ebook img

Representations of reductive normal algebraic monoids PDF

0.08 MB·
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 Representations of reductive normal algebraic monoids

Representations of reductive normal algebraic monoids StephenDoty 4 1 0 2 n a J 7 DedicatedtoLexRennerandMohanPutcha ] T AbstractTherationalrepresentationtheoryofareductivenormalalgebraicmonoid R (withone-dimensionalcenter)formsahighestweightcategory,inthesenseofCline, . Parshall, and Scott. This is a fundamental fact about the representation theory of h reductivenormalalgebraicmonoids.We surveyhow this resultwas obtained,and t a treatsomenaturalexamplescomingfromclassicalgroups. m [ 1 Introduction v 7 0 LetM beanaffinealgebraicmonoidoveranalgebraicallyclosedfieldK.See[10, 4 1 13, 12] for generalsurveysand backgroundon algebraic monoids.Assuming that . Misreductive(itsgroupGofunitsisareductivegroup)whatcanbesaidaboutthe 1 representationtheoryofMoverK? 0 4 Recallthatanyaffinealgebraicgroupissmoothandhencenormal(asavariety). 1 The normality of the algebraic group plays a significant role in its representation : v theory,forinstanceintheproofofChevalley’stheoremclassifyingtheirreducible i representations.Thusitseemsreasonableintryingtoextend(rational)representa- X tiontheoryfromreductivegroupstoreductivemonoidstolookfirstatthecasewhen r a themonoidM isnormal.Furthermore,evenincaseswhereagivenreductivealge- braicmonoidisnotnormal,onemayalwayspasstoitsnormalization,whichshould becloselyrelatedtotheoriginalobject. Renner[11]hasobtainedaclassificationtheoremforreductivenormalalgebraic monoids under the additional assumptions that the center Z(M) is 1-dimensional andthatMhasazeroelement.Renner’sclassificationtheoremdependsonanalge- LoyolaUniversityChicago,DepartmentofMathematicsandStatistics,ChicagoIL60660USA, e-mail:[email protected] 1 2 StephenDoty braic monoidversionof Chevalley’sbig cell, whichholdsforany reductiveaffine algebraicmonoid(withnoassumptionsaboutitscenterorazero).Asacorollaryof itsconstruction,Rennerderivesaveryusefulextensionprinciple[11,(4.5)]which isakeyingredientintheanalysis. 1 Reductive normal algebraicmonoids Let M be a linear algebraicmonoid overan algebraicallyclosed field K. In other words,M is a monoid(with unitelement1∈M) whichisalso anaffine algebraic varietyoverK,suchthatthemultiplicationmapm :M×M→M isamorphismof varieties.WeassumethatMisirreducibleasavariety.HencetheunitgroupG=M× (thesubgroupofinvertibleelementsofM)isaconnectedlinearalgebraicgroupover K andGisZariskidenseinM. 1.1. Associated with M is its affine coordinate algebra K[M], the ring of regular functionsonM.ThereexistK-algebrahomomorphisms D :K[M]→K[M]⊗ K[M], e :K[M]→K K called comultiplication and counit, respectively. For a given f ∈K[M], we have e (f)= f(1);furthermore,ifD (f)=(cid:229) r f ⊗f′then f(m m )=(cid:229) r f(m )f′(m ), i=1 i i 1 2 i=1 i 1 i 2 forallm ,m ∈M.ThemapsD ,e makeK[M]intoabialgebraoverK.Thismeans 1 2 thattheysatisfythebialgebraaxioms: (1) (id⊗D )◦D =(D ⊗id)◦D , (2) (e ⊗id)◦D =id=(id ⊗e )◦D wherej ⊗j ′denotesthemapa⊗a′7→j (a)j ′(a′). We note that the commutative bialgebra (K[M],D ,e ) determines M, as the set Hom (K[M],K) of K-algebra