ebook img

Rankin-Selberg methods for closed string amplitudes PDF

0.38 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 Rankin-Selberg methods for closed string amplitudes

Rankin-Selberg methods for closed string amplitudes 4 1 0 Boris Pioline 2 l u Abstract. Afterintegratingoversupermoduliandvertexoperatorpositions, J scattering amplitudes insuperstringtheory atgenus h≤3arereduced toan 0 integralofaSiegelmodularfunctionofdegreehonafundamental domainof 1 the Siegelupper halfplane. Adirectcomputation isingeneral unwieldy, but becomesfeasibleiftheintegrandcanbeexpressedasasumoverimagesunder ] asuitablesubgroupoftheSiegelmodulargroup: ifso,theintegrationdomain h canbeextendedtoasimplerdomainattheexpenseofkeepingasingletermin t eachorbit–atechnique knownastheRankin-Selbergmethod. Motivatedby - p applications to BPS-saturated amplitudes, Angelantonj, Florakis and I have e appliedthistechniquetoone-loopmodularintegralswheretheintegrandisthe h productofaSiegel-Narainthetafunctiontimesaweakly,almostholomorphic [ modular form. I survey our main results, and take some steps in extending thismethodtogenus greaterthanone. 2 v 5 6 2 1. Introduction 4 . Accordingtothebasicpostulateofsuperstringtheory,scatteringamplitudesof 1 nexternalstatesath-thorderinperturbationtheoryaregivenbycorrelationfunc- 0 tions of n vertex operators,integrated over the moduli space M of n-punctured 4 h,n 1 super-Riemann surfaces Σ of genus h [1, 2]. After integrating over the fermionic : moduli and the locations of the punctures,1 such amplitudes reduce to an integral v over the moduli space of ordinary Riemann surfaces without marked point. i h,0 X The latter is a quotienMt of the Teichmu¨ller space by the mapping class group h T r Γ . For 1 h 3, is isomorphic (via the period map, Σ Ω Ω +iΩ , away a h ≤ ≤ Th 7→ ≡ 1 2 from suitable divisors) to the degree h Siegel upper-half plane , while Γ is h h identifiedwiththeSiegelmodulargroupSp(h,Z),actingon byHfractionallinear h H transformations. Thus, scattering amplitudes at genus h are ultimately written as modular integrals h+1 (1.1) h =R.N. dµhFh(Ω) , dµh = Ω2 − 2 dΩ1dΩ2 A | | ZFh where = Γ is a fundamental domain of the action of Γ on , F(Ω) is h h h h h F \H H a function on invariant under Γ , Ω = det[ (Ω)] > 0, dµ is the standard h h 2 h H | | ℑ Contribution to the proceedings of the String-Math conference, June 17-21, 2013, Simons CenterforGeometryofPhysics,StonyBrook. CERN-PH-TH/2014-007,arXiv:1401.4265v2. 1For h > 2 there is no canonical way of performing these integrations, due in part to the non-projectivenessofsupermodulispace[3, 4],butdifferentprescriptionsareexpectedtoleadto thesameintegrandonMh,0,uptototalderivatives. 1 2 BORISPIOLINE invariant measure on , and R.N. denotes a suitable infrared renormalization h H prescription. For h > 3, the Teichmu¨ller space embeds as a subvariety of codi- h T mensionone or greaterin the Siegelupper-half plane , anddµ is replacedby a h h H suitablemeasurewithsupporton –a complicationwhichweshallnotconfront. h F In general,integralsof the type (1.1) are untractable (except, perhaps, numer- ically), due to the complicated nature of the modular function F , but also to the h unwieldy shape of any fundamental domain: for h = 2, it takes no less than 25 