ebook img

Elementary subgroup of an isotropic reductive group is perfect PDF

0.16 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 Elementary subgroup of an isotropic reductive group is perfect

Elementary subgroup of an isotropic reductive group is perfect 0 1 A. Luzgarev∗, A. Stavrova† 0 2 December 31, 2009 n a J 7 ] 1 Introduction G A Let R be a commutative ring with 1, and let G be an isotropic reductive algebraic group . over R. In [7] Victor Petrov and the second author introduced a notion of an elementary h t subgroup E(R) of the group of points G(R). In this note we prove that, as one might a expectfromthe split case(e.g., [10])as wellas fromthe fieldcase (e.g.,[12]), under natural m assumptions the elementary subgroup of a reductive group is perfect. [ More precisely, assume that G is isotropic in the following strong sense: it possesses a 1 parabolic subgroup that intersects properly any semisimple normal subgroup of G. Such v a parabolic subgroup P is called strictly proper. Denote by E (R) the subgroup of G(R) P 5 generated by the R-points of the unipotent radicals of P and of an opposite parabolic 0 subgroup P−. The main theorem of [7] states that E (R) does not depend on the choice 1 P 1 of P, as soon as for any maximal ideal M of R all irreducible components of the relative 1. root system of GRM (see [6, Exp. XXVI, §7] for the definition) are of rank ≥2. Under this assumption,wecall E (R)the elementary subgroupofG(R)anddenote it simply byE(R). 0 P 0 In particular, E(R) is normal in G(R). This definition of E(R) generalizes the well-known 1 definition of an elementary subgroup of a Chevalley group (or, more generally, of a split : reductive group), as well as several other definitions of an elementary subgroup of isotropic v i classical groups and simple groups over fields [2, 12, 13, 14, 3]. X By the structure constants of a root system we mean the structure constants of the r correspondingsemisimple complexLiegroup,or,inotherwords,constantsappearinginthe a Chevalley commutator formulas for the corresponding Chevalley group. They are among ±1, ±2, ±3. Theorem 1. Let G be an isotropic reductive algebraic group over a commutative ring R. Assume that for any maximal ideal M of R all irreducible components of the relative root system of G are of rank ≥2, and, if one of the irreducible components of the (usual) root RM system of G is of type B =C or G , that theresidue field R /MR is not isomorphic RM 2 2 2 M M to F . Then E(R)=[E(R),E(R)]. 2 Observethatthefirstconditionofthetheoremensuresthatthetheelementarysubgroup E(R)ofG(R)iscorrectlydefined,whilethesecondoneessentiallyeliminatesthewell-known cases where the elementary subgroup of a split reductive group is not perfect. Thus, the result is the strongest possible. One should note that, if we assume only that the rank of relative root systems of G is ≥1, the question whether individual subgroups E (R) are RM P perfect is of separate interest. ∗Theauthorissupported byRFBR09-01-00784, RFBR09-01-00878andRFBR09-01-90304. †Theauthorissupported byRFBR09-01-00878andRFBR09-01-90304. 1 The proofofTheorem1 is basedonthe notionofrelativerootsubschemes(with respect to a parabolic subgroup) of an isotropic group introduced in [7], the generalized Chevalley commutator formula [7, Lemma 9], and localization in the Quillen—Suslin style [11]. To shortenthe proof,wealsomakeuseoftheclassificationofTitsindicesofisotropicreductive groups over local rings obtained in [8]. 