A First Approach to

N = 1 Supersymmetry

Daniel Farquet

Physics department
EPFL

Presented to
Claudio Scrucca

Special thanks to
L. Brizi and C. Andrey

Abstract
This document provides an introduction to supersymmetry. The no-go theo-
rems and grading are introduced first. We then introduce the notion of superfield
and superspace. Constrained fields and their corresponding actions as well as
matter coupling in non-abelian gauge theory are discussed in detail. Not only
is supersymmetry breaking discussed but also a mass formula and particular re-
sults are given in the case of a U(N) and S U(N) gauge group. For example we
prove that if the gauge group is U(N) in its fundamental representation, then it
forbids the possibility of having a non-trivial gauge invariant superpotential. The
non-linear sigma model and non-linear gauge invariant sigma model are treated.
A very general mass formula is proved for the latter. All needed mathematical
results are expounded in the appendix.
 1

Preliminary remarks: This paper has been written in the spirit of introduc-
ing supersymmetry in field theory. The necessary background are the basics of
classical field theory and a bit of quantum field theory (QFT). We also suppose
that the reader knows spinor algebra and the associated spinors identities. They
can be found in sections 10.1 and 10.2 of the appendix as a reminder. We advise
the reader to go quickly through those two sections in order to know what the
conventions used throughout this document are. Basically the conventions we
use are those of Wess and Bagger, [2], excepting for σmn (which has an i so it is
self-adjoint).
We will sometimes refer to the appendix. This appendix presents the various
mathematical notions that the reader needs to fully understand this paper. It will
be explicitly stated in the text when one needs to read the appendix in order to
fully understand.
This document might seem quite dense for a first read in supersymmetry but
it is meant to be-self contained. All calculations are either expounded upon in
detail or their main steps are given.

Contents
1 Introduction 2

2 No-Go Theorems and Grading 2

2.1 No-Go Theorems . . . . . . . . . . . . . . . . . . . . . . . . . 2
2.2 Graded Algebra . . . . . . . . . . . . . . . . . . . . . . . . . . 4
2.3 Graded matrices . . . . . . . . . . . . . . . . . . . . . . . . . . 5

3 The Algebra 7

4 Superfields and Superspace 8

4.1 Superfields and Charge Representation . . . . . . . . . . . . . . 9
4.2 Covariant Derivatives . . . . . . . . . . . . . . . . . . . . . . . 11
4.3 Component Fields . . . . . . . . . . . . . . . . . . . . . . . . . 14

5 Constrained Fields and Corresponding Actions 15

5.1 Chiral Fields . . . . . . . . . . . . . . . . . . . . . . . . . . . . 15
5.2 Vector Field . . . . . . . . . . . . . . . . . . . . . . . . . . . . 17

6 Matter Coupling and Non-abelian Gauge Theory 21

6.1 Construction of the Lagrangian . . . . . . . . . . . . . . . . . . 21
6.2 Calculation of the Component Lagrangian . . . . . . . . . . . . 22

7 Supersymmetry Breaking 25
7.1 General Features . . . . . . . . . . . . . . . . . . . . . . . . . 25

Daniel Farquet EPFL - Physics

 2

7.2 The Mass Formula . . . . . . . . . . . . . . . . . . . . . . . . 27

7.3 The U(N) Gauge Group . . . . . . . . . . . . . . . . . . . . . . 30
7.4 The S U(N) Gauge Group . . . . . . . . . . . . . . . . . . . . . 30
7.5 Examples . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 32

8 The Non-linear Chiral Sigma Model 33

9 The Non-linear Gauge Invariant Sigma Model 34

9.1 The Mass Formula . . . . . . . . . . . . . . . . . . . . . . . . 36
9.2 The Linear Case . . . . . . . . . . . . . . . . . . . . . . . . . . 38

10 Appendix 39
10.1 Convention and Reminder on Spinors . . . . . . . . . . . . . . 39
10.2 Spinor identities . . . . . . . . . . . . . . . . . . . . . . . . . . 42
10.3 Independence of θ and θ̄ . . . . . . . . . . . . . . . . . . . . . 42
10.4 Grassmann derivative . . . . . . . . . . . . . . . . . . . . . . . 43
10.4.1 General theory . . . . . . . . . . . . . . . . . . . . . . 43
10.4.2 The Weyl Spinor Derivative . . . . . . . . . . . . . . . 44
10.5 Grassmann Integration . . . . . . . . . . . . . . . . . . . . . . 45
10.6 The Fermionic Mass Square Matrix . . . . . . . . . . . . . . . 47
10.7 U(N) and S U(N) in the Fundamental Representation . . . . . . 48
10.8 Kähler Geometry . . . . . . . . . . . . . . . . . . . . . . . . . 49

Daniel Farquet EPFL - Physics

 3

1 Introduction
This document is meant to introduce supersymmetry in field theory. The point of
view that is taken to introduce supersymmetry algebra is the one that considers
the possible symmetry group of the S-matrix. In fact we will shortly discuss
the theorem of Coleman-Mandula and the theorem of Lopuszanski-Sohnius that
forbid or allow a possible extension of the Poincaré group as a symmetry of the
S-matrix.
Supersymmetry is a deep and fascinating subject and its hypothetical role in
the fundamental theory of nature has encouraged people to study it in great detail.
Several types of supersymmetric theories can be considered, depending on the
number of ‘charges’ the theory carries. In this document we only discuss the
N = 1 case. There is already a lot to study at this level: superspace, superfields,
supersymmetric Lagrangians, supersymmetry breaking, the mass formula, the
non-linear sigma model, etc...
Since a very good introductory discussion is given in [3] we will not para-
phrase that paper and will allow the interested reader to read it as an introduction.
Therefore we will go directly to the technical part.

2 No-Go Theorems and Grading

2.1 No-Go Theorems
In this section we will briefly discuss the theorems of Coleman-Mandula and
Lopuszanski-Sohnius. These two theorems are of great importance since they al-
low us, or forbid us, to extend the Poincaré algebra to a bigger group of symmetry
of the S matrix.

Theorem 2.1. (Coleman-Mandula) Let G be a connected symmetry group of the

S-matrix, i.e. a group whose generators commute with the S-matrix. We make the
assumptions:

1. Lorentz invariance: G contains a subgroup which is locally isomorphic to

the Poincaré group.

2. Particle finiteness: All particle types correspond to positive-energy repre-

sentations of the Poincaré group. For any finite mass M, there is only a
finite number of particles with mass less than M.

3. Weak elastic analyticity: Elastic scattering amplitudes are analytic func-

tions of centre-of-mass energy squared s and invariant momentum transfer
squared t in some neighborhood of the physical region, except at normal
thresholds.

Daniel Farquet EPFL - Physics

 4

4. Occurence of scattering: Let |pi and |p0 i be any two one particle momen-
tum eigenstates, and let |p, p0 i be the two-particle state constructed from
them. Then
T |p, p0 i , 0
where T is related to the S matrix by
S = 1 − i(2π)4 δ4 (pm − p0m )T,
except, maybe, for certain isolated values of S . In layman’s term this as-
sumption means: two plane waves scatter at almost any energy.
5. Technical assumption: The generators of G, considered as integral opera-
tors in momentum space, have distributions for their kernel.
Then the group G is locally isomorphic to the direct product of a compact sym-
metry group and the Poincaré group.
Thus this result shows that the most general Lie algebra of symmetries of the
S -matrix contains the energy momentum operator Pm , the Lorentz rotation and
boosts generator Mmn and a finite number of Lorentz scalar operators B p , i.e.
[Pm , B p ] = 0 [Mmn , B p ] = 0,
where the B p constitute a Lie algebra
[B p , Bq ] = icrpq Br ,
and the crpq are the structure constants of the algebra.
We can relax the hypothesis of the last theorem by including algebraic sys-
tems whose defining relations also involve, in addition to the usual commutators,
anticommutators. These kind of algebra are called superalgebra, or graded Lie
algebra.
Let us assume that we have a set of generators called spinor charges QαA ,
A = 1, · · · , N which transform according to some representation of a compact Lie
group G which represents the internal symmetry group. The generators of G are
the Lorentz scalars Br . Consider the relations
{QαA , QβB } = 0
{Q̄α̇A , Q̄β̇B } = 0
{QαA , Q̄β̇B } = 2δAB σm P
αβ̇ m
[QαA , Pm ] = 0 (2.1)
[Q̄α̇A , Pm ] = 0
[QαA , Bl ] = iS lAB QαB
β
[QαA , M mn ] = (σmn )α QβA
[Bl , Bm ] = icklm Bk

Daniel Farquet EPFL - Physics

 5

where the S lAB are matrices in the hermitian representation containing the charges
QαA . We have the following result.

Theorem 2.2. (Lopuszanski-Sohnius) The maximal symmetry of the S -matrix is

the direct product of an internal symmetry with the superalgebra given by the
relations (2.1). The only allowed extension is the possible appearance of so-
called central charges in the anticommutator of two undotted (or two dotted)
spinors:
{QαA , QβB } = εαβ Z AB , Z AB = −Z BA . (2.2)
Furthermore, the Z’s commute with all generators:

[Z AB , anything] = 0, (2.3)

which is the reason why they are called central charges.

From now on we will restrict ourselves to the case where there is only one
charge Qα , this is called N = 1 supersymmetry, or unextended supersymmetry.
We speak of extended supersymmetry if N > 1.
We are now left with the non-trivial task of finding a way to represent the
algebra on a well-suited space. This will lead us to the notion of superfields.
Before studying superfields, we will briefly discuss a way of constructing the
algebra (2.1) by referring to the notion of grading.

2.2 Graded Algebra

The notion of graded Lie algebra grants us an abstract generalisation of the alge-
bra (2.1).

Definition 2.1. A linear algebra L := span{Xm } is given a Z2 grading if L is the

direct sum of two subspaces L0 and L1 , i.e. L = L0 ⊕ L1 where we set

L0 = span{Ei }, i = 1, · · · , dim L0
L1 = span{Qa }, a = 1, · · · , dim L1

on which a composition law

# : L×L → L

acts as follow

L0 # L0 ⊂ L0 L0 # L1 ⊂ L1 L1 # L1 ⊂ L0

and we add the constraint that

Ei # E j = −E j # Ei , Ei # Qa = −Qa # Ei , Qa # Qb = Qb # Qa .

Daniel Farquet EPFL - Physics

 6

If we take the matrix commutator to be the composition law for L0 # L0 and

L0 # L1 and the matrix anticommutator to be the composition law for L1 # L1 , it
can be shown without too much difficulty that instead of one Jacobi identity, the
graded algebra bequeaths four identities:

[Ei , [E j , Ek ]] + [Ek , [Ei , E j ]] + [E j , [Ek , Ei ]] = 0

[Ei , [E j , Qa ]] + [Qa , [Ei , E j ]] + [E j , [Qa , Ei ]] = 0 (2.4)
[Ei , {Qa , Qb }] − {Qb , [Ei , Qa ]} − {Qa , [Qb , Ei ]} = 0
[Qa , {Qb , Qc }] + [Qc , {Qa , Qb }] + [Qb , {Qc , Qa }] = 0

where Ei , E j , Ek ∈ L0 and Qa , Qb , Qc ∈ L1 . It is important to observe that the

commutators and anticommutators in the previous equations are consistent with
the composition law for graded algebra.
From the Jacobi identities, we can show that the matrix elements for the com-
mutators of an E and a Q have to form a representation of the algebra of L0 . This
gives us an important piece of information: There must exist a representation of
L0 whose dimension is equal to the number of charges Qa , i.e. the dimension of
L1 .
Let us take L0 to be the Poincaré group and L1 = {Q1 , Q2 } since we work
with N = 1 supersymmetry. From the last comment, we know that we must look
for a two-dimensional representation of the Poincaré group in order to obtain the
coefficients for [QαA , Pm ] and [QαA , M mn ]. This is exactly what is achieved when
we choose
β
[QαA , Pm ] = 0, [QαA , M mn ] = (σmn )α QβA (2.5)
The other commutators and anticommutators of relation (2.1) ensue from the
generalised Jacobi identities (2.4) and the last equality.
The grading procedure has given us an abstract way of constructing the super-
algebra. Nonetheless, one must keep in mind that it is the Lopuszanski-Sohnius
theorem which is the powerful tool here as it states that it is impossible to find
other algebras that are symmetries of the S -matrix.

2.3 Graded matrices

Before entering into the study of the algebra in question, we will introduce the
notion of graded matrices. We will do that at this stage since it is quite discon-
nected from the other topics. However even if this notion is not explicitly used in
this paper it can be useful in some situations.

Definition 2.2. A graded matrix (or supermatrix) is a matrix M given by

!
A B
M= ,
C D

Daniel Farquet EPFL - Physics

 7

with A = (Ai j ), B = (Bia ), C = (Cai ), D = (Dab ) i, j = 1, ..., n and a, b = 1, ..., m. To

this bloc defined matrix we add the condition that
BiaCb j = −Cb j Bia ,
i.e. the parameters composing the matrices B and C are Grassmann numbers but
the one composing A and D are just ordinary numbers.
For these kind of matrices, we want to recover the usual properties of the trace
and determinant. The simplest way to construct a trace wich inherits interesting
properties is the following
Definition 2.3. Let M be a graded matrix. We define the supertrace by
STr M := Tr A − Tr D. (2.6)
This definition is a shrewd one since it gives us the cyclic property of the
supertrace (if we had chosen a plus sign instead of a minus sign it would not have
been the case):
Proposition 2.1. Let M1 and M2 be two graded matrices. The supertrace is
cyclic, i.e.
STr(M1 M2 ) = STr(M2 M1 ). (2.7)
Proof. The proof is straightforward since it just involves the use of the cyclic
property of the normal trace and the anticommuting nature of B and C.
By complete analogy with linear algebra, we can define the determinant using
the exponential.
Definition 2.4. The determinant of a supermatrix M, the superdeterminant, is
defined to be
Sdet M := exp(STr ln M) (2.8)
Finally, one must note that we recover the usual multiplication property of the
determinant.
Proposition 2.2. Let M1 and M2 be two graded matrices. The superdeterminant
satisfies
Sdet(M1 M2 ) = Sdet(M1 ) · Sdet(M2 ). (2.9)
Proof. As the supertrace is cyclic, the supertrace of a commutator vanishes. Us-
ing the Baker-Campbell-Hausdorff formula, we can write
( )
1
Sdet(M1 M2 ) = Sdet exp ln M1 + ln M2 + [ln M1 , ln M2 ]
2
= exp {STr(ln M1 + ln M2 )} = exp {STr ln M1 } exp {STr ln M2 }
= Sdet M1 · Sdet M2 .