inequalities to define ! [5]Yet, the computation ofsuch integralsis anunavoid- 2 F able step in extracting any phenomenologicalprediction of superstring theory, and ininvestigatingsomeofits structuralpropertiessuchasinvarianceunderdualities. The daunting task of integrating (1.1), however, can be considerably simplified if the integrand can be written as a sum over images (or Poincar´e series) of a given function f (Ω), h (1.2) F (Ω)= f (Ω) h h γ | γ∈ΓXh,∞\Γh wheref (Ω)=f (γ Ω)andf (Ω)isinvariantunderasubgroupΓ Γ . Ifthe h γ h h h,∞ h | · ⊂ sumisabsolutelyconvergent,exchangingitwiththeintegralextendstheintegration domaintoalargerfundamentaldomain =Γ ,whilerestrictingthesum h,∞ ∞,h h F \H to a single coset, (1.3) =R.N. dµ f (Ω) . h h h A ZF∞,h This‘unfoldingtrick’,attheheartoftheRankin-Selbergmethodinnumbertheory, is expedient if both and f are simpler than and F . h,∞ h h h F F Aspecialclassofamplitudeswherethismethodisadvantageousariseswhenthe integrand F (Ω) factorizes as Φ(Ω) Γ , where Φ(Ω) is a (non-holomorphic, h d+k,d,h × in general) Siegel modular form of weight k/2 and Γ is the Siegel-Narain d+k,d,h − theta series (1.4) Γd+k,d,h(G,B,Y;Ω)= Ω2 d/2 e−πΩ2,αβM2(pα,pβ)+2πiΩ1,αβhpα,pβi | | pα∈Λ X where the sum runs over h-tuples of vectors pα, α = 1...h in an even self-dual latticeofsignature(d+k,d)(hencek 8Z)withquadraticform , 2Zequipped ∈ h· ·i∈ with positive definite quadratic form 2(, ). The latter is parametrized by the M · · Grassmannian (1.5) G =[SO(d+k) SO(d)] SO(d+k,d) , d+k,d × \ whichcanbe coordinatizedby a realpositive definite symmetric matrix G ,a real ij antisymmetric matrix B and a real rectangular matrix Y (i,j = 1...d,a = ij i,a 1...k). This type of theta series arises in any string vacua which involve a d- dimensional torus with constant metric G and Kalb-Ramond field B , equipped ij ij withaU(1)kflatconnectionwithholonomiesY . Importantly,Γ isinvariant i,a d+k,d,h under Γ O(d+k,d,Z), where the last factor is the automorphism group of the h × lattice Λ (also known as T-duality group), acting by right-multiplication on the coset (1.5). In such cases, the integral (1.6) =R.N. dµ Γ (G,B,Y;Ω)Φ(Ω) . h h d+k,d,h A ZFh RANKIN-SELBERG METHODS FOR CLOSED STRING AMPLITUDES 3 can be computed by expressing Γ as a sum of Poincar´eseries under Γ , and d+k,d,h h applyingthe unfolding tricktoeachoneofthem. This‘lattice unfoldingtechnique’ has been the method of choice for one-loop amplitudes in the physics literature [6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17],andhasbeenveryusefulinextracting asymptoticexpansionsat particularboundarycomponents ofG . However,its d+k,d main drawback is that the various terms in the Poincar´e series decomposition are notinvariantunder SO(d+k,d,Z), eventhoughthe sumis. As a result,the result of the unfolding trick is not manifestly invariant under SO(d+k,d,Z). Another option, advocatedin [18]2 and further developed in [21, 22] (see [23] for a complementary survey), is to represent the other factor Φ(Ω) in the inte- grand (or the full integrand, in the absence of any Siegel-Narain theta series) as a Poincar´e series, and use it to unfold the integration domain. This technique is of course limited by our ability to find absolutely convergent Poincar´e series