A REMARK ON SOME BASES IN THE HECKE ALGEBRA R.VIRK 1 Contents 1 0 1. Introduction 1 2 2. The Hecke algebra 2 n 3. The Ext form 4 a 4. Dualities between the bases 5 J References 6 8 ] O C h. 1. Introduction t a InthisbriefnoteweconsidersomebasesintheHeckealgebraandexhibitcertain m dualities between them. Our results and arguments are completely combinatorial. [ However,toexplainthe motivationandputthecontentsofthisnoteinperspective we need to discuss some results from geometric representation theory. 2 LetGbeareductivegroup,B ⊆GaBorelsubgroupandletBbetheflagvariety v 4 of G. Then the Iwahori-Hecke algebra of G realizes the Grothendieck group of B- 2 equivariantperversesheavesonB.Multiplication inthe Heckealgebracorresponds 9 to convolution of sheaves. 1 In his streamlined treatment of the Hecke algebra, Soergel [So1] considers sev- . 2 eral pairwise commuting automorphisms (see [So1, Thm. 2.7]. The geometric sig- 1 nificance of some of these is well known. For instance, the automorphism d (2.3.1) 0 corresponds to Verdier duality on sheaves. This note was motivated by trying to 1 : understand the geometric significance of the automorphism b (3.4.1). I will now v outline the contents of this note from this perspective. i X A cornerstone of Kazhdan-Lusztig theory is the identification of IC-complexes r (=simple objects) on B with the Kazhdan-Lusztig basis of the Hecke algebra. In a Thm. 3.7 we identify/construct the basis corresponding to projective objects. In Thm.4.2weexplainhowbconnects the projectivebasiswith the Kazhdan-Lusztig basis. Actually, Thm. 4.2 implies that b switches the Kazhdan-Lusztig basis with the basis corresponding to tilting objects (see [Virk, §9] for the precise statement). Now,Soergel’s‘tiltingduality’[So2]switchestiltingobjectswithprojectiveobjects and the Koszul duality of [BGS] switches projective objects with simple objects. It follows that b is the composition of these two dualities. Another way to see this is to use [So3, Thm. 4.4] to directly deduce that b switches the Kazhdan-Lusztig basis with the tilting basis. The aforementioned Koszul duality does not commute with convolution (it doesn’t preserve the monoidal unit). In particular, it does not descendtoaringautomorphismofthe Heckealgebra.However,Thm.4.2isacom- binatorial shadow of the statement that the tilting duality intertwines convolution and Koszul duality (cf. [BG, Conjecture 5.18, Thm. 5.24, Thm. 6.10]). 1 2 R.VIRK Acknowledgments I am grateful to W. Soergel for some very helpful correspon- dence and to Arun Ram for his comments on a preliminary version of this note. This work is supported by the NSF grant DMS-0652641. 2. The Hecke algebra 2.1. Let(W,S) be aCoxetersystem, ℓ: W →Z the correspondinglength func- ≥0 tion and let ≤ denote the Bruhat order on W. In particular, x < y means x ≤ y and x6=y. The identity element in W is denoted by e. 2.2. The Hecke algebra H is the free Z[v,v−1]-module Z[v,v−1]T with x∈W x Z[v,v−1]-algebra structure given by L T T =T if ℓ(xy)=ℓ(x)+ℓ(y), (2.2.1) x y xy (T +1)(T −v−2)=0 for all s∈S. (2.2.2) s s Set H =vℓ(x)T for all x∈W. Then (2.2.2) implies that each H is invertible. In x x x particular, if s∈S, then H2 =1−(v−v−1)H . Hence, s s H−1 =H +(v−v−1). (2.2.3) s s Note that T =1. e 2.3. Define a ring involution d: H→H by d(v)=v−1, d(H )=H−1 . (2.3.1) x x−1 We will often write H instead of d(H). We call C ∈H self-dual if C =C. For each s∈S set C =H +v. (2.3.2) s s Then, by (2.2.3), each C is self-dual. In the sequel we will need the following well s known formulae: 2.4. Lemma. Let s∈S and let x∈W be arbitrary. Then H +vH if sx>x; sx x C H = s x (Hsx+v−1Hx if sx<x. Proof. This is a direct computation and is left to the reader. (cid:3) The following classical result is due to Kazhdan and Lusztig [KL], the proof presented here is stolen from [So1]. 