homomorphismsfrom K[M] into K. The multi- K−alg plication on this set is defined by j ·j ′ =(j ⊗ j ′)◦D and the identity element isjustthecounite .OneeasilyverifiesthatthisreconstructsM fromitscoordinate bialgebraK[M]. Moregenerally,givenanycommutativebialgebra(A,D ,e )overK,onedefineson thesetM(A)=Hom (A,K)analgebraicmonoidstructurewithmultiplication K−alg m (j ,j ′)=j ·j ′=(j ⊗j ′)◦D .Thisgivesafunctor {commutativebialgebrasoverK}→{algebraicmonoidsoverK} which is quasi-inverse to the functor M 7→ K[M]. Thus the two categories are antiequivalent. 1.2. SinceGisdenseinM,therestrictionmapK[M]→K[G](givenby f 7→ f )is |G injective,sowemayidentifyK[M]withasubbialgebraoftheHopfalgebraK[G]of regularfunctionsonG. Representationsofreductivenormalalgebraicmonoids 3 1.3. AssumethatM isreductive;i.e.,itsunitgroupG=M× isreductiveasanal- gebraic group. Fix a maximal torus T in G. (Up to conjugationT is unique.) Let X(T)=Hom(T,K×)bethecharactergroupofT;thisistheabeliangroupofmor- phismsfromT intothemultiplicativegroupK×ofK.LetX∨(T)=Hom(K×,T)be theabeliangroupofcocharactersinto T. LetR⊂X(T)betherootsystem forthe pair(G,T)andR∨⊂X∨(T)thesystemofcoroots.Accordingtotheclassification of reductivealgebraicgroups,the reductivegroupG is uniquelydeterminedup to isomorphismbyitsrootdatum(X(T),R,X∨(T),R∨). 1.4. We now add the assumptionthat M is normalas a variety.LetD=T be the ZariskiclosureofT in M. ThenT ⊂D is anaffine torusembedding.LetX(D)= Hom(D,K) be the monoid of algebraic monoid homomorphisms from D into K. The restriction c of any c ∈X(D) is an element of X(T), so restriction defines |T a homomorphismX(D)→X(T).Since T isdenseinD,thismapisinjective,and thuswemayidentifyX(D)withasubmonoidofX(T).Rennerhasshownthatthe additionaldatumX(D)isallthatisneededtodetermineMuptoisomorphism,under theadditionalhypotheses(probablyunnecessary)thatthecenter Z(M)={z∈M:zm=mz, forallm∈M} is 1-dimensionaland that M has a zero element. (One can always add a zero for- mally,sothelastrequirementisinsubstantial.) Itturnsoutthattheset X(D)also determinesthe rationalrepresentationtheory ofthereductivenormalalgebraicmonoidM,inasensemadepreciseinSection3. 1.5. Notethatitiseasytoconstructreductivealgebraicmonoids.Startwitharatio- nal representationr :G→End (V) of a reductivegroupG in some vectorspace K V with dim V =n<¥ . Theimager (G)isa reductiveaffinealgebraicsubgroup K ofEnd (V)≃M (K),themonoidofalln×nmatricesunderordinarymatrixmul- K n tiplication. Desiring our monoid to have a center of at least dimension 1, we in- cludethescalarsK×asscalarmatrices,definingG tobethesubgroupofEnd (V) 0 K generated by r (G) and K×. Now we set M = G , the Zariski closure of G in 0 0 End (V)≃M (K).Thisisareductivealgebraicmonoid. K n Forexample,ifG=SL (K)andV isitsnaturalrepresentationthenG ≃GL (K) n 0 n and M=M (K). (In general,to obtaina monoidM closely related to the starting n groupG, oneshouldpickV to bea faithfulrepresentation.)Thereisnoguarantee thatthisprocedurewillalwaysproduceanormalreductivemonoid,butifnotthen onecanalwayspasstoitsnormalization. 