On the central value of symmetric square \(L\)-functions (Q957901)

This paper presents results regarding the first and second moments of the symmetric square and Rankin-Selberg square \(L\) function, when averaged over a basis \(\mathcal{B}_k(N, \chi),\) for the space \(S_k(N, \chi)\) of holomorphic cuspforms of weight \(k\geq 2,\) level \(N,\) and character (nebentypus) \(\chi,\) subject to the following constraints: (1) the level, \(N,\) is squarefree throughout; in the treatment of the second moment, it is assumed to be prime. (2) the character, \(\chi,\) is assumed to be real and primitive modulo \(N.\) Condition (2) ensures that \(S_k(N, \chi)\) has no oldforms. The basis \(\mathcal{B}_k(N, \chi)\) consists of normalized Hecke newforms. For \(f\in \mathcal{B}_k(N, \chi),\) with Hecke eigenvalues \(\lambda_f(n).\) The symmetric square \(L\) function of \(f\) is given by \[ L(s, \text{sym}^2 f) = \zeta^{(N)}(2s) \sum_{m |N^{\infty}} \tau(m) \sum_{(n, N)=1} \frac{\lambda_f(( mn)^2)}{(mn)^s}, \] where \[ \tau(m) := \sum_{d\mid n} 1, \qquad \zeta^{(N)}(s) = \prod_{p\nmid N} (1-p^{-s})^{-1}, \] and \(m \mid N^\infty\) means that \(m\mid N^r\) for some \(r.\) The Rankin-Selberg square \(L\) function is given by \(L(s, f\otimes f)=L(s, \chi) L(s, \text{sym}^2 f),\) where \(\chi\) is the character of \(f.\) The result concerning the first moment states that for \((\ell, N) = 1,\) and any \(\varepsilon >0,\) \[ \begin{multlined} \frac{\Gamma(k-1)}{(4\pi)^{k-1}} \sum_{f \in \mathcal{B}_k(N, \chi)} \frac{1}{\|f\|^2} \lambda_f( \ell^2) L \left( \frac 12, \text{sym}^2 f\right)\\ = \frac1{\sqrt{\ell}} \left( 2 \gamma + \frac{L_\infty'( \frac 12, \text{sym}^2 f)}{L_\infty(\frac 12,\text{sym}^2 f)}+ \log \frac{\sqrt{N}}\ell + \sum_{p \mid N} \frac{2\log p}{p-1} \right) + O_{k,\varepsilon}\left( ( \ell N)^\varepsilon \left( \frac{\ell}{N^{\frac 34}} + \frac1{N^{\frac 14}} \right)\right), \end{multlined} \] as \(N \to \infty.\) Here, \(\gamma= \lim_{s\to 1} \zeta(s) - \frac 1{s-1}\) is Euler's \(\gamma\) constant, and \(L_\infty(s, \text{sym}^2 f)\) is the gamma factor of \(L(s, \text{sym}^2 f),\) given by \[ L_\infty(s, \text{sym}^2 f) = 2^{-s} \pi^{-\frac 32s} \Gamma( s+k-1) \Gamma\left( \frac{s+ 1-\nu}2\right) \qquad \nu = \begin{cases} 1,&k \text{ odd}, \\ 0, & k \text{ even.} \end{cases} \] A similar asymptotic formula is given for the second moment \[ \frac{\Gamma(k-1)}{(4\pi)^{k-1}} \sum_{f \in \mathcal{B}_k(N, \chi)} \frac{1}{\|f\|^2} \lambda_f(\ell^2) \left| L\left( \frac 12, \text{sym}^2 f\right)\right|^2, \] under the additional hypothesis that \(N\) is prime. This new condition is for the sake of simplicity and not intrinsic to the method. It is noted that ``similar formulae hold for any point \(s\) on the critical line and the error term depends polynomially on \(\text{Im}\,s\).'' As an application, it is shown that there is a constant \(C\) depending on \(k\) such that \[ \frac{\# \left\{ f \in \mathcal{B}_k(N, \chi)\mid L(1/2, \text{sym}^2 f) \neq 0\right\} } {\# \mathcal{B}_k(N, \chi)} \geq C, \] for all sufficiently large primes \(N.\) Multiplying by \(L(1/2, \chi)\) for the first moment or \(L(1/2, \chi)^2\) for the second, yields analogous results for Rankin-Selberg square \(L\) functions.