A density of ramified primes (Q2065744)

\textit{J. B. Friedlander} et al. [Invent. Math. 193, No. 3, 697--749 (2013; Zbl 1296.11150)] considered totally real cyclic extensions \(K/\mathbb Q\) in which every totally positive integer is a square and showed that if \(\sigma\) denotes a generator of \(\operatorname{Gal}(K/\mathbb Q)\), then the set of principal prime ideals \(\pi \mathbb Z_K\) which split in \(K(\sqrt{\sigma(\pi)})\) is of density \(1/2\). A certain generalization has been later provided by \textit{P. Koymans} and \textit{D. Milovic} [Duke Math. J. 170, No. 8, 1723--1755 (2021; Zbl 1478.11109)]. The authors deal with totally real Galois extensions \(K/\mathbb Q\) of odd degree \(n\) with odd narrow class number in which \(2\) is inert. Let \(p\) be a prime splitting in \(K\) and for prime ideals \(\mathfrak p\) lying over \(p\) denote by \(R_\mathfrak p^+\) the maximal abelian extension of \(K\) unramified at all prime ideals \(\ne \mathfrak p\). Let \(K_\mathfrak p\) be the unique quadratic extension of \(K\) lying in \(R_\mathfrak p^+\) and let \(K(p)\) be the composite of all fields \(K_\mathfrak p\). For \(\mu\in\{-1,1\}\) denote by \(S_\mu\) the set of all primes \(p\equiv\mu\) mod \(4\), splitting in \(K\) and by \(F_\mu\) the subset of \(S_\mu\) consisting of primes whose prime divisors in \(K(p)\) are of the first degree. Denote by \(F_\mu(x),S_{\mu}(x)\) the counting functions of the sets \(F_\mu\), resp. \(S_{\mu}\). The main result (Theorem 1) shows that if for all real characters \(\chi\ne \chi_0=1\) of conductors \(q\le Q\) and all integers \(M\) and \(N\le Q^\eta\) with \(\eta=2/n(n-1)\) one has for every \(\varepsilon>0\) \[ \left|\sum_{a=M+1}^N\chi(n)\right|\le cQ^{\eta(1-\delta)+\varepsilon} \] with some \(\delta>0\) and \(c=c(\varepsilon)\) (this is a variant of the conjecture stated by Friedlander et al. [loc. cit.]), then there is an explicit formula for the value of the limit \(\lim_{x\to\infty}F_\mu(x)/S_\mu(x)\), which is positive. In the case \(n=3\) this result is shown unconditionally in Theorem 2, the limit being equal to \(1/8\) for \(\mu=1\) and to \(3/8\) for \(\mu=-1\).