Lecture notes on

Advanced Supersymmetry

MMathPhys, University of Oxford, HT2020

Cyril Closset

Contents

I Dynamics of 4d N = 1 gauge theories 4

1 Quantum aspects of 4d gauge theories 4

1.1 Anomalies for gauge and global symmetries . . . . . . . . . . . . . . 4
1.1.1 Global symmetry, background gauge fields and gauging . . . 4
1.1.2 Anomaly as an obstruction to gauging a global symmetry . . 5
1.1.3 Three types of anomalies . . . . . . . . . . . . . . . . . . . . 6
1.1.4 The ’t Hooft anomaly matching condition . . . . . . . . . . . 8
1.2 Instantons, θ angle and chiral anomalies . . . . . . . . . . . . . . . . 8
1.2.1 Instantons and θ angle: executive summary . . . . . . . . . . 9
1.2.2 θ-term and chiral anomaly . . . . . . . . . . . . . . . . . . . . 11

2 General aspects of supersymmetric gauge theories 12

2.1 Vacuum and beta functions . . . . . . . . . . . . . . . . . . . . . . . 12
2.1.1 Classical scalar potential and vacuum manifold . . . . . . . . 12
2.1.2 β-function and chiral anomalies . . . . . . . . . . . . . . . . . 14
2.2 Renormalisation of the holomorphic gauge coupling . . . . . . . . . . 16
2.3 The “exact” β-function . . . . . . . . . . . . . . . . . . . . . . . . . 17
2.3.1 Rescaling of the chiral superfields . . . . . . . . . . . . . . . . 18
2.3.2 Rescaling of the vector superfield . . . . . . . . . . . . . . . . 19
2.3.3 Looking for non-trivial fixed points . . . . . . . . . . . . . . . 19

1
2 Contents

3 SQCD: Lagrangian, symmetries and classical moduli space 20

3.1 Anomalies and anomaly-free R-symmetry . . . . . . . . . . . . . . . 21
3.2 The classical vacuum moduli space of SQCD and gauge-invariant
operators . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 22
3.2.1 The case Nf < Nc . . . . . . . . . . . . . . . . . . . . . . . . 23
3.2.2 The case Nf ≥ Nc . . . . . . . . . . . . . . . . . . . . . . . . 24
3.2.3 The case Nf = Nc . . . . . . . . . . . . . . . . . . . . . . . . 25
3.2.4 The case Nf = Nc + 1 . . . . . . . . . . . . . . . . . . . . . . 26
3.3 The IR phases of SQCD . . . . . . . . . . . . . . . . . . . . . . . . . 26
3.4 Infrared phases of gauge theories . . . . . . . . . . . . . . . . . . . . 26
3.5 Aspects of the quantum vacuum of SQCD . . . . . . . . . . . . . . . 27
3.5.1 Nf = 0: SYM theory . . . . . . . . . . . . . . . . . . . . . . . 27
3.5.2 0 < Nf < Nc : runaway supersymmetry breaking . . . . . . . 28
3.5.3 Are there low-energy σ-models for Nf ≥ Nc ? . . . . . . . . . 29
3.5.4 Nf = Nc : deformed moduli space . . . . . . . . . . . . . . . . 30
3.5.5 Nf = Nc + 1: A σ-model . . . . . . . . . . . . . . . . . . . . 31
3.5.6 Nf > Nc + 1: SCFTs, free theories, and duality . . . . . . . . 32

II Dynamics of 4d N = 2 gauge theories 35

4 Gauge theories with N = 2 supersymmetry 35

4.1 N = 2 supersymmetry and R-symmetry . . . . . . . . . . . . . . . . 36
4.1.1 The massless vector multiplet . . . . . . . . . . . . . . . . . . 37
4.1.2 Massive multiplets and BPS bound . . . . . . . . . . . . . . . 37
4.2 The classical Coulomb branch . . . . . . . . . . . . . . . . . . . . . . 38
4.3 ’t Hooft-Polyakov monopoles on the Coulomb branch . . . . . . . . . 40

5 Theories of abelian vector multiplets 42

5.1 Stating the problem: “solving N =2 SYM”? . . . . . . . . . . . . . . 42
5.2 U (1) gauge fields and electro-magnetic duality . . . . . . . . . . . . . 42
5.3 The N =2 prepotential for an abelian vector multiplet . . . . . . . . 42
5.4 Semi-classical dynamics: anomalies and one-loop running . . . . . . 42
5.5 Abelian vector multiplet coupled to hypermultiplet . . . . . . . . . . 42

6 The Seiberg-Witten solution for pure SU (2) 42

Contents 3

Introduction

In this set of lectures, we will build on the formalism developed in the Supersym-
metry and Supergravity lectures. There, we learned how to write down classical
field-theory Lagrangians with supersymmetry, and we discussed important aspects
of the quantum theory in the simple case of theories with chiral multiplets only.
In the present lectures, we will explore aspects of the quantum dynamics of
supersymmetric gauge theories in 4d. These are close cousins to the gauge theories
that we use in Particle Physics. Supersymmetry leads to more technical countrol,
leading to many exact results that we could only dream off in ‘real-world’ QFT with-
out supersymmetry. It is thus of great theoretical interest to study these models in
details, for instance as ‘toy models’ for the much harder problem of understanding,
say, QCD at strong coupling from first principles.

These short lectures will have two parts:

In part I, we will study the dynamics supersymmetric gauge theories with N = 1
supersymmetry, The main question is simple to state: given an asymptotically-
free gauge theory, we have a well-defined, perturbative, weakly-coupled QFT in the
ultraviolet (UV), while the gauge coupling becomes large towards the infrared (IR).
Can we say anything useful about the IR physics? We will focus our attention on
a supersymmetric version of QCD, called N = 1 SQCD.
In part II, we will consider gauge theories with N = 2 supersymmetry. The extra
supercharges give us additional analytic control, and we will see that, in some
sense, one can entirely ‘solve’ for the IR physics in terms of the UV data. This
is the celebrated Seiberg-Witten solution of N = 2 gauge theories. We will only
be able to cover some basics, but it will hopefully be enough to display the great
beauty of the subject, and to give you the motivation to go and learn more about
it on your own.
Lecture notes last updated on: April 22, 2020.

References and further reading:

• Again, the lecture notes by Argyres: http://homepages.uc.edu/~argyrepc/
cu661-gr-SUSY/index.html.
• The classic lectures by Intriligator and Seiberg [1] on 4d N = 1 dynamics.
• The lectures by Tachikawa on N = 2 dynamics [2], which also covers some
more modern aspects.
• The original papers by Seiberg [3, 4] (N = 1) and Seiberg and Witten [5]
(N = 2). (There hasn’t been quite anything like the year 1994 for the study
of supersymmetric QFT...)
4 1 Quantum aspects of 4d gauge theories

Part I
Dynamics of 4d N = 1 gauge theories
1 Quantum aspects of 4d gauge theories
In this section, we discuss some general aspects of (gauge) anomalies in QFT, as
well as the role of the θ angle in gauge theories. (This is by no means self-contained,
but should amply suffice for our purpose in these lectures.)

1.1 Anomalies for gauge and global symmetries

Let us first make some general comments about quantum anomalies in gauge and
global symmetries.

1.1.1 Global symmetry, background gauge fields and gauging

Consider some fermions ψ and bosons φ that transform into some representation
Rψ and Rφ , respectively, of some global symmetry group G.e Then, by Noether
µ
theorem, we have some conserved currents ja , with the index a = 1, · · · , dimG e
running over the generator of the Lie algebra e
g.
Whenever we have such conserved currents, we can introduce some sources for
them:
L [A] = L0 + Aaµ jaµ + · · · . (1.1)
Here, L0 is the Lagrangian of the theory with a global symmetry, Aaµ is the
source for the current operator jaµ (x), and the ellipsis denotes higher-order terms
in the sources. The source is nothing but a background gauge field—that is, a
non-dynamical vector field Aµ , which is introduced to keep track of the conserved
current. The path integral in terms of the sources takes the form:
Z  Z 
4
Z[A] = [Dφ][Dψ] exp i d xL [A] , (1.2)

for any fixed background Aµ .

Gauging a global symmetry. Given a global symmetry G, e a natural operation

in QFT is to gauge it. In path-integral language, gauging a global symmetry means
that we first introduce background gauge fields as in (1.2), and then integrate over
all possible gauge fields:
Z
Z = [DA] eiSYM [A]+iStop [A] Z[A]
Z  Z  (1.3)
= [DA][Dφ][Dψ] exp i d x (L [A] + LYM [A] + Ltop [A]) .
4
1.1 Anomalies for gauge and global symmetries 5

Figure 1: The triangle one-loop diagram that determines the gauge anomaly in four
dimensions, with fermions running in the loop. The wiggly lines correspond to the
external gauge fields; equivalently, one inserts a current jaµ at each vertex.

Here, we weighted each gauge-field configuration by the Yang-Mills action (the

standard kinetic) term, as well as by the topological term, which we will discuss
further in the next subsection:
1 θ
LYM [A] = − tr Fµν F µν , Ltop [A] = − tr (µνρσ Fµν Fρσ ) . (1.4)
4g 64π 2

1.1.2 Anomaly as an obstruction to gauging a global symmetry

The gauging of a global symmetry can only be done consistently if the functional
(1.2) is itself gauge invariant—infinitesimally:

δα log Z[A] = 0 , (1.5)

for δα Aµ = Dµ α. The symmetry G

e has an anomaly if and only if:

δα log Z[A] 6= 0 . (1.6)

This can be taken as our definition of a gauge anomaly. In the presence of an

anomaly for G e in the original theory L0 , the global symmetry cannot be gauged,
and the corresponding gauge theory (1.3) does not exist as a consistent QFT.
While the classical Lagrangian L [A] itself is gauge invariant by construction,
the path-integral measure in (1.2) might not always be—this is precisely the ori-
gin of those “quantum anomalies.” In renormalisable gauge theories with scalar
and fermion matter fields only, anomalies arise exclusively from chiral fermions in
complex representations of the gauge group (i.e. R 6= R̄).
In some appropriate regularisation scheme, the anomaly takes the explicit form:
Z
iAabc
δα log Z[A] = − d4 x αa Fµν
b c µνρσ
Fρσ  , (1.7)
64π 2
6 1 Quantum aspects of 4d gauge theories

(R )
schematically, with α(x) = αa (x)Ta ψ the gauge-transformation functions acting
on the fermions; the sum over the e g-indices a, b, c is understood. The anomaly
coefficients Aabc are given by:
 
(R ) (R ) (R )
Aabc = tr Ta ψ {Tb ψ , Tc ψ } . (1.8)

These are also called cubic anomalies, since they are cubic in the “charges” under
G.
e Physically, they can be extracted from the three-point function of the currents
jaµ in the original theory with Lagrangian L0 :

jaµ (p)jbµ (q)jcµ (−p − q) ∼ Aabc . (1.9)

The anomaly coefficient (1.8) is entirely determined by the one-loop contribution

to these observables, which are triangle diagrams of the type depicted in Figure 1.
We should also mention the existence of some more subtle quantum anomalies
in 4d, the linear anomalies:
(Rψ )
Aa = tr(Ta ). (1.10)

They are “mixed anomalies” between gauge invariance and diffeomorphism invari-
ance. They are often called “gravitational anomalies,” because they contribute to
the right-hand-side of (1.7) in the presence of a non-trivial metric gµν (with non-
zero curvature)—they arise at one-loop from the same triangle diagram as in Fig 1,
but with one gauge field and two gravitons for the external legs. We see from (1.10)
that linear anomalies can only be non-zero for abelian symmetries, the U (1) factors
inside G.
e

