• Keine Ergebnisse gefunden

Web calculi in representation theory

N/A
N/A
Protected

Academic year: 2022

Aktie "Web calculi in representation theory"

Copied!
31
0
0

Wird geladen.... (Jetzt Volltext ansehen)

Volltext

(1)

Web calculi in representation theory

Or: the diagrammatic presentation machine Daniel Tubbenhauer

5 2 6 1 7

6 6 7 2

5

7 1 8

2 3

5

1 6

Joint work with David Rose, Antonio Sartori, Pedro Vaz and Paul Wedrich

August 2015

Daniel Tubbenhauer August 2015 1 / 30

(2)

History of diagrammatic presentations in a nutshell

Rumer, Teller, Weyl (1932), Temperley-Lieb, Jones, Kauffman, Lickorish, Masbaum-Vogel, ... (≥1971):

Uq(sl2)-tensor category generated byC2q. Kuperberg (1995):

Uq(sl3)-tensor category generated byV1qC3q∼=C3q andV2qC3q. Cautis-Kamnitzer-Morrison (2012):

Uq(slN)-tensor category generated by VkqCNq. Sartori (2013),Grant (2014):

Uq(gl1|1)-tensor category generated by VkqC1|1q . Rose-T. (2015):

Uq(sl2)-tensor category generated bySymkqC2q.Thus,Uq(sl2)-Mod.

Link polynomials:Queffelec-Sartori (2015); “algebraic”:Grant (2015):

Uq(glN|M)-tensor category generated by VkqCN|Mq . T.-Vaz-Wedrich (2015):

Uq(glN|M)-tensor category generated by VkqCN|Mq andSymkqCN|Mq . Sartori-T. (maybe! 2015):

Uq(so2N+1,sp2N,so2N)-tensor categories generated by VkqC2N(+1)q .

Daniel Tubbenhauer August 2015 2 / 30

(3)

1 The story forsl2

Graphical calculus via Temperley-Lieb diagrams The full story forsl2

Proof? Symmetric Howe duality!

2 ExteriorglN-web categories Its cousins: theN-webs Proof? Skew Howe duality!

3 As far as we can go in typeAN−1

Even more cousins: the green-redN-webs Proof? Super Howe duality!

4 The machine in action – yet again

What happens in typesBN, CN andDN? This!

Promise: no moreq’s from now on. But you can insert them everywhere if you like.

Daniel Tubbenhauer August 2015 3 / 30

(4)

The 2-web space

Definition(Rumer-Teller-Weyl 1932)

The 2-web spaceHom2-Web(b,t) is the freeC-vector space generated by non-intersecting arc diagrams withb,t bottom/top boundary points modulo:

Circle

removal: 1 =−2.

Isotopy relations:

1 1

=

1 1

=

1 1

Daniel Tubbenhauer Graphical calculus via Temperley-Lieb diagrams August 2015 4 / 30

(5)

The 2-web category

Definition(Kuperberg 1995)

The 2-web category 2-Webis the (braided) monoidal,C-linear category with:

Objects are vectors~k = (1, . . . ,1) and morphisms areHom2-Web(~k,~l).

Composition◦:

1 1

1 1

= 1 ,

1 1

1 1

=

1 1

1 1

Tensoring⊗:

1 1

1 1

1 1

=

1 1

1 1 1

1

Daniel Tubbenhauer Graphical calculus via Temperley-Lieb diagrams August 2015 5 / 30

(6)

Diagrams for intertwiners

Observe that there are (up to scalars) uniqueU(sl2)-intertwiners cap: C2⊗C2։C, cup:C֒→C2⊗C2, projectingC2⊗C2 ontoCrespectively embeddingCintoC2⊗C2.

Letsl2-Modbe the (braided) monoidal,C-linear category whose objects are tensor generated byC2. Define a functor Γ : 2-Web→sl2-Mod:

~k = (1, . . . ,1)7→C2⊗ · · · ⊗C2,

1 1

7→cap ,

1 1

7→cup

Theorem(Folklore)

Γ : 2-Web→sl2-Modis an equivalence of (braided) monoidal categories.

Daniel Tubbenhauer Graphical calculus via Temperley-Lieb diagrams August 2015 6 / 30

(7)

The symmetric story

Aredsl2-web is a labeled trivalent graph locally made of

capk =

k k

, cupk =

k k

, mk+lk,l =

k+l

k l

, sk,lk+l=

k+l

k l

Herek,l,k+l∈ {0,1, . . .}.

