the one-form differential dλ (aka Seiberg-Witten differential) can be easily
computed
∂dλ dx dx (x − u)dx
= =p −→ dλ = , (12.68)
∂u y (x − Λ4 )(x − u)
2 y

up to exact forms. Using the above definition of A and B cycles, and deforming
them so to lie entirely along the cuts between −Λ2 and Λ2 and between Λ2 and u,
respectively, one can express the integrals (12.67) as
√ Z Λ2 √
2 x−u
a(u) = dx √ (12.69)
π −Λ2 x2 − Λ 4
√ Z u √
2 x−u
aD (u) = dx √ , (12.70)
π Λ2 x2 − Λ 4
where the overall normalization has been fixed by requiring that for u → ∞ one
recovers the (semi)classical result (12.55). Using the identity
Z 1
Γ(γ)
F (α, β, γ; z) = dx xβ−1 (1 − x)γ−β−1 (1 − zx)−α , (12.71)
Γ(β)Γ(γ − β) 0
one can finally recast (12.69) and (12.70) in terms of hypergeometric functions
√ 1 1 2
a(u) = 2(Λ2 + u)1/2 F (− , , 1; ) (12.72)
2 2 1 + u/Λ2
Λ − u/Λ 1 1 1 − u/Λ2
aD (u) = i F ( , , 2; ). (12.73)
2 2 2 2
One can invert (12.72) to obtain u(a) and insert the result into (12.73) to obtain
aD (a). Integrating with respect to a yields F(a). Equivalently, deriving with respect
to a yields τ (a) and, hence, the exact expression of the low energy effective action
(12.45)!
Let us emphasize again that the expression one gets for F(a) is not globally
defined on the moduli space, and different analytic continuations should be used in
different patches. For example, near u = Λ2 , better to use S-dual coordinates, where
the role of what is electric and what is magnetic is inverted. This is represented in
figure 12.7.
As a check that the result we got describes the coupling τ entering the effective
Lagrangian (12.45), one can expand (12.73) and (12.72) around u = 0, Λ2 and −Λ2
and show agreement with the expected (singular) behavior for aD and a, including
the monodromies (12.57) and (12.60).

322
 ✓ ◆
i 2 a2
F(a) = a log + ...
2⇡ ⇤2
1

F̃D = FD (a aD )
⇤2
⇤2

i 2 ⇣a ⌘
D
FD (aD ) = aD log + ...
2⇡ ⇤

Figure 12.7: The quantum moduli space Mq of pure SYM with G = SU (2) rep-
resented as a sphere, obtained by adding the point at infinity to the complex u
plane. The space is covered by three distinct regions where a local, weakly coupled
Lagrangian can be written using appropriate coordinates, i.e. the appropriate du-
ality frame. No local Lagrangian exists which would be globally defined on Mq .
What are classical and strongly coupled regions is not an invariant concept, since it
depends on the coordinate frame.

√ i√
• For u → ∞ we have a ∼ u and aD ∼ π
u log Λu2 ∼ πi a log Λa .
This reproduces the (semi)classical result (12.55) and so also the correct mon-
odromy (12.57). Note that there is no choice of (nm , ne ) giving a vanishing
mass, in agreement with the fact that at u → ∞ there are no extra massless
particles in the spectrum.

• For u → Λ2 we have a ∼ πi aD log aΛD and aD ∼ (u − Λ2 ).

So we see that a is singular at u ∼ Λ2 while aD vanishes. This is the correct
behavior for a magnetic monopole with charge nm becoming massless at u =
Λ2 , in agreement with what we previously found. Again, the monodromy MΛ2
in eq. (12.60) is correctly reproduced.

• For u → −Λ2 we have a − aD ∼ (u + Λ2 ) and a ∼ πi (aD − a) log aDΛ−a .

This shows that at u = −Λ2 we have a singularity where a = aD , which gives
a massless dyon with opposite electric and magnetic charges, ne = −nm , again
in agreement with previous results, including the monodromy M−Λ2 .

There are other non-trivial checks one can make. For example, one can expand
τ (u), eq. (12.65), in (inverse) powers of u, at large u, and compare with (12.46),

323
using u = a2 . This gives perfect agreement with the instanton coefficients d1 and
d2 , which have been independently calculated.
As anticipated, an inspection of the exact solution (12.72) shows that for no
values of u the scalar field VEV a becomes 0. So, the point a = 0 is not part of the
quantum exact moduli space, as anticipated. This is consistent with the claim that
nowhere on the moduli space extra massless gauge bosons arise.
To sum-up, at a generic point of the moduli space the effective theory is a N = 2
abelian free theory. At two special points, u = Λ2 , −Λ2 the effective theory is N = 2
SQED with one massless flavor. Just, to have a local description one should use a
dual frame, since monopoles couple locally to an effective Lagrangian written in
terms of the dual gauge field AD and dyons with charge (1, −1) couple locally to an
effective Lagrangian written in terms of the gauge field AD − A. In this sense the
word ”monopole” or ”dyon” is just conventional and adapted to the large u region,
where the theory is semiclassical and local in terms of the gauge field A, in which
monopoles and dyons appear as non-perturbative states (for instance, in the AD
frame it is an electron which looks as a non-perturbative state).

