Lecture Notes on

Supersymmetry
Kevin Zhou
[email protected]

These notes cover supersymmetry, closely following the Part III and MMathPhys Supersymmetry
courses as lectured in 2017/2018 and 2018/2019, respectively. The primary sources were:

• Fernando Quevedo’s Supersymmetry lecture notes. A short, clear introduction to supersym-

metry covering the topics required to formulate the MSSM. Also see Ben Allanach’s revised
version, which places slightly more emphasis on MSSM phenomenology.

• Cyril Closset’s Supersymmetry lecture notes. A comprehensive set of supersymmetry lecture

notes with more emphasis on theoretical applications. Contains coverage of higher SUSY, and
spinors in various dimensions, the dynamics of 4D N = 1 gauge theories, and even a brief
overview of supergravity.

• Aitchison, Supersymmetry in Particle Physics. A friendly introductory book that covers the
basics with a minimum of formalism; for instance, the Wess–Zumino model is introduced
without using superfields. Also gives an extensive treatment of subtleties in two-component
spinor notation. The last half covers the phenomenology of the MSSM.

• Wess and Bagger, Supersymmetry and Supergravity. An incredibly terse book that serves as
a useful reference. Most modern sources follow the conventions set here. Many pages consist
entirely of equations, with no words.

We will use the conventions of Quevedo’s lecture notes. As such, the metric is mostly negative,
and a few other signs are flipped with respect to Wess and Bagger’s conventions. The most recent
version is here; please report any errors found to [email protected].
2 Contents

Contents
1 Introduction 3
1.1 Motivation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3
1.2 The Poincare Group . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6
1.3 Spinors in Four Dimensions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9
1.4 Supersymmetric Quantum Mechanics . . . . . . . . . . . . . . . . . . . . . . . . . . . 13
1.5 Spinors in Various Dimensions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 15

2 SUSY Algebra and Representations 18

2.1 The SUSY Algebra . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 18
2.2 N = 1 SUSY Representations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 20
2.3 Extended SUSY . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 24
2.4 SUSY in Various Dimensions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 29

3 Superspace and Superfields 31

3.1 Superspace . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 31
3.2 The Scalar Superfield . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 33
3.3 Chiral and Vector Superfields . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 36

4 Supersymmetric Lagrangians 41
4.1 N = 1 Supersymmetry . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 41
4.2 The Wess–Zumino Model . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 46
4.3 Non-Renormalization Theorems . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 47
4.4 Extended Supersymmetry . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 49

5 The MSSM 51
5.1 SUSY Breaking . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 51
5.2 Particles and Interactions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 55
3 1. Introduction

1 Introduction
1.1 Motivation
We begin with a review of the Standard Model and its problems.

• A spacetime symmetry is one that acts explicitly on the spacetime coordinates,

xµ → x′µ (xν )

and include Poincare transformations in special relativity, and more generally, general coordinate
transformations in general relativity.

• An internal symmetry corresponds to transformations of the different fields in a field theory,

Φa (x) → M ab Φb (x).

If M is constant, the symmetry is global, and if M = M (x) it is local.

• Symmetries constrain the interactions between fields. For example, most quantum field theories
of vector bosons are non-renormalizable, but gauge theories are renormalizable.

• Symmetries may also be spontaneously broken. This is important phenomenologically because

it naturally introduces an energy scale in the system, determined by the VEV, and allows for
more complex fundamental symmetries than we observe at low energies.

• The SM has Poincare symmetry and a gauge SU (3)C ×SU (2)L ×U (1)Y symmetry, spontaneously
broken to SU (3)C × U (1)A .

• The hierarchy problem is the result

mh
∼ 10−17
Mp
which is technically unnatural; there is nothing protecting mh from receiving O(Mp ) quantum
corrections. Similarly, the cosmological constant problem is

(Λ/Mp )4 ∼ 10−120 .

A related issue is how the 20 free parameters of the SM are determined. Finally, the SM does
not account for dark matter.

Next, we turn to historical motivations for supersymmetry.

• In the 1960’s, much progress was made by classifying hadrons into multiplets, and there were
attempts to enlarge the symmetry groups by including spacetime symmetries.

• The Coleman–Mandula theorem (1967) states that spacetime and internal symmetries cannot
be combined nontrivially in a relativistic theory with nontrivial scattering, a mass gap, and
finitely many particles. More precisely, the symmetry group of the S-matrix must be a direct
product of the Poincare group and an internal symmetry group. (Conformal field theories
(CFTs) evade this theorem because they don’t have a mass gap, allowing the larger spacetime
symmetry group SO(2, d) in d spacetime dimensions.)
4 1. Introduction

• In 1971, Gelfand and Likhtman extended the Poincare algebra by adding generators that trans-
formed like spinors and satisfied anticommutation relations, thus inventing SUSY; this evaded
the Coleman–Mandula theorem because the symmetry was described by a Lie superalgebra
rather than a Lie algebra. Note that the spin-statistics theorem ensures that in all dimensions,
SUSY generators must be spinors.

• Simultaneously, Ramond, Neveu, and Schwarz found that string theory extended with fermions
was a two-dimensional supersymmetric theory on the worldsheet, inventing superstring theory.
The string worldsheet also has conformal symmetry, making it a superconformal field theory
(SCFT).

• In the 1970’s, neutrinos were thought to be massless. In 1973, Volkov and Akulov proposed
that neutrinos were Goldstone fermions, called Goldstinos, due to the spontaneous breaking of
SUSY.

• In 1974, Wess and Zumino wrote down the first example of an interacting four-dimensional quan-
tum field theory with linearly realized SUSY. Simultaneously, Salam and Strathdee invented
the tools of superfields and superspace, coining the term ‘supersymmetry’.

• In 1975, Haag, Lopuszanski, and Sohnius generalized the Coleman–Mandula theorem to essen-
tially state that the most general symmetry possible was a direct product of the super Poincare
group and internal symmetries.

• Making Poincare symmetry local yields general coordinate transformations and hence general
relativity. In 1976, Friedman, van Niewenhuizen, and Ferrara, and Deser and Zumino made
SUSY local, yielding supergravity. The superpartner of the graviton was the spin 3/2 gravitino.

• From 1977 to the 1980’s, SUSY phenomenology was developed. It was demonstrated that SUSY
could solve the hierarchy problem in a natural way, though this is less relevant today.

• Simultaneously, in 1977 Gliozzi, Scherk, and Olive demonstrated how to remove the tachyon
from the Ramond-Neveu-Schwarz model, and conjectured the resulting theory had spacetime
supersymmetry. From 1981 to 1984, Green and Schwarz proved this conjecture, discovering
an anomaly cancellation mechanism for superstring theory in d = 10 and starting the first
superstring revolution.

• In 1991, LEP performed precision tests of the SM. It was found that gauge coupling unification
did not occur for the SM, presenting problems for GUTs, but would happen for the MSSM as
long as superpartners had masses in the range 100 GeV to 10 TeV.

• In 1994, Seiberg and Witten investigated N = 2 superstring theory nonperturbatively, discov-

ering M-theory and starting the second superstring revolution.

• In 1996, Strominger and Vafa counted the microstates of a black hole in superstring theory to
confirm the Bekenstein-Hawking formula S = A/4.

• In 1998, the AdS/CFT duality was proposed by Maldacena, showing that certain CFTs in d
dimensions are dual to quantum gravity theories in AdS space in d + 1 dimensions. The best
studied instances of the AdS/CFT duality involve SCFTs which are dual to superstring theories
in AdS, making supersymmetry a useful tool for studying quantum gravity.
5 1. Introduction

Next, we discuss the hierarchy problem in more detail.

• Consider a Higgs potential of the form

λ † 2
V = −µ2 ϕ† ϕ + (ϕ ϕ) .
4
Then the Higgs vev, which sets the weak scale, is
√ √
⟨ϕ⟩ = 2µ/ λ.

For perturbation theory to apply, λ should not be too large, so µ ≲ ⟨ϕ⟩.

• The issue is not that µ is small, but that quantum effects give large corrections to µ. This
doesn’t happen for gauge boson masses, which are held at zero by gauge symmetry, or for spinor
masses, because chiral symmetry is restored when the mass vanishes. Then δm ∼ m log Λ,
which is reasonably small even when Λ is the Planck scale.

• On the other hand, the one-loop contribution due to the Higgs is

Z Λ
2 d4 k 2
δµ ∼ λ 2 ∼ λΛ
k 2 − MH

so if Λ is the Planck scale, µ2 must be fine-tuned to get an acceptable observed value of µ2phys ,
the coefficient of ϕ† ϕ in the 1PI effective action. Thus to avoid fine tuning there must be new
physics around the TeV scale.

• One solution is to postulate that spontaneous symmetry breaking occurs ‘dynamically’. In a

technicolor theory, the Higgs is a composite of fermions, analogous to Cooper pairs in BCS
theory, so the theory above is only an effective field theory valid up to the TeV scale. However,
this theory has issues with giving masses to fermions.

• Another, more radical solution is to set the Planck scale to the TeV scale; this is consistent if
there are large extra dimensions. We’ll put these ideas aside and focus on SUSY.

• The one-loop contribution to δµ2 above can also be canceled by fermion contributions. Consider
a fermion with Yukawa coupling gf to the Higgs. Then
Z Λ
2 2 1
δµ ∼ −gf d4 k tr ∼ −4gf2 Λ2
(k/ − mf )2
where the minus sign comes from the fermion loop. Then the quadratic divergence cancels if,
for every boson, there is a fermion whose coupling to the Higgs is related; this is guaranteed by
SUSY.

• Even given this cancellation, there is still a logarithmic divergence,

δµ2 ∼ λ(MH
2
− m2f ) log Λ

where we dropped all numerical factors, which depends quadratically on the particle masses.
This is completely generic; we would even get a contribution of m2 from a particle of mass m
that didn’t couple directly to the Higgs, by a multi-loop diagram. Hence the parameter µ2 is
quadratically sensitive to any scale associated with new physics.
6 1. Introduction

• Thus to avoid fine tuning, new physics should arise at the TeV scale. Moreover, if that physics
is SUSY, the superpartner masses should generally be around the TeV scale. This is especially
important in SUSY GUTs, where there are many very heavy particles.

• In the MSSM, applying naturalness to the coefficients shows that the Higgs should be no heavier
than 140 GeV. By contrast, in the SM there is no constraint, unless we count perturbative
unitarity, which bounds the Higgs mass by a few hundred GeV.

Note. A cartoon explanation of the Coleman–Mandula theorem. Essentially, the theorem states
that conserved charges from internal symmetries can’t have Lorentz indices; the only such charges
are momentum Pµ and angular momentum Mµν which arise from spacetime symmetries. Suppose
we had another such charge Qµν . By Lorentz invariance,

Qµν |p⟩ = (αpµ pν + βgµν )|p⟩.

Now consider a two-particle state. We suppose that Qµν values are additive, conserved, and act on
only one particle at a time. Then

Qµν |p(1) , p(2) ⟩ = (α(p(1) (1) (2) (2) (1) (2)

µ pν + pµ pν ) + 2βgµν )|p , p ⟩.

Therefore, in a 1 + 2 → 3 + 4 scattering process we have

p(1) + p(2) = p(3) + p(4) , p(1) (1) (2) (2) (3) (3) (4) (4)
µ pν + pµ pν = pµ pν + pµ pν .

However, these conditions are so restrictive that there are no nontrivial solutions! We can only have
forward or backwards scattering.

Note. A preview of the SUSY algebra. We will have a spinorial generator Qa with a = 1, 2,
which relates bosons and fermions; then the above argument fails at the first step, since we cannot
superpose the two. The Qa satisfy anticommutation relations among themselves, and commute
with H. Then we have
[{Qa , Qb }, H] = 0.
Now, {Qa , Qb } should be a “spin one’ object’, so in a relativistic field theory it should be a four-vector.
The only conserved four-vector is Pµ , so

{Qa , Qb } ∼ Pµ .

Hence supersymmetry transformations inevitably relate internal and spacetime symmetries. They
function as a kind of
√ “square root” of translations, and hence take us from ordinary space to
superspace like how −1 takes us from the real line to the complex plane. Note that the spin 1/2
generator Qa is the only exception to the Coleman–Mandula theorem, i.e. the Haag–Lopuszanski–
Sohnius rules out spin 3/2 generators Qµa , and so on.

1.2 The Poincare Group

We begin by reviewing the Poincare algebra.

• The Poincare group corresponds to the basic symmetries of special relativity. It acts on spacetime
coordinates by
xµ → x′µ = Λµν xµ + aµ
7 1. Introduction

where the Lorentz transformations Λ satisfy

ΛT ηΛ = η, η = diag(1, −1, −1, −1).

We will focus on the proper orthochronous Lorentz group SO(1, 3)↑ , while the full Lorentz
group is O(3, 1) = {1, ΛP , ΛT , ΛP T } × SO(1, 3)↑ . Below we’ll just write SO(1, 3) for SO(1, 3)↑ .

• Infinitesimally, we have
Λµν = δνµ + ω µν , aµ = ϵµ
where ωµν = −ωνµ . If the Poincare group is represented by U (Λ, a) on a Hilbert space, then
infinitesimally we define
i
U (1 + ω, ϵ) = 1 − ωµν M µν + iϵµ P µ .
2
A useful explicit expression is

(M ρσ )µν = −i(η µσ δνρ − η ρµ δνσ ).

Note that we use the same notation for the abstract Poincare algebra elements and their
representations on a Hilbert space, since we will use the latter constantly. By the definition of
a representation, the commutator on the latter is the bracket on the former.

• We now find the Poincare algebra. Since translations commute in the Hilbert space, we have

[Pµ , Pν ] = 0.

• Since P µ is a vector, it transforms under the Lorentz group as

1
P σ → Λσρ P ρ = P σ + ωαρ (η σα P ρ − η σρ P α ).
2
On the other hand, we can compute the transformation of P µ explicitly in a representation of
the Poincare group on a Hilbert space, where the operator P µ transforms as
   
σ † σ i µν σ i µν i
P → U P U = 1 + ωµν M P 1 − ωµν M = P σ − ωµν (P σ M µν − M µν P σ )
2 2 2
from which we read off the commutation relation

[M µν , P σ ] = i(P µ η νσ − P ν η µσ )

• By similar reasoning, we have

[M µν , M ρσ ] = i(M µσ η νρ + M νρ η µσ − M µρ η νσ − M νσ η µρ ).

As above, this just states the tensorial transformation properties of M µν infinitesimally.

• We define the Hermitian and anti-Hermitian generators of rotations and boosts by

1
Ji = ϵijk Mjk , Ki = M0i
2
where ϵ123 = ϵ123 = 1. Note that the indices on M here are not lowered by the metric,
M0i = M 0i . The Lorentz algebra is

[Ki , Kj ] = −iϵijk Jk , [Ji , Kj ] = iϵijk Kk , [Ji , Jj ] = iϵijk Jk .

8 1. Introduction

• We now define the linear combinations

1 1
Ai = (Ji + iKi ), Bi = (Ji − iKi )
2 2
which satisfy the su(2) ⊕ su(2) commutation relations

[Ai , Aj ] = iϵijk Ak , [Bi , Bj ] = iϵijk Bk , [Ai , Bj ] = 0.

Hence we conclude so(3, 1) ∼ = su(2) ⊕ su(2) as complex Lie algebras. Note that J = A + B,
and under parity J → J and K → K, so A and B are interchanged. This leads to the usual
classification of representations of the Lorentz group.

• However, there is another route. There is a homomorphism SL(2, C) → SO(1, 3) as SL(2, C)

is the universal/double cover of SO(1, 3). For a four-vector X, we define
 
µ µ x0 + x3 x1 − ix2
X = xµ e = (x0 , x1 , x2 , x3 ), x̃ = xµ σ = , σ µ = (1, σ).
x1 + ix2 x0 − x3

However, if we use mostly positive signature, we instead must define σ µ = (−1, σ).

• Now, SO(1, 3) and SL(2, C) act on these spaces by

X → ΛX, x̃ → N x̃N † , Λ ∈ SO(1, 3), N ∈ SL(2, C)

so we can construct the homomorphism by mapping back and forth. Explicitly, it is

1
Λµν = tr σ µ N σν N † .
2

• This map is well-defined and surjective since the only constraint on the x̃ transformations is

det x̃ = x10 − x21 − x22 − x23 = constant

while the only constraint on the Lorentz transformations is

|X|2 = x10 − x21 − x22 − x23 = constant.

It is a double cover since both N = ±1 correspond to Λ = 1.

