ebook img

Normality and non-normality of group compactifications in simple projective spaces PDF

0.27 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 Normality and non-normality of group compactifications in simple projective spaces

NORMALITY AND NON-NORMALITY OF GROUP COMPACTIFICATIONS IN SIMPLE PROJECTIVE SPACES PAOLOBRAVI,JACOPOGANDINI,ANDREAMAFFEI,ALESSANDRORUZZI Abstract. If G is a complex simply connected semisimple algebraic group and if λ is a dominant weight, we consider the compactification Xλ ⊂ P(cid:0)End(V(λ))(cid:1) obtained as the closure of the G×G- orbitofthe identity andwegive necessaryand sufficientconditions onthe supportof λsothat Xλ is normal;aswell,wegivenecessaryandsufficientconditions onthesupportofλsothatXλ issmooth. 1 1 0 2 Introduction n Consider a semisimple simply connected algebraic group G over an algebraically closed field k of a J characteristiczero. If λis a dominantweight(with respectto afixed maximaltorus T anda fixedBorel 7 subgroup B ⊃T) and if V(λ) is the simple G-module of highest weight λ, then End V(λ) is a simple 1 G×G-module. Let I ∈End V(λ) be the identity map and consider the variety X ⊂P End(V(λ)) λ λ(cid:0) (cid:1) ] given by the closure of the G(cid:0)×G-o(cid:1)rbit of [Iλ]. In [Ka], S. Kannan studied for which λ t(cid:0)his variety i(cid:1)s G projectively normal, and this happens precisely when λ is minuscule. In [Ti], D. Timashev studied the A more generalsituation of a sum of irreducible representations,giving necessary and sufficient conditions . h for the normality and smoothness of these compactifications; however the conditions for normality are t a not completely explicit. In this paper we give an explicit characterizationof the normality of X , which λ m allows to simplify the conditions for the smoothness as well. [ To explain our results we need some notation. Let ∆ be the set of simple roots (w.r.t. T ⊂ B) and 2 identify ∆ with the vertices of the Dynkin diagram. Define the support of λ as the set Supp(λ)={α∈ v ∆ : hλ,α∨i6=0}. 8 7 4 TheoremA(seeTheorem13). ThevarietyXλ isnormalifandonlyifλsatisfiesthefollowing property: 2 (⋆) For every non-simply laced connected component ∆′ of ∆, if Supp(λ)∩∆′ contains a long root, . 5 then it contains also the short root which is adjacent to a long simple root. 0 0 1 In particular, if the Dynkin diagram of G is simply laced then Xλ is normal, for all λ. In the paper : we will prove the theorem in a more general form, for simple (i.e. with a unique closed orbit) linear v i projective compactifications of an adjoint group (see section 1.4). We will make use of the wonderful X compactification of G , the adjoint group of G, and of the results on projective normality of these r ad a compactificationsprovedbyS.Kannanin[Ka]. Theseresultsholdinthemoregeneralcaseofasymmetric variety; however our method does not apply to this more general situation (see section 4.2). Theorem B (see Theorem 22). The variety X is smooth if and only if λ satisfies property (⋆) of λ Theorem A together with the following properties: i) For every connected component ∆′ of ∆, Supp(λ)∩∆′ is connected and, in case it contains a unique element, then this element is an extreme of ∆′; ii) Supp(λ) contains every simple root which is adjacent to three other simple roots and at least two of the latter; iii) Every connected component of ∆rSupp(λ) is of type A. Theorem B can be generalized to any simple and normal adjoint symmetric variety. Following a criterionofQ-factorialityforsphericalvarietiesgivenbyM.Brionin[Br],propertiesi)andii)characterize 1 the Q-factoriality of the normalization of X (see Proposition 20), while property iii) arises from a λ criterion of smoothness given by D. Timashev in [Ti] in the case of a linear projective compactification of a reductive group. As a corollary of Theorem B, we get that X is smooth if and only if its normalization is smooth. λ The paper is organizedas follows. In the first section we introduce the wonderful compactification of G and the normalization of the variety X . In the second section we prove Theorem A, and in the ad λ third section Theorem B. In the last section we discuss some possible generalizations of our results. 