An analogue of Artin's conjecture for Abelian extensions (Q788048)

For each prime \(q\) let \(L_ q\) be an algebraic number field, finite and normal over \({\mathbb Q}\); and for square-free \(k\) let \(L_ k\) be the compositum of \(L_ q\) for \(q| k\). If \(n(k)=[L_ k:{\mathbb Q}]\) then \textit{L. J. Goldstein} [Analytic number theory, Proc. Symp. Pure Math. 24, 103--110 (1973; Zbl 0289.12013)] has conjectured that the set of rational primes which do not split completely in any \(L_ q\) has density \(\sum \mu(k)/n(k)\). Artin's conjecture on primitive roots is the case \(L_ q={\mathbb Q}(1^{1/q},a^{1/q}).\) Goldstein's conjecture is false in general, as \textit{P. J. Weinberger} [Proc. Am. Math. Soc. 35, 49--52 (1972; Zbl 0259.12009)] has shown. However, for the case \(L_ q/{\mathbb Q}\) Abelian, Goldstein (loc. cit.) proved the conjecture on the Generalized Riemann Hypothesis (GRH -- i.e. for Dedekind Zeta-functions) assuming certain conditions on the rates of growth of the degree and discriminant of \(L_ k\). The present paper removes the condition of GRH, and also notes various errors in Goldstein's work, including the statement of his result. The proof uses estimates for \(\pi(x,L)\) -- the number of rational primes \(p\leq x\) which split completely in \(L\). One can use GRH to estimate \(\pi(x,L_ k)\), but, as the paper shows, it suffices to use weaker unconditional results for ''small'' \(k\), and the Brun-Titchmarsh theorem for ''large'' \(k\). Unfortunately, it seems to the reviewer that the paper uses the erroneous bound \(\pi(x,f,a)\ll x/f\), for \(f\geq x\), implicitly, in the very last paragraph.