A two-variable Artin conjecture (Q5928011): Difference between revisions

Artin's conjecture considered those primes~\(p\) for which a specified number~\(a\) is a primitive root, and gives a value for the density of this set inside the set of all primes. In this form the conjecture was shown by \textit{C. Hooley} [J. Reine Angew.\ Math. 225, 209-220 (1967; Zbl 0221.10048)] to follow from a generalisation of the Riemann Hypothesis. The authors consider the related density \(\delta(a,b)\) of those~\(p\) for which a specified~\(b\) is in the multiplicative subgroup generated by \(a,\bmod p\), where they allow \(a\) and~\(b\) to be rational numbers whose numerators and denominators are not divisible by~\(p\). A preprint by one of the authors considered the case when \(a\) and~\(b\) satisfy a non-trivial multiplicative relation \(a^x b^y = 1\), where it can be unconditionally proved that the density \(\delta(a,b)\) exists. Accordingly the authors suppose that no such relation holds. It is then a case of a theorem of \textit{G.~Pólya} that there are infinitely many primes of the type considered, but this argument does not show that the proportion of such primes has positive density. They deduce an expression of the shape \(c_{a,b}C_0\) for the density \(\delta(a,b)\) from the Generalised Riemann Hypothesis. Here \(C_0\) is given by a certain infinite product over primes, but an explicit formula for~\(c_{a,b}\) would be rather complicated. However, in the case when the group \({\mathbb Q}^*/\langle-1,a,b\rangle\) is torsion-free (so that \(\pm a^x b^y\) is not an \(n\)th power in~\({\mathbb Q}^*\) when \(x\) and~\(y\) are not both divisible by~\(n\)) they give a one-line expression of \(c_{a,b}\) in terms of discriminants of the relevant quadratic fields. The case of their theorem in which \(a\) and~\(b\) are coprime integers and not perfect powers occurs in a paper of \textit{P.~J.~Stephens} [J. Number Theory 8, No.~3, 313-332 (1976; Zbl 0334.10018)], but the authors do not agree with Stephens's evaluation of the constant~\(c_{a,b}\). A corollary of the authors' theorems deduces similar results for the set of primes dividing some member of a sequence given by a second order linear recurrence \(x_{k+2}=r x_{k+1}-sx_k\), provided \(X^2-rX+s\) splits in \({\mathbb Q}[X]\) and the sequence does not satisfy a first order recurrence.