ebook img

The Eynard--Orantin recursion for simple singularities PDF

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

Preview The Eynard--Orantin recursion for simple singularities

THE EYNARD–ORANTIN RECURSION FOR SIMPLE SINGULARITIES TODORMILANOV 5 1 0 Abstract. Accordingto[9]and[19],theancestorcorrelatorsofanysemi-simple 2 cohomologicalfieldtheorysatisfylocalEynard–Orantinrecursion. Inthispaper, n we prove that for simple singularities, the local recursion can be extended to a a globalone. Thespectralcurveoftheglobalrecursionisaninterestingfamilyof J RiemannsurfacesdefinedbytheinvariantpolynomialsofthecorrespondingWeyl 5 group. Wealsoprovethatforgenus0and1,thefreeenergiesintroducedin[10] 1 coincideuptosomeconstantfactorswithrespectivelythegenus0and1primary ] potentialsofthesimplesingularity. G A . h t Contents a m 1. Introduction 1 [ 2. Analyticextensionofthekernel ofthelocal Eynard–Orantinrecursion 8 1 3. From localto global 11 v 4. Thefree energies andtheprimarypotentialsingenus0 and 1 15 7 7 AppendixA. Thespaceofholomorphic1-forms 21 6 References 23 3 0 . 1 0 5 1. Introduction 1 : v 1.1. Motivation. According to [9] and [19], the ancestor correlators of any semi- i X simple cohomological field theory satisfy local Eynard–Orantin recursion. The r term local refers to thefact that thespectral curveis justa disjointunionof several a discs. If we are interested in computing specific ancestor Gromov–Witten (GW) invariants in terms of Givental’s R-matrix, then the local recursion is all that we need. However,ifwewanttounderstandthenatureofthegeneratingfunctionfrom the point of view of representations of vertex algebras (see [2]) and integrable sys- tems (see [15]), then it is important to extend the local recursion to a global one, i.e., extend the spectral curve and the recursion kernel to global objects (see [3]). The appropriate spectral curve however, looks quite complicated in general, since 2000Math. Subj.Class. 14D05,14N35,17B69. Keywordsandphrases: periodintegrals,Frobeniusstructure,Gromov–Witteninvariants,vertex operators. 1 2 TODORMILANOV it is parametrized by period integrals. In particular, finding whether an appropriate generalization of the global Eynard–Orantin recursion [10, 3] exists in the settings of semi-simple cohomological field theories is a very challenging and important problem. In this paper we would like to solve the above problem for simple singularities. Inthiscase,thespectralcurveturnsouttobeaclassicalRiemannsurfacedefinedby the invariant polynomials of the monodromy group of non-maximal degree, while theinvariantpolynomialofmaximaldegreedefinesabranchedcoveringofP1. This brunched covering was also studied by K. Saito (unfortunately he did not write a text),becauseitisacoveringofwhathecalledaprimitivedirectionin thespaceof miniversaldeformationsofthesingularity. I think that the spectral curve for simple singularities is important also in the representation theory of the corresponding simple Lie algebras. For example, one canobtainasimpleproofofthewellknownfactthattheorderoftheWeylgroupis theproductofthedegrees