1.1.3 Three types of anomalies

In general, the symmetry group G e of the theory L0 might be a product of many

simple groups and U (1) factors. Let us consider:

e = GF × G .
G (1.11)

Here, we would like to gauge the factor G ⊂ G,

e while GF will remain as a global
symmetry (often called a “flavor symmetry”) of the gauge theory with gauge group
G. The cubic anomalies (1.8) can then take the schematic form:

Tr(GGG) , Tr(GGGF ) , Tr(GGF GF ) , Tr(GF GF GF ) , (1.12)

while we have Tr(G) and Tr(GF ) for the linear anomalies. There are thus, in fact,
three types of anomalies that concern us when G is gauged, from the lethal to the
innocuous:
1.1 Anomalies for gauge and global symmetries 7

(i) Gauge anomalies. A gauge anomaly is an anomaly of the form:

Tr(GGG) , (1.13)

for three currents of the gauge group G. If it were non-zero, the theory would be
inconsistent quantum-mechanically. Therefore, we have the anomaly-free condi-
tion:
Aabc = 0 , (1.14)
where a, b, c runs over the generators of G only. For a simple Lie algebra g, we
have:
1
Aabc = A R(Rψ dabc = 0 ,


(1.15)
2
which is a non-trivial condition only for SU (N ) with N > 2. (See below equation
(7.28) in the Susy&Sugra lectures.) Note also that, if the gauge group is a product:
Y Y
G= Gi × U (1)k ,
i j

with Gi some simple factors (as in the Standard Model, for instance), we also have
non-trivial constraints:

tr(Gi Gi U (1)k ) = 0 , tr(U (1)k U (1)l U (1)m ) = 0 , (1.16)

which translate to:

qk [ψ]T (Rψ
X X
i )=0 , qk [ψ]ql [ψ]qm [ψ] = 0 , (1.17)
ψ ψ

where T (Rψ ψ
i ) is the quadratic index of the representation Ri of Gi under which ψ
transforms, and qk [ψ] denotes the U (1)k charges of the fermions.

(ii) Anomalous global symmetries. We could also have so-called “mixed anoma-
lies:”
Tr(GGGF ) , Tr(GGF GF ) , (1.18)
between the gauge-symmetry and the global-symmetry currents. If G is semi-
simple, only Tr(GGGF ) can be non-trivial. Let a, b, · · · run over the generators Ta
of the gauge group G, and let α, β, · · · run over the generators Tα of GF . Then,
any anomaly:
Aabα 6= 0 , or Aaαβ 6= 0 , (1.19)
signals that the currents jαµ , or jαµ and jβµ , part of the naive symmetry GF , are
actually not conserved in the gauge theory. This is an anomalous global symmetry,
which is then not a symmetry of the quantum system.
We will mostly focus on G = SU (N ) and GF ⊃ U (1)A , with a particular
flavor symmetry U (1)A that can be anomalous. This non-conservation of U (1)A is
generally called a chiral anomaly.
8 1 Quantum aspects of 4d gauge theories

(iii) ’t Hooft anomalies. Finally, consider an anomaly-free theory with gauge

group G and a non-anomalous global symmetry group GF (with generators Tα ).
In general, the anomalies involving only the global symmetry currents can be (and
generally are) non-zero:
Aαβγ 6= 0 . (1.20)
These are called ’t Hooft anomalies. They are an obstruction to gauging the sym-
metry GF —per our general definition of anomalies—, but they are otherwise in-
nocuous. In the absence of GF background gauge fields, the currents jαµ are still
conserved. There is a mild modification of the Ward identities for GF that follows
from the anomalous variations (1.7), but does not change their essential meaning—
the symmetry GF still implies all the usual selection rules, in particular.

1.1.4 The ’t Hooft anomaly matching condition

While ’t Hooft anomalies are innocuous, they carry some interesting information
about the QFT. The reason is that they are invariant under RG flow, in the fol-
lowing sense. Let us consider some UV theory with some global symmetry GF
and a set of ’t Hooft anomalies AGF for that global symmetry. Then, consider an
RG flow starting from the UV theory, preserving the symmetry GF , that flows to
some effective field theory in the infrared. That IR effective theory could look very
different from the UV description, since the RG flow does not need to be perturba-
tive. Nonetheless, we claim that the ’t Hooft anomalies in the UV and the IR must
match [6]:
AGF (TU V ) = AGF (TIR ) . (1.21)
This is because we can always add some free fermions in the UV theory, TU V to
saturate the anomalies—that is, we can add some free fermions that transform in
some representation of GF exactly so that the anomaly vanishes in the enlarged
theory TU V ⊗ Tfree . We can then consistently gauge the enlarged theory TU V ⊗
Tfree with some arbitrarily weakly coupled gauge field. Upon RG flow the free-
free fermion sectors then behaves as a spectator, and therefore we obtain another
consistent theory TIR ⊗ Tfree very weakly coupled to GF -gauge fields in the IR. It
must now be true that the GF anomaly associated to TIR exactly cancels the one
associated to Tfree . Therefore, (1.21) must hold.
This ’t Hooft anomaly matching condition provides a strong constraint on RG
flows, even at strong coupling, as we will see in some supersymmetric examples.

1.2 Instantons, θ angle and chiral anomalies

The action of a Yang-Mills theory contains a so-called topological term:
Z
θ
Stop = − d4 x tr (µνρσ Fµν Fρσ ) , (1.22)
64π 2
It is also known as the instanton density, or Pontryagin density.
1.2 Instantons, θ angle and chiral anomalies 9

1.2.1 Instantons and θ angle: executive summary

The meaning of this topological term is best described in Euclidean signature. Let
us also replace space-time by a general Riemannian four-manifold M4 :

4 √
Z Z
θ µνρσ θ
Stop = − i d x g tr ( F F
µν ρσ ) = − i tr F ∧ F , (1.23)
64π 2 M4 16π 2 M4

where we used the form notation, with F = 21 Fµν dxµ ∧ dxν in local coordinates. It
is clear from the right-hand-side of (1.23) that Stop is independent of the metric on
M4 . In fact, the integrand is also a total derivative:
Z Z  
2i
tr F ∧ F = d tr A ∧ dA − A ∧ A ∧ A , (1.24)
M4 M4 3

and thus we naively expect the action to vanish.

However, the topological action can be non-trivial in the presence of a non-trivial
gauge-field configuration, which are called instantons. Indeed, consider M4 = R4 .
A physically sensible gauge field does not need to vanish at infinity; instead, it
should only be pure gauge:

lim Aµ = ig(x)∂µ g(x)−1 , (1.25)

|x|→∞

3 , at infinity:
with g(x) an gauge transformation on the three-sphere, S∞

g(x) 3
: S3 → G . (1.26)
S∞

This map can have non-trivial winding number k ∈ Z, corresponding to an element

of the homotopy group π3 (G). One can show that the topological action precisely
computes the winding number:
Z Z  
2i
tr F ∧ F = tr A ∧ A − A ∧ A ∧ A = 16π 2 k . (1.27)
M4 3
S∞ 3
R
Mathematically, the quantity M4 F ∧ F is a so-called characteristic class, known
as the Pontraygin class, which captures some of the non-trivial topology of a non-
trivialisable G-bundle over a four-manifold M4 —a non-trivial gauge-field configu-
ration on M4 . It is a non-trivial mathematical fact that this topological invariant
is always an integer: Z
1
tr F ∧ F = k ∈ Z . (1.28)
16π 2 M4
Let us insist on this amazing fact: while the action S[ϕ] in general depends on the
detailed value of the fields ϕ(x), the topological term (1.28) gives an integer for any
gauge field Aµ (x). Thus, the quantity eiStop = eiθk factors out of the path integral
for each “topological sector” at fixed k.
10 1 Quantum aspects of 4d gauge theories

Given a gauge field Aµ with field-strength Fµν , define the dual (or “magnetic”)
field-strength as:
1
Feµν ≡ µνρσ F ρσ . (1.29)
2
Definition: An (anti)-instanton is a gauge field configuration on M4 which is
(anti)-self dual:
Fµν = ±Feµν , (1.30)
with the + and − signs for the instanton and anti-instanton, respectively. In QFT,
we consider, in particular, M4 = R4 , which can be compactified to a sphere S 4 .
One physical significance of instantons is that they are the non-trivial classical
saddles of the Yang-Mills action. Indeed, we have:
Z Z Z
1 2 2
|F + F | = |F | ± F ∧ F ≥ 0 ,
e (1.31)
2
schematically, which implies that the YM action in always larger or equal to the
topological action, the sense that:
Z Z
|F |2 ≥ F ∧ F = 16π 2 |k| , (1.32)

with the inequality saturated for the (anti)-instanton configuration. We have k > 0
for an instanton and k < 0 for an anti-instanton. On any (anti)-instanton back-
ground, we have:

4 √ 4 √ µνρσ 8π 2
Z Z
1 µν 1
d x gFµν F = ± d x g Fµν Fρσ = |k| . (1.33)
4g 2 8g 2 g2
(k)
Thus, a k-instanton gauge field, Aµ , is weighted by a numerical factor:
8π 2 |k|
−SYM [A(k) ] −
e =e g2 , (1.34)

in the path integral. Note the similarlity to the (one-loop) strong-coupling scale:
8π 2
−
Λ = µe b0 g 2 (µ) , (1.35)

which we mentioned in previous lectures. Each topological sector, for each k ∈ Z—

that is, the set of all gauge fields with a non-zero instanton number (1.28)—gives a
non-perturbative contribution to the path integral. For g 2 very small and k 6= 0, the
instanton factor (1.34) is extremely small, and can be neglected for most purposes
(hence the name, non-perturbative 1 ). However, as g 2 grows under RG flow (in
1
Note also that the function:
− 12
f (g) = e g

is non-differentiable at g = 0. The apparence of such a factor is always the hallmark of a non-

perturbative correction.
1.2 Instantons, θ angle and chiral anomalies 11

the IR, for an asymptotically-free gauge theory), the instantons will start giving
increasingly important contributions to the dynamics. At strong coupling, we would
even expect them to dominate.
While the trivial saddle Aµ = 0 (corresponding to k = 0) is the starting point
of ordinary perturbation theory, one can (and should) do a similar perturbative
expansion around all saddles. The computation of any given observable will a
priori receive contributions from all saddles (plus perturbative fluctuations). For a
simple, simply-connected gauge group, such as SU (N ) (for simplicity), we have:
Stop = −iθk , (1.36)
in Euclidean signature, in the presence of a k-instanton. Then, the Yang-Mills path
integral takes the schematic form:
8π 2 |k|
Z
iθk − g2
X
iSYM +iStop
Z = [DAµ ] e = e e Zk , (1.37)
k∈Z

where Zk is the perturbative contribution of each topological sector. Note that,

since a shift θ → θ + 2π does not modify the exponentiated action, θ is indeed an
angle, with period 2π. We can also write (1.37) as:
X X
Z = Z0 + q k Zk + q̄ −k Zk , q ≡ e2πiτ , (1.38)
k>0 k<0

with the holomorphic gauge coupling τ defined in previous lectures:

θ 4πi
τ≡ + 2 . (1.39)
2π g

1.2.2 θ-term and chiral anomaly

Consider a chiral anomaly—namely, a simple gauge group G and an anomalous
flavor current U (1)A , the anomalous chiral symmetry. Denoting by qA the U (1)A
charge, we have the anomaly coefficients:
X X
AAab = 2 qA [ψ] Ta [Rψ ]Tb [Rψ ] = δab qA [ψ] T (Rψ ) , (1.40)
ψ ψ

where the sum is over all the chiral fermions ψ with non-zero U (1)A charges q[ψ].
Each such ψ sits in a representation Rψ of the gauge group G, with quadratic index
T (Rψ ). Let us define, then, the chiral anomaly coefficient:
X
AU (1)A ≡ qA [ψ] T (Rψ ) . (1.41)
ψ

Under a chiral symmetry transformation (with symmetry parameter α ∈ R, a

constant), we have:
iα AU (1)A
Z
δα log Z = − d4 x tr (µνρσ Fµν Fρσ ) , (1.42)
64π 2
12 2 General aspects of supersymmetric gauge theories

as a special case of (1.7). But this is equivalent to a shift of the θ angle:

θ → θ + α AU (1)A . (1.43)

Thus, even though the axial symmetry U (1)A does not exists in the quantum theory
if AU (1)A 6= 0, one can keep track of it (and extract physical consequences) by
assigning the U (1)A transformation (1.43) to θ, viewing θ itself as a background
field.

