Classical modular forms p-adic and overconvergent modular forms Explicit computations A Gentle Introduction to Overconvergent Modular Forms L. J. P. Kilford University of Bristol 27 July 2009 L. J. P. Kilford (University of Bristol) A Gentle Introduction to Overconvergent Modular Forms Classical modular forms p-adic and overconvergent modular forms Explicit computations Congruence subgroups We recall the definition of the modular group SL2 (Z): SL2 (Z) := ca db : a, b, c, d ∈ Z such that ad − bc = 1 . A congruence subgroup Γ of SL2 (Z) is a subgroup which contains Γ(N) := ca db ∈ SL2 (Z) such that ca db ≡ ( 10 01 ) mod N for some integer N ≥ 2. We say that Γ has level N. L. J. P. Kilford (University of Bristol) A Gentle Introduction to Overconvergent Modular Forms Classical modular forms p-adic and overconvergent modular forms Explicit computations Modular forms A classical modular form f of weight k for the congruence subgroup Γ is classically described as a holomorphic function from the Poincaré upper half plane to C which satisfies az + b a b k = (cz + d) f (z) for ∈ Γ. f c d cz + d A famous example of a modular form for the full modular group is the ∆-function, given by ∆(q) = q ∞ Y n 24 (1 − q ) n=1 = ∞ X τ (n)q n , n=1 where q := exp(2πiz) and τ (n) is the Ramanujan τ function. L. J. P. Kilford (University of Bristol) A Gentle Introduction to Overconvergent Modular Forms Classical modular forms p-adic and overconvergent modular forms Explicit computations The Eisenstein series We can define modular forms of even weight k ≥ 4 for SL2 (Z) by Ek (z) := X c,d∈Z (c,d)=1 1 . (cz + d)k These can be shown to transform correctly under the two 0 1 generators of SL2 (Z) ( −1 0 and ( 10 11 )) and to have no poles on the upper half-plane. We can show that the Fourier expansion at ∞ of the Eisenstein series is ∞ ζ(1 − 2k) X + σk−1 (n)q n . Ek (q) := 2 n=1 L. J. P. Kilford (University of Bristol) A Gentle Introduction to Overconvergent Modular Forms Classical modular forms p-adic and overconvergent modular forms Explicit computations What about k = 2? If we make the same definition for k = 2, then we find that we can’t rearrange the terms of the series defining E2 , so we fail to obtain a modular form (this is called a quasi-modular form in the literature). We can obtain a Fourier series, though, which is E2 (q) := 1 − 24 ∞ X σ1 (n)q n . n=1 We will encounter E2 again later on. L. J. P. Kilford (University of Bristol) A Gentle Introduction to Overconvergent Modular Forms Classical modular forms p-adic and overconvergent modular forms Explicit computations A complete answer? We can in fact write every modular form for SL2 (Z) as a graded polynomial in E4 and E6 ; it can be shown that M = C[E4 , E6 ] and S = ∆ · M, where E43 − E62 , 1728 and a similar result holds for other congruence subgroups. ∆= However, this does not tell the whole story. L. J. P. Kilford (University of Bristol) A Gentle Introduction to Overconvergent Modular Forms Classical modular forms p-adic and overconvergent modular forms Explicit computations The j-invariant The modular j-invariant is an important modular function (as opposed to modular form). It has weight 0 and is defined by j(q) := 1 E4 (q)3 = +744+196884q+21493760q 2 +864299970q 3 +· · · ∆(q) q It has a single simple pole at ∞, and satisfies several interesting √ properties; for instance, because τ = 1+ 2−163 is an element of a field with class number 1, j(τ ) is an integer; this means that eπ √ 163 = 262, 537, 412, 640, 768, 743.999 999 999 999 25 . . . ≈ 640, 3203 + 744. It is also closely connected to the Monster group, the largest of the sporadic finite simple groups (which has an irreducible representation of dimension 196883). L. J. P. Kilford (University of Bristol) A Gentle Introduction to Overconvergent Modular Forms Classical modular forms p-adic and overconvergent modular forms Explicit computations Theta series Another standard example of modular forms are provided by theta series. We can take a quadratic form and use it to create a modular form; for instance, we have X 2 2 θ(q) := q x +y = 1 + 4q + 4q 2 + 4q 4 + 8q 5 + · · · m,n∈Z This is a modular form of weight 1 of level 4. One motivation for studying modular forms is to give bounds on the number of ways an integer can be represented by a quadratic form (if one is very lucky, one can derive formulae for them a la Glaisher). L. J. P. Kilford (University of Bristol) A Gentle Introduction to Overconvergent Modular Forms Classical