oftheinvariantpolynomials(seeAppendixA). Finally, after a small modification our argument should work also for all finite reflection groups. The spectral curve is a certain family of Hurwitz covers of P1 parametrized by an open subset in the space of orbits of the corresponding reflec- tion group. It would be interesting to obtain the Frobenius structure on the space of orbits of the reflection group (see [8, 22]) via the construction of a Frobenius structureon themodulispaceofHurwitzcovers(see [8], Lecture5). 1.2. Singularity theory. Let f C[x ,x ,x ] be a weigthed-homogeneous poly- 1 2 3 ∈ nomial that has an isolated critical point at 0 of ADE type. Such polynomials cor- respond to the ADE Dynkin diagrams and are listed in Table 1, where we have included also the Coxeter number h and the Coxeter exponents of the correspond- ingsimpleLiealgebra. Table1. Simplesingularities Type f(x) Exponents h A xN+1+x2+x2 1,2,...,N N+1 N 0 1 2 D xN 1+x x2+x2 1,3,...,2N 3,N 1 2N 2 N 0− 0 1 2 − − − E x4+x3+x2 1,4,5,7,8,11 12 6 0 1 2 E x3x +x3+x2 1,5,7,9,11,13,17 18 7 0 1 1 2 E x5+x3+x2 1,7,11,13,17,19,23,29 30 8 0 1 2 Wefix aminiversaldeformation N F(t,x) = f(x)+ tv(x), t = (t ,...,t ) B := CN, (1) i i 1 N X ∈ i=1 THEEYNARD–ORANTINRECURSIONFORADESINGULARITIES 3 where v(x) N isaset ofweighted-homogeneouspolynomialsthatrepresent a ba- { i }i=1 sisoftheJacobialgebra H = C[x ,x ,x ]/(f , f , f ). 1 2 3 x1 x2 x3 Theformω := dx dx dx isprimitiveinthesenseofK.Saito[21,24]andthespace 1 2 3 B inherits a Frobenius structure (see [16, 23]). For somebackground on Frobenius structureswereferto[8]. TheFrobeniusmultiplicationonT Bisobtainedfromthe t multiplicationintheJacobialgebraof F(t, )viatheKodaira-Spencerisomorphism · T B (cid:27) C[x ,x ,x ]/(F (t,x),...,F (t,x)), ∂/∂t ∂F/∂t. t 1 2 3 x1 x3 i 7→ i WhiletheFrobeniuspairing (, )on TBistheresiduepairing 1 φ (x)φ (x) (φ (x),φ (x)) := 1 2 dx ...dx , 1 2 t (2π√−1)3 ZΓ Fx1(t,x)···Fx3(t,x) 1 N where the cycle Γ is a disjoint union of sufficiently small tori around the critical points of F defined by equations of the type F = = F = ǫ. In particular, | x1| ··· | x3| wehavethefollowingidentifications: T B (cid:27) TB (cid:27) B T B (cid:27) B H, ∗ 0 × × wherethefirstisomorphismisgivenbytheresiduepairing,thesecondbytheLevi– Civitaconnectionoftheflatresiduepairing,andthelastoneistheKodaira–Spencer isomorphism T B (cid:27) H, ∂/∂t ∂ F mod(f ,..., f ). (2) 0 i 7→ ti t=0 x1 x3 (cid:12) Let B B be the subset of semi-simple(cid:12)points, i.e., points t B such that the ss (cid:12) ⊂ ∈ critical valuesof F(t, )forma coordinatesysteminaneighborhoodoft. Forevery · t B ,usingGivental’shigher-genusreconstructionformalism[12,13],wedefine ss ∈ ancestorcorrelationfunctionsofthefollowingform (c.f. [19]) a ψk1,...,a ψkn (t), a H, k Z (1 i n). (3) h 1 1 n n ig,n i ∈ i ∈ ≥0 ≤ ≤ A priory, each correlator depends analytically on t B , but it might have poles ss ∈ along the divisor B B . According to [20] the correlation functions (3) extend ss \ analyticallytotheentiredomain B. 1.3. The period vectors. Put X = B C3 and S = B C. Let Σ S be the × × ⊂ discriminantofthemap ϕ : X S, ϕ(t,x) := (t,F(t,x)). → RemovingthesingularfibersX = X ϕ 1(Σ)weobtainasmoothfibrationX S , ′ − ′ ′ where S = S Σ, known as the Mil\norfibration. Let us denote by X = ϕ →1(t,λ) ′ t,λ − the fiber over (\t,λ) S . The vector spaces H2(X ;C) and H (X ,C) form the so ′ t,λ 2 t,λ ∈ called vanishing cohomology and homology bundles. They are equipped with flat Gauss–Maninconnections. 