2 Examples: symplecticandorthogonal monoids The paper [4] considered some more substantial examples of reductive algebraic monoids coming from other classical groups. LetV =Kn with its standard basis {e ,...,e }.Puti′=n+1−iforanyi=1,...,n. 1 n 4 StephenDoty 2.1. The orthogonal monoid. Assume the characteristic of K is not 2. Define a symmetricnondegeneratebilinearformh, ionV byputting (a) hei,eji=d i,j′ forany1≤i,j≤n. Hered isKronecker’sdeltafunction.LetJbethematrixofh, iwithrespecttothe basis{e ,...,e }.ThentheorthogonalgroupO(V)isthegroupoflinearoperators 1 n f ∈End (V)preservingtheform: K (b) O(V)={f ∈End (V):hf(v),f(v′)i=hv,v′i, allv,v′∈V}. K LetAbethematrixof f withrespecttothebasis{e ,...,e }.Thenwemayidentify 1 n O(V)withthematrixgroup T (c) O (K)={A∈M (K):A JA=J}. n n Thisis containedin thelargergroupGO (K), thegroupoforthogonalsimilitudes n (seee.g.,[9])definedby (d) GO (K)={A∈M (K):ATJA=cJ, somec∈K×}. n n NotethatGO (K)isgeneratedbyO (K)andK×.Wedefinetheorthogonalmonoid n n OM (K)tobe n (e) OM (K)=GO (K), n n theZariskiclosureinM (K).Thesemonoids(fornodd)werestudiedbyGrigor’ev n [8].In[4]thefollowingresultwasproved. 2.2Proposition. TheorthogonalmonoidOM (K)isthesetofallA∈M (K)such n n T T thatA JA=cJ=AJA ,forsomec∈K. Note that the scalar c∈K in the above is allowed to be zero, and the “extra” T T condition cJ =AJA is necessary. If c6=0 then it is easy to see that A JA=cJ T isequivalenttocJ=AJA ,butwhenc=0thisequivalencefails.Theequivalence meansthatwecouldjustaswellhavedefinedGO (K)by n GO (K)={A∈M (K):ATJA=cJ=AJAT, somec∈K×} n n whichisperhapsmoresuggestiveforthedescriptionofOM (K)givenabove. n 2.3. Thesymplecticmonoid.Assumethatn=dim V iseven,sayn=2m.Define K anantisymmetricnondegeneratebilinearformh, ionV byputting (a) hei,eji=eid i,j′ forany1≤i,j≤n. where e is 1 if i≤m and −1 otherwise. Let J be the matrix of h , i with respect i to the basis {e ,...,e }. Then the symplectic group Sp(V) is the group of linear 1 n operators f ∈End (V)preservingthebilinearform: K Representationsofreductivenormalalgebraicmonoids 5 (b) Sp(V)={f ∈End (V):hf(v),f(v′)i=hv,v′i, allv,v′∈V}. K LetAbethematrixof f withrespecttothebasis{e ,...,e }.Thenwemayidentify 1 n Sp(V)withthematrixgroup T (c) Sp (K)={A∈M (K):A JA=J}. n n ThisiscontainedinthelargergroupGSp (K),thegroupofsymplecticsimilitudes, n definedby (d) GSp (K)={A∈M (K):ATJA=cJ, somec∈K×}. n n NotethatGSp (K)isgeneratedbySp (K)andK×.Asintheorthogonalcase,we n n couldjustaswellhavedefinedGSp (K)by n GSp (K)={A∈M (K):ATJA=cJ=AJAT, somec∈K×} n n T T thankstotheequivalenceoftheconditionsA JA=cJandcJ=AJA incasec6=0. WedefinethesymplecticmonoidSpM (K)tobe n (e) SpM (K)=GSp (K), n n theZariskiclosureinM (K).In[4]thefollowingwasproved. n 2.4Proposition. ThesymplecticmonoidSpM (K)isthesetofallA∈M (K)such n n T T thatA JA=cJ=AJA ,forsomec∈K. Notethatthescalarc∈K in theaboveisallowedtobezero,andthecondition T cJ=AJA isnecessary,justasitwasintheorthogonalcase. 