When is a polarised abelian variety determined by its $\boldsymbol{p}$-divisible group?
Tomoyoshi Ibukiyama, Valentijn Karemaker, Chia-Fu Yu
TL;DR
The paper addresses when the supersingular locus ${\mathcal{S}}_g$ in the Siegel moduli space is geometrically irreducible and when a principally polarised supersingular abelian variety is uniquely determined by its polarised $p$-divisible group (singleton central leaf). It builds a bridge between the arithmetic of definite quaternion Hermitian lattices and the geometry of ${\mathcal{A}}_g$ by exploiting mass formulae, automorphism group analyses, and Ekedahl-Oort stratification, including a detailed study in genus $g=4$. The main results give an explicit list of irreducible cases for ${\mathcal{S}}_g$ and classify singleton central leaves (in low genus) via class-number one problems, linking geometric questions to Gauss-type lattice problems. Overall, the work provides a computer-free, mass-formula-based approach that unifies arithmetic lattice theory with the geometry of supersingular abelian varieties and their central leaves, yielding precise irreducibility and uniqueness criteria with broad implications for moduli of abelian varieties in characteristic $p$.
Abstract
We study the Siegel modular variety $\mathcal{A}_g \otimes \overline{\mathbb{F}}_p$ of genus $g$ and its supersingular locus $\mathcal{S}_g$. As our main result we determine precisely when $\mathcal{S}_g$ is irreducible, and we list all $x$ in $\mathcal{A}_g \otimes \overline{\mathbb{F}}_p$ for which the corresponding central leaf $\mathcal{C}(x)$ consists of one point, that is, for which $x$ corresponds to a polarised abelian variety which is uniquely determined by its associated polarised $p$-divisible group. The first problem translates to a class number one problem for quaternion Hermitian lattices. The second problem also translates to a class number one problem, whose solution involves mass formulae, automorphism groups, and a careful analysis of Ekedahl-Oort strata in genus $g=4$.
