• Ingen resultater fundet

Character formulae and tensor ideals

In document Afhandling (Sider 35-45)

This chapter presents the character formulae for indecomposable quantum tilting mod-ules. The characters are expressed by certain basis change coefficients in a Hecke algebra module. We introduce this module, the Hecke module, first and then go on to state So-ergel’s character formula.

This chapter contains also a classification of tensor ideals of quantum tilting modules.

The proof relies on Hecke algebra combinatorics, and may be seen as an application of the right cell theory of Chapter 4 and of the character formula for indecomposable quantum tilting modules. We conclude the chapter by introducing weight cells.

Assumep hthroughout this chapter.

The Hecke module

The material in this section may be found in (Soergel 1997).

The finite Weyl groupW0is a parabolic subgroup ofW, so each right coset ofW0 W has a unique representative of minimal length. We denote the set of these representatives byW0; multiplication then gives a bijectionW0 W0 W. The setW0is described thoroughly by Lemma 4.19. In particular, if we associate to each element x W0 the alcovexCas in figure 2,3, and 4 then we get all alcoves in the dominant Weyl chamber.

LetH0denote the Hecke algebra ofS0W0. It is a subalgebra ofH. We have a surjec-tive vv 1-algebra homomorphism,φ

v:H0

vv 1, mapping each generatorHsi, si S0, to v. This gives vv 1 aH0-module structure, and by induction we obtain a rightH-module

N vv 1 H0H

This module is the Hecke module. In Lemma 5.5 below we show that N may also be constructed as a quotient ofH with a right cell ideal.

We denote byεthe canonical surjection

ε:H N

H 1 H

Note that forz W0,x W0 we haveεHzx vlzNx. N has a basis consisting of

Nx 1 Hx x W0 and the action ofH is given by

NxHs

Nxs! vNx xs" xandxs W0; Nxs! v 1Nx; xs# x;

0 xs$ W0;

(5.1)

The first and second line follows from equation (4.3). The assumption in the last line forces xs txfor somet S0, which gives the formula.

We define an involution onN bya H a H. This involution isH-skewlinear (i.e.

NH N HforN NH H). We say thatNis selfdual ifN N. As was the case with H, we want to replace the canonical basis Nx 1 Hx x W0 with one of selfdual elements.

35

36 5. CHARACTER FORMULAE AND TENSOR IDEALS

THEOREM5.1. For each x W0there is a unique selfdual element NxinN so that Nx Nx

y xvvNy

PROOF. This theorem is proved by the same method as Theorem 4.1. In particular, the proof is constructive and gives an algorithm to obtain the selfdual elementNx.

To begin, note thatNe 1 Heis selfdual; this shows the existence part of the theorem forxof length 0.

Givenxof length 1 we chooses Sso thatxs x. By induction we may assume the existence of the selfdualNxs Nxsy xsnyxsNy, withnyxs vv. ThenNxsHsis selfdual. We write

NxsHs Nx

y x

nyNy (5.2)

wherenyv by (5.1). So constant coefficients may appear inNxsHs. We fix this with Nx NxsHs

y xny0 Ny which is selfdual, and this shows the existence part forx.

We leave uniqueness to the reader.

LEMMA5.2.

εHx

Nx x W0 0 x W0

PROOF. The first line follows from the fact thatεHx is selfdual andεHx Nx

y xvvεHy Nxz xvvNz. For the second line, note thatεHsi εHsi v 0 forsi s0. Whenx W0there is asi s0so thatx six. Therefore we may prove the

second line by induction via Remark 4.4.

DEFINITION5.3. Let nyx Nx:Nyv denote the coefficients of the new selfdual basis expressed in the old basis, so that

Nx

y xnyxNy

REMARK5.4. The mapεreveals that the polynomials nyxand hyxare related by the formula

nyx

w W0

vlwhwyx (5.3)

Recall that we denote the linear coefficient of hyxby µyx . The equation (5.3) shows that the linear coefficient of nyxis equal to µyx for xy W0.