rep- resentations for (in general, non-holomorphic) Siegel modular forms. For genus one, it turns out that any almost, weakly holomorphic modular form of negative weight under Γ =SL(2,Z) (or congruence subgroups thereof) can be represented 1 as a linear combination of certain absolutely convergent Poincar´e series, first in- troduced by Niebur [24] and Hejhal [25] and revived in recent mathematical work [26, 27, 28, 29]. Almost, weakly holomorphic integrands Φ(Ω) are non-generic, but do occur for certain classes of ’BPS-saturated’ amplitudes, which play a cen- tral role for determining threshold corrections to gauge couplings and for testing non-perturbative dualities (see e.g. [30] for a review). As we shall see, the unfold- ing trick produces a sum over lattice vectors with fixed integer norm, manifestly invariantunder T-duality. Physically,itcanbe interpretedasa sumoffield-theory type amplitudes,with BPSstatesrunning inthe loop. Inparticular,itexposesthe singularitiesoftheamplitude,originatingfromBPSstatesbecomingmassless. The pricetopayisthatthebehaviorattheboundarycomponentsisobscured,although itcanberecoveredinsomecases,showingagreementwith–anduncoveringhidden structure in – the result of the usual lattice unfolding technique. A third, and most radical option, is to represent 1 as a Poincar´e series, and use it to unfold the integral. Indeed, it is well-known that 1 is a residue of non- holomorphic Siegel-Eisenstein series, the simplest conceivable example of Poincar´e series. This trick, which we refer to as the Rankin-Selberg-Zagier method, is very usefultoevaluateintegralsofthe type (1.6)withΦ=1(hence k =0). Itexpresses the result, for any genus, as a Langlands-Eisenstein series of SO(d,d,Z), verifying a conjecture put forward in [31]. We begin our survey of various applications of the Rankin-Selberg method (or unfolding trick) to closed string amplitudes in 2 by computing one-loop in- § tegrals of symmetric lattice partition functions by means of a non-holomorphic Eisenstein series insertion. In 3 we move on to more general modular integrals of § non-symmetriclatticepartitionfunctionsagainstharmonicellipticgenus,whichwe computebyrepresentingthe latteraslinearcombinationofNiebur-Poincar´eseries. In the last section 4 we take steps towards extending the Rankin-Selberg-Zagier § method to higher genus. 2TheideaofapplyingRankin-Selberg-typemethodstocomputeone-loopmodularintegrals, albeitinaratherdifferentwayfrom[18],was firstputforwardin[19]. Steps towardsextending themtohighergenusweretakenin[20]. 4 BORISPIOLINE Acknowledgements: I wish to thank C. Angelantonj and I. Florakis for a very enjoyablecollaborationontheresultsreportedin 2-3,K.BringmannandD.Zagier § for valuable advice during the course of this project, R. Donagi and R. Russo for commentsonanearlierversionofthis manuscript,andthe organizersofthe String Math 2013 conference for their kind invitation to speak. 