Classification of Harish-Chandra modules over the W-algebra W(2, 2) ∗ 8 0 0 Dong Liu 2 Department of Mathematics, Huzhou Teachers College n a Zhejiang Huzhou, 313000, China J Linsheng Zhu 8 1 Department of Mathematics, Changshu Institute of Technology ] Jiangsu Changshu, 215500, China T R . h t a Abstract m [ In this paper, we classify all irreducible weight modules with finite dimen- 2 sional weight spaces over the W-algebra W(2,2). Meanwhile, all indecompos- v able modules with one dimensional weight spaces over the W-algebra W(2,2) 1 are also determined. 0 6 2 Keywords: W-algebra, irreducible weight modules . 1 0 Mathematics Subject Classification (2000): 17B56; 17B68. 8 0 : v i X 1 Introduction r a The W-algebra W(2,2) was introduced in [ZD] for the study of classification of vertex operatoralgebras generated by weight 2 vectors. By definition, the W-algebra W(2,2) is the Lie algebra L with C-basis {L ,I ,C,C |m ∈ Z} subject to the m m 1 following relations. Definition 1.1. The W-algebra L = W(2,2) is a Lie algebra over C (the field of complex numbers) with the basis {x ,I(n),C,C |n ∈ Z} n 1 and the Lie bracket given by n3 −n [x ,x ] = (m−n)x +δ C, (1.1) n m n+m n,−m 12 ∗ This work is supported in part by NSF of China (No. 10671027,10701019and 10571119). 1 n3 −n [x ,I(m)] = (m−n)I(n+m)+δ C , (1.2) n n,−m 1 12 [I(n),I(m)] = 0, (1.3) [L,C] = [L,C ] = 0. (1.4) 1 The W-algebra W(2,2) can be realized from the semi-product of the Virasoro algebra Vir and the Vir-module A of the intermediate series in [OR]. In fact, 0,−1 let W = C{x | m ∈ Z} be the Witt algebra (non-central Virasoro algebra) and m V = C{I(m) | n ∈ Z} be a W-module with the action x ·I(n) = (n−m)I(m+n), m then W(2,2) is just the universal central extension of the Lie algebra W ⋉V (see [OR] and [GJP]). The W-algebra W(2,2) studied in [ZD] is the restriction for C = C of W(2,2) in our paper. 1 The W-algebra W(2,2) can be also realized from the so-called loop-Virasoro algebra (see [GLZ]). Let C[t,t−1] be the Laurents polynomial ring over C, then the loop-Virasoro algebra V˜L is the universal central extension of the loop algebra Vir⊗C[t1,t−1] and W(2,2) = V˜L/C[t2]. The W-algebra W(2,2) is an extension of the Virasoro algebra and is similar to the twisted Heisenberg-Virasoro algebra (see [ADKP]). However, unlike the case of the later, the action of I(0) in W(2,2) is not simisimple, so its representation theory is very different from that of the twisted Heisenberg-Virasoro algebra in a fundamental way. The W(2,2) and its highest weight modules enter the picture naturally during our discussion onL(1/2,0)⊗L(1/2,0).The W-algebraW(2,2)isanextension of the Virasoro algebra and also has a very goodhighest weight module theory (see Section 2). Its highest weight modules produce a new class of vertex operator algebras. Contrast to the Virasoro algebra case, this class of vertex operator algebras are always irrational. The present paper isdevoted to determining allirreducible weight modules with finite dimensional weight spaces over L from the motivations in [LZ] and [LJ]. More precisely we prove that there are two different classes of them. One class is formed by simple modules of intermediate series, whose weight spaces are all 1-dimensional; the other class consists of the highest(or lowest) weight modules. The paper is arranged as follows. In Section 2, we recall some notations and collect known facts about