CATEGORY O AND sl LINK INVARIANTS k JOSHUASUSSAN 7 0 Contents 0 2 1. Introduction 1 n a 2. Representation Theory of sl 3 k J 3. Categorification of sl Modules 7 1 k 4. Diagram Relations 14 ] A 5. Graded Category O 39 Q 6. Crossings 61 h. References 74 t a m [ 1. Introduction 1 v The program of categorification via category O was introduced by J. Bernstein, I. Frenkel, and M. Kho- 5 vanovin[BFK]wheretensorpowersofthestandardtwodimensionalrepresentationofsl wererecognizedas 4 2 0 GrothendieckgroupsofcertainsubcategoriesofOforvariousgln.Theyhadtwodifferentconstructions. One 1 was based on studying certain blocks with singular generalized central characters. The other was based on 0 examining the trivial regular block but by considering various parabolic subcategories. In the first case the 7 actionofsl wascategorifiedbyprojectivefunctorsactingonthesesingularcategories. Theintertwinerswere 0 2 categorified by derived Zuckerman functors acting on the derived category O. In the latter case, the action / h of the Lie algebra was lifted to Zuckerman functors and the intertwiners became projective functors. These t a two constructions are related by Koszul duality where the projective functors get exchanged for Zuckerman m functors and visa-versa[Rh], [MOS]. : Several conjectures posed in [BFK] were solved in [Str2]. The most important result of that work was v i that the Jones polynomial was lifted to a functorial invariant of tangles. Natural transformations between X these functors became invariantsof2-tanglesactingascobordismsbetweentwodifferent tangles[Str4]. One r of the principle goals of categorification is to obtain invariants for n+1 dimensional topological objects by a homological realizations of classical n dimensional invariants. Thereisaparallelapproachtocategorificationoflinkinvariantsdue toKhovanov. In[Khov],abi-graded homology theory was constructed whose graded Euler characteristic is the Jones polynomial. M. Khovanov andL.Rozanskyextended this workto a homologytheory categorifyingthe HOMFLYPT polynomial[KR]. The theory of matrix factorization is their primary tool. H. Murakami, T. Ohtsuki, and S. Yamada have an intepretationoftheHOMFLYPTpolynomialthroughapolynomialassignedtocertainplanargraphs[MOY]. Khovanov and Rozansky categorifiedthese polynomials which is a crucial step in the categorificationof the HOMFLYPT polynomial. Thegoalofthispaperistocategorifythesl tangleinvariantviacategoryO.Itisimportanttounderstand k all of the planar graphs representation theoretically. The planar graphs give a graphical interpretation of intertwiners between various tensor products Λi1Vk−1 ⊗···⊗ΛirVk−1 of fundamental slk representations. Thefirststepistocategorifytensorpowersofthestandardk−dimensionalrepresentationofsl .Inorderto k accomplishthis, we lookatmoregeneralsingularblocksofcategory⊕dOd(gln).Here, ddenotesa k− tuple Date:February2,2008. 1 (dk−1,...,d0).ThenOd(gln)isablockcorrespondingtoanintegraldominantweightwhosestabalizerunder the dot action of the Weyl group is S ×···×S . Parabolic subcategories of these blocks, Op(gl ) will dk−1 d0 d n provide a categorification of tensor products of fundamental representations. The action of the Lie algebra sl will be lifted to projective functors. These are functors given by tensoring with a finite dimensional k representationandthenprojectingontoacertainblock. MorphismsbetweenvariousΛi1Vk−1⊗···⊗ΛirVk−1 will become inclusion and derived Zuckerman functors between various parabolic subcategories of these blocks. Relationsbetweentheintertwinersbecomefunctorialisomorphismsbetweenvariouscompositionsof inclusion and derived Zuckerman functors. It should be possible to do this construction in the Koszul