\(p\)-adic Tate conjectures and abeloid varieties (Q2332341)

The authors study a conjecture of Raskind's, which may be thought of as an analogue of the Tate conjecture. In order to state it, let us first fix some notations: suppose \(X/K\) is a smooth and proper variety over a local field \(K/\mathbb{Q}_p\). Let \(\bar{K}\) denote an algebraic closure of \(K\), and \(G_K\) the Galois group \(\mathrm{Gal}(\bar{K}/K)\). We first consider codimension one cycles, and consider the cycle class map \[ c: \mathrm{NS}\otimes_{\mathbb{Z}} \mathbb{Q}_p\rightarrow H^2_{\text{ét}}(X_{\bar{K}}, \mathbb{Q}_p(1))^{G_K}, \] where \(\mathrm{NS}\) denotes the Néron-Severi group, and the target denotes the Galois invariants in (Tate-twisted) étale cohomology. For varieties over a finite field or a number field, the map \(c\) is predicted to be an isomorphism by the Tate conjecture. Although formally similar, the current setting of \(X\) over local fields is in fact rather different; here, the map \(c\) is injective but easily shown to be not surjective, and so if one wants an isomorphism it is necessary to impose further conditions on \(X\). This is what Raskind's conjecture aims to do: Conjecture (Raskind). The map \(c\) is surjective if \(X\) has totally degenerate reduction. The phrase \textit{totally degenerate reduction} needs to be made precise, but essentially means that \(X\) has bad reduction, and the components of the special fiber (as well as their intersections, and the intersections of those, and so on) have Chow groups as simple as possible. It can be thought of as a maximal unipotent monodromy condition. Raskind's conjecture is in fact more general and deals with cycles of higher codimensions, but we will abusively refer to the above conjecture as Raskind's conjecture in this review. One key difference between Tate-type conjectures and, for example, the Hodge conjecture to note is that, while the Hodge conjecture would imply that one can pin down the \(\mathbb{Q}\)-vector space of algebraic cycles inside cohomology, the Tate conjecture only allows one to do so after tensoring to \(\mathbb{Q}_{\ell}\). This observation will be relevant in what follows. The following is one of the main results of this paper Theorem. There exist abelian surfaces \(B/K\) for which Raskind's conjecture is false. Let us give an outline of the proof; we will be brief here since this is nicely explained in the Introduction. Using one of Fontaine's equivalence of categories from \(p\)-adic Hodge theory, one can write \[ H^2_{\text{ét}}(X_{\bar{K}}, \mathbb{Q}_p(1))^{G_K} \cong H \cap \mathrm{Fil}^1_{dR}, \] where \(H\) is a \(\mathbb{Q}_p\)-vector space (a subspace of \(H^2_{\text{log-cris}}(X)\)), and the intersection is with the Hodge filtration after comparing to de Rham cohomology. On the other hand, \(H\) has a natural \textit{rational structure}, being spanned by algebraic cycles of the special fiber, and Raskind's conjecture amounts to comparing the intersection of this rational structure with \(\mathrm{Fil}^1_{dR}\), and the a priori larger intersection \(H \cap \mathrm{Fil}^1_{dR}.\) Since one has an explicit description of \(H\) and its rational structure, this gives a strategy to prove the Theorem, and the authors explicitly find very clean counterexamples in this way. The authors also consider the analogue of Raskind's Conjecture for homomorphisms between abelian varieties and find counterexamples using the same strategy as above; they also prove positive results for abelian varieties isogenous to products of Tate elliptic curves. Throughout the paper the authors work in the more general setting of abeloid varieties. The paper under review is very well written, with the ideas clearly presented.