Download as pdf or txt
Download as pdf or txt
You are on page 1of 67

Geometric Langlands Theory

A Bridge between Number Theory and Physics

Philsang Yoo

Seoul National University

Colloquium
April 28, 2022

1/18
Plan of the Talk

1. Introduction to the Langlands program


2. Transition to Geometric Langlands Correspondence
3. Physical Mathematics and Geometric Langlands Theory
4. Outlook

2/18
Fermat’s Last Theorem

One of the most famous problems in mathematics was so-called Fermat’s


Last Theorem:
Theorem (Fermat’s Last Theorem; Wiles, Taylor–Wiles 1995)
The equation x n + y n = z n has no nontrivial rational (or integral)
solutions if n > 2.

3/18
Fermat’s Last Theorem

One of the most famous problems in mathematics was so-called Fermat’s


Last Theorem:
Theorem (Fermat’s Last Theorem; Wiles, Taylor–Wiles 1995)
The equation x n + y n = z n has no nontrivial rational (or integral)
solutions if n > 2.

One may think that the equation is so random and not of any significant
importance for mathematics. However, it turns out to be relevant to
seemingly different subjects of independent interest; in fact, the theorem
was first proved by studying fundamental objects such as elliptic curves
and modular forms.

3/18
Elliptic Curves
An elliptic curve E over Q is defined by an equation

y 2 + c1 xy + c3 y = x 3 + c2 x 2 + c4 x + c6 with ci ∈ Q.

Again we would like to find rational solutions but it is too difficult.

4/18
Elliptic Curves
An elliptic curve E over Q is defined by an equation

y 2 + c1 xy + c3 y = x 3 + c2 x 2 + c4 x + c6 with ci ∈ Q.

Again we would like to find rational solutions but it is too difficult.


Instead, for each prime p, one can find the number of solutions
(x, y ) ∈ (Z/p)2 satisfying the equation modulo p, denoted by Solp . We
record this by setting ap (E ) := p − Solp .

4/18
Elliptic Curves
An elliptic curve E over Q is defined by an equation

y 2 + c1 xy + c3 y = x 3 + c2 x 2 + c4 x + c6 with ci ∈ Q.

Again we would like to find rational solutions but it is too difficult.


Instead, for each prime p, one can find the number of solutions
(x, y ) ∈ (Z/p)2 satisfying the equation modulo p, denoted by Solp . We
record this by setting ap (E ) := p − Solp .
Example
Consider the elliptic curve E defined by y 2 + y = x 3 − x 2 . E.g., for
p = 5, we have (0, 0), (0, 4), (1, 0), and (1, 4), so Sol5 = 4.

4/18
Elliptic Curves
An elliptic curve E over Q is defined by an equation

y 2 + c1 xy + c3 y = x 3 + c2 x 2 + c4 x + c6 with ci ∈ Q.

Again we would like to find rational solutions but it is too difficult.


Instead, for each prime p, one can find the number of solutions
(x, y ) ∈ (Z/p)2 satisfying the equation modulo p, denoted by Solp . We
record this by setting ap (E ) := p − Solp .
Example
Consider the elliptic curve E defined by y 2 + y = x 3 − x 2 . E.g., for
p = 5, we have (0, 0), (0, 4), (1, 0), and (1, 4), so Sol5 = 4.

p 2 3 5 7 11 13 17 19 23 29 31 ···
Solp 4 4 4 9 - 9 19 19 24 29 24 ···
ap (E) −2 −1 1 −2 - 4 −2 0 −1 0 7 ···
The elliptic curve E has a bad reduction at 11.

4/18
Modular Forms

A modular form of weight k is a holomorphic function f : H → C, where


H = {τ ∈ C | Im(τ ) > 0} is the upper-half plane, such that
! " ! "
aτ + b ! a b
f (γ · τ ) = f = (cτ + d)k f (τ ) for γ = ∈ SL2 (Z),
cτ + d c d

satisfying a growth condition.

5/18
Modular Forms

A modular form of weight k is a holomorphic function f : H → C, where


H = {τ ∈ C | Im(τ ) > 0} is the upper-half plane, such that
! " ! "
aτ + b ! a b
f (γ · τ ) = f = (cτ + d)k f (τ ) for γ = ∈ SL2 (Z),
cτ + d c d

satisfying a growth condition. #


Since f (τ + 1) = f (τ ), one can write

q = e 2πiτ and f (τ ) = f (q) = n=−∞ bn (f )q n with bn (f ) ∈ C.

5/18
Modular Forms

A modular form of weight k is a holomorphic function f : H → C, where


H = {τ ∈ C | Im(τ ) > 0} is the upper-half plane, such that
! " ! "
aτ + b ! a b
f (γ · τ ) = f = (cτ + d)k f (τ ) for γ = ∈ SL2 (Z),
cτ + d c d

satisfying a growth condition. #Since f (τ + 1) = f (τ ), one can write



q = e 2πiτ and f (τ ) = f (q) = #n=−∞ bn (f )q n with bn (f ) ∈ C. The

growth condition says f (q) = n=0 bn (f )q n .