The Hecke algebra acts on the selfdual basis by the following formulae; to see this use εon the equations in Proposition 4.3.

NxHs !"

""""

#"""""$

Nxs

y% Rxy W0

µyxNy xs xandxs W0;

y% R

xy W0

µyx Ny xs xandxs W0;

v v& 1Nx xs x

(5.4)

We conclude this section with an alternative description ofN. Recall the right cell idealsIxfrom Chapter 4.

LEMMA5.5. εinduces an isomorphism of rightH-modules betweenH i

'

( 0Isiand N.

PROOF. We have∑i(' 0Isi*) x+W0vv& 1Hxand this is the kernel ofεby Lemma

5.2.

SOERGELS THEOREM 37

We will need the following analogue of Lemma 4.5.

LEMMA5.6. When xs x and ys y we have nysx vnyx.

Soergels Theorem

Our main interest in the Hecke module is Theorem 5.7 below. It relates the structure of quantum tilting modules with combinatorics of the corresponding Hecke algebra. Recall that for a moduleQwith a Weyl filtration, Q:Vqλ denotes the number of timesVqλ appears as a quotient in the filtration. Clearly, the numbers Q:Vqλ for allλdetermine the character ofQ. WhenQis a quantum tilting module we letQ:Tqλ denote the num-ber of timesTqλ appears as a summand ofQ; we say that Q:Tqλ is the multiplicity ofTqλ inQ.

THEOREM5.7. Let p h. For a weightλ C and xyz W0we have (i) Tqxλ :Vqyλ nyx1

(ii) If xs x then ΘsTqxλ :Tqzλ NxHs:Nz.

The proof of (i) may be found in (Soergel 1998). This result of Soergel relies on an equivalence of categories between affine Lie algebra modules and quantum group modules established in (Kazhdan and Lusztig 1993), (Kazhdan and Lusztig 1994) together with results from (Lusztig 1994) and (Kashiwara and Tanisaki 1996). In some types these results impose mild restrictions onp.

We will use the second part of Theorem 5.7 extensively. It is a straightforward conse-quence of the first part as shown in the proof below.

PROOF OF(II). Consider a moduleQwith character chQy;yλ X ayλχyλ. Ap-plying the wallcrossing functor we obtain a module with character given by chΘsQ

y;yλ X ayλχyλ χysλ.

Now consider a Weyl factorVqyλ inTqxλ. Suppose first thatys W0; then by the first part of the theorem,

nyx1 nysx1 Tqxλ :VqyλTqxλ :Vqysλ

ΘsTqxλ :Vqyλ

z xs

ΘsTqxλ :TqzλTqzλ :Vqyλ

z xs

ΘsTqxλ :Tqzλnyz1 In the first equation below, see Equation (5.1).

NxHs y;

ys

y ysW0

nyxNys vnyxNy nysxNy v 1nyxNys

NxHs

z xs

NxHs:Nz

y;ys ynyzNy nyszNys

Considering the coefficientNxHs:Ny and evaluating inv 1, we find that nyx1 nysx1

z xs

NxHs:Nznyz1

Ifys W0thenΘsTqxλ :Vqyλ 0 NxHs:Ny. So for allyz xswe have

z

xs

ΘsTqxλ :Tqzλnyz1

z xs

NxHs:Nznyz1

It is now a matter of linear algebra to conclude thatΘsTqxλ :Tqzλ NxHs:Nz

for allz xs.

38 5. CHARACTER FORMULAE AND TENSOR IDEALS

The Hecke module atv 1

We continue this chapter with a construction of tensor ideals of quantum tilting mod-ules. This is based on the paper (Ostrik 1997). It illustrates the strength of right cells when applied to the representation theory of quantum groups at a root of unity, through Theorem 5.7. This section prepares the ground for Ostriks tensor ideals in the next section.

