The rôle of Coulomb branches in 2D gauge theory

Constantin Teleman∗

January 31, 2018

arXiv:1801.10124v1 [math.AG] 30 Jan 2018

Abstract
I give a simple construction of certain Coulomb branches C3,4 (G; E) of gauge theory in 3 and
4 dimensions defined by Nakajima et al. [N, BFN] for a compact Lie group G and a polarisable
quaternionic representation E. The manifolds C (G; 0) are abelian group schemes (over the bases of
regular adjoint GC -orbits, respectively conjugacy classes), and C (G; E) is glued together from two
copies of C (G; 0) shifted by a rational Lagrangian section εV , the Euler class of the index bundle of a
polarisation V of E. Extending the interpretation of C3 (G; 0) as “classifying space” for topological
2D gauge theories, I characterise functions on C3 (G; E) as operators on the equivariant quantum
cohomologies of M × V , for all compact symplectic G-manifolds M . The non-commutative version
has a similar description in terms of the Γ-function of V , appearing to play the role of Fourier
transformed J-function of the gauged linear Sigma-model V /G.

1. Introduction
Associated to a compact connected Lie group G and a quaternionic representation E, there are ex-
pected to be “Coulomb branches C3,4 (G; E) of N = 4 SUSY gauge theory in dimensions 3 and 4, with
matter fields in the representation E.” They ought to be components of the moduli space of vacua,
representing solutions of the monopole equations with singularities. Following early physics leads
[SW, CH] and more recent calculations [CHMZ], these spaces were given a precise construction in the
series of papers [N, BFN] by Nakajima and collaborators in the case when E is polarisable (isomorphic
to V ⊕ V ∨ for some complex V ). The case of abelian G was handled independently by Bullimore,
Dimofte and Gaitto [BDG] from a physics perspective, while the case of the zero representation had
been developed in [BFM], although only later recognised as such [T1, T2].
The C3,4 are expected to be hyperkähler (insofar as this makes sense for singular spaces), with C3
carrying an SU(2) hyperkähler rotation. They are constructed in [BFN] as algebraic Poisson spaces,
with C× -action in the case of C3 . We shall rediscover them as such in a simpler construction, which
also illuminates their relevance to 2-dimensional gauge theory. We will build the C3,4 for polarised E
from their more basic versions for the zero representation E = 0; those are always smooth. I should
mention that the main reconstruction results, Theorems 1, 2 and 4, are more elementary than their 2D
gauge theory interpretation, but it is the latter which seems to give them meaning. In compromise, I
have attempted to isolate gauge theory comments into paragraphs whose omission does not harm the
remaining mathematics. Similarly, I have separated the story of non-commutative deformations into
the final section, as that has not yet been cogently meshed with quantum cohomology.

Acknowledgements. The author has learned about Coulomb branches from A. Braverman, T. Di-
mofte and H. Nakajima. The work was partially supported by NSF grant DMS–1406056 and by
All Souls College, Oxford, and was first presented at a 2016 Clay workshop in Oxford.
∗
UC Department of Mathematics, 970 Evans Hall, Berkeley, CA 94720. teleman@berkeley.edu

1
2. Overview and example
The complex-algebraic symplectic manifold C3 (G; 0) was introduced for general G in [BFM]; for
G = SUn , it had been studied in [AH], as the space of SU2 monopoles of charge n. The description
most relevant for us is Spec H∗G (ΩG; C), the conjugation-equivariant homology of the based loop group
ΩG, with its Pontryagin product. From here, its rôle as a classifying space for topological 2-dimensional
gauge theories was developed in [T1, T2], where the space was denoted BFM(G∨ ). As we now recall,
this virtue of C3 (G; 0) must be read in the sense of semiclassical symplectic calculus, and not in the
sense of a spectral theorem à la Gelfand-Naimark.
(2.1) Relation with quantum cohomology. A partial summary of the classifying property of C3 (G; 0)
is that its regular functions act on the equivariant quantum cohomologies QHG ∗ (M ) of compact G-

Hamiltonian symplectic manifolds M , and that the E2 structure on QHG ∗ (M ) (and on its co-chain

version) is compatible with the E3 structure on C3 defined by the symplectic form. This lays out
QHG ∗ (M ) as a sheaf over C , with Lagrangian support. (The Lagrangian property is an equivariant
3
generalisation of the flatness of the pencil of quantum connections.) This generalises Seidel’s construc-
tion [S] of shift operators on QH ∗ and their equivariant extension, and is the mirror description of the
gauged A-model in 2 dimensions. A similar narrative applies to C4 (G; 0) and equivariant quantum K-
theory (minding the orbifold nature of C4 , see §3), even though the general framework for K-theoretic
mirror symmetry is incompletely understood.
(2.2) Coulomb branches with matter. The universal property of the C (G; 0) leaves the new spaces
C (G; E) in search of a rôle. Their new characterisation will address this riddle, which is partially
rooted in an omission in [T2]. Namely, the Seidel shift operators act on QH ∗ (M ) only when M is
compact; for more general spaces, the most we expect is an action on the symplectic cohomology, when
the latter is defined [R]. Equivariant symplectic cohomology SHG ∗ (X) is sometimes a localisation of

QHG ∗ (X), in which case the space C (G; 0) will capture a dense open part of QH ∗ (X), with portions
3 G
lost at infinity. Notably, this is the case when X is a linear G-space V , or a product M × V with M
compact. The lost part of QHG ∗ (M × V ) can be captured in a second chart of C (G; 0), shifted from

the original by the effect of the functor M 7→ M × V .