2. One-loop modular integrals with trivial elliptic genus We startwith the simplestcase ofone-loopmodular integralsofthe form(1.6) with Φ =1 (hence k = 0). Such integrals were computed by the ‘lattice unfolding technique’in[6, 7]ford=1,[8]ford=2,[16]ford 3,andaconjecturalrelation ≥ to constrained Epstein series was put forward in [31]. In this section, we shall calculatethem insteadbyinserting by handanon-holomorphicEisensteinseriesin the integrand, computing the integral by the unfolding trick and taking a suitable residue at the end. We use the standard notations τ = τ +iτ , q = e2πiτ for the 1 2 periodΩ andmodulusofanellipticcurve,andwriteΓ=Γ =SL(2,Z), = , 11 1 1 H H etc. 2.1. Non-holomorphic Eisenstein series . The non-holomorphic Eisen- stein series for Γ=SL(2,Z) is defined by the sum over images τs E⋆(s;τ)=ζ⋆(2s) τs γ = 1ζ⋆(2s) 2 , (2.1) 2| 2 cτ +d2s γ∈XΓ∞\Γ (cX,d)=1| | where Γ is the subgroup of upper-triangular matrices in Γ, and ∞ (2.2) ζ⋆(s) π−s/2Γ(s/2)ζ(s)=ζ⋆(1 s) ≡ − is the completed Riemann zeta function. The sum convergesabsolutely for (s)> ℜ 1,andhasameromorphiccontinuationtoall s. Thenormalizationin(2.1)ensures that E⋆(s;τ) invariant under s 1 s, and has only simple poles at s = 0 and 7→ − s=1. The key point for our purposes is that the residue at s=1 is constant – in agreement with the fact that E⋆(s;τ) is an eigenmode of the Laplacian ∆ on , H H with vanishing eigenvalue at s=0 or 1, (2.3) ∆ 1s(s 1) E⋆(s;τ)=0 , ∆ =2τ2∂ ∂ . H− 2 − H 2 τ τ¯ Moreprecisely,(cid:2)the Laurentexpan(cid:3)sionats=1is givenby the firstKroneckerlimit formula, 1 (2.4) E⋆(s;τ)= + 1 γ log(4πτ η(τ)4) + (s 1) , 2(s 1) 2 E − 2| | O − − (cid:0) (cid:1) where η = q1/24 (1 qn) is the Dedekind eta function and γ is Euler’s n≥1 − E constant. All these statements are easy consequences of the Fourier series repre- Q sentation of E⋆(s;τ), or Chowla-Selberg formula, E⋆(s;τ)=ζ⋆(2s)τs+ζ⋆(2s 1)τ1−s 2 − 2 (2.5) +2 |N|s−12σ1−2s(N)τ21/2Ks−12(2π|N|τ2)e2πiNτ1, N6=0 X where σ (N) = dt is the divisor function and K (z) is the modified Bessel t d|N t function of the second kind. Below we shall denote the first line of (2.5), which P dominates the behavior at τ , by E⋆(s;τ). 2 →∞ 0 RANKIN-SELBERG METHODS FOR CLOSED STRING AMPLITUDES 5 ForanymodularfunctionF(τ)ofrapiddecay(suchasthemodulussquare ψ 2 | | of a holomorphic cusp form), we consider the modular integral (also known as the Rankin-Selberg transform) (2.6) ⋆(F;s)= dµE⋆(s;τ)F(τ) . R ZF For (s) > 1, the sum over cosets in (2.1) can be exchanged with the integral, so ℜ that the integration domain is extended to the strip (2.7) =Γ Γ= τ >0, 1 <τ 1 , S ∞\ { 2 −2 1 ≤ 2} at the expense of retaining the contribution of the unit coset only, ∞ (2.8) ⋆(F;s)=ζ⋆(2s) dµτsF(τ)=ζ⋆(2s) dτ τs−2F (τ ) . R 2 2 2 0 2 ZS Z0 The last equality expresses ⋆(F;s) as a Mellin transform of the zero-th Fourier R coefficient F (τ )= 1/2 dτ F(τ). At the same time, ⋆(F;s) inherits the mero- 0 2 −1/2 1 R morphicity and invariance under s 1 s satisfied by E⋆. In the case where R 7→ − F = ψ 2, ⋆(F;s) is proportional to the L-series a 2n−s, whose analyticity n | | R | | and functional equation are thereby determined. This is one of the main uses of P the Rankin-Selberg method in number theory [32]. More importantly for our purposes, the fact that the residue of E⋆(s;τ) at s = 1 is constant implies that the residue of ⋆(F;s) at s = 1 is proportional to R the modular integral of F, (2.9) Res ⋆(F;s)= 1 dµF . s=1R 2 ZF Unfortunately,thisstatementonlyholdsforfunctionsF ofrapiddecay,whichrules out the interesting case F = Γ . In the next section, following [33] we discuss d,d,1 how the unfolding trick can nevertheless be used after proper regularization. 