Laços de Wilson em Teorias Super-Simétricas

Universidade de São Paulo
Instituto de Física
Laços de Wilson em teorias super-simétricas

de Chern-Simons com matéria
Marcia Rodrigues Tenser
Orientador: Prof. Dr. Diego Trancanelli
Tese de doutorado apresentada ao Instituto de Física
da Universidade de São Paulo, como requisito parcial
para a obtenção do título de Doutor em Ciências.
Banca Examinadora:
Prof. Dr. Diego Trancanelli - Orientador (IF-USP)
Prof. Dr. Victor Rivelles (IF-USP)
Profa. Dra. Silvia Penati (UNIMIB, Itália)
Prof. Dr. Diego Correa (IF-UNLP, Argentina)
Prof. Dr. Alberto Faraggi (Ciencias Exactas/UNAB, Chile)
São Paulo
2021
FICHA CATALOGRÁFICA
Preparada pelo Serviço de Biblioteca e Informação
do Instituto de Física da Universidade de São Paulo
Tenser, Marcia Rodrigues
Laços de Wilson em teorias super-simétricas de Chern-Simons com

matéria / Wilson Loops in supersymmetric Chern-Simons-matter theories.
São Paulo, 2021.
Tese (Doutorado) – Universidade de São Paulo. Instituto de Física.

Depto. de Física Matemática
Orientador: Prof. Dr. Diego Trancanelli

Área de Concentração: Física Matemática.
Unitermos: 1. Teoria de campos; 2. Teoria quântica de campo; 3.

Física teórica
USP/IF/SBI-055/2021
University of São Paulo
Physics Institute
Wilson Loops in supersymmetric

Chern-Simons-matter theories
Marcia Rodrigues Tenser
Supervisor: Prof. Dr. Diego Trancanelli
Thesis submitted to the Physics Institute of the Uni-
versity of São Paulo in partial fulllment of the require-
ments for the degree of Doctor of Science.
Examining committee:
Prof. Dr. Diego Trancanelli - Orientador (IF-USP)
Prof. Dr. Victor Rivelles (IF-USP)
Profa. Dra. Silvia Penati (UNIMIB, Itália)
Prof. Dr. Diego Correa (IF-UNLP, Argentina)
Prof. Dr. Alberto Faraggi (Ciencias Exactas/UNAB, Chile)
São Paulo
2021
À minha avó Cynthia
Agradecimentos
Antes de mais nada, gostaria de agradecer o Conselho Nacional de Desenvolvi-
mento Cientíco e Tecnológico (CNPq) pelo apoio nanceiro.
Dentre os inúmeros aprendizados colecionados durante o desenvolvimento desse
trabalho, há naturalmente uma parcela que diz respeito à física. Apesar dessa
parcela não representar nem de longe a maior parte do que aprendi, praticamente
todas as páginas dessa tese serão dedicadas à ela. Reservo portanto esse espaço
para falar um pouco sobre algumas das partes que cam de fora mas devem ser
mencionadas.
O que pretendo incluir aqui gira em torno de uma só ideia: as pessoas que
dividiram comigo essa trajetória. São elas que representam o cerne do que levo
desse doutorado e quero mencionar então algumas em particular.
Para começar, os meus pais e o meu irmão, por serem as pessoas que me
apoiaram nas minhas escolhas e me deram suporte para encarar essa empreitada.
O meu orientador, Diego, por ter me recebido no seu grupo de pesquisa e ter
me acompanhado do início ao m, apesar de todos os percalços.
Os alunos do IFUSP, Daniel, Gabriel, Felipe, Levy, Carlos, Caio, Giulio, Alex,
Laura e Renato, e a Gabi e a Simone, por tornarem acolhedores até mesmo os
corredores estéreis do Departamento de Física Matemática. Em particular os que
dividiram comigo não só a mesma sala, mas também toda a montanha russa de
sentimentos que foram esses anos: Daniel, Gabriel e Felipe. Vocês foram os meus
escudeiros éis e olhando em retrospecto vejo que a base do que construí em São
Paulo certamente começou como um tripé. Por m, o Levy, por ter corrido ao
meu lado e me proporcionado doses semanais de alívio em meio à pandemia.
A minha amiga Renata, que a princípio dividiria comigo o apartamento 16, mas
que acabou por dividir muito mais e por car ao meu lado em todos os aspectos
possíveis e imagináveis.
À vocês os meus profundos agradecimentos. O que dividimos é o que levo de
mais precioso dessa jornada.

Abstract
The goal of this thesis is to study Wilson loops in 3 dimensional Chern-Simons-
matter theories aiming at applications in holography. In particular, we study
operators preserving some amount of supersymmetry in ABJ(M) theory and in
N =4 Chern-Simons-matter theories. In both cases, we seek for a systematic
construction of these operators to understand the moduli space of the theory. In
the case of ABJ(M), we also make a perturbative detour aiming at some universal
information as, for example, the Bremsstrahlung function.
Key-Words: supersymmetry, Wilson loops", Chern-Simons-matter theories.

Resumo
O objetivo desta tese é estudar laços de Wilson em teoria supersimétrica de Chern-
Simons com matéria em 3 dimensões visando aplicações em holograa. Em par-
ticular, nós estudamos operadores preservando alguma quantidade de supersime-
tria em teoria de ABJ(M) e em teorias com N = 4. Em ambos os casos, nossa
intenção é obter uma construção sistemática desses operadores para entender o
espaço de módulos da teoria. No caso de ABJ(M), nós também fazemos cálcu-
los perturbativos visando informações universais como, por exemplo, a função de
Bremsstrahlung.
Palavras-Chave: supersimetria, laços de Wilson", teorias supersimétricas de
Chern-Simons com matéria.

Contents
1 Introduction 1
2 Basic toolkit 5
2.1 Supersymmetric quiver theories . . . . . . . . . . . . . . . . . . . . 5
2.2 Supersymmetric Wilson loops . . . . . . . . . . . . . . . . . . . . . 7
2.3 Quivers and Wilson loops . . . . . . . . . . . . . . . . . . . . . . . 9
3 N =6 Chern-Simons-matter theory 14
3.1 Supersymmetric Wilson loops . . . . . . . . . . . . . . . . . . . . . 16
3.1.1 Circular loops . . . . . . . . . . . . . . . . . . . . . . . . . . 16
3.1.2 Latitude loops . . . . . . . . . . . . . . . . . . . . . . . . . 21
3.2 From 1-node to 2-node loops . . . . . . . . . . . . . . . . . . . . . 23
3.3 New 1=6 BPS operator . . . . . . . . . . . . . . . . . . . . . . . . . 40
3.4 Perturbative detour: deforming the 1/2 BPS loop . . . . . . . . . . 48
3.4.1 1-loop contribution . . . . . . . . . . . . . . . . . . . . . . . 49
3.4.2 2-loops contribution . . . . . . . . . . . . . . . . . . . . . . 55
4 N = 4 Chern-Simons-matter theories 60
4.1 1-node Wilson loops . . . . . . . . . . . . . . . . . . . . . . . . . . 64
4.2 Hyperloops at =0 . . . . . . . . . . . . . . . . . . . . . . . . . . 66
4.2.1 Moduli spaces . . . . . . . . . . . . . . . . . . . . . . . . . . 70
4.3 Hyperloops at 6= 0 . . . . . . . . . . . . . . . . . . . . . . . . . . 73
4.3.1 Construction . . . . . . . . . . . . . . . . . . . . . . . . . . 73
4.3.2 A closer look at the hyperloops . . . . . . . . . . . . . . . . 77
4.3.3 Further examples . . . . . . . . . . . . . . . . . . . . . . . . 79
4.4 A matrix model proposal . . . . . . . . . . . . . . . . . . . . . . . 81
4.5 Comments . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 82
4.6 New hyperloops . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 84
Appendices 89
Chapter 1
Introduction
Wilson loops (WLs) are fundamental objects in any gauge theory. Mathematically,
they represent the holonomy of the gauge connection A around a closed path ,
WR [ ] = Tr R P exp i A : (1.1)
The symbol P stands for the path-ordering of the operator. The gauge eld
takes value on the Lie algebra of the corresponding gauge group and the trace is
taken with respect to the representation R of the gauge eld. From the physical
point of view, these loops give information, for example, about the Schwinger pair
production probability [1] and about connement [2].
In the context of supersymmetric theories, Wilson loops may acquire an extra
and special feature. When these operators are made supersymmetric, their ex-
pectation value may be computed exactly. This is extremely valuable since exact
results are rare in quantum eld theory (QFT). Such results are obtained using
a technique called localization [35]. In fact, localization can be used to compute
the expectation value of any supersymmetry preserving operator and it relies on
the fact that in some cases path integrals are exactly equal to their semiclassical
approximations [6, 7]. Roughly speaking, a supersymmetric theory admits a Lie
superalgebra of symmetries. This superalgebra contains odd generators and, for
a theory with N supersymmetry, there are 4N of those. An operator O that

is annihilated by at least one odd element Q of this superalgebra is named after
Evgeny Bogomolny, M. K. Prasad, and Charles Sommereld as BPS and denoted
as O BPS . S of the theory by

The spirit of localization is then to deform the action
a term generated by the action of Q on some eld conguration A, called Q-exact
term, such that
S ! S + tQA : (1.2)
1
2
For Q2-invariant A, this deformation does not change the vacuum expectation
value of O . In terms of the Feynman path integral we have that the vacuum
BPS
expectation value of O is given by

BPS

hO i = D[a] e
BPS
S [a]+tQA O
BPS ; (1.3)
where a are the elds of the theory. Finally, for suciently well-behaved A, the
expectation value is independent of t so this path integral can be evaluated in the
limit t ! 1, where the integral is dominated by the saddle points of the deformed
term.
These exact results provide a rich arena to test the conjectured duality between
string theory in anti-de Sitter (AdS) space and conformal eld theories (CFTs),
the so called AdS/CFT correspondence [8, 9]. The duality is an explicit realization
of the holographic principle [10, 11] and for this reason is sometimes referred to
as holography. More precisely, it states a weak-strong coupling duality where,
on the one hand, we have a string theory dened in terms of three fundamental
quantities: the radius L d + 1 dimensional AdS space, the string length ls

of a
and the string coupling constant gs . On the other hand, we have a d dimensional
eld theory living in the boundary of AdS with coupling constant g and SU (N ) YM
gauge symmetry, such that the 't Hooft coupling constant is NgY M . For this
2
reason subscripts are often included to specify the dimensions as AdSd+1 /CFTd
and we often refer to the respective theories as bulk and boundary theories. The
correspondence then states that quantities from both sides are related through

L 4
= ; g2 = 2g :
l
s
YM s
Holography has been mostly implemented in the limit where the number N of
elds involved is very large and the 't Hooft coupling is kept xed and large.
In this limit, referred to as large N limit, only planar diagrams in perturbation
theory survive and the gauge theory is substantially simplied [12]. From the
bulk theory point of view, this implies that the string coupling constant is small,
gs 1, such that the string theory is weakly coupled and non-perturbative eects
can be safely neglected. Moreover, having 1, implies that the radius L of AdS
is large. Thus the AdS curvature becomes small and string theory is eectively
described by classical supergravity so this side of the duality is also often
referred to as the gravity side. Therefore one use of the AdS/CFT correspondence
is the possibility to use an easier description rather than the other for computing
3
interesting physical observables, since questions on one side of the duality can be
reformulated in the language of the other side.
Being a weak-strong coupling duality makes the AdS/CFT correspondence very
powerful, but it also makes it hard to be tested. Testing it requires a computation
of the expectation value of some observable at all orders in the coupling constant
on one side of the correspondence. This is where the aforementioned exact results
become particularly relevant. Holographycally, Wilson loop operators are mapped
to fundamental strings and their exact results allow us to probe the duality at
both weak and strong coupling.
This non-trivial check of the duality was rst done in four-dimensional N =4

super Yang-Mills (SYM), which is dual to type IIB strings on AdS5 S5 [9].
The circular 1/2 BPS loop was computed exactly through localization [3, 4]. The
weak coupling expansion of the exact result matches the perturbative computation
whereas the strong coupling limit matches the fundamental string result [13].
In the same spirit, this non-trivial check can be extended to three-dimensional
supersymmetric theories. In this work we will focus on the particular cases of
N =6 Chern-Simons-matter (CSm) theory [14, 15], known as ABJ(M) theory,
and N = 4 Chern-Simons-matter theories [1619]. Even though most of our
attention will be driven to the construction of supersymmetric operators itself, it
is clear how this ts into the holography program.
The thesis is organized as follows. In Chapter 2 we begin by covering some
common ground concerning generic CSm theories. Hopefully it provides, in a
pedagogical way, the required tools to later discuss the particular cases of N =6
and N = 4 theories.
The case of N = 6 is addressed in Chapter 3. We begin it with a review of
the theory on R3 and the known BPS WLs. We explain how one can reformulate
circular operators and in this way address their moduli space. This is based
on results presented in the second chapter of [20], but here we provide much
more detail. Then we introduce a new 1=6 BPS operator that to our knowledge
was not known in the literature. Finally, we make an attempt of a perturbative
computation concerning a deformed version of the 1=2 BPS loop.

Chapter 4 is then dedicated to N =4 theories on S3 and it is mostly based
in [21]. We begin by introducing the theory and presenting BPS operators that
couple to a single node. These are presented as great circles that are continuously
connected by marginal deformations to the usual bosonic 1=4 BPS Gaiotto-Yin
WLs [22]. Extending the proposal of [20] to N = 4, we use these operators to

4
derive their multiple nodes counterparts and then we study their moduli spaces.
Finally, we include some preliminary results of an ongoing project [23] that started
from our studies of the N = 4 analogue of the new 1=6 BPS operator derived in
ABJ(M).
All details about supersymmetry transformations, Feynman rules and other
technicalities are presented in three appendices.

Chapter 2
Basic toolkit
The main content of this work is divided into two parts, one about ABJ(M) theory
and one about N =4 CSm theories. There is, however, some overlap between
these two, since both consist on supersymmetric quiver theories and in both cases
we are interested in WLs preserving some amount of supersymmetry. We dedicate
this chapter to cover this common ground and equip the reader with a basic toolkit
for what follows.
2.1 Supersymmetric quiver theories

Quiver diagrams are pictorial representations that illustrate the eld content and
symmetries of a theory in a simple way. For supersymmetric theories, the kind of
quiver we can draw varies depending on the amount of supersymmetry and, once
augmented by the superpotential, mass terms and Fayet-Iliopoulos (FI) terms, the
quiver diagram is enough to completely x the Lagrangian of a particular theory.
The general idea is to associate a symbol to each element of the theory: nodes
to gauge group factors and arrows to matter elds. Within each node we have
a number NI specifying the U (NI ) gauge group. An arrow connecting a node I

and a neighbor node I + 1 represents bifundamental elds in the fundamental of
the starting node and anti-fundamental representation of the target node, i.e., in
(;
) of U (NI ) U (NI +1 ). It will not be the case for the particular theories
considered in the following chapters, but for completeness let us also consider
the possibility of having global (avor) symmetry and, in this case, U (MI ) global
factors are associated to square nodes.
Take for instance the simplest case where N = 1. As explained above, vector
superelds are represented as nodes and chiral superelds as arrows. To each
vector multiplet V transforming in the adjoint representation of the gauge group
5
2.1 Supersymmetric quiver theories 6
we associate a circle node. To each global symmetry we associate a square node.
To each chiral supereld we associate an arrow.

Going one step further, we have N = 2 supersymmetry. Its vector multiplet
can be decomposed in terms of N = 1 superelds as the direct sum of a vector
and a chiral multiplet,
N = 2 vector multiplet ! V ;
whereas its hypermultiplet can be split into the direct sum of two N =1 chiral
multiplets,
N = 2 hypermultiplet ! ~ :
Thus one way of drawing a quiver for a theory with N = 2 is simply by following
the same rules used for the N = 1 case. However, we can also simplify these
rules by considering the decomposition above. In particular, every circle node
(associated to V ) has an arrow (associated to ) that starts and ends in the same
circle node, so we can remove arrows that start and end on the same circle node.
Furthermore, every arrow starting on node I and ending on node I + 1 (associated

to ) has a counterpart, i.e., an arrow starting on node I + 1 and ending on node
I (associated to ~ ). So for N =2 we can replace two arrows between nodes by
an edge with no orientation.
To clarify the ideas explained so far, let us consider as an example 3d N =4

SQED with N avors. Since this theory can be obtained from 4d N = 2 SQED
through dimensional reduction, we can apply the set of rules explained above and
see how one would draw the corresponding quiver. The eld content of 4d N =2
SQED is composed by one vector multiplet and one hypermultiplet. Being it
a supersymmetric version of Quantum Electrodynamics, there is an U (1) gauge
symmetry and, in addition to it, we also include an U (N ) avor symmetry. The
quiver diagrams representing the theory are shown in Figure 1, where in (a) the
quiver is drawn in N =1 language whereas in (b) we use the simplied rules of
N = 2.
Let us pause for a moment to comment on the consistency of these diagrams.
First, note that and ~ transform in the same representations of both gauge and
avor groups, as expected from the N = 2 hypermultiplet decomposition. There-

fore and ~ transform in conjugate representations, namely (;
) and (
; ) of
U (1) U (N ). This is captured by the orientation of the arrows representing them
in Figure 1a. Also, note that belongs to the adjoint representation of U (1), as
can be seen from the N = 2 vector multiplet decomposition. This is captured by
2.2 Supersymmetric Wilson loops 7
~
1 N 1 N
(a) (b)
Figure 1: Quiver diagram of 4d N =2 SQED in (a) N =1 and (b) N =2

languages. In addition to the quiver, we also need the superpotential to completely
x the Lagrangian of the theory.
the fact that its arrow starts and ends on the circle node in Figure 1a. Of course,
once you are familiarized with the theory and its eld content, it is easy to see
that the N =2 quiver in Figure (1b) is enough to capture all essential features
present in Figure 1a.
Throughout this work, we will make a minor change to the rules presented
above. In order to follow the standard notation in the literature of supersymmetric
WLs, instead of writing NI inside each node corresponding to a U (NI ) gauge group
factor, we will place the gauge eld AI itself inside the node. Nevertheless, it will
soon become clear how quiver diagrams are useful tools to study supersymmetric
operators.
2.2 Supersymmetric Wilson loops

In supersymmetric theories, a Wilson loop operator may preserve some amount
of supersymmetry once we allow it to couple not only to the gauge eld but also
to other elds of the theory. For example, in four-dimensional N =4 SYM in
Euclidean space, this is achieved by coupling the loop to scalar elds I , I =

1; ; 6, such that
!
WR [ ; ] = Tr R P exp i (A x_ ijx_ j I I) ds : (2.1)
Note that, in comparison with (1.1), the operator now depends on the coupling
.1
Generically, this coupling varies along the contour along which the loop is
supported. For a circular loop, we nd that the particular form I = (1; 0; 0; 0; 0; 0)
notation WR [ ; ] is precise because it makes explicit the dependence of the loop on three
1 The
features: the representation R of the gauge group, the contour along which it is supported and
the matter coupling . Nevertheless, to avoid cluttering, from now on we will denote the operator
simply as W and the trace as Tr . Specications will be included only when needed.
is such that
(A x_ ijx_ j I I ) = 0 ; (2.2)
for 8 out of the 16 supercharges of the theory. Therefore, for this choice of coupling,
the resulting loop is 1/2 BPS.
For supersymmetric loops that couple to the gauge eld and scalars, as in
(2.1), the argument of the integral is usually referred to as the dressed bosonic
connection and denoted as A. In general, the corresponding WLs are called
bosonic and written as

!
W = Tr P exp i A ds : (2.3)
As in (2.2), the condition they need to fulll to be supersymmetric is that, under
the action of some supersymmetry transformation, the variation of A is null.

There is, however, the possibility of allowing the loop to also couple to fermions,
giving rise to operators that are called fermionic. These are constructed by pro-
moting the dressed bosonic connection to a superconnection L that includes
gauge, scalar and fermi elds of the theory. In this case, the loop is written as
W = Tr P exp i L ds ; (2.4)
where L is a supermatrix and the trace is used loosely, in the sense that it could be
a supertrace or something more intricate, depending on the particular structure
of L. Actually, in the rst construction of a fermionic loop [24], the authors
obtained that, in order to preserve supersymmetry, the operator should be traced.
However, tracing supermatrices does not seem like a reasonable operation. Indeed,
later in [25] the authors argued that the operator should be written in terms of a
supertrace. We postpone the details of this discussion to the next chapter.
In contrast to the bosonic case, supersymmmetry invariance of fermionic loops
is more subtle. In particular, the condition A = 0 is replaced by the weaker
requirement that L translates into a supergauge transformation, such that the
traced/supertraced loop operator is invariant. To wit, a fermionic loop is super-
symmetric if
L = Ds G @s G + i[L; G ] : (2.5)
Here Ds G represents the supercovariant derivative constructed out of the super-

2.3 Quivers and Wilson loops 9
connection L acting on a supermatrix G .

Historically, the rst fermionic loop was constructed in ABJ(M) theory in [24].
The proof of supersymmetry invariance was carried out explicitly, i.e., by expand-
ing the path-ordered exponential in (2.4) and showing that terms from dierent
orders of the expansion non-trivially cancelled each other. Later in [26], it was
understood that this was actually a consequence of having equation (2.5) satis-
ed. This allowed the authors of [25] to construct loops supported along arbitrary
curves on S 2.
2.3 Quivers and Wilson loops

When reviewing the literature of supersymmetric WLs, we usually come across a
narrative that, as the one of the previous section, presents bosonic and fermionic
loops as distinct objects. Such a distinction is comprehensible if we look at the
developments within the eld through chronological lenses. However, here we
would like to follow a dierent approach. In this section our goal is to give a tour
that starts from A in (2.3) and goes to L in (2.4) in a intuitive way, hopefully
oering a unied overview of the underlying physics. To make our journey as
smooth as possible, let us put aside particularities of a theory and focus on the
narrative itself.
Suppose we have a certain three-dimensional Chern-Simons-matter theory with
multiple gauge elds and matter elds, scalars and fermions, charged under them.
In N = 2 language, such theory would be schematically described by the quiver
diagram of Figure 2.
AI 1 AI AI +1
Figure 2: Schematic representation of a quiver gauge theory in N = 2 language.
Now imagine that within this theory, we have supersymmetric WLs with the
same structure as the one in (2.3). Since our hypothetical theory has multiple
nodes, we already have the generalization where there is an operator like that for
each node of index I,

! !
W = Tr P exp i AI ds = Tr P exp i AI x_ + ds :

Let us focus on what could arise in the ellipsis piece. Whatever that is, it should
feature some of the properties of the gauge eld: it should have mass dimen-
sion equal to 1 and belong to the adjoint representation of U (NI ). Recall that
matter elds in this theory belong either to the the bifundamental or to the anti-
bifundamental representation. So if we want to include them, it should be done as
a bilinear combination belonging to the same representation as AI . Since in 3 di-
mensions scalars have dimension 1/2 and fermions have dimension 1, we conclude
that scalar bilinears should be the perfect t for the ellipsis above, giving
W = Tr P exp i AI x_ + scalar bilinears ds : 2
In particular, such linear combination can be made for all scalars that are charged
under the I -th node, either the ones to the left or to the right of it. At this point it
N = 1 and N = 2 quivers and open each edge

is enlightening to mix languages for
in Figure 2 as the corresponding two arrows of N = 1. This results in the quiver
diagram shown in Figure 3 where, in addition to turning edges into arrows, we also
highlight the matter elds involved in red. From it we see that we could think of
AI as starting from the I -th node and traveling through (scalar) red arrows such
that by the end of the day the path takes us back to the starting point.
AI 1 AI AI +1
Figure 3: Schematic representation of a quiver gauge theory mixing N =2 and

N = 1 languages where we highlight matter elds involved in the denition of the
1-node WL.
So far we have the possibility of dening an operator that couples to one node
of the quiver theory and, in particular, we have an operator like that for each node
I. We can then think of each of these as one Lego piece and, in the same spirit
as the game, try to explore the possibility of gluing two of these pieces together.
Physically, this would correspond to allowing the resulting operator to couple to
two nodes of the quiver theory, giving rise to something that could be sketched as
0 1
W = Tr P exp i @
AI 0 A
!
ds :
0 AI +1
Whatever this object is, we already see that it has dimensions (NI + NI +1 ) (NI +
2 The
analogous analysis holds for (2.1). The theory is four-dimensional, thus scalars have mass
dimension 1. These belong to the adjoint representation, so they appear linearly in A.
NI +1 ). Its diagonal blocks areNI NI and NI +1 NI +1 matrices, since these are

in the adjoint of U (NI ) and U (NI +1 ).
Consequently, its o-diagonal entries are
NI NI +1 and NI +1 NI dimensional. Now recall that matter elds connecting

nodes I and I + 1 have these exact matrix dimensions, since they belong to (; )
and (
; ) of U (NI ) U (NI +1 ). So we can consider linear appearances of them
as suitable candidates to be placed in the o-diagonal entries of this matrix. By
dimensional analysis, such candidate should have mass dimension 1 and this is the
precise case for fermions in 3d. So these represent a perfect t, giving rise to an
operator that looks like
0 1
W = Tr P exp i @
AI fermions
A
!
ds :
fermions AI +1
This motivates the form of L in (2.4), showing that its o-diagonal blocks are
Grassmann odd and therefore L is actually a (NI +NI +1 )(NI +NI +1 ) supermatrix.
As in the 1-node case, looking at the quiver diagram, we could think of this
operator as diagonal entries consisting of paths starting and ending at each node.
In particular, we would have the red path of Figure 3 for AI and the analogous
one for AI +1, as show in Figure 4a. These are augmented by o-diagonal entries
consisting of open paths, the (fermionic) arrows, connecting neighbor nodes. These
are denoted in blue in Figure 4b. The top-right entry of the operator is represented
by the blue arrow that goes from node I to node I + 1 whereas the bottom-left
entry is the conjugate arrow pointing in the opposite direction.
AI 1 AI AI +1
(a)
AI 1 AI AI +1
(b)
Figure 4: Highlighted paths illustrating the new pieces that are now part of L. In
(a) we highlight the structure of AI +1. In (b) we highlight the structure of the
o-diagonal entries of the superconnection.
As long as we follow the consistency conditions outlined above, we could keep
generalizing the procedure by, for example, gluing more than two pieces together.
This would lead to operators that couple to three or more nodes of the quiver
theory and written in terms of a superconnection with the generic form
0 1
..
.
B C
B
B
B
AI 1 fermions
C
C
C
L= B
B
B fermions AI fermions
C
C
C :
AI +1
B C
B C
B fermions C
@ A
..
.
In the same spirit, another possibility would be to include some of the nodes more
than once. Suppose we did so by including the I -th node twice. This would give
us an operator with
0 1
..
.
B C
B
B
B
AI 1 fermions
C
C
C
L=
B
B
B fermions AI C
C
C
;
AI
B C
B C
B fermions C
B C
B
B
@
fermions AI +1 C
C
A
..
.
which seems to be the most general structure we can construct using the building
blocks outlined so far.
There is, however, a further possibility that we can consider. That is the
possibility of placing scalar bilinears above and below the fermionic entries in
L [27]. As previously argued, such bilinears automatically t the dimensional
analysis. Now in order to t the matrix representation they should consist of
scalars belonging to neighboring edges of the quiver. To illustrate this possibility in
more detail, let us consider for instance a 3-node operador with a superconnection
of the form
AI
0 1
B
1 fermions scalar bilinears
C
L= B
B
@
fermions AI fermions
C
C
A
:
scalar bilinears fermions AI +1
Above we colored the new structures appearing in the superconnection. These
are denoted with the respective colors in Figure 5. Note that, as in the case of
fermions, their respective highlighted paths are open since these scalar bilinears
connect consecutive nodes.