2.5. Lemma. For each x ∈ W there exists a self dual element C ∈ H such that x C ∈H + vZ[v]H . x x y<x y Proof. ProcPeed by induction on the Bruhat order. Certainly we can start our in- duction with C = H = 1. Now let x ∈ W be given and suppose we know the e e existence of C for all y < x. If x 6= e find s ∈ S such that sx < x. Then by y the induction hypothesis C C = H + h H for some h ∈ Z[v]. Set s sx x y<x y y y C = C C − h (0)C . By construction C satisfies the required condi- tioxns. s sx y<x y y P x (cid:3) P 2.6. Lemma. Suppose C ∈ vZ[v]H is self dual. Then C =0. x x Proof. Write C = xhxHxPfor hx ∈vZ[v,v−1]. Using (2.2.3) one sees that P Hy ∈Hy+ Z[v,v−1]Hz z<y X forally ∈W.Picky maximalsuchthath 6=0.ThenC =C impliesthath =h . y y y Thus, h =0, contradicting our assumption. So C =0. (cid:3) y A REMARK ON SOME BASES IN THE HECKE ALGEBRA 3 2.7. Theorem ([KL]). For each x ∈ W there exists a unique self-dual element C ∈H such that C ∈H + vZ[v]H . x x x y y Proof. The existenceis givenbPyLemma 2.5,the uniquenessis providedbyLemma 2.6. (cid:3) 2.8. For each x,y ∈W define polynomials h ∈Z[v] by y,x C = h H . x y,x y y X The following result is well known. It will be key in the sequel. 2.9. Lemma ([KL]). Let s∈S and let x∈W. If sx<x then C =C C − µ(y,sx)C x s sx y y<sx, sXy<y where µ(y,sx) is the coefficient of v in h . y,sx Proof. We have C C =C (H + h H s sx s sx y,sx y y<sx X =C H + h H + h H s sx y,sx y y,sx y y<sx, y<sx, (cid:0) sXy<y sXy>y (cid:1) =H +vH + h (H +v−1H )+ h (H +vH ). x sx y,sx sy y y,sx sy y y<sx, y<sx, sXy<y sXy>y We need to look slightly more carefully at the construction of C in Lemma 2.5. x Note that for y <sx each h ∈vZ[v]. So, by the proof of Lemma 2.5, y,sx C =C C − µ(y,sx)C . (cid:3) x s sx y y<sx, sXy<y 2.10. Proposition. Let s ∈ S and let x ∈ W. For each y ∈ W let µ(y,x) denote the coefficient of v in h , then y,x Csx+ y<x,µ(y,x)Cy if sx>x; C C = sy<y s x ((v+v−P1)Cx if sx<x. Proof. Suppose sx<x, then by Lemma 2.9 C =C C − µ(y,sx)C . x s sx y y<sx, sXy<y Further, C2 = (v +v−1)C by (2.2.3) and so the result follows by the induction s s hypothesis. If sx>x, then once again by Lemma 2.9 C =C C − µ(y,x)C , sx s x y y<x, sXy<y whence the result. (cid:3) The following is originally due to Kazhdan-Lusztig [KL], I learnt the proof pre- sented here from [So1]. 4 R.VIRK 2.11.Proposition([KL]). SupposeW is finite.Letw ∈W bethelongestelement. 0 Then C = vℓ(w0)−ℓ(x)H . w0 x x∈W X Proof. LetH = h H ,h ∈Z[v,v−1],beanelementofHsuchthatb(C )H =0 y y y y s for all s∈S. Then, by Lemma 2.4 P h (H −v−1H )+ h (H −vH )=0 y sy y y sy y sy>y sy<y X X for all s ∈ S. Consequently, h = v−1h for all s ∈ S. Proceeding by induction we s e see that H = h vℓ(y)H . By Prop. 2.10 C C = (v+v−1)C for all s ∈ S. e y y s w0 w0 The result follows. (cid:3) P 3. The Ext form 3.1. Define a symmetric Z[v,v−1]-bilinear form h·,·i: H⊗H→Z[v,v−1] by hH ,H i=δ . (3.1.1) x y x,y 3.2. Remark. This form correspondsto the form consideredin the proofof [BGS, Thm. 3.11.4]. Namely, it corresponds to (−1)idim Exti(M,DNhni)v−n, i,n X where M,N are perverse sheaves on the flag variety, D is Verdier duality, hni is shift + (half) Tate twist in the derived category and v is a formal variable. 