2.5. Sketch of proof. I want to briefly sketch the ideas involved in the proof of Propositions2.2and2.4.Fulldetailsareavailablein[4].Themethodofproofworks for any infinite field K (except that characteristic 2 is excluded in the orthogonal case).Wecontinuetoassumethatn=2miseveninthesymplecticcase. LetG=GO (K) or GSp (K) andlet M=OM (K) or SpM (K), respectively. n n n n LetT bethemaximaltorusofdiagonalelementsofG.Thenwehaveinclusions (a) T ⊂G⊂M andwedesiretoprovethatthelatterinclusionisactuallyanequality.Toaccomplish this, we consider the action of G×G on M given by (g,h)·m=gmh−1. Suppose thatwecanshowthateveryG×GorbitisoftheformGaG,forsomea∈T.Then itfollowsthat (b) M= GaG⊂G a∈T S and this givesthe opposite inclusion that provesPropositions2.2 and 2.4. In fact, as it turned out, the distinct a∈T in the above decompositioncan be taken to be certainidempotentsinT. 6 StephenDoty Thissuggeststheprogramthatwascarriedoutin[4],whichintheendleadsto additionalstructuralinformationonM: (i)ClassifyallidempotentsinT. (ii)ObtainanexplicitdescriptionofT. (iii)DeterminetheG×GorbitsinM. Part(i)iseasy.Forpart(ii)oneexploitstheactionofT onT byleftmultiplication and determines the orbits of that action. Part (iii) involves developing orthogonal andsymplecticversionsofclassicalGaussianelimination. 2.6. The normality question. It is clear from the equalities in Propositions 2.2 and2.4thatOM (K)andSpM (K)bothhaveone-dimensionalcentersandcontain n n zero.Whatisnotclear,andnotaddressedin[4],iswhetherornottheyarenormal asalgebraicvarieties. This question was recently settled in [6], where it is shown that SpM (K) is n alwaysnormal,whileOM (K)isnormalonlyin caseniseven.Moreprecisely,it n isshownin[6]thatwhenn=2miseven,OM+(K)andOM−(K)arebothnormal n n varieties.Here (a) OM (K)=OM+(K)∪OM−(K) n n n isthedecompositionintoirreduciblecomponents,whereOM+(K)isthecomponent n containingtheunitelement1. 3 Representation theory FromnowonweassumethatMisanarbitraryreductivenormalalgebraicmonoid, withunitgroupG=M×.Wewishtodescribesomeresultsof[5].Themainresult isthatthecategoryofrationalM-modulesisahighestweightcategoryinthesense ofCline–Parshall–Scott[1]. 3.1. We work with a fixed maximaltorusT ⊂G, and setD=T. We assume that dimZ(M)=1and0∈M.RecallthatrestrictioninducesaninjectionX(D)→X(T), sowemayidentifyX(D)withasubmonoidofX(T).WefixaBorelsubgroupBwith T ⊂B⊂GandletthesetR−ofnegativerootsbedefinedbythepair(B,T).Weset R+=−(R−),thesetofpositiveroots.WehaveR=R+∪R−.Let X(T)+={l ∈X(T):(a ∨,l )≥0, foralla ∈R+} betheusualsetofdominantweights.Wedefine X(D)+=X(T)+∩X(D). 3.2. By a (left) rationalM-modulewe meana linearaction M×V →V suchthat the coefficientfunctionsM→K of the actionare allin K[M]. Thisis the same as Representationsofreductivenormalalgebraicmonoids 7 havinga(right)K[M]-comodulestructureonV.Thismeansthatwehaveacomodule structuremap (a) D :V →V⊗ K[M]. V K SinceK[M]⊂K[G]themapD inducesacorrespondingmapV →V⊗ K[G]mak- V K ingV into a K[G]-comodule;i.e., a rational G-module. Thus, rationalM-modules mayalsoberegardedasrationalG-modules.AnyrationalM-moduleissemisimple whenregardedasarationalD-module,withcorrespondingweightspacedecompo- sition (b) V = l ∈X(D)Vl whereVl ={v∈V :d·v=l (d)v, allLd∈D}. RecallthatanyrationalG-moduleV is semisimplewhenregardedasa rational T-module,withcorrespondingweightspacedecomposition (c) V = l ∈X(T)Vl whereVl ={v∈V :t·v=l (t)v, allLt ∈T}.IfV isarationalM-modulethenthe weight spaces relative to T are the same as the weight spaces relative to D. So the weights of a rational M-module all belong to X(D). Conversely, we have the following. 