4 TODORMILANOV We fix (0,1) S as a reference point and denote by h := H2(X ,C). The dual 0,1 space h = H (X∈ ,C) is equipped with a non-degenerate intersection pairing and ∗ 2 0,1 wedenoteby( )thenegativeoftheintersectionpairing,sothat(αα) = 2forevery | | vanishing cycle α. The set R of all vanishing cycles together with the pairing ( ) | is a root systemof type ADE. Moreover, according to the Picard–Lefschetz theory (see[1]) theimageofthemonodromyrepresentation π (S ) GL(h ) (4) 1 ′ ∗ → is theWeyl group of R, i.e., the monodromytransformation s along a simpleloop α around the discriminant corresponding to a path along which the cycle α vanishes isthefollowingreflection s (x) = x (αx)α, α R, x h . α ∗ − | ∈ ∈ Letusintroducethenotationd ,where x = (x ,...,x )isacoordinatesystemon x 1 m some manifold, for the de Rham differential in the coordinates x. This notation is especiallyusefulwhenwehavetoapplyd tofunctionsthatmightdependonother x variablesas well. Themainobject ofourinterestare thefollowingperiodintegrals I(k)(t,λ) = d (2π) 1∂k+1 d 1ω T B (cid:27) H, (5) α − t − λ Z −x ∈ t∗ αt,λ where α h is a cycle from the vanishing homology, α H (X ,C) is the t,λ 2 t,λ parallel tra∈nsportofαalongareference path,and d 1ω is any2∈-form η Ω2 such that d η = ω. The periods are multivaluedanalytic−xfunctions in (t,λ) B∈ CC3 with x polesalongthediscriminantΣ. ∈ × 1.4. The period isomorphism. Let us fix a coordinate system t = (t ,...,t ) on 1 N B defined by a miniversal unfolding of f of the type (1). We may assume that v (x) = 1 and denote by t λ1 the point with coordinates (t ,t ,...,t λ). Note N 1 2 N that X = X ,so thepe−riodvectorshavethefollowingtranslationsy−mmetry t,λ t λ1,0 − I(k)(t,λ) = I(k)(t λ1,0). (6) α α − Sometimeswe restrict theperiod integralsto λ = 0 and itwill beconvenientto use as a reference point 1 B. Note that this choice is compatible with the choice of the other reference−po∈int (0,1) B C in a sense that the values of the period ∈ × vectorsat thesetwopointsare identifiedviathetranslationsymmetry(6). NowwecanstatethefollowingresultthatgoesbacktoLooijenga[18]andSaito [21]. Themonodromycoveringspaceof B := S Bisthecovering B of B corre- ′ ′ ′ ′ ∩ sponding to the kernel of the monodromy representation (4). It can be constructed e as the set of equivalence classes of pairs (t,C), where t B and C is a path in B ′ ′ ∈ from the reference point 1 to t and the equivalence relation (t ,C ) (t ,C ) is 1 1 2 2 t = t andC C 1 isint−hekernelofthemonodromyrepresentation(4∼). Notethat 1 2 1◦ 2− THEEYNARD–ORANTINRECURSIONFORADESINGULARITIES 5 the period integrals are by definition functions on B . In particular we have a well ′ defined periodmap e Φ : B h , Φ(C,t),α := (I( 1)(t,0),1), ′ → ′ h i α− where h is the coempleement in h oef the reflection hyperplanes of the roots R, i.e., ′ h = x h α,x , 0 α R . The first statement is that Φ is an analytic ′ isomo{rph∈ism.