This shift is implemented as follows. There is a symmetric monoidal structure on 2-dimensional
TQFTs with G-gauge symmetry, the tensor product of theories (with diagonal G-action). This is
mirrored in the classifying space BFM(G∨ ) by multiplication along an abelian group structure over
the ground ring H∗G (point) (whose spectrum is isomorphic to the space greg C /GC of regular adjoint
∗
orbits). The sheaf QHG (V ) is (the structure sheaf of) a certain rational Lagrangian section εV of this
group scheme,1 and the operation QHG ∗ (M ) 7→ QH ∗ (M × V ) is represented by multiplication by ε .
G V
The statement of the main theorems requires preparation, and is postponed to the end of §3; the
remainder of the introduction develops the following
(2.3) Key example: G = U1 , with the standard representation V = C. We have

C3 (U1 ; 0) = Spec H∗U1 (ΩU1 ; C) = C × C× ∼

= T ∨ C× , (2.4)

with C× dual to U1 : the coordinates τ and z on the two factors generate H 2 (BU1 ) and π1 U1 . The
natural symplectic form dτ ∧ dz/z has an intrinsic topological definition (cf. §3.1 and §5.3 below). One
commonly defines the “toric mirror” of the space V as the function (“super-potential”) ψ(z) = z on
the space C× . The differential dψ defines the Lagrangian εV := {τ = z} ⊂ T ∨ C× . View εV instead
as the rational section τ 7→ z = τ of the projection T ∨ C → C, and note in passing the Legendre
transform ψ ∗ (τ ) = τ (log τ − 1) of ψ (in the sense that εV = exp dψ ∗ ).
Functions on εV are identified with C[τ ± ], the U1 -equivariant symplectic cohomology of V , rather
than its quantum cohomology C[τ ]. However, gluing onto (2.4) a second copy T ∨ C× with coordinates
1
This is the gauged B-model mirror of V : see Remark 3.13.

2
τ and z ′ = z/τ leads to the space C2 \ {0}, with coordinates (x, y) = (z, 1/z ′ ), living over the line
τ = xy. The section εV is now the line y = 1, identified with the full τ -axis by projection.
In [BDG, BFN], C3 (U1 ; V ⊕ V ∨ ) is taken to be the affine completion C2 = Spec C[x, y]. The
following characterisation is obvious:

2.5 Proposition. C[x, y] is the subring of regular functions f (τ, z) on T ∨ C× with the property that
f (τ, zτ ) is also regular.

Our Lagrangian z = τ is the “Euler class of the index bundle” in the following sense. Denote
by Pic(P1 ) the moduli stack of holomorphic line bundles on P1 ; its equivariant homotopy type is the
stack BU1 × ΩU1 implied in (2.4). Over P1 × Pic(P1 ) lives the universal line bundle, associated to the
standard representation V . Its index along P1 (with vanishing constraint
at a single marked point) is
a virtual bundle IndV over Pic(P1 ), with Euler class eV ∈ H ∗ Pic(P1 ) [τ −1 ]. Specifically, eV = τ n
on the component n ∈ Z = π1 U1 of Pic. The following is clear.

2.6 Proposition. Multiplication by εV on T ∨ C× , z 7→ τ z, corresponds to the cap-product with eV on

H∗U1 (ΩU(1); C) [τ −1 ].

These propositions also capture the rôle of C (G; V ⊕ V ∨ ) in quantum cohomology: the Euler class
condition picks out precisely those generalised Seidel operators which act on QHU∗ (V ).
1

2.7 Proposition. The subring C[x, y] ⊂ C[τ, z ± ] acts on QHU ∗ (M × V ) for any compact U -
1 1
Hamiltonian symplectic manifold M , and it is the largest subring with that property.

Proof. The subring C[τ ] ∼

= H∗U1 (point) acts in the natural way. Recall now that the Seidel element σn
n
going with z is a “twisted 1-point function”, namely the element in QHU ∗ (X) defined by evaluation
1
at ∞ of sections of the X-bundle over P1 associated to O(−n). All is well when X is compact: σn
is a unit with inverse σ−n . For X
= M × V though, equivariant integration along the fibres of the
evaluation H 0 P1 ; O(−n) ⊗C× V → V∞ can acquire an Euler class denominator (a factor of τ n for
n < 0, in our case). The extra factor τ in y = τ z −1 precisely cancels this denominator.

(2.8) Final detail. Propositions 2.5–2.7 generalise to all G and V , as Theorems 1 and 3 below; Theo-
rem 2 is the K-theory analogue. I shall also describe non-commutative versions in §5. The one change
needed everywhere is the inclusion, in the ground ring, of an additional equivariant parameter µ, from
the natural C× -scaling of V . There is one such parameter2 for each irreducible G-component of V ,
but the overall scaling suffices for the theorems. The need for such scaling is easily seen if we replace
the standard representation V , in the example above, by V ⊕ V ∨ : there is a false cancellation in the
Euler n
 class, eV ⊕V ∨ = eV ∪ eV ∨ = (−1) on the nth component. With the scaling, we get instead
n
µ+τ
µ−τ . The latter is the homogenised total Chern class of the index bundle, rather than the Euler
class, and is more naturally defined for virtual bundles, such as the index bundle over G ⋉ ΩG.

