The asymptotic distribution of Elkies primes for reductions of abelian varieties is Gaussian
Alexandre Benoist, Jean Kieffer
TL;DR
This work generalizes Elkies primes from elliptic curves to abelian varieties with real multiplication and proves that, under GRH and a large adelic Galois image, the count of Elkies primes in specified ranges for reductions of A distributes as a Gaussian in the limit. The authors develop a two-pronged approach: (i) a density-theoretic analysis via an explicit Chebotarev framework applied to torsion fields, and (ii) a combinatorial analysis of split matrices in $ ext{GSp}_{2h}$ to quantify Elkies-conditions through characteristic polynomials. They show convergence of all moments to those of the standard Gaussian, derive explicit asymptotics for the even moments, and obtain an asymptotic Gaussian behavior for the Elkies-prime counts on average over reductions, with implications for the SEA algorithm on abelian surfaces. The results extend prior work on elliptic curves to higher dimensions, providing new insights into the distribution of Elkies primes and confirming the practical viability of SEA-like point counting in the RM setting. The numerical experiments corroborate the Gaussian pattern and motivate further exploration of the constants $oldsymbol{}_h$ across RM families.
Abstract
We generalize the notion of Elkies primes for elliptic curves to the setting of abelian varieties with real multiplication (RM), and prove the following. Let $A$ be an abelian variety with RM over a number field whose attached Galois representation has large image. Then the number of Elkies primes (in a suitable range) for reductions of $A$ modulo primes converges weakly to a Gaussian distribution around its expected value. This refines and generalizes results obtained by Shparlinski and Sutherland in the case of non-CM elliptic curves, and has implications for the complexity of the SEA point counting algorithm for abelian surfaces over finite fields.
