Coefficients of half-integral weight modular forms modulo \(\ell^{ j }\) (Q707440): Difference between revisions

For a half-integral weight cusp form \(F(z)\) with Fourier series \[ \sum_{n \geq 1} a(n)q^n, \] where the \(a(n)\) are integers, the authors say that the coefficients \(a(n)\) are \textit{well-distributed} modulo \(M\) if for every integer \(r\), \[ \#\{ 1 \leq n \leq X : a(n) \equiv r \pmod{M} \} \gg_{r,M} \begin{cases} \frac{\sqrt{X}}{\log X}, & {\text{if \(r \not \equiv 0 \pmod{M}\)}}, \\ X, & {\text{if \(r \equiv 0 \pmod{M}\)}}. \end{cases} \] The principal result of this paper is that if \(\ell \geq 5\) is a prime not dividing the level of \(F(z)\), if \(F(z) \not \equiv 0 \pmod{\ell}\), and if there are finitely many square-free integers \(n_1,n_2,\dots,n_t\) such that \[ F(z) \equiv \sum_{i=1}^t \sum_{m \geq 1} a(n_im^2)q^{n_im^2} \pmod{\ell}, \] then there are some restrictions on the weight \(\lambda + 1/2\) of \(F(z)\), in the form of upper bounds on the residue class of \(\lambda\) modulo \(\ell-1\). The connection between this result and forms whose coefficients are well-distributed is the fact (which the authors prove) that if the coefficients of \(F(z)\) are not well-distributed modulo \(\ell^j\), then \(F(z)\) has the special form above modulo \(\ell\). The main application of the principal result is a proof that for every prime \(\ell \geq 5\), for every integer \(j \geq 1\), and for every integer \(r\), there are infinitely many natural numbers \(n\) such that the number of partitions of \(n\) is congruent to \(r\) modulo \(\ell^j\). In an addendum, the authors give an example of a cusp form contradicting one of their conjectures as well as a conjecture of Balog, Darmon, and Ono.