| hIn pairticula∀r, th∈ere i}s an induced isomorphism Φ : B h /W. The 2nd statement is that Φ extends analytically across the discreimin′an→t and′ the extension provides an analytic isomorphism B (cid:27) h/W := Spec(S(h )W). Using the ∗ isomorphismΦ and thenatural projection B B we can thinkof thecoordinates ′ ′ → t as W-invariant holomorphic functions on h . The 2nd statement is equivalent i e e ′ to saying that each coordinate t extends holomorphically through the reflection i mirrors, the extension is in fact a W-invariant polynomial in h, and the ring of all W-invariantpolynomialsisS(h )W = C[t ,...,t ].Wereferto[18,21]fortheproof ∗ 1 N ofallthesestatements. 1.5. The spectral curve. Let us fix a set of simple roots α N h and denote by x = (x ,...,x ) the coordinatesystem in h correspondin{git}oi=t1he⊂ba∗sis of funda- 1 N mentalweights ω N h, i.e., { i}i=1 ⊂ N x = xω, x = α,x . i i i i X h i i=1 As explained above t C[x ,...,x ]W are invariant polynomials and since the i 1 N ∈ periodmappingisweighted-homogeneous,t arehomogeneouspolynomialsofcer- i tain degrees d. Let us assume that the degrees are in an increasing order, then the i numbers 1 = d 1 d 1 ...d 1 =: h 1 are known as the Coxeter 1 2 N exponents(see Ta−ble1≤). Giv−en s≤ CN 1 w−edefine th−ealgebraiccurveV PN − s ∈ ⊂ t(X ,...,X ) = sXdi, 1 i N 1. i 1 N i 0 ≤ ≤ − As we will see later on if s B , then V is non-singular. In fact, the points s for ss s ∈ which V has singularities are precisely the caustic B B . I am not aware if the s ss familyofalgebraiccurvesV ,s B hasanofficialnam−eattached,butsinceitwill s ss ∈ be the spectral curve for the EO recursion, we will refer to it as the spectral curve ofthesingularityorjustthespectralcurvewhenthesingularityisunderstoodfrom thecontext. Thereisanatural projection λ : V P1, [X ,X ,...,X ] [Xh,t (X ,...,X )], (7) s → 0 1 N 7→ 0 N 1 N which is a branched covering of degree W , where A denotes the number of ele- ments of the set A. The branching point|s a|re λ = u| ,|...,u , , where u are the 1 N i ∞ critical values of F(s,x). By definition, the period integral (I( 1)(s,λ),1) defines − 6 TODORMILANOV locallynearanon-branchingpointλ P1 asectionofthebranched covering(7). It ∈ followsthat theset oframificationpoints λ 1(u), 1 i N − i ≤ ≤ isprecisely theintersectionsofV and thereflection mirrors s N α,X = α,ω X = 0, α R , i i + h i Xh i ∈ i=1 whereR isthesetofpositiveroots. Theremainingramificationpointsareλ 1( ). + − ∞ They correspond to eigenvectors of the Coxeter transformations with eigenvalue η := e2π√ 1/h: − [X ,X ,...,X ] λ 1( ) 0 1 N − ∈ ∞ if and only if X = 0 and N Xω h is an eigenvector with eigenvalue η for a 0 i=1 i i ∈ Coxeter transformation. ItPis easy to see that the ramification index of any point in λ 1(u)is2, whiletheramificationindexofapointinλ 1( )ish. − i − ∞ 1.6. TheEynard–Orantinrecursion. Wemakeuseofthefollowingformalseries fα(t,λ;z)= I(k)(t,λ)( z)k, φα(t,λ;z) = I(k+1)(t,λ)( z)kdλ. X α − X α − k Z k Z ∈ ∈ Note that φα(t,λ;z) = d fα(t,λ;z). Given n cycles α ,...,α and a semi-simple λ 1 n point s B wedefinethefollowingn-pointsymmetricforms ss ∈ ωgα,1n,...,αn(s;λ1,...,λn) = φα+1(s,λ1;ψ1),...,φα+n(s,λn;ψn) (s), (8) g,n D E where the + means truncation of the terms in the series with negativepowers of z. Thefunctions(8)willbecalledn-pointseriesofgenusgorsimplycorrelatorforms. The ancestor correlators (3) are known to be tame (see [14]), which by definition means that they vanish if k + + k > 3g 3 + n. Hence the correlator (8) is 1 n ··· − a polynomial expression of the components of the period vectors (5). Thanks to the translation symmetry, we may assume that s = 0, then (8) is a meromorphic N function on the spectral curve V V with possible poles at the ramification s s ×···× pointsofthecovering(7). Let us fix s = (s ,...,s ) CN 1 and denote by γ h an arbitrary cycle, s.t., 1 N 1 − ∗ (γα) , 0 for all α R. We−defi∈ne a set ofsymmetricm∈eromorphicdifferentialson | ∈ Vn withpolesalongtheramificationpointsofV s s ω (s;p ,...,p ) := ωγ,...,γ(s;λ ,...,λ ), (9) g,n 1 n g,n 1 n where the RHS is defined by fixing a reference path for each (s,λ) S , s.t., i ′ p = (I( 1)(s,λ),1). Ourmain resultcan bestatedas follows. ∈ i − i THEEYNARD–ORANTINRECURSIONFORADESINGULARITIES 7 Theorem 1.1. If s B , then the forms ω , 2g 2 + n > 0, satisfy the Eynard– ss g,n ∈ − Orantinrecursion associatedwith thebranchedcovering (7) and themeromorphic function f : V P1, f (x) := γ,x . γ s γ → h i Therecursion willberecalled lateron (seeSection 3.3). Wewouldlikehowever toemphasizethattheEynard–Orantinrecursioninourcasediffersfromthestandard onebytheinitialcondition: ω (x,y) = (γwγ)B(x,wy), (10) 0,2 X | w W ∈ where B(x,y) is the Bergman kernel of V and for x = wy one has to regularize s the RHS by removing an appropriate singular term (see Section 3.3). Let us point out that while the Bergman kernel depends on the choice of a Torelli marking of V , i.e., a symplectic basis of H (V ;Z), our initial condition is independent of the s 1 s Torellimarking(seeCorollary2.5). Thisfactcouldbeprovedalsodirectlybyusing the explicit formula (24) for the holomorphic 1-forms and some standard facts for W-invariant polynomials. Finally, let us point out that the set of correlators (9) determinestheset(8),because bydefinition ωw1γ,...,wNγ(s;λ ,...,λ ) = ω (s;w 1p ,...,w 1p ), g,n 1 N g,n −1 1 −N N wherew ,...,w W are arbitrary. 1 N ∈ The branched covering (7) and the meromorphic function f determine a bi- γ rational model of V in C2. Following [10] we can introduce the tau-function s Z(~,s) := Z (V ,λ, f ) of the birational model of the spectral curve. It has the ~ s γ form ∞ Z(~,s) = exp ~g 1F(g)(s) , − (cid:16)Xg=0 (cid:17) where F(g)(s) is called the genus-g free energy of V . It is very natural to compare s F(g)(s) with Givental’sprimary genus-g potentials F(g)(s). Unfortunately we could notsolvethisproblemingeneral, butonlyfor g = 0 and g = 1 W W F(0)(s) = (γγ)| |F(0)(s), F(1)(s) = | |F(1)(s) = 0, − | N 2 where the first identity is valid up to quadratic terms in the Frobenius flat coordi- nates of s B, whilethesecond oneup