A onedimensional theory for Higgs branch operators
Accepted: March
onedimensional theory for Higgs branch operators
Mykola Dedushenko 0 1 2 3
Silviu S. Pufu 0 1 2
Ran Yacoby 0 1 2 4
0 Rehovot 76100 , Israel
1 45248 1200 E California Blvd , Pasadena CA 91125 , U.S.A
2 Princeton , NJ 08544 , U.S.A
3 Walter Burke Institute for Theoretical Physics, California Institute of Technology , USA
4 Department of Physics and Astrophysics, Weizmann Institute of Science
We use supersymmetric localization to calculate correlation functions of halfBPS local operators in 3d N = 4 superconformal eld theories whose Lagrangian descriptions consist of vectormultiplets coupled to hypermultiplets. The operators we primarily study are certain twisted linear combinations of Higgs branch operators that can be inserted anywhere along a given line. These operators are constructed from the hypermultiplet scalars. They form a onedimensional noncommutative operator algebra with topological correlation functions. The 2 and 3point functions of Higgs branch operators in the full 3d N = 4 theory can be simply inferred from the 1d topological algebra. After conformally mapping the 3d superconformal eld theory from threesphere, we preform supersymmetric localization using a supercharge that does not belong to any 3d N = 2 subalgebra of the N = 4 algebra. The result is a simple model that can be used to calculate correlation functions in the 1d topological algebra mentioned above. This model is a 1d Gaussian theory coupled to a matrix model, and it can be viewed as a gauge xed version of a topological gauged quantum mechanics. Our results generalize to nonconformal theories on S3 that contain real mass and FayetIliopolous parameters. We also provide partial results in the 1d topological algebra associated with the Coulomb branch, where we calculate correlation functions of local operators built from
Extended Supersymmetry; Supersymmetric Gauge Theory; Conformal Field

A
the vectormultiplet scalars.
Theory
ArXiv ePrint: 1610.00740
Actions with vectormultiplets and hypermultiplets
Closure of the supersymmetry transformations
Nonconformal supersymmetry algebra on S3
Central extension of nonconformal supersymmetry algebra
5.5 Integration cycle from Morse theory
6
Properties of twisted Higgs branch theory
Brief summary
Topological gauged quantum mechanics
The conformal case
Nonvanishing mass and FI parameters
6.3
Correlators of twisted Higgs branch operators
6.3.1
The conformal case
Twisted operators on S3 by stereographic map
Interpretation in terms of su(2j1)`
5.2 3d Gaussian theory coupled to a matrix model
5.3 1d Gaussian theory for twisted Higgs branch operators
5.4 1d theory from localization of hypermultiplet
Localizing with QKWY
Localizing with Q
Localizing with Q
C
H
O shell closure 1d action from BPS equations Hypermultiplet 1loop determinant Integration cycle from localization
1 Introduction
2 3d N = 4 theories on S3
3
Cohomology in SCFTs
SCFT in at space SCFT on the sphere
4
5
2.1
2.2
2.3
2.4
3.1
3.2
4.1
4.2
6.1
6.2
2.3.1
2.3.2
2.3.3
3.2.1
3.2.2
5.2.1
5.2.2
5.2.3
5.4.1
5.4.2
5.4.3
5.4.4
6.2.1
6.2.2
{ i {
2 and 3point correlators of Higgs branch operators of the SCFT
Star product, Higgs branch chiral ring, and deformation quantization
Operator mixing
SQED with N charged hypermultiplet avors
N node quiver
U(2) with adjoint hypermultiplet and fundamental hypermultiplet
8.2 Introducing mass parameters
SQED with nonzero FI parameter
N node quiver with nonzero FI parameters
Massdeformed N node quiver
Massdeformed SQED
50
51
rems, such as in [1, 2] who showed that the nonperturbative answer for 3point functions
result.1 The conformal bootstrap approach also allows one to calculate certain correlators
of BPS operators in some particular supersymmetric conformal eld theories (SCFTs) with
8 Poincare supersymmetries in various dimensions [4{8]. Other examples use the technique
of supersymmetric localization (for recent reviews, see [9{25] and references therein)
allowing for the calculation of twopoint functions of conserved avor or Rsymmetry currents in
3d N = 2 superconformal eld theories (SCFTs) [26, 27], and of Coulomb branch operators
in 4d N = 2 SCFTs [28]. (See also [29{32].)
Our goal here is to provide more instances of such exact computations of
correlation functions of local operators.
We focus on 3d quantum
eld theories with N = 4
functions of certain 12 BPS operators. The derivation of these formulas also relies on
supersymmetric localization, albeit using a di erent supercharge from the one used in previous
supersymmetric localization studies of such theories.
The 12 BPS operators whose correlation functions we compute fall within two classes
of more general operators. The rst class, referred to as Higgs branch operators, consists
of gauge invariant operators constructed from the scalar elds in the hypermultiplet, while
the second class, referred to as Coulomb branch operators, contains operators constructed
from the scalars in the vectormultiplet as well as from 12 BPS scalar monopole operators.
The naming of the two classes re ects that, when the QFT is de ned on R3, these operators
acquire nonzero expectation values on either the Higgs or Coulomb branch of the moduli
space of supersymmetric vacua.
We restrict our attention to the origin of the moduli space, where the N = 4 theories
we consider ow in the IR to SCFTs. Let us denote the gauge group of such a theory by G
and assume, without loss of generality, that there is only one hypermultiplet transforming
in a unitary representation R of G, where R may be reducible.3 In a nutshell, we have
three main results:
We present a relatively simple matrix model coupled to a 1d Gaussian theory,
ZHiggs =
1 Z
jWj Cartan
)]
Z
`
Z
d det 0adj [2 sinh(
DQ DQe exp
1See also [3] for a proof, and [4] for a generalization to 4d N = 2 theories.
2Such theories have been extensively studied in the literature starting with refs. [33, 34]. For string
theory constructions of such theories, see, for instance, [35{38].
3Let NH be the total number of hypermultiplets in the absence of gauging. The 2NH complex scalars
and 2NH complex two component fermions transform in the pseudoreal fundamental representation of the
avor symmetry USp(2NH ). We consider the U(NH ) subgroup of USp(2NH ) under which the fundamental
of USp(2NH ) decomposes as NH
NH , and we take the gauge group G to be a subgroup of U(NH ). At
the level of Lie algebras, we have that a subalgebra g of u(NH ) is gauged, and we de ne the representation
map R : g ! u(NH ) from the gauge algebra into NH
xed point. In addition, (1.1) can be used to calculate npoint correlators
of certain twisted Higgs branch operators of the SCFT. These twisted operators
are speci c positiondependent linear combinations of Higgs branch operators, to
be de ned precisely in section 3, obtained by contracting the various Rsymmetry
components of Higgs branch operators with positiondependent polarization vectors.
Let us describe (1.1). jWj is the order of the Weyl group of G. The variable
is the matrix degree of freedom and takes values in the Cartan of the Lie algebra
g = Lie(G).4 It is coupled to a 1d Gaussian theory de ned on a circle parameterized
; ) whose degrees of freedom are the antiperiodic scalar elds Q(') and
Qe(') which transform in the representations R and R of g, respectively. The path
integral over Q and Qe is over a middledimensional integration cycle in the space of
complexvalued elds Q and Qe. We will discuss this integration cycle in more detail
in section 5. Lastly, ` is a parameter with dimensions of length whose meaning we
will explain momentarily.
The twisted Higgs branch operators of the 3d SCFT whose correlators can be
calculated using (1.1), can be inserted anywhere along a line in R3 or, equivalently, along a
great circle on S3 that maps to this line under the stereographic projection. In (1.1),
the angle ' parameterizes the great circle on S3, and `
4 r is proportional to the
radius of S3. Moreover, the twisted Higgs branch operators are represented in the
1d model (1.1) by gaugeinvariant polynomials in Q(') and Qe('). From the 2 and
3point functions of the twisted Higgs branch operators, one can extract in a simple
way the 2 and 3point functions of the most general Higgs branch operators.
The 1d sector consisting of the twisted Higgs branch operators of a general 3d N = 4
SCFT was previously studied in [4, 7, 8] abstractly, using only properties of the
superconformal algebra.5 These properties imply that the correlation functions of the
twisted Higgs branch operators are topological, in the sense that they do not depend
on the relative separation between the insertion points, but do depend on the ordering
of the insertions. Moreover, the 1d OPE algebra of this sector is an associative
noncommutative algebra obeying certain very special properties. In some cases, refs. [7, 8]
used bootstraptype arguments to show that these properties determine the 1d OPE
algebra uniquely up to a
nite number of parameters. The model (1.1) provides
a complementary approach to the analysis in [7, 8] whereby (1.1) can be used to
calculate explicitly the structure constants of the 1d OPE algebra.
We provide a partial result toward a similar computation of correlation functions of
Coulomb branch operators. In particular, we only consider Coulomb branch operators
4Very roughly, the theory (1.1) can be interpreted as a 1d gauged quantum mechanics with gauge group
G, in the gauge where A' = . The determinant factor in (1.1) is precisely the FadeevPopov determinant
corresponding to the gauge
5At this abstract level, there is no di erence between the 1d sector associated with the Higgs branch
and that associated with the Coulomb branch.
{ 3 {
that are not monopole operators. Such nonmonopole operators are given by
gaugeinvariant polynomials in the vectormultiplet scalars. At the SCFT
xed point, the
2 and 3point correlators of nonmonopole Coulomb branch operators, as well as
npoint functions of their twisted analogs, can be calculated by inserting gaugeinvariant
polynomials in
into the matrix model
ZCoulomb =
1 Z
as a result of a supersymmetric localization computation that uses only N
supersymmetry, and with the goal of calculating expectation values of BPS Wilson
loop operators. Its relation to the Higgs branch theory (1.1) is that one obtains (1.2)
after integrating out Q and Qe in (1.1).
As was the case with Higgs branch operators, the twisted Coulomb branch operators
whose correlation functions can be computed from (1.2) are part of the 1d
topological Coulomb branch sector studied abstractly in [4, 7, 8]. In terms of the elds of
the 3d SCFT, the twisted Coulomb branch operators represented by gaugeinvariant
polynomials in
correspond to positiondependent linear combinations of
polynomials in the vectormultiplet scalars. A more complete analysis that includes monopole
operators is left for future work.
The above results can be generalized to nonconformal N = 4 QFTs on S3 that
are obtained by introducing real mass and FayetIliopolous (FI) parameters. The
real mass parameters are introduced in (1.1){(1.2) by shifting
exponent of (1.1) and denominator of (1.2), where m is a real mass matrix taking
value in the Cartan of the avor symmetry algebra of the hypermultiplet.6 For each
abelian factor of the gauge group, one can introduce an FI parameter a by including
!
+ mr in the
in (1.1){(1.2) an additional factor
(1.2)
HJEP03(218)
e 8 2ir tr
:
(1.3)
In (1.3), tr
P
a a a with the sum taken over all abelian factors in g, while a
are the corresponding FI parameters and a are the components of
that take values
in those abelian factors.
The correlators of twisted Higgs (Coulomb) branch operators that are not charged
under the
avor (topological) symmetries associated with nonzero real mass (FI)
6As in footnote 3, let NH be the total number of hypermultiplets in the absence of gauging. The full
avor symmetry group GbF of the NH hypermultiplets is de ned as the normalizer of the gauge group inside
USp(2NH ) modulo the gauge group [40]. However, we take our gauge group G to be contained in a U(NH )
subgroup of USp(2NH ), and when de ning the
avor symmetry we also consider GF = GbF \ U(NH ). We
sometimes refer to GF as the avor symmetry of the hypermultiplet. The embedding of GF into U(NH )
induces a map F : gF ! u(NH ), where gF = Lie(GF ).
By
+ mr in the main text we then mean R( ) + rF(m), where R : g ! u(NH ) is the representation
map from the gauge algebra into NH
parameters are still topological. We will show that when a real mass parameter
ow between two SCFTs, the correlation functions in the 1d theory
interpolate between topological correlators in the UV and the IR, even though in the
intermediate regime these correlation functions may be positiondependent.