irreducible, indecomposable modules over the classical Vi- rasoro algebra. In Section 3, we determine all irreducible (indecomposable) weight modulesofintermediateseriesoverL,i.e., irreducible(indecomposable) weight mod- ules with all 1-dimensional weight spaces. In Section 4, we determine all irreducible uniformly bounded weight modules over L which turn out to be modules of interme- diate series. In Section 5, we obtain the main result of this paper: the classification of irreducible weight L-modules with finite dimensional weight space. As we men- tioned, they are irreducible highest, lowest weight modules, or irreducible modules of the intermediate series. 2 2 Basics In this section, we collect some known facts for later use. For any e ∈ C, it is clear that [x +neI(n), x +meI(m)] = (m−n)(x +(m+n)eI(n+m)), ∀n 6= −m, n m n+m n3 −n [x +neI(n), x +(−n)eI(−n)] = −2nx + C. n −n 0 12 So {x + enI(n),C|n ∈ Z} spans a subalgebra Vir[e] which is isomorphic to the n classical Virasoro algebra. In many cases, we shall simply write Vir[0] as Vir. Introduce a Z-grading on L by defining the degrees: deg x =deg I(n)=n and n deg C = 0. Set L = (Cx +CI(n)), L = (Cx +CI(n)), + n − n Xn≥0 Xn≤0 and L = Cx +CI(0)+CC. 0 0 AnL-moduleV is calledaweight moduleif V isthesum ofallits weight spaces. For a weight module V we define Supp(V) := {λ ∈ C | V 6= 0}, λ which is generally called the weight set (or the support) of V. A nontrivial weight L-module V is called a weight module of intermediate series if V is indecomposable and any weight spaces of V is one dimensional. A weight L-module V is called a highest (resp. lowest) weight module with highest weight (resp. highest weight) λ ∈ C, if there exists a nonzero weight vector v ∈ V such that λ 1) V is generated by v as L-module; 2) L v = 0 (resp. L v = 0). + − Remark. For a highest (lowest) vector v we always suppose that I v = c v for 0 0 some c ∈ C although the action of I is not semisimple. 0 0 Obviously, if M is an irreducible weight L-module, then there exists λ ∈ C such that Supp(M) ⊂ λ+Z. So M is a Z-graded module. If, in addition, all weight spaces M of a weight L-module M are finite dimen- λ sional, the module is called a Harish-Chandra module. Clearly a highest (lowest) weight module is a Harish-Chandra module. 3 Let U := U(L) be the universal enveloping algebra of L. For any λ,c ∈ C, let I(λ,c,c ,c ) be the left ideal of U generated by the elements 0 1 {x ,I(i) | i ∈ N} {x −λ·1,C −c·1,I −c ·1,C −c ·1}. i 0 0 0 1 1 [ Then the Verma module with the highest weight λ over L is defined as M(λ,c,c ,c ) := U/I(λ,c,c ,c ). 0 1 0 1 It is clear that M(λ,c,c ,c ) is a highest weight module over L and contains a 0 1 unique maximal submodule. Let V(λ,c,c ,c ) be the unique irreducible quotient of 0 1 M(λ,c,c ,c ). 0 1 The following result was given in [ZD]. Theorem 2.1. [ZD] The Verma module M(λ,c,c ,c ) is irreducible if and only if 0 1 m2−1c +2c 6= 0 for any nonzero integer m. 12 1 0 In the rest of this section, we recall some known facts about weight represen- tations of the classical Virasoro algebra which can be considered as a subalgebra of L: Vir := span{x ,C|n ∈ Z}. n Kaplansky-Santharoubane [KS] in 1983 gave a classification of Vir-modules of the intermediate series. There are three families of indecomposable modules of the intermediate series (i.e nontrivial indecomposable weight modules with each weight space is at most one-dimensional) over the