New worlds for Lie Theory

math.columbia.edu/~okounkov/icm.pdf
Lie groups, continuous symmetries, etc. are among the main building
blocks of mathematics and mathematical physics

Since its birth, Lie theory has been constantly expanding its scope
and its range of applications.
Simple finite-dimensional Lie groups have been classified by the
1890s. Their elegant structure and representation theory in many
ways shaped the development of mathematical physics in the XX
century
In return, very concrete questions prompted many fruitful directions of
research in Lie theory
In return, very concrete questions prompted many fruitful directions of
research in Lie theory

The Virasoro Lie algebra, Affine Lie algebras (a special case of Kac-
Moody Lie algebras), and their relatives, play a key role in the study
of 2-dimensional critical phenomena described by the Conformal
Field Theories
In return, very concrete questions prompted many fruitful directions of
research in Lie theory

The Virasoro Lie algebra, Affine Lie algebras (a special case of Kac-
Moody Lie algebras), and their relatives, play a key role in the study
of 2-dimensional critical phenomena described by the Conformal
Field Theories

Their quantum group analogs underlie integrable lattice

discretizations of CFT
Today, I want to talk about a more recent set of ideas that links
mathematical physics with Lie theory in a new, much expanded sense. It
originates in the study of supersymmetric QFT, in particular, susy gauge
theories in <4 (especially 3) space-time dimensions
Today, I want to talk about a more recent set of ideas that links
mathematical physics with Lie theory in a new, much expanded sense. It
originates in the study of supersymmetric QFT, in particular, susy gauge
theories in <4 (especially 3) space-time dimensions

I want to share with you my excitement

about a subject that is still forming. We don’t
see yet its true logical boundaries and our
definitions, technical foundations, etc. are
improving in real time. I will try to stick to
what we know for certain and not try to be
too visionary.
What I understand about the subject owes a great deal to Nikita
Nekrasov and Samson Shatashvili, as well as to Mina Aganagic, Roman
Bezrukavnikov, Hiraku Nakajima, Davesh Maulik, and many others
What I understand about the subject owes a great deal to Nikita
Nekrasov and Samson Shatashvili, as well as to Mina Aganagic, Roman
Bezrukavnikov, Hiraku Nakajima, Davesh Maulik, and many others

One of the guiding stars in the subject has been a certain powerful
duality that generalizes Langlands duality to this more general setting.
It goes back to Intrilligator and Seiberg, and has been studied by many
teams of researchers, in particular, by Davide Gaiotto, Hiraku
Nakajima, Ben Webster, and their collaborators
It may be easier to explain what is new by explaining which
highlights of the late XX century Lie theory are being generalized

Before, it may be helpful to remind ourselves what is a Weyl group,

Hecke algebra, etc. as generalizations of these objects will be essential
in what follows
Finite and discrete reflection groups W of a
Euclidean space Rn appear in Lie theory as
finite and affine Weyl groups and play a central
role in classification and representation theory
Finite and discrete reflection groups W of a
Euclidean space Rn appear in Lie theory as
finite and affine Weyl groups and play a central
role in classification and representation theory

To every W one can associate a braid group

and a Hecke algebra, in which the generators

satisfy a generalization of S2=1
Back to

... It may be easier to explain what is new by explaining which

highlights of the late XX century Lie theory are being generalized
Macdonald-Cherednik theory

Irreducible Lie group characters and, more generally, spherical

functions, are eigenfunctions of invariant differential operators, that
is, solutions to certain linear differential equations. In MC theory,
these are generalized to certain q-difference equations associated to
root systems and involving additional parameters. Solutions of these
equations are remarkable multivariate generalizations of q-
hypergeometric functions, whose terminating cases are known as the
Macdonald polynomials.

Numerous applications of those in combinatorics, number theory,

probability theory, algebraic geometry etc. have been found.
Macdonald-Cherednik theory

The algebraic backbone of the theory is a certain double version of

the affine Hecke algebra constructed by Cherednik. A fundamental
symmetry of this doubling yields an amazing label-argument
symmetry in Macdonald polynomials, of which

is a kindergarten example. It plays a key role in applications and is

a preview of the general duality statements.
Kazhdan-Lusztig theory

in its simplest form, describes the characters of irreducible highest

weight modules over a Lie algebra in terms of the combinatorics of the
associated finite Hecke algebra. The proof of the original KL conjectures
by Beilinson-Bernstein and Brylinski-Kashiwara is, perhaps, one of
highest achievements in all of Lie theory, with further contributions
by Ginzburg, Soergel, Bezrukavnikov, Williamson, and many, many
others. In characteristic p ≫ 0, there is a version with affine Hecke
algebra.