Daniel Farquet EPFL - Physics

 8

3 The Algebra
The following equations summarise results of the previous sections for N = 1
supersymmetry and serve as a reminder of the Poincaré algebra:

[Pm , Pn ] = 0
[Pm , Mlp ] = i(ηml P p − ηmp Pl )
[Mmn , Mlp ] = i(ηnl Mmp − ηnp Mml − ηml Mnp + ηmp Mnl )
[R, Pm ] = [R, Mmn ] = 0
[Q̄α̇ , Pm ] = [Qα , Pm ] = 0
β
[Qα , M mn ] = (σmn )α Qβ (3.1)
[Q̄α̇ , M mn ] = (σ̄mn )α̇β̇ Q̄β̇
[Qα , R] = Qα
[Q̄α̇ , R] = −Q̄α̇
{Qα , Qβ } = {Q̄α̇ , Q̄β̇ } = 0
{Qα̇ , Q̄β̇ } = 2σm P
αβ̇ m

The attentive reader has surely noticed the fact that the B’s have disappeared and
have been replaced by a single R, a U(1) generator. In fact a quick glance at (2.1)
will convince the reader that there can only exist one B in the case of N = 1. By
convention it is named R. Thus R is the only internal symmetry of the theory.
Now that the algebra is finally completely defined, we can derive an important
relation which holds also for N , 1 supersymmetry.

Proposition 3.1. Every representation of the supersymmetry algebra (3.1), and

more generally (2.1), contains as many bosonic states as fermionic states.

Proof. We introduce a fermion number operator (−1)NF . When applied to a state

|si it acts as
(−1)NF |si = (−1)nF |si
if the state |si contains nF fermions. Since Qα and Q̄α̇ are spinors, they annihilate
or create fermions when promoted to operators. Thus

(−1)NF Qα = −Qα (−1)NF . (3.2)

For any finite dimensional representation such that the trace is well-defined, we
have
h i h i
Tr (−1)NF {Qα , Q̄β̇ } = Tr (−1)NF (Qα Q̄β̇ + Q̄β̇ Qα )
h i
= Tr (−1)NF Qα Q̄β̇ + Qα (−1)NF Q̄β̇
= 0

Daniel Farquet EPFL - Physics

 9

by (3.2) for the second equality and by the cyclic property of the trace for the
third. On the other hand using (3.1) we obtain
h i  
0 = Tr (−1) {Qα , Q̄β̇ } = Tr (−1) 2σαβ̇ Pm
NF NF m

= 2σm P Tr(−1)NF .
αβ̇ m

For a non-zero momentum this gives Tr(−1)NF = 0, i.e.

X X
hB|(−1)NF |Bi + hF|(−1)NF |Fi = NB − NF = 0
| {z } | {z }
B =1 F =−1

where F and B stand for fermions and bosons respectively. This demonstrates the
proposition.
This proposition tells us that the bosonic number of degrees of freedom is
always equal to the fermionic number of degrees of freedom. When we construct
representations it can help us to see whether the result we obtain is consistent or
not.

4 Superfields and Superspace

Very much like the Lorentz covariance formalism was constructed to permit us
to easily write invariants in quantum field theory, we are looking for a formalism
that could help us write supersymmetric invariant models. In order to obtain such
a formalism we can extend the Minkowski space to superspace. Elements of
superspace are called supercoordinate and consist of the usual four Minkowski
spacetime coordinates xm and four independent Grassmann numbers (see annex,
section 10.3):
{θα }α=1,2 , {θ̄β̇ }β̇=1̇,2̇ .
These Grasmann numbers are supposed to anticommute with everything fermionic,
including the Q’s, and to commute with everything else. The use of anticommut-
ing parameters allows use to rewrite the anticommutators appearing in the algebra
(3.1) with commutators in the following sense:

Proposition 4.1. We have that

[θQ, θ̄ Q̄] = 2θσm θ̄Pm

[θQ, θQ] = [θ̄ Q̄, θ̄ Q̄] = 0 (4.1)
[Pm , θQ] = [Pm , θ̄ Q̄] = 0

Daniel Farquet EPFL - Physics

 10

Proof. We demonstrate the first equality because the others can be shown by the
same technique. We have

[θα Qα , θ̄α̇ Q̄α̇ ] = [θα Qα , θ̄α̇ ]Q̄α̇ + θ̄α̇ [θα Qα , Q̄α̇ ]

= θ A {Qα , θ̄α̇ } − {θα , θ̄α̇ }Qα Q̄α̇
 

+ θ̄α̇ θ A {Qα , Q̄α̇ } − {θα , Q̄α̇ }Qα

 

= θ̄α̇ θα 2εα̇β̇ σm P = 2θσm θ̄Pm .

αβ̇ m

At this stage we see that we are facing an ordinary algebra if we consider θQ

and θ̄ Q̄ instead of Q and Q̄. Knowing that in quantum field theory a field φ(x)
can be written as
φ(x) = exp(−ixm Pm )φ(0) exp(ixm Pm )
where xm Pm form an ordinary algebra with λmn M mn , it is very intuitive to define
a superfield using the same procedure.

4.1 Superfields and Charge Representation

We start by giving a precise definition of a superfield.

Definition 4.1. A superfield described by the independent coordinates xm , θ and

θ̄ is defined by

φ(x, θ, θ̄) := L(xm , θα , θ̄α̇ )φ(0, 0, 0)L−1 (xm , θα , θ̄α̇ ) (4.2)

where
L(xm , θα , θ̄α̇ ) := exp −ixm Pm + iθQ + iθ̄ Q̄ .
 
(4.3)

Let us remark that here Pm , Qα and Q̄α̇ stand for operators which act on
functions, i.e. they are representation of the algebra (3.1) and must therefore
respect the commutation and anticommutation relations of the algebra. Let us
also note that the dimension of θ and θ̄ is − 21 in energy. Indeed Q and Q̄ have
a dimension of + 12 as can be seen from their anticommutation relation in (3.1)
(because they give a Pm ) and because we want θQ and θ̄ Q̄ to be dimensionless
since they appear in an exponential.

Proposition 4.2. Under a supersymmetry transformation T (y, ξ), which acts on

φ as a usual field transformation, the field transforms as

T (y, ξ)φ(x, θ, θ̄) = φ(x + y + iθσξ̄ − iξσθ̄, θ + ξ, θ̄ + ξ̄). (4.4)

Hence the parameters θ and θ̄ undergo just a translation in superspace.

Daniel Farquet EPFL - Physics

 11

Proof. By definition of the action of T ,

T (y, ξ)φ(x, θ, θ̄) = L(y, ξ, ξ̄)φ(x, θ, θ̄)L−1 (y, ξ, ξ̄)
= [L(y, ξ, ξ̄)L(x, θ, θ̄)]φ(0, 0, 0)[L(y, ξ, ξ̄)L(x, θ, θ̄)]−1 .
On the other hand
L(y, ξ, ξ̄)L(x, θ, θ̄) = exp{−i(x + y)P + i(θ + ξ)Q
+ i(θ̄ + ξ̄)Q̄ − ξσm θ̄Pm + θσm ξ̄Pm }
= L(x + y − iξσθ̄ + iθσξ̄, θ + ξ, θ̄ + ξ̄).
For the first equality we used exp A exp B = exp(A + B+ 12 [A, B] + ...) and the equa-
tions (4.1) implie that the expansion stops already after the first commutator. The
second equality follows from the definition of L. We therefore have the expected
result.
Proposition 4.3. An infinitesimal supersymmetry transformation defined with an
infinitesimal ξ by
δ s φ := T (0, ξ)φ(x, θ, θ̄) − φ(x, θ, θ̄) (4.5)
is also given by
δ s φ = i[ξQ + ξ̄ Q̄]φ(x, θ, θ̄). (4.6)
Proof. Developing to the first order in ξ we get
δ s φ = exp(iξQ + iξ̄ Q̄)φ(x, θ, θ̄) exp(−iξQ − iξ̄ Q̄) − φ(x, θ, θ̄)
= i[ξQ, φ(x, θ, θ̄)] + i[ξ̄ Q̄, φ(x, θ, θ̄)] + . . .
Supposing that the representation of the Q’s obeys the Leibniz rule (this is co-
herent since this proposition will lead us to a representation given by differential
operators), for any function f in superspace we have
i[ξQ, φ(x, θ, θ̄)] f = iξ[Qφ(x, θ, θ̄)] f
This last relation concludes the proof.
Before announcing the next proposition, we advise the reader who does not
feel comfortable with spinor derivatives to read section 10.4 in the annex. The
notation used is also explained in the annex.
Proposition 4.4. The charges Pm , Qα and Q̄α̇ are represented by differential
operators on functions of the form φ(x, θ, θ̄). In the notation ∂α = ∂θ∂α , ∂¯ α̇ = ∂∂θ̄α̇
and ∂m = ∂x∂m they are given by
Pm = −i∂m
 
m β̇
Qα = −i ∂α − iσαβ̇ θ̄ ∂m (4.7)
 
α̇ α̇ α m β̇α̇
Q̄ = −i ∂ − iθ σαβ̇ ε ∂m
¯

Daniel Farquet EPFL - Physics

 12

Proof. Using (4.4) and (4.5) we see that to the first order in ξ

δ s φ = φ(x + iθσξ̄ − iξσθ̄, θ + ξ, θ̄ + ξ̄) − φ(x, θ, θ̄)

∂ ∂
!
= ξ + ξ̄ + i(θσ ξ̄ − ξσ θ̄)∂m + . . . φ(x, θ, θ̄)
m m
∂θ ∂θ̄

As ξ and ξ̄ can be considered independent, identifying the last equality to equation

(4.6) gives the explicit form of Q and Q̄ announced above. For the representation
of Pm it is a straightforward exercise to go through the same procedure to find that
we indeed have Pm = −i∂m . One could also put forward that Pm must have the
same representation as in quantum field field theory, as it is clear from (4.2).
Finally, as a consistenty check, it can be verified that the representation of the
charges respect the right algebra.

Proposition 4.5. The charges Qα and Q̄α̇ verify

{Qα , Qβ } = {Q̄α̇ , Q̄β̇ } = 0 (4.8)

{Qα , Q̄β̇ } = 2σm P .
αβ̇ m

Proof. We only demonstrate the second equality for they can all be demonstrated
using the same technique. First of all we need to find the charge Q̄α̇ because we
only have Q̄α̇ .

Q̄α̇ = εα̇β̇ Q̄β̇ = −i(−∂¯ α̇ − iθα σm γ̇β̇ α m

αγ̇ ε εα̇β̇ ∂m ) = i(∂α̇ − iθ σαα̇ ∂m )
¯

Thus

{Qα , Q̄β̇ } = {∂α , ∂¯ β̇ } − i{∂α , θγ σnγβ̇ ∂n } − i{σm γ̇ m γ̇ γ n

αγ̇ θ̄ ∂m , ∂β̇ } − {σαγ̇ θ̄ ∂m , θ σγβ̇ ∂n }
¯
γ γ̇
= −iδα σnγβ̇ ∂n − iσm
αγ̇ δβ̇ ∂m = −2iσαβ̇ ∂m = 2σαβ̇ Pm .
m m

4.2 Covariant Derivatives

When working with the classical theory of gauge fields, it is very convenient to
introduce a derivative which is covariant, i.e. transforms in a nice way under
the gauge transformation. Similarly in our case, it would be very convenient to
have a covariant derivative with respect to supersymmetry transformations. The
requirement for such a derivative Dα is that it verify [Dα , δ s ](any superfield)= 0.
Having such a derivative in hand will allow us to write manifestly supersymmetric
invariant Lagrangians, so it is very important to find a covariant derivative.

Daniel Farquet EPFL - Physics

 13

Definition 4.2. The covariant derivatives for any superfield φ(x, θ, θ̄) are defined
by
Dα := ∂α + iσm
αβ̇
θ̄β̇ ∂m (4.9)
Dα := εαβ Dβ = −∂α − iθ̄γ̇ σ̄mγ̇α ∂m (4.10)
and
D̄α̇ := −∂¯ α̇ − iθβ σm
βα̇ ∂m (4.11)
D̄α̇ := εα̇β̇ D̄β̇ = ∂¯ α̇ + iσmα̇γ θγ ∂m (4.12)
∂ ∂ ∂
where ∂α = ∂θα , ∂¯ α̇ = ∂θ̄α̇
and ∂m = ∂xm .
As stated before, we have the following favourable result.
Proposition 4.6. The covariant derivatives are invariant under supersymmetry
transformations, i.e.
[Dα , δ s ]φ(x, θ, θ̄) = 0 (4.13)
[D̄α̇ , δ s ]φ(x, θ, θ̄) = 0. (4.14)
where φ(x, θ, θ̄) is any possible superfield.
Proof.
[Dα , δ s ]φ(x, θ, θ̄) = [Dα , iξQ + iξ̄ Q̄]φ
= −iξβ {Dα , Qβ }φ − iξ̄β̇ {Dα , Q̄β̇ }φ
using the fact that {Dα , ξβ } = {Dα , ξ̄β̇ } = 0 because they are all fermionic. On the
other hand,
{Dα , Qβ } = −i{∂α + iσm
αβ̇
θ̄β̇ ∂m , ∂β − iσnβγ̇ θ̄γ̇ ∂n } = 0
because of the properties of spinor derivatives (see annex). It is also easy to show
that {Dα , Q̄β̇ } = 0. Hence the first equality is proved.
The second equality can be shown using the same technique.
The algebra verified by the covariant derivative is the same as the one verified
by the charges Qα , Q̄α̇ . As the proof can be achieved in the exact same manner,
we only give the algebra without proving it.
Proposition 4.7. The following algebra holds:
{Dα , Dβ } = {D̄α̇ , D̄β̇ } = 0
D3α = D̄3α̇ = 0
{Dα , D̄β̇ } = 2σm P
αβ̇ m
(4.15)
{Dα , D̄β̇ } = 2σ̄mβ̇α Pm
[Dα , ∂m ] = [D̄α̇ , ∂m ] = 0.

Daniel Farquet EPFL - Physics

 14

In addition to the last proposition, covariant derivatives verify several other

properties that can be very useful for calculations.
Proposition 4.8. We have that:
α̇
[Dα , D̄2 ] = −4iσm
αα̇ D̄ ∂m (4.16)
[Dα , D̄2 ] = 4iD̄γ̇ σ̄mγ̇α ∂m (4.17)
[D̄α̇ , D ] =
2
4iDα σm αα̇ ∂m (4.18)
α̇ mα̇α
[D̄ , D ] 2
= −4iσ̄ Dα ∂m (4.19)
D̄σ̄m D = −Dσm D̄ + 4i∂m (4.20)
[D2 , D̄2 ] = −8iDσm D̄∂m − 16 (4.21)
[D̄2 , D2 ] = −8iD̄σ̄m D∂m − 16 (4.22)
Dα D̄2 Dα = D̄α̇ D2 D̄α̇ . (4.23)
Proof. Let us calculate the first commutator.
[Dα , D̄α̇ D̄α̇ ] = {Dα , D̄α̇ }D̄α̇ − D̄α̇ {Dα , D̄α̇ }
α̇ α̇β̇
= 2σm
αα̇ Pm D̄ − D̄α̇ ε 2σαβ̇ Pm
m

α̇ α̇
= 4σm
αα̇ D̄ Pm = −4iσαα̇ D̄ ∂m .
m

The second equality can be obtained with the use of ε. We do not show the third
and the fourth equality for the proof is exactly the same as for the first and the
second. For (4.20) we consider
D̄α̇ σ̄mα̇α Dα = (−2iσnαα̇ ∂n − Dα D̄α̇ )σ̄mα̇α
= −Dα D̄α̇ εαβ εα̇β̇ σm
ββ̇
− 2iσnαα̇ σ̄mα̇α ∂n
= −Dσm D̄ − 2i Tr(σn σ̄m )∂n
= −Dσm D̄ + 4i∂m
The next identity is
[Dα Dα , D̄2 ] = Dα [Dα , D̄2 ] + [Dα , D̄2 ]Dα
= −Dα 4iσm α̇
αα̇ D̄ ∂m + 4i D̄β̇ σ̄
mβ̇α
∂m Dα
= −4iDσm D̄∂m + 4iσ̄mβ̇α ∂m (−Dα D̄β̇ + 2σnαβ̇ Pn )
= −8iDσm D̄∂m + 8i Tr(σ̄m σn )∂m Pn
= −8iDσm D̄∂m − 16
Equation (4.22) can be shown in the same way. Using the first two equations of
this proposition, the last can be shown to hold. Indeed, we have
Dα (Dα D̄2 − D̄2 Dα ) = −4iDα σm α̇
αα̇ D̄ ∂m
(D̄α̇ D2 − D2 D̄α̇ )D̄α̇ = 4iDα σm α̇
αα̇ D̄ ∂m .