Virasoro algebra. They are Vir-modules ”without central charges”. (1) A = Cv : x v = (a+i+bm)v ; a,b i∈Z i m i m+i P (2) A(a) = Cv : x v = (i+m)v if i 6= 0, x v = m(m+a)v ; i∈Z i m i m+i m 0 m P (3) B(a) = Cv : x v = iv if i 6= −m, x v = −m(m + a)v , for i∈Z i m i m+i m −m 0 some a,b ∈ C. P When a ∈/ Z or b 6= 0,1, it is well-known that the module A is simple. In a,b the opposite case the module contains two simple subquotients namely the trivial module and C[t,t−1]/C. Denote the nontrivial simple subquotients of A , A(a), a,b B(a) by A′ , A(a)′, B(a)′ respectively. They are all Harish-Chandra modules of a,b the intermediate series over the Virasoro algebra. (These facts appear in many references, for example in [SZ]). We shall use T to denote the 1-dimensional trivial module, use V′(0,0) to denote the unique proper nontrivial submodule of V(0,1) (which is irreducible). An indecomposable module V over Vir is said to be an extension of the Vir- module W by the Vir-module W if V has a submodule isomorphic to W and 1 2 1 V/W ≃ W . 1 2 Theorem 2.2. [MP2] Let Z be an indecomposable weight Vir-module with weight spaces of dimension less than or equal to two. Then Z is one of the following: 4 1) an intermediate series module; 2) an extension of A , A(a) or B(a) by themselves; α,β 3) an extension of A by A , where β −β = 2,3,4,5,6; α,β1 α,β2 1 2 4) an extension of A by W, where β = 2,3,4,5 and W is one of: A , A , 0,β 0,0 0,1 A′ , A′ ⊕T, T, A(a) or B(a); 0,0 0,0 5) an extension of T by A or A(a), ; 0,0 6) an extension of A′ by A or A(1); 0,0 0,0 7) an extension A by A or B(0); 0,1 0,0 8) an extension of A by A or A(a); 0,0 0,1 9) an extension of A(0) by A ; 0,0 10) an extension of B(a) by A ; 0,1 11) the contragredient extensions of the previous ones; where α,β,β ,β ,a ∈ C. 1 2 Remark. In the above list, there are some repetitions, and not all of them can occur. Theorem 2.3. [MP2] Thereare exactlytwo indecomposableextensionsV = span{v , α+i v′ |i ∈ Z} of A by A (α ∈/ Z) given by the actions x v = (α+n)v ,∀i ∈ α+i α,0 α,0 i α+n α+n+i Z and a) x v′ = (α+n)v′ −iv , for all i,n ∈ Z; or i α+n α+n+1 α+n+1 b) x v′ = (α+ n)v′ , x v′ = (α +n)v′ , x v′ = (α + n)v′ + 1 α+n α+n+1 −1 α+n α+n−1 2 α+n α+n+2 1 v , x v′ = (α+n)v′ − 1 v ,where{v ,v′ } (α+n+2)(α+n+1) α+n+2 −2 α+n α+n−2 (α+n−2)(α+n−1) α+n−2 α+n α+n forms a base of V for all n ∈ Z. α+n We also need the following result from [MP1]. Theorem 2.4. [MP1] Suppose that V is a weight Vir-module with finite dimensional weight spaces. Let M+ (resp. M− ) be the maximal submodule of V with upper (resp. lower) bounded weights. If V and V* (the contragredient module of V) do not contain trivial submodules, there exists a unique bounded submodule B such that V = B M+ M−. L L 3 Irreducible weight modules with weight multi- plicity one In this section we determine all irreducible and indecomposable weight modules over L with weight multiplicity one. 5 Let V = ⊕Cv be a Z-graded L-module. Then the action of I(0) is semisimple i and we can suppose that I(0)v = λ v for some λ ∈ C. Moreover, by [I(0),x ] = i i i i m mI(m) we have mI(m)v = (λ −λ )x v . Hence V is an irreducible (indecom- i m+i i m i posable) L-module if and only if V is an irreducible (indecomposable) Vir-module. Denoted A , A(a,0), B(a,0) by the L-module induced from the Vir-module a,b,0 A , A(a), B(a) with the trivial actions of I(n) for any n ∈ Z, respectively. More- a,b over we also denote