An Omega theorem on the general asymmetric divisor problem (Q1363099)

Let \(E(x) = E(a_1,\ldots ,a_p;x)\) denote the error term in the asymptotic formula for the summatory function of the (general divisor) function \(d(a_1,\ldots ,a_p;n)\), where for fixed integers \(a = a_1 \leq \ldots \leq a_p\) \[ \sum_{n=1}^\infty d(a_1,\ldots ,a_p;n)n^{-s} = \zeta(a_1s)\ldots\zeta(a_ps) \qquad(\text{Re} s > 1/a). \] The author proves that \[ E(x) = \Omega_*\left\{x^\theta(\log x)^{a\theta}(\log\log x)^{p-1+ca\theta} \exp\left(-A\sqrt{\log\log\log x} \right)\right\},\leqno(1) \] where \(A > 0\) is a constant, \(\theta = {p-1\over2(a_1 + \ldots + a_p)},\quad c = \max\Big\{ p\log2-1, p\log(a+1)-a\Big\},\) and \(* = \pm\) for \(p \geq 4\), whereas \(* = +\) for \(p = 2,3\). As usual \(f(x) = \Omega_\pm(g(x))\) means that both \(f = \Omega_+(g)\) (meaning \(\limsup f/g > 0\)) and \(f = \Omega_-(g)\) (meaning \(\liminf f/g < 0\)) hold. The proof skillfully uses the methods of \textit{J. L. Hafner} [Invent. Math. 63, 181-186 (1981; Zbl 0458.10031)], who obtained (1) in the classical case of the Dirichlet divisor problem (\(p = 2, a_1 = a_2 = 1\)), and \textit{W. G. Nowak} [Abh. Math. Semin. Hamb. 65, 259-276 (1995; Zbl 0854.11048)], who obtained (1) without ``\(ca\theta\)'' in the exponent of \(\log\log x\) and without \(\exp(-A\ldots)\). In fact the present work may be considered as a continuation of Nowak's paper.