The distribution of square-full integers (Q5906539)

Numbers that are products of squares and cubes are said to be square- full, and their counting function has an asymptotic formula with an error term \(\Delta(x)\), which was shown to be \(o(x^{1/6})\) by \textit{P. T. Bateman} and \textit{E. Grosswald} [Ill. J. Math. 2, 88-98 (1958; Zbl 0079.071)]. They also pointed out that the exponent \({1\over 6}\) can be lowered if and only if the zero-free region for \(\zeta(s)\) can be properly enlarged. This prompted the study of \(\Delta(x)\) under the assumption of the Riemann hypothesis, and \textit{D. Suryanarayana} and \textit{R. Sitaramachandra Rao} [Ark. Mat. 11, 195-201 (1973; Zbl 0266.10037)] showed that, assuming the said hypothesis, \(\Delta(x)\ll x^{\theta+ \varepsilon}\), where \(\theta= 13/81\) and \(\varepsilon>0\). It was announced in a paper by \textit{R. Balasubramanian}, \textit{K. Ramachandra} and \textit{M. V. Subbarao} [Acta. Arith. 50, 107-118 (1988; Zbl 0652.10033)] that the first two authors had obtained \(\theta= 11/72\). Here the author shows that \(\theta= 5/33\) is admissible. The innovation is an application of D. R. Heath-Brown's method to tackle a ``type I'' exponential sum. Perhaps we should add that all the results concerning \(\theta\) here are subject to the Riemann hypothesis, since this point has been omitted in the abstract of the paper.