5/18
Modular Forms

A modular form of weight k is a holomorphic function f : H → C, where


H = {τ ∈ C | Im(τ ) > 0} is the upper-half plane, such that
! " ! "
aτ + b ! a b
f (γ · τ ) = f = (cτ + d)k f (τ ) for γ = ∈ SL2 (Z),
cτ + d c d

satisfying a growth condition. #Since f (τ + 1) = f (τ ), one can write



q = e 2πiτ and f (τ ) = f (q) = #n=−∞ bn (f )q n with bn (f ) ∈ C. The

growth condition says f (q) = n=0 bn (f )q n .

In other words, a modular form is



$
f (q) = bn (f )q n
n=0

where bn (f )’s are all related by the SL2 (Z) symmetry.

5/18
Modular Forms

A modular form of weight k is a holomorphic function f : H → C, where


H = {τ ∈ C | Im(τ ) > 0} is the upper-half plane, such that
! " ! "
aτ + b ! a b
f (γ · τ ) = f = (cτ + d)k f (τ ) for γ = ∈ SL2 (Z),
cτ + d c d

satisfying a growth condition. #Since f (τ + 1) = f (τ ), one can write



q = e 2πiτ and f (τ ) = f (q) = #n=−∞ bn (f )q n with bn (f ) ∈ C. The

growth condition says f (q) = n=0 bn (f )q n .

In other words, a modular form is



$
f (q) = bn (f )q n
n=0

where bn (f )’s are all related by the SL2 (Z) symmetry. Note that a
random sequence {bn } wouldn’t give a modular form.

5/18
Modularity Theorem

[Taniyama, Shimura 1956-1957] formulated the celebrated


Taniyama–Shimura conjecture.

Theorem (Modularity Theorem; W, TW, BCDT 2001)


Every elliptic curve defined over Q is modular: for E , there is a (unique)
modular form fE of weight 2 with ap (E ) = bp (fE ) for (almost) all prime p.

6/18
Modularity Theorem

[Taniyama, Shimura 1956-1957] formulated the celebrated


Taniyama–Shimura conjecture.

Theorem (Modularity Theorem; W, TW, BCDT 2001)


Every elliptic curve defined over Q is modular: for E , there is a (unique)
modular form fE of weight 2 with ap (E ) = bp (fE ) for (almost) all prime p.

That modularity theorem implies FLT was known earlier by the work of
Ribet which is in turn based on the work of Frey and Serre.

6/18
Modularity Theorem

[Taniyama, Shimura 1956-1957] formulated the celebrated


Taniyama–Shimura conjecture.

Theorem (Modularity Theorem; W, TW, BCDT 2001)


Every elliptic curve defined over Q is modular: for E , there is a (unique)
modular form fE of weight 2 with ap (E ) = bp (fE ) for (almost) all prime p.

That modularity theorem implies FLT was known earlier by the work of
Ribet which is in turn based on the work of Frey and Serre.

Theorem (Ribet)
If a, b, and c are positive integers such that ap + b p = c p for p ≥ 5, then
the elliptic curve defined by y 2 = x(x − ap )(x + b p ) is not modular.

6/18
Modularity Theorem

[Taniyama, Shimura 1956-1957] formulated the celebrated


Taniyama–Shimura conjecture.

Theorem (Modularity Theorem; W, TW, BCDT 2001)


Every elliptic curve defined over Q is modular: for E , there is a (unique)
modular form fE of weight 2 with ap (E ) = bp (fE ) for (almost) all prime p.

That modularity theorem implies FLT was known earlier by the work of
Ribet which is in turn based on the work of Frey and Serre.

Theorem (Ribet)
If a, b, and c are positive integers such that ap + b p = c p for p ≥ 5, then
the elliptic curve defined by y 2 = x(x − ap )(x + b p ) is not modular.

In other words, if there existed a solution to Fermat’s equation, then we


could construct an elliptic curve which cannot possibly exist!

6/18
An Explicit Example of Modularity Theorem
For the elliptic curve E defined by y 2 + y = x 3 − x 2

p 2 3 5 7 11 13 17 19 23 29 31 ···
Solp 4 4 4 9 - 9 19 19 24 29 24 ···
ap (E) −2 −1 1 −2 - 4 −2 0 −1 0 7 ···
one can find a modular form
%∞
fE (q) = q (1 − q n )2 (1 − q 11n )2 = q(1 − q)2 (1 − q 11 )2 (1 − q 2 )2 · · ·
n=1
= q − 2q 2 − q 3 + 2q 4 + q 5 + 2q 6 − 2q 7 − 2q 9 − 2q 10 + q 11
−2q 12 + 4q 13 + 4q 14 − q 15 − 4q 16 − 2q 17 + 4q 18 + 2q 20 + 2q 21
−2q 22 − q 23 + 4q 25 − 8q 26 + 5q 27 − 4q 28 + 2q 30 + 7q 31 + · · ·

That is, one has ap (E) = bp (fE ) for all primes except for p = 11. This is
not at all possible for a random sequence {ap }; we crucially needed deep
and surprising relations between elliptic curves and modular forms.