Example

5 2 6 1 7

6 6 7 2

5

7 1 8

2 3

5

1 6

Daniel Tubbenhauer The full story forsl2 August 2015 7 / 30

(8)

Let us form a category again

Define the (braided) monoidal,C-linear category 2-Webrby using:

Definition

Thered 2-web space Hom2-Webr(~k,~l) is the freeC-vector space generated byred 2-webs modulo the circle removal, isotopies and:

glm “ladder”

relations :

l k

l k

l+1 k−1

1 1

k l

k l

k+1 l−1

1 1

= (k−l)

l k

l k

Dumbbell

relation : 2

1 1

1 1

=−

1 1

1 1

+

1 1

1 1

2

Daniel Tubbenhauer The full story forsl2 August 2015 8 / 30

(9)

Diagrams for intertwiners

Observe that there are (up to scalars) uniqueU(sl2)-intertwiners

capk:SymkC2⊗SymkC2։C, cupk:C֒→SymkC2⊗SymkC2,

mkk+l,l :SymkC2⊗SymlC2։Symk+lC2, skk,l+l: Symk+lC2֒→SymkC2⊗SymlC2 given by projection and inclusion.

Letsl2-Mods be the (braided) monoidal,C-linear category whose objects are tensor generated bySymkC2. Define a functor Γ : 2-Webr→sl2-Mods:

~k= (k1, . . . ,km)7→Symk1C2⊗ · · · ⊗SymkmC2,

k k

7→capk ,

k k

7→cupk ,

k+l

k l

7→mkk+l,l ,

k+l

k l

7→skk,l+l

Theorem

Γ : 2-Webr →sl2-Mods is an equivalence of (braided) monoidal categories.

Daniel Tubbenhauer The full story forsl2 August 2015 9 / 30

(10)

“Howe” to prove this?

Howe: the commuting actions ofU(glm) andU(glN) on SymK(Cm⊗CN)∼= M

k1+···+km=K

(Symk1CN⊗ · · · ⊗SymkmCN)

introduce anU(glm)-actionf on the right term with~k-weight space Sym~kCN. In particular, there is a functorial action

Φmsym: ˙U(glm)→glN-Mods,

~k 7→Sym~kCN, X ∈1~lU(glm)1~k 7→f(X)∈HomglN-Mods(Sym~kCN,Sym~lCN).

Howe: Φmsym isfull. Or in words:

relations in ˙U(glm) + kernel of Φmsym relations inglN-Mods.

Daniel Tubbenhauer Proof? Symmetric Howe duality! August 2015 10 / 30

(11)

The diagrammatic presentation machine

Theorem

Define2-Webrsuch there is a commutative diagram U(gl˙ m) Φ

m

sym //

Υm

$$

❏❏

❏❏

❏❏

❏❏

gl2-Mods

2-Webr Γ

99

ss ss ss ss ss

with

Υm(Ei1~k)7→

ki ki+1

ki+1−1 ki+1

1 , Υm(Fi1~k)7→

ki ki+1

ki−1 ki+1 +1 1

Υm glm “ladder” relations, ker(Φmsym) dumbbell relation.

Daniel Tubbenhauer Proof? Symmetric Howe duality! August 2015 11 / 30

(12)

Exempli gratia

Theglm “ladder” relations come up as follows:

EF1~k−FE1~k= (k−l)1~k

l k

l k

l+1 k−1

1 1

k l

k l

k+1 l−1

1 1

= (k−l)

l k

l k

The dumbbell relation comes up as follows:

C2⊗C2∼=V2C2⊕Sym2C2∼=C⊕Sym2C2

2

1 1

1 1

=−

1 1

1 1

+

1 1

1 1

2

Daniel Tubbenhauer Proof? Symmetric Howe duality! August 2015 12 / 30

(13)

It is even better than expected!

The hardestglm “ladder” relations, e.g. Serre relations as Ei2Ei+11~k−2EiEi+1Ei1~k+Ei+1Ei21~k= 0

h k l

h+2 k−1 l−1

1 1

1 a

k

b

−2

h k l

h+2 k−1 l−1

1 1

1

a k

c

+

h k l

h+2 k−1 l−1

1 1

1

a

c

d = 0

do not have to be forced to hold, but are consequences. This pattern repeats in for other web categories.

