Diophantine problems and \(p\)-adic period mappings (Q785943)

In the paper under review, the authors introduce a novel \(p\)-adic approach for establishing Zariski non-density results for varieties over number fields. They illustrate their methods by giving new proofs of the finiteness of solutions of the \(S\)-unit equation and of Falting's theorem (Mordell's conjecture). Moreover, they produce a result concerning the moduli space of degree \(d\) hypersurfaces in \(\mathbb P^n\), for sufficiently large \(d\) and \(n\). Note that the last result is well beyond the \(1\) dimensional case. The method introduced in this paper has many similarities in spirit with Kim's non-abelian Chabatau [\textit{M. Kim}, Invent. Math. 161, No. 3, 629--656 (2005; Zbl 1090.14006); Publ. Res. Inst. Math. Sci. 45, No. 1, 89--133 (2009; Zbl 1165.14020)]. Both methods use \(p\)-adic Hodge theory and \(p\)-adic period maps in an essential way, albeit in a different manner. The rough idea is as follows: Let \(Y/K\) be a smooth variety over a number field and \(S\) a finite set of places of \(K\). In place of Kim's non-abelian cohomology, the authors use a suitable smooth proper covering \(\pi:X\to Y\). Assume that \(\pi\) extends to a smooth proper morphism of smooth \(\mathcal O_{K,S}\)-schemes, and choose a place \(v\notin S\), with residue characteristic \(p\), which satisfies some minor conditions. Fix \(y_0\in Y(K_v)\). For each \(K_v\)-point of \(Y\) the de Rham cohomology of its fiber carries the structure of a filtered \(\phi\)-module over \(K_v\). This gives rise to a local period map with source the residue disc of \(Y(K_v)\) at \(y_0\). The authors relate the local to the complex period map, by using the fact that in both situations the identification of the cohomology of nearby fibres can be given by the Gauss-Manin connection which can be defined over \(K\). Using this relation, the authors can control the fibers of the local period map under the assumption that the centraliser of the crystalline Frobenius is small in an appropriate sense which is related with the topological monodromy group which needs to be `big'. Each \(S\)-integral point, corresponds to a \(K\)-point and the \(p\)-adic cohomology of the geometric fibre at that point gives a global Galois representation. Consider now the \(S\)-integral points in the residue disc at \(y_0\), whose associated global Galois representation is semisimple. The image under the local period map of these points, is contained in a finite union of orbits of the centraliser of the crystalline Frobenius on the `local period domain'. Once one has enough control on the fibers of the local period map and the image of the \(S\)-integral points, the method gives Zariski non-density results. For the method to go through, one needs to exhibit a family, such that (a) one can show that the centraliser of the crystalline Frobenius is small in an appropriate sense and (b) one can control the \(S\)-integral points for which the associated global representation in not semisimple (conjecturally there are no such points). The authors successfully implement their methods for the three results mentioned in the first paragraph by using a variant of the Legendre family for the \(S\)-unit equation, by constructing a suitable family they call Kodaira-Parshin family for the Mordell conjecture, whereas for the moduli space they use the universal family. The methods introduced in this paper have many more potential applications. On the effectivity side, an interesting question is whether and to what extent these methods can become explicit and algorithmic.