modular forms p-adic and overconvergent modular forms Explicit computations Finite-dimensionality We can prove that given a fixed level and weight, the space of modular forms of that level and weight is finite-dimensional as a complex vector space; this allows us to perform computations effectively. o o B A OO OO .T (P) .P OO 33 C . D ω . E i F ++ H . G −ω _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ 1 L. J. P. Kilford (University − of2 Bristol) 1 2 A Gentle Introduction to Overconvergent Modular Forms Classical modular forms p-adic and overconvergent modular forms Explicit computations Number-theoretic identities Using the finite-dimensionality of spaces of modular forms, we can prove theorems like σ7 (n) = σ3 (n) + 120 n−1 X σ3 (i) · σ3 (n − i); i=1 this follows because M4 (SL2 (Z)) = CE4 and M8 (SL2 (Z)) = CE8 , so we can prove it by noting that E8 = E42 and comparing their Fourier coefficients. L. J. P. Kilford (University of Bristol) A Gentle Introduction to Overconvergent Modular Forms Classical modular forms p-adic and overconvergent modular forms Explicit computations Hecke operators We can define linear operators called Hecke operators Tp (called Up if p | N) on modular forms in many intrinsic ways. The effect on Fourier expansions is given by (Tp f )(q) = ∞ X X apn + χ(p)p k−1 an/p q n = bn q n , n=0 n=0 where χ is a Dirichlet character and an/p = 0 if n/p is non-integral. We can find examples of forms which are eigenforms for all of these Hecke operators; examples are ∆, θ, and the Eisenstein series Ek . L. J. P. Kilford (University of Bristol) A Gentle Introduction to Overconvergent Modular Forms Classical modular forms p-adic and overconvergent modular forms Explicit computations Slopes of modular forms If f is an eigenform for Tp , then we say that the p-slope of f is ap vp . a1 For instance, ∆(q) = q − 24q 2 + 252q 3 + O(q 4 ), so the 2-slope of ∆ is 3. The slopes will play an important role in the p-adic theory we will be introducing later, but they can be understood purely in terms of the classical forms. L. J. P. Kilford (University of Bristol) A Gentle Introduction to Overconvergent Modular Forms Classical modular forms p-adic and overconvergent modular forms Explicit computations Example of slopes The T2 operator acting on the space S42 (SL2 (Z)) has characteristic polynomial x 3 + 344688x 2 − 6374982426624x − 520435526440845312; the Newton polygon of this is: • |• || || || || || || || || || || || || || || || || || || || || || || || || || || || | | || || || || || || || || || || l• lll lll lll lll lll lll lll lll lll lll lll lll lll lll lll lll lll lll lll l •_ll_ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ the slopes here are [4, 9, 12]. L. J. P. Kilford (University of Bristol) A Gentle Introduction to Overconvergent Modular Forms Classical modular forms p-adic and overconvergent modular forms Explicit computations Two Conjectures Maeda’s Conjecture says that the characteristic polynomials of the Hecke operators Tp acting on Sk (SL2 (Z)) are irreducible. This is still open (it has been checked for many choices of p and k, and density results are also known). It is also not known whether the positive Fourier coefficients of ∆(q) = q ∞ Y (1 − q n )24 = n=1 ∞ X τ (m)q m m=1 are all nonzero! L. J. P. Kilford (University of Bristol) A Gentle Introduction to Overconvergent Modular Forms Classical modular forms p-adic and overconvergent modular forms Explicit computations p-adic modular forms If p is a prime number, then we can define the q-expansion of a p-adic modular form of weight κ to be the p-adic limit of q-expansions of classical modular forms of weights k which tend p-adically to κ. For instance, it is well-known that Epr (p−1) ≡ 1 mod p r +1 · (p − 1) for r ≥ 1, and using this formulation means therefore that we can think of these Eisenstein series as being p-adic approximations to the constant form 1 in weight 0. They form a “p-adic family”. L. J. P. Kilford (University of Bristol) A Gentle Introduction to Overconvergent Modular Forms Classical modular forms p-adic and overconvergent modular forms Explicit computations Fitting E2 into the picture We recall the quasi-modular form E2 that we saw earlier. This is not a modular form, but given any p we can write it as a p-adically convergent sum of p-adic modular forms, so it is a p-adic modular form for any p. If we define Ek∗ (z) := Ek (z) − p k−1 Ek (pz), for any Eisenstein series Ek , then we can expand Ek as Ek = Ek∗ (z)+p