Note. The topology of SL(2, C). To see it, use the polar decomposition

N = eH U

where H is Hermitian and U is unitary. We may parametrize them as

   
a b + ic x + iy z + iw
H= , U=
b − ic −a −z + iw x − iy

where x2 + y 2 + z 2 + w2 = 1. The set of H is R3 while the set of U is S 3 , so SL(2, C) ∼

= R3 × S 3 ,
while the Lorentz group mods out S 3 by Z2 .
9 1. Introduction

1.3 Spinors in Four Dimensions

As shown above, we can use the representation theory of SL(2, C) to find the projective representa-
tions of SO(1, 3). This is especially useful for the fundamental spinor representations. Here we will
use the tensor methods explained in the notes on Group Theory.

• The fundamental representation transforms as

ψα → Nα β ψβ
and contains left-handed Weyl spinors. The conjugate representation

χα̇ → N ∗ α̇β̇ χβ̇

contains right-handed Weyl spinors, where N ∗ is the conjugate of N .
• These representations also have dual/contravariant representations
−1
ψ α → ψ β (N −1 )β α , χα̇ → χβ̇ (N ∗ )β̇ α̇ .

• We are using a redundant notation: the ψ and χ don’t matter, but dotted indices are associated
with bars. This is useful because we can then write expressions unambiguously without indices.
• For the matrices N , dotted indices always accompany a conjugate, so they’re redundant as we
always write the conjugate explicitly. We simply assign indices to N so that the indices match
up properly; note that the first index is always down.
• The invariant tensors in SL(2, C) are delta functions δβα and δβ̇α̇ and the Levi–Civita symbols

ϵαβ = ϵα̇β̇ = −ϵαβ = −ϵα̇β̇ , ϵ12 = 1

where the minus sign ensures ϵαβ ϵβγ = δγα . They are invariant because
ϵαβ → Nα ρ Nβ σ ϵρσ = ϵαβ det N = ϵαβ
with similar proofs for the others. Then the Levi–Civita can be used to invert matrices,
ϵσδ Nδ β ϵβα = (N −1 )ασ

• The Levi–Civitas can be used to raise or lower indices. This is a bit tricky because ϵαβ is not
symmetric; by convention we always contract the second index. We define

ψ α = ϵαβ ψβ , χα̇ = ϵα̇β̇ χβ̇ , ψα = ϵαβ ψ β , χα̇ = ϵα̇β̇ χβ̇ .

All these objects transform as their index placement would suggest.
• We can also compute a transformation for the Pauli matrices (σ µ )αα̇ which have mixed indices.
The equation x̃ → N x̃N † implies x̃ has one dotted and one undotted index, so
x̃ = (xµ σ µ )αα̇ → Nα β (xν σ ν )β γ̇ N ∗ α̇γ̇ = Λµν xν (σ µ )αα̇
which gives the transformation rule
(σ µ )αα̇ = Nα β N ∗ α̇γ̇ Λµν (σ ν )β γ̇
which is exactly what we would expect from the index structure.
10 1. Introduction

• It is also useful to define

(σ µ )α̇α ≡ (σ µ )αα̇ = ϵαβ ϵα̇β̇ (σ µ )β β̇ = (1, −σ)

which obeys a similar transformation law; note that if we didn’t ‘swap the indices’ then the
matrix σ 2 would have the wrong sign.

• There are some useful identities for σ and σ. They form a Clifford algebra, as

(σ µ σ ν + σ ν σ µ )βα = 2η µν δαβ .

We may think of σαµα̇ as a set of Clebsch–Gordan coefficients for the identity (1/2, 0) × (0, 1/2) =
(1/2, 1/2). The completeness of both bases is expressed by

(σ µ )αβ̇ (σ µ )γ̇δ = 2δαδ δβ̇γ̇ , tr σ µ σ ν = 2η µν .

Specifically, we can swap back and forth as

1
Vαα̇ = σαµα̇ Vµ ↔ V µ = (σ µ )α̇α Vαα̇ .
2
Here, two irreps can multiply to another irrep because the Lorentz group is not semi-simple.

Next, we construct the generators of SL(2, C) for the spinor representations.

• Just as the Dirac spinor is built from the Clifford algebra of gamma matrices, we have
i i
(σ µν )βα = (σ µ σ ν − σ ν σ µ )βα , (σ µν )β̇α̇ = (σ µ σ ν − σ ν σ µ )α̇β̇ .
4 4
Then the matrices σ µν , and the matrices σ µν , satisfy the Lorentz algebra,

[σ µν , σ λρ ] = i(η µρ σ νλ + η νλ σ µρ − η µλ σ νρ − η νρ σ µλ ).

They obey the identity

1
tr σ µν σ κτ = (η µκ η ντ − η µτ η νκ + iϵµνκτ ).
2

• The (σ µν )βα can also be used to project out the (1, 0) representation in the product

(1/2, 1/2) × (1/2, 1/2) = (1, 1) + (1, 0) + (0, 1) + (0, 0).

That is, Vµ Wν (σ µν )βα transforms in the (1, 0). On the other hand, we also know that (1, 0) is the
symmetric product of two (1/2, 0)’s, which implies (σ µν )βα is symmetric in α and β. Similarly,
(σ µν )β̇α̇ projects out (0, 1).

• One can show that the left-handed and right-handed spinors transform as
 β  
i µν α̇ β̇ i µν
ψα → exp − ωµν σ ψβ , χ → χ exp − ωµν σ .
2 α 2
In terms of the usual classification of Lorentz irreps we can show
1 i
ψα :(A, B) = (1/2, 0), Ji = σi , Ki = − σi ,
2 2
1 i
χα̇ :(A, B) = (0, 1/2), Ji = σi , Ki = σi .
2 2
11 1. Introduction

• We also have the self-duality and anti self-duality identities

1 µνρσ 1 µνρσ
σ µν = ϵ σρσ , σ µν = − ϵ σ ρσ .
2i 2i
This ensures the transformations above are specified by 3 complex parameters, not 6. Here
we define ϵ0123 = 1, ϵ0123 = −1 as is natural in general relativity, i.e. we use the opposite sign
convention to SL(2, C).

Next, we show how to multiply Weyl spinors.

• The contraction of Weyl spinors requires an ordering convention because the Levi–Civita is
antisymmetric. As motivated below, we define
α̇
χψ ≡ χα ψα = −χα ψ α , χψ ≡ χα̇ ψ = −χα̇ ψ α̇ .

That is, indices contract ↘ for undotted indices and ↗ for dotted indices.

• In particular, we have
ψψ = ψ α ψα = ϵαβ ψβ ψα = ψ2 ψ1 − ψ1 ψ2 .
This appears to vanish classically, but since spinors are inherently anticommuting we choose to
represent them as Grassmann numbers classically. Then
1
ψψ = 2ψ2 ψ1 , ψα ψβ = ϵαβ (ψψ).
2
This also implies that contraction is symmetric, χψ = ψχ and χψ = ψχ, and
1 1
(θχ)(θξ) = − (θθ)(χξ), (θχ)(θξ) = − (θθ)(χξ)
2 2

• One can conjugate a representation by just conjugating the vectors. That is, we define
α̇
ψ α̇ = ψα† , ψ = ψ α†

where the dagger simply stands for complex conjugation. Complex conjugation is defined to
reverse the order of Grassmann numbers, (θ1 θ2 )∗ = θ2∗ θ1∗ , which implies

(χψ)† = ψχ = χψ, (ψσ µ χ)† = χσ µ ψ

where we used ((σ µ )αβ̇ )∗ = ((σ µ )αβ̇ )T = (σ µ )β α̇ since the σ µ are Hermitian.

• Two-component spinor notation can be used to deal with tensor products of Lorentz represen-
tations. For example, we have the Fierz identity
1
ψα χα̇ = (ψσµ χ)σαµα̇ , (1/2, 0) × (0, 1/2) = (1/2, 1/2)
2
showing that a left-handed and right-handed spinor yield a vector ψσµ χ.
12 1. Introduction

• Defining (σ µν )αγ ϵγβ = (σ µν ϵT )αβ and using the identity

(σ µν )αβ (σµν )γ δ = ϵαγ ϵβδ + δαδ δγβ

we have the Fierz identity

1 1
ψα χβ = ϵαβ (ψχ) + (σ µν ϵT )αβ (ψσµν χ), (1/2, 0) × (1/2, 0) = (0, 0) + (1, 0)
2 2
where ψχ is a scalar and ψσµν χ is a self-dual tensor, which has the desired 3 degrees of freedom.
The same kind of decomposition works for two dotted spinors.

• Another useful set of Fierz identities is

1 1
(θψ)(χη) = − (θσ µ η)(χσ µ ψ), (θσ µ θ)(θσ ν θ) = η µν (θθ)(θθ).
2 2
To use these identities, it is useful to ‘reorder’ fields. We have

θσ µ χ = −χσ µ θ, θσ µ σ ν χ = χσ ν σ µ θ

which implies that

ψσ µν χ = −χσ µν ψ.
The pattern continues with alternating signs for more σ’s, i.e. with everything reversed in order
with σ and σ interchanged.

Next, we make contact with four-component Dirac spinors.

• A Dirac spinor Ψ is the direct sum of two Weyl spinors ψ and χ of opposite chirality,
 
ψα
Ψ= .
χα̇
Here the left-handed component is on top and the right-handed component is on the bottom.

• The analogue of the matrices σ µ are the Clifford matrices

0 σµ
 
µ
γ =
σµ 0
which also form a Clifford algebra. Then they similarly yield a representation of the Lorentz
group, with the generators  µν 
µν i µν σ 0
Σ = γ =
4 0 σ µν
which is naturally block-diagonal.

• We define the chiral matrix  

−1 0
γ 5 = iγ 0 γ 1 γ 2 γ 3 =
0 1
giving the projection operators
1 1
PL = (1 − γ 5 ), PR = (1 + γ 5 ).
2 2
We can also see this works because {γ 5 , γ µ } = 0, so [γ 5 , Σµν ] = 0 and Lorentz transformations
preserve chirality.
13 1. Introduction

• We define the Dirac conjugate Ψ and the charge conjugate ΨC by

! !
T χ α ϵαβ 0
Ψ = (χα , ψ α̇ ) = Ψ† γ 0 , ΨC = CΨ = α̇ , C= .
ψ 0 ϵα̇β̇

Then charge conjugation simply exchanges χ and ψ. Majorana spinors have ψ = χ and hence
are mapped to themselves under charge conjugation.

• Note that we have the gamma matrix identities

i
Σµν = ϵµνρσ γ 5 Σρσ , tr γ 5 γ µ γ ν γ ρ γ σ = −4iϵµνρσ .
2

1.4 Supersymmetric Quantum Mechanics

In this section we’ll give a preview of the following results for the case of a one-dimensional quantum
field theory, i.e. quantum mechanics.

• Formally, a superalgebra over C is a Z2 graded vector space

A = A0 ⊕ A1 ,

whose two components are called the bosonic and fermionic subalgebras respectively, with a
bilinear multiplication operator so that

a0 a′0 ∈ A0 , a0 a1 ∈ A1 , a1 a′1 ∈ A0

if a0 , a′0 are bosonic and a1 , a′1 are fermionic.

• A supersymmetry algebra over R3 is a superalgebra which contains the d-dimensional Poincare

symmetry algebra iso(1, d − 1) as a subalgebra of its bosonic subalgebra; below we will restrict
to the case where iso(1, d − 1) is precisely the bosonic subalgebra.

• In one dimension, the Poincare algebra has a single generator,

d
E = −i .
dt
We introduce N supersymmetry generators QI . In one dimension, these generators do not need
an additional spinor index. It turns out that

[QI , E] = 0

which indicates that all the states in an irrep of the supersymmetry algebra (a supermultiplet)
are degenerate.

• The anticommutators between the SUSY generators are generally

{QI , QJ } = 2Eδ IJ + Z IJ

where the central charge Z IJ is real and symmetric. Central charges are not elements of the
original Lie algebra, and are taken to commute with all elements of it. They arise naturally
when we allow for projective representations, as we encountered for the Galilean group in the
notes on Group Theory.
14 1. Introduction

One simple N = 1 supersymmetric model is a theory with D free 1D bosons X µ (t) and D free 1D
fermions ψ µ (t), all of vanishing mass.

• The Lagrangian is
1
L = Ẋµ Ẋ µ + iψµ ψ̇ µ
2
where the fermions ψ µ are Grassmann-valued. If the indices above are raised and lowered with
the Minkowski metric, then this system represents a relativistic spinning massless particle in
flat D-dimensional spacetime.
• The conjugate momenta are
ΠµX = Ẋ µ , Πµψ = iψ µ .
In canonical quantization, we hence have
[X µ , Ẋ ν ] = iη µν , {ψ µ , ψ ν } = η µν .
The latter result tells us the spin degrees of freedom match those of a Dirac spinor.
• The Lagrangian has a 1D N = 1 supersymmetry, acting on the fields as
δX = 2iϵψ, δψ = −ϵẊ
where ϵ is a Grassmann parameter. Under this transformation the Lagrangian changes by a
total derivative,
d  
δL = iϵ ψµ Ẋ µ .
dt
Applying Noether’s theorem, we find the conserved charge and supersymmetry generator
Q = ψµ Ẋ µ .

• The energy operator is

1
E = −P0 = Ẋ 2 .
2
The operators Q and E hence form a supersymmetry algebra, obeying the appropriate anti-
commutation relations
{Q, Q} = {ψµ , ψν }Ẋ µ Ẋ ν = Ẋ 2 = 2E.

• We can also confirm this relation at the field level, defining δ = ϵQ. Since we have
δ 2 X = −2iϵ2 Ẋ, δ 2 ψ = −2iϵ2 ψ̇
then we have {Q, Q} = 2E when acting on the fields.
• Upon quantization, the state space is a tensor product of spin and spatial degrees of freedom.
The operators ψ µ serve as spin raising and lowering operators, with the total dimensionality of
this space matching the degrees of freedom of a single Dirac spinor particle.
• The operators X µ and Ẋ µ act on L2 (RD−1 ). Their equations of motion have solutions
X µ (t) = X µ (0) + Ẋ µ (0)t, Ẋ µ (t) = Ẋ µ (0).
Algebraically, they behave like the usual operators in nonrelativistic quantum mechanics. That
is, while the fields X µ are formally massless, they represent the position of a nonrelativistic
particle with unit mass.
15 1. Introduction

• Like other symmetries, we may find the irreducible representations of the symmetry algebra;
the Hilbert space will then be built out of these “supermultiplets”.
√ Since [Q, H] = 0, we may
restrict to states of energy E. Defining the rescaling b = Q/ 2E, the N = 1 SUSY algebra is

{b, b} = 1, b = b† .

The irreducible representations are all one-dimensional, so this gives us little information.

• For N = 2n supersymmetry, we get nontrivial information. Define the complex supercharges

i
Qi = Qi + iQn+i , Q = Qi − iQn+i , i = 1, . . . , n.

The supersymmetry algebra then takes the form

j i j
{Qi , Q } = 4Eδ ij , {Qi , Qj } = 2iZ ij , {Q , Q } = −2iZ ij , Z ij = Z i,n+j

In the case of vanishing central charges, if we define

Qi Qi
ai = √ , a†i = √
2 E 2 E
then we have n independent fermionic QHOs, and hence a supermultiplet with 2n states.

We will see much of this again in more detail for d = 4, though there will be additional complications,
such as spinor indices.

1.5 Spinors in Various Dimensions

Though we will mostly focus on four dimensions, it is useful to set conventions for spinors in arbitrary
dimension. We’ve already covered spinor representations of so(n) in the notes on Group Theory,
but here will focus on spinor representations in Minkowski space, which differ in several respects.

• We focus on so(1, d − 1), where the signature is mostly positive. We take the generators
Mµν = −Mνν to satisfy the algebra

[Mµν , Mρσ ] = i(ηµσ Mνρ + ηνρ Mµσ − ηµρ Mνσ − ηνσ Mµρ ).

• A Clifford algebra is a set of d matrices satisfying

{γ µ , γ ν } = 2η µν .

The Dirac spinor representation of so(1, d − 1) is the one with representation matrices

i
Mµν = [γµ , γν ]
4
where the γµ have the minimum possible dimension while still representing the Clifford algebra
faithfully. Note that various factors of i may differ depending on the source.
16 1. Introduction

• To see the dimension of the Dirac spinor representation, let d = 2(k + 1) and define
1 1
γ 0± = (±γ 0 + γ 1 ), γ a± = (γ 2a ± iγ 2a+1 ), a = 1, . . . k.
2 2
This defines a set of n + 1 independent fermionic QHOs,
{γ a+ , γ b− } = δ ab , {γ a+ , γ b+ } = {γ a− , γ b− } = 0.
Assuming the Clifford algebra is represented faithfully and irreducibly, this gives 2k+1 states.
• Given a Clifford algebra in d = 2n dimensions, we can automatically construct a Clifford algebra
in d = 2n + 1 dimensions by adding
γ 2n+1 = (−i)n+1 γ 0 γ 1 . . . γ 2n−1 .
Note that this means the index µ in γ µ takes values {0, 1, . . . , 2n − 1, 2n + 1}, which is unfortu-
nately conventional.
• Given a Clifford algebra γ µ in d = 2n dimensions, there are various ways to construct a Clifford
algebra Γµ in d = 2n + 2 dimensions by tensor product. One way to do this is
Γµ = γ µ ⊗ σ 1 , Γ2n+1 = γ 2n+1 ⊗ σ 1 , Γ2n+2 = 1 ⊗ σ 2 .

