Arxiv:1603.09556V1 [Math.NT]

Home , Weil conjectures

arXiv:1603.09556v1 [math.NT] 31 Mar 2016 lsia -ucinaddnt by denote and ∆-function classical hsppr ecnie h aeo iglmdlrfr fgenus of form modular Siegel of case the consider we paper, conjectur this Weil the of geometry. proof algebraic from Deligne’s methods from complicated follows (1.1) estimate The upfr fweight of form cusp hscnetr a engnrlzdfrgnrl oiieintegra positive general, for generalized so-called been The has conjecture This hr r oneeape oigfo it c.[13]). (cf. lifts from coming counterexamples are there iglmdlrfrso ml weight. small of forms modular Siegel For gemn .352 AQSER. results - these 335220 to Framework n. Seventh leading Union’s agreement research European the the under and Council Research foundation Krupp the of where (1.2) ihFuircoefficients Fourier with ttsta,for that, states k arx hnacnetr fRsio n adã[6 asthat says Saldaña [16] and Resnikoff of conjecture a Then matrix. (1.1) upfr nacnrec ugop hn as then, subgroup, congruence a on form cusp hr r ayrltdcnetrsfrmr opiae ye o types complicated more for conjectures related many are There For orcl hti nw o eu ,lt∆( let 1, genus for known is coeffi what Fourier recall for bounds To existing improve to is paper this of goal The h eerho h uhrwsspotdb h lre rp Pr Krupp Alfried the by supported was author the of research The MRVDBUD O ORE OFIINSO SIEGEL OF COEFFICIENTS FOURIER FOR BOUNDS IMPROVED g hsi xcl h aaua-eeso ojcue o high For conjecture. Ramanujan-Petersson the exactly is this 1 = g > k ,tebs nw siaeis estimate known best the 1, + aaua-eeso conjecture Ramanujan-Petersson p prime, k 1. ∈ nrdcinadsaeeto results of statement and Introduction N c a g a ( ihrsett h iglmdlrgopΓ group modular Siegel the to respect with a ( T := ( T T ,where ), a ) ( )        n ≪ ≪ ) ε,F OUA FORMS MODULAR 1 2 4 36 13 ≪ 1 ε,F g AHI BRINGMANN KATHRIN τ + ε,f det( ( det( n T | t ore offiins The coefficients. Fourier its ) n τ 1 sapstv ent ymti half-integral symmetric definite positive a is T k ( T − − 2 p ) 1 ) ) k 2 + ≤ | 1 g k 2 − ε − 1 c g ( α ttsta if that states g + 2 τ +1) 4 g ε p n := ) 11 2 + ∞ → . ε if if if g g > g ( q ( > ε Q > ε [3] 3 = [13] 2 = rgam F/0721)/ECGrant ERC / (FP/2007-2013) Programme , a eevdfnigfo h European the from funding received has ( n ∞ > ε [2]. 3 =1 0) 0) f (1 , ( . z o on nvriyTeachers University Young for ize 0) τ = ) , − . , > g q egtmdlrforms. modular weight l n uoopi om.In forms. automorphic f P ) g aaua conjecture Ramanujan .Frti,let this, For 1. 24 s[,8,uighighly using 8], [7, es =Sp := n ∞ inso ihrgenus higher of cients =1 rgenus er ( q a ( := g n ( ) Z q e ) n 2 g πiτ ⊂ saweight a is however, , ethe be ) GL F g 2 g ea be × ( Z g ) Here g 1 2 α−1 := 4(g 1)+4 − + . g − 2 g +2 In [4] and [5] it was shown that (1.2) still holds for k = g + 1 and k = g, respectively. Moreover, for (g + 3)/2

k − 1− 1 α +ε (1.3) a(T ) det(T ) 2 ( g ) g . ≪ε,F In this paper we improve (1.3) and obtain Theorem 1.1. We have for g/2+1

k + g−k − 1 − 1− 1 α +ε a(T ) det(T ) 2 2g(g−2) 2g ( g ) g . ≪ε,F Remark. Theorem 1.1 is indeed an improvement since g k 1 − < 0. 2g(g 2) − 2g − Our proof follows the idea of [2] using a Jacobi decomposition of Siegel modular forms. Our main achievement is an improved bounds for Kloosterman sums. The paper is organized as follows. In Section 2 we recall basic facts about Jacobi forms and their relation to Siegel modular forms. In Section 3 we bound higher dimensional Kloosterman sums. Section 4 is devoted estimating coeﬃcients of Poincar´eseries, in Section 5 we then conclude our main theorem. Acknowledgments The author thanks Winfried Kohnen and Mike Woodbury for comments on an earlier version of this paper.