k−1 Ek∗ (z)+p 2(k−1) Ek∗ (pz)+· · ·+p m(k−1) Ek∗ (p m z)+· · · , and each of these terms is a classical modular form. If we put in k = 2 then we have our result. L. J. P. Kilford (University of Bristol) A Gentle Introduction to Overconvergent Modular Forms Classical modular forms p-adic and overconvergent modular forms Explicit computations Caveats about the p-adics The previous definition is known as “p-adic forms in the sense of Serre”; this definition is fairly clear but fails to encompass everything. A more comprehensive but rather more abstract definition was given by Katz; this is in terms of a rule sending p-adic test objects to elements of a p-adic ring; the rule must satisfy conditions similar to those defining classical modular forms. The definition also uses the theory of canonical subgroups of elliptic curves. Also, the definition in terms of Serre does not give the whole picture for very small primes (such as 2 and 3). L. J. P. Kilford (University of Bristol) A Gentle Introduction to Overconvergent Modular Forms Classical modular forms p-adic and overconvergent modular forms Explicit computations Problems with the p-adics Unfortunately the p-adic modular forms as defined above are not optimal for computations, because although we can define analogues of the Hecke operators for them, these extended Hecke operators are not compact. Here is an example of the sort of 1 0 0 1 0 0 .. .. . . L. J. P. Kilford (University of Bristol) thing that goes wrong: 0 ... 0 ... 1 ... .. . . . . A Gentle Introduction to Overconvergent Modular Forms Classical modular forms p-adic and overconvergent modular forms Explicit computations Overconvergence We are therefore motivated to find a large subspace of the p-adic modular forms where the Up operator is compact. The formal definition of overconvergence is rather complicated, but here is an idea. If the space of p-adic modular forms of weight 0 is given by sums of the form ∞ X ai j i , where ai ∈ Cp , i=0 then the overconvergent modular forms of weight 0 are ∞ X ai j i , where ai ∈ Cp , and |ai |p → 0 as i → ∞. i=0 L. J. P. Kilford (University of Bristol) A Gentle Introduction to Overconvergent Modular Forms Classical modular forms p-adic and overconvergent modular forms Explicit computations Handwavy overconvergence The idea of overconvergence is to consider the definition of p-adic modular forms as sections of structure sheaves on a modular curve viewed as a rigid analytic space, and extend it slightly. The problems one has to deal with is that there are bad areas on the modular curve that we don’t want to define our forms in, but if we exclude too much of the curve then we get way too many functions (these are the p-adic modular forms). L. J. P. Kilford (University of Bristol) A Gentle Introduction to Overconvergent Modular Forms Classical modular forms p-adic and overconvergent modular forms Explicit computations Details for experts We also need to be careful about the positions of the poles of the classical Eisenstein series we are using. If we (very carefully) extend the range of definition of our forms a little way into the bad areas, then this excludes a lot of the more badly-behaved forms and gives us a nicer ring of forms to deal with that have good arithmetic (the overconvergent modular forms). The area we are extending into is the locus of not-too-supersingular curves; these are the supersingular curves for which there is a defined canonical subgroup. L. J. P. Kilford (University of Bristol) A Gentle Introduction to Overconvergent Modular Forms Classical modular forms p-adic and overconvergent modular forms Explicit computations Motivation for the overconvergent forms Why do we care about these obscure objects that one needs to know about structure sheaves and rigid geometry to even define? 1. The classical modular forms inject into the overconvergent forms; if we can prove results about the latter we can intuit stuff about the former. 