1. Preliminaries 1.1. Notation. Recall that G is semisimple and simply connected. Fix a Borel subgroup B ⊂ G, a maximal torus T ⊂ B and let U denote the unipotent radical of B. Lie algebras of groups denoted by upper-case latin letters (G,U,L,...) will be denoted by the corresponding lower-case german letter (g,u,l,...). Let Φ denote the set of roots of G relatively to T and ∆ ⊂ Φ the basis associated to the choiceofB. Forallα∈∆lete ,α∨,f beansl(2)-tripleofT-weightsα,0,−α. LetΛdenotetheweight α α lattice of T and Λ+ the subset of dominant weights. For all α ∈ ∆, denote by ω the corresponding α fundamental weight. If λ∈Λ, recall the definition of its support: Supp(λ)={α∈∆ : hλ,α∨i6=0}. If I ⊂∆, define its border ∂I, its interior I◦ and its closure I as follows: ∂I ={α∈∆rI : ∃β ∈I suchthathβ,α∨i6=0}; I◦ =Ir∂(∆rI); I =I ∪∂I. For λ∈Λ, denote by L the line bundle on G/B whose T-weight in the point fixed by B is −λ. For λ λ dominant, V(λ) = Γ(G/B,L )∗ is an irreducible G-module of highest weight λ; when we deal with λ different groups we will use the notation V (λ). G Denote by Π(λ) the set of weights occurring in V(λ) and set Π+(λ)=Π(λ)∩Λ+. Let λ7→λ∗ be the linear involution of Λ defined by (V(λ))∗ ≃V(λ∗), for any dominant weight λ. The weight lattice Λ is endowed with the dominance order 6 defined as follows: µ6λ if and only if λ−µ ∈ N∆. If β = n α ∈ Z∆, define its support over ∆ (not to be confused with the previous α∈∆ α one) as follows: P Supp (β)={α∈∆ : n 6=0}. ∆ α We introduce also some notations about the multiplication of sections. Notice that, for all λ,µ ∈ Λ, L ⊗L = L . Therefore, if λ,µ are dominant weights and n ∈ N, the multiplication of sections λ µ λ+µ defines maps as follows: m :V(λ)×V(µ)→V(λ+µ) and mn :V(λ)→V(nλ). λ,µ λ We will also write uv for m (u,v) and un for mn(u). Since G/B is irreducible, m and mn induce λ,µ λ λ,µ λ the following maps at the level of projective spaces: ψ :P(V(λ))×P(V(µ))→P(V(λ+µ)) and ψn :P(V(λ))→P(V(nλ)). λ,µ λ The following lemma is certainly well known; however we do not know any reference. Lemma 1. Let λ,µ be dominant weights. i) If Supp(λ)∩Supp(µ) = ∅, then the map ψ : P(V(λ))×P(V(µ)) → P(V(λ+µ)) is a closed λ,µ embedding. ii) For any n>0, the map ψn :P(V(λ))→P(V(nλ)) is a closed embedding. λ 2 Proof. i). Fix highest weight vectors v ∈V(λ), v ∈V(µ) and v =v v ∈V(λ+µ). λ µ λ+µ λ µ IfV isirreducible,thenP(V)hasauniqueclosedorbit,namelythe orbitofthe highestweightvector. Consequently, since P(V(λ))×P(V(µ)) has a unique closed orbit, in order to prove the claim it suffices to prove that ψ is smooth in x = ([v ],[v ]) and that the inverse image of [v ] is x. The second λ,µ λ µ λ+µ claim is clear for weight reasons. In order to prove that ψ is smooth in x, consider T-stable complements U ⊂V(λ), V ⊂V(µ) and λ,µ W ⊂V(λ+µ) of kv , kv and kv . So in a neighbourhood of x the map ψ can be described as λ µ λ+µ λ,µ ψ :U ×V −→W where ψ(u,v)=uv +v v+uv. µ λ The differentialof ψ in x is then givenby the differential of ψ in (0,0),thus it is described as follows: λ,µ dψ (u,v)=uv +v v. x µ λ Supposethatdψ isnotinjective. SinceitisT-equivariant,consideramaximalweightη ∈Π(λ+µ)r{λ+ x µ} such that there exists a couple of non-zero T-eigenvectors (u,v) ∈ kerdψ with weights respectively x η−µ and η−λ. Suppose that η−µ∈Π(λ)r{λ} is not maximalandtake α∈∆ suchthat η−µ+α∈ Π(λ)r{λ} and e u6=0: then α (e u)v +v (e v)=e (uv +v v)=0 α µ λ α α µ λ and η+α∈Π(λ+µ)r{λ+µ}, againstthe maximality of η. Thus η−µ is maximalin Π(λ)r{λ} and similarly η−λ is maximal in Π(µ)r{µ}. Therefore, on