2. Preliminaries 2.1. Basic facts on Jacobi forms. Here we recall some basic facts about Jacobi cusp forms; for details we refer the reader to [9] and [19]. The Jacobi group ΓJ := SL (Z) ⋉ (Zg Zg) g 2 × acts on H Cg in the usual way by (( a b ) SL (Z), (λ,µ) Zg) × c d ∈ 2 ∈ a b aτ + b z + λτ + µ , (λ,µ) (τ, z) := , . c d ◦ cτ + d cτ + d Note that throughout vectors are viewed as columns unless noted otherwise. Let k N, m a b ∈J be a positive definite symmetric half-integral g g matrix, γ = (( c d ) , (λ,µ)) Γg , and φ : H Cg C. Then we define the following Jacobi× slash action ∈ × → φ γ(τ, z) := (cτ +d)−ke c cτ + d)−1m[z + λτ + µ]+ m[λ]τ +2λT mz φ(γ (τ, z)), |k,m − ◦ where e(w) := e2πiw ( w C) , and where A[B] := BT AB for matrices A and B of compatible sizes. ∀ ∈ A holomorphic function φ : H Cg C is called a Jacobi cusp form of weight k and index J × →J m with respect to Γg , if, for all γ Γg , we have φ k,mγ = φ, and φ has a Fourier expansion of the form ∈ | φ(τ, z)= c(n, r)e nτ + rT z , D>0 X 2 2n rT g cusp where D := det r 2m with n N and r Z . We denote by Jk,m the vector space of Jacobi cusp forms. ∈ ∈ cusp The space Jk,m is a finite dimensional Hilbert space with the Petersson scalar product

−1 k J φ, ψ := φ(τ, z)ψ(τ, z) exp 4πm[y] v v dVg , ˆ J g h i Γg \H×C − · J −g−2 where dVg := v dudvdxdy, τ = u + iv, and z = x + iy. 2.2. Jacobi Poincaréseries. We next recall certain Jacobi Poincaréseries, as considered in [2]. For n Z, r Zg, and m a positive definite symmetric half-integral g g matrix such that 4n >∈ m−1[r∈], define a Poincaréseries of exponential type by ×

n,r (2.1) Pk,m;(n,r)(τ, z) := e γ(τ, z), k,m J J γ∈ΓXg,∞ Γg

n,r 2πi(nτ+rT z) J 1 n g where e (τ, z) := e and Γg,∞ := (( 0 1 ), (0,µ)) n Z,µ Z is the stabilizer n,r cusp{ | ∈ ∈ } group of e . For k>g + 2, Pk,m;(n,r) Jk,m and the Petersson coefficient formula holds cusp ∈ (φ J with Fourier coefficients c ), ∈ k,m φ (2.2) φ,Pk,m;(n,r) = λk,m,Dcφ(n, r), where − g g −k+ g +1 k− g+3 −k+ g +1 λ := 2 2 Γ k 1 (2π) 2 det(2m) 2 D 2 . k,m,D − 2 − For k g + 2 the Poincaréseries (2.1) diverge. However there is a way to analytically continue≤ them, using the so-called Hecke trick. We denote the corresponding functions again by Pk,m;(n,r). We have [2, 4, 5]: cusp Proposition 2.1. For k >g/2+2, the functions Pk,m;(n,r) are elements of Jk,m . We have the Fourier expansions ± ′ ′ ′ ′T Pk,m;(n,r)(τ, z)= gk,m;(n,r)(n ,r )e n τ + r z , n′∈Z,r′∈Zg DX′>0 ′ 2n′ r′T where D := det r′ 2m and ± ′ ′ ′ ′ k ′ ′ g (n ,r ) := gk,m;(n,r)(n ,r )+( 1) gk,m;(n,r)(n , r ) k,m;(n,r) − − with

k g 1 ′ 2 − 4 − 2 ′ ′ ′ ′ k − 1 D (2.3) g (n ,r ) := δ (n, r, n ,r )+2πi det(2m) 2 k,m;(n,r) m D ′ T −1 ′ ′ ′ 2π√DD − g −1 e2c r m r Hm,c(n, r, n ,r )Jk− g −1 c 2 . × 2 det(2m)c c≥1 ! X 2πix Here ec(x) := e c ,