Let us explain in more detail the procedure by which one arrives at the aforementioned
a Q
results. While our results hold in SCFTs de ned on any conformally
at space, we nd
it convenient to rst consider more general nonconformal 3d N = 4 QFTs on a round
S3 of radius r. Due to the large amount of supersymmetry, placing the N = 4 theories
on S3 is unambiguous. The S3 Lagrangians we consider are curved space generalizations
of the usual at space ones: they contain kinetic terms for the hypermultiplets, a
YangMills term for the vectormultiplet with YangMills coupling gYM, and, optionally, real
mass and FI parameters, all containing certain curvature couplings that vanish in the limit
r ! 1. Setting to zero the real mass and FI parameters and taking both gYM; r ! 1,
the correlation functions computed in an N = 4 theory on S3 approach those of the deep
infrared limit of the same N = 4 theory de ned on R3. In the examples we consider, such
a deep infrared limit is a nontrivial interacting SCFT. Alternatively, we can
rst take
gYM ! 1 at xed r and then conformally map from S3 to R3. As we will see, the S3
correlators we study are independent of gYM, so the limit gYM ! 1 is taken trivially.
After placing the theories of interest on S3, we perform supersymmetric localization
with an appropriately chosen supercharge. The choice of supercharge is guided by the
SCFTs on R3 in [7, 8]. In particular, the authors of [7, 8] identi ed a supercharge Q
cohomological construction of [4], which was elaborated upon in the context of 3d N = 4
H in the
N = 4 superconformal algebra whose cohomology classes are represented by the twisted
Higgs branch operators mentioned in the rst bullet point above. A similar construction
for Coulomb branch operators involves the cohomology of a di erent supercharge QC .
We perform supersymmetric localization using precisely the supercharge Q
mapped to S3 using the stereographic map. At rst, this statement may seem puzzling for
H or QC ,
the following reason. In performing supersymmetric localization, one adds to the action
H;C exact localizing term. The standard localizing term for the vectormultiplet is
usually constructed by acting with two supercharges on fermion bilinears, and it thus has
scaling dimension 4 and breaks conformal invariance. However, both supercharges Q
C were constructed in [7, 8] as speci c linear combinations of Poincare and conformal
supercharges on R3, so conformal symmetry seemed important. Consequently it seems
H and
confusing why the standard vectormultiplet localizing term could even be invariant under
will show, however, that the supercharges Q
H;C belong to an N = 4 supersymmetry algebra
su(2j1)r, which, upon mapping to S3, can be seen to contain only the isometries
of S3 and a U(1)2 Rsymmetry, without any conformal generators.7 Thus, Q
H;C invariant
7We stress that upon contraction r ! 1 the supercharges QH;C 2 su(2j1)`
su(2j1)r we de ne on S3
reduce to ordinary Poincare supercharges on R3, and not to the supercharges constructed in [7, 8]. The
latter were also denoted by QH;C above, in a slight abuse of notation.
{ 5 {
theories on S3 are not necessarily conformal invariant; they include the more general
nonconformal QFTs on S3 mentioned above.
It is worth commenting on the relation between the localization computation using
H;C and that preformed by Kapustin, Willett and Yaakov (KWY) in [39] for N
theories that yielded the matrix model (1.2). This computation was later generalized to
3
N
= 2 theories in [41, 42]. The supercharge Q
also resides in an N = 2 subalgebra, namely su(2j1)
KWY used for localization in [39] thus
su(2), of the full N = 4 algebra
localizing with Q
described above.
C or Q
C do not reside in any such N = 2
subalgebra, but are instead di erent linear combinations of supercharges in the two su(2j1)
factors of the N = 4 superalgebra. In spite of these di erences, we nd that the results of
H are very much related to the KWY matrix model, as was brie y
Concretely, our calculation proceeds as follows. Just as in [42], we nd that the
YangMills action is Qexact (with respect to Q
H or QC ) and can be added with a large coe cient,
thus localizing the N = 4 vectormultiplet in precisely the same way as in [39, 42]. In other
words, the localization of the vectormultiplet is realized by taking the theory to small gauge
coupling. At any point on the vectormultiplet localization locus, the hypermultiplet is thus
free, but massive, with its mass matrix depending on the precise location on the localization
locus. In this weakly coupled theory, the correlation functions of the hypermultiplet can be
computed by rst using Wick's theorem at a xed point on the vectormultiplet localization
locus, and then integrating over it with a measure given by the determinant of uctuations
of all the elds. If we focus our attention on Q
are twisted Higgs (or Coulomb) branch operators inserted along a great circle of S3, a
standard argument shows that these correlation functions are independent of the
YangH;C closed operators, which as we show
Mills coupling.
We conclude that once the vectormultiplet has been localized, calculating correlators of
twisted Higgs branch operators does not require a further localization of the hypermultiplet.
Indeed, the hypermultiplet action in the background of the localized vectormultiplet is
Gaussian, and thus the remaining path integral is trivially solvable. For localization with
QH , however, it is instructive to also localize the hypermultiplet, which leads to the explicit
description (1.1) of the correlators in terms of the 1d Gaussian theory coupled to the matrix
model obtained from localizing the vectormultiplet.
Localization of the hypermultiplet with Q
H has several features that are worth
mentioning. Unlike the N
= 4 vectormultiplet, the supersymmetry algebra does not close
o shell on the hypermultiplet. Since the supersymmetric localization arguments require
a supercharge that does close o shell, the rst step is to add a number of auxiliary elds
and modify the supersymmetry transformation rules such that the algebra generated by
H does close o shell on the hypermultiplet elds.8 The next step is to add to the action
a QH exact term whose bosonic part is positivede nite. To describe the localization locus,
let us think of S3 as a circle bered over a disk with the circle shrinking at the boundary
8A similar construction in the case of N = 4 supersymmetric YangMills theory in 4d was used in [43, 44]
following the work of [48].
{ 6 {
of the disk. We nd that the hypermultiplet localizes on eld con gurations that are
independent of the coordinate parameterizing the circle and that obey additional di erential
constraints in the disk directions. These constraints imply that the hypermultiplet action,
when evaluated on the localization locus, becomes a total derivative on the disk and
reduces to a boundary term. This boundary term, living on the boundary of the disk, is
the Gaussian action for our localized theory. We will argue that the oneloop determinant
of uctuations around this con guration equals 1, so there is no additional determinant
factor coming from the hypermultiplet. The situation presented here for the localization
of the hypermultiplet is very similar to the one encountered in [44] in the case of 4d N = 4
We apply (1.1){(1.2) in a few examples where we calculate explicitly several correlation
functions of twisted Higgs and Coulomb branch operators, with the main focus on the Higgs
branch. As in [8], we interpret the 1d OPE algebra as a noncommutative star product
on the Higgs branch chiral ring,9 and we compute explicitly the values of the parameters
that the bootstrap analysis of [8] left undetermined. As we will discuss, in calculating
correlation functions using (1.1){(1.2), one should be aware of the possibility of operator
mixing on S3. The mixing can be removed by diagonalizing the matrix of 2point functions
using the GramSchmidt procedure. A similar approach was taken in [28] for the Coulomb
branch operators of 4d N = 2 theories.
The rest of this paper is organized as follows. In section 2 we introduce the N = 4 QFTs
on S3 we will study. In section 3 we review the cohomological construction of [4, 7, 8] in the
case of N = 4 SCFTs in
In section 4 we generalize this construction to QFTs on S3 that do not necessarily possess
conformal symmetry. Section 5 contains a description of the localization computation that
leads to the results (1.1){(1.2) summarized above. In section 6 we describe in general terms
the various properties of the 1d theory (1.1) and its applications. Sections 7 and 8 contain
applications of our results to speci c theories. We end with a brief discussion in section 9.
at space and explain how it is mapped stereographically to S3.
2
3d N
= 4 theories on S3
In this section we will review the construction of N
= 4 supersymmetric Lagrangians
using vectormultiplets and hypermultiplets on S3. We rst provide the supersymmetry
transformation rules and the supersymmetric actions. We then discuss the supersymmetry
algebras preserved by these actions.
2.1
Actions with vectormultiplets and hypermultiplets
The components of the vectormultiplets and hypermultiplets carry Lorentz indices as well
as su(2)C
su(2)H Rsymmetry indices. Explicitly, the components of the vectormultiplet
9The twisted Higgs branch operators are in 1to1 correspondence with chiral Higgs branch operators.
While a generic Higgs branch operator corresponds to a function on the Higgs branch, a chiral Higgs branch
These operators form the Higgs branch chiral ring. Similar statements hold for Coulomb branch operators.
V transform in the adjoint representation of the gauge group G and will be denoted by
The vector A in (2.1) is the gauge eld, the spinor
aa_ is the gaugino, and
a_ b_ and Dab
are scalars, transforming in the (1; 1), (2; 2), (3; 1) and (1; 3) irreps of su(2)C
respectively.10 The hypermultiplet H transforms in some unitary representation R of G
and has components
V = (A ;
aa_ ; a_ b_ ; Dab) :
H = (qa; qea;
a_ ; e a_ ) ;
where qa, qea are complex scalars transforming in (1; 2) and (1; 2) irreps of the Rsymmetry
and representations R and R of G, and
a_ , e ;a_ are their spinor superpartners, which
transform, respectively, in the (2; 1) and (2; 1) irreps of the Rsymmetry, and in the R and
R representations of G. The multiplets V and H also have \twisted" versions, in which the
roles of su(2)C and su(2)H are interchanged, though we will not consider them in this paper.
On S3, superconformal transformations are generated by spinors aa_ in the (2; 2) irrep
of the Rsymmetry, which satisfy the conformal Killing spinor equations
r
aa_ =
a0a_ ;
r
particular, the transformation rules for the vectormultiplet elds (2.1) are11
A
=
ab_ =
a_ b_ = c
i ab_
2
i
2
"
Dab =
(a_ jcjb_) ;
iD ( (a
c_
ab_ ;
ab_ F
and those of the hypermultiplet (2.2) are
Dac cb_
i
c_
a D
c_b_ + 2i b_ c_ a0c_ +
i
_
2 ad_[ b_ c_; c_d] ;
b)c_)
2i (0ac_
b)c_ + i[ (ac_ b)d_; c_d_] ;
qa = ab_ _ ;
b
e
qa = ab_ eb_ ;
a_ = i
ea_ = i
a a_D qa + i a0a_ qa
i ac_ c_a_ qa ;
a a_D qea + iqea a0a_ + i ac_qe
a c_a_ :
One can check that acting twice with the transformation rules presented above realizes the
superconformal algebra osp(4j4) up to gauge transformations and fermionic equations of
motion. We will return to this point with more details shortly.
10We label the doublet irrep of su(2)rot. frame rotations by indices
; ; : : : = 1; 2, of su(2)H by
a; b; : : : = 1; 2, and of su(2)C by a_ ; b_; : : : = 1; 2. See appendix A for more details on our conventions.
11The
eld strength is F
[A ; A ], and D
= r
iA
is the space and gauge
covariant derivative. Brackets () enclosing indices denote an average over their permutations.
{ 8 {
(2.1)
(2.2)
With the supersymmetry transformation rules in hand, one can construct
supersymmetric actions. The action of a hypermultiplet coupled to a vectormultiplet is
This action is invariant under the full osp(4j4) algebra. Indeed, one can check that it is
invariant under the transformations (2.4){(2.9) provided that (2.3) is obeyed. This action
could have been deduced from the analogous
at space action by simply covariantizing
all derivatives and introducing the conformal mass term 43r2 qeaqa for the hypermultiplet
Multiplets V and H as de ned above have too many bosonic components and, in the
path integral, have to be integrated over the middledimensional cycle determined by the
following reality conditions on bosons:
HJEP03(218)
e
qa = (qa) ;
(AI T I ) = AI (T I ) ;
( Ia_ b_ T I ) =
(DaIbT I ) =
Ia_ b_ (T I ) ;
DIab(T I ) ;
where, for the vectormultiplet elds, we made the representation matrices T I by which they
act on R explicit. In Lorentzian signature, the fermions would also obey reality constraints,
namely ea_ would be the hermitian conjugate of
a_ and
ab_ would be the hermitian
conjugate of
ab_ , but in the Euclidean signature,
the representations R and R of G respectively, and
a_ and e a_ are independent spinors in
ab_ do not obey any constraints either.