• When d is even, the Dirac spinor representation is reducible. Conceptually, this is because the
generators Mµν are all built from an even number of γ matrices, so they preserve the parity of
the number of fermionic QHO excitations. Concretely, we can extract the irreps, called Weyl
spinors, using the projection operators (1 ± γ 2n+1 )/2. In all dimensions, we call these two Weyl
spinors left-chiral and right-chiral respectively; they correspond to the two special roots at the
end of the Dynkin diagram for so(2n). Going up to d = 2n + 1 combines these irreps into one.
• So far, all representations have been complex, in the mathematician’s sense. However, the Dirac
spinor is not complex, in the physicist’s sense; it is either real or pseudoreal. In the case it is
real (in the physicist’s sense), there is a chance we can extract a real representation (in the
mathematician’s sense) from it. Such a spinor is called a Majorana spinor, and it turns out to
be possible if d ≡ 0, 1, 2, 3, 4 (mod 8).
• The Weyl spinors may be self-conjugate, or conjugate to each other, depending on the dimension.
In the case where they are self-conjugate, there is a chance we can extract a real representation
from them, yielding a Majorana–Weyl spinor. This turns out to be possible if d ≡ 2 (mod 8).
• The facts above are summarized in the below table.
d dim γ Majorana Weyl Majorana–Weyl min. dim.
2 2 yes self yes 1
3 2 yes 2
4 4 yes complex 4
5 4 8
6 8 self 8
7 8 16
8 16 yes complex 16
9 16 yes 16
10 32 yes self yes 16
11 32 yes 32
12 64 yes complex 64
17 1. Introduction

The columns are the total spacetime dimension, the dimension of the (complex) gamma matrices,
the presence of Majorana, Weyl, and Majorana–Weyl spinors, and the real dimension of the
smallest possible representation.

Example. Spinors in low dimensions. For d = 2, we may choose

   
0 2 0 −1 1 1 0 1
γ = −iσ = , γ =σ = .
1 0 1 0

This gives a two-dimensional Dirac spinor which decomposes into one-dimensional “left-moving”
and “right-moving” Weyl spinors. To move to d = 3 we may add
 
3 3 1 0
γ =σ =
0 −1

which also may be used to project out the left-movers and right-movers. Of course, the cases d = 4
and hence d = 5 are familiar, though we won’t get the usual chiral representation if we use our
inductive scheme above.
18 2. SUSY Algebra and Representations

2 SUSY Algebra and Representations

2.1 The SUSY Algebra
Next, we deduce the SUSY algebra.

• The SUSY algebra is a graded Lie algebra. The operators in such an algebra obey

[Oa , Ob ]± ≡ Oa Ob − (−1)ηa ηb Ob Oa = iC eab Oe

where the gradings ηa take the form

(
0 Oa bosonic
ηa =
1 Oa fermionic.

Below we won’t use the [·, ·]± notation, but instead will make (anti)commutators explicit.

• The ‘super-Jacobi identity’ is

(−1)ηa ηc [Oa , [Ob , Oc ]± ]± + (−1)ηb ηa [Ob , [Oc , Oa ]± ]± + (−1)ηc ηb [Oc , [Oa , Ob ]± ]± = 0.

Note that the signs do nothing unless exactly two of the operators are fermionic.

• For the SUSY algebra, the generators are the Poincare generators P µ and M µν and the spinor
A A †
generators QA
α and Qα̇ = (Qα ) where A = 1, . . . , N . For N = 1, we have simple SUSY,
while for N > 1 we have extended SUSY. Here we focus on simple SUSY, which is the most
phenomenologically relevant.

• The Poincare algebra still holds, so by the grading we must find

[Qα , M µν ], [Qα , P µ ], {Qα , Qβ }, {Qα , Qβ̇ }.

We now consider these four in turn.

• The logic for the first is like that for [P σ , M µν ]. Since Qα is a spinor, it transforms as
  β   β
′ i µν i µν
Qα = exp − ωµν σ Qβ ≈ 1 − ωµν σ Qβ .
2 α 2 α

On the other hand, for a representation of the super-Poincare algebra on a Hilbert space,
   
i i i
Q′α = U † Qα U ≈ 1 + ωµν M µν Qα 1 − ωµν M µν = Qα − ωµν [M µν , Qα ].
2 2 2
Comparing the two, we conclude

[Qα , M µν ] = (σ µν )αβ Qβ

which is just the statement Qα is a spinor. By similar reasoning,

α̇ β̇
[Q , M µν ] = (σ µν )α̇β̇ Q .

More generally, for a SUSY generator transforming in an arbitrary spinor representation, σ µν

would be replaced with the representation matrices Mµν .
19 2. SUSY Algebra and Representations

• The spinors are translationally invariant, so on intuitive grounds

α̇
[Qα , P µ ] = [Q , P µ ] = 0.

To derive this more formally, note that we must have

α̇
[Qα , P µ ] = c(σ µ )αα̇ Q

by index structure and linearity on the right-hand side. The adjoint of this equation is
α̇
[Q , P µ ] = c∗ (σ µ )α̇β Qβ

and the Jacobi identity for P µ , P ν , and Qα reduces to

|c2 |(σ ν σ µ − σ µ σ ν )αβ Qβ = 0

which can only hold in general if c = 0.

• Next, consider {Qα , Qβ }. This transforms in the Lorentz representation (1/2, 0) × (1/2, 0) =
(1, 0) + (0, 0), but the (0, 0) piece vanishes because the Qα are anticommuting. Then

{Qα , Qβ } = k(σ µν )αβ Mµν

for an arbitrary constant k, where σ carries the appropriate SL(2, C) indices and Mµν is the only
thing that can absorb its Lorentz indices. By the Jacobi identity, the left-hand side commutes
with P µ but the right-hand side does not unless k = 0, so we must have

{Qα , Qβ } = 0.

• Finally, for {Qα , Qβ̇ } we have (1/2, 0) × (0, 1/2) = (1/2, 1/2), so we must have (why not the
Pauli–Lubanski vector?)
{Qα , Qβ̇ } = t(σ µ )αβ̇ Pµ .
There is no way to fix t. If we set t = 0, the algebra is trivial since the spinor and Poincare
parts are completely independent. Then by convention we set t = 2 for

{Qα , Qβ̇ } = 2(σ µ )αβ̇ Pµ .

Remarkably, this means that QQ is a translation! That is, if we start with a bosonic/fermionic
state and act with QQ, we get back a translated bosonic/fermionic state.

• Finally, let Ti generate an internal symmetry. Then usually we must have [Qα , Ti ] = 0. The
exception is the U (1) automorphism of the supersymmetry algebra called R symmetry,

Qα → eiλ Qα , Qα̇ → e−iλ Qα̇ .

If R generates this symmetry then

[Qα , R] = Qα , [Qα̇ , R] = −Qα̇ .

A specific SUSY theory may or may not have this R-symmetry, depending on the Lagrangian;
furthermore R-symmetry may be anomalous. As we’ll see later, postulating an R-symmetry is
useful for constraining the MSSM.
20 2. SUSY Algebra and Representations

2.2 N = 1 SUSY Representations

Next, we turn to representations of the SUSY algebra. We begin by reviewing Wigner’s classification.

• Recall that for su(2), we have a Casimir operator J 2 which labels the irreps; states in the irreps
are labeled by Jz .

• In the Poincare algebra, the Pauli–Lubanski vector is a “generalized spin”,

1
Wµ = ϵµνρσ P ν M ρσ
2
and it obeys the commutation relations

[Wµ , Pν ] = 0, [Wµ , Mρσ ] = i(ηµρ Wσ − ηµσ Wρ ), [Wµ , Wν ] = −iϵµνρσ W ρ P σ .

The first two simply say that Wµ is a translationally-invariant vector, while the third indicates
it does not form a closed algebra. In verifying these results it is useful to use the identity
a1 a
ϵa1 ...ap cp+1 ...cn ϵb1 ...bp cp+1 ...cn = −p!(n − p)!δ[b1
. . . δbpp] .

• As a result, the Poincare Casimirs are

C1 = P µ P µ , C2 = W µ Wµ .

The eigenvalue of C1 is written as m2 , where m is the mass of the particle.

• Next, we find the irreps using the little group. To use this method, we fix a reference momentum
pµ and look at the subalgebra that preserves the momentum; in the Poincare group the only
such operators are the Wµ , which take the form

1
Wµ = ϵµνρσ pν M ρσ .
2

• In the massive case, we have

pµ = (m, 0, 0, 0), W0 = 0, Wi = −mJi .

Then the little group is SO(3), and C2 indexes the spin.

• In the massless case, we have

pµ = (E, 0, 0, E), Wµ = E(J3 , −J1 + K2 , −J2 − K1 , −J3 )

which have the commutation relations

[W1 , W2 ] = 0, [W3 , W1 ] = −iEW2 , [W3 , W2 ] = iEW1

of the Euclidean group in two dimensions, E2 , which has infinite-dimensional irreps which are
not seen in nature. Concentrating on the finite-dimensional representations, the translations
must act trivially, leaving SO(2). The irreps are labeled by the helicity λ where W µ = λP µ ,
and projective representations allow half-integer λ.
21 2. SUSY Algebra and Representations

Next, we extend these results to the SUSY algebra.

• Since the SUSY generators commute with P µ , C1 remains a Casimir operator, so all particles
in a SUSY multiplet have the same mass. Now, for N = 1 SUSY, we have

[Wµ , Qα ] = −iPν (σ µν )αβ Qβ

which means that C2 is no longer a Casimir operator. This is as expected, as we can have
particles of different spin inside a SUSY multiplet.

• Instead, we define the operators

1
Bµ = Wµ − Qα̇ (σ µ )α̇β Qβ , Cµν = Bµ Pν − Bν Pµ
4
which yields a Casimir operator, called the superspin,

C̃2 = Cµν C µν .

• Next, we claim that in any SUSY multiplet the number nB of bosonic states equals the number
nF of fermionic states. Consider the fermion number operator (−)F , defined by

(−)F |B⟩ = |B⟩, (−)F |F ⟩ = −|F ⟩.

This operator anticommutes with Qα as, e.g. we have

(−)F Qα |F ⟩ = (−)F |B⟩ = |B⟩ = Qα |F ⟩ = −Qα (−)F |F ⟩.

Next we consider the trace

tr(−)F {Qα , Qβ̇ } = tr(−)F Qα Qβ̇ + tr(−)F Qβ̇ Qα = 0

by the anticommutation relation and the cyclic property of the trace. On the other hand,

tr(−)F {Qα , Qβ̇ } = 2 tr(−)F (σ µ )αβ̇ Pµ = 2(σ µ )αβ̇ pµ tr(−)F

where in the last step we restricted to states with momentum pµ . This can only hold if
X X X X
0 = tr(−)F = ⟨B|(−)F |B⟩ + ⟨F |(−)F |F ⟩ = ⟨B|B⟩ − ⟨F |F ⟩ = nB − nF .
B F B F

• There is an exception to this reasoning. If the supersymmetry is not broken, then the vacuum
states have pµ = 0, as we’ll see below. Then the trace tr(−)F evaluated over the entire Hilbert
space may be nonzero; it is called the Witten index. It is important because it is preserved
under certain deformations of the theory.

We now construct the massless SUSY multiplets. These are the most relevant phenomenologically
as almost all particles in the SM are ‘really’ massless, only acquiring mass from the Higgs.

• We take the reference momentum to be pµ = (E, 0, 0, E) and consider states in this irrep
with the reference momentum |pµ , λ⟩, where λ stands for all quantum numbers. The Casimirs
C1 = P µ Pµ and C̃2 = Cµν C µν are both zero. We already know that the Poincare generators
don’t give any new states, so we focus on the spinors.
22 2. SUSY Algebra and Representations

• Note that among the states |pµ , λ⟩,

 
1 0
{Qα , Qβ̇ } = 2(σ µ )αβ̇ Pµ = 2E(σ 0 + σ 3 ) = 4E .
0 0 αβ̇

Therefore, we have
⟨pµ , λ|{Q2 , Q2̇ }|pµ , λ⟩ = 0
which can only hold if Q2 |pµ , λ⟩ = 0.

• Meanwhile, the Q1 satisfy {Q1 , Q1̇ } = 4E, so defining

Q1 Q
a= √ , a† = √1̇ , {a, a† } = 1, {a, a} = {a† , a† } = 0
2 E 2 E
which are the commutation relations for a fermionic harmonic oscillator.

• Using the SUSY algebra, we may show that

1 α̇ 1 β̇
[Qα , Ji ] = (σi )β α Qβ , [Q , Ji ] = (σi )α̇β̇ Q
2 2
where the σi with indices in these positions are just ordinary Pauli matrices. Then
1 1
[a† , J 3 ] = (σ 3 )22 a† = − a† .
2 2
2̇ 1̇
Here, care must be taken with the index positions, noting that Q1̇ = −Q and Q2̇ = Q .

• Thus, we see that a† raises the J 3 eigenvalue by 1/2. Since the particle is moving in the −z
direction, it lowers the helicity λ by 1/2.

• We let |Ω⟩ = |pµ , λ⟩ be the state of highest helicity. Then we get just one other state,

a† |Ω⟩ = |pµ , λ − 1/2⟩.

As before, CPT flips λ, so we get irreps where the states have helicities {±λ, ±(λ − 1/2)}.

• We have chiral multiplets with λ = 0, 1/2, examples being

(squark, quark), (slepton, lepton), (Higgs, Higgsino)

along with vector/gauge multiplets with λ = 1/2, 1, examples being

(photino, photon), (gluino, gluon), (Wino, W ), (Zino, Z).

In general, the λ = 1/2 components are called gauginos. The SM matter fields can’t be gauginos,
because both particles in a vector multiplet transform the same way under SU (3)c × SU (2)L ×
U (1)Y , and vector particles must be created by gauge fields, which transform in the adjoint.

• As in Wigner’s classification, we need only consider integer or half-integer λ. We could have

λ = (1, 3/2), but it turns out to be phenomenologically disallowed, leaving only

λ = (3/2, 2) : (gravitino, graviton).

23 2. SUSY Algebra and Representations

Next, we consider massive supermultiplets, which are somewhat more complicated.

• In the massive case we have pµ = (m, 0, 0, 0) with Casimirs

C1 = P µ Pµ = m2 , C̃2 = 2m4 Y i Yi

where Yi is the superspin,

1 Bi
Yi = Ji − Qσ i Q = , [Yi , Yj ] = iϵijk Yk .
4m m
The eigenvalues of Y 2 = Y i Yi are thus y(y + 1), and we label irreps by m and y.

• Again restricting to states in an irrep with momentum pµ , we have

 
µ 0 1 0
{Qα , Qβ̇ } = 2(σ )αβ̇ Pµ = 2m(σ )αβ̇ = 2m .
0 1 αβ̇

Therefore, we have two sets of fermionic ladder operators,

Q1,2 Q1̇,2̇
a1,2 = √ , a†1,2 = √ , {ap , a†q } = δpq
2m 2m
which means that, starting from a vacuum state, we can build 4 states instead of 2.

• We let |Ω⟩ be a ‘vacuum’ state, annihilated by a1,2 . Then

1 √
Yi |Ω⟩ = Ji |Ω⟩ − Qσ i 2ma|Ω⟩ = Ji |Ω⟩
4m
so for a vacuum state the spin j and superspin y coincide,

|Ω⟩ = |m, j = y, pµ , j3 ⟩.

We can get all j3 values by Lorentz transformations, so there are 2y + 1 vacuum states.

• Since the SUSY generators carry spin 1/2, they act on spin j = y states to yield states of spin
j = y ± 1/2. Using the same relations as above, we find the SUSY generators change j3 by
1 1
[a†1 , J 3 ] = − a†1 , [a†2 , J 3 ] = − a†2
2 2
so that a†1 raises J3 as in the massless case and a†2 lowers it. Then it can be shown that
α̇ 3 α̇ β̇
[J 2 , Q ] = Q − (σi )α̇β̇ Q Ji , [J3 , a†1 a†2 ] = [J 2 , a†1 a†2 ] = 0.
4
The last identity states that acting with both a†1 and a†2 does not change J 2 .