the nontrivial simple subquotients of A , A(a), B(a) by A′ , a,b a,b A′(a), B′(a) respectively. Clearly, A′(a,0) ∼= B′(a,0) ∼= A′ Now we shall prove 0,0,0 that the above three kinds modules are all indecomposable modules with weight multiplicity one. Lemma 3.1. Let V = Cv be an Z-graded L-module such that x v = (a + i∈Z i n i i+ bn)vi+n for all n,i ∈PZ and some a,b ∈ C. If a+ bn 6= 0 for any n ∈ Z, then I(m)v = 0. i Proof. Since V is a module of the Virasoro algebra Vir= ⊕m∈ZL(m), it is clear that C = 0 (cf. [SZ] for example). Suppose that I(n)v = f(n,t)v for all n,t ∈ Z. t n+t From [L(n),I(m)] = (m−n)I(n+m)+ 1 δ (n3 −n)C , we see that 12 n+m,0 1 f(m,t)L(n)v −f(m,n+t)(a+t+bn)v t+m n+t+m 1 = (m−n)f(n+m,t)v + δ (n3 −n)C v . (3.1) n+m+t n+m,0 1 t 12 In this case, 1 f(m,t)(a+t+m+bn)−f(m,n+t)(a+t+bn) = (m−n)f(n+m,t)+ δ (n3−n)C . n+m,0 1 12 (3.2) Let t = 0 in (3.2), then 1 f(m,n)(a+bn) = (a+m+bn)f(m,0)−((m−n)f(n+m,0)+ δ (n3−n)C ). n+m,0 1 12 (3.3) Let m = n in (3.2), then f(n,t)(a+(b+1)n+t)−f(n,n+t)(a+bn+t) = 0. (3.4) Let t = 0 in (3.4), then f(n,0)(a+(b+1)n)−f(n,n)(a+bn) = 0. (3.5) Let t = −n in (3.4), then f(n,−n)(a+bn) = f(n,0)(a+(b−1)n). (3.6) Setting n = −m in (3.3), then 1 (a−mb)f(m,−m) = (a+m−bm)f(m,0)−2mf(0,0)− (m3 −m)C . (3.7) 1 12 6 From (3.6) and (3.7) we obtain that a+bm 1 f(m,0) = (f(0,0)− (m2 −1)C ), m 6= 0. (3.8) 1 a 24 Setting t = n and m = 0 in (3.2), we have (f(0,2n)−f(0,n))(a+(b+1)n) = nf(n,n). (3.9) Let m = 0 in (3.3), then f(0,n)(a+bn) = (a+bn)f(0,0)+nf(n,0). (3.10) From (3.10) and (3.8) we have a+n 1 f(0,n) = f(0,0)− (n3 −n)C . (3.11) 1 a 24a Applying (3.11) to (3.9), we have a+(b+1)n 1 f(n,n) = (f(0,0)− (7n2 −1)C ). (3.12) 1 a 24 Combining (3.5) and (3.11) we obtain C = 0. 1 Therefore a+bm f(m,0) = f(0,0), a for all m ∈ Z. By (3.3) we obtain that a+bm+n f(m,n) = f(0,0) = c(a+bm+n), (3.13) a for all m,n ∈ Z and some c ∈ F. However, by [I(m),I(n)] = 0 we have c = 0. So f(m,n) = 0 for any m,n ∈ Z. Remark. In the following cases, we can also deduce that C = C = 0 as in Lemma 1 1 7.3. So in the following discussions, we always assume that C = C = 0. 1 1 Lemma 3.2. If V ≃ A as Vir-module, then V ≃ A as L-module. α,β α,β,0 Proof. (I.1) Suppose that a 6∈ Z and a+bn 6= 0 for all n ∈ Z, then f(m,n) = 0 for all m,n ∈ Z by Lemma 3.1. (I.2) a 6∈ Z and a = bp for some p ∈ Z\{0}. So b 6= 0,1. Therefore f(m,n) = 0 if n+p 6= 0 by (3.4). It follows from (3.4) that f(0,−p) = 0. (3.14) 7 Setting i = −p, m = 0 in (3.2) and using (3.14, we have f(n,−p) = 0, n 6= 0. Therefore f(m,n) = 0 for all m,n ∈ Z. (I.3) a ∈ Z. Since A ∼= A , so we can suppose that x v = (i+bn)v for a,b 0,b n i n+i all n,i ∈ Z. (I.3.1) b 6= 0,1, then a+bn 6= 0 for all n 6= 0. So f(m,n) = 0 by Lemma 3.1. Therefore V is isomorphic to A . 0,b,0 (I.3.2) b = 1. In this case (3.3) still holds and (3.2) and (3.3) becomes f(m,i)(i+m+n)−f(m,n+i)(i+n) = (m−n)f(n+m,i) (3.15) and nf(m,n) = (n−m)f(n+m,0)+(n+m)f(m,0) (3.16) Let m = i = 0 in (3.15) we have f(0,n) = f(n,0)+f(0,0), n 6= 0. (3.17) Replacing t by n and letting n = −m in (3.15), we have nf(m,n) = (n−m)f(m,n−m)+2mf(0,n). (3.18) Replacing n by n−m in (3.16), we have (n−m)f(m,n−m) = (n−2m)f(n,0)+nf(m,0). (3.19) From (3.16)-(3.19), we have (n−m)f(n+m,0) = nf(n,0)−mf(m,0)+2mf(0,0). (3.20) Let n = 0 in (3.20), we have f(0,0) = 0. So (3.20) becomes (n−m)f(n+m,0) = nf(n,0)−mf(m,0). We deduce that there exists c,d ∈ C such that f(n,0) = c+dn, (3.21) for all n ∈ Z and n 6= 0. Applying (3.20) to (3.16) and using (3.21), we have f(m,n) = f(0,0), n 6= 0. Therefore f(m,n) = 2d+c(m+n), n 6= 0. By (3.15) we have d = 0. 