Inspired by Theorem 5.7 we begin with the specialization ofN atv 1. Consider as anvv 1-module withvacting as multiplication by 1. We then obtain a rightH-module by

N1 v

v 1 N From thevv 1-bases ofN we get two bases ofN1:

N1x 1 Nx x W0 N1x 1 Nx x W0

As endomorphisms ofN1we haveHxHy Hxyeven whenlx ly lxy. Further, the generators act by the formulae below; the first equation follows from the defining relations (4.1) and (4.2) ofH and the second equation follow from equation (5.4).

Nx1Hs

Nxs1 ifxs W0;

N1x ifxs W0; (5.5)

N1xHs !

N1xs

y" Rx y# W0

µyxN1y xs$ x xs W0;

y" R

x y# W0

µyxN1y xs$ x xs W0;

2Nx xs x%

(5.6)

A right cell submodule ofN1 is a right H-submodule with a basis consisting of

Ny y Y for some subsetY ofW. The following lemma describes all of them.

LEMMA5.8. LetC & W0denote a right cell, and let x C. Then JC ('

y) RC y# W0

*vv 1N1y

is a right cell submodule andJC is the smallest right cell submodule to contain N1x. A right cell submodule is a sum of suchJC, and a sum of suchJC is a right cell submodule.

REMARK 5.9. Note that the set y W0 y+ R C is a union of right cells; it is equal to the union of allC,-+ R C.

PROOF OFLEMMA5.8. Equation (5.6) shows thatJC is stable under the generators HsofH. So eachJC is a right cell submodule.

Suppose that y W0 andy+ R C. We will show thatN1y belongs to any right cell submodule containingN1x. There is a chainy w0. R w1/0/1/. R wn x, wherewi W to begin with. But in factwi W0, asLwi2& Ly3 s0 by Lemma 4.13. Nowwi. R wi4 1means thatHwi

5 1Hs:Hwi76 0 for somes; further

Hwi

5 1Hs:Hwi N1wi

5 1Hs:N1wi

by equation (5.6). This shows that each N1wi must be part of any basis of a right cell submodule containingN1x. We have shown thatJC is the smallest right cell submodule containingN1x.

We leave the remaining assertions to the reader. 8

OSTRIK’S TENSOR IDEALS 39

Ostrik’s tensor ideals

Inside the categoryCqof finite dimensionalUq-modules we have the full subcategory of tilting modulesTq.

DEFINITION5.10. A subsetτ Tqis a tensor ideal if it satisfies the following condi-tions.

(i) For any Q1inτand any tilting module Q, the tensor product Q1 Q belongs to τ.

(ii) If Q is a summand in Q1and Q1belongs toτ, then Q belongs toτ.

These two properties are sufficient to ensure that tensor ideals are stable under the translation functors.

Consider a tensor idealτ Tq; it follows from the defining properties thatτis spanned by a set of indecomposable tilting modules Tqλ λ Λ (Λdenoting a subset of the dominant weights) in the sense that all modules inτare direct sums of these indecompos-ables. What can we say aboutΛ? What are the possible subsetsΛ X ?

LEMMA5.11. Letτ Tqdenote a tensor ideal. Then Tqx0 belongs toτif and only if Tqxλ belongs toτwhenever xλis a dominant weight in the lower closure of the alcove xC.

PROOF. Recall that tensor ideals are stable under the translation functors. The claim follows from (Soergel 1997, Remark 7.2 2.): Whenxλbelongs to the lower closure of the alcovexC, we have

T0λTqx0 TqxλWλ Tλ0Tqxλ Tqx0

HereWλdenotes the stabilizer ofλ.

It follows that all tensor ideals are spanned by sets Tqλ λ Λ whereΛis the union of the set of dominant weights in lower closures of alcoves.

COROLLARY5.12. To each tensor idealτinTqwe may associate a subsetCτ W0 so that the indecomposable tilting modules inτare all Tqλ withλin the lower closure of one of the alcoves yC, y Cτ.

PROPOSITION5.13. Letτdenote a tensor ideal inTqand letCτ W0denote the associated set of Weyl group elements. Then x Cτ N1x is a right cell submodule inN1, andCτ is a union of right cells in W0.

