Hilbert irreducibility, the Malle conjecture and the Grunwald problem (Q6073550)

This article has a nice 7-page introduction that I will try to summarize here. It deals with two interesting topics in Inverse Galois Theory. The first topic is the Malle Conjecture, which establishes an estimate of the number \(N(K,G,y)\) of different Galois extensions \(L/K\) of a fixed number field \(K\) with fixed Galois group \(G\) and bounded discriminant \(N_{K/\mathbb{Q}}(d_{L/K})\leq y\). More precisely, it states that there exists a constant \(c_1\) and, for every \(\varepsilon >0\), there exist constants \(c_2,y_0\) such that \[ c_1y^{a(G)}\leq N(K,G,y)< c_2y^{a(G)+\varepsilon},\quad\text{for all }y\geq y_0, \] where \(a(G):=(|G|(1-1/\ell))^{-1}\) and \(\ell\) is the smallest prime divisor of \(|G|\). The second topic is the Grunwald Problem, which concerns the realization of finitely many fixed local Galois extensions \(L^{\mathfrak{p}}/K_{\mathfrak{p}}\), with Galois group embedding into a fixed group \(G\), by a single global \(G\)-extension \(L/K\). Here \(K_{\mathfrak{p}}\) denotes of course the completion of \(K\) at the prime \(\mathfrak{p}\). Both topics are related in the sense that a control on the amount of solutions to a given Grunwald Problem immediately gives lower bounds on the number \(N(K,G,y)\) from the Malle Conjecture. The main result of this article (Theorem AB) is a generalization to an arbitrary number field \(K\) of a result by \textit{P. Dèbes} [Isr. J. Math. 218, 101--131 (2017; Zbl 1425.11179)], which deals with the case \(K=\mathbb{Q}\). They both count \(G\)-extensions obtained by \textit{specialization} from a regular \(G\)-extension \(F/K(T)\), giving lower bounds for such a number. More precisely, given a regular \(G\)-extension \(F/K(T)\), the result establishes a lower bound of the form \(y^{(1-1/|G|)/\delta}\) for the number \(N(F/K(T),y,\mathcal{F}_y)\) of \(G\)-extensions obtained from \(F/K(T)\), where \(y\) is big enough and bounds the discriminant as before, and \(\mathcal{F}_y\) denotes fixed local conditions on a finite set of primes depending on \(y\) and \(\delta\). This immediately yields a ``Malle type'' lower bound for \(G\) whenever it is a regular Galois group over \(K\) (Theorem A). Under the same hypothesis, it also yields a quantitative solution to some Grunwald problems (Theorem B), improving results from \textit{P. Dèbes} and \textit{N. Ghazi} [Ann. Inst. Fourier 62, No. 3, 989--1013 (2012; Zbl 1255.14022)], where they proved the existence of solutions without regard to their quantity. \textit{P. Dèbes}' result in [Isr. J. Math. 218, 101--131 (2017; Zbl 1425.11179)] used an explicit version of Hilbert's Irreducibility Theorem due to \textit{Y. Walkowiak} [Acta Arith. 116, No. 4, 343--362 (2005; Zbl 1071.12002)]. Thus, a generalization as the one given here needs a generalization of Walkowiak's result. And this is the other main result of this article (Theorem C). Given an irreducible polynomial \(P(X_1,X_2)\in\mathcal O_K[X_1,X_2]\), monic on \(X_2\), it bounds by above the number \(N(P,B)\) of roots \((x_1,x_2)\in \mathcal O_K^2\) of \(P\) that have bounded height \(H(x_i)\leq B\) for some suitably defined height function \(H\). Along with some other tools, this allows to bound by below the number of specializations on one variable that give an irreducible polynomial on the other variable, which yields the intended application. The proof of such a statement relies on yet another generalization (Theorem 3.4) to arbitrary number fields of a result by \textit{D. R. Heath-Brown} [Ann. Math. (2) 155, No. 2, 553--598 (2002; Zbl 1039.11044)].