ebook img

On modules for double affine Lie algebras PDF

0.2 MB·English
Save to my drive
Quick download
Download
Most books are stored in the elastic cloud where traffic is expensive. For this reason, we have a limit on daily download.

Preview On modules for double affine Lie algebras

5 1 0 On modules for double affine Lie algebras 2 n Naihuan Jing and Chunhua Wang* a J 8 Abstract. ImaginaryVermamodules,parabolicimaginaryVermamod- 2 ules, and Vermamodules at levelzerofor doubleaffineLie algebras are constructed using three different triangular decompositions. Their re- ] lations are investigated, and several results are generalized from the T affine Lie algebras. In particular, imaginary highest weight modules, R integrable modules, and irreducibility criterion are also studied. . h t a m 1. Introduction [ Let ˆg be the untwisted affine Lie algebra associated to a complex finite 1 v dimensional simple Lie algebra g with Cartan subalgebra h. The highest 7 weight irreducible ˆg-module L(λ) of the highest weight λ can be studied 8 with the help of the Verma module V(λ), which is an induced moduleof the 1 7 one-dimensional module C1λ of the Borel subalgebra bˆ = hˆ+nˆ+ such that 0 nˆ 1 = 0. If one partitions the affine root system using the loop realization + λ . 1 of ˆg, the associated imaginary Verma module [8] behaves quite differently. 0 For example, the new Verma module can have infinite dimensional weight 5 subspaces. 1 : Double affine Lie algebras are certain central extensions of maps from v a 2-dimensional torus to the Lie algebra g. They are analogous to ˆg but i X with two centers, and first appeared in I. Frenkel’s work [7] on affinization r of Kac-Moody algebras. Moody and Shi [12] have shown that the root a systems have different properties from those of the affine root systems. For example, some roots can not be spanned positively or negatively by the “simple” ones. Thus a usual highest weight module would blow up beyond control. In this paper, we generalize imaginary Verma modules (IVM) to double affine Lie algebras and use them to produce irreducible modules. Tostudytheseimaginary Vermamodules,weconsidergeneralized imag- inary Verma modules, which bare some similarity to the parabolic Verma modules in classical Lie theory. The structure of IVM’s is characterized by 2010MathematicsSubjectClassification. Primary: 17B67;Secondary: 17B10,17B65. Key words and phrases. double affine Lie algebras, Verma modules, irreducibility, Weylmodules. *Corresponding author. 1 2 NAIHUANJINGANDCHUNHUAWANG* the generalized IVM’s, and we show that they are irreducible if the second center is nonzero. When the second center is zero, they are similar to the evaluation modules of the affine Lie algebras, for which we also generalize several results from the Tits-Kantor-Koecher algebras [1]. To further understand the situation of trivial centers, we adopt Chari and Pressley’s technique [4, 3] of Weyl modules to modules of double affine Lie algebras, which has played an important role for loop algebras (see [2] forasurvey). Weremarkthatthetriangular decompositionemployed inour case is