12.3.2 Intermezzo: confinement by monopole condensation

Before discussing generalizations of this model, there is one (very nice) consistency
check one can do.
Let us start from N = 2 SU (2) SYM and add a mass m to the chiral superfield Φ
belonging to the N = 2 vector multiplet, that is W = m TrΦ2 . This breaks explicitly
N = 2 to N = 1. For m >> Λ we can use the UV Lagrangian, integrate Φ out
and end-up with pure N = 1 SU (2) SYM at low energy, which admits two isolated
supersymmetric vacua with charge confinement and mass gap. As we discussed in
chapter 10, by supersymmetry this same scenario should hold even if m << Λ. In
this regime, the low-energy N = 2 effective description we discussed before should be
approximately valid and we should use it, adding to it the small mass perturbation.
But how the moduli space can be lifted giving back just two isolate (gapped) vacua?
How can the otherwise massless photon get a mass, since, see our discussion in §
12.1, there are no light charged fields? As we are going to show below, the results
we got in the previous section contain the answer.
The addition of a mass term with m << Λ makes the effective theory becoming
an N = 1 abelian gauge theory with a massive (neutral!) chiral multiplet Φ. Let us

324
dub U = TrΦ2 the chiral superfield whose lowest component VEV u parametrizes
the (original) N = 2 moduli space. At a generic value of u, there are no massless
(or nearly massless) chiral superfields other than U so we easily see that the F-term
equation we have to impose on the space of D-flat directions
∂W
=0 (12.74)
∂U
cannot be satisfied for m 6= 0. Therefore, there are no (N = 1) supersymmetric
vacua, in contradiction with what expected. In fact, we have learnt that at special
points of the complex u-plane (two points, actually), there are extra massless degrees
of freedom. One such points is u = Λ2 where a massless magnetically charged
hypermultiplet is present and should hence be included in the effective theory. Let
us describe our theory near u = Λ2 . There, better to use S-dual variables, for which
the superpotential reads
√
W = 2 H1 ΦD H2 + m U , (12.75)

where ΦD is the S-dual of Φ and U should be thought of as a function of ΦD , now.

The D-term equations from the coupling to the (magnetic dual) U (1) gauge field
imply that |H1 | = |H2 | (recall that H1 and H2 have conjugate internal quantum
numbers and hence are oppositely charged under the U (1) gauge symmetry), while
the F-term equations read
√ du
2H1 H2 + m = 0 , aD H1 = aD H2 = 0 . (12.76)
daD
Since du/daD 6= 0 (u is a good global coordinate on u!) we get the following answer

m = 0 : H1 = H2 = 0 , aD = any
 1/2
m du
m 6= 0 : H1 = H2 = − √ , aD = 0 . (12.77)
2 daD aD =0
For m = 0 we recover (tautologically) the N = 2 moduli space. For m 6= 0, since
H1 and H2 are (magnetically) charged, their VEVs break U (1) and give a mass to
the abelian gauge field (to all the N = 1 gauge multiplet, in fact). So, we end up
with a supersymmetric vacuum with a mass gap. This same reasoning holds also at
u = −Λ2 , where the role of the massless magnetic monopole is played by a massless
dyon. So, the N = 2 moduli space is fully lifted but at two points, where there are
supersymmetric vacua with mass gap. Exactly what we expect. The exact answer

325
we got for the original N = 2 model is just right to give what should hold for the
N = 1 massive theory under study!
This is nice, but, superficially, there is still a point of concern. The two N = 1
supersymmetric vacua are confining ones. Here, instead, the dynamics is more of
a Higgs-like mechanism, where a scalar monopole or a dyon field condenses. The
point is that the Higgs mechanism taking place is not the usual condensation of an
electrically charged field, but of a magnetically (or dyonically) charged one.
To understand what that means let us recall some basics of the usual Higgs
mechanism, where electrically charged fields condense.
The condensation of the electric charge has the effect that any background elec-
tromagnetic field gets screened. This implies that electric sources in the theory are
(almost) free, since their electric fields can be absorbed by the vacuum condensate
and their interaction energy drops off exponentially. Magnetic charges behave very
differently. The magnetic field lines have no condensate where to, say, end on. The
result is that magnetic field lines tend to be expelled from the vacuum (this is the
well-known Meissner effect taking place in superconductors). The minimum energy
configuration is for the magnetic field to be confined to a thin flux tube connecting
opposite magnetic charges. Therefore, in the Higgs mechanism, electric charges are
screened and magnetic charges are confined (note: this is strict confinement).
In the model above what condenses (let us focus, momentarily, on the vacuum
at u = Λ2 ) is not an electric charged state, but a magnetically charged one. By
electromagnetic duality it follows that here magnetic charges are screened, while
electric charges are confined. So, eventually, we do have a confining vacuum (due to
a magnetic dual of the Meissner effect)! This is a concrete realization of an old idea,
due to ’t Hooft and others, that confinement in non-Abelian gauge theories maybe
due to monopole condensation. The point u = −Λ2 , where a dyonic field condense,
is just related to the latter by a different electro-magnetic duality rotation. There,
both electric and magnetic charges are confined but dyonic charges proportional to
(1, −1) won’t, they will just be screened. This is known as oblique confinement, also
proposed by ’t Hooft long ago.
The result we got is beautiful from several point of views. First, it shows that
the presence of magnetically charged solitons becoming massless somewhere on the
moduli space is necessary to match N = 2 dynamics with N = 1 via holomorphic
decoupling, one of our guiding principles all along this course. Second, it gives an
a posteriori consistency check about the claim that two and only two singularities

326
should be there on M. Finally, it shows (at least in this softly broken N = 2 model)
that confinement is due to monopole condensation, providing a concrete realization
of the old idea that this could in fact be the way electric charge gets confined.