one hand it must be η−µ=λ−α with α∈Supp(λ), while on the other hand it must be η−λ=µ−β with β ∈Supp(µ). Since Supp(λ)∩Supp(µ)=∅, this is impossible and shows that, if (u,v)∈kerdψ , x thenitmustbeu=0orv =0. Supposenowthat(u,0)∈kerdψ : thenuv =0andbytheirreducibility x µ of G/B also u=0. A similar argument applies if v =0. ii). Suppose that v,w ∈ V(λ) are such that vn = wn: then v = tw for some t ∈ k. Thus ψn is λ injective. Let us show now that ψn is smooth; it is enough to show it in x = [v ] where v ∈ V(λ) is λ λ λ a highest weight vector. Let V ⊂ V(λ) be the T-stable complement of kv , identified with the tangent λ space T P(V(λ)). If v ∈V, the differential d(ψn) is described as follows x λ x d(ψn) (v)=nvn−1v. λ x λ Thus d(ψn) is injective and ψn is smooth. (cid:3) λ x λ 1.2. The variety X . If λ is a dominantweight,denote by E(λ) the G×G-module End(V(λ)) and set λ X the closure of the G×G-orbit of [I ]∈P(E(λ)). More generally if λ ,...,λ are dominant weights λ λ 1 m we define X =G×G([I ],...,[I ])⊂P(E(λ ))×···×P(E(λ )). λ1,...,λm λ1 λm 1 m Since E(λ) is an irreducible G×G-module of highest weight (λ,λ∗), as a consequence of Lemma 1 we get that if λ and µ have non-intersecting supports and if n∈N then X ≃X and X ≃X . λ+µ λ,µ nλ λ As a consequence we get the following proposition: Proposition 2. Let λ,µ be dominant weights. Then X ≃ X as G×G-varieties if and only if λ and λ µ µ have the same support. Moreover, if Supp(λ)={α ,...,α } then 1 m X ≃X . λ ωα1,...,ωαm 3 Proof. By the discussion above we have to prove only that the condition is necessary. This follows by noticing that if X and X are G×G-isomorphic then also their closed G×G-orbits are isomorphic, λ µ which is equivalent to the fact that λ and µ have the same support. (cid:3) 1.3. The wonderful compactification of G and the normalization of X . When λ is a regular ad λ weight(i.e.Supp(λ)=∆)the varietyX iscalledthe wonderfulcompactificationofG andithasbeen λ ad studied by C. De Concini and C. Procesi in [DP]. We will denote this variety by M: it is smooth and the complement of its open orbit is the union of smooth prime divisors with normal crossings whose intersection is the closed orbit. The closed orbit of M is isomorphic to G/B×G/B and the restriction of line bundles determines an embedding of Pic(M) into Pic(G/B×G/B), that we identify with Λ×Λ as before; the image of this map is the set of weights of the form (λ,λ∗). Therefore Pic(M) is identified with Λ and we denote by M a line bundle on M whose restriction to G/B ×G/B is isomorphic to λ Lλ⊠Lλ∗. If D ⊂ M is a G×G-stable prime divisor then the line bundle defined by D is of the form M , where α is a simple root. The map D 7→α defines a bijection between the set of G×G-stable αD D D prime divisors and ∆, and we denote by M the prime divisor which corresponds to a simple root α. α We denote by s a section of M whose associated divisor is M ; notice that such a section is G×G- α α α invariant. More generally if ν = α∈∆nαα ∈ N∆, set sν = α∈∆snαα ∈ Γ(M,Mν). Then, given any λ∈Λ, the multiplication by sν injects Γ(M,M ) in Γ(M,M ). P λ−ν Qλ Ifλis adominantweight,the mapG −→P(E(λ)) extends to amapq :M −→P(E(λ)) (see[DP]) ad λ whose image is X and such that M = q∗(O (1)). If we pull back the homogeneous coordinates λ λ λ P(E(λ)) of P(E(λ)) to M, we get then a submodule of Γ(M,M ) which is isomorphic to E(λ)∗; by abuse of λ notation we will denote this submodule by E(λ)∗. If λ∈Λ, in [DP, Theorem 8.3] the following decomposition of Γ(M,M ) is given: λ Γ(M,M )= sλ−µE(µ)∗. λ µ∈ΛM+:µ6λ Consider the graded algebra A(λ) = ∞ A (λ), where A (λ) = Γ(M,M ), and set X = n=0 n n nλ λ ProjA(λ). We have then a commutative diagram as follows: L e pλ // // M X A λ A A A qλ AAAA e(cid:15)(cid:15)(cid:15)(cid:15) rλ X λ In[Ka],ithasbeenshownthatA(λ)isgeneratedindegree1andin[D]thatr =r isthenormalization λ of X . Notice that