The generalized Franchetta conjecture for some hyper-Kähler varieties (Q2274096)

Let \(\mathcal{F}_g\) be the moduli stack of polarized \(K3\) surfaces of genus \(g\), and let \(\mathcal{S}\rightarrow \mathcal{F}_g\) be the universal family of \(K3\) surfaces. \textit{K. G. O'Grady} proposed in [J. Math. Pures Appl. (9) 100, No. 5, 701--718 (2013; Zbl 1325.14054)] the following analoque of the Franchetta conjecture which concerns about the universal family of the smooth projective curves of genus \(g\): For any algebraic cycle \(z\in \mathrm{CH}^2(\mathcal{S})\) and any point \(b\in\mathcal{F}_g\), the restriction of \(z\) to the fiber \(\mathcal{S}_b\) is a multiple of the Beauville-Voisin class [\textit{A. Beauville} and \textit{C. Voisin}, J. Algebr. Geom. 13, No. 3, 417--426 (2004; Zbl 1069.14006)] of \(\mathcal{S}_b\). This conjecture is verified in [\textit{N. Pavic} et al., Int. Math. Res. Not. 2017, No. 16, 4971--4983 (2017; Zbl 1405.14017)] for \(K3\) surfaces of genus \(g\leq 10\) and \(g=12,13,16,18,20\). The main goal of the present paper is to investigate a higher-dimensional analogue of O'Grady's conjecture, and proves the following result: Let \(\mathcal{M}\) be the moduli stack of smooth \(K3\) surfaces of genus 3,4, or 5, and let \(\mathcal{S}\rightarrow\mathcal{M}\) be the universal family. Let \(\mathcal{X}\) be \(\mathcal{S}\times_{\mathcal{M}}\mathcal{S}, \mathcal{S}\times_{\mathcal{M}}\mathcal{S}\times_{\mathcal{M}}\mathcal{S}, \mathrm{Hilb}_{\mathcal{M}}^2\mathcal{S}, \mathcal{S}\times_{\mathcal{M}}\mathrm{Hilb}_{\mathcal{M}}^2\mathcal{S}\), or \(\mathrm{Hilb}_{\mathcal{M}}^3\mathcal{S}\). Then for any algebraic cycle \(z\in \mathrm{CH}^*(\mathcal{X})_{\mathbb{Q}}\) and any point \(b\in\mathcal{M}\), the restriction of \(z\) to the fiber \(\mathcal{X}_b\) is zero if and only if it is numerically trivial. The authors also draw many consequences in the direction of the Beauville-Voisin conjecture as well as Voisin's refinement involving coisotropic subvarieties.