Approximation of L-functions associated to Hecke cusp eigenforms (Q7019153)

Let \(\Gamma = \mathrm{SL}_2(\mathbb{Z})\), and let \(f\) be a modular form of weight \(k\) and level \(C\) for the group \(\Gamma\). To this \(f\) is associated a completed \(L\)-function defined by the Euler product \(\Lambda(s) = g(s)\prod_{p=2}^\infty L_p(s)\) in a right half-plane and extended to the rest of \(\mathbb{C}\) by analytic continuation, which is entire and satisfies \(\Lambda(s) = \pm \Lambda(k-s)\). Here, \(L_p(s)\) and \(g(s)\) are the local \(L\)-factors at each finite place and at infinity, respectively, and each finite place factor \(L_p(s)\) can be written as the reciprocal of a quadratic polynomial in \(p^{-s}\). To construct the approximation, we start with a truncation \(\Lambda^{\mathrm{Euler}}_N (s) = g(s)\prod^{p_N}_{p=2} L_p(s)\) of the Euler product. Then, we regularize the truncated Euler product by defining \(\Lambda^{\mathrm{ingoing}}_N (s)=\Lambda^{\mathrm{Euler}}_N (s)- \Lambda^{pp}_N (s)\), where \(\Lambda^{pp}_N (s)\) is defined as the sum of the principal parts of the Laurent expansion of \(\Lambda^{\mathrm{Euler}}_N (s)\) at each of its (infinitely many) poles. Along the way, it is shown that the convergence of the infinite sum defining \(\Lambda^{pp}_N\) is closely related to the equidistribution of a sequence involving ratios of logarithms of primes. The final step is to impose a functional equation on the approximation, by setting \(\Lambda_N (s)=\lambda^{\mathrm{ingoing}}_N (s) \pm \Lambda^{\mathrm{ingoing}}_N (k-s)\) according to the sign in the functional equation of \(\Lambda\). The resulting function \(\Lambda_N\) is entire, satisfies the same functional equation as \(\Lambda\), and contains the data of the first \(N\) Euler factors of the original \(L\)-function. The authors establish an analogous integral formula for \(\Lambda\) using just its functional equation and decay properties, and together the integral formulas lead to an error formula for the approximations. Let \(s_0\in \mathbb{C}\). For \(\sigma > \max(\Re(s_0), \Re(k-s_0), \frac{k+1}{2} )\), \N\[\N\Lambda(s_0)- \Lambda_N(s_0)=\int_{\Re(s)=\sigma} (\Lambda(s) - \Lambda^{\mathrm{Euler}}_N (s))\left(\frac{1}{s-s_0}\pm \frac{1}{s -k + s_0}\right) ds.\N\]\NThen, it is shown that for any \(s_0 \in \mathbb{C}\), \(\Lambda_N (s_0)\) converges to \(\Lambda(s_0)\) as \(N \to \infty\). Finally, the paper gives an interpretation of the contour integral formulas in terms of Mellin transforms and thus derives a convolution-type formula for the completed \(L\)-function as well as the approximations: For all \(s\in \mathbb{C}\), \N\[\N\Lambda_N (s) =\int_{1}^\infty(t^{s-1}\pm t^{k-1-s})f_N\left(\frac{it}{\sqrt{C}}\right)dt, \N\]\Nwhere \(f_N\) is a certain subseries of the Fourier expansion of \(f\).