′ ′ g ′ ′ 1 if D = D and r r 2mZ , δm(n, r, n ,r ) := − ∈ (0 otherwise 3 and the Kloosterman sums

′ ′ T ′ ′T Hm,c(n, r, n ,r ) := ec (m[λ]+ r λ + n)d + n d + r λ , λ (mod c) d (modX c)∗ where by λ (mod c), we mean that all components run (mod c) and d (mod c)∗ sums only over d (mod c) which are coprime to c. Moreover formula (2.2) holds. Remark. Note that in [2] the Kloosterman sums have a slightly different normalization. cusp Proposition 2.1 gives that for k >g/2 + 2 the Pk,m;(n,r) are a generating system of Jk,m . We easily obtain, just using the Cauchy-Schwarz inequality Lemma 2.2. For k >g/2+2 and φ J cusp with Fourier coefficients c (n, r), we have ∈ k,m φ k g 1 1 2 − 4 − 2 2 D cφ(n, r) k bn,r Pk,m;(n,r) φ . k − 1 (g+3) | |≪ det(2m) 2 4 k k Thus, to get bounds for the Fourier coefficients of Jacobi forms, one only has to bound the Fourier coefficients of the Poincaréseries which are explicitly given in Proposition 2.1. However, we also bound coefficients of Siegel modular forms, which requires estimating φ . The connection between Siegel modular forms and Jacobi forms is described in the nextk k subsection.

2.3. Relation to Siegel modular forms. Let Hg be the usual Siegel upper half space and τ zT g−1 ′ write Z H as Z = ′ with τ H, z C , and τ H . Then F S (Γ ), the ∈ g z τ ∈ ∈ ∈ g−1 ∈ k g space of Siegel cusp forms of weight k for Γg, has a so-called Fourier Jacobi expansion of the form 2πi tr(mτ ′) F (Z)= φm(τ, z)e , m>0 X where tr denotes the trace of a matrix and where m runs through all positive definite symmetric half-integral (g 1) (g 1) matrices. It is well-known, that the coefficients of φm are Jacobi cusp forms.− So bounds× − for the Fourier coefficients of Siegel modular forms follow from the understanding of the coefficients of Jacobi forms.

3. Bounding Kloostermann sums A first step in bounding Fourier expansions of Poincaréseries is to estimate certain higher- dimensional Kloosterman sums which occur when restricting the Fourier coefficients of Jacobi Poincaréseries to the diagonal (n′,r′)=(n, r). To be more precise, we set H± (n, r) := H (n, r, n, r). m,c m,c ± To bound these, we require well-known evaluations of (generalized) Gauss sums

