Dirichlet series whose partial sums of coefficients have regular variation (Q1769181)

Let \(p(n)\) be a nonnegative valued arithmetic function and define another function \(a(n)\) by the formula \[ \sum _{n=1}^\infty a(n)n^{-s} = \prod _{j=2}^\infty (1-j^{-s})^{-p(j)} , \] so if \(p(n)\) is the number of ``primes'' of size \(n\), then \(a(n)\) is the number of ``integers'' of size \(n\). Let \(A(x)=\sum _{n\leq x} a(n)\) and \(P(x)=\sum _{n\leq x} p(n)\) be the corresponding counting functions. The main results of the paper are direct theorems of the following kind. If for some \(0<t<1\) we have \(P(tx)/P(x) \to 1\), then \(A(tx)/A(x) \to 1\). If with some \(\alpha >0\) we have \(P(tx)/P(x) \to t^\alpha \) and \( \liminf ( \log x ) P(x)x^{-\alpha } > \alpha \), then \(A(tx)/A(x) \to t^\alpha \) and \(A(x) \sim x^{\alpha } A_0(x)\) with a nondecreasing \(A_0\), \(A_0(x)\to \infty \). (Theorem 1.) Theorem 1 ends with a sentence about the case when \( \limsup ( \log x ) P(x)x^{-\alpha } < \alpha \). This sentence is confused and the reviewer did not find any proof about this case; an example is given to show that the asymptotic given for \(A(x)\) may not hold in this case. The results are applied to prove first-order limit laws in the sense of \textit{S. N. Burris} [Number theoretic densit an logical limit laws. Mathematical Surveys and Monographs. 86. Providence, RI: American Mathematical Society (AMS) (2001; Zbl 0995.11001)]. Conjecture 10.7 of this book is established.