LPTHE-00-04 Dual Baxter equations and quantization of Affine Jacobian. 0 0 0 2 n F.A. Smirnov0 a J 1 2 LaboratoiredePhysiqueThe´oriqueetHautesEnergies1 1 Universite´PierreetMarieCurie,Tour16,1er e´tage,4placeJussieu v 75252ParisCedex05,France 2 3 0 1 0 0 0 Abstract. A quantumintegrablemodelis consideredwhichdescribesa quantization / of affinehyper-ellipticJacobian. Thismodelis shown to possess the propertyof du- h p ality: a dual model with inverse Planck constant exists such that the eigen-functions - ofits Hamiltonianscoincidewith theeigen-functionsofHamiltoniansofthe original h model. Weexplainthatthisdualitycanbeconsideredasdualitybetweenhomologies t a andcohomologiesofquantizedaffinehyper-ellipticJacobian. m : v i X r a 0MembreduCNRS 1Laboratoireassocie´auCNRS. 1 1 Introduction. TherelationbetweentheaffineJacobianandintegrablemodelsiswellknown(cf. [1]). Inthepaper[2]wehaveshownthatthealgebraoffunctionsontheaffineJacobianis generatedbyactionofhamiltonianvectorfieldsfromfinitenumberoffunctions. The latter functions are coefficients of highest non-vanishing cohomologies of the affine Jacobian.Actually,theideathatsuchdescriptionofthealgebraoffunctionsispossible appearedfromthepaper[3]whichconsidersthestructureofthealgebraofobservables forquantumandclassicalTodachain. Inthepresentpaperwegivequantumversionof[2].Aquantummechanicalmodel is formulated which gives a quantization of the affine Jacobian. As usual in Quan- tum Mechanicswe can describe notthe variety itself but the algebra of functionson it(observables). We needtoshowthatthe quantumalgebraofobservablespossesses essentialpropertyofcorrespondingclassicalalgebraoffunctions.Inourcasethisprop- ertyisthepossibilityofcreatingeveryobservablefromafinitenumberofobservables (cohomologies)byactionofHamiltonians. In the process of realization of this program we find the Baxter equations which describethespectrumofthemodel.Ithappensthattheseequationspossesstheproperty ofduality:thereisdualmodelwithinversePlanckconstantforwhichtheeigen-vectors are the same. The algebras of observables of two dual models commute. The next ingredientofourstudyisthemethodofseparationofvariablesdevelopedbySklyanin [4]. Using thismethodwe presentthematrixelementsofanyobservablein termsof certainintegrals. WeshowthattheintegralsinquestionareexpressesintermsofdeformedAbelian integrals(cf.[3,5]).Theobservablesforbothdualmodelsaredefinedintermsofcoho- mologies.Themostbeautifulfeatureofourconstructionisthatinthesecohomologies enter the integrals for matrix elements in such a way that the cohomologies of dual modelplayroleofhomologiesfororiginaloneandviseaversa.Weconsiderthisrela- tionbetweenweek-strongdualityinquantumtheorywithdualitybetweenhomologies andcohomologiesasthemostimportantconclusionofthispaper. 2 Affine Jacobian. Inthissectionwebrieflysummarizenecessaryfactsconcerningrelationbetweeninte- grablemodelsandalgebraicgeometryfollowingthepaper[2].Thereasonforrepeating certainfactsfrom[2]isthatweshallneedtheminslightlydifferentsituation. Consider2 2matrixwhichdependspolynomiallyontheparameterz: × a(z) b(z) m(z)= c(z) d(z) (cid:18) (cid:19) 2 wherethematrixelementsarepolynomialsoftheform: a(z)=zg+1+a zg+ +a , (1) 1 g+1 ··· b(z)=zg+b zg−1+ +b , 1 g ··· c(z)=c zg+c zg−1 +c , 2 3 g+2 ··· d(z)=d zg−1+d zg−2 +d 2 3 g+1 ··· InthetheaffinespaceC4g+2withcoordinatesa , ,a ,b , ,b ,c , ,c , 1 g+1 1 g 2 g+2 ··· ··· ··· d , ,d considerthe(2g+1)-dimensionalaffinevariety definedasquadric 2 g+1 ··· M f(z) a(z)d(z) b(z)c(z)=1 (2) ≡ − Weconsiderthissimplestsituation,butinprincipleitispossibletoputarbitrarypoly- nomialofdegree2ginRHS. Onthequadric letusconsiderthesectionsJ (t)definedbytheequations: aff M a(z)+d(z)=t(z) (3) wheret(z)isgivenpolynomialoftheform: t(z)=zg+1+zgt + +t , (4) 1 g+1 ··· The notation J (t) stands for affine Jacobi variety. The definition of affine Jacobi aff varietyanditsequivalencetoJ (t)describedabovearegivenintheAppendixA.We aff includeAppendixAbecausethereisminordifferencewiththesituationconsideredin [1] and [2]. The variety is foliated into the affine Jacobians J (t). Mechanical aff M modeldescribedbelowprovidesacleverwayofdescribingthisfoliation. Wewouldliketounderstandthegeometricalmeaningofquantumintegrablemod- els. The general philosophy teaches that in order to describe the quantization of a manifoldonehasto deformthe algebraoffunctionsonthismanifoldpreservingcer- tainessentialpropertiesofthisalgebra. Theclassicalalgebramustallowthe Poisson structureinorderthatquantizationispossible. CertainPoissonbracketsforthecoefficientsofmatrixm(z)canbeintroduced.We donotwritethemdownexplicitly,ifneededtheycanbeobtainedtakingclassicallimit ofthecommutationrelations(14).Thealgebra =C[a , ,a ,b , ,b ,c , ,c ,d , ,d ] 1 g+1 1 g 2 g+2 2 g+1 A ··· ··· ··· ··· becomesaPoissonalgebra.ThemostimportantpropertiesofthisPoissonstructureare b thefollowing.First,thecoefficientsofthedeterminantf(z)belongtothecenterofthe Poissonalgebra,so,theequation(2)isconsistentwithPoissonstructure. Second,the tracet(z)generatescommutativesub-algebra: t(z),t(z′) =0 { } It can be shown, actually, that the coefficientt of the trace belongsto the center, g+1 it is convenientto put t = ( 1)g+12. The subject of our study is the algebra of g+1 − functionsonthe : M = A A f(z)=1, t =( 1)g+12 g+1 { − } b 3 onwhichthePoissonstructureiswelldefined. ThePoissoncommutativealgebrageneratedbythecoefficientst , ,t iscalled 1 g ··· thealgebraofintegralsofmotion.Introducethecommutingvector-fields ∂ g = t ,g , i=1, g i i { } ··· Thevector-fields∂ describemotionalongthesub-varietiesJ (t). j aff Onecanthinkofthesevector-fieldsas ∂ ∂ = j ∂τ j where τ are ”times” correspondingto the integrals of motion t . Define the ring of j j integralsofmotion =C[t , ,t ] (5) 1 g T ··· IntroducethespaceofdifferentialformsCk withbasis xdτ dτ , x i1 ∧···∧ ik ∈A andthedifferential d=∂ dτ j j ConsidercorrespondingcohomologiesHk. Inthepaper[2]theargumentsaregivenin favourofthefollowing Conjecture1. ThecohomologiesHk arefinite-dimensionaloverthering ,theyare T isomorphictothecohomologiesoftheaffinevarietyJ (t)withtingenericposition. aff On the algebra and on the spaces Ck one can introduce degree [2]. Take the ba- A sis of Hg considered as a vector space over which is composed of homogeneous T representatives Ω =g dτ dτ α α 1 g ∧···∧ whereαtakesfinitenumberofvalues. Thefactoffoliationof intovarietiesJ (t) aff M correspondstothefollowingstatementconcerningthealgebra [2] A Proposition1. Everyelementxof canbepresentedintheform A x= p (∂ , ,∂ )g (6) α 1 g α ··· α X wherep (∂ , ,∂ )arepolynomialsof∂ , ,∂ withcoefficientsin . α 1 g 1 g ··· ··· T Therepresentation(6)isnotunique,theequations p (∂ , ,∂ )g =0 (7) α 1 g α ··· α X arecountedbyHg−1[2]. The formula (6) can be useful only if we are able to control the cohomologies. Concerningthesecohomologiesweadoptseveralconjecturesfollowing[2]. 