As far as we know, it is not possible to write down other superconformal actions on
S3 with just vectormultiplets and hypermultiplets.12 It is possible, however, to write down
actions that are invariant under half the supersymmetries in osp(4j4), which anticommute
to the isometries of S3 without any conformal transformations. Such actions provide curved
space analogs of the YangMills action or of the actions corresponding to real masses and
FI terms, all of which are not conformally invariant on R3, and therefore cannot be mapped
to S3 using the stereographic map. The projection from 16 supersymmetries in osp(4j4)
to the 8 under which these actions on S3 are invariant is described in terms of two su(2)
matrices, hab and ha_ b_ ,
hab 2 su(2)H ;
a_
h b_ 2 su(2)C ;
(2.12)
normalized such that hachcb = ab and ha_ c_hc_b_ = a_ b_ , and obeying the tracelessness
condition haa = ha_ a_ = 0. We interpret these matrices as representing the Cartan elements of
12A ChernSimons action for the vectormultiplet would be conformal, but preserving N = 4
supersymmetry would require the presence of twisted hypermultiplets which we do not consider here.
{ 9 {
a_ b_ qa
(2.10)
(2.11)
this fashion.
The YangMills action is given by
The nonconformal supersymmetric actions depend explicitly on the matrices (2.12).
Under a decomposition of N = 4 into an N = 2 subalgebra, one can show that (2.14) is
nothing but the S3 action of an N = 2 vectormultiplet plus an adjoint chiral of Rcharge
1. For each U(1) factor in G we can introduce an FIterm:
SFI[V] = i
Z
d3xpg habDba
r
1 a_
_
h b_ ba_ :
su(2)H
su(2)C . Then one can restrict the S3 Killing spinors (2.3) to those obeying the
further condition
This condition reduces the number of independent aa_ by a factor of two. We will interpret
it shortly in terms of generating a subalgebra of osp(4j4), but let us
actions on S3 that are invariant under the 8 supersymmetry transformations restricted in
rst present the
(2.13)
(2.14)
(2.15)
(2.16)
(2.17)
(2.18)
Note that while on R3 the FI parameters ab take value in the (1; 3) irrep of su(2)C
only the single component
= hab ab invariant under the Cartan of su(2)H survives on
S3. Finally, one can introduce mass terms for the hypermultiplets by coupling them to
background vectormultiplets Vb.g. in the Cartan of the avor symmetry. In order to preserve
supersymmetry all the components of Vb.g. are set to zero except for
2
1 ha_ b_ ( b.g.)ba_ =
_
r
2 hab(Db.g.)ba ;
and the supersymmetry variations (2.8) and (2.9) have to be deformed accordingly to
account for the masses. As happened with the FI terms, out of the su(2)C triplet of mass
parameters that exist on R3 for each avor group Cartan element, only one survives on S3.
In the remainder of this paper, in order to conform with the conventions of [7] we will
sometimes choose
hab =
2
;
ha_ b_ =
3
:
2.2
Closure of the supersymmetry transformations
Irrespective of the actions presented above, the superconformal transformations (2.4){(2.7)
of V close o shell into
f ; ~gV =
where G^ is a gauge transformation with parameter
= ( ~ca_ cb_ ) a_ b_
i( ~aa_
de ned as
aa_ )A ;
while K^ ;~ generates bosonic symmetries in osp(4j4) and is written explicitly in terms of
and ~ in appendix B. The transformations of the scalars qa in H also close o shell as
in (2.18), but those of the fermions a_ do not. Instead, one nds
f ; ~g a_ =
f ; ~g ea_ =
where the equations of motion operators are given by
eom
a_
eom
e a_
h
i D= a_ +
_
a_ b b_ + aa_ q
ai ;
h
i D= ea_
_
eb_ ba_
ea a a_ :
q
i
These are precisely the equations of motion following from the hypermultiplet action (2.10).
That the supersymmetry algebra closes only up to the fermion equations of motion will be
important in section 5.4.1, since the supercharge used for localization of the hypermultiplet
has to be closed o shell.
2.3
Nonconformal supersymmetry algebra on S3
Let us now return to the projection condition (2.13) and interpret it from the point of view
of which supersymmetry algebra it is that the actions (2.14){(2.15) as well as the mass
terms introduced via (2.16) are invariant under.
Let us assume for now that we have not introduced any mass terms and that we
set all possible FI parameters to zero. The anticommutator of two supersymmetries
restricted by (2.13) does not produce all the bosonic generators of the superconformal algebra
osp(4j4), but only a subset of them. This was to be expected, because we have argued that
the YangMills action (2.14) is invariant under supersymmetries obeying (2.13), and since
the YangMills action is not conformal, it must be that the anticommutator of
supersymmetries (2.13) does not generate any conformal transformations.
Judiciously working out all possible (anti)commutators, one can check that (2.13)
parameterize the 8 supersymmetry transformations of the algebra
(See also [49].) The bosonic generators of this algebra consist of the so(4) = su(2)` su(2)r
isometries of S3 as well as two u(1) Rsymmetries that we will denote by u(1)` and u(1)r,
re ecting which su(2j1) factor they belong to. The u(1)`
u(1)r is a subalgebra of su(2)H
su(2)C . That (2.24) contains 8 supersymmetries means it is an N = 4 supersymmetry
algebra.
The algebra (2.24) will be central in our work, so let us describe it in more detail. Let
us denote the generators by J (`), R`, and Q
(` ) for su(2j1)` and J (r), Rr, and Q
(r ) for
su(2j1)r. Abstractly, the algebra obeyed by J (`), R`, and Q
(` ) is
[Ji(`); Jj(`)] = i ijkJk(`) ;
[J (`); Q(` )] =
Q
(` ) + "
Q
where
The generators of su(2j1)r obey the same relations with ` ! r.
To be more concrete, let us explain how these generators act on the various operators
in the theory. We will take this opportunity to set up some of the notation we will use later.
2.3.1
Action of S3 isometries
The commutators with S3 isometries act on a gaugeinvariant operator O as the Lie
derivative
J (`)
(J1(`) + iJ2(`))
J (`)
3
J (`)
3
J (`)
1
iJ2(`)
!
:
we can use the parameterization
in terms of the coordinates
vectors in (2.28) are
v1` =
v2` =
v3` =
v1r =
v2r =
v3r =
i
2
i
2
i
2
i
2
i
2
i
2
( sin(
tan( ) sin( + ')@ + cot( ) sin( + ')@') ;
2 [0; 2 ], ; ' 2 [
; ]. In this parameterization, the Killing
with respect to the Killing vectors vi` and vir that obey the su(2) algebra, [vi`; vj`] = i"ijkvk`,
In an explicit description where the threesphere of radius r is embedded in R4 via
[Ji(`); O] =
Lvi` O ;
[Ji(r); O] =
Lvir O
X12 + X22 + X32 + X42 = r2 ;
X1 + iX2 = r cos ei ;
X3 + iX4 = r sin ei'
which manifests S3 as a U(1) bration over D2, with the bers being \warped" by w and
HJEP03(218)
shrinking to zero size at the boundary of the disk.
2.3.2
Action of Rsymmetries
The action of R` and Rr on the elds of the previous section depends on the precise
embedding of u(1)` and u(1)r into su(2)C
su(2)H given in terms of the matrices h and h
in (2.12) as follows. Let us rst de ne the operators
RH =
1
2 habRba ;
RC =
2
1 ha_ b_ Rb_ a_ ;
where Rba and Rb_ a_ are the generators of su(2)H and su(2)C respectively. In our conventions,
we then have
R` = RH + RC ;
Rr = RH
RC :
This equation provides an identi cation of R` and Rr with linear combination of the Cartan
elements RH and RC of the Rsymmetry of the superconformal algebra. In terms of their
action on elds, it is su cient to describe how they act on su(2)H and su(2)C fundamental
operators. We have13
(2.32)
(2.33)
(2.34)
(2.35)
(2.36)
We will make signi cant use of the parameterization (2.30) in the remainder of this paper.
The metric in these coordinates is
ds2(S3) = r2(d 2 + cos2( )d 2 + sin2( )d'2):
Coordinates
and ' parametrize a disk with the metric ds2(D2) = r2(d 2 + sin2( )d'2),
where
is the radial coordinate of the disk. The sphere metric then becomes:
ds2(S3) = ds2(D2) + w2d 2;
w = r cos ;
[RH ; Oa] =
1
2 habOb ;
[RC ; Oa_ ] =
1 b_
2
h a_ Ob_ ;
with a straightforward generalization to operators with multiple su(2)H
su(2)C indices.
For instance, [RH ; Oabc] = 12 hadOdbc + 12 hbdOad
c
12 hdcOabd. The action of R` and Rr on
operators can then be inferred from simply combining (2.35) and (2.36).
2.3.3
Action of supersymmetries
The action of the odd generators of su(2j1)l
su(2j1)r on operators in general multiplets
can be quite complicated. As mentioned above, on the vector and hypermultiplet operators
their action is just a particular subset of the transformation rules (2.4){(2.9). The precise
(r ) is given in (C.9) and (C.21).
correspondence between the various aa_ obeying (2.13) and the supercharges Q
(` ) and
13Note that in our conventions hab = hba, and similarly for h.
2.4
The discussion in section 2.3 was restricted to the case of vanishing mass and FI parameters.
Introducing these parameters amounts to central extensions of the algebra (2.24), as we
will now describe.
It is not hard to see, using Jacobi identity, that one cannot introduce central charges
in (anti)commutators between left and right algebras, so one can only separately centrally
extend su(2j1)` and su(2j1)r. Each of these algebras admits only one nontrivial central
extension, so in total we have two central charges.
We denote the centrally extended
algebras with a tilde, so the supersymmetry algebra of our theories is s^u(2j1)`
s^u(2j1)r.
Denoting central charges of the left and right subalgebras by Z` and Zr respectively, the
only place where they appear are the following anticommutators:
Physically, the central charges Z` and Zr correspond to turning on real masses and FI
parameters. We turn on masses by coupling to background vectormultiplets in the Cartan
of the avor symmetry, as explained in section 2. The only components of these background
in (2.16).
_
multiplets which are nonzero are ha__ ( b:g:)ba_ =
b
( b:g:)1_2_ and hab(Db:g:)ba, as explained
The supersymmetry algebra has a gauge transformation on the right, as written in
eq. (2.18), with the gauge parameter
of (2.19). In gauge theories, dynamical gauge elds
force us to consider only operators which are not charged under the corresponding gauge
symmetry. For such operators, the gauge transformation in the SUSY algebra vanishes.
For background gauge
elds, this is not so.
We can have operators which are charged
under the corresponding global symmetry (which would be gauged if the gauge eld were
dynamical), and for them, such gauge transformations in the algebra will generate central
charges. A simple computation, using the expression (2.19) for , shows that:
1
r
(Z` + Zr) = i( b:g:)1_2_ = im ;
(Z`
Zr) = 0 :
1
r
Here mb = diag(mI ), where mI are real masses for hypers qaI, I being the avor index.
Analogously, FI parameters correspond to background twisted vectormultiplets in the
Cartan of the gauge group. They similarly generate central charges with:
1
r
(Z` + Zr) = 0 ;
(Z`
Zr) = ib:
Here, b = I tI acts nontrivially only on operators charged under the topological symmetry,
where I is the FI parameter and tI is the corresponding topological charge. Examples of
such operators are monopole operators.
(2.38)
(2.39)
b
1
r
Our aim in this section and the next is to describe a procedure that generalizes the
cohomological truncation of [4, 7, 8] from N = 4 SCFTs to the more general nonconformal
N = 4 theories on S3 that were described in section 2. The construction of [4, 7, 8] was
based on identifying two supercharges Q1H and Q2H in N = 4 SCFTs on R3, such that
the OPE restricted to their cohomology gives a certain quantization of the Higgs branch.
It was also possible to
nd another pair of supercharges, Q1C and Q2C , whose cohomology
similarly leads to a quantization of the Coulomb branch, though this second possibility was
not explored in detail. We will generalize both cases to nonconformal theories on S3, but,
just as in [4, 7, 8], our main focus will also be the Higgs branch.
We will nd that local operators in the cohomology, both for QiH and for QiC , can
only be inserted along a great circle S1
S3.14 The circle is the xed point locus of the
U(1) isometry that appears in the anticommutator fQ1H ; Q2 g or fQ1C ; Q2C g. In the case of
H
QiH , the operators that can be inserted on S1 will be referred to as \twisted Higgs branch
operators", because, as we will see, they are in 1to1 correpsondence with Higgs branch
chiral ring operators. Similarly, operators in QiC cohomology will be referred to as \twisted
Coulomb branch operators".