12.4 Seiberg-Witten theory: generalizations

Till now we have been focusing on pure N = 2 SYM with gauge group SU (2). The
story can be generalized to gauge groups with higher rank and/or coupled to matter
fields.
We are not going to discuss these generalizations in detail and refer the interested
reader to the references at the end of this lecture. Still, we want to make a few
remarks and discuss in some detail one (instructive) example.
For G = SU (2) and no matter we have learnt that the moduli space, whose
complex dimension is one, is the complex plane u with two singularities at u =
±u0 (beside the singularity at infinity), exchanged by the residual Z2 R-symmetry.
At these two singularities magnetically charged objects, a monopole and a dyon,
respectively, become massless. The metric on the moduli space can be described via
an auxiliary elleptic curve, a Riemann surface Γ(u) of genus n = 1 (a torus), whose
modulus τ can be explicitly computed and corresponds (in fact, its imaginary part
does) to the metric itself, the only unknown in the low energy effective action.
Gauge groups with higher ranks mean moduli spaces M with complex dimen-
sion n, locally Cn /WG , where WG is the Weyl group of the gauge group G, n the
gauge group rank and uI , I = 1, . . . , n, gauge invariant coordinates on M, see e.g.
eqs. (12.5). As already discussed, these theories are generalizations of the (N = 2
supersymmetric version of the) Giorgi-Glashow model. As such, they admit several
types of charged soliton-like solutions which, in the BPS limit, satisfy the BPS mass
formula (12.35). The low energy effective action is form-invariant under electro-
magnetic duality rotations, which are generated by Sp(2n, Z). Again, one finds
that on the quantum exact moduli space there are singularities where magnetically
charged states become massless. The prepotential is not globally defined and dif-
ferent charts (duality frames) should be used. In order to extract the metric on the
moduli space, one can again make use of auxiliary elliptic curves Γ(uI ), generaliza-
tions of the curve (12.61), which are genus n Riemann surfaces, now, and can be
described as a double-sheeted x plane with n + 1 branch cuts. The period matrix τ
of these genus n two-dimensional surfaces, which is defined in analogy to the mod-

327
ulus of a torus and whose imaginary part is in fact positive, gets identified with the
gauge coupling matrix τ and hence determines the metric on the moduli space and,
in turn, the exact low energy effective action (12.41).
A genus n Riemann surface, figure 12.8, can be characterized in terms of n pairs
of homology cycles AI and BI and, correspondingly, period integrals defined as
I I
J
ωIL = ΩL , ωD L = ΩL I, J, L = 1, . . . , n , (12.78)
AI BJ

where ΩL are n independent holomorphic one-forms, generalizations of the unique

holomorphic differential Ω defined on genus 1 surfaces, eqs. (12.64). The period
matrix is given by
τ = ωD · ω −1 (12.79)
(to be understood as a matrix equation), which ensures that Im τ > 0. The identi-
fication goes as before. In particular, the period matrix (12.79) gets identified with
the gauge coupling matrix and the periods with the u derivative of aI and aDJ , that
is I I
∂aI ∂dλ J ∂aDJ ∂dλ
ωIL = = , ωD L = = (12.80)
∂uL AI ∂uL ∂uL B I ∂uL
and I I
aI = dλ , aDJ = dλ , (12.81)
AI BJ

where dλ is again the Seiberg-Witten differential.

A1
>

>
>

A2 An 1
An
>

B1 B2 ….. Bn 1 Bn
> >
> >

Figure 12.8: A genus n Reimann surface with the basis of homology cycles AI , BJ .

What about adding matter? The presence of matter fields opens-up the possi-
bility for Higgs branches. However, in what follows we will focus on the Coulomb
branch only, since, as already emphasized, that is the only component of the moduli
space which gets modified at the quantum level. Note, also, that since we want
to keep the theory UV-free, matter is constrained, there cannot be too much. For

328
instance, adding matter to SU (2) SYM one can add up to three hyermutiplets in
the fundamental representation or, if allowing a vanishing β function, four hyper-
multiplets in the fundamental representation or one in the adjoint (the latter case
corresponds to N = 4 SYM).
While matter fields do not change the dimension of the Coulomb branch (and
so the genus of the corresponding Seiberg-Witten curve), they do change its singu-
larity structure. One way to see this is to notice that hypermultiplets enjoy two
contributions to their (effective) mass: the bare mass mi and, whenever the adjoint
chiral superfield Φ gets a VEV, the one inherited from the N = 2 supersymme-
try preserving cubic coupling, eq. (12.32). So, one expects new singularities on
the moduli space wherever the two mass contributions cancel each other and the
(charged) hypermultiplets become effectively massless. Notice, further, that in order
to understand the monodromy associated to these singularities, one should use the
generalized BPS mass formula
XF
1
Z = a · n e + aD · n m + √ mi Si , (12.82)
i=1
2
in place of (12.35).
In order to construct the curves Γ(uI ) concretely, one should follow the same
logical steps we discussed for the SU (2) theory, some of the guiding principles being
matching their singularity structure with the appearance of massless particles in
the N = 2 theory spectrum, R-symmetry (and in this case also flavor symmetry)
considerations as well as agreement, by holomorphic decoupling or scale matching,
with curves with less flavors or smaller gauge groups. In any event, one ends up
with equations like
y 2 = f (x, uI , mi , Λ) , (12.83)
where uI , the moduli space coordinates, parametrize the period matrix τ of the
curves, eq. (12.79), mi are hypermultiplet bare masses (where i = 1, . . . , F ) and Λ
is the strong coupling scale (we are assuming, for simplicity, that the gauge group
G is simple).
It should be remarked that different parametrizations can be used to represent
the curve associated to the moduli space of a given theory. Some may be more useful
than others, depending what one wants to look at. For this reason, in the literature
(and in the references at the end of this lecture) different parametrizations can be
found. Needless to say, they are all physically equivalent, as they should. We will
see one such example shortly.