2 An abstract definition of relative roots Inthis sectionwerecallthe notionofan(abstract)systemofrelativerootsintroducedin[7] and prove a technical lemma. Let Φ be a reduced root system in a Euclidean space with a scalar product (−,−). Let Π = {α ,...,α} be a fixed system of simple roots of Φ; if Φ is irreducible, we assume 1 l that the numbering follows Bourbaki [5]. Let D be the Dynkin diagram of Φ. We identify nodes of D with the corresponding simple roots in Π. For a subgroup Γ ⊆ Aut(D) and a Γ-invariant subset J ⊆Π, consider the projection π =π : ZΦ−→ZΦ/hΠ\J; α−σ(α), α∈J, σ ∈Γi. J,Γ The set Φ = π(Φ)\{0} is called the system of relative roots corresponding to the pair J,Γ (J,Γ). The rank of Φ is the rank of π(ZΦ) as a free abelian group. J,Γ ItisclearthatanyrelativerootA∈Φ canberepresentedasauniquelinearcombina- J,Γ tion ofrelative roots fromπ(Π). We say that A∈Φ is a positive (resp. negative) relative J,Γ root,if it is a non-negative (respectively, a non-positive) linear combinationof the elements of π(Π). The sets of positive and negative relative roots will be denoted by Φ+ and Φ− J,Γ J,Γ respectively. By the level lev(A) of a relative root A we mean the sum of coefficients in its decomposition. Observe that Γ acts on the set of irreducible components of the root system Φ. If this action is transitive, the system of relative roots Φ is irreducible. Clearly, any system J,Γ of relative roots Φ is a disjoint union of irreducible ones; we call them the irreducible J,Γ components of Φ . J,Γ We will need the following lemma. Lemma 1. Let Φ be a root system with a fixed set of simple roots Π, Γ be a subgroup of Aut(D), andJ beaΓ-invariant subsetofΠ. Ifarelativeroot A∈Φ lies in an irreducible J,Γ component of rank ≥ 2, then there exist such non-collinear B,C ∈ Φ that A = B +C J,Γ and all relative roots iB+jC ∈ Φ , i,j > 0, (i,j) 6= (1,1), have the same sign as A and J,Γ satisfy |lev(iB+jC)|>|lev(A)|. Proof. WecanassumethattherootsystemΦisirreducible,andthatAisapositiverelative root, that is, π−1(A)⊆Φ+. Assume first that A = kπ(α ), where α ∈ Π is a simple root and k > 0 is a positive r r integer. Let α ∈ J be a simple root such that the Γ-orbits of α and α are distinct, s s r and α is at the least possible distance from α on the Dynkin diagram. It is easy to see s r that for any α ∈ π−1(A) there exists β ∈ π−1(α ) such that (α,β) < 0, and, consequently, s α+β ∈Φ. Indeed, we have m (α)=0 by definition, thus we can take for β the sum of all s simplerootsintheDynkindiagramchainbetweenα andthenearestsimplerootappearing s in the decomposition of α. Now set B = π(α+β) and C = π(−β). It is clear that any root in π−1(iB +jC), i,j > 0, contains the summand iα in its decomposition, and thus r is a positive root. Then iB +jC is a positive relative root for any i,j > 0. Moreover, one sees that lev(iB +jC) = lev(A) if and only if i = j = 1. Since π(α) = kπ(α ), and r π(−β)=−π(α ), the roots B and C are non-collinear. s Consider the case where A6=kπ(α ) for any α ∈J. For any α∈π−1(A) there exists a r r sequenceofsimplerootsβ ,...,β ∈Πsuchthatα=β +...+β andβ +...+β ∈Φfor 1 n 1 n 1 i 2 any 1 ≤ i ≤ n. Let i be the least possible index such that β ,...,β ∈ ∆. Then β ∈ J i+1 n i and π(β +...