2 G(a, b; c) := ec an + bn . n (mod c) X Lemma 3.1. Let p be prime, a, b Z, ν N, and α := ordp(a). ∈ ∈ 4 (1) For α ν, we have ≥ pν if b 0 (mod pν), G (a, b; pν)= ≡ (0 otherwise. (2) For 0 α < ν, G(a, b; pν)=0 unless b 0 (mod pα) in which case we have the following evaluations:≤ ≡ (i) If p =2 and b 0 (mod pα), then 6 ≡ α ν α+ν a/p 2 4a G (a, b; p )= p 2 ε ν−α e ν+α b , p pν−α p − pα ν+a where ℓ denotes the inverse of ℓ (mod p ) and εj = 1 or i depending on whether j 1 (mod 4) or j 3 (mod 4), respectively. (ii) If≡p =2 and b 0 (mod≡ pα), then G(a, b; pν) equals ≡ 2ν if α = ν 1 and b 0 (mod2ν), ν+α ν−α − 6≡ −2 a 2 a α+1 2 2 α ε (1 + i)e2ν+α+2 b α if b 0 (mod2 ) and ν α (mod 2),  a/2 2α − 2 ≡ ≡  0 otherwise,  where ℓ denotes the inverse of ℓ (mod 2ν+α+2). We are now ready to bound the higher-dimensional Klosterman sums. Lemma 3.2. We have ± g+1 1 H (n, r) (D,c) c 2 det(2m) 2 . m,c ≪ Proof: Our proof closely follows the one in [2]. There it was shown on page 507 that, for c = c1c2 with (c1,c2) = 1, ± ± ± Hm,c(n, r)= Hc1m,c2 (nc1,r)Hc2m,c1 (nc2,r), where c1 and c2 are inverses of c1 and c2 modulo c2 and c1, respectively. Thus we may assume that c = pν with p prime and ν N and for simplicity we for now restrict to p = 2. The modifications required for p = 2 follow∈ along the same lines as in [2]. 6 Since a non-degenerate binary quadratic form over Z (p = 2) is diagonalizable, we may p 6 assume that m = diag(m1,...,mg) is a diagonal matrix. Set µj := ordp(mj) (1 j g). We assume without loss of generality that ν µ for 1 j ℓ and ν>µ for ℓ +1≤ j≤ g. ≤ j ≤ ≤ j ≤ ≤ Write r =(r1,...,rg). From (18) of [2], we conclude that g ± 2 H ν (n, r)= e ν n d + d e ν m λ + r λ d r λ . m,p p p j j j j ± j j d (mod pν )∗ j=1 λ (mod pν ) X Y j X ν The sum on λj equals G(mj,rj(d 1); p ) and we may use Lemma 3.1 to evaluate it. For 1 j ℓ, we have ± ≤ ≤ pν if r d 1 0 (mod pν ), G m ,r d 1 ; pν = j ± ≡ j j ± r d pν . (0 if j 1 0 (mod ) 5 ± 6≡ For ℓ +1 j g, the Gauss sum equals ≤ ≤ ν+µj µj 2 mj /p 2 mj νj p 2 ε ν−µj ν−µ e ν+µj r d 1 4 µ if r d 1 0 (mod p ), p p j p − j ± p j j ± ≡ νj (0 if rj d 1 0 (mod p ). ± 6≡ ± Thus Hm,pν (n, r) becomes (3.1) g νj ν µ mj/p + j 2 mj νℓ 2 2 p ε ν−µj p e ν+µj rj d 1 4 . p pν−µj p − ± pµj d (mod pν ) j=ℓ+1 X ν Y rj (d±1)≡0 (mod p )(1≤j≤ℓ) µj rj (d±1)≡0 (mod p )(ℓ+1≤j≤g) We now consider whether pν D or not. If pν D, then we have (D,p| ν)= pν. We bound (3.1) trivially, yielding | ν 1 g νg νℓ 1 g ± νℓ ν (g−ℓ) Pj ℓ µj ν + + Pj ℓ µj H ν (n, r) p p p 2 p 2 = +1 =(D,p )p 2 2 2 = +1 . m,p ≤ · · Now νg + ν g +1 νℓ + 1 Pg µ 1 Pg µ 1 − ν p 2 2 = c 2 , p 2 2 j=ℓ+1 j p 2 j=1 j det(2m) 2 , p 2 1, ≤ ≤ ≤ giving the claim in this case. If pν ∤ D, then we use that 1 (3.2) D = det(2m) 4n m−1[r] , 2 − which follows from the Jacobi decomposition. This gives that pν divides at most one of the ν mj. There are two cases to distinguish depending on whether p divides one of the mj or none. We first assume ν>µj for 1 j g and let λ := ordp(D). In (27) of [2] it was shown that ≤ ≤ ± ν(g+1) ν g+1 ν H (n, r) 2p 2 (D,p ) c 2 (D,p ) . m,c ≤ ≪ This implies the claim in this case. ν Finally we consider the case that p divides exactly one mj and we may assume without loss of generality that µg ν. Let κ := max 0, ν ρg,µ1 ρ1, ,µg−1 ρg−1 , where ρ = ord(r )(1 j g). It≥ is shown in the first{ displayed− formula− ··· on page 509− of [2]} that j j ≤ ≤ g−1 ± 2ν−κ 1 (ν+µ ) ν(g+1) +ν−κ 1 Pg−1 µ (3.3) H (n, r) p p 2 j = p 2 p 2 j=1 j . m,c ≤ j=1 Y We next analyze (3.2). Since m−1 = diag(m−1,...,m−1), we obtain, since pν m , 1 g | g g g g g−1 D =2g+1n m 2g−1 r2 m 2g−1r2 m (mod pν). j − j j ≡ − g j j=1 j=1 ℓ=1 j=1 Y X Yℓ6=j Y Thus, since λ < ν, g−1