3. Definitions and Main results

Let us recall the construction and properties of the Coulomb branches. Up to the introduction of the
Euler Lagrangians in §3.11, this is mostly a condensation of material from [BFM, BF, BFN]. I will
write C3,4 for C3,4 (G; 0) when no confusion arises. Denote by H ⊂ G a maximal torus and by H ∨ , G∨
the Langlands dual groups.
2
These are the complex components of the mass parameters in physics, [BDG].

3
(3.1) The basic Coulomb branches [BFM]. The space C3 := Spec H∗G (ΩG; C) is an affine symplectic
resolution of singularities of the Weyl quotient T ∨ HC∨ /W . It arises by adjoining to T ∨ HC∨ the functions
∨
(eα − 1)/α for all root-coroot pairs α, α∨ of G. The C× -action, from the homology grading, scales
the symplectic form. The underlying Poisson structure is the leading term of a non-commutative
deformation over C[h] = H ∗ (BR), obtained by incorporating equivariance under the loop-rotation
circle R in the said homology. The rotation is revealed by writing ΩG ∼ = LG/G. (It helps to consider
the two-sided groupoid G ⋉ LG ⋊ G, with a Hecke-style product: see Remarks 3.2 and 3.9 below.)
For simply connected G, the spectrum of K∗G (ΩG; C) is also a symplectic manifold, resolving
∨
(HC × HC∨ )/W by adjoining the functions (eα − 1)/(eα − 1) before Weyl division. However, this
space has singularities when π1 G has torsion. Write G = G̃/π for the torsion subgroup π ⊂ π1 G,
H = H̃/π. Being a subgroup of Z(G̃), π acts by automorphisms on KG̃ (X) ⊗ C of any G-space
X, multiplying each π-eigenbundle by the corresponding character. We adopt the smooth symplectic
orbifold π ⋉ Spec K∗G̃ (ΩG; C) as the correct definition of C4 .
3.2 Remark (Sphere topology). Some properties of C3,4 are explained by Chas-Sullivan theory: the
underlying topological object is the mapping space from S 2 to the target stack BG, and has a natural
E3 structure. This is the Poisson form. Enforcing the loop rotation symmetry cuts this down to E1 ,
because rotating 2-spheres may be strung together linearly as beads, but no longer moved around each
other in the ambient 3-dimensional space.
(3.3) Group scheme structure. The Hopf algebra structures of H∗G (ΩG), K∗G (ΩG) over the ground
∗ , K of a point lead to relative abelian group structures
rings HG G

χ κ
C3 (G; 0) −−→ hC /W, C4 (G; 0) −−→ π ⋉ (H̃C /W ); (3.4)

when π1 G has torsion, the second base is an affine orbifold whose ring of functions is KG (point).
These maps define integrable systems: χ is a partial completion of the classical Toda system [BF];3 κ
is its finite-difference version.
3.5 Remark (Adjoint and Whittaker descriptions). As an algebraic symplectic manifold, C3 is the
∨ G∨ //G∨ of the regular (in the fibre) cotangent bundle under con-
algebraic symplectic reduction Treg C C
jugation. There is a simliar description of C4 using the Langlands dual Kac-Moody group (not the
loop group of G∨ ), capturing the holomorphic (but not the right algebraic) symplectic structure.
The space C3 has another description as the two-sided symplectic reduction of T ∨ G∨
C by N , at the
regular nilpotent character. This is easily seen to be algebraic symplectic; much less obviously, it is
hyper-Kähler, thanks to work of Bielawski [B]. The non-commutative deformation has a corresponding
description in terms of N × N monodromic differential operators on G∨ C [BF].

(3.6) Coulomb branches for E = V ⊕V ∨ . To build the spaces C3,4 (G; E), we follow [BFN], to which we
refer for full details, and replace ΩG in the original C by a linear space LV → Ωa G, a stratified space
whose fibres are vector bundles over the Schubert cells of the algebraic model Ωa G := GC ((z))/GC [[z]]
of ΩG. The fibre of LV over a Laurent loop γ ∈ Ωa G is the kernel of the difference
Id−γ
LV |γ → V [[z]] ⊕ V [[z]] −−−→ V ((z)). (3.7)

Projection embeds LV in either factor V [[z]] with finite co-dimension, bounded on any finite union of
cells in Ωa G. More precisely, the complex (3.7) descends to G[[z]]\Ωa G, with the left and right copies
of G[[z]] acting on the respective factors V [[z]], and with the left one alone acting on V ((z)). Over any
finite union of cells, LV contains two sub-bundles of finite co-dimension, coming from a left and a right
z n V [[z]], for sufficiently large n. This cellular finiteness lets one define the Borel-Moore (K-)homologies
BM H∗G (LV ), BM K∗G (LV ), renormalising the grading as if the dimension of V [[z]] were zero.
3
This was rediscovered in [T2], with a simple proof; I thank H. Nakajima for pointing me to the original reference.