7/18
The Main Diagram of the Langlands Program
[Langlands 1967] proposed a generalization of this picture:
Algebraic Geometry
Motives❈
① ❈❈
①①① ❈❈
① ❈❈
①①① ! ❈❈
① L-functions ❈❈
①①① ❧❧" # $$$ ❈❈
① ❧ $
①① ❧❧❧❧ $$$ ❈❈


① ❧❧❧❧ ❧ $$$$ ❈❈❈
①① ❧❧❧ $$$$ ❈❈❈
$
Harmonic Analysis
Number Theory
Representation Theory
Galois Rep’n
Automorphic Rep’n

8/18
The Main Diagram of the Langlands Program
[Langlands 1967] proposed a generalization of this picture:
Algebraic Geometry
Motives❈
① ❈❈
①①① ❈❈
① ❈❈
①①① ! ❈❈
① L-functions ❈❈
①①① ❧❧" # $$$ ❈❈
① ❧ $
①① ❧❧❧❧ $$$ ❈❈


① ❧❧❧❧ ❧ $$$$ ❈❈❈
①① ❧❧❧ $$$$ ❈❈❈
$
Harmonic Analysis
Number Theory
Representation Theory
Galois Rep’n
Automorphic Rep’n

Langlands
{automorphic representations of G } % $ {Gal(F /F ) → Ǧ }
& &

"! modularity
"!
{modular forms} % $ {elliptic curves}

with a number field F , reductive group G and its dual group Ǧ ; the
example is when F = Q and G = Ǧ = GLn with n = 2. 8/18
Three Step Variants
1 [Weil’s Rosetta Stone; Analogy]
◮ number field F ↔ algebraic curve C over Fq ;
◮ (unramified) automorphic representations are Fun (BunG (C )(Fq ));
◮ (unramified) Galois representations are {π1ét (C , x) → Ǧ }/Ǧ ;

9/18
Three Step Variants
1 [Weil’s Rosetta Stone; Analogy]
◮ number field F ↔ algebraic curve C over Fq ;
◮ (unramified) automorphic representations are Fun (BunG (C )(Fq ));
◮ (unramified) Galois representations are {π1ét (C , x) → Ǧ }/Ǧ ;
◮ [Drinfeld for G = GL2 , L. Lafforgue for G = GLn ] for a Ǧ -local
system σ on C , we want a function fσ on BunG (C )(Fq ).

9/18
Three Step Variants
1 [Weil’s Rosetta Stone; Analogy]
◮ number field F ↔ algebraic curve C over Fq ;
◮ (unramified) automorphic representations are Fun (BunG (C )(Fq ));
◮ (unramified) Galois representations are {π1ét (C , x) → Ǧ }/Ǧ ;
◮ [Drinfeld for G = GL2 , L. Lafforgue for G = GLn ] for a Ǧ -local
system σ on C , we want a function fσ on BunG (C )(Fq ).

2 [Grothendieck’s Function-Sheaf Correspondence; Geometrization]


◮ An ℓ-adic (perverse) sheaf F on X over k = Fq gives a function fF
on X (k) by fF (x) = Tr(Frobx , Fx ) where Frobx acts on
Fx = x ∗ (F) for x : Spec Fq → Spec Fq → X ;

9/18
Three Step Variants
1 [Weil’s Rosetta Stone; Analogy]
◮ number field F ↔ algebraic curve C over Fq ;
◮ (unramified) automorphic representations are Fun (BunG (C )(Fq ));
◮ (unramified) Galois representations are {π1ét (C , x) → Ǧ }/Ǧ ;
◮ [Drinfeld for G = GL2 , L. Lafforgue for G = GLn ] for a Ǧ -local
system σ on C , we want a function fσ on BunG (C )(Fq ).

2 [Grothendieck’s Function-Sheaf Correspondence; Geometrization]


◮ An ℓ-adic (perverse) sheaf F on X over k = Fq gives a function fF
on X (k) by fF (x) = Tr(Frobx , Fx ) where Frobx acts on
Fx = x ∗ (F) for x : Spec Fq → Spec Fq → X ;
◮ [Deligne, Drinfeld, Laumon, Frenkel–Gaitsgory–Vilonen] for a
Ǧ -local system σ, we want a sheaf Fσ on BunG (C ).

