Subgroups of profinite surface groups. (Q2428806)

The authors of the present article study closed subgroups of the étale fundamental group \(\Pi=\pi_1(X)\), where \(X\) is a smooth projective curve over an algebraically closed field of characteristic~\(0\) that has genus \(g\geq 2\). Their main result for the profinite surface group \(\Pi\) is an analog of the diamond theorem which, in its original form, applies to free profinite groups. We recall that the Nielsen-Schreier theorem for abstract free groups does not have a straightforward analog in the profinite setting. However, the diamond theorem yields a criterion for deciding that a closed subgroup of a free profinite group is again free in the category of profinite groups; [see \textit{L.\ Bary-Soroker}, Math. Ann. 336, No. 4, 949-961 (2006; Zbl 1111.20029)]. Let \(\Pi\) be a profinite surface group of genus \(g\geq 2\), i.e.\ the profinite completion of an abstract surface group with presentation \[ \langle x_1,\dots,x_g,y_1,\dots,y_g\mid\prod_{i=1}^g[x_i,y_i]\rangle, \] and consider a closed subgroup \(N\) of \(\Pi\). A necessary condition for the profinite group \(N\) to be free is that it is projective or, equivalently, that \(|\Pi:N|=\prod_pp^\infty\) as supernatural numbers, where \(p\) runs over all primes. Moreover, once we assume that \(N\) is projective, it is known that \(N\) is free if and only if \(N\) is semi-free, i.e.\ if every finite split embedding problem for \(N\) is properly solvable. The latter means that for every pair of epimorphisms \(f\colon N\to B\) and \(\alpha\colon A\to B\), where \(A,B\) are finite groups and \(\alpha\) splits, there exists an epimorphism \(\psi\colon N\to A\) such that \(f=\alpha\circ\psi\). Let \(N\) be a closed subgroup of \(\Pi\) as above and suppose, in addition, that the index \(|\Pi:N|\) is infinite. Then the main theorem of the article under review states: if there exist closed normal subgroups \(K_1,K_2\) of \(\Pi\) whose intersection is contained in \(N\) but which themselves are not contained in \(N\), then \(N\) is semi-free of countable rank. A direct consequence of this result is that, if \(N\) is normal in \(\Pi\), then every proper open subgroup of \(N\) is semi-free. The proofs involve a number of different techniques, for instance, a method described as Haran-Shapiro induction which links a given embedding problem \(\mathcal E\) for \(N\) to an embedding problem \(\mathcal E_{\text{ind}}\) for \(\Pi\). Substantial use is also made of the notion of sparse and abundant subgroups of~\(\Pi\). A closed subgroup \(M\leq\Pi\) of infinite index is called sparse if for every \(n\in\mathbb N\) there exists an open subgroup \(K\leq\Pi\) with \(M\subseteq K\) such that for every proper open subgroup \(L\leq K\) with \(M\subseteq L\) one has \(|K:L|\geq n\). It is shown that every sparse subgroup of \(\Pi\) is semi-free of countable rank. This covers one of several cases arising in the proof of the main theorem. An interesting problem highlighted in one of the remarks is the following: if \(N\) is normal in \(\Pi\) such that \(\Pi/N\) is hereditarily just infinite, does this imply that \(N\) is semi-free?