2.2. Rankin-Selberg-Zagiermethod. Letusnowconsideramodularfunc- tion F with polynomial growth at the cusp, (2.10) F(τ) ϕ(τ ) , ϕ(τ )= c τα . ∼ 2 2 α 2 α X In order to regulate infrared divergences in the integral (2.6), we truncate the integration domain to = τ > and define the renormalized integral as T 2 F F ∩{ T} (2.11) R.N. dµF(τ)= lim dµF(τ) ϕˆ( ) , ZF T→∞(cid:20)ZFT − T (cid:21) where ϕˆ( ) is the anti-derivative of ϕ(τ ), 2 T α−1 (2.12) ϕˆ( )= c T +c log . α 1 T α 1 T α6=1 − X On the other hand, we define the Rankin-Selberg transform of F as the Mellin transform of the zero-th Fourier coefficient, minus its leading behavior, ∞ (2.13) ⋆(F;s)=ζ⋆(2s) dτ τs−2 (F ϕ) . R 2 2 0− Z0 6 BORISPIOLINE The renormalized integral (2.11) is then related to the Rankin-Selberg transform (2.13) via a generalizationof (2.9) [33] (2.14) R.N. dµF(τ)=2Res [ ⋆(F;s)+ζ⋆(2s)h (s)+ζ⋆(2s 1)h (1 s)] ϕˆ( ), s=1 T T R − − − T ZF where h (s) is the meromorphic function of s defined by T T α+s−1 (2.15) h (s)= dτ ϕ(τ )τs−2 = c T . T 2 2 2 α α+s 1 Z0 α − X Notethattheright-handsideof(2.14)isbyconstructionindependentoftheinfrared cut-off . The derivation of (2.14) is based on the generalized unfolding trick for T modular integrals on the truncated fundamental domain, (2.16) dµF f = dµF f dµF f , γ γ | − | ZFT γ∈XΓ∞\Γ ZST ZF−FT γ∈XγΓ6=∞1\Γ where is the truncated strip τ < , 1/2 < τ < 1/2 . Applying this ob- T 2 1 S { T − } servation to the regulated integral R⋆(F;s) dµFE⋆(s;τ) and reorganizing T ≡ FT terms gives [33, Eq. (27)] R ⋆(F;s)=R⋆(F;s)+ dµ (F E⋆(s;τ) ϕE⋆(s;τ )) (2.17) R T ZF−FT − 0 2 ζ⋆(2s)h (s) ζ⋆(2s 1)h (1 s) , T T − − − − fromwhich(2.14)follows. Anotherconsequenceof (2.17)isthattheRankin-Selberg transform ⋆(F;s)hasameromorphiccontinuationins,invariantunders 1 s, R 7→ − and analytic away from s = 0,1,α ,1 α . A particularly pleasant feature of the i i − renormalization prescription (2.11) is that the Rankin-Selberg transform ⋆(F;s) R coincides with the renormalized integral (2.18) ⋆(F;s)=R.N. dµF(τ)E⋆(s;τ) . R ZF Moreover,ifF isconstant, ⋆(F;s)vanishesandthereforeR.N. dµE⋆(s;τ)=0. R F R 2.3. Constrained Epstein series. The Rankin-Selberg-Zagier method dis- cussedintheprevioussubsectionappliesimmediatelytomodularintegralsofSiegel- Narain theta series for even self-dual lattices of signature (d,d), (2.19) Γ (G,B;τ)=τd/2 e−πτ2M2(mi,ni)+2πiτ1mini , d,d 2 (mi,Xni)∈Z2d where 2 is the positive definite quadratic form M (2.20) 2(m ,ni)=(m +B nk)Gij(m +B nl)+niG nj i i ik j jl ij M and(G ,B )=G , B )parametrizetheGrassmannianG . Eq. (2.19)defines ij ij ji ji d,d − a modular function F on of polynomial growth characterized by H d d (2.21) ϕ(τ2)=τ2d2 , hT(s)= sT+s+d2−11, ϕˆ(τ2)=τlo22g−τ1/(d2 −1) iiff dd=6=22 . 