9/18
Three Step Variants
1 [Weil’s Rosetta Stone; Analogy]
◮ number field F ↔ algebraic curve C over Fq ;
◮ (unramified) automorphic representations are Fun (BunG (C )(Fq ));
◮ (unramified) Galois representations are {π1ét (C , x) → Ǧ }/Ǧ ;
◮ [Drinfeld for G = GL2 , L. Lafforgue for G = GLn ] for a Ǧ -local
system σ on C , we want a function fσ on BunG (C )(Fq ).

2 [Grothendieck’s Function-Sheaf Correspondence; Geometrization]


◮ An ℓ-adic (perverse) sheaf F on X over k = Fq gives a function fF
on X (k) by fF (x) = Tr(Frobx , Fx ) where Frobx acts on
Fx = x ∗ (F) for x : Spec Fq → Spec Fq → X ;
◮ [Deligne, Drinfeld, Laumon, Frenkel–Gaitsgory–Vilonen] for a
Ǧ -local system σ, we want a sheaf Fσ on BunG (C ).

3 [Beilinson–Drinfeld; Categorical Statement]


◮ Furthering the analogy, consider an algebraic curve C over C or a
Riemann surface C ;
◮ in this setting, we want more than a set-theoretical statement,
encoding information for a family of such correspondences.
9/18
Categorical Formulation

Goal: find an equivalence of categories using algebraic geometry in which


a sheaf on BunG (C ) and Ǧ -local system on C naturally sit, respectively.

10/18
Categorical Formulation

Goal: find an equivalence of categories using algebraic geometry in which


a sheaf on BunG (C ) and Ǧ -local system on C naturally sit, respectively.
◮ {Perverse sheaves} ↩→ {D-modules} by the Riemann–Hilbert
correspondence;
◮ {π1ét (C , x) → Ǧ }/Ǧ ⇝ FlatǦ (C )

10/18
Categorical Formulation

Goal: find an equivalence of categories using algebraic geometry in which


a sheaf on BunG (C ) and Ǧ -local system on C naturally sit, respectively.
◮ {Perverse sheaves} ↩→ {D-modules} by the Riemann–Hilbert
correspondence;
◮ {π1ét (C , x) → Ǧ }/Ǧ ⇝ FlatǦ (C )

Meta-Conjecture (Best Hope)


There is an equivalence of categories

D-mod(BunG (C )) ≃ QCoh(FlatǦ (C ))

10/18
Categorical Formulation

Goal: find an equivalence of categories using algebraic geometry in which


a sheaf on BunG (C ) and Ǧ -local system on C naturally sit, respectively.
◮ {Perverse sheaves} ↩→ {D-modules} by the Riemann–Hilbert
correspondence;
◮ {π1ét (C , x) → Ǧ }/Ǧ ⇝ FlatǦ (C )

Meta-Conjecture (Best Hope)


There is an equivalence of categories

D-mod(BunG (C )) ≃ QCoh(FlatǦ (C ))

Remark
• category of flat bundles on space of vector bundles vs
category of vector bundles on space of flat bundles

10/18
Categorical Formulation

Goal: find an equivalence of categories using algebraic geometry in which


a sheaf on BunG (C ) and Ǧ -local system on C naturally sit, respectively.
◮ {Perverse sheaves} ↩→ {D-modules} by the Riemann–Hilbert
correspondence;
◮ {π1ét (C , x) → Ǧ }/Ǧ ⇝ FlatǦ (C )

Meta-Conjecture (Best Hope)


There is an equivalence of categories

D-mod(BunG (C )) ≃ QCoh(FlatǦ (C ))

Remark
• category of flat bundles on space of vector bundles vs
category of vector bundles on space of flat bundles
• The conjecture is known to be wrong as soon as G is non-abelian.

10/18
Why Geometric Langlands?

• An insurmountable problem in number theory might become


accessible when phrased in terms of language of the other realms.

11/18
Why Geometric Langlands?

• An insurmountable problem in number theory might become


accessible when phrased in terms of language of the other realms.
• [Ngo B. C.] proved the fundamental lemma in the original Langlands
program using tools from geometric Langlands (2010 Fields medal)

11/18
Why Geometric Langlands?

• An insurmountable problem in number theory might become


accessible when phrased in terms of language of the other realms.
• [Ngo B. C.] proved the fundamental lemma in the original Langlands
program using tools from geometric Langlands (2010 Fields medal)
• [V. Lafforgue] used geometric Langlands in a fundamental way for
his breakthrough work for the function field Langlands
correspondence in automorphic to Galois direction for general G
(2019 Breakthrough prize).

11/18
Why Geometric Langlands?

• An insurmountable problem in number theory might become


accessible when phrased in terms of language of the other realms.
• [Ngo B. C.] proved the fundamental lemma in the original Langlands
program using tools from geometric Langlands (2010 Fields medal)
• [V. Lafforgue] used geometric Langlands in a fundamental way for
his breakthrough work for the function field Langlands
correspondence in automorphic to Galois direction for general G
(2019 Breakthrough prize).
• [P. Scholze] used ingredients of geometric Langlands and ideas of
Beilinson–Drinfeld to revolutionize arithmetic geometry (2018 Fields
medal).