329
 Admittedly, despite the clear logical steps and guiding principles one can follow,
some amount of educated guesswork is usually needed to find the correct curves.
That is to say, there is not an overall recipe to construct the Seiberg-Witten curve
for an arbitrary theory.
This said, a systematic way to construct a large class of Seiberg-Witten curves
does exist and relies on M-theory, where a physical meaning can be given to the
Riemann surfaces (the curves) themselves. Four-dimensional N = 2 theories can be
engineered from M 5 branes wrapped on suitably chosen two-dimensional compact
surfaces. At low energy, smaller than the typical size of the surface, the theory
becomes effectively four-dimensional and preserves N = 2 supersymmetry. Such
Riemann surfaces are nothing but the Seiberg-Witten curves! This makes several
properties of the low energy effective theory having a geometrical interpretation
which often helps.
As already stressed, there are a number of consistency checks one can make on
the curves (12.83). For example, by making one hypermultiplet massive, that is
taking its mass mi large, eventually mi → ∞, one can integrate the hypermultiplet
out and end on the theory with one flavor less, and then show that the limit of the
corresponding curve agrees with the curve with one less hypermultiplet. Similarly,
by letting one of the vacuum expectation values becoming large, one obtains a limit
in which the gauge group is higgsed at high energy, e.g. SU (N ) → SU (N − 1).
Again, the corresponding limit of the SU (N ) curve should agree with the curve for
SU (N − 1).

12.4.1 A case study: N = 2 SU (2) SQCD with one flavor

To make the above discussion a bit more concrete, we will now work out one of
the simplest generalizations of the original Seiberg-Witten model, namely N = 2
SQCD with gauge group SU (2) and one massive hypermultiplet in the fundamental
representation. The proposed corresponding Seiberg-Witten curve reads

y 2 = x2 (x − u) − Λ61 + 2mΛ31 x , (12.84)

where, with obvious notation, Λ1 is the strong coupling scale.

Let us first check that upon holomorphic decoupling one recovers the curve for
the unflavored SU (2) gauge theory. If we send m → ∞ keeping mΛ31 fixed, the
flavor decouples and, using the scale-matching relation Λ4 = mΛ31 , one ends up with

330
the curve
y 2 = x2 (x − u) + 2Λ4 x . (12.85)
This is the curve for pure SYM with gauge group SU (2), although in a different
parametrization (and normalization) with respect to eq. (12.61). Let us see how
these two parametrizations are related.
First, as previously noticed, in a normalization where charged fields transforming
in the adjoint representation of the gauge group have integer charges, those of fields
transforming in the fundamental are half integers. Hence, in making a comparison
between gauge theory with and without matter, it is convenient to first change the
normalization we used to treat SU (2) pure SYM and multiply, in eq. (12.30), ne by
2, so to ensure ne to still be an integer, and divide a by 2, so to ensure that (12.30) is
unchanged. This change of conventions corresponds to the following transformation
on the vector (aD , a) ! ! !
aD 1 0 aD
→ , (12.86)
a 0 1/2 a
which changes monodromy matrices as
! ! ! ! !
a b 1 0 a b 1 0 a 2b
→ = . (12.87)
c d 0 1/2 c d 0 2 c/2 d

With these normalizations, the monodromy matrices (12.57) and (12.60) become
! ! !
−1 4 1 0 −1 4
M∞ = , M Λ2 = , M−Λ2 = (12.88)
0 −1 −1 1 −1 3

The corresponding elleptic curve reads

1
y 2 = x2 (x − u) + Λ4 x , (12.89)
4
which, by a constant rescaling of Λ, coincides with (12.85).
As compared to (12.61), the above expression makes it less transparent the points
on the complex plane u where singularities arise, namely where two branch points
collide and the curve degenerates. To this aim, regardless the parametrization one is
Q
using, it suffices to compute the discriminant of the x-poynomial, ∆ = i<j (αi −αj )2
(where αi are the roots of the polynomial), and find the values of u such that some
roots coincide. For a cubic polynomial of the form

x3 + bx2 + cx + d , (12.90)

331
we have ∆ = b2 c2 − 4c3 − 4b3 d + 18bcd − 27d2 which applied to eq. (12.89) gives
1 8 2
∆= Λ (u − Λ4 ) , (12.91)
16
which vanishes at u = ±Λ2 , in agreement with our previous analysis. More generally,
the roots αi of the x-polynomial (i.e. the branch points) are functions of the parame-
ters of the theory which, in the pure SYM SU (2) are just u and Λ, αi = αi (u, Λ). The
explicit form of these functions (and their number!) depends on the parametrization
chosen to describe the elliptic curve. For example, in the parametrization (12.61)
they are α1 = Λ2 , α2 = −Λ2 , α3 = u, while in the parametrization (12.89) they are
√ √
α1 = 0, α2 = 1/2(u + u2 − Λ4 ), α3 = 1/2(u − u2 − Λ4 ). What does not (and can-
not) change, instead, are the number and the locations of the singularities on the u
plane (namely the values of u for which two or more roots coincide), nor the nature
of the particles becoming massless there, since these are physical information.
Let us now go back to the curve (12.84), and look at the singularity structure
of the moduli space of N = 2 SQCD with F = 1 more carefully. As compared to
pure SYM, we expect now three singularities at finite distance in the u plane. This
can be seen as follows. Let us first suppose the mass m to be very large. Then, at
roughly the scale |m| the hypermultiplet decouples and, below that scale the theory
behaves, effectively, as SU (2) SYM. This suggests that the structure of the moduli
space in the region |u| << |m2 | should be the same of the pure SU (2) theory, with
two singularities at u ' ±Λ2 . Moreover, following our general discussion, we expect
a third singularity where the hypermultiplet becomes effectively massless. For m
large, this could only happen in the large VEV region where u = a2 . One can then
easily see that a balance, and hence a potential cancellation, between the bare mass
and the one coming from the cubic term in the superpotential, eq. (12.32), will occur
at u ' m2 . Consistently with holomorphic decoupling, sending m → ∞ this third
singularity is pushed all the way to infinity and one recovers, correctly, the pure
SU (2) SYM moduli space of figure 12.5.
Let us now consider the other extreme case, namely m = 0. While a mass term
for the hypermultiplet completely breaks the R-symmetry, in the massless case there
is a preserved Z6 R-symmetry at the quantum level. The u coordinate has R-charge
4 and one can then easily see that each point on the u plane preserve a Z2 symmetry.
Hence, in the massless case we expect three strong coupling singularities, related by
the broken Z3 generators.
The above conclusions can be checked analytically from the curve (12.84), by