+β +β ) = A. Set B = π(β +...+β ) and C = π(β ). Since B and 1 i−1 i 1 i−1 i C are positive relative roots, we have lev(iB+jC) > lev(A) for any i,j > 0 distinct from i=j =1. The relative roots B and C are non-collinear since otherwise we would have had A=kπ(β ) for some k >0. i 3 Isotropic reductive groups and relative root subschemes Inthis sectionwerecallsomebasic notionspertainingto reductivegroupsoverrings;see[6, 7, 9] for more detailed exposition. Let R be a commutative ring with 1, and let G be a reductive group scheme, or reduc- tive group for short, over R (see [6] for the definition). We denote by Gad and Gsc the corresponding adjoint and simply connected semisimple groups, respectively. Any reductive algebraic group G over R is split locally in the fpqc topology on SpecR. If G is of constant type over R (that is, the root system of G is the same at any point of SpecR), then G is a twisted form of a split reductive algebraic group G over R, given 0 by a cocycle ξ ∈ H1 (R,Aut(G )). Recall that the connected component of Aut(G ) fpqc 0 0 is precisely Gad. The group G is of inner type, if ξ is in the image of the natural map 0 H1 (R,Gad)→H1 (R,Aut(G )). One can always find a finite Galois extension S of R fpqc 0 fpqc 0 such that G is of inner type. The Galois group Gal(S/R) acts on the Dynkin digram of S eachG , where k(s) is the algebraic closure of the residue field of a point s∈SpecR, via k(s) a ∗-action (see [8, 9]). Recall that G is called isotropic, if it contains a proper parabolic subgroup P over R. Recall that we call a parabolic subgroup P of G strictly proper, if it does not contain any semisimple normal subgroup of G. We set EP(R)=hUP(R),UP−(R)i, whereP− isanyparabolicsubgroupofGoppositetoP,andUP andUP− aretheunipotent radicals of P and P− respectively. The main theorem of [7] states that E (R) does not P depend on the choice ofa strictly proper parabolicsubgroup P, as soonas for any maximal ideal M in R all irreducible components of the relative root system of G are of rank RM ≥ 2. Under this assumption, we call E (R) the elementary subgroup of G(R) and denote P it simply by E(R). Let P =P+ be a parabolic subgroup of G, and P− be an opposite parabolic subgroup. LetL=P+∩P− betheircommonLevisubgroup. Itwasshownin[7]thatwecanrepresent Spec(R) as a finite disjoint union m Spec(R)= Spec(R ), a i i=1 so that the following conditions hold for i=1,...,m: • for any s∈SpecR the root system of G is the same; i k(s) • for any s∈SpecR the type of the parabolic subgroup P of G is the same; i k(s) k(s) • if S /R is a Galois extension of rings such that G is of inner type, then for any i i Si s ∈ SpecR the Galois group Gal(S /R ) acts on the Dynkin diagram D of G via the i i i i k(s) same subgroup of Aut(D ). i From here until the end of this section, assume that R = R for some i (or just extend i the base). Denote by Φ the root system of G, by Π a set of simple roots of Φ, by D the corresponding Dynkin diagram. Then the ∗-action on D is determined by a subgroup Γ of Aut D. Let J be the subset of Π such that Π\J is the type of P (that is, the set of k(s) 3 simple rootsof the Levisugroup L ). ThenJ is Γ-invariant. The systemofrelativeroots k(s) Φ is called the system of relative roots corresponding to P and denoted also by Φ . If R J,Γ P is a local ring and