4
 The normalised grading is compatible with the multiplication defined by the following correspon-
dence diagram on the fibres of LV , which lives over the multiplication of two loops γ, δ ∈ Ωa G:
LV |γ ⊕ LV |δ ֋ LV |γ ⊕ LV |δ ֌ LV |γ·δ ; (3.8)
V [[z]]

the sum in the middle is fibered over the right component of LV |γ and the left one of LV |δ , while
the right embedding is the projection to the outer V [[z]] summands. The wrong-way map in homology
along the first inclusion is well-defined, over γ, δ in a finite range of Schubert cells, after modding out
by a common subspace z n V [[z]], and the result is independent of n. The operation has degree zero,
after our renormalisation.
As before, a non-commutative deformation of these spaces arises by including the loop rotation
action by the circle group R on ΩG and on V [[z]]; its leading term is a Poisson structure.
3.9 Remark (E3 Hecke property). A Laurent loop defines a transition function for a principal GC -
bundle over the non-separated disk −:− with doubled origin. There is a Hecke interpretation of our
multiplications from correspondences on G ⋉ ΩG and LV , all of them induced, left-to-right, by the
maps relating non-separated disks with doubled and tripled centres:
(−:−) .
։ (−..−) ֋ (−:−)
(−:−)
The left map identifies the bottom sheet of the first with the top sheet of the second; the right inclusion
hits the top and bottom centres of the triple-centred disk. The E3 property follows by sliding the
multiple centres around, as in Chas-Sullivan sphere topology. Once we insist on equivariance under
loop rotation, the freedom to slide disappears and we are reduced to an E1 multiplication; this is the
non-commutative deformation of the Coulomb branches.
On G-bundles, the Hecke-induced operation is represented by multiplication of transition functions,
once we identify, on the left side, the top bundle on its bottom sheet with the bottom bundle on its
top sheet. Next, associated to the representation V is a vector bundle over −:−, and LV is its space
of global sections. The correspondence (3.8) arises by retaining those pairs of global sections on the
left which match on the said pair of sheets, and then restricting them to the top and bottom sheets
of the triple-centred disk.
(3.10) Scaling. Introduce the enhanced Coulomb branches in which S 1 ⊂ C× scales the fibres of LV ,
1
C3◦ (G; E) := Spec BM H∗G×S (LV ; C), projecting to hC /W × C,
G̃×S 1
C4◦ (G; E) := π ⋉ Spec BM K∗ (LV ; C), projecting to π ⋉ (H̃C /W ) × C× ;
the projections to the enhanced Toda bases are defined as in (3.4), and denoted by χ(µ), κ(m), with
generators µ ∈ H 2 (BS 1 ), m± ∈ KS 1 (point). The scaling action is trivial when E = 0 and LV = Ωa G,
but it will appear in the Euler class of the index bundle over G ⋉ ΩG.
The notation is subtly abusive: the C ◦ depend on the polarisation V and not just on E. For
instance, switching V ↔ V ∨ leads an isomorphic space only if we also change the orientation of the
rotating circle. This V -dependence disappears at µ = 0 or m = 1. We will see in §4 that the C ◦ (G; E)
are flat over C[µ], C[m± ], so the C3,4 (G; E) do appear by specialising to µ = 0 and m = 1.
(3.11) The Euler Lagrangians. Consider the graphs of the following rational maps from hC × C and
HC × C× to HC∨ , defined in terms of the weights ν of V , included with their multiplicities:

ν
(µ + hν|ξi)ν ,
Y Y
εV : (ξ, µ) 7→ λV : (x, m) 7→ 1 − (mxν )−1 . (3.12)
ν ν

(For w ∈ C× , wν
:= exp(ν log w) is a point in HC∨ .)
The graphs are regular away from a co-dimension 2
locus over their domains and define Lagrangians εV ⊂ C3◦ (G; 0) and λV ⊂ C4◦ (G; 0), over their
respective ground rings C[µ], C[m± ], as the closure of their regular parts (cf. Remark 3.13 below).
Their Euler class origin is spelt out in the next section.

5
3.13 Remark. For generic µ and m (most meaningfully, near µ, m = ∞), the maps (3.12) are the
exponentiated differentials (in ξ ∈ gC and x ∈ GC ) of the functions

ξ 7→ TrV [(ξ + µ) · (log(ξ + µ) − 1)] , x 7→ TrV Li2 ((mx)−1 ).

The first function appeared in the abelian GLSM as the “Σ log Σ Landau-Ginzburg model” for the
gauged B-model mirror of V , see [W] and cf. also Remark 5.9. The Lagrangian λV and associated
primitive appeared, for the adjoint representation, in the index formula for Kähler differentials over
the moduli of G-bundles on curves [TW, Eqn. 6.2 and Thm. 6.4]. The relation with Coulomb branches
was not known at the time; today, we would express that index formula in terms of Lagrangian calculus
in C4◦ (G, 0), namely the intersection of λV with graphs of isogenies HC → HC∨ . (Those correspond to
the Theta line bundles, and are the discrete Toda twists of the unit section of C4 .)
(3.14) Main Theorems. Multiplication by the sections εV , respectively λV gives a rational symplecto-
morphism of C3◦ , C4◦ , relative to the Toda projections of §3.10.

