On common values of \(\varphi(n)\) and \(\sigma(m)\). II (Q1950990)

Let \({\mathcal V}_f\) denote the image of an arithmetic function \(f\) and \({\mathcal V}_f(x)={\mathcal V}_f\cap[1,x]\), \(V(x)=\#{\mathcal V}_f(x)\), where in this paper \(f\equiv\varphi\) or \(\sigma\) with \(\varphi\) denoting Euler's function and \(\sigma\) the sum of divisors function, and let \({\mathcal V}_{\varphi,\sigma}={\mathcal V}_\varphi\cap{\mathcal V}_\sigma\). Estimates for \(V_f(x)\) when \(f\in\{\varphi,\sigma\}\) have been obtained by several authors, including P. Erdős who in 1935 showed that \(V_f(x)={x\over(\log x)^{1+o(1)}}\). More recently, the first author [Ramanujan J. 2, No. 1--2, 67--151 (1998; Zbl 0914.11053)] established the precise order of magnitude of \(V_f(x)\). Erdős asked in 1959 whether \({\mathcal V}_{\varphi,\sigma}\) is infinite, and this was shown unconditionally to be so by the first author, \textit{F. Luca} and \textit{C. Pomerance} [Bull. Lond. Math. Soc. 42, No. 3, 478--488 (2010, Zbl 1205.11010)]. In [Acta Math. Hung. 133, No. 3, 251--271 (2011; Zbl 1265.11092)] the authors proved, under the assumption of a strong uniform version of Dickson's prime \(k\)-tuples conjecture, that as \(x\to\infty\), \[ V_{\varphi,\sigma}(x)= {x\over (\log x)^{1+o(1)}}. \] The aim of the paper under review is to compare the size of \(V_{\varphi,\sigma}(x)\) with \(V_\varphi(x)+ V_\sigma(x)\) by establishing that as \(x\to\infty\), \[ V_{\varphi,\sigma}(x)\leq {V_\varphi(x)+ V_\sigma(x)\over (\log\log x)^{{1\over 2}+o(1)}}, \] and the authors remark that they were unable to increase the exponent \({1\over 2}\). This result shows that almost all \(\varphi\)-values are not \(\sigma\)-values and vice versa. The proof is extremely complicated and utilizes the structure and lemmas in Ford's 1998 paper (loc.cit.) as well as other results from the literature, and it also depends on a fundamental sieve estimate. To aid the reader's understanding, at each stage of the proof the authors give an informal sketch of the ideas behind their argument. They achieve their result by concentrating on specific subsets \({\mathcal A}_f\) of pre-images of the \(f\)-values for \(f\in\{\varphi,\sigma\}\) and verifying that the remaining \(f\)-values with preimages not in \({\mathcal A}_f\) can be discarded.