differentfrom that used in previous work on toroidal Lie algebras (cf. [9, 14, 15]). In our modules, the Borel subalgebras are defined by carving out the second imaginary part to control the growth of the IVM’s. 2. Double affine Lie algebras Let g be a complex finite dimensional simple Lie algebra of simply laced type with h its Cartan subalgebra. Let ∆ be the root system generated by the simple roots α (i = 1,...,s) and let α∨ ∈ h (i = 1,...,s) be the i i corresponding simple coroots such that hα∨,α i = a , the entries of the j i ij s Cartan matrix of g. Denote by θ = k α the longest root of g, and θ∨ the P i i i=1 corresponding dual element in h. For any positive root α = c α ∈ ∆, Pi i i s we denote its height by ht(α) = c . Pi=1 i The double affine Lie algebra T is the central extension of the 2-loop algebra defined by T = g⊗C[t1,t−11,t2,t−21]⊕Cc1⊕Cc2, whereC[t1,t−11,t2,t−21] is the ring of Laurent polynomials in two commuting variables t and t . Writing x⊗tmtn as x(m,n), the Lie bracket is given by 1 2 1 2 (2.1) [x(m ,n ),y(m ,n )]= [x,y](m +m ,n +n ) 1 1 2 2 1 2 1 2 +(x |x )δ δ (m c +m c ), 1 2 m1,−n1 m2,−n2 1 1 2 2 (2.2) [c ,x(m ,n )]= [c ,x(m ,n )] = 0, 1 1 1 2 1 1 where (m1,m2),(n1,n2) ∈ Z2, x1,x2,x ∈ g, and (x1|x2) is the g-invariant bilinear form. Adjoining the derivations, we define the extended double affine Lie algebra T as T = g⊗C[t1,t−11,t2,t−21]⊕Cc1⊕Cc2⊕Cd1⊕Cd2. The derivations act on T via (2.3) [d ,x(m ,m )]= m x(m ,m ), [d ,c ] = 0, i,j = 1,2. i 1 2 i 1 2 i j Let hˆ = h⊕Cc1⊕Cc2⊕Cd1⊕Cd2 be the Cartan subalgebra of T, and hˆ∗ be the dual space. For β ∈ hˆ∗ the root subspace T = {x ∈ T|[h,x] = β β(h)x, ∀h ∈ hˆ} is defined in the usual way. We then define the root system ∆ to be the set of all nonzero β ∈ hˆ∗ such that T 6= 0. It is known that T β ON MODULES FOR DOUBLE AFFINE LIE ALGEBRAS 3 [12] the root system is different from the usual root system in that a root is no longer a positive or negative sum of the “simple” roots. Tobespecific,weletδ ∈hˆ∗ suchthatδ (h) = δ (c )= 0,andδ (d )= δ i i i j i j ij for i,j = 1,2. Then the extended root system ∆T = {α+Zδ1+Zδ2|α ∈ ∆}∪({Zδ1 +Zδ2}\{0}), where the first and second subsets form the real and imaginary roots, de- noted by ∆re and ∆im respectively. The corresponding root subspaces are T T T =h⊗tmtn, (m,n) 6= (0,0), mδ1+nδ2 1 2 Tα+mδ1+nδ2 =gα⊗tm1 tn2, (m,n) ∈Z×Z. Then one has the root space decomposition T = hˆ⊕ M Tβ. β∈∆T Obviously the extended double affine Lie algebra T contains two affine Lie algebras as subalgebras: ˆgi = g⊗C[ti,t−i 1]⊕Cci⊕Cdi, i= 1,2. Recall that [11] α = δ −θ,α ,...,α are the simple roots of gˆ . Similarly, 0 1 1 s 1 the roots α = δ −θ,α ,...,α are simple roots for ˆg . Following [12] we −1 2 1 s 2 call the elements α ,α ,α ,...,α the “simple” roots of T. However, not −1 0 1 s all roots can be represented as non-negative or non-positive linear combina- tion of the simple roots. Thecorrespondingcoroots are α∨ ,α∨,α∨,...,α∨, −1 0 1 s where α∨ = c −θ∨ ⊗1 and α∨ = c −θ∨ ⊗1. Consequently the Cartan −1 2 0 1 subalgebra hˆ is spanned by α∨ ,α∨,α∨,...,α∨,d ,d . −1 0 1 s 1 2 3. Imaginary Verma modules