How large is \(A_g(\mathbb F_q)\)? (Q1712811)

Let \(A(p, g)\) denote the number of isomorphism classes of principally polarized abelian varieties of dimension \(g\) over the finite field \(\mathbb{F}_p\). In the present paper, the authors prove that \(\log A(p, g)\) grows at least as quickly as a multiple of \(g^2 \log g\). On the other hand, they also show that if one forgets the polarization and simply counts the number \(B(p, g)\) of abelian varieties over \(\mathbb{F}_p\), then the rate of growth of \(\log B(p, g)\) is much slower. In fact, the abundance of principal polarizations on a fixed abelian variety is the primary reason for the large lower bound on \(A(p, g)\). The extremely large gap between the lower bound for \(A(p, g)\) and the upper bound for \(B(p, g)\) implies some statistically counterintuitive behavior for abelian varieties of large dimension over a fixed finite field. Conditionally, they show that principally polarized abelian varieties do not obey the Cohen-Lenstra heuristics. Unconditionally, they prove that most principally polarized abelian varieties correspond to abelian varieties with a repeated isogeny factor.