2 General aspects of supersymmetric gauge theories

In the following, we would like to study some aspects of renormalisable gauge
theories with N = 1 supersymmetry in four dimensions.
Let us take the gauge group G to be a simple compact Lie group, and consider
some matter field in chiral multiplet, Φ, in some (generally reducible) representa-
tion R. The full Lagrangian for the vector and chiral multiplets can be written
compactly, in superspace, as:
Z Z Z
−2V τ τ̄
L = 2 2
d θd θ̄ Φ̄e Φ− 2
d θ tr(WW) + d2 θ̄ tr(W̄ W̄)
16πi 16πi
Z Z (2.1)
2 2
+ d θW (Φ) + d θ̄W̄ (Φ̄) .

Here, W (Φ) is some gauge invariant holomorphic polynomial in Φ (which we take

to be at most cubic in the renormalisable theory).
Note that the Lagrangian of any renormalisable 4d N = 1 supersymmetric gauge
theory is fully determined by the data of:
θ 4πi
• The gauge group G with gauge coupling(s) τ = 2π + g2
.

• The representation R for the chiral multiplets.

• The superpotential W (Φ).

All the various interactions terms are then determined by the combination of gauge
invariance and supersymmetry, as well as by W .

2.1 Vacuum and beta functions

2.1.1 Classical scalar potential and vacuum manifold
By looking at the classical Lagrangian in components, it is easy to study the classical
scalar potential of the gauge theory. The adjoint-valued auxiliary field D enters as:

1 2
L ⊃ D − φ̄Dφ , (2.2)
2g 2
2.1 Vacuum and beta functions 13

in the WZ gauge. The equations of motions for the auxiliary fields D = Da Ta give:

Da = g 2 φ̄Ta(R) φ , a = 1, · · · , dim(G) . (2.3)

Integrating out D, we then find the scalar potential:

2 dim(G)
X ∂W g 2 X  (R) 2
V0 = + φ̄Ta φ . (2.4)
∂φ 2
φ a=1

The first term is the contribution from the superpotential, which we discussed in
previous sections, while the second term can be viewed as a contribution from the
gauge interactions themselves. The real operators:

µa (φ, φ̄) ≡ φ̄Ta(R) φ (2.5)

are often called the “moment map operators.”

Since the scalar potential is again a sum of perfect squares, the classical vacuum
equations of a supersymmetric gauge theory are:

∂φ W = 0 , ∀φ , µa (φ, φ̄) = 0 , ∀a . (2.6)

Any two solutions to (2.6) related by a (constant) gauge transformations are physi-
cally equivalent. So, we introduce the equivalence relation on the space of constant
field values:
a T (R)
φ0 ∼ φ if ∃ (αa ) ∈ Rdim(G) such that φ0 = eiα a
φ. (2.7)

The constant values of the scalar field φ ∈ Φ span the vector space:

VR ∼
= Cn , n ≡ dim(R) , (2.8)

on which the representation R acts. Then, the vacuum manifold of the gauge theory
takes the general form:

M = {φ ∈ VR | ∂φ W = 0 , µa = 0}/G , (2.9)

where the quotient by the gauge group corresponds to the equivalence relation
(2.7). In our discussion of theories with only chiral multiplets, we saw that the
vacuum moduli space was a purely algebraic object—in particular, everything was
holomorphic in φ. This is apparently not the case in a gauge theory, since the
formula (2.9) is non-holomorphic in two ways: the moment maps µa are real, and
the gauge equivalence (2.7) is in terms of real gauge parameters αa .
Nonetheless, there is a simple-looking (although by no mean obvious) way to
rewrite (2.9) more algebraically. It turns out that imposing the vanishing of the
14 2 General aspects of supersymmetric gauge theories

moment maps, µa = 0, and then dividing by G, is equivalent to dividing by the

complexified gauge group:

M = {φ ∈ VR | ∂φ W = 0}/GC . (2.10)

In this approach, we are considering the space of complexified gauge orbits (or,
more precisely, their closure), under the GC action:
a T (R)
φ0 ∼ φ if ∃ (ω a ) ∈ Cdim(G) such that φ0 = eiω a
φ. (2.11)

The fact that the two approaches (2.9) and (2.10) reproduce the same moduli space
was shown explicitly in [7]. 2
Conceptually, this was to be expected: the fact that we only divide by real gauge
transformations in (2.9) is an artefact of the WZ gauge. The supersymmetric gauge
transformations on chiral superfields,

δΩ Φ = eiΩ Φ , (2.12)

are really GC -valued gauge transformations. More generally, the F -term contribu-
tions to the Lagrangian of any supersymmetric gauge theory are invariant under
the complexified gauge group GC , while the total Lagrangian (in particular, the
D-term kinetic term for matter fields) is only G-invariant.
Finally, it is non-obvious but nonetheless true that the vacuum moduli space
M of a gauge-theory is also a Kähler manifold, just like in the case without gauge
fields.

2.1.2 β-function and chiral anomalies

The one-loop β function of the YM coupling, in our supersymmetric gauge theory,
is given by:  
1 b0 3 1
β 2
= 2 , b0 = T (adj) − T (R) . (2.13)
g 8π 2 2
Here, we simply specialised the general result (see eq.(7.21) in the Susy&Sugra lec-
tures) to the supersymmetric matter content, taking into account that the gaugino
transforms in the adjoint representation.

Anomalous symmetries (with W = 0). Consider the axial symmetry U (1)A

that gives a charge 1 to all chiral multiplets. This is a symmetry of the theory
without superpotential (W = 0), which is however anomalous, since:

AU (1)A = T (R) . (2.14)

2
Mathematically, it is a non-trivial equivalence between Kähler quotients (corresponding to
(2.9)) and Geometric Invariant Theory (GIT) quotients (corresponding to (2.10)).
2.1 Vacuum and beta functions 15

More generally, we should decompose R into irreducible representations:

R = ⊕i Ri , (2.15)
and define the symmetry U (1)i , which charge qi , that acts only on Φi (in the
representation Ri ) with charge 1:
qi [Φj ] = δij . (2.16)
The symmetry is anomalous, with chiral anomaly Ai = T (Ri ), which acts as:
Φj → eiδij α Φj , θ → θ + T (Ri ) α , (2.17)
including the anomalous shift of the θ angle. Another symmetry which is anomalous
is the “reference” R-charge R0 with charges:
R0 [V ] = 0 , R0 [Φi ] = 0 ⇒ R0 [λ] = 1 , R0 [ψi ] = −1 . (2.18)
Any mixing of an R-symmetry with a non-R (flavor) symmetry gives another
R-symmetry. Let us introduce a generic R-symmetry:
X
R = R0 + ri qi , ri ∈ R , (2.19)
i

where the parameters ri are the R-charges of the chiral multiplets, R[Φi ] = ri (so
the corresponding chiral fermions have R[ψi ] = ri − 1). One can then often find a
non-anomalous R-symmetry, by choosing the R-charges ri such that:
X
AU (1)R = T (adj) + (ri − 1)T (Ri ) = 0 . (2.20)
i

A digression: The reader might wonder about the fact that the same quadratic
indices of the gauge representations appear in the expression for the YM β function
(2.13) and for the chiral anomalies. In fact, one can easily define an R-symmetry:
2 2X
Rc ≡ R0 + A = R0 + qi , (2.21)
3 3
i

which is such that the chiral anomaly is exactly proportional to the β function:
2
AU (1)Rc = b0 . (2.22)
3
The R-charge Rc assigns rc = 23 to all chiral multiplets, which is compatible with
a classically-marginal superpotential, W = Φ3 (schematically). In fact, in the far
UV, such a super-Yang-Mills (SYM) theory is classically conformal (in particular,
scale invariant), and the R-charge combines with the supersymmetry current and
the energy-momentum tensor into a larger algebraic structure (known as N = 1
superconformal multiplet). Quantum corrections break both conformal invariance
(giving the running of g 2 , which is a “quantum anomaly” of scale invariance) and
the R-symmetry (through the chiral anomaly), but supersymmetry relates these
two “quantum anomalies” exactly as in (2.22).
16 2 General aspects of supersymmetric gauge theories

2.2 Renormalisation of the holomorphic gauge coupling

In a supersymmetric gauge theory, we now have a “generalised superpotential:”

τ (µ0 )
Wµ0 (W, Φ, τ, λ) = − tr W α Wα + W (Φ, λ; µ0 ) . (2.23)
16πi
Here, λ denote the ordinary superpotential coupling. This F -term must depend
holomorphically on both τ and λ, which restricts the possible quantum corrections
to W, at least in some appropriate “holomorphic scheme.” We also introduced the
UV scale µ0 explicitly in (2.23).
Perturbatively in g 2 , the holomorphy in τ gives some very severe restriction
when combined with the fact that θ is an angle with period 2π, which means that
the theory is invariant under:
τ ∼τ +1 . (2.24)
This implies that τ can only appear linearly in Wµ , at any scale µ, and only in the
precise form:
τ (µ)
Wµ ⊃ − tr W α Wα . (2.25)
16πi
Indeed, in that case θ multiplies the topological term as in (1.23). Since it is a
topological invariant, it cannot depend on µ at all. The one-loop running of the
holomorphic coupling τ is given by:

b0 µ
τ (µ) = τ (µ0 ) − log . (2.26)
2πi µ0

It is exact in perturbation theory. This is because any higher-order terms in β(τ )

would be given in terms of Im(τ ) = 4π g2
, which is incompatible with holomorphy. 3
One can also rule out other perturbative corrections to the superpotential, like in
the case without gauge interactions. This leaves the possibility of non-perturbative
corrections.
It is useful to introduce a complexified dynamically-generated scale, generalizing
(1.35), defined as:
2πiτ (µ0 )
i bθ
Λ=e 0 |Λ| = µ0 e b0 . (2.27)
Then, we have:
Λb0 = µb0 e2πiτ (µ) , (2.28)
which is a well-defined quantity, invariant under the shift θ ∼ θ + 2π. Any non-
perturbative effects would appear as:
∞
τ (µ) X
Wµ = − tr W α Wα + W (Φ, λ; µ) + Λb0 k gn (Φ, λ, · · · ) , (2.29)
16πi
k=1
3
One could also expect perturbative corrections in the superpotential couplings λ in β(τ ), but
that can be ruled out by considering the weak-coupling limit g 2 → 0.
2.3 The “exact” β-function 17

since the theory should be regular in the limit Λ → 0. Note the expression:

b0 Λ
τ (µ) = log . (2.30)
2πi µ

One could still constraint more carefully the form of the possible quantum cor-
rections to W. This sort of analysis, however, is only reliable at weak coupling.
As we RG flow from an asymptotically-free theory in the UV toward µ ∼ |Λ|, the
theory become strongly coupled and we need new methods to explore the infrared
physics. As we will see in a particular example (SQCD), supersymmetry and gen-
eral symmetry arguments sometimes are exceptionally powerful in order to “guess”
what the infrared physics is.