P is a minimal parabolic subgroup of G, then Φ can be identified with the relative root system of G in the sense of [6, Exp. XXVI §7] (see also [4] for the field case), as was shown in [7, 9]. To any relative root A∈Φ one associates a finitely generatedprojective R-module V P A and a closed embedding X :W(V )→G, A A where W(V ) is the affine group scheme over R defined by V , which is called a relative A A root subscheme of G. These subschemes possess several nice properties similar to that of elementaryrootsubgroupsofa split group,see [7,Th. 2]. In particular,they aresubjectto certain commutator relations which generalize the Chevalley commutator formula. Moreprecisely,assumethatA,B ∈Φ satisfymA6=−kB foranym,k ≥1. Thenthere P exists a polynomial map N : V ×V →V , ABij A B iA+jB homogeneous of degree i in the first variable and of degree j in the second variable, such that for any R-algebra R′ and for any for any u∈V ⊗ R′, v ∈V ⊗ R′ one has A R B R [X (u),X (v)]= X (N (u,v)) (1) A B Y iA+jB ABij i,j>0 (see [7, Lemma 9]). In a strict analogy with the split case, for any R-algebra R′ we have E(R′)=hX (V ⊗ R′), A∈Φ i A A R P (see [7, Lemma 6]). We will aslo use the following statement which is a slight extension of [7, Lemma 10]. Lemma 2. Consider A,B ∈Φ satisfying A+B ∈Φ and mA6=−kB for any m,k ≥1. P P Denote by Φ an irreducible component of Φ such that A,B ∈π(Φ ). 0 0 (1) In each of the following cases (a) structure constants of Φ are invertible in R (for example, if Φ is simply laced); 0 0 (b) A6=B and A−B 6∈Φ ; P (c) Φ is of type B , C , or F , and π−1(A+B) consists of short roots; 0 l l 4 (d) Φ is of type B , C , or F , and there exist long roots α ∈ π−1(A), β ∈ π−1(B) 0 l l 4 such that α+β is a root; the map N :V ×V →V is surjective. AB11 A B A+B (2) If A−B ∈Φ and Φ 6=G , then P 0 2 im N +im N + im(N (−,v))=V , AB11 A−B,2B,1,1 X A−B,B,1,2 A+B v∈VB where im N =0 if 2B 6∈Φ . A−B,2B,1,1 P Proof. (1) By [7, Lemma 4] any γ ∈ π−1(A+B) decomposes as γ = α+β, α ∈ π−1(A), β ∈ π−1(B). Let S be any member of an affine fpqc-covering SpecS → SpecR that τ τ ` splits G. Set Ψ={iA+jB | i,j >0, (i,j)6=(1,1), iA+jB ∈Φ }. P Then in the notation of [7, Th. 2], over S the commutator [X (e ),X (e )], computed τ A α B β modulo the subgroup hX (V ), C ∈Ψi, is of the form x (±c) = X (±ce ), where c = C C γ A+B γ ±1,±2,±3is the corresponding structure constant. If (a) holds, then c is invertible. If (b), (c) or (d) holds, then c necessarily equals ±1. Indeed, in the only dubious case (d) one 4 should note that, due to the transitive action of the Weyl group of the Levi subgroup on the rootsofthe sameshape (see [1]), any longrootγ ∈π−1(A+B) decomposesasa sumof longroots. Hencec is alwaysinvertible. Thisimplies thatim(N ) =V ⊗S . Since AB11 τ A+B τ im N is a submodule of V defined over the base ring, we have im N =V . AB11 A+B AB11 A+B (2) See [7, Lemma 10]. Lemma 3. Suppose that Φ is an irreducible component of Φ such that Φ ∼= C , l >2, P 0 0 l is a parabolic subgroup of type Π\J, where J ={α ,α }, 2i=l (α is short, α is long), so i l i l that Φ ∼=C . Denote π(α )=A and π(α )=A . Then 0,P 2 i 1 l 2 im(0,N )+ im f =V ⊕V , A1,A1+A2,1,1 X v A1+A2 2A1+A2 v∈VA where f =(N (v,−),N (v,−)): V →V ⊕V . v A1,A2,1,1 A1,A2,2,1 A2 A1+A2 2A1+A2 Proof. Let S be any member of an affine fpqc-covering SpecS →SpecR that splits G. τ τ Suppose that γ ∈ π−1(2A +A ) is a short root. We c`an find short roots α ∈ π−1(A ), 1 2 1 β ∈π−1(A +A ) such that γ =α+β. Therefore, 1 2 [X (e ),X (e )]=x (±1)=X (±e ). A1 α A1+A2 β γ 2A1+A2 γ Hence, e ∈ im(N ) . Now let γ ∈ π−1(2A +A ) be a long root. Take α ∈ γ A1,A1+A2,1,1 τ 1 2 π−1(A ), β ∈ π−1(A ) such that β 6= α and γ = 2α+β (note that α is short, β is long). 