dual situationaswell. InthiscaseprojectivefunctorscategorifytheintertwinersandderivedZuckermanfunctors categorify the action of the Lie algebra. This construction only gives rise to a categorification of sl − modules. In order to get a categorification k of the quantum group, we consider the above categories to be a category of graded modules. Then in the Grothendieckgrouptheshiftfunctorwilldescendtotheactionofmultiplicationbythequantumparameterq. ThetheoryofgradedcategoryOisduetoW.SoergelandhasbeenstudiedextensivelybyC.Stroppel[Str1]. Projectivefunctorshavegradedlifts. ThisallowsustogiveacategorificationofthequantumSerrerelations. Zuckermanfunctorsandinclusionfunctorsalsohavegradedlifts. Gradedfunctorsmaynowbeassignedtoall flattangles. Inordertoextendthisconstructiontotanglewithcrossings,weconsideradjunctionmorphisms between shifted compositions of inclusion and Zuckermanfunctors for a minimal parabolic and the identity functor. Cones of these morphisms are assigned to the crossings. In this graded setup it is easy to see that these cones satisfy the sl skein relation. A cobordism between two tangles should give rise to a natural k transformationbetweenthe functorsassignedto the tangles. This naturaltransformationshouldbecomean invariant of 2-tangles as in [Str4], [Khov3], [Jac]. In section 2 we review the necessary finite dimensional representation theory of sl . A categorification k of the fundamental representations will be given in section 3. An equivalence of categories between certain parabolicsubcategorieswhichgeneralizescorollary5in[BFK]isalsogiveninthissection. Thisequivalenceis usedinthedefinitionofthefunctorsassignedtotheflattangles. Insection4weprovidefunctorialanalogues of the graphic relations presented in section 2. Each of the graphical relations will give rise to a functorial isomorphism. NaturaltransformationsareconstructedandshownthatwhenrestrictedtogeneralizedVerma modules, they are isomorphisms. Most of the section is devoted to calculations of the relevant functors on generalized Verma modules. In section 5 graded category O will be discussed and all of the results in the previoussections willbe lifted to the gradedcase. We definein section6 ourfunctor valued tangleinvariant F and prove that it satisfies the Reidemeister relations. Suppose that tangles T and T′ are morphisms from r points to r′ points labeled from the set {1,k−1}. Denote by n and n′ the sum of the labels for the r and r′ points respectively. These compositions of n and n′ naturally give rise to parabolic subalgebrasp and p′ of gl and gl . Theorem 7 is the main result of this n n′ work. Theorem. If tangles T and T′ are ambient isotopic tangles, then F(T) and F(T′) are isomorphic as derived functors from Db(⊕dOdp(gln)) to Db(⊕d′Odp′′(gln′)). The functorial isomorphisms of section 4 play a crucial role in proving this theorem. When restricted to (0,0)− tangles, the invariant is a complex of graded vector spaces and thus homology groups may be assigned to links. It follows immediately from the definition of F that on the Grothendieck group, the following equality holds: Theorem. Let Ω and Π be the functors assigned to the crossings. Then qk[Ω ]−q−k[Π ]=(−1)k−1(q1− i i i i q−1)[Id] An explicit relationship between the functorial tangle invariant introduced by Khovanov in [Khov5] and thecategoryOinvariantof[Str2]wasgivenin[Str5]. Stroppelconsideredthesubcategoryofthetrivialblock ofO(sl )ofmoduleslocallyfinitewithrespecttoaparabolicsubalgebrawhosereductivepartisgl ⊕gl and 2n n n whoseprojectivepresentationonlyconsistsofprojective-injectivemodules. Thissubcategorycategorifiesthe space of invariantsin tensor products of the standardtwo dimensionalof sl . This