In section 3.1, we start by reviewing the construction of [4, 7, 8] in at space, and
then in section 3.2 we translate this construction to S3. In section 4 we describe the
generalization of this cohomology directly based on the su(2j1)`
su(2j1)r algebra.
3.1
SCFT in at space
Consider theories living on a threedimensional Euclidean space R
3 with the standard
coordinates ~x = (x1; x2; x3). The bosonic subalgebra of the osp(4j4) superconformal
algebra is so(4)
sp(4), where the generators of sp(4) are rotations M
, translations P and
special conformal transformations K , and the generators of the so(4) = su(2)H
su(2)C
Rsymmetry are denoted by Rab and Ra_ b_ and act on the Higgs and Coulomb branches,
respectively. The fermionic generators are Q aa_ and S aa_ , denoting Poincare and conformal
supercharges, respectively. The detailed description of this algebra can be found in appendix C.1.
De ne the two supercharges Q1H and Q2H by
Q1H = Q112_ +
Q2H = Q211_ +
1
2r
S121_ :
are nilpotent, i.e., (Q1H )2 = (Q2H )2 = 0, and their anticommutator is given by
In (3.1), r is some arbitrary parameter with dimensions of length.15 The supercharges Q1;2
H
Z =
ir
H H
4 fQ1 ; Q2 g =
M12 + R1_ 1_ :
Whether we consider the cohomology of Q1H or Q2H , the above equation implies that it
can be represented by elements from the Z = 0 subspace. In order to satisfy Z = 0, local
operators with zero R_ 1_ charge can only be inserted at the xed point locus of the M12
1
rotation, i.e., at the line x1 = x2 = 0.
15When we map (3.1) to S3, we will interpret r as the radius of the sphere.
14There is some freedom in choosing QiH;C, which corresponds precisely to the choice of great circle on S3.
(3.1)
(3.2)
There are Q1H;2exact twisted translation and dilatation given by
L
b
Lb0 =
=
=
1
i
i
1
1
4 fQ1H ; Q221_ g =
4 fQ2H ; Q122_ g = P3 +
8 fQ1 ; Q1Hyg =
H
1
8 fQ2 ; Q2Hyg =
H
8 fQ1H ; 2rQ211_
S121_ g
8 fQ2H ; 2rQ112_ + S222_ g =
i
2r
The twisted translation generated by Lb can be used to move cohomology classes along the
line x1 = x2 = 0. In particular, every cohomology class de ned at the origin can be
twistedtranslated to the whole line x1 = x2 = 0. It is those observables on the line which where
referred to before as twisted operators. Because Lb
is Q1H  and Q2H exact, this twisted
translation is a trivial operation at the level of the cohomology of Q1H or Q2H . Therefore,
to characterize the local operators in cohomology completely, it is su cient to consider
them inserted at the origin. By the stateoperator map, this corresponds to studying
the state cohomology. Using (3.4), Lb0 =
1 Hy
8 fQ1H ; Q1 g =
8 fQ2H ; Q2Hyg, the standard
1
Hodge theory argument proves that the cohomologies of Q1H and Q2H are identical and are
represented by the kernel of Lb0.
R3
As shown in [7, 8], these representatives are given by local operators Oa1 an (~0)
transforming in the (n + 1; 1) irrep of the su(2)H
su(2)C Rsymmetry and of conformal
dimension
= n=2. When translated with Lb they give the twisted operator:
R3
O(s) = Oa1 an ~x=(0;0;s)uaR13
uaRn3 ;
uR3
1;
s
2r
;
which de nes a nontrivial cohomology class on the line x1 = x2 = 0.
Local operators in the cohomology form a certain algebraic structure under the OPE
of the full theory. In particular, because Lb
is zero in cohomology, the OPE of operators
in the cohomology does not depend on their positions on the line, but it can depend on
their ordering. By moving operators to one point, we then de ne a product of cohomology
classes.
This way we get an algebra in the cohomology, which is associative but not
necessary commutative.
As explained in [8], the operators O(s), when inserted at the origin s = 0, are just
the Higgs branch chiral ring operators. However, as we move away from the origin, they
become mixed with antichiral operators, because of the twisting factor uR3 = (1; 2sr ).
This twisting factor can be thought of as an sdependent choice of the Cartan generator
of su(2)H given by 11+ss22==((22rr))22 3 + 1+s2=(2r)2 1
s=r
. A twisted operator O(s) is in the su(2)H
highest weight state with respect to this sdependent Cartan generator. The fact that the
twisted operators are not chiral with respect to a xed Cartan generator is responsible for
the fact that the algebraic structure we get is not a chiral ring, but rather its deformation
quantization.16 The deformation parameter is 21r , which was denoted by
in [8].
In fact, it turns out to be slightly more convenient to study the cohomology of a linear
combination Q1H +
Q2H with some generic . This operator squares to the bosonic
transformation Z, and so it plays a role of the equivariant di erential. The cohomology problem
16More precisely, the Higgs (or Coulomb) branch chiral ring has a natural Poisson structure, since it
corresponds to the ring of holomorphic functions on the moduli space, which for N
= 4 theories is a
hyperkahler cone. The algebraic structure we obtain is the deformation quantization of this Poisson algebra.
(3.3)
(3.4)
(3.5)
for this operator therefore involves two steps: one has to restrict to the Z = 0 subspace rst,
and then compute the cohomology there. The idea to treat a supercharge as an equivariant
di erential (with respect to spacetime symmetries) in SUSY gauge theories is an old one
and goes back, e.g., to [45, 46], while the associated quantization was rst discussed in [47].
Recall that Z is a sum of rotation in the (x1; x2) plane and a certain Rsymmetry
transformation. The condition Z = 0 then implies that geometrically, the con guration of
operators should be invariant under this rotation. In particular, local operators, as well
as line operators, can only be inserted at the line x1 = x2 = 0, which is the
locus of this rotation. Surface operators, on the other hand, can only span the orthogonal
(x1; x2) plane and correspond to some xed value of x3.
Including line operators would of course change the answer, and the cohomology of local
operators located at the line defect at x1 = x2 = 0 would give a di erent protected algebra.
Surface operators, on the other hand, are expected to give some modules for the protected
algebra to act on. They would describe point defects on the line x1 = x2 = 0, acted on
by the local operators. This action simply corresponds to merging local operators and
the defect together. Including extended operators gives an interesting direction for further
explorations, and it would potentially allow one to extract more dynamical information
about the theory. However, in this paper, we do not consider any extended operators and
study only the protected algebra of local operators.
3.2
SCFT on the sphere
Now let us identify the counterpart of the above construction on the sphere. After
describing it in some detail, we will be able to see that it generalizes to nonconformal theories in
a straightforward fashion.
Mills theory to an S2.
Using the stereographic map, one can place any conformal theory on S3. Under this
map, the line x1 = x2 = 0 maps to a great circle S1
S3, along which the cohomology
classes of local operators described in the previous subsection will be inserted. The rotation
in Z now becomes a U(1) isometry of the sphere, whose xed point locus is precisely this S1.
As mentioned in section 2.3.1, it will be useful to represent S3 as a U(1)
bration
over the disk D2, with
bers shrinking at its boundary @D2 = S1. This boundary S1,
parameterized by the angle ' at
= 2 , is the great circle mentioned above along which
local operators in cohomology can be inserted. The situation here is similar to that in [44],
where an analogous representation of S4 was used in the localization of 4d N = 4
Yang3.2.1
Twisted operators on S3 by stereographic map
(0; 0; si). Let us map them on S3:
In R3, we were interested in correlators of twisted operators Oi(si) inserted at points
hO1(s1)
Ok(sk)iR3 = hO1('1)
Ok('k)iS3 :
(3.6)
Operators on the right are the sphere counterparts of the at space twisted operators,
S3
and are given by contraction of the S3 operators Oa1 an (')
, inserted on the great
= 2
HJEP03(218)
O
S3 , with
O(') = cosn '
circle at
have O
R3 =
=2, with u = (1; x2r3 ) = (1; tan '2 ). For every operator of dimension
being the conformal factor, which evaluates to
= cos2 '2 at
=
=2. The de nition (3.5) then implies
S3
2 Oa1 an = 2
uaR13
S3
uaRn3 = Oa1 an = 2
uaS13
uaSn3 ;
(3.7)
where uS3 = uR3 cos '2 = (cos '2 ; sin '2 ). Note that the twisted operators O do not transform
with a conformal factor in going from R3 to S3, and this is why they do not carry an R
3
or S3 superscript and why there is no conformal factor in (3.6). We will now interpret this
construction in a more intrinsic way using the theory on S3 only.
In section 2 and appendix C, we chose an embedding of the su(2j1)`
su(2j1)r superalgebra
in osp(4j4), such that su(2)`
su(2)r
sp(4) corresponds to isometries of the sphere and
su(2j1)r by Q
be found in appendix C.2.
so(4)R = su(2)H
su(2)C was a Cartan subalgebra of the Rsymmetry
algebra. The choice of Cartan subalgebra was parametrized by the matrices h and h. To
be more precise, h parameterizes the Cartan generator RH in su(2)H and h parameterizes
the Cartan generator RC in su(2)C . The generators R` and Rr of u(1)` and u(1)r are given
by (2.35). The supercharges of su(2j1)` were denoted by Q
(` ), and the supercharges of
(r ). Their expressions in terms of conformal supercharges Q aa_ and S aa_ can
Using this embedding, it is easy to identify our supercharges Q1H and Q2H as:
Each of these supercharges is of course nilpotent, and
Q1H = Q1
(`+) + Q1
(r );
Q2H = Q2
(J3(`) + J3(r)) is the translation acting as P
= i@ on gaugeinvariant
where P
operators.
Using the su(2j1)` su(2j1)r algebra, we can check that:
n H r
Q1 ; 4i
Q2
(` )
Q2
(r+) o = n
r
Q2; 4i
Q1
(`+)
Q1
(r ) o = P' + RH
Pb' :
(3.10)
Here P' = J (r)
3
3
J (`) is simply the ' translation isometry of S3 acting on gauge invariant
operators as P' = i@'. The generator Pb' de ned above is a new twistedtranslation, which
is de ned on the sphere purely in terms of the su(2j1)`
su(2j1)r superalgebra.
Let Oa1 an be some local operator in the SCFT on the sphere,17 in the spinn=2 irrep
of su(2)H . If O11 1, when inserted at the point
=
=2, ' = 0 (which corresponds to the
origin of R3 upon stereographic projection), is in the cohomology of Q1H and Q2H , (recall
17From now on, we drop the superscript or subscript S3 present in the previous subsection.
Integrating out the Q's and X's we get the matrix model [39]
Z =
d 1d 2
sinh2( ( 1
2))
The U(2) gauge theory with a fundamental and an adjoint hypermultiplet is believed
to ow to the same IR xed point as the N = 8 U(2) YangMills theory. The IR
SCFT has two N = 8 stress tensor multiplets, one of which corresponds to a free sector
and one to an interacting sector. Intuitively, the free sector corresponds to the IR limit of
the diagonal U(1) in the YangMills description, while the interacting sector corresponds
to the IR limit of SU(2) YangMills theory, as will be made more precise shortly.
It was shown in [7] that upon decomposition to N = 4 SCFT notation, the 1d Higgs
branch theory has a avor su(2)F symmetry that is a subgroup of the so(8) Rsymmetry.
Under su(2)F , (Xe ; XT ) form a doublet.33 In order to match the notation in [7], let us
introduce polarization variables ya, a = 1; 2, and denote the operators in the 1d theory by
O2jF ('; y) = Oa1:::a2jF ya1
yajF ;
(7.41)
where jF is the spin of the su(2)F representation.
We will identify 3 operators in the 1d theory and compute their correlation functions:
The twisted Higgs branch representative of the N
= 8 free
eld multiplet. The
N = 8 free
eld multiplet consists of 8 scalar operators of scaling dimension 1=2
and 8 spin1=2 operators of scaling dimension 1. Under the decomposition to N = 4
supersymmetry, 4 of the scalar operators are interpreted as Higgs branch operators
(transforming under su(2)H
su(2)F as (2; 2), while the other 4 are Coulomb branch
operators. From the 4 Higgs branch operators one can construct the twisted Higgs
branch operator O1;free('; y).