of T Verma modules of affine Lie algebras are defined with the help of a tri- angular decomposition, which is constructed by a choice of the base in the root system. Since the root system for the extended double affine algebras cannotbedividedintopositiveandnegative rootsintheusualsense,weuse a closed subset to partition the root system. For affine Lie algebras Futorny [8] studied a new class of Verma modules called the imaginary Verma mod- ules (IVM) [5, 6] associated with a closed subset defined by a function. For finite dimensional simple Lie algebras, the partition derived from the closed subset coincides with the usual partition of positive and negative roots. In this section, we introduce the imaginary Verma modules for the extended double affine Lie algebras and derive their important properties. s 3.1. Imaginary Verma modules. Set ϕ = α∗ −ht(θ)α∗ , where P i −1 i=0 α∗ ∈ hˆ such that α∗(α ) = δ (i,j = −1,0,...,s). Clearly, ϕ(α ) = 1 for i i j ij i i = 0,1,...,s, ϕ(δ ) = 0, and ϕ(δ ) = 1+ht(θ). Let I = {α ∈ ∆ |ϕ(α) > 2 1 T 0}∪Nδ2. Clearly it is closed under the addition. 4 NAIHUANJINGANDCHUNHUAWANG* Recall [11] that the root system of ˆg1 is ∆ˆg1 = {α + Zδ1|α ∈ ∆} ∪ ({Zδ1}\{0}), anditspositiverootsare∆ˆg1+ = {±α+Nδ1|α ∈ ∆}∪∆+∪Nδ1. Note that for any α ∈ ∆ , ϕ(α) = ht(α). Then I can be written as I = ˆg1+ {α+Zδ2|α ∈ ∆ˆg1+}∪Nδ2. Subsequently I∪(−I)=∆T and I∩(−I)= ∅. Denote by Q(I) the Z-span of I. Let T = ⊕ T and T = ⊕ T , then I β∈I β −I β∈−I β (3.1) T = T ⊕hˆ⊕T −I I is a triangular decomposition of T associated with the closed subset I. The Poincar´e-Birkhoff-Witt theorem implies that the universal enveloping alge- bra U(T) = U(T )⊗U(hˆ)⊗U(T ). −I I Let b = hˆ⊕T be the imaginary Borel subalgebra, which is a solvable I I Lie algebra. A vector space V is called a weight module of T if V = M Vµ, µ∈ˆh∗ where V = {v ∈ V|hv = µ(h)v, ∀h ∈ hˆ}. Set P(V) = {µ ∈ hˆ∗|V 6= 0}. µ µ We say that λ ≥ µ (λ,µ ∈ hˆ∗) with respect to ϕ if λ−µ is a non-negative linear combination of roots in I. For simplicity we will omit the reference to ϕ if no confusion arises from the context. For example, both T and U(T) are weight modules for T. Definition 3.1. Let λ ∈ hˆ∗. A nonzero vector v is called an imaginary highest vector with weight λ if T .v = 0 and h.v = λ(h)v for all h ∈ hˆ. If I V = U(T)v then V is called a highest weight module of highest weight λ. If V is a highest weight module of weight λ, then (3.2) V = U(T−I)v = M Vλ−η, η∈Q(I)+ where Q(I)+ = Z+-span of I. For λ ∈ hˆ∗, let C1λ be the one-dimensional bI−module defined by (x+ h).1 = λ(h) · 1 (x ∈ T , h ∈ ˆh). The imaginary Verma module is the λ λ I induced module M(λ) = U(T)⊗U(bI)C1λ. 3.2. Properties of IVM. Based on the theory of the standard Verma modules [11] and IVM’s for the affine Lie algebras [8], the following result can be proved similarly. Proposition 3.1. For any λ ∈hˆ∗, one has that (1) M(λ) is a U(T )−free module generated by the imaginary highest −I vector 1⊗1 of weight λ. λ (2) (a) dimM(λ) = 1; λ (b) 0 < dimM(λ) < ∞ for every positive integer k; λ−kδ2 ON MODULES FOR DOUBLE AFFINE LIE ALGEBRAS 5 (c) If M(λ) 6= 0 and