Sieving by large prime factors (Q749601): Difference between revisions

Let Q be a set of primes with the following property: There are constants \(\delta\), \(0\leq \delta <1\), and \(A>0\) such that \[ \sum_{p\leq x,\quad p\in Q}\log p=\delta x+O(x \log^{-A}x). \] By \(\psi\) (x,y;Q) we denote the number of all positive integers \(\leq x\) which do not have prime- divisors \(>y\) and \(\in Q\). \textit{D. A. Goldston} and \textit{K. S. McCurley} [J. Number Theory 28, 94-115 (1988; Zbl 0639.10028)] have shown that \[ \psi (x,y;Q)=\frac{e^{\gamma \delta}}{\Gamma (1- \delta)}x\prod_{y<p\leq x,\quad p\in Q}(1-\frac{1}{p})(1+o(1))\text{ for } u:=\frac{\log x}{\log y}\to \infty. \] We prove that for the unspecified quantity o(1) one may take \(1/u+R(x)\) where \(R(x)=(\log \log x)\log^{-1}x\) if \(A=1\) and \(R(x):=\log^{-B}x\), \(B:=\min \{A,1\}\), else.