Daniel Farquet EPFL - Physics

 15

And thus
Dα D̄2 Dα = D̄α̇ D2 D̄α̇ .

Finally we calculate how the covariant derivative acts after some particular
change of variable. Since the proof is straightforward and there are no subtleties,
we allow ourselves to not expound it.
Proposition 4.9. Let us define ym := xm + iθσm θ̄ and zm := xm − iθσm θ̄. For any
superfiled φ(x, θ, θ̄) we have that

m α̇ ∂
!
Dα φ(x, θ, θ̄) = ∂α + 2iσαα̇ θ̄ φ(y, θ, θ̄) (4.24)
∂ym
D̄α̇ φ(x, θ, θ̄) = −∂¯ α̇ φ(y, θ, θ̄) (4.25)
and
Dα φ(x, θ, θ̄) = ∂α φ(z, θ, θ̄) (4.26)
∂
!
D̄α̇ φ(x, θ, θ̄) = − ∂¯ α̇ + 2iθα σm
αα̇ m φ(z, θ, θ̄) (4.27)
∂z
A final comment needs to be made at this stage. As it may have appeared
to the reader, it seems that there is a link between Dα and D̄α̇ . Indeed, the † on
the covariant derivative, understood as the usual matrix † when acting on the part
containing the spinors and the sigma matrix and only as the complex conjugation
when acting on the spinor derivative, is the link between the two. It can be verified
that
(Dα )† = D̄α̇ . (4.28)

4.3 Component Fields

From definition (4.2), we see that any superfield can be understood in terms of
a series expansion in θ and θ̄ (via the Hausdorff formula). The most general
expansion is
φ(x, θ, θ̄) := f (x) + θφ(x) + θ̄χ̄(x) + (θθ)m(x) + (θ̄θ̄)n(x) + (θσm θ̄)Vm (x)
+ (θθ)θ̄λ̄(x) + (θ̄θ̄)θψ(x) + (θθ)(θ̄θ̄)d(x) (4.29)
where f (x), φ(x), χ̄(x), m(x), n(x), Vm (x), λ̄(x), ψ(x), d(x) are called the com-
ponent fields and are defined via (4.29). Using equation (4.5), we are likely to
define that
δ s φ(x, θ, θ̄) = δ s f (x) + θδ s φ(x) + θ̄δ s χ̄(x) + (θθ)δ s m(x)
+ (θ̄θ̄)δ s n(x) + (θσm θ̄)δ s Vm (x) + (θθ)θ̄δ s λ̄(x)
+ (θ̄θ̄)θδ s ψ(x) + (θθ)(θ̄θ̄)δ s d(x) (4.30)
:= i[ξQ + ξ̄ Q̄]φ(x, θ, θ̄)

Daniel Farquet EPFL - Physics

 16

with of course the Q and Q̄ represented by (4.7). The transformation laws for
the various fields may be found by comparing the different powers of θ and θ̄ in
the last equation. In [1] this calculation is carried out with great care and the
variation of every component field is calculated. As we will not directly need this
last result (we will calculate supersymmetry transformations for particular cases)
we do not expound it here. Nevertheless let us mention that it is always the case
that under a supersymmetry transformation the highest component (in θ and θ̄)
of a superfield transforms into a total derivative. In our case the very result is
that d(x) → ∂µ (something). It is a property that must be kept in mind because it
is exactly by using the highest components of superfields that we will construct
Lagrangians.
From these definitions, it is straightforward to see that any linear combination
of superfields, or even product of superfields, are still superfields. This is of
course a consequence of the fact that Q and Q̄ are differential operators.
Unfortunately this very general representation is highly reducible. We shall
see in the next section that by considering constrained fields irreducible represen-
tations emerge and are of great importance.
For the sake of clarity and simplicity, detailed calculations will not be carried
out every time from now on. We will still give the main steps but avoid trivial
algebraic manipulations in order to render the discussion lighter.

5 Constrained Fields and Corresponding Actions

5.1 Chiral Fields
The first constraint we impose defines the set of chiral superfields. It is the set of
superfields such that
D̄α̇ φ(x, θ, θ̄) = 0. (5.1)
This constraint is hopefully invariant under supersymmetry since [D, δ s ] = 0. Us-
ing the y variable defined above, the constraint becomes

−∂¯ α̇ φ(y, θ, θ̄) = 0, (5.2)

√ expressed via y does not depend on θ̄. Thus the

which tells us that the superfield
most general solution is (the 2 is just a convenient normalisation)
√
φ(y, θ, θ̄) = φ(y, θ) = A(y) + 2θψ(y) + θθF(y) (5.3)

or, if we expand in x.
1
φ(x, θ, θ̄) = A(x) + iθσm θ̄∂m A(x) + θθθ̄θ̄A(x)
4
√ i
+ 2θψ(x) − √ θθ∂m ψ(x)σm θ̄ + θθF(x). (5.4)
2

Daniel Farquet EPFL - Physics

 17

Using the y representation, the supersymmetry transformation can be easily cal-

culated and is given in x by
√
δ s A(x) = 2ξψ(x) (5.5)
√ √
δ s ψα (x) = i 2(σ ξ̄)α ∂m A(x) + 2ξα F(x)
m
(5.6)
√
δ s F(x) = i 2ξ̄σ̄ ∂m ψ
m
(5.7)

We notice that the last component, i.e. F(x), transforms as a total derivative under
supersymmetry transformations. It tells us that if we take F to be a Lagrangian
the corresponding action would be invariant under susy (supersymmetry) trans-
formations.
The products of chiral superfields are still chiral superfields, as it might be
expected. For example if φi and φ j are two chiral superfields, then φi φ j is still a
chiral superfield. If for every i = 1, .., n φi is chiral, we conclude that any function
W(φ) := (φ1 , ..., φn ) is chiral as well. It follows that the θθ component of W(φ)
transforms into a total derivative.
This discussion can be redone for antichiral superfields denoted by φ† . They
obey the relation
Dα φ† (x, θ, θ̄) = 0. (5.8)
Taking the † of (5.1) yields
 †
D̄α̇ φ(x, θ, θ̄) = Dα (φ† ) = 0.

Thus any antichiral superfield is the † of a chiral superfield. The results previously
found can be extended to antichiral fileds by just taking the † of the previous
result, for example
√
φ† (z, θ, θ̄) = A∗ (z) + 2θ̄ψ̄(z) + θ̄θ̄F ∗ (z) (5.9)

or √
δ s A∗ (x) = 2ξ̄ψ̄(x).
The product of a chiral and an antichiral field is neither chiral nor antichi-
ral but has its θθθ̄θ̄ components transformed into a total derivative under susy
transformations. For example for φ†i and φ j are given by

1 1 1
φ†i (x, θ, θ̄)φ j (x, θ, θ̄)
= Fi∗ F j + A∗i A j + A∗i A j − ∂m A∗i ∂m A j +i∂m ψ̄i σ̄m ∂m ψ j
θθθ̄θ̄
4 4 2
(5.10)
where we have dropped the dependance in x for the sake of simplicity. In order
to lighten the notation, from now on, if the dependance is not written explicitly it
means that it is a dependance in x. In calculating this result one cannot use the
variables y or z to simplify it. Indeed y = x + iθσθ̄ and z = x − iθσθ̄ and thus they
carry both θ and θ̄ implying that φ†i (x, θ, θ̄)φ j (x, θ, θ̄) , φ†i (z, θ̄)φ j (y, θ) .
θθθ̄θ̄ θθθ̄θ̄

Daniel Farquet EPFL - Physics

 18

Equation (5.10) looks very much like a kinetic term of a theory containing
two different scalars and one fermion. If we let Lkin = φ†i (x, θ, θ̄)φi (x, θ, θ̄) we
θθθ̄θ̄
have
Lkin = i∂m ψ̄i σ̄m ψi − ∂m A∗i ∂m Ai + Fi∗ Fi + 4-div (5.11)
which is a true kinetic term of a theory involving 2n scalars and n fermions, if
i = 1, .., n. A more general susy Lagrangian can be obtained by adding terms of
the form W(φ)|θθ andR W̄(φ† )|θ̄θ̄ . Using projection tehchniques (cf. annex), we
know that W(φ)|θθ = d2 θW(φ) = − 14 D2 W(φ)|0 , that is

1 α ∂W(φ)
!
1 α
W(φ)|θθ = − D Dα W(φ)|0 = − D Da φi
4 4 ∂φi
! 0
1 ∂ W(φ) α
2 ∂W(φ) 2
= − D φ j Dα φi + D φi
4 ∂φ j ∂φi ∂φi 0
1h i 1
= − Wi j 2ψi ψ j − Wi 4Fi = Wi Fi − Wi j ψi ψ j
4 2

where we have introduced the notation Wi := ∂W(A) ∂ W(A) 2

∂Ai , Wi j := ∂Ai ∂A j and used the
explicit form of φi . W(A) is the function W evaluated on the field A(x) and is
what was needed since the calculation was done with θ = θ̄ = 0.
Applying the same procedure for φ† with the notation W̄ī := ∂W̄(A
∗)
∂A , W̄ī j̄ :=
∗
i
∂2 W̄(A∗ )
∂A∗i ∂A∗j we find that
1
W̄(φ† )|θ̄θ̄ = W̄ī Fi∗ − W̄ī j̄ ψ̄i ψ̄ j
2
Hence, the most general susy Lagrangian involving at most two derivatives for
chiral and antichiral fields is given by

L = φ†i φi + W(φ)|θθ + W̄(φ† )|θ̄θ̄

θθθ̄θ̄
= i∂m ψ̄i σ̄m ψi − ∂m A∗i ∂m Ai + Fi∗ Fi (5.12)
1 1
+Wi Fi + W̄ī Fi∗ − Wi j ψi ψ j − W̄ī j̄ ψ̄i ψ̄ j
2 2
From the explicit form of L we see that the F’s are auxiliary fields for no deriva-
tives of F are involved.

5.2 Vector Field

The set of vector superfields is the set of fields V such that

V(x, θ, θ̄) = V † (x, θ, θ̄). (5.13)

Daniel Farquet EPFL - Physics

 19

As the operator i[ξQ + ξ̄ Q̄] is hermitian, the reality condition is not spoiled under
susy transformations so the set of vector superfields is also an invariant subspace.
The most general vector superfield has the expansion
i i
V(x, θ, θ̄) = C(x) + iθχ(x) − iθ̄χ̄(x) + θθ[M(x) + iN(x)] − θ̄θ̄[M(x) − iN(x)]
 2i  2 i 
−θσ θ̄Vm (x) + iθθθ̄ λ̄ + σ̄ ∂m χ(x) − iθ̄θ̄θ λ(x) + σm ∂m χ̄(x)
m m

" #2 2
1 1
+ θθθ̄θ̄ D(x) + C(x) (5.14)
2 2

where we have chosen a very particular combination of fields as coefficients for

later convenience. The component fields C, D, M, N and Vm must all be real in
order for V to be a vector superfield.
The vector superfield carries redundant information, because for any chiral
field φ, we have that φ(x, θ, θ̄) + φ† (x, θ, θ̄) is a vector superfield. In order to get
rid of these spurious degrees of freedom we introduce the susy generalisation of
gauge transformations. The abelian gauge transformation of V is defined by

V → V 0 = V + φ + φ† . (5.15)
√
If we choose φ to be such that φ(y, θ, θ̄) = A(y) + 2θψ(y) + θθF(y), we find that

C(x) → C 0 (x) = C(x) + A(x) + A∗ (x)

√
χ(x) → χ0 (x) = χ(x) − i 2ψ
M(x) + iN(x) → M 0 (x) + iN 0 (x) = M(x) + iN(x) − 2iF(x)
Vm (x) → Vm0 (x) = Vm (x) − i∂m (A(x) − A∗ (x))
λ(x) → λ0 (x) = λ(x)
D(x) → D0 (x) = D(x)

We observe that λ and D are gauge invariant because of our choice of components
for the expansion of V. There is a special gauge, called the Wess-Zumino (WZ)
gauge, in which we choose C, χ, M and N to be equal to zero. This gauge still
allows us to perform a usual gauge transformation on the vector field Vm because
the imaginary part of A is not fixed by the WZ gauge, i.e. A − A∗ = a is not fixed
and Vm → Vm − i∂m a.
In this gauge V has quite a simple form and powers of V are easy to calculate:
1
V(x, θ, θ̄) = −θσm θ̄Vm (x) + iθθθ̄λ(x) + θθθ̄θ̄D(x) (5.16)
2
1
V (x, θ, θ̄) = − θθθ̄θ̄Vm (x)V (x)
2 m
(5.17)
2
V 3 (x, θ, θ̄) = 0. (5.18)

Daniel Farquet EPFL - Physics

 20

We must point out that the WZ gauge is not susy invariant, i.e. a susy transfor-
mation will take us out of the WZ gauge. It is also important to note that a susy
transformation sends the D component of the field into a total derivative. This
result is general for one could show that the highest component of any super-
field, i.e. the θθθ̄θ̄ component, always transforms into a total derivative as already
stated.
As we now know, λ and λ̄ are the lowest dimensional components of V which
are gauge invariant. They are actually also the lowest components of the field
strength Wα and W̄α̇ defined by
1
Wα = − D̄2 Dα V (5.19)
4
1 2
W̄α̇ = − D D̄α̇ V. (5.20)
4
These fields are chiral and antichiral respectively (because D3 = D̄3 = 0) and
gauge invariant for
α̇
D̄2 Dα V(φ + φ† ) = D̄2 Dα Vφ = (Dα D̄2 + 4iσm
αα̇ D̄ ∂m )φ = 0

where we used (4.16) and the chirality of φ. The component expansion of Wα in

the WZ gauge is easier to calculate in the y representation of V. We first show
that
1
V(y, θ, θ̄) = −θσm θ̄Vm (y) + iθθθ̄λ̄(y) − iθ̄θ̄θλ(y) + θθθ̄θ̄[D(y) − i∂m V m (y)].
2
Thus
∂
!
α̇
Dα V(y, θ, θ̄) := Vα (y, θ, θ̄) = ∂α + 2iσm
αα̇ θ̄ V(y, θ, θ̄)
∂ym
= −Vm (y)(σm θ̄)α + 2iθα θ̄λ̄(y) − iθθλa (y)

β i m n β 
+θ̄θ̄ δα D(y) − (σ σ̄ )α Vmn (y) θβ + θθ(σn ∂n λ̄(y))α
2
with Vmn = ∂m Vn − ∂n Vm . Expressing D̄α̇ in terms of the variable y gives
1 1
Wα (y, θ) = − D̄2 Dα V(y, θ, θ̄) = − (−∂¯ α̇ ∂¯ α̇ )Vα (y, θ, θ̄)
4 4
hence

β i β

Wα (y, θ) = −iλα (y) + δα D(y) − (σm σ̄n )α Vmn (y) θβ + θθ(σn ∂n λ̄(y))α . (5.21)
2
and do not depend on θ̄ as expected (because it is a chiral field). Using the fact
that V = V † and D̄† = D we see that W̄α̇ = (Wα )† yielding the expansion of W̄α̇
from the one of Wα .

