The cohomology of the general stable sheaf on a K3 surface

TL;DR

This paper classifies when the general sheaf in moduli spaces on K3 surfaces has simple cohomology, providing algorithms and bounds, and offers new insights into Ulrich bundles and sheaf generation.

Contribution

It introduces a classification of weak Brill-Noether failure for sheaves on K3 surfaces, with explicit algorithms and bounds, utilizing Bridgeland stability conditions.

Findings

Finitely many counterexamples for each rank on Picard rank one K3 surfaces.

Explicit classification and cohomology calculations for these counterexamples.

Sharp bounds on parameters ensuring weak Brill-Noether holds.

Abstract

Let $X$ be a K3 surface with Picard group $\mathrm{Pic}(X)\cong\mathbb{Z} H$ such that $H^2=2n$. Let $M_{H}(\mathbf{v})$ be the moduli space of Gieseker semistable sheaves on $X$ with Mukai vector $\mathbf{v}$. We say that $\mathbf{v}$ satisfies weak Brill-Noether if the general sheaf in $M_{H}(\mathbf{v})$ has at most one nonzero cohomology group. We show that given any rank $r \geq 2$, there are only finitely many Mukai vectors of rank $r$ on K3 surfaces of Picard rank one where weak Brill-Noether fails. We give an algorithm for finding the potential counterexamples and classify all such counterexamples up to rank 20 explicitly. Moreover, in each of these cases we calculate the cohomology of the general sheaf. Given $r$, we give sharp bounds on $n$, $d$, and $a$ that guarantee that $\mathbf{v}=(r,dH,a)$ satisfies weak Brill-Noether. As a corollary, we obtain another proof of the classification of Ulrich bundles on K3 surfaces of Picard rank one. In addition, we discuss the question of when the general sheaf in $M_H(\mathbf{v})$ is globally generated. Our techniques make crucial use of Bridgeland stability conditions.