On the divisor function in short intervals (Q652227): Difference between revisions

Let \(d(n)\) denote the number of divisors function and, as usual, \[ \Delta(x) := \sum_{n\leqslant x}d(n) - x(\log x+2\gamma-1) \] denotes the error term in the Dirichlet divisor function, and \(\gamma = - \Gamma'(1)\) is the Euler constant. The author conjectures the following: Let \(h = x^\theta\) with \(\theta >0, Y = X^\alpha\), where \(\alpha\;(\theta < \alpha \leq 1)\) is given and let \(\varepsilon >0\) be arbitrarily small. Let \(k\geq 1\) be given. Then there exists \(X_0 >0\) such that \[ \int_X^{X+Y}\left(\Delta(x+h(x))-\Delta(x)\right)^kdx \ll (h(x))^{k/2}YX^\varepsilon\leqno(1) \] uniformly for \(X\geq X_0\). \textit{M. Jutila} [Ann. Univ. Turku., Ser. A I 186, 23--30 (1984; Zbl 0536.10032)] proved unconditional results involving (1) when \(k=2\) and conjectured (1) when \(k =4\) and \(\alpha =1\). The reviewer [Ramanujan J. 19, No. 2, 207--224 (2009; Zbl 1226.11086)] proved (1) for \(k =4, \alpha=1, \theta > 3/8\). Assuming the above conjecture (1), the author proves that \[ \sum_{x<n\leq x + h(x)}d(n) \sim h(x)\log x\qquad(x\to\infty)\leqno(2) \] for \(2\alpha/(k+2) < \theta < \alpha\), if \(h = x^\theta, k\geq 1, 0 < \alpha < 1\). At the end of the paper the author states how the above conjecture can be somewhat relaxed to still yield (2).