1 2 l Therefore [X (e ),X (e )]=x (±1)x (±1)=X (±e )X (±e ). A1 α A2 β α+β 2α+β A1+A2 α+β 2A1+A2 γ Finally, any γ ∈ π−1(A + A ) is a short root, so there exist short roots α ∈ π−1(A ), 1 2 1 β ∈ π−1(A ) such that α+β = γ, hence [X (e ),X (e )] = X (±e ). Combining 2 A1 α A2 β A1+A2 γ these results and noting that the modules in question are defined overthe base ring, we are done. 4 Proof of Theorem 1 Let R be a commutative ring with 1, G be a reductive groupoverR, P be a strictly proper parabolic subgroup of G. For any ideal I ⊆R we write EP(I)=hUP(I),UP−(I)i≤G(R). We denote by R[Y,Z] a ring of polynomials in two variables Z and Y over R. Lemma 4. Let G be a reductive group scheme over a commutative ring R, and let P and P′ be two strictly proper parabolic subgroups of G such that P ≤ P′ or P′ ≤ P. Then for any integer m>0 there exists an integer k >0 such that EP(ZkR[Z])≤EP′(ZmR[Z]). Proof. Without loss of generality, we can assume that over R we have two sets of relative root subschemes XA(VA), A ∈ ΦP, and XB(VB), B ∈ ΦP′, corresponding to P and P′ respectively. Then,ifP ≤P′,by[7,Lemma12]thereexistsanintegerk >0suchthatforanyA∈Φ P andanyv ∈VA one canfind relativerootsBi ∈ΦP′, elements vi ∈VBi, andintegersni >0 (1≤i≤m), such that m X (Ykv)= X (Yniv ), A Y Bi i i=1 5 and hence XA(Ykv)∈EP′(YR[Y]). Substituting R by R[Z] and Y by Zm, we obtain EP(ZkR[Z])≤EP′(ZmR[Z]). If, conversely, P′ ≤ P, we have UP ≤ UP′. Let Ψ± ⊆ Φ+ be two closed sets of roots corresponding to UP±. Then we have π(Ψ±) ⊆ Φ±P′, where π : Φ → ΦP′ is the canonical projection. By [7, Lemma 6] the map XΨ: W(cid:16) M VA(cid:17)→UP±, (vA)A 7→YXA(vA), A∈π(Ψ±) A where the product is takeninany fixedorderrespecting the levelofrelativeroots inΦP′, is an isomorphism of schemes over R. Therefore, UP±(ZmR[Z])≤UP′±(ZmR[Z]). Lemma 5. In the setting of Theorem 1, assume moreover that R is a local ring. Then for any integer m>0 there exists an integer k >0 such that for any R-algebra R′ one has E (ZkR′[Z])⊆[E (ZmR′[Z]),E (ZmR′[Z])]. P P P Proof. Letder(G)bethealgebraicderivedsubgroupofthereductivegroupschemeG(see[6, Exp. XXII, 6.2]). Then, clearly, E (R) ⊆ der(G)(R). Since der(G) is a semisimple group, P we can assume that G is semisimple. Moreover, since the canonical projection Gsc → G, where Gsc is the simply connected semisimple group corresponding to G, is surjective on U±(R),wecanassumethatGissimplyconnected. Anysimplyconnectedsemisimplegroup is a direct product of simply connected semisimple groups that cannot be decomposed into a productofsmaller semisimple groups. These groupsG , i=1,...,n, areWeil restrictions i ofcertainsimplereductivegroupsG′ overafinite´etaleextensionS ofR: G =R (G′)[6, i i S/R i Exp. XIV Prop. 5.10]. Note that the group of R-points G (R) is canonically isomorphic to i thegroupofS-pointsG′(S). ThisisomorphismalsorespectstheembeddingP (R)→G (R), i i i for any parabolic subgroup P of G . Then, clearly, we can assume from the very