λ = µj +2ρg. j=1 X 6 Moreover, from the deﬁnition of κ, we obtain that κ ν ρ . Thus, by (3.3), ≥ − g ± ν (g+1)+ 1 Pg−1 µ +ρ ν (g+1)+ λ g+1 1 g+1 H (n, r) p 2 2 j=1 j g = p 2 2 = c 2 D 2 c 2 (D,c). m,c ≤ ≤ This ﬁnishes the proof.

4. Bounding coefficients of Poincare´ series

In this section, we estimate the Fourier coefficients bn,r of Pk,m;(n,r). This is of independent interest for obtaining bounds for Fourier coefficients of Jacobi forms. Theorem 4.1. Assume that k N satisfies (g + 3)/2

g +ε D 2 1 k−g−1 −k+g+1+ g−1 (−k+g+1)+ε bn,r Pk,m;(n,r) 1+ g+1 1+ D det(2m) . ≪ det(2m) 2 ! Proof: We use the explicit representation of bn,r given in Proposition 2.1. The ﬁrst term in (2.3) yields the ﬁrst term in the bound in Theorem 4.1. Thus, we have to bound

± 2πD − g −1 fm(n, r) := H (n, r) Jk− g −1 c 2 . m,c 2 det(2m)c c≥1 X We may rewrite

g A − 2 −1 ± fm(n, r)= (cd) H (n, r) Jk− g −1 , m,dc 2 c d|D c≥1 X c,XD =1 ( d ) 2πD where A = Ad := d det(2m) . To bound the inner sum, we split it into three pieces: a part with c A, a contribution from A c B, and a piece with c B, with B to be determined later.≤ Note that the range of≤ any≤ of these sums is allowed≥ to be empty. We require the bounds for Kloosterman sums from Section 3 as well as the following estimates (4.1) H± (n, r) cg+ε(D,c), m,c ≪ − 1 ℓ (4.2) J (t) min t 2 , t . ℓ ≪ℓ The bound (4.1) can be found in [2] whereas (4.2)n is standard.o To bound the part with c A, we use (4.1) with dc instead of c and the ﬁrst estimate in (4.2). This gives the contribution≤ for the sum on c g 2 +ε − 1 g +ε g − 1 +ε g +ε g +ε D A 2 d 2 c 2 2 A 2 d 2 . ≪ ≪ det(2m) 1≤c≤A X − 1 Upon multiplying by det(2m) 2 , this yields the second summand in the bound in Theorem 4.1. Next we estimate the part with A c B. For this, we use (4.1) and the second estimate in (4.2). This gives the contribution,≤ using≤ that k

k− g −1 g +ε −k+g+ε k− g −1 g +ε −k+g+1+ε A 2 d 2 c A 2 d 2 B . ≪ A≤c≤B X 7 Finally, we estimate the piece with c B. For this, we use Lemma 3.2 and the second estimate in (4.2). This gives the bound ≥

k− g −1 1 1 g+1 −k g+3 −k k− g −1 1 1 A 2 d 2 det(2m) 2 c 2 B 2 A 2 d 2 det(2m) 2 ≪ c≥B X since k > (g + 3)/2. Now, to minimize the error, we choose B such that the second and third error agree (up to ε exponents). One can show that this is the case for

1 B = d−1 det(2m) g−1 .

− 1 Plugging back in gives the claim after multiplying by det(2m) 2 .

5. Proof of theorem 1.1

In this section we use the previous bounds with g g 1 and φ = φm, where φm comes from the Jacobi coeﬃcients of a Siegel modular form.7→ We− recall the following bound from Proposition 2 of [2].