Daniel Farquet EPFL - Physics

 21

Constructing a gauge invariant and susy Lagrangian is now an easy task since
we know that the W’s are chiral and gauge invariant. As we still want Lorentz
invariance, we take the θθ component of W α Wα to be part of the Lagrangian for
the spinor indices must be contracted. Let us remark that for any chiral super-
field φ(x, θ, θ̄)|θθ = φ(y|0 = x, θ)|θθ as can be checked from the explicit expansions
(5.3) and (5.4). The field Wα is a chiral field and hence W α Wα is chiral as well.
Recalling that by W α Wα we mean W α (x, θ, θ̄)Wα (x, θ, θ̄) we see that in this case
we also have W α Wα |θθ = W α (y|0 = x, θ)Wα (y|0 = x, θ)|θθ . This means that during
our calculations of the Lagrangian we can use expression (5.21) for taking the θθ
component and replace y by x. So the susy formulation of a gauge invariant free
vector field is given by
1
L = (W α Wα |θθ + W̄α̇ W̄ α̇ |θ̄θ̄ ) (5.22)
4
which can be found very easily. Indeed it is straightforward that
!
1 1 i m n 
αβ i l p αβ 
L = −2iλσ ∂m λ̄ + εαβ D − (σ σ̄ )αβ Vmn ε D − (σ σ̄ ) Vlp + h.c.
m
4 2 2 2
!
1 2 1 m n l p αβ
= 2i∂m λσ λ̄ + D − (σ σ̄ )αβ (σ σ̄ ) Vmn Vlp + i · (something real) + h.c.
m
4 8
!
1 2 1 ml np
= 2i∂m λσ λ̄ + D − η η Vmn Vlp + i · (something real) + h.c.
m
4 2
!
1 2 1 mn 1
= 2i∂m λσ λ̄ − 2iλσ ∂m λ̄ + D + D − Vmn V − Vmn V
m m 2 mn
4 2 2
which finally gives
1 1
L = i∂m λσm λ̄ + D2 − Vmn V mn . (5.23)
2 4
We recognise the theory of a free massless vector field, a free massless Majorana
spinor and a non-dynamical field D. If this Lagrangian were to be used alone,
the non-dynamical field could be left out since its equation of motion is D = 0.
The latter Lagrangian is not the most general supersymmetric abelian action. In
fact we know that the D component of the vector field V transforms into a total
derivative and is gauge invariant. Therefore the most general action involving a
single vector field V in the WZ gauge is
Z Z Z
1 α 1 α̇
SV = d xd θW Wα +
4 2
d xd θ̄W̄α̇ W̄ + ξ d4 xd2 θV (5.24)
4 4
4 4
Z " #
4 1 α α̇
= d x (W Wα |θθ + W̄α̇ W̄ |θ̄θ̄ ) + ξD (5.25)
4
Z !
1 2 1
= d x i∂m λσ λ̄ + D − Vmn V + ξD
4 m mn
(5.26)
2 4
where ξ is a real parameter.

Daniel Farquet EPFL - Physics

 22

6 Matter Coupling and Non-abelian Gauge Theory

6.1 Construction of the Lagrangian
In what follows we shall discuss how to couple a gauge vector field to a super-
symmetric multiplet of matter. As there is no fundamental difference between the
abelian and non-abelian case we will only consider the latter.
Let us consider a chiral superfield φ, i.e. D̄α φ = 0. The simplest gauge sym-
metry is a U(1) symmetry that can be generalised to

φ → φ0 = e−iΛ φ with Λ = Λa T a (6.1)

The matrices T a are the hermitian generators of the gauge group in the represen-
tation defined by the chiral field φ. Now φ is a column vector and φ† becomes a
row vector. The notation † now makes complete sense because it has to be under-
stood as a real matrix †, i.e. transposed and complex conjugate. Each component
field becomes a vector as well. For example, A(x) is a column vector whilst A† (x)
is its complex transposed version. The Λa ’s must be chiral superfields in order
for φ0 to still be a chiral superfield. (In QFT the Λa ’s are just functions of x. Here
we have to promote then to chiral fields or φ0 is not a chiral field anymore.) We
normalise our generators by

Tr T a T b = kδab , k>0 (6.2)

k depending on the chosen representation. With this convention, the structure

constants tabc
[T a , T b ] = itabc T c (6.3)
are completely antisymmetric. The kinetic part of the Lagrangian of the chiral
field is not invariant under this tranformation:
†
φ† φ → φ0† φ = φ† eiΛ e−iΛ φ

since Λ†a , Λa . The invariance can be recovered by use of the gauge field V. In
this case we extend the abelian transformation law of V to the non-abelian gauge
transformation
0 †
eV → eV = e−iΛ eV eiΛ (6.4)
where V = Va T a . In its infinitesimal form, it turns out that
Λ + Λ†
" #
δV = V − V = iΛ − iΛ + i V,
0 †
+... (6.5)
2

and we see that we recover the abelian case of the last section if tabc = 0 and
φ = iΛ. The new invariant kinetic term for the chiral field is given by

L = φ† eV φ|θθθ̄θ̄ (6.6)

Daniel Farquet EPFL - Physics

 23

as it is straightforward to verify. The susy field strength is generalised to

1
Wα = − D̄2 e−V Dα eV (6.7)
4
1 2 −V
W̄α̇ = − D e D̄α̇ eV (6.8)
4
and we will see later on that it reduces to the abelian case if tabc = 0. It is not
difficult to verify that
Wα → Wα0 = e−iΛ Wα eiΛ (6.9)
that is to say Wα transforms in the adjoint representation. An invariant Lagrangian
for the field V can be taken to be

L = Tr W α Wα |θθ + W̄α̇ W̄ α̇ |θ̄θ̄

 
(6.10)

where the trace is understood to be the trace over the matrices T a (remember that
V, W, φ and Λ are matrices). It is clear that this Lagrangian is invariant under
susy transformations because we select the θθ component of a chiral field and the
θ̄θ̄ component of a antichiral field. The Lagrangian is also invariant under gauge
transformations as it is clear from (6.9) using the cyclic property of the trace.
We are thus ready to write down the Lagrangian of an interacting non-abelian
gauge theory:
1 
α α̇

L= Tr W W |
α θθ + W̄ α̇ W̄ |θ̄θ̄ + φ e φ|θθθ̄θ̄ + W(φ)|θθ + W̄(φ )|θ̄θ̄ + ξ V |θθθ̄θ̄
† V † a a
16kg2
(6.11)
The normalisation of the gauge kinetic term is chosen to recover the canonical
normalisation for the component action after a rescaling of V, to be seen later on.
In order to satisfy gauge invariance the potential W(φ) cannot be any function
of φ. This is of course the same for its complex conjugate version W̄(φ† ). The
last term, ξa Da in the WZ gauge, always transforms into a total derivative under
susy transformations. Unfortunately by (6.5) the D component is gauge invariant
only in the abelian case. Therefore the ξa ’s are parameters that are also present
to restrict the Da ’s to the abelian sector of the theory. This term is called the
Fayet-Iliopoulos term.

6.2 Calculation of the Component Lagrangian

To understand the physics behind the Lagrangian (6.11) it is much more conve-
nient to consider its component expression. Let us first of all focus on the gauge
field part. From equation (6.5) we deduce that it is possible to impose the WZ
gauge at the local level. Hence eV = 1 + V + 21 V 2 . In this gauges Wα becomes

1 1
Wα = − D̄2 Dα V + D̄2 [V, Dα V].
4 8

Daniel Farquet EPFL - Physics

 24

The first part is just the same as equation (5.19) and its component expression is
given by (5.21) because we did the calculation in such a way as to never commute
any quantity. In the y representation the commutator is
1
[V, Dα V] = [Vm , Vn ](y)(σn σ̄m θ)α θ̄θ̄ − i[Vm , (σm λ̄)α ](y)θθθ̄θ̄
2
and then
1 2 1 1 i
D̄ [V, Dα V] = − ∂α̇ ∂α̇ [V, Dα V](y) = − [Vm , Vn ](y)(σn σ̄m θ)α + [Vm , (σm λ̄)α ](y)θθ
8 8 4 2
These two contributions can be added to the non-abelian one and we end up with

β i m n β 
Wα = −iλα (y) + δα D(y) − (σ σ̄ )α Vmn (y) θβ + θθ(σn Dn λ̄(y))α (6.12)
2
where
1 1
Dn λ̄a := ∂n λ̄a − tabc Vmb λ̄c a
Vmn := ∂m Vna − ∂n Vma − tabc Vmb Vnc . (6.13)
2 2
As stated above, it is clear that this expression reduces to the abelian one when
tabc = 0. The calculation of Tr(W α Wα |θθ ) can be carried out in the same way that
we found (5.23), replacing ∂n by Dn and taking Vmn to be given by (6.13). Using
Tr T a T b = kδab we have
!
α a a 1 a
Tr(W Wα |θθ ) = k D D − Vmn V amn
− 2iλ σ Dm λ̄ + i(something real)
a m a
2
and hence
1 
α α̇
 1 1 a a 1 a amn
2
Tr W Wα |θθ + W̄α̇ W̄ |θ̄ θ̄ = D D − Vmn V (6.14)
16kg 4g2 2 4

−iλ σ Dm λ̄
a m a

As it is clear from (6.13) and (6.14) the gauge field part of the Lagrangian does
not depend on the representation chosen for φ, reminding us of Yang-Mills theory
in QFT.
The kinetic part of the chiral Lagrangian given by φ† eV φ|θθθ̄θ̄ produces the
coupling between the matter fields in φ and the gauge fields. It is important to
remember that φ is a column vector here. We have previously learnt that this term
must be calculated in the x representation. Thus
1 † m √ √ 
φ† Vφ|θθθ̄θ̄ = i[A V ∂m A − ∂n A† V n A] + ψ̄σ̄m Vm ψ + 2iA† λψ + A† DA − 2iψ̄λ̄A
2
and
1 † 2 1
φ V φ|θθθ̄θ̄ = − A† Vm V m A.
2 4
Daniel Farquet EPFL - Physics
 25

Using the fact that we work in the WZ gauge, i.e.

1
φ† eV φ = φ† φ + φ† Vφ + φ† V 2 φ
2
and equation (5.10) for the first term, we get
i
φ† eV φ|θθθ̄θ̄ = F † F − ∂m A† ∂m A − iψ̄σ̄m ∂m ψ + √ (A† λψ − ψ̄λ̄A)
2
1 † 1 m i † m i
+ A DA + ψ̄σ̄ Vm ψ + A V ∂m A − ∂n A† V n A
2 2 2 2
1 †
− A Vm V m A (6.15)
4
In order to have a canonical normalisation for the gauge field part, we rescale the
gauge field V to 2gV. This means

D → 2gD
λ → 2gλ
Vm → 2gVm .

Collecting all the terms and using the indices i, j to run from 1, . . . , N where N is
the dimension of the representation of φ, the Lagrangians reads
1 a amn 1
L = − Vmn V − iλa σm Dm λ̄a + Da Da
4 √ 2 ∗ a
+gD Ai T i j A j + gξ D + i 2g(Ai T i j ψ j λa − λ̄a ψ̄i T iaj A j )
a ∗ a a a

−iψ̄i σ̄m Dm ψi − Dm A∗i Dm Ai + Fi∗ Fi (6.16)

1 1
+Wi Fi + W̄ī Fi∗ − Wi j ψi ψ j − W̄ī j̄ ψ̄i ψ̄ j
2 2
with

Dm A = ∂m A + igVma T a A
Dm ψ = ∂m ψ + igVma T a ψ (6.17)
Dm λa = ∂m λa − gtabc Vmb λc
a
Vmn = ∂m Vna − ∂n Vma − gtabc Vmb Vnc

It is clear that the fields Da and Fi are auxiliary. Their equations of motion are

Fi∗ = −Wi
Da = −gA∗i T iaj A j − gξa . (6.18)

Daniel Farquet EPFL - Physics

 26

When substituted into the Lagrangian, we have

1 a amn
L = − Vmn V − iλa σm Dm λ̄a
4
−iψ̄i σ̄m Dm ψi − Dm A∗i Dm Ai
√
+i 2g(A∗i T iaj ψ j λa − λ̄a ψ̄i T iaj A j ) (6.19)
1 1
− Wi j ψi ψ j − W̄ī j̄ ψ̄i ψ̄ j − V(A, A∗ )
2 2
where the scalar potential is
X 1X a 2 X 1 X
V(A, A∗ ) = |Fi |2 + (D ) = Wi W̄ī + g2 (A∗i T iaj A j + ξa )2 . (6.20)
i
2 a i
2 a

The first two lines of the Lagrangian are just kinetic terms that have nothing
particular. The third line exhibits an interaction between the fields Ai , ψi and the
‘gauginos’ λa (susy partner of Vma ) whose strength is given by the gauge coupling
constant g.

7 Supersymmetry Breaking
7.1 General Features
In the algebra (3.1) we saw that Pm commutes with the charges Q yielding that
P2 commutes with the charges Q. Since P2 is nothing but the mass, we must
conclude that the susy algebra implies that all the masses are the same inside
a multiplet. As we have not observed any supersymmetric particles up to now,
susy must be broken in some way in order to allow a mass spectrum that can be
realistic, or at least non-contradictory. The algebra (3.1) also gives

{Qα , Q̄β̇ } = 2σm P

αβ̇ m

which multiplied by σ̄nβ̇α turns to

{Qα , Q̄β̇ }σ̄nβ̇α = −4Pn .