332
computing the discriminant and the three roots expanding the result for large, re-
spectively small m. The discriminant of the x-polynomial (12.84) is

∆ = 4m2 u2 Λ61 − 32m3 Λ91 − 4u3 Λ61 − 27Λ12 9

1 + 36umΛ1 . (12.92)

Let us consider first the large mass regime, m >> Λ1 , eventually m → ∞ keeping
mΛ31 = Λ4 fixed. One can compute ∆ at u ' m2 and find ∆ = 0 + O[(Λ1 /m)3 )].
Similarly, at u2 = ±8Λ4 , one finds ∆ ∼ 0 ± O[(Λ1 /m)3/2 )]Λ4 , as expected. In the
massless case, m = 0, the discriminant reduces to ∆ = −4u3 Λ61 − 27Λ12 1 and one
easily sees that the three singularities are instead located at
 1/3
27 6
u = − Λ1 , (12.93)
4
which shows, as anticipated, that they all are in the strong coupling region and get
transformed one another by Z3 rotations. Clearly, one can interpolate between these
two extreme cases by continuously increasing (decreasing) m. Figure 12.9 shows how
the moduli space changes as we vary the hypermultiplet mass.

u u 2 u
m
x

e2/3⇡i ⇤21 add mass

x
! !
x x
mass x x
in nity x x

e4/3⇡i ⇤21 ⇤21 ⇤20 ⇤20 ⇤20 ⇤20

Figure 12.9: The moduli space of N = 2 SQCD with F = 1 as we vary the hyper-
multiplet mass. In the limit of infinite mass one recovers the pure SU (2) moduli
space, rightmost figure.
fi

Interestingly, the nature of the hypermultiplet becoming massless at the three

singularities depends on m. When m is large, the two strong coupling singulari-
ties (and the associated monodromies) are basically the same as the pure theory,
and the corresponding massless particles a monopole and a dyon, respectively. At
u ' m2 the (elementary) hypermultiplet becomes massless, which is an electrically
charged object. So, for m → ∞ the (p, q) charge of the massless particles are
(1, 0), (1, −1), (0, 1). As we decrease |m| the three singularities become closer and
closer and more and more similar to those of the massless theory, which are related

333
by Z3 rotations. So there is less and less a clear distinction between hypermultiplets
coming from solitons or from elementary objects. In fact, in the massless case the
(p, q) charge of the massless particles can be easily computed to be (1, 0), (1, 1), (1, 2),
which is very different from the massless spectrum in the large mass case. This does
not hurt. The mass is a UV parameter and, as we change it, we do change the
theory. Hence, that the physics changes should not come as a surprise.
Notice, finally, that at any given singularity, regardless the value of m, there
always exists a dual frame where the massless particle is ”electric” and the effective
theory is then nothing but N = 2 SQED with one massless flavor.

Argyres-Douglas theories

In this model there is a choice of UV parameters which makes one of the singular
points of the u-plane special. Let us first recap. The roots of the x-polynomial are
functions of the Coulomb branch parameter u, the strong coupling scale Λ1 and the
hypermultiplet bare mass m, αi = αi (u, Λ1 , m). There exists three different ways to
collide two branch points, α1 = α2 , α1 = α3 or α2 = α3 , which correspond to the
three singularities of the u-plane discussed above, where the A or B or A + B cycles
collapse and one hypermultiplet (no matter its nature) becomes massless. This is
what happens for generic values of m. Can something more singular happen?
Let us choose the mass m to be m = (3/2) ωΛ1 with ω 3 = 1. With this choice
the Seiberg-Witten curve (12.84) becomes (we set for simplicity Λ1 = 1 and ω = 1)

y 2 = x2 (x − u) + 3x − 1 , (12.94)

while the discriminant (12.92) reads

∆ = 9u2 − 108 − 4u3 + 54u − 27 . (12.95)

Singularities on the u-plane occur at those values of u for which the discriminant
vanishes. In the present case it so happens that two singularities on the u-plane
merge. There is one singularity at u = −15/4 where two roots of the x-polynomial
coincide and another one at u = 3 where all three roots coincide. Indeed, when
u = −15/4 we get y 2 = (x + 2)2 (x − 1/4) while for u = 3 the Seiberg-Witten curve
becomes y 2 = (x − 1)3 and all branch points coincide at x = 1. This means that at
u = 3 both A and B cycles shrink and all three (mutually non local!) hypermultiplets
become massless.