beginning i i that G is a simple reductive group, and the root system Φ of G is irreducible. Note that by Lemma 4 we can substitute P by any other strictly proper parabolic subgroup P′ of G such that P ≤ P′ or P′ ≤ P. Further, over R we have a set of relative root subschemes X (V ), A∈Φ , corresponding to P. A A P Wearegoingtoshowbyinductionon|levA|thatforanyA∈Φ thereexistsaninteger P k =k(A)>0 such that for any R-algebra R′ and any v ∈V ⊗ R′ one has A R X (Zkv)∈[E (ZR′[Z]),E (ZR′[Z])]. (2) A P P The claim of the lemma then follows by substituting Z by Zm and R′ by R[Z], and taking the final k to be the maximum of all k(A), A∈Φ . P Recall that by Lemma 1 there exist non-collinear relative roots B,C ∈ Φ such that P A = B +C and and all roots iB+jC ∈ Φ , i,j > 0, (i,j) 6= (1,1), have the same sign J,Γ as A and satisfy |lev(iB+jC)| > |lev(A)|. Assume that the map N : V ×V →V BC11 B C A is surjective. Then by the generalized Chevalley commutator formula (1) we have that for any R-algebra R′, and any v ∈V ⊗ R′[Z], A R X (Zkv)=[X (Zu ),X (Zk−1u )]· X (Zi+j(k−1)u ) A B 10 C 01 Y iB+jC ij i,j>0; (i,j)6=(1,1) for some u ∈V ⊗ R′[Z], i,j >0, (i,j)6=(1,1). Then, by the inductive hypothesis, ij iB+jC R (2) holds for k large enough. 6 Now for anyrelativerootA∈Φ (for a suitable choice ofthe parabolicsubgroupP)we P either show that for any decomposition A = B+C, where B and C are non-collinear, the map N is surjective; or provide an explicit decomposition of X (Zkv), v ∈ V ⊗ R′, BC11 A A R into a product of commutators in E (ZR′[Z]), so that (2) is satisfied for k large enough. P Assume first that all structure constants of the root system Φ of G are invertible in R; this includes the case where Φ is simply laced. Then by Lemma 2 the map N is BC11 surjective for any decomposition A=B+C, where B and C are non-collinear. Consider the case where Φ = Φ = C , so G is split. Let Π = {A ,A }, Φ+ = P 2 1 2 {A ,A ,A +A ,2A +A }. Let M be the maximal ideal of R. By the hypothesis of 1 2 1 2 1 2 Theorem1, R/M 6∼=F , hence we cantakeε∈R\M suchthat ε2−ε∈R\M =R∗. If the 2 root A∈Φ is long, we can assume that A=2A +A . Let P 1 2 g (s,t)=[X (s),X (t)]=X (st)X (s2t) 1 A1 A2 A1+A2 2A1+A2 and g (s,t,u)=[X (u),[X (s),X (t)]]=X (−stu)X (−s2t2u). 2 A2 A1+A2 −A2 A1+A2 2A1+A2 Therefore, g (Z2,−Zk−4ε(ε2−ε)−1v)·g (Z,Zε,−Zk−4(ε2−ε)−1v)=X (Zkv). 1 2 2A1+A2 If the root A∈Φ is short, we can assume that A=A +A , hence P 1 2 g (Z,Zk−1v)·X (−Zk+1v)=X (Zkv). 1 2A1+A2 A1+A2 This means that (2) holds for these roots for any k ≥5. Consider the case where Φ = Φ = G , so G is split. Let Π = {A ,A }, Φ+ = P 2 1 2 {A ,A ,A +A ,2A +A ,3A +A ,3A +2A }. BythehypothesisofTheorem1, R/M 6∼= 1 2 1 2 1 2 1 2 1 2 F , hence we can take ε∈R\M such that ε2−ε∈R\M =R∗. If the root A is long, we 2 can assume that A=3A +2A . Then 1 2 [X (Zv),X (Zk−1)]=X (Zkv). A2 3A1+A2 3A1+2A2 Therefore,(2) holds for long roots for any k ≥2. If the rootA is short,we can assume that A=2A +A . Then 1 2 [X (s),X (t)]=X (st)·X (s2t)·X (s3t)·X (s3t2). A1 A2 A1+A2 2A1+A2 3A1+A2 3A1+2A2 Hence, [X (Zε),X (−(ε2−ε)−1Zk−2v)]−1·[X (Z),X (−ε(ε2−ε)−1Zk−2v)]= A1 A2 A1 A2 X (Zkv)X ((ε+1)Zk+1v)X (ε(ε2−ε)−1Z2k+1v), (3) 2A1+A2 3A1+A2 3A1+2A2 and the roots 3A +A and 3A +2A are long. This means that (2) holds for short roots 1 2 1 2 for any k≥3. We areleft with the casewhere Φ is of type B ,C , or F . Recallthat by the hypothesis l l 4 ofTheorem1 Gcontainsa splittorus ofrank≥2. Hence inthe F casethe classificationof 4 Tits indices over localrings [8] says that G is a split group. Hence we can assume that P is a BorelsubgroupofG, andΦ =Φ is arootsystemoftype F . Then if the rootA∈Φ is P 4 P short, by Lemma 2 the map N is surjective for any non-collinear B,C ∈Φ such that BC11 P A=B+C. If this rootis long, then it belongs to the long rootsubsystem of F , whichhas 4 typeD . Then(forexample,byLemma1)wecanfindtwolongrootsB,C ∈Φ , suchthat 4 P B+C = A, and necessarily, iB+jC is not a root for any i,j > 0 distict from i = j = 1. Then X (Zkv)=[X (Zu ),X (Zk−1u )] A B 10 C 01 7 for some u ∈V ⊗ R′[Z] and u ∈V ⊗ R′[Z]. 10 B R 01 C R Consider the case where Φ is of type B , l≥3. By the classification of Tits indices over l local rings [8], we can assume that P is a parabolic subgroup of type Π\J, where Π is a set of simple roots of Φ and J = {α ,α }. Then Φ can be identified with a root system 1 2 P of type B . One readily sees, using the fact that l ≥ 3, that for any relative root A ∈ Φ 2 P andany pair B,C ∈Φ satisfyingA=B+C, we canfind a pair oflong roots β ∈π−1(B), P γ ∈π−1(C) such that β+γ is a root (one can assume that A is one of two simple roots of Φ , due to the lifting of the relative Weyl group [6, Exp. XXVI Th. 7.13 (ii)]). Then by P Lemma 2 the map N is surjective. BC11 It remains to consider the case where Φ is of type C , l ≥ 3. First assume that P is a l parabolic subgroup of type Π\J, where J = {α ,α } for two short simple roots α ,α of i j i j Φ. Then Φ can be identified with a root system of type BC . One readily sees that for P 2 all extra-short and short relative roots A ∈ Φ the set π−1(A) consists of short roots, and P hence by Lemma 2 the map N is surjective for any decomposition A = B+C. Let A BC11 be a long root. Let A and A be a short and an extra-short simple roots of Φ . We can 1 2 P assume without loss of generality that A = 2A +2A . Take k ≥ 4. Then by Lemma 2 1 2 (2) and by the generalized Chevalley commutator formula, for any R-algebra R′, and any v ∈V ⊗ R′[Z], we have A R X (Zkv)=[X (Zu ),X (Zk−2u )]·X (Zk−1u ) A A1 1 2A2 2 A1+2A2 3 for some u ∈V ⊗ R′[Z], u ∈V ⊗ R′[Z], and u ∈V ⊗ R′[Z]. Further, by 1 A1 R 2 2A2 R 3 A1+2A2 R Lemma 2 (1) and by the generalized Chevalley commutator formula, since π−1(A +2A ) 1 2 consists of short roots, we have X (Zk−1u )=[X (Zu ),X (Zk−3u )] A1+2A2 3 A1+A2 4 A2 5 forsomeu ∈V ⊗ R′[Z]andu ∈V ⊗ R′[Z]. Hence(2)holdsforAforanyk ≥4. 4 A1+A2 R 5 A2 R By the classification of Tits indices over local rings [8], the only remaining case is when P is a parabolic subgroup of type Π\J for J ={α ,α}, where l =2i. Now α is short, α i l i l is long, and Φ can be identified with a root system of type B = C . As in Lemma 3, we P 2 2 put A =π(α ), A =π(α ). Then if the root A∈Φ is short, by the lifting of the relative 1 i 2 l P Weyl group [6, Exp. XXVI Th. 7.13 (ii)]), we can assume that A=A +A . By Lemma 3 1 2 for any R-algebra R′, and any v ∈V ⊗ R′[Z], we have A R X (Zkv)= [X (Zv ),X (Zk−1u )] A Y A1 i A2 1 i for some u ∈V ⊗ R′[Z], v ∈V ⊗ R′[Z]. If the root A∈Φ is long, we can assume 1 A2 R i A1 R P that C =2A +A . By Lemma 3 for any R-algebraR′, and any v ∈V ⊗ R′[Z], we 1 2 2A1+A2 R have X (Zkv)=[X (Zu ),X (Zk−1u )]· [X (Zv ),X (Zk−2u )] A A1 1 A1+A2 2 Y A1 i A2 3 i for some u ∈V ⊗ R′[Z], u ∈V ⊗ R′[Z], u ∈V ⊗ R′[Z], v ∈V ⊗ R′[Z]. 