2 −  2  RANKIN-SELBERG METHODS FOR CLOSED STRING AMPLITUDES 7 Its Rankin-Selberg transform is ∞ d ⋆(Γ ;s)=ζ⋆(2s) dτ τs+2−2 e−πτ2M2(mi,ni) R d,d 2 2 (2.22) Z0 (mi,ni)X∈Z2d\(0,0) mini=0 = SO(d,d),⋆(s+ d 1;G,B) EV 2 − where SO(d,d),⋆(s) is the completed, constrained Epstein series defined by [31] EV SO(d,d),⋆(s)=ζ⋆(2s)ζ⋆(2s+2 d) SO(d,d)(s) EV − EV 1 (2.23) SO(d,d)(s;G,B)= [ 2(m ,ni)]−s , EV ζ(2s) M i (mi,ni)X∈Z2d\(0,0) mini=0 the sum being absolutely convergent for (s) > d. The results of 2.2 show that ℜ § SO(d,d),⋆(s)admitsameromorphiccontinuationins,invariantunders d 1 s, EV 7→ − − with simple poles at s = 0,d 1,d,d 1 (or double poles at s = 0 and s = 1 if 2 − 2 − d=2). For d=2, the residues at s= d or s= d 1 produce the modular integral 6 2 2− of interest: (2.24) R.N. dµΓ =2Res SO(d,d),⋆(s;G,B)=ζ⋆(d 2) SO(d,d) d 1;G,B , d,d s=dEV − EV 2 − ZF 2 (cid:0) (cid:1) rigorously proving a conjecture in [31]. Physically, the integral (2.24) computes (among other things) the one-loop contribution to 4 couplings in type II strings R compactified on a torus Td [34, 16]. The right-hand side is interpreted as a sum of one-loop contributions from particles of momentum m and winding ni along i the torus, with mass 2(m ,ni), satisfying the BPS constraint m ni = 0. It is i i manifestly invariant uMnder the T-duality group SO(d,d,Z), under which m ,ni i transform in the vector (defining) representation. For s 1 the s-dependent → generalization(2.22)canbethoughtofasthedimensionallyregularized3amplitude, i.e. the amplitude in D = 10 (d+2s 2) non-compact dimensions. The case − − s=2alsocomputesD4 4 couplingsatone-loopintypeIIstringtheoryonTd[38]. R Mathematically, (2.23) is recognized as the degenerate Langlands-Eisenstein series of SO(d,d) with infinitesimal character ρ 2sλ (where ρ is the Weyl vector and 1 − λ theweightofthevectorrepresentation)[39, 40]. TheresidueofthisLanglands- 1 Eisensteinseriesats= d yieldsthe minimalthetaseriesassociatedtothe minimal 2 representationof SO(d,d) [41]. Ford=2,theGrassmannianG reducestoaproductoftwoupperhalfplanes 2,2 ,whereT andU aretheK¨ahlermodulusandcomplexstructuremodulus, T U Hresp×ecHtively, while the T-duality group decomposes into SL(2,Z) SL(2,Z) ⋉ T U × σ ,whereσ isaninvolutionexchangingT andU. TheBPSconstraintm n1+ T,U T,U 1 m n2 =0canbesolvedexplicitly,allowingtorewritetheconstrainedEpsteinseries 2 as a product of two non-holomorphic Eisenstein series [18], (2.25) SO(2,2),⋆(s;T,U)=2E⋆(s;T)E⋆(s;U) . EV 3Otherversionsofdimensionalregularizationinstringtheorywerediscussedin[35,36,37]. 8 BORISPIOLINE Extracting the residue at s=0 or 1 by means of the Kroneckerlimit formula (2.4) leads to (2.26) Γ (T,U;τ)dµ= log T U η(T)η(U)4 +cte , 2,2 2 2 − | | ZF whichagreeswith[8],uptoarenormalizat(cid:0)ionscheme-depend(cid:1)entadditiveconstant. The reader familiar with [8] may appreciate the elegance of the Rankin-Selberg- Zagier method compared with the lattice unfolding method. 3. One-loop modular integrals with harmonic elliptic genus While the integrand Φ in type II one-loop string amplitudes is of polynomial growth at the cusp τ i , this is not the case for heterotic strings, due to the → ∞ tachyoninthespectrumbeforeimposingtheGSOprojection. Instead,Φisamod- ular form of negative modular weight w = k/2 with a first order pole at the − cusp. The Rankin-Selberg-Zagier method described in 2.2 is not directly appli- § cable, however the unfolding trick could still be used provided Φ had a uniformly convergentPoincar´eseries representation (3.1) Φ= f γ w | γ∈XΓ∞\Γ with suitable Γ -invariant seed f(τ). f should grow as 1/qκ at τ if (3.1) ∞ 2 → ∞ is to represent a modular form with a κ-th order pole at the cusp, but uniform convergencerequires f(τ) τ1−w2 as