• Therefore for y > 0, we have

a†1 |j = y, j3 ⟩ = k1 |j = y + 1/2, j3 + 1/2⟩ + k2 |j = y − 1/2, j3 + 1/2⟩

and
a†2 |j = y, j3 ⟩ = k3 |j = y + 1/2, j3 − 1/2⟩ + k4 |j = y − 1/2, j3 − 1/2⟩
where we suppress m and pµ and the ki are Clebsch–Gordan coefficients.
24 2. SUSY Algebra and Representations

• Finally, we have spin j states of the form |Ω′ ⟩ = a†2 a†1 |Ω⟩. These are not proportional to |Ω⟩,
since the ai annihilate |Ω⟩ but not |Ω′ ⟩. Thus we have

two particles of spin j = y, one particle each of spin j = y ± 1/2

which gives nF = nB = 2(2y + 1) as expected. The total number of physical states in the
multiplet is 4(2y + 1), i.e. 4 times the 2y + 1 vacuum states.

• The case y = 0 is slightly different. In this case we have y ⊗ 1/2 = 1/2, i.e. we have one particle
of spin 1/2 and two particles of spin 0. Again we have 4(2y + 1) = 4 states, but only three
SO(3) irreps.

• Finally, we consider parity transformations. Parity exchanges the Lorentz representations

(1/2, 0) and (0, 1/2), and since {Qα , Qβ̇ } = 2(σ µ )αβ̇ Pµ we must have

β̇ α̇ ˙
PbQa Pb−1 = ηP (σ 0 )αβ̇ Q , PbQ Pb−1 = ηP∗ (σ 0 )αβ Qβ

where ηP is a phase factor, and we are using the identity matrices σ 0 and σ 0 just to make the
indices match up. As a result,

PbP µ Pb−1 = (P 0 , −P), Pb2 QPb−2 = −Q.

• Heuristically, parity exchanges Q and Q. Note that the ai annihilate |Ω⟩ and the a†i annihilate
|Ω′ ⟩. Then parity exchanges the highest and lowest states. The states with definite parity are

|±⟩ = |Ω⟩ ± |Ω′ ⟩, P |±⟩ = ±|±⟩

where |+⟩ is scalar and |−⟩ is pseudoscalar.

Note. A common confusion with supersymmetry is the following: “Supersymmetry is a symmetry

that swaps fermions and bosons. But it can’t be true, because a state with two identical bosons
can’t map to anything, by the Pauli exclusion principle.” The reason this is misguided is that the
intuition of symmetry transformations leaving a system the same only applies for symmetry groups.
We never work with a “SUSY group”, only with the SUSY algebra. Applying an operator in a
symmetry algebra (such as Q, but also L± ) represents an infinitesimal change in state, rather than
a new state related by symmetry. The reasoning above really tells us why we can’t exponentiate
the SUSY algebra into a SUSY group: the resulting transformations wouldn’t be invertible.

2.3 Extended SUSY

Next, we turn to the case of extended SUSY, N > 1.

• The SUSY algebra remains the same, but the anticommutation relations between SUSY gen-
erators are modified. Before, we only showed that {QA B
α , Qβ } could not depend on Pµ or Mµν .
(In fact, in general dimension, the anticommutator can also depend on Pµ .) However, for N > 1
it is consistent to include a central charge,

{QA µ A
α , Qβ̇B } = 2(σ )αβ̇ Pµ δB , {QA B
α , Qβ } = ϵαβ Z
AB
.

The central charges Z AB are bosonic and antisymmetric; they commute with all of the generators
and with each other. Thus they form an abelian invariant subalgebra of internal symmetries.
25 2. SUSY Algebra and Representations

• Specifically, if G is the set of internal symmetries, define the R-symmetry group H ⊂ G to be

the set that do not commute with the supersymmetry generators, i.e. the ones that change the
QAα nontrivially. By the Coleman–Mandula theorem, they do not affect the Poincare generators.

• In the case Z AB = 0, the R-symmetry group is U (N ), e.g.

A B
QA A B
α → U B Qα , Qα̇ → (U ∗ )AB Qα̇

generalizing the U (1) symmetry found earlier. That is, the QA transform in the fundamental
A
and the Q in the antifundamental.

• The full R-symmetry need not be realized, depending on the theory. For example, the maximal
R-symmetry in N = 4 is U (4), but the actually realized symmetry in N = 4 SYM is SU (4).

Next, we proceed to the massless irreps.

• Again taking pµ = (E, 0, 0, E), we have

 
1 0
{QA
α , Qβ̇B } = 4E δA
0 0 αβ̇ B

which again implies that QA µ

2 |p , λ⟩ = 0, and hence Z
AB |pµ , λ⟩ = 0.

• We now define N sets of fermionic creation and annihilation operators,

A
A† QA Q
a = √1 , a = √1̇ ,
A
{aA , a†B } = δB
A
2 E 2 E
where we flip the convention here for convenience. We start with the vacuum |Ω⟩ with helicity
λ0 . Then the N states
†
aA |Ω⟩
all have helicity λ0 + 1/2. More generally we may act with k creation operators, giving Nk

states with helicity λ0 + k/2, for a total of 2N states.

• For example, for N = 2 and λ0 = 0 we have the vector multiplet

λ = 0, 2 × λ = 1/2, λ = 1.

We could also keep track of the R-symmetry; in this case the λ = 1/2 transform in the 2 of
U (2) while the others transform in the 1.

• Upon restriction to N = 1, the N = 2 vector multiplet decomposes into an N = 1 vector and

chiral multiplet. Note that these multiplets depend on which of the two supersymmetries we
restrict to! In a CPT symmetric theory, the vector multiplet must be accompanied with its
CPT conjugate.

• Next, for N = 2 and λ0 = −1/2, we have the hyper multiplet

λ = −1/2, 2 × λ = 0, λ = 1/2

which decomposes into two N = 1 chiral multiplets.

26 2. SUSY Algebra and Representations

• It’s tempting to conclude this multiplet could be its own CPT conjugate (as long as, e.g. there
are no additional complex internal quantum numbers). However, it turns out that the λ = 0
sector must be a real representation of the R-symmetry group, while in this case it is pseudoreal.
(Central charges make little difference here, as they only break the R-symmetry group from
U (2) to SU (2).) Hence the hyper multiplet contains two copies of the helicities above, while
only one copy is a “half hyper multiplet”.

• Next, for N = 4 and λ0 = −1, we have the vector multiplet

λ = −1, 4 × λ = −1/2, 6 × λ = 0, 4 × λ = 1/2, λ = 1.

This is the only N = 4 multiplet where |λ| ≤ 1. Restricting to N = 2, we get two N = 2 vector
multiplets and hypermultiplets. Restricting to N = 1, we get two N = 1 vector multiplets and
six N = 1 chiral multiplets. Since the whole N = 4 multiplet is symmetric under λ → −λ, and
the 6 of U (4) is real, it could be its own CPT conjugate, in which case the submultiplets pair
up under CPT.

• Finally, for N = 8 and λ0 = −2, we have the ‘maximum multiplet’ or ‘gravity multiplet’

λ = ±2, 8 × λ = ±3/2, 28 × λ = ±1, 56 × λ = ±1/2, 70 × λ = 0.

Since the 28 of U (8) is real, this multiplet again could be its own CPT conjugate. The general
rule is that this holds when the maximum helicity is N /4, so we’ve exhausted the realistic
self-CPT conjugate possibilities above.

We now comment on the physical properties of the particles in these multiplets.

• Note that renormalizable field theories must have |λ| ≤ 1, since otherwise the propagator does
not fall off fast enough, so we require N ≤ 4 for renormalizability. This doesn’t mean such
theories are physically irrelevant, as gravity isn’t renormalizable either.

• Generally, a massless particle with |λ| ≥ 1 must couple to a conserved current, i.e. a conserved
vector for λ = ±1 (as in electromagnetism) and a conserved tensor for λ = ±2 (as in gravity).
This is necessary to remove the growth in the propagators, which would otherwise violate
perturbative unitarity.

• There aren’t conserved tensors of higher rank by the Coleman–Mandula theorem and its gener-
alizations. Thus, |λ| > 2 is forbidden, and we can only have one particle with λ = 2 because
all such particles must act like gravitons. Then N = 8 is the maximum realistic number of
supersymmetries.

• Theories with massless particles of helicity |λ| > 2 are called higher-spin theories and are rather
exotic. To be realized in quantum field theory, they must either be free, or contain an infinite
tower of particles of essentially every spin.

• In light of the above, N = 4 is the ‘nicest’ for gauge theory and N = 8 is the ‘nicest’ for gravity,
explaining why N = 4 SYM and N = 8 SUGRA are so well studied.

• However, N > 1 supersymmetry is ‘non-chiral’, in contradiction with the Standard Model.

First, note that with the sole exception of the N = 2 hypermultiplet, all such multiplets contain
λ = ±1 particles.
27 2. SUSY Algebra and Representations

• It can be argued generally that these particles must be gauge bosons; as a result, they must
transform in the adjoint representation, which is in general a real representation. (For matrix
Lie groups, this is easy to see, since it is essentially the fundamental times the antifundamental.)
Since internal symmetries commute with SUSY transformations, the λ = ±1/2 particles must
also transform in the adjoint.

• If the multiplet contains both λ = ±1/2, then these particles transform in the same representa-
tion, so we cannot get a chiral theory; this accounts for the N = 2 hypermultiplet.

• On the other hand, if the multiplet contains only, say, λ = 1/2, then by CPT there must be
another multiplet with λ = −1/2 which transforms in the conjugate representation. Since the
adjoint is real, the λ = −1/2 particle also transforms in the adjoint, and we again don’t have a
chiral theory.

• Helicity λ = 3/2 is also somewhat exotic. It turns out that it must couple to the supersymmetry
current (i.e. the current whose charge corresponds to the SUSY generators) and be associated
with a field with gauge symmetry, where the gauge symmetry is local supersymmetry. This
necessitates local super-Poincare gauge symmetry, i.e. supergravity, so a theory with a λ = 3/2
particle (called a gravitino) must also have a λ = 2 particle.

Finally, we consider the massive multiplets.

• We consider pµ = (m, 0, 0, 0) which gives

 
1 0 A
{QA
α , Qβ̇B } = 2m δ .
0 1 B

We first consider the case where all the central charges vanish. Then we have 2N pairs of
fermionic creation and annihilation operators,
A
QA † Q
aA
α =√α , aA
α̇ = √ α̇
2m 2m
which yields a much larger multiplet containing 22N states for each vacuum state, for a total
of (2y + 1)22N . As before, each of the raising operators changes the spin by 1/2, while the
† A†
combination aA1̇
a2̇ does not change the spin.

• For example, for N = 2 with a spin 0 vacuum, at each level of raising we have

1 × spin 0 → 4 × spin 1/2 → 3 × spin 0, 3 × spin 1 → 4 × spin 1/2 → 1 × spin 0

which gives 16 total states, with 5 spin 0 particles, 4 spin 1/2 particles, and 1 spin 1 particle.
There are 8 fermionic states and 8 bosonic states, as expected. However, note that the number
of fermionic Poincare irreps does not match the number of bosonic Poincare irreps.

• In the case Z AB =
̸ 0, the size of the multiplets depends on the central charges. It is simplest
to begin with N = 2, where the central charge has one degree of freedom. Then we may take
 
A 1 0 A A B
{Qα , Qβ̇B } = 2m δB , {QA B
1 , Q2 } = 2Zϵ
AB
, {Q1̇ , Q2̇ } = 2ZϵAB
0 1

with all other anticommutators vanishing.

28 2. SUSY Algebra and Representations

• We need to perform a change of basis to yield fermionic QHOs. To do this, we take

1 2 1 2 1 2 1 2
a1 = √ (Q11 + α1 Q2̇ ), a2 = √ (Q12 − α2 Q1̇ ), b1 = √ (Q11 − α1 Q2̇ ), b2 = √ (Q12 + α2 Q1̇ )
2 2 2 2
† A
along with the conjugates, using (QA
α ) = Qα̇ . These are constructed so that all anticommutators
vanish except for

{ai , a†i } = 2m + αi Z + αi Z, {bi , b†i } = 2m − αi Z − αi Z, {ai , b†i } = αi Z − αi Z

for i = 1, 2.

• In order to get an independent set of fermionic QHOs, we may choose

α1 = α2 = ei arg Z

upon which the nonzero anticommutators become

{ai , a†i } = 2m + 2|Z|, {bi , b†i } = 2m − 2|Z|.

Since the left-hand sides are positive definite, we have the BPS bound

|Z| < m.

• If the BPS bound is not saturated, then upon a rescaling, we have four independent fermionic
QHOs and get 16 degrees of freedom as before. However, Z = m, then bi must be realized
trivially on the multiplet, so we only get 4 degrees of freedom.

• More generally, for even N we can diagonalize Z AB to 2 × 2 blocks of the above form, with
values Z1 through ZN /2 . There is a BPS bound for each individual block, 2m ≥ Zi .

• If none of these bounds are saturated, we get a “long multiplet” of 22N states. If k of them
are, we get a “short multiplet” of 22(N −k) states, by the same logic as the N = 2 case. If all of
them are, we get an “ultra-short multiplet” of 2N states.

• Historically, BPS bounds and states were first found for soliton/monopole solutions of the Yang–
Mills equations. The BPS states are stable because they are the lightest charged particles.

• Extremal black holes are also BPS states in extended supergravity theories. They are stable,
as they are the endpoints of Hawking radiation. In string theory, some D branes are BPS.

• BPS states are important for understanding strong/weak coupling dualities, because they are
distinguished by short multiplets, and multiplets can’t change size as the coupling continuously
changes from weak to strong.

Note. We can see how the Higgs mechanism would work in a supersymmetric field theory by
looking at the structure of the multiplets. Without supersymmetry, a helicity ±1 Poincare irrep
“eats” a helicity 0 irrep to gain mass, forming a spin 1 irrep. Similarly, for N = 1 a vector multiplet
eats a chiral multiplet. Accounting for their CPT conjugates as well, this forms the y = 1/2 massive
multiplet. For N = 2 a vector multiplet again eats a chiral multiplet. Accounting for their CPT
conjugates, this forms the y = 0 massive multiplet.
29 2. SUSY Algebra and Representations

2.4 SUSY in Various Dimensions

Now we consider SUSY in various dimensions.

• In general, the SUSY generators QI (for I = 1, . . . , N ) are taken to transform in the minimal
spinor representation of so(1, d − 1). We define N to be the ratio of the number of real
supercharges NQ to the real dimension of the minimal spinor representation. However, it turns
out that generally the important variable is NQ , not N .

• A summary of the degrees of freedom for low dimension is shown below.

d spinors NQ max R-sym SUSY

Nmax SUGRA
Nmax
1 N Majorana N SO(N ) – –
2 (nL , nR ) Maj–Weyl nL + nR SO(nL ) × SO(nR ) (8, 8) (16, 16)
3 N Majorana 2N SO(N ) 8 16
4 N Weyl 4N U (N ) 4 8
5 N Dirac 8N Sp(N ) 2 4
6 (nL , nR ) Weyl 8(nL + nR ) Sp(nL ) × Sp(nR ) nL + nR = 2 nL + nR = 4
7 N Dirac 16N Sp(N ) 1 2
8 N Weyl 16N U (N ) 1 2
9 N Majorana 16N SO(N ) 1 2
10 (nL , nR ) Maj–Weyl 16(nL + nR ) SO(nL ) × SO(nR ) nL + nR = 1 nL + nR = 2
11 N Majorana 32N SO(N ) 0 1

Note that in d = 2, 6, 10 there are two distinct “minimal” representations, which are not
conjugate; hence N really must be described by two independent numbers. Similarly, in d = 4
(and others) the Weyl and Majorana spinors have the same real dimension, so one may take
the supercharges to be either; the two are equivalent as real representations.

• As for d = 4, Nmax
SUSY is the largest amount of supersymmetry where we can have massless
SUGRA is the same for helicity |λ| > 2.
particles without requiring helicity |λ| > 1, while Nmax
(typo in 11 row for number of degrees of freedom? shouldn’t things be called
Majorana?)

• Looking at higher dimensions, there is no rigid supersymmetry beyond d = 10, and N = (1, 0)
SYM in d = 10 is closely related to N = 4 SYM in d = 4.

• Furthermore, there is no supergravity beyond d = 11. The N = (1, 1) and N = (2, 0) SUGRA
theories in d = 10 are low-energy limits of type IIa and IIb superstring theory, respectively.
The N = 1 supergravity theory in d = 11 is thought to be the low-energy limit of M -theory.