By definition P0 is the Hamiltonian density H. Taking the n = 0 component of

the latter relation gives
1
H = (Q1 Q̄1 + Q2 Q̄2 + Q̄1 Q1 + Q̄2 Q2 ). (7.1)
4
This equation tells us that hψ|H|ψi ≥ 0 for any state |ψi. Hence, in supersymmetry
the energy is always positive. Moreover, it tells us that states with a vanishing
energy density are the supersymmetric ground states of the theory. Indeed they

Daniel Farquet EPFL - Physics

 27

are clearly ground states, but more particularly susy ground states since h0|H|0i =
0 implies that Q|0i = 0 and Q̄|0i = 0 yielding δ s |0i = 0, i.e. supersymmetry is
not broken. On the other hand ground states |φ0 i with a non-vanishing energy
density must break supersymmetry because it must be the case that Q|φ0 i , 0 and
Q̄|φ0 i , 0.
The question that arises now is the following: How can we know whether
supersymmetry is broken or not for the Lagrangian (6.19)? This Lagrangian is
highly non-linear and the study of its dynamics would not be a piece of cake.
The trick here is to use perturbation theory around the vacuum of the theory
(susy or not). The standard choice is to allow only constant values of the Lorentz
invariant fields for this vacuum configuration (because Lorentz invariance cannot
be broken). In other words the only fields that can acquire a vacuum expectation
value (vev) are the Ai ; their vev is denoted by hAi i and we have ∂m hAi i = hψi i =
hλa i = hVma i = 0. The theory results in evaluating the Lagrangian on these various
vev which gives L = −V(hAi, hA∗ i). Hence, the equations of motion are given by
∂V
= 0. (7.2)
∂Ai Ai =hAi i

Furthermore, we want to impose that the theory be a stable theory. Two crite-
ria are to be considered. The first is that hAi i is a local minimum of the potential,
not necessarily a global one. The local stability imposes that the matrix

 ∂ V ∗ ∂2 V 
 2 
 ∂Ai ∂A j ∂Ai ∂A j 
 ∂2 V ∂2 V   , (7.3)
∂Ai ∂A j∗ ∂Ai ∂A j
∗ ∗

which is the mass square matrix of scalar fields Ai as it will be clear later on, has
only positive or zero eigenvalues. At the quantum level, global stability has to be
imposed because there could be some false vacuum decay. It means that

V(hAi, hA∗ i) ≤ V(A, A∗ ). (7.4)

Depending on the dynamics and the potential it is possible that the false vacuum
has a life-time comparable to the age of the universe and is then an acceptable
vacuum. We will not discuss this point and suppose that local stability is enough.
In this setup, the Hamiltonian of the theory is simply H = V(hAi, hA∗ i). Hence
the vacuum energy is given by V(hAi, hA∗ i). Therefore susy is spontaneously bro-
ken if and only if the minimum of the potential is greater than 0 and equivalently
by a non-zero expectation value of the auxiliary fields hFi i and/or hDa i.
Before moving up to a more quantitative analysis, we briefly discuss the exis-
tence of the Goldstone particle associated to susy breaking. On general grounds
we know that a massless particle appears when a symmetry is spontaneously bro-
ken. In QFT this particle is know as the Goldstone boson. In susy this particle
cannot be a boson but is a fermion called the goldstino, as we shall now prove.

Daniel Farquet EPFL - Physics

 28

Let |Ωi be the ground state of the theory and G the Goldstone particle (i.e field)
associated to susy breaking. The condition of breaking can be stated by saying
that the vev of the susy transformation of G is different from zero:

hΩ|δ s G|Ωi , 0. (7.5)

The field G cannot be a boson because its variation are fermions and a vev of
fermions breaks Lorentz invariance. Therefore it must be a fermion.
Since the goldstino is massless the property that all the masses are the same
inside a multiplet is no longer valid. One of the signatures of a susy breaking is
therefore the fact that the mass spectrum is modified.

7.2 The Mass Formula

A very particular feature of supersymmetric theories is the existence of a mass
formula valid for all possible vacua, breaking spontaneously or preserving super-
symmetry, relating the various masses of all fields present in the theory.
The first mass formula that can be considered is the one where susy is not
broken: all states belonging to a given supermultiplet have the same mass. In
susy theories we know that the number of bosonic states is equal to the number
of fermionic states inside a multiplet. We could therefore redefine the supertrace
in such a way that

STr M 2 = M 2 (number of bosonic states − number of fermionic states) = 0 (7.6)

In our case this is achieved by setting

STr M 2 := 3 Tr M12 − 2 Tr M1/2

2
+ Tr M02 (7.7)

with MN/22 the square mass matrix of spin N/2 particles. A very powerful result
is that this formula almost holds for any arbitrary vacuum, including those who
break supersymmetry. The appearance of the goldstone boson modifies the mass
spectrum but in a way that satisfies a certain formula. This is the object of the
next theorem.

Theorem 7.1. For supersymmetric theories whose Lagrangian is given by (6.19)

we have the following formula around any kind of vacuum (supersymmetric or
not)
STr M 2 = −2ghDa i Tr(T a ) (7.8)
where the sum over a is understood.

Proof. The goal of this proof is the calculation of the mass square matrix for
the various spins. As explained above the only fields that can acquire a vev are
the Ai and thus A∗i . The vev of the other fields vanish. The first very crude

Daniel Farquet EPFL - Physics

 29

approximation was to say that L = −V(hAi, hA∗ i). In order to obtain masses we
have to consider small fluctuations around vacuum, that is to say that we will
work with the Lagrangian (6.19) bearing in mind that Vma , ψi and λa are small
fluctuations around their vev, i.e. 0. Keeping the bilinear part of the Lagrangian
will allow us to calculate the matrices.
We begin with the spin 1 matrix. In general when scalar fields acquire a vev,
the gauge bosons become massive. It is related to the presence of
−g2 Vma V bm hA† iT a T b hAi
in Dm A∗i Dm Ai . The bilinears in the vector fields in the Lagrangian are
1 a amn 2 † a b
− Vmn V + g hA iT T hAiVma V bm
4
which gives
(M12 )ab = g2 hA† iT a T b hAi + g2 hA† iT b T a hAi
because the mass matrix has to be symmetric in order to be diagonalisable. Thus
Tr M12 = 2g2 hA† iT a T a hAi
For the spin 1/2 mass matrix we collect all terms that are bilinear in fermionic
fields with possible vev hAi i and hA∗i i. We get
√ 1 1
−iλa σm ∂m λ̄a −iψi σm ∂m ψ̄i +i 2(hA† iT a ψλa − λ̄a ψ̄T a hAi)− hWi j iψi ψ j − hW̄ī j̄ iψ̄ī ψ̄ j̄ .
2 2
This can be rewritten as
1h a m
−iλ σ ∂m λ̄a + i∂m λa σm λ̄a − iψi σm ∂m ψ̄i + i∂m ψi σm ψ̄i
2 √ √ i
−hWi j iψi ψ j − hW̄ī j̄ iψ̄i ψ̄ j + i2 2ghA† iT a ψλa − i2 2gλ̄a ψ̄T a hAi (7.9)
or more conveniently
  √ ∗ iT b 
 ! 
1     hW i j i −i 2ghA ψ j
kinetic terms − ψi λa  √ +
j ji 
 
 
 λb h.c.  (7.10)
−i 2ghA∗i iT iaj

2 0

Notice that the matrix has been written in a symmetric form in order to be easily
identified with the mass matrix M1/2 ,
√
−i 2ghA∗j iT bji 
 
hWi j i
M1/2 =  √  .

−i 2ghA∗i iT iaj 0

The complex conjugate is

 √ 
†
 hW̄ī j̄ i i 2gT ibj hA j i
M1/2 =  √  .
i 2gT aji hAi i 0

Daniel Farquet EPFL - Physics

 30

2 =M †
Since the fermionic mass square matrix is M1/2 1/2 M1/2 , cf. appendix section
10.6, we find
2
Tr M1/2 = hWi j ihW̄ī j̄ i + 4g2 hA† iT a T a hAi
The last thing we need is the scalar square mass matrix. The corresponding
part of the Lagrangian is
−∂m A∗i ∂m Ai − V(A, A∗ ).
Near the minimum we can expand Ai around hAi i via hAi i + ∆Ai with ∆Ai small.
The bilinear terms are in ∆Ai
 
1 ∂2V ∂2V 1 ∂2V
− ∂m ∆A∗i ∂m ∆Ai + h i∆Ai ∆A j + h i∆Ai ∆A∗j + h ∗ ∗ i∆A∗i ∆A∗j  .
 
2 ∂Ai ∂A j ∂Ai ∂A j
∗ 2 ∂Ai ∂A j

The equations of motion for ∆Ai are

!  h ∂2 V ∗ i h ∂2 V i 
 
∆Ai  ∂Ai ∂A j
∗  ∂Ai ∂A j  ∆A∗!
j = 0.
 −


∆Ai h ∂∗2 V ∗ i h ∂∗2 V i ∆A j
∂A ∂A ∂A ∂A j
| i j {z i }
=M02

Therefore near vacuum Tr M02 = 2h ∂A∂i ∂A V 2

∗ i. The derivatives of V are not very
i
difficult to calculate. Using (6.18) and (6.20) we find
∂V
= hWi j ihW̄ j̄ i − ghDa ihA∗j iT aji
∂Ai
∂ V
2
= hWi j ihW̄ī j̄ i + g2 hA† iT a T a hAi − ghDa i Tr T a .
∂Ai ∂A∗i
By collecting the various expressions and calculating equation (7.8) one has
STr M 2 = −2ghDa i Tr T a ,
as had to be shown.
Let us briefly discuss how one can interpret the latter result. If STr M 2 = 0
it means that there must exist at least one susy particule which is lighter than
its partner. As this fact has never been observed in experiments we know that
it cannot be the case in nature. Therefore it would be very convenient to have
STr M 2 , 0. If one is to consider the gauge group to be either S U(N) or U(1), as
is the case in the standard model, one is forced to conclude that the gauge group is
a U(1) group because Tr T a = 0 for the generators of S U(N). Such models have
been constructed but are not satisfactory regarding phenomenology. A deeper
discussion of this issue is presented in [4].
Regarding gauge groups other than U(1) there are several non-trivial results
for the theory we have constructed so far.

Daniel Farquet EPFL - Physics

 31

7.3 The U(N) Gauge Group

In this little section we want to show that the potential W(φ) cannot be gauge
invariant and depend on the fields φi if we were to choose the gauge group to be
U(N) in its fundamental representation. In other words W(φ) = 0 (or a constant
that can be chosen to be 0 without loss of generality) for such a gauge group.
This statement is not trivial at all but the proof is very simple.
Theorem 7.2. Let the gauge group be U(N), N > 1, in the fundamental represen-
tation. Then imposing the gauge invariance of W(φ) forces W(φ) = 0.
Proof. Let us first recall how φi transforms under a gauge transformation:
δg φi = −iΛa T iaj φ j .
(Note that we sometimes use upper and lower indices for the gauge generators
and parameters. This has no real significance but allows us to write more read-
able expressions.) The statement of the gauge invariance of W can be written
δg W(φ) = 0, or
∂W ∂W
δg φi = 0 =⇒ T iaj φ j = 0 ∀a
∂φi ∂φi
Multiplying by T kla (the sum over a is understood) and using the identity (10.61)
proved in the appendix in the case N > 1 (Be careful with the notation because it
is a bit sloppy. Here we use a that runs over all the generators of the U(N) group
but we used A for U(N) in the appendix. This was so only in order to simplify
the proof of this relation), we have
∂W ∂W
T kla T iaj φ j = 0 =⇒ φk =0
|{z} ∂φi ∂φl
kδil δ jk

Since gauge invariance must be valid everywhere we conclude that ∂W

∂φl = 0 for all
l, yielding that W(φ) =constant.
Only the θθ component of W(φ) appears in the Lagrangian, therefore without
loss of generality we can chose W(φ) = 0.

7.4 The S U(N) Gauge Group

In the case where the gauge group is S U(N) in the fundamental representation
the behavior of the theory when susy is spontaneously broken is very much con-
strained: the F terms control the breaking. Let us first announce a theorem and
then discuss it.
Theorem 7.3. If the gauge group is S U(N), N > 1, in the fundamental repre-
sentation and the Lagrangian is given by (6.19) (with ξa = 0), then around any
minimum of the scalar potential V(A, A∗ ) we have
X X
k (Da )2 ≤ |Fi |2 . (7.11)
a i

Daniel Farquet EPFL - Physics

 32

Proof. We previously found that

(M12 )ab = g2 hA† iT a T b hAi + g2 hA† iT b T a hAi
Da = −ghA† iT a hAi
where we set ξa = 0. To these relations we add
|Da (M12 )ab Db | = |2gDa F † T a F|
that will be proved later on in a more general context, see (9.17). The trick of the
proof is to calculate |Da (M12 )ab Db | using the latter equality and see that it leads to
the desired inequality.
1. Calculation of |Da (M12 )ab Db |: Using the explicit expression for Da and
(M12 )ab but not Db we find

|Da (M12 )ab Db | = |ghA∗i iT iaj hA j i2g2 hA∗k iT kla T lm

b
hAm iDb |.

In the appendix, equation (10.60), we show that T iaj T kla = k(δil δ jk − N1 δi j δkl )
for S U(N), N > 1, in the fundamental representation. It leads to

2kg2 †
|Da (M12 )ab Db | = 2kg2 hA† Ai (−g)hA† iT b hAi Db − hA Ai (−g)hA† iT b hAi Db
| {z } N | {z }
Db Db

Bearing in mind that A† A is nothing but an euclidian complex scalar prod-

uct we denote A† A = |A|2 . With this notation we can write
!
1 X
|Da (M1 ) Db | = 2kg 1 −
2 ab 2
|hAi|2 (Da )2 .
N a

2. Calculation of |2gDa hF † iT a hFi|: In components this quantity becomes

|2gDa hF † iT a hFi| = |2g2 hA∗i iT iaj hA j iFk∗ T kla Fl |
2g2 †
= 2g2 hA† iFF † hAi − hA AiF † F .
N

Using A† F ≤ |A||F|, a banal property of the scalar product, we have

!
† a 2 1
|2gDa hF iT hFi| ≤ 2g 1 − |hAi|2 |F|2
N

Using the fact that N > 1 we can conclude that

X X
k (Da )2 ≤ |F|2 = |Fi |2 .
a i

Daniel Farquet EPFL - Physics

 33

In the last section we found that susy is spontaneously broken if and only
if the scalar potential V(A, A∗ ) = i |Fi |2 + 21 a (Da )2 is different from zero. The
P P
previous result states that the breaking of supersymmetry is completely controlled
by the F terms because if they are zero so are the Da . Taking the converse, it is
never possible to have even a single non-zero Da whilst the Fi are all zero.

7.5 Examples
Up to now we have dealt with a superpotential W(φ) that could take any functional
form. However, if one is looking for renormalisable models the potential cannot
be arbitrary. In order to be renormalisable the couplings must have a positive or
zero dimension in energy.
We have already derived that the Grassmann variables have dimension −1/2.
What about φi ? As we know, the first component of φi is a scalar field Ai which
surely has dimension one, hence φi has dimension one. In order for the La-
grangian to be a number W(φ) has to have dimension three (d4 x has dimension
−4 and d2 θ dimension +1 ). Therefore the most general possible formulation of
W(φ) is
W(φ) = λk φk + mik φi φk + gi jk φi φ j φk . (7.12)
For such a potential the equations of motion for Fk∗ are

Fk∗ = −(λk + mik Ai + gi jk Ai A j ).

Supposing we worked in a theory where there was no gauge field at all (simply
put all fields appearing in V a to 0 in the Lagrangian (6.19)) the scalar potential
would be V(A, A∗ ) = Fk∗ Fk . Breaking supersymmetry would amount to finding
parameters such that
0 = λk + mik Ai + gi jk Ai A j
does not have a solution in Ai . Such models are called O’Raifeartaigh models.
The simplest one is realised with three chiral fields and the potential

W(φ) = λφ1 + mφ2 φ3 + gφ1 φ2 φ2 .