334
 This is something not specific to SU (2) SQCD with one flavor, but it can happen
whenever the gauge group rank is larger than one or, as in the present case, if matter
is added to pure SYM: there exist special points on the moduli space, known as
Argyres-Douglas points, where mutually non-local objects become simultaneously
massless. This means that there does not exist a duality frame in which all (light)
fields are electric and that the theory cannot be as simple as SQED coupled to
massless flavors. What that can be?
It is believed that at points where mutually non-local objects becomes simul-
taneously massless the theory enjoys an interacting (as opposed to free) conformal
phase. At first sight this might sound surprising. Coleman-Gross theorem states
that in four dimensions any theory of scalars, spinors and abelian gauge fields is
IR-free. Our low energy effective theory is abelian and, as we already emphasized,
there are no points whatsoever on the quantum moduli space where extra gauge
bosons (extra with respect to those associated to U (1)n , where n = RankG) become
massless. So there seems not to be room for an interacting fixed point.
Actually, as we already observed, what is special about these points is that
cycles having non-vanishing intersections (like the A and B cycles of the two-torus)
simultaneously shrink there. Physically, this corresponds to, e.g. a dyon and a
monopole, or a dyon and an electrically charged object becoming simultaneously
massless. This is a situation where the Coleman-Gross theorem cannot be proven,
since the theory lacks a Lagrangian description. As we now discuss, it turns out
that at such points the theory enjoys an interacting abelian Coulomb phase.
Let us consider a CFT in four dimensions and focus on fields’ scaling dimensions
∆ and Lorentz spins. The latter can be represented as (j+ , j− ), the two eigenvalues
being SU (2) quantum numbers. Unitarity and conformal symmetry provide lower
bounds on the scaling dimensions of various operators. For instance, for a ”chiral”
primary operator (an operator annihilated by special conformal generators Kµ and
for which either j+ or j− vanish) the following inequality holds

∆ ≥ j+ + j− + 1 . (12.96)

For non-chiral primaries, instead

∆ ≥ j+ + j− + 2 . (12.97)

Equality in the above equations holds for free fields. Take the field strength operator
Fµν with given conformal dimension ∆. This is the sum of two conformal primary

335
operators, schematically F ± = F ±∗ F , whose Lorentz spins are (1, 0) and (0,1),
respectively. One can show that the states associated to the conserved currents
Jµ± = ∂ ν Fµν
±
= ∗ dF ± satisfy the equation ||Jµ± i|2 = 2(∆ − 2). So we see that ∆ = 2
if and only if the currents are null vectors, J ± = 0, which are nothing but the
Bianchi identity and the equation of motion of a free Maxwell theory. Conversely,
if F is not free ∆ > 2 and both J + and J − are different from zero. Since they
are descendants of different primary fields, they are linearly independent and, in
turn, this implies that both the electric current Je = J + + J − and the magnetic
current Jm = J + − J − are non zero. So we conclude that in a CFT any interacting
field strength must couple both to electric and magnetic charged objects. In other
words having both elementary monopoles and electric charges allow QED to have
a non-trivial fixed point, while QED without elementary monopoles cannot have a
non-trivial fixed point (in agreement with Coleman-Gross theorem).
We have seen that in our theory (for a proper choice of the UV mass parameter)
at the u = 3 singularity mutually non-local objects become massless and so, by the
above argument, the IR dynamics is believed to be described by an interacting fixed
point, a so-called Argyres-Douglas theory.
In all this discussion supersymmetry did not play any role. Not surprisingly,
however, the extra constraints imposed by N = 2 supersymmetry let one get more
clues on the property of the interacting fixed point, for example by providing the
exact scaling dimension of some CFT operators. As discussed in § 6.3, a new feature
of the superconformal algebra as compared with the conformal algebra is that there is
another symmetry generator under which operators transform, the R-symmetry. In
the N = 2 superconformal algebra this is U (2)R and operators are also characterized
by the U (1)R charge R and SU (2)R ”spin” I. From the superconformal algebra one
can show that a chiral primary operator (a state with j− = 0 and annihilated by
i
the supercharge Qα̇ ) satisfies the relation
1
∆ = 2I + R ≥ 2I + j+ + 1 . (12.98)
2
Let us now consider the N = 2 vector superfield. In N = 2 superfield formalism
this is a scalar superfield U satisfying the chiral constraint

Dα̇i U = 0 where i = 1, 2. (12.99)

Using N = 2 unconstrained superfield formalism one can show that in N = 1

language this has the same field content of a chiral superfield Φ and a gaugino

336
superfield Wα . The lowest component of U , that we dub u, has I = 0 and Lorentz
spin (0, 0), so ∆(u) = 12 R(u) ≥ 1. This implies that ∆(F + ) ≥ 2 and, when the
equality is saturated we have that dF + = 0 (recall previous discussion). When
R(u) = 2 (and hence ∆(u) = 1) the superfield U satisfies also the equation

Dα(i Dαj) U = 0 . (12.100)

which is then equivalent to say that the field is free and there is a null state, dF + = 0.
For an interacting vector multiplet eq. (12.100) does not hold, but just (12.99), there
are no null states and electric and magnetic currents cannot vanish.
Let us now consider the Coulomb branch of N = 2 SU (2) SQCD. At a generic
point we have a free Maxwell theory described by a free N = 2 U (1) vector multiplet
U with ∆(u) = 1. What changes in the IR theory if we add a relevant operator in
the UV theory, like a mass term for the one flavor?
If we shift the elementary hypermultiplet UV mass m, the prepotential of the
effective theory is modified. The leading order operator at a given point on the
moduli space can be obtained by expanding the variation of the prepotential: the
R
constant term would not contribute, the linear one, d4 θU , using eqs. (12.99) and
(12.100) can be shown to be a total space-time derivative so the leading term is
proportional to U 2 and provides a shift of the effective coupling τ . The same holds at
singular points of the Coulomb branch where one hypermultiplet becomes massless,
since a mass term can be absorbed in a shift of U and again the linear term in U
does not contribute.
At an interacting fixed point, instead, U is chiral but eq. (12.100) is not satisfied
anymore. Hence, at such special point, if it exists, the leading effect is the linear
one Z
d4 θ mU (12.101)

which does not vanish now, implying that

∆(m) + ∆(U ) = 2 . (12.102)

This shows that m is a source for the CFT operator U .