The twisted Higgs branch representatives of the free and of the interacting N =
8 stress tensor multiplets. Any N = 8 stress tensor multiplet contains 35 scalar
operators of scaling dimension 1, 9 of which being Higgs branch operators from an
N = 4 point of view. From them, one can construct twisted Higgs branch operators
O2('; y). We will denote the operator corresponding to the free stress tensor multiplet
by O2;free('; y) and the one corresponding to the interacting stress tensor multiplet
by O2;int('; y).
7.3.1
Free N = 8 multiplet
The free multiplet operator O1;free('; y) is
O1;free('; y) = y1 tr Xe (') + y2 tr X(') :
(7.42)
by
33Because of the Dterm relations, we may construct operators only from X and Xe. Indeed, the equations
of motion for the auxiliary eld Dab imply QejQi+XejkXik
XekiXkj = 0, so every pair QejQi can be replaced
XejkXik + XekiXkj. Since gaugeinvariant operators can only contain an equal number of Q's and Qe's
such replacements yield expressions depending only on X and Xe.
From (7.38), we see that tr Xe and tr X only appear in the kinetic term, so computing
correlation functions of these operators can be performed using the propagator
htr X('1) tr Xe ('2)i =
are necessary to establish (7.43).) Using this expression and (7.42),
one obtains
hO1;free('1; y1)O1;free('2; y2)i =
hy1; y2i sgn('1
'2) :
1
`
where the angle bracket notation is de ned by
Higher point functions of O1;free('; y) can be computed using Wick contractions
using (7.44).
Free N = 8 stress tensor multiplet
There are two su(2)F triplets of linearly independent operators that are quadratic in X
corresponding to the two stress tensor multiplets of the theory. It is easy to identify the
one corresponding to the free N = 8 multiplet because this is the only one appearing in
the OPE of O1;free
O1;free: it is simply the square of the free N = 8 operator O1;free('; y),
We can then easily see that
(7.43)
(7.44)
(7.45)
(7.46)
(7.47)
(7.48)
(7.49)
(7.50)
O2;free('; y) = (y1)2(tr Xe )2 + 2y1y2(tr Xe )(tr X) + (y2)2(tr X)2 :
Again using (7.43) gives
hO2;free('1; y1)O2;free('2; y2)i =
2
`2 hy1; y2i2 :
7.3.3
Interacting N = 8 stress tensor multiplet
The interacting stress tensor multiplet must be orthogonal to the free one. To obtain O2;int,
we rst compute the matrix of 2point functions
h(tr X)2(') (tr Xe )2(0)i h(tr X)2(') (tr Xe 2)(0)i!
h(tr X2)(') (tr Xe )2(0)i h(tr X2)(') (tr Xe 2)(0)i
=
1
1
+ (y2)2 (tr X2)
(tr X)2
:
1
2
1
2
1
2
(tr Xe 2)(')
(tr Xe )2(') (tr X)2(0)
= 0 ;
which implies that the (y1)2 component of O2;int('; y) is (tr Xe 2)(')
12 (tr Xe )2(') up to an
overall normalization factor of our choice. The su(2)F symmetry then implies
O2;int('; y) = (y1)2 (tr Xe 2)
(tr Xe )2
+ 2y1y2 (tr XXe T )
(tr X)(tr Xe )
1
2
Computing the twopoint function of O2;int is more challenging, as one now has to use the
nontrivial propagators coming from (7.39). A careful calculation shows that the twopoint
function is
hO2;int('1; y1)O2;int('2; y2)i = hy1; y2i2 1 Z
sinh2(
d 1d 2 16 cosh4( 12) cosh( 1) cosh( 2)
12) [5 + cosh(2
4`2
Z
=
2
We can use the formalism we have developed to calculate the 4point functions of
O2;free('; y) and O2;int('; y) and compare with [7]. In [7] it was found that the 4point
function of an operator O2('; y) corresponding to an N = 8 stress tensor multiplet (which
could be any linear combination of O2;free('; y) and O2;int('; y)) is
hO2('1; y1)O2('2; y2)O2('3; y3)O2('4; y4)i
= C2hy1; y2i2hy3; y4i2 1+
Here, w is de ned as
the constant C is given by the normalization of the operator,
w
hy1; y2ihy3; y4i ;
hy1; y3ihy2; y4i
hO2('1; y1)O2('2; y2)i = Chy1; y2i2 ;
and (2B;2), (2B;+), and s2tress are the squares of the various OPE coe cients of N = 8
superconformal multiplets appearing in the OPE of the N = 8 stress tensor multiplet with itself.
The fourpoint function of O2;free('; y) does not require any integrals, as it again only
uses (7.43). When '1 < '2 < '3 < '4, we obtain
hO2;free('1; y1)O2;free('2; y2)O2;free('3; y3)O2;free('4; y4)i = `4 hy1; y2i2hy3; y4i2 6+2w 2w2
4
w2
Obtaining O2;int('; y) is slightly more complicated. The nal result is
hO2;int('1; y1)O2;int('2; y2)O2;int('3; y3)O2;int('4; y4)i =
Comparing these expression with (7.52), we nd
15`4 hy1; y2i2hy3; y4i2 4 + w
8
w2
Free stress tensor:
Interacting stress tensor:
2
stress = 16 ;
2
stress = 12 ;
2
(B;+) = 16 ;
2
(B;+) =
64
5
;
2
(B;2) = 0 ;
2
(B;2) = 0 :
(7.51)
(7.52)
(7.53)
(7.54)
;
:
(7.55)
w2
(7.56)
(7.57)
HJEP03(218)
The expressions for the OPE coe cients of the free N = 8 stress tensor multiplet match
the result of [7] in the free N = 8 theory (of 8 free massless scalars and 8 free massless
Majorana fermions), while the corresponding expressions obtained for the interacting stress
tensor match those obtained in [7] for the U(2)2
U(1) 2 ABJ theory. The former theory
is the infrared limit of N = 8 super YangMills theory with gauge group U(1), while the
latter theory is the infrared limit of N = 8 YangMills theory with gauge group SU(2).
These results show quite explicitly how, at the level of the N = 4 Higgs branch theory, the
IR limit of N = 8 U(2) YangMills theory (or the U(2) gauge theory with one fundamental
and one adjoint hypermultiplet) is a product between a free theory and the IR limit of
= 4 QFTs on S3 with nonvanishing mass and FI
SU(2) YangMills theory.
8
Applications to N
parameters
either m or
nonvanishing.
The S3 partition function is
Let us now present a few examples of correlation functions in nonconformal theories with
Deformation by FI parameters
SQED with nonzero FI parameter
Turning on a nonzero FI parameter in SQED is easily implemented by replacing Z in (7.1)
and subsequent formulas in section 7.1 by
Z
=
e2 i `
N
2
i `
2 (N
N2 + i `
1)!
:
The twopoint function of J
(p) is then still given by (7.11). The integrals evaluate to
(8.1)
(8.2)
(8.3)
:
(8.4)
Comparing with eq. (1.4) of [58] we see that (8.4) agrees with the bilinear form of the
generalized higher spin algebra hs (sl(N )) with parameter
=
2i `=N .
(p)('1; y1; y1)J (p)('2; y2; y2)i
( 1)p (N )
N
2
=
i ` + p
N
2
i `
N2 + i ` + p
N2 + i `
(y1 y2)p(y1 y2)p :
These twopoint functions can be combined into a single formula upon using the de
nition (7.14). We have
hJ ('1; y1; y1)J ('2; y2; y2)i = 3F2
N
2
N
2
i `;
+ i `; 1;
N N + 1
2
2
;
(y1 y2)(y1 y2)
4
=
N
N
X
j=1
j = 0 :
Because the j sum to zero, it is possible to write them as
Z = Y
N
j=1
e2 i` j j
j = !j 1
!j ;
for some !j , and then using summation by parts one can write
One can also compute 3point functions. We have, for example,
hJ
(1)('1; y1; y1)J (1)('2; y2; y2)J (1)('3; y3; y3)i
1
2i `
N + 2
(y1 y2)(y2 y3)(y3 y1) ;
(y1 y3)(y3 y2)(y2 y1)
(8.5)
(8.6)
(8.7)
(8.8)
(8.9)
JI J ? JK
L = JIK JL +
K JI L + ILJK
J
J
( IJ JK
L + KL JI J )
2
N
2
N (N + 1)
L J
I K
2` K JI L +
N
4`2(N + 1)
L J
I K
In the limit ` ! 1, (8.6) reduces to the commutative product on the deformed Higgs
branch chiral ring. Indeed, using (7.6) and (7.8), one can check that the multiplication of
J
JI JK
L yields the ` ! 1 limit of (8.6) provided that the relation QeI QI = i is satis ed,
as appropriate for the deformed Higgs branch chiral ring.
8.1.2
N node quiver with nonzero FI parameters
The N node quiver has gauge group U(1)N =U(1) containing N
1 Abelian factors. Conse
quently, there are N
1 linearly independent FI parameters that can be introduced. Let us
introduce an FI parameter j for each one of the N gauge group factors with the constraint
The deformation to nonzero j 's is realized by modifying the expression of Z in (7.19) to
which matches eq. (1.4) of [58] upon making the identi cation
=
2i `=N . (See
footnotes 31 and 32.) The star product of the generators of the chiral ring becomes
N
X
j=1
N
X
j=1
N
j=1
X !j ( j
j j =
j (!j 1
!j ) =
j+1) :
(8.10)
we obtain
condition
This expression can be substituted into (8.8). Upon performing the Fourier transform to
the j coordinates using now
e 2 i`!j( j j+1)
=
Z
e2 i`( j j+1) j
d j 2 cosh( ( j + `!j))
;
Z
Z =
QjN=1 [2 cosh( ( + `!j))]
:
The S3 partition function (8.12) agrees with that of SQED with N hypermultiplets with
real masses !j, as required by mirror symmetry. (See [59] where this equivalence was rst
shown at the level of the S3 partition function.) Note that an overall shift in !j can be
\gauged away" by shifting the integration variable . We will thus impose a gauge xing
In the presence of the FI terms, we can use a modi ed de nition of the operators (7.22):
X = Q1Q2
QN ;
Y = Qe1Qe2
QeN ;
Z = Qe1Q1 i!1 = Qe2Q2 i!2 = : : : = QeN QN
i!N :
They obey the classical relation X Y = (Z + i!1)(Z + i!2)
the deformation of the Kleinian singularity X Y = Z
N with parameters !j.
(Z + i!N ), corresponding to
With the de nition in (8.14), we have
hZ('1)Z('2)i = `2 Z
d
1 1 Z
(i )2
QjN=1 [2 cosh( ( + `!j))]
:
Then the second equation in (7.27) still holds, and we have
hX ('1)Y('2)i =
More generally, de ning
1 1 Z
`N Z
d
QjN=1 i( + `!j)
1
2
QjN=1 [2 cosh( ( + `!j))]
p
j=1
Zcp = Y(QejQj i!j)
we nd that for '1 < '2 < '3 we have
hZcp('1)Zcq('2)i = `p+q Z
hZcp('1)X ('2)Y('3)i =
1 1 Z
1 1 Z
`N+p Z
d
d
(i )p+q
QjN=1 [2 cosh( ( + `!j))]
(i )p QjN=1 i( + `!j)
QjN=1 [2 cosh( ( + `!j))]
;
1
2 :
From these expressions, it is straightforward to extract the corresponding star product
deformed by the parameters !j.
N
j=1
X !j = 0 :
(8.11)
(8.12)
(8.13)
(8.14)
(8.15)
(8.17)
(8.18)
;
for '1 < '2 :
(8.16)
The N node quiver has a U(1) avor symmetry under which the Qi carry charge +1=N
while Qei carry charge
operators X and Y carry charges +1 and
1, respectively. 1=N . This normalization of the U(1) charge is such that the We can introduce a real mass term associated with this avor symmetry by adding to the exponent of (7.18). This amounts to replacing Z in (7.19) by
The partition function is given by the equation
`
Z
d'
mr
N QeiQi
Z = Y
N
j=1
1
2 cosh ( ( j
j+1 + mr=N ))
:
Z =
N
2
imr
2 (N
N2 + imr
1)!
;
which, upon the replacement mr !