the talk by Geordie Williamson

 Yang-Baxter equation and quantum groups

In vertex models of 2D statistical mechanics, the degrees of freedom live

in vector spaces Vi attached to edges of a grid and their interaction is
described by a matrix R of weights attached to each vertex
Yang-Baxter equation and quantum groups

Baxter noted the importance of the YB equation

for exact solvability, with further important insights by the Faddeev

and Jimbo-Miwa-Kashiwara schools. This gives rise to the whole theory
of quantum groups (Drinfeld, … ), associated knot invariants, et cetera,
et cetera
Note by Reshetikhin et al the quantum group may be reconstructed
from matrix elements of the R-matrix, or as the algebra behind the
braided tensor category constructed from R.

Particularly important are R-matrices with a

spectral parameter that correspond to quantum
loop groups. By Baxter, these contain large
commutative subalgebras that become quantum
integrals of motion in vertex models and
associated quantum spin chains

Many brilliant minds worked on diagonalization of these algebras, a

problem known as the “Bethe Ansatz”
 new variable

More generally, R-matrices with a

spectral parameter define an action of an
affine Weyl group of type A by q-
difference operators, the lattice part of
which are the quantum Knizhnik-
Zamoldchikov equations of Frenkel and
Reshetikhin. These are among the most
important linear equations in
mathematical physics; solving them
generalizes the Bethe Ansatz problem
 For knot theory and other topological applications, limits of R(u) are important

YB equation becomes a Reidemeister move

 Example:

associated to T*P1, classical

hypergeometry, etc. Self-dual!
modern formulas are more suitable downloads than slides
A 3-dimensional supersymmetric
Quantum Field Theory is a lot of data,
of which we will be using only a very
small piece - the susy states in the
Hilbert space associated to a given time
slice, a Riemann surface B with,
maybe, boundary and marked points.
Even narrower, we will focus on the

Index = Even fermion number - Odd

as a virtual representation of all symmetries and as a virtual vector

bundle over the moduli of B, as in the talk by Rahul Pandharipande.
Even narrower, we will focus on the

Index = Even fermion number - Odd

as a virtual representation of all symmetries and as a virtual vector

bundle over the moduli of B, as in the talk by Rahul Pandharipande.

This (Witten) index being deformation invariant, it can be studied

using any of the different description of the QFT in various corners of
its parameter space.
At lowest energies (that is, for very large B), the states of
a QFT may be described as modulated vacuum, that is,
a map f from B to the moduli space X of vacua of the
theory. The amount of supersymmetry that we want
makes X, ideally, a hyperkähler manifold and f a
holomorphic map. Finer details of the theory will
become important at the singularities of X or f which
are, in general, unavoidable.
Mathematically, this becomes a problem in the spirit of enumerative
geometry. Susy states are holomorphic maps f from B to X, which is a
symplectic algebraic variety, or stack, or … The index is the Euler
characteristic of a certain coherent sheaf (a virtual Â-genus, like for
the index of a Dirac operator) on the moduli space of such. This index
is graded by the action of Aut(X). The additional grading on this
index by the degree of the map may be viewed as a character of the
Kahler torus
Mathematically, this becomes a problem in the spirit of enumerative
geometry. Susy states are holomorphic maps f from B to X, which is a
symplectic algebraic variety, or stack, or … The index is the Euler
characteristic of a certain coherent sheaf (a virtual Â-genus, like for
the index of a Dirac operator) on the moduli space of such. This index
is graded by the action of Aut(X). The additional grading on this
index by the degree of the map may be viewed as a character of the
Kahler torus

To make the index nontrivial, we require that the symplectic form

ωX is scaled by Aut(X) with a nontrivial weight ħ
For example, susy gauge theories contain gauge fields for a compact form
of a Lie group G, matter fields in a symplectic representation M of G, and
their superpartners. In this case

and moduli spaces in question are “stable quasimaps” f: B -> X. There

are generalizations with critical loci of functions etc. If
For example, susy gauge theories contain gauge fields for a compact form
of a Lie group G, matter fields in a symplectic representation M of G, and
their superpartners. In this case

and moduli spaces in question are “stable quasimaps” f: B -> X. There

are generalizations with critical loci of functions etc. If

many

then X is a Nakajima quiver variety. The

quiver is the generalization of the Dynkin
diagram from before
 For instance, for this quiver this theory is:

(1) the K-theoretic Donaldson-Thomas theory of Y3=rank 2 bundle

over B, which together with its sister theories eventually determines
the K-theoretic DT counts in all threefolds (not just CY). These capture
deeper information than the cohomological DT and Gromov-Witten
counts

(2, conjecturally) the theory on the worldsheet of the M2 brane of M-

theory
The physical diversity of operator insertions and boundary conditions
translates into different flavors of evaluation maps from such moduli
spaces to X, or … As function of B, these define a K-theoretic analog of
CohFT with a state space K(X).