AI 1 AI AI +1
Figure 5: Highlighted paths illustrating the possibility of including scalar bilinears
outside of A.
Though we have intentionally skipped details and considered a generic theory,
this game sketches the idea behind all known examples of supersymmetric WLs in
three-dimensional Chern-Simons-matter theories. By now, we hope that the reader
is convinced that quiver diagrams can oer an intuitive view on supersymmetric
loops. In the following chapters we will focus on the particular case of ABJ(M)
and N = 4 theories. In particular, in Chapter 4 we will associate a diagram to the

Wilson loop operator itself, allowing us to increment this Lego game with extra
features, providing an even richer overview.

Chapter 3
N =6 Chern-Simons-matter theory
N =6 Chern-Simons-matter theory is a three-dimensional quiver gauge theory
constructed from two copies of Chern-Simons actions coupled to matter, one with
gauge groups U (N1 ) at level k and the other with gauge group U (N2 ) at level
k. The construction of such theory was rst done in 2008 by Aharony, Bergman,
Jaeris and Maldacena [14] for the case where N1 = N2 and it was named ABJM
theory. Later in the same year the generalisation for N1 6= N2 was done by
Aharony, Bergman and Jaeris [15] and it was named ABJ theory. In both cases,
the gauge theory is dual to type IIA strings on AdS4 CP3 . Here, as mostly done
in the literature, we consider gauge groups with dierent ranks and refer to the
theory as ABJ(M).
The eld content of ABJ(M) is depicted in the quiver diagram of Figure 6. It
includes two gauge elds, A1 and A2 , in the adjoint representation of the respec-
tive gauge groups U (N1 ) and U (N2 ). The matter sector has SU (4) R-symmetry
and is composed by scalars CI and fermions , I = f1; 2; 3; 4g, in the (; )

I
representation of U (N1 ) U (N2 ). By conjugation there are also the elds C I and
I in (
; ) of U (N1 ) U (N2 ).
CI I
A1 A2
C I I
k k
Figure 6: Quiver diagram of ABJ(M) theory. Below each node we include the
level of the respective copy of the Chern-Simons action.
In Euclidean space, the theory is described by the action
S =S +S CS matter +S ;
gf (3.1)
14
15
1
where
" #
ik 2 2
S = d3 x Tr A1 @ A1 + iA1 A1 A1 A2 @ A2 iA A A ;
CS
4 3 3 2 2 2
" #
Smatter = d3 x Tr D CI D C I + i I D I +S ; int
" #
k 1 1
S = d3 x Tr ( @ A )2 + @
1 cD c ( @ A )2
2 @ c^D c^ :
gf
4
The covariant derivatives are dened as
D CI = @ CI + i(A1 CI CI A2 ) ; D C I = @ C I i(C I A1 A2 C I ) ;

(3.2)
D I = @ I + i(A1 I I A2 ) ; D I = @ I i( I A1 A2 I ) :
N = 6 supersymmetry and is invariant under 24 supersymme-

The theory has
try transformations generated by QIJ and SIJ . Generators Q come accompanied
by their respective parameters and generate super-Poincaré transformations.
Analogously, generators S come accompanied by parameters and generate su-
perconformal transformations. The fact that the theory is invariant under these
supersymmetry transformations means that
S = (Q + S )S = IJ QIJ S + IJ SIJ S = 0 ;
for any Q and S or, equivalently, any and . Note that indices I and J go from
43
1 to 4 and are anti-symmetric, giving rise to =6 parameters. In addition
2
to that, there are two possibilities for the spinor index , which we denote as +
and . Thus there is a total of 6 2 = 12 parameters and 12 parameters that
leave S invariant, giving the 24 supercharges preserved by the theory.
In what follows we will focus on supersymmetric WLs within this theory. So we
should clarify that for any BPS operator O

BPS , its invariance under supersymmetry
transformations is shown to hold only for certain linear combinations of generators
Q and S . So, in contrast to the notation above, we denote a particular preserved
linear combination of generators as Q. In this case we have that, under the action
of a preserved supercharge Q, the operator satises
O BPS
QO
= BPS = 0:
1 Theinteraction terms in Sint will only play a role in the perturbative computation and, in
particular, through the interation vertices in (A.3) and (A.4). Thefore we do not write it explicitly.
Also, the piece Sgf will play no role since in our perturbative computation we use Feynman rules
in Landau gauge and therefore ghost elds ; c; c^ are removed.
The paramater
is read from Q through the appropriate index contractions,
namely IJ QIJ . Moreover, for any preserved supercharge Q, the corresponding

parameter is a conformal Killing spinor. In particular, gauge and matter elds
transform under supersymmetry transformations generated by these parameters
as (A.2).
3.1 Supersymmetric Wilson loops

As explained in the previous chapter, one can construct supersymmetric Wilson
loops within this theory and, in particular, there are two main ways of doing so.
One way is to allow the loop to couple only to one node of the quiver in Figure 6,
such that it contains vector and scalar elds [2830]. The other way is to allow
the loop to couple to the two nodes of the theory, including not only vector and
scalar elds, but also fermions [24, 31].
To avoid cluttering, we intentionally skipped the details in the narrative of
Chapter 2, but recall that every loop is dened along some particular contour. In
what follows we specialise to either the circle C,
C = (0; cos ; sin ) ; (3.3)
or the latitude ,

= (sin 0 ; cos 0 cos ; cos 0 sin ) with
2
0 ;
2
(3.4)
as shown in Figure 7. We will mainly state the form of the operators and include
the corresponding references for more details.
Before proceeding, let us point that we do not include normalisation factors
in the denition of the operators. These are usually included to guarantee that,
at zero order in the perturbative computation, the expectation value of W is
normalised to 1. Since a perturbative analysis is restricted to Section 3.4, we
remove the normalisation everywhere else.
3.1.1 Circular loops
When dened along the circle C, the 1-node loop of ABJ(M) is invariant under
4 out of the 24 supercharges and therefore is 1=6 BPS. This operator was rst
discovered as the 1/2 BPS loop of the N = 2 version of Chern-Simons-matter
0 C
Figure 7: Circle and latitude curves with their respective parameters.
theory [22]. It takes the form
W = Tr P exp i A d ; A Ax_ ijx_ j ; (3.5)
where, in accordance with [24], the path ordering is taken to be left-to-right. In
N = 2 theory, the eld is the auxiliary eld in the vector multiplet. In N =6

theory, the loops coupling to each node can be dened independently as
! !
WN = Tr N P exp
1 1 i A1 d and WN = Tr N P exp
2 2 i A2 d ;
(3.6)
where
A1 = A1x_ i1jx_ j and A2 = A2x_ i2jx_ j : (3.7)
Scalar bilinears are then obtained by integrating out the auxiliary elds 1 and
2 . In particular, in [2830] these were xed to be
2 2
1 = MJI CI C J and 2 = MJI C J CI ; (3.8)
k k
with M = diag( 1; 1; +1; +1).
The scalar coupling matrix M dictates the form in which scalar bilinears ap-
pear in the loop. With the particular form above, the resulting loops preserve
four supercharges parametrised by 12 and 34 , accompanied by superconformal
supersymmetries that are xed to

3 12 3 34
12 = i 34 = i
jxj and
jxj : (3.9)
We can explicitly check that the loops WN 1 and WN 2 are invariant under super-
symmetry transformations generated by these parameters.
For instance, let us take WN . 1 We begin by introducing projectors
0 1

1
1
x_ 1 1
!
= @ i
ie i
:
jx_ j
A (3.10)
2 2 ie 1
Then, (3.9) implies that each Killing spinor can be written as

12 = 2+ 12 = 212

and 34 = 2 34 = 234 + ;
(3.11)
since the projectors satisfy ( ) = ( ) .

2
Plugging these into the variation
of each eld, according to the rules of Appendix A, we obtain
" #
4i 12 3 C 4 4 C 3
A1 = ( ) C1 2 C2 1 +
k
" #
4i 34
+ ( ) C3 4 C4 3 + 1 C 2 2 C 1 ; (3.12)
k
CI = 2 12 12IJ J + 2 34 34IJ J ; (3.13)
C J = 2 JK K : (3.14)
Now, the particular combination
2i
A1 = A1x_ k
jx_ jMJI CI C J (3.15)
is such that
" #
4i 12
A1 = (1 + x_ )
C1 2 C2 1 +
3 C 4 4 C 3
k
" #
4i 34 1 2 2 1
(1 x_ ) C3 4 C4 3 + C C
k
" #
8i 12
= (+ ) C1 2 C2 1 + C
3 4 C
4 3
k
" #
8i 34 1 2 2 1
( ) C3 4 C4 3 + C C :
k
2 Asexplained in Appendix A, spinor indices are omitted as in (A.1) and are raised and lowered
12 = 2(+ ) 12 = 2( ) 12 = 212 ( ) =
with the appropriate symbol. Explicitly,
2 ( ) .
12
From (3.11), it follows that
12 (+ )
=0 and 34 ( )
= 0; (3.16)
since projectors satisfy + = + = 0. A1 = 0 and the loop

Consequently
WN indeed preserves 4 supercharges parametrised by 12 and 34 accompanied by

1
special parameters xed as in (3.9). Thus the operator is indeed 1=6 BPS. An
analogous computation can be carried for WN , giving A2 = 0. 2
It is worth noticing that in this check of supersymmetry invariance, it is crucial
to write
in terms of the projectors and to have the specic form of M=
diag( 1; 1; +1; +1). These ingredients appear with mild variations throughout
this work. We should also comment that, as can be seen from the structure of M,
the loop has a residual SU (2) SU (2) R-symmetry. This residual symmetry is
relevant because it should be also manifest in the dual setting, i.e., in the string
theory side the corresponding semi-classical string surface should have SU (2)
SU (2) bosonic symmetry.
This gives the simplest example of a circular supersymmetric loop. Its vev was
computed exactly through localization in [5]. On the string theory side, this 1=6
BPS solution matched a CP3 smeared string. This was the rst non-trivial check
of the AdS/CFT duality in the context of ABJ(M).
The 2-node version of this loop is based on a superconnection L that is a
(N1 + N2 ) (N1 + N2 )-dimensional supermatrix. For a loop supported along a
particular curve in Euclidean space, L has the generic form

A1 = A1x_ jx_ jMJI CI C J ;
0 q 1
L= @
A1
q
i k jx_ jI
2 I
; A
2i
k
i 2k jx_ j I I A2 A2 = A2x_ 2i
k jx_ jMJI C J CI :
(3.17)
The fermi elds appear through fermionic couplings and that are Grassman
even quantities, such that the o-diagonal blocks of L are Grassmann odd and L
is indeed a supermatrix.
When the loop is supported along the circle C , the couplings

0 1
ie i=2
I ( ) = ei=2 ie i=2 I1 ; I ( ) = @ A 1I ; (3.18)
ei=2
and
M = diag( 1; +1; +1; +1) ;

make the traced loop,
W = Tr P exp i L( ) d ; (3.19)

C
invariant under half of the supersymmetries of the the theory [24]. In particu-
lar, this operator is invariant under the action of supersymmetry transformations
parameterised by 1I and IJ , I; J 6= 1, with superconformal parameters xed to

3 1I 3 IJ
1I = i IJ = i I; J 6= 1 :
jxj and
jxj ; (3.20)
The parametrisation (3.20) is such that 1I = 2+ 1I

and IJ = 2 IJ ,

I; J 6= 1. In terms of these, the variation of each component of the loop is given
by
8i 4i
A1 = CI 1 1I + IJ 1 C K ;
k k 1IJK
8i 1I + 4i 1IJK
A2 = 1 CI IJ C K 1 ;
k k (3.21)
1 ( ) = D (4i1 CI ) ;
1 1I
1 1 ( ) = D ( 2i1IJK IJ 1 C K ) :

The covariant derivatives above are written in terms of the dressed bosonic con-
nections, i.e., instead of the usual denition in (3.2) we have
D CI x_ @CI + i(A1CI CI A2) ;

(3.22)
D C I x_ @C I i(C I A1 A2C I ) :
Plugging (3.21) into the variation of the superconnection,
0 q 1
A1 i 2 jx_ jI I ;
L = @ q k A (3.23)
i 2k jx_ j I I A2
we nd that (2.5) is satised with
s 0 1
2 @ 0 2 1I CI A
G=2 k : (3.24)
1IJK IJ C K
This loop is dual to a fundamental CP1 localized string solution preserving
1/2 of the supercharges of the vacuum and U (1) SL(2; R) SU (3) bosonic
symmetry. Accordingly, note that the scalar coupling M = diag( 1; +1; +1; +1)
of this 2-node loop has precisely SU (3) residual symmetry. Chronologically, this
string solution was known before the construction of the corresponding loop on
the gauge theory side. Finding the gauge theory dual relied on allowing, for the
rst time, the loop to couple to fermions.
3.1.2 Latitude loops
When dened along the latitude in (3.4) the 1-node loop of ABJ(M) is invariant
under 2 out of the 24 supercharges and is 1/12 BPS. As in the circular case, it can
be dened independently for each gauge group and the respective dressed bosonic
connections are
2i
A1 = A1x_ k
jx_ jMJI CI C J ;
2i
A2 = A2x_ k
jx_ jMJI C J CI ;
with 0
i p 1
e 1 2 0 0C
B p
B
B ei 1 2 0 0CC
M= B C :
0 0 1 0C
B C
B
@ A
0 0 0 1
Here the M matrix is written in terms of what is called the eective latitude
parameter,
sin 2 cos 0 ; (3.25)
where is an arbitrary parameter that can be freely chosen in [0; 2 ] and 0 is
the latitude angle. The operator is invariant under the action of supersymmetry
transformations parameterised by two spinorial parameters !i , i = 1; 2. The non-
vanishing components of IJ
are
i0 p i0 p
+13 = e 1 + !1 ;
2 13
+ = ie 2 1 + !1 ;
i p i p
23 = ie 1 !1 ; 2
0
23 = e 2
0
1 !1 ;
i p i p
+14 = e 2
0
1 !2 ; 14
+ = ie 2
0
1 !2 ;
i p i p
24 = ie 1 + !2 ;
2
0
24 = e 1 + !2 :
2
0
Going to its 2-node version, we nd that for a superconnection L of the form
(3.17) with couplings
0 p 1
1+
B p C
epi
2
B
B 1 e i C
C
I ( ) = B C
1 ie i ;
2 B
B
@
0
C
C
A
0 I
0 1
I ( ) =
ep
i
2 p p i
I 1
;
1+ 1 e 0 0 @ A (3.26)
2 iei
0
i p 1
e 1 2 0 0C
B p
B
ei 1 2 0 0C
MJI
B C
= B C ;
0 0 1 0C
B C
B
@ A
0 0 0 1
3
the supertraced operator
! !
W = sTr P exp i L d T (3.27)
is 1=6 BPS. Here the matrix T , called twist matrix, is explicitly inserted in the
supertrace to guarantee invariance of W under supergauge transformations. In
the latitude case, it is given by
0 1
i
e IN 0
T
2
= @
1
A : (3.28)
0 e i IN
2
2
In particular, the operator in (3.27) is invariant under the action of super-
symmetry transformations parameterised by four constants !i , i = 1; ; 4. The
non-vanishing components of IJ
are
i0 p i0 p i0 p i0 p

+13 = e 2 1 !1 + e 1+ 2 !2 ; 13
+ = ie 1 2 !1 ie 1 + !2 ;
2
i p i p i p i p
23 = ie 1 + !1 ie
2
0
1 2
0
!2 ; 23 = e 1 2
0
!2 e 1 + !1 ;
2
0
i p i p i p i p
+14 = e 2
0
1 !3 + e 1+ 2
0
!4 ; 14
+ = ie 1 2
0
!3 ie 1 + !4 ;
2
0
i p i p i p i p
24 = ie 1 + !3 ie
2
0
1 2
0
!4 ; 24 = e 1 2
0
!4 e 1 + !3 :
2
0
This operator is actually a particular example within a larger family of loops
dened along arbitrary curves on S 2 [25]. For = 0, the limit ! 0 corresponds

to the Zarembo-like loop [32]. For = =4, we have the ABJ(M) analogue of
the DGRT loops of N =4 SYM [33]. For more details, we refer to [34].
3 The
supertrace sums the rst and subtracts the second diagonal entry: given some diagonal
matrix diag(a; b), its supertrace gives a b.
3.2 From 1-node to 2-node loops 23
3.2 From 1-node to 2-node loops

With the 1-node and 2-node circular loops properly introduced, we have the re-
quired ingredients to address what was developed in the second chapter of [20]. In
this collaboration we show how these two operators can be connected through a
continuous deformation. As every chapter of [20] consists on a short review, here
we provide more details on the construction and results we found.
The starting point of the project was the formulation of the 1/2 BPS circular
Wilson loop according to [25, 34]. In contrast with the original formulation of [24],
there the authors argue that a supersymmetric WL operator should be dened in
terms of a supertrace instead of a trace, since the operator is constructed with
supermatrices and only supertraces are invariant under cyclic permutations of
their arguments. So they rewrite (3.19) as
" ! #
W = sTr P exp i L( ) d T ; (3.29)

C
where T = idiag(IN ; IN ) is a twist matrix that, for short, we denote as T =

1 2
i3 with the understanding that each non-zero entry of 3 is actually a block
diagonal matrix with the required dimensions. Note that this twist matrix is
recovered from the latitude one in (3.28) when ! 1.

As mentioned before, the twist matrix is included in the denition of the WL
to enforce that the supertraced operator is invariant under supergauge transfor-
mations. Let us dedicate a few words to explain what we mean by that. Recall
that under supersymmetry transformations preserved by the loop, the supercon-
nection transforms as the supercovariant derivative of some supermatrix G , i.e.,

L = D G . First, if we look at the structure of L we see that derivative terms
should appear only in its o-diagonal entries, since we only expect derivative terms
to arise from the variation of fermions, see (A.2). Therefore G is always block o-
diagonal. Furthermore, since we integrate this supercovariant derivative along the
loop, we have to take into account its boundary conditions.
In the circular case, for instance, we have G given by (3.24) and, due to the
anti-periodicity of the fermionic couplings in (3.18), it satises G (2 ) = G (0).
Therefore under nite supergauge transformations U = exp( iG ), we have
U (2) = T U (0)T 1;
with T = i3 . Under the action of a preserved supercharge, the path ordered
exponential in (3.29) transforms as
!
w exp i L( ) d 7! U (0)wU 1 (2 ) ;

C
such that the operator itself transforms as
sTr (wT ) 7! sTr (U (0)wU 1 (2 )T ) = sTr (U (0)wT T 1 U 1 (2 )T )

= sTr (U (0)wT U 1 (0))
= sTr (wT ) :
Thus the supertraced operator sTr (wT ) is indeed invariant supergauge transfor-
mations generated by G and is supersymmetric.
The circular loop has a twist matrix whose eect can be seen as multiplying
the upper diagonal block by 1 and the lower by -1, thus taking the supertrace in
(3.29) back to original trace of (3.19). So by looking only at this particular case,
it seems that the two formulations are straightforwardly equivalent to each other.
Nevertheless, the argument above carries over to all loops and, in particular, to the
latitude case where G now satises G (2) = T G (0)T 1 with the non-trivial twist
matrix (3.28) due to the boundary conditions of the fermionic couplings (3.4).
The punchline is that, due to the fact that and are not periodic around
the loop, neither is G and we need to enforce periodicity by adding a twist matrix.
Keeping this in mind, let us go back to the supertraced operator (3.29). Our initial
idea when we started developing the second chapter of [20] was to make a gauge
transformation parameterised by
(2 )
=
8
3 ;
where once more we apply the shorthand notation 3 = diag (IN ; IN ). I.e.,
1 2
we made abelian gauge transformations U 1 = IN e 1

i 8 )
(2
2 U (N1 ) and U2 =
IN e
2
i 8 )
(2
2 U (N2), where =8 accounts for dierent i factors we introduced for
; compared to the original formulation. Under these gauge transformations the
elds transform as
A1 ! U1 A1 U1 1 iU1 dU1 1 = A1 +
jx_ j ;
4jxj
A2 ! U1 A2 U2 1 iU2 dU2 1 = A2
jx_ j ;
4jxj
p p
CI ! U 1 CI U 2 1 = ie i=2 C
I C I ! U2 C I U1 1 = iei=2 C I ;
I ! U1 I U2 1 =
p p
ie i=2 I ;
I ! U2 I U1 1 = iei=2 I :
The discontinuity of at 2 yields a delta function term which can be integrated
to cancel the twist matrix T . Indeed, is not a periodic function on [0; 2 ]. It
has a discontinuity, which we account for such that
!
1
@ =
4 2
( 2 ) 3 : (3.30)
We can account for this delta function contribution via
2+ ! !

P exp i L d = exp i 3 = i 3 = T
2
1; (3.31)
2
such that in the new gauge the Wilson loop is
!
W = sTr P exp i L0( ) d ; (3.32)

C
with
0 s 1
2i 1 2i 0 I
B A1 x_ jx_ jMJI CI C J + I C
L0 = BB sk 4 k C
C : (3.33)
B
2i 0I 2i 1 C
@
k I A2 x_
k
jx_ jMJI C J CI 4
A
The M matrix is unchanged, since scalars C and C change oppositely, whereas
the fermionic couplings become
0 1
1
I0

= 1 ie i I1 and 0I ( ) = @ A 1I : (3.34)
iei
The advantage of the formulation (3.32) is that it provides a manifestly (su-
per)gauge invariant operator without the aid of a twist matrix, since the couplings
(3.34) are periodic along the circle. This comes, of course, at the expense of in-
troducing the constant 1=4 pieces in the dressed bosonic connections A1 and
A2 that can be viewed as the coupling to a background eld. Moreover, it also
gives a reparametrisation invariant formulation of the couplings by allowing us to
write them in terms of eigenstates of the projector + . This happens in contrast
to all previous formulations, where the couplings contained extra phase factors
depending on the parametrisation of the contour along which the loop is dened.
To be precise, the new formulation gives fermionic couplings that are expressed in
terms of + as
I0 ( ) = (s 2+ ) I1 and 0I ( ) = (2+ s) 1I ;
whereas the ones in the original formulation are written as
I = ei=2 (s 2+ ) I1 and I = ie i=2 (2

+ s) 1I ;

for a constant spinor s = 1 0 4
. Since the observable should depend on its
contour, but not on how we choose to parametrise it, reparametrisation invariance
of the couplings and, consequently, of the loop is naturally expected.
Furthermore, this new formulation allowed us to go one step further and view
the circular 2-node loop as a deformation of the 1-node one. The idea behind this
deformation procedure is to start from the 1=6 BPS loops in (3.6) and construct
a composite bosonic connection Lbos containing the constant 1=4 shifts,
0 1
L =@
A1 0A
+
jx_ j :
bos
0 A2 4jxj 3
We want to add some deformation L to L bos such that the resulting supercon-
nection,
L=L bos + L ;
varies as a supercovariant derivative of some supermatrix. The resulting operator,
!
W = sTr P exp i L d ; (3.35)

C
should then be supersymmetric and couple also to fermions. In particular, the
deformation L that we propose is given by

0 1
0 i Ci A
L = iQG + G 2 ; where G= @ ;
i 0
iC
i, i , i = 1; 2, are constant complex parameters that are not necessarily complex

conjugates and Q is some supercharge that we will dene in what follows.
Before moving on to the details of the construction, let us discuss the general
idea behind the expression above. First, the G 2 piece adds to L bos some diagonal
contribution of scalar bilinears. To that we also add some o-diagonal contribu-
tion composed by fermions coming from QG . In the same spirit as the fermionic
4 This formulation was extended to DGRT-like loops in Chapter 7 of [20].

variations in (3.21), this piece should be such that its variation gives a covariant
derivative of scalars written in terms of dressed bosonic connections. Namely we
should have
(QG ) D (G ) ;
which in terms of Q translates to
Q2G D G :
We choose Q to be given by a linear combination of the supercharges preserved by
the 1-node loops that we started up with, such that the resulting operator should
also be invariant under its action. At the end of the day we expect the following
to hold,
L = (L
bos + L) = (L) ! D (G ) = @ (G ) + i[L; G ] ; (3.36)
such that the resulting operator is supersymmetric. Note that at this point we
are constantly referring to three dierent types of derivatives, so let us review
these before proceeding. First we have the usual covariant derivative of the theory
D, dened in (3.2). Then we have its version with dressed bosonic connections
denoted as D, see (3.22). Finally, we have the supercovariant derivative D in (2.5)

which, in contrast to the previous two, acts on supermatrices instead of individual
elds.
With this scenario in mind, we are ready to move on to the details of the
computation. Our goal is to explicitly check that the resulting superconnection
L is invariant under all supersymmetry transformations preserved by L . We bos
begin by writing the operator L explicitly.
Recall that the 1-node loops containing A1 and A2 are invariant under super-
symmetry transformations generated by 12 and 34 with superconformal param-
eters xed as (3.9). We choose the supercharge Q, with respect to which we will
build our deformation L, to be parametrised by +12 and 34 , xing superconfor-
mal parameters to
12 12
+ = i+ and 34 = i34 :
We also rescale this supercharge by a factor of p1 , such that

5
2 2
Q = 2p1 2 (Q+12 iS12+ ) + (Q34 iS34) :

According to (3.11), this choice of parameters xes conformal Killing spinors
to
0 1 0 1
1 ie i
12 = @
A +12 and 34 =
@ A 34 :
iei 1
Equivalently,

12 = iei
1 +12 and 34 = 1
ie i 34 :
Also, for later convenience, we write these in terms of as
12 = (2+ ) s^ +12 ;
12 = s^ (2 )
+12 ;
34 = (2 ) s 34 ;
34 = s (2+ )
34 ;
where s^ and s are constant spinors,

0 1 0 1
1 0A
s^ @ A and s @ :
0 1
In terms of these Killing spinors, the variation of scalars is given by
C1 = 2 34 2 ; C2 = 2 34 1 ;

(3.37)
C 1 = 2 12 2 ; C 2 = 2 12 1 :
Thus, recalling the factor of p1 in the denition of Q, its action on each scalar
2 2
is
QC1 = p12 (s 2+) 2 ; QC2 = p12 (s 2+) 1 ;

QC 1 = p12 (^s 2 ) 2; QC 2 = p12 (^s 2 ) 1;
prefactor in Q is precisely what is required to have iQ2 G = D G . This will become evident
5 The
after we derive (3.45).

giving
pi
0 1
0 (s 2+ )( 1 2 2 1 )
iQG = B
B
2 C
C
:
@ i
B C
p (^s 2 )( 1 2 2 1) 0 A
2
Therefore we can write L explicitly as
pi
0 1
B A0 1
2
(s 2+ )( 1 2 2 1 ) C
L= B C
;
pi (^s 2
B C (3.38)
@
2
)( 1 2 2 1) A02 A
where
A01 A1 + i j CiC j + 41 and A02 A2 + i j C j Ci 1