subcategoryis equivalent 2 to a category of modules over an associative algebra which is isomorphic to the algebra constructed in 2 [Khov5]. A naturalquestionis to find aconnectionbetweenthe functorialsl invariantin this workandthe k invariantconstructedviamatrixfactorizationin[KR]orSoergelbimodulesin[Khov4]. Oneshouldprobably look for a subcategory of O that categorifies the space of invariants in tensor products of the standard k− dimensional representation of sl and then study the associative algebra governing that subcategory. k Acknowledgements: Theauthorisgratefultohisadvisor,IgorFrenkelforhisencouragementandsupport throughout the development of this project. In addition, the author is very thankful to Mikhail Khovanov, Raphael Rouquier, Catharina Stroppel, and Gregg Zuckerman for helpful conversations and comments on preliminary versions of this work. 2. Representation Theory of sl k 2.1. Basic Definitions. We begin by reviewing the finite dimensional representation theory of sl . Recall k that sl is the Lie algebra of k×k matrices with entries in C. Denote the matrix with only 1 in the (i,j) k entry by e . There is the triangular decomposition sl =n−+h+n+ where n− is the subalgebra of lower i,j k triangular matrices, n+ is the subalgebra of upper triangular matrices and h is the Cartan subalgebra of diagonal matrices. The dual h∗ of the Cartan subalgebra has a basis α = e∗ −e∗ ,...,α = e∗ −e∗ ,...,α = 1 1,1 2,2 i i,i i+1,i+1 k−1 e∗ −e∗ where e∗ (e )=δ δ . These α are called the positive simple roots for sl . k−1,k−1 k,k i,j r,s i,r j,s i k A linear map λ: h→C may be written in coordinates λ=λ e∗ +···+λ e∗ . 1 1,1 k k,k The finite dimensional irreducible representations of sl are indexed by these weights λ. There exists a k unique irreducible representation of highest weight λ =λ e∗ +···+λ e∗ when λ ≥λ ≥···≥ λ . For 1 1,1 k k,k 1 2 k our purposes, one of the most important representations is the standard k− dimensional vector space Ck where sl acts on it naturally as matrices. Call this space V . The basis for this vector space is given by k k−1 e ,...,e and e (e )=δ e . Notice that the highest weight of this representation is e∗ . We denote the 1 k ij m j,m i 1,1 ith exterior power of a module V by ΛiV. Proposition 1. Let V be the module defined above. Then ΛiV is an irreducible module of highest k−1 k−1 weight e∗ +···+e∗ . 1,1 i,i Proof. See [FH] page 221. (cid:3) Now we define the algebra U (sl ) and its fundamental representations. q k Definition1. ThequantumgroupU (sl )istheassociativealgebraoverC(q)withgeneratorsE ,F ,K ,K−1 q k i i i i for i=1,...,k−1 satisfying the following conditions: (1) K K−1 =K−1K =1 i i i i (2) K K =K K i j j i (3) KiEj =qci,jEjKi (4) KiFj =q−ci,jFjKi (5) E F −F E =δ Ki−Ki−1 i j j i i,j q−q−1 (6) E E =E E if |i−j|>1 i j j i (7) F F =F F if |i−j|>1 i j j i (8) E2E −(q+q−1)E E E +E E2 =0 i i±1 i i±1 i i±1 i (9) F2F −(q+q−1)F F F +F F2 =0 i i±1 i i±1 i i±1 i where c =2 if j =i i,j −1 if j =i±1 0 if |i−j|>1. 3 ThemostbasicrepresentationofU(sl )isV .Itisthek-dimensionalvectorspacewithbasisv ,...,v . k k−1 0 k−1 The algebra acts on this space as follows: E v =0 if j 6=i−1 i j E v =v if j =i−1 i j i F v =0 if j 6=i i j F v =v if j =i i j i−1 K±1v =q±1v if j =i i j i K±1v =q∓1v if j =i−1 i j i−1 K±1v =v if j 6=i−1,i. i j j There are severalintertwinersbetween various representationsthat will be importantfor later. There is the map ΛkV →V ⊗Λk−1V given by k−1 k−1 k−1 k−1 vk−1∧···∧v0 →X(−1)jqk−j−1vj ⊗(vk−1∧···∧vˆj ∧···∧v0), j=0 where vˆ means that the term is omitted from the expression. j There is the map ΛkV →Λk−1V ⊗V given by k−1 k−1 k−1 k−1 vk−1∧···∧v0 →X(−1)k−1−jqj(vk−1∧···∧vˆj ∧···∧v0⊗vj). j=0 There is a map