On divisors of Lucas and Lehmer numbers (Q392795): Difference between revisions

For non-zero \(m\in\mathbb{Z}\) let \(P(m)\) and \(\omega(m)\) denote the greatest prime divisor of \(m\) and the number of distinct prime divisors of \(m\), respectively. Further, let \(\alpha,\beta\in\mathbb{C}^\times\) be such that \((\alpha+\beta)^2,\alpha\beta\in\mathbb{Z}\) and \(\alpha/\beta\) not a root of unity; then define \(\Phi_n(\alpha,\beta):=\beta^{\varphi(n)}\Phi_n(\alpha/\beta)\in\mathbb{Z}\setminus\{0\}\) for any integer \(n>2\), where \(\Phi_n(X)\) denotes the \(n\)-th cyclotomic polynomial and \(\varphi(n)\) its degree. The main result of the present paper states that, under the previous conditions, there exists a \(C\in\mathbb{R}_+\), effectively computable in terms of \(\omega(\alpha\beta)\) and the discriminant of \(\mathbb{Q}(\alpha/\beta)\), such that \[ P(\Phi_n(\alpha,\beta))>n\exp((\log n)/(104\log\log n))\quad\text{for any }n>C. \tag{*} \] In particular, if \(a,b\in\mathbb{Z}\) with \(a>b>0\), then \(P(a^n-b^n)\) exceeds the right-hand side of (*) if \(n\) is sufficiently large in terms of \(\omega(ab)\). This quantitatively settles the conjecture \(P(2^n-1)/n\to\infty\) as \(n\to\infty\) of \textit{P. Erdős} [Lect. Modern Math. 3, 196--244 (1965; Zbl 0132.28402)]. Similar lower bounds for \(P(u_n),P(\tilde{u}_n)\) are given if \((u_n)_{n\ge 0}\) or \((\tilde{u}_n)_{n\ge 0}\) are a Lucas or a Lehmer sequence. These are defined by \(u_n:=(\alpha^n-\beta^n)/(\alpha-\beta)\), or by \(\tilde{u}_n:=u_n\) if \(n\) is odd and \(\tilde{u}_n:=(\alpha^n-\beta^n)/(\alpha^2-\beta^2)\) if \(n\) is even, respectively. The author's initial proof of Erdős' above conjecture was mainly based on a lower bound for linear forms in \(p\)-adic logarithms established by \textit{K. Yu} [Forum Math. 19, No. 2, 187--280 (2007; Zbl 1132.11038)]. In order to treat here also Lucas and Lehmer numbers, he had to apply a more refined estimate obtained again by \textit{K. Yu} [Acta Math. 211, No. 2, 315--382 (2013; Zbl 1362.11071)].