8 Therefore f(m,n) = c(m+n). By [I(m),I(n)] = 0 we have c = 0. So f(m,n) = 0 for any m,n ∈ Z. (I.3.3) b = 0. (3.2) becomes (i+m)f(m,i)−tf(m,n+i) = (m−n)f(m+n,i). (3.22) Let i = 0 and n = −m in (3.22), then f(m,0) = f(0,0). Let i = 0 and n = m in (3.22), then f(m,0) = 0,m 6= 0. Hence f(n,0) = 0,∀n ∈ Z. (3.23) Setting i = −n in (3.22) and using (3.23), we have f(m,−n) = f(m+n,−n), m 6= n. (3.24) Let i = 1 in (3.22) then f(m,n+1) = (m+1)f(m,1)−(m−n)f(m+n,1). (3.25) Setting n = −1 in (3.25) and using (3.23), we have f(m,1) = f(m−1,1),m 6= −1. (3.26) So f(m,1) = c for some c ∈ F and for any m ≥ −1. f(m,1) = c for some c ∈ F 1 1 2 2 and for any m ≤ −2. Hence (3.25) becomes f(m,n) = c n, m ≥ −1; f(m,n) = c n, m ≤ −2. (3.27) 1 2 By (3.24) we have c = c = c for some c ∈ F. 1 2 1 Therefore f(m,n) = nc for any m,n ∈ Z. By [I(m),I(n)] = 0 we have c = 0. So f(m,n) = 0 for any m,n ∈ Z. Lemma 3.3. If V ∼= A(α) or B(α) as Vir-module, then V ∼= A(α,0) or B(α,0). Proof. If V ∼= A(α), then x v = (i+n)v if t 6= 0, x v = n(n+a)v for some n i n+i n 0 n ∼ a ∈ C. We can deduce that f(m,i) = 0 for all n,i ∈ Z. Therefore V = A(a,0). ∼ If V = A(α), then x v = tv if t 6= −n, x v = −n(n + a)v , for some n i n+i n −n 0 a ∈ F. We can deduce that f(m,i) = 0 for all n,i ∈ Z. Then V is isomorphic to B(a,0). From Lemma 3.2 and Lemma 3.3 we have 9 Theorem 3.4. Suppose that V is a nontrivial irreducible weight L-module with weight multiplicity one. Then we have V ≃ A or V ≃ A′ for some α,β ∈ α,β,0 0,0,0 C. Meanwhile, the three kinds modules listed in the before of Lemma 3.1 are all indecomposable weight L-module with weight multiplicity one. 4 Uniformly bounded irreducible weight modules In this section, we assume that V is a uniformly bounded nontrivial irreducible weight module over L. So there exists α ∈ C such that Supp(V) ⊂ α + Z. From representation theory of Vir, we have C = 0 and dimV = p for all α+n 6= 0. If α+n α ∈ Z, we also assume that α = 0. Consider V as a Vir-module. We have a Vir-submodule filtration 0 = W(0) ⊂ W(1) ⊂ W(2) ⊂ ··· ⊂ W(p) = V, where W(1),···,W(p) are Vir-submodules of V, and the quotient modules W(i)/W(i−1) have weight multiplicity one for all nonzero weights. Choose v1,···,vp ∈ V such that the images of vi + W(i−1) form a basis of n n α+n n (W(i)/W(i−1)) for all α+n 6= 0. We may suppose that α+n x (v1,···,vp) = (x v1,···,x vp) = (v1 ,···,vp )A , i n n i n i n n+i n+i i,n where A are upper triangular p×p matrices, and A (j,j) = α+n+iβ . Denote i,n i,n j I(i)(v1,···,vp) = (v1 ,···,vp )F , (4.1) n n n+i n+i i,n where F are p×p matrices. i,n The Lie brackets give F F −F F = 0, (4.2) i,j+n j,n j,i+n i,n A A −A A = (j −i)A , (4.3) i,j+n j,n j,i+n i,n i+j,n 1 A F −F A = (j −i)F + δ (i3 −i)C I , (4.4) i,j+n j,n j,i+n i,n i+j,n i,−j 1 p 12 where the last three formulas have the restriction (α+n)(α+n+i)(α+n+j)(α+ n+i+j) 6= 0. We shall denote the (i,j)-entry of a matrix A by A(i,j). Lemma 4.1. If all nontrivial irreducible sub-quotient Vir-modules of V are isomor- phic to A′ , then V ≃ A′ . 0,0 0,0,0 Proof. Nowwe cansuppose thatdim(W(1)) ≤ 1 (IfV containsatrivialsubmodule 0 Cv , then the span{u1 = I(k)v |k ∈ Z} is a Vir-submodule, which can be chosen as 0 k 0 W(1)). Claim. The (k,1)-entry F (k,1) = 0 for all k ≥ 2, n 6= 0 and j +n 6= 0. j,n 10