Theorem 1. The space C3◦ (G; E) → hC /W × C is the affinisation of two copies of C3◦ (G; 0) glued
together by means of εV -multiplication.

The automorphism is regular in co-dimension one, so affinisation is well-defined: regular functions

on C ◦ (G; E) are those regular functions on C ◦ (G; 0) whose composition with εV -translation remains
regular. For Abelian G, Theorem 1 is evident from the presentations in [N, BDG]. The result for C4
is completely analogous:

Theorem 2. The orbifold C4◦ (G; E) → π ⋉ (H̃C /W ) × C× is the relative affinisation of two copies of
C4◦ (G; 0) glued together by means of λV -multiplication.

The proofs, in the next section, are straightforward. Non-commutative generalisation will be given in
§5. The next theorem, characterising the rings of regular functions C[C ◦ (G; E)], follows now with the
same proof as Proposition 2.7.

Theorem 3. C[C3◦ (G; E)] comprises those regular functions on C3◦ which act on the equivariant
quantum cohomologies QHG×S ∗
1 (M × V ), for compact Hamiltonian G-manifolds M .

C[C4 (G; E)] comprises those regular functions on C4◦ which act on the equivariant quantum K-theories
◦
∗
QKG×S 1 (M × V ), for compact Hamiltonian G-manifolds M .

4. Proof of Theorems 1 and 2

The proof uses the Schubert stratification of Ωa G. Even-dimensionality collapses the associated spec-
tral sequences and leads to ascending filtrations on the rings C[C ◦ (G; E)]. The associated graded
components are easily described (§4.7 below), and are locally free over the Toda bases. This makes
the original rings locally free as well; in particular, they are flat over C[µ], C[m± ].
I write out the proof for C3◦ ; the K-theory case is entirely parallel. Call RV the ring, implied
in Theorem 1, of regular functions on C3◦ which remain regular after εV -translation. We will see
from topology how this last operation is compatible with the Schubert filtration, so that we can also
define the subring GV ⊂ Gr C[C3◦ ] of symbols which remain regular after εV -translation. Clearly,
GrRV ⊂ GV . The theorem will follow from two observations:

(i) C[C3◦ (G; E)] ⊂ RV ;

(ii) Gr C[C3◦ (G; E)] = GV .

6
(4.1) The index bundle. Over the stack BunG (P1 ) of principal GC -bundles over P1 , there lives the
virtual index bundle IndV (of the sheaf of sections of V over P1 , with simple vanishing condition at
one marked point), an element in KG×S 0
1 (ΩG). We include the equivariance under scaling of V , and
call eV its equivariant Euler class. It is the µ-homogenised total G-equivariant Chern class of IndV .
The following two propositions must be understood after suitable localisation on the base hC /WC × C.
1
4.2 Proposition. Translation by εV on C3 corresponds to cap-product with eV on H∗G×S (ΩG).
Proof. Localise on the base hC /W × C to the complement of the walls to reduce, by the fixed-point
theorem, to the case of a torus, where Proposition 2.6 applies.
4.3 Corollary. The Schubert filtration is preserved by εV -translation.
4.4 Remark. Cap-product with eV must correspond to translation by some rational section: the index
bundle is additive for the sphere multiplication in G ⋉ ΩG, so its Euler class is multiplicative, and
therefore represents a (rational) section of the group scheme C ◦ over its Toda base.
(4.5) Embedding C[C3◦ (G; E)] twice. Referring to the constructions in §3.6 and Remark 3.9, the stack
G[[z]]\Ωa G of GC -bundles over the double-centered disk −:− maps to BunG (P1 ), by performing the
Hecke construction at the marked point. (This map is an equivariant homotopy equivalence, if we
only retain the maximal compact subgroups of the stabilisers.) We now observe that, when restricted
to G[[z]]\Ωa G, IndV is the “left minus right” copy of V [[z]].
More precisely, note the two inclusions ιl,r : LV ֒→ V [[z]], and recall that over any finite union of
cells, LV contains a finite co-dimension sub-bundle. Quotienting it out regularises the difference of
V [[z]]-bundles into a class in KG×R (ΩG). A moment’s thought identifies this with IndV , as the index
of the Hecke transform of the trivial V -bundle on P1 , minus that of the trivial V -bundle.
Each inclusion ιl,r defines a graded ring homomorphism ϕl,r : C[C3◦ (G; E)] ֌ C[C3◦ ], by intersecting
with the zero-section in the ambient bundle. Per our discussion, ϕl = eV ∩ ϕr . Using ϕr to pin down
C3◦ (G; E), Proposition 4.2 now settles Observation (i).
(4.6) Working out GV . For a 1-parameter subgroup z η ∈ ΩH, with Schubert cell Cη and Levi cen-
traliser Z(η) ⊂ G, split V = V+ ⊕ V0 ⊕ V− following the sign of the η-eigenvalue. The index bundle IV
splits then as I+ ⊖ I− , with the ν-weight space ofQV± appearing ±hν|ηi times in I± . The Euler class
eV factors correspondingly as e+ · e−1
− , with e± = ν (µ + ν)
|hν|ηi| , including the scaling weight µ. The

η-graded component of GV is then the submodule

Z(η)×S 1 Z(η)×S 1 1
e− ∩ H∗ (point) ⊂ H∗ (point) ∼
= BM H∗G×S (Cη ) = Grη C[C ◦ (G; 0)].
(4.7) Working out Gr C[C3◦ (G; E)]. Collapse of the filtration spectral sequence implies that
1
Grη C[C3◦ (G; E)] = BM H∗G×S ( LV |η ).
Now, the the homology group is generated by the fundametal class of the total space of LV over Cη ,
and its complement in the right V [[z]] of (3.7) is precisely I− ; therefore
Grη C[C3◦ (G; E)] = e− ∩ Grη C[C ◦ (G; 0)],
in agreement with (GV )η as determined above. This settles Observation (ii).