Example. SUSY in d = 2. The Lorentz group is SO(1, 1) ∼ = R, so all representations are one-
dimensional and we may classify them by their SO(1, 1) charge, which we call the spin. In the case
N = (1, 1), there are right-moving and left-moving real supercharges Q± with spin ±1/2. The boost
generator M transforms in the trivial representation of SO(1, 1), while the translation operators
P µ decompose into spin ±1. The algebra contains

[P 0 , M ] = P 1 , [P 1 , M ] = P 0 , [Q± , M ] = ±Q±
30 2. SUSY Algebra and Representations

with all other commutators vanishing. As for the anticommutation relations between the SUSY
charges, representation theory fixes

{Q+ , Q+ } = P+ , {Q− , Q− } = P− , P± = P 0 ± P 1 .

Next, since {Q+ , Q− } carries spin 0, it could contain M or a central charge. By the same argument
as in d = 4, it can’t contain M , so
{Q+ , Q− } = Z.
For N = (2, 0), the result is similar, but there would be no room to include central charges.

Example. SUSY in d = 3. In this case, the SUSY generators are two-component Majorana spinors.
(An easy way to see this is to note that so(2, 1) ∼
= sl(2, R), and the 2 of sl(2, R) is manifestly real.)
We take the gamma matrices to be

(γ µ )αβ = (−iσ 2 , σ 1 , σ 3 ).

Spinor indices may be raised and lowered using ϵαβ as in d = 4. We have

γ µ γ ν = η µν − ϵµνρ γρ

where ϵ012 = 1 and we use mostly positive signature. Then the spinor representation matrices are
i
Mµν = − ϵµνρ γ ρ
2
which are pure imaginary as required, which implies
i
[Mµν , QIα ] = ϵµνρ (γ ρ QI )α .
2
Since 2 × 2 = 1 + 3, the general anticommutator of the SUSY charges has the form
µ
{QIα , QJβ } = c0 γαβ Pµ δ IJ + ϵαβ Z IJ
µ
for an antisymmetric central charge Z IJ , as γαβ is symmetric in α and β. (finish this)
31 3. Superspace and Superfields

3 Superspace and Superfields

3.1 Superspace
So far we’ve only been working with particle states, but in quantum field theory we would like
to understand them as excitations generated by fields. We hence turn to the problem of writing
supersymmetric Lagrangians. This is difficult, because supersymmetry introduces strong constraints.

• Before supersymmetry, we constructed Lorentz-invariant Lagrangians using fields φ(xµ ) which

were functions of coordinates xµ in Minkowski space, transforming under a definite representa-
tion of the Lorentz group.

• For example, we could have introduced the four fields of the Dirac spinor as separate objects,
but then Lorentz invariance would have strongly constrained the couplings. It’s much more
convenient to work with the Dirac spinor as one object.

• Similarly, in supersymmetry, we work with superfields Φ(X) which transform under a definite
representation of the super-Poincare group. We will see this requires X to be in ‘superspace’,
Minkowski space with extra Grassmann dimensions.

• Also note that, in contrast to the particle case, symmetries for the fields should ideally hold
off-shell. That is, the Lagrangian must be SUSY-invariant independent of the equations of
motion. We will find below that we must introduce auxiliary fields to achieve this.

To motivate superspace, we require the basics of group actions.

• Every Lie group G has a group manifold MG .

– For G = U (1), the elements are g = eiα with α ∈ [0, 2π], so MG = U (1).
– For G = SU (2), the elements are
 
α β
g= , |α|2 + |β|2 = 1
−β ∗ α∗

which implies MG = S 3 .
– For G = SL(2, C), we’ve already seen MG = R3 × S 3 .

• We can do the same reasoning for cosets G/H.

– Consider G/H = SU (2)/U (1) ∼ = SO(3)/SO(2). The U (1) factor can be taken to be
diag(eiγ , e−iγ ), which can be used to make the parameter α above real. Then MG/H = S 2 .
– More generally, MSO(n+1)/SO(n) = S n .
– We have Poincare/Lorentz = {ω µν , aµ }/{ω µν } = {aµ } = Minkowski.

• This last result motivates us to define N = 1 superspace as the coset

super Poincare/Lorentz = {ω µν , aµ , θa , θα̇ }/{ω µν }.

Explicitly, elements of the super Poincare group take the form

 
α̇
g = exp i(ω µν Mµν + aµ Pµ + θα Qα + θα̇ Q )
32 3. Superspace and Superfields

where the θα and θα̇ transform like Weyl spinors and hence must be Grassmann numbers by
spin-statistics. Note that this implies
β̇ β̇
{Qα , Qα̇ } = 2(σ µ )αα̇ Pµ , [θα Qα , θ Qβ̇ ] = 2θα (σ µ )αβ̇ θ Pµ .
Superspace is ordinary space augmented with the Grassmann dimensions θα , θα̇ .
We now review properties of Grassmann numbers.
• For a single Grassmann number θ, an arbitrary function f (θ) can be expanded as
f (θ) = f0 + f1 θ
and define df /dθ = f1 . Integrals are defined as
Z Z
dθ = 0, dθ θ = 1

which implies that the ‘Dirac delta’ is δ(θ) = θ. Note that the integral is equal to the derivative.
• Now consider spinors of Grassmann numbers θα , θα̇ . Their squares as defined, as earlier, by
α̇
θθ = θα θα , θθ = θα̇ θ
which gives the identities
1 α̇ β̇ 1
θα θβ = − ϵαβ θθ, θ θ = ϵα̇β̇ θθ.
2 2
• Derivatives are defined by
β̇
∂θβ β̇ ∂θ
∂α θ ≡ α = δαβ ,
β
∂ α̇ θ ≡ α̇
= δα̇β̇ .
∂θ ∂θ
However, note that by raising and lowering indices, this implies that
α̇
∂ α θβ = −δβα , ∂ θβ̇ = −δβ̇α̇ .
In index-free notation we thus have
(ψ∂)(θχ) = ψχ, (ψ∂)(θχ) = −ψχ.

• For multiple integrals, we have Z Z

dθ1 dθ2 θ2 θ1 = 1

but we also note that θ2 θ1 = (1/2)θθ. Then it is convenient to define

Z Z Z
1 1
d2 θ = dθ1 dθ1 , d2 θ = − dθα dθβ ϵαβ
2 4
which gives the simple result Z
d2 θ θθ = 1.

Similarly, we define Z
2 1 α̇ β̇
d θ = dθ dθ ϵα̇β̇ , d2 θ θθ = 1.
4
• Integration can again be related to differentiation,
Z Z
2 1 αβ 1
d θ = ϵ ∂α ∂β , d2 θ = − ϵα̇β̇ ∂ α̇ ∂ β̇ .
4 4
33 3. Superspace and Superfields

3.2 The Scalar Superfield

We now define the N = 1 scalar superfield. First, we review ordinary scalar fields.

• Consider a scalar field φ(xµ ). It is an element of the function space F which is a representation
of the Poincare group. Let P µ be the representation of P µ on F. Then

φ(xµ ) → exp(iaµ P µ )φ(xµ ) = φ(xµ + aµ ), Pµ = −i∂µ .

Similarly, we may define the action of M µν on F. Since φ is a scalar, we have

Mµν = −i(xµ ∂ν − xν ∂µ ).

If φ transformed in a nontrivial Lorentz representation, this expression would have extra terms,
as the Lorentz transformation would act on the field indices.

• More generally, Lorentz transformations are defined as vector fields under spacetime, and the
changes of fields under these transformations are given by Lie derivatives. In the case of a
scalar, this is just the vector field acting on the scalar, as we see above.

• Upon quantization, φ is an operator in a Hilbert space, and

φ → exp(−iaµ P µ )φ exp(iaµ P µ ).

Comparing our two expressions, at first order in aµ , the change in φ under translation is

δφ = i[φ, aµ P µ ] = iaµ Pµ φ = aµ ∂µ φ.

Note that these results are not specific to scalar fields.

Next, we turn to the scalar superfield. More complicated superfields which are not scalar-valued
can also appear, but the scalar superfield and restrictions of it will suffice for our purposes.

• An N = 1 scalar superfield is a function on superspace S(xµ , θα , θα̇ ). Performing a Taylor

expansion in the Grassmann variables, we get a finite number of terms,

S(xµ , θα , θα̇ ) = φ(x) + θψ(x) + θχ(x) + θθ M (x) + θθ N (x)

+ (θσ µ θ)Vµ (x) + (θθ) θλ(x) + (θθ) θρ(x) + (θθ)(θθ) D(x).

The scalar superfield contains fields that are not Lorentz scalars, such as ψ. The spinor fields
are Grassmann; these Grassmann variables are independent of the superspace variables θ, and
come in via the path integral measure.

• It’s clear that the terms up to second order are the most general possible. We could write more
third-order and fourth-order terms using σ µ , but they would be redundant with our existing
terms by Fierz identities.

• Since S is a scalar superfield, we know how Poincare transformations act on it, so we focus on
‘supertranslations’,

S(xµ , θα , θα̇ ) → e−i(ϵQ+ϵQ) S(xµ , θα , θα̇ )ei(ϵQ+ϵQ)

µP µP
= e−i(ϵQ+ϵQ) e−i(x µ +θQ+θQ)
S(0, 0, 0)ei(x µ +θQ+θQ)
ei(ϵQ+ϵQ) .

The second step is, at this point, a reasonable ansatz that we will verify holds.
34 3. Superspace and Superfields

• This expression may be simplified using the Baker-Campbell-Hausdorff formula

eA eB = eA+B+[A,B]/2+...
where all higher-order terms are zero because in this case,
[A, B] = [xµ Pµ + θQ + θQ, ϵQ + ϵQ] = (−i(ϵσ µ θ) + i(θσ µ ϵ))Pµ
which commutes with both A and B. Therefore,
S(xµ , θα , θα̇ ) → e−i((x+δx)P +(θ+δθ)Q+(θ+δθ)Q) S(0, 0, 0)ei((x+δx)P +(θ+δθ)Q+(θ+δθ)Q)
where
δxµ = −i(ϵσ µ θ) + i(θσ µ ϵ), δθα = ϵα , δθα̇ = ϵα̇ .
That is, a translation in superspace induces a translation in real space.
• Again, we can also think of φ classically as an element of a function space, where
S(xµ , θα , θα̇ ) → ei(ϵQ+ϵQ) S(xµ , θα , θα̇ ) = S(xµ − i(ϵσ µ θ) + i(θσ µ ϵ), θ + ϵ, θ + ϵ)
where the second equality comes from our result above. Thus we have
β̇
Qα = −i∂α − (σ µ )αβ̇ θ ∂µ , Qα̇ = i∂ α̇ + θβ (σ µ )β α̇ ∂µ , Pµ = −i∂µ .

We can then verify that Qα and Qα̇ satisfy the supersymmetry algebra,
{Qα , Qα̇ } = 2(σ µ )αα̇ Pµ , {Qα , Qβ } = 0.
Note that this is sometimes phrased in terms of the variations δϵ = iϵQ, δϵ = iϵQ, in which case
[δϵ , δϵ ] = −2(ϵσ µ ϵ)Pµ .

• Note that the ‘extra’ terms cancel out in the supertranslation operators,
α̇
θQ + θQ = −iθα ∂α − iθ ∂ α̇
which retroactively justifies our ansatz. By comparing our two expressions to first order in ϵ,
δS = i[S, ϵQ + ϵQ] = i(ϵQ + ϵQ)S, i(ϵQ + ϵQ) = ϵ∂ − i(ϵσ µ θ)∂µ − ϵ∂ + i(θσ µ ϵ)∂µ .

• We now tabulate how each of the individual pieces transform.

δφ = ϵψ + ϵχ
δψ = 2ϵM + σ µ ϵ(i∂µ φ + Vµ )
δχ = 2ϵN − ϵσ µ (i∂µ φ − Vµ )
i
δM = ϵλ − ∂µ ψ∂ µ ϵ
2
i
δN = ϵρ + ϵσ µ ∂µ χ
2
i
δVµ = ϵσµ λ + ρσµ ϵ + (∂ ν ψσµ σ ν ϵ − ϵσ ν σµ ∂ ν χ)
2
i
δλ = 2ϵD + (σ ν σ µ ϵ)∂µ Vν + iσ µ ϵ∂µ M
2
i
δρ = 2ϵD − (σ ν σ µ ϵ)∂µ Vν + iσ µ ϵ∂ν N
2
i
δD = ∂µ (ϵσ µ λ − ρσ µ ϵ).
2
35 3. Superspace and Superfields

Note that δD is a total derivative.

We now make some general remarks on superfields.

• A superfield is a function on superspace that transforms under supertranslations as δS =

i(ϵQ + ϵQ)S.

• If S1 and S2 are superfields, then so is S1 S2 , because

δ(S1 S2 ) = (δS1 )S2 + S1 (δS2 ) = (i(ϵQ + ϵQ)S1 )S2 + S1 (i(ϵQ + ϵQ))S2 .

But this is equal to i(ϵQ + ϵQ)S1 S2 , because Q and Q obey the Leibniz rule.

• Similarly, linear combinations of superfields are also superfields.

• The field ∂µ S is a superfield, but ∂α S is not, because

δ(∂α S) = ∂α (δS) = i∂α (ϵQ + ϵQ)S ̸= i(ϵQ + ϵQ)(∂α S)

because ∂α and ϵQ + ϵQ do not commute. This makes sense because ∂α S has terms only up to
linear order in θ, while S has terms up to quadratic order.

• We can alternatively derive this with commutators/Poisson brackets. We have

δ(∂α S) = i[∂α S, ϵQ + ϵQ] = i∂α [S, ϵQ + ϵQ] = i∂α (ϵQ + ϵQ)S

as above. Here, Q is not a vector field; it is instead the Noether charge associated with
supertranslation under Q. Since the charge is integrated over superspace, ∂α Q = 0.

• Instead, we define the covariant derivatives

β̇
Dα = ∂α + i(σ µ )αβ̇ θ ∂µ , Dα̇ = ∂ α̇ + iθβ (σ µ )β α̇ ∂µ

which satisfy
{Dα , Qβ } = {Dα , Qβ̇ } = {Dα̇ , Qβ } = {Dα̇ , Qβ̇ } = 0.
To verify these results, we need to use the fact that ϵ is anticommuting. Note that these
covariant derivatives have nothing to do with gauge fields. The covariant derivatives are very
similar to Q and Q, but not quite the same.

• We can further verify that

{Dα , Dβ̇ } = 2i(σ µ )αβ̇ ∂µ , {Dα , Dβ } = {Dα̇ , Dβ̇ } = 0, [Dα , ϵQ + ϵQ] = 0

The last result shows that if S is a superfield, so is Dα S. The first result means that the
connection ∇ = (∂µ , Dα , Dα̇ ) on flat superspace has nontrivial torsion.

Note. Our derivations above are somewhat heuristic. On a deeper level, given a Lie group G with
a subgroup H, there is an induced action of G on G/H by left multiplication; this is the action of
the supersymmetry generators we found heuristically above. The SUSY covariant derivatives above
are defined by the action of G on G/H by right multiplication, which is why they anticommute
with the Qα . The SUSY covariant derivatives (along with ∂µ ) define a connection on superspace,
and the fact that {Dα , Dβ } =
̸ 0 indicates the connection has nontrivial torsion.
36 3. Superspace and Superfields

The scalar superfield is not an irreducible representation of supersymmetry, so it can be reduced.

• Suppose only the “body” φ is nonzero. Then under a supertranslation we pick up ψ and χ
terms unless ∂µ φ = 0, so we need φ to be constant, which is not interesting. We might also
attempt to set only ψ nonzero, but then we pick up φ terms. Instead, we build our constraints
using covariant derivatives.

• The chiral and antichiral superfield satisfy

Dα̇ Φ = 0, Dα Φ = 0.

Intuitively, these only depend on θ and θ, respectively, so they contain only left-chiral and
right-chiral spinor fields, respectively. As we’ll see below, these generate the particles in a chiral
multiplet.

• The vector or real superfield satisfies V = V † , where the dagger is a complex conjugate at the
classical level. It is the SUSY analogue of a real vector gauge field. These generate the particles
in a vector multiplet.

• The linear superfield L is a vector superfield satisfying DDL = 0. Since it contains the constraint
∂µ V µ = 0, it is the SUSY analogue of a conserved current.

Note. What about superspace beyond N = 1 SUSY? Though such formalisms do exist, they are
generally not useful. The problem is that for NQ real supercharges, we require NQ Grassmann
directions and hence 2NQ = 24N components for a general superfield. This is much larger than
even a long multiplet, which has only 22N states. The constraints required to eliminate this many
degrees of freedom become quite complicated, and in some cases cannot be imposed consistently.
For general dimension, a rule of thumb is that a superspace formalism is only useful for NQ ≤ 4.