2.3 The “exact” β-function

We saw that, at the level of the F -terms, the holomorphic gauge coupling τ is
exact at one-loop (up to, possibly, non-perturbative corrections), and is given by
(2.30). This is true in the holomorphic scheme—that is, when we choose to preserve
the holomorphy of the F-terms. On the other hand, the D-terms in (2.1) are
renormalised non-trivially, with:
Z
SD-term,µ = d2 θd2 θ̄ ZΦ Φ̄e−2V Φ + · · · ,


(2.31)

where the ellipsis denotes contributions from higher-dimensional operators, and the
wave function renormalisation factor depends on all the coupling constants:

ZΦ = ZΦ (g, |λ|; µ) , (2.32)

and can be computed, in principle, at any order in perturbation theory. Moreover,

recall that the physical gauge coupling, is the one obtained by rescaling:

Aµ → gc Aµ , (2.33)

so that positive powers g appears in interactions vertices (including through the

(c)
covariant derivative Dµ = ∂µ − igc Aµ ). Here, for the moment, we write gc for this
‘physical coupling’ to distinguish it from g that appears holomorphically in τ . In
the holomorphic scheme, we have:
Z Z  
−2V iθ 1
Lµ = d θd θ̄ ZΦ (µ)Φ̄e
2 2
Φ+ d θ 2
− W 2 + h.c. , (2.34)
32π 2 4g 2 (µ)
schematically. Here, we set W = 0 for simplicity of notation. Note also that there
is really one distinct wavefunction renormalisation factor ZΦi for each irreducible
gauge representation Ri . We would like to define the “physical” fields:
p 1
ΦiR = ZΦi Φi , Vc = V . (2.35)
gc
18 2 General aspects of supersymmetric gauge theories

As we did for theories of chiral multiplets. It turns out, however, that these field
redefinitions are anomalous in the presence of the gauge interactions—that is, the
change of variable in the path integral gives a non-trivial Jacobian [8]. Heuristi-
cally, this can be understood as follows. The symmetry group of the F -terms is
complexified—we mentioned this fact before for the gauge group G, but that is true
of any global symmetries as well. The rescaling (2.35), in particular, can be under-
stood as a complexified “chiral rotation.” Since that chiral symmetry is anomalous,
the rotation shifts the θ angle by an imaginary amount. This, effectively, shift g12
by a real quantity.

2.3.1 Rescaling of the chiral superfields

It is useful to break down the field redefinition (2.35) in two steps. First, consider:
p
ΦiR = ZΦi Φi , (2.36)
without touching V . This change of variable can be achieved by some complexified
chiral rotations,
i
ΦiR = eiαi Φi , αi = − log ZΦi , (2.37)
2
acting on each Φi independently. Then, according to (2.17), this gives a shift to
the θ angle by:
X iX
θ → θ0 = θ + αi T (Ri ) = θ − T (Ri ) log ZΦi . (2.38)
2
i i

We then find the F -term Lagrangian:

Z  
2 iθ 1
d θ − 2 W α Wα , (2.39)
32π 2 4gR
with:
1 1 1 X
2 = − T (Ri ) log ZΦi . (2.40)
gR g 2 16π 2
i
2 has a β-function:
This effective coupling gR
b0 + 21 i T (Ri )γφi
  P
1
β 2 =
gR 8π 2
(2.41)
1  X 
= 3T (adj) − T (Ri )(1 − γ φ i) ,
16π 2
i

where we made use of the definition:

∂
γφ = −µ log Zφ , (2.42)
∂µ
for the anomalous dimensions γφ of Φ. This receives contributions from every loop
order in perturbation theory, but only through γφ , similarly to our discussion of
the ‘physical superpotential couplings’ in previous lectures.
2.3 The “exact” β-function 19

2.3.2 Rescaling of the vector superfield

Now, start from (2.39) and introduce the canonically-normalised vector multiplet:

V = gc Vc , (2.43)

to obtain the canonically normalised gauge field. This change of variables also has
a non-trivial Jacobian, which gives [8]:
Z  
2 iθ 1 T (adj)
d θ − 2 + log(gc ) W α Wα . (2.44)
32π 2 4gR 32π 2

Then, equating the real coefficient inside the parenthesis with the canonically-
normalised coupling − 4g12 , we get:
c

 
1 1 T (adj) 1
= 2 + log , (2.45)
gc gR 16π 2 gc2

which gives gc in terms of gR , which is itself defined in (2.40) in terms of the

“holomorphic coupling” g. In particular, we find the β-function:

1 P
b0 + i T (Ri )γφi
 
1 2
β = 1 . (2.46)
gc2 8π 2 − 2 T (adj)gc
2

Note that the denominator is the same as in (2.41). This is the famous NSVZ
β-function, which was first derived by completely different methods [9]. It is some-
times called “the exact β-function,” in the sense that it depends only on the anoma-
lous dimensions of the fields. Note that the anomalous dimensions themselves de-
pend on gc2 (and on any other superpotential couplings), so it just tells us that the
β function is known is we know the exact anomalous dimensions. The latter depend
on the details of the Kähler potential, and thus we cannot have exact formulas for
them. This is thus similar to the “exact” β-function for the superpotential coupling
that we discussed previously, 4 namely:

eR ) = − 3 +
X 1

e
β(λ 1 + γ φi d i λR . (2.47)
2
i

2.3.3 Looking for non-trivial fixed points

The physical significance of the denominator in (2.46) is not entirely clear, because
perturbation theory becomes unreliable before the denominator can have any im-
portant effect on the RG running. At the level of the present discussion, we should
simply view the NSVZ β function as a beautiful example of supersymmetry leading
to huge simplifications in the analysis of perturbative RG flows.
4
See the “Supersymmetry and Supergravity” HT2020 lectures, section 6.3.
20 3 SQCD: Lagrangian, symmetries and classical moduli space

One classic use that has been made of these “exact” results is to look for
perturbatively-exact fixed points of the RG flow. Combining the results for the
gauge-coupling and superpotential coupling constants, (2.46) and (2.47), we can
ask whether it is possible to find an solution to the equations:
 
1
β 2 =0, β(λ
eR ) = 0 . (2.48)
gR

Here we used gR , since its β-function has the same zeros as gc . The existence of such
fixed points should be completely scheme-independent. The equations (2.48) give
strong constraints on the anomalous dimensions γφ that can arise at any candidate
fixed point, even at strong coupling.

3 SQCD: Lagrangian, symmetries and classical moduli

space
Let us now focus on some particularly nice 4d N = 1 gauge theory, supersymmet-
ric QCD (a.k.a. SQCD). This is an SU (Nc ) N = 1 supersymmetric gauge theory
with Nf flavors. This means that we have Nf chiral multiplets in the fundamen-
tal representation of SU (Nc ), and Nf chiral multiplets in the anti-fundamental
representation. We denote them by:

Qi , i = 1, · · · , Nf , ej ,
Q j = 1, · · · , Nf , (3.1)

respectively. For obvious reasons, the numbers Nc and Nf are called the number
of colors and flavors, respectively, and the scalar fields Q, Q e are called the squarks.
The supersymmetric Lagrangian takes the form:
 
Z Nf Nf
X X ¯e −2V e j 
L = d2 θd2 θ̄  Q̄i e−2V Qi + Q je Q
i=1 j=1 (3.2)
Z Z
τ τ̄
− d2 θ tr(WW) + d2 θ̄ tr(W̄ W̄) .
16πi 16πi
More precisely, this is massless SQCD, with vanishing superpotential. We could
also consider adding Dirac masses for the quarks, through a superpotential:

W = µi j Q
e j Qi , (3.3)

with µi j the mass matrix. Note that all gauge indices are implicit. We denote by
a = 1, · · · , Nc the gauge indices in the fundamental representation. Then,
e j Qi ≡ Q
Q e ja Qai ,

which is obviously gauge invariant, and similarly for the contraction of the gauge
indices in (3.2).
3.1 Anomalies and anomaly-free R-symmetry 21

SU (Nc ) SU (Nf ) SU (Nf ) U (1)A U (1)B U (1)R

(Qai ) (Nc ) (Nf ) (1) 1 1 r
e ja )
(Q (Nc ) (1) (Nf ) 1 −1 r

Table 1: Gauge and global symmetry representations of the chiral multiplets Q

e and
Q of SQCD.

The symmetries of the classical Lagrangian of massless SQCD are:

e F = SU (Nf ) × SU (Nf ) × U (1)A × U (1)B ,
G (3.4)

times an R-symmetry U (1)R . The U (1)A and U (1)B factors are called the axial
symmetry and the baryonic symmetry, respectively. The charges of the chiral su-
perfields under the gauge and global symmetries are summarised in Table 1. The
R-charge shown if for an R-charge:

R = R0 + rA , (3.5)

with A the generator of U (1)A and R0 the “reference” R-charge under which the
chiral multiplets are neutral.

3.1 Anomalies and anomaly-free R-symmetry

One can easily check that the gauge group SU (Nc ) of SQCD is anomaly free—that
is the reason we needed as many fundamental as antifundamental chiral multiplets.
As a special case of the discussion in section 2.1.2, we see that U (1)A suffers
from a chiral anomaly:
X
AU (1)A = T (Ri ) = 2Nf , (3.6)
i

and so does U (1)R , for a general mixing parameter r ∈ R:

AU (1)R = 2Nc + 2Nf (r − 1) . (3.7)

However, there is a unique choice of r such that the U (1)R -SU (Nc )2 anomaly (3.7)
vanishes, namely:
Nc
r =1− . (3.8)
Nf
The other symmetries are non-anomalous. Thus, the global symmetry of SQCD,
at the quantum level, is:

GF × U (1)R , GF ≡ SU (Nf ) × SU (Nf ) × U (1)B , (3.9)

with this particular choice of R-charge.