5. Non-commutative deformation
Incorporating the loop rotation circle R leads to non-commutative deformations N C3,4 ◦ (G; E) of the
∗ ±
Coulomb branches over the ground rings C[h] = H (BR) and C[q ] = KR (point), respectively. The
geometric objects exist in the formal neighbourhoods of h = 0 and q = 1; away, only the function
rings survive, which we denote by A3,4 , but keeping the conversational pretence of an underlying space
N C is convenient. The arguments in §4 for their description will apply mutatis mutandis: we are
missing the statements, which we summarise below before spelling out the calculation.

7
(5.1) Summary. The integrable abelian group structure of the C ◦ over their Toda bases deforms to
a symmetric tensor structure on A -modules4 induced from the diagonal inclusion ΩG ֌ ΩG × ΩG.
On underlying modules over the Toda base, this is the ordinary tensor product, and the tensor unit
O1 becomes the structure sheaf. For C3◦ in the Whittaker presentation (Remark 3.5), the operation
comes from convolution of D-modules on the Langlands dual group G∨ .
The Lagrangians εV , λV deform to modules EV , ΛV over A3,4 , and the (rational) automorphisms
of C defined by εV , λV -translation become the functors of convolution with EV , ΛV on A -modules.
The Hamiltonian nature of the translations renders these functors (genercially) trivialisable by inner
automorphisms of A . We describe the Coulomb branches N C ◦ (G; E) as the subrings of elements of
A which survive these inner automorphisms (remain regular thereunder; Theorem 4).
While our description of the N C ◦ (G; E) is uniform, a distinction arises between formal and
genuine deformations. Formally, the modules EV and ΛV are generically invertible, being analogous
to flat line bundles with regular singular connections. However, explicit holomorphic solutions for the
(inverse) modules are easy to find, once the deformation parameters are truly turned on. (Solutions
are morphisms to O1 .) For C3 , the solution is the reciprocal Γ-function of V ; a q-analogue solves C4 .
These solutions define the inner automorphisms just mentioned, and are our primary objects: away
from the formal limit, the modules EV , ΛV are defined from the solutions.
In the holomorphic case we can say more. The solutions make the modules EV , ΛV invertible (and
trivialisable). The inverse module EV−1 is presented as a quotient5 of A3 by the annihilator IV of the
solution Γ−1 ,
∼
EV−1 := A3 /IV −−−→ O1 = A3 /I1 .
Γ−1
∪
O1 ⊂ A3
We can then also characterise N C ◦ (G; E) as the subring of elements A3 which preserves the inclusion
O1 ⊂ EV−1 , identified above with Γ−1 O1 ⊂ O1 . There is a parallel story for N C4 .
5.2 Remark. In the case of C3 for a torus, the relevant module EV is the Laplace transform of the
quantum D-module for the G-gauged quantum cohomology of V ; the (inverse) Γ-function could be
regarded as the Fourier transform for the small J-function (Remark 5.9). One can formulate a rather
precise conjecture here, but I do not know a general proof.
Before spelling out the full story, let us review the
(5.3) Key example: U1 with its standard representation. The symplectic space T ∨ C× = Spec H∗U1 (ΩU1 )
has a natural non-commutative deformation, realised topologically by the Pontryagin ring H∗U1 ×R (ΩU1 ).
Indeed, on π1 U1 , z is the shift n 7→ n + 1, at which point the R-rotation collects an extra weight of
U1 . We compute from here the Pontryagin ring as C[h]hz ± , τ i with relation zτ = (τ + h)z. We now
identify the non-commutative Coulomb branch H∗U1 ×R (LV ) for the standard representation V :

5.4 Lemma. N C3 (U1 ; V ⊕ V ∨ ) is the subring of H∗U1 ×R (ΩU1 ) generated over C[h] by z, z −1 τ .

5.5 Remark. If X = z, Y = z −1 τ , the ring is C[h]hX, Y i/([X, Y ] − h), as could also be inferred from
the Poisson relation {x, y} = h in C[x, y].

Proof. Using the definitions in §3 and the right inclusion in §4.5 to embed the ring, we find at the
winding mode n ≥ 0 the generator z n over C[h, τ ]; whereas at a negative winding mode (−n), we find

z −n e− = z −n τ (τ + h) · · · · · (τ + (n − 1)h) = (z −1 τ )n ,

from the Euler class e− of I− , which is the summand missing from the right copy of V [[z]].
4
I thank D. Ben-Zvi for pointing out the generality of this statement.
5
The quotient should be understood as a sheaf over the Toda base, because of the infinitely many poles of Γ.