11/18
Why Geometric Langlands?

• An insurmountable problem in number theory might become


accessible when phrased in terms of language of the other realms.
• [Ngo B. C.] proved the fundamental lemma in the original Langlands
program using tools from geometric Langlands (2010 Fields medal)
• [V. Lafforgue] used geometric Langlands in a fundamental way for
his breakthrough work for the function field Langlands
correspondence in automorphic to Galois direction for general G
(2019 Breakthrough prize).
• [P. Scholze] used ingredients of geometric Langlands and ideas of
Beilinson–Drinfeld to revolutionize arithmetic geometry (2018 Fields
medal).
• [L. Fargues] used geometric Langlands to reshape the view of the
local Langlands correspondence.

11/18
Enumerative Geometry
The aim of the subject is to count the number of solutions satisfying a
geometric condition. The simplest is
Q. How many straight lines are there passing two given points
on a plane?
A. 1

12/18
Enumerative Geometry
The aim of the subject is to count the number of solutions satisfying a
geometric condition. The simplest is
Q. How many straight lines are there passing two given points
on a plane?
A. 1
One can make all sorts of variation. Of historical importance is
Q. How many rational curves of degree d are there on a generic
quintic threefold?
where
• a (generic) quintic threefold is
X = {[x0 , · · · , x4 ] ∈ P4 | F (x0 , · · · , x4 ) = 0} for a (generic) degree 5
homogeneous polynomial F ;
• a rational curve of degree d in P4 is [α0 , · · · , α4 ] where αi is of
degree d in coordinates [u, v ] of P1 , that is, of the form

αi (u, v ) = ai,d u d + ai,d−1 u d−1 v + · · · + ai,1 uv d−1 + ai,0 v d .

12/18
Rational Curves of Degree d on a Quintic Threefold
Finding a rational curve on X ⊂ P4 means imposing the condition
F (α0 (u, v ), · · · , α4 (u, v )) = 0; we would like to find {(ai,j )0≤i≤4, 0≤j≤d }
which solves the equation. As it is a polynomial equation of degree 5d in
u and v , it is of the form

b5d (ai,j )u 5d + b5d−1 (ai,j )u 5d−1 v · · · + b0 (ai,j )v 5d = 0.

Therefore, we obtain 5d + 1 relations on 5(d + 1) variables, i.e.,


bk (ai,j ) = 0. Quotienting out by automorphisms of P1 and rescaling, we
expect a finite number of such curves, denoted by nd .

13/18
Rational Curves of Degree d on a Quintic Threefold
Finding a rational curve on X ⊂ P4 means imposing the condition
F (α0 (u, v ), · · · , α4 (u, v )) = 0; we would like to find {(ai,j )0≤i≤4, 0≤j≤d }
which solves the equation. As it is a polynomial equation of degree 5d in
u and v , it is of the form

b5d (ai,j )u 5d + b5d−1 (ai,j )u 5d−1 v · · · + b0 (ai,j )v 5d = 0.

Therefore, we obtain 5d + 1 relations on 5(d + 1) variables, i.e.,


bk (ai,j ) = 0. Quotienting out by automorphisms of P1 and rescaling, we
expect a finite number of such curves, denoted by nd .

◮ [Schubert, 1886] n1 = 2875; [Katz, 1986] n2 = 609250; [Ellingsrud


and Strømme, 1991] n3 = 317206375;

13/18
Rational Curves of Degree d on a Quintic Threefold
Finding a rational curve on X ⊂ P4 means imposing the condition
F (α0 (u, v ), · · · , α4 (u, v )) = 0; we would like to find {(ai,j )0≤i≤4, 0≤j≤d }
which solves the equation. As it is a polynomial equation of degree 5d in
u and v , it is of the form

b5d (ai,j )u 5d + b5d−1 (ai,j )u 5d−1 v · · · + b0 (ai,j )v 5d = 0.

Therefore, we obtain 5d + 1 relations on 5(d + 1) variables, i.e.,


bk (ai,j ) = 0. Quotienting out by automorphisms of P1 and rescaling, we
expect a finite number of such curves, denoted by nd .

◮ [Schubert, 1886] n1 = 2875; [Katz, 1986] n2 = 609250; [Ellingsrud


and Strømme, 1991] n3 = 317206375;
◮ [Candelas, de la Ossa, Green, and Parkes, 1991] gave an answer for
all d at once!