4
:
To check whether the variation of L is equal to the aforementioned superco-

variant derivative, we have to derive how each piece of L varies under Q. In what
1
follows we drop its p factor since it appears equally in every piece of the calcu-
2 2
lation and therefore its is meaningless. Let us begin by computing the variation
of the o-diagonal components of the superconnection. In particular, let us start
from the top-right component containing . Using (A.2), we have
4i 12 I
1 = 2i 12 D C2 (C2 C CI CI C I C2 )
k
8i 12 1
(C1 C C2 C2 C 1 C1 ) + 34 (C3 C 1 C4 C4 C 1 C3 ) 2i12 C2 ;
k
4i
2 = 2i 12 D C1 + 12 (C1 C I CI CI C I C1 )
k
8i 12 2
(C1 C C2 C2 C 2 C1 ) + 34 (C3 C 2 C4 C4 C 2 C3 ) + 2i12 C1 :
k
(3.39)
In L these appear multiplied by (s 2+), so it is worth listing the following prop-

erties of the projectors,
= x_ ; ( )2 = and + = + = 0 ;

to derive the relations
(s 2+ ) 12 = (s 2+ ) (2+ s^)+12 = 2(s 2+ s^)x_ +12 = 2x_ +12 ;
(s 2+ ) 12 = (s 2+ )(2+ s^)+12 = 2(s 2+ s^)+12 = 2+12 ;
(3.40)
(s 2+ ) 34 = (s 2+ )(2 s)34 = 0 ;
(s 2+ )12 = 12 12
+ = i+ :
With these in hand, we nd
1 2
(s 2+ ) = 4i+12 x_ D C2 (C2 C I CI CI C I C2 )
k
)
4 i
( C1 C 1 C 2 C2 C 1 C 1) + C2 ;
k 2
(
2
(s 2+ ) 2 = 4i+12 x_ D C1 (C1 C I CI CI C I C1 )
k
)
4 i
+ ( C1 C 2 C 2 C2 C 2 C 1) + C1 ;
k 2
such that the variation of the top-right component of L can be written as
p (
pi (s 2+ )( 1 2 2 1 ) =2 2+12 i D Ci

2
k
(Ci C I CI CI C I Ci )
2 )
4 i
+ (Ci C j Cj Cj C j Ci ) + Ci :
k 2
The expression above can be put in the form:
( )
2 4 i
i D Ci (Ci C I CI CI C I C i) + ( Ci C j C j Cj C j Ci ) + Ci
k k 2
( !
2i 1
= i @ Ci + i A1 x_ (CI C I 2Cj C j ) + Ci
k 4
!)
2i 1
iCi A2 x_ (CI C I 2Cj C j )
k 4
( !
2i 1
= i @ Ci + i A1 x_ (CI C I 2Cj C j ) + j k Cj C k + Ci
k 4
!)
2i 1
iCi A2 x_ (CI C I 2Cj C j ) + j k C k Cj :
k 4
From the rst to the second expression, we simply open up the denition of the
supercovariant derivative. Then, from the second to the third, we include the con-
tractions j k Cj C k and j k C k Cj since they cancel each other when multiplying

i Ci from the left and from the right, respectively. Finally, note that
(CI C I 2Cj C j ) = MJI CI C J and (C I CI 2C j Cj ) = MJI C J CI :
Thus we nd that the top-right component of L is such that

i
iQ( i Ci ) = p (s 2+ )( 1 2

2 1 )
2
p ( )
=2 2+12 i @ Ci + iA01 Ci iC A0i 2

(3.41)
p
= D 2 2+12 i Ci ;
where, as anticipated, the derivative is written in terms of the new dressed bosonic
connections of L.
We can carry out the same computation for the bottom-left entry containing
. In this case, from (A.2), we have
34 D C 2 + 4i
= 2i
1 34 (C 2 CI C I C I CI C 2 )
k
8i 12 3 4 4 3 34 1 2
+ (C C1 C C C1 C ) + (C C1 C C 2 C1 C 1 ) 2i34 C 2 ;
k
4i 34 1 I I 1
2 = 2i 34 D C 1 ( C CI C C CI C )
k
8i 12 3 4 4 3 34 1 2
+ (C C2 C C C2 C ) + (C C2 C C 2 C2 C 1 ) + 2i34 C 1 :
k
(3.42)
In L these appear multiplied by (^s 2 ), so we use the following relations
(^s 2 ) 34 = (^s 2 ) (2 s)34 = 2(^s 2 s)x_ 34 = 2x_ 34 ;

(^s 2 ) 34 = (^s 2 )(2 s)34 = 2(^s 2 s)34 = 234 ;
(3.43)
(^s 2 ) 12 = (^s 2 )(2+ s^)+12 = 0 ;
(^s 2 )34 = 34 = i34 ;
to write
(
2 2 I
(^s 2 ) 1 = 4i34 x_ D C 2 ( C CI C C I CI C 2 )
k
)
4 1 2 i 2
( C C1 C C 2 C 1 C 1 ) C ;
k 2
(
2 1 I
(^s 2 ) 2 = 4i34 x_ D C 1 ( C CI C C I CI C 1 )
k
)
4 i 1
+ (C 1 C2 C 2 C 2 C 2 C 1 ) C :
k 2
Then the bottom-left component of L varies as
p 34 (
pi (^s 2 )( 1 2 2 1) = 2 2 i i D C
2 i I
k
( C CI C C I CI C i )
2 )
4 i
+ (C i Cj C j C j Cj C i ) Ci :
k 2
Doing the analogous algebraic steps we did for , we nd
i
iQ( i C i ) = p (^s 2 )( 1

2 2 1)
2
p ( )
= 2 234 i @ C i iC i A01 + iA02 C i (3.44)
p
= D 2 234 ( i C i ) ;
and again the derivative is written in terms of the new dressed bosonic connections
of L.
Putting (3.41) and (3.44) together, we have that the variation of the o-
diagonal components of the superconnection are given by
0 p 1 0 1
0
(iQG ) = @ p 34 i
D 2 2+12 i Ci A @
0 D g : A (3.45)
D 2 2 i C 0 D g 0
Note that if we keep the p1 factor in the variation above, we recover precisely
2 2
(iQG ) = D (G ) and equivalently iQ2 G = D G .
Now we can move to the variation of the diagonal entries of L. In this case,
we already know by construction that for the supercharges we are considering
A1 = A2 = 0. Thus we only need to compute the variation of the piece arising
from G 2, namely

i j (Ci C j ) = i j (Ci )C j + Ci ( C j ) ;

i j (C j Ci ) = i j )Ci + C j (Ci ) :
j ( C
Using (3.37), we write them as

i j (Ci C j ) = 2 i j ik 34 k C j + Ci jk 12 k ;

i j (C j Ci ) = 2 i jk 12 C + C j 34 k :
j k i ik
Focusing on the rst expression, for instance, we can plug in the particular form
of the Killing spinors and anti-commute Grassmann-odd quantities to rewrite it
as

i j (Ci C j ) = 2 i j ik (s 2+ )34 k C j + Ci jk (^s 2 )+12 k
" !
p
pi (s 2+)ik i k

=i 2 234 j C j
2
p !#
pi

2 2+12 i Ci (^s 2 )jk j k :
2
We organized the expression above to make it easier to identify the quantities
!
p
pi

(s 2+ )ik i k = iQ( i Ci ) and 2 234 j C j = g ;
2
as well as
p !
pi

2 2+12 i Ci = g and (^s 2 )jk j k = iQ( j C j ) ;
2
to write
" #
i j ( Ci C j ) = i iQ ( i C i) g g iQ( j C j ) : (3.46)
Analogously, for the second expression, we nd
" #
i j Ci ) = i iQ( j C j ) g
j (C g iQ ( i Ci ) : (3.47)
Therefore, for the superconnection L in (3.38), we nd that its variation gives
0 ! 1
B i iQ ( i C i) g g iQ( j C j ) D g C
L =
B C
B
B
!C
C ;
@
D g i iQ( j C j ) g g iQ ( i Ci ) A
which exactly matches the supercovariant derivative of
0 1
0 gA
@ : (3.48)
g 0
Indeed,
0 1 0 1 " 0 1#
0 gA 0 gA 0 gA
D @ = @ @ + i L; @
g 0 g 0 g 0
0 1 "0 1 0 1#
=@
0 @ g A
+i @
A01 iQ( i Ci )A 0 g
;@ A
@ g 0 iQ( i C i ) A02 g 0
0 ! 1
B
B
i iQ ( i Ci ) g g iQ( j C j ) @ g + i (A0 g 1 gA0 )2 C
C
= B
B
!C
C
@
@ g + i (A02 g gA01 ) i iQ( j C j ) g g iQ( i Ci ) A
0 ! 1
B
B
i iQ ( i C i ) g g iQ( j C
j ) D g C
C
= B
B
!C
C :
@
D g i iQ( j C j ) g g iQ
( i C i)
A
The computation above was performed for supersymmetry transformations
generated by +12 and 34 with superconformal parameters xed as in (3.11). It
shows us that L preserves Q+12 iS12+ and Q34 iS34 separately, since for the
former we get
0 1
0 g
L = D @ A ;
0 0
whereas for the latter 0 1
0 0
L = D @ A :
g 0
Nevertheless, recall that our goal is to show that L is invariant under all super-
symmetry transformations preserved by L bos . Thus we still need to check whether
L is invariant under supersymmetry transformations generated by 12 and +34 .

As we did above, we will do this by carrying out a single computation to show
invariance of L under both Q12 + iS12 and Q+34 + iS34+ . Namely, we will use

Q0 = Q12 + iS12 + Q+34 + iS34
+ : (3.49)
For clarity, we will refer to the corresponding variations as 0.

First, turning on 12 and +34 implies on superconformal parameters 12 = i12
and 34 34
+ = i+ , such that the Killing spinors are given by
0 1 0 1
ie i 1
12 = @
A 12 and 34 = @
A +34 : (3.50)
1 iei
Equivalently,

12 = 1
ie i 12 and 34 =
iei 1 +34 : (3.51)
In terms of spinors s and s^, these can be written as

12 = (2+ ) s 12 ;
12 = s (2 ) 12 ;

34 = (2 ) s^ +34 ;
34 = s^ (2+ ) +34 :

With these in hand, we are ready to reproduce the analogues of (3.40) and (3.43).
For the Killing spinors above, we nd
(s 2+ ) 12 = (s 2+ ) (2+ s)12 = 2(s 2+ s)x_ 12 = 2ie i x_ 12 ;
(s 2+ ) 12 = (s 2+ )(2+ s)12 = 2(s 2+ s)12 = 2ie i 12 ;
(3.52)
(s 2+ ) 34 = (s 2+ )(2 s)+34 = 0 ;
(s 2+ )12 = ie i 12 = e i 12 :
while for the second we have
(^s 2 ) 34 = (^s 2 ) (2 s^)+34 = 2(^s 2 s^)x_ +34 = 2iei x_ +34 ;

(^s 2 ) 34 = (^s 2 )(2 s^)+34 = 2(^s 2 s^)+34 = 2iei +34 ;
(3.53)
(^s 2 ) 12 = (^s 2 )(2+ s)12 = 0 ;
(^s 2 )34 = iei 34 i 34
+ = e + :
Then, using (3.39), we have
(
2
(s 2+ ) 0 1 = 412 e i x_ D C2 (C2 C I CI CI C I C2 )
k
)
4 i
(C1 C 1 C2 C2 C 1 C 1) C2 ;
k 2
(
2
= 412 x_ D (e i C
2) (e i C
2 )C
IC
I CI C I (e i C2 )
k
)
4
i
C1 C 1 (e i C 2) (e i C
2 )C 1 C 1 + (e i C
2) ;
k 2
and
(
2
(s 2+ ) 0 2 = 412 e i x_ D C1 (C1 C I CI CI C I C1 )
k
)
4 i
+ ( C1 C 2 C 2 C2 C 2 C 1) C1
k 2
(
2
= 412 x_ D (e i C1 ) (e i C
1 )C
IC
I CI C I (e i C1 )
k
)
4
2 C2

i
+ (e i C
1 )C C2 C 2 (e i C
1) + (e i C
1) :
k 2
Note that, when going from the rst to the second equality, in both expressions,
we included the phase factor e i inside the derivatives, leading to the change of
sign of the i=2 factor. Thus, following the same steps as before, we obtain that
the top-right component of L varies as

i p
0 iQ( i Ci ) = p (s 2+ )( 1 0 2

2 0 1 ) = D 2 2i12 i (e i C )
i :
2
Furthermore, using (3.42), we have
(
2 2 I
(^s 2 ) 0 1 = 4ei 34 + x_ D C 2 ( C CI C C I CI C 2 )
k
)
4 1 2 i
( C C1 C C 2 C1 C 1 ) + C 2 ;
k 2
(
2 i 2 I
= 4+34 x_ D (ei C 2 ) ( e C ) CI C C I CI (ei C 2 )
k
)
4 1
i
C C1 (ei C 2 ) (ei C 2 )C1 C 1 (ei C 2 ) ;
k 2
and
(
2 1 I
(^s 2 ) 0 2 = 4ei 34 + x_ D C 1 ( C CI C C I CI C 1 )
k
)
4 i
+ (C 1 C2 C 2 C 2 C2 C 1 ) + C 1
k 2
(
2
= 4+34 x_ D (ei C 1 ) (ei C 1 )CI C I C I CI (ei C 1 )
k
)
4
i
+ (ei C 1 )C 2 C 2 C 2 C 2 (ei C 1 ) (ei C 1 ) :
k 2
Then the bottom-left component of L varies as
i p
0 iQ( i C i ) = p (^s 2 )( 1 0

2 2
0
1) = D 2 2i+34 i (ei C i ) :
2
Moving on to the diagonal entries of L, we have that

i j 0 (Ci C j ) = 2 i j ik 34 k C j + Ci jk 12 k ;
now gives

i j 0 (Ci C j ) = 2 i j ik (^s 2+ )+34 k C j + Ci jk (s 2 )12 k
" !
p
pi

=i (s 2+ )ik i k 2 2iei +34 j C j
2
p !#
pi

2 2ie i 12 i C (^s 2 )jk :
i j k
2
Above we used that (^s 2+ ) = iei (s 2+ ) and (s 2 ) = ie i (^
s 2 ). Also,
we organized the expression to make it easier to identify the quantities
!
p
pi

(s 2+ )ik i k = iQ( i Ci ) and 2 2iei +34 j C j g0 ;
2
as well as
p !
pi

2 2ie i 12 i C
i g0 and
2
(^s 2 )jk j k = iQ( j C j ) ;
to write
" #
i 0 j ) = i iQ( i Ci ) g0
j (Ci C g0 iQ( j C j ) : (3.54)
Analogously, for i j 0 (C j Ci ), we nd
" #
i j 0 (C j Ci ) = i iQ( j C j ) g 0 g0 iQ( i Ci ) : (3.55)
Thus we have that for this second pair of supercharges the analysis holds in
complete analogy to the one dervied for the rst pair. In this case, instead of
(3.48), we have that 0 L equates to the supercovariant derivative of

0 1
0 g0A
G 0 @ 0 : (3.56)
g 0
Therefore we nd that, as claimed from the beginning, the 2-node loop we con-
structed is 1=6 BPS and is indeed invariant under the same supercharges as the
1-node loops we started from.
Having carried out the explicit computation of supersymmetry invariance of
L, we hope to have convinced the reader about the power of the deformation
proposal. It does not only give a well dened procedure to nd fermionic operators,
but it also spares the long calculation above. That is, instead of performing the
explicit supersymmetry check of the superconnection, we simply need to nd Q

and G satisfying iQ2G = D G . Invariance of L under Q is then straightforwardly
guaranteed. On the one hand we have

L = L bos + iQG + G 2 = (iQG ) + f G ; Gg ;
and on the other we have
D (G ) = @ (G ) + i[L; G ] = @ (G ) + i[Lbos + iQG + G 2 ; G ]

= @ (G ) + i[L ; G ] [QG ; G ] :
bos
If iQ2 G = D G , it follows that (iQG ) = @ (G ) + i[L ; G ]. bos Then the two
equations above are indeed equal, since
f G ; Gg = G G + G G = QG G + G QG = QG G + G QG = [QG ; G ] ;

and [G 2 ; G ] = 0.
As for Q0, invariance of L under it follows because we have
0 L = 0 (iQG ) + f 0 G ; Gg ;
and
D (G 0 ) = @ (G 0 ) + i[Lbos ; G 0 ] [QG ; G 0 ] + i[G 2 ; G 0 ]:

What happens in this case is that 0 G = A G and G 0 = B G with
p
0 1 0 1
iei 0 ie i 0
A=@ i
A ; and B = 2 2@ A :
0 ie 0 iei
The matrices A and B are such that AG = G B and A G = G B . We then have
p
f 0G ; Gg = A G G + G A G = 2 2(AQG G + G AQG )
p
= 2 2( QG B G + B G QG )
= QG G 0 + G 0 QG
= [QG ; G 0 ] :
Together with the fact that [G 2 ; G 0 ] = 0, we see that the second pair of super-
charges is also preserved. More directly, one could already conclude this by noting
that G 0 is related to G by a gauge transformation.

In addition to what was outlined above, there are aspects about the construc-
tion that are worth mentioning. First, we nd that the resulting family of 1=6
i,
BPS loops is parametrised by i; i = 1; 2, and it can be explicitly written as
(3.35) with superconnection (3.38). We can identify the moduli space of 1=6 BPS
deformations by noting that any rescaling of and such that their product is
invariant can be absorbed by a constant gauge transformation. Thus the result-
ing manifold is spanned by four complex parameters minus the trivial action of
a constant gauge transformation acting on the superconnection as C . This is
the space of singular complex matrices C 4 ==C , which is the conifold.

6
Also, this
construction reproduces previous loops presented in the literature. In particular,
it matches Class I of [31], while Class II is obtained by breaking the other SU (2),
i.e., by coupling to C 4 in G . Therefore, the corresponding two
C3 , C 3 , C4 and
i
branches are spanned by constants i ; with i begin either 1; 2 or 3; 4. These
intersect at the origin singularity L , since for f i ; g ! 0 they approach the
i
bos
same point of the moduli space.
Through this deformation procedure, the scalar coupling of the bosonic con-
nections, MJI = JI 2Ji iI , gain an extra contribution of

! 1
2i
Mji k
i j :
In particular, when M has eigenvalues 2 and 0, the full matrix M + M has
enhanced SU (3) symmetry. Commuting the 4 preserved supercharges with this
SU (3) symmetry gives rise to 12 supercharges, so these operators are 1/2 BPS.
More precisely, they correspond to the loops of [24] apart from a gauge transfor-
mation = (2 )
8 3, as explained in the beginning of this section.
When we wrote the second chapter of [20], one question that was raised was
whether the deformation L = iQG + G 2 covers the full moduli space of 1=6 BPS
loops. To that we can already say that the answer is no. The reason is that we
found a new example of 1=6 BPS loop that is not connected to L bos through L.
We dedicate the next section to present this new operator.
6 We leave a more detailed discussion about these conical spaces to Section 4.2.1.
3.3 New 1=6 BPS operator 40
3.3 New 1=6 BPS operator

After extensively studying the moduli space of circular loops, which resulted in
the content presented in the previous section, we tried constructing a 1=3 BPS
WL. The motivation for this comes from the fact that Wilson loop operators are
expected to become vortex operators under mirror symmetry [35]. A 1=3 BPS loop
in ABJ(M) was never found, even though vortex operators preserving the same
amount of supersymmetry have already been constructed [36]. While seeking for
such operator, we ended up nding a new 1=6 BPS one instead.

The content of this section follows closely the original construction of the 1=2
BPS loop of [24]. We begin by dening the operator supported along the line in
7
Lorentzian space-time. Then we move to the one supported along the circle in
Euclidean space. These are not built in the language of Section 3.2, i.e., not as a
deformation of a composite bosonic connection. Quite oppositely, the construction
relies heavily on a particular guess of ansätze for the couplings. In this section
we simply derive the operators and leave a proper analysis of them to the next
chapter, in particular to Section 4.6 where we study their N = 4 analogues.
Line
Here we take the theory to be dened in Lorentzian space-time. The operator is
supported along the time-like line x = (; 0; 0) and it is written as

!
W = Tr P exp i L d :
The superconnection has the generic form
0 q 1
A1 + 2 M I C C J 2 I
L= @ q k
2
J I
I
k I

MIC JC
A :
k I
A 2 + 2
k J I
For the fermionic couplings we take the following ansatz
0 1

I 1I A
I = I1 I2 and = @ ;
2I
7 So far we only considered operators dened along closed curves , but recall that under a
conformal transformation a line is transformed into a circle. It is actually interesting that, even
though this is a symmetry of the theory, it changes the topology of the curve and, as it turns out,
also the expectation value of the WL. In the case of the 1=2 BPS WLs of N = 4 SYM, for example,
one nds that, whereas the straight line has trivial expectation value, the circular loop depends
on the coupling constant of the theory.
where and are constants. The matrix M M = diag( 1; +1; 0; 0).

is given by
At the beginning we expected +13 , +14 , 23 , 24 out

this operator to preserve
If this was the case, the set + , + , ,

13
of the 12 Poincaré supercharges.
14 23 24
would also be preserved since the operator is dened along the line, Poincaré
and superconformal parameters are preserved separately. Thus there would also
be 4 out of 12 superconformal supercharges and the operator would be 1/3 BPS.
However, in the end what we found was that the loop is 1=6 BPS. This fact will
become clear after we go through the computation and see that some constraints
between these parameters arise.
First, we need to compute the variation of the loop when the parameters +13 ,
+14 , 23 and 24 are turned on. We use the rules of (A.2), with the caveat that
Killing spinors
are replaced by Poincaré parameters and, in the variation of
fermions, we lose the linear term. For superconformal transformations, we have
the same rules with now replaced by . Thus in the following we perform the
and the proof for follows straightforwardly.
computation for
We begin by computing the variation of the top-right entry of the supercon-
nection, f . Using (A.2), it takes the following form
( )
4i 1I 8i IJ
f = 2i1I ( )+ D CI (C C J CJ CJ C J CI ) C C 1 CJ
k + I k + I
( )
4i 2I J 8 i
2i2I ( ) D CI (CI C CJ CJ C CI )
J CI C CJ :
IJ 2
k k
The equation above holds for any Poincaré parameter. Now looking at the set of
parameters we want to turn on, namely +13 , +14 , 23 and 24 , the equation above
becomes
( )
4i 8i
f = +1j 2i( )+ + D Cj (Cj C J CJ CJ C J Cj ) (C1 C 1 Cj Cj C 1 C1 )
k k
( )
4i J 8i 2
2j 2i( ) D Cj ( C C CJ CJ C J Cj ) (C C C Cj C C2 ) ;
2
k j k 2 j
where lower case indices i; j belong to the set f3; 4g. From now on, adopt this
notation throughout the section. Note that (i ) +

+ = 0 and (i ) = 0 , so
( )
4i 8i
f = +1j 2D0 Cj ( Cj C J C J CJ C J Cj ) ( C1 C 1 C j Cj C 1 C 1)
k k
( )
4i 8i
2j 2D0 Cj (Cj C J CJ CJ C J Cj ) (C2 C 2 Cj Cj C 2 C2 )
k k
( ! " ! !
2i
= 2 +1i + 2i D0 Ci +1i + 2i Ci ( C 1 C 2 C2 ) +
1+C +1i 2i
k
# " ! ! #)
2i
Ci (C j C j) + ( C1 C 1 + C2 C 2 ) +1i + 2i Ci + (Cj C j ) +1i 2i Ci :
k
To combine this as a covariant derivative with a dressed bosonic connection
we impose that
+1i = 2i : (3.57)
This means that the loop is actually 1=6 BPS, since it is supersymmetric if we
turn on +13 ; +14 ; 23 ; 24 and also impose the constraint above. In particular, in this
case it follows that

( ! ! )
2i 2i
f = 4+1i D0 Ci Ci C 1 C1 + C 2 C2 + C1 C 1 + C2 C 2 Ci :
k k
Using the denition of the covariant derivative with dressed bosonic connection,
D0Ci = @0Ci + i(A1Ci CiA2), we have simply

f = D0 4+1i Ci : (3.58)
For the bottom-left entry of the superconnection, f , we go through the

same steps as f and nd that its variation gives

f = D0 4 2i C i : (3.59)
For the diagonal entries, we start from the top-left one. We have A1 A1 +
2
MJI CI C J and its variation gives as
k
"
4i 2j + i
A1 = (j Mji )Ci 2+ 1j (ji + Mji )Ci 1 1+i (ji + Mji ) +1 C j
k
#
+ 2i (ji Mji ) 2 C j :
Using the fact that Mji = 0 and imposing (3.57), it becomes

" #
4i 1i
A1 = + Ci ( +
1 2 ) 2i ( +1 2 )C i : (3.60)
k
Analogously, for A2 A2 + 2k MJI C J CI the variation is given by

" #
4i 1i
A2 = + ( +
1 2 ) Ci 2i C i ( 1
+
2 ) : (3.61)
k
With (3.58) and (3.59) in hand, we see that L should be equal to the super-
covariant derivative of
0 q 1
0 4 2 1i C
@ q k + iA :
4 2k 2i C i 0
This means that we expect L to be equal to D0 acting on this matrix, i.e.,
0 q 1 0 q 1
A1 2 f
k i 2 (fg
k gf ) D0 g
@q A = @ q A : (3.62)
2 f
k A2 D0g i 2 (fg
k gf)
q q
g = 4 2 1i C g = 4 2 i.
with
k + i and
k 2i C O-diagonal equalities are thus au-
tomatically satised. As for the diagonal ones, we plug g and g into the equations
for A1 and A2 ,
s s
2 2
A1 = i (f g gf ) and A2 = i (fg gf) ;
k k
1
and it gives us the condition = .
2
Summarizing, we nd that the superconnection for this line operator is given
by
0 q 1
A1 + 2 M I C C J 2 I
L= @ q k
2
J I
I
k I

MIC JC
A ;
k I
A 2 + 2
k J I
with
! !
I = I1 + 2
I ; I = I +
1 I
2 ; MJI = diag( 1; +1; 0; 0) ;
1
and = . The resulting loop is invariant under supersymmetry transfor-
2
mations generated by +13 , +14 , 23 and 24 subject to the constraint +1i = 2i .
Therefore the operator is 1=6 BPS.
Circle
To construct the loop supported along the circle we go to Euclidean space. The
operator is supported along the curve C and it is dened as
W = Tr P exp i L d
where
0 q 1
2i I J 2 I
L = @A1x_ q 2k MJ CI I C
i k I

2i M I C JC
A :
i k I A2 x_ k J I
For the couplings we make the following ansantz,
I = 1 ( )(+ )+ I1 2 ( )( ) I2 ;

I = i1 ( )(+ ) + 1I i2 ( )( ) 2I ;
MJI = diag( 1; +1; 0; 0) ;
where 1 ( ) and 2 ( ) are functions of .