22 3 SQCD: Lagrangian, symmetries and classical moduli space

3.2 The classical vacuum moduli space of SQCD and gauge-invariant

operators
In the absence of superpotential, SQCD can have a large vacuum moduli space.
e ja the scalars in the corresponding superfields, and also
Let us denote by Qai and Q
their VEVs. We then have:
n o
M = (Qai , Q e j ) ∈ C2Nc Nf µ = 0 /SU (Nc ) .
a (3.10)

Here, the “D-term constraints” can be written as the vanishing of the traceless
Nc × Nc matrices: 5
Nf  Nf 
δba δba
 X 
†i a †i †a e j † ej
X
µab ≡ Q b Qi − tr(Q Qi ) − Qj Qb −
e tr(Qj Q ) = 0 . (3.11)
e
Nc Nc
i=1 j=1

Note that the matrix µab is obviously SU (Nc ) × SU (Nc ) × U (1)B -invariant.
For any fixed number of “colors” Nc , the structure of the SU (Nc ) SQCD moduli
space changes as we vary the number of flavors, Nf . The basic physical reason is
the Higgs mechanism. A non-zero VEV for a single fundamental scalar Q = (Qa )
of SU (Nc ) breaks the gauge group as:

hQa i =
6 0 ⇒ SU (Nc ) → SU (Nc − 1) . (3.12)

By a gauge transformation, we can take the vector hQa i to be (q1 , 0, 0, · · · , 0), which
is obviously preserved by the SU (Nc − 1) subgroup of SU (Nc ). More generally, a
generic VEV for the Nf squarks breaks the gauge group according to:

hQai i =
6 0, i = 1, · · · , Nf ⇒ SU (Nc ) → SU (Nc − Nf ) . (3.13)

For Nf ≥ Nc , the SU (Nc ) gauge group is entirely broken at a generic point on the
vacuum moduli space. The VEVs also have to satisfy the D-term conditions (3.11)
in order to preserve supersymmetry.
Another approach to analysing the moduli space is to use the description (2.10).
In the absence of superpotential, this tells us that:
n o
M = (Qai , Qe ja ) ∈ C2Nc Nf /SL(Nc , C) . (3.14)

This space can be constructed algebraically by building all the possible gauge in-
variant chiral operators, X, and then imposing relations between the fields X that
follow from their definition—these are known as syzygies. This is a classic problem
in invariant theory. We will present explicit examples below.
5
Here we take:
1 a c
(T a )ab = (Tdc )ab = δbc δda −
δb δd ,
Nc
for the fundamental of SU (Nc ), and then (T̄ a )ab = −(T a )ab for the anti-fundamental.
3.2 The classical vacuum moduli space of SQCD and gauge-invariant operators 23

3.2.1 The case Nf < Nc

If Nf < Nc , we can pick the VEVs of the fundamental squarks to be:
 
q1 0 · · · 0
 0 q2 · · · 0 
 
 .. . . 
 . . 
a
 
(Qi ) =  0 0 · · · qNf 

 . (3.15)
0 0 ··· 0 
 
 .. 
 . 
0 0 ··· 0

Plugging this into (3.11), we see that this solves the D-term condition if and only:
 
qe1 0 · · · 0 0 ··· 0
 0 qe2 · · · 0 0 · · · 0
e ia ) = 
(Q  .. . .  ,

with |qi | = |e
qi | , ∀i . (3.16)
 . . 0 · · · 0
0 0 ··· qeNf 0 ··· 0

This is not the most general solution to (3.11), but any other allowed VEV can be
reached by considering the SU (Nf ) × SU (Nf ) orbit of (3.15)-(3.16). It turns out
that:
dimC (M) = Nf2 , (3.17)
as a complex space. This can be understood easily in terms of the Higgs mechanism.
At a generic point on M, the gauge symmetry is broken as in (3.12) and so:

NW = (Nc2 − 1) − ((Nc − Nf )2 − 1) = 2Nf Nc − Nf2 , (3.18)

gauge bosons get a mass, by each “eating” a complex scaler φ. Thus, out of the
2Nf Nc complex scalars in the UV, 2Nf Nc − NW = Nf2 scalars survive on the IR
moduli space, matching the counting (3.17).
We can also see this in the purely algebraic description (3.14). In this language,
we should construct all the gauge invariant scalars build out of the fundamental
e There are only N 2 of them, which we denote by:
squarks Q and Q. f

Mji ≡ Q
e ja Qai . (3.19)

They are usually called the SQCD mesons, since they are made of two fundamental
squarks, in analogy with the mesons of real-world QCD which are made of two
quarks. They are the natural coordinates on the moduli space for Nf < Nc , with:
2
M∼ N
=C f . (3.20)

Note also that the case Nf = Nc − 1 is special, since the gauge group is com-
pletely broken. (The “SU (1)” group is trivial.)
24 3 SQCD: Lagrangian, symmetries and classical moduli space

SU (Nf ) SU (Nf ) U (1)B U (1)R U (1)A

M (Nf ) (Nf ) 0 2r 2
B (⊗N c
A Nf ) (1) Nc rNc Nc
B
e (1) (⊗Nc
A Nf ) −Nc rNc Nc
Λb0 (1) (1) 0 2Nc + 2Nf (r − 1) 2Nf

Table 2: Global symmetry representations the gauge-invariant chiral operators of

SQCD. Here, U (1)A is anomalous, and so is U (1)R unless we choose the R-charge as
in (3.8). This induces A and R charges for the SQCD scale Λb0 , with b0 = 3Nc −Nf .

3.2.2 The case Nf ≥ Nc

For Nf ≥ Nc , the gauge group is completely Higgsed at a generic point on the
moduli space. By the same Higgs-mechanism argument as above, we should have:

dimC (M) = 2Nf Nc − Nc2 + 1 . (3.21)

Let us see how this comes about in the language of gauge-invariant chiral fields.
Now, in addition to the mesons (3.19), we can build other gauge-invariants, by
making use of the fully-antisymmetric invariant tensor of SU (Nc ). These are the
so-called “baryons:”
a
Bi1 i2 ···iNc ≡ a1 a2 ···aNc Qai11 Qai22 · · · QiNNc , (3.22)
c

and the “anti-baryons:”

e j1 j2 ···jNc ≡ a1 a2 ···aNc Q

B e ja1 Q e jaNc .
e ja2 · · · Q (3.23)
1 2 Nc

Again, the name comes from QCD, where a baryon is a gauge-invariant combination
of 3 fermions in the fundamental of the gauge group SU (3).
For future reference, we collect the symmetry charges of the mesons and baryons
in Table 2. In total, there are:
 
2 Nf
NX ≡ Nf + 2 , (3.24)
Nc

gauge-invariant operators, since the baryons B transform in Nc -index antisymmet-

ric representations of SU (Nf ), and similarly for the anti-baryons B.
e This is larger
than the expected dimension of the moduli space, (3.21).
The reason is that they are relations amongst the so-called generators of M
seen as an algebraic variety:
X = (M, B, B) e , (3.25)
3.2 The classical vacuum moduli space of SQCD and gauge-invariant operators 25

which follow directly from their definition in terms of fundamental fields. One can
easily check that:

j ···j [j
e j1 j2 ···jNc Bi i ···i − M 1 · · · M Nc = 0 , j ]
Pi11···iNNc (X) ≡ B 1 2 Nc i1 iN (3.26)
c c

where the square bracket denotes anti-symmetrization of the indices. Moreover,

since the anti-symmetrisation of n > Nc squarks must vanish, we have:

i ···iNf j i ···iNf
Q Nc +2 (X) ≡  1 Bi1 ···iNc MijN =0, (3.27)
c +1

if Nf > Nc , and similarly:

j1 ···jNc j
Q
ej
Nc +2 ···jNf i (X) ≡ j1 ···jNf B Mi Nc +1 = 0 . (3.28)

The moduli space is then given explicitly in terms of generators and relations, as:
n o
M = X ∈ CNX P (X) = 0 , Q(X) = 0 , Q(X)
e =0 . (3.29)

Mathematically, this is known as a affine variety (the zero set of a some polynomials
in Cn ).

3.2.3 The case Nf = Nc

Consider the case Nf = Nc in more detail. According to the description (3.29), we

have:
NX = Nc2 + 2 (3.30)

generators, while the moduli space is of dimension Nc2 + 1. Indeed, the generators
are:
Mij , B, Be , (3.31)

where we defined:

B = i1 ···iNc Bi1 ···iNc , e j1 ···jNc ,

Be = j1 ···jNc B (3.32)

using the SU (Nf ) × SU (Nf )-invariant -symbols (with Nf = Nc ). There is a single

relation, of the form (3.26), amongst the generators, which can be written as:

det(M ) − BB
e =0. (3.33)

The resulting affine variety M is an example of a hypersurface in algebraic geometry.

26 3 SQCD: Lagrangian, symmetries and classical moduli space

3.2.4 The case Nf = Nc + 1

In this case, we have:
NX = Nf2 + 2Nf (3.34)
generators, the mesons and baryons:

Mij , B i0 = i0 i1 ···iNc Bi1 ···iNc , e j1 ···jNc .

Bej0 = j0 j1 ···jNc B (3.35)

On the other hand, the expected dimension of the moduli space is:

dim(M) = 2Nf (Nf − 1) − (Nf − 1)2 + 1 = Nf2 . (3.36)

The relations amongst the generators take the form:

B i Bej − Minor(M )ij = 0 , Mij B i = 0 , Bej Mij = 0 . (3.37)

Note that there are NP = Nf2 + 2Nf relations amongst the NX = Nf2 + 2Nf
generators, but:
NP > NX − dim(M) . (3.38)
Thus, the relations cannot be all independent, but nonetheless there does not exist
a smaller set of relations. This is a common feature of algebraic varieties. (Variety
whose dimensions is given by the number of generators minus the number of rela-
tions are called complete intersections. The SQCD moduli space for Nf > Nc is
not a complete intersection.)

3.3 The IR phases of SQCD

In this section, we discuss how the classical picture of the SQCD vacuum is modified
quantum mechanically. We will only be able to touch upon the subject, for lack of
space and time, but hopefully it will be enough to give you some the desire to go
and learn more about this beautiful chapter of mathematical physics.

3.4 Infrared phases of gauge theories

Asymptotically-free gauge theories run to strong coupling at low energy, and per-
turbation theory breaks down. Therefore, exploring the infrared physics of gauge
theories is a very challenging problem, theoretically.
Given an RG flow from any weakly coupled UV theory (in particular, a gauge
theory), there are roughly three possibilities for what may happen in the infrared:

• Mass gap. The theory might be gapped. That is, there are no excitations
below a finite energy E0 , and therefore the far-infrared physics is trivial.
For instance, pure Yang-Mills theory in 4d is expected to have a mass gap.
(Proving that conjecture is a Millennium Prize Problem, literally and figura-
tively worth $1,000,000.)
3.5 Aspects of the quantum vacuum of SQCD 27

• IR free. The infrared theory consists of free massless particles.

This happens, for instance, in QED below the electron mass, where we only
have a free photon. This also happens in theories with spontaneous symmetry
breaking, where the Goldstone bosons are the massless particles.

• Non-trivial fixed point (CFT). The infrared theory may be at a non-

trivial fixed point of the renormalisation group flow. In that case, the IR
theory is a non-trivial conformal field theory (CFT). 6
While there are no obvious examples of this kind of RG flow in real-world
particle physics, there are plenty of examples in condensed-matter physics.

Note that the vacuum at a generic point of the classical SQCD moduli space is
an IR-free theory, consisting of n free chiral multiplets, with n = dim(M). On the
other hand, it is much more challenging to understand what happens at the origin
of the moduli space, where the gauge group SU (Nc ) is unbroken and we expect the
strongly-coupled gauge dynamics at scales µ ≤ Λ to be dominant.

3.5 Aspects of the quantum vacuum of SQCD

3.5.1 Nf = 0: SYM theory

In the case of super-Yang-Mills theory (Nf = 0), we have a ordinary pure YM

theory SU (Nc ) coupled to a fermion λ in the adjoint representation. In the UV,
the theory has a U (1)R symmetry classically. Due to the chiral anomaly:

AU (1)R = 2Nc , (3.39)

from the gaugino, the R-symmetry is broken to a discrete subgroup:

U (1)R → Z2Nc . (3.40)

It is expected that the theory confines and develops a mass gap in the IR, just
like pure YM theory. Moreover, the theory has Nc distinct vacua, in which the R-
symmetry Z2Nc is spontaneously broken to Z2 , due to the appearance of a gaugino
condensate in the supersymmetry-preserving vacuum:
2πin
hλα λα i = Λ3 e Nc , n = 1, · · · , Nc . (3.41)

The fact that the theory has (at least) Nc distinct vacua can also be inferred from
the Witten index of the theory, which is equal to Nc [10].
6
Free massless particles are CFTs too, but trivial ones.
28 3 SQCD: Lagrangian, symmetries and classical moduli space

3.5.2 0 < Nf < Nc : runaway supersymmetry breaking

In the case 0 < Nf < Nc , the classical moduli space is spanned by the Nf2 mesons
Mij . This vacuum structure is preserved in perturbation theory, but might be
modified by non-perturbative effects.
Any such non-perturbative correction should appear as new operators in the
superpotential:
Weff = W (M, Λ) . (3.42)
Here, Weff can only be an holomorphic function of the complex scale Λ and of Mji
(instead of Q and Qe individually, by gauge invariance). In order to preserve the
symmetries, with charges given in the Table 2, we must have:

Weff = α(det M )c1 Λb0 c2 , (3.43)

with the dependence on det M only, to preserve SU (Nf ) × SU (Nf ), and the coef-
ficients:

c1 2Nf + c2 2Nf = 0 , 2rc1 Nf + c2 (2Nc + 2Nf (r − 1)) = 2 , (3.44)

for consistency with U (1)A × U (1)R . This gives:

1
Λ3Nc −Nf
 N
c −Nf
Weff = α , (3.45)
det M
up to some undetermined dimensionless coupling constant. In fact, it is a famous
result by Affleck, Dine and Seiberg (ADS) [11] that this superpotential is generated
by a one-instanton effect (that is, at first order in Λb0 ) when Nf = Nc − 1, giving:

Λ3Nc −Nf
WADS,Nf =Nc −1 = . (3.46)
det M
By consistency with various decoupling limits, this fixes:
1
Λ3Nc −Nf
 N
c −Nf
WADS = (Nc − Nf ) , (3.47)
det M

for any 0 < Nf < Nc .

