The invariants of the Weil representation of \(\mathrm{SL}_2(\mathbb{Z})\) (Q6950173)

A discriminant form, namely a finite abelian \(D\) with a non-degenerate quadratic form \(q:D\to\mathbb{Q}/\mathbb{Z}\), gives rise to a Weil representation \(\rho_{D}\) of the double cover of \(\operatorname{SL}_{2}(\mathbb{Z})\) on the space \(\mathbb{C}[D]\). This representation shows up in numerous technical constructions in the theory of modular forms, theta functions, and theta lifts, as when \(D\) is the discriminant form of an even lattice \(L\), the natural setting for the theta function of \(L\) is the vector-valued one, where the function is of level 1 and representation \(\rho_{L}\).\N\NScalar-valued modular forms of level 1 appear as components of vector-valued ones, once the vector that they multiply is invariant under \(\operatorname{SL}_{2}(\mathbb{Z})\). This is one of the motivations for investigating the subspace of \(\mathbb{C}[D]\) on which \(\operatorname{SL}_{2}(\mathbb{Z})\) acts trivially. The first main result about these invariants was first proved in the book appearing as reference number 15 in this paper using the language of codes, and later in reference number 3 in the current terminology, asserts that when \(D\) contains a self-dual isotropic subgroup, then the invariants are generated by the images of 1 under the arrow operators corresponding to the self-dual isotropic subgroups. When this happens, \(D\) has signature 0 and square size.\N\NThe current paper shows how to construct the invariants in general. A first, trivial observation is that when invariants exist, the signature of \(D\) is even, and the representation is indeed of \(\operatorname{SL}_{2}(\mathbb{Z})\), without the need to use the double cover. Moreover, one can decompose \(D\) according to its \(p\)-parts, yielding a decomposition of \(\mathbb{C}[D]\) as a tensor product, and thus it suffices to consider the case where \(D\) is a \(p\)-group.\N\NNow, for any discriminant form that is a \(p\)-group, there is a minimal sub-quotient, which is thus anisotropic and hence belongs to the a finite family. The family consists of the trivial discriminant form, the two discriminant forms \(p^{\pm1}\) as well as \(p^{\varepsilon2}\) for one choice of \(\varepsilon\) when \(p\) is odd, and when \(p=2\) there are the discriminants \(2^{\pm1}_{\pm2}\) and \(2^{-2}_{II}\) of size 4 as well as \(2^{\pm1}_{t}4^{\pm1}_{s}\) of size 8 (there are also a few anisotropic ones with odd signatures, but they are not related to invariants).\N\NTo describe the main result of this paper in a unified manner, let \(D_{p}^{\mathrm{an}}\) denote one of these anisotropic discriminant forms of even signature. Then the authors attach to it a unique discriminant form \(D_{p}^{\min}\), which yields \(D_{p}^{\mathrm{an}}\) as a sub-quotient, and such that the space of invariants inside the representation \(\mathbb{C}[D_{p}^{\min}]\) is 1-dimensional. The main result is that whenever \(D\) is a discriminant of \(p\)-power level whose minimal sub-quotient is \(D_{p}^{\mathrm{an}}\), the space of invariants is generated by the images, under all the possible arrow operators from \(D_{p}^{\min}\) to \(D\), of the 1-dimensional space of invariants inside \(D_{p}^{\min}\).\N\NAs an application of the main result, the authors work out a formula for the dimension of cusp forms of weight 2 and the Weil representation \(\rho_{p^{\varepsilon n}}\) for even \(n\), as well as for Jacobi forms of singular weights.\N\NTo make the main result explicit, we go over the list of possible discriminant forms for \(D_{p}^{\mathrm{an}}\), and indicate that the corresponding \(D_{p}^{\min}\) is trivial, \(p^{\pm3}\), \(2^{\pm1}_{\pm2}4^{+2}_{II}\), \(2^{-4}_{II}\), and \(2^{+1}_{1}4^{\varepsilon}_{r}8^{+2}_{II}\) for the corresponding \(r\) and \(\varepsilon\) respectively. We note that the images in the main theorem are usually not linearly independent, and even in the classical case where \(D_{p}^{\mathrm{an}}\) and \(D_{p}^{\min}\) there are trivial (this is the case arising from self-dual isotropic subgroups) linear relations between these images in general.\N\NThe method used in the paper concerns an averaging map \(\operatorname{inv}_{D}\), which serves as a projection from \(\mathbb{C}[D]\) onto its space of invariants. This can be done since \(\rho_{D}\) is known to factor through a finite quotient, and is thus a representation of a finite group. Moreover, when the level of \(D\) is a power of a prime \(p\), the images of the basis vectors under \(\operatorname{inv}_{D}\) can be made more explicit, which allows for a better understanding of what the invariant space may be, as well as produce some dimension formulae that can be used to verify when the full space of invariants is accounted for. As usual in the theory of Weil representations, the case where \(p=2\) is more delicate and involves more cases.\N\NThe paper is divided into 8 sections. Section 1 is the Introduction, with the statement of the results. Section 2 provides the basic definitions involving discriminant forms, and Section 3 contains the main properties of Weil representations and the maps \(\operatorname{inv}_{D}\). Section 4 produces specialized results about \(\operatorname{inv}_{D}\) in the prime power level case, and Section 5 considers the extra delicacies when the prime is 2. Section 6 introduces the arrow operators (also called induction in this paper), and Section 7 establishes the main result. Finally, Section 8 discusses the applications to dimensions of spaces of cusp forms and Jacobi forms.