the projective coordinate ring of X ⊂ P(E(λ)) is given by the graded subalgebra λ λ B(λ)= ∞ B (λ) of A(λ) generated by E(λ)∗ ⊂Γ(M,M ). n=0 n λ L 1.4. The variety X . We consider now a generalization of the variety X . Let Σ be a finite set of Σ λ dominant weights and denote E(Σ) = E(µ); let x = [(I ) ] ∈ P(E(Σ)) and define X as the µ∈Σ Σ µ µ∈Σ Σ closureofthe G×G-orbitofx inP(E(Σ)). IfΣ={λ},thenwe getthe varietyX , whileif Σ=Π+(λ) Σ L λ we getits normalizationX . Notice that the diagonalactionofGfixes the pointx so we haveaG×G λ Σ equivariantmapG−→X givenbyg 7−→(g,1)x . ThismapinducesamapfromG toX ifandonly Σ Σ ad Σ e if the action of the center of G×G on E(λ) is the same for all λ ∈ Σ or equivalently if Σ is contained in a coset of Λ modulo Z∆. In this case we say that X is a semi-compactification of G . If G is a Σ ad ad simple group and and Σ 6= {0} then X is a compactification of G , while if G is not simple we can Σ ad ad only say that is a compactification of a group which is a quotient of G . ad We say that Σ is simple if there exists λ ∈ Σ such that Σ ⊂ Π+(λ) or equivalently if Σ contains a unique maximal element with respect to the dominance order 6. Notice also that if λ∈Σ is such that for all µ ∈Σ different from λ the vector µ−λ is not in Q [∆] then is easy to construct a cocharacter >0 4 χ : k∗ −→ G×G such that lim χ(t)x is the highest weight line in P(E(λ)). In particular X is a t→0 Σ Σ simple G×G semi-compactification of G if and only if Σ is simple. ad By the description of the normalization of X is Σ is simple and λ∈Σ is the maximal element, then λ we get r // // Xλ XΣ Xλ In particular, it follows that r=r :X −→X is the normalization of X . Σ λe Σ Σ If Σ is simple, denote B(Σ)= ∞ B (Σ) the projective coordinate ring of X ⊂P(E(Σ)): it is the n=0 n Σ subalgebra of A(λ) generated by E(Σ)e∗ ⊂Γ(M,M ). L λ Remark 3. The discussion above and the fact that in P(E(λ)) there is only one point fixed by the diagonal action of G (the line of scalar matrices) proves that any G×G linear projective compactifi- cation of G is of the form X . A projective G×G-variety X is said to be linear if there exists an ad Σ equivariant embedding X ⊂P(V) where V is a finite dimensional rational G×G-module. In particular as a consequence of Sumihiro’s Theorem (see for example [KKLV, Corollary 2.6]) all normal projective compactifications are linear. In this paper we study only linear compactifications. 2. Normality In this section we determine for which simple Σ the variety X is normal, proving in particular Σ Theorem A. In the following, by λ we will always denote the maximal element of Σ. Let ϕ ∈ E(λ)∗ be a highest weight vector and set X◦ ⊂ X the open affine subset defined by the λ Σ Σ non-vanishingofϕ . Inparticular,wesetX =X andnoticethatX◦ =r−1(X◦). NoticethatX◦ λ λ Π+(λ) λ Σ Σ is B×B-stable and, since it intersects the closed orbit, it intersects every orbit: therefore X is normal Σ if and only if XΣ◦ is normal if and only if thee restriction r Xe◦ :Xλ◦ →XΣ◦eis an isomorphism. Denote by λ B¯(Σ) the coordinate ring of X◦ and by A¯(λ) the coordina(cid:12)te ring of X◦; then we have Σ (cid:12) e λ ϕ ϕ A¯(λ)={ : ϕ∈A (λ)}⊃{ : ϕ∈B (Σ)}=B¯(Σ) ϕn n ϕn n e λ λ andX isnormalifandonlyifA¯(λ)=B¯(Σ). TheringsA¯(λ)andB¯(Σ)arenotG×G-modules,however Σ since X◦ is an open subset of X we still have an action of the Lie algebra g⊕g on them. Σ Σ By [Ka], A¯(λ) is generated by the elements of the form ϕ/ϕ with ϕ∈A (λ). In particular we have λ 1 the following lemma. Lemma 4. The variety X is normal if and only if for all µ ∈ Λ+ such that µ 6 λ there exists n > 0 Σ such that sλ−µE(µ+(n−1)λ)∗ ⊂B (Σ). n Proof. Let ϕ ∈ sλ−µE(µ)∗ be a highest weight vector and suppose that X is normal. Then, by the µ Σ descriptions of A¯(λ) and B¯(Σ), for every dominant weight µ6λ there exist n>0 and ϕ∈B (Σ) such n that ϕ/ϕn = ϕ /ϕ or equivalently ϕ = ϕ ϕn−1 ∈ B (Σ). Since ϕ is a highest weight vector of the λ µ λ µ λ n module sλ−µE(µ+(n−1)λ)∗ the claim follows. Conversely assume that for every dominant weight µ6λ there exists n such that sλ−µE(µ+(n−1)λ)∗ ⊂B (Σ); n inparticularϕ=ϕ ϕn−1 ∈B (Σ). Let’sprovethatϕ/ϕ ∈B¯(Σ)foreveryϕ∈sλ−µE(µ)∗;thisimplies µ λ n λ the thesis since A¯(λ) is generated in degree one. If ϕ = ϕ this is clear. Using the action of the Lie µ algebra g⊕g on B¯(Σ), let’s show that if ϕ/ϕ ∈B¯(Σ) then f (ϕ)/ϕ ∈B¯(Σ): indeed we have λ α λ f (ϕ) ϕ ϕ f (ϕ ) α α λ =f ( )+ · α ϕ ϕ ϕ ϕ λ λ λ λ and the claim follows since f (ϕ )∈E(λ)∗ ⊂B (Σ). (cid:3) α λ 1 5 We can describe the set B (Σ) more explicitly. Indeed, as in [D] or in [Ka], it is possible to identify n sectionsofalinebundleonM withfunctionsonGandusethedescriptionofthemultiplicationofmatrix coefficients. RecallthatasaG×G-modulewehavek[G]= E(λ)∗ ≃ V(λ)∗⊗V(λ). More λ∈Λ+ λ∈Λ+ explicitly if V is a representation of G, define c :V∗⊗V −→k[G] as usual by c (ψ⊗v)(g)=hψ,gvi. V L L V If we multiply functions in k[G] of this type then we get c (ψ⊗v)·c (χ⊗w)=c (ψ⊗χ)⊗(v⊗w) : V W V⊗W in particularwe getthat the imageof the multiplicationE(cid:0)(λ)∗⊗E(µ)∗ −→(cid:1)k[G] is the sum ofall E(ν)∗ with V(ν)⊂V(λ)⊗V(µ). As a consequence we obtain the following Lemma: Lemma 5 ([Ka, Lemma 3.1] or [D]). Let ν,ν′ be dominant weights, then the image of E(ν)∗ ⊗E(ν′)∗ in Γ(M,Mν+ν′) via the multiplication map is sν+ν′−µE(µ)∗. V(µ)⊂VM(ν)⊗V(ν′) Proof. Indeed let π : G →M be the map induced by the inclusion G ⊂ M. Then any line bundle on ad G can be trivializedso that the image of π∗ :E(λ)∗ ⊂Γ(M,M )−→k[G] is the image of c andthe ν V(λ) claim follows from previous remarks. (cid:3) Together with Lemma 4, this gives the following Proposition 6. The variety X is normal if and only if, for every µ∈Λ+ such that µ6λ, there exist Σ n>0 and λ ,...,λ ∈Σ such that 1 n V(µ+(n−1)λ)⊂V(λ )⊗···⊗V(λ ). 1 n 2.1. Remarks on tensor products. By Proposition 6, in order to establish the normality (or the non-normality) of X , we need some results on tensor product decomposition. Σ Lemma 7. Let λ,µ,ν be dominant weights and let ∆′ ⊂ ∆ be such that Supp (λ+µ−ν) ⊂ ∆′; let ∆ L⊂G be the standard Levi subgroup associated to ∆′. If π ∈Λ+, denote by V (π) the simple L-module L of highest weight π. Then V(ν)⊂V(λ)⊗V(µ) ⇐⇒ V (ν)⊂V (λ)⊗V (µ). L L L Proof. If a is any Lie algebra, denote U(a) the corresponding universal enveloping algebra. Suppose that V (ν) ⊂ V (λ)⊗V (µ); fix maximal vectors v ∈ V (λ) and v ∈ V (µ) for the Borel L L L λ L µ L subgroupB∩L⊂Landfix p∈U(l∩u−)⊗U(l∩u−)suchthatp(v ⊗v )∈V (λ)⊗V (µ) isamaximal λ µ L L vector of weight ν. Since V (λ)⊗V (µ) ⊂ V(λ)⊗V(µ), we only need to prove that p(v ⊗v ) is a L L λ µ maximal vector for B too. If α∈∆′ then we have e p(v ⊗v )=0 by hypothesis. On the other hand, α λ µ if α ∈ ∆r∆′, notice that e commutes with p, since by its definition p is supported only on the f ’s α α with α∈∆′. Since v ⊗v is a maximal vector for B, then we get λ µ e p(v ⊗v )=pe (v ⊗v )=0; α λ µ α λ µ thus p(v ⊗v ) generates a simple G-module of highest weight ν. λ µ Assume conversely that V(ν) ⊂ V(λ)⊗V(µ) and fix p ∈ U(u−)⊗U(u−) such that p(v ⊗v ) ∈ λ µ V(λ)⊗V(µ) is a maximal vector of weight ν. Since Supp (λ+µ−ν) ⊂ ∆′, we may assume that the ∆ only f ’s appearing in p are those with α∈∆′; therefore p(v ⊗v )∈V (λ)⊗V (µ) and it generatesa α λ µ L L simple L-module of highest weight ν. (cid:3) Lemma 8. Fix λ,µ,ν ∈Λ+ such that V(ν)⊂V(λ)⊗V(µ). Then, for any ν′ ∈Λ+, it also holds V(ν+ν′)⊂V(λ+ν′)⊗V(µ). 