The effect of the ADS superpotential is quite dramatic. While there is a large
moduli space of vacua classically, and at all orders in perturbation theory, a non-
2 2
perturbative correction—a “tiny” correction of order e−8π /g —destabilises the vac-
uum. Indeed, the vacuum equations that follow from this superpotential are:

∂Q WADS = 0 ∂Qe WADS = 0 . (3.48)

These equations have no solution except in the limit:

hM i → ∞ . (3.49)
3.5 Aspects of the quantum vacuum of SQCD 29

Thus, supersymmetry is spontaneously broken—since the breaking is due to dy-

namical effect, this is an example of what is known as dynamical supersymmetry
breaking. In fact, there is still a supersymmetry vacuum asymptotically, “at infin-
ity” in field space.
Exercice: Consider the case of massive SQCD, with the tree-level Dirac mass term
(3.3). We still have an ADS superpotential generated at one-loop, so that:

Weff = µij Mij + WADS . (3.50)

Show that, in that case, we have a finite number (equal to Nc ) of supersymmetric

vacua. In the massless limit µ → 0, these vacuum are pushed to infinity in field
space. This is known as runaway supersymmetry breaking.

3.5.3 Are there low-energy σ-models for Nf ≥ Nc ?

Consider now the case Nf ≥ Nc . Far away on the moduli space, as field distances
hφi  Λ, the physics is the one of the weakly-coupled Higgs mechanism, which
gave us a simple way to compute dim(M) in (3.21). The low energy excitations are
given in terms of mesons and baryons—even though there are constraints relating
those gauge-invariant fields, we can always solve those constraint locally at a generic
point on M, to keep n = dim(G) coordinates.
The hard question is to understand what happens quantum-mechanically near
the origin of the moduli space, at field distances:

hφi ≤ Λ . (3.51)

At φ = 0, classically, we have the massless bosons of the SU (Nc ) vector multiplet,

but their dynamics is strongly coupled in the infrared. Assuming that supersym-
metry is not broken, one hypothesis is that the low-energy effective theory can be
written entirely in terms of gauge-invariant massless states, the mesons and baryons
X, sitting in chiral multiplets, and interacting through a superpotential.
The simplest form of this hypothesis is that there the low energy theory is just
a supersymmetric σ-model of chiral multiplets:
Z Z Z
Lσ-model = d θd θ̄ K(X̄, X) + d θW (X) + d2 θW (X) ,
2 2 2
(3.52)

for the “meson” and “baryon” fields X in (3.25), which we now view as the “fun-
damental fields” in the low-energy description. The effective Lagrangian at scale
µ  Λ is then given in terms of some unknown (and presumably complicated)
Kähler potential K, and some holomorphic superpotential W (X).
We can test this hypothesis using the ’t Hooft anomaly matching condition.
There are many non-trivial ’t Hooft anomalies for global symmetry group GF ×
U (1)R of the UV theory, massless SQCD. For instance:

tr(SU (Nf )3+ ) = −Nc , (3.53)

30 3 SQCD: Lagrangian, symmetries and classical moduli space

for the SU (Nf ) factor under which Qi transforms. We also have:

tr(SU (Nf )2 U (1)B ) = Nc , (3.54)

and:
Nc4
tr(U (1)R ) = −Nc2 − 1 , tr(U (1)3 ) = Nc2 − 2 −1 , (3.55)
Nf2
where we used the non-anomalous R-charge (3.8). There are still more ’t Hooft
anomaly coefficients, whose computation is left as an exercice.
On the other hand, the naive theory (3.52) would have:
 
tr(SU (Nf )3+ ) = −Nf + A ⊗N A
c
Nf , (3.56)

where the first term is the contribution from the mesons M , and the second term is
the contribution from the baryons B to the cubic anomaly. In particular, we have:

tr(SU (Nf )3+ ) = −Nf , tr(SU (Nf )3+ ) = −Nf + 1 , (3.57)

Nf =Nc Nf =Nc +1

for Nf = Nc or Nf = Nc = 1. From this and other ’t Hooft anomalies, we see that

the naive σ-model does not reproduce the ’t Hooft anomalies of SQCD for generic
Nf .
In fact, one can easily show that all anomalies match if and only if Nf = Nc + 1. In
that case, the anomalies match almost miraculously. For instance, the cubic U (1)R
’t Hooft anomaly in the σ-model is given by:
(Nf − 2Nc )3 (Nc Nf − Nf − Nc2 )3 Nf
 
3
tr(U (1) ) = +2 . (3.58)
Nf Nf3 Nc

This does match the corresponding SQCD anomaly in (3.55) for Nf = Nc + 1,

namely:
2Nc4
tr(U (1)3 ) =− + Nc2 − 1 . (3.59)
Nf =Nc +1 (Nc + 1)2

3.5.4 Nf = Nc : deformed moduli space

For Nf = Nc , we just argued that the low energy theory at the origin of the moduli
space cannot be simply a theory of massless mesons and baryons. To make a long
story short, what happens in this case is that the origin of the moduli space does
not exist quantum mechanically.
Recall that the classical moduli space is described as an hypersurface:

det(M ) − BB
e =0. (3.60)

Quantum mechanically, this relation is deformed to [3]:

e = Λb0 ,
det(M ) − BB b0 = 2Nc . (3.61)
3.5 Aspects of the quantum vacuum of SQCD 31

This is again a non-perturbative (one-instanton) effect. Note that the deformation

is fully compatible with the (anomalous) symmetries from Table 2 with Nf = Nc .
Algebraically, this is an example of a deformation of a singularity—while the
hypersurface (3.60) had a singularity at the origin, the deformed space (3.61) is
smooth. 7

3.5.5 Nf = Nc + 1: A σ-model

For Nf = Nc + 1, we saw above that we could saturate the ’t Hooft anomalies with
our naive σ-model of mesons and baryons. This cannot be the full description,
however, since the moduli space has a lower dimension than the number of fields,
NX = Nf2 +2Nf . Instead, we should have a superpotential to impose some relations
amongst the fields X. By symmetry, we can only have:

det M B i Mij Bej

W =α +β (3.62)
Λb0 Λb0

By various decoupling limits, one can fix α = −β = −1. We thus claim that the
correct superpotential is:

B i Mij Bej − det M

W = . (3.63)
Λb0

This is a rather strange result, since it does not seem to behave well in the classical
limit, Λ → 0. However, the numerator would also vanish in this limit, due to the
classical constraints (3.37)—in particular, det M = 0 classically since it is a matrix
of rank Nc < Nf .
In the low-energy description, the F -term equations that follow from (3.63),
when treating M and B, Be as fundamental fields, give us:

∂W
= B i Bej − (M −1 )ij det M = 0 ,
∂Mij
∂W
= Mij B i = 0 , (3.64)
∂B i
∂W
= Bej Mij = 0 .
∂ Bej

These are precisely the constraints (3.37) that define the classical moduli space.
Thus, for Nf = Nc + 1, the low-energy description seems to be in terms of a
σ-model whose vacuum moduli space is exactly the same as in the UV description.
7
Here, deformation is also a technical term; more precisely, we have a “complex structure
deformation” of an algebraic variety.
32 3 SQCD: Lagrangian, symmetries and classical moduli space

3.5.6 Nf > Nc + 1: SCFTs, free theories, and duality

For Nf > Nc + 1, the low-energy physics is more interesting. It was elucidated
in 1994 by Seiberg [4]. Here, we only give some brief summary of that beautiful
subject, and we refer to the classic lecture notes by Intriligator and Seiberg [1] for
further reading; see also Argyres’ lecture notes (link in introduction) for a detailed
and pedagogical account.
In order to describe what happens in SQCD with:

Nc + 1 < Nf ≤ 3Nc , (3.65)

It is perhaps easiest to first discuss what happens near the upper bound Nf = 3Nc .
At precisely Nf = 3Nc , the Yang-Mills β-function vanishes at one-loop. We thus
have have a fixed-point of the RG flow, at first order—a four-dimensional CFT.
Consider the “exact” β function (2.46), which reads:
 
1 3Nc − Nf (1 − γφ )
β 2
= . (3.66)
gc 8π 2 − Nc gc2
for SQCD; here, we used the global symmetries to equate all the anomalous dimen-
sions,
γQi = γQej ≡ γφ . (3.67)
For Nf = 3Nc , there is a fixed point, at all order in perturbation theory, if γφ = 0.
That is, if the squarks retain their classical dimensions, ∆ = 12 . Therefore, it seems
that massless SQCD at Nf = 3Nc is an “almost free” conformal field theory.
Now, for Nf < 3Nc , we could try to obtain a zero of the β function:
 
1 Nf − 3Nc
β =0 ⇔ γφ = , (3.68)
gc2 Nf
where we used the global symmetries to equate all the anomalous dimensions. Con-
sider, in particular, the case of large number of colors, Nc  1 and 2Nc − Nf very
small; then, the anomalous dimensions are arbitrarily small and one can understand
the fixed point perturbatively. Such a fixed point is called a Banks-Zaks fixed point;
it also occurs in QCD-like theories without supersymmetry [12]. All we need is a
β-function of the form:
 
1 b0
β 2
= 2 − c0 Nf g 2 + O(g 4 ) , (3.69)
g 8π
at two-loop order, with c0 > 0 a positive numerical constant. Then, we have a
perturbative fixed point with a coupling constant:
1 b0
g∗2 = 1, (3.70)
8π 2 c0 Nf
if b0 is smaller than c0 Nf .
3.5 Aspects of the quantum vacuum of SQCD 33

SCFTs. The claim is that there is a non-trivial fixed point in the IR of SQCD in
the full range:
3Nc
≤ Nf ≤ Nc . (3.71)
2
This is called the SQCD conformal window. The gauge coupling gc2 at the fixed
point is small near the upper limit (for Nc and Nf sufficiently large, giving a Banks-
Zaks fixed point), but becomes strong (with g∗2 of order one) as we lower Nf , at fixed
Nc . The lower bound on the conformal window comes about as follows. Any fixed
point preserving supersymmetry necessarily enjoys a larger space-time symmetry
algebra, called the N = 1 superconformal algebra. It has generators: 8
Pµ ,
Qα , Q̄α̇ ,
∆, Mµν , R, (3.72)
Sα , S̄α̇ ,
Kµ ,
generalising the super-Poincaré algebra. Here, ∆ is the dilation operator, whose
eigenvalues are the quantum dimensions (or just “conformal dimensions”) of the
operators, by definition, and R is the U (1)R charge. The R-charge is now a non-
trivial part of the algebra. In 4d N = 1 superconformal field theories (SCFTs), the
scalar chiral operators Φ satisfy a BPS-type relation tying up their R-charges and
dimensions:
2
R[Φ] = ∆[Φ] . (3.73)
3
Note that this relation is compatible with a classically-marginal (that is, conformally-
invariant) superpotential. Another general fact about 4d CFTs (with or without
supersymmetry) is that the dimension of any well-defined (gauge-invariant) opera-
tor must satisfy:
∆(O) ≥ 1 , (3.74)
and the operator is free if and only if this so-called unitary bound is satisfied.
Consider the quantum dimension (3.68) for the squarks chiral superfields. Using
(3.73), that implies:
 