3.3 Chiral and Vector Superfields

We now consider the chiral superfield in detail.

• For convenience, we define

y µ = xµ + iθσ µ θ
and consider Φ = Φ(y(x, θ, θ), θ, θ). The covariant derivative is
∂Φ ∂y µ
Dα̇ Φ = ∂ α̇ Φ + + iθβ (σ µ )β α̇ ∂µ Φ = ∂ α̇ Φ − iθβ (σ µ )β α̇ ∂µ Φ + iθβ (σ µ )β α̇ ∂µ Φ = ∂ α̇ Φ
∂y µ ∂θα̇
where we picked up a minus sign from anticommuting a Grassmann derivative. That is, in the
(y, θ, θ) variables, the covariant derivatives act as
∂
Dα̇ = ∂ α̇ , Dα = ∂α + 2i(σ µ θ)α .
∂y µ

• Therefore, a chiral superfield obeys ∂ α̇ Φ = 0 when expressed in terms of y, θ, and θ, so

√
Φ(y, θ) = φ + 2θψ + θθF

where we suppressed the position argument y of the fields on the right. Here, φ and F are
complex scalar fields, and ψ is a left-chiral Weyl spinor.
37 3. Superspace and Superfields

• The field F is auxiliary, and hence determined by the others on-shell. Then on-shell there are
2 spin 0 degrees of freedom and 2 spin 1/2 degrees of freedom, obeying nB = nF .

• Note that off-shell, there are 4 bosonic degrees of freedom, since φ and F are complex, and
4 fermionic degrees of freedom. This is convenient, because it means SUSY is manifest even
off-shell, and is essentially the reason we need the F field.

• Expressing Φ in terms of xµ and Taylor expanding, we have

√ i 1
Φ(x, θ, θ) = φ + 2θψ + (θθ)F + i(θσ µ θ)∂µ φ − √ (θθ)∂µ ψσ µ θ − (θθ)(θθ)∂µ ∂ µ φ
2 4

where we suppressed x-dependence. Note that this is not an approximation; higher-order terms
in the Taylor expansion are just all identically zero.

• Under a supersymmetry transformation δΦ = i(ϵQ + ϵQ)Φ, the components change as

√ √ √ √
δφ = 2 ϵψ, δψ = i 2 σ µ ϵ∂µ φ + 2 ϵF, δF = i 2 ϵσ µ ∂µ ψ.

In particular, note that δF is a total derivative, just as δD was for a general superfield. It is
straightforward to check manually that this satisfies the supersymmetry algebra. Later on, the
transformation rules for the anti-chiral multiplet will also be useful; they are
√ √ √ √
δφ = 2 ϵψ, δψ = i 2 σ µ ϵ∂µ φ + 2 ϵF , δF = i 2 ϵσ µ ∂µ ψ.

• Note that the product of chiral superfields is also a chiral superfield, with
√ √
Φ1 Φ2 = (φ1 + 2θψ1 + θθF1 )(φ2 + 2θψ2 + θθF2 )

giving
φ′ = φ1 φ2 , ψ ′ = ψ1 φ2 + ψ2 φ1 , F ′ = F1 φ2 + F2 φ1 − ψ1 ψ2 .
This is one of the key benefits of superfields; we may easily take products of them.

• More generally, any holomorphic function f (Φ) is also chiral, but Φ = Φ† is antichiral. The
fields Φ† Φ and Φ† + Φ are real, but neither chiral nor antichiral.

• We can further constrain chiral superfields. For example, let X be a nilpotent chiral superfield,
so X 2 = 0 and Dα̇ X = 0. Renaming some of the fields, we have
√
X(y, θ) = x + 2θψx + θθFx

and squaring this gives

√
X 2 = x2 + 2 2 xθψx + (2xFx − ψx2 )(θθ) = 0.

The final term vanishes if x = ψx2 /2Fx , and this makes the first two terms automatically vanish
as well, because they are proportional to ψx4 and ψx3 , and ψx is a two-component spinor. We see
the scalar field is a ‘composite’ of the fermion. However, it is only well-defined if Fx is nonzero,
which we will see indicates SUSY breaking.
38 3. Superspace and Superfields

• In the absence of a mass term, the chiral superfield corresponds to the particles of an N = 1
chiral multiplet. With a mass, it corresponds to the particles of an N = 1 massive multiplet
with superspin y = 0.

Next, we turn to the vector superfield.

• The most general vector superfield has the form

i i
V (x, θ, θ) = C + iθχ − iθχ + θθ(M + iN ) − θθ(M − iN ) + θσ µ θVµ
 2   2   
i i 1 1
+ iθθ θ −iλ + σ µ ∂µ χ − iθθ θ iλ − σ µ ∂µ χ + (θθ)(θθ) D − ∂µ ∂ µ C
2 2 2 2

where we have shifted some fields with respect to their definitions in the general superfield for
convenience. Off shell, there are eight bosonic components, as C, M , N , D, and V µ are all real,
and eight fermionic components, from the complex χ and λ.

• Just as in the non-supersymmetric case, we want to impose a gauge symmetry to get rid of the
unwanted degrees of freedom in the vector. If Λ is a chiral superfield, then i(Λ − Λ† ) is a vector
superfield with components
√ 1
C = i(φ − φ† ), χ= 2 ψ, (M + iN ) = F, Vµ = −∂µ (φ + φ† ), λ = D = 0.
2
We may define a generalized gauge transformation on vector superfields by
i
V → V − (Λ − Λ† ).
2
This generalizes the ordinary notion of a gauge transformation as it acts on Vµ by

Vµ → Vµ + ∂µ Re(φ) ≡ Vµ − ∂µ α.

• We may use this gauge freedom to remove some of the vector superfield components. In
Wess–Zumino gauge, which we use exclusively, we set C = χ = M = N = 0, giving

1
VWZ (x, θ, θ) = (θσ µ θ)Vµ + (θθ)(θλ) + (θθ)(θλ) + (θθ)(θθ)D.
2

This leaves only the usual gauge freedom of Vµ . We hence have a vector field which yields gauge
bosons, spinor fields which yield their superpartners, and another auxiliary field D.

• Wess–Zumino gauge is not SUSY invariant; if we perform a SUSY transformation we must

perform a further gauge transformation to return to Wess–Zumino gauge. This is like how some
QED gauges are not Lorentz invariant.

• Powers of VWZ are given by

2 1
VWZ = (θθ)(θθ)V µ Vµ
2
with all higher powers equal to zero.
39 3. Superspace and Superfields

• Given the gauge symmetry above, the vector superfield corresponds to the massless particles
of an N = 1 vector multiplet. In the supersymmetric analogue of the Higgs effect, a chiral
superfield couples to the vector superfield, and in terms of the particles, the vector multiplet
‘eats’ the chiral multiplet to become a massive y = 1/2 multiplet.

• Again, the number of degrees of freedom balance off-shell, since the gauge field Vµ has 3 bosonic
degrees of freedom and the real scalar D has 1, so an auxiliary field is again required.

Next, we introduce the abelian field strength superfield.

• Recall that a complex scalar field φ and a U (1) gauge field Vµ have the gauge symmetry

φ → eiqα φ, Vµ → Vµ + ∂µ α

where α is a real-valued field that specifies the gauge transformation. Starting with a free φ
field, we may minimally couple it to the gauge field by a covariant derivative,

Dµ φ = ∂µ φ − iqAµ φ, L ⊃ Dµ φ(Dµ φ)∗ .

The kinetic term for the gauge field is written using the gauge invariant field strength
1
Fµν = ∂µ Vν − ∂ν Vµ , L ⊃ Fµν F µν .
4

• Similarly, in supersymmetry, we let a chiral and vector superfield have the gauge symmetry
i
Φ → eiqΛ Φ, V → V − (Λ − Λ† )
2
where Λ(x) is a chiral superfield that specifies the gauge transformation; this is necessary to
ensure that Φ remains a chiral superfield upon gauge transforming it. Note the term

Φ† exp(2qV )Φ

is gauge invariant and can serve as an interaction term.

• For a generic superfield, the spinor-valued superfield

1
Wα ≡ − (DD)Dα S
4
is a chiral superfield, because applying D will give three powers of D. The three indices on the
factors of D must be totally antisymmetric, since the D anticommute with each other, but this
implies they are identically zero.

• In the case where S is the vector superfield V , the result

1
Wα = − (DD)Dα V
4
is also invariant under generalized gauge transformations; it is called the field strength superfield.
We can similarly define an anti-chiral field strength superfield W α̇ .
40 3. Superspace and Superfields

• To get an explicit expression, it’s most useful to work in terms of y rather than x, giving

β̇
Wα (y, θ) = λα + θα D + (σ µν θ)α Fµν − i(θθ)(σ µ )αβ̇ ∂µ λ .

Since W is invariant, the fields λ, D, and Fµν are all separately gauge invariant. Note that
since we have a spinor-valued superfield, the coefficient of θβ is a bispinor, which naturally
decomposes into a scalar (the field D) and a self-dual two-form (the field Fµν ).

• We have the identity

α̇
Dα Wα = Dα̇ W
which contains the Bianchi identity ∂[µ Fνρ] = 0 as a subset.

• This result can also be generalized to non-abelian gauge fields, in which case we pick up extra
terms from the structure constants, and our gauge invariant fields become gauge covariant.
41 4. Supersymmetric Lagrangians

4 Supersymmetric Lagrangians
4.1 N = 1 Supersymmetry
Now we write supersymmetric Lagrangians, beginning with the simplest case of a chiral superfield.

• A theory with Lagrangian L(Φ) is supersymmetric if the variation δL under supersymmetry

transformations is a total derivative. Now recall that for a general scalar superfield S,
i
S ⊃ (θθ)(θθ)D, δD = ∂µ (ϵσ µ λ − ρσ µ ϵ)
2
and for a general chiral superfield Φ,
√
Φ ⊃ (θθ)F, δF = i 2 ϵσ µ ∂µ ψ.

Therefore, the Lagrangian is supersymmetric if it is built from the D terms of superfields and
the F terms of chiral superfields. This is not surprising, since the terms in Qα and Qα̇ that
multiply by Grassmann numbers come with factors of ∂µ .

• Note that for the Lagrangian to be real, the D terms come from real superfields and the F
terms of chiral superfields are paired with the F terms of antichiral superfields. In particular,
for a theory with a single chiral superfield Φ, the most general possibility is

L = K(Φ, Φ† )|D + (W (Φ)|F + h.c.) .

Here the Kahler potential K is a real function of Φ and Φ† , the conjugate Φ† is an anti-chiral
superfield, and the superpotential W is a holomorphic function of Φ, and hence a chiral superfield.
Here K|D just means the coefficient of (θθ)(θθ) in K.

• The action may thus be written as a superspace integral

Z Z Z Z 
4 4 4 2
S = d x d θK + d x d θ W + h.c. .

Note that not every term is integrated over all of superspace. This is perfectly acceptable and
also occurs in string theory, where objects may be confined to branes.

• Next, we perform dimensional analysis to determine renormalizability. Ultimately, we just have

a complicated collection of scalar and fermion fields, which must have the usual dimensions,
3
[φ] = 1, [ψ] = .
2
On the other hand, the chiral superfield is
√ 1 1
Φ=φ+ 2θψ + (θθ)F, [Φ] = 1, [θ] = − , [dθ] = , [F ] = 2.
2 2
Here F does not have the usual dimensions of a scalar field, because it is an auxiliary field.
42 4. Supersymmetric Lagrangians

• The Kahler potential and superpotential take the form

K ⊃ (θθ)(θθ)KD , W ⊃ (θθ)WF

and renormalizability requires the operators in KD and WF to have dimensions at most 4, so

[K] ≤ 2, [W ] ≤ 3.

Therefore, the Kahler potential is at most quadratic and the superpotential is at most cubic.
However, the terms Φ + Φ† and ΦΦ + h.c. do not have D terms, so the most general Kahler
potential is K = Φ† Φ, where we rescaled to remove the coefficient.

• For one chiral superfield, the general Lagrangian is known as the Wess–Zumino model,

LWZ = Φ† Φ|D + (W (Φ)|F + h.c.) .

Evaluating the Kahler potential term is straightforward and yields the kinetic terms along with
F F ∗ . The superpotential provides interactions; to evaluate it, we perform a Taylor expansion
in the Grassmann variables,

∂W 1 ∂2W ∂W ∂W
W (Φ) = W (φ) + (Φ − φ) + (Φ − φ)2 , ≡
∂φ 2 ∂φ2 ∂φ ∂Φ Φ=φ

where the linear term contributes θθF and the quadratic term contributes (θψ)(θψ). Then

1 ∂2W
   
∂W
LWZ = ∂ µ φ∗ ∂µ φ − iψσ µ ∂µ ψ + F F ∗ + F + h.c. − ψψ + h.c. .
∂φ 2 ∂φ2

Historically this was the first nontrivial four-dimensional supersymmetric model, and it was
originally written without the benefit of superspace and superfields.

• The portion of the Lagrangian that depends on F is

∂W ∂W ∗ ∗
LF = F F ∗ + F+ F .
∂φ ∂φ∗
There is no kinetic term, confirming F is an auxiliary field. Since it is quadratic, it is straight-
forward to eliminate F . Setting δSF /δF = 0, we find

∂W ∂W ∗
F∗ + = 0, F+ = 0.
∂φ ∂φ∗
Substituting this back in,
2
∂W
LF = − ≡ −VF (φ).
∂φ
That is, these terms simply yield a positive semi-definite scalar potential for φ.

• We can eliminate the auxiliary field F from the Wess–Zumino model by plugging in its equations
of motion. The cost of doing this is that the Lagrangian becomes SUSY invariant only on-shell;
thus we prefer to keep it explicit if we’re not doing practical calculations.
43 4. Supersymmetric Lagrangians

• We may regard our N = 1 Lagrangian as a special N = 0 Lagrangian. Field redefinitions can

be used to remove the linear term in W , and the constant term doesn’t matter, so
m 2 g 3
W = Φ + Φ .
2 3
We note that both the complex scalar φ and spinor ψ have mass m, and the Yukawa coupling
is the same as the scalar self-coupling, L ⊃ g(φψψ) + g 2 |φ|4 . As shown earlier, this ensures
that divergences cancel in perturbation theory.
†
• We now generalize to multiple chiral superfields. We have a Kahler potential K(Φi , Φj ) and
superpotential W (Φi ). Expanding about Φi = φi ,

∂2K
Kij = ≡ ∂i ∂j K
∂φi ∂φj∗
where a bar denotes a conjugated field. Here, Kij may be regarded as a metric in a space with
coordinates φi which is a complex manifold, specifically a Kahler manifold because the metric
can be derived by differentiating a potential. For a renormalizable theory this is rather trivial,
as Kij is constant, but if we allow the Kahler potential to be arbitrary (e.g. viewing the theory
as an effective theory) we can get more complicated manifolds.

• The kinetic terms become

 
µ j∗ i j µ i i j∗
L ⊃ Kij ∂ φ ∂µ φ − iψ σ ∂µ ψ + F F + ...

where a general Kahler potential produces extra fermion interaction terms. We say the Kahler
potential is canonical if Kij is diagonal and constant, which can be achieved by a U (n) field
redefinition in the renormalizable case.

• Restricting to renormalizable theories, the superpotential is expanded as before, resulting in

LF = Kij F i F j∗ + (∂i W )F i + (∂i W ∗ )F i∗ .

Varying with respect to F , the auxiliary field is

Kij F j∗ + ∂i W = 0, Kij F i + ∂j W ∗ = 0.

Plugging these results in gives a contribution to the scalar potential of

VF = Kij F i F j∗ = K ij ∂i W ∂j W ∗

where the Kahler metric with raised indices is the inverse of the original Kahler metric.

Note. Keeping track of the R-symmetry. The U (1) R-symmetry in such theories takes the form

Φi → eiri α Φi
i
where ri = R[Φi ] is the U (1)R charge of the field Φi , and R[Φ ] = −ri . By definition, we have

R[Qα ] = −1, R[Qα̇ ] = 1

44 4. Supersymmetric Lagrangians

which implies that the superspace coordinates are charged as well,

R[θ] = 1, R[θ] = −1, R[dθ] = −1, R[dθ] = 1.

This implies that the components of a chiral superfield have different R-charges,

R[φi ] = ri , R[ψ i ] = ri − 1, R[F i ] = ri − 2.

The Kahler potential is invariant under U (1)R provided that it does not mix fields of different
R-charges. The superpotential is only invariant if R[W ] = 2, and hence it generically breaks the
R-symmetry. We also note that the R-symmetry may be ambiguous, as we may combine it with
flavor symmetries which rotate the Φi individually.

