Homomorphisms of multiplicative groups of fields preserving algebraic dependence (Q2326956)

From the introduction: ``In this paper we formulate and prove a version of the Grothendieck Section Conjecture [GSC for short] . For function fields of algebraic varieties over algebraically closed ground fields, this conjecture states, roughly, that the existence of group-theoretic sections of homomorphisms of their absolute Galois groups implies the existence of geometric sections of morphisms of models of these fields (...) This raises the question of functoriality, i.e., the reconstruction of rational morphisms between algebraic varieties from continuous homomorphisms of absolute Galois groups of their function fields. This general fundamental question was proposed by Grothendieck and lies at the core of the Anabelian Geometry Program.'' The main open problem concerns a Galois-theoretic criterium for the existence of rational sections of fibrations. More precisely, let $\pi:X \to Y$ be a fibration of integral algebraic varieties over an algebraically closed field $k$, with geometrically irreducible generic fiber of dimension at least $1$ over a base $Y$ of dimension $\geq 2$. This defines a field embedding $\pi^*:K := k(Y) \hookrightarrow L := k(X)$ such that the image of $L$ is algebraically closed in $K$. Dually, this induces a surjective restriction homomorphism of absolute Galois groups $G_K \to G_L$. Fix a prime $\ell \neq char(k)$ and write $\mathcal{G}$ for the maximal pro-$\ell$-quotient of $G$. The previous restriction then induces homomorphisms $\mathcal{G}_K^a \to \mathcal{G}_L^a$ and $\mathcal{G}_K^c \to \mathcal{G}_L^c$, where $\mathcal{G}^a$ (resp. $\mathcal{G}^c$) denotes the first quotient in the derived (resp. central) descending series of $\mathcal{G}$. The authors present here what they call a ``minimalistic'' version of GSC. Let $\pi_a:\mathcal{G}_K^c \to \mathcal{G}_L^a$ be the natural surjection, and $\Sigma_K = \Sigma(\mathcal{G}_K^c)$ the set of topologically non cyclic subgroups $\sigma \subset \mathcal{G}_K^a$ whose preimages $\pi_a^{-1}(\sigma) \subset \mathcal{G}_K^c$ are abelian. Assume that $\pi_a$ admits a section $\xi_a:\mathcal{G}_L^a \to \mathcal{G}_K^a$ such that $\xi_a(\Sigma_L) \subset \Sigma_K$. Then there exists a finite purely inseparable extension $\iota^* : L \hookrightarrow L' =k(Y')$ and a rational map $\xi:Y' \to X$ such that $\iota^*.\pi^*(L) = \iota^*(L)\subset L'$. Thus $\xi(Y')$ is a section over $Y$, modulo purely inseparable extensions. The main result of this paper establishes a link between the minimalistic GSC above and homomorphisms of multiplicative groups of fields preserving algebraic independence (as announced in the title). Note that $\mathcal{G}_K^a \cong \mathrm{Hom}(K^\times, \mathbb{Z}_l(1))$ by Kummer theory, $\xi_a$ induces the dual homomorphism $\hat{\psi}:\hat{K}^\times \to \hat{L}^\times$ of pro-$\ell$-completions, and the inclusion $\xi_a(\Sigma_L) \subset \Sigma_K$ says that $\hat{\psi}$ respects the skew-symmetric pairings on $\hat{K}^\times$ and $\hat{L}^\times$, with values in the second Galois cohomology groups of the corresponding fields (with $\ell$-torsion coefficients). If the restriction $\psi$ of $\hat{\psi}$ to $K^\times /k^\times$ satisfies $\psi : K^\times /k^\times \subseteq L^\times /k^\times \subset \hat{L}^\times$, then $\psi$ respects algebraic dependence, i.e. it maps algebraically dependent elements in $K^\times$ to algebraically dependent elements of $L^\times$ modulo $k^\times$ (the latter notion is intuitively clear, even if its formal definition can be cumbersome). The authors' central theorem reads as follows. From now on, $K$ will be an arbitrary field, and $v$ a non archimedean valuation on $K$, with the usual pertaining notations (valuation ring $\mathcal{O}_{K,v}$, maximal ideal $\mathfrak{m}_{K,v}$, residue field $\mathbf{K}_v$, etc.) Let us consider a tower of extensions $k \subseteq \tilde{k} \subseteq \tilde{k}_a \subset K$, where $k$ is the prime subfield of $K$, $\tilde{k}_a$ the algebraic closure of $\tilde{k}$ in $K$, and a second tower $l \subseteq \tilde{l} = \tilde{l}_a \subset L$ extending the first one. Give a homomorphism $\psi:K^\times /k^\times \to L^\times /\tilde{l}^\times$ such that: (i) $\psi$ preserves algebraic dependence w.r.t. $\tilde{k}$ and $\tilde{l}$; (ii) there exist $y_1,y_2 \in \psi(K^\times /k^\times)$ such that $y_1,y_2$ are algebraically independent modulo $\tilde{l}^\times$. Suppose moreover that $\psi$ satisfies a certain technical assumption called (AD) (which holds in non null characteristic but is conjecturally superfluous). Then three cases and only three can occur: (P) there exists a field $F \subset K$ such that $\psi$ factors through $K^\times /k^\times \twoheadrightarrow K^\times /F^\times$; (V) there exists a non trivial $v$ on $K$ such that the restriction of $\psi$ to $\mathcal{O}_{K,v}^\times\mathcal{O}_{k,v}^\times$ is trivial on $(1+\mathfrak{m}_{K,v})^\times/\mathcal{O}_{k,v}^\times$ and it factors through the reduction map $\mathcal{O}_{K,v}^\times\mathcal{O}_{k,v}^\times \twoheadrightarrow \mathbf{K}_v^\times /\mathbf{k}_v^\times \to L^\times /\tilde{l}^\times$; (VP) there exists a non trivial valuation $v$ on $K$ and a field $\mathbb{F}_v \subset \mathbf{K}_v$ such that the restriction of $\psi$ to $\mathcal{O}_{K,v}^\times\mathcal{O}_{k,v}^\times$ factors through $\mathcal{O}_{K,v}^\times\mathcal{O}_{k,v}^\times \twoheadrightarrow K \times\mathbf{K}_v^\times /\mathbf{k}_v^\times \to L^\times /\tilde{l}^\times$. The idea of the proof of the main theorem is to reduce the problem to a question in plane projective geometry over the prime subfield $k$, viewing $\mathbb{P}(K) := K \times /k \times$ as a projective space over $k$ and describing the homomorphisms $\psi:\mathbb{P}(K) \to \mathbb{P}(L)$ which preserve algebraic dependence. Indeed it suffices to show the existence of a subgroup $\mathfrak{U} \subset \mathbb{P}(K)$ such that, for every projective line $\mathfrak{l} \subset \mathbb{P}(K)$, $\mathfrak{U} \cap \mathfrak{l}$ is either: (i) the line $\mathfrak{l}$; (ii) a point $\mathfrak{q} \in \mathfrak{l}$; (iii) the affine line $\mathfrak{l} \setminus \mathfrak{q}$; (iv) if $k = \mathbb{Q}$, a set projectively equivalent to $\mathbb{Z}_{(p)}$ in $\mathbb{A}_1(\mathbb{Q}) \subset \mathbb{P}^1(\mathbb{Q})$. Actually such a subgroup is necessarily either $F^\times / k^\times$ for some subfield $F$ of $K$, or some $\mathcal{O}_{K,v}$ associated to a valuation $v$ of $K$. In case (V) (resp. (P)), $\psi$ factors through a valuation (resp. a subfield); case (VP) corresponds to valuations composed with projections.