6 Table 1. type of Φ highest short root A α +···+α =ω +ω r 1 r 1 r B α +···+α =ω r 1 r 1 C α +2(α +···+α )+α =ω r 1 2 r−1 r 2 D α +2(α +···+α )+α +α =ω r 1 2 r−2 r−1 r 2 E α +2α +2α +3α +2α +α =ω 6 1 2 3 4 5 6 2 E 2α +2α +3α +4α +3α +2α +α =ω 7 1 2 3 4 5 6 7 1 E 2α +3α +4α +6α +5α +4α +3α +2α =ω 8 1 2 3 4 5 6 7 8 8 F α +2α +3α +2α =ω 4 1 2 3 4 4 G 2α +α =ω 2 1 2 1 Proof. Fix a maximal vector vν′ ∈V(ν′) and consider the U-equivariantmap φ: V(λ)⊗V(µ) −→ V(λ+ν′)⊗V(µ) w1⊗w2 7−→ mλ,ν′(w1,vν′)⊗w2 Theclaimfollowssince,ifv ∈V(λ)⊗V(µ)isaU-invariantvectorofweightν,thenφ(v )∈V(λ+ν′)⊗ ν ν V(µ) is a U-invariant vector of weight ν+ν′. (cid:3) We nowdescribesomemoreexplicitresults. Whenwedealwithexplicitirreduciblerootsystems,un- lessotherwisestated,wealwaysusethenumberingofsimplerootsandfundamentalweightsofBourbaki [Bo]. In order to describe the simple subsets Σ ⊂ Λ+ which give rise to a non-normal variety X , we will Σ make use of following lemma. Lemma 9. (1) Let G be of type B . Then, for any n, V((n−1)ω )6⊂V(ω )⊗n. r 1 1 (2) Let G be of type G . Then, for any n, V(ω +(n−1)ω )6⊂V(ω )⊗n. 2 1 2 2 Proof. We consider only the first case, the second is similar. Fix a highest weight vector v ∈ V(ω ). 1 1 If α is any simple root and if 1 6 s 6 r, notice that f acts non-trivially on f ···f v if and only α αs−1 α1 1 if α = α . The T-eigenspace of weight 0 in V(ω ) is spanned by v = f ···f v , and similarly the s 1 0 αr α1 1 T-eigenspaceof weight (n−1)ω in V(ω )⊗n is spanned by v⊗i−1⊗v ⊗v⊗n−i, where 16i6n. Since 1 1 1 0 1 the vectors e (v⊗i−1⊗v ⊗v⊗n−i)=v⊗i−1⊗(e v )⊗v⊗n−i αr 1 0 1 1 αr 0 1 are linearly independent, there exists no maximal vector of weight (n−1)ω in V(ω )⊗n. (cid:3) 1 1 Dual results will be needed to describe the subsets Σ which give rise to a normal variety X , but Σ before we need to introduce some further notation. If Φ is an irreducible root system and ∆ is a basis for Φ we will denote by η the highest root if Φ is simply laced or the highest short root if Φ is not simply laced. For the convenience of the reader we list the highest short root of every irreducible root system in Table 1. Recall the condition (⋆) defined in the introduction: a dominant weight λ satisfies (⋆) if, for every non-simply laced connected component ∆′ ⊂ ∆, if Supp(λ)∩∆′ contains a long root then it contains also the short root which is adjacent to a long simple root. Definition 10. If ∆′ ⊂ ∆ is a non-simply laced connected component, order the simple roots in ∆′ = {α ,...,α } starting from the extreme of the Dynkin diagram of ∆′ which contains a long root and 1 r denote α the first short root in ∆′. If λ is a dominant weight such that α 6∈ Supp(λ) and such that q q 7 Supp(λ)∩∆′ containsalongroot,denoteα thelastlongrootwhichoccursinSupp(λ)∩∆′;forinstance, p if ∆′ is not of type G , then the numbering is as follows: 2 q qq qq pppppppppppppppppppp qq q α1 αp αq αr The little brother of λ with respect to ∆′ is the dominant weight λl∆b′ =λ−i=qpαi =( λλ−+ωω1p−+1ω−2ωp+ωq+1 iofthGeriws iosfetype G2 X where ω is the fundamental weight associated to α if 1 6 i 6 r, while ω = ω = 0. The set of i i 0 r+1 the little brothers of λ will be denoted by LB(λ); notice that LB(λ) is empty if and only if λ satisfies condition (⋆) of Theorem A. For convenience, define LB(λ)=LB(λ)∪{λ}, while if ∆ is connected and non-simply laced set λlb =λlb. ∆ Lemma 11. Assume G to be simple and let λ∈Λ+r{0}. Denote η the highest root of Φ if the latter is simply laced or the highest short root otherwise. (1) If λ satisfies the condition (⋆) then V(λ)⊂V(η)⊗V(λ). (2) If λ does not satisfy the condition (⋆) and if λlb is the little brother of λ then V(λ)⊂V(η)⊗V(λlb). Proof. If ∆ is simply laced, then V(η) ≃ g is the adjoint representation: in this case the claim follows straightforward by considering the map g⊗V(λ) → V(λ) induced by the g-module structure on V(λ), which is non-zero since λ is non-zero. Suppose now that ∆ is notsimply laced. If λ satisfies condition(⋆), then by Lemma 8 it is enoughto study the case λ=ω where α is a short simple root: α Type B : V(ω )⊂V(ω )⊗V(ω ). r r 1 r Type C : V(ω )⊂V(ω )⊗V(ω ), with i<r. r i 2 i Type F : V(ω )⊂V(ω )⊗V(ω ) and V(ω )⊂V(ω )⊗V(ω ). 4 3 4 3 4 4 4 Type G : V(ω )⊂V(ω )⊗V(ω ). 