Next, we turn to the vector superfield Lagrangian, i.e. the theory of “super QED”.

• We can deduce QED by demanding a local U (1) symmetry for a complex scalar field φ,
parametrized by a scalar field α. To get super QED we replace the scalar fields with chi-
ral superfields, with the gauge transformation

Φ → exp(iqΛ)Φ.

As in QED, the naive kinetic term K = Φ† Φ is no longer gauge invariant, but

i
K = Φ† exp(2qV )Φ, V → V − (Λ − Λ† )
2
is gauge invariant, as shown above. The kinetic term for the vector superfield/gauge field V is

Lkin = f (Φ)(W α Wα )|F + h.c.

where renormalizability requires the ‘gauge kinetic function’ f to be a constant, f = τ .

• A new feature of super QED is a new gauge-invariant term, the Fayet–Iliopoulos (FI) term
1
LFI = ξV |D = ξD
2
where ξ is a constant. The FI term only appears for an abelian gauge theory, because the gauge
field is not charged under U (1). More generally the gauge field would be charged, which would
make D charged and the FI term not gauge invariant.

• Therefore, the renormalizable Lagrangian of super QED is

  
† 1 α
L = (Φ exp(2qV )Φ) D + W (Φ) + W Wα + h.c. + ξV D
4 F

where we have set τ = 1/4 to get a canonically normalized photon field.

• Note that for the superpotential W (Φ) to be gauge invariant, it must contain terms like Φ1 · · · Φn
where the charges of the terms in the product add to zero. If there is only one charged Φ field,
the superpotential must vanish.
45 4. Supersymmetric Lagrangians

• Next, we write out the first term explicitly. Taking Wess–Zumino gauge and Taylor expanding,

exp(2qV ) = 1 + 2qV + 2q 2 V 2

where all higher terms vanish. We thus find

(Φ† exp(2qV )Φ) D

= F ∗ F + ∂µ φ∂ µ φ∗ − iψσ µ ∂µ ψ + qV µ (−ψσ µ ψ + iφ∗ ∂µ φ − iφ∂µ φ∗ )
√
+ 2q(φλψ + φ∗ λψ) + q 2 (D + qVµ V µ )|φ|2 .

We could make this manifestly gauge invariant by grouping terms into covariant derivatives.

• Next, we consider the gauge kinetic term. We have

1 i
W α Wα |F = D2 − Fµν F µν − 2iλσ µ ∂µ λ − Fµν F̃ µν , F̃µν ≡ ϵµνρσ F ρσ
2 4
where, to get the Fµν terms, we used the identity
1
tr σ µν σ κτ = (η µκ η ντ − η µτ η νκ + iϵµνκτ ).
2
• Then the gauge kinetic term is
1 α 1 1
W Wα + h.c. = D2 − Fµν F µν − iλσ µ ∂µ λ.
4 F 2 4

More generally, we could take τ complex, yielding an F F̃ term. However, this term is a total
derivative, so it makes no difference perturbatively.

• Collecting the terms involving D, we have

1 1
LD = qD|φ|2 + D2 + ξD
2 2
so D is an auxiliary field like F . Setting δSD /δD = 0 gives
ξ
D = − − q|φ|2 .
2
Substituting this back in gives
 2
1 1 ξ
LD = − D2 = − + q|φ|2 ≡ −VD (φ)
2 2 2
so we again get a contribution to the scalar potential. The total scalar potential is

∂W 2 1 ξ
 2
2
V (φ) = VF (φ) + VD (φ) = + + q|φ| ≥ 0.
∂φ 2 2
We see that the FI term could be responsible for spontaneous symmetry breaking.

• Finally, the general action can be written as a superspace integral as

Z Z Z Z 
4 4 4 2 α
S[K, W, f, ξ] = d x d θ (K + ξV ) + d x d θ (W + f W Wα ) + h.c. .

We will not consider the non-abelian case, where there are many complications.
46 4. Supersymmetric Lagrangians

• The results easily generalize for more fields. Assuming the gauge group remains U (1), we don’t
get more vector superfields, but we can have multiple chiral superfields Φi . We find
1 1 ξ
LD = qDKij φi φj∗ + D2 + ξD, D = − − qKij φi φj∗
2 2 2
so the total scalar potential is
 2
1 1 ξ
i i
V (φ ) = Kij F F j∗
+ D2 = K ij ∂i W ∂j W ∗ + + qKij φi φj∗ .
2 2 2

• To get the supersymmetric analogue of ordinary QED, with a single Dirac fermion, we require
two chiral superfields Φ+ and Φ− with opposite charges, corresponding to the electron and
positron. The Dirac mass term comes from the superpotential W = mΦ+ Φ− . A single massive
chiral superfield instead corresponds to a Majorana fermion, which cannot be charged.

4.2 The Wess–Zumino Model

In this section, we consider the renormalization of the Wess–Zumino model in detail.

• We take the superpotential to have the standard form given above. To establish notation, the
Lagrangian has kinetic terms
m
L0 = −∂µ ϕ∂ µ ϕ − m2 ϕϕ − iψσ µ ∂µ ψ − (ψψ + ψψ)
2
where we are using (− + ++) signature, so the propagators are: where the momentum p flows
to the right.

• The interaction terms are

2 2
Lint = −λ2 ϕ ϕ2 − mλ(ϕ ϕ + ϕϕ2 ) − λ(ψψϕ + ψψϕ).

For the scalar, this leads to cubic and quartic vertices of −imλ and −iλ2 respectively, as well
as the Yukawa interactions:

• We could have also treated the fermion mass as an interaction term, in which case only ⟨ψψ⟩
and ⟨ψψ⟩ would be nonzero, and we would have an interaction term −im which reversed the
chirality.

• At one loop, the total scalar tadpole vanishes, as the two diagrams cancel: In fact, this cancel-
lation holds at all orders in perturbation theory, and it implies that vevs are not renormalized.

• Next, consider the scalar self-energy at zero external momentum. This directly determines the
coefficient of the mass term ϕϕ in the effective action, and three diagrams contribute: The total
contribution is
2
tr(−σ µ σ ν )qµ qν

−i −i
Z Z Z
2 21 2
(−imλ) d̄q − (−iλ) d̄q + (−iλ) d̄q 2 = 0.
q 2 + m2 2 (q 2 + m2 )2 q + m2

Hence the mass term isn’t renormalized either, and this again holds to all orders in perturbation
theory. One can check the same holds for all of the vertices.
47 4. Supersymmetric Lagrangians

• We could also formulate the perturbation theory with the auxiliary field,
1 1
L0 = −∂µ ϕ∂ µ ϕ − iψσ µ ∂µ ψ + F F + m(F ϕ + F ϕ − ψψ − ψψ)
2 2
which has a simpler set of interactions,
2
Lint = λ(F ϕ2 + F ϕ − ψψϕ − ψψϕ).

• To compute the scalar propagators, write the kinetic terms as (ϕ, F )M (ϕ, F )T and invert M ,
to find
−i ip2 im
⟨ϕϕ⟩ = 2 2
, ⟨F F ⟩ = 2 2
, ⟨ϕF ⟩ = ⟨ϕF ⟩ = 2 .
p +m p +m p + m2
In particular, there is a nontrivial, though unusual propagator for F .

• Upon computation of the effective action, we find no effects at one loop except for the same
field strength renormalization factor

ZΦ ≡ Zϕ = Zψ = ZF .

This continues at all orders in perturbation theory. The only effect is the anomalous dimension
γ = d log ZΦ /d log µ. If we account for renormalization effects by working in terms of the
1/2
renormalized fields ΦR = ZΦ Φ, and γ is given, then we can compute the RG flow of the
renormalized couplings exactly.

4.3 Non-Renormalization Theorems

We have seen that theories of chiral and vector superfields are determined by the functions K, W ,
f , and the parameter ξ. We now investigate how they are renormalized in general.

• In 1977, Grisaru, Siegel, and Rocek showed using “supergraphs” that, except for one-loop
corrections to f , quantum corrections only come in the form
Z Z
d x d4 θ . . . .
4

Then W and ξ are not renormalized in perturbation theory at all, while K is.

• In 1993, Seiberg used symmetry and holomorphicity arguments to establish this result nonper-
turbatively in a simple and elegant way; we will follow this proof here. The intuition is that
since the superpotential is holomorphic, it is determined by its singularities and its asymptotics;
this allows us to pin it down in general using its weak coupling limit.

• In order to keep track of symmetries, we introduce spurion superfields

X = (x, ψx , Fx ), Y = (y, ψy , Fy )

so the action becomes

Z Z Z Z 
4 4 4 2 α
S = d x d θ (K + ξV ) + d x d θ (Y W + XW Wα ) + h.c. .

Here we note that the integrand of the d2 θ integral is holomorphic. These spurion fields have
no dynamics; they are just a way to rewrite numerical coupling constants in the action to make
symmetries manifest. We might think of them as being very heavy fields with fixed vevs.
48 4. Supersymmetric Lagrangians

• Specifically, the action has a U (1)R symmetry with charges

Φi V X Y θ θ Wα
0 0 0 2 1 −1 1

where we note that if θ → eiα θ, then dθ → e−iα dθ because dθ θ = 1, and the charge for Wα
R

can be found from its definition in terms of covariant derivatives, using ∂θ → e−iα ∂θ .

• The action also has a shift symmetry,

X → X + ir, r∈R

because the contribution of X to the action is

XW α Wα ⊃ Re(x)Fµν F µν + Im(x)Fµν Feµν

and hence the shift contributes a total derivative, which does not affect perturbation theory.
This symmetry is also called a Peccei–Quinn symmetry, making X an ‘axion-like field’.

• Now consider the Wilsonian action SΛ attained by integrating out all degrees of freedom above
Λ. We must have
Z Z
SΛ = d4 x d4 θ J(Φ, eV , X, Y, D, . . .) + ξ(X, Y )V

Z Z
+ d4 x d2 θ H(Φ, X, Y, W α ) + h.c.

where H is a holomorphic function, and J and ξ are not.

• By U (1)R invariance we must have

H = Y h(X, Φ) + g(X, Φ)W α Wα .

Moreover, we must still have invariance under shifts in X. Hence the only term involving X
that is allowed takes the form XW α Wα , so

H = Y h(Φ) + (αX + g(Φ))W α Wα .

Now, in the limit Y → 0, we must have h(Φ) = W (Φ), because any higher order corrections to
h(Φ) would be higher order in Y and hence negligible. Thus we must have h(Φ) = W (Φ) for
all Y , so the superpotential is not renormalized!

• We claim the gauge kinetic function is only renormalized at one loop. Since the gauge kinetic
term appears as XW α Wα , the gauge field propagator is proportional to 1/x and the three-point
gauge vertex is proportional to x. (don’t understand) Then the number of powers of x at
L loops is 1 − L, so αX is the tree-level contribution and g(Φ) is the one-loop contribution.
In practice, this means that one can compute divergent higher-loop corrections for the gauge
kinetic function, but they all miraculously cancel.
49 4. Supersymmetric Lagrangians

• We cannot constrain the Kahler potential nearly as much, because it is not holomorphic. How-
ever, the FI term must be a constant to maintain gauge invariance under

V → V + i(Λ − Λ† ).
P
Moreover, the contributions correcting ξ are proportional to i qi where the qi are the U (1)
charges. This vanishes if the gravitational anomaly vanishes, so ξ is not renormalized.

Note. The above derivation is a bit complicated; it’s easier to see the key ideas in the Wess–Zumino
model. Take the superpotential at scale µ0 to be
m 2 λ 3 m̃ λ
Wµ0 = Φ + Φ ≡ µ 0 + Φ3
2 3 2 3
where we defined the dimensionless coupling m̃. The free theory at W = 0 has a U (1) rotation
symmetry for Φ and a U (1)R symmetry, with charges

Φ : (1, 1), m̃ : (−2, 0), λ : (−3, −1).

Then the most general allowed for the effective superpotential at scale µ is
 
λΦ µ
Wµ = µm̃Φ2 f ,
µm̃ µ0
where f is holomorphic in its first argument. Since the limit λ → 0 is regular, we can expand it in
a Taylor series,
∞
X λn
Wµ = cn (µ/µ0 ) Φn+2 .
(µm̃)n−1
n=0
Now we also need a regular m̃ → 0 limit, which means terms with n > 1 are disallowed. Then

Wµ = c0 (µ/µ0 )µm̃Φ2 + c1 (µ/µ0 )λΦ3 .

But now we can take the weak coupling limit: we know that in the limit λ → 0 no nontrivial
renormalization occurs at all, so c0 = 1/2 and c1 = 1/3. Then
m̃(µ) 2 λ 3
Wµ = µ Φ + Φ
2 3
which shows that the superpotential is not renormalized.

4.4 Extended Supersymmetry

Next, we briefly look at extended supersymmetry. We will write the results in N = 1 language,
i.e. the actions will just be N = 1 actions with constraints.

• The simplest case in N = 2 is the vector multiplet, whose particles are created by a massless
chiral superfield Φ and massless vector superfield V , just as in super QED.

• Here, the N = 2 supersymmetry imposes the constraint W = 0, ensuring Φ is massless, and

∂2F
 
† 1 † ∂F
f (Φ) = , K(Φ, Φ ) = Φ exp(2V ) − h.c.
∂Φ2 2i ∂Φ
where F(Φ) is a holomorphic function called the prepotential, where F(Φ) = Φ2 at tree level.
50 4. Supersymmetric Lagrangians

• It can be shown that F(Φ) only receives one-loop corrections in perturbation theory. It also
receives nonperturbative corrections which can be written in terms of an “instanton expansion”
2
P
k ak exp(−kc/g ) found by Seiberg and Witten in 1994.

• There are other combinations of fields that produce N = 2 multiplets, but they are much more
complicated.

• In N = 4, we consider the vector multiplet, which consists of an N = 2 vector multiplet and

an N = 2 hypermultiplet. Here there are no free functions at all, only a single free parameter
Θ 4πi
f =τ = + 2 .
2π g
The theory is finite, i.e. has no UV divergences, and the beta function vanishes, yielding
conformal invariance. This is the theory of N = 4 super Yang–Mills.

• The AdS/CFT correspondence relates a gravitational theory in AdS space to a conformal field
theory without gravity in one fewer dimension. The prime example of this correspondence is
between AdS in five dimensions and N = 4 super Yang–Mills in four dimensions.
51 5. The MSSM

5 The MSSM
5.1 SUSY Breaking
We now review the basics of supersymmetry breaking, since it must occur in our universe.

• Classically, suppose that fields transform under a symmetry as

φi → exp(iαa T a )ij φj , δφi = iαa (T a )ij φj .

The symmetry is said to be broken if the vacuum is not invariant,

αa (T a )ij (φvac )j ̸= 0.

• For example, for a complex field φ = ρeiθ with a U (1) internal symmetry,

δφ = iαφ, δρ = 0, δθ = α.

When the vacuum satisfies ⟨ρ⟩ =

̸ 0, the symmetry is broken, and θ is a Goldstone boson.

• Using the SUSY commutation relations, note that

˙
(σ ν )βα {Qα , Qβ̇ } = 2(σ ν )β̇α (σ µ )αβ̇ Pµ = 4η µν Pµ = 4P ν .

In particular, taking the ν = 0 component, we have

2
˙ X
0 βα
(σ ) {Qα , Qβ̇ } = (Qα Q†α + Q†α Qα ) = 4P 0 = 4E.
α=1

Since the left-hand side is positive definite, E ≥ 0 for any state.

• In particular, consider the vacuum state |Ω⟩. If the vacuum is SUSY invariant, Qα |Ω⟩ = 0, then
the left-hand side vanishes and E = 0. If SUSY is broken, then the vacuum energy is positive.

• This might seem a bit strange, because we are used to the vacuum energy being indefinite in
quantum field theory; for example, we change it by normal ordering the Hamiltonian. The
difference is that the SUSY commutation relations involve the Hamiltonian itself, H ∼ |Q|2 .
Hence requiring a SUSY algebra at the quantum level constrains the operator ordering.

• As before, SUSY breaking is caused by fields acquiring vevs that are not SUSY invariant. Recall
that in the case of a chiral superfield Φ,
√ √ √ √
δφ = 2ϵψ, δψ = 2ϵF + i 2σ µ ϵ∂µ φ, δF = i 2ϵσ µ ∂µ ψ.

The field ψ cannot have a vev, as this would violate Lorentz invariance. Similarly, we must
have ⟨∂µ φ⟩ = 0. Then the transformation reduces to
√
δ⟨φ⟩ = δ⟨F ⟩ = 0, δ⟨ψ⟩ = 2ϵ⟨F ⟩.