in the other direction: V ⊗Λk−1V →ΛkV given by k−1 k−1 k−1 v ⊗(v ∧···vˆ ∧···∧v )→(−1)k−j−1q−jv ∧···∧v . j k−1 j 0 k−1 0 There is also the map Λk−1V ⊗V →ΛkV given by k−1 k−1 k−1 (v ∧···vˆ ∧···∧v )⊗v →(−1)jqj−k+1v ∧···∧v . k−1 j 0 j k−1 0 The inclusion map Λ2V →V ⊗V is determined by k−1 k−1 k−1 v ∧v →v ⊗v −q1v ⊗v i j i j j i where i>j. The projection map V ⊗V →Λ2V is given by k−1 k−1 k−1 v ⊗v →q−1v ∧v if i>j i j i j −v ∧v if i<j. i j More generally, let π ⊗···⊗π : V⊗(r1+···+rt) →Λr1V ⊗···⊗ΛrtV r1 rt k−1 k−1 k−1 be the canonical projection map and i ⊗···⊗i : Λr1V ⊗···⊗ΛrtV →V⊗(r1+···+rt) r1 rt k−1 k−1 k−1 be the canonical inclusion map. Since we will not need these more general intertwiners, we will omit the precise formulas. 2.2. Graphical Calculus. There is a graphical description of the intertwiners between tensor products of fundamental representations via colored trivalent graphs. A line segment labeled by an integer i where 1≤ i≤k,willdepicttherepresentationΛiV .Notethatanedgelabeledbykdepictsthetrivialrepresentation k−1 and thus such an edge may be added or removed at will. The figure below depicts the canonical projection map ΛiV ⊗ΛjV →Λi+jV . k−1 k−1 k−1 The inclusion Λi+jV →ΛiV ⊗ΛjV is given by the diagram below. k−1 k−1 k−1 There are relations between these maps. Graphically, the relations are depicted by the following five diagrams. 4 Figure 1. Figure 2. Figure 3. Diagram Relation 1 Figure 4. Diagram Relation 2 Ifthetrivalentgraphhasexternaledgeslabeledonlybyk,thenitrepresentsacompositionofintertwiners from the the trivial representationto itself so a Laurent polynomial may be assigned to it. We will say that such a graph is closed. Theorem 1. There is a Laurent polynomial hDi in Z[q,q−1] which may be assigned to closed, colored k trivalent graphs D which satisfy the five relations above. It is invariant under ambient isotopy of R2. Proof. See sections 1 and 2 of [MOY]. (cid:3) The Laurent polynomial in the theorem above coincides with the representation-theoretic polynomial. A version of the HOMFLYPT for links may then be defined. Each crossing may be resolved in the following two ways: Let D ,D ,D be identical link diagrams except near a crossing as given by the figure below. + − 0 Definition 2. Let D be a planar projection of a link. Let (1) hD i =q1hD i −hD i , + k 0 k ∗ k (2) hD i =q−1hD i −hD i , − k 0 k ∗ k 5 Figure 5. Diagram Relation 3 Figure 6. Diagram Relation 4 Figure 7. Diagram Relation 5 Figure 8. (3) P (D)=q(k)(−w(D))hDi where w(D) is the difference between the number of positive and negative k k crossings. Theorem 2. (1) P (D) satisfies the Reidemeister moves and thus is an invariant of a link L which has k planar projection D. (2) The invariant above satisfies the skein relation for the one variable sl specialization of the HOM- k FLYPT polynomial: qkP (D )−q−kP (D )=(q1−q−1)P (D ). k + k − k 0 6 Figure 9. Proof. This is theorem 3.2 of [MOY]. (cid:3) 3. Categorification of sl Modules k The most important representations to categorify are the tensor powers V⊗n. When k =2, the categori- k−1 fications were constructed in [BFK]. The authors worked on the most singular blocks of category O which correspondto maximal subgroups of the symmetric group. They also suggested what to do for the sl case: k consider other blocks of category O. 3.1. Categorification of Λi1Vk−1⊗···⊗ΛirVk−1. Definition 3. (1) Denote the Grothendieck groupof an abelian or triangulated categoryC by [C]. It is the free abelian group generated by the symbols [M] where M is an object of C. The only relations in this group are of the form [N]=[M]+[P] when there is a short exact sequence or distinguished triangle of the form 0→M →N →P →0. (2) TheimageofanobjectM