Further enriched by the data of an arbitrary Aut(X)-bundle over B.

Of paramount importance are the vertex functions, that is, counts for
B= complex plane, with boundary conditions imposed at infinity (this
is formalized as maps from P1 nonsingular at ∞). Like for Nekrasov
counts of instantons on R4, these make sense equivariantly for the
action of
Of paramount importance are the vertex functions, that is, counts for
B= complex plane, with boundary conditions imposed at infinity (this
is formalized as maps from P1 nonsingular at ∞). Like for Nekrasov
counts of instantons on R4, these make sense equivariantly for the
action of

A fundamental feature of the theory are linear q-difference equations

in all variables, Kahler Z or equivariant T ⊂ Aut(X), satisfied by the
vertex functions. The operators in these equations are certain counts for
B=P1 with insertions at both 0 and ∞
Main point:

These q-difference equations generalize what we have seen before,

or if one prefers abstract statements to special functions, then

The whole enumerative theory may be described using certain

new geometric representation theory
 Main point:

These q-difference equations generalize what we have seen before,

or if one prefers abstract statements to special functions, then

The whole enumerative theory may be described using certain

new geometric representation theory

In general, it will involve algebras that are not Hopf, but for
Nakajima quiver varieties we get new quantum loops groups and their
entire package
[Maulik-O,12] gave a geometric construction of solutions of the YB and
related equations using their theory of stable envelopes. This associates
a new quantum loop group Uħĝ to any quiver so that K(X) is a weight
space in a Uħĝ-module. The corresponding Lie algebra g is a
generalization of the Kac-Moody Lie algebra constructed geometrically
by Nakajima
[Maulik-O,12] gave a geometric construction of solutions of the YB and
related equations using their theory of stable envelopes. This associates
a new quantum loop group Uħĝ to any quiver so that K(X) is a weight
space in a Uħĝ-module. The corresponding Lie algebra g is a
generalization of the Kac-Moody Lie algebra constructed geometrically
by Nakajima

[O., 15] q-difference equations in certain equivariant variables are the

qKZ equations for Uħĝ
[Maulik-O,12] gave a geometric construction of solutions of the YB and
related equations using their theory of stable envelopes. This associates
a new quantum loop group Uħĝ to any quiver so that K(X) is a weight
space in a Uħĝ-module. The corresponding Lie algebra g is a
generalization of the Kac-Moody Lie algebra constructed geometrically
by Nakajima

[O., 15] q-difference equations in certain equivariant variables are the

qKZ equations for Uħĝ

[Smirnov-O., 16] the affine dynamical groupoid of Uħĝ gives the q-

difference equations in the Kahler variables
[Maulik-O,12] gave a geometric construction of solutions of the YB and
related equations using their theory of stable envelopes. This associates
a new quantum loop group Uħĝ to any quiver so that K(X) is a weight
space in a Uħĝ-module. The corresponding Lie algebra g is a
generalization of the Kac-Moody Lie algebra constructed geometrically
by Nakajima

[O., 15] q-difference equations in certain equivariant variables are the

qKZ equations for Uħĝ

[Smirnov-O., 16] the affine dynamical groupoid of Uħĝ gives the q-

difference equations in the Kahler variables

Last 2 statements generalize what was proven in cohomology in [MO]

a dynamical groupoid is a collection of
operators of the form

for every wall w of a periodic hyperplane

arrangement. Must satisfy
a dynamical groupoid is a collection of
operators of the form

for every wall w of a periodic hyperplane

arrangement. Must satisfy
 a dynamical groupoid is a collection of
operators of the form

for every wall w of a periodic hyperplane

arrangement. Must satisfy

Generalize YB equation, braid groups,

and give flat q-difference connections.
Constructed for every Uħĝ in [OS]
 a dynamical groupoid is a collection of
operators of the form

for every wall w of a periodic hyperplane

arrangement. Must satisfy

Generalize YB equation, braid groups,

and give flat q-difference connections.
Constructed for every Uħĝ in [OS] Bezrukavnikov
This gives a complete control over quantities of prime enumerative
interest, such as those from K-theoretic DT theory of threefolds.
Conversely, with geometric tools one can really control the analytic
theory of these equations
This gives a complete control over quantities of prime enumerative
interest, such as those from K-theoretic DT theory of threefolds.
Conversely, with geometric tools one can really control the analytic
theory of these equations