2 1 Nc
R[Q] = R[Q] =
e 1 + γφ = 1 − =r, (3.75)
3 2 Nf
precisely the anomaly-free R-charge of SQCD. Then, for the gauge-invariant meson
operators M = QQ e to satisfy the unitarity bound, we must have:

e =3−3 Nc 3
∆(QQ) ≥1 ⇔ Nf ≥ Nc . (3.76)
Nf 2
That explains the lower-bound on the conformal window.
8
Here the generators are organised according to their conformal dimensions, from ∆[Pµ ] = 1
to ∆[Kµ ] = −1. The Poincaré supercharges Qα , Q̄α̇ have dimension 21 , and the special conformal
supercharges Sα , S̄α̇ have dimension − 12 .
34 3 SQCD: Lagrangian, symmetries and classical moduli space

IR-free phase. Finally, we should discuss what happens in the window:

3
Nc + 1 < Nf < Nc . (3.77)
2
We saw that the naive mesons and baryons cannot give a good description of the
origin of the moduli space. One heuristic reason is that the only superpotential
term allowed for the mesons is of the form:
1
W ∼ (det M ) Nf −Nc , (3.78)

which is singular at the origin. Such singularities in the effective action typically
hint at the presence of light particles, which we forgot to take into account in the
Wilsonian action. The extraordinary claim, due to Seiberg, is that one should
describe the low energy physics in terms of some IR-free gauge theory with gauge
group:
SU (Nf − Nc ) (3.79)
and Nf flavors in chiral multiplets q i and qej , coupled to some additional Nf2 gauge-
singlets caled Mij , with a cubic superpotential:

W = tr qej Mij q i . (3.80)

This SQCD-like theory, which we call the “Seiberg-dual theory” of SQCD, has a
β-function coefficient:

bD
0 = 3(Nf − Nc ) − Nf = 2Nf − 3Nc , (3.81)

which is negative in the window (3.77). Thus, indeed, it becomes a free theory in
the infrared (and needs to be defined with a UV cut-off, at the scale Λ).
Let us repeat the claim: the low-energy theory for asymptotically-free SQCD
in the window (3.77) is given in terms of an IR-free gauge theory, with gauge group
SU (Nf − Nc ). The gauge bosons and matter fields of this “dual theory” have
nothing to do with the original fundamental fields of SQCD in the UV. Nonetheless,
one can check that all ’t Hooft anomalies match between SQCD and the proposed
IR description! The proposal, in fact, passes many other consistency checks, which
we will not discuss here.

Seiberg duality. Even more amazingly, this relation between two different gauge
theories, known as “Seiberg duality,” extends all the way into the conformal window,
where both the SU (Nc ) and the “dual” SU (Nf −Nc ) gauge group are asymptotically
free. In that case, we have two well-defined asymptotically free gauge theories in
the UV, written schematically as:

SU (Nc ) , Nf , W =0 ↔ SU (Nf − Nc ) , Nf , W = qeM q . (3.82)

 35

They are certainly two different theories, with different numbers of degrees of free-
dom in the UV. The claim is that, in the conformal window, they both flow to the
same SCFT in the infrared. Moreover, when one description is strongly coupled,
the other is weakly coupled—that fact, as you can imagine, can be very useful.
We should point out that the above intricate picture of the quantum vacuum struc-
ture of SQCD has no definite proof for Nf ≥ Nc , to this day, but it passes so many
highly non-trivial consistency checks that its correctness is beyond any reasonable
doubt.

Part II
Dynamics of 4d N = 2 gauge theories
We now turn our attention to gauge theories with minimally extended (N = 2)
supersymmetry. Our aim will be to introduce the most important elementary facts
about N = 2 supersymmetric gauge theories, and then to discuss the Seiberg-
Witten solution for their low-energy dynamics. For simplicity, we shall mainly
focus on the simplest case of a single SU (2) N = 2 vector multiplet.

4 Gauge theories with N = 2 supersymmetry

By way of introduction, let us consider an N = 1 gauge theory consisting of an
N = 1 vector multiplet V for a gauge group G, and of a single chiral superfield Φ
in the adjoint representation of G.
In N = 1 superspace notation, the Lagrangian reads:
 Z  Z
τ 1
L = Im − d θ tr W Wα + 2 d2 θd2 θ̄ tr Φ̄e−2[V,−] Φ ,
2 α
(4.1)
8π g

where V acts on Φ in the adjoind representation, as denoted by the commutator,

and the holomorphic gauge coupling τ is as defined in (1.39). Here, we introduced
a seemingly arbitrary normalization factor 1/g 2 in front of the kinetic term for Φ.
This is because of the (non-obvious) fact that this action actually preserves N = 2
supersymmetry, and the normalization is the natural one in that case. To see some
evidence for the presence of the extended supersymmetry, let us write down the
action in components:

1  1 1
L = tr − Fµν F µν − iλ̄σ̄ µ Dµ λ + D2 − Dµ φ̄Dµ φ
g2 4 2
√ √ 
− iψ̄σ̄ µ Dµ ψ + F̄ F − φ̄[D, φ] − i 2φ̄[λ, ψ] + i 2[λ̄, ψ̄]φ (4.2)
θ 
µνρσ

− tr  Fµν ρσ .
F
64π 2
36 4 Gauge theories with N = 2 supersymmetry

We claim that this is the action of a N = 2 vector multiplet. The N = 2 vector

multiplet contains a complex scalar φ, two gauginos λI (I = 1, 2), a vector field Aµ ,
and three real auxiliary fields DIJ = DJI :

VN =2 = φ , φ̄ , Aµ , λI , λ̄I , DIJ ,


(4.3)

for a total of 8 + 8 off-shell degrees of freedom, with all fields transforming in the
adjoint representation of g = Lie(G). In the N = 1 notation above, we have:

(λI ) = (λ, ψ) , DIJ ∼ (D, F, F̄ ) . (4.4)

Let us integrate out the auxiliary fields. Then, the Lagrangian (4.2) becomes:

1  1
LN =2 SYM = tr − Fµν F µν − Dµ φ̄Dµ φ − iλ̄I σ̄ µ Dµ λI
g2 4
(4.5)
1 i i 
− [φ̄, φ]2 − √ φ̄IJ [λI , λJ ] + √ IJ [λ̄I , λ̄J ]φ ,
2 2 2

plus the topological term (with coupling constant θ), which is separately super-
symmetric. This Lagrangian is classically invariant under a SU (2)R × U (1)r R-
symmetry, with I = 1, 2 the SU (2)R index. We will explore this important point
momentarily.
Note also the form of the scalar potential that appears in (4.5):

1
tr [φ̄, φ]2 .


V0 (φ, φ̄) = 2
(4.6)
2g

If the theory were only N = 1 supersymmetric, we could have considered adding

a superpotential term W (Φ) for the adjoint scalar φ, and that would contribute
additional terms to the scalar potential. The simple form (4.6) is dictated by the
requirement of extended supersymmetry.

4.1 N = 2 supersymmetry and R-symmetry

The N = 2 supersymmetry algebra takes the form:

{QIα , Q̄β̇J } = 2σαµβ̇ Pµ δ I J ,

{QIα , QJβ } = 2εαβ IJ Z̄ , (4.7)
{Q̄α̇I , Q̄β̇J } = 2εα̇β̇ IJ Z ,

with I, J = 1, 2 and 12 = 12 = 1 for the SU (2)R indices. The maximal R-
symmetry of the N = 2 supersymmetry algebra is:

U (2) ∼
= SU (2)R × U (1)r , (4.8)
4.1 N = 2 supersymmetry and R-symmetry 37

with the charge assignements:

SU (2)R
SU (2)R × U (1)r T3 U (1)N
R
=1

QIα (2)−1 ± 12 I
−δ 1
(4.9)
Q̄α̇J (2̄)1 ∓ 12 +δ I 1
Z (1̄)2 0

The presence of a supercharge Z 6= 0 breaks the U (1)r symmetry explicitly. Re-

gardless, one can keep track of the symmetry by assigning r(Z) = 2 to the central
charge, as indicated.
Here we also considered an N = 1 subalgebra of the N = 2 superalgebra,
corresponding to the supercharges QI=1 , Q̄I=1 . The corresponding the N = 1
U (1)R symmetry is simply:
1 (R)
RN =1 = r − T3 , (4.10)
2
where r is the U (1)r charge, and T3 is the generator in the Cartan of SU (2)R .

4.1.1 The massless vector multiplet

We can easily build the massless one-particle multiplets (see section 3.3.1 of ‘Su-
persymmetry and Supergravity’). The N = 2 vector multiplet takes the form:
E a† 1 E a† E
I I
λ = −1; (1)0 −→ λ = − ; (2̄)1 −→ λ = 0; (1)2 ,
2 (4.11)
E a† 1 E a† E
I I
λ = 0; (1)−2 −→ λ = ; (2̄)−1 −→ λ = 0; (1)0 ,
2
where we indicated the helicities and the R-charges of the states. The gauge field
Aµ corresponds to the |λ| = 1 states, which must be neutral under the R-symmetry.
This then determines the R-charges of the other states. Note 2 ∼ = 2̄, so that the
1
fermions, with helicity |λ| = 2 , form a CPT-invariant set. Their free-field realization
is in term of the N = 2 gaugino λIα , λ̄α̇I :

Aµ λI λ̄I φ φ̄ ,
(4.12)
SU (2)R × U (1)r (1)0 (2)1 (2̄)−1 (1)2 (1)−2

Note that the complex scalar φ has U (1)R charge 2.

4.1.2 Massive multiplets and BPS bound

Consider massive particle states. A generic massive one-particle multiplet contains
8 + 8 on-shell degrees of freedom. On the other hand, in the presence of non-trivial
central charges, there can short representations for massive particle, called the BPS
states. 9 Let us see how this comes about.
9
Note that ‘BPS’ stands for Bogomolnyi-Prasad-Sommerfield.
38 4 Gauge theories with N = 2 supersymmetry

On massive one-particle states, with Pµ = (−M, 0, 0, 0), the supersymmetry

algebra takes the form:

{QI1 , Q̄1̇J } = {QI2 , Q̄2̇J } = 2M δ I J ,

{QI1 , QJ2 } = −2Z̄IJ , (4.13)
{Q̄1̇I , Q̄2̇J } = −2ZIJ ,

with all the other anticommutators vanishing. Let us define the operators:
1 1
aI = √ (QI1 + α(I) IJ Q̄2̇J ) , a†I = √ (Q̄1̇I + ᾱ(I) IJ Q̄J2 ) ,
2 2
(4.14)
1 1
bI = √ (QI1 − α(I) IJ Q̄2̇J ) , b†I = √ (Q̄1̇I − ᾱ(I) IJ Q̄J2 ) ,
2 2
with α1 , α2 some pure phases. One can check that the only non-zero commutators
amongst these operators are:

{aI , a†J } = δ I J 2M − α(I) Z̄ − ᾱ(I) Z ,

{bI , b†J } = δ I J 2M + α(I) Z̄ + ᾱ(I) Z ,

(4.15)
{aI , b†J } = δ I J − α(I) Z̄ + ᾱ(I) Z ,

{a†I , bJ } = δ I J α(I) Z̄ − ᾱ(I) Z .

