The asymptotic distribution of prime numbers on the average (Q4375222)

This paper, published after the death of Prof. S. B. Stechkin, brings an exhaustive study of power moments of the well-known functions \[ R(x) := \psi(x) - x = \sum_{p^\alpha\leq x}\log p - x,\quad R_1(x) := \int_0^x R(u) \text{ d}u \] from prime number theory. The key idea is that the moments of \(R(x)\) and \(R_1(x)\), at least under the Riemann Hypothesis (RH), can be viewed as almost periodic functions in the sense of Bohr or Besicovitch. Most of the 19 theorems given in the paper depend on \(\theta\), the supremum of the real parts of zeros of the Riemann zeta-function \(\zeta(s)\). Thus \(\theta = 1/2\) is the RH, but at the present state of knowledge even \(\theta = 1\) cannot be excluded. One of the main results of the paper, all of which cannot be stated in this review, (cf. Theorem 1) is that there exists an effective constant \(C > 0\) such that, for all \(x,b \geq 1\), \[ \int_1^x | R(u)| ^b \text{ d}u \leq (Cb)^{2b}x^{1+b\theta}. \] This result is of course trivial if \(\theta = 1\), but it is not if \(\theta < 1\), in which case it shows (like most of the other results), that for any \(b > 1\) the function \(| R(u)| ^b\) is ``estimated on the average as \(O(u^{b\theta})\)''. Theorem 2 says that \(R(x) \ll_\theta x^\theta\), while if \(\zeta(\theta + it) \not = 0\) for any real \(t\), then even \(R(x) = o(x^\theta)\) as \(x \to \infty\). Theorem 4 says that unconditionally there is a constant \(D > 0\) such that, for \(x > D\), \[ \int_x^{2x} | R(u)| \text{ d}u > {x^{3/2}\over 200}, \] and Theorem 6 says that \(D\) is effective under the RH. To prove such precise results the authors first approximate \(R(x)\) by \(-\sum{x^\rho\over \rho}\), where \(\rho\) runs over complex zeros of \(\zeta(s)\). After the evaluation of the power moments in question, the problem is reduced to the estimation of various (multiple) exponential sums over the zeros \(\rho\). Here, besides the results from the theory of almost periodic functions, the authors make use of sharp explicit estimates (see e.g. Lemma 3 and Lemma 4) for various functions connected with Fourier transforms of suitable test-functions. It is not stated whether one can obtain better results (numerical constants) by other choices of test functions, but presumably such an attempt would necessarily lead to much more cumbersome calculations.