With this potential and the kinetic term φ† eV φ the Lagrangian is renormalisable.
The various mass matrices can be easily calculated using the explicit form of
W(φ) because we have already derived a generic expression for them.
Another famous example is the Fayet-Iliopoulos model. This is the theory of
two U(1) gauged fields φ1 and φ2 with charge +e and −e repectively (i.e. super
QED) where we add the Fayet-Iliopoulos term. The Lagrangian is
1
L = (Wα W α |θθ + W̄ α̇ W̄α̇ |θ̄θ̄ ) + φ†1 eeV φ1 |θθθ̄θ̄ + φ†2 e−eV φ2 |θθθ̄θ̄
4
+m(φ1 φ2 |θθ + φ†1 φ†2 |θ̄θ̄ ) + 2κV|θθθ̄θ̄

Daniel Farquet EPFL - Physics

 34

whose component expansion can be easily found by comparing to (6.11) and

using (6.19). This Lagrangian spontaneously breaks supersymmetry with its D
and F terms.
These two examples are treated in [1] and [2] as well as other examples. The
respective mass spectra are calculated depending on the parameters appearing in
the Lagrangians. Since these books already treat examples in detail we do not
expound them here.

8 The Non-linear Chiral Sigma Model

As we have seen, renormalisability imposes strong constraints on the possible
forms of a supersymmetric theory. If one insists that supersymmetry is not a
fundamental theory but only an effective theory, then renormalisability is not a
guideline anymore.
This means that we can go further and generalise our previous developments.
In order to stick to standard notations, we will define φ̄ to be equal to φ† , i.e.
φ̄ := φ† . We recall that φ is a column vector yielding φ̄ as a row vector, thus φ̄φ
is a scalar product. We also replace the ∗ by a bar to mean complex conjugation
and use an index with a bar when it is complex conjugation, for example Fī∗ = F̄ī
or A∗ī = Āī .
In this section we will restrict ourselves to the development of a Lagrangian
involving chiral fields only. We generalise to the situation where a gauge cou-
pling is present in the next section. Let K(φ, φ̄) be any real function of φ and φ̄,
W(φ) the potential and W̄(φ̄) its complex conjugate. The most general Lagrangian
involving φ alone and only up to two derivatives of the fields is
Z Z Z
L = d θ K(φ, φ̄) + d θ W(φ) + d2 θ̄ W̄(φ̄)
4 2
(8.1)

The potential part has already been calculated. The part arising from K is not
known yet. In order to calculate it one can use projection techniques as we did
with the potential part, i.e.
Z
1
d4 θ K(φ, φ̄) = D̄2 D2 K(φ, φ̄)|0
16
where we dropped boundary terms. Since the calculation does not give any new
insight into the theory but is rather just an unpleasant technical development, we
will not expound it. Before giving the result we need to point out a detail that was
not important before but becomes crucial here. The fields φi have carried lower
indices up to now but, form now on, we change our convention and use an upper
index instead, φi . At first sight this seems to be of no importance because these
indices do not refer to any metric and can be put wherever one likes. This will not
be the case at the end of the calculation for we will introduce a metric for such

Daniel Farquet EPFL - Physics

 35

indices. In other words lower and upper will have a meaning and we cannot put
the index wherever we like.
In components the kinetic term K is
Z
1 1 1 ¯
d4 θ K(φ, φ̄) = Ki j̄ F i F̄ j̄ − Ki j̄k̄ F i ψ̄ j̄ ψ̄k̄ − Kī jk F̄ ī ψ j ψk + Ki jk̄l¯ψi ψ j ψ̄k̄ ψ̄l
2 2 4
−Ki j̄ ∂m Ai ∂m Ā j̄ − iKi j̄ ψ̄ j̄ σ̄m ∂m ψ − iKi jk̄ ψ̄k̄ σ̄m ψi ∂m A j
∂K ∂K
where again Ki = ∂A i , Kī = ∂ Āi and so on. Using the notation of Kähler geometry
(appendix, section 10.8) and adding the potential terms we find
" # " #
i j̄ 1 i k j̄ l¯ i 1 l¯ ī 1
L = gi j̄ F F̄ + gi j̄,kl¯ψ ψ ψ̄ ψ̄ − F gil¯Γ j̄k̄ ψ̄ ψ̄ − Wi − F̄ glī Γ jk ψ ψ − W̄ī
j̄ k̄ l j k
4 2 2
1 1
−gi j̄ ∂m Ai ∂m Ā j̄ − igi j̄ ψ̄ j̄ σ̄m Dm ψi − Wi j ψi ψ j − W̄ī j̄ ψ̄ī ψ̄ j̄
2 2
with Dm ψi = ∂m ψi + Γijk ∂m A j ψk a contravariant tensor in the Kähler manifold (see
proposition 10.3 in the appendix). The equation of motion for the auxiliary fields
are
1
gi j̄ F i − gk j̄ Γkml ψm ψl + W̄ j̄ = 0
2
leading to the final component form
1 ¯
L = −gi j̄ ∂m Ai ∂m Ā j̄ − igi j̄ ψ̄ j̄ σ̄m Dm ψi + Ri j̄kl¯ψi ψk ψ̄ j̄ ψ̄l
4
1 1
− ∇i W j ψi ψ j − ∇ī W̄ j̄ ψ̄ī ψ̄ j̄ − gi j̄ Wi W̄ j̄ . (8.2)
2 2
The use of Kähler geometry leads to a clear interpretation. The scalar fields are
the coordinates in the Kähler manifold. If one suppose that the fermions are
tensors in the tangent space then the derivative Dm ψi is covariant and it is clear
that the Lagrangian is invariant under a change of coordinates. Moreover it is
also clearly Lorentz invariant.
We finally remark that under a Kähler transformation defined by

K(φ, φ̄) → K(φ, φ̄) + F(φ) + F̄(φ̄), (8.3)

with F a function of φ, the Lagrangian is invariant as well.

9 The Non-linear Gauge Invariant Sigma Model

We now want to introduce some gauge fields V a and couple them with the chiral
fields. The most general Lagrangian involving at most two derivatives is
Z Z " #
1
L = d θ K(φ, φ̄, V) + ξa V + d θ W(φ) + fab (φ)W Wα + hc. (9.1)
4  a 2 aα b
4

Daniel Farquet EPFL - Physics

 36

In this context K is the real Kähler potential, W a holomorphic superpotential, ξa

the Fayet-Iliopoulos constants, fab the holomorphic gauge kinetic functions and
the matrix Wα is still given by (6.7). In order to simplify a lot of the analysis of
the above Lagrangian, we restrict ourselves to the case where fab (φ) = δab . The
general case is treated in [8]. Our Lagrangian is then
Z Z " #
1 α
L = d θ K(φ, φ̄, V) + ξa V + d θ W(φ) + Tr(W Wα ) + hc.
4  a 2
(9.2)
4

Up to now we have stuck to the case of linear gauge transformations. Extending

to the case of non-linear transformations we define

δg φi = Λa Xai (φ) (9.3)

i 1
δg V a = − (Λa − Λ̄a ) + tabc (Λb + Λ̄b )V c + O(V 2 ) (9.4)
2 2
The Λa are the chiral fields that parametrise the transformation and Λ̄a is the
associated antichiral field. The Xai (φ) are functions of φ and allow non-linear
transformations. Note that the variation of V a has, by definiton, an opposite sign
with respect to section 6.5. Imposing gauge invariance of the Lagrangian imposes
strong constraints, particularly on K. In fact we impose that the holomorphic
terms must be strictly invariant, whilst the non-holomorphic part is at most a
Kähler transformation (remember that it is sufficient for the Lagrangian to be
invariant)
Λa fa (φ) + Λ̄a f¯a (φ̄).
For example, the gauge invariance of W(φ) reads δg W(φ) = Wi δg φi yielding

Xai Wi = 0. (9.5)

On the other hand the variation of K is δg K = Ki δg φi + Kī δg φī + Ka δg V a , where

Ka signifies the derivative with respect to V a . We recall that the Fayet-Iliopoulos
term is present and already gauge invariant if and only if tabc = 0. The constraint
of gauge invariance of K reads
i 1
Λa fa + Λ̄a f¯a = Λa Xai Ki + Λ̄a X̄aī Kī − (Λa − Λ̄a )Ka + tabc (Λb + Λ̄b )V c Ka + O(V 2 ).
2 2
Identifying the Λ leads to
i
fa = Xai Ki − Ka + O(V) (9.6)
2
Moreover, using gi j̄ := ∇ j̄ Ki |V=0 and applying ∇ j̄ to the last equation gives

i
Xai = gi j̄ ∇ j̄ Ka + O(V). (9.7)
2
Daniel Farquet EPFL - Physics
 37

This proves that Xai is a Killing vector for the Killing potential − 21 Ka (see [2]
for an introduction to isometries and Killing vectors for Kähler manifolds). This
provides an interpretation of the Fayet-Iliopoulos constants: they are the freedom
of adding a constant to the Killing potential for each U(1) factor, as it is clear
from the Lagrangian in the WZ gauge, see below.
To work out the component Lagrangian, it is convenient to use the WZ gauge:
Z " #
a 1
L = d θ K(φ, φ̄) + (Ka + ξa )V + Kab V V
4 a b
2
Z " #
1
+ d2 θ W(φ) + Tr(W α Wα ) + hc.
4

Taking the derivative of (9.6) with respect to Vb gives Kab = −2iXai Kib + O(V),
and since K is real Kab = 2iX̄aī Kīb + O(V). The use of equation (9.7) allows us
j̄
to write Kab = 4gi j̄ X̄a Xbi + O(V). With this relation at hand, the Lagrangian can
be expanded in components. The steps are pretty much the same as the ones in
section 6.2 but are more involved. After eliminating the auxiliary fields one has
1 a amn ¯
L = −gi j̄ Dm Ai Dm Ā j̄ − Vmn V / ψ̄ j̄ + Γ ¯j̄ D
− igi j̄ ψi (D / Āl ψ̄k̄ )
4 l k̄
i a a
− λ D / λ̄ + hc. − VS − VF . (9.8)
2
with
1
VS = gi j̄ Wi W̄ j̄ + (Ka + ξa )(Ka + ξa ) (9.9)
8
1 √ j̄ 1 ¯
VF = ∇i W j ψi ψ j + hc. + 2gi j̄ X̄a ψi λa + hc. − Ri j̄kl¯ψi ψk ψ̄ j̄ ψ̄l (9.10)
2 4
∂X i
where we have defined Dm Ai = ∂m Ai + Vma Xai , Dm ψi = ∂m ψi + Vma ∂Aaj ψ j , Dm λa =
∂m λa + tabc Vmb λc and Vmn m n n m m n / stands for
a = ∂ V a − ∂ V a + tabc V b V c . The notation D

the contraction of Dm with the sigma matrices. The auxiliary fields were
1
F i = −gi j̄ W̄ j̄ + Γijk ψ j ψk (9.11)
2
1
Da = − (Ka + ξa ). (9.12)
2

9.1 The Mass Formula

We want to derive a mass formula as we have already done in a less general case.
For simplicity we leave out the Fayet-Iliopoulus terms. Around a minimum the
only fields that are allowed to take a vev are the scalars. The auxiliary fields

Daniel Farquet EPFL - Physics

 38

become

F i = −W̄ i
1
Da = − Ka .
2
As before a very generic formula is valid.

Theorem 9.1. For supersymmetric theories whose Lagrangian is given by (9.8)

we have the following formula around any kind of vacuum (supersymmetric or
not)
STr M 2 = 2Ri j̄ F i F̄ j̄ + 2i∇i Xai Da . (9.13)

Proof. For the scalars we need to calculate Tr M02 := 2hgi j̄ ∂i V ī i. Using the vari-
2

∂A ∂Ā
ous properties of Kähler geometry, we have

∂2 V ¯
= ∇i Wk ∇ j̄ W̄ k − Ri j̄kl¯F k F̄ l + X̄ai Xa j̄ + i∇i Xa j̄ Da .
∂Ai ∂Ā j̄
Omitting the average for the sake of clarity we get

Tr M02 = 2∇i W j ∇i W̄ j + 2Ri j̄ F i F̄ j̄ + 2X̄ai Xai + 2i∇i Xai Da .

The two blocks of the fermionic mass matrix are easily identified by compar-
ing it with (7.9) and (7.10). We have

(M1/2 )i j = ∇i W j
√
(M1/2 )ia = 2X̄ai .
2 = Tr M †
1/2 M1/2 = (M1/2 )i j (M1/2 ) + 2(M1/2 )ia (M1/2 ) a , hence
The trace is Tr M1/2 ∗ ij ∗ i

2
Tr M1/2 = ∇i W j ∇i W̄ j + 4X̄ai Xai .

The vector mass matrix comes from −gi j̄ Dm Ai Dm Ā j̄ and is given by

(M12 )ab = X̄bi Xai + X̄ai Xbi (9.14)

since we need to symmetrise it so that it is diagonalisable. Its trace is

Tr M12 = 2X̄ai Xai

Thus the supertrace defined by STr M 2 = 3 Tr M12 − 2 Tr M1/2

2 + Tr M 2 is
0

STr M 2 = 2Ri j̄ F i F̄ j̄ + 2i∇i Xai Da

Daniel Farquet EPFL - Physics

 39

9.2 The Linear Case

The theory developed in the previous sections is rather general. If one let K(φ, φ̄, V) =
φ̄eV φ it is clear that the metric is gi j̄ = δi j̄ and the space is flat, i.e. the Rie-
mann tensor vanishes. Apart from some normalisation factors and signs that are
slightly different here, the results found in section 6 can be recovered if the gauge
transformation is linearised. For instance if instead of δg φi = Λa Xai (φ) one takes
δg φi = −iΛa (T a )i j φ j , i.e. Xai = −i(T a )i j φ j , the supertrace reads

STr M 2 = 2Da Tr T a (9.15)

which is nothing but the supertrace formula of equation (7.8) up to a normalisa-
tion (we give an explanation below for the sign and the factor g).
In section 7.4 when dealing with the special case of a S U(N) gauge group,
we used the relation |Da (M12 )ab Db | = |2gDa F † T a F| while postponing its proof.
With the formalism at hand it is now a comfortable enterprise to prove this rela-
tion. Near the vacuum of the theory the scalar potential is stationary. Using the
complex conjugate of (9.7) the condition VS i = ∇i VS = 0 gives
∇i Wk F k + iX̄bi Db = 0.
Multiplying this equality by 2Da Xai leads to
Da (M12 )ab Db = 2iDa Xai ∇i W j F j
where we used (9.14). We now differentiate the gauge invariance condition of W
namely Xai Wi = 0 to get ∇ j Xai Wi + Xai ∇ j Wi = 0. The identity ∇i W j = ∇ j Wi and the
last equation finally allows us to write that
Da (M12 )ab Db = 2i∇ j Xaī Da F̄ ī F j . (9.16)
This is just a more general form of the equation we are looking for. If the space
is flat, we can forget about the bar on the indices as well as the upper and lower
structure. Under the assumption of a linear gauge transformation ∇ j Xai = −iT ai j ,
yielding
Da (M12 )ab Db = 2Da F † T a F.
The final subtlety to be considered is the change of normalisation between the
Lagrangians (6.19) and (9.8). The important difference is that in the Lagrangian
(6.19) a coupling constant g appears whilst it does not in (9.8). This change
of normalisation only appears in the derivatives Dm of the different fields. One
can check that in order to recover this factor g and the right signs in (6.17) it
is sufficient to send T a to −gT a and tabc to −gtabc in this section. (This sign
difference comes from the sign difference in the definition of the gauge variation
of V a ) Thus the equality with the right normalisation is
|Da (M12 )ab Db | = |2gDa F † T a F|, (9.17)
as had to be shown.