13/18
Rational Curves of Degree d on a Quintic Threefold
Finding a rational curve on X ⊂ P4 means imposing the condition
F (α0 (u, v ), · · · , α4 (u, v )) = 0; we would like to find {(ai,j )0≤i≤4, 0≤j≤d }
which solves the equation. As it is a polynomial equation of degree 5d in
u and v , it is of the form

b5d (ai,j )u 5d + b5d−1 (ai,j )u 5d−1 v · · · + b0 (ai,j )v 5d = 0.

Therefore, we obtain 5d + 1 relations on 5(d + 1) variables, i.e.,


bk (ai,j ) = 0. Quotienting out by automorphisms of P1 and rescaling, we
expect a finite number of such curves, denoted by nd .

◮ [Schubert, 1886] n1 = 2875; [Katz, 1986] n2 = 609250; [Ellingsrud


and Strømme, 1991] n3 = 317206375;
◮ [Candelas, de la Ossa, Green, and Parkes, 1991] gave an answer for
all d at once!
◮ It was soon proved by independent works [Givental, Lian–Liu–Yau in
1996] in a rigorous manner.

13/18
Rational Curves of Degree d on a Quintic Threefold
Finding a rational curve on X ⊂ P4 means imposing the condition
F (α0 (u, v ), · · · , α4 (u, v )) = 0; we would like to find {(ai,j )0≤i≤4, 0≤j≤d }
which solves the equation. As it is a polynomial equation of degree 5d in
u and v , it is of the form

b5d (ai,j )u 5d + b5d−1 (ai,j )u 5d−1 v · · · + b0 (ai,j )v 5d = 0.

Therefore, we obtain 5d + 1 relations on 5(d + 1) variables, i.e.,


bk (ai,j ) = 0. Quotienting out by automorphisms of P1 and rescaling, we
expect a finite number of such curves, denoted by nd .

◮ [Schubert, 1886] n1 = 2875; [Katz, 1986] n2 = 609250; [Ellingsrud


and Strømme, 1991] n3 = 317206375;
◮ [Candelas, de la Ossa, Green, and Parkes, 1991] gave an answer for
all d at once!
◮ It was soon proved by independent works [Givental, Lian–Liu–Yau in
1996] in a rigorous manner.

Q. How is that possible???


13/18
Mirror Symmetry
Physicists had a totally different way of looking at the counting problem:

14/18
Mirror Symmetry
Physicists had a totally different way of looking at the counting problem:
1 Curve counting problem is natural “invariants” of a 2-dimensional
physical theory, say labelled by (X , A).

14/18
Mirror Symmetry
Physicists had a totally different way of looking at the counting problem:
1 Curve counting problem is natural “invariants” of a 2-dimensional
physical theory, say labelled by (X , A).
2 A physical theory (X , A) is dual to another theory (X̌ , B); meaning
that they have different descriptions but in terms of invariants they
are indistinguishable.

14/18
Mirror Symmetry
Physicists had a totally different way of looking at the counting problem:
1 Curve counting problem is natural “invariants” of a 2-dimensional
physical theory, say labelled by (X , A).
2 A physical theory (X , A) is dual to another theory (X̌ , B); meaning
that they have different descriptions but in terms of invariants they
are indistinguishable.
3 Natural “invariants” of (X̌ , B) are easier to compute.

14/18
Mirror Symmetry
Physicists had a totally different way of looking at the counting problem:
1 Curve counting problem is natural “invariants” of a 2-dimensional
physical theory, say labelled by (X , A).
2 A physical theory (X , A) is dual to another theory (X̌ , B); meaning
that they have different descriptions but in terms of invariants they
are indistinguishable.
3 Natural “invariants” of (X̌ , B) are easier to compute.
4 As the invariants should be the same, from the easier computation,
we make the conjectural answer for the curve counting problem!
This is called Mirror Symmetry.

14/18
Mirror Symmetry
Physicists had a totally different way of looking at the counting problem:
1 Curve counting problem is natural “invariants” of a 2-dimensional
physical theory, say labelled by (X , A).
2 A physical theory (X , A) is dual to another theory (X̌ , B); meaning
that they have different descriptions but in terms of invariants they
are indistinguishable.
3 Natural “invariants” of (X̌ , B) are easier to compute.
4 As the invariants should be the same, from the easier computation,
we make the conjectural answer for the curve counting problem!
This is called Mirror Symmetry.
Once a physics context is found for the problem, it not only enriched the
initial problem, but also completely reshaped algebraic geometry and
symplectic geometry ever since.
There is an even better invariant in terms of a category of “branes” or
“boundary conditions”:
Conjecture (Homological Mirror Symmetry, Kontsevich 1994)
There is an equivalence of categories Fuk(X ) ≃ Coh(X̌ ).
14/18
Kapustin–Witten “Geometric Langlands is S-duality!”
Recall the geometric Langlands correspondence

D-mod(BunG (C )) ≃ QCoh(FlatǦ (C )

15/18
Kapustin–Witten “Geometric Langlands is S-duality!”
Recall the geometric Langlands correspondence

D-mod(BunG (C )) ≃ QCoh(FlatǦ (C )

Kapustin and Witten argued at the physical level of rigor that by thinking
of a certain duality, called S-duality, one can obtain the geometric
Langlands correspondence. In other words, even without knowing
anything about number theory or various technicalities, one could have
found the statement! It sounds mind-blowing.