PROOF. Letx Cτ and assume thatN1xHs:N1y

0. We must show thaty Cτ, as this demonstrates that x Cτ N1xis stable under the generatorsHsofH. By definition ofCτ we know thatTqx0 belongs to the tensor idealτ. To show the proposition we need only verify thatTqy0 belongs toτ. This is immediate from

ΘsTqx0 :Tqy0 NxHs:Ny N1xHs:N1y

The first equality above follows from Theorem 5.7 and the second by comparing the equa-tions (5.4) and (5.6). The last assertion in the proposition follows from Lemma 5.8 and

Remark 5.9.

By now we have seen that a tensor ideal inTqcorresponds to a right cell submodule.

Further the right cell submodules separate tensor ideals, in the sense that two tensor ideals are different if and only if they correspond to different right cell submodules. This takes us some way toward a classification of tensor ideals inTqas Lemma 5.8 provide us with a classification of right cell submodules in terms of right cells.

It remains to construct a tensor ideal from each right cell submodule. It suffices to do this for the minimal right cell submodulesJC. This occupies the rest of this section.

40 5. CHARACTER FORMULAE AND TENSOR IDEALS

We begin by connecting the ring of characters with the moduleN1. Recall that the characters of the Weyl modules χλ λ X form a basis of the character ring. Then set

α0: XW0 N1 χλ

0 ifλ W0

Nx1 ifλ x0 for somex W0 REMARK5.14.

(i) This map is well defined as0is a regular weight when p h.

(ii) By Theorem 5.7 we haveα0Tqx0 N1x. Note that we should really write α0chTqx0, but this should not cause confusion.

The following lemmata give two important properties of the mapα0. The first lemma is almost trivial, but the second is rather technical.

LEMMA5.15. Let w W . Then

α0χw0 α0χ0Hw

PROOF. Decomposew w0xwithw0 W0andx W0satisfyinglw lw0 lx . Then, using formula (5.5) in the second line

α0χw0 α0 1lw0χx0 ! 1lw0Nx1

α0χ0Hw Ne1Hw0x ! 1lw0Ne1Hx ! 1lw0Nx1 "

The following lemma from (Ostrik 1997) is the key result needed in the proof of Proposition 5.18; there we show that a minimal right cell submoduleJC corresponds to a tensor ideal.

LEMMA5.16. Letλ#µ C and assume that zλis a dominant weight. For any module Q inCq, there is an HQ H so that

α0Tµ0Tqzλ%$ Q α0Tλ0TqzλHQ

PROOF. Write chTqzλ&w' Wawχwλ. Then by (Jantzen 1987, II.7.5b) we have chprµTqzλ($ Q

w'W

aw

ν)ν

λ' W*µ

dimQνχw+λ ν

Consider the following set of Weyl group elements:

P, x W ν λ xµ# dimQν- 0

This set is clearly stable under multiplication from the right withw1 Wµ. It is also stable under multiplication from the left byw2 Wλ: We havew2xµ w2.λ ν& w2λ w/ν for somew/ in the finite Weyl group, and dimQν dimQw0ν.

The actions of the groupsWµ(from the right) andWλ(from the left) are simply transi-tively. The orbits of theWµ-action are indexed by the weights ofQ, so we choose for each weightνofQan elementwνso thatwνWµµ λ ν. Letwi,i 1#1#ldenote a set of representatives from the orbits of theWλ-action. ThenPmay be described as

P, wνw1 w1 Wµ# dimQν- 02, w2wi w2 Wλ# i 1#1#l3 We may now continue the calculation. We have

chprµTqzλ($ Q

w' W

aw

ν)ν

λ4 wν*µ

dimQν χwwνµ5

This gives

chTµ0Tqzλ($ Q

w' W

aw

ν)ν λ4 wν*µ

dimQν

w1' Wµ

χwwνw105

OSTRIK’S TENSOR IDEALS 41

We use Lemma 5.15 in the following calculation.

