Note on the mean value of the Erdős-Hooley delta-function (Q6936096)

The Erdős-Hooley \(\Delta\)-function, introduced in a problem in 1973 by \textit{P. Erdős} [Canad. Math. Bull. 16, 621--622 (1973)] and studied in 1979 by \textit{C. Hooley} [Proc. Lond. Math. Soc., III. Ser. 38, 115--151 (1979; Zbl 0394.10027)] is defined for integer \(n\geq 1\) by\N\[\N\Delta(n):=\max_{u\in\mathbb{R}}\left|\{d:d\mid n,\,e^u<d\leq e^{u+1}\}\right|.\N\]\NEstimating the average order of \(\Delta\) is a serious and useful problem in analytic number theory, and meets several improvements since Hooley, ibid. Let\N\[\NS(x):=\sum_{n\leq x}\Delta(n).\N\]\NRecently, \textit{D. Koukoulopoulos} et al. [Proc. Lond. Math. Soc. (3) 127, No. 6, 1865--1885 (2023; Zbl 1536.11142)] showed that \(S(x)\ll x(\log\log x)^{\alpha}\) with \(\alpha=11/4\). In the present paper, the authors improve on this approximation by proving it with \(\alpha=5/2\). Moreover, the authors give a short proof of the following known [\textit{R. R. Hall} and \textit{G. Tenenbaum}, Divisors, Cambridge University Press (1988; Zbl 0653.10001)] approximation\N\[\Nx\log x\ll\sum_{n\leq x}\Delta(n)^2\ll x(\log x)(\log\log x)^2.\N\]\NThe authors pose a challenging open question, to determine the true exponent of \(\log\log x\) in the above second moment approximation.