`, can be seen to agree with eq. (8.2) of the partition
function of SQED with N charged hypers and FI parameter . Indeed, under mirror
symmetry the real masses and FI parameters are interchanged.
Eqs. (7.27) and (7.28) still hold, with the only change that j(j+1) is replaced by
j(j+1) + mr=N . We obtain, for instance, that
1 1 Z
`2 Z
d
e2 imr
More generally, we can de ne the operators Zcp with p
functions is given by
N , whose matrix of two point
hZcp('1)Zcq('2)i =
1 1 Z
`p+q Z
d
e2 imr
)]N (i )p+q :
where
m ! 1:
The mixing of these operators can be removed by performing a GramSchmidt
procedure as was the case for SCFTs. For example, we can remove the mixing with the identity
operator by subtracting the expectation values of the operators. Explicitly, (8.22){(8.23)
imply that the connected correlation function of Z is
hZ('1)Z('2)i
hZ('1)ihZ('2)i =
(1) N
2
imr +
(1) N2 + imr
4 2`2
;
(8.24)
(n)(z) is the polygamma function. One can see that this function vanishes as
(8.19)
(8.20)
(8.21)
(8.22)
(8.23)
The SQED theory with N charged hypermultiplets has an SU(N ) avor symmetry. One
can introduce N
by adding
1 real mass parameters corresponding to the U(1)N 1 Cartan of SU(N )
N
I=1
`
Z
d'
X mI r QeI Q ;
I
N
I=1
X mI = 0
to the exponent of the second equation in (7.1). The condition PN
that the mI are real masses for the Cartan of SU(N ). The expression for the S3 partition
function in (7.1) gets replaced by
Z
Z =
d Z ;
Z =
1
QIN=1 2 cosh( ( + mI r))
:
The S3 partition function agrees with that of the N node quiver (8.12) upon the
replacement mI r ! `!I , in agreement with mirror symmetry.
While it is possible to perform computations for arbitrary N , for simplicity let us give
an example in the case N = 2 where we take m1 =
m2 = m. The partition function
in (8.26) evaluates to
Z = mr csch(2 mr)
in this case. Let us de ne the quadratic operators
J3 =
1
2
Qe1Q
1
Qe2Q
2 ;
2
J+ = Qe1Q ;
1
= Qe2Q :
The operator J3 is neutral under the U(1) Cartan of avor SU(2) symmetry, so it's
correlation functions are independent of position. We obtain, for instance,
hJ3('1)J3('2)i hJ3('1)ihJ3('2)i =
hJ3(')i =
1 2 mr coth(2 mr)
4 mr`
1+8 2m2r2 cosh(4 mr) csch2(2 mr)
32 2m2r2`2
:
(8.29)
On the other hand, the operators J
carry charges
2 under the Cartan of the avor
SU(2). Their expectation values must vanish because they cannot mix with the identity
operator. Their correlation functions, however, do depend on position as in (6.16) with
FJ (mr) =
2mr. We obtain
hJ+('1)J ('2)i =
e2mr('1 '2) [1
2mr coth(2mr )] [coth(2mr )
One can see that both (8.29) and (8.30) interpolate between a nontrivial topological
expression at mr = 0 and they both tend to zero as mr ! 1. Indeed, if we interpret mr as
the RG scale, then at small mr we are probing the UV SCFT, while at large mr we are
probing the infrared.
(8.25)
(8.26)
(8.27)
(8.28)
:
(8.30)
In this paper we used supersymmetric localization to derive a 1d theory coupled to a
matrix model, given in (6.1), that can be used to calculate correlation functions of twisted
Higgs branch operators of N
= 4 QFTs on S4. In the case of N
= 4 SCFTs, this
theory provides a Lagrangian realization of the protected Higgs branch topological sector
discussed in [7, 8]. The immediate practical application of (6.1) is to the computation of
2 and 3point functions of Higgs branch operators.
Our results can be used to perform more detailed tests of mirror symmetry. We have
seen, for instance, that in the N node necklace quiver, the twisted operator Z has the
1 1 Z
`2 Z
d
1
This theory is mirror dual to SQED with N
avors. One expects the twisted Higgs branch
operator Z in the N node quiver to be mirror dual to the twisted Coulomb branch operator
constructed from the vectormultiplet scalars in SQED. In section 5.2.2 we explained that
the 2point function of
can be computed by replacing each insertion of
by 2 =r in the
KWY matrix model, thus obtaining
(9.1)
(9.2)
h ('1) ('2)i =
64 2 1 Z
`2 Z
d
1
Comparing (9.1) and (9.2), we can thus identify Z in the N node quiver with
i =(8 )
in SQED. A similar exercise shows that, at least for p
p in the N node quiver
can be identi ed with [ i =(8 )]p in SQED. These are, of course, rather simple tests of
mirror symmetry. It should be possible to perform more nontrivial tests in nonAbelian
gauge theories.
There are a few generalizations of our results that we have left for the future. One
such generalization is to N = 4 gauge theories that include twisted vectormultiplets and
twisted hypermultiplets, which would then open the possibility of including ChernSimons
interactions. Another such generalization would be to complete the Coulomb branch
localization computation by allowing for insertions of monopole operators. Yet another such
generalization would be to Higgs branch operators in theories with 8 supercharges de ned
in a di erent number of spacetime dimensions. We hope to report on these questions in
)]N
N , Z
2
:
future publications.
Acknowledgments
We thank Chris Beem, Cyril Closset, Thomas Dumitrescu, Jaume Gomis, Bruno Le Floch,
Wolfger Peelaers, Herman Verlinde, and Edward Witten for useful discussions. The work
of SSP was supported in part by the US NSF under Grant No. PHY1418069, and that of
RY by NSF Grant No. PHY1314198. Work of MD was supported in part by Walter Burke
Institute for Theoretical Physics and the U.S. Department of Energy, O
ce of Science,
O ce of High Energy Physics, under Award Number DESC0011632, as well as Sherman
Fairchild foundation. { 69 {
Conventions
Curved space vector indices are denoted by ; ; : : :, while frame indices are denoted by
i; j; : : : = 1; 2; 3. We label the doublet (spinor) representation of the SU(2)rot. frame rotation
group by ; ; : : : = 1; 2, of SU(2)H by a; b; : : : = 1; 2, and of SU(2)C by a_; b_; : : : = 1; 2.
Spinor indices are raised and lowered from the left with the antisymmetric tensors "
and " , where "12 =
SU(2)C
SU(2)H indices (e.g., aa_
"12 =
their contraction is de ned with the convention:
1. The same conventions are used for raising and lowering
"ab"a_ b_
bb_ ). When SU(2) spinor indices are suppressed
(
) :
(A.1)
HJEP03(218)
In particular, for any three spinors x, y and z (either commuting or anticommuting) we
have the Fierz identity:
x (yz) + (xy)z + x y z = 0 :
We will always take variation spinors , as in (2.4){(2.9), to be commuting, while the
symbol itself to be anticommuting.
The at space gamma matrices are the usual Pauli matrices, ( i)
i, which satisfy
i j = ij + i"ijk k
;
( i) ( i)
= 2
:
("123 = 1) ;
Given a Euclidean metric g
an orthonormal frame is de ned by
and R = R
, respectively.
(A.2)
(A.3)
(A.4)
(A.5)
(A.6)
(A.7)
(A.8)
(A.9)
(A.10)
(A.11)
A spin connection ! ij =
! ji is then xed from the conditions
The Riemann tensor is
while the Ricci tensor and scalar are de ned by R
With this de nition R = 6 for a round unit 3sphere.
The space covariant derivative of spinors is de ned as
g
= ei ej
ij ;
ij = g ei ej :
dei + !ij ^ ej = 0 ;
e
ei dx :
) ;
= R
while the Lie derivative L^v along v acting on scalars , spinors , and vector elds A , is
given by
r
! ij ijk k) ;
v r
r v
i
We will hereby summarize various details on di erential geometry on S3 that are used in
the main text. Let S3 be the radius r 3sphere embedded into C2 as
Each point on S3 can be represented by an SU(2) element
jz1j2 + jz2j2 = 1 ;
r~z 2 C2 :
g =
z2 iz1!
iz1 z2
:
dei(`) +
dei(r)
1r ijke(`)
1r ijke(r)
k ^ e(j`) = 0 ;
k ^ e(jr) = 0 ;
The su(2)valued left/right invariant 1forms !(`=r), and the frame 1forms e(`=r) associated
with them are de ned as
i e(`) i
r i
dgg 1 =
i e(r) i
r i
:
They satisfy the MaurerCartan equations
(A.12)
(A.13)
(A.14)
(A.15)
(A.16)
(A.17)
(A.18)
(A.19)
(A.20)
i
ds2 = e(`)ie(`) = e(r)ie(r) = r2(d 2 + cos2( )d 2 + sin2( )d'2) :
34Li generates the left SU(2) action Lig =
12 ig, while Ri generates the right action Rig = 12 g i.
from which the spinconnections can be directly reado by using (A.6).
The su(2)`
and Rj , which are dual to the 1forms ei(r) and ei(`) up to proportionality constants that
su(2)r isometries of S3 are generated, respectively, by the vector elds Li
ei(`)(Rj ) =
ir j
2 i ;
ei(r)(Lj ) =
ir j
2 i :
[Li; Lj ] = i"ijk k
L ; [Ri; Rj ] = i"ijk k
R :
They satisfy the su(2) algebra
the metric on S3 is given by
In the round coordinates
z1 = cos( )ei ;
z2 = sin( )ei' ;
and the vectors Li = Li @
Ri = R
tan( ) sin( + ')@ + cot( ) sin( + ')@') ;
tan( ) cos( + ')@ + cot( ) cos( + ')@') ;
L
L
L
R
R
R
1 =
2 =
3 =
1 =
2 =
3 =
2
2
2
2
2
2
( sin(
')@ + tan( ) cos(
It will also be useful to introduce stereographic coordinates. Let
rz1 = X1 + iX2 ;
rz2 = X3 + iX4 :
The stereographic coordinates xi (i = 1; 2; 3) are de ned as
X1;2 =
x1;2 =
1 + 4xr22 ;
2X1;2
1 + X3=r
X4 =
x3 =
2X4
1 + X3=r
X3 = r
x2
1 + 44xrr222 ;
;
x
2
x21 + x22 + x32 :
In our de nition the origin ~x = (0; 0; 0) is mapped to X~ = (0; 0; r; 0). The induced metric
on S3 is conformally at
and we de ne the stereographic frame as
g
= e2
e =
1
1 + 4xr22 ;
e
i = e
i :
Let us summarize how Killing spinors on S3 look in the di erent frames that we
introduced. The spinor covariant derivatives in the left and right invariant frames are
given, respectively, by
r
left inv.
i
2r
right inv.
2r
:
Let (`) and (r) be spinors satisfying
(`) ;
(r) =
(r) :
Then in the left invariant frame (`) is some constant spinor (`), while (r) is some constant
spinor (r) in the right invariant frame. In the stereographic frame one can check that
(`) = e =2 1
(`) ;
(r) = e =2 1 +
(r) :
(A.34)
r
(`) =
2r
2r
i i
x i
r
r
2r
2r
i i
x i
(A.21)
(A.22)
(A.23)
(A.24)
(A.25)
(A.26)
(A.27)
(A.28)
(A.29)
(A.30)
(A.31)
(A.32)
(A.33)
Closure of superconformal algebra
For any two spinors aa_ and ~aa_ satisfying (2.3), the anticommutator of supreconformal
transformations (2.4){(2.9) acting on any eld
closes up to equations of motion into
The operator K^ ;~ is de ned as
where
f ; ~g
K^ ;~ + G^
+ e.o.m. :
K^ ;~
L^v + R^C + R^H +
^ ;
Lv is the Lie derivative along v
i ~aa_
aa_ .
R^C=H is an su(2)C=H transformation, acting on doublets with the matrices
Ra_ b_
Rab
i( ~c(a_ j0cjb_) + c(a_ ~j0cjb_)) ;
i( ~(ac_ 0
b)c_ + (ac_ ~b0)c_) ;
for triplets: R^H Dab = RacDcb + RbcDac, etc.
such that, e.g., R^C a_ = Ra_ b_ b_ , R^H qa = Rabqb, and with the obvious generalization
is the dilation transformation parameter
= i( ~ab_ a0b_ + ab_ ~a0b_ ) :
= ( ~c a_ cb_ ) a_ b_
v A ;
and those of the hypermultiplet (2.2) have ^ [H] = ( 12 ; 1).
