Local Galois theory in dimension two (Q2577009)

Let \(k\) be any field, \(K=k((x,t))\) the fraction field of the power series ring \(k[[x,t]]\). The main result of the paper says that the absolute Galois group \(G_K\) of such a field \(K\) is quasi-free of rank equal to the cardinality of \(K\). Thus every finite group is a Galois group over \(K=k((x,t))\). This result can be regarded as a higher dimensional version of Shafarevich's Conjecture saying that the absolute Galois group of the maximal cyclotomic extension of a global field is a free profinite group of countable rank. The paper contains five sections. The first one is the introduction to the subject. In Section 2 the authors define the notion of quasi-free profinite group. An embedding problem \(\mathcal{E}\) is a pair \((\alpha \colon \Pi \to G, f \colon \Gamma \to G)\) of epimorphisms of profinite groups. \(\mathcal{E}\) is finite if \(\Gamma\) is finite, it is non-trivial if \(\ker(f)\) is nontrivial, and it is split if \(f\) has a section. A weak solution to \(\mathcal{E}\) is a homomorphism \(\lambda \colon \Pi \to \Gamma\) such that \(f \lambda = \alpha\). A proper solution to \(\mathcal{E}\) is a weak solution in which \(\lambda\) is surjective. A profinite group \(\Pi\) is quasi-free of rank \(m\) if every non-trivial split embedding problem for \(\Pi\) has exactly \(m\) proper solutions. The first result proved by the authors asserts that for an infinite cardinal \(m\) a profinite group is free of rank \(m\) if and only if it is quasi-free of rank \(m\) and projective. Moreover, if \(m\) is countable, then \(\Pi\) is a free profinite group if and only if \(\Pi\) is projective and every finite split embedding problem for \(\Pi\) has a proper solution. In Section 3 the authors study the models of curves over complete discrete valuation rings in order to extend the results of \textit{F. Pop} [Ann. Math. 144, 1--34 (1996; Zbl 0862.12003)] and \textit{D. Haran} and \textit{M. Jarden} [Forum Math. 10, 329--351 (1998; Zbl 0903.12003)] on split embedding problem for curves over large fields. This is done in Section 4 which is the core of the paper. Namely, the authors prove: one can choose the solutions of split embedding problem over curve to be totally split over any finite set of points; there are many such solutions; and, in the case of a large field of the form \(k((t))\), the solution can be chosen to be totally ramified over a specified closed point on a model over \(k[[t]]\). This is applied in the final Section 5 to prove the main result of the paper: every non-trivial finite split embedding problem for \(K=k((x,t))\) has \(\mathrm{card}\, K\) proper solutions. The result follows from its geometric version stated in terms of covers which relies on the split embedding problems over curves considered in Section 4.