3.3. Proposition. If s∈S, then hC H,H′i=hH,C H′i s s for all H,H′ ∈H. Proof. It suffices to prove the assertion for H = H and H′ = H with x,y ∈ W. x y By Lemma 2.4 we have δ +vδ if sx>x; sx,y x,y hC H ,H i= s x y (δsx,y+v−1δx,y if sx<x; and δ +vδ if sy >y; x,sy x,y hH ,C H i= x s y (δx,sy+v−1δx,y if sy <y. So there are four cases to consider: if sx > x and sy > y, then the assertion is evident; if sx > x and sy < y, then certainly x 6= y and so the assertion holds; if sx<xandsy >y,thenonceagainx6=y andthe assertionholds;finally,ifsx<x and sy <y, then the assertion is evident. (cid:3) 3.4. Define a ring involution b: H→H by b(v)=−v−1, b(H )=H . (3.4.1) x x Then b commutes with d. Furthermore: 3.5. Proposition. If s∈S, then hb(C )H,H′i=hH,b(C )H′i s s for all H,H′ ∈H. Proof. Since b(C )=C −(v+v−1), the result follows from Prop. 3.3. (cid:3) s s A REMARK ON SOME BASES IN THE HECKE ALGEBRA 5 3.6. Lemma. Assume W is finite. Then for each x ∈ W there exists a P ∈ x H + vZ[v]H such that hP ,C i=δ for all z ∈W. x y>x y x z x,z Proof.PProceed by induction. Let w be the longest element in W. We start our 0 induction with P = H . Now let x ∈ W be given and suppose we know the w0 w0 existence of P for all y > x. If x 6= w find s ∈ S such that sx > x. Then by the y 0 induction hypothesis b(C )P ∈H + h H s sx x y y y>x X for some h ∈Z[v]. For each y >x let p =hb(C )P ,C i. Set y y s sx y P =b(C )P − p P . x s sx y y y>x X Let’s show that P ∈ H + vZ[v]H . To do this we must demonstrate that x x y>x y (a) p ∈ Z[v] and (b) p (0) = h (0), for each y > x. Both of these follow from y y y P the fact that each h ∈ Z[v] and each C ∈ H + vZ[v]H . By construction y y y z z hP ,C i=δ for all z ∈W. (cid:3) x z x,z P 3.7. Theorem. Assume W is finite. Then for each x ∈ W there exists a unique P ∈H such that hP ,C i=δ for all y ∈W. x x y x,y Proof. Existence was established in the previous Lemma. Uniqueness follows from the evident fact that the form h·,·i is non-degenerate. (cid:3) 4. Dualities between the bases 4.1. Lemma. Assume W is finite. Let w be the longest element in W. Then for 0 all x∈W, P =CxH for some self-dual Cx ∈H. x w0 Proof. Proceed by induction. We start our induction with P = H . Now let w0 w0 x ∈ W be given and suppose we know that P = CyH , with Cy = Cy, for all y w0 y > x. If x 6= w find s ∈ S such that sx > x. Then, by the proof of Lemma 3.6 0 and the induction hypothesis, P =b(C )P − p P =b(C )CsxH − p CyH , x s sx y y s w0 y w0 y>x y>x X X where p = hb(C )P ,C i. To complete the proof it suffices to show that p ∈ y s sx y y Z[v+v−1] for each y. By Prop. 3.5 hb(C )P ,C i=hP ,b(C )C i. s sx y sx s y If sy < y, then b(C )C = (C −(v+v−1))C = 0 by Prop. 2.10. If sy > y, then s y s y once again using Prop. 2.10, b(C )C =(C −(v+v−1))C =C −(v+v−1)C + m C s y s y sy y z z z X for appropriate integers m ∈Z. The result follows. (cid:3) z 4.2. Theorem. Assume W is finite. Let w be the longest element in W. Then 0 b(C )H =P x w0 xw0 for all x∈W. Proof. By Lemma 4.1 P H−1 is self-dual. On the other hand, by Lemma 3.6, xw0 w0 P H−1 ∈H H−1+ vZ[v]H H−1. xw0 w0 xw0 w0 y w0 y>Xxw0 6 R.VIRK Using the fact that H H−1 =H H−1H−1 =H−1 and applying d we obtain z w0 z z w0z−1 w0z−1 P H−1 ∈H + v−1Z[v−1]H . xw0 w0 x yw0 y>Xxw0 By the uniqueness of C (Thm. 2.7) we must have that xw0 P H−1 =b(C ). (cid:3) xw0 w0 x References [BG] A. Beilinson, V. Ginzburg,Wall-crossing functors and D-modules, Represent. Theory 3 (1999), 1-31(electronic). [BGS] A.Beilinson,V.Ginzburg,W.Soergel,KoszuldualitypatternsinRepresentationThe- ory,JournalofA.M.S.9(1996), 473-527. [KL] D. Kazhdan, G. Lusztig, Representations of Coxeter groups and Hecke algebras, Invent. Math.53(1979), 165-184. [So1] W. Soergel,Kazhdan-Lusztig polynomials and a combinatoric for tilting modules, Repre- sentationTheory1(1997), 37-68(electronic). [So2] W. Soergel, Character formulas for tilting modules over Kac-Moody algebras, Represen- tationTheory2(1998), 432-448(electronic). [So3] W.Soergel,Andersenfiltrationand hard Lefschetz,Geom.Funct.Anal.17(2008),no.6, 2066-2089. [Virk] R.Virk,A note on Hecke patterns incategory O,arXiv:1012.2416v1. DepartmentofMathematics,University of California,Davis, CA95616 E-mail address: [email protected]