2 1 1 1 If λ does not satisfy condition (⋆), by Lemma 8 we can assume that λ=ω with α a long root: α Type B : V(ω )⊂V(ω )⊗V(ω ), if 1<i<r, and V(ω )⊂V(ω )⊗V(0). r i 1 i−1 1 1 Type C : V(ω )⊂V(ω )⊗V(ω ). r r 2 r−2 Type F : V(ω )⊂V(ω )⊗V(ω ) and V(ω )⊂V(ω )⊗V(ω +ω ). 4 1 4 4 2 4 1 4 Type G : V(ω )⊂V(ω )⊗V(ω ). 2 2 1 1 The above mentioned inclusion relations for tensor products are essentially known: let us treat the case of type C with λ=ω and i<r, the other cases are easier or can be checked directly. r i Let v be a highest weight vector of V(ω ) and w be a highest weight vector of V(ω ). Let f be the 0 2 0 i following product (in the universal enveloping algebra U(u−)) f =f ···f ·f ···f ·f ···f , αi α1 αi+1 αr−1 αr α2 and consider all the factorizations f =p·q such that p,q ∈U(u−). If β ,...,β ∈∆, set 1 j r(f ···f )=(−1)j2δf ···f , β1 βj βj β1 where δ equals 0 (resp. 1) if α occurs an even (resp. odd) number of times in {β ,...,β }. Then it is i 1 j easy to check that the vector p.v ⊗ rq.w 0 0 p·q=f X is a U-invariant vector in V(ω )⊗V(ω ) of T-weightω . (cid:3) 2 i i 8 IftheDynkindiagramofGisnotsimplylacedwewillneedsomefurtherpropertiesoftensorproducts. If ∆ is connected but not simply laced, we will denote by α the short simple root that is adjacent S to a long simple root α ; moreover, we will denote the associated fundamental weights by ω and ω . L S L Finally, define ζ as the sum of all simple roots and notice that ω +ζ is dominant. S Lemma 12. Let λ be a non-zero dominant weight. (1) If G is of type F or C (r >3) and if Supp(λ) contains a long root then 4 r V(λ+ω )⊂V(ζ +ω )⊗V(λ). S S (2) If G is of type G and if λ does not satisfy (⋆) then 2 V(λ+ω )⊂V(ω )⊗V(λlb). 1 2 (3) If G is of type G and if α ∈Supp(λ) then 2 S V(λ+ω )⊂V(ω )⊗V(λ). 1 2 Proof. By Lemma 8 it is enough to check the statements for λ=ω with α a long root in the first two α cases and α=α in the last case. S Type C : by Lemma 8 it is enough to check that V(ω )⊂V(ω )⊗V(ω ). r r−1 1 r Type F : we have λ=ω or λ=ω and ω +ζ =ω +ω . 4 1 2 S 1 4 Type G : we have λ=ω and λlb =ω in point (2) and λ=ω in point (3). (cid:3) 2 2 1 1 2.2. Normality and non-normality of X . We are now able to state the main theorem. Σ Theorem 13. Let Σ be a simple set of dominant weights and let λ be its maximal element. The variety X is normal if and only if Σ⊃LB(λ). Σ Theorem A stated in the introduction follows immediately by considering the case Σ = {λ}. The remaining part of this section will be devoted to the proof of Theorem 13. The general strategy will be basedonProposition6andwillproceedbyinductiononthedominanceorderofweights. Theingredients of this induction will be the results provedin section2.1 together with the descriptionof the dominance order given by J. Stembridge in [St]: the dominance order between dominant weights is generated by pairs which differ by the highest short root for a subsystem of the root system. If K is a subset of ∆, denote Φ ⊂ Φ the associated root subsystem and, in case K is connected, K denote by η the correspondinghighest shortroot. Moreover,if β = n α, set β| = n α. K α∈∆ α K α∈K α The result of [St] that we will use is the following. P P Lemma 14 ([St, Lemma 2.5]). Let λ,µ be two dominant weights with λ > µ; set I = Supp (λ−µ). ∆ Let Φ be an irreducible subsystem of Φ (where K ⊂I). K I (a) If h(λ−µ) ,α∨i>0 for all α∈K∩Supp(µ), then µ+η 6λ. K K (b) If in addition hµ+η ,α∨i>0 for all α∈IrK, then µ+η ∈Λ+. (cid:12) K K (cid:12) The next two lemmas are the main steps of our induction. Lemma 15. Suppose that Φ is irreducible; let λ,µ ∈ Λ+ such that λ > µ and Supp (λ−µ) = ∆. ∆ Assume that either Φ is simply laced, or there exists a short root α∈Supp(λ) such that hλ−µ,α∨i>0, or α ∈/ Supp(µ). Then there exists a connected subset K of ∆ such that S i) µ+η 6λ; K ii) µ+η ∈Λ+; K iii) K∩Supp(λ)6=∅. Proof. Set K = {α ∈ ∆ : hλ−µ,α∨i > 0}. Since λ > µ we have that K ∩Supp(λ) is non-empty. 