Maximality of Galois actions for abelian and hyper-Kähler varieties (Q2178465)

This article makes significant progress on the question of maximality of images of compatible systems of \(\ell\)-adic Galois representations over number fields in at least two ways: (1) It proposes an approach that shifts part of the difficulty the arithmetic geometry side to group theory. (2) New maximal image results are obtained for certain systems of Galois representations arising from hyper-Kähler varieties. The key novelty of the article is that the maximality of the Galois image can be detected and characterised by an equality of dimensions, which establishes a passage between \(\ell\)-adic and mod~\(\ell\) information. In order to describe the results some notation need be introduced. Let \(K\) be a subfield of \(\mathbb{C}\) which is finitely generated over \(\mathbb{Q}\) and let \(X\) be a proper smooth variety over it. Fix an integer \(i \ge 0\) and, for each prime number \(\ell\), consider the \(i\)-th étale cohomology of \(X\) with coefficients in \(\mathbb{Q}_\ell\). The action of the absolute Galois group \(G_K\) of \(K\) on it is \(\mathbb{Q}_\ell\)-linear and gives rise to a linear \(\ell\)-adic representation \(\rho_\ell\) of \(G_K\). Let \(n\) be its dimension, which is independent of \(\ell\). An important class of examples is provided by abelian varieties over number fields. Serre proved that the images of \(\rho_\ell\) are maximal (in a suitable sense) for large enough \(\ell\) under some assumptions on the dimension of the abelian varieties; that work was completed by \textit{J. P. Wintenberger} [J. Reine Angew. Math. 553, 1--16 (2002; Zbl 1048.14008)] who made crucial use of cases of Tate's conjecture due to Faltings. The maximality of images is stated in a way introduced by one of the authors in [\textit{M. Larsen}, Duke Math. J. 80, No. 3, 601--630 (1995; Zbl 0912.11026)]. More precisely, let \(\Gamma_\ell = \rho_\ell(G_K)\) be the image of the Galois representation and, furthermore, let \(\mathbf{G}_\ell\) be its Zariski closure in \(\mathrm{GL}_{n,\mathbb{Q}_\ell}\). Consider the universal covering group \(\mathbf{G}_\ell^{\mathrm{sc}}\) of the quotient \(\mathbf{G}_\ell^{\mathrm{ss}}\) of the connected component \(\mathbf{G}_\ell^\circ\) by its radical. Accordingly, let \(\Gamma_\ell^\circ = \Gamma_\ell \cap \mathbf{G}_\ell^\circ(\mathbb{Q}_\ell)\) and \(\Gamma_\ell^{\mathrm{sc}}\) be the preimage under \(\mathbf{G}_\ell^{\mathrm{sc}}(\mathbb{Q}_\ell) \twoheadrightarrow \mathbf{G}_\ell^{\mathrm{ss}}(\mathbb{Q}_\ell)\) of the image of \(\Gamma_\ell^\circ\) in \(\mathbf{G}_\ell^{\mathrm{ss}}(\mathbb{Q}_\ell)\). The maximality considered is the statement that the \(\ell\)-adic Lie group \(\Gamma_\ell^{\mathrm{sc}}\) is a hyperspecial maximal compact subgroup of \(\mathbf{G}_\ell^{\mathrm{sc}}(\mathbb{Q}_\ell)\). Conjecturally, this should hold for large enough \(\ell\). The main application of the techniques of the article allow the authors to prove this conjecture for abelian varieties and for hyper-Kähler varieties (for \(i=2\)). The former constitutes a new proof of Wintenberger's result cited above, whereas the latter is new. These results are derived from the main theorem of the paper, characterising the maximality of the image by the mentioned equality of dimensions in the case of a number field \(K\) and under the assumption that \(\mathbf{G}_\ell\) is connected. Theorem. For all large enough primes \(\ell\), the following statements are equivalent: (i) \(\Gamma_\ell^{\mathrm{sc}}\) is a hyperspecial maximal compact subgroup of \(\mathbf{G}_\ell^{\mathrm{sc}}(\mathbb{Q}_\ell)\). (ii) \(\dim_{\mathbb{Q}_\ell} \big(\mathrm{End}_{\rho_\ell^{\mathrm{ss}}(G_K)}(\mathbb{Q}_\ell^n)\big) = \dim_{\mathbb{F}_\ell} \big(\mathrm{End}_{\overline{\rho}_\ell^{\mathrm{ss}}(G_K)}(\mathbb{F}_\ell^n)\big)\), where \(\rho_\ell^{\mathrm{ss}}\) is the semi-simplification of \(\rho_\ell\) and \(\overline{\rho}_\ell^{\mathrm{ss}}\) is the semi-simplification of any reduction of a lattice in \(\rho_\ell\). The main ingredient in the proof of this theorem is a purely group theoretic result (Theorem 1.5 in the paper), the proof of which uses Nori's theory, Bruhat-Tits theory as well as significant earlier work of the authors.