This loop is constructed based on the 1=6 BPS line. For the line, we found
that the parameters that should be turned on were +13 ; +14 ; 23+ ; 24+ subject to
the constraint +1i = 2i . For the circle, we will turn on the same supercharges but
with both chiralities, i.e., we will turn on 13 ; 14 ; 23 ; 24 . As happened for the
line, for the circular case we also expect some constraint between these parameters
to arise.
First, let us go through the preliminary choices. We begin by choosing the
following form for the superconformal parameters:
1i = i3 1i and 2i = i3 2i : (3.63)
With this choice, it follows that
1i = 2(+ )1i
and 2i = 2( )2i :
(3.64)
Since we will need to use it later, it is also useful to write down
1i = i1i 3 ; 2i = i2i 3 and 1i = 21i (+ ) ; 2i = 22i ( ) :
Moreover, we will need the following property
( ) ( ) = ( )x_ : (3.65)
Now let us compute the variation of the loop when the parameters 1i and 2i
are turned on. Following the same steps taken for the line operator, we will sketch
each piece of the variation separately. We begin with the variation of f and
see that it takes the following form
f = 1 ( )(+ )+ 1 2 ( )( ) 2
(
1I ( ) D CI 4i 1I J
= 1 ( )(+ )+ 2i (CI C CJ CJ C J CI )
k
) (
8i IJ 1 2I ( ) D CI
CI C CJ 2i1I C I 2 ( )( ) 2i
k
)
4i 2I J 8i IJ 2
(CI C CJ CJ C J CI ) CI C CJ 2i2I C I :
k k
Using (3.63), (3.64) and (3.65), the expression above becomes
(
8i 1i
f = 1 ( )(+ )+ 4i1i x_ D Ci (Ci C J CJ CJ C J Ci )
k
)
16i 1i
(C1 C 1 C i Ci C 1 C 1) 2i( i3 ) 1i C i
k
(
8i 2i
2 ( )( ) 4i2i x_ D Ci (Ci C J CJ CJ C J Ci )
k
)
16i 2i
C 2 C C 2 C 2 ) 2i(i3 ) Ci :
(C2 Ci 2i
i
k
To be able to sum everything as a single covariant derivative we impose
1 ( )(+ )+ 1i = 2 ( )( ) 2i : (3.66)
This is our rst constraint on the circle. Once satised, we have
( !)
2
f = 8i1 ( )(+ )+ 1i D Ci C( C 1 C + C 2 C
2 ) ( C1 C 1 + C2 C 2 )C
k i 1 i
21 ( )(+ )+ (3 ) 1i Ci 22 ( )( ) (3 ) 2i Ci

! !
= 8iD 1 ( )(+ )+ 1i Ci 8i@ 1 ( )(+ )+ 1i Ci 21 ( )(+ )+ (3 )

1i Ci 22 ( )( ) (3 ) 2i Ci :
By requiring that

8i@ 1 ( )(+ 1i )+ Ci + 21 ( )(+ 3 1i )+ Ci + 22 ( )( 3 2i ) Ci = 0 ;
(3.67)
we have
!
f = 8iD 1 ( )(+ )+ 1i Ci : (3.68)
Equation (3.67) is our second constraint on the circle. One possible solution to it
is
1 ( ) = ei=2 ; 2 ( ) = e i=2 ; 1i = i2i and +1i = i+2i : (3.69)
Note that this solution is consistent in the sense that it does not violate (3.66).
Moving on to the derivation of f ( ), we follow the same steps we did

for f. We come across an analogous version of (3.66),
1i (+ ) + 1 ( ) = 2i ( ) 2 ( ) : (3.70)
We also come across and analogous version of (3.67),

8i@ (1i + )+ 1 ( ) C i 2(1i 3 + )+ 1 ( )C i 2(2i 3 ) 2 ( )C i = 0 :
These conditions are automatically satised as a consequence of (3.69), since it
implies that
1 ( ) = e i=2 ; 2 ( ) = ei=2 ; 1+i = i2+i and 1i = i2i :
Thus we nd
!
f = 8D 1 ( )1i (+ ) + i

C : (3.71)
2i
Now the variation of A1 A1x_ k
MJI CI C J can be written in terms of 1i
and 2i as
" # " #
8i 8i
A1 = Ci 1 (+ )
1i
2 ( ) +
2i 1i (+ ) 1 2i ( ) 2 C i : (3.72)
k k
2i
Analogously, the variation of A2 A2x_ k
MJI C J CI can be written as
" # " #
8i 8i i
A2 = 1 (+ )1i 2 ( )2i Ci + C 1i (+ ) 1 2i ( ) 2 : (3.73)
k k
With all variations in hand, we can check whether L equates to the superco-
variant derivative of the block o-diagonal supermatrix
0 1 s s
0 gA 2 2
@ ; with g=8 1 ( )(+ )+ 1i Ci ; g = 8i 1 ( )1i (+ ) + C i :
g 0 k k
(3.74)
This means checking that the diagonal entries of
0 q 1 0q 1
A1 i 2 f
k
2 (fg
k gf ) D0 g
@ q A =@ q A ;
i 2k f A2 D0g 2 (fg
k gf)
hold. We nd that
s s
2 2
A1 = (f g gf ) and A2 = (fg gf) ;
k k
hold automatically if we use (3.66) and (3.70).
Summarizing, the superconnection for the circular Wilson loop operator is
0 q 1
2i I J 2 I
L = @A1x_ q 2k MJ CI I C i k I

2i M I C
A ; (3.75)
i k I A2 x_ k J
JC
I
with
I = ei=2 (+ )+ I1 e i=2 ( ) I2 ;

I = ie i=2 (+ ) + 1I iei=2 ( ) 2I ;
MJI = diag( 1; +1; 0; 0) :
By imposing 1i = i3 1i , 2i = i3 2i , 1i = i2i and +1i = i+2i , the supercon-
nection is such that L = D ( G ) with
s 0 1
2 @ 0 ei=2 (+ )+ 1i Ci A
G = 8 ;
k ie i=2 1i (+ ) + C i 0
where the variation is taken with respect to the parameters 1i . Therefore the loop
is invariant under the action of supersymmetry transformations generated by,
(Q+1i iS1+i ) + i(Q+2i + iS2+i ) and (Q1i + iS1i ) i(Q2i iS2i ) ; (3.76)
with i = 3; 4 and it is 1=6 BPS. As in the case of the line, we see from the scalar
coupling matrix M that the loop has a residual U (1) U (1) R-symmetry. In the
context of the AdS/CFT correspondence, this operator should then have a dual
description on the string theory side with U (1) U (1) bosonic symmetry.
Finally, it is worth mentioning that this new operator does not belong to any
previously known examples and, in particular, not to any classication of [49].

3.4 Perturbative detour: deforming the 1/2 BPS loop 48
3.4 Perturbative detour: deforming the 1/2 BPS

loop
The 1/2 BPS loop of ABJ(M) theory was presented in Section 3.1, see (3.19). Here
our goal is to compute the expectation value of a deformed version of this operator
perturbatively. We should clarify that by deformation we no longer mean the one
built out of G , namely L = iQG + G 2, but actually the one where we change the
contour along which the loop is supported.
The deformation we consider is analogous to the one introduced in [37] in the
context of N = 4 SYM. The idea is to deform the circle C by making small
perturbations parameterised by a real valued function g ( ). Then the closed path
C~ is given by
C~ = (0; eg( ) cos( ); eg( ) sin( )) : (3.77)
We carry the perturbative computation by expanding the observable not only in
the t'Hooft coupling = N=k, but also in the deformation parameter g ( ). In
particular, we write the deformation parameter in terms of its Fourier series,
1
X
g ( ) = bn ein : (3.78)
n= 1
The Wilson loop operator we are interested in is dened in terms of the super-
connection
0 q 1
L( ) = @ q
A1 i k jx_ jI
2 I
; A (3.79)
i 2k jx_ j I I A2
with dressed bosonic connections
2i 2i
A1 = A1x_ k
jx_ jMJI CI C J and A2 = A2x_ k
jx_ jMJI C J CI : (3.80)
In the case of the perfect circle, i.e., when g( ) = 0, the fermionic couplings are
0 1
ie i=2
I ( ) = ei=2 ie i=2 I1 and I ( ) = @ A 1I : (3.81)
ei=2
The scalar coupling is M = diag( 1; 1; 1; 1). These couplings are such that the
traced loop is 1/2 BPS [24]. Here we are interested on a generically deformed
circle, where g( ) 6= 0. We keep the scalar coupling as the original one, but
consider new fermionic couplings given by
0 1
i=2
g( )=2 @ie

I ( ) = eg( )=2 ei=2 ie i=2 I1 and I ( ) = e A 1I : (3.82)
ei=2
Since the couplings depend on the loop's contour, it is natural to expect them
to change when the contour is deformed. However, it is not completely clear
why they should take the form in (3.82). The motivation to make this particu-
lar choice comes from the Feynman diagram computation, which we will present
subsequently. We expect hW i at second order in the deformation parameter g( )

(and consequently at second order in bn ) to behave as
1
hW ijb n(n2 1)jbn j2 :
X
2
n=2
This particular form is expected because it gives what is called the Bremsstrahlung
function, denoted as B (). We will explain the meaning of B () at the end of
this section, but for now let us just blindly assume that the behavior above should
hold. We nd that this is obtained for couplings in (3.82) and we rely on this
result to justify our choice. Nevertheless, a deeper understanding of its structure
is still lacking.
With the given denitions the expectation value of the Wilson loop reads
!
hW i = N +1 N ; ] e
D[A1 ; A2 ; C; C; S Tr P exp i L( ) d : (3.83)
1 2 C~
In the following we present the results obtained so far up to two loops, i.e., evalu-
ating hW i up to second order in 1 = N1=k and 2 = N2=k. We use the Feynman

rules shown in Appendix A.
3.4.1 1-loop contribution
A 1-loop computation of the Wilson loop requires expanding the path-ordered
exponential in (3.83) up to second order:
! !
Tr P exp i L( ) d = Tr 1 i d1 L(1 ) + ( i)2 d1>2 L(1 )L(2 ) :

(3.84)
Here the trace is taken over the corresponding representation. In particular, the
superconnection L is a (N1 + N2) (N1 + N2) dimensional supermatrix. The upper

left entry has dimensions N1 N1 and the bottom right N2 N2 . These diago-
nal blocks belong to the adjoint representation of U (N1 ) and U (N2 ) respectively.
Moreover, the term quadratic in L reads
0 s !1
2 2
B
B
A1(1)A1(2) k
jx_ 1jjx_ 2j( )1( )2 i
k
jx_ 2jA1(1)( )2 + jx_ 1j( )1A2(2) C
C
B s ! C,
B
2 2 C
@
i
k
jx_ 1j( )1A1(2) + jx_ 2jA2(1)( )2 A2(1)A2(2) k
jx_ 1jjx_ 2j( )1( )2 A
where we have introduced the shorthand notation ( )i I (i ) I (i ) and ( )i
I (i ).
I (i )
Plugging (3.84) in (3.83), we get three non-vanishing correlators. In what
follows, we focus on the upper N1 N1 block contribution of the supermatrix,
which is the U (N1 ) sector. The result for the N2 N2 block can be obtained from
that simply by exchanging N1 $ N2 in the nal result.
For the N1 N1 block, we have three dierent diagrams coming from Wick
contracting the elds. These are shown in Figure 8.
(a) (b) (c)
Figure 8: One loop diagrams: (a) the scalar tadpole, (b) the single gluon exchange
and (c) the single fermion exchange. The blue wavy line represents the contour,
the black solid line is the scalar propagator, the black wavy line is the gauge
propagator and the black dashed line is the fermion propagator.
The rst diagram represents the exchange of a single scalar coming from A.
This term contains C (x)C (x), which generates a scalar tadpole graph. The
second one is the exchange of a gauge boson coming from the term quadratic on
A. Schematically we have
A(1)A(2) AA x_ (1)x_ (2) + scalars ; (3.85)
and the gauge eld Wick contraction comes from the rst term on the right-hand
side, generating the single gauge eld exchange. Finally, the third diagram is the
single fermion exchange that comes from Wick contracting ( )1 ( )2 . This term
arises from the piece that is quadratic in L.

For a circular contour, the contribution of each diagram was computed in [38].
There the integrals were computed using dimensional regularisation with a di-
mensional reduction scheme (DRED). This scheme preserves gauge invariance and
supersymmetry of Chern-Simons theories up to two loops [39] and we will also use
it as one of the techniques to compute our integrals.
In the DRED scheme, we perform tensor manipulations and apply the re-
spective Feynman rules in three dimensions. Then, once the loop integrals are
obtained, we promote them to 2) dimensions. There is a prescription to

(3
contract three dimensional object with (3 2) dimensional ones. For example,
whenever we have a three dimensional metric
contracted with a (3 2) di-
mensional one ^ , we assign the following rules:
= 3 ; ^ ^ = 3 2 ; ^ = ^ : (3.86)
Let us briey resume the results obtained for the circular case, i.e., for the
non-deformed contour. The scalar tadpole vanishes in dimensional reduction.
The single gluon exchange also vanishes, not due to DRED, but due to the anti-
symmetry of the tensor contracted with three vectors lying on the plane. The
single fermion exchange is a bit more intricate, but in the end of the day it gives
no nite contribution.
For the deformed circle, both scalar tadpole and single gauge exchange dia-
grams must still evaluate to zero. In the rst case because the contour does not
change the tadpole nature of the scalar contraction and in the second because the
deformation keeps the contour lying on the equatorial plane. Therefore, the only
possibly non-vanishing contribution must come from the single fermion exchange.
The single fermion exchange contribution is given by
* +
2
Tr d1>2 jx_ 1 jjx_ 2 j( )1 ( )2 ; (3.87)
k
where
( )2 = ( I )îj (2 ) I (2 ) ; (3.88)

^
( )1 = J (1 )( J )jk (1 ) ; (3.89)
and we introduced the short-hand notation xi x(i ). Using the Feynman rules
of Appendix A, we have
* +
2
Tr d1>2 x_ 1 x_ 2 J (1 )( J )^jk (1 )( I )îj (2 )I (2 )
j jj j
k

2 ( 3 ) (x(1 ) x(2 ))
= iN1 N2 2 d1>2 jx_ (1 )jjx_ (2 )j(1 2 ) ;
3.4 Perturbative detour: deforming the 1/2 BPS loop
blocks belong to the fundamental representation of U (N1 ) and U (N2 ) respectively.

50
k 2
3
2 [(x(1 ) x(2 ))2 ]3=2
( x) U (N 2 )
0 s !1
2 2
B
B
A1(1)A1(2) k
jx_ 1jjx_ 2j( )1( )2 i
k
jx_ 2jA1(1)( )2 + jx_ 1j( )1A2(2) C
C
where is the Euler Gamma function. The sector contributes with

B s C
C,
!
B
2 2
@
i jx_ j( )1A1(2) + jx_ 2jA2(1)( )2
k 1
A2(1)A2(2) k
jx_ jjx_ j( ) ( )
1 2 1 2
A
I (i ).
I (i )
Plugging (3.84) in (3.83), we get three non-vanishing correlators.
exactly same integral, so the single fermion exchange diagram is twice the result
In what
U (N1 ) sector. N2 N2 block can be obtained from

above:
which is the The result for the

N1 N1

For the block, we have three dierent diagrams coming from Wick
( 3 )
4 (x(1 ) x(2 ))

= iN1 N2 2 d1>2 jx_ (1 )jjx_ (2 )j(1 2 ) :
(a) (b) (c)
k 2
3
2 [(x(1 ) x(2 ))2 ]3=2
g( ) that deform the

We want to evaluate this integral for small perturbations

C~ ,
The rst diagram represents the exchange of a single scalar coming from A.
C (x)C (x),
This term contains
circle. We consider the contour
which generates a scalar tadpole graph.
The
assume the couplings (3.82) and expand the
A(1)A(2) AA x_ (1)x_ (2) +integrand
scalars ; (3.85) in powers of g ( ) . The 0-th order term gives us an integral of the form

1
( )1 ( )2 .
d1>2
single fermion exchange that comes from Wick contracting This term
;

sin2 1 2 2
that matches the one obtained for the circular contour, see equation (3.2) of [38].
As mentioned before, when DRED is used, it gives a contribution proportional
to the regularisation parameter and it goes to zero for ! 0. The 2-nd order
expansion in g( ) gives a sum of terms of the form

ein +in
1 1 2 2
d1>2 ; (3.90)
(ei ei )4
1 2
for integer valued n1 and n2 . This structure appears also in [37] and there the
authors propose a recursive method to solve this kind of integral. We refer to this
method as the Ap method and a review of it can be found in Appendix B.

Thus we can compute the contribution of the single fermion exchange diagram
either through DRED, as in [38], or using the Ap method, as in [37]. Using DRED,
we nd that (c) gives a nite result proportional to
1
n(n2 1)jbn j2 ;
X
(3.91)
n=2
wheren are the Fourier modes and bn the Fourier coecients of the deformation
p
parameter g ( ). To double check this result, we also use the A method, namely
rule (B.4), to compute (c). In this case, we nd a result proportional to
1 1
1)jbn j + 16 J4(2) 1)jbn j2 ;
X X
n(n2 2 (n 2 (3.92)
n=2 n=2
where J4(2) is the arbitrary and possibly innite constant Jn(2)+n 1 2

for n1 + n2 = 4 that
arises through the Ap method. Note that, compared to the result found through
DRED, there is an extra piece proportional to J4(2) that was not expected.
What is expected from the Ap method is a nal result that is independent of
constants. These J 's account for innite contributions and can appear individually
for each diagram but they are expected to cancel after summing all diagrams
arising in the perturbative computation. So, to investigate the appearance of J4(2) ,

we went one step back in our analysis and reconsidered the 1-loop diagrams (a)
and (b). The idea was to check whether they could contribute with a constant
piece that would cancel the one present at the single fermion exchange result. For
D E
the scalar tadpole (a), we have a contraction of the form
C (x)C (x) , where
blocks belong to the fundamental representation of U (N1 ) and U (N2 ) respectively.
1
hC (x)C (y)i :
0 s !1
2 2
B
B
A1(1)A1(2) k
jx_ 1jjx_ 2j( )1( )2 i
k
jx_ 2jA1(1)( )2 + jx_ 1j( )1A2(2) C
C
[(x y )2 ]
B s C
C,
!
2 2
1
B
@
i jx_ j( )1A1(2) + jx_ 2jA2(1)( )2
k 1
A2(1)A2(2) jx_ jjx_ j( )1( )2
k 1 2
A
2
I (i ).
I (i )
y = x times jx_ j, i.e.,

The integral for the tadpole U (is

N ) given by
N N this propagator at
which is the 1 sector. The result for the 2 2 block can be obtained from

N1 N1
2
For the block, we have three dierent diagrams coming from Wick
jx_ j
d :
[(x x)2 ]
1
(a)
0
(b) (c)
2
We choose to rewrite it as
2 A.
jx_ j
The rst diagram represents the exchange of a single scalar coming from
!
1 2 1 (x2 x1 )
jxx__1j + jxx__2j jx_ 1jjx_ 2j d2d1 ;
d
0 [(x(0) x)2 ]
1
2 2 0 0 [(x2
A(1)A(2) AA x_ (1)x_ (2) + scalars ;
x1 ) ]
2 1
(3.85)
3
22
single fermion exchange that comes from Wick contracting ( )1 ( )2 . This term
L.
giving two integrals that we need to solve:
arises from the piece that is quadratic in
2
I0
jx_ j d ;
[(x(0) x)2 ]
1
0 2
2 !
( x 2 x1 )
I1
1
jxx__1j + jxx__2j jx_ 1jjx_ 2j d2d1 :
1
2 0 0 [(x2 x1 )2 ] 3
2 1 2
For the single gluon exchange (b) we still get a null result due to a propagator
containing the contraction (x y) that vanishes for the planar contour C~ we
are considering so we do not worry about it.
Note that I0 is a single integral, whereas the Ap method solves double integrals.
p
To solve it we developed another recursive method, named B , which is the single
p
integral analogue of the A method. The I1 term, on the other hand, is a double
p
integral, so in principle it could be solved using A symbols. However, at second
order in g ( ), we nd terms of the form (3.90) with half-integer valued n1 and n2 .
p
So we generalised the A method for half-integers to solve I1 . These developments
are all presented in Appendix B. In particular, using (B.10) and (B.6) we nd
(quite non-trivial) nite results for both I0 and I1 that precisely cancel each other,
" ! !#
(1 + 4n2 ) 1 1
I0 = HarmonicNumber n + HarmonicNumber +n
4 2 2
log(2) n2 ( 3 + log(16)) = I1 :
J4(2) appearing
Therefore we nd no innite constants that could possibly cancel
p
in the single fermion exchange diagram when computed through the A method,
see (3.92).
Since reconsidering the scalar tadpole diagram was not enough to clarify the
presence of this constant, we took yet another step back and applied these recursive
methods to compute the integrals for the non-deformed loop. In this case, we nd
! detour: deforming the 1/2 BPS loop

1
3.4 Perturbative 50
I0 = 2 PolyGamma 0; ;blocks belong to the fundamental representation of
L reads
U (N1 ) and U (N2 ) respectively.
2
Moreover, the term quadratic in
0 s !1
2 2
B
B
A1(1)A1(2) k
jx_ 1jjx_ 2j( )1( )2 i
k
jx_ 2jA1(1)( )2 + jx_ 1j( )1A2(2) C
C
" ! !#
B s C
C,
!
2 2
1 3
B
@
i jx_ j( )1A1(2) + jx_ 2jA2(1)( )2
k 1
A2(1)A2(2) jx_ jjx_ j(
k 1 2
)1 ( )2 A
I1 = 2 2 K2(1) + 2 where we have introduced the shorthand notation
PolyGamma 0; ( )i I (i ) I (i ) and ( )i

PolyGamma 0; ;
2 2
I (i )
I (i ) .

follows, we focus on the upper
50
N1 N1 block contribution of the supermatrix,
which is the 1 sector. The result for the 2 U (N )

2 block can be obtained from N N
1 and 2 respectively. U (N ) U (N )
N $N
blocks belong to the fundamental representation of
L
that simply by exchanging 1 2 in the nal result.
N N
Moreover, the term quadratic in reads
For the 1 1 block, we have three dierent diagrams coming from Wick
giving
0 s !1
2 2contracting the
B
B
A1(1)A1(2) k
jx_ 1jjx_ 2j( )1( )2 i
k
jx_ 2jA1(1)( )2 +elds.
jx_ 1j( These
)1 A2 (2are
) C
C
shown in Figure 8.
B s ! C,
2 2
= 2 2 K2(1) ;
B C
@
i jx_ j( )1A1(2) + jx_ 2jA2(1)( )2
k 1
A2(1)A2(2) jx_ jjx_ j( )1( )2
k 1 2
A
I (i ).
I (i )
follows, we focus on the upper N1 N1 (a)

block contribution of the supermatrix,
(b) (c)
U (N1 ) sector. 2 N2 block

NFigure
whereas for the single fermion exchange we nd
which is the The result for the 8: Onecan bediagrams:
loop obtained from
(a) the scalar tadpole, (b) the single gluon exchange
N $N
that simply by exchanging 1 2 and
in the nal (c) the single fermion exchange. The blue wavy line represents the contour,
result.
For the 1 N N the black solid line is the scalar propagator, the black wavy line is the gauge
1 block, we have three dierent diagrams coming from Wick
A.
= 2 2 J2(1) :
The rst diagram represents the exchange of a single scalar coming from
(a) (b) (c) A(1)A(2) AA x_ (1)x_ (2) + scalars ; (3.85)
K2(1) and J2(1) are the ones arising from the respective substitution rule,
Constants singlepropagator.
propagator and the black dashed line is the fermion fermion exchange that comes from Wick contracting

( )1 ( )2 . This term
The rst diagram represents the exchange of aFor a circular

single scalar contour, the contribution
coming from . of each diagram was computed in [38]. A
C ( x ) C ( x)
(B.5) and (B.3). Therefore, not even for the circular loop we recover a nite result
This term contains , which generates a scalar tadpole graph.
The
that is free from constants.