Morally: web categories have avery economic presentation!

Daniel Tubbenhauer Proof? Symmetric Howe duality! August 2015 13 / 30

(14)

Replace red by green and add orientations

AgreenN-web is an oriented, labeled, trivalent graph locally made of

mk+lk,l =

k+l

k l

, sk,lk+l=

k+l

k l

k,l,k+l ∈N

(and some caps, cups and signs that I skip today).

Example

5 2 6 1 7

6 6 7 2

5

7 1 8

2 3

5

1 6

Daniel Tubbenhauer Its cousins: theN-webs August 2015 14 / 30

(15)

Let us form a category again

Define the (braided) monoidal,C-linear categoryN-Webg by using:

Definition(Cautis-Kamnitzer-Morrison 2012)

ThegreenN-web space HomN-Webg(~k,~l) is the freeC-vector space generated by greenN-webs modulo isotopies and:

glm “ladder”

relations :

l k

l k

l+1 k−1

1 1

k l

k l

k+1 l−1

1 1

= (k−l)

l k

l k

Exterior

relation : k = 0 , if k>N.

Daniel Tubbenhauer Its cousins: theN-webs August 2015 15 / 30

(16)

Diagrams for intertwiners

Observe that there are (up to scalars) uniqueU(glN)-intertwiners

mk+lk,l : VkCNVlCN։Vk+lCN , sk,lk+l:Vk+lCN֒→VkCNVlCN given by projection and inclusion.

LetglN-Mode be the (braided) monoidal,C-linear category whose objects are tensor generated byVkCN. Define a functor Γ :N-Webg→glN-Mode:

~k = (k1, . . . ,km)7→Vk1CN⊗ · · · ⊗VkmCN,

k+l

k l

7→mk+lk,l ,

k+l

k l

7→sk,lk+l

Theorem(Cautis-Kamnitzer-Morrison 2012)

Γ :N-Webg →glN-Mode is an equivalence of (braided) monoidal categories.

Daniel Tubbenhauer Its cousins: theN-webs August 2015 16 / 30

(17)

“Howe” to prove this?

Howe: the commuting actions ofU(glm) andU(glN) on

VK(Cm⊗CN)∼= M

k1+···+km=K

(Vk1CN⊗ · · · ⊗VkmCN)

introduce anU(glm)-actionf on the right term with~k-weight space V~kCN. In particular, there is a functorial action

Φmskew: ˙U(glm)→glN-Mode,

~k7→V~kqCN, X ∈1~lU(glm)1~k 7→f(X)∈HomglN-Mode(V~kqCN,V~lqCN).

Howe: Φmskewis full. Or in words:

relations in ˙U(glm) + kernel of Φmskew relations inglN-Mode.

Daniel Tubbenhauer Proof? Skew Howe duality! August 2015 17 / 30

(18)

Define the diagrams to make this work

Theorem(Cautis-Kamnitzer-Morrison 2012)

DefineN-Webg such there is a commutative diagram U(gl˙ m) Φ

m

skew //

Υm

%%

❏❏

❏❏

❏❏

❏❏

❏❏

glN-Mode

N-Webg Γ

88

rr rr rr rr rr r

with

Υm(Ei1~k)7→

ki+1 ki

ki+1−1 ki+1

1 , Υm(Fi1~k)7→

ki ki+1

ki−1 ki+1 +1 1

Υm glm “ladder” relations, ker(Φmskew) exterior relation.

Daniel Tubbenhauer Proof? Skew Howe duality! August 2015 18 / 30

(19)

Could there be a pattern?

Agreen-redN-webis a colored, labeled, trivalent graph locally made of

mk+lk,l =

k+l

k l

, mk+lk,l =

k+l

k l

, mk+lk,1 =

k+ 1

k 1

, mk+lk,1 =

k+ 1

k 1

And of course splits and some mirrors as well!

Example

5 2 6 1 7

6 6 7 2

5

7 1 8

2 3

5

1 6

Daniel Tubbenhauer Even more cousins: the green-redN-webs August 2015 19 / 30

(20)

The green-red N-web category

Define the (braided) monoidal,C-linear categoryN-Webgrby using:

Definition

Given~k ∈Zm+n

≥0 ,~l ∈Zm+n

≥0 . Thegreen-red N-web space HomN-Webgr(~k,~l) is the freeC-vector space generated byN-webs modulo isotopies and:

glm+gln

“ladder”

relations

: same as before, but now ingreenandred!

