Three conjectures on \(P^+(n)\) and \(P^+(n+1)\) hold under the Elliott-Halberstam conjecture for friable integers (Q2660304)

Let \(P^+(n)\) denote the largest prime factor of an integer \(n\). An old question of Erdős and Turan asks if the set \[ E_1=\{\,n\ge1 : P^+(n)<P^+(n+1)\,\} \] has natural density \(\mathbf{d}(E_1)=1/2\)? Restricting considerations to friable (or smooth) numbers \(n\) gives rise to aligned problems. Here we need to recall the asymptotic formula for the counting function of \(y\)-friable numbers \(n\le x\), \(\Psi(x,y)\), namely \[ \Psi(x,y)\sim\rho(u)x, \] as \(x\to\infty\), where \(\rho\) is the Dickman function and \(u=\log x/\log y\). Thus \textit{P. Erdős} and \textit{C. Pomerance} [Aequationes Math. 17, 311--321 (1978; Zbl 0379.10027)] conjectured that the set \[ E_2=\{\,n\ge1 : P^+(n)\le n^s\text{ and }P^+(n+1)\le (n+1)^t\,\}, \] where \(0<s,t<1\) are fixed, also has natural density and it is given by \(\mathbf{d}(E_2)=\rho(1/s)\rho(1/t)\). The paper under review shows that both of these conjectures (and another closely related but more technical conjecture) follow from the Elliott-Halberstam conjecture for friable integers. More precisely, the author uses the following weak form of the Elliott-Halberstam conjecture. The inequality \[ \sum_{q\le x^{1-\varepsilon}}\max_{(a,q)=1} \bigg|\sum_{\substack{n\le x,\, P^+(n)\le y\\ n\equiv a\pmod q}} 1 -\frac1{\varphi(q)} \sum_{\substack{n\le x,\, P^+(n)\le y\\ (n,q)=1}} 1 \bigg| \ll_{\varepsilon, A}\frac x{(\log x)^A} \] is valid for every \(\varepsilon>0\) and \(A>0\). Finally, a review of a paper on this topic would be incomplete without mentioning the recent advances of \textit{J. Teräväinen} [Forum Math. Sigma 6, Paper No. e10, 41 p. (2018; Zbl 1469.11385)] and \textit{T. Tao} and \textit{J. Teräväinen} [Algebra Number Theory 13, No. 9, 2103--2150 (2019; Zbl 1476.11127)]. In particular, we now know unconditionally that the sets \(E_1\) and \(E_2\) certainly have (correct) logarithmic densities.