15/18
Kapustin–Witten “Geometric Langlands is S-duality!”
Recall the geometric Langlands correspondence

D-mod(BunG (C )) ≃ QCoh(FlatǦ (C )

Kapustin and Witten argued at the physical level of rigor that by thinking
of a certain duality, called S-duality, one can obtain the geometric
Langlands correspondence. In other words, even without knowing
anything about number theory or various technicalities, one could have
found the statement! It sounds mind-blowing.
There is a 4d theory (G , Ψ) for G and Ψ ∈ CP1 and there is a duality
(G , 0) ↔ (Ǧ , ∞).

15/18
Kapustin–Witten “Geometric Langlands is S-duality!”
Recall the geometric Langlands correspondence

D-mod(BunG (C )) ≃ QCoh(FlatǦ (C )

Kapustin and Witten argued at the physical level of rigor that by thinking
of a certain duality, called S-duality, one can obtain the geometric
Langlands correspondence. In other words, even without knowing
anything about number theory or various technicalities, one could have
found the statement! It sounds mind-blowing.
There is a 4d theory (G , Ψ) for G and Ψ ∈ CP1 and there is a duality
(G , 0) ↔ (Ǧ , ∞). Upon compactification on C , where for M = Σ × C we
declare only Σ to be a new spacetime, it becomes
(T ∗ BunG (C ), A) ≃ (LocǦ (C ), B):

15/18
Kapustin–Witten “Geometric Langlands is S-duality!”
Recall the geometric Langlands correspondence

D-mod(BunG (C )) ≃ QCoh(FlatǦ (C )

Kapustin and Witten argued at the physical level of rigor that by thinking
of a certain duality, called S-duality, one can obtain the geometric
Langlands correspondence. In other words, even without knowing
anything about number theory or various technicalities, one could have
found the statement! It sounds mind-blowing.
There is a 4d theory (G , Ψ) for G and Ψ ∈ CP1 and there is a duality
(G , 0) ↔ (Ǧ , ∞). Upon compactification on C , where for M = Σ × C we
declare only Σ to be a new spacetime, it becomes
(T ∗ BunG (C ), A) ≃ (LocǦ (C ), B):

Fuk(T ∗ BunG (C )) Coh(LocǦ (C ))

D-mod(BunG (C )) QCoh(LocǦ (C ))
15/18
Motivations, Quotes, Result
• Mathematical understanding of Kapustin–Witten’s work

16/18
Motivations, Quotes, Result
• Mathematical understanding of Kapustin–Witten’s work
Here is an interesting quote from Witten’s interview on the occasion
of his 2014 Kyoto Prize reception:
I felt like I discovered the meaning of life and couldn’t explain it.

16/18
Motivations, Quotes, Result
• Mathematical understanding of Kapustin–Witten’s work
Here is an interesting quote from Witten’s interview on the occasion
of his 2014 Kyoto Prize reception:
I felt like I discovered the meaning of life and couldn’t explain it.
I wanted to explain the meaning of life.

16/18
Motivations, Quotes, Result
• Mathematical understanding of Kapustin–Witten’s work
Here is an interesting quote from Witten’s interview on the occasion
of his 2014 Kyoto Prize reception:
I felt like I discovered the meaning of life and couldn’t explain it.
I wanted to explain the meaning of life.
• Capturing algebraic structures

16/18
Motivations, Quotes, Result
• Mathematical understanding of Kapustin–Witten’s work
Here is an interesting quote from Witten’s interview on the occasion
of his 2014 Kyoto Prize reception:
I felt like I discovered the meaning of life and couldn’t explain it.
I wanted to explain the meaning of life.
• Capturing algebraic structures
Experts were not too excited because the geometric Langlands
theory is for an algebraic curve or a Riemann surface C while
Kapustin–Witten’s proposal is for a topological surface.

16/18
Motivations, Quotes, Result
• Mathematical understanding of Kapustin–Witten’s work
Here is an interesting quote from Witten’s interview on the occasion
of his 2014 Kyoto Prize reception:
I felt like I discovered the meaning of life and couldn’t explain it.
I wanted to explain the meaning of life.
• Capturing algebraic structures
Experts were not too excited because the geometric Langlands
theory is for an algebraic curve or a Riemann surface C while
Kapustin–Witten’s proposal is for a topological surface. Here is a
quote from Drinfeld’s email on this point:
... this gave me an excuse not to study their work (and probably
I am not alone) ... Due to your work, I don’t have this excuse
anymore, so I should try to eventually understand what you are
doing ...

