Some remarks on the \(q\)-exponents of generalized modular functions (Q539131)

From the introduction: In this short note we discuss some properties of the \(q\)-exponents of certain generalized modular functions (GMFs) \(f\) on the Hecke congruence subgroup \(\Gamma_0(N)\) \((N\in\mathbb N)\). We suppose that the weight of \(f\) is zero and \(f = 0\). Recall that \(f\) is a holomorphic function on the complex upper half-plane \(\mathcal H\), meromorphic at the cusps, that transforms under \(\Gamma_0(N)\) like a usual modular function, with the sole exception that the character \(\chi\) need not be unitary. We suppose that \(\chi(\gamma) = 1\) for all parabolic \(\gamma \in\Gamma_0(N)\) of trace 2. Each GMF \(f\) has a product expansion \[ f(z) = q^h\prod_{n\geq 1} (1-q^n)^{c(n) }\quad (0 < |q| < \varepsilon), \] where \(h\in\mathbb Z\), the \(c(n)\) \((n\geq 1)\) are uniquely determined complex numbers and \(q =e^{2\pi iz}\) \((z\in\mathcal H)\). It was proved in [the first author and \textit{Y. Martin}, Ramanujan J. 15, No. 1, 103--107 (2008; Zbl 1141.11022)] that for each squarefree \(N\geq 11\) one can find a GMF \(f\) on \(\Gamma_0(N)\) such that \(f\) has no zeros on \(\mathcal H\) (in fact, \(\mathrm{div}(f ) = \emptyset\)) and the \(q\)-exponents \(c(n)\) \((n\geq 1)\) take infinitely many different values. This result and the proof given are actually easily seen to be valid for arbitrary \(N\geq 11\) and to hold for any non-constant \(f\), if one exploits the fact proved in [\textit{M. I. Knopp} and \textit{G. Mason}, J. Number Theory 99, No. 1, 1--28 (2003; Zbl 1074.11025)] that GMFs \(f\) with \(\mathrm{div}(f ) = \emptyset\) correspond to cusp forms of weight 2 by taking logarithmic derivatives. The assertion for any non-constant \(f\) also follows from [the first author, Funct. Approximatio, Comment. Math. 43, No. 1, 23--29 (2010; Zbl 1275.11076), cf. Corollary 1] where more generally it was shown that a GMF \(f\) on \(\Gamma_0(N)\) has empty divisor if and only if \(c(n)\ll_{f,\varepsilon} n{-1/2+\varepsilon}\) \((\varepsilon> 0)\). Our first result sharpens the above statement under certain conditions on \(f\). If \(\mathrm{div}(f ) = \emptyset\), \(f\) is not constant and has algebraic Fourier coefficients, then we prove that there exist already infinitely many primes \(p\) such that the \(c(p)\) take infinitely many different values, provided that certain ``norm functions'' of \(f\) and \(f | W_N\) (where \(W_N\) is the Fricke involution) are constant. The proof heavily depends on a positive proportion result regarding values not taken by the Fourier coefficients of cusp forms lying in the subspace of newforms, due to \textit{K. Ono} and \textit{C. Skinner} [Ann. Math. (2) 147, 453--470 (1998; Zbl 0907.11017)]. The latter in turn is a consequence of the existence of the \(\ell\)-adic Galois representation attached to Hecke eigenforms. Our second result shows that under the hypothesis \(\mathrm{div}(f ) = \emptyset\), the \(c(n)\) \((n\geq 1)\) change signs infinitely often, provided that they are real numbers. The proof uses Landau's theorem on Dirichlet series with non-negative coefficients, coupled with the fact that the abscissa of absolute convergence of the Hecke \(L\)-function of a nonzero cusp form of weight 2 is exactly \(\frac 32\).