Mollification and non-vanishing of automorphic \(L\)-functions at the primary level (Q643534)

The special values of automorphic \(L\)-functions are of great interest, in particular at the central point \(s=1/2\). In the paper under review the author obtains a sharp lower bound for a proportion of primitive new forms which are non-zero at the critical point, where the level \(q\) is a high power of a fixed prime. The result generalizes previously known results for levels which are primes or square-free integers. Let \(S^*_k(q)\) be the space of newforms of weight \(k\) and level \(q\) with respect to \(\Gamma_0(q)\), and let \(H^*_k(q)\) be its orthogonal basis consisting of normalized Hecke eigenforms. Any \(f\in H^*_k(q)\) has a Fourier expansion at \(\infty\) in the form \(f(z)=\sum_{n\geq 1}\lambda_f(n)n^{(k-1)/2}e^{2\pi i nz}\) and a corresponding automorphic \(L\)-function \(L(s,f)=\sum_{n\geq 1}\lambda_f(n)n^{-s}\) for \(\text{Re}(s)>1\). More generally, the function \(\Lambda(s,f)=\sqrt{q}/(2\pi)\Gamma\left(s+(k-1)/2\right)L(s,f)\) has an analytic continuation on \(\mathbb{C}\) and satisfies a functional equation \(\Lambda(s,f)=\varepsilon_f\Lambda(1-s,f)\) with \(\varepsilon_f=\pm 1\). The main theorem of this paper states that for even integer \(k\geq 2\), a prime \(p\) and any \(\varepsilon>0\), there exists a constant \(\nu_0=\nu_0(k,p,\varepsilon)\) such that for \(\nu\geq \nu_0\) and \(q=p^{\nu}\) we have \[ \mathop{{\sum}^h}_{\substack{ f\in H^+_k(q)\\ L(1/2,f)\neq 0}} 1\geq \frac{p-1}{6p}-\varepsilon\,. \] Here \({\sum}^h\) denotes the harmonic sum \[ \mathop{{\sum}^h}_{f\in A} \alpha_f:=\frac{\Gamma(k-1)}{(4\pi)^{k-1}}\sum_{f\in A}\frac{\alpha_f}{\langle f,f\rangle} \] for any set \(A\) of newforms, \(\langle f,f\rangle\) is the Petersson norm and \(H^+_k(q)\) is the set of those eigenforms in \(H^*_k(q)\) with \(\varepsilon_f=1\). This theorem improves previous result of the author [Acta Arith. 147, No. 1, 1--32 (2011; Zbl 1219.11074)] giving, for the same level \(q=p^{\nu}\), a weaker non-sharp lower estimate for a similar harmonic sum. The proof uses a certain trace formula from the above-cited article and the approximate functional equation of Theorem 5.3 from [\textit{H. Iwaniec} and \textit{E. Kowalski}, Analytic number theory. Providence: AMS (2004; Zbl 1059.11001)]. These are used to compute asymptotically the weighted (mollified) first and second moment of \(L(1/2,f)\). After optimizing the mollifiers the bounds obtained for the moments are plugged in a certain Cauchy-Schwarz inequality. This idea is described for example in Chapter 26 of Iwaniec-Kowalski's book.