µ 6= λ−kδ for any nonnegative integer k, µ 2 then we have dimM(λ) = ∞. µ (3) Any imaginary highest weight T−module of highest weight λ is a quotient of M(λ). (4) M(λ) has a unique maximal submodule J. (5) If µ ∈ hˆ∗, then any nonzero homomorphism M(λ) → M(µ) is injective. We denote by L(λ) the irreducible quotient M(λ)/J. 3.3. Irreducibility criterion for IVM. Futorny [8] found that the affine imaginary Verma module is irreducible if and only if the center acts nontrivially. It turns out that a similar result can beobtained for M(λ) (see Theorem 3.5). Lemma 3.2. Let M = ∞ M(λ) . For any nonzero v ∈ M(λ), Lj=0 λ−jδ2 U(T)v∩M 6= 0. Proof. Write λ = λ−rδ , where λ is the component of ∆ . We can 2 ˆg1 assume that r is minimum so that v 6= 0, otherwise we can replace λ by λ λ′ and M(λ) = M(λ′) where λ ≡ λ′mod(Zδ2). Therefore ht−2nvλ 6= 0 for s any n ≥ 0. Assume that v ∈ M(λ)λ−µ, where µ = Pniαi +kδ2, ni ∈ Z+, i=0 s and k ∈ Z. Define the height ht(µ) = Pni and use induction on ht(µ). If i=0 ht(µ) = 0, the weight of v is λ−kδ2 for some k ∈ Z+ by assumption, so the result clearly holds. Let e ,f ,α∨(0 ≤ i ≤ s) be the Chevalley generators of the derived i i i affine Lie algebra ˆg . If ht(µ) > 0, there exists i ∈ {0,··· ,s} such that 1 0 e v 6= 0. In fact, if ht(µ) = 1, say v = f t−kv . If k < 0, then α∨tk−1v = i0 i 2 λ i 2 −2f t−1v 6= 0. Then let v′ = α∨tk−1v and e v′ = −2α∨t−1v 6= 0. The i 2 λ i 2 i i 2 λ case of ht(µ) ≥ 2 is treated similarly. Moreover, e (h⊗t−m).v′ 6= 0 for all i0 2 m ≥ 0. Hence T .v′ 6= 0, and any of its nonzero element has weight αi0−mδ2 λ−µ+α −(m+1)δ . As ht(µ−α )= ht(µ)−1, by inductive hypothesis i0 2 i0 we have U(T)(T .v′)∩M 6= 0. Since U(T)(T .v′) ⊂ U(T)v, it αi0−mδ2 αi0−mδ2 follows that U(T)v∩M 6= 0. (cid:3) Let M(λ)+ = {v ∈ M(λ)|T .v = 0} be the space of extremal vectors. I Clearly M(λ)+ is ˆh−invariant. For any nonzero element v ∈ M(λ)+, U(T).v generates a submodule of M(λ). The following result describes the form of extremal vectors. Corollary 3.3. M(λ)+ ⊂ M. Proof. Supposeonthecontrarythatthereexistsanonzerov ∈ M(λ)+∩ s s M(λ) such that µ = n α +kδ and ht(µ) = n > 0. Note that λ−µ P i i 2 P i i=0 i=0 6 NAIHUANJINGANDCHUNHUAWANG* U(T)v = U(T )v, so the weight of any homogeneous vector in U(T)v is −I λ−µ−ν for ν ∈ Q(I)+. As ht(µ) > 0, λ−µ−ν 6= λ(modZδ2). Hence U(T)v∩M = 0, which contradicts with Lemma 3.2 on M(λ) . (cid:3) λ−µ Define the Heisenberg subalgebra Hˆ2 = Ln∈Z×h⊗tn2 +Cc2, then the space M = ∞ M(λ) is a Verma modulefor Hˆ . Thefollowing result Lj=0 λ−jδ2 2 is well-known from Stone-Von Neumann’s theorem. Lemma 3.4. M is irreducible as a Verma module for Hˆ if and only if 2 λ(c )6= 0. 2 Theorem 3.5. The imaginary Verma module M(λ) is irreducible if and only if λ(c )6= 0. 2 Proof. Let v be the highest weight vector of M(λ). By definition λ M(λ)isirreducibleifandonlyifthespaceofextremalvectorsM(λ)+ = Cvλ. Suppose that M(λ) is irreducible but λ(c ) = 0. Since M is reducible 2 as an Hˆ -module by Lemma 3.4, there exists w 6= 0 with weight λ − kδ 2 2 (k > 0) such that T .w = 0 for any l > 0. It is clear that T .w = 0 lδ2 β for all β ∈ {α ∈ ∆ |ϕ(α) > 0}, because the weight of T .w is larger than T β λ. Thus w ∈/ Cvλ is also an imaginary highest vector in M(λ), which is a contradiction. So we must have λ(c ) 6= 0. 