A( )A( ) AA x_ ( )Since
x_ ( ) + ; 1 2 1 2 scalars
J 's and K 's are not related to each other,
(3.85)
we
cannot arm that the contributions

( ) ( ) from (a) and (c) cancel each other and give
single fermion exchange that comes from Wick contracting 1 2. This term

a null result. Thus we argue that setting the constants to zero should be scheme
equivalent to DRED. This way we recover hW i 1-loop

= 0 for the circular loop and
for the deformed loop we nd
1
hW i n(n2 1)jbn j2 :
X
1-loop
n=2
3.4.2 2-loops contribution
A computation of the expectation value of the Wilson loop at 2-loops requires
expanding the path-ordered exponential up to fourth order:
! "
Tr P exp i L( ) d
= Tr 1 + ( i) d L( ) + ( i)2 d1>2 L(1 )L(2 )
#
+( i)3 d1>2>3 L(1 )L(2 )L(3 ) + ( i)4 d1>2>3>4 L(1 )L(2 )L(3 )L(4 ) :
In particular, the N1 N1 block has the following form

" !
2
Tr 1 i d A1 d1>2 A1 (1 )A2 (2 ) jx_ 1j( )1jx_ 2j( )2
k

2
+i d1>2>3 A1 (1 )A1 (2 )A1 (3 ) jx_ 1j( )1jx_ 2j( )2A1(3)
k
!
2 2
k
A1(1)jx_ 2j( )2jx_ 3j( )3 k
jx_ 1j( )1A2(2)jx_ 3j( )3

2 2
+ d1>2>3>4 A1 (1 )A1 (2 )A1 (3 )A1 (4 ) + ( )1 ( )2 ( )3 ( )4
k
2 2
A ( )A ( )jx_ j( )3jx_ 4j( )4
k 1 1 1 2 3
A ( )jx_ j( )2A2(3)jx_ 4j( )4
k 1 1 2
2 2
jx_ j( )1A2(2)A2(3)jx_ 4j( )4
k 1
A ( )jx_ j( )2jx_ 3j( )3A1(4)
k 1 1 2 !#
2 2
jx_ j( )1A2(2)jx_ 3j( )3A1(4)
k 1
jx_ j( )1jx_ 2j( )2A1(3)A1(4) :
k 1
In this case, there are six dierent diagrams to consider and these are represented
in Figure 9. Three of them are purely bosonic and we will refer to those as the
bosonic sector. The other three contain fermions and will be referred to as the
fermionic sector.
Bosonic sector
Diagram (a) in Figure 9 comes from the third order expansion of the WL con-
8
tracted with the gauge cubic vertex . Its value depends only on the topology of
the loop. Since our deformation preserves the unknot structure of the contour, we
can borrow the result from the literature [40] and, for the N1 N1 block, we have
N1 (N1 2 1) 2
:
N1 + N2 6k2
8 The interaction vertices appearing in the action are shown in Appendix A
(a) (b) (c)
(d) (e) (f)
Figure 9: Bosonic (a-c) and fermionic diagrams (d-f) from the 2-loop expansion.
Considering also the N2 N2 block, we get that the rst diagram gives
N2 (N2 2 1) + N1 (N1 2 1) 2
= :
(a)
N1 + N2 (b)
6k2 (c)
Diagram (b) comes from the bosonic part of the quadratic term in the WL
expansion with a one-loop gauge propagator inserted. More specically, it comes

(d) (e) (f)
from the one-loop insertion in Figure 9: Bosonic (a-c) and fermionic diagrams (d-f) from the 2-loop expansion.
N2 N2 block, we get that the rst diagram gives

Considering also the
N2"
(N2 2 1) + N1 (N1 2 1) 2 #
= :
N1 + N2 6k 2

Tr d1>2 A1 (x1 )x_ 1 A1 (x2

)x_
2 ;
from the one-loop insertion in
" #
Tr d1>2 A1 (x1 )x_ 1 A1 (x2 )x_ 2 ;
which gives
which gives ) 2 x_ 1 x_ 2

2( 1
x_ 1 x_ 2
2 N N d1>2 2 :
k 2 1 2 1 2
2( 1
[(x1 x2 )2 ]1
Summing the
)
N2 N2 contribution to it, we get
2
2 Nd N
d1>2 2 :
k2N+ N1 k2 1[(x x ) 2]
N N + N N ( )
2 2x_ x_
2 1
[(x1 x2 )2 ]1
= 1 2 2 1 2 : 1 2
2 1 2 1>2 2 1 2 (3.92)
1 2 1 2
As in the single fermion exchange diagram, the second order expansion of the
Summing the N2 N2 contribution to it, we get

56
N12 N2 + N22 N1 2 ( 12 ) x_ 1 x_ 2
= d1 >2 2 : (3.93)
(a) (b)
N1 + N2 k2 1 2(c)
[(x1 x2 )2 ]1
integrand in the deformation parameter g( ) gives

(d) (e) (f)
1
n(n2 1)jbn j2 :
X
=
N2 (N2 2 1) + N1 (N1 2 1) 2
:
(3.94)
n=2
N1 + N2 6k 2
" #
Tr d1>2 A1 (x1 )x_ 1 A1 (x2 )x_ 2 ;
which gives

2( 1 ) 2 x_ 1 x_ 2
2 N N d1>2 2 :
k 2 1 2 1 2 [(x1 x2 )2 ]1

N12 N2 + N22 N1 2 ( 21 ) x_ 1 x_ 2
= d1>2 2 : (3.92)
N1 + N2 k2 1 2 [(x1 x2 )2 ]1
We were able to obtain this result through the Ap method and also through the
DRED regularisation scheme after setting the constants to zero.
Diagram (c) comes from the bosonic part of the quadratic term in the WL
expansion with contractions between scalars:
!2 " #
2i
k
Tr j j
d1>2 x_ 1 MJ I CI C J jx_ 2jMK LC K :
LC
It gives

2 ( 12 ) 2 2 jx_ 1jjx_ 2j
! " #
N1 N2 d1>2 MJ I MI J :
k 43=2 [(x1 x2 )2 ]1 2
Using that MJ I MI J = 4 and also summing the N2 N2 contribution, we nd


jx_ 1jjx_ 2j
!2 " #
2 ( 12 ) N1 2 N2 + N2 2 N1
= d1>2 4 : (3.95)
(a) (b) (c)
k 43=2 N1 + N2 [(x1 x2 )2 ]1 2
Up to second order in the deformation parameter g ( ), the integral above gives
zero. This result was obtained both through the Ap method and DRED.
(d) (e) (f)
N2 N2 block, we get that the rst diagram gives
Fermionic sector
Considering also the
N2 (N2 2 1) + N1 (N1 2 1) 2
= :
N1 + N2 6k 2
Diagram (d) comes from the fermionic part of the second order expansion of the
" #
Tr d1>2 A1 (x1 )x_ 1 A1 (x2 )x_ 2 ;
which gives
2( 1 ) 2 WL with a one-loop contraction between two fermions, i.e., it comes from the
x_ 1 x_ 2
2 N N d1>2 2 :
k 2 1 2 1 2 [(x1 x2 )2 ]1
N12 N2 + N22 N1 2 ( 21 )
one-loop insertion in
x_ 1 x_ 2
= d1>2 2 : (3.92)
N1 + N2 k2 1 2 [(x1 x2 )2 ]1
! " #
2
TrN1 d1>2 jx_ 1 jjx_ 2 j( )1 ( )2 :

k
It gives

jx_ 1jjx_ 2j I (x1) I (x2) :
!2
2 2( 1 )
i detour: deforming the 1/22 3BPS loop2 (N2 N1 )N1 N2 d1>2
k 16 [(x1 x2 )2 ]1 2
3.4 Perturbative 56
Taking into account the contribution from both blocks, we get

(a) (b) (c)

d1>2 [(x jx_ 1xjjx_)22j]1

2 I (x1 )I (x2 ) I (x2 )I (x1 ) :

(d)
1 2
(e) (f)
Note that I (x1 )I (x2 ) = I (x2 )I (x1 ), so diagram (d) gives zero.
N2 (N2 2 1) + N1 (N1 2 1) 2
= :
N1 + N2 6k 2
Diagram (e) comes from the double contraction of fermions in the fourth order
" #
Tr d1>2 A1 (x1 )x_ 1 A1 (x2 )x_ 2 ;
which gives
2( 1 ) 2 x_ 1 x_ 2
2 N N d1>2 2 :
k 2 1 2 1 2 [(x1 x2 )2 ]1

N12 N2 + N22 N1 2 ( 21 ) x_ 1 x_ 2
= d1>2 2 : (3.92)
N1 + N2 k2 1 2 [(x1 x2 )2 ]1
3.4 Perturbative detour: deforming the 1/2 BPS loop 3.4 Perturbative detour: deforming the 1/2 BPS loop 56 58
expansion of the WL,

(a) (b) (c)

1 2 2 2 32
= d1>2>3>4 jx_ 1 jjx_ 2 jjx_ 3 jjx_ 4 j
(
N 1 + N 2 k 4 3 2
( x1 x2 ) ( x 3 x4 )
(d) (e) (f)
h i
N2 N1 2 (1 2 )(3 4 ) + N2 2 N1 (2 1 )(4 3 )
[(x1 x2 )2 (x3 x4 )2 ] )
Considering also the N2 N2 block, we get that the rst diagram gives 3
N2 (N2 2 1) + N1 (N1 2 1) 2 2
= :
N1 + N2 6k 2
h i
( x 1 x4 ) ( x 2 x3 )
+ N2 2 N1 (1 4 )(3 2 ) + N2 N1 2 (4 1 )(2 3 ) ;
[(x1 x4 )2 (x2 x3 )2 ]
3
" #
Tr d1>2 A1 (x1 )x_ 1 A1 (x2 )x_ 2 ; 2
3.4 Perturbative which

detour: deforming the
gives 1/2 BPS loop 56
2( 1 ) 2 x_ 1 x_ 2
2 N N d1>2 :
k 2 1 2 1 2 [(x1 x2 )2 ]1 2
and diagram (f) comes from the six gauge-fermion mixed terms in the third order

N12 N2 + N22 N1 2 ( 21 ) x_ 1 x_ 2
= d1>2 2 : (3.92)
N1 + N2 k2 1 2 [(x1 x2 )2 ]1
(a) (b)
expansion of the WL,
(c)
(
1 2 D E
= d1>2>3 Tr 2I 3J A1 (1 ) 2I 3J x_ 1 jx_ 2 jjx_ 3 j + 2I 3J
N1 + N2 k
D E D E
A2 (1 ) 2I 3J x_ 1 jx_ 2 jjx_ 3 j + 3I 1J 1I A1 (2 ) 3J jx_ 1 jx_ 2 jx_ 3 j + 3I 1J
(d) (e) (f)
D E D E
1J A2 (2 ) 3I jx_ 1 jx_ 2 jx_ 3 j + 1I 2J 1I 2J A1 (3 ) jx_ 1 jjx_ 2 jx_ 3 + 1I 2J
N2 (N2 2 1) + N1 (N1 2 1) 2
= :
N1 + N2 6k 2

)
D E
1I 2 A2 (3 ) jx_ 1 jjx_ 2 jx_ 3 :

Tr
"
d1>2 A1 (x1 )x_ 1 A1 (x2 )x_ 2 ;
#
J
which gives

2( 1 ) x_ 1 x_ 2
2 N1 2 N2
d1>2 2 :
k2 1 2 [(x1 x2 )2 ]1

N12 N2 + N22 N1 2 ( 21 ) x_ 1 x_ 2
=
N1 + N2 k2 1 2
d1>2
[(x1 x2 )2 ]1 We still
: have to compute these last two diagrams to nish the 2-loop computation
2 (3.92)
of the Wilson loop. This is work in progress.
Comments
The O(b2n) terms appearing in the computation of the expectation value of W

give what is known as the wavy-line approximation. This approximation was
considered in the context of N =4 SYM theory in [41]. There the authors work
with the WL supported along a wavy line, which deviates by a small amount from
an innite straight line. They parameterise it as
x (s) = (s; (s)); s 2 ( 1; +1) ;
where the three-dimensional vector (s) is a smooth function of the curve param-
eter s with small magnitude. They found that, relying on symmetry properties of
the theory, a universal term of the form

(_(s) _(s0 ))2
ds ds0
(s s0 )2
(3.96)
appears in the computation of the expectation value of the Wilson loop up to sec-
ond order in (s). They also found that this universal term comes accompanied by
a non-trivial function of the coupling constant, which should receive contributions
from all orders in perturbation theory. Later in [42] this non-trivial function of
the coupling constant was named Bremsstrahlung function B (). It was shown
that the B function appears not only in the two point function of the displacement
operator,
hh Di(t)Dj (0) ii = 12Bt4 ij ;

where the double brackets denote the expectation value evaluated on the Wilson
line, but also in the the energy emitted by a moving quark in the small velocity
limit (which justies the name Bremsstrahlung),

E = 2B dt (v_ )2 ;
and at the derivative of the cusp anomalous dimension cusp () at zero angle 9
1
B = @2 () :
2 cusp
=0
In fact, the appearance of B in these three observables holds for any conformal
gauge theory with a Wilson loop operator. Here, since we have a deformed Wilson
loop in N = 6 Chern-Simons-matter theory, we expect the O(b2n) contribution to

be proportional to the Bremsstrahlung function. Moreover, since we are computing
hW i up to two loops, we expect to nd B () up to second order in the coupling

constant . In particular, recall that the t'Hooft couplings are 1 = N1 =k and
2 = N2 =k. In the large N limit, we have 1 2 and

1
hW i(2) B () n(n2 1)jbn j2 :
X
(3.97)
n=2
After computing the remaining diagrams listed above, we expect to recover a
closed formula at two loops for the function B ( ). This will allow us to compare
our result to the ones of [34]. If these results match, not only our claim that the
fermionic couplings should take the form (3.18) would be strongly supported, but
also the use of the substitution methods of Appendix B with constants set to zero.
With this settled, we could for instance use the results to do the perturbative
computation at fourth order in the deformation parameter g, which has not been
considered before in the literature.
9 For an operator supported along a straight line, a cusp is a region where the line makes a
sudden turn by an angle . Then the cusp anomalous dimension is the logarithmic divergence
that arises in the expectation value of the Wilson line due to the sudden turn [43].
Chapter 4
N = 4 Chern-Simons-matter theories
This chapter is mostly based on [21]. At the end we also include some preliminary
results of an ongoing project that should be presented in the near future [23].
An N = 4 Chern-Simons-matter theory is a quiver theory that can be either

circular or linear. In this chapter most of our discussion is carried for a generic
theory and we do not worry about the endpoints of the quiver. We consider nodes
labeled by an index I and these represent the respective gauge eld AI . An edge
connecting nodes I and I + 1 represents a hypermultiplet (qIa ; I a_ ), whereas an

edge connecting I and I 1 represents a twisted hypermultiplet (~qI 1 a_ ; ~Ia 1 ).
This structure is carried over along the quiver in an alternated fashion. The eld
content is summarized in the quiver diagram of Figure 10, where the thick solid
lines represent the matter elds.
qI 2 a ; Ia_ 2 q~I 1 a_ ; ~Ia 1 qI a ; Ia_ q~I +1 a_ ; ~Ia+1

AI 1 AI AI +1
qIa 2 ; I 2 a_ q~Ia_ 1 ; ~I 1a qIa ; I a_ q~Ia_+1 ; ~I +1 a
k k k
Figure 10: The quiver and eld content of the N = 4 theory.
The (twisted) hyper multiplets can be decomposed into pairs of chiral mul-
tiplets. Figure 11 shows the chiral scalar in this decomposition explicitly. Here
and throughout chiral elds are denoted as solid arrows and when needed, the
anti-chiral elds are represented by dashed arrows. The orientation of the arrows
stands for the eld's representation. For example, the elds qI2 is in the (; ) of
U (NI ) U (NI +1 ) and qI 1 is in the conjugate representation.
60
61
qI 2;1 q~I 1;1_ qI;1 q~I +1;1_
AI 1 AI AI +1
q2I 2 q~2_
I 1 q2I q~I2_+1
k k k
Figure 11: The decomposition of the N = 4 matter multiplets into pairs of chiral
multiplets. Only the chiral scalar of each multiplet is indicated explicitly.
The SU (2)L SU (2)R R-symmetry indices a; b = 1; 2 and a;

_ b_ = 1_ ; 2_ are raised
a ab
and lowered using the appropriate epsilon symbols: v = vb and va = ab v with
b
12 = 21 = 1, and similarly for the dotted indices.

The covariant derivatives acting on scalars, for example, are dened as
D' qIa = @' qIa iA';I qIa + iqIa A';I +1 ;

(4.1)
D' qI a = @' qI a iA';I +1 qI a + iqI a A';I :
Naturally, for twisted scalars we have
D' q~I 1 a_ = @' q~I 1 a_ iA';I 1 q~I 1 a_ + iq~I 1 a_ A';I ;

D' q~Ia_ 1 = @' q~Ia_ 1 iA';I q~Ia_ 1 + iq~Ia_ 1 A';I 1 :
For later convenience we dene moment maps and currents, which are bilinears
of the hypermultiplets and twisted hypermultiplets in the adjoint representation
of U (NI ), as follows
1 a c
I ab = qIa qI b q q ; jIab_ = qIa Ib_ ac b_ c_ I c_ qI c ;
2 b I Ic
1 a_ c_
~I a_ b_ = q~Ia_ 1 q~I 1 b_ q~ q~ 1 c_ ; |~Ib_ a = q~Ib_ 1 ~Ia b_ c_ ac ~I 1 c q~I 1 c_ ;
2 b_ I 1 I 1
I = qIa qI a ; ~I = q~Ia_ 1 q~I 1 a_ :
Similar bilinears (with the appropriate replacement of hypermultiplets and twisted
hypermultiplets) exist also for the other nodes. For example, for the I +1 node one
can dene I +1 = qIa qIa . Note that in this notation the index of the moment maps
represents the node under which they are charged, rather than the elds they are
made of, as is the case in the notation of [17, 44]. In particular, I +1 ab is made
of the same elds qI a and qIb as I b

a, but it is charged under a dierent group.
The moment maps are triplets of the respective SU (2) R-symmetry group and are
used below to construct the basic 1-node Wilson loops. The moment maps can
be thought of as the generalization to N = 4 of the N = 2 scalar , though the
latter is an auxiliary eld in an o-shell formulation while we work in an on-shell
formulation. We provide some details on this correspondence in Appendix C.

62
We dene the theory on S3 and the Wilson loops we construct are supported
along the equator of this sphere. The corresponding N = 4 supersymmetry trans-

formations are given in the appendix in (C.1), along with details about how they
relate to the ones for the N = 2 theory [45, 46]. The theory is invariant under 16
supersymmetry transformations generated by parameters ab_ . Each parameter is a
linear combination of four Killing-spinors on S 3 such that, together with the two
possible values fora and b_ , we indeed have a total of 2 2 4 = 16 parameters.
l l r r
We label each Killing-spinor as , , , , according to their chiralities. They
obey
r l;l = 2i l;l ; r r;r = 2i r;r : (4.2)
Along the circle we take ' = 3 and these reduce to [35]
0 1 0 1 0 1 0 1
1 0 e i' 0A
l = @ A ; l = @ A ; r = @ A ; r = @ : (4.3)
0 1 0 ei'
Compared to ABJ(M) theory, in N =4 the allowed set of theories is wider
and dierent loops that are equivalent under the SO(6) R-symmetry of ABJ(M)
may be on disconnected branches of the moduli space in the presence of only
SO(4) R-symmetry. Operators preserving 4 and 8 out of the 16 supercharges were
considered in [47]. The 1=2 BPS ones were found to present a nite degeneracy
that was recognized in [44] and was soon realized holographically in [48]. As for
less supersymmetric operators [31, 49, 27, 50], the study of their degeneracy has
blossomed into an independent research program whose full scope is still unclear.
In CSm theories in general, the past several years were marked by more and
more examples of BPS loops being found. In the last couple of years, in partic-
ular, some initial steps to reorganize the subject were taken. First, the roadmap
paper [20] reviewed what was known at the time about BPS Wilson loops in
three dimensions, introduced some new formalism and, for the rst time, prop-
erly addressed the moduli spaces of the 1=6 BPS Wilson loops in ABJ(M) theory.
Second, the moduli spaces of BPS Wilson loops in N = 2 theories were stud-
ied in [51] and identied with quiver varieties. Most recently, the symmetries of
BPS line operators in diverse dimensions were analyzed in [52], where the natu-
ralness of marginal defect couplings in three dimensions was stressed. Our work
in [21] aims to continue on that path, focusing on N =4 Chern-Simons-matter
theories [1619].
We study circular Wilson loops in three-dimensional N = 4 theories that are

63
continuously connected by marginal deformations to the usual 1=4 BPS Gaiotto-

Yin (bosonic) Wilson loop [22]. One type of deformation arising in quiver gauge
theories involves couplings to the matter elds in bifundamental representations
of the gauge groups. This follows closely previous studies of the moduli spaces of
loops in N = 2 theories and in ABJ(M) [20, 51]. The other deformation is often
called the latitude deformation and is similar to the four-dimensional Wilson
loops in [53] and to the ABJM latitudes [27]. This construction uses that N =4
theories have triplets of bifundamental bilinears, the moment-maps that generalize
the scalar coupling of the Gaiotto-Yin Wilson loop.
The usual 1=4 BPS loop involves couplings to scalar bilinears of both the hyper
and twisted hyper elds of the three-dimensional Chern-Simons-matter theories.
These couplings break the SO(4) ' SU (2)L SU (2)R to U (1)L U (1)R , while
the latitude deformation further breaks one of the U (1)'s (we choose it to be
U (1)R ). This is a continuous deformation with a parameter and for generic

values of this parameter the loop preserves 1=8 of the supercharges. Given this as
a starting point, we can deform the bosonic loops by introducing couplings to more
bifundamental elds through superconnections and get fermionic operators, in
the spirit of [24, 20]. This produces richer moduli spaces of BPS loops, which also
include 1=16 BPS operators. As this construction involves in an intimate way
the hypermultiplet elds and their SU (2) R-symmetries, we name these operators
hyperloops.
As mentioned above, we dene our theories on S 3 and our loops are supported
along great circles of this space. What we call latitude deformation is then
a slight misnomer, as for us this deformation only aects the internal space of
moment map couplings and not the geometric contour on which these operators
are dened. As a consequence, the two supercharges preserved by the latitude
loops are not a subset of the four preserved by the Gaiotto-Yin loop, but this
can be resolved by a conformal transformation mapping the original circle at the
equator to an actual latitude of the S 3. The alternative formulation would follow
the four-dimensional construction in [33] or the ABJ(M) analog in [25] with the
operators dened from the start along latitudes and with the matter eld couplings
dictated by this choice of geometric contour. In that case, the latitude loops would
be a subset of loops with arbitrary shapes on S 2 S 3 , all preserving a xed subset

of the supercharges. So, barring the fact that we have made the choice to keep
the circle xed in space, we can view all the loops presented here as continuous
deformations of the Gaiotto-Yin loop preserving a subset of the supercharges.

4.1 1-node Wilson loops 64
The classication of the hyperloops depends on (i ) some discrete data and (ii )
some continuous parameters. The discrete data is a choice of vector elds that ap-
pear in the (super)connection (and their multiplicities) and a subset of the matter
elds that we allow to couple to the loop. This information can be conveniently
conveyed by certain quiver diagrams, as we explain below. Restricting to only
half of the matter elds enhances supersymmetry and gives hyperloops that are
(at least) 1=8 BPS, so there are two choices of which half of these elds to include,
with each option spanning a separate branch of the moduli space. One can also
allow a coupling to all the elds at the price of preserving less supersymmetry and
obtaining 1=16 BPS operators. The continuous parameters are the latitude angle
mentioned above, an azimuthal angle '0 (which we mostly ignore) and the con-
tinuous couplings to the matter elds. In the simplest case, there are two or four
complex parameters per edge in the quiver, depending on how many supercharges
one wants to preserve. These parameters are subject to a global gauge symmetry,
reducing the moduli space to a cone, similar to the conifold in the case of ABJ(M)
theory [20].
4.1 1-node Wilson loops

We start by constructing Wilson loops coupling to a single gauge eld and hence
suppress the I index on the elds. The most symmetric such loop is a circle
coupling to both the untwisted and twisted moment maps through the connection
A = A' ki (11 22 + ~1_ 1_ ~2_ 2_ ): (4.4)
This choice of scalar coupling can be motivated by considering what is the natural
generalization to the N =4 case of the scalar appearing in the Wilson loop
in N = 2 Chern-Simons-matter theory [22]. Since the moment maps are triplets

of the R-symmetry group for N = 4, the coupling above corresponds to picking
the moment maps along the `third' direction of the R-symmetry triplet, so can be
denoted alternatively as 3 and ~3 .

From (C.1) one nds the supersymmetry transformations of the moment maps
1 a cc_ 1
I ab = bc_ jIac_ j ; ~I a_ b_ = cb_ |~Iac_ + ba__ cc_ |~Icc_ ;
2 b cc_ I
(4.5)
2
so the variation of the connection in (4.4) is
i i
A = ab_ ' (j ab_ |~b_ a ) (1b_ j 1b_ 2b_ j 2b_ a1_ |~1_ a + a2_ |~2_ a ) ; (4.6)
k k
4.1 1-node Wilson loops 65
with ' along the equatorial circle on which the loop is supported. There are no
solutions to A = 0 for 12_ and 21_ , while for the other components one nds the
conditions
11_ (1 ' ) = 22_ (1 + ' ) = 0 : (4.7)
The resulting Wilson loop is hence 1=4 BPS, and it is in fact the same as the
1=2 BPS loop in N = 2 theories [22] or the 1=6 BPS loop in ABJ(M) theory, as
in [2830].
Equation (4.7) restricts 11_ to the two chiralities l and r, while 22_ is a linear
combination of l and r. We can write the corresponding four supersymmetries
as
Q1l_ 1 ; Q1r_ 1 ; Q2l_ 2 ; Q2r_ 2 : (4.8)
In the subsequent sections we construct families of Wilson loop operators that
preserve all or particular linear combinations of those supercharges. These loops
couple to two or more nodes of the quiver and involve combining several gauge
connections and couplings to more bifundamental elds. However, there is still a
family of Wilson loops involving just a single node that preserve only two super-
charges, being therefore 1=8 BPS. Following the logic of the 1=4 BPS Wilson loops
in N = 4 SYM in 4d [53], the connection (4.4) is naturally generalized introducing
a latitude
1
angle and an azimuthal angle '0 , as follows
A = A' ki 11 22 + cos (~1_ 1_ ~2_ 2_ ) + sin (e

~1_
i(' '0 )
2_ + e
i(' ~2_
'0 )
1_ ) :
(4.9)
Notice that only the couplings to the twisted hypermultiplets are modied, via
their moment maps. This operator is now coupled to three dierent moment
maps, which can alternatively be written as ~3 and ~ . This deformation is not
possible in theories with only N = 2 symmetry, with only one eld in the
vector multiplet. An analogous construction where the untwisted moment maps
are modied and the twisted ones remain as in (4.4) also works, but deforming
both at the same time does not give BPS operators. This deformation involves
two parameters, and '0 , which dene an S 2 . For simplicity we set '0 = 0 in the
following.
1 Note
that this is only a latitude in the internal space of scalar couplings, while the loop is still
an equator of the S 3 .
4.2 Hyperloops at = 0 66
The supersymmetry variation of (4.9) gives
i
A = ab_ ' (j ab_ |~b_ a ) (1b_ j 1b_ 2b_ j 2b_ )
k
+ cos (a1_ |~1_ a a2_ |~2_ a ) + sin (e i' |~1_ a + ei' |~2_ a )
a2_ a1_ :
Requiring this to vanish and collecting terms according to the components of the
currents, one nds conditions on the supersymmetry parameters which are all
solved imposing
1a_ (1 ' ) = 2a_ (1 + ' ) = 0 ;

a1_ ( ' cos ) = sin e i' a2_ : (4.10)
The rst line of (4.10) sets 1l a_ = 1ra_ = 2l a_ = 2ra_ = 0. The second line also
r l l r l r r l
eliminates _ = _ = _ = _ = 0. The remaining _ , _ , _ , _ are related by
a2 a2 a1 a1 11 12 21 22
sin r sin l
1l 1_ = 12_ = cot 1r2_ ; 2r1_ = 22_ = tan 2l 2_ : (4.11)
1 cos 2 1 + cos 2
We nd nally the two independent supercharges preserved by the loop
Q1 = cos 2 Q1l_ 1 + sin 2 Q2r_ 1 ; Q2 = cos 2 Q2l_ 2

sin Q1r_ 2 ;
2
(4.12)
which is then 1/8 BPS, as advertised. Notice that the supercharges in (4.12) are
not a subset of those in (4.8). The reason for this is that we kept the circle on the
equator of S 3 . Were we to follow the logic of [25] and place the loop at a latitude
angle of =2 , which can be done via a conformal transformation on the S 3 , the
1_ 1 2_ 2
resulting loops would preserve Q and Ql , which are indeed a subset of (4.8).
l
4.2 Hyperloops at = 0
In this and the next section we construct what we dub hyperloops: BPS Wilson
loops involving multiple gauge elds and couplings to the hypermultiplets beyond
their bilinears. We apply the deformation procedure that was introduced in [20],
and explained in detail in Section 3.2, based on the 1-node loop (4.4). We also
explain what the moduli space of the resulting operator is and its relation to quiver
varieties according to [51].
We begin by dening two linear combinations of the four supercharges pre-

served by the 1=4 BPS 1-node loop,
QL+ = Q1l_ 1 + Q2l_ 2 ; QL = Q1l_ 1 Q2l_ 2 :

These are such that, when acting twice on scalars, give back their covariant deriva-
tives. Therefore, these are appropriate candidates to construct a deformation
L = iQ G + G 2 for some matrix G.

In the context of ABJM, the matrix G is block 2 2 because the theory is a
2-node quiver theory and the resulting loops always couple to these 2-node. Here,
as anticipated in Chapter 2, our loop is allowed to couple to multiple nodes and
consequently the matrix G may have a richer structure.