3.3Lemma. IfV isarationalG-modulesuchthat {l ∈X(T):Vl 6=0}⊂X(D) thenV extendsuniquelytoarationalM-module. ThisisprovedasanapplicationofRenner’sextensionprinciple,whichisaver- sionofChevalley’sbigcellconstructionforalgebraicmonoids. 3.4Remark. Aspecialcaseofthelemma(forthecaseM=M (K))canbefound n in[7]. 3.5. Nextoneneedsanotionofinductionforalgebraicmonoids,i.e.,aleftadjoint torestriction.Theusualdefinitionofinducedmoduleforalgebraicgroupsdoesnot work for algebraic monoids. Instead, we use the following definition. LetV be a rationalL-modulewhereLisanalgebraicsubmonoidofM.WedefineindMV by L indMV ={f ∈Hom(M,V): f(lm)=l·f(m), alll∈L,m∈M}. L Thisisviewedasa rationalM-moduleviarighttranslation.Onecancheckthatin caseL,Marealgebraicgroupsthenthisisisomorphictotheusualinducedmodule. ItiswellknownthattheBorelsubgroupBhasthedecompositionB=TU,where U isits unipotentradical.Givena characterl ∈X(T)oneregardsK asa rational T-modulevial ;thisisoftendenotedbyKl .OneextendsKl toarationalB-module bylettingU acttrivially.Similarly,wehavethedecompositionB=DU.Ifl ∈X(D) 8 StephenDoty thenwehaveKl asabove,andagainwemayregardthisasarationalB-moduleby lettingU acttrivially. NowwecanformulatetheclassificationofsimplerationalM-modules. 3.6Theorem. LetM be a reductivenormalalgebraicmonoid.Letl ∈X(D)and letKl betherationalB-moduleasabove.Then (a)indMKl 6=0ifandonlyifl ∈X(D)+. B (b)IfindMKl 6=0thenitssocleisasimplerationalM-module(denotedbyL(l )). B (c)ThesetofL(l )withl ∈X(D)+givesacompletesetofisomorphismclasses ofsimplerationalM-modules. Letl ∈X(T).Let(cid:209) (l )=indGBKl .Itiswellknownthat(cid:209) (l )6=0ifandonlyif l ∈X(T)+.Thefollowingisakeyfact. 3.7Lemma. Ifl ∈X(D)+ thenindMBKl =(cid:209) (l )=indGBKl . 3.8. Nowweconsidertruncation.Letp ⊂X(T)+.GivenarationalG-moduleV,let Op V betheuniquelargestrationalsubmoduleofV withthepropertythatthehighest weightsofallitscompositionfactorsbelongtop .The(leftexact)truncationfunctor Op wasdefinedbyDonkin[2]. RecallthatX(T)ispartiallyorderedbyl ≤m ifm −l canbewrittenasasumof positiveroots;thisissometimescalledthedominanceorder.Asubsetp ofX(T)+is saidtobesaturatedifitispredecessorclosedunderthedominanceorderonX(T). Inotherwords,p issaturatedifforanym ∈p andanyl ∈X(T)+,l ≤m implies thatl ∈p . In order to show that the category of rational M-modules is a highest weight category,wearegoingtotakep =X(D)+.Weneedthefollowingobservation. 3.9Lemma. Thesetp =X(D)+isasaturatedsubsetofX(T)+. For l ∈X(T)+, let I(l ) be the injective envelope of L(l ) in the category of rationalG-modules.Forl ∈X(D)+ letQ(l )betheinjectiveenvelopeofL(l )in thecategoryofrationalM-modules.Thefollowingrecordstheeffectoftruncation onvariousclassesofrationalG-modules. 3.10Theorem. Letp =X(D)+.Foranyl ∈X(T)+ wehavethefollowing: (cid:209) (l ) ifl ∈p (a) Op (cid:209) (l )= (0 otherwise. Q(l ) ifl ∈p (b) Op I(l )= (0 otherwise. (c) Op K[G]=K[M]. Note that K[M] is regarded as a rational M-module via right translation. A (cid:209) - filtrationforarationalG-moduleV isanascendingseries 0=V ⊂V ⊂···⊂V ⊂V =V 0 1 r−1 r Representationsofreductivenormalalgebraicmonoids 9 ofrationalsubmodulessuchthatforeach j=1,...,r, thequotientV /V isiso- j j−1 morphictosome(cid:209) (l ).WheneverV isarationalG-modulewitha(cid:209) -filtration,let j (V :(cid:209) (l )) bethe numberofl forwhichl =l . Thisnumberis independentof j j thefiltration. The proof of the above theorem, which relies on results of [3], also shows the followingfacts. 