Groups of points on abelian threefolds over finite fields (Q2422195)

If \(X\) is an abelian threefold defined over some \(\mathbb{F}_q\) then the \(l\)-primary part of the torsion is determined by the action of Frobenius on the Tate module, but the list of possibilities can be complicated if the characteristic polynomial is inseparable. For threefolds, some cases were described by \textit{S. Rybakov} [Mosc. Math. J. 15, No. 4, 805--815 (2015; Zbl 1382.14019)]. This paper completes the threefold case. The polynomials concerned are \(P(t)^2Q(t)\) with \(P\) and \(Q\) both separable of degree \(2\), and \(P(t)(t\pm\sqrt{q})^2\) with \(P\) separable of degree~\(4\) or \(P=Q^2\) with \(Q\) separable. In these cases \(X\) is decomposible and the resulting exact sequence induces an exact sequence of Tate modules as \(\mathbb{Z}_l[T]\) modules, with \(T\) acting by \(1-\mathrm{Frob}\). The question then is to determine which exponents (Smith invariants) can occur in the middle term. This is achieved by using Littlewood-Richardson rules, applying a modified version of results of \textit{W. Fulton} [Bull. Am. Math. Soc., New Ser. 37, No. 3, 209--249 (2000; Zbl 0994.15021)] and \textit{A. P. Santana} et al. [Linear Multilinear Algebra 46, No. 1--2, 1--23 (1999; Zbl 0935.05093)] on eigenvalues of sums of hermitian matrices. Editorial remark: The reviewer points out that frequently the term ``Liouville-Richardson'' appears in the paper which seems to be a misprint for ``Littlewood-Richardson''.