8
 Recall the h-periodic Gamma-function Γ(w; h) := hw/h−1 Γ(w/h), satisfying Γ(w+h; h) = wΓ(w; h).
The relation zτ z −1 = τ + h gives

Γ(τ ; h) · z · Γ(τ ; h)−1 = τ −1 z. (5.6)

5.7 Corollary. Generically in τ , Γ(τ ; h) conjugates the module N C3 /(z − 1) into N C3 /(z − τ ).

Recall that the solution space of a D-module M is RHomD (M , O). Extrapolating this language here
and using for O the tensor unit module N C3 /(z − 1) exhibits the reciprocal function Γ(τ ; h)−1 as
a solution to the module N C3 /(z −1 − τ ), the quantisation of the inverse of εV . Holomorphy of the
solution is the reason to prefer the inverse module to the original.

5.8 Proposition. N C3 (U1 ; V ⊕ V ∨ ) is the subring of elements of H∗U1 ×R (ΩU1 ) which remain regular
under conjugation by Γ(τ ; h)−1 .

Proof. Lemma 5.4 gives the left-to-right inclusion. The other way, we specialise to a negative z-degree
(−n) in H∗U1 ×R (ΩU1 ). Reordering factors expresses a ring element uniquely in monomials of the form

(z −1 τ )n τ m , m ≥ 0, and (z −1 τ )a z a−n , 0 ≤ a < n.

The former are in N C3 . The latter Γ−1 -conjugate to z −a (τ z)a−n , and these are not regular in any
C[h]-linear combination, or else a right multiplication by (τ z)n would lead to a dependence among the
monomials
z −a (τ z)a = (τ − h) · · · · · (τ − ah), 0≤a<n
−n m n m
z τ (τ z) = (τ − nh) · (τ − h) · · · · (τ − nh), m≥0

which is pre-empted by reason of τ -degree. So no monomials of the second type can appear in a ring
element that survives conjugation.

5.9 Remark. As h → 0 with arg(τ /h) bounded away from ±π/2, Stirling’s approximation gives
τ 3 1
log Γ(τ ; h) = (log τ − 1) − log h + log(2πτ ) + o(1),
h 2 2
and we spot in the leading h−1 term the Legendre transform ψ ∗ (τ ) of ψ(z) = z. The Legendre
correspondence quantises to the Laplace transform: viewing N C3 as the ring of Dh -modules on C× ,
∂
with τ = h · z ∂z , we find that the function exp(−z/h) on C× is the solution to the module Dh /(τ + z),
the Laplace transform of the one in Proposition 5.7. Notice that exp(−z/h) is the quantum J-function
associated to the gauged linear sigma-model V /G.
(5.10) The space N C3◦ . For the general version of Proposition
Q 5.8, we promote Γ to a multiplicative
characteristic class of complex vector bundles: Γ(F ; h) := ρ Γ(ρ; h), over the Chern roots ρ of F .
A priori, this is ill-defined, because Γ has a pole at 0. The reciprocal 1/Γ is entire holomorphic, but
its vanishing at 0 leads to an unstable class. It can be regularised by including our equivariant mass
parameter µ, scaling F ; the resulting meromorphic calculus suffices for us.
First, we note, in the formal neighbourhood of h = 0:

5.11 Proposition. Γ(V ; h) conjugates the unit module of N C3◦ into a module EV with support εV .

Proof. This follows from the abelian calculation in the example, by localising to a dense open set in
the in the Toda base.

9
As before, we prefer the module EV−1 (inverse with respect to convolution), which has the entire
holomorphic solution
Γ(V ; h)−1 : hC /W × C → C, (ξ, µ) 7→ detV Γ−1 (ξ ⊕ µ; h);
EV−1 is the quotient of A3 by the annihilator of Γ(V ; h)−1 , and maps to O1 by sending the unit in A3
to the solution Γ(V ; h)−1 .
1 ×R
Theorem 4. The non-commutative deformation N C3◦ (G; E), defined by H∗G×S (LV ), comprises
1
those elements of H∗G×S ×R (ΩG) which survive conjugation by Γ(V ; h)−1 .
Proof. With respect to the embeddings ϕl,r in §4.5 (now with the R-action included), I claim that
conjugation switches ϕr to ϕl . This embeds N C3 (G; E) into the surviving subring described. The
moral reason is that Γ(V ; h) is a regularised reciprocal Euler class of V [[z]]: conjugation therefore
undoes the intersection of LV with the zero-section in the right bundle and intersects in the left
bundle instead.
To verify the claim, we exploit the collapse of the Schubert spectral sequence and the absence
of zero-divisors, and confine ourselves to a single Schubert cell Cη , where we identify the symbols
Grη N C3◦ (G; E) with e+ · Grη N C3◦ in the left embedding and with Grη N C3◦ · e− in the right one.
Formula (5.6) generalises easily to
Γ(V ; h)−1 · ([Cη ]e− ) · Γ(V ; h) = e+ [Cη ],
equating the conjugation by Γ−1 with the operation of capping6 with the equivariant index class.
The proof is completed by noting that Grη N C3◦ (G; E) exhausts the part of Grη N C3◦ which
survives conjugation, by elaborating the argument used in Proposition 5.8. (Alternatively, we can
work from the commutative case, showing surjectivity on increasing powers of h, and exploiting the
polynomial nature of our ring elements to ensure the termination of this process, since deg h = 2.)

