Some remarks on almost rational torsion points (Q2433575)

Given a commutative group scheme \(G\) over a perfect field \(K\) one says that a point \(P\in G(\overline{K})\) is almost rational if, whenever \(s\) and \(t\) are elements of the absolute Galois group of \(K\), the equality \(s(P)+t(P)=2P\) implies \(s(P)=t(P)=P\). One denotes by \(G^{\mathrm{ar}}_{\mathrm{tors},K}\) the set of almost rational points of \(G\) over \(K\) which are moreover torsion. This notion was introduced by Ribet (when \(G\) is the Jacobian of an algebraic curve over a number field) as a key tool for his new proof of Raynaud's theorem, previously known as Manin-Mumford conjecture. In this paper, the authors go further in the study of this notion in various settings. They first show a kind of ``universal boundedness conjecture'' (in the sense of Merel's theorem for elliptic curves) for almost rational points, when the base field \(K\) is a number field and \(G\) is a torus, or an abelian variety with complex multiplication, or in some cases a (non CM) elliptic curve. More precisely, they show that in any of the three preceding cases, for any pair of integers \((d,g)\), there is an integer \(U_{d,g}\) such that if \(K\) has degree \(d\) and \(G\) has dimension \(g\), then \(G^{\mathrm{ar}}_{\mathrm{tors},K}\) is killed by \(U_{d,g}\). They also prove results of opposite spirit when the base field \(k\) is finite or a \(p\)-adic field. In those cases they show indeed that given \(k\), except for an explicit finite set of \(G\)s over \({\mathbb F}_2\) or \({\mathbb F}_4\), the set of almost rational torsion points is infinite. They deduce from this result that for any abelian variety \(G\) over a \(p\)-adic field \(K\), there is a finite extension of \(K\) such that \(G^{\mathrm{ar}}_{\mathrm{tors},K}\) is infinite.