to a constant independent of s. It is known ∈ thatthegenus-1potentialoftheFrobeniusstructureishomogeneousofdegree0,so itmustvanishinthecaseofasimplesingularity. Accordingto[10],Z(~,s)satisfies Hirotabilinearequations. Accordingto[15],thetotalancestorpotential (~;q)of s A anADEsingularitysatisfiestheHirotabilinearequationsofthecorrespondinggen- eralized KdV hierarchy. It willbeinterestingto clarify therelation between F(g)(s) and F(g)(s) for g 2, as well as to determine whether Givental’s primary ancestor potential (~;0)≥alsosatisfies Hirotabi-linearequations. s A 8 TODORMILANOV 2. Analytic extension ofthekernel ofthelocal Eynard–Orantin recursion It was proved in [19] (see also [9]) that the correlator forms (8) satisfy a local Eynard–Orantin (EO) recursion, whose kernel is defined by the symplecticpairing of certain period series. In this section, we will prove that these symplectic pair- ings are convergent and can be extended to the entire spectral curve V . Moreover, s the corresponding extensions can be expressed in an elegant way via the so called Bergmankernel ofV . s 2.1. The kernel ofthelocal recursion. Recall thesymplecticpairing Ω(f(z),g(z)) = Res (f( z),g(z))dz, f,g H((z 1)). z=0 − − ∈ Thelocalrecursion isdefined intermsofthesymplecticpairings Ω(φα(s,λ;z),fβ(s,µ;z)) = dλ ∞ ( 1)k+1(I(k+1)(s,λ),I( k 1)(s,µ)), + − X − α β− − k=0 where the infinite series is interpreted formally in a neighborhood of a point µ = u(s), s.t., the cycle β vanishes over µ = u(s). We are going to prove that this i s,µ i infiniteseries expansionis convergenttoa meromorphicfunctionon V V . s s Tobeginwith,letusrecallthattheperiodssatisfythefollowingsystem×ofdiffer- entialequations ∂ I(n)(s,λ) = v I(n+1)(s,λ) a a s − • ∂ I(n)(s,λ) = I(n+1)(s,λ) λ 1 (λ E )∂ I(n)(s,λ) = θ n I(n)(s,λ), (11) s λ − • − − 2 (cid:16) (cid:17) where ∂ := ∂/∂s , E is the Euler vector field E = N deg(t)t∂ , and θ is the a a i=1 i i ti Hodge-gradingoperator P θ : H H, v (D/2 deg(v ))v , a a a → 7→ − where D = deg(Hess(f)) = 1 2/h is the conformal dimension of the Frobenius − structure. The key to proving the convergence is the so called phase 1-form (see [14, 2]) (s,ξ) = I(0)(s,ξ) I(0)(s,0) T S, Wα,β α • β ∈ s∗ wheretheperiodvectorsareinterpretedaselementsinT S andthemultiplicationin s∗ T S isinducedbytheFrobeniusmultiplicationviathenaturalidentificationT S (cid:27) s∗ s∗ T S. The dependence on the parameter ξ is in the sense of a germ at ξ = 0, i.e., s Taylor’sseries expansionaboutξ = 0. Thephaseformisapowerseriesinξ whose coefficients are multivalued1-forms on B . ′ THEEYNARD–ORANTINRECURSIONFORADESINGULARITIES 9 Lemma 2.1. Wehave (αβ) = ι (s,0) = (I(0)(s,0),E I(0)(s,0)). | − EWα,β − α • β Thisisawell knownfact dueoriginallytoK. Saito[21]. Lemma 2.2. The phaseformisweighted-homogeneousof weight0, i.e., (ξ∂ +L ) (s,ξ) = 0, ξ E α,β W where L istheLiederivativewithrespect to thevector field E. E Proof. Notethat (s,ξ) = (I(0)(s,ξ),dI( 1)(s,0)). Wα,β α β− Itiseasytocheckthat isaclosed1-form,sousingtheCartan’smagicformula α,β L = d ι +ι d , wherWeι isthecontractionbythevectorfield E, weget E s E E s E L = d (I(0)(s,ξ),(θ+1/2)I( 1)(s,0)) = d ((θ 1/2)I(0)(s,ξ),I( 1)(s,0)). EWα,β s α β− − s − α β− Weused thatθ isskew-symmetricwithrespect totheresiduepairingand that ι d I( 1)(s,0) = EI( 1)(s,0)) = (θ+1/2)I( 1)(s,0), E s β− β− β− where the last equality comes from the differential equation (11) with n = 1 and λ = 0. Furthermore, usingtheLeibnitzruleweget − ((θ 1/2)d I(0)(s,ξ),I( 1)(s,0)) ((θ 1/2)I(0)(s,ξ),d I( 1)(s,0)). − − s α β− − − α s β− Thefirst residuepairing is (AI(1)(s,ξ),(θ+1/2)I( 1)(s,0)) = (AI(1)(s,ξ),E I(0)(s,0)), (12) α β− − α • β whereweusedthatθisskew-symmetricandthatd I(0) = AI(1)withA = N (∂/∂s )ds . s α − α a=1 a• a Similarly,the2ndresiduepairingbecomes P ((ξ∂ +E)I(0)(s,ξ),d I( 1)(s,0)) = ξ∂ (s,ξ)+(E I(1)(s,ξ),AI(0)(s,0)). (13) ξ α s β− ξWα,β • α β On the otherhand, since theFrobenius multiplicationis commutative,[A,E ] = 0, sotheterms(12)and (13)add upto ξ∂ (s,ξ),whichcompletesthepro•of. (cid:3) ξ α,β W Let us define the following meromorphic 1-forms on V V . Given (x,y) s s × ∈ V V ,s.t., x,yarenotramificationpoints,thereareuniquepairs(C,λ)and(C ,µ), s s ′ × s.t., x = Φ(C,s λ1) and y = Φ(C ,s µ1), where C and C are paths in B ′ ′ ′ − − connectingrespectively s λ1and s µ1 withthereference point. Put e − e − dλ K (s,x,y) := (I(0)(s,λ),(θ+1/2)I( 1)(s,µ)), α,β λ µ α β− − 10 TODORMILANOV where the branches of I(0) and I( 1) are determined respectively by the pathsC and α β− C . This definition extends analytically across the ramification points and the form ′ has apoleonlyalong thedivisor (x,y) V V : λ(x) = λ(y) = x = wy , s s w W { ∈ × } ∪ ∈ { } wheretheactionofW on his inducedfromtheactionon h , i.e., ∗ α,wx = w 1α,x , α h , x h. − ∗ h i h i ∈ ∈ Proposition2.3. Thesymplecticpairing Ω(φα(s,λ;z),fβ(s,µ;z)) = K (s,x,y), + α,β − where x = Φ(C,s λ1), y = Φ(C,s µ1) andC is thepath that specifies thevalue − − ofthesymplecticpairing. e e Proof. Usingthedifferentialequationsfortheperiods, itiseasy to verifythat d Ω(φα(s,λ;z),fβ(s,µ;z)) = dλI(1)(s,λ) I(0)(s,µ) = d (s µ1,λ µ). s + − α •s β λWα,β − − Accordingto Lemma2.2 wehave 1 ∂ (s ,λ µ) = d ι (s ,λ µ) , (14) λ α,β ′ s E α,β ′ W − − ′ λ µ W − (cid:16) (cid:17) − whichby definitionis 1 d (I(0)(s ,λ µ),(θ+1/2)I( 1)(s ,0) . s′ λ µ α ′ − β− ′ (cid:16) (cid:17) − Integrating(14)withrespect to s alongashortpath from s := s u(s)1 to s µ1 ′ 0 i − − and usingthat I( 1)(s ,0)vanishesas s s , weget β− ′ ′ → 0 dλ Ω(φα(s,λ;z),fβ(s,µ;z)) = (I(0)(s,λ),(θ+1/2)I( 1)(s,µ)). (cid:3) (15) + λ µ α β− − − 2.2. ThelocalkernelandtheBergmankernel. Nowweareinapositiontoprove the key result in this paper. Let us fix a symplectic basis , g of H (V ;Z), {Ai Bi}i=1 1 s s.t., = δ . Thereisauniquesymmetricdifferential B(x,y) Ω1 ⊠Ω1 (2∆) Ai◦B| i,j ∈ Vs Vs whichisholomorphiconV V exceptforapoleoforder2withnoresiduealong s s thediagonal∆ V V , no×rmalizedby s s ⊂ × B(x,y) = 0, 1 i N I ≤ ≤ y i ∈A and dλ(x)dλ(y) B(x,y) = + (λ(x) λ(y))2 ··· − for any local coordinate λ : U C (U V ) and for all x,y U U. The s differential B(x,y) is called the B→ergman ke⊂rnel. We refer to [10] f∈or m×ore details and references.

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.