2 If λ(c ) 6= 0, Lemma 3.4 implies that the Hˆ −module M is irreducible. 2 2 Consider any nonzero submodule U(T)v, where v ∈ M(λ). Lemma 3.2 says that U(T)v ∩M 6= 0. Note that U(T)v ∩M is then a non-trivial Hˆ - 2 submoduleof M, consequently U(T)v∩M = M, thus M ⊂ U(T)v. Because v ∈ M, we have that U(T)v = M(λ), i.e., M(λ) is irreducible. (cid:3) λ 4. Generalized IVM M(λ,A) and highest weight modules In this section, we give a new class of modules M(λ,A) generalizing the previousIVM’s to studythestructureof IVM. Theyare similar to parabolic Verma modules. 4.1. Definition of M(λ,A). Let A ⊂ B = {0,1,2,...,s}, the index set of ˆg . We denote A∗ = A\{0} and B∗ = B\{0}. Define f = α∗− 1 A P i i∈B\A ( k )α∗ for A =6 B and f = 0, then f (δ ) = 0. Set Q(A) = {α ∈ P i −1 B A 2 i∈B∗\A∗ ∆ |f (α) ≥ 0} which is closed under addition. Then T A Q(A)= I ∪{−Xliαi+Zδ2|li ≥ 0,Yli 6= 0}∪{−Nδ2}. i∈A i Note that I $ Q(A). The following result is clear. Proposition 4.1. Q(A)∩(−Q(A))= {P Zαi +Zδ2}∩∆T. i∈A ON MODULES FOR DOUBLE AFFINE LIE ALGEBRAS 7 RecallthattheCartanmatrixisgivenbyhα∨,α i= a fori = −1,0,1,...,s, j i ij j = 0,1,...,s. Let hˆ ⊂ hˆ be the space spanned by α∨ ( i ∈ A). Consider A i the subspaces T and T , where Q(A) = Q(A)\(−Q(A)). Then T Q(A) −Q(A) decomposes itself as: T = T ⊕hˆ⊕T . −Q(A) Q(A) Let λ ∈ hˆ∗ such that λ(hˆA ⊕Cc2) = 0. Let C1λ be the one-dimensional hˆ ⊕ T −module such that (x + h).1 = λ(h) · 1 (x ∈ T , h ∈ hˆ). Q(A) λ λ Q(A) Define the induced T-module associated with I, A and λ: M(λ,A) = U(T)⊗U(ˆh⊕TQ(A))C1λ. 4.2. Properties of M(λ,A). The following result is similar to Propo- sition 3.1. Proposition 4.2. For any λ ∈ hˆ∗ such that λ(hˆA ⊕Cc2) = 0, one has that (1) M(λ,A) is a U(T )−free module generated by 1⊗1 . −Q(A) λ (2) dimM(λ,A)µ = 0,1 for µ = λ − P kiαi − αj + kδ2, ki ∈ Z+, i∈A k ∈Z, and j ∈ B\A. Otherwise, dimM(λ,A)µ = ∞. (3) The T−module M(λ,A) has a unique irreducible quotient L(λ,A). (4) Let A′′ ⊂ A. Then there exists a chain of surjective homomor- phisms M(λ) → M(λ,A′′) → M(λ,A). (5) Letλ,µ ∈ hˆ∗. TheneverynonzeromapinHom (M(λ,A),M(µ,A)) T is injective. (6) Let A ⊂ B. Then the module M(λ,A) is irreducible if and only if λ(α∨) 6= 0 i for any i ∈ A′ \A, A & A′, i.e. A is the maximal set such that λ(α∨) = 0. i Remark 4.1. If A = B, then M(λ,A) = L(λ,A) is a trivial one- dimensional module. Corollary 4.2. Let λ ∈ hˆ∗, A ⊂ B, and hˆA⊕Cc2 ⊂ kerλ. Also assume that λ(α∨) 6= 0 for any i ∈ A′ \ A, A & A′. Then L(λ) ∼= M(λ,A) = i L(λ,A)∼= L(λ,A′′) for every A′′ ⊂A. Proof. Item (6) in Proposition 4.2 implies M(λ,A) = L(λ,A). Mean- while, item (4) in Proposition 4.2 implies L(λ,A) ∼= L(λ) and L(λ,A′′) ∼= L(λ,A) for every A′′ ⊂ A. This completes the proof. (cid:3) Corollary 4.3. Let λ ∈ hˆ∗ and λ(c ) = 0. If λ(α∨)6= 0 for any i∈ B, 2 i then M(λ)+ = M. 8 NAIHUANJINGANDCHUNHUAWANG* Proof. By Corollary 3.3, it suffices to show M ⊂ M(λ)+. If λ(α∨) 6= 0 i for any i ∈ B, then L(λ) ∼= M(λ,∅) = L(λ,∅). The result follows by comparing the definition of L(λ) and that of M(λ,∅). (cid:3) The following result is a consequence of Corollary 4.3. Theorem 4.4. Suppose that λ ∈ ˆh∗, λ(c ) = 0 and λ(α∨) 6= 0 for any 2 i i∈ B. Then one has that (1) M(λ) has infinitely many proper submodules: M(λ) ⊃ M(λ−δ ) ⊃ M(λ−2δ )⊃ ··· 2 2 where dimM(λ − kδ ) = dimM(λ) = m are finite. 