4 3 Conjectured form of cohomologies. TheaffinevarietyJ (t)allowsthefollowingdescription. Considerthehyper-elliptic aff curveX ofgenusg: w2 t(z)w+1=0 (8) − Thiscurvehastwopointsoverthepointz = whichwedenoteby ±. ∞ ∞ Consideramatrixm(z)satisfying(2).Takethezerosofb(z): g b(z)= (z z ) j − j=1 Y and w =d(z ) j j Obviouslyz ,w satisfytheequationofthecurveX (8).Thusm(z)definesapoint j j P (divisor)onthesymmetrizedg-thpowerX[g]ofthecurveX. Thedivisor consists P ofthepointsp =(z ,w ) X. j j j ∈ Oppositelyone can reconstructm(z)starting formthe divisor . Corresponding P mapissingular,thesingularitiesbeinglocatedon D = p =σ(p )forsomei,jorp = ± forsomei (9) i j i {P | ∞ } Whereσishyper-ellipticinvolution.ThusthealternativedescriptionofJ (t)is aff J (t)=X[g] D aff − ConsiderthemeromorphicdifferentialsonX withsingularitiesat ±. Wechose ∞ thefollowingbasisofthesedifferentials: dz µ (p)=zg+k , g k 0 k y − ≤ ≤ d dz µ (p)= y (zk−g−1y) , k 1 (10) k dz ≥ y ≥ (cid:2) (cid:3) where p = (z,w), y = 2w t(z), [ ] means that only non-negative degrees of ≥ − Laurentseriesinthebracketsaretaken. Theform µ = µ (p ) k k i i X is viewed as a form on J (t). Iteis easy to see that the forms µ (hence µ ) with aff k k k g+1areexact.ConsiderthespaceWm withthebasis: ≥ e Ω =µ µ k1,···,km k1 ∧···∧ km where g k g. Following[2]weadoptthe − ≤ j ≤ e e 5 Conjecture2. Wehave Wm Hm = (11) σ Wm−2 ∧ where g σ = µ µ j −j ∧ j=1 X Accordingto(9)thesingularitiesofdiffereentialeformsoccureitheratpi =σ(pj)orat p = ±. Thenon-trivialessenceofConjecture2isthatthefirstkindofsingularities i ∞ can be eliminated by adding exact forms. There are (g 1)-forms singular at p = i − σ(p )suchthatthesesingularitiesdisappearafterapplyingd. Thisistheoriginofthe j spaceσ Wk−2[2]. ∧ Considerbrieflythedualpicture. Ontheaffinecurvewithpuncturesat ± there ∞ are 2g + 1 non-trivial cycles δ with k = g, ,g. The cycles δ , k < 0 are k k − ··· a-cycles, the cycles δ , k > 0 are b-cycles and δ is the cycle around +. One k 0 ∞ definesthecyclesδ onthesymmetricalpoweroftheaffinecurve. The -operationis k ∧ introducedforthesecyclesbydualitywithcohomologies.Thenon-trivialconsequence ofConjecture2 isethateverycycle onJ (t) canbeconstructedbywedgingδ . The aff k formuladualto(11)is W e H = m (12) m σ′ W m−2 ∧ whereW isspannedby m ∆ =δ δ k1,···,km k1 ∧···∧ km and ge e σ′ = δ δ j −j ∧ j=1 X Weneedtofactorizeoverσ′ W becauesethee2-cycleσ′intersectswithD. m−2 ∧ LetusreturntotherelationofHg tothealgebra . Noticethat A dτ dτ µ µ Ω 1 g 1 g ∧···∧ ≃ ∧···∧ ≡ Thefunctions x =Ωe−1Ω e k1,···,kg k1,···,kg are symmetricpolynomialsofz , ,z . Recall thatb , ,b arenothingbutele- 1 g 1 g ··· ··· mentarysymmetricpolynomialsofz , ,z .Hencethecoefficientsofcohomologies 1 g ··· havetheform: g =g (b , ,b ) α α 1 g ··· ThedimensionofHg isdeterminedbyConjecture2: α=1, , 2g+1 2g+1 , ··· g − g−2 Theequations(7)areconsequencesofth(cid:0)efollo(cid:1)win(cid:0)gone(cid:1)s g ∂ Ω−1(µ Ω ) =0, Ω Wg−1 (13) k −k∧ k1,···,kg−1 ∀ k1,···,kg−1 ∈ k=1 X (cid:0) (cid:1) 6 4 Quantization of affine Jacobian. Let us consider a quanization of algebra . The parameter of deformation (Planck A constant)isdenotedbyγ,weshallalsouse