(5.12) The space N C4◦ . In the Key Example of U1 , the Pontryagin ring K∗U1 ×R (ΩU1 ) is the standard
non-commutative (complexified) torus, C[q ± ]ht± , z ± i with relation zt = qtz. To proceed, we need
Jackson’s p-Gamma function [J]: in terms of p-Pochhammer symbols (x; p)∞ = n≥0 (1 − xpn ),
Q
convergent for |p| < 1, that function is
(ph ; ph )∞ 1 − pw
Γp (w; h) = (1 − p)1−w/h , satisfying Γp (w + h; h) = Γp (w; h).
(pw ; ph )∞ 1−p
The requisite version of Γp arises in the limit p, h → 0 while the expansion variables q = ph and
pw remain finite. Expressed for us in the group element t = p−τ , Γ0 (t) := (q; q)∞ /(t−1 ; q)∞ . The
conjugation replacing (5.6) is
Γ0 (t)−1 · z · Γ0 (t) = (1 − t−1 )z;
in particular, Γ0 (t) conjugates the unit module z = 1 to the λV -supported module z = (1 − t−1 ).
5.13 Remark. In full analogy with Remark 5.9, the Laplace
  transform of our solution Γ0 is expressed
in terms of the q-exponential function eq , namely eq 1−q = (z; q)−1
z
∞.
From here, define the multiplicative class Γ0 for vector bundles, valued in localised equivariant
K-theory, as the product for their line bundle (virtual) summands; the following is then clear.
5.14 Proposition. Γ0 (V ) conjugates the unit module of N C4◦ into a module ΛV with support λV .
Finally, the argument used for N C3 applies, after working locally on the Toda base, to give
1
Theorem 5. The non-commutative deformation N C4◦ (G; V ⊕ V ∨ ), defined by K∗G×S ×R (LV ), com-
1
prises those elements of K∗G×S ×R (ΩG) which remain regular after conjugation by Γ0 (V )−1 .
6 ∗
Caution: left and right multiplication by classes in HG×R differ at z η , because the left and right pull-backs from
B(G × R) do. Of course, the cap-product operation is canonical.

10
References
[AH] M.F. Atiyah, N. Hitchin: The geometry and dynamics of magnetic monopoles. M.B. Porter
Lectures. Princeton, 1988

[B] R. Bielawski: Hyperkähler structures and group actions. J. London Math. Soc. (2) 55 (1997),
400–414

[BDG] M. Bullimore, T. Dimofte, D. Gaiotto: The Coulomb branch of 3D N = 4 theories.

Comm. Math. Phys. 354 (2017), 671–751

[BF] R. Bezrukavnikov, M. Finkelberg: Equivariant Satake category and Kostant-Whittaker reduc-

tion. Mosc. Math. J. 8 (2008), 39–72

[BFM] R. Bezrukavnikov, M. Finkelberg, I. Mirkovic: Equivariant homology and K-theory of affine

Grassmannians and Toda lattices. Compos. Math. 141 (2005), 746–768

[BFN] A. Braverman, M. Finkelberg, H. Nakajima: Towards a mathematical definitoin of Coulomb

branches of 3-dimensional N = 4 gauge theories II. https://arxiv.org/abs/1601.03586

[CH] B. Chalmers, A. Hanany: Three-dimensional gauge theories and monopoles. Nuclear Phys. B
489 (1997), 223–244

[CHMZ] S. Cremonesi, A. Hanany, M. Noppadol, A. Zaffaroni: Coulomb branch Hilbert series and
Hall-Littlewood polynomials. J.H̃igh Energy Phys. 2014, 178–

[J] The Rev. F.H. Jackson: The basic Gamma-function and the Elliptic functions. Proc. Royal
Soc. London, Series A, 76 (1905), 127–144

[N] H. Nakajima: Towards a mathematical definitoin of Coulomb branches of 3-dimensional N = 4

gauge theories I. Adv. Theor. Math. Phys. 20 (2016), 595–669

[R] A. Ritter: Floer theory for negative line bundles via Gromov-Witten invariants. Adv. Math.
262 (2014), 1035–1106

[SW] N. Seiberg, E. Witten: Monopoles, duality and chiral symmetry breaking in N = 2 supersym-
metric QCD. Nuclear Phys. B 431 (1994), 484–550

[S] P. Seidel: π1 of symplectic automorphism groups and invertibles in quantum homology rings.
Geom. Funct. Anal. 7 (1997), 1046–1095

[T1] C. Teleman: Langlands duality and Mirror Symmetry. Perspectives in Geometry Lecture Series,
UT Austin, November 2011.
https://www.ma.utexas.edu/rtgs/geomtop/rtg/perspectives.html

[T2] C. Teleman: Gauge theory and Mirror Symmetry. Proc. Internat. Congress Math., Seoul 2014

[TW] C. Teleman and C.T. Woodward: The index formula on the moduli of G-bundles. Ann. Math.
170 (2009), 495–527

[W] E. Witten: Dynamics of Quantum Field Theory, Lecture 13. In: Quantum fields and strings:
a course for mathematicians, Vol. 2, Amer. Math. Soc., Providence, RI, (1999)

11