Integers, without large prime factors, in arithmetic progressions. I (Q1309360)

For \((a,q)=1\) let \(\Psi(x,y;a,q)\) denote the number of integers \(n\leq x\) in the congruence class \(n\equiv a\pmod q\) such that \(p\leq y\) for every prime factor \(p\) of \(n\). Let \(\Psi_ q(x,y)\) denote the number of such integers with \((n,q)=1\) rather than \(n\equiv a\pmod q\). One can estimate \(\Psi_ q(x,y)\) using the Dickman function, via the methods of \textit{A. Hildebrand} and \textit{G. Tenenbaum} [Trans. Am. Math. Soc. 296, 265-290 (1986; Zbl 0601.10028)]. One expects that \[ \Psi(x,y;a,q) \sim \Psi_ q(x,y)/ \varphi(q) \qquad (x\to\infty) \] as soon as the primes up to \(y\) generate the full multiplicative group modulo \(q\). \textit{E. Fouvry} and \textit{G. Tenenbaum} [Proc. Lond. Math. Soc., III. Ser. 63, 449-494 (1991; Zbl 0745.11042)] have shown that \[ \Psi(x,y;a,q)= {{\Psi_ q(x,y)} \over {\varphi(q)}} \{1+O(\exp(-c \log y))\} \] uniformly for \(\exp(c(\log\log x)^ 2)\leq y\leq x\), \(q\leq\log^ A x\). This is now replaced by \[ \Psi(x,y;a,q)= {{\Psi_ q(x,y)} \over {\varphi(q)}} \left\{ 1+O \left( {{\log q} \over {\log y}} \right)\right\} \] uniformly for \(y\leq x\), \(q\leq\min(x,y^ A)\). Any extension of this to give an asymptotic result for \(y=q^ N\) with \(N\to\infty\) would of course answer Vinogradov's problem on least quadratic non-residues. Thus the present result is, in one sense, as good as could be hoped for. A second result due to Fouvry and Tenenbaum, of `Bombieri-Vinogradov type' is also improved, and upper and lower bounds for \(\Psi(x,y;a,q)\) are considered.