oranexactfunctorF intheGrothendieckgroupwillbe indicatedby[M] and [F] respectively. (3) Given an abelian category C, let Db(C) denote the corresponding bounded derived category. (4) Denote by Hj: Db(C)→C, the jth cohomologyfunctor. If F is a left exact functor, the jth derived functor RjF is defined to be Hj ◦RF. If F is right exact, the jth derived functor L F is defined to j be H−j ◦LF. The truncation functors τ≤n and τ≥n are well defined on the derived category. Proposition 2. Let X be an object in Db(C). The following triangles exist in the derived category: τ≤nX →X →τ≥n+1X τ≤n−1X →τ≤nX →Hn(X)[−n] Proof. See proposition 4.1.8 of [Sch]. (cid:3) Definition 4. Let O(gl ) be the category of gl modules which satisfy the following properties: n n (1) Finitely generated as U(gl )− modules. n (2) Diagonalizable under the action of the Cartan subalgebra h. (3) Locally finite under the action of the Borel subalgebra b=h+n+. This category decomposes into a direct sum of subcategories corresponding to the generalized central characters. Definition 5. (1) Let O(dk−1,dk−2,...,d0) = Od be the block of O(gln) for the central character corre- sponding to the weight k−1 di ie∗ −ρ, where Pi=0 Pj=1 d0+···+di−1+j n−1 n−3 1−n ρ= e + e +···+ e 1 2 n 2 2 2 is half the sum of the positive roots. (2) Let M(a ,...,a ) be the Verma module with highest weight a e +···+a e −ρ. 1 n 1 1 n n (3) Let L(a ,...,a ) be the Verma module with highest weight a e +···+a e −ρ. 1 n 1 1 n n There are d terms in the weight from the definition above with coefficient i. Note that k−1d =n. i Pj=0 j Proposition 3. Assume that the following direct sum is over all d such that the entries are non-negative integers and the sum of the entries is n. Then C⊗Z[⊕dOd]∼=V⊗n. k−1 7 Proof. The image of the Verma module [M(a ...,a )] gets mapped to v ⊗···⊗v . (cid:3) , n a1 an Thispropositionisthefirststeptowardscategorificationofsl −modules. Nextwewouldliketocategorify k the actionof the Lie algebra. The desired functors come directly from [BFK]. It is essentially the projective functoroftensoringwiththen−dimensionalrepresentationV .Oneonlyhastobecarefulaboutprojecting n−1 onto the various blocks. This is done next. Definition 6. (1) Let E : O →O be the functor defined by i (dk−1,dk−2,...,d0) (dk−1,...,di+1,di−1−1,...,d0) E M =proj (V ⊗M). i (dk−1,...,di+1,di−1−1,...,d0) n−1 (2) Let F : O →O by i (dk−1,dk−2,...,d0) (dk−1,...,di−1,di−1+1,...,d0) F M =proj (V∗ ⊗M). i (dk−1,...,di−1,di−1+1,...,d0) n−1 (3) Let H : O →O be Id⊕(di−di−1). i (dk−1,dk−2,...,d0) (dk−1,dk−2,...,d0) Theorem 3. Functorial Serre relations are satisfied on ⊕dO(gln). (1) H H ∼=H H . i j j i (2) If d >d then E F ∼=F E ⊕H . i i−1 i i i i i (3) If d =d then E F ∼=F E . i i−1 i i i i (4) If d <d then F E ∼=E F ⊕H . i i−1 i i i i i (5) If i6=j, then E F ∼=F E . i j j i (6) If d −d ≥0, then H E ∼=E H ⊕E⊕2. i i−1 i i i i i (7) If d −d ≤−2, then E H ∼=H E ⊕E⊕2. i i−1 i i i i i (8) If d −d =−1, then H E ∼=E H . i i−1 i i i i (9) If d −d ≥2, then F H ∼=H F ⊕F⊕2. i i−1 i i i i i (10) If d −d ≤0, then H F ∼=F H ⊕F⊕2. i i−1 i i i i i (11) If d −d =1, then H F ∼=F H . i i−1 i i i i (12) If |i−j|>1, then H E ∼=E H . i j j i (13) If |i−j|>1, then H F ∼=F H . i j j i (14) If |i−j|>1, then E E ∼=E E . i j j i (15) If |i−j|>1, then F F ∼=F F . i j j i (16) If j =i+1, then E E E ⊕E E E ∼=E E E ⊕E E E . i i j j i i i j i i j i (17) If j =i+1, then F F F ⊕F F F ∼=F F F ⊕F F F . i i j j i i i j i i j i (18) If d −d −1≥0, then H E ⊕E ∼=E H . i i−1 i i+1 i+1 i+1 i (19) If d −d ≤0, then E H ⊕E ∼=H E . i i−1 i+1 i i+1 i i+1 (20) If d −d −1≥0, then H E ⊕E ∼=E H . i i−1 i i−1 i−1 i−1 i (21) If d −d ≤0, then E H ⊕E ∼=H E . i i−1 i−1 i i−1 i i−1 (22) If