Let us now go back to the special point u = 3, m = 3/2 of our model where the
IR dynamics we now know is a SCFT, and let us expand the curve (12.94) around
such point in terms of shifted variables (M, ũ, x̃) defined as
3 1
m= + M , u = 3 + 2M + ũ , x = u + x̃ . (12.103)
2 3
337
We get
4
y 2 = x̃3 − 2(M + ũ)x̃ − (ũ + M 2 ) + . . . , (12.104)
3
where the dots are higher order terms in M and ũ, which are both small in a
neiborhood of the fixed point. From the above equation we see that in order to see
the cubic singularity at ũ = M = 0, meaning that nothing should dominate against
x̃3 in (12.104) - we have to assign the following relative scaling

∆(x̃) : ∆(M ) : ∆(ũ) = 1 : 2 : 3 . (12.105)

So in eq. (12.104) we can drop ũ x̃ and M 2 terms and get the simplified expression

y 2 = x̃3 − 2M x̃ − ũ (12.106)

at the SCFT point. Remarkably, from the scaling dimensions of the coefficients of
the curve (12.106) one can extract the scaling dimensions of CFT operators. The
polynomial (12.106) determines the scaling dimensions of the various couplings, up
to an overall scaling. This can be fixed recalling that a ∼ (ũ/y)dx̃ gives the mass of
BPS particle and hence should have scaling dimension one. From (12.105) we have
3
[y] = [x̃] , [ũ] = 3[x̃] , [M ] = 2[x̃] (12.107)
2
and using that [a] = 1 we finally get
2 6 4 3
[x̃] = , [ũ] = , [M ] = , [y] = . (12.108)
5 5 5 5
From these relations we learn a few interesting facts:

• We get the exact scaling dimension of ũ, and this is larger than 1, showing we
are at an interacting fixed point, as expected. Notice, further, that since in
the UV u = 12 Trφ2 has dimension ∆ = 2, we see that u acquires an order one
anomalous dimension along the RG flow, in agreement with the idea that we
are at a strongly coupled fixed point.

• ∆(ũ) + ∆(M ) = 2 in agreement with the idea that M is the dual coupling to
R
Ũ , Ũ is a genuine operator and d4 θM U e a deformation out the fixed point
theory.
R
• Since ∆(ũ) < 2, d4 θM U e is in fact a relevant deformation.

338
 As already remarked, Argyres-Douglas fixed points do not exist only for the
theory we have been considering. For instance, as far as SU (2) SQCD, there exist
Argyres-Douglas CFTs for any F ≤ 3. The same happens for the other minimal
generalization of the original Seiberg-Witten model, namely pure SYM with gauge
group SU (3) (which is actually the first instance where this phenomenon was dis-
covered) and several generalizations thereof.

12.5 N = 4: Montonen-Olive duality

In view of discussing N = 4 SYM, it is useful to do a step back and consider
again the original Georgi-Glashow model, which, as already stressed, finds a natural
embedding both into N = 2 and N = 4 SYM.
In the BPS limit, all states of the Giorgi-Glashow model satisfy the BPS mass
formula (12.22). Combined with Dirac quantization condition, it implies that states
carrying magnetic charge are very heavy at weak coupling and states carrying electric
charge are heavy at strong coupling, and viceversa. Therefore, one could imagine
that at strong coupling the rôles of electric (fundamental) and magnetic (solitonic)
sources are interchanged, and that the theory at strong coupling is a theory of light
monopoles. This idea was put forward by Montonen and Olive which suggested
that the Giorgi-Glashow model could have two completely equivalent descriptions,
related by a S-duality transformation, one in terms of electric sources and one in
terms of magnetic sources, the two being exchanged under S duality (which indeed
interchanges electric and magnetic couplings).
There are two non-trivial evidences in favor of such duality in the Giorgi-Glashow
model:

• The BPS mass formula (12.22) is S-duality invariant.

• One can explicitly show that there is no interactions between two monopoles,
while there is a non-vanishing interaction between a monopole and an anti-
monopole. If duality is correct, since upon S transformations monopoles and
gauge bosons are exchanged, the same should hold for the W + and W − bosons
in the Giorgi-Glashow model. This has been shown to be the case. Basically,
in the BPS limit the (massless) Higgs field contributes exactly the opposite
to the photon in the interactions between W ’s with equal charge, and exactly
the same to the photon in the interactions between W ’s with opposite charge.

339
These convincing evidences, however, are not enough to conclude that Montonen-
Olive duality is realized in the Giorgi-Glashow model. For one thing, as already
noticed, there is no guarantee that the (semi)classical BPS mass formula (12.22)
holds at quantum level (differently from theories with extended supersymmetry).
Second, another necessary condition for the duality to hold is that the W ± bosons
and monopoles carry the same spin. Both these two crucial requirements cannot be
verified in the Giorgi-Glashow model (and most likely are not met).
What about the supersymmetric version of this story?
The persistence of (12.22) at the quantum level ensured by the N = 2 super-
symmetry algebra, could suggest that exact S duality could be realized in N = 2
theories. However, the other necessary condition, namely that monopoles should
have the same quantum numbers of massive gauge bosons, does not hold. Mag-
netically charged states sit in hypermultiplets, which do not accomodate spin one
particles. In fact, we have seen that in N = 2 theories a quite different duality is
realized, which is not an exact duality but rather a electro-magnetic duality which
holds at the level of the IR effective theory. As we are going to discuss below, N = 4
SYM, instead, is believed to realize exact S duality.
There are several facts which suggest this to be plausible. We enumerate them
in turn.
First, we know from the representation of the supersymmetry algebra that in
N = 4 massive representations cannot be anything but BPS multiplets containing
spin one particles. This implies that monopoles and dyons sit in vector multiplets, as
gauge bosons do, differently from N = 2 SYM. In fact, in N = 4 SYM all physical
states (massive or massless) sit in BPS saturated vector multiplets, which is the
only possible supersymmetric representation in theories without gravity, see § 3.1
and 3.2.
Second, differently from N = 2, in N = 4 the BPS bound (12.22) does not only
hold true at quantum level, but non-renormalization theorems guarantee that the
quantities therein are classically exact, see lecture 6. Hence, the U (1) couplings en-
tering the effective Lagrangian (12.1) do not renormalize, i.e. τIJ are free parameters
and they are all proportional to the (non-abelian!) UV-coupling τ