Suppose that our loop couples to 3 nodes, say I 1, I and I + 1. Then G
I , as well as twisted ones q~I 1 , q~I 1 . Moreover, suppose
should contain scalars qI , q
we want to include some of the nodes more than once and choose, for example,
node I to appear twice. Then our composite bosonic connection is
AI
0 1
B
1 + 21 0 0 0 C
L =
B
B
B
0 AI 0 0 C
C
C ; (4.13)
bos B
B
@
0 0 AI 0
C
C
A
0 0 0 AI +1 + 12
where we already include the constant shifts. Note that instead of adding 1=4 to
every node, as in Section 3.2, we chose to move these around in such a way that
they appear in an alternated fashion. This is harmless to the derivation since what
matters is the dierence between two neighboring nodes, but it will be useful in
the analysis that follows. Next we dene the matrix
0 1
B
0 1I 1 q~I 1 2_ 2I 1 q~I 1 2_ 0 C
B 1 2_
1 q~I 0 0 1q2C
G= B
B I
2_
1 I IC
C
; (4.14)
1 q~I
B
B 2 0 0 2q2C
C
@ I 1 I IA
0 1I qI 2 2I qI 2 0
which is parameterised by constants and that are complex, but not necessarily
complex conjugates of each other. These carry the capital node index I and a
i
lower-case multiplicity index . From this matrix we construct the superconnection
L ; = L bos + iQL+ G + G 2 : (4.15)

2
The hyperloop is then
" #
W ; = sTr P exp i L ; jx_ j ds ;

and it preserves all four supersymmetry transformations generated by (4.8).
Another possibility is to consider the composite bosonic connection with swapped

3
constant shifts,
AI
0 1
B
1 0 0 0 C
L =
B
B
B
0 AI + 12 0 0 C
C
C : (4.16)
bos B
B
@
0 0 AI + 12 0
C
C
A
0 0 0 AI +1
In this case we nd that for
0 1
I 1 q~I 1 1_ I 1 q~I 1 1_
0 1 2 0
B C
1_ I1 qI1 C
G = BBBB I2 1q~I1_
B 1
1 0 0 C
C ; (4.17)
q~ 0 0 I2 qI1 C
C
@ I 1 I 1 A
I qI 1 I qI 1
0 1 2 0
we can follow the construction exactly as above with a superconnection and hy-
perloop that could be denoted as L ; and W ; (with the parameters and

being complex, but, again, not complex conjugates of each other).
Applying the analysis of [21], these two hyperloops can be visualized by quiver
diagrams, which may include some or all of the nodes and edges of the original
quiver dening the gauge theory. Let us pause for a moment to explain the ma-
chinery to associate a quiver diagram to a hyperloop. First, we include a node
for each vector multiplet the loop couples to, solid arrows for the chiral elds and
dashed arrows for the anti-chirals. Some of the nodes are denoted by squiggly
circles and some by unsquiggly ones. This represents that the connection of the
gauge eld in the squiggly nodes has an extra shift by 1=2, 4

and supersymmetry
requires alternating squiggly and unsquiggly nodes. Each node is decorated by
2 Note the +i factor in the path-ordered exponential, in contrast with i from (2.4). The
dierence is due to the conventions we follow in this chapter. See for instance the sign dierence in
the covariant derivatives (3.2) and (4.1). This sign dierence is also manifest in the supercovariant
derivative denition, which here becomes D' () @' () i[L; ()].
3 To avoid cluttering the notation too much, we denote the bosonic connections and the matrices
G of all these examples with the same symbols: Lbos and G . We always refer to explicit equations,
so this should hopefully not lead to confusion.
4 Here and throughout we set the radius of the sphere to R = 1. For a generic radius the shifts
would be 21R , and similarly for the shifts in the next section.
integers pI , indicating the multiplicity of the gauge eld in the hyperloop. We
indicate the chiral elds coupling to the hyperloop (according to the decomposi-
tion in Figure 11) by solid arrows and the anti-chirals by dashed arrows. Arrows
between the I and I+1 nodes are decorated by pI pI +1 complex parameters
I, I for chirals and I , for anti-chirals, though they are not complex conju-
gates. Applied to W ; and W ; , these rules give the quiver diagrams drawn in
Figure 12 and we see that the supertrace in the denition of the operators treats
I
the unsquiggly nodes ( ) as even and the squiggly ones ( I 1) as odd.
pI 1 I 1 pI I pI +1
I 1 I
(a)
I 1 I
pI 1 I 1 pI I pI +1
(b)
Figure 12: Quiver diagrams representing 1=4 BPS Wilson loops (a) W ; and (b)
W ; .
Other hyperloops are also allowed and at this point the power of these quiver
diagrams starts manifesting. Say that we want to use a G that couples to all
scalars, instead of only those with indices 1; 1_ or 2; 2_ as above. In this case, what
we would be seeking is a diagram containing edges from Figure 12a and Figure
12b at the same time. However, by looking at these diagrams, we see that we need
to be careful due to their dierent squigglyness. This means that we need to
combine scalars to match the constant shifts of a single bosonic connection. To be
more explicit, suppose that we want to use L bos in (4.13). In this case, we can use
G in (4.14) but for the one in (4.17) we would have to include some extra phases
0 1
B
0 ei' 1
I 1 q~I 1 1_ ei' 2
I 1 q~I 1 1_ 0 C
i' 1 q 1_ e i' I1 qI1 C
G! e I 1 ~I 1 0 0
B
B C
B C; (4.18)
e i' 2 q 1_ e I qI C
i' 2 1
I 1 ~I 1 0 0
B
B C
@ A
0 ei' I1 qI 1 ei' I2 qI 1 0

5
in order to preserve supersymmetry . This would lead to a new G given by
0 1
B
0 1I 1 q~I 1 2_ + ei' I1 1 q~I 1 1_ 2I 1 q~I 1 2_+ ei' I2 1 q~I 1 1_ 0 C
B
B I 1 q~I
1 2_ 1+e
i' 1 q 1_
I 1 ~I 1 0 0 1 q 2 + +e i' 1 q 1 C
I I I IC
B C:
B
B 2
I
2_
1 q~I 1+e
i' I 1 q~I 1
2 1_
0 0 I qI + e
2 2 i' I qI C
2 1 C
@ A
0 I qI 2 + ei' I1 qI 1

1 2I qI 2 + ei' I2 qI 1 0
(4.19)
The rest of the construction follows as before where the superconnection and the
hyperloop would now carry indices ; ; ; . However, the resulting operator

will only be invariant under Q+ so it is 1=16 BPS. We can also use Q to build
L L
our hyperloop and in this case only this supercharge would be preserved. The
resulting operator can be represented graphically as in Figure 13.
I 1 I
pI 1 I 1 I 1 pI I I pI +1
I 1 I
Figure 13: A quiver diagram for a 1=16 BPS hyperloop.
For any quiver diagram we can always exchange squiggled and unsquiggled
nodes (i.e. grading) at the price of adding phases to the matter elds, as in
(4.18). This leads to the same loop operator, but in a dierent gauge. Note that
in the case of1=16 BPS loops shown in Figure 13, the two gradings of the quiver
are gauge equivalent. This is not the case for families of 1=4 BPS loops, where the
two gradings also represent which of the chiral elds are included, as illustrated
in Figure 12.
These diagrams represent 1=4 BPS hyperloops, preserving all four supercharges
6
in (4.8) if all solid arrows point into squiggly nodes. If the decorated quiver
contains solid arrows pointing both in and out of the squiggly circles, the loop
preserves only one linear combination of the four supercharges in (4.8), either QL+
or QL according to how we choose to build the hyperloop, and is 1=16 BPS.
4.2.1 Moduli spaces
The hyperloops we construct are built upon a connection L bos to which we add
bifundamental couplings encoded in the matrix G. The deformation is thus de-
5 Fromthe point of view of the chiral decomposition of N = 4 to N = 2 theories, the need

to include the extra phases can be seen as matching the behavior of the chiral scalar q1 and the
anti-chiral scalar q2 .
6 In a dierent gauge, they all point out of squiggly nodes.
scribed by the parameters in this matrix, which are a set of complex numbers ,
, etc. This description is however redundant, so the moduli space of hyperloops
is a quotient of the space of these parameters, as described below.
Wilson loops are intimately related to gauge invariance, and in particular are
I U (NI )
Q
gauge invariant observables. The invariance under gauge transforma-
tions are built into the denition of the hyperloops, but they posses a larger
(global) gauge symmetry. L can be thought of as a superconnection, which in the

example above is in the superalgebra GL(NI 1 + NI +1 j2NI ). We do not expect
local-gauge symmetry under this group, as it is not a symmetry of the theory, but
constant gauge transformations do not require extra gauge elds. Of those, the
transformations that preserve our formulation and are not the gauge symmetries
of the quiver theory are the centralizer of Lbos , which in our examples is
(C )2 GL(2; C) :
Concretely this symmetry acts by conjugation
0 1 0 1
B
x 1 0 0 0 C B
x 0 0 0C
B C B C
B 0 0 C B 0 0C
L ! B
B S 1 C
C L B
bos B S C = L
C
; S 2 GL(2; C) :
0 0 0 0C
bos bos
B C B
B C B C
@ A @ A
0 0 0 y 1 0 0 0 y
The action on G and consequently on L ; is generally non-trivial, mapping super-

connections with dierent parameters to each-other. Note though that matrices
in the center of GL(4; C), i.e. where y = x and S = diag(x; x) do commute with
all matrices, so should be excluded, thus the gauge symmetry is really
S ((C )2 GL(2; C)) : (4.20)
Remaining with our examples above, we describe the 1=4 BPS loops with
eight complex parameters, see (4.14) and (4.17), and the 1=16 BPS loops with
sixteen, see (4.19). As argued, this means that the moduli space of 1=4 BPS loops
correspond to two copies of
C8==S ((C)2 GL(2; C)) ;

while for 1=16 BPS operators we have a single copy of
C16==S ((C)2 GL(2; C)) :

These are 3- and 11-(complex) dimensional conical spaces, respectively.
These spaces are the usual quiver varieties associated to the quiver representa-
tions in Figures 12 and 13, see [5458]. We use the double slash notation mirroring
the concept of geometric invariant theory, in order to point out that we need to be
careful when identifying the singular orbits of the resulting manifolds. In particu-
lar, hyperloops with o-diagonal components that are exclusively upper or lower
triangular are identical as quantum operators for all values of and therefore
should be identied.
The analysis can be carried out for more general quivers as follows. In the
case of 1=4 BPS loops, each edge between nodes of multiplicities pI and pI +1 has
2pI pI +1 complex parameters. For 1=16 BPS operators, this becomes 4pI pI +1 . From
these we need to remove the symmetries, which amount to a factor of GL(pI ; C)
for each node with multiplicity pI , apart for the trivial action of the center, as
explained above (4.20).
Therefore, for a linear quiver of length L we nd the moduli space of 1=4 BPS
loops to be two copies of
C2p p C2p p C2pL

1 2 2 3 1 pL ==S (GL(p
1) GL(p2) GL(pL)) : (4.21)
For the 1=16 BPS loops, we loose the second copy and we have the same as above
where the 2s in the exponents become 4s.
For a circular quiver, the moduli space of 1=4 BPS loops is two copies of
C2p p C2p p C2pL

1 2 2 3 1 pL C2pLp ==S (GL(p1) GL(p2) GL(pL)) :
1
(4.22)
For 1=16 BPS loops, we again loose the second copy and the manifold is the one
above with 2s replaced by 4s.
All these moduli spaces are the quiver varieties associated to the quiver rep-
resentations that the hyperloops furnish. Before moving on to the next section,
where we study the 6= 0 versions of these hyperloops, let us specialize to the
moduli space of 1=6 BPS loops in ABJ(M) theory with p1 = p2 = 1. As it is a
circular two-node quiver, we set L = 2 in (4.22), giving two copies
C2 C2==S ((C) (C)) = C4==C :

This quotient is the conifold, as already found in [20].
4.3 Hyperloops at 6= 0 73
4.3 Hyperloops at =
6 0
Having reviewed the formulation of the loops in terms of quiver representations
and implemented it for theories with N =4 supersymmetry, we generalize it to
loops based on the bosonic connections deformed as in (4.9). We provide here
much more detail, as most of these loops are novel.
4.3.1 Construction
To construct hyperloops with 6= 0 we apply the same logic of [20], involving a

deformation built from the supercharges in (4.12). More specically, we consider
the combinations Q = Q1 Q2. Note that when = 0 it follows that Q = QL.

The whole procedure relies on being able to write (Q ) G = D' G , for some matrix
2
of matter elds G and an appropriate covariant derivative D' . Since (Q1 ) =

2
(Q2 )2 = 0, this is the same as acting with fQ1 ; Q2 g on G . In Appendix C we

compute the explicit form of this double transformation acting on the untilded
scalar elds qIa and on a specic spatially-dependent rotation of the tilded ones,
dened by
i' sin
r~I 1 1_ cos q~I 1 1_ + e q~I 1 2_ ; r~I 1 2_ cos q~I 1 2_ ei' sin q~I 1 1_ :
2 2 2 2
(4.23)
These combinations are nice because the connection in (4.9) can be written com-
pactly in their terms as
AI = A';I ki (I 11 I 22 + r~I1_ 1r~I 1 1_ r~I2_ 1 r~I 1 2_ ) : (4.24)
Moreover, one can show that the double transformation acting on qIa and r~I 1 a_
can be recast as a covariant derivative, see (C.6) and (C.7). These elds are then
the natural ingredients to write down G.

The connections appearing in the double deformations are as in (4.24), with
extra shifts of 21 and 12 cos that can be viewed as the coupling to a background
eld on the sphere. This is implemented by shifting the original connections, as
in the previous section. Let us note that the eect of these shifts is to introduce
phases like ei cos in the denition of the Wilson loop, which are compensated
for in the denition of the trace. Recall that in the original formulation of the 1/2
BPS Wilson loop of [24] a trace was required to make them gauge invariant, but
in the gauge introduced in [20] and in Section 4.2, this is replaced with a more
natural supertrace. For gauge invariance of hyperloops at 6= 0, the denition
of supertrace should include 1 and ei cos gradings. The generic Wilson loop
with both and couplings is then dened as
" #
W ; ; ; = sTr P exp i L ; ; ; jx_ j ds ;

where sTr includes the generalized gradings mentioned above to compensate for
the eect of the shifts.
To be concrete, we illustrate this explicitly in an example of a hyperloop in-
volving the three nodes I 1, I and I + 1, as before. For pI 1 = pI = pI +1 = 1,

we now have
AI
0 1
1
B
1+2 0 0 C
L0 = B
B
@
0 AI
0 C :
C
A
(4.25)
0 0 AI+1 + 12 cos
The explicit expression for AI1 is written in (C.5). We associate to this choice
of connection the quiver diagram in Figure 12a, though there are now some dif-
ferences: the loop preserves only two supercharges, so it is 1/8 BPS; the right
node has a shift of

1
2 cos instead of 1
2 and the left arrows are couplings to the ro-
tated elds r~I 1 2_ and their conjugates, rather than to the q~'s. The corresponding
hyperloop is
" # " #

1
W0 = sTr P exp i L0jx_ jds
Tr P exp i A +
I 1 jx_ jds
2
" #
+ Tr P exp i A jx_ jds

I
" #

1
+e i cos Tr P exp i A +
I +1 2
cos jx_ jds :
To proceed with the deformation, we dene as before the matrix
0 1
B
0 I 1 r~I 1 2_ 0 C
G= B
B
@
I 1 r
~I2_ 1 0 I qI
2 C
C
A
: (4.26)
0 I qI 2 0
Crucially, this matrix is such that i(Q+ )2 G = @' G i[L ; G ]. bos

The expression in
(4.26) can then be used to construct a deformed superconnection
L ; = L bos
+ iQ+ G + G 2 : (4.27)
Under the action of Q+

Q+L ; = i(Q+)2G + fQ+G ; Gg = @'G i[L ; G ] + fQ+G ; Gg : bos
(4.28)
This is easily seen to correspond to the supercovariant derivative
D'G @'G ifL ; ; G ] ;

so that the hyperloop with superconnection L ; transforms as a total derivative
under Q+.
To analyse the behavior under the action of Q it is useful to split G = G + G
with
0 1 0 1
B
0 0 0 C B
0 I 1 r~I 1 2_ 0 C
G=B
B
@
I 1 r
~I2_ 1 0 I qI
2 C
C
A
and G = B
B
@
0 0 0 C
C
A
; (4.29)
0 0 0 0 I qI 2 0
which satisfy Q G = Q+G and Q G = Q+G . One has then

Q L ; = i(Q+)2( G + G ) + fQ+(G G ); Gg
= @' ( G + G ) i[L0 ; G + G ] + fQ+ (G G ); Gg
= D' ( G + G ) + fQ+ (G G ); Gg fQ+ G ; G + G g + i[G 2 ; G + G ] ;
(4.30)
where we attempted to write the last line as a covariant derivative with a L ;

superconnection. For the loop to be invariant under Q , the three extra (anti)com-
mutators should vanish and while there are some cancellations among them, the
remaining terms are eliminated when G and G are nilpotent of degree 2 as in
(4.29). This implies that for generic and these Wilson loops are 1=8 BPS,
preserving exactly the same supercharges as L bos

.
Next we turn to the hyperloops with the quiver diagram in Figure 12b, but
7
again we take multiplicity equal to one in the central node. In this case
AI
0 1
B
1 0 0 C
L0 = B
B
@
0 AI + 21 0 C
C
A
: (4.31)
0 0 AI+1 + 1 cos
2
Note that now the connection in the bottom right corner is shifted by (1 cos )=2,
7 Again, we always use the same symbols L0 and G and refer to explicit equations to avoid
confusion.
8
though the corresponding node in the diagram is not squiggly. This does not
exactly match the shifts, which are no longer all equal to 1/2. However, we retain
the same notation in the gure to indicate the allowed couplings to matter elds
that preserve the supercharges. These are now organized in the matrix
0 1
B
0 I 1 r~I 1 1_ 0 C
G= B
B
@
I 1 r~I1_ 1 0 I qI1 C
C
A
; (4.32)
0 I qI 1 0
and the construction proceeds exactly as before to produce another family of 1=8
BPS loops.
To construct 1=16 BPS loops we again gauge transform the connection in (4.31)
to bring it to the form in (4.25). This results in the extra phases in the corre-
sponding G,
0 1
i'
B
0 I 1 e r~I 1 1_ 0 C
G! B
B
@
I 1 e i' r
~I1_ 1 0 I e i' cos q 1
I
C
C
A
: (4.33)
i' cos q
0 Ie I1 0
We can now add this expression to the G in (4.26) to get

0 1
I 1 r~I i'
B
0 1 2_ + I 1 e r~I 1 1_ 0 C
G= B
B
@
I 1 r
~I2_ 1 + I 1 e i' r
~I1_ 1 0 I qI
2 + I e i' cos q 1
I
C
C
A
:
0 I qI 2 + I ei' cos qI 1 0
(4.34)
This is similar to the sum of the two matrices in (4.19), but there is an obstruction
to using this G to construct a deformation, as the combination I qI2 + I e i' cos q 1

I
is not periodic. This is worse than the non-periodicity in (4.33), as it cannot be
xed by a gauge transformation. The only way to overcome this is to set either
I = I = 0 or I = I = 0, giving two branches of 1=16 BPS Wilson loops.

For I = I = 0 we can represent it by the quiver in Figure 14.
I 1
1 I 1 I 1 1 I 1
I 1 I
Figure 14: A quiver diagram for a 1=16 BPS Wilson loop.
particular shift of (1
8 The cos )=2 appears because supersymmetry requires a relative shift
between nodes I and I + 1 of ( cos )=2.
In the case of I = I = 0, it is a bit nicer to employ the gauge in (4.31) to
avoid the awkward phase multiplying qI1 . The diagram is in Figure 15.
I 1 I
1 I 1 I 1 1 I 1
I 1
Figure 15: The other class of 1=16 BPS loops.
4.3.2 A closer look at the hyperloops
Let us examine in detail the hyperloops with superconnection L ; in (4.27). To do

so we need the explicit expressions ofQ+ acting on the scalar elds qIa and r~I 1 a_ .
For the latter, we get back components of ~I 1 as (the on the ~ represent spinor
a
indices)
Q+r~I 1 1_ = ~I1 1;+ ; Q+r~I 1 2_ = ~I2 1; ;

Q+r~I1_ 1 = ~ ; Q+r~I2_ = ~I 1 2;+ :
I 1 1; 1
For the action on the qIa elds it is convenient to dene rotated fermions I a_ via
i' sin
I 1_ cos I 1_ + e I 2_ ; I 2_ cos I 2_ ei' sin I 1_ ; (4.35)
2 2 2 2
such that
Q+qI1 = I 1_ ; ; Q+qI2 = I 2_ ;+ ; Q+qI 1 = 1I;_ + ; Q+qI 2 = 2I;_ :

Hence we nd
0 1
1 I 1 r~I 1 2_ r
~I2_ 1 i I 1 ~I2 1; 1 I r~I 1 2_ qI C
I I 2
L ; = L
B

+B i I 1 ~I 1 2;+ 1 I 1 r
~I2_ 1 r~I 1 2_ + I I qI qI 2 i I I 2_ ;+ :
2 C
bos
B
@
I C
A
I 1 qI 2 r
I 2_
~I 1 i I 2I;_ I I qI 2 qI2
(4.36)
An interesting question is whether there are any points of enhanced supersym-
metry along our moduli space. A simple guide is to look for points of enhanced
bosonic symmetry that does not commute with the preserved supercharges, so
either SU (2)L or SU (2)R .

Looking to impose the SU (2)R symmetry, recall that it acts on the dotted
indices of r~ and ~. Examining (4.36), we immediately see that we should impose
I = I = 0 to eliminate the o-diagonal entries where dotted indices appear. This

has the eect of eliminating the entire third row and column of the supermatrix.
The corresponding diagonal entry in L0, which is AI+1, is also clearly not SU (2)R
symmetric, as can be seen from (C.5). So we should remove it from L0 as well and

focus on a loop coupling to only two nodes. Of course, it is possible to generalize
this to any even number of nodes, but with only couplings between pairs.
Focusing then on the upper-left 22 block of (4.36), there are still explicit
dotted indices in the diagonal parts, but those appear also in L0. Let us write
them together, say for the I -th node:

i a_ f b_ i a b
AI + I 1 I 1 r
2_
~I 1 r~I 1 2_ + I I qI qI 2 2 = A';I r~ M r~
k I 1 a_ I 1 b_ q M q ;
k I a Ib
(4.37)
with
0 1 0 1
f b_
1 0 1 0
Ma_ = @ A and Ma b = @ A : (4.38)
0 ik I 1 I 1 1 0 ik I I 1
M
We chose to write the matrix f in the basis of the twisted r~, but regardless of
the basis, in order to preserve SU (2)R , it has to be proportional to the identity,
so ik I 1 I 1 = 2.
The exact same structure follows for the top left entry, which is now also
q
symmetric. Recalling the C symmetry, we can further x I 1 = I 1 = 2i=k
and the full form of the connection becomes
0 1
1 a_ q ~2
A';I 1 + q~I ~Ia_ 1 + qI 1 q 1 qI 2 q 2 + 21 1p i
L = @
~
k I 1; A :
1p i
k I 1 2;+ AI;' + q~Ia_ 1 q~I 1 a_ + q1 qI 1 q2 qI 2
Note that the -dependence completely dropped out of this expression, so it is
within the class of 1/4 BPS operators presented in Section 4.2, but with extra
SU (2)R symmetry, so this is in fact the 1/2 BPS loop of [44], now adapted to the
3-sphere.
Imposing SU (2)L symmetry is similar, but the results are dierent. Looking
for undotted indices in (4.36), we see that now we should take I 1 = I 1 = 0.

This eliminates the top line and left column from the matrix and, as before, we
should remove the top left entry in L bos

as well.
We should again also examine the diagonal blocks, as we again have the ex-
pression in (4.37). Now f M = diag(1; 1) and we want

q
M to be proportional to
the identity, so we set ik I 1 = 2, or I = I = 2i=k. The resulting 2 2
connection is
0 1
AI;' + r~I1_ 1 r~I 1 1_ r~I2_ 1 r~I 1 2_ + q a qIa 1p i
k I 2_ ;+
L = @
1_
A:
1p i 2_
k I; AI +1;' + r~I 1 1_ r~I 1 r~I 1 2_ r~I2_ 1 + qIa qa + 21
(4.39)
The fermions in the o-diagonal entries are dened in terms of the original elds
in (4.35) and the scalar bilinears are
0 10 1
cos e i' sin A @q~I 1 1_ A
r~1_
I 1 r~I 1 1_ r~2_
I 1 r~I 1 2_ = q~I1_ 1 q~I2_ 1
@ ;
ei' sin cos q~I 1 2_
We recognize the same structure as the fermionic latitude of [25], so this is its
generalization to arbitrary N =4 Chern-Simons-matter theories. For =0 it
becomes another class of 1=2 BPS loop of [44].