Daniel Farquet EPFL - Physics

 40

10 Appendix
10.1 Convention and Reminder on Spinors
The following metric is used ηmn = diag(−1, 1, 1, 1). Roman indices always re-
fer to Lorentz transformations, i.e. they are reserved for transformations under
S O(1, 3). On the other hand, spinor indices will always be given by greek letters,
e.g. α or β etc...
Let M be a matrix in S L(2, C). We have the following representations of M,
denoted by D(M):

• The self representation of dimension 2, D1 (M) := M, ∀ M ∈ S L(2, C).

β
The transformation is written ψ0α = Mα ψβ where ψ belongs to F, a 2-
dimensional C vector space and α, β = 1, 2. Elements transforming accord-
ing to the latter relation are called left-handed Weyl spinors.

• The complex conjugate self representation defined by D2 (M) := M ∗ , ∀ M ∈

S L(2, C), where * stands for complex conjugation. The representation C
vector space is called Ḟ and its elements are denoted ψ̄. The transformation
β̇
reads ψ̄0α̇ = (M ∗ )α̇ ψ̄β̇ with α̇, β̇ = 1̇, 2̇. The elements ψ̄ are called right-
handed Weyl spinors.

Two other simple representations can be found, namely D3 (M) = M −1T and
D4 (M) = M −1† where † means the adjoint in the matricial sense. We recall that
the representation D3 can be found by conjugation starting from D1 and that D4
can be found by conjugation from D2 . If we introduce the matrices ε such that
! !
0 −1 αβ 0 1
(εαβ ) = , (ε ) = (εαβ ) =
−1
(10.1)
1 0 −1 0

and ! !
0 −1 α̇β̇ 0 1
(εα̇β̇ ) = , (ε ) = (εα̇β̇ ) −1
= (10.2)
1 0 −1 0
it follows that
γ
εαβ εβγ = δαγ , εαβ εβγ = δα (10.3)
and
γ̇
εα̇β̇ εβ̇γ̇ = δα̇γ̇ , εα̇β̇ εβ̇γ̇ = δα̇ . (10.4)
By defining that
ψα := εαβ ψβ thus ψα = εαβ ψβ (10.5)
and
ψα̇ := εα̇β̇ ψβ̇ thus ψα̇ = εα̇β̇ ψβ̇ (10.6)

Daniel Farquet EPFL - Physics

 41

it implies that the transformation properties are given by

β
ψ0α = Mα ψβ (10.7)
α
ψ0α = M −1T ψβ

(10.8)
β
∗
β̇
ψ̄α̇ = M α̇ ψ̄β̇
0
(10.9)
α̇
ψ̄0α̇ = M −1† ψ̄β̇

(10.10)
β̇
(10.11)

We now introduce a linear map φ from F to C:

φ:F→C (10.12)

such that for all ψ ∈ F

φ(ψ) := φα ψα ∈ C. (10.13)
It must be pointed out that such a bilinear form is invariant under S L(2, C). Ac-
cording to our notation, we must interpret two-component spinors with upper
indices as elements of the dual F ∗ of F. From equation (10.5), we conclude that
the ε matrix may be considered as a map form F to F ∗ . In very much the same
way, we can repeat this procedure for spinors with dotted indices, while taking
note that with our convention the ‘product’ is defined by ψ̄φ̄ = ψ̄α̇ φ̄α̇ . We come
to the conclusion that spinors with upper indices are elements of Ḟ ∗ and that the
ε matrix with dotted indices is a map from Ḟ to Ḟ ∗ . Like before, such a bilinear
form is invariant under S L(2, C).
Let us define a set of four matrices σm with

~ ) = (σ0 , σ
(σm ) := (−12×2 , σ ~) (10.14)

where σ0 is the 2 × 2 unit matrix and the (σi )i=1,2,3 are the usual Pauli matrices,
i.e. ! ! !
0 1 0 −i 1 0
σ =
1
σ =
2
σ =
3
. (10.15)
1 0 i 0 0 −1
which obey the following relations
n o
σi , σ j = 2δi j 12×2 (10.16)
h i
σi , σ j = 2iεi jk σk (10.17)
(10.18)

i, j, k = 1, 2, 3. The spinor index structure of the four matrices σm is such that

σm := (σmαα̇ ). (10.19)

Daniel Farquet EPFL - Physics

 42

The indices may be raised by application of the ε tensor leading to a new set of
matrices σ̄m , i.e.
(σ̄m )α̇α := εαβ εα̇β̇ (σm )ββ̇ . (10.20)
It is straightforward to verify that in matrix form we have

σ̄0 = σ0 and σ̄i = −σi . (10.21)

The matrices σ0 and σ̄0 allow us to write a map from F ∗ to Ḟ and back and also
from F to Ḟ ∗ and back:

ψα = −ψ̄∗α̇ (σ̄0 )α̇α (10.22)

ψ̄α̇ = −ψα∗ (σ0 )αα̇ (10.23)
ψ̄α̇ = −(σ̄0 )α̇α ψ∗α (10.24)
ψα = −(σ0 )αα̇ ψ̄α̇∗ . (10.25)

One usually uses shorthand notation which consists of leaving out the σ0 and σ̄0
matrices in the latter expressions, e.g. ψα = ψ̄∗α̇ . The dagger symbol can thus be
defined in a consistent way by

ψ̄α̇† := ψα and ψβ† := ψ̄β̇ . (10.26)

We remind the reader that if either the dagger or complex conjugation is applied
to a product of spinors, the order of the spinors has to be reversed.

Postulate We postulate that the components of spinors are Grassmann vari-

ables. We demand that:

{ψα , ψβ } = {ψα , ψβ } = {ψα , ψβ } = 0 (10.27)

β̇ α̇ β̇
{χ̄α̇ , χ̄ } = {χ̄α̇ , χ̄β̇ } = {χ̄ , χ̄ } = 0 (10.28)

and that all mixed anticommutators vanish too, e.g

{ψα , χ̄β̇ } = 0. (10.29)

We finally define
i
σmn = (σm σ̄n − σn σ̄m ) (10.30)
4
i
σ̄mn = (σ̄m σn − σ̄n σm ). (10.31)
4

Daniel Farquet EPFL - Physics

 43

10.2 Spinor identities

The following properties can be shown to hold
ψχ = χψ (10.32)
ψ̄χ̄ = χ̄ψ̄ (10.33)
1
θα θβ = − εαβ θθ (10.34)
2
1
θα θβ = εαβ θθ (10.35)
2
1 α̇β̇
θ̄α̇ θ̄β̇ = ε θ̄θ̄ (10.36)
2
1
θ̄α̇ θ̄β̇ = − εα̇β̇ θ̄θ̄ (10.37)
2
including the Fierz reordering formulae:
1 1
(θφ)(θψ) = − (φψ)(θθ) = − (θθ)(φψ) (10.38)
2 2
1
(φψ)χ̄α̇ = − (φσm χ̄)(ψσm )α̇ (10.39)
2
We also have
Tr[σm σ̄n ] = −2ηmn (10.40)
σm σ̄n + σn σ̄m = −2ηmn 12×2 (10.41)
σ̄m σn + σ̄n σm = −2ηmn 12×2 (10.42)
β̇β β β̇
σm
αα̇ σ̄m = −2δα δα̇ (10.43)
φσm χ̄ = −χ̄σ̄m φ (10.44)
(φσm χ̄)† = (χσm φ̄) (10.45)
(θφ)† = θ̄φ̄ (10.46)
α̇ 1
σm
αα̇ θ̄ (θσ θ̄) =
n
− ηmn θα (θ̄θ̄) (10.47)
2
1 mn
(θσm θ̄)(θσn θ̄) = − η (θθ)(θ̄θ̄) (10.48)
2
1 m n
(σm θ̄)α (θσn θ̄) = (σ σ̄ θ)α θ̄θ̄ (10.49)
2
(σmn )† = σ̄mn (10.50)
φσmn χ = −χσmn φ (10.51)

10.3 Independence of θ and θ̄

As stated above, we want to consider θ and θ̄ as independent variables. We know
that this is actually not the case since there is a map, composed of ε and σ0 , that
relates θ and θ̄. Therefore, how can we address the problem?

Daniel Farquet EPFL - Physics

 44

The variables θ and θ̄ are complex Grassmann variables in our case. In order
to understand how this problem can be formally solved we can restrict ourselves
to usual complex variables. Let f be a complex function of two variables, i.e.
f (u, v) with u, v ∈ C. If we suppose that f is a derivable function in the complex
sense, the two following quantities are well-defined:
∂ f (u, v) ∂ f (u, v)
, .
∂u ∂v
Suppose one wants to work in a ‘section’ of C2 where v = u∗ . The last equation
becomes
∂ f (u, u∗ ) ∂ f (u, u∗ )
, .
∂u ∂v
and is perfectly mathematically well-defined as well. Under the assumption that
we always work in this section of C2 , we can use the formal notation
∂ f (u, u∗ ) ∂ f (u, u∗ )
, .
∂u ∂u∗
Hence it is clear that one can think of u and u∗ as the independent variables of the
problem. To summarise we can consider u and u∗ as independent variables and
take derivatives with respect to them only if the function f is a derivable complex
function of its two arguments.
Identifying z with θ and z∗ with θ̄, it is now clear what we mean by the in-
dependence of θ with respect to θ̄ (again, this is just a restriction to a section of
a bigger space where the variables are truly independent, this will be used when
we introduce the derivative in θ and θ̄). Being sure of this kind of reasoning, it
will be used from now on.

10.4 Grassmann derivative

10.4.1 General theory
We consider a set of discrete anticommuting numbers ai , i = 1, ..., N. We can
define a formal differential calculus with
∂ai
:= δi j .
∂a j
In order to take into account the anticommuting nature of the a’s, the product rule
has to be redefined:
∂
(ai . . . ain ) = δmi1 ai2 . . . air − δmi2 ai1 ai3 . . . air + . . . + (−1)r−1 δmir ai1 . . . air−1 .
∂am 1
Proposition 10.1. The next equation can be shown to hold
∂
( )
, ak = δmk .
∂am

Daniel Farquet EPFL - Physics

 45

Proof. Let f be a function of any a, i.e. f (a) := f (a1 , ..., an ). Then

∂ ∂ ∂
( )
, ak f (a) = (ak f (a)) + ak f (a)
∂am ∂am ∂am
∂ak ∂ f (a) ∂ f (a)
= f (a) − ak + ak
∂am ∂am ∂am
= δmk f (a), ∀ f (a).

Where we used the product rule applied to ak f (a) in the second step.
At this stage, we do not need any more information in order to construct a
differential calculus on Weyl spinors.

10.4.2 The Weyl Spinor Derivative

Throughout the document we use the notation
∂ ∂
∂α := ∂α := (10.52)
∂θα ∂θα
∂ ∂
∂¯ α̇ := α̇ ∂¯ α̇ := α̇ (10.53)
∂θ̄ ∂θ̄
As a consequence of the general theory of Grassmann differentation and of the
independence of θ with respect to θ̄ we impose that

{∂α , ∂β } = 0, {∂¯ α̇ , ∂¯ β̇ } = 0

and
{∂¯ α̇ , θβ } = 0, {∂α , θ̄β̇ } = 0.
The last equation is equivalent to

∂¯ α̇ θβ = 0, ∂α θ̄β̇ = 0,

which is nothing but the independence of θ with respect to θ̄. The defining prop-
erties of the derivatives are
β
∂α θβ = δα , ∂α θβ = δαβ
β̇
∂¯ α̇ θ̄β̇ = δα̇ , ∂¯ α̇ θ̄β̇ = δα̇β̇ .

and we add that the derivatives must be linear with respect to ordinary commuting
numbers. Quantities like ∂α θβ can be calculated in this way:

∂α θβ = ∂α (εβγ θγ ) = εβγ ∂α θγ = εβα .

Two important results must be kept in mind.

Daniel Farquet EPFL - Physics

 46

Proposition 10.2. Indices on the derivative can be raised and lowered with ε up
to a minus sign. More explicitly

εαβ ∂β = −∂α , ∂α = −εαβ ∂β

∂¯ α̇ = −εα̇β̇ ∂¯ ,
β̇ ∂¯ α̇ = −ε ∂¯ β̇
α̇β̇

Under complex conjuguation a minus sign has to be taken into account, i.e.

(∂α )∗ = −∂¯ α̇ .

Proof. Concerning the first set of properties, it is sufficient to show the very first
one because the second one is obtained by applying εαβ and the two others can
be obtained by complex conjugation as soon as the second property is proved.
We have
εαβ ∂β θγ = εαγ ,
and
−∂α θγ = −∂α εγδ θδ = −εγδ δαδ = −εγα = εαγ = εαβ ∂β θγ .
Since this last relation is true for a single θ it must be true for any function of
theta, as can be seen from a Taylor expansion in θ.
Regarding complex conjugation, we start by calculating
β
∂α θβ θ̄γ̇ = δα θ̄γ̇ ,

and then remembering that the order of the θ’s has to be reversed under complex
conjugation and that lower indices become upper indices and vice-versa we have

δα̇β̇ θγ = (∂α θβ θ̄γ̇ )∗ = (∂α )∗ θγ θ̄β̇ = (∂α θ̄γ̇ )∗ θ̄β̇ − θγ (∂α )∗ θ̄β̇ = −θγ (∂α )∗ θ̄β̇ .

Hence we must have δα̇β̇ = −(∂α )∗ θ̄β̇ . On the other hand ∂¯ α̇ θ̄β̇ = δα̇β̇ , and by identi-
fying we obtain −(∂α )∗ = ∂¯ α̇ which is the relation we were looking for.

10.5 Grassmann Integration

Defining an integral over anticommuting variables will allow us to write down in
a very beautiful manner manifestly susy Lagrangians. As we already know, the
highest component in θ and θ̄ of a superfield transforms into a total derivative un-
der susy transformations. Therefore taking the highest component of a superfield
automatically gives an invariant action. It is then natural to define the integral
such that it is proportional to the highest component of the field, i.e.
Z
dθ1 dθ2 dθ̄1 dθ̄2 f (x, θ, θ̄) ∝ f (x, θ, θ̄)|θθθ̄θ̄ ,

Daniel Farquet EPFL - Physics

 47

where f (x, θ, θ̄)|θθθ̄θ̄ means the θθθ̄θ̄ component of the superfield f (x, θ, θ̄).
If we only consider chiral superfields, we can define an integral over the θ (or
θ̄) variable alone such that the θθ component (or θ̄θ̄) component is selected since
it is the one that goes into a total derivative under susy transformations. This
means that we want to define the integral such that
Z
d2 θφ(x, θ, θ̄) := φ(x, θ, θ̄)|θθ , (10.54)

where of course d2 θ means the integral over θ1 and θ2 . At this stage, we note
R

that the highest component of a chiral superfield can be obtained with a covariant
derivative. Indeed, if φ(x, θ, θ̄) is given by (5.4)

∂α ∂α φ(x, θ, θ̄) = −4F(x).