d −d ≥0, then F H ⊕F ∼=H F . i i−1 i+1 i i+1 i i+1 (23) If d −d +1≥0, then H F ⊕F ∼=F H . i i−1 i i+1 i+1 i+1 i (24) If d −d ≥0, then F H ⊕F ∼=H F . i i−1 i−1 i i−1 i i−1 (25) If d −d +1≥0, then H F ⊕F ∼=F H . i i−1 i i−1 i−1 i−1 i OneonlyhastocomputeintheGrothendieckgrouptochecktheserelations[BG].Detailswillbeprovided later when a functorial version of the quantum Serre relations is proved. Our next goal is to categorify tensor products of exterior powers of V . The case k = 2 was addressed k−1 in [BFK]. When k=2, Λ2V is just a one dimensional vector space. The main ingredient in the categori- k−1 fication of V⊗n → V⊗(n−2) is the Zuckerman functor and parabolic subcategories. Thus for general k, we 1 1 expect a locally finite subcategory to categorify a tensor product of fundamental representations. We now give a definition of various parabolic subalgebras of sl and the corresponding locally finite subcategories k first introduced by A. Rocha-Caridi in [Ro]. Definition 7. (1) The subalgebra p is the parabolic subalgebra whose reductive subalgebra is (r1,...,rt) gl ⊕···⊕gl , where r +···+r =n. (2) Der1note by Odprtthe full su1bcategorytof Od of modules locally finite with respect to the subalgebra p. By abuse of notation, let O(r1,...,rt) be the category Op(r1,...,rt). d d 8 It is important to know some objects which are in these locally finite categories. Definition 8. (1) Let S denote the subset of simple roots defining the parabolic subalgebra p. (2) Let P+ ={λ∈h∗|hλ,αi∈N,∀α∈S}. p Given such a λ ∈ P+, we may define the generalized Verma module Mp(λ) = U(g)⊗ E(λ), where p U(p) E(λ) is the simple p− module with highest weight λ. Now it is easy to give conditions for which modules in these singular blocks are locally finite. Let λ = a e∗+···+a e∗. We need to give a condition for λ to be in this set. Since hρ,h i = 1, the condition 1 1 n n i that hλ−ρ,h i ≥ 0 simply becomes a > a . We now could state which simple modules and generalized i i i+1 Verma modules are in the locally finite subcategories. Lemma 1. (1) L(a ,...,a ) is a simple module in Op if a >a whenever α is a simple root of p. 1 n d i i+1 i (2) Mp(a ,...,a ) is a generalized Verma module in Op if a >a whenever α is a simple root of p. 1 n d i i+1 i Proof. This follows directly from above by evaluating at all simple roots α which define the parabolic i subalgebra p. (cid:3) We will often group coefficients in the highest weight of a generalized module such as Mp(a ,...,a ,a ,...,a ,...,a ) 1 i−1 i r n | {z } to stress that a >···>a so that it is locally finite with respect to a certain subalgebra p. i r Proposition 4. C⊗Z[⊕dOd(r1,...,rt)]∼=Λr1Vk−1⊗···⊗ΛrtVk−1. Proof. It suffices to show that this is an isomorphism on the basis of generalized Verma modules. Let p be the subalgebra given above. The isomorphism sends [Mp(a ,...,a )] to 1 n (v ∧···∧v )⊗···⊗(v ∧···∧v ). a1 r1 ar1+···+rt−1+1 ar1+···+rt This is clearly a bijection. (cid:3) Remark 1. If any of the r above is larger than k, then the category contains no non-trivial objects. i The Zuckerman functor plays an obvious role in this setup. It categorifies projection maps of these modules. Definition 9. (1) Let the Zuckerman functor Γ(r1,...,rt): Od → Od(r1,...,rt) be the functor of taking the maximal locally p finite submodule. (r1,...,rt) (2) Let the dual Zuckerman Z(r1,...,rt): Od → Od(r1,...,rt) be the functor of taking the maximal locally p finite quotient. (r1,...,rt) (3) Let ǫ(r1,...,rt): Od(r1,...,rt) →Od be the natural inclusion functor. Remark 2. The dual Zuckerman functor is also known as the Bernstein functor. See [KV]. Ifthereisaninclusionofparabolicsubalgebrasq⊂p,thereisanobviousgeneralizationofthesedefinitions