On prime numbers of special kind on short intervals (Q2473699): Difference between revisions

Let \[ \psi_1(x)=\sum_{\substack{ n\leq x\\ \{(1/2)n^{1/c}\}<1/2}} \Lambda(n),\quad 1<c\leq 2. \] Under the Riemann Hypothesis, the author shows that the short-interval estimate \[ \psi_1(N+H)-\psi_1(N)={{H}\over {2}}\biggl(1+O\biggl({{1}\over {N^{\varepsilon}}}\biggr)\biggr)\quad \text{holds~for }~H>N^{1/2+10\varepsilon}. \] For the proof, following an idea of \textit{Yu. V. Linnik} [Dokl. Akad. Nauk. SSSR 47, 7--9 (1945; Zbl 0063.03586)], the author notes that under the Riemann Hypothesis, it suffices to show that for the expression \[ Q(N,H)=\sum_{{{1}\over {2}}N^{1/c}<x\leq {{1}\over {2}} (N+H)^{1/c}} \bigl(\psi((2x+2)^c)-2\psi((2x+1)^c)) +\psi((2x)^c), \] the estimate \(Q(N,H)=O(H/N^{\varepsilon})\) holds in the given range of \(H\) versus \(N\). Here, \(\psi(x)\) is the Chebyshev function. In turn, this last estimate is proved by using the representation of the Chebyshev function as a sum over the nontrivial zeros of the Riemann zeta-function. This last sum is estimated by first splitting it into various subsums, then by using Abel's transformation to bound such sums in terms of certain integrals, and finally by using known bounds on such integrals in terms of the first and second derivatives of the integrand.