The cuspidalisation of sections of arithmetic fundamental groups (Q436131)

This paper concerns Grothendieck's section conjecture in anabelian geometry: let \(X\) be a smooth, proper, geometrically connected algebraic curve of genus at least 2 defined over a field \(k\) of characteristic zero. Denote by \(G_k\) the absolute Galois group of \(k\), by \(\pi_1(X)\) the étale fundamental group of \(X\), and by \(\pi_1(\overline{X})\) the étale fundamental group of \(\overline{X}:=X\times_k\overline{k}\). These groups naturally sit in an exact sequence \[ 1\rightarrow \pi_1(\overline{X})\rightarrow \pi_1(X) \rightarrow G_k\rightarrow 1. \] Every \(k\)-rational point \(x\in X(k)\) defines a continuous group-theoretic section \(s_x:G_k\rightarrow \pi_1(X)\) of the projection map \(\mathrm{pr}:\pi_1(X)\twoheadrightarrow G_k\), determined up to conjugation by the geometric fundamental group \(\pi_1(\overline{X})\). Grothendieck's Anabelian Section Conjecture (GASC) states that if \(k\) is finitely generated over \(\mathbb{Q}\), then this association yields a bijection between \(X(k)\) and the set of conjugacy classes of sections of \(\mathrm{pr}:\pi_1(X)\twoheadrightarrow G_k\). There is also a birational version of the conjecture (BGASC), in which the fundamental groups \(\pi_1(X)\) and \(\pi_1(\overline{X})\) are replaced by the absolute Galois groups \(G_X:=\text{Gal}(K_X^{\mathrm{sep}}/K_X)\) and \(\overline{G_X}:=\text{Gal}(K_X^{\mathrm{sep}}/K_X\overline{k})\) (here \(K_X\) is the function field of the curve \(X\)). Finally, there are \(p\)-adic versions of both the section conjecture and its birational analogue, where \(k\) is taken to be a finite extension of \(\mathbb{Q}_p\) for some prime \(p\). The \(p\)-adic BGASC is now a theorem due to \textit{J. Koenigsmann} [J. Reine Angew. Math. 588, 221--235 (2005; Zbl 1108.14021)] and (in strengthened form) \textit{F. Pop} [Compos. Math. 146, No. 3, 621--637 (2010; Zbl 1210.11072)]. By work of \textit{S. Mochizuki} [Invent. Math. 138, No. 2, 319--423 (1999; Zbl 0935.14019)], distinct rational points of \(X\) yield distinct conjugacy classes of sections of \(\mathrm{pr}:\pi_1(X)\rightarrow G_k\) (in both the original and \(p\)-adic settings). But the remaining question of whether every conjugacy class of sections is point-theoretic (i.e. arises from a \(k\)-rational point) is wide open. The idea of the present paper is to attempt a reduction of the GASC and \(p\)-adic GASC to their respective birational versions. To that end, observe that there is a natural surjection \(G_X\twoheadrightarrow\pi_1(X)\), so one may ask whether a given section \(s:G_k\rightarrow\pi_1(X)\) may be lifted to a section \(\tilde{s}:G_k\rightarrow G_X\). This is called the \textit{cuspidalisation problem} for the section \(s\), and if \(s\) is point-theoretic, then such a lifting \(\tilde{s}\) always exists. In this paper, the author investigates a variant of the cuspidalization problem, where the full absolute Galois group \(G_X\) is replaced by a certain quotient \(G_X^{c-ab}\) called the \textit{maximal cuspidally abelian quotient} of \(G_X\). Section 1 of the paper is devoted to specifying a necessary cohomological condition for a section \(s\) to be point-theoretic: the étale cycle class of \(s\) must be uniformly orthogonal to \(\text{Pic}^\wedge\). The main result of the paper (Theorem 2.3.5) says that if the field \(k\) satisfies a certain cohomological finiteness condition, then a section \(s:G_k\rightarrow \pi_1(X)\) has cycle class uniformly orthogonal to \(\text{Pic}^\wedge\) if and only if it can be lifted to a section \(\tilde{s}:G_k\rightarrow G_X^{c-ab}\). In fact, the author works more generally over fields \(k\) of arbitrary characteristic, and with geometrically pro-\(\Sigma\) fundamental groups, where \(\Sigma\) is a nonempty set of prime numbers not containing \(\text{char}(k)\). The third and final section of the paper presents examples of non-point-theoretic sections of geometric pro-\(\Sigma\) fundamental groups in the case of a \(p\)-adic field \(k\), provided that \(p\not\in\Sigma\).