τ 0. The naive choice f(τ)=1/qκ is fine ≪ 2 2 → for weight w>2 but fails for w 2. ≤ 3.1. Selberg-Poincar´e and Niebur-Poincar´e series. A first, natural op- tion for regulating the sum is to insert a non-holomorphic convergence factor `a la Hecke-Kronecker, i.e. choose a seed f(τ) = τs−w2 q−κ. The resulting Selberg- 2 Poincar´e series (3.2) E(s,κ,w;τ) 1 (cτ +d)−wτ2s−w2 e−2πiκcaττ++db ≡ 2 cτ +d2s−w (cX,d)=1 | | converges absolutely for (s) > 1, but analytic continuation to the desired value ℜ s = w is non-trivial, as it depends on deep properties of Kloosterman sums is 2 tricky, and in general non-holomorphic [42, 43]. Another undesirable feature of the Selberg-Poincar´e series (3.2) is that it is not an eigenmode of the weight w Laplacian on , rather4 H 1 (3.3) ∆ + (s w)(1 w s) E(s,κ,w)=2πκ(s w)E(s+1,κ,w) , H,w 2 − 2 − 2 − − 2 (cid:20) (cid:21) sothattheanalyticcontinuationtos= w isnotguaranteedtoyieldaholomorphic 2 result, nor even harmonic [44]. A much more convenient choice, which does not require analytic continuation, is the Niebur-Poincar´e series5 κτ (3.4) F(s,κ,w;τ)= 21 (cτ +d)−wMs,w − cτ +2d2 e−2πiκℜ(caττ++db) (cX,d)=1 (cid:18) | | (cid:19) 4Here∆H,w =2Dw−2D¯w whereDw,D¯w aredefinedin(3.8). Noticethechangeofconven- tioncomparedto[21]. 5TherelationbetweenthePoincar´eseries(3.2)and(3.4)canbefoundin[21,App. B]. RANKIN-SELBERG METHODS FOR CLOSED STRING AMPLITUDES 9 first introduced by Niebur [24] (for weight zero) and Hejhal [25] and revived in recent mathematical work on Mock modular forms[26, 27, 28, 29]. The seed f(τ)= s,w( κτ2)e−2πiκτ1, where M − (3.5) Ms,w(y)=|4πy|−w2 Mw2sgn(y),s−12 (4π|y|) , is proportional to the Whittaker function M (z), is uniquely determined by the λ,µ requirements that (s,κ,w;τ) be an eigenmode of the Laplacian on , F H (3.6) ∆ + 1(s w)(1 w s) (s,κ,w;τ)=0 , H,w 2 − 2 − 2 − F and that f(τ) ha(cid:2)s the desired growth at τ2 (cid:3) and τ2 0, →∞ → Γ(2s) (3.7) f(τ)∼τ2→∞ Γ(s+ w)q−κ , f(τ)∼τ2→0 |4πκτ2|s−w2e−2πiκτ1 . 2 The last property ensures that (s,κ,w) converges absolutely for (s) >1, while F ℜ a more detailed argument based on the Fourier expansion of (s,κ,w) (which can F be found in [29, 21]) shows that (s,κ,w) is holomorphic for (s) > 3 [45, 46]. F ℜ 4 Besides being an eigenmode of ∆ , (s,κ,w) also transforms in a simple way H,w under the raising, lowering and Hecke oFperators D,D¯,H defined by m iw (3.8) D = i ∂ , D¯ = iπτ2∂ , w π τ − 2τ w − 2 τ¯ (cid:18) 2(cid:19) aτ +b (3.9) (H Φ)(τ)= d−wΦ , m · d a,d>0bmodd (cid:18) (cid:19) X X ad=m namely [21] D (s,κ,w;τ)=2κ(s+ w) (s,κ,w+2;τ), w·F 2 F 1 D¯ (s,κ,w;τ)= (s w) (s,κ,w 2;τ). (3.10) w·F 8κ − 2 F − H (s,κ,w;τ)= d1−w (s,κm/d2,w;τ) . m ·F F d|X(κ,m) The decisive advantage of Niebur’s Poincar´e series over Selberg’s, however, is thatthevalues=1 w,degeneratewiththevalues= w undertheLaplacian(3.6), −2 2 lies in the convergence domain (s)> 1 (except for w = 0, which requires a more ℜ careful treatment). Eigenmodes of the Laplacian (3.6) with s = w or equivalently 2 s=1 w areknownas weak harmonic Maass forms (WHMS), and havea Fourier − 2 expansion near τ =i of the form6 ∞ ∞ ∞ (3.11) Φ= a qm+ (4πτ2)1−w¯b + mw−1¯b Γ(1 w,4πmτ )q−m m w−1 0 m − 2 m=−κ m=1 X X Weak holomorphic modular forms are a special case of WHMS, where the neg- ative frequency coefficients ¯b vanish. Mock modular forms are defined as the m analytic part Φ+ = ∞ a qm of a WHMS. Acting on any WHMS Φ of m=−κ m weight w with the lowering operator D¯ produces the complex conjugate of a holo- P morphic modular form Ψ = b qm of weight 2 w (the shadow) while m≥1 m − the iterated raising operator D1−w produces a weakly holomorphic modular form P 6Herewerestricttothecasewheretheshadowisregularatτ =i∞,see[46]forthegeneral expansion. 