Dumbbell relation : 2

1 1

1 1

=

1 1

1 1

2 +

1 1

1 1

2

Exterior

relation : k = 0 , ifk >N.

Daniel Tubbenhauer Even more cousins: the green-redN-webs August 2015 20 / 30

(21)

Diagrams for intertwiners - Part 4

Observe that there are (up to scalars) uniqueU(glN)-intertwiners

mk+1k,1 : VkCN⊗CN։Vk+1CN, mkk+1,1 :SymkCN⊗CN ։Symk+1CN plus others as before.

LetglN-Modes be the (braided) monoidal,C-linear category whose objects are tensor generated byVkCN,SymkCN. Define a functor Γ :N-Webgr→glN-Modes:

~k = (k1, . . . ,km,km+1, . . . ,km+n)7→Vk1CN⊗ · · · ⊗Symkm+nCN,

k+1

k 1

7→mk+1k,1 ,

k+1

k 1

7→mk+1k,1 , · · ·

Theorem

Γ :N-Webgr→glN-Modes is an equivalence of (braided) monoidal categories.

Daniel Tubbenhauer Even more cousins: the green-redN-webs August 2015 21 / 30

(22)

Super gl

m|n

Definition

Thegeneral linear superalgebraU(glm|n) is generated byHi andFi,Ei subject the some relations, most notably, thesuper relations:

Em2 = 0 =Fm2, Hm+Hm+1=FmEm+EmFm, 2EmEm+1Em−1Em=EmEm+1EmEm−1+Em−1EmEm+1Em

+Em+1EmEm−1Em+EmEm−1EmEm+1 (plus an F version).

There is a Howe pair (U(glm|n),U(glN)) with~k = (k1, . . . ,km+n)-weight space under theU(glm|n)-action onVK(Cm|n⊗CN) given by

Vk1CN⊗ · · ·VkmCN⊗Symkm+1CN⊗ · · · ⊗Symkm+nCN.

An aside: everything works forgreen-redU(glN|M)-webs as well, with the Howe pair (U(glm|n),U(glN|M)).

Daniel Tubbenhauer Proof? Super Howe duality! August 2015 22 / 30

(23)

Define the diagrams to make this work - yet again

Theorem

DefineN-Webgr such there is a commutative diagram U˙q(glm|n) Φ

m|n

su //

Υm|nsu▲▲▲▲▲▲▲%%

▲▲

glN-Modes

N-Webgr Γ

88

rr rr rr rr rr r

with

Υm|nsu (Em1~k)7→

km+1 km

km+1−1 km+1

1 , Υm|nsu (Fm1~k)7→

km km+1

km−1 km+1 +1 1

Υm|nsu “glm|n ladder” relations, ker(Φm|nsu ) the exterior relation.

Daniel Tubbenhauer Proof? Super Howe duality! August 2015 23 / 30

(24)

Another meal for our machine

Howe: the commuting actions ofU(so2m) andU(so2N(+1)) on

VK

(Cm⊗C2N(+1))∼= M

k1+···+kn=K

V~kC2N(+1)

introduce anU(so2m)-actionf with~k-weight spaceV~kC2N(+1). In particular, there is a functorial action

Φmso: ˙U(so2m)→so2N(+1)-Mode,

~k 7→V~kC2N(+1), etc..

Howe: Φmsois full. Or in words:

relations in ˙U(so2m) + kernel of Φmso relations inso2N(+1)-Mode.

Daniel Tubbenhauer What happens in typesBN,CNandDN? August 2015 24 / 30

(25)

And another one

Howe: the commuting actions ofU(sp2m) andU(sp2N) on

VK(Cm⊗C2N)∼= M

k1+···+kn=K V~kC2N

introduce anU(sp2m)-actionf with~k-weight spaceV~kC2N. In particular, there is a functorial action

Φmsp: ˙U(sp2m)→sp2N-Mode,

~k 7→V~kC2N, etc.

Howe: Φmsp isfull. Or in words:

relations in ˙U(sp2m) + kernel of Φmsp relations insp2N-Mode.