2. We can consider many overconvergent weights at once, 3. This is in some way a more “natural” way to think about modular forms than the complex approach. L. J. P. Kilford (University of Bristol) A Gentle Introduction to Overconvergent Modular Forms Classical modular forms p-adic and overconvergent modular forms Explicit computations The different sorts of modular forms Let us consider the different types of modular forms by considering the example of the quasimodular form E2 . E2 E2∗ E2 E2 classical classical p-adic overconvergent no yes yes no In fact, E2 is transcendental over the ring of overconvergent modular forms. L. J. P. Kilford (University of Bristol) A Gentle Introduction to Overconvergent Modular Forms Classical modular forms p-adic and overconvergent modular forms Explicit computations The Arithmetic Approach Although modular forms are normally defined (at least to begin with) as complex functions, and the proof that the spaces are finite-dimensional relies on their complex definition, most number theorists care about them because they have an interesting arithmetic theory. “. . . the reason that modular forms enter into number theory at all, in some sense, is exactly because modular curves can be defined over schemes other than Spec C” — Wikipedia talk page for Modular form L. J. P. Kilford (University of Bristol) A Gentle Introduction to Overconvergent Modular Forms Classical modular forms p-adic and overconvergent modular forms Explicit computations Choice of basis The definition given can be refined to allow us to choose a specific basis for the overconvergent forms; for instance, when p = 2, if we define g := ∆(q 2 )/∆(q) then a suitable Banach basis is 6 1, 2 g , 212 g 2 , 218 g 3 , . . . . We can do this because there are results which guarantee that a wide range of choices of power of p will work here and give the same characteristic power series for the Hecke operator Up . L. J. P. Kilford (University of Bristol) A Gentle Introduction to Overconvergent Modular Forms Classical modular forms p-adic and overconvergent modular forms Explicit computations Using Magma Magma does not natively support overconvergent modular forms, but it can work with Fourier expansions, so one can investigate them using it. It does support classical modular forms well, which is a good reason to choose it. Sage could also be used for much of the work (and the newest version of Sage now does support overconvergent modular forms natively (this is work of Loeffler)). Although overconvergent modular forms are defined over the p-adic field, which is an inexact ring (and some instantiations of which can actually fail to be associative!), a lot of the calculations can in fact be taken to be over finite extensions of Q, which is much better for computing. L. J. P. Kilford (University of Bristol) A Gentle Introduction to Overconvergent Modular Forms Classical modular forms p-adic and overconvergent modular forms Explicit computations An explicit example We say that a classical modular form f of weight k has character χ if it satisfies az + b a b k f = χ(d) · (cz + d) f (z) for ∈ Γ0 (4). c d cz + d Let τ : (Z/4Z)× → C× be the nontrivial Dirichlet character. The slopes of the Hecke operator U2 acting on modular cusp forms of character τ and odd weight are Weight 5 7 9 11 .. . L. J. P. Kilford (University of Bristol) Slopes 2 2,4 2,4,6 2,4,6,8 .. . A Gentle Introduction to Overconvergent Modular Forms Classical modular forms p-adic and overconvergent modular forms Explicit computations The theorem We can in fact prove: Theorem Let k be an odd integer. The slopes of the U2 operator acting on overconvergent modular forms of weight k and conductor τ are {2i}i∈N . This can be proved by considering one weight first (weight 1) and then using a result of Coleman to move to all other odd weights. Similar results hold for other characters of conductor 2n (for an integer n ≥ 2); this is joint work with Buzzard. This result also tells us about the field of definition of the basis of eigenforms. L. J. P. Kilford (University of Bristol) A Gentle Introduction to Overconvergent Modular Forms Classical modular forms p-adic and overconvergent modular forms Explicit computations Extensions to other primes Similar results have been proved for other (“small”) primes: Prime 2 3 5 7 K., Buzzard-K. Roe, Jacobs K. K.-McMurdy (in progress) L. J. P. Kilford (University of Bristol) A Gentle Introduction to Overconvergent Modular Forms Classical modular forms p-adic and overconvergent modular forms Explicit computations More complicated cases When the level is 1, the slopes can be more involved. Here are the slopes of T2 acting on weight k classical forms: Weight 12 14 16 18 20 22 24 26 28 30 Slopes 3 3 4 3 5 3,7 4 3,8 6,6 Weight 32 34 36 38 40 42 44 46 48 50 Slopes 3,7 4,8 3,9 5,8 3,7,11 4,9,12 3,8,11 6,6,13 3,7,12,15 4, 8, 13 (Emerton, Clay, Smithline, and Buzzard-Calegari, Loeffler. . . ) L. J. P. Kilford (University of Bristol) A Gentle Introduction to Overconvergent Modular Forms