Lemma 5.1. If φm is the mth Fourier-Jacobi coefficient of a Siegel modular form F , then k −α +ε φ det(2m) 2 g (ε> 0). || m|| ≪ε,F We are now ready to prove Theorem 1.1 Proof of Theorem 1.1: By Lemma 2.2 and Lemma 5.1, 1 k − g − 1 g + 1 −α +ε a(T ) b P 2 D 2 4 4 det(2m) 4 2 g . ≪ n,r k,m;(n,r) Theorem 4.1 then yields 1 − g 2 k − g − 1 g + 1 −α +ε A(T ) det(2m) 2 f(m, D) D 2 4 4 det(2m) 4 2 g , ≪ where g g−1 +ε k−g −k+g+ 1 (−k+g)+ε f(m, D) := det(2m) 2 + D 2 1+ D det(2m) g−2 . Define m (T ) := min T [U] U GL (Z) , g−1 { |g−1 | ∈ g } where T [U] g−1 denotes the determinant of the leading (g 1)-rowed submatrix of T [U]. Since both| sides of the bound in Theorem 1.1 are invariant− under replacing T by T [U] T (U GL (Z)), we may assume that T = n r /2 with det(m) = m (T ). Now, by ∈ g r/2 m g−1 reduction theory, 1− 1 det(m)= m (T ) D g . g−1 ≪ It is easy to see that the powers of det(m) in f(m, D) are all non-negative. So we may 1− 1 replace det(m) by D g in this expression. One can then show that the last term in f(m, D) is dominant. This gives g−k 1 −α +ε k − 1 + +ε a(T ) det(2m) 2 g D 2 2 2g(g−2) . ≪ This yields the claim of the theorem since 1/2 αg > 0. 8 − References [1] N. Abramowitz and I. Stegun, Handbook of mathematical functions. Dover Books on Mathematics, New York (1965). [2] S. Böcherer and W. Kohnen, Estimates for Fourier coefficients of Siegel cusp forms. Math. Ann. 297 (1993), 499-517. [3] S. Breulmann, Abschätzungen für Fourierkoeffizienten Siegelscher Spitzenformen vom Geschlecht 3. Doktorarbeit, Universität Heidelberg (1996). [4] K. Bringmann, Estimates of Fourier coefficients of Siegel cusp forms for subgroups and in the case of small weight. J. London Math. Soc. 73 (2006), 31-47. [5] K. Bringmann and T. Yang, On Jacobi Poincaréseries of small weight. Int. Math. Res. Not. IMRN, (2007). [6] J. Cassels, Rational quadratic forms. L.M.S. Monographs, No.13. (1978) London, New York, San Fran- cisco: Academic Press. [7] P. Deligne, La conjecture de Weil. I. Publ. Math., Inst. Hautes Etud.´ Sci. 43 (1973), 273-307. [8] P. Deligne, J.-P. Serre, Formes modulaires de poids 1. Ann. Sci. Ec.´ Norm. Supér., IV. Sér. 7 (1974), 507-530. [9] M. Eichler and D. Zagier, The theory of Jacobi forms. Progress in Mathematics, 55. (1985) Boston- Basel-Stuttgart: Birkhäuser. [10] E. Freitag, Siegelsche Modulfunktionen. Berlin Heidelberg New York: Springer (1983). [11] E. Hecke, Mathematische Werke. Hrsg. im Auftr. der Akademie der Wissenschaften in Göttingen. 3., durchges. Aufl. Göttingen: Vandenhoeck & Ruprecht (1983). [12] W. Kohnen,On the average growth of Fourier coefficients of Siegel cusp forms of genus 2. Pac. J. Math. 179 (1997), 119-121. [13] W. Kohnen, Estimates of Fourier coefficients of Siegel cusp forms of degree two. Compos. Math. 87 (1993), 231- 240. [14] W. Kohnen, A Maass space in higher genus. Compos. Math. 141 (2005), 313-322. [15] A. Krieg, A Dirichlet series for modular forms of degree n. Acta Arith. 59 (1991), No.3, 243-259. [16] H. Resnikoff, Some properties of Fourier coefficients of Eisenstein series of degree two. J. Reine Angew. Math. 265 (1974), 90-109. [17] M. Sato and T. Shintani, On zeta functions associated with prehomogeneous vector spaces. Ann. Math. (2) 100 (1974), 131-170. [18] T. Yamazaki, Rankin-Selberg method for Siegel cusp forms. Nagoya Math. J. 120 (1990), 35-49. [19] C. Ziegler, Jacobi forms of higher degree. Abh. Math. Semin. Univ. Hamb. 59 (1989), 191-224.

Mathematical Institute, University of Cologne, Weyertal 86-90, 50931 Cologne, Ger- many E-mail address: [email protected]