Daniel Tubbenhauer What happens in typesBN,CNandDN? August 2015 25 / 30

(26)

The definition of the diagrams is already determined

Theorem

DefineN-BDWebg such there is a commutative diagram U(so˙ 2m) Φ

2m

so //

Υmso

&&

▼▼

▼▼

▼▼

▼▼

▼▼

so2N(+1)-Mode

N-BDWebg Γ

77

♥♥

♥♥

♥♥

♥♥

♥♥

♥♥

DefineN-CWebgsuch there is a commutative diagram U(sp˙ 2m) Φ

2m

sp //

Υmsp

%%

▲▲

▲▲

▲▲

▲▲

▲▲

sp2N-Mode

N-CWebg Γ

88

♣♣

♣♣

♣♣

♣♣

♣♣

Υmso so2m “ladder” relations, Υmsp sp2m“ladder” relations etc.

Daniel Tubbenhauer What happens in typesBN,CNandDN? August 2015 26 / 30

(27)

Green type BCD-webs

Greenwebs in types BN andDN are generated by

k k

,

k+l

k l

,

k+l

k l

,

k k

,

k k

Greenwebs in type CN are generated by

k k

,

k+l

k l

,

k+l

k l

, •

2k

, •

2k

The lanterns reflect the fact thatVkC2N is notirreducible in typeCN:

2k

slantern:VkC2N։C, •

2k

plantern:C֒→VkC2N

Daniel Tubbenhauer This! August 2015 27 / 30

(28)

Red type BCD-webs

There are also Howe pairs (U(sp2m),U(so2n(+1))) and (U(so2m),U(sp2n)) acting now on the symmetric tensors. Guess what comes out:red webs!

Redwebs in type CN are generated by

k k

,

k+l

k l

,

k+l

k l

,

k k

,

k k

Redwebs in types BN andDN are generated by

k k

,

k+l

k l

,

k+l

k l

, •

2k

, •

2k

The lanterns reflect the fact thatSymkC2N(+1)isnotirreducible in typesBN,DN:

2k

slantern:Sym2kC2N ։C, •

2k

plantern:C֒→Sym2kC2N

Daniel Tubbenhauer This! August 2015 28 / 30

(29)

I do not have tenure. So I have to bore you a bit more.

Some additional remarks.

Homework: feed the machine with yourfavorite duality.

Everything quantizes without too many difficulties. The quantized version sheds new light on HOMFLY-PT, Kauffman and Reshetikhin-Turaev polynomials: their symmetries can beexplainedrepresentation theoretical.

Some parts even work in thenon-semisimple case (e.g. at roots of unities).

The whole approach seems to be amenable to categorification.

Relations to categorifications of the Hecke algebra using Soergel bimodules or categoryO need to be worked out.

This could lead to a categorification of ˙Uq(glm|n) (since the “complicated”

super relations are build in the calculus).

A “green-red-foamy” approach could shed additional light on colored Khovanov-Rozansky homologies.

Daniel Tubbenhauer This! August 2015 29 / 30

(30)

There is stillmuchto do...

Daniel Tubbenhauer This! August 2015 30 / 30

(31)

Thanks for your attention!

Daniel Tubbenhauer This! August 2015 30 / 30

Referenzen

ÄHNLICHE DOKUMENTE

Daniel Tubbenhauer Colored Jones and HOMFLY-PT polynomials October 2015 26 / 28.. I do not

Daniel Tubbenhauer The virtual Khovanov homology July 2013 11 / 29.+. the complex is an invariant up to chain

Theorem 3.11 implies that if V is a finite dimensional representation of a finite group G, then the complex conjugate representation V (i.e., the same space V with the same addition

The author conjectures that the whole construction can be used as a “blueprint” for a cate- gorification of the virtual sl 3 polynomial (as explained in Section 1.3), since

•...in the context of the representation theory of Hopf algebras? Fusion rules i.e. tensor products rules... •...in the context of categories? Morphisms of representations and

2-representation theory is a modern version of representation theory and has connections to quantum groups, Hopf and Lie algebras.. It grow out of and runs in parallel to the study

2-representation theory is a modern version of representation theory and has connections to quantum groups, Hopf and Lie algebras.. It grow out of and runs in parallel to the study

it is unknown whether there are always only finitely many 2-simples (probably not)... An additive, k-linear, idempotent complete, Krull–Schmidt category C is called finitary if it