On the number of positive integers \(\leq x\) and free of prime factors \(>y\) (Q1064343): Difference between revisions

Let \(\Psi(x,y)\) denote the number of positive integers \(\leq x\), all whose prime factors are \(\leq y\), and let \(\rho(u)\) be the Dickman function, defined by the conditions \(\rho(u)=1\) \((0\leq u\leq 1)\), \(\rho'(u)=-u\rho(u-1)\) \((u>1)\), \(\rho(u)\) continuous at \(u=1\). The main result of this paper is that the estimate \[ \Psi(y^ u,y)=y^ u \rho(u)(1+O_{\varepsilon}(\log (u+1)/\log y)), \] which has been originally established by \textit{N. G. de Bruijn} [Ned. Akad. Wet. Proc. Ser. A 54, 50--60 (1951; Zbl 0042.04204)] for the range \(1\leq u\leq (\log y)^{3/5-\varepsilon}\), remains valid in the range \[ 1\leq u\leq \exp ((\log y)^{3/5-\varepsilon}).\tag{*} \] The proof proceeds by a careful induction argument, using the identity \[ \Psi (x,y) \log x=\int^{x}_{1}(\Psi (t,y)/t)\,dt+\sum_{p^ m\leq x, p\leq y}\Psi (xp^{-m},y) \log p\quad (x,y\geq 1) \] and the trivial estimate \(\Psi (x,y)=x+O(1)\) (1\(\leq x\leq y)\) to initiate the induction process. The same method is employed to prove two further results on \(\Psi(x,y)\), namely a short interval estimate valid in the range (*) and a lower bound valid uniformly in \(x\geq y\geq 1\).