[Aganagic-O., 16] Monodromy of these q-difference equations computed

in terms of elliptic R-matrices (<- elliptic cohomology generalization of
stable envelopes). This generalizes the Kohno- Drinfeld computation of
the monodromy of KZ equations, among other things
This gives a complete control over quantities of prime enumerative
interest, such as those from K-theoretic DT theory of threefolds.
Conversely, with geometric tools one can really control the analytic
theory of these equations

[Aganagic-O., 16] Monodromy of these q-difference equations computed

in terms of elliptic R-matrices (<- elliptic cohomology generalization of
stable envelopes). This generalizes the Kohno- Drinfeld computation of
the monodromy of KZ equations, among other things

[Aganagic-O., 17] Integral representation of solutions which, in

particular, solves the corresponding generalization of the Bethe Ansatz
problem in the q->1 limit
Beyond Nakajima varieties
q-diff equations exist for abstract reasons
q-diff equations exist for abstract reasons

The Kähler ones should be given by a similar groupoid,

now with roots of X (a finite subset of effective curve
classes) instead of the roots of g.
q-diff equations exist for abstract reasons

The Kähler ones should be given by a similar groupoid,

now with roots of X (a finite subset of effective curve
classes) instead of the roots of g.

This is a work in progress [Halpern-Leistner+Maulik+O] for rather general

gauge theories. In other cases, one is still looking for the right moduli spaces.
q-diff equations exist for abstract reasons

The Kähler ones should be given by a similar groupoid,

now with roots of X (a finite subset of effective curve
classes) instead of the roots of g.

This is a work in progress [Halpern-Leistner+Maulik+O] for rather general

gauge theories. In other cases, one is still looking for the right moduli spaces.

Same for shifts in the equivariant variables in A ⊂ Aut(X,ω). The roots

here are the normal weights of A at XA and the wall operators come from
stable envelopes
q-diff equations exist for abstract reasons

The Kähler ones should be given by a similar groupoid,

now with roots of X (a finite subset of effective curve
classes) instead of the roots of g.

This is a work in progress [Halpern-Leistner+Maulik+O] for rather general

gauge theories. In other cases, one is still looking for the right moduli spaces.

Same for shifts in the equivariant variables in A ⊂ Aut(X,ω). The roots

here are the normal weights of A at XA and the wall operators come from
stable envelopes

The duality should exchange the Kahler torus Z and A, and the two
groupoids. In some limited generality, this is indeed shown in a work
in progress with M. Aganagic
Unlike classical Langlands duality, the Kähler and equivariant roots live in
spaces of apriori different dimension, making the duality more dramatic

Kähler roots for Xv

Equivariant roots for Xv

Kähler roots for X

Equivariant roots for X

Where is the Kazhdan-Lusztig theory ?

The braid group limit Sw(0), Sw(∞) of the Kähler groupoid gives the
right analog of the Hecke algebra for quantizations of X over a field of
characteristic p ≫ 0 as shown by [Bezrukavnikov-O] for a list of
theories that includes all Nakajima varieties.
 Where is the Kazhdan-Lusztig theory ?

The braid group limit Sw(0), Sw(∞) of the Kähler groupoid gives the
right analog of the Hecke algebra for quantizations of X over a field of
characteristic p ≫ 0 as shown by [Bezrukavnikov-O] for a list of
theories that includes all Nakajima varieties.

Better still than such limit, one should study the full elliptic theory of
[Aganagic-O]. It controls the roots of unity analogs of characteristic p ≫ 0
quantization questions for finite p. It categorifies to equivalences between
different descriptions of the category of boundary conditions in the QFT,
which is where the different roads of categorification in Lie theory
should converge.
Any formulas today ?
 Any formulas today ?

I haven’t put up any beyond the X=T*P1 example, but these new
worlds are full of e.g. remarkable q-diff equations whose solutions
contain a treasure of geometric, representation-theoretic,
combinatorial, and no doubt number-theoretic information.

Explicit formulas for e.g. stable envelopes [Smirnov], Bethe eigen-

functions, etc. contain, as a special case, answers to many old questions.
back to Earth
back to Earth
Добро пожаловать в Санкт-Петербург на ICM2022!