2 λ−kδ2 λ−kδ2 k Moreover L(λ) = M(λ)/M(λ−δ ). 2 (2) The root multiplicities dimM(λ) = m for all k ≥ 0, and λ−kδ2 k M(λ−kδ ,∅) exhaust all irreducible subquotients of M(λ). 2 (3) dimHom (M(λ−kδ ),M(λ)) = m for any integer k ≥ 0. T 2 k One can describe general highest weight modules as follows. Corollary 4.5. Let V be a highest weight T-module of highest weight λ. If c acts trivially and λ(α∨) 6= 0for i = 0,··· ,s, then V ≃ M(λ)/M(λ− 2 i kδ ) for some k. 2 5. Highest weight modules of T In this section, we construct another class of highest weight T-modules byslightlymodifyingthetriangulardecompositionforIVM.Usingthemethod of [1], we generalize some results of TKK modules to the highest weight T- modules under the condition that λ(c )= λ(c ) = 0. 1 2 These centerless modules are introduced to understandour earlier IVMs and parabolic IVMs. We remark that our construction differs from [14, 15] in that the Cartan subalgebra is purely generated by the imaginary root δ . 2 Let Φ+ = I\Nδ2, Φ− = −Φ+, and Φ0 = Zδ2. Correspondingly, the root spaces are T = T and T = T . Obviously, T = 0 Lα∈Φ0 α ± Lα∈Φ+ ±α T ⊕T ⊕T , and ∆ = Φ ∪Φ ∪(Φ \{0}). + − 0 T + − 0 WedefineanewmodulestructureonC1λ suchthath.1λ = λ(h)·1λ (h ∈ T ), and T .1 = 0. Similarly, we define the induced module M(λ) of T: 0 + λ M(λ) = U(T)⊗U(T0⊕T+)C1λ. The following result describes the relations among IVM’s, generalized IVM’s, and highest weight modules. Proposition 5.1. For any λ ∈hˆ∗, one has that (1) M(λ) is a U(T )−free module generated by 1⊗1 := v . − λ λ (2) M(λ) has a unique irreducible quotient L(λ). (3) Let λ ∈ hˆ∗, hˆA ⊕Cc2 ⊂ kerλ, and A′′ ⊂ A. Then there exists a chain of surjective homomorphisms M(λ) → M(λ) → M(λ,A′′)→ M(λ,A). ON MODULES FOR DOUBLE AFFINE LIE ALGEBRAS 9 (4) Let µ ∈ hˆ∗. Then every nonzero element of Hom (M(λ),M(µ)) is T injective. (5) Let A ⊂ B, and hˆA⊕Cc2 ⊂ kerλ. If λ(α∨i ) 6= 0 for any i∈ A′\A, ′ ∼ ∼ A & A , then L(λ) = L(λ) = L(λ,A). Remark 5.1. (1) If A = ∅, then M(λ) = M(λ,A). (2) L(0) is one-dimensional. 5.1. Irreducible module L(λ). In this subsection, we study integra- bility of L(λ) and its weight subspaces. Definition 5.2. A T-module M is called integrable if M is a weight module and xα(m,n) ∈ T (α ∈ ∆;m,n ∈ Z) are locally nilpotent on every nonzero v ∈ M, i.e., there exists a positive integer N = N(α,m,n) such that x (m,n)N.v = 0. α We write x(α,n) = x ⊗tn for α ∈ ∆ . The following result is clear. α 2 ˆg1 Lemma 5.3. T± is generated by {x(±αi,n)|i = 0,1,...,s;n ∈ Z} re- spectively. For an arbitrary Lie algebra g, we recall the following results. Proposition 5.2. [11] Let v ,v ,... be a system of generators of a g− 1 2 module V, and suppose that each x ∈ g is locally ad-nilpotent on g and xNi(vi) = 0 for some positive integers Ni (i = 1,2,...). Then