Unified treatment of explicit and trace formulas via Poisson-Newton formula (Q2018319): Difference between revisions

Let \[ f(s) = 1+ \sum_{n \geq 1} a_n e^{-\lambda_n s} \] be a non-constant Dirichlet series. Here \(a_n \in \mathbb C\), and \[ 0<\lambda_1<\lambda_2< \cdots \to \infty. \] Assume that the abscissa of absolute convergence is finite, i.e., \( \sum_{n \geq 1} |a_n| e^{-\lambda_n \sigma}<\infty, \) for some \(\sigma< \infty\). Assume further that \(f\) is meromorphically continuable of finite order to the whole complex plane. Denote by a \(\rho\) zero or a pole of \(f\), and \(n_\rho\) the multiplicity of \(\rho\). Then the main result of this paper is the following Poisson-Newton formula associated with the Dirichlet series \(f\) \[ \sum_\rho n_\rho e^{\rho t} = \sum_{{\mathbf k}} \langle {\mathbf \lambda}, {\mathbf k} \rangle b_{{\mathbf k}} \delta_{\langle {\mathbf \lambda}, {\mathbf k} \rangle}, \] where the convergence of infinite series and the equality holds in the sense of distributions on \(\mathbb R_+\). Here \({\mathbf 0} \neq {\mathbf k} = (k_n)_{n \geq 1} \in \oplus_n \mathbb{N} \) runs over sequences of non-negative integers whose all but finitely many elements are zero, \(\langle {\mathbf \lambda}, {\mathbf k} \rangle = \sum \lambda_n k_n\), and \(b_{\mathbf k}\) are determined by the formula \(-\log f(s) = \sum_{\mathbf k} b_{\mathbf k} e^{-\langle {\mathbf \lambda}, {\mathbf k} \rangle}\). That formula is a generalization of the following classical Poisson formula in Fourier analysis, \[ \sum_{n \in \mathbb Z} e^{i \frac{2\pi}{\lambda} n t} = \lambda \sum_{k \in \mathbb Z} \delta_{\lambda k}, \] which is related to the Dirichlet series \(f(s) = 1 - e^{-\lambda s}\). Some consequences including explicit formulas for zeros of the Riemann zeta function and Selberg trace formula are investigated.