In addition to the supercharges Q, this loop is also invariant under the ones
obtained by swapping undotted indices,

cos Q1l_ 2 + sin Q2r_ 2 ; cos Q2l_ 1 sin Q1r_ 1 :
2 2 2 2
The examples thus far were for the grading in Figure 12a. The story for the
second grading in Figure 12b is analogous, with the parameters coupling to
the remaining elds. In particular, the matrices M

f
and M in (4.38) would have
the upper left corner shifted by bilinears of . To preserve SU (2)R we now set
I = I = 0, and again have to focus on a superconnection involving only the

I 1 and I nodes with M as before and M f
= diag( 1; 1). This is the second
SU (2)R invariant 1/2 BPS Wilson loop described in [44]. Likewise, the SU (2)L
invariant hyperloop on this branch is another version of the fermionic latitude
loops described in [25] (in the notation of that paper it has l = 1).
The analysis here relies on a bosonic symmetry to indicate enhanced supersym-
metry. In principle there could be further points with accidental or more subtle
supersymmetry enhancement. We leave the study of that to the future.
4.3.3 Further examples
So far we focused on hyperloops involving three nodes of a long quiver. The
simplest generalization arising already in that case is taking multiple copies of
the connections, i.e. pJ > 1 in Figure 12, forming larger L

bos
connections and
involving more and/or parameters in the appropriate matrices G . This mirrors

the examples in Section 4.2, where the central node had multiplicity 2.
In the case of a linear quiver one may worry about the nal nodes which couple
either to hypermultiplets or to twisted hypermultiplets, although both kinds of
elds appear in (4.9) and (C.5). This turns out not to be a problem, and the
construction proceeds as before with the missing moment maps removed. For
example, suppose that there is no hypermultiplet to the left of the I 1 node
in the quiver of the underlying theory. When constructing the hyperloop as in
(4.25), this would mean that AI 1 would loose the contribution and become
i 1_ 2_
A';I 1 (~r 1 1_ r~I 1 r~I 1 2_ r~I 1 ) ;
k I
but the construction of the hyperloop would follow exactly as before. Likewise
if the underlying theory had no twisted hypermultiplet to the right of the node
I + 1, then AI+1 would become

i
A';I +1 ( 1 I +1 2 2 ) ;
k I +1 1
and again the construction would follow as before.
Of course, if we have only a 2-node quiver and no twisted hypermultiplets, there
would not be a way to -deform their couplings. In that case there would only
be the analog deformation of , which is completely parallel to the constructions
based on
~.
Next we examine what happens for longer quivers and for circular quivers like
ABJ(M) theory. Consider rst the hyperloop coupling to 4 nodes as in Figure 16.
pI 1 I 1 pI I pI +1 I +1 pI +2
I 1 I I +1
Figure 16: A quiver diagram for a 4-node 1/8 BPS hyperloop.
The story proceeds exactly as before, except that we should remember that
the shifts are relative to neighboring nodes, so in this case (for all pJ = 1) the
starting point is
AI
0 1
B
1 + 21 0 0 0 C
L =
B
B
B
0 AI 0 0 C
C
C :
bos B
B
@
0 0 AI +1 + 21 cos 0
C
C
A
0 0 0 AI +2 + cos2 1
4.4 A matrix model proposal 81
It should be clear how to deform this loop by adding couplings to the fermions and
also construct the loop corresponding the quiver with the second possible grading.
In the case of circular quivers, turning on = 6 0 poses a challenge. In the

simplest case of ABJ(M) we have two nodes, so L = 2. Then in our gures and
expressions for L , etc. we can take I = 2 and identify the nodes I

bos 1 = 1 and
I + 1 = 3. If we consider a hyperloop coupling only to one edge in the quiverso
the quiver for the hyperloop is a linear 2-node quiverthere is no problem. If we
want to couple to matter from both nodes, we face the fact that while we identify
nodes 1 and 3, the shift of A1 is 21 and that of A3 it should be 21 cos , see (4.25).
The solution to this problem was already anticipated in [51] (in other contexts
where the shifts were not 1=2) and it amounts to taking a cover of the original
gauge theory quiver. So we can couple the Wilson loop to both edges of the
original quiver as long as we consider a 3-node hyperloop with L

bos
given by
A1 + 21
0 1
B
0 0 C
L0 = B
B
@
0 A2 0 C
C
A
: (4.40)
0 0 A1 + 12 cos
Here A1 appears twice with dierent shifts. One can continue further with another
copy of A2 with a shift of (cos 1)=2, and so on.
6 0 the connection in (4.40) has (C)3 symmetry (with one copy acting
For =
trivially), so the moduli space of 1=4 BPS loops according to (4.21) is
C2 C2==C C ' C2 :
The = 0 case has enhanced symmetry (C )3 ! GL(2; C) C . The moduli space
according to (4.22) is now four-dimensional
C6==GL(2; C) :
In addition to the very special case of = 0, for any rational cos , a hyperloop
based on a long enough quiver will have also some enhanced symmetry.
4.4 A matrix model proposal

The construction of the hyperloops is based on a deformation of a bosonic loop and
all the loops of xed are cohomologically equivalent under the supercharge Q+
used to dene them. This means that any localization computation for any loop
4.5 Comments 82
on the moduli space is immediately applicable to any one. The proof following
[24, 44] requires expanding the exponentials and checking order-by-order that the
dierence between the dierent operators is Q+-exact. We do not reproduce this
computation here since it is essentially identical.
We propose now a matrix model that we hope captures the expectation value
of our operators. The matrix model partition function can be motivated by con-
sidering the usual ingredients due to the vector multiplets (hyperbolic sines) and
hypermultiplets (hyperbolic cosines) at each node [5] and the proposal in [59] on
how to introduce the -deformation:

Y
NI d NI
1 Ii i 4kI 2Ii Y ( Ij ) Ii Ij
4 sinh Ii
Y
Z= e sinh
I NI ! i=1 2 i<j 2 2
0
N N
1 1
I +1
I Y ( 1)I ( I +1;j ) A
Ii
Y
@ 2 cosh :
i=1 j =1 2

The explicit value of the parameter can be xed by the comparison with a
p
perturbative computation and turns out to be given, in our notation, by = cos
[59]. Note that twisted and untwisted hypermultiplets contribute dierently, with
either in the numerator or in the denominator of the argument of the hyperbolic

cosines. The expectation value of the -deformed hyperloops is given by inserting
P PN
I Ii .
in the partition function above I i=1 e
4.5 Comments
So far we reorganized the space of known Wilson loop operators in N = 4 Chern-
Simons-matter theories in three dimensions, which we now call hyperloops, and
generalized it considerably to include loops preserving 1, 2, 4 and 8 supercharges.
Our ndings clarify and elaborate the intricate structure of the supersymmetric
line operators and their moduli spaces. The strategy, adapted from [20, 51], is to
start with a choice of diagonal (bosonic) superconnection, which is at the apex of a
conical moduli space, and describes all other superconnections as deformations of
it. The operators we have thus obtained are classied in terms of quiver diagrams
encoding which gauge elds are involved and the couplings to the matter elds.
This, together with a latitude parameter , completes the set of data necessary
for the classication.
More concretely, we need to choose some or all of the vector elds and the
number of times they are represented in the Wilson loop. The next step is to
4.5 Comments 83
choose a grading, which plays two roles: it indicates the constant shifts in some
of the diagonal connections and it implies which half of the chiral elds we couple
to, in order to get hyperloops with 4 preserved supercharges. The two possible
gradings then also give the two branches of the moduli space. Finally, for each of
the included chiral (antichiral) elds we have couplings I, I ( I , I ) subject to

a global gauge invariance, reducing the moduli space to a quotient of Cp for some
p. These moduli spaces are known as quiver varieties and generically are conical,
as rst observed in [20] for the conifold in the case of the 1=6 BPS loops of the
ABJ(M) theory.
The hyperloops with = 0 were previously found in [51], as well as in [27, 44,
4749], but our description is much more algorithmic and their moduli space was
never studied in such detail. Loops with 6= 0 were only studied in ABJ(M) theory
and not in theories with N = 4 supersymmetry. Furthermore, the constructions
focused on the bosonic loops and on the analogues of the 1/4 BPS loops (which
are 1=6 BPS in ABJ(M)) presented in Section 4.3.2. The continuous family of
hyperloops interpolating between those two cases and all the other directions in
the moduli space have not been previously described.
As mentioned in Section 4.1, the Gaiotto-Yin loop breaks SU (2)L SU (2)R !

U (1)L U (1)R . The parameter further breaks U (1)R by including extra cou-
plings to the moment maps arising from the twisted hypermultiplets. All the
examples that are presented in the preceding sections have analogues with the
roles of hypermultiplets and twisted hypermultiplets reversed.
There are many possible directions that can be pursued from here. Among the
most obvious ones is to attempt the complete exploration of the full moduli space
of line operators in three-dimensional Chern-Simons-matter theories. In addition
to the tried and tested approach of making ansätze and restricting them to be BPS,
we can try to extend the point of view introduced in [20] and employed here, of
constructing the loops as deformations of previously identied ones. One should
rst verify whether there are further line operators involving only single nodes.
Then whether there are any further deformations of them with more complicated
forms than considered here. Finally, one should examine other points along the
moduli space to see whether there are other branches that may intersect those
points, but not pass through the origin we employ here.
Still, these explorations cannot answer the question of what is the full space of
BPS line operators, which would require new tools to address. Moreover, there are
other types of line operators, known as vortex loops [36, 60, 61]. In some cases they
4.6 New hyperloops 84
are known to be dual under mirror symmetry to Wilson loops [35], so it should
be exciting to understand their moduli spaces as well.
In Section 4.3.2 we looked at some special examples of these loops which have
enhanced supersymmetry. It is not clear that the ones identied there, which
were all previously known, are the only points of enhanced supersymmetry on the
moduli space.
Also in this spirit of discovering new hyperloops would be the construction
of operators supported along generic curves on a S 2 S 3, following the four-
dimensional example of [62, 63, 33] or the ABJ(M) analog of [25].
Another question worth asking is what happens to these operators at the quan-
tum level and whether the classical moduli spaces described here receive correc-
tions. The analysis of [52] suggests that moduli spaces are natural for line operators
in three dimensions, but it does not predict their dimensions, as we found here,
nor that their classical structure is not subject to quantum corrections. The heroic
9
3-loop calculation of [64] suggests that the degeneracy among pairs of 1/2 BPS
Wilson loops of [44] may sometimes get lifted. Such a perturbative analysis would
also be useful to test the matrix model proposal we put forward in Section 4.4.
Yet another angle is to study these moduli spaces as defect conformal manifolds
in the context of defect CFT. Explicit analysis of this type for line operators in
four-dimensional N = 4 theory include [6668] and in three dimensions [69].

Wilson loops are also interesting in the holographic context, for they provide a
rich dictionary between gauge theory and string theory objects. It would then be
interesting to understand the holographic realization of the operators constructed
here. Very little has been done in this direction since the original proposal for
the holographic dual in [2830]. Proposals for the holographic duals of 1=2 BPS
loops in some N = 4 theories were put forward in [48] and a rst examination of
a possible moduli space of 1=6 BPS loops in ABJ(M) theory was done in [70].
4.6 New hyperloops

To conclude this chapter, we present some preliminary results concerning more
general hyperloops inN = 4 CSm theories. This is part of an ongoing project [23]
that started from the N = 4 analogue of the loops presented in Section 3.3. So
let us begin by presenting such operators and, in particular, by translating the
supercharges in (3.76) to our setting. To do so we have to map not only R-
9 See also [65] for a previous attempt limited to a two-loop computation.

symmetry indices from SU (4) to SU (2)L SU (2)R but also supercharges in R3 to

S 3. We map R-symmetry indices as
f1; 2; 3; 4g 7! f1_ ; 2_ ; 1; 2g ;
and supercharges as
Q+ iS + Ql ; Q iS Ql ; Q + iS Qr ; Q+ + iS + Qr ;
to nd that the N =4 analogue should be invariant under 4 supercharges that
are linear combinations of Q1l_ 1 and Q2r_ 1 , Q1r_ 1 and Q2l_ 1 , Q1l_ 2 and Q2r_ 2 , Q1r_ 2 and Q2l_ 2 .
The null entries of the scalar coupling matrix appearing in (3.75) translate to
N = 4 as the absence of scalars from one of the matter multiplets. Through the
map of R-symmetry indices above, this should be the untwisted multiplet, such
that the dressed bosonic connection appears as
AI = AI;' + ki (~I 1_ 1_ ~I 2_ 2_ ) : (4.41)
As for the o-diagonal entries we nd that the two fermi elds appearing through
and should then translate to fermi elds of N = 4 carrying dotted indices, so

the operator should couple to fermions in the hypermultiplet and not to ones in
the twisted hypermultiplet. At the end we nd a superconnection given by
0
i 1_ _ i i'

1
A';I (~ ~I 2_ )

2
I e
L= i B
k I 1_ 2 I I 1_ I 2_ + C
C ;
i
B
@
I1_ + I e i' I2_

_ _ 1 A
(4.42)
I A';I +1 (~I +1 1_ ~I +1 2_ )
1 2
2 k 2
and, for constants f I ; I g and f I ; I g satisfying I I = I I = 2i=k, the
resulting loop is invariant under the action of the following supercharges
I I I I
Q1l_ 1 Q2r_ 1 ; Q1l_ 2 Q2r_ 2 ; Q1r_ 1 Q2l_ 1 ; Q1r_ 2 Q2l_ 2 : (4.43)
I I I I
One novelty here is that the set of preserved supercharges carry this unusual
; dependence. To try to understand the structure of this operator and its pre-
served supercharges, we attempted to employ the logic proposed in [20]. However,
note that supercharges in (4.43), are not linear combinations of the ones preserved
by the 1-node loop, see (4.8). So we should not expect to connect L above with
L bos through L = iQG + G 2 .
What we nd is that this operators can actually be seen as a composition of
two hyperloops that are 1=2 BPS. These two are dened in terms of
0
i 1
A';I ( I + ~I 1_ 1_ ~I 2_ 2_ ) i I I 1_

L ; = B
k C
i 1 ;
B C
i I I1_+
@ A
A';I +1 ( + ~ 1_ ~I +1 2_ 2_ )
k I +1 I +1 1_ 2
and
0
i 1 1
A';I (I + ~I 1_ 1_ ~I 2_ 2_ ) i I I 2_ +

L ; = B
k 2 :
C
i
B C
i I I2_
@ A
A';I +1 (I +1 + ~I +1 1_ 1_ ~I +1 2_ 2_ )
k
The corresponding operators, to which we refer below respectively as W ; and
W ;, are SU (2)L symmetric. They preserve all supersymmetry transformations
preserved by L bos plus the ones obtained by exchanging undotted indices, namely,
Q1l_ 1 ; Q1r_ 1 ; Q2l_ 2 ; Q2r_ 2 and Q1l_ 2 ; Q1r_ 2 ; Q2l_ 1 ; Q2r_ 1 : (4.44)
In particular, they match what in [44] was named 1 -loop and 2 -loop. Employing
the analysis of [20] and [51], they can be obtained from L bos
+
using Q and L
0 1 0 1
0 I qI1 0 I qI A
2
G ; = @ A or G ; = @ ;
I qI 1 0 I qI 2 0
together with the respective constraints I I = 2i=k or I I = 2i=k, responsible

for rendering the loop SU (2)L symmetric. To them we can associate the quiver
diagrams in Figure 17.
I
pI I pI +1
(a)
pI I pI +1
I
(b)
Figure 17: Quiver diagrams representing 1=2 BPS Wilson loops (a) W ; and (b)
W ;.
Now the superconnection in (4.42) can be seen as a composition of L ; with

the following gauge transformed version of L ;,

0 1
A';I i
k (I + ~I 1_ 1_ ~I 2_ 2_ ) i I e
i' _
I 2+
L0 ; = B
i 1A :
C
i I e i' I2_
@
A';I +1 (I +1 + ~I +1 1_ 1_ ~I +1 2_ 2_ )
k 2
Indeed, we precisely have that it corresponds to

L ; + L0 ; .
2
The reasoning behind the particular choice of L ; and L0 ; and supercharges
in (4.44) is that, under the variation generated by these, the superconnections
vary as the supercovariant derivative of precisely the same supermatrix. More

I
precisely, under Q1l_ 1 Q2r_ 1 , for example, we have
I
L ; = D' H1 = @' H1 i[L ; ; H1 ] ;

(4.45)
L0 ; = D' H1 = @' H1 i[L0 ; ; H1 ] ;
with 0 1
0 I ( l ) qI1
H1 = @ A :
I ( r )+ qI 2 0
For the other three supercharges we have analogous versions of (4.45). In particular
we have that under the variation generated by
I I I
Q1l_ 2 Q2r_ 2 ; Q1r_ 1 Q2l_ 1 ; Q1r_ 2 Q2l_ 2 ;
I I I
we obtain the corresponding supercovariant derivatives of
0 1 0 1
0 I ( l ) qI2 0 I ( r) qI1
H2 = @ A ; H3 = @ A ;
I ( r )+ qI 1 0 I ( l )+ qI 2 0
and 0 1
0 I ( r) qI2
H4 = @ A ;
I ( l )+ qI 1 0
respectively.
Thus we nd that supersymmetry of L in (4.42) automatically holds. Indeed,

for the four supercharges in (4.8), we nd
!
1 1
L = ( L ; + L0; ) = @' Hi i[L ; ; Hi ] + @' Hi i[L0 ; ; Hi ]
2 2
= @' H i i
L ; + L ;
0 "
; Hi
#
2
= D' H i ;
with i = 1; 2; 3; 4. From this derivation we were able to see that we can actually
relax the condition on L ; and L0; appearing with the same weights in L. We
nd that, as long as theirs weights sum to one, the derivation above holds. I.e.,
for some constant parameter , the operator built out of
L = (1 ) L ; + L0; ; (4.46)
is invariant under all four supercharges since, following the same steps as above,
we obtain
h i
L = @' Hi i (1 ) L ; + L0; ; Hi = D' Hi ;
with i = 1; 2; 3; 4. Thus we nd a one parameter family of operators parameterised
by that generically is 1=4 BPS. At the particular points = 0 and = 1 the

corresponding operators are 1=2 BPS.
Of course the presentation within this section is mostly based on the construc-
tion introduced in Section 3.3 in the context of ABJ(M) theory. Consequently
it also relies heavily on a particular guess of ansätze for the couplings. However,
the discussion here is neater specially because we are on S 3. This spares us from
working with projectors and all the spinorology we painfully faced in Section 3.2.
Fortunately, it allowed us to understand further aspects of the operators we found
and in particular nd the one parameter family in (4.46). Furthermore, it opened
our eyes to the possibility of generalizing the deformation procedure introduced
in [20]. As exemplied by this new 1=4 BPS operator, not all supersymmetric
loops are continuously connected to Lbos through L = iQG + G 2 . Indeed, the
example above points towards the direction of using the superconnection of a 1=2
BPS operator as the starting point upon which we should build our deformation.
Since a 1=2 BPS operator furnishes a wider set of preserved supercharges than
Lbos , we should be able to recover from it a wider spectrum of supersymmetric
loops. This is currently under investigation and should be presented soon in [23].
Appendix A
ABJM(M) theory
We denote the coordinates as x = fx0 ; x1 ; x2 g. Spinor indices = +; are
lowered and raised as (

) = ( ) where

0 1 0 1
0 1A 0 1A
=@ and =@ :
1 0 1 0
When spinor indices are omitted, the contraction is understood as
= ;: (A.1)
In Lorentzian space R1;2 the metric is given by g = diag( 1; 1; 1) and the
gamma-matrices are chosen as ( ) = f i 3 ; 1 ; 2 g.

In Euclidean space, the metric is g = diag(1; 1; 1) and the gamma-matrices
are chosen as ( ) = f 3 ; 1 ; 2 g.
The supersymmetry transformations are as written in [25]:
!
4i IJ 1
A = ( ) CI J + IJKL K C L ;
k 2
!
4i IJ 1
A^ = ( ) J CI + IJKL C L K ;
k 2
4i IJ
I = 2i IJ ( ) D CJ (CJ C K CK CK C K CJ )
k
8i JK I
CJ C CK 2iIJ CJ ;
k (A.2)
2i
I = i KL IJKL ( ) D C J + KL IJKL (C J CM C M
k
4i
C M CM C J ) + KL KLMN C M CI C N iKL IJKL C J ;
k
KL
CI = IJKL ; J
C I = 2 IJ J ;
89
90
where IJ = IJ (x )IJ is a conformal Killing spinor. The Poincaré ( IJ ) and
superconformal parameters ( IJ ) satisfy the relations
1
IJ = JI ; IJ = IJKL KL ;
2
1
IJ = JI ; IJ = IJKL KL :
2
In Lorentzian space, they obey the reality condition IJ = (IJ ) and IJ = (IJ ) .
In Euclidean space, there is no reality condition, i.e., 6= y and 6= y .
From the Euclidean action (3.1) one has the following Feynman rules in Landau
gauge:
Tree-level vector propagators:

D E(0)
(x y )
2i
! 3
2
Aa (x)Ab (y) = ab ;
k 2 [(x y )2 ]
3
2
3
2
D
â ^b
E(0)
ab 2i
! 3
2 (x y )
A (x)A (y) = ;
k 2 [(x y )2 ]
3
2
3
2
One-loop vector propagators:

D E(1)
2
!2 2 3
"
(x y )2
#
Aa (x)Ab (y) = ab N2 23 2 @ @ ;

k 4 [(x y )2 ]1 2 4(1 + 2)

D E(1)
2 2
! 2 3 " (x y )2
#
Aâ (x)A^b (y) = ab N1 2

@ @ ;
k 4 3 2 [(x y )2 ]1 2 4(1 + 2)
Scalar propagator:

^j
(0) 1 1
(CI )i (x)(C J )(y )lk^ = IJ il k^^j 2
;
4 [(x y )2 ]
3 1
2 2
Tree-level fermion propagator:

D E(0) ( ) (x y )3
( I )^ji (x)( J )^lk (y ) = iIJ î^l kj 2
;
2 [(x y )2 ]
3 3
2 2
One-loop fermion propagator:

D E(1)
2
! 2 1 1
( I )^ji (x)( J )^lk (y ) = iIJ î^l kj (N2 N1 ) 2
:
k 16 3 2 [(x y )2 ]1 2
The interaction piece of the action, S int , contains the following vertices:
91
Gauge cubic vertex

ik
d3 x f abc Aa Ab Ac ; (A.3)
12
where f abc is the structure constant of U (N1 );
Gauge-fermion cubic vertex
!
d3 x Tr I i A I A^ I : (A.4)
Appendix B
Ap and B p methods
The Ap method was proposed in [37] as a recursive algorithm to compute integrals
of the form
2 1
1 ein +in
1 1 2 2
Apn1 ;n2 = d1 d2 ; (B.1)
4 2 0 0 (ei ei )p
1 2
where p 2 Z+ and n1 ; n2 2 Z. For p = 0 this is simply A0n ;n = 0;n 0;n .

1 2 1 2
For p > 0, instead of performing the integrals, we view them as formal objects
satisfying the parity condition and recurrence relations, arising from combinations
of integrands with factorisable numerators,
0 1
p p
X
p
Apn ;n = ( 1)p Apn ;n ; @ A ( 1)k A
n +p k;n2 +k = A0n ;n : (B.2)
k=0 k
1 2 2 1 1 1 2
For p = 2 this is solved by

1
A2n ;n = jn1 n2 j2;n +n + Jn(1)+n ; (B.3)
1 2
4 1 2 1 2
with arbitrary Jn(1) . For p = 4 we have instead

1
A4n ;n =
1 2
jn
96 1
n2 j((n1 n2 )2 4)4;n +n + Jn(2)+n (n1 n2 )2 + Jn(3)+n :
1 2 1 2 1 2
(B.4)
The arbitrary J 's (Jn(1) , Jn(2) , Jn(3) and so on) should contain the divergences of the
integrals.
For half-integer valued n1 and n2 , we have that Apn ;n 1 2

should still satisfy par-
ity condition and recurrence relations outlined in (B.2), but now with boundary
condition
1 1
A0n ;n = :
1 2
2 n1 n2
92
93
In this case, for p = 2 we nd

(1 n2 )PolyGamma[0; n2 ] + (1 n1 )PolyGamma[0; 2 n1 ]
A2n ;n = + Kn(1)+n ;
1 2
( 2 + n 1 + n 2 ) 2 1 2
(B.5)
and for p = 4 we have instead

"
1
A4
n1 ;n2 = ( 3 + n2 )( 2 + n2 )( 1 + n2 ) PolyGamma [0; 4 n2 ]
12(4 n1 n2 ) 2
!
+ PolyGamma[0; n2 ] + ( 3 + n1 )( 2 + n1 )( 1 + n1 ) PolyGamma [0; 4 n1 ]

!#
3n2 2 + 3n1 2 3(n2 + n1 ) 4n2 n1
+ PolyGamma[0; n1 ] +
24 2
+ Kn(2)+n (n1
1 2
n2 )2 + Kn(3)+n ;1 2
(B.6)
with arbitrary Kn(i) .

In analogy to the derivation above, the Bp method we propose is a recursive
algorithm to compute integrals of the form
2
1 ein
Bnp = d ; (B.7)
2 0 (ei 1)p
for p 2 Z+ . Here we view Bnp symbols as formal objects satisfying the recurrence
relations
0 1
pp
X
@ A ( 1)k+p B
p 0
n+k = Bn ; (B.8)
k=0 k
i
where Bn0 = . For p = 1 we nd
n
(Sign[n] + (n 1)n;0 )
8
>
>
< ;n2Z
Bn1 = > iPolyGamma 2 ;
[0; n] (B.9)
>
: ; n + 12 2 Z

while for p = 3 we have
8
>
>
<
jn
1j(n 2)
+ n T (0) + n2 T (1) ;n2Z
3 4
Bn = > (6n 2n2 4)PolyGamma[0; n] ;
>
: + nR + n R
(0) 2 (1) ; n+ 2 2Z
1
4i
(B.10)
94
with arbitraryT (i) and R(i) .

Constants J , K , T and R that arise in the recursive methods outlined above
are not related to each other.

Appendix C
N = 4 theories
The supersymmetry transformations of the N =4 Chern-Simons-matter theory
on S 3 can be seen to be given by

i
A I = ab_ (jIab_ |~Ib_ a ) ;
k
qI = ab_ I b_
a
qI a = ab_ Ib_ ;

q~I 1 b_ = ab_ ~Ia 1 ;
q~Ib_ 1 = ab_ ~I 1 a ;
i 2i
I a_ ba_ (I qIb qIb I +1 ) + bc_ ~I c_ a_ qIb qIb ~I +1 a_ c_ ;
= i ba_ D qIb + iba_ qIb
k k
i 2i
I = i D qI b + i qI b
a
_ b a
_ b a
_ (qI b I I +1 qI b ) + bc_ qI b ~I a_ c_ ~I +1 c_ a_ qI b ;
b a
_
k k
i
~Ia 1 = i ab_ D q~I 1 b_ i ab_ q~I 1 b_ + ab_ (~qI 1 b_ ~I ~I 1 q~I 1 b_ )
k
2i bc_
(~qI 1 c_ I ab I 1 ba q~I 1 c_ ) ;
k
i
~I 1 a = i ab_ D q~Ib_ 1 iab_ q~Ib_ 1 + ab_ (~I q~Ib_ 1 q~Ib_ 1 ~I 1 )
k
2i b c_
bc_ I a q~I 1 q~I 1 I 1 ab :
c_
k
(C.1)
where ab_ = 31 r ab_ . More specically, from (4.2) one nds al;b_l = 2i al;b_l and
ar;b_r = 2i ar;b_r. We work in Euclidean signature and take the gamma-matrices,
( ) , to be given by the Pauli matrices. As usual, the spinor contractions are
such that
1 2 1 2; = +2 1 ; 1 2 1 ( ) 2; = 2 1 ; ; = :
95
96
Then it follows that the Killing spinors on l l = l l = 1 and

S3 of (4.3) satisfy
l l = l l = ' , and similarly for the contractions involving r and r.

The expressions in (C.1) can be motivated by relating them to the transforma-
tions of the N =2 Chern-Simons-matter theory on S3 written down in [45, 46].
The o-shell supersymmetry transformation of the physical elds from the vec-
tor multiplets ( A , , D, , ) and chiral multiplets ( , , F) in the (;

)
bifundamental and their conjugates are
i
A I = ( I I ) ;
I = I ; I = I ;
2
i
I = i D I + i(I I I I +1 ) + D I + FI ; (C.2)
3
i
I = i D I i(I +1 I I I ) + I D + FI :

3
To match with N = 4 theories we go on-shell, so use the actions 1

p
" #
kI 2i
(I )
SCS = d3 x gTr p (A @ A A A A )
I I I I I I I + 2DI I ;
4 g 3

p
"
(I ) 3i
iS = d3 x gTr iD I D I + I D I + I (I I I +1 )
matter
4l2 I I I
I (I I I I +1 ) + I ( I I I +1 )

I I (DI I I
#
DI +1 ) + iI (I I I 2I I I +1 + I I +1 I +1 ) + iFI FI ;
to integrate out the auxiliary elds , , , D, F and F .