q =eiγ Considerthe2 2matrixm(z)withnoncommutingentries. Supposethatthedepen- × dence on the spectral parameter z is exactly the same as in classical case (1). The variablesa ,b ,c ,d aresubjecttocommutationrelationswhicharesummarizedas j j j j follows: r (z ,z )m (z )k (z )s m (z )k (z )= 21 1 2 1 1 12 1 12 2 2 21 2 =m (z )k (z )s m (z )k (z )r (z ,z ) (14) 2 2 21 2 21 1 1 12 1 12 1 2 where usual conventionsare used: the equation (14) is written in the tensor product C2 C2, a = a I, a = I 2, a = Pa P where P is the operation of 1 2 21 12 perm⊗uattions.TheC⊗-numbermatri⊗cesr,k,sare: z qz z +qz r (z ,z )= 1− 2 (I I)+ 1 2 (σ3 σ3)+ 12 1 2 1 q ⊗ 1+q ⊗ − +2(z σ− σ++z σ+ σ−), 1 2 ⊗ ⊗ k (z)=I (I σ3)+ q−σ3 +z(q2 1)σ− (I +σ3), 12 ⊗ − − ⊗ s =I I (q q−1)σ(cid:16)− σ+ (cid:17) (15) 12 ⊗ − − ⊗ These commutation relations are important because they respect the form of matrix m(z)prescribedby(1),weshallexplainhowtheyarerelatedtomoreusualr-matrix relationsinthenextsection. Definethepolynomials: t(z)=qa(z)+q2d(z) z(q2 1)b(z) − − f(z)=qd(z)t(zq−2) q2d(z)d(zq−2) qb(z)c(zq−2) (16) − − The algebra (q) is generated by a , ,a , b , ,b , c , ,c , 1 g+1 1 g 2 g+2 A ··· ··· ··· d , ,d ,Thepolynomialf(z)belongstothecenterof (q). Thecoefficientsof 2 g+1 ··· A t(z)arecommbuting,actually,t belongstothecenterof (q). Wedefine: g+1 A b (q) b (q)= A A f(z)=1, t =( 1)g+12 g+1 { − } b The non-commutativealgebra (q) defines a quantizationof the algebra of function A onthequadric . However,wecannotdefinedirectlythequantizationofthealgebra M offunctionsontheaffineJacobianbecausethecoefficientsoft(z)arenotinthecenter of (q). Whatwe candois todescribethequantumversionofProposition1 andof A descriptionofcohomologies.Theexpositionwillbemoredetailedthanintheclassical case. 7 Likeinthepaper[3]weacceptthefollowing Conjecture3. Thealgebra (q)isspannedaslinearspacebyelementsoftheform: A x=p (t , ,t )g(b , ,b )p (t , ,t ) (17) L 1 g 1 g R 1 g ··· ··· ··· wherep (t , ,t ), g(b , ,b ), p (t , ,t )arepolynomials. L 1 g 1 g R 1 g ··· ··· ··· Wewerenotabletoprovethisstatement,however,sincethealgebra (q)isgradedwe A cancheckitdegreebydegree.Thishasbeendoneuptodegree8. Noticethesimilarity between the representation (17) and the representation for spin operators proved in [6]. Conjecture 3 implies that certain generalization of the results of [6] is possible. In fact the formula (17) is similar to the formula (6): we can either symmetrize or anti-symmetrize t in (17) which correspondsin classics to multipilation by t or to j j applying∂ . Inordertohavecompleteagreementwithclasicalcasewehavetoshow j thatonlyfinitelymanydifferentpolynomialsg(b , b )(cohomologies)createentier 1 g ··· algebra (q). A Noticethatthecommutationrelations(14)implyinparticularthat [b(z),b(z′)]=0 whichmeansthatwehavethecommutativefamilyofoperatorsz definedby j b(z)= (z z ) j − Y So, every polynomialsg(b , ,b ) can be considered as symmetric polynomialof 1 g ··· z andviceversa. j Itisveryconvenienttouse thefollowingformaldefinitions. Considerthering T definedin(5).By kwedenotethespaceofanti-symmetricpolynomialsofkvariables V suchthattheirdegreeswithrespecttoeveryvariableisnotlessthan1withcoefficients in . Inotherwords k isthespacespannedbythepolynomials: T ⊗T V p h p p (t , ,t )h(z , ,z )p (t′, ,t′) L· · R ≡ L 1 ··· g 1 ··· k R 1 ··· g wherehisanti-symmetric,vanishingwhenoneofz vanishes. Thefollowingopera- j tionscanbedefined. 