1 A1 R 2 A1+A2 R 3 A2 R i A1 R Hence (2) holds for A for any k ≥3. Proof of Theorem 1. RecallthatwecanrepresentSpec(R)asafinitedisjointunionSpec(R)= n n Spec(R ), so that E(R)= E(R ) and for any 1≤i≤n we have i i ` Q i=1 i=1 E(R )= X (V ), A∈Φ , i (cid:10) A A PRi(cid:11) for a set of root subschemes X , A∈Φ , over R . Hence we can assume that A PRi i E(R)=hX (V ), A∈Φ i A A P 8 from the very beginning. We show that [E(R),E(R)] contains any X (v), A ∈ Φ , v ∈ V , by induction on A P A |levA|. Take I ={s∈R | X (tsv)∈[E(R),E(R)] ∀ t∈R}. A By [7, Th. 2] for any u,u′ ∈V we have A X (u+u′)=X (u)X (u′) X (u ) A A A Y iA i i>0 for some u ∈ V . Hence by the inductive hypothesis I is an ideal of R. If I 6= R, let M i iA be a maximal ideal of R containing I. Let F denote both the localizationhomomorphism M R →R and the induced homomorphism G(R[Y,Z])→G(R [Y,Z]). By Lemma 5 there M M is an m > 0 such that we can represent the element F (X (ZmYv)) of G(R [Y,Z]) as a M A M product n F (X (ZmYv))= [h (Y,Z),g (Y,Z)], M A Y i i i=1 for some h (Y,Z),g (Y,Z)∈E(ZR [Y,Z]), 1≤i≤n. By [7, Lemma 15] there exist i i M h′(Y,Z), g′(Y,Z)∈E (R[Y,Z],ZR[Y,Z])≤E (R[Y,Z])=E(R[Y,Z]) i i P P and s ∈ R\M such that F (h′(Y,Z)) = h (Y,sZ) and F (g′(Y,Z)) = g (Y,sZ) for all i. M i i M i i Hence n F X ((sZ)mYv) =F [h′(Y,Z),g′(Y,Z)] . M(cid:0) A (cid:1) M(cid:16)Y i i (cid:17) i=1 Then by [7, Lemma 14] there exists t∈R\M such that n X ((tsZ)mYv)= [h′(Y,tZ),g′(Y,tZ)]. A Y i i i=1 Substituting Z by 1 and Y by an arbitrary element of R, we see that (ts)m ∈ I. But (ts)m ∈R\M, a contradiction. TheauthorsaresincerelygratefultoVictorPetrovandNikolaiVavilovfortheirinspiring comments on the subject of this paper. References [1] H.Azad,M.Barry,G.Seitz,Onthestructureof parabolic subgroups,Comm.inAlgebra 18 (1990), 551–562. [2] H. Bass, K-theory and stable algebra, Publ. Math. I.H.E´.S. 22 (1964), 5–60. [3] A. Bak, N. Vavilov, Structure of hyperbolic unitary groups I. Elementary subgroups, Algebra Colloquium 7 (2000), 159–196. [4] A. Borel, J. Tits, Groupes r´eductifs, Publ. Math. I.H.E´.S. 27 (1965), 55–151. [5] N. Bourbaki, Groupes et alg`ebres de Lie. Chapitres 4–6, Hermann, Paris, 1968. [6] M. Demazure, A. Grothendieck, Sch´emas en groupes, Lecture Notes in Mathematics, Vol. 151–153,Springer-Verlag,Berlin-Heidelberg-New York, 1970. 9 [7] V. Petrov, A. Stavrova,Elementary subgroups of isotropic reductive groups, St. Peters- burg Math. J. 20 (2009), 625–644. [8] V. Petrov, A. Stavrova, Tits indices over semilocal rings, Preprint POMI, N 2 (2009), 1–22. [9] A. Stavrova,Stroenije isotropnyhreduktivnyhgrupp,PhD thesis, St. Petersburg State University, 2009. [10] M.R. Stein, Generators,relations,andcoveringsofChevalleygroupsovercommutative rings, Amer. J. Math. 93 (1971), 965–1004. [11] A.A. Suslin, On the structure of the special linear group over polynomial rings, Math. USSR Izv. 11 (1977), 221–238. [12] J. Tits, Algebraic and abstract simple groups, Ann. of Math. 80 (1964), 313–329. [13] L.N.Vaserstein,Normal subgroups of orthogonal groups over commutative rings,Amer. J. Math. 110 (1988), 955–973. [14] L.N.Vaserstein,Normal subgroups of symplectic groups over rings,K-Theory2(1989), 647–673. 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.