On a class of relatively prime sequences (Q1254272)

For each \(n\geq 1\) let \(a_0(1)=n\) and define \(a_{i+1}(n)> a_i(n)\) inductively as the least integer coprime to \(a_j(n)\) for \(0\leq j\leq i\). Let \(g(n)\) be the largest \(a_i(n)\) which is neither a prime \(n\) or the square of a prime. It is shown here that \(g(n)\thicksim n\) and that \(g(n)-n \gg m^{1/2}\log n\). The true order of magnitude of \(g(n)-n\) remains unsettled, and some relevant computations are discussed. Other results on the sequence \(a_i(n)\) are given, extending work of \textit{P.Erdős} [Math. Mag. 51, 238-240 (1978; Zbl 0391.10004)] . The following result occurs incidentally in one of the proofs: if \(n\) is large enough \([n/p]\) is composite for some prime \(p< n^{1/2}\).