Hence, SUSY is broken if and only if ⟨F ⟩ ≠ 0, and ψ is said to be a Goldstone fermion, or

goldstino; note that it is not the SUSY partner of a Goldstone boson. There is no Goldstone
boson, as SUSY, being a fermionic symmetry, breaks an assumption of Goldstone’s theorem.
52 5. The MSSM

• Recalling that the contribution to the scalar potential is

VF = Kij F i F j∗

we see that SUSY is broken if ⟨VF ⟩ > 0, corresponding to a positive vacuum energy at the
classical level, just as at the quantum level. Equivalently, a SUSY-invariant vacuum is one
where ∂W/∂φ = 0.

Note. Reformulating the above in mathematical language. As we’ve seen above, the scalar fields
are a map from spacetime to a Kahler manifold; we can think of the φi as local coordinates on
this manifold. The vacuum equations ∂W/∂φi = 0 are a set of holomorphic algebraic equations
and hence define a subvariety of this manifold. If the subvariety is trivial, SUSY is necessarily
spontaneously broken. If there are a continuum of solutions, we say there is a vacuum moduli space.

Example. The Wess–Zumino model. We take the canonical Kahler potential and superpotential
m 2 λ 3
W = Φ + Φ .
2 3
The condition for a vacuum solution is
∂W
= mφ + λφ2 = 0
∂φ
which gives the two discrete supersymmetric vacua
m
φ = 0, φ=− .
λ
Example. Consider a theory with three chiral superfields, the canonical Kahler potential, and

W = Φ1 Φ2 Φ3 .

The vacuum equations are ϕ1 ϕ2 = ϕ1 ϕ3 = ϕ2 ϕ3 = 0, which defines a subvariety; note that the
moduli space is not a manifold in this case.

Example. The O’Raifertaigh model. We consider three chiral superfields Φ1 , Φ2 , Φ3 with

K = Φ†i Φi , W = gΦ1 (Φ23 − m2 ) + M Φ2 Φ3 , M ≫ m.

Using the equations of motion for the auxiliary field F ,

∂W ∂W ∂W
−F 1∗ = = g(φ23 − m2 ), −F 2∗ = = M φ3 , −F 3∗ = = 2gφ1 φ3 + M φ2 .
∂φ1 ∂φ2 ∂φ3
Since we cannot have Fi∗ = 0 for all i simultaneously, this superpotential necessarily breaks SUSY.
The scalar potential is

VF (φi ) = F 1∗ F 1 + F 2∗ F 2 + F 3∗ F 3 = g 2 |φ23 − m2 |2 + M 2 |φ3 |2 + |2gφ1 φ3 + M φ2 |2 .

Hence, since M is large, the potential is minimized at

⟨φ2 ⟩ = ⟨φ3 ⟩ = 0, ⟨φ1 ⟩ arbitrary, ⟨V ⟩ = g 2 m4 > 0.

53 5. The MSSM

The only nonzero F field is thus ⟨F1 ⟩ ̸= 0. For simplicity, we take ⟨φ1 ⟩ = 0. The fermion mass
terms are then  
 2  0 0 0
∂ W
ψ i ψ j = 0 0 M  ψ i ψ j .
∂φi ∂φj
0 M 0
Hence we have two fermions with mass M and one massless fermion ψ1 , which is the Goldstino
corresponding to the nonzero vev of F1 . The quadratic terms in the scalar potential expanded about
the vev are
VF (φi ) ⊃ −m2 g 2 (φ23 + φ∗3 2 ) + M 2 |φ3 |2 + M 2 |φ2 |2 .
Hence the φ1 field is massless, since it corresponds to a flat direction in the scalar potential, and
the φ2 has mass M . For the φ3 , expand φ3 = a + bi to find

m2a = M 2 − 2g 2 m2 , m2b = M 2 + 2g 2 m2 .

We define the supertrace as the trace with an extra minus sign for bosons,
X
STr(M 2 ) ≡ (−1)2j+1 (2j + 1)m2j = 0.
j

This result is generic for tree-level SUSY breaking.

Note. We can show that the supertrace vanishes at tree level for arbitrarily many chiral superfields.
First note that the fermion mass matrix is

(MF )ij = ⟨∂ i ∂ j W ⟩, tr MF† MF = ⟨∂ i ∂ j W ⟩Kii Kjj ⟨∂ i ∂ j W ∗ ⟩.

Now, the scalar potential is

V = Kij ⟨∂ i W ⟩⟨∂ j W ∗ ⟩
which means the scalar mass terms take the form
 
1 1 1
L ⊃ − φ∗j φi ∂ i ∂ j V + φi φj ∂ i ∂ j V + φ∗i φ∗j ∂ i ∂ j V = − φ† MB2 φ
2 2 2

where we consider φi and φ∗i as independent real fields, and

   
φ ∂∂V ∂∂V
φ= , MB2 = .
φ∗ ∂∂V ∂∂V

Hence we have
 
tr MB2 = 2∂∂V = 2Kii ∂ i ∂ i Kjj ⟨∂ j W ⟩⟨∂ j W ∗ ⟩ = 2⟨∂ i ∂ j W ⟩Kii Kjj ⟨∂ i ∂ j W ∗ ⟩ = 2 tr MF† MF .

Since each fermionic field contains two degrees of freedom, Str(M 2 ) = 0 as desired.

Note. We have shown that W is not renormalized to all orders in perturbation theory; hence if
SUSY is unbroken at tree level, it is unbroken in perturbation theory. Moreover, if SUSY is broken
at tree level, the supertrace of M 2 vanishes, implying that the superpartners cannot be too much
heavier. Since this appears to be experimentally ruled out, SUSY must be broken nonperturbatively.
54 5. The MSSM

Example. For a vector superfield V = (λ, Aµ , D) in Wess–Zumino gauge, we must have ⟨λ⟩ =
⟨Aµ ⟩ = 0 by Lorentz invariance. However, D can acquire a vev, and since

δλ ∝ ϵD

we see that SUSY can be broken when D acquires a vev, where λ is the Goldstino. This is called
D-term SUSY breaking, in contrast with F -term SUSY breaking above. Since the contribution to
the scalar potential is proportional to ⟨D⟩2 , SUSY is broken if ⟨VD ⟩ > 0.
In the very simplest case, we consider a single chiral superfield with U (1) charge q and trivial
superpotential, where q > 0 and ξ ≥ 0. The scalar potential is
 2
1 ξ 2 ξ
V (φ) = + q|φ| , ⟨φ⟩ = 0, ⟨D⟩ = −
2 2 2
which means that SUSY is broken when ξ > 0. Since ⟨φ⟩ = 0, the U (1) symmetry is unbroken, so
the λ and Vµ remain massless. Since the superpotential is trivial, the ψ remains massless. Finally,
using the scalar potential, we see
ξ
V (φ) ⊃ q|φ|2 , m2φ = qξ/2.
2
On the other hand, if q > 0 and ξ < 0, then we have
ξ
|⟨φ⟩|2 = − , ⟨D⟩ = 0
2q
which indicates that SUSY is not broken, but the U (1) symmetry is. Then the λ and Vµ fields
acquire mass by the ordinary Higgs effect by interacting with the vev of φ.
Note. In the case of D-term breaking, the supertrace sum rule is slightly modified; it turns out
to be proportional to the sum of all U (1) charges. However, this quantity must vanish to ensure
anomaly cancellation.
Finally, we briefly discuss SUSY breaking in supergravity.

• In supergravity, there is a new auxiliary field Fg , which can break SUSY by acquiring a vev.
Specifically, the F -term is

F ∝ DW, Di W ≡ ∂i W + (∂i K)W

where we have set Mpl = 1.

• The scalar potential has a negative gravitational contribution,

 
V = eK K ij Di W Dj W ∗ − 3|W |2 .

This is important because it allows ⟨V ⟩ = 0 even after SUSY breaking, which avoids an
unacceptably large cosmological constant. However, this does not solve the cosmological constant
problem, because ⟨V ⟩ is generically large and negative.

• There are “no-scale” supergravity models where the Kahler potential and superpotential are
chosen so that ⟨V ⟩ = 0, but these are not regarded as a solution of the cosmological problem
either, because there is no reason the form of the potentials should be preserved by quantum
corrections.
55 5. The MSSM

• In the process of SUSY breaking, the gravitino field, which is the gauge field of N = 1 su-
pergravity, “eats” the goldstino and gains mass. This is called the super Higgs effect, and
should not be confused with the supersymmetric extension of the ordinary Higgs effect, where
a massless vector superfield eats a chiral superfield to gain mass.

5.2 Particles and Interactions

Next, we discuss the matter content of the MSSM.

• The MSSM has N = 1 SUSY with gauge group SU (3)C × SU (2)L × U (1)Y . The matter fields
are the same as in the SM, with spinor fields promoted to chiral superfields. Note that some
conjugations are necessary, since chiral superfields only contain left-chiral spinors.

• Specifically, we have quarks and squarks,

Qi = (3, 2, 1/6), uci = (3, 1, −2/3), dci = (3, 1, −1/3)

including, e.g., the stop squark, as well as leptons and sleptons,

Li = (1, 2, 1/2), eci = (1, 1, 1),

including, e.g. the selectron sneutrino.

• The Higgs acquires a superpartner, the Higgsino. Since the Higgsino contributes to the U (1)Y
anomaly, a second Higgs field with opposite hypercharge is required to cancel it. We have

H1 = (1, 2, −1/2), H2 = (1, 2, 1/2)

where the H1 field is not present in the SM. Hence the MSSM is a two Higgs doublet model.
Another reason a second Higgs is required is that in the SM, we must use the Higgs conjugate
field for some of the Yukawa terms, but we can’t do that here since the superpotential is
holomorphic.

• The gauge bosons correspond to vector superfields, giving gluons and gluinos, W bosons and
winos, and B bosons and binos,

G = (8, 1, 0), W = (1, 3, 0), B = (1, 1, 0).

The neutral winos, binos, and Higgsinos mix to form Majorana fermions called neutralinos, the
lightest of which could serve as a dark matter candidate; typically this candidate is mostly bino.
The charged winos, binos, and Higgsinos form charginos.

We now consider the interactions in the MSSM.

• As in super QED, we have interactions by the chiral superfield kinetic term. However, the FI
term must be zero, as otherwise the scalar potential for squarks and sleptons would yield a
vacuum breaking U (1)A and SU (3)C symmetry.

• Rescaling the gauge fields, the gauge kinetic terms fa = τa have Re τa = 4π/ga2 , specifying the
gauge couplings.
56 5. The MSSM

• The most general renormalizable superpotential is given by

W = y1 QH2 uc + y2 QH1 dc + y3 LH1 ec + µH1 H2 + λ1 LLec + λ2 LQdc + λ3 uc dc dc + µ′ LH2

where we have suppressed generation indices; properly every coefficient is a matrix in generation
space. The first three terms yield standard Higgs Yukawa couplings to matter, while the fourth
is a mass term for the two Higgs fields.

• The last four terms break either U (1)B or U (1)L . They are not allowed phenomenologically,
because if the parameter values were natural, then protons would decay in seconds.

• The simplest way to forbid these terms is to impose R-parity,

(
3(B−L)+2s +1 all observed particles,
R ≡ (−1) =
−1 superpartners

where s is the spin. This has the additional benefit that the lightest superpartner (LSP) is
stable, and hence can serve as a candidate for cold weakly interacting dark matter. In collider
experiments, one can search for LSP pair production by ‘missing energy’.

• Note that it would have been completely equivalent to define R to be (−1)3(B−L) , because all
interaction terms are Lorentz scalars, so the spins of the fields involve must sum to an integer.
Our definition of R is just slightly nicer. R-parity can be realized by having a U (1)B−L gauge
symmetry spontaneously broken at high energies in an appropriate way.

Note. The imposition of R-parity is a useful general model building tool, outside of SUSY. In a
generic extension of the Standard Model with an R-parity-like symmetry (which we call T -parity),
every vertex has an even number of new particles, and these must be connected up to yield a
contribution to a Standard Model operator, so the leading contributions are at loop level. This
gives one more room to avoid bounds, and reduces contributions to the Higgs mass. Theories with
T -parity also tend to have a dark matter candidate, namely the lightest T -odd particle (LTOP).

Next, we discuss mechanisms for SUSY breaking.

• As shown above, naive SUSY breaking wouldn’t work, because STr(M 2 ) vanishes, and the
superpartners would be too light. Instead, we introduce a hidden sector which breaks SUSY.
The hidden sector may obey the sum rule, but it isn’t ruled out because it doesn’t interact
directly with the MSSM fields; instead it interacts through a messenger sector. Typically, the
gauge group is enlarged by another factor G, under which all MSSM fields are singlets.

• One possible SUSY breaking mechanism is gaugino condensation. Here an asymptotically free
gauge coupling g becomes large at some energy scale M . If we start with a cutoff Λ, then

M = Λ exp(g −2 (Λ)/β)

so it is easily possible to have M ≪ Λ. Here SUSY is broken dynamically and nonperturbatively,

so the sum rule doesn’t apply. This is analogous to dimensional transmutation in QCD, which
explains why ΛQCD ≪ Mpl .
57 5. The MSSM

• Next, we need to specify the messenger sector. For example, the mediating field could simply
be the graviton. Then couplings are suppressed by Mpl , so by dimensional analysis

M2
∆m =
Mpl

where ∆m describes the size of the mass splittings in the MSSM. Setting ∆m ∼ 1 TeV and
Mpl ∼ 1018 GeV gives M ∼ 1011 GeV. This scenario requires a gravitino, which acquires a mass
∆m by the super Higgs mechanism.

• Another situation is gauge mediation. Here the messenger fields are charged under both G and
the SM gauge group, and the SUSY breaking is transmitted by loops. Then
M
∆m ∼
16π 2
which means M must also be around the TeV scale. Then the gravitino mass is on the order of
M 2 /Mpl ∼ eV so it is the LSP.

• To work phenomenologically, we integrate out the messenger sector and hidden sector to yield
a Lagrangian for the MSSM with SUSY breaking terms. Generically, we get all possible “soft
SUSY breaking terms”, i.e. renormalizable terms that do not reintroduce the hierarchy problem
(quadratic sensitivity to Λ2 ), such as mass terms for superpartners and additional interactions.
This is the source of the many (> 100) parameters in the MSSM.

• Almost all of the MSSM parameter space is ruled out, because SUSY particles could heavily
mix in general, and this mixing would be transferred to quarks by loops, causing flavor changing
neutral currents.

• Specific high scale models provide relations between the parameters. For example, the con-
strained MSSM (CMSSM), which may arise from string theory has only three free parameters.

• Extra structure is needed to account for neutrino masses. One may also add an additional
singlet Higgs (and its superpartner), which resolves some theoretical tensions. The result is the
next-to-minimal extension of the SM, the NMSSM.

Finally, we revisit the hierarchy problem.

• We may split the hierarchy problem into two parts: why Mew ≪ Mpl at tree level, and why this
is stable under quantum corrections. These are qualitatively distinct applications of naturalness,
and both are challenging.

• Note that SUSY introduces new scalar particles, but they don’t create new hierarchy problems
because they are superpartners of fermions, which are naturally light.

• As argued before, SUSY cancels the quadratic divergences in the Higgs self-energy. We retain
logarithmic divergences of the form ∆m log(M/∆m), which may naturally give a small result
as long as ∆m is around the TeV scale, motivating low-scale SUSY. An independent argument
for TeV scale SUSY is gauge coupling unification. A third argument comes from the WIMP
miracle, which is that TeV scale SUSY can account for dark matter by the LSP.
58 5. The MSSM

• A more sophisticated way to understand these cancellations comes from the non-renormalization
theorems above. The Higgs mass term comes from the superpotential, and we have shown it is
not renormalized.

• In principle, SUSY could also solve the cosmological constant problem. However, it is broken at
far too high a scale; we would have MΛ ∼ ∆m, while in reality MΛ ≪ ∆m. At present, there
is no satisfactory solution to this problem.

Note. Gauge coupling unification only makes sense in the context of grand unification. The
coupling constants for non-abelian gauge theories are normalized by normalizing the generators T a
so that, e.g. tr(T a T b ) = δ ab /2. However, there’s no canonical way to normalize the U (1) coupling;
by different choices of this normalization one can make gauge coupling unification happen for any
theory. The only way to resolve this ambiguity is to determine how U (1) is embedded in the GUT
group, which fixes its normalization.

Note. The technical naturalness of the smallness of the fermion masses can be seen nicely by
spurions. Let the mass parameter be m. Then the Lagrangian maintains chiral symmetry if m is
charged under it, which implies corrections to it must take the form δm = mf (|m|2 ). In general,
whenever a symmetry is restored when a parameter vanishes, it can be maintained for nonzero
values of that parameter by promoting it to a spurion charged under that symmetry.