Generalizations of results of Friedman and Washington on cokernels of random $p$-adic matrices

TL;DR

This paper studies the distribution of cokernels of polynomial evaluations of random p-adic matrices, showing they follow Cohen-Lenstra distributions and generalize previous results.

Contribution

It extends Friedman and Washington's results to multiple polynomials and modules, establishing independence and distribution properties for cokernels of random p-adic matrices.

Findings

Probabilities of cokernels are asymptotically independent for different polynomials.

Cokernel distributions follow Cohen-Lenstra type distributions.

Results verify new cases of a conjecture by Cheong and Huang.

Abstract

Let $p$ be prime and $X$ be a Haar-random $n \times n$ matrix over $\mathbb{Z}_{p}$, the ring of $p$-adic integers. Let $P_{1}(t), \dots, P_{l}(t) \in \mathbb{Z}_{p}[t]$ be monic polynomials of degree at most $2$ whose images modulo $p$ are distinct and irreducible in $\mathbb{F}_{p}[t]$. For each $j$, let $G_{j}$ be a finite module over $\mathbb{Z}_{p}[t]/(P_{j}(t))$. We show that as $n$ goes to infinity, the probabilities that $\mathrm{cok}(P_{j}(X)) \simeq G_{j}$ are independent, and each probability can be described in terms of a Cohen-Lenstra distribution. We also show that for any fixed $n$, the probability that $\mathrm{cok}(P_{j}(X)) \simeq G_{j}$ for each $j$ is a constant multiple of the probability that that $\mathrm{cok}(P_{j}(\bar{X})) \simeq G_{j}/pG_{j}$ for each $j$, where $\bar{X}$ is an $n \times n$ uniformly random matrix over $\mathbb{F}_{p}$. These results generalize work of Friedman and Washington and prove new cases of a conjecture of Cheong and Huang.