x is locally nilpotent on V. Proposition 5.3. [13] Let π : g → gl(V) be a representation of g. If both adx and π(x) are locally nilpotent for any x ∈ g, then for any y ∈ g, (5.1) π(exp(adx)(y)) = (expπ(x))π(y)(expπ(x))−1. For a real root γ ∈ ∆ˆrg1e = {α+nδ1|α 6= 0, n ∈ Z}, define the reflection r on hˆ∗ by γ r (λ) = λ−λ(γ∨)γ, γ where γ∨ =α∨+nc if γ = α+nδ . Let W be the affine Weyl group of ˆg 1 1 a 1 generated by the reflections r , γ ∈ ∆re. Then W is a Coxeter group. γ ˆg1 a Lemma 5.4. Suppose that x(−αi,n)(i = 0,1,...,s,n ∈ Z) are locally nilpotent on the highest weight vector vλ in L(λ). Then all x(m,n) (m ∈Z) are locally nilpotent on L(λ). Proof. Since x(−α ,n)(i = 0,1,...,s) are locally nilpotent on v , they i λ act locally nilpotent on any x(m,n) via the adjoint representation. By Prop. 5.2 the elements x(−α ,n) are locally nilpotent on L(λ). So L(λ) i is integrable for each of the sl -algebras: < x(α ,n),x(−α ,−n),α∨ > , 2 i i i i= 1,2,...,s as well as <x(α ,n),x(−α ,−n),α∨ = −θ∨ > (Note that we 0 0 0 assume that c ,c act as zero). 1 2 Supposeβ = α+mδ +nδ ∈ ∆re istherootofx(m,n). Letγ = α+mδ , 1 2 T 1 then γ = β−nδ ∈ ∆re. For any i∈ {0,...,s}, there exists a w ∈ W such 2 ˆg1 a 10 NAIHUANJINGANDCHUNHUAWANG* that w(γ) = α [11]. Since w(δ ) = δ , we have w(β) = α +nδ . Let s be i 2 2 i 2 w the linear automorphism of T associated with w. Up to a nonzero constant, we have s (x(m,n)) = Y for Y ∈ T . It follows from Prop. 5.3 that w αi+nδ2 all x(m,n) are locally nilpotent on L(λ). (cid:3) We now recall Weyl modules [4] for the loop algebra sˆl2(C) = sl2 ⊗ C[t,t−1]. Let a1,...,an ∈ C× and λ1,...,λn ∈ Z+ with |λ| = Piλi. We define B(a,λ) to be the cyclic sˆl2(C)−module generated by w such that e(m)w = f(0)|λ|+1w = 0, ∀m, n h(m)w = Xλjamj w, ∀m. j=1 The following result was proved by Chari and Pressley. Proposition 5.4. [4] The sˆl2(C)−module B(a,λ) is finite dimensional. IfB′ isafinitedimensional sˆl2(C)-modulegeneratedbyw′ suchthatdim U(α∨⊗ C[t,t−1])w′ = 1, then B′ is a quotient of some B(a,λ) constructed above. We also need the following remarkable formula proved by Garland. Lemma 5.5. [10] Let β = α+r1δ1 ∈∆ˆg1+, r1 ∈ Z. Then for any t ≥ 1, we have t x(β,±1)tx(−β,0)t+1 = X x(−β,±m)Λ±(β∨,t−m)+X, m=0 x(β,±1)t+1x(−β,0)t+1 =Λ±(β∨,t+1)+Y, where X and Y are in the left ideal of T generated by the subalgebra T and + Λ±(β∨,j) is the coefficient of uj in Λ±(β∨,u) = exp(− ∞ β∨t±2juj). P j j=1 Remark 5.6. By definition of Λ±(β∨,u) it follows that every element h ⊗ tm (h ∈ h,m ∈ Z×) is a polynomial in the variables Λ±(α∨,j) (i = 2 i 1,2,...,s;j ∈ N). Theorem 5.7. For each p = 1,...,s, let λp,i ∈ Z+,ap,i ∈ C×,i = 1,··· ,k . If λ satisfies p kp (5.2) λ(α∨p(0,n)) = Xλp,ianp,i i=1 and λ(c ) = λ(c ) = 0, then 1 2 (1) L(λ) is integrable. (2) L(λ) = U(nˆ1− ⊗C[t2]).vλ, where nˆ1− is the negative nilpotent sub- algebra of ˆg . 1 ′ Moreover if L(λ) and L(λ ) are nonzero irreducible modules, then (5.3) L(λ) ∼= L(λ′) ⇐⇒ λ = λ′.

See more

The list of books you might like

Most books are stored in the elastic cloud where traffic is expensive. For this reason, we have a limit on daily download.