On the Tate and Mumford-Tate conjectures in codimension 1 for varieties with \(h^{2,0}=1\) (Q523411): Difference between revisions

This paper concerns one of the big challenges in algebraic and arithmetic geometry: the Tate Conjecture for divisor classes on varieties which are defined over a finitely generated field of characteristic zero. This is known for abelian surfaces [\textit{G. Faltings}, Invent. Math. 73, 349--366 (1983; Zbl 0588.14026)], \(K3\) surfaces [\textit{Y. André}, Math. Ann. 305, No. 2, 205--248 (1996; Zbl 0942.14018)], [\textit{S. G. Tankeev}, Math. USSR, Izv. 37, No. 1, 191--208 (1991; Zbl 0722.14024); translation from Izv. Akad. Nauk SSSR, Ser. Mat. 54, No. 4, 846--861 (1990)], hyperkähler varieties [André, loc. cit.], and Hilbert modular surfaces (Harder-Langlands-Rapoport and others), but beyond this remains wide open. The present paper proves the Tate conjecture for divisor classes, and in fact an important generalization, the Mumford-Tate conjecture in cohomology degree two (which is open even for abelian varieties), for a complex variety \(X\) (smooth and complete) with Hodge number \(h^{2,0}(X)=1\) under the following natural moduli assumption: \(X\) sits in a smooth projective family with nonsingular irreducible base such that the underlying variation of Hodge structure is not isotrivial. In fact, the theorem is valid for any submotive of the second cohomology satisfying the above assumptions as the arguments only use an irreducible transcendental part \(\mathbb V\) of the Hodge structure as input. Crucially the author considers the endomorphism algebra \(E\) of \(\mathbb V\) which is either totally real or CM by results of Zarhin. In the totally real case (outside rank \(4\)), the proof proceeds roughly as in André's work in loc.cit. Key novelties consist in a variant of the Kuga-Satake construction geared towards the impact of non-trivial endomorphisms from \(E\), and the systematic use of norm functors to overcome the non-linearity in the Kuga-Satake construction, especially in the first step to form the even Clifford algebra. This formalism is valid generally in any Tannakian category of Hodge structures, motives, Galois representations etc (building on work of \textit{D. Ferrand} [Bull. Soc. Math. Fr. 126, No. 1, 1--49 (1998; Zbl 1017.13005)]) and thus should be of independent interest. In the CM case, the author follows a different line of argument which gives a new interpretation of \textit{B. van Geemen}'s half twists from [J. Math. Soc. Japan 53, No. 4, 813--833 (2001; Zbl 1074.14509)], valid in families. This step involves twisting by some one-dimensional motive which, while not being proved to be trivial, has trivial Hodge and \(\ell\)-adic realizations so that results of Faltings apply. The remaining totally real case of rank \(4\) requires a special treatment because the algebraic monodromy may be smaller than the generic Mumford-Tate group. To overcome this, the author develops a combination of the arguments from the above two cases to reduce to the Mumford-Tate conjecture for certain abelian varieties which can then be treated by fairly direct arguments (but which do not seem to be covered by the existing literature). The paper concludes by applying the main result to the prototype case of algebraic surfaces with geometric genus \(p_g=1\). Here the non-isotriviality of the variations of Hodge structures are known for many moduli components (such as \(K3\) surfaces and Todorov surfaces), but the author also sketches how to treat some further cases, essentially covering all components with a nice description available.