3.11Corollary. (a)Letl ∈p =X(D)+.ThemoduleQ(l )hasa(cid:209) -filtration.Fur- thermore,itsatisfiesthereciprocityproperty (Q(l ):(cid:209) (m ))=[(cid:209) (m ):L(l )] foranym ∈X(D)+,where[V :L]standsforthemultiplicityofasimplemoduleLin acompositionseriesofV. (b)ThemoduleK[M]hasa(cid:209) -filtration.Moreover,(K[M]:(cid:209) (l ))=dim (cid:209) (l ) K foreachl ∈X(D)+. 3.12. Fromtheseresultsoneobtainstheimportantfactthatthecategoryofrational M-modulesisahighestweightcategory,inthesenseof[1].Inparticular,onealso seesthatdim Q(l )isfinite,foranyl ∈X(D)+.(Incontrast,itiswellknownthat K dim I(l )isinfinite.) K 3.13. Itisalsoshownin[5],exploitingtheassumptionthatZ(M)isone-dimensional, thatthecategoryofrationalM-modulessplitsintoadirectsumof‘homogeneous’ subcategories each of which is controlled by a finite saturated subset of X(D)+. Fromtheresultsof[1]itthenfollowsthatthereisafinitedimensionalquasihered- itaryalgebraineachhomogeneousdegree,whosemodulecategoryispreciselythe homogeneoussubcategoryin thatdegree.Details are givenin [5], where it is also shown that the quasihereditary algebras in question are in fact generalized Schur algebrasinthesenseofDonkin[2]. References 1. E.Cline,B.Parshall,andL.Scott:Finitedimensionalalgebrasandhighestweight categories,J.ReineAngew.Math.391(1988),85–99. 2. S.Donkin:OnSchuralgebrasandrelatedalgebrasI,J.Algebra104(1986),310–328. 3. S.Donkin:Goodfiltrationsofrationalmodulesforreductivegroups,Proc.SymposiaPure Math.(PartI)47(1987),69–80. 4. S.Doty:Polynomialrepresentations,algebraicmonoids,andSchuralgebrasofclassical type,J.PureAppliedAlgebra123(1998),165–199. 5. S.Doty:Polynomialrepresentations,algebraicmonoids,andSchuralgebrasofclassical type,J.PureAppl.Algebra123(1998),165–199. 6. S.DotyandJunHu:Schur–Weyldualityfororthogonalgroups,Proc.Lond.Math.Soc.(3) 98(2009),679–713. 7. E.FriedlanderandA.Suslin,Cohomologyoffinitegroupschemesoverafield,Invent. Math.127(1997),209–270. 10 StephenDoty 8. D.Ju.Grigor’ev,AnanalogueoftheBruhatdecompositionfortheclosureoftheconeofa Chevalleygroupoftheclassicalseries(Russian),Dokl.Akad.NaukSSSR257(1981), 1040–1044. 9. M.–A.Knus,A.Merkurjev,M.Rost,andJ.–P.Tignol:Thebookofinvolutions,American MathematicalSocietyColloquiumPublications,44,AmericanMathematicalSociety, Providence,RI,1998. 10. M.Putcha:LinearAlgebraicMonoids,LondonMath.Soc.LectureNotesSeries133, CambridgeUniv.Press,1988. 11. L.Renner:Classificationofsemisimplealgebraicmonoids,Trans.Amer.Math.Soc.292 (1985),193–223. 12. LexE.Renner:Linearalgebraicmonoids,EncyclopaediaofMathematicalSciences,134; InvariantTheoryandAlgebraicTransformationGroups,V.Springer–Verlag,Berlin,2005. 13. L.Solomon:Anintroductiontoreductivemonoids,inSemigroups,formallanguagesand groups(York,1993),pp.295–352,NATOAdv.Sci.Inst.Ser.CMath.Phys.Sci.,466, KluwerAcad.Publ.,Dordrecht,1995.

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.