The monodromy of cyclic Pryms

TL;DR

This paper investigates the monodromy groups of cyclic Prym varieties, showing they are often large and related to unitary groups, with applications to the distribution of Selmer groups of certain elliptic surfaces over finite fields.

Contribution

It establishes that the geometric monodromy of cyclic Prym families is contained between a unitary group and its derived subgroup, with implications for Selmer group distributions.

Findings

Geometric monodromy is sandwiched between a unitary group and its derived subgroup.

Large monodromy results imply deviations from standard heuristics for Selmer groups.

Average size of $l$-Selmer groups for certain elliptic surfaces is explicitly computed.

Abstract

The {\em Prym} of a cyclic covering of smooth projective curves is the ``new'' part of the Jacobian: the quotient of the Jacobian of the covering curve by the Jacobians of the intermediate covers. Given a family of such coverings, the fundamental group of the base of the family acts on the Tate modules of the Pryms, and the image of this representation is a key ingredient in answering arithmetic statistics questions about the distribution of the group structure of the $L$-torsion of a random Prym in the family. (Over ${\mathbb{F}}_q$, the action of Frobenius is roughly uniformly distributed over the {\em arithmetic} monodromy, a coset of the image of the fundamental group of the base change to $\bar{\mathbb{F}}_q$ (the {\em geometric} monodromy).) In the present note, we show for a number of natural families that (with limited exceptions) the geometric monodromy is sandwiched between a certain unitary group and its derived subgroup. In particular, this holds for the one-parameter families obtained by starting with any fixed cover and varying one (tame) ramification point. As an application, we deduce analogous largeness results for the monodromy of the Selmer groups of elliptic surfaces with $j=0$ or $j=1728$, by relating them to cyclic covers of degree 6 or 4 respectively, implying that their Selmer groups do not satisfy the standard heuristics. For instance, for eliptic surfaces with $j=0$ of sufficiently large height over ${\mathbb{P}}^1_{{\mathbb{F}}_q}$, the average size of the $l$-Selmer group is $l+3+o_q(1)$ when $l$ (fixed) and $q$ (large) are both 1 mod 3, compared to $l+1+o_q(1)$ for general elliptic surfaces.