τIJ = CIJ τ , (12.109)

where CIJ is the Cartan matrix of the gauge group G. This suggests that, unlike
for N = 2 theories, the electro-magnetic duality of the effective theory may propa-

340
gate all the way to the UV. Indeed, the Sp(2n, Z) transformations (12.34) - which
leave the low energy effective action form-invariant - contain transformations which
would act on the UV coupling τ , via eq. (12.109), as τ → τ + 1 and τ → −1/τ ,
which generate the group Sp(2, Z). This implies that theories with UV couplings
τ related by Sp(2, Z) transformations are physically equivalent. That is to say, not
only theories where electrically charged, respectively magnetically charged states are
the fundamental degrees of freedom are physically equivalent. All theories whose
fundamental degrees of freedom are dyonic states related to purely electric ones by
a Sp(2, Z) transformation
! !
aτ + b d −b a b
τ→ , (nm , ne ) → (nm , ne ) with ∈ Sp(2, Z) , (12.110)
cτ + d −c a c d

are. Notice how different this duality is with respect to the IR dualities discussed
previously, which act at the level of free IR effective U (1)’s theories. Here we are
claiming that different interacting theories, with different UV couplings and different
UV degrees of freedom, are physically equivalent!
Third, if Montonen-Olive duality is an exact symmetry, not only monopoles and
dyons should carry the same Lorentz representations as gauge bosons: the whole
spectrum of the theory should be duality invariant. In particular, given that massive
gauge bosons are BPS states with charges (nm , ne ) = ±(0, 1), there should also be
in the theory all BPS states which can be obtained acting on ±(0, 1) with Sp(2, Z)
transformations (one state for all relatively prime choice of electric and magnetic
charges). Several evidence and consistency checks were given showing this to be the
case for N = 4 SYM.
Finally, we cannot resist saying that the strongest evidence for Montonen-Olive
duality in N = 4 SYM comes, in fact, from string (and M -) theory. There, it
exists an intricate set of dualities between different string theories which implies, as
a by-product, Montonen-Olive duality of N = 4. The self-consistency of this web of
dualities has passed many tests and it is regarded as an independent indication for
the S duality of N = 4 SYM.

341
12.6 Exercises
1. Consider the three following parametrizations of the elliptic curve for pure
N = 2 SU (2) SYM

y 2 = (x − Λ2 )(x + Λ2 )(x − u)
1
y 2 = x2 (x − u) + Λ4 x
4
y 2 = (x2 − u)2 − Λ4

Show that they are physically equivalent. In particular, that the number and
the locations of the singularities on the u plane as well as the nature of the
particles becoming massless there, do not depend on which curve one considers.

References

[1] P. Argyres, Non-perturbative dynamics of four-dimensional supersymmetric

field theories, http://homepages.uc.edu/ argyrepc/cu661-gr-SUSY/fgilec.pdf

[2] S. Coleman, Aspects of Symmetry: Selected Erice Lectures, Chapter 6, Cam-

bridge University Press, 1985.

[3] N. Seiberg and E. Witten, Electric - magnetic duality, monopole condensation,

and confinement in N=2 supersymmetric Yang-Mills theory, Nucl. Phys. B 426
(1994) 19 Erratum: [Nucl. Phys. B 430 (1994) 485] [hep-th/9407087].
N. Seiberg and E. Witten, Monopoles, duality and chiral symmetry breaking in
N=2 supersymmetric QCD, Nucl. Phys. B 431 (1994) 484 [hep-th/9408099].

[4] E. D’Hoker and D. H. Phong, Lectures on supersymmetric Yang-Mills theory

and integrable systems, Chapters 2.10, 3 and 4, hep-th/9912271.

[5] M. E. Peskin, Duality in supersymmetric Yang-Mills theory, Chapters 5 and 6,

arXiv:hep-th/9702094.

[6] A. Bilal, Duality in N=2 SUSY SU(2) Yang-Mills theory: A Pedagogical intro-
duction to the work of Seiberg and Witten, Chapters 3 and 5, hep-th/9601007.

[7] L. Alvarez-Gaume and S. F. Hassan, Introduction to S duality in N=2 super-

symmetric gauge theories: A Pedagogical review of the work of Seiberg and
Witten, Chapter 3, Fortsch. Phys. 45 (1997) 159 [hep-th/9701069].

342
 [8] P. Di Vecchia, Duality in N=2, N=4 supersymmetric gauge theories, Chapters
12, 16 and 17, hep-th/9803026.

[9] Y. Tachikawa, N=2 supersymmetric dynamics for pedestrians, Lect. Notes Phys.
890 (2014) [arXiv:1312.2684 [hep-th]].

[10] P. C. Argyres and M. R. Douglas, New phenomena in SU(3) supersymmetric

gauge theory, Nucl. Phys. B 448 (1995), 93-126 [arXiv:hep-th/9505062 [hep-
th]].

[11] P. C. Argyres, M. R. Plesser, N. Seiberg and E. Witten, New N=2 super-

conformal field theories in four-dimensions, Nucl. Phys. B 461 (1996), 71-84
[arXiv:hep-th/9511154 [hep-th]].

343