forcategorieslocallyfinite withrespectto these algebras. Forexample,one maytakea module locallyfinite with respect to p and apply the Zuckerman functor Γp. q The Zuckerman functor is left exact. One usually studies its right derived functor and its cohomology functors. Taking the derived functor is important in categorification so that it becomes exact as a functor on the derived category. We denote its right derived functor shifted by j by RΓ[j]. The dual Zuckerman functoris rightexactandsimilarly,oneshouldconsideritsleft derivedfunctor. Onthe Grothendieckgroup, LZ and RΓ descend to the canonical map from Vk⊗−n1 to Λr1Vk−1⊗···⊗ΛrtVk−1. Proposition 5. (1) [LZ(r1,...,rt)]=πr1 ⊗···⊗πrt. (2) [ǫ [−Σt ir(ir−1)]]=i ⊗···⊗i . (r1,...,rt) r=1 2 r1 rt Proof. By proposition 5.5 of [ES], one may easily prove the first part by computing on the basis of Verma modules. The second part follows from the generalized BGG resolution. (cid:3) 9 Let d be the codimension of q in p. Lemma2. ThederivedZuckermanfunctorRΓp andtheinclusionfunctorǫp satisfythefollowingadjointness q q properties in the derived category: (1) Hom(ǫqpX,Y)∼=Hom(X,RΓpqY) (2) Hom(X,ǫqp[2d]Y)∼=Hom(RΓpqX,Y). Proof. See the proof of theorem 5 in [BFK]. (cid:3) Lemma 3. The derived dual Zuckerman functor LZp and the inclusion functor ǫp satisfy the following q q adjointness properties in the derived category: (1) Hom(ǫqp[−2d]X,Y)∼=Hom(X,LZqpY) (2) Hom(X,ǫqpY)∼=Hom(LZqpX,Y). Corollary 1. The derived Zuckerman and dual Zuckerman functors are related by RΓpq[2d]∼=LZqp. Proof. By the previous two lemmas these functors are adjoints of the same functor so they are the same up to isomorphism. (cid:3) 3.2. Equivalences of Categories. We look to generalize section 3.2.2 of [BFK] to the subcategories with generalizedcentralcharacterconsideredhere. Inordertomakethe notationmorecompact,wewillhavethe following notation for several important subalgebras. Definition 10. Let (1) p to be the parabolic subalgebra corresponding to (1,...,1,k,1,...,1). i | i{−z1 } (2) q to be the parabolic subalgebra corresponding to (1,...,1,k−1,1,...,1). i | {iz } (3) r be the parabolic subalgebra corresponding to (1,...,1,k−2,1,...1). i | i{−z1 } (4) s be the parabolic subalgebra corresponding to (1,...,1,2,1,...1). i | i{−z1 } (5) t be the parabolic subalgebra corresponding to (1,...,1,3,1,...1). i | i{−z1 } Clearly p ⊃ q and p ⊃ q . Recall from earlier the definitions of the functors ǫqj, ǫqj−1, LZpj, and j j j j−1 pj pj qj LZpj . Note that the derived Zuckerman functors in this setup could have non-zero cohomology functors qj−1 onlyindegrees0through2(k-1). Asin[BFK],themiddlecohomologyfunctorplaysasignificantrole. Given ageneralizedVermamoduleMpj(α),wewouldliketocomputeitsimageunderthefunctorL(k−1)Zqpjj+1◦ǫqpjj. The main result of this subsection is there are equivalences of categories: Opj ∼=Opj+1 . (dk−1,...,d0) (dk−1,...,d0) The equivalences are compositions of inclusions into larger categories and the middle cohomology of the derived Zuckerman functor. The plan of the proof is exactly that from [BFK]. First the action of these functors on generalizedVerma modules is computed. Then after some exactness andadjointness statements are proved, the equivalence will follow from lemma 2 of [BFK]. First we recall the generalized BGG resolution. Let S be a subset of simple roots defining a parabolic subalgebrap andW the correspondingWeyl group. Denote byWp the set of shortestcosetrepresentatives p in W/W and ρ half the sum of the positive roots of the subalgebra p. p S Theorem 4. Let µ be dominant integral. For all j =0,...,dimu, define CS =⊕ Mp(w(µ+ρ)−ρ ). j w∈Wp,l(w)=j S Then there is an exact sequence 0→CS →···→CS →L(µ+ρ)→0 dimu 0 10