10 BORISPIOLINE Ξ= ∞ m1−wa qm of weight 2 w (the ghost ?) such that Φ+ is an Eichler m=−κ m − integralofΞ. In the caseof (1 w,κ,w), the shadow is the usualPoincar´eseries P F − 2 Ψ P( κ,2 w) = qκ , while the ghost is the Niebur-Poincar´e series γ,2−w ∝ − − | Ξ (1 w,κ,2 w). In particular, Ψ is a cusp form of weight 2 w, so must van∝isFh for−w2= 0, −2, P4, 6, 8, 12. Indeed, for these values, (1− w,κ,w) is − − − − − F − 2 an ordinary weak holomorphic modular form, e.g. (3.12) (1,1,0)=J +24 , (2,1, 2)=3!E4E6 , (7,1, 12)=13!/∆ ,... F F − ∆ F − whereE ,E aretheusualEisensteinseriesofweight4,6underSL(2,Z),∆=η24is 4 6 themodulardiscriminantandJ = E43 744=1/q+ (q)istheusualHauptmodul. ∆ − O Incontrast,forw= 10, (1,1, 10)isa genuineWHMS, withirrationalpositive − F − frequency Fourier coefficients and non trivial shadow, proportional to ∆ [47]. It is worth noting that for s = 1 w, the seed of the Niebur-Poincar´e series − 2 simplifies to −w (4πκτ )ℓ (3.13) f(τ)=Γ(2 w) q−κ q¯κ 2 , − − ℓ! ! ℓ=0 X whichplainlyshowstheimprovedultravioletbehaviorcomparedtothenaivechoice f(τ) q−κ. ∼ Using the Niebur-Poincar´eseries (s,κ,w) with s=1 w, we can now repre- F − 2 sent any weakly holomorphic modular form Φ of weight w 0 as a linear combi- ≤ nation of Niebur-Poincar´e series7 1 (3.14) Φ= a (1 w, m,w;τ)+a′ δ Γ(2 w) mF − 2 − 0 w,0 − −κ≤m<0 X where the coefficients are read off from the polar part Φ = a qm + −κ≤m≤0 m (1) at the cusp τ i . Indeed, the difference between the left and right-hand O → ∞ P sides of (3.14) is a harmonic Maass form of negative weight which is exponentially suppressed at the cusp (for suitable choice of a′ if w =0), hence vanishes [26]. In 0 particular, while each term in (3.14) may be a WHMF with non-trivial shadow, the shadow cancels in the linear combination (3.14). More generally, using the fact that almost, weakly holomorphic modular forms of weight w < 0 are linear combinations8 of iterated derivatives DnΦ of weakly holomorphic modular w−2n forms of weight w 2n, along with (3.10), we can similarly represent any almost, − weakly holomorphic modular form of weight w <0 as a linear combination n (3.15) Φ= a(p)(m) (1 w +p, m,w;τ) +a′ δ F − 2 − 0 w,0 p=0−κ≤m<0 X X wherenisthedepth(i.e. themaximalpowerofEˆ ). Asanexamplerelevantforthe 2 computation of threshold corrections to gauge couplings in heterotic string theory 7Similarly, modular forms under congruence subgroups of SL(2,Z) can be represented as linear combinations of Niebur-Poincar´e series attached to all cusps, see [22] for the example of theHeckecongruence subgroupΓ0(N). 8Thisisnotthecaseforalmost,weaklyholomorphicmodularformsofweightw>0,(Eˆ2)nJ beinga counter-example. Such cases can be treated by considering s-derivatives of F(s,κ,w)at s=w/2[48].

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.