This yields the on-shell transformations
" #
2i
A I = (I 1 I 1 I I ) (I I I 1 I 1 ) ;
kI
I = I ;
I = I ;
" #
2i i
I =i D
I + (I 1 I
1 I I )I I (I +1 I +1 I I ) + D I ;
kI 3
" #
2i i
I = i D
I ( ) I (I 1 I I I ) + I D :
kI I +1 I +1 I I I 1
3
(C.3)
Now we use the chiral decomposition of N =4 hypermultiplets and twisted
hypermultiplets in Figure 11. The chiral elds in this representation are q2 and
1 Note that for the N = 4 theory the CS-levels are alternating, kI = ( 1)I k.
C.1 Double transformations of the elds 97
q~1_ . Extending to the other elds in the multiplets we match (C.1) to (C.3) with
the replacements
q~1_ ; q2 ! ; ~1 ! ; q~1_ ; q2 ! ; 2_ ; ~ ! :

2_ ; 1
where we also identied the supersymmetry parameters as 11_ ! and 22_ ! .
A mismatch by 2 in the non-linear terms can be xed by rescaling the elds.
The other elds transform in the conjugate (
; ) representation, and with the
same choice of supersymmetry parameter identication they would be matched to
N = 2 elds in this representation according to

q~2_ ; q1 ! ; 1_ ; ~ ! ; q~2_ ; q1 ! ; ~2 ! :
2 1_ ;
C.1 Double transformations of the elds

The main ingredient in the construction of fermionic Wilson loops are the double
transformations of scalar elds. Using (C.1), we write these as
[1 ; 2 ]qIa = i(1ab_ 2bb_

2ab_ 1bb_ )D qIb i(1ab_ 2bb_ 2ab_ 1bb_ )qIb
i 2i aa_
+ (1ab_ 2bb_ 2ab_ 1bb_ )(I qIb qIb I +1 ) (1 2bb_ 2aa_ 1bb_ )(~I b_ a_ qIb qIb ~I +1 a_ b_ )
k k
[1 ; 2 ]~qI 1 a_ = i(1aa_ 2 2aa_ 1 )D q~I 1b_ i(1aa_ 2ab_ 2aa_ 1ab_ )~qI 1b_
ab_ a _
b
i
+ (1aa_ 2ab_ 2aa_ 1ab_ )(~qI 1b_ ~I ~I 1 q~I 1b_ )
k
2i
( bb_ 2aa_ 1bb_ )(~qI 1b_ I ab I 1 ab q~I 1b_ ) :
k 1aa_ 2
(C.4)
We specialize the double transformations to the supercharges (4.12), whose
corresponding parameters are given by
! !
l
1;ab_ = 1
a b1__ cos + b2__ sin r ; 2;ab_ = 2
a b2__ cos l b1__ sin r :
2 2 2 2
The Killing spinors in (4.3) obey (4.2), from which one sees that the second term
in (C.4) becomes
1
i(1ab_ 2bb_ 2ab_ 1bb_ )qIb = cos 1a qI1 2a qI2 :
2
Combining with the rest of (C.4) one nds
i 1
ifQ1 ; Q2 gqI1 = D' qI1 cos qI1 + (I qI1 qI1 I +1 )
2 k
1 1_
1
r~I 1 r~I 1 1_ qI1 + qI1 r~I +1 1_ r~I1_+1 r~I +1 2_ r~I2_+1 ;
r~I2_ 1 r~I 1 2_
k k
i 1
ifQ1 ; Q2 gqI2 = D' qI2 + cos qI2 ( q 2 q 2 )
2 k I I I I +1
1 1_
1
r~I 1 r~I 1 1_ r~I2_ 1 r~I 1 2_ qI2 + qI2 r~I +1 1_ r~I1_+1 r~I +1 2_ r~I2_+1 ;
k k
with elds r~ dened in (4.23). Noting that
I qI1 qI1 I +1 = (I 11 I 22 )qI1 + qI1 (I +1 11 I +1 22 ) ;

I qI2 qI2 I +1 = (I 11 I 22 )qI2 qI2 (I +1 11 I +1 22 )
we can write
i
ifQ1 ; Q2 gqIa = D' qIa cos 1a qI1 2a qI2 i(AI A';I )qIa + iqIa (AI+1 A';I +1 ) :
2
where AI+1 is given by the natural generalization of (4.24) to the I + 1-th node
AI1 = A';I 1 ki (I 111 I 122 + r~I 1 1_ r~I1_1 r~I 1 2_ r~I2_1) : (C.5)
The covariant derivative is D' qIa = @' qIa iA';I qIa + iqIa A';I +1 , so that the double
transformation can be recast as a total covariant derivative with respect to the
A -connection, including the coupling to a background eld

i
ifQ1 ; Q2 gqI1 = D' qI1 @' qI1 cos qI1 iAI qI1 + iqI1 AI+1 ;
2 (C.6)
i
ifQ1 ; Q2 g I = D' I @' I +
q2 q2 q2 cos q 2 I iA I I A
q 2 + iq 2 :
I I +1
2
Note thatqI1 and qI2 are charged oppositely with respect to the background eld.
We repeat the computation for the r ~a_ 's arriving at
i 1
ifQ1 ; Q2 gr~I = D' r~I 1 1_ + r~I 1 1_
1 1_ (~r ~ ~ 1 r~I 1 1_ )
2 k I 1 1_ I I
1
+ r~I 1 1_ (I 1 1 I 2 2 ) (I 1 1 1 I 12
2 )~
r I 1 1_ ;
k
i 1
ifQ1 ; Q2 gr~I 1 2_ = D' r~I 1 2_ r~I 1 2_ + (~rI 1 2_ ~I ~I 1 r~I 1 2_ )
2 k
1
+ r~I 1 2_ (I 1 I 2 ) (I 1 1 1
1 2 I 12
2 )~
r I 1 2_ :
k
It is easy to see that these are also given by total derivatives
i
ifQ1 ; Q2 gr~I 1 1_ = D' r~I 1 1_ @'r~I
+ r~I
1 1_ 1 1_ iAI 1 r~I 1 1_ + ir~I A
1 1_ I ;
2 (C.7)
i
ifQ1 ; Q2 gr~I 1 2_ = D' r~I 1 2_ @' r~I 1 2_ r~ 1 2_ iA r~
I 1 I 1 2_ + ir~I 1 2_ I ; A
2 I
with, again, the two components oppositely charged with respect to the back-
ground eld. We also need the corresponding expressions for the conjugate elds,
which are given by
i
ifQ1 ; Q2 gqIa = D' qIa @' qIa ( 1)a cos qIa iAI+1 qIa + iqIa AI ;
2 (C.8)
i
ifQ1 ; Q2 gr~Ia_ 1 = D' r~Ia_ 1 @' r~Ia_ 1 + ( 1)a_ r~Ia_ 1 iAI r~Ia_ 1 + ir~Ia_ 1 AI 1 :
2
Bibliography
[1] J. Schwinger, On gaunge invariance and vacuum polarization, Physical

Review (U.S.) Superseded in part by Phys. Rev. A, Phys. Rev. B: Solid
State, Phys. Rev. C, and Phys. Rev. D .
https://www.osti.gov/biblio/4422049. 1
[2] K. G. Wilson, Connement of Quarks, Phys. Rev. D 10 (1974)

24452459. 1
[3] N. Drukker and D. J. Gross, An exact prediction of N = 4 SUSYM theory

for string theory, J. Math. Phys. 42 (2001) 28962914, hep-th/0010274.
1, 3
[4] V. Pestun, Localization of gauge theory on a four-sphere and
supersymmetric Wilson loops, Commun. Math. Phys. 313 (2012) 71129,

arXiv:0712.2824. 1, 3
[5] A. Kapustin, B. Willett, and I. Yaakov, Exact results for Wilson loops in
superconformal Chern-Simons theories with matter, JHEP 03 (2010) 089,

arXiv:0909.4559. 1, 19, 82
[6] M. F. Atiyah and R. Bott, The Moment map and equivariant cohomology,
Topology 23 (1984) 128. 1
[7] N. Berline and M. Vergne, Classes caractéristiques équivariantes. formule
de localisation en cohomologie équivariante, CR Acad. Sci. Paris 295

no. 2, (1982) 539541. 1
[8] A. M. Polyakov, String theory and quark connement, Nuclear Physics B

- Proceedings Supplements 68 no. 1-3, (Nov, 1998) 1â8.
http://dx.doi.org/10.1016/S0920-5632(98)00135-2. 2
[9] J. Maldacena, Wilson loops in largeneld theories, Physical Review

Letters 80 no. 22, (Jun, 1998) 4859â4862.
http://dx.doi.org/10.1103/PhysRevLett.80.4859. 2, 3
[10] G. 't Hooft, Dimensional reduction in quantum gravity, Conf. Proc. C

930308 (1993) 284296, arXiv:gr-qc/9310026. 2
100
BIBLIOGRAPHY 101
[11] L. Susskind, The World as a hologram, J. Math. Phys. 36 (1995)

63776396, arXiv:hep-th/9409089. 2
[12] G. 't Hooft, A Planar Diagram Theory for Strong Interactions, Nucl.
Phys. B 72 (1974) 461. 2
[13] N. Drukker, D. J. Gross, and H. Ooguri, Wilson loops and minimal
surfaces, Phys. Rev. D60 (1999) 125006, hep-th/9904191. 3
[14] O. Aharony, O. Bergman, D. L. Jaeris, and J. Maldacena, N=6
superconformal Chern-Simons-matter theories, M2-branes and their gravity
duals, JHEP 10 (2008) 091, arXiv:0806.1218 [hep-th]. 3, 14
[15] O. Aharony, O. Bergman, and D. L. Jaeris, Fractional M2-branes, JHEP

11 (2008) 043, arXiv:0807.4924 [hep-th]. 3, 14
[16] D. Gaiotto and E. Witten, Janus congurations, Chern-Simons couplings,
and the -angle in N = 4 super Yang-Mills theory, JHEP 06 (2010) 097,

arXiv:0804.2907. 3, 62
[17] Y. Imamura and K. Kimura, N = 4 Chern-Simons theories with auxiliary

vector multiplets, 10 (2008) 040, arXiv:0807.2144. 3, 61, 62
JHEP
[18] K. Hosomichi, K.-M. Lee, S. Lee, S. Lee, and J. Park, N = 4
superconformal Chern-Simons theories with hyper and twisted hyper
multiplets, JHEP 07 (2008) 091, arXiv:0805.3662. 3, 62
[19] N. Hama, K. Hosomichi, and S. Lee, SUSY gauge theories on squashed
three-spheres, JHEP 05 (2011) 014, arXiv:1102.4716. 3, 62
[20] N. Drukker et al., Roadmap on Wilson loops in 3d Chern-Simons-matter
theories, J. Phys. A 53 no. 17, (2020) 173001, arXiv:1910.00588. 3, 23,
24, 26, 39, 62, 63, 64, 66, 72, 73, 82, 83, 85, 86, 88
[21] N. Drukker, M. Tenser, and D. Trancanelli, Notes on hyperloops in N=4
Chern-Simons-matter theories, arXiv:2012.07096 [hep-th]. 3, 60, 62, 68
[22] D. Gaiotto and X. Yin, Notes on superconformal Chern-Simons-Matter
theories, JHEP 08 (2007) 056, arXiv:0704.3740. 3, 17, 63, 64, 65
[23] N. Drukker, M. Tenser, D. Trancanelli, M. Probst, and Z. Kong, To
appear,. 4, 60, 84, 88
[24] N. Drukker and D. Trancanelli, A supermatrix model for N = 6 super

Chern-Simons-matter theory, JHEP 02 (2010) 058, arXiv:0912.3006. 8,
9, 16, 17, 20, 23, 39, 40, 48, 63, 73, 82
[25] V. Cardinali, L. Griguolo, G. Martelloni, and D. Seminara, New
supersymmetric Wilson loops in ABJ(M) theories, Phys. Lett. B718

(2012) 615619, arXiv:1209.4032. 8, 9, 22, 23, 63, 66, 79, 84, 89
BIBLIOGRAPHY 102
[26] K.-M. Lee and S. Lee, 1=2-BPS Wilson loops and vortices in ABJM
model, JHEP 09 (2010) 004, arXiv:1006.5589. 9
[27] A. Mauri, S. Penati, and J.-j. Zhang, New BPS Wilson loops in N =4
circular quiver Chern-Simons-matter theories, JHEP 11 (2017) 174,
arXiv:1709.03972. 12, 62, 63, 83
[28] N. Drukker, J. Plefka, and D. Young, Wilson loops in 3-dimensional N =6

supersymmetric Chern-Simons Theory and their string theory duals,
JHEP 11 (2008) 019, arXiv:0809.2787. 16, 17, 65, 84
[29] B. Chen and J.-B. Wu, Supersymmetric Wilson loops in N = 6 super

Chern-Simons-matter theory, Nucl. Phys. B825 (2010) 3851,
arXiv:0809.2863. 16, 17, 65, 84
[30] S.-J. Rey, T. Suyama, and S. Yamaguchi, Wilson loops in superconformal
Chern-Simons theory and fundamental strings in anti-de Sitter supergravity
dual, JHEP 03 (2009) 127, arXiv:0809.3786. 16, 17, 65, 84
[31] H. Ouyang, J.-B. Wu, and J.-j. Zhang, Novel BPS Wilson loops in
three-dimensional quiver Chern-Simons-matter theories, Phys. Lett. B753

(2016) 215220, arXiv:1510.05475. 16, 39, 62
[32] K. Zarembo, Supersymmetric Wilson loops, Nucl. Phys. B 643 (2002)

157171, arXiv:hep-th/0205160. 22
[33] N. Drukker, S. Giombi, R. Ricci, and D. Trancanelli, Supersymmetric
Wilson loops on S 3 , JHEP 05 (2008) 017, arXiv:0711.3226. 22, 63, 84
[34] M. S. Bianchi, L. Griguolo, M. Leoni, S. Penati, and D. Seminara, Bps
wilson loops and bremsstrahlung function in abj(m): a two loop analysis,
Journal of High Energy Physics 2014 no. 6, (Jun, 2014) .

http://dx.doi.org/10.1007/JHEP06(2014)123. 22, 23, 59
[35] B. Assel and J. Gomis, Mirror symmetry and loop operators, JHEP 11
(2015) 055, arXiv:1506.01718. 40, 62, 84
[36] N. Drukker, J. Gomis, and D. Young, Vortex loop operators, M2-branes
and holography, JHEP 03 (2009) 004, arXiv:0810.4344. 40, 83
[37] M. Cooke, A. Dekel, N. Drukker, D. Trancanelli, and E. Vescovi,
Deformations of the circular wilson loop and spectral (in)dependence,
Journal of High Energy Physics 2019 no. 1, (Jan, 2019) .

http://dx.doi.org/10.1007/JHEP01(2019)076. 48, 52, 92
[38] M. S. Bianchi, G. Giribet, M. Leoni, and S. Penati, The 1/2 BPS Wilson
loop in ABJ(M) at two loops: The details, JHEP 10 (2013) 085,

arXiv:1307.0786 [hep-th]. 50, 52
BIBLIOGRAPHY 103
[39] W. Chen, G. W. Semeno, and Y.-S. Wu, Two-loop analysis of non-abelian
chern-simons theory, Phys. Rev. D 46 (Dec, 1992) 55215539.

https://link.aps.org/doi/10.1103/PhysRevD.46.5521. 51
[40] E. Guadagnini, M. Martellini, and M. Mintchev, Wilson Lines in
Chern-Simons Theory and Link Invariants, Nucl. Phys. B 330 (1990)

575607. 55
[41] G. W. Semeno and D. Young, Wavy Wilson line and AdS /CFT, Int. J.
Mod. Phys. A20 (2005) 28332846, hep-th/0405288. 58
[42] D. Correa, J. Henn, J. Maldacena, and A. Sever, An exact formula for the
radiation of a moving quark in N=4 super Yang Mills, JHEP 06 (2012)

048, arXiv:1202.4455 [hep-th]. 59
[43] A. M. Polyakov, Gauge elds as rings of glue, Nucl.Phys. B164 (1980)

171188. 59
[44] M. Cooke, N. Drukker, and D. Trancanelli, A profusion of 1=2 BPS Wilson

loops in N = 4 Chern-Simons-matter theories, JHEP 10 (2015) 140,
arXiv:1506.07614. 61, 62, 78, 79, 82, 83, 84, 86
[45] N. Hama, K. Hosomichi, and S. Lee, Notes on SUSY gauge theories on
three-sphere, JHEP 03 (2011) 127, arXiv:1012.3512. 62, 96
[46] Y. Asano, G. Ishiki, T. Okada, and S. Shimasaki, Large- N reduction for
N = 2 Chern-Simons theories and localization, Int. J. Mod. Phys. Conf.

Ser. 21 (2013) 175176. 62, 96
[47] H. Ouyang, J.-B. Wu, and J.-j. Zhang, Supersymmetric Wilson loops in
N = 4 super Chern-Simons-matter theory, JHEP 11 (2015) 213,

arXiv:1506.06192. 62, 83
[48] M. Lietti, A. Mauri, S. Penati, and J.-j. Zhang, String theory duals of
Wilson loops from Higgsing, JHEP 08 (2017) 030, arXiv:1705.02322. 62,
83, 84
[49] H. Ouyang, J.-B. Wu, and J.-j. Zhang, Construction and classication of
novel BPS Wilson loops in quiver Chern-Simons-matter theories, Nucl.

Phys. B910 (2016) 496527, arXiv:1511.02967. 47, 62, 83
[50] A. Mauri, H. Ouyang, S. Penati, J.-B. Wu, and J. Zhang, BPS Wilson
loops in N 2 superconformal Chern-Simons-matter theories, JHEP 11

(2018) 145, arXiv:1808.01397. 62
[51] N. Drukker, BPS Wilson loops and quiver varieties, J. Phys. A 53 no. 38,
(2020) 385402, arXiv:2004.11393. 62, 63, 66, 81, 82, 83, 86
BIBLIOGRAPHY 104
[52] N. B. Agmon and Y. Wang, Classifying superconformal defects in diverse
dimensions part I: superconformal lines, arXiv:2009.06650. 62, 84
1=4 BPS circular loops, unstable world-sheet instantons and

[53] N. Drukker,
the matrix model, JHEP 09 (2006) 004, hep-th/0605151. 63, 65
[54] H. Nakajima, Lectures on Hilbert Schemes of Points on Surfaces.

University Lecture Series. American Mathematical Society, 1999. 72
[55] A. Hanany and E. Witten, Type IIB superstrings, BPS monopoles, and
three-dimensional gauge dynamics, Nucl. Phys. B492 (1997) 152190,

hep-th/9611230. 72
[56] W. Crawley-Boevey, Lectures on representations of quivers.
http://www.math.uni-bielefeld.de/~wcrawley/quivlecs.pdf. 72
[57] V. Ginzburg, Lectures on nakajima's quiver varieties, arXiv:0905.0686.

72
[58] A. Kirillov, Quiver Representations and Quiver Varieties. Graduate
studies in mathematics. American Mathematical Society, 2016. 72
[59] M. S. Bianchi, L. Griguolo, A. Mauri, S. Penati, and D. Seminara, A
matrix model for the latitude Wilson loop in ABJM theory, JHEP 08
(2018) 060, arXiv:1802.07742. 82
[60] N. Drukker, T. Okuda, and F. Passerini, Exact results for vortex loop
operators in 3d supersymmetric theories, JHEP 07 (2014) 137,

arXiv:1211.3409. 83
[61] A. Kapustin, B. Willett, and I. Yaakov, Exact results for supersymmetric
abelian vortex loops in 2+1 dimensions, JHEP 06 (2013) 099,

arXiv:1211.2861. 83
[62] N. Drukker, S. Giombi, R. Ricci, and D. Trancanelli, More supersymmetric
Wilson loops, Phys. Rev. D 76 (2007) 107703, arXiv:0704.2237. 84
[63] N. Drukker, S. Giombi, R. Ricci, and D. Trancanelli, Wilson loops: From
four-dimensional SYM to two-dimensional YM, Phys. Rev. D 77 (2008)

047901, arXiv:0707.2699. 84
[64] M. S. Bianchi, L. Griguolo, M. Leoni, A. Mauri, S. Penati, and D. Seminara,
The quantum 1/2 BPS Wilson loop in N = 4 Chern-Simons-matter

theories, JHEP 09 (2016) 009, arXiv:1606.07058. 84
[65] L. Griguolo, M. Leoni, A. Mauri, S. Penati, and D. Seminara, Probing
Wilson loops in N = 4 Chern-Simons-matter theories at weak coupling,

Phys. Lett. B 753 (2016) 500505, arXiv:1510.08438. 84
BIBLIOGRAPHY 105
[66] M. Cooke, A. Dekel, and N. Drukker, The Wilson loop CFT: Insertion
dimensions and structure constants from wavy lines, J. Phys. A50 no. 33,
(2017) 335401, arXiv:1703.03812. 84
[67] S. Giombi, R. Roiban, and A. A. Tseytlin, Half-BPS Wilson loop and
AdS2 /CFT1 , Nucl. Phys. B922 (2017) 499527, arXiv:1706.00756. 84
[68] P. Liendo, C. Meneghelli, and V. Mitev, Bootstrapping the half-BPS line
defect, JHEP 10 (2018) 077, arXiv:1806.01862. 84
[69] L. Bianchi, G. Bliard, V. Forini, L. Griguolo, and D. Seminara, Analytic
bootstrap and Witten diagrams for the ABJM Wilson line as defect CFT1 ,
JHEP 08 (2020) 143, arXiv:2004.07849 [hep-th]. 84
[70] D. H. Correa, V. I. Giraldo-Rivera, and G. A. Silva, Supersymmetric mixed
boundary conditions in AdS2 and DCFT1 marginal deformations, JHEP

03 (2020) 010, arXiv:1910.04225. 84

Laços de Wilson em Teorias Super-Simétricas

Enviado por

Direitos autorais:

Formatos disponíveis

Laços de Wilson em Teorias Super-Simétricas

Enviado por

Dados do documento

Título original

Direitos autorais

Formatos disponíveis

Compartilhar este documento

Compartilhar ou incorporar documento

Opções de compartilhamento

Você considera este documento útil?

Este conteúdo é inapropriado?

Direitos autorais:

Formatos disponíveis

Laços de Wilson em Teorias Super-Simétricas

Enviado por

Direitos autorais:

Formatos disponíveis

Universidade de São Paulo

Laços de Wilson em teorias super-simétricas

Marcia Rodrigues Tenser

Orientador: Prof. Dr. Diego Trancanelli

Tese de doutorado apresentada ao Instituto de Física

da Universidade de São Paulo, como requisito parcial

para a obtenção do título de Doutor em Ciências.

Prof. Dr. Diego Trancanelli - Orientador (IF-USP)

Prof. Dr. Victor Rivelles (IF-USP)

Profa. Dra. Silvia Penati (UNIMIB, Itália)

Prof. Dr. Diego Correa (IF-UNLP, Argentina)

Prof. Dr. Alberto Faraggi (Ciencias Exactas/UNAB, Chile)

Tenser, Marcia Rodrigues

Laços de Wilson em teorias super-simétricas de Chern-Simons com

Tese (Doutorado) – Universidade de São Paulo. Instituto de Física.

Orientador: Prof. Dr. Diego Trancanelli

Unitermos: 1. Teoria de campos; 2. Teoria quântica de campo; 3.

Wilson Loops in supersymmetric

Marcia Rodrigues Tenser

Supervisor: Prof. Dr. Diego Trancanelli

Thesis submitted to the Physics Institute of the Uni-

versity of São Paulo in partial fulllment of the require-

ments for the degree of Doctor of Science.

Prof. Dr. Diego Trancanelli - Orientador (IF-USP)

Prof. Dr. Victor Rivelles (IF-USP)

Profa. Dra. Silvia Penati (UNIMIB, Itália)

Prof. Dr. Diego Correa (IF-UNLP, Argentina)

Prof. Dr. Alberto Faraggi (Ciencias Exactas/UNAB, Chile)

Antes de mais nada, gostaria de agradecer o Conselho Nacional de Desenvolvi-

mento Cientíco e Tecnológico (CNPq) pelo apoio nanceiro.

Dentre os inúmeros aprendizados colecionados durante o desenvolvimento desse

desse doutorado e quero mencionar então algumas em particular.

me acompanhado do início ao m, apesar de todos os percalços.

Laura e Renato, e a Gabi e a Simone, por tornarem acolhedores até mesmo os

corredores estéreis do Departamento de Física Matemática. Em particular os que

meu lado e me proporcionado doses semanais de alívio em meio à pandemia.

À vocês os meus profundos agradecimentos. O que dividimos é o que levo de

mais precioso dessa jornada.

The goal of this thesis is to study Wilson loops in 3 dimensional Chern-Simons-

matter theories aiming at applications in holography. In particular, we study

operators preserving some amount of supersymmetry in ABJ(M) theory and in

N =4 Chern-Simons-matter theories. In both cases, we seek for a systematic

construction of these operators to understand the moduli space of the theory. In

information as, for example, the Bremsstrahlung function.

Key-Words: supersymmetry, Wilson loops", Chern-Simons-matter theories.

O objetivo desta tese é estudar laços de Wilson em teoria supersimétrica de Chern-

Simons com matéria em 3 dimensões visando aplicações em holograa. Em par-

ticular, nós estudamos operadores preservando alguma quantidade de supersime-

tria em teoria de ABJ(M) e em teorias com N = 4. Em ambos os casos, nossa

intenção é obter uma construção sistemática desses operadores para entender o

espaço de módulos da teoria. No caso de ABJ(M), nós também fazemos cálcu-

los perturbativos visando informações universais como, por exemplo, a função de

Palavras-Chave: supersimetria, laços de Wilson", teorias supersimétricas de

Chern-Simons com matéria.

2.2 Supersymmetric Wilson loops . . . . . . . . . . . . . . . . . . . . . 7

2.3 Quivers and Wilson loops . . . . . . . . . . . . . . . . . . . . . . . 9

3.1.1 Circular loops . . . . . . . . . . . . . . . . . . . . . . . . . . 16

3.1.2 Latitude loops . . . . . . . . . . . . . . . . . . . . . . . . . 21

3.2 From 1-node to 2-node loops . . . . . . . . . . . . . . . . . . . . . 23

versity of São Paulo in partial fulllment of the require-

mento Cientíco e Tecnológico (CNPq) pelo apoio nanceiro.

me acompanhado do início ao m, apesar de todos os percalços.

Key-Words: supersymmetry, Wilson loops", Chern-Simons-matter theories.

Simons com matéria em 3 dimensões visando aplicações em holograa. Em par-

Palavras-Chave: supersimetria, laços de Wilson", teorias supersimétricas de

Chern-Simons com matéria.

4.2 Hyperloops at =0 . . . . . . . . . . . . . . . . . . . . . . . . . . 66

production probability [1] and about connement [2].

a technique called localization [35]. In fact, localization can be used to compute

a term generated by the action of Q on some eld conguration A, called Q-exact

described by classical supergravity so this side of the duality is also often

This non-trivial check of the duality was rst done in four-dimensional N =4

and N = 4 Chern-Simons-matter theories [1619]. Even though most of our

is clear how this ts into the holography program.

quiver diagram is enough to completely x the Lagrangian of a particular theory.

superelds are represented as nodes and chiral superelds as arrows. To each

To each chiral supereld we associate an arrow.

can be decomposed in terms of N = 1 superelds as the direct sum of a vector

symmetry and, in addition to it, we also include an U (N ) avor symmetry. The

quiver is drawn in N =1 language whereas in (b) we use the simplied rules of

avor groups, as expected from the N = 2 hypermultiplet decomposition. There-