α0 Tµ0Tq zλ Q

wW

aw

νν λ wνµ

dim Qν

w1 Wµ

α0χ wwνw10

α0

w Wawχ w0

νν λ wνµ

dim Qν

w1 WµHwνw1

α0

w W

awχ w0

i

w2 Wλ

dim QνHw2wi

α0

w2 Wλ

w W

awχ ww20

i

dim Qν Hw

i

α0 T0λTq zλ

i dim QνHwi

Here the last line follows as

chTλ0Tq zλ

w W

aw

w2 Wλ

χ ww20

REMARK5.17. In Lemma 5.16 we may replace Tq zλ by any other module Q with prλQ Q.

Proposition 5.13 and Corollary 5.12 states that each tensor ideal inTqdetermines a right cell submodule in N1, and that two different tensor ideals are mapped to different right cell submodules. In the other direction, Proposition 5.18 shows that each of the minimal right cell modulesJC from Lemma 5.8 corresponds to a tensor ideal.

PROPOSITION 5.18. LetC W0 denote a right cell. The indecomposable tilting modules with highest weight in the lower closure of an alcove xC, x R C span a tensor ideal.

PROOF. Letτ C denote the set of all indecomposable tilting modules with highest weight in the lower closure of one of the alcovesxC,x R C. Recall the mapα0from the ring of characters toN1defined above. We first show that the mapα0reveals whether a given indecomposable tilting module belongs toτ C or not. Suppose thatTq yµ is such a module, withµ Candychosen so thatyµbelongs to the lower closure of the alcove yC. Then

α0 Tµ0Tq yµ α0 Tq y0 N1y

We see thatTq yµ belongsτ C if and only if it is mapped toJC N1byα0 Tµ0

. Now assume thatx R C. LetQdenote an arbitrary tilting module and consider the tensor productTq xλ Q. We must show that each summandTq yµ belongs toτ C. By the discussion above this is equivalent toα0 Tµ0 Tq xλ Q JC for allµ C. Now Lemma 5.16 shows that

α0 Tµ0 Tq xλ Q α0 Tλ0Tq xλH

for some H H. As α0 Tλ0Tq xλH N1xH and asJC is a H-module we find that

α0 Tµ0 Tq xλ Q JC as needed. We are done.

Denote byτ C the tensor ideal corresponding to the right cell C, i.e. letτ C de-note the tensor ideal spanned by all indecomposable quantum tilting modules with highest weight in the lower closure of an alcovexCwithx R C. We may now classify the tensor ideals inTq.

THEOREM5.19. LetC W0denote a right cell and let x C. The tensor idealτ C is the smallest tensor ideal to contain Tq x0. A tensor ideal is a union of suchτ C, and a union of suchτ C is a tensor ideal.

42 5. CHARACTER FORMULAE AND TENSOR IDEALS

PROOF. Let us consider the smallest tensor ideal containingTq x0. By Corollary 5.12 and Proposition 5.13 we may associate to it a subsetY W0, so that y Y N1yis a right cell submodule ofN1. This submodule containsN1x, so it contains alsoJC, where C W0is the right cell containingx. We have shown thatC Y. Hence any tensor ideal containingTq x0 must containτ C. Further,τ C is a tensor ideal by Proposition 5.18;

we see thatτ C is the smallest tensor ideal to containTq x0.

The last assertions are clear.

Weight cells

In the previous sections we have seen that there is a close connection between right cell ideals inH and right cell submodules ofN1on one side and tensor ideals of quantum tilting modules on the other side. Weyl group elements that generate the same right cell ideal or right cell submodule are said to belong to the same right cell. There is a similar notion in the category of quantum tilting modules. The highest weights in tilting modules that generate the same tensor ideal are said to belong to the same weight cell.

DEFINITION 5.20. (Ostrik 2001) Write µ Tq λif Tq λ Q:Tq µ 0for some quantum tilting module Q.