1 1 Notice also that Supp(µ)⊃∆rK . Define K as follows: 1 9 a) If Φ is simply laced, let K be a connected component of K which intersects Supp(λ). 1 b) If α ∈ Supp(λ) is a short root such that hλ−µ,α∨i > 0 let K be the connected component of K containing α. 1 c) If Φ is not simply laced and there does not exist a short root α as in b), let K be a connected component of K which intersects Supp(λ). 1 Properties i) and iii) are then easily verified by Lemma 14(a) and by construction. To prove ii) notice that, if Φ is not simply laced, by the construction of K it follows that if α ∈K L then α ∈ K as well: indeed, K is a connected component of K and if there is no short root α as in S 1 b) then α 6∈Supp(µ) implies α ∈K . By the description of highest short roots in Table 1 we deduce S S 1 that, if α∈KrK◦, then the respective coefficient in η is 1: hence hη ,α∨i=−1 for all α∈∂K and, K K since Supp(µ)⊃∆rK ⊃∂K, we get µ+η ∈Λ+. (cid:3) 1 K In order to proceed with the induction, in the next lemma we will need to consider the condition (⋆) also for a Levi subgroup of G. If K ⊂∆ let L be the associated standard Levi subgroup; we say that K λ ∈ Λ+ satisfies condition (⋆ ) if, for every non-simply laced connected component K′ of K such that K Supp(λ)∩K′ contains a long root, Supp(λ)∩K′ contains also the short root adjacent to a long root. Notice that if λ satisfies (⋆) then it also satisfies (⋆ ) for all K ⊂∆. K Similarlywecanalsodefinethelittlebrotherofadominantweightw.r.t.theLevisubgroupL : ifK′ K is a connectedcomponentofK suchthat λdoes not satisfy(⋆K′), define the little brotherλlKb′ w.r.t. K′ as in Definition 10 and denote by LB (λ) the set of little brothers of λ constructed in this way. Notice K that if K′ is a connected component of K such that λ does not satisfy (⋆K′) and if ∆′ is the connected component of ∆ containing K′, then λ does not satisfy (⋆∆′) as well and λlKb′ = λl∆b′. In particular LB (λ)⊂LB(λ). K Lemma16. AssumeGtobesimpleandletλ,µbetwodominantweights suchthatλ>µandSupp (λ− ∆ µ)=∆. Then there exist µ′ ∈Λ+ and λ′ ∈LB(λ) such that µ<µ′ 6λ and V(µ+λ)⊂V(µ′)⊗V(λ′). Proof. Suppose first that either Φ is simply laced or α ∈/ Supp(µ) or there exists a short root α in S Supp(λ) such that hλ−µ,α∨i > 0. Take K as in Lemma 15 and set µ′ = µ+η : then by Lemma 11 K together with Lemma 8 we get V (µ+λ)⊂V (µ′)⊗V (λ′) LK LK LK with λ′ ∈LB (λ). The claim follows by Lemma 7 together with the inclusion LB (λ)⊂LB(λ). K K SupposenowthatΦisnotsimplylaced,thatα ∈Supp(µ)andthatthereisnoshortrootα∈Supp(λ) S such that hλ−µ,α∨i > 0. Since λ > µ there exists α ∈ Supp(λ) such that hλ−µ,α∨i > 0: therefore, Supp(λ) contains at least a long root. Set µ′ = µ+ζ; notice that µ′ 6λ and that µ′ is dominant. The claim follows then by Lemma 11 and by Lemma 8 if Φ is of type B, while if Φ is of type C, F or G it 4 2 follows by Lemma 12 and by Lemma 8. (cid:3) Proof of Theorem 13. We prove first that the condition is necessary. Assume that there exists a little brother µ = λlb of λ which is not in Σ. We prove that for every positive n and for every choice of ∆′ weights λ ,...,λ ∈Σ the module V(µ+(n−1)λ) is not contained in V(λ )⊗···⊗V(λ ). 1 n 1 n We proceed by contradiction. Assume there exist weights λ ,...,λ as above and notice that any of 1 n them satisfies µ 6 λ 6 λ: indeed, λ−µ = nλ−(µ+(n−1)λ) > nλ− λ > λ−λ for every i. i i i Therefore Supp ( λ −(µ+(n−1)λ))⊂Supp (λ−µ). By Definition 10 together with Lemma 7, it ∆ i ∆ P is enough to analyse the case G of type B and Supp(λ) = {α } or G of type G and Supp(λ) = {α }. P r 1 2 2 We analyse these two cases separately. Type B : we have λ = aω , µ = (a −1)ω and µ+ (n−1)λ = (na−1)ω . If a = 1 we notice r 1 1 1 that there are no dominant weights between λ and µ. So the only possibility is λ = λ = ω for all i 1 10

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.