The components of the vectormultiplet (2.1) appear with dimensions ^ [V] = (0; 32 ;1; 2),
G^ is a gauge transformation with parameter
such that, e.g., G^ A
= D
, G^
a_ b_ = i[ ; a_ b_ ], G^ qa = i qa, G^ qea =
i qea, etc.
C
C.1
N
= 4 algebras
Superconformal algebra
The 3d N = 4 superconformal algebra is osp(4j4), and its bosonic subalgebra so(3; 2)
su(2)H consists of conformal and Rsymmetry transformations. In at space, the
conformal symmetry generators can be divided into translations P , rotations M
,
dilatations D and special conformal transformations K . The su(2)C and su(2)H Rsymmetry
generators will be denoted by Rab and Ra_ b_ , respectively. The corresponding subalgebra is
P
K
M
K
M
P ;
K
; [D; P ] = P
; [D; K
K
;
[P ; K ] = 4 (
( M )
) + 4 (
) D ;
[Rab; Rcd] =
adRcb + cbRad ; [Ra_ b_ ; Rc_d_] =
a_ dRc_b + c_bRa_ d_ ;
_ _ _
(B.1)
(B.2)
(B.3)
(B.4)
(B.5)
(B.6)
(C.1)
(C.2)
(C.3)
(C.4)
(C.5)
HJEP03(218)
P
( )
P ;
K
( )
K ;
M
The algebra (C.1){(C.5) is represented on a dimension
in the (2; 2) irrep of su(2)C
su(2)H as
M
:
(C.6)
scalar primary operator Oaa_ (x)
[P ; Oaa_ (x)] = i@ Oaa_ (x) ;
2x (x @) 2 x )Oaa_ (x) ;
[M
[Rab; Occ_(x)] = cbOac_
2 abOcc_ ;
[D; Oaa_ (x)] = (x @ + )Oaa_ (x) ;
[R a_b_ ; Occ_(x)] = c_bOca_
_
1
_
2 a_bOcc_ ;
(C.7)
(2.9) as follows. The solution of the conformal Killing spinor equation (2.3) on R3 is
The transformations of the odd generators in osp(4j4) can be read from the variations (2.4){
We then de ne the action of the Poincare supercharges Q aa_ and conformal supercharges
S aa_ by
algebra is
and also
aa_ = aa_ + xi i aa_ ;
a0a_ = aa_ :
2
O
[ aa_ Q aa_ +
aa_ S aa_ ; O] :
The commutators of the odd generators Q aa_ and S aa_ then follow by matching the action
of K^ ;~ de ned in appendix B with (C.7). The resulting oddodd and evenodd part of the
fQ aa_ ; Q bb_ g = 4"ab"a_ b_ P
fS aa_ ; S bb_ g = 4"ab"a_ b_ K
;
; Q aa_ ] = i
S aa_ +
Q aa_
Q aa_ ;
[M
; S aa_ ] =
S a a_+
S aa_ ;
1
2
1
2 abQ cc_ ;
1 _
2 a_ bQ cc_ ;
(C.8)
(C.9)
(C.10)
(C.12)
(C.13)
(C.14)
(C.15)
(C.16)
(C.17)
1
2
[D; S aa_ ] =
S aa_ ;
[Rab; S cc_] = cbS ac_
[Ra_ b_ ; S cc_] = c_b_ S ca_
1
2
1
2 abS cc_ ;
1 _
b
2 a_ S cc_ ;
i
:
fQ aa_ ; S bb_ g = 4i h"ab"a_ b_ M
D
"a_ b_ Rab + "abRa_ b_
C.2
Nonconformal N = 4 algebra on S3
We will now construct the S3
N = 4 algebra su(2j1)`
su(2j1)r explicitly, as a subalgebra
be decomposed into commuting spinors (\twistors") ua and ua_ as
of the osp(4j4) superconformal algebra de ned in (C.1){(C.5) and (C.10){(C.16). The
matrices hab and ha_ b_ in (2.13) are traceless and square to 1. Therefore, they can always
hab = u+aub + u aub+ ;
ha_ b_ = ua+_u b_ + ua_ u+b_ ;
where (u+u ) = (u+u ) = 1. The decomposition to twistors (C.17) simpli es the
construction of the nonconformal subalgebra, because it eliminates the need to carry the
su(2)H indices. The twistors u and u are simply the eigenvectors of hab and ha_ b_ :
ua hab =
ub ;
Let us parameterize the Cartan of su(2)C
ha_ b_ u
b_ =
ua_ :
su(2)H as35
3
RH
1
2 habRba = (u+Ru ) ;
2
1 ha_ b_ Rba_ = (u+Ru ) :
_
R` = RH3 + RC3 ;
3
Rr = RH
RC3 :
The generators of the u(1)`
su(2)C
are then de ned in terms of (C.19) to be
su(2)H Rsymmetry of the S3
N = 4 algebra
HJEP03(218)
su(2j1)r are given by
Q
(` )
2
1+i ua ua_
2r
Q aa_ +
Q
(r )
2
1+i ua ua_
2r
Q aa_
S aa_ :
(C.21)
stant by demanding that Q
commutators are
The relative coe cients between Q a a_ and S aa_ in (C.21) can be xed up to one
con(` ) anticommute with Q
(r ). The only nontrivial oddodd
where J (`) and J (r) are the su(2)`
su(2)r isometry generators of S3 de ned by
fQ
g =
fQ
4i
4i
J (`) +
J (r) +
1
2
1
2
J (`) =
J (r) =
2
2
P
P
1
4r2
K
1
4r2
K
M
1
1
M
R` ;
Rr ;
In particular, if we denote their components as
then the only nontrivial eveneven commutators are
J (`)
J (`) J +(`) !
3
J (`)
3
(J (`))
; (J (r))
J (r)
J (r) J +(r) !
3
J (r)
3
;
[J3(`); J (`)] =
[J3(r); J (r)] =
J (`) ; [J +(`); J (`)] = 2J3(`) ;
J (r) ; [J +(r); J (r)] = 2J3(r) :
35Recall that (u+u )
ua+u a and (u+Ru )
ua+Rabu b, etc.
(C.18)
(C.19)
(C.20)
(C.22)
(C.23)
(C.24)
(C.25)
(C.26)
(C.27)
(C.28)
The action of the generators J (`=r) on a scalar operator O(x) on S3 is given by
[J3(`); O(x)] =
[J (`); O(x)] =
L3O(x) ;
[J3(r); O(x)] =
iL2)O(x) ; [J (r); O(x)] =
R3O(x) ;
iR2)O(x) ;
where Li and R
i were de ned in (A.21){(A.26).
Finally, the nontrivial evenodd commutators are
(` )
[Rr; Q(r )] =
Q
(` ) ; [(J (r)) ; Q(r )] =
Q
(r ) ;
Q
(r )
1
2
For the choice of hab =
2 and ha_ b_ =
3, we can take
ua+ =
1
2
1
1 + i
u
a =
1
1
i !
ua+_ =
1
u
a_ =
1
One then nds that (C.29) (C.30) (C.31)
Q
(r ) :
(C.32)
(C.33)
(C.34)
(C.35)
(D.1)
(D.2)
(D.3)
(r ) + Q1
(`+) = Q112_ +
Q2
(r+) = Q211_ +
1
2r
1
2r
S121_ ;
are the nilpotent supercharges used to de ne the Higgs branch cohomology in [7, 8].
D
1d Green's function from 3d theory
The Green's function (5.22) of the fundamental twisted Higgs branch operators (4.14)
inserted on the
= 2 circle in S3, can be calculated directly from the 3d Gaussian
action (5.10). Without loss of generality let us consider a U(1) gauge theory with one
hypermultiplet. The bosonic part of the action is
where the operator Dab(x) is de ned by
Sfree hyper =
Z
d3x pg qa(x)Dab(x)qb(x) ;
Dab(x)
2
2 + 43r2 + r2
r2
r2
2
2 + 43r2 + r2
!
hqa(x)qeb(x0)i by solving the di erential equation
It is a straightforward exercise to determine the twopoint function Gab(x; x0) =
Dac(x)Gcb(x; x0) =
b
a
3(x
x0) :
Gab(x; x0) = hqa(x)qeb(x0)i =
1
8 r cosh(
sin( =2)
cos( =2)
) sinh(
) cosh(
cos( =2)
sin( =2)
) !
where
is the relative angle between the points x and x0. In the coordinates ( ; ; ') used
previously, it is given by
cos
= cos cos 0 cos(
0) + sin sin 0 cos('
'0) :
In particular, when both x and x0 belong to the circle at
=2, we have
Using the de nition (4.14) of Q(') and Qe(') in terms of the elds qa(x) and qa(x) evaluated
= j'
HJEP03(218)
(D.5)
(E.1)
H
c_b_ ;
(E.2)
(E.3)
(E.4)
(E.5)
hQ(')Qe(0)i =
cos
2 G12('; 0)
sin
2 G22('; 0) ;
which, when using (D.4), can be seen to agree precisely with (5.22).
E
Q
H BPS equations
In this section we will study the full set of BPS equations
aa_ =
; ea_ = 0 ;
where the transformations were de ned in (2.5), (5.32) and (5.33), the Killing spinor
=
is de ned in (5.5), and aa_ satis es (5.34).36 Here, we study the consequences of (E.1) before
the reality conditions are imposed on the elds.
Let us unpack the contents of these equations. The gaugino BPS equations, can be
used to solve for the auxiliary elds Dab. This solution can be written as
iDab =
1
4
v
+ ac( aa_
_
bb)D
a_ b_ + 2 ac( ca_ b0b_ ) a_ b_
1
2 ab a_ b_ a_
where the symmetric matrices ab and
a_ b_ are given by
and v is the Killing vector generating translations along :
ab = ( ac_ bc_) ;
a_ b_ = ( ca_ cb_ ) ;
det( ab) = det( a_ b_ ) =
2 cos2( ) ;
v
i aa_
aa_ :
The remaining gaugino BPS equations imply that the elds are independent of
up
to a eld dependent gauge transformation. The result is more conveniently expressed in
terms of the twisted elds
e1_1_
i
1_1_ ;
e2_2_
e i
2_2_ ;
36In this section, we will always write for the particular spinor H de ned in (5.5) to avoid clutter.
squares to Z. Let us also de ne a modi ed connection D
? as37
which satisfy [Z; e1_1_ ] = [Z; e2_2_ ] = 0, up to a gauge transformation, where Z was de ned
in (5.36). The BPS con gurations are naturaly expressed in terms of (E.5) since Q
H
D
? = D
cos( )
ir
1
e1_ 1_ +
e2_2_ ;
D?;' = D ;' :
Using the de nitions (E.5) and (E.6), one can show that
ab_ = 0 implies that
F ? = F'? = D?Dab = D? e1_1_ = D? e2_2_ = D
? 1_2_ = 0 ;
where F ? = i[D? ; D?]. As implied by (E.7), the modi ed connection (E.6) actually satis es
D
It then follows that all elds in V are similarly independent.
= 0, and so is literally independent of
up to a gauge transformation.
The analysis of the H0 hypermultiplet BPS equations
; a_ =
; ea_ = 0, is similar.
One rst solves for the auxiliary elds:
Ga =
Gea =
1 (ad) h( da_
1 (ad) h( da_
ba_ )D q
b
( da_ bc_) c_a_ qb + ( da_ 0ba_ )qb ;
ba_ )D qeb + ( da_ bc_)qeb c_a_ + ( da_ 0ba_ )qeb ;
i
where we de ned
( ac_ bc_). The remaining equations then imply that
D?Ga = D?Gea = D?qa = D?qea = 0 :
Note that the solutions (E.8) and (E.9) for the auxiliary elds depend on the spinors aa_ .
Nevertheless, the conditions these spinors satisfy (5.34) can be shown to imply that (E.10)
holds for any choice of aa_ .