Considering (4.9), it is obvious that we have

1 1
− ∂α ∂α φ(x, θ, θ̄) = − D2 φ(x, θ, θ̄)|0
4 4
where the notation |0 means that we evaluate the expression at θ = θ̄ = 0. Finally,
we can write down the following important expression
Z
1
d2 θφ(x, θ, θ̄) = φ(x, θ, θ̄)|θθ = − D2 φ(x, θ, θ̄)|0 . (10.55)
4
The replacement of the integral, i.e. taking the θθ component, by a covariant
derivative is called the projection technique since it projects out every compo-
nent except for θθ. Analogously, one can check (by taking the † of the latter
expression, or by going through the same steps that we have just taken) that for
the antichiral field φ† we have
Z
1
d2 θ̄φ† (x, θ, θ̄) := φ† (x, θ, θ̄)|θ̄θ̄ = − D̄2 φ(x, θ, θ̄)|0 . (10.56)
4

Let us come back to the case where we wanted to select the θθθ̄θ̄ component
of a superfield. One can straightforwardly check that
1
D̄α̇ D2 D̄α̇ θθθ̄θ̄ = 1.
16
Taking the notation d4 θ to express the fact that the integration is performed over
R

θ1 , θ2 , θ̄1 and θ̄2 , we have

Z
1
d4 θ f (x, θ, θ̄) := f (x, θ, θ̄)|θθθ̄θ̄ = D̄α̇ D2 D̄α̇ f (x, θ, θ̄)|0 . (10.57)
16

Daniel Farquet EPFL - Physics

 48

1 α 2
The identity (4.23) also permits us to perform the projection with 16 D D̄ Dα , i.e.
Z
1
d4 θ f (x, θ, θ̄) = f (x, θ, θ̄)|θθθ̄θ̄ = Dα D̄2 Dα f (x, θ, θ̄)|0 . (10.58)
16
Let us note that, with the help of proposition 4.7, we could also write that
1 2 2 1
f (x, θ, θ̄)|θθθ̄θ̄ = D̄ D f (x, θ, θ̄)|0 + 4-div = D2 D̄2 f (x, θ, θ̄)|0 + 4-div. (10.59)
16 16
Thus if f (x, θ, θ̄)|θθθ̄θ̄ is integrated over x, the 4-div term is of no importance and
can be omitted. Such a case might appear when the action is considered since it
is an integral over xm of a Lagrangian.
The use of covariant derivatives in order to project out components can be
very useful in some situations as can be seen in section 5. The various identities
involving covariant derivatives that we proved in proposition 4.7 are of great help
if one uses projection technique to calculate integrals.
The manner in which we chose to present integration over Grassmann vari-
ables is probably not the most common one. In books such as [1] or [3] they
prefer to introduce the so-called Berizinian integration which is more or less an
integral in the common sense of the word. As we did not need to know the de-
tails of this theory but only a way of projecting out components, it was sufficient
to restrict ourselves to the notion of projection. As one may notice throughout
this paper, we could always replace Grassmann integrals by |θθ , |θ̄θ̄ or |θθθ̄θ̄ . We
nonetheless chose to introduce it because it is used in some books and articles.

10.6 The Fermionic Mass Square Matrix

We want to show that the mass square matrix is not given by m2 for two com-
ponents spinors but rather by m† m. We slightly change our convention here and
choose ηmn = diag(1, −1, −1, −1). The Dirac Lagrangian in terms of a four com-
ponents spinor ψ is
LD = ψ(i∂/ − M)ψ
0 σm
!
where ∂/ = γ ∂m with γ = m
m m and M is a 4 × 4 matrix. The associated
σ̄ 0
equation of motion is
(i∂/ − M)ψ = 0.
If we multiply it by (−i∂/ − M) and we use ∂/ ∂/ = , we find
( + M 2 )ψ = 0,
hence M 2 is the mass! square matrix. What does it turn into in terms of Weyl
ϕ
spinors? Let ψ = α̇α with ϕ and φ two two-component Weyl spinors. We know
φ̄
that the Lagrangian has to be written as
β
LD = iϕ̄σ̄m ∂m ϕ + iφσm ∂m φ̄ − φα mα ϕβ − ϕ̄α̇ m∗α̇β̇ φ̄β̇ .

Daniel Farquet EPFL - Physics

 49

β β
If we define the 2 × 2 mass matrix m by mα and take the notation −φα mα ϕβ −
ϕ̄α̇ m∗α̇β̇ φ̄β̇ = −φmϕ − ϕ̄m† φ̄, the Lagrangian becomes

 0 m† ! ϕ!
LD = ψ̄i∂/ ψ − ϕ̄ φ

,
m 0 φ̄
that is to say
0 m†
!
M= .
m 0
The mass square matrix is then given by

m† m
!
0
M = 2
.
0 mm†

For two-component spinors it is clear that the mass square matrix is m† m and not
m2 . This is good news because m† m is always diagonalisable with real eigenval-
ues, the masses.

10.7 U(N) and S U(N) in the Fundamental Representation

In this section we want to prove two identities that are used in the context of susy
breaking. They involve the generators of the U(N) and S U(N) group. We start
by recalling the definitions of those two groups:

U(N) := {g ∈ GL(N, C)|g† g = 1} , S U(N) := {g ∈ U(N)|detg = 1}.

They are related by

U(N) = U(1) × S U(N)
and the dimension of U(N) is N 2 whilst the dimension of S U(N) is N 2 − 1. These
two groups are generated by a Lie algebra. The Lie algebra of S U(N) is

su(N) = {M ∈ M(N, C)|M + M † = 0, Tr M = 0}.

The case N = 1 is trivial because the generator is the identity matrix, therefore we
suppose that N > 1. We denote by {T a }a=1,...,N 2 −1 the set of generators of S U(N),
i.e. a basis of su(N). We extend our notation to A = 0, ..., N 2 − 1 by defining
T 0 := √1 1. This extension is very convenient because it allows us to choose the
N
set {T A }A=0,...,N 2 −1 as the set of generators of U(N).
Let αA with A = 0, ..., N 2 − 1 be real numbers. Then the matrix

M := α0 1 + αa T a

is in U(N) if α0 , 0 and in S U(N) if α0 = 0. Using Tr T a T b = kδab it follows that

Tr M = Mll = α0 N and Tr(MT a ) = Mlk T kla = kαa ,

Daniel Farquet EPFL - Physics

 50

yielding that Mi j can be written as

Mll 1
Mi j = δi j + Mlk T kla T iaj
N k
or
Mlk 1
Mlk δil δ jk = δkl δi j + Mlk T kla T iaj .
N k
If the parameter αa , 0 then Mi j , 0 for all i, j. This comes from the fact that
the generators of S U(N) in the fundamental representation can be constructed by
miming the structure of the Pauli matrices (fundamental of S U(2)) which clearly
give non-zero matrix elements for M. We end up with
!
1
T i j T kl = k δil δ jk − δi j δkl
a a
(10.60)
N
which is the desired relation for S U(N). Regarding U(N) it is sufficient to notice
that N1 δi j δkl = T i0j T kl0 to find

T iAj T klA = kδil δ jk . (10.61)

10.8 Kähler Geometry

In this section we give a very short, but sufficient, introduction to Kähler mani-
folds. A Kähler manifold is a complex analytic manifold that can be parametrised
by complex coordinates Ai and Āī with i = 1, ..., n. Under an analytic coordinate
transformation given by

Ai = Ai (A0 ) and Āī = Āī (Ā0 )

the differentials transform in the usual way, for example

∂A0i j
dA0i = dA .
∂A j
Tensors also transform according to the usual rules, e.g.
∂A j
Vi0 (A0 , Ā0 ) = V j (A, Ā)
∂A0i
∂Ā j̄
Vī (A , Ā ) =
0 0 0
V j̄ (A, Ā)
∂Ā0ī
where we introduced the notation ī for some tensors. This notation will be jus-
tified shortly. The metric on a Kähler manifold is hermitian and given by the
second derivative of an analytic function (the Kähler potential)
∂ ∂
gi j̄ := K(A, Ā)
∂Ai ∂Ā j̄

Daniel Farquet EPFL - Physics

 51

which is invariant under analytic shifts

K(A, Ā) → K(A, Ā) + F(A) + F̄(Ā).

The metric allows us to raise and lower indices by

Vi = gi j̄ V j̄ V j̄ = gi j̄ V i
V i = gi j̄ V j̄ V j̄ = gi j̄ Vi

and of course consistency requires that

gik̄ g jk̄ = δij and gkī gk j̄ = δīj̄ .

We now understand why some tensor indices are written with a bar. On such
manifolds we impose that the analytic structure be respected by the covariant
derivatives yielding that Γk̄i j = Γk̄ī j = 0 where Γ is the Christoffel symbol. It is
also always possible to impose that Γkīj vanish. Thus, the covariant derivatives of
tensors are defined to be
∂
∇i V j := V j − Γkij Vk
∂A i
∂
∇ī V j := Vj
∂Āī
In order to assure that the covariant derivatives are indeed covariant, the transfor-
mations law of the connection is defined by

∂Al ∂Am ∂A0k n ∂2 An ∂A0k

Γ0k
ij = Γ + .
∂A0i ∂A0 j ∂An lm ∂A0i ∂A0 j ∂An
The constraint of compatibility between the connection and the metric is also
imposed:
∇k gi j̄ = 0 ∇k̄ gi j̄ = 0.
The first constraint can be easily solved for the Christoffel symbols and gives

¯ ∂
Γkij = gkl g ¯.
∂Ai jl
From this expression it is obvious that Γkij = Γkji . Before defining the Riemann
tensor in question let us make a comment on the covariant derivative. The for-
malism developed so far is similar to the one of general relativity. For example if
W(A) is a scalar (i.e. a function) its covariant derivative is defined to be equal to
the normal derivative:
∂W
∇i W = i = Wi .
∂A

Daniel Farquet EPFL - Physics

 52

It follows straightforwardly that Wi is a tensor. Hence, this time, its covariant

derivative is given by
∂2 ∂
∇i W j = W − Γkij k W.
∂A ∂A
i j ∂A
By construction ∇i W j is a rank two tensor and its covariant derivative is defined
as usually, and so on and so forth.
Let us move on to the curvature. The usual Riemann tensor vanishes : [∇i , ∇ j ]Vk =
Ri jk Vl = 0. On the other hand, by extending our definition of R to
l

¯
[∇i , ∇ j̄ ]Vk := Rli j̄k Vl = Ri j̄kl¯V l

where we also defined Ri j̄kl¯ := gml¯Rm i j̄k

, we end up with a non-vanishing Riemann
tensor. Indeed it is not difficult to prove that
∂ m ∂ ∂ ∂ ∂
! !
Ri j̄kl¯ = gml¯ Γik = i g ¯− gmn̄
gml¯ gkn̄ .
∂Ā j̄ ∂A ∂Ā j̄ kl ∂Ā j̄ ∂Ai
One can prove that the Riemann tensor has the following symmetries:

Ri j̄kl¯ = −Ri j̄lk¯ = −R j̄ikl¯ = R j̄ilk¯ .

Finally we define the Ricci tensor by

¯
Ri j̄ := −gkl Rkli¯ j̄ .

There is still a proposition that is needed in order for the discussion of our
required properties to be complete.
Proposition 10.3. Suppose the fields ψi are tensors in the tangent space of the
Kähler manifold. Let Ai and Āī be the coordinates of the Kähler manifold. Then
the object defined by Dm ψi = ∂m ψi + Γijk ∂m A j ψk is:
1. A space-time covariant derivative
2. A contravariant tensor in the Kähler manifold
Proof. The proof of the first statement is obvious because under space-time trans-
formations ∂m is covariant. The second statement is also not difficult to prove
but it requires a bit of work. We consider the following change of coordinates:
A0i = A0i (A). If we define M i j := ∂A
0i
∂A j
, then

ψ0i = M i j ψ j ,

because the ψi are contravariant. Thus

(∂m ψi )0 = ∂m ψ0i = M i j ∂m ψ j + (∂m M i j )ψ j .

Daniel Farquet EPFL - Physics

 53

The transformation of ∂m A j is ∂m A0 j = ∂A ∂A 0j k j
∂Ak ∂xm
= M k ∂m Ak , i.e. it is a true tensor.
Since ψk is also a tensor, the only non-tensorial part that comes from Γijk ∂m A j ψk
is the one which comes from Γijk , that is to say

∂2 An ∂A0i
(Dm ψi )0 = M i j Dm ψ j + (∂m M i j )ψ j + (∂m A0 j )M kl ψl .
∂A0 j ∂A0k ∂An
∂A0i
We now introduce N in = (M −1 )in = ∂An that allows us to write

∂ n
!
(Dm ψ )
i 0
= M j Dm ψ + (∂m M j )ψ +
i j i
N M in (∂m A0 j )M kl ψl
j
∂A0 j k
∂
!
= M j Dm ψ + (∂m M j )ψ − N k
i j i j n
M (∂m A0 j ) M in ψl
k
∂A0 j l
| {z }
∂m M kl

= M i j Dm ψ j + (∂m M i j )ψ j − M in N nk ∂m M kl ψl
| {z }
δik

= M i j Dm ψ j .

j j j
Where we used the fact that (∂N i j )M k = −N i j (∂M k ) which comes from ∂(N i j M k ) =
∂δik = 0.

Daniel Farquet EPFL - Physics

 54

References
[1] Müller-Kirsten, Harald J. and Wiedemann Franz. Introduction to super-
symmetry, World scientific lecture notes in physics, Vol. 80, second edition,
2009.

[2] Wess, Julius and Bagger, Jonathan. Supersymmetry and Supergravity, sec-
ond edition, Princeton University Press, 2002.

[3] Sohnius, Martin F. Introducing Supersymmetry, Elsvier Science Publishers

B.V., Physics reports 128, Nos. 2 & 3 (1985) 39-204.

[4] Derendinger, Jean-Pierre. Lecture Notes on Globally Supersymmetric The-

ories in Four and Two dimensions, Physics Institute, University of Neucha-
tel, Switzerland.

[5] Maggiore, Michele. A Modern Introduction to Quantum Field Theory, Ox-

ford Master Series, 2009.

[6] Peskin, Michael and Schroeder, Daniel. An introduction to quantum field

theory.Westview Press, 1995.

[7] Witten, Edward and Bagger, Jonathan. The Gauge Invariant Supersymmet-
ric Nonlinear Sigma Model. Physics Letters, Volume 118B, 2 December
1982.

[8] Jacot, Jean-Claude and Scrucca, Claudio. Metastable supersymmetry

breaking in N=2 non-linear sigma-models.arXiv:1005.2523v2 [hep-th]

Daniel Farquet EPFL - Physics