REMARK5.21.Note that Tqis a preorder, since Tq λ Q1:Tq µ 0and Tq ν Q2:Tq λ 0gives Tq ν Q1 Q2:Tq µ 0. Note that Q1 Q2is tilting because a tensor product of tilting modules is tilting.

DEFINITION 5.22. Let Tq be the equivalence relation defined by Tq. The equiv-alence classes of Tq are calledweight cells. The preorder Tq induces a partial order (also denoted Tq) on the set of weight cells in the natural way.

To ease the notation, we write ˇAfor the lower closure of an alcoveA.

REMARK5.23.

(i) If µ λ νand all three weights are dominant then µ Tqλas weight consid-erations show that Tq µ is a summand of Tq λ Tqν.

(ii) If Tq µ is a summand in a translation or a wallcrossing of Tq λ then µ Tqλ.

(iii) Let x W0and assume thatλ, µ xˇC X . Then (the proof of) Lemma 5.11 shows thatλ Tq µ. So a weight cell consists of all dominant weights in the lower closure of xC, x in some subset of W0.

(iv) If xy W0and y R x then y0 Tq x0since ΘsTq x0 :Tq y0 NxHs: Ny 0. This immediately give us that y R x implies y0 Tq x0and that y R x implies y0 Tqx0

THEOREM5.24. There is a one-to-one correspondence between the right cells in W0 and weight cells in X . The image of a right cell is given by

C "!$#

x C

xˇC X

This correspondence is order preserving.

PROOF. We first show that % x CxˇC X is a weight cell, for a right cellC. Letz, y Cand letyλ yˇC X ,zµ zˇC X . We have

zµ Tqz0 Tqy0 Tqyλ

where the first and last Tq follows from Remark 5.23 (iii) and the second follows from (iv) asx R y.

To complete the proof we must show thatyλ yˇC,zµ zˇCwithzµ Tqyλgives z R y. First of all, we havez0 Tqy0 as before. This means thatTq z0 belongs to all tensor ideals containingTq y0; by Theorem 5.19 we conclude thatz R y. We get in the same wayy R zand hencez R y.

WEIGHT CELLS 43

Finally, it follows from Remark 5.23 (iv) that this mapping respects the order of the sets.

We will denote the weight cell corresponding to the ’first’ right cell e byc1; note that c1 C. The weight cell corresponding to the second right cellCs0 is similarly denoted byc2; it is the second largest weight cell. In the theorem below we identify the minimal weight cell which we will denote bycSt. Figure 5, 6, and 7 in Chapter 6 hold pictures of the weight cells in typeA2,B2, andG2.

THEOREM5.25. (Andersen 2001a)St X is a weight cell. It is minimal.

PROOF. The proof relies on a complete description of the injective indecomposable tilting modules inTq, given in (Andersen 2001b). We have

Tqλ injective λ St X

Let us first see thatTqSt is in every tensor ideal of tilting modules. The dual of a tilting module is tilting, since the dual of Weyl module is an induced module and vice versa.

Further, the dual of a indecomposable module is indecomposable, as EndQ EndQ for all modules. We thus find thatTλ T w0λ, since the highest weight ofTλ is

w0λ. Then

HomTqStTqλ Tq w0λ St HomTqSt Tq w0λTq w0λ St The last space is nonzero since the second module is a summand in the first by weight considerations. ThenTqSt is a summand inTqλ Tq w0λ St asTqSt is simple and injective. It follows that St Tqλfor all dominant weights. So St is a minimal element in the preorder Tq.

We have seen that St belongs to the minimal weight cell. By Remark 5.23 (i) we find that St X is contained in this minimal weight cell. It remains to see thatλ TqSt implies λ St X .

Suppose thatTqλ is a summand in aTqSt QwithQinTq. The tensor product is injective since TqSt is. ThenTqλ is necessarily injective as it is a summand in an injective module. It follows thatλ St X .

CHAPTER 6

In document Afhandling (Sider 35-45)