Hilbert squares of genus 16 K3 surfaces
Junyu Meng
TL;DR
The article advances the understanding of genus $16$ K3 surfaces by proving that the Hilbert scheme $S^{[2]}$ is isomorphic to the moduli space $M_{S'}(2,h',7)$ for the FM partner $S'$, and by constructing a natural embedding of a projectivized bundle $X'= extbf{P}_{S'}(F'^*)$ into both hyper-Kähler fourfolds in a way compatible with the Fourier–Mukai correspondence. It provides an explicit geometric realization of this isomorphism via a short exact sequence involving the rank-$8$ bundle $E_8'$, yielding a concrete description of the isomorphism in terms of stable sheaves and Brill–Noether loci, and it situates these objects within Mukai’s projective models of genus $16$ K3s. The paper further connects to the Debarre–Voisin program by explicitly constructing a trivector $t_1$ on $V_{10}$ that should realize $S^{[2]}$ as a Debarre–Voisin fourfold $DV(t_1)$, thereby offering a geometric explanation for the isomorphism and linking the genus $16$ hyper-Kähler geometry to modern projective constructions. Overall, it provides a coherent geometric picture tying together Hilbert schemes, moduli of sheaves, FM partners, and Debarre–Voisin varieties in the genus $16$ setting, with explicit, verifiable structures.
Abstract
We consider the geometry of a general polarized K3 surface $(S,h)$ of genus 16 and its Fourier-Mukai partner $(S',h')$. We prove that $S^{[2]}$ is isomorphic to the moduli space $M_{S'}(2,h',7)$ of stable sheaves with Mukai vector $(2,h',7)$ and describe the embeddings of the projectivization of the stable vector bundle of Mukai vector $(2,-h',8)$ over $S'$ into these two isomorphic hyper-Kähler fourfolds. Following the work of FrédÉric Han in arXiv:2501.16013, we explicitly construct an interesting 3-form $t_1\in \wedge^3 V_{10}^*$ which potentially gives an isomorphism between $S^{[2]}$ and the Debarre-Voisin fourfold in $G(6,V_{10})$ associated to $t_1\in \wedge^3 V_{10}^*$. This would provide a geometric explanation of the existence of such an isomorphism, which was proved in arXiv:2102.11622 by a completely different argument.
