Finiteness of Mordell-Weil groups of Kuga fiber spaces of abelian varieties (Q2367110)

Let \(f : {\mathcal X}\to M\) be a projective abelian scheme over an arithmetic quotient of a hermitian symmetric domain \(M = \Gamma \backslash {\mathcal D}\), constructed from a symplectic representation of the associated algebraic group \(G_ \mathbb{Q}\). Such an abelian scheme is called a Kuga fiber space of abelian varieties. The Mordell-Weil group \(\text{MW} ({\mathcal X}/{\mathcal M})\) of the Kuga fiber space \(f : {\mathcal X} \to M\) is defined to be the group of rational sections of \(f\). In the paper, we show a finiteness theorem of the Mordell-Weil groups of Kuga fiber spaces of abelian varieties arising from standard \(\mathbb{Q}\)- primary symplectic representations. We remark that standard \(\mathbb{Q}\)- symplectic representations are defined and classified by Satake, and each of them can be decomposed into \(\mathbb{Q}\)-primary ones. The class of standard representations are very natural and it includes most of interesting and important cases such as elliptic modular cases, Siegel modular cases, Shimura's moduli families of abelian varieties and so on. Therefore our theorem is a generalization of Shioda's results on elliptic modular surfaces [\textit{T. Shioda}, J. Math. Soc. Japan 24, 20-59 (1972; Zbl 0226.14013)], and Silverberg's results on some important classes of Kuga fiber spaces [\textit{A. Silverberg}, Invent. Math. 81, 71-106 (1985; Zbl 0576.14020)]. The main idea of the proof is a generalization of Silverberg's cohomological criterion for finiteness of Mordell-Weil groups of Kuga fiber spaces [cf. Duke Math. J. 56, No. 1, 41-46 (1988; Zbl 0645.14019)]. Let \(\Gamma \to GL (W_ \mathbb{C})\) be the symplectic representation associated to a Kuga fiber space, and let \(H^ \bullet (\Gamma, W_ \mathbb{C})\) denote the Eilenberg-MacLane cohomology group. Then Silverberg's criterion says that if \(\dim M>1\) or \(M\) is compact and \(H^ q (\Gamma, W_ \mathbb{C}) = 0\) for \(q=0,1\), the Mordell-Weil group MW\(({\mathcal X}/M)\) is finite. Therefore vanishing theorems for \(H^ q (\Gamma, W_ \mathbb{C})\) imply the finiteness of the Mordell-Weil groups. In case when the \(\mathbb{R}\)-rank of \(G_ \mathbb{R}\) is greater than one, we can show that \(H^ q (\Gamma, W_ \mathbb{C}) = 0\) for \(q \leq 1\) by using the intersection cohomology group, Zucker conjecture proved by Looijenga and by Saper and Stern, and Borel-Wallach vanishing theorem. In case when \(\text{rank}_ \mathbb{R} G_ \mathbb{R} = 1\) which includes the case of elliptic modular surfaces, we do have examples with \(H^ 1 (\Gamma, W_ \mathbb{C}) \neq 0\). Therefore we have to generalize the Silverberg criterion by using (mixed) Hodge theory on \(H^ 1 (\Gamma, W_ \mathbb{C})\). The generalized criterion says that the Mordell-Weil group is finite if and only if \(H^ 1 (\Gamma, W_ \mathbb{Q}) \cap H^{0,0} = 0\) where \(H^{0,0}\) is the Hodge component of \(H^ 1 (\Gamma, W_ \mathbb{C})\) of type (0,0). (In case when \(M\) is not compact, we will consider the intersection cohomology and its pure Hodge structure.) Then, though we have examples with \(H^{0,0} \neq 0\), we can show that \(H^ 1 (\Gamma, W_ \mathbb{Q}) \cap H^{0,0} = 0\), hence the Mordell-Weil group is finite.