On the minimum of the absolute value of the sum for a Dirichlet series with bounded sequence of exponents (Q1905318): Difference between revisions

Consider a Dirichlet series \[ F(z)= \sum^\infty_{n= 0} a_n \exp(z \lambda_n),\tag{1} \] where \(0< \lambda_n\uparrow+ \infty\). When (1) converges absolutely in \(\mathbb{C}\), the author [Dokl. Akad. Nauk Ukr., Ser. A 1984, No. 11, 22-24 (1984; Zbl 0564.30003)] obtains a necessary and sufficient condition for \[ F(x+ iy)= (1+ o(1))a_{\nu(x)}\exp(x+ iy)\lambda_{\nu(x)}\tag{2} \] uniformly in \(y\in \mathbb{R}\) as \(x\to +\infty\) on a set of finite measure, where \[ \nu(x)= \max\{n\mid (a_n)\exp(x \lambda_n)= \max\{(a_k)\exp(x\lambda_k)|k\in \mathbb{N}\}\}. \] He proves in this paper that when (1) converges absolutely only in \(\text{Re }z< 0\), (2) holds as \(x\to- 0\) outside a set of finite logarithmic measure if \[ \sum^\infty_0 (1/(\mu_{n+ 1} - \mu_n))<+ \infty,\tag{3} \] where \(\{\mu_n\}\) is the sequence \(\{\ln(a_n)\}\) arranged in ascending order. If the increasing sequence \(\{\mu_n\}= \{\ln(a_n)\}\), the condition (3) is also necessary. Under the condition of a bounded sequence of exponents \[ 0\leq \lambda_n< \sup\{\lambda_j\mid j\geq 0\}= \Delta\qquad (n\geq 0), \] when (1) converges absolutely in \(\mathbb{C}\), there are analogous results.