The distribution of r-free numbers (Q1122620)

Let \(A_ r(x)\) denote the number of r-free integers not exceeding x. \textit{H. L. Montgomery} and \textit{R. C. Vaughan} [Recent progress in analytic number theory, Vol. 1, 247-256 (1981; Zbl 0462.10029)] obtained \(R_ r(x):=A_ r(x)-x/\zeta (r)\ll x^{1/(r+1)+\epsilon}\) for a fixed integer \(r\geq 2\) under the truth of the Riemann hypothesis. This followed easily from their estimate \[ R_ r(x)=-S_ r(x,N)+O(x^{1/2+\epsilon} N^{(1-r)/2})+O(N^{1/2+\epsilon}), \] \[ S_ r(x,N)=\sum_{n\leq N}\mu (n)\psi (xn^{-r}),\quad \psi (t)=t-[t]-1/2. \] The present authors undertake an extensive treatment of \(S_ r(x,N)\). They skilfully apply various techniques from the theory of exponential sums to estimate \(S_ r(x,N)\), necessitating altogether 17 lemmas. In this way they prove, assuming the Riemann hypothesis, \(R_ r(x)\ll x^{D(r)+\epsilon}\), where \[ D(r)=7/(8r+6)\quad if\quad 2\leq r\leq 5,\quad D(r)=67/514\quad if\quad r=6, \] \[ D(r)=11(r-4)/(12r^ 2-37r- 41)\quad if\quad 7\leq r\leq 12,\quad D(r)=23(r-1)/(24r^ 2+13r-37)\quad if\quad 13\leq r\leq 20. \] Estimates for \(r>20\) are also obtained, but the formulas are more complicated.