The solutions (E.2), (E.8), (E.9) for the auxiliary
elds in terms of the dynamical
ones, together with the independence conditions (E.7) and (E.10), comprise the full set
of restrictions that follow from the BPS equations (E.1) without imposing additional reality
conditions on the elds. These conditions are su cient in order to show that the action
Sh0yper[H0] de ned in (5.31) localizes to the 1d action (5.43). Indeed, after dimensional
reduction on , plugging (E.2), (E.8) and (E.9) in Sh0yper[H0], one can show that
(E.6)
(E.8)
(E.10)
Sh0yper[H0]
Q
H BPS
Z
D2
d2xpgD2 r K ;
cb_ )qebD qc +
a_ b_ ( ba_
0cb_ )qebqc +
ac( ca_
(E.11)
_
bb)qa a_ b_ qb ;
(E.12)
where
K
r cos( )
completes the derivation.
imaginary.
where
runs over the coordinates
and ' of D2. Using the explicit form of (E.12), one
can check that the boundary term left from (E.11) is precisely the 1d action (5.43). This
37Note that in our conventions ( 1_1_)y =
2_2_, so the connection (E.6) is complex unless
is pure
Open Access.
This article is distributed under the terms of the Creative Commons
Attribution License (CCBY 4.0), which permits any use, distribution and reproduction in
any medium, provided the original author(s) and source are credited.
[1] S. Lee, S. Minwalla, M. Rangamani and N. Seiberg, Three point functions of chiral operators
in D = 4, N = 4 SYM at large N , Adv. Theor. Math. Phys. 2 (1998) 697 [hepth/9806074]
[2] E. D'Hoker, D.Z. Freedman, S.D. Mathur, A. Matusis and L. Rastelli, Extremal correlators
in the AdS/CFT correspondence, hepth/9908160 [INSPIRE].
[3] M. Baggio, J. de Boer and K. Papadodimas, A nonrenormalization theorem for chiral
primary 3point functions, JHEP 07 (2012) 137 [arXiv:1203.1036] [INSPIRE].
[4] C. Beem, M. Lemos, P. Liendo, W. Peelaers, L. Rastelli and B.C. van Rees, In nite Chiral
Symmetry in Four Dimensions, Commun. Math. Phys. 336 (2015) 1359 [arXiv:1312.5344]
017 [arXiv:1404.1079] [INSPIRE].
[5] C. Beem, L. Rastelli and B.C. van Rees, W symmetry in six dimensions, JHEP 05 (2015)
[6] C. Beem, W. Peelaers, L. Rastelli and B.C. van Rees, Chiral algebras of class S, JHEP 05
[7] S.M. Chester, J. Lee, S.S. Pufu and R. Yacoby, Exact Correlators of BPS Operators from the
3d Superconformal Bootstrap, JHEP 03 (2015) 130 [arXiv:1412.0334] [INSPIRE].
[8] C. Beem, W. Peelaers and L. Rastelli, Deformation quantization and superconformal
symmetry in three dimensions, Commun. Math. Phys. 354 (2017) 345 [arXiv:1601.05378]
[9] Y. Tachikawa, A brief review of the 2d=4d correspondences, J. Phys. A 50 (2017) 443012
[10] T.T. Dumitrescu, An introduction to supersymmetric eld theories in curved space, J. Phys.
A 50 (2017) 443005 [arXiv:1608.02957] [INSPIRE].
[11] D.R. Morrison, GromovWitten invariants and localization, J. Phys. A 50 (2017) 443004
[arXiv:1608.02956] [INSPIRE].
443016 [arXiv:1608.02968] [INSPIRE].
[12] S. Pasquetti, Holomorphic blocks and the 5d AGT correspondence, J. Phys. A 50 (2017)
[13] S. Kim and K. Lee, Indices for 6 dimensional superconformal eld theories, J. Phys. A 50
[14] V. Pestun and M. Zabzine, Introduction to localization in quantum eld theory, J. Phys. A
(2017) 443017 [arXiv:1608.02969] [INSPIRE].
50 (2017) 443001 [arXiv:1608.02953] [INSPIRE].
[arXiv:1608.02963] [INSPIRE].
443007 [arXiv:1608.02959] [INSPIRE].
[15] K. Zarembo, Localization and AdS/CFT Correspondence, J. Phys. A 50 (2017) 443011
[16] M. Marin~o, Localization at large N in ChernSimonsmatter theories, J. Phys. A 50 (2017)
(2017) 443013 [arXiv:1608.02965] [INSPIRE].
[20] T. Dimofte, Perturbative and nonperturbative aspects of complex ChernSimons theory, J.
Phys. A 50 (2017) 443009 [arXiv:1608.02961] [INSPIRE].
[21] J. Qiu and M. Zabzine, Review of localization for 5d supersymmetric gauge theories, J. Phys.
A 50 (2017) 443014 [arXiv:1608.02966] [INSPIRE].
[23] F. Benini and B. Le Floch, Supersymmetric localization in two dimensions, J. Phys. A 50
[25] L. Rastelli and S.S. Razamat, The supersymmetric index in four dimensions, J. Phys. A 50
[26] C. Closset, T.T. Dumitrescu, G. Festuccia, Z. Komargodski and N. Seiberg, Contact Terms,
Unitarity and FMaximization in ThreeDimensional Superconformal Theories, JHEP 10
(2012) 053 [arXiv:1205.4142] [INSPIRE].
[27] C. Closset, T.T. Dumitrescu, G. Festuccia and Z. Komargodski, Supersymmetric Field
Theories on ThreeManifolds, JHEP 05 (2013) 017 [arXiv:1212.3388] [INSPIRE].
[28] E. Gerchkovitz, J. Gomis, N. Ishtiaque, A. Karasik, Z. Komargodski and S.S. Pufu,
Correlation Functions of Coulomb Branch Operators, JHEP 01 (2017) 103
[arXiv:1602.05971] [INSPIRE].
[29] K. Papadodimas, Topological AntiTopological Fusion in FourDimensional Superconformal
Field Theories, JHEP 08 (2010) 118 [arXiv:0910.4963] [INSPIRE].
[30] M. Baggio, V. Niarchos and K. Papadodimas, Exact correlation functions in SU(2) N = 2
superconformal QCD, Phys. Rev. Lett. 113 (2014) 251601 [arXiv:1409.4217] [INSPIRE].
[31] M. Baggio, V. Niarchos and K. Papadodimas, tt equations, localization and exact chiral
rings in 4d N = 2 SCFTs, JHEP 02 (2015) 122 [arXiv:1409.4212] [INSPIRE].
[32] M. Baggio, V. Niarchos and K. Papadodimas, On exact correlation functions in SU(N )
N = 2 superconformal QCD, JHEP 11 (2015) 198 [arXiv:1508.03077] [INSPIRE].
[33] N. Seiberg and E. Witten, Gauge dynamics and compacti cation to threedimensions, in The
mathematical beauty of physics: A memorial volume for Claude Itzykson. Proceedings,
Conference, Saclay, France, June 57, 1996, pp. 333{366 (1996) [hepth/9607163] [INSPIRE].
[34] K.A. Intriligator and N. Seiberg, Mirror symmetry in threedimensional gauge theories, Phys.
Lett. B 387 (1996) 513 [hepth/9607207] [INSPIRE].
[35] J. de Boer, K. Hori, H. Ooguri and Y. Oz, Mirror symmetry in threedimensional gauge
theories, quivers and Dbranes, Nucl. Phys. B 493 (1997) 101 [hepth/9611063] [INSPIRE].
[36] A. Hanany and E. Witten, Type IIB superstrings, BPS monopoles and threedimensional
[38] D. Gaiotto and E. Witten, Sduality of Boundary Conditions In N = 4 Super YangMills
Theory, Adv. Theor. Math. Phys. 13 (2009) 721 [arXiv:0807.3720] [INSPIRE].
[39] A. Kapustin, B. Willett and I. Yaakov, Exact Results for Wilson Loops in Superconformal
ChernSimons Theories with Matter, JHEP 03 (2010) 089 [arXiv:0909.4559] [INSPIRE].
[40] M. Bullimore, T. Dimofte and D. Gaiotto, The Coulomb Branch of 3d N = 4 Theories,
Commun. Math. Phys. 354 (2017) 671 [arXiv:1503.04817] [INSPIRE].
[41] D.L. Ja eris, The Exact Superconformal RSymmetry Extremizes Z, JHEP 05 (2012) 159
[arXiv:1012.3210] [INSPIRE].
03 (2011) 127 [arXiv:1012.3512] [INSPIRE].
[42] N. Hama, K. Hosomichi and S. Lee, Notes on SUSY Gauge Theories on ThreeSphere, JHEP
[43] V. Pestun, Localization of gauge theory on a foursphere and supersymmetric Wilson loops,
Commun. Math. Phys. 313 (2012) 71 [arXiv:0712.2824] [INSPIRE].
[44] V. Pestun, Localization of the fourdimensional N = 4 SYM to a twosphere and 1=8 BPS
Wilson loops, JHEP 12 (2012) 067 [arXiv:0906.0638] [INSPIRE].
[45] N.A. Nekrasov, SeibergWitten prepotential from instanton counting, Adv. Theor. Math.
Phys. 7 (2003) 831 [hepth/0206161] [INSPIRE].
[46] A. Lossev, N. Nekrasov and S.L. Shatashvili, Testing SeibergWitten solution, in proceedings
of Strings, branes and dualities, NATO Advanced Study Institute, Cargese, France, May
26{June 14, 1997, pp. 359{372 [hepth/9801061] [INSPIRE].
[47] N.A. Nekrasov and S.L. Shatashvili, Quantization of Integrable Systems and Four
Dimensional Gauge Theories, in Proceedings, 16th International Congress on Mathematical
Physics (ICMP09), Prague, Czech Republic, August 3{8, 2009, pp. 265{289
[48] N. Berkovits, A Tendimensional superYangMills action with o shell supersymmetry, Phys.
Lett. B 318 (1993) 104 [hepth/9308128] [INSPIRE].
[49] B. Assel and J. Gomis, Mirror Symmetry And Loop Operators, JHEP 11 (2015) 055
[arXiv:1506.01718] [INSPIRE].
[50] O. Aharony, A. Hanany, K.A. Intriligator, N. Seiberg and M.J. Strassler, Aspects of N = 2
supersymmetric gauge theories in threedimensions, Nucl. Phys. B 499 (1997) 67
[hepth/9703110] [INSPIRE].
[51] S. Giombi and V. Pestun, Correlators of local operators and 1=8 BPS Wilson loops on S2
from 2d YM and matrix models, JHEP 10 (2010) 033 [arXiv:0906.1572] [INSPIRE].
[52] S. Giombi and V. Pestun, Correlators of Wilson Loops and Local Operators from
MultiMatrix Models and Strings in AdS, JHEP 01 (2013) 101 [arXiv:1207.7083] [INSPIRE].
[53] N. Drukker and J. Plefka, Superprotected npoint correlation functions of local operators in
N = 4 super YangMills, JHEP 04 (2009) 052 [arXiv:0901.3653] [INSPIRE].
anomaly in quantum mechanics, JHEP 10 (2010) 054 [arXiv:1007.5292] [INSPIRE].
formula, Commun. Math. Phys. 212 (2000) 591 [math/9902090] [INSPIRE].
structure constants, JHEP 05 (2014) 103 [arXiv:1401.7977] [INSPIRE].
Dualities, JHEP 10 (2010) 013 [arXiv:1003.5694] [INSPIRE].
Counting and T (U(N )), JHEP 12 (2011) 077 [arXiv:1105.2817] [INSPIRE].
theories, JHEP 04 (1999) 021 [hepth/9902033] [INSPIRE].
[22] V. Pestun , Review of localization in geometry , J. Phys. A 50 ( 2017 ) 443002 [24] S.S. Pufu , The FTheorem and FMaximization , J. Phys. A 50 ( 2017 ) 443008 [17] B. Willett , Localization on threedimensional manifolds , J. Phys. A 50 ( 2017 ) 443006 [18] J.A. Minahan , Matrix models for 5d super YangMills , J. Phys. A 50 ( 2017 ) 443015 [19] K. Hosomichi , N = 2 SUSY gauge theories on S4 , J. Phys. A 50 ( 2017 ) 443010 gauge dynamics , Nucl. Phys. B 492 ( 1997 ) 152 [ hep th/9611230] [INSPIRE].
[37] J. de Boer , K. Hori , H. Ooguri , Y. Oz and Z. Yin , Mirror symmetry in threedimensional theories , SL (2; Z ) and Dbrane moduli spaces , Nucl. Phys. B 493 ( 1997 ) 148