1. Multiplicationbyt andt′. j j 2. Operation : k l k+l whichisdefinedasfollows: ∧ V ⊗V →V (p h p ) (p′ h′ p′ )=p p′ (h h′) p p′ L· · R ∧ L· · R L L· ∧ · R R where (h h′)(z , ,z )= 1 k+l ∧ ··· 1 = ( 1)π h(z , ,z )h′(z , ,z ) k!l! − π(1) ··· π(k) π(k+1) ··· π(k+l) π∈XSk+l Wehaveamap g χ (q) V −→A 8 definedonthebasiselementsas h(z , ,z ) χ(p h p )=p (t , ,t ) 1 ··· g p (t , ,t ) L· · R L 1 ··· g z (z z ) R 1 ··· g i i<j i− j andcontinuedlinearly. TheConjecture3QstatesQthatthismapissurjective. Wewantto describethekernelofthemapχ. First, consider the space 1. The elements of this space are polynomials of one V variable z with coefficients in . In Appendix B we describe certain basis in T ⊗ T 1 considered as a linear space over . The basis in question consists of the V T ⊗ T polynomials: s with k g such that the degree of s with respect to z equals k k ≥ − g+k+1. Thekernelofχisthejointofthreesub-spaces,letusdescribethem. 1. Fork g+1wehave: ≥ χ s g−1 =0 (18) k ∧V (cid:0) (cid:1) 2. Considerc 2definedas ∈V g c= s s j −j ∧ j=1 X wehave χ c g−2 =0 (19) ∧V (cid:0) (cid:1) 3. Considerd 1definedas ∈V d=(t t′)s j − j −j wehave χ d k−1 =0 (20) ∧V Theconstructionofthespace (cid:0) (cid:1) g V (q) Ker(χ) ≃A is in complete correspondence with the classiccal case. In classics we start with all the 1-forms µ . Imposing (18) corresponds to throwing away the exact forms and k working with µ for k = g ,g only. Imposing (19) correspondsto factorizing k − ··· overσ Wk−2 inclassics. Finally,(20)correspondstotheequation(13). e ∧ Theoriginoftheequations(18),(19),(20)willbeexplainedintheSection9.There e shouldbepurelyalgebraicmethodofproovingtheseequations,butwedonotknowit. Itisimportantto mentionthatacceptingConjecture3 weare forcedto concludethat thekernelofχiscompletelydescribedbytheequations(18),(19),(20).Thisisproved bycalculationofcharacterssimilarlytothatof[3]. 9 5 The realization of A(q). Wewanttodescribearealizationofthealgebra (q)inaspaceoffunctions.Consider A thequantummechanicalsystemdescribedbytheoperatorsx withj =1, ,2g+2 j ··· andy(zeromode). Theoperatorsx andy areself-adjoint,theysatisfythecommuta- j tionrelations: x x =q2x x k <l, k l l k yx =q2x y k k k ∀ Thehamiltonianofthesystemis 2g+2 h=q−1 x x−1 k k−1 k=1 X where x qyx 2g+3 1 ≡ PhysicallythismodeldefinesthesimplestlatticeregularizationofthechiralBosefield withmodifiedenergy-momentumtensor. It is usefulto double the numberof degreesof freedom. Consider the algebra A generatedbytwooperatorsuandvsatisfyingthecommutationrelations: uv =qvu TakethealgebraA⊗(2g+2) theoperatorsu , v (j = 1, ,2g+2)aredefinedasu j j ··· andvactinginj-thtensorcomponent.Theoriginaloperatorsx areexpressedinterms i ofu ,v asfollows: i i k−1 2g+2 x =v u−2, y = u k k j j j=1 j=1 Y Y Considerthe“monodromymatrix” a(z) b(z) m(z)= =l (z) l (z) (21) 2g+2 1 c(z) d(z)! ··· e e wherethel-operatorfsare e e 1 zu qvu l(z)= − (22) √z zv−1u−1 0 (cid:18) (cid:19) This is a particular case of more genaral l-operator l(z,κ) in which the last matrix elementisnot0butκzu,themodelcorrespondingtothelatterl-operatorisasubject ofstudyinaseriesofpapers[7]. Thematrixelementsofthematrixm(z)satisfythecommutationrelations: r (z ,z )m (z )m (z )=m (z )m (z )r (z ,z ) (23) 12 1 2 1 1 2 2 2 2 1 1 12 1 2 f f f f 10