For any fixed Z, we can choose α(I) = ei arg(Z) , so that the RHS of the last two
lines in (4.15) vanish. We then find the interesting conditions:

{aI , a†I } = 2 (M + |Z|) ≥ 0 , {bI , b†I } = 2 (M − |Z|) ≥ 0 , (4.16)

(no summation on I here), since {aI , a†I } and {bI , b†I } are positive-definite. These
so-called BPS inequalities are very important in the study of 4d N = 2 quantum
field theories.
We can easily study the supermultiplets of one-particle states, as before. For
M 6= ±|Z|, we have an ordinary massive multiplet, also known as a long multiplet.
It has 24 = 16 states, 8 bosonic and 8 fermionic. When the so-called BPS condition:

M = |Z| , (4.17)

is satisfied, on the other hand, we have a short multiplet, with half the number of
components, since bI , b†I in (4.16) are then realised trivially on one-particle states.

4.2 The classical Coulomb branch

The N = 2 SYM theory (4.5) has a classical vacuum moduli space determined by
(4.6), namely:
[φ, φ̄] = 0 , (4.18)
4.2 The classical Coulomb branch 39

modulo G gauge transformations. Semi-classically, this can be described simply,

as follows. The vanishing of the commutator (4.18) implies that φ and φ̄ can be
diagonalized simultaneously, by a g transformation—that is, φ can be conjugated
to the Cartan subalgebra h ⊂ g. This fixes the gauge up to gauge transformations
in the Weyl group WG of G. This classical moduli space thus takes the form

MC = hC /WG . (4.19)

It is called the Coulomb branch, for reason that will be clear momentarily. Its
complex dimension is the rank of G.
We will focus on G = SU (2). Then, the scalar φ can be conjugated to:
 
a 0
φ= , a∈C. (4.20)
0 −a

The Weyl group of SU (2) is S2 = Z2 , and it acts on a as a sign invertion:

WSU (2) = Z2 : a → −a . (4.21)

It is clear that the VEV (4.20) with a 6= 0 breaks the gauge group according to the
pattern:
SU (2) → U (1) , (4.22)
by the Higgs mechanism. For a general G, we have:

G → U (1)r0 , r0 ≡ rank(G) . (4.23)

at a generic point on the Coulomb branch. Thus, classically, the low-energy dynam-
ics is dictated by the Higgs mechanism. The strict IR limit contains only massless
U (1)
abelian gauge fields Aµ , and their N = 2 superpartners in abelian vector mul-
U (1)
tiplets VN =2 . Any vacuum moduli space of a supersymmetric theory with this
property is called a Coulomb branch, because of the presence of long-range inter-
actions between charged particle in the IR through the ‘Coulombic’ interactions
mediated by the abelian gauge fields. 10
On the Coulomb branch (4.20) of the SU (2) theory, we also have a perturbative
massive vector field, the W-boson, of mass:

MW = 2|a| . (4.24)

At large distance on the Coulomb branch, that is with |a| → ∞, the semi-classical
Higgs mechanism gives the full physics. On the other hand, we will see that, at
finite distance on the Coulomb branch, strong coupling effects will change the low
energy dynamics significantly.
10
In addition, we also have attractive forces from the exchange of the massless scalars. For a
BPS particle, the repulsive Coulombic forces and the scalar forces cancel out exactly.
40 4 Gauge theories with N = 2 supersymmetry

4.3 ’t Hooft-Polyakov monopoles on the Coulomb branch

In our discussion below, we will need a key fact about the classical SU (2) gauge
theory spontaneously broken to U (1), which is that it allows for topologically non-
trivial solitons, called ’t Hooft-Polyakov monopoles. For further reference, see the
lectures [13]. 11
First, some basic definitions, to avoid any possible confusions. In any (classical)
relativistic field theory:

• an instanton is a non-trivial solution of the classical equations of motion,

ϕ = ϕ0 (x), that has finite action, S[ϕ0 ] < 0.

• a soliton is a non-trivial solution of the classical equations of motions, ϕ =

ϕ0 (x), that has finite energy, E < 0.

We discussed Yang-Mills instantons before; they are solutions of the Yang-Mills

equations, setting all other fields in the theory to zero. By contrast, monopoles
are solitons that exist in theories with both gauge fields and scalar. Consider the
G = SU (2) theory with an adjoint hermitian scalar ϕ (that is, ϕ† = ϕ), with action:
Z
1 4
 1
µν 1 µ

S = 2 d x tr − Fµν F − Dµ ϕD ϕ . (4.25)
g 4 2

The energy of any classical field configuration reads:

Z
1  
E[B, ϕ] = 2 d3 x tr Bi Bi + Di ϕDi ϕ (4.26)
2g

Here, we defined the electric and magnetic field in the usual way:

F0i = Ei , Fij = ijk B k . (4.27)

Now, we have the following trick:

Z
1 3 2


E[B, ϕ] = d x tr(Bi ∓ D i ϕ) ± 2 tr(B i Di ϕ)
2g 2
Z Z (4.28)
1 3 1
≥ ± 2 d x tr Bi Di ϕ = ± 2 dni tr(B i ϕ)
g g S∞ 2

In the second line, we used the fact that the term |Bi ∓ Di ϕ|2 is positive definite,
and that the remaining term is a total derivative, so that the final integral is over
the sphere at spatial infinity. In this way, we learn that the energy of any field
configuration is bounded by the magnetic charge of the gauge field times the VEV
of the scalar ϕ at spatial infinity.
11
Like for our discussion of instantons in previous lectures, we cannot do justice to this beautiful
subject here. I encourage you to read more broadly about solitons in classical field theories.
4.3 ’t Hooft-Polyakov monopoles on the Coulomb branch 41

The lower bound (4.28) on the soliton energy is also called ‘the BPS bound.’
The bound is saturated by solutions that satisfy the Bogolmonyi equations:
Bi = Di ϕ . (4.29)
This is essentially the condition for the solution for the bosonic background (Aµ , φ)
of an N = 2 SU (2) vector multiplet to preserve half of the supersymmetry.
The ’t Hooft-Polyakov monopole is a specific solution for an adjoint scalar with
boundary conditions set at spatial infinity, and the solution can be chosen to satu-
rates the BPS bound. Let us pick the VEV (4.20) for the complex scalar φ. By a
U (1)r rotation, we can choose a = |a| ∈ R, for simplicity. This is thus equivalent
to:  
|a| 0
ϕ= , (4.30)
0 −|a|
up to some unimportant constant factor. This breaks SU (2) to U (1), and the
profiles for the SU (2) gauge field compatible with the VEV are then of the form:
!
U (1)
Aµ 0
Aµ = U (1) , (4.31)
0 −Aµ
up to gauge transformations. The monopole solution is then indexed by the the
magnetic flux of this abelian gauge field at spatial infinity:
Z
1 U (1)
m= dni Bi . (4.32)
2π S∞ 2

and the VEV (4.30) for the scalar field. It has energy: 12

4π
E = Mmonopole = |a|m . (4.33)
g2

In conclusion, on the Coulomb branch of the SU (2) N = 2 SYM theory, there exists
monopoles in the classical theory. Note that their mass goes like g12 , and therefore
they are very massive in the perturbative regime, g 2  1 (much more massive than
the W-boson, whose mass is given by (4.24)).
In the quantum theory, we should view the monopoles as particle excitations,
on par with the W-boson. Due to the BPS bound, the monopole with the lowest
magnetic charge is a stable particle. Note that, while the W-boson is electrically
charged under the U (1) gauge symmetry in the IR, the monopole is magnetically
charged.
Simply comparing the semi-classical masses MW ∼ |a| and Mmonopole ∼ |a| g2
, we
may suspect that the W-bosons and the monopoles become much more similar in
the strong coupling regime. This is indeed the case. As we will see next, in the
full quantum field theory, the gauge coupling becomes strong near the origin of the
Coulomb branch, and the monopoles play a role very similar to the W-bosons in
the full quantum theory.
12
Up to some numerical factor we did not keep track of.
42 References

5 Theories of abelian vector multiplets

5.1 Stating the problem: “solving N =2 SYM”?
5.2 U (1) gauge fields and electro-magnetic duality
5.3 The N =2 prepotential for an abelian vector multiplet
5.4 Semi-classical dynamics: anomalies and one-loop running
5.5 Abelian vector multiplet coupled to hypermultiplet

6 The Seiberg-Witten solution for pure SU (2)

References
[1] K. A. Intriligator and N. Seiberg, “Lectures on supersymmetric gauge
theories and electric-magnetic duality,” Nucl. Phys. Proc. Suppl. 45BC
(1996) 1–28, arXiv:hep-th/9509066 [hep-th]. [,157(1995)].

[2] Y. Tachikawa, N=2 supersymmetric dynamics for pedestrians, vol. 890. 2014.
arXiv:1312.2684 [hep-th].

[3] N. Seiberg, “Exact results on the space of vacua of four-dimensional SUSY

gauge theories,” Phys. Rev. D49 (1994) 6857–6863, arXiv:hep-th/9402044
[hep-th].

[4] N. Seiberg, “Electric - magnetic duality in supersymmetric nonAbelian gauge

theories,” Nucl. Phys. B435 (1995) 129–146, arXiv:hep-th/9411149
[hep-th].

[5] N. Seiberg and E. Witten, “Electric - magnetic duality, monopole

condensation, and confinement in N=2 supersymmetric Yang-Mills theory,”
Nucl. Phys. B426 (1994) 19–52, arXiv:hep-th/9407087 [hep-th].
[Erratum: Nucl. Phys.B430,485(1994)].

[6] G. ’t Hooft, “Naturalness, chiral symmetry, and spontaneous chiral symmetry

breaking,” NATO Sci. Ser. B 59 (1980) 135–157.

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

gauge theories,” Phys. Rev. D53 (1996) 3399–3405, arXiv:hep-th/9506098
[hep-th].

[8] N. Arkani-Hamed and H. Murayama, “Holomorphy, rescaling anomalies and

exact beta functions in supersymmetric gauge theories,” JHEP 06 (2000)
030, arXiv:hep-th/9707133 [hep-th].
References 43

[9] V. A. Novikov, M. A. Shifman, A. I. Vainshtein, and V. I. Zakharov, “Beta

Function in Supersymmetric Gauge Theories: Instantons Versus Traditional
Approach,” Phys. Lett. 166B (1986) 329–333. [Yad. Fiz.43,459(1986)].

[10] E. Witten, “Dynamical Breaking of Supersymmetry,” Nucl. Phys. B188

(1981) 513.

[11] I. Affleck, M. Dine, and N. Seiberg, “Dynamical Supersymmetry Breaking in

Supersymmetric QCD,” Nucl. Phys. B241 (1984) 493–534.

[12] T. Banks and A. Zaks, “On the Phase Structure of Vector-Like Gauge
Theories with Massless Fermions,” Nucl. Phys. B196 (1982) 189–204.

[13] J. A. Harvey, “Magnetic monopoles, duality and supersymmetry,” in Fields,

strings and duality. Proceedings, Summer School, Theoretical Advanced Study
Institute in Elementary Particle Physics, TASI’96, Boulder, USA, June 2-28,
1996. 3, 1996. arXiv:hep-th/9603086.