16/18
Motivations, Quotes, Result
• Mathematical understanding of Kapustin–Witten’s work
Here is an interesting quote from Witten’s interview on the occasion
of his 2014 Kyoto Prize reception:
I felt like I discovered the meaning of life and couldn’t explain it.
I wanted to explain the meaning of life.
• Capturing algebraic structures
Experts were not too excited because the geometric Langlands
theory is for an algebraic curve or a Riemann surface C while
Kapustin–Witten’s proposal is for a topological surface. Here is a
quote from Drinfeld’s email on this point:
... this gave me an excuse not to study their work (and probably
I am not alone) ... Due to your work, I don’t have this excuse
anymore, so I should try to eventually understand what you are
doing ...

• [Elliott–Y.] developed a framework of physics where one can capture


(de Rham) geometric Langlands program and in particular derive the
geometric Langlands correspondence from physics.
16/18
What do we get?
• [Elliott–Y.] developed a physics framework to capture Geometric
Langlands Theory; found new conjectures and prove new theorems
in the subject; surprising and unexpected results from a traditional
perspective

17/18
What do we get?
• [Elliott–Y.] developed a physics framework to capture Geometric
Langlands Theory; found new conjectures and prove new theorems
in the subject; surprising and unexpected results from a traditional
perspective
• [Hilburn–Y.] found a new construction in geometric representation
theory considerably generalizing the seminal work of
[Braverman–Finkelberg–Nakajima]

17/18
What do we get?
• [Elliott–Y.] developed a physics framework to capture Geometric
Langlands Theory; found new conjectures and prove new theorems
in the subject; surprising and unexpected results from a traditional
perspective
• [Hilburn–Y.] found a new construction in geometric representation
theory considerably generalizing the seminal work of
[Braverman–Finkelberg–Nakajima]
• began new subject “(de Rham) 3d Homological Mirror Symmetry”
(cf. [Braverman–Finkelberg])

17/18
What do we get?
• [Elliott–Y.] developed a physics framework to capture Geometric
Langlands Theory; found new conjectures and prove new theorems
in the subject; surprising and unexpected results from a traditional
perspective
• [Hilburn–Y.] found a new construction in geometric representation
theory considerably generalizing the seminal work of
[Braverman–Finkelberg–Nakajima]
• began new subject “(de Rham) 3d Homological Mirror Symmetry”
(cf. [Braverman–Finkelberg])
• united symplectic representation theory
[Braden–Licata–Proudfoot–Webster, Bezrukavnikov, Losev,
Okounkov] and geometric Langlands theory

17/18
What do we get?
• [Elliott–Y.] developed a physics framework to capture Geometric
Langlands Theory; found new conjectures and prove new theorems
in the subject; surprising and unexpected results from a traditional
perspective
• [Hilburn–Y.] found a new construction in geometric representation
theory considerably generalizing the seminal work of
[Braverman–Finkelberg–Nakajima]
• began new subject “(de Rham) 3d Homological Mirror Symmetry”
(cf. [Braverman–Finkelberg])
• united symplectic representation theory
[Braden–Licata–Proudfoot–Webster, Bezrukavnikov, Losev,
Okounkov] and geometric Langlands theory
• [Ben-Zvi–Sakellaridis–Venkatesh] used the idea to find new structure
in the original Langlands program

17/18
What do we get?
• [Elliott–Y.] developed a physics framework to capture Geometric
Langlands Theory; found new conjectures and prove new theorems
in the subject; surprising and unexpected results from a traditional
perspective
• [Hilburn–Y.] found a new construction in geometric representation
theory considerably generalizing the seminal work of
[Braverman–Finkelberg–Nakajima]
• began new subject “(de Rham) 3d Homological Mirror Symmetry”
(cf. [Braverman–Finkelberg])
• united symplectic representation theory
[Braden–Licata–Proudfoot–Webster, Bezrukavnikov, Losev,
Okounkov] and geometric Langlands theory
• [Ben-Zvi–Sakellaridis–Venkatesh] used the idea to find new structure
in the original Langlands program
• [Raghavendran–Y.] developed a framework for string theory and
created a source to generate new examples of S-duality which in
turn will lead to interesting mathematics
17/18
Summary

◮ Langlands program makes a surprising conjecture that enables one


to translate fundamental problems in number theory to
representation theory (and vice versa) in a precise way.
◮ Geometric Langlands program attempts to understand its more
accessible analogues but in a more structured way; it has recently
paid off for the original program.
◮ It was widely believed that Kapustin–Witten’s proposal for capturing
geometric Langlands in terms of QFT can only see the topological
version, but we show otherwise.
◮ Using further ideas of QFT, we identify new structures in geometric
Langlands theory.
◮ Other subjects in geometric representation theory and mathematical
physics can also be further related.

Thank you for your attention!


18/18

You might also like