Footnotes to papers of O'Grady and Markman (Q2118187)

The author generalizes a result of O'Grady proved for hyper-Kähler manifold of generalized Kummer type to any hyper-Khähler manifold \(X\) with \(b_3(X) \neq 0\). O'Grady [\textit{K. G. O'Grady}, Int. Math. Res. Not. 2021, No. 16, 12356--12419 (2021; Zbl 1490.53068)] and Markman consider an hyper-Kähler manifold \(X\) of generalized Kummer type and consider the intermediate Jacobian \(J^{3}(X)\) which is the complex torus built on the Hodge structure on \(H^{3}(X,\mathbb{Z})\) which is of level 1 since \(H^{3,0}(X)=0\). When \(X\) is projective, the complex torus \(J^3(X)\) is an abelian variety of dimension \(4\) because \(b_3(X)=8\). O'Grady proved that \(J^3(X)\) is a Weil abelian fourfold and that for a very general projective deformation of \(X\), the Kuga-Satake abelian variety \(KS(X)\) of \((H^2(X,\mathbb{Q})_{tr},(, ))\) is isogenous to a power of \(J^3(X)\). The last fact is a consequence of certain universality property of the Kuga-Satake construction proved by Charles [\textit{F. Charles}, Math. Z. 300, No. 4, 3623--3643 (2022; Zbl 1490.14018)] and by another result of O'Grady which states that if \(X\) is an hyper-Kähler 2n-folds of generalized Kummer deformation type with \(n \geq 2\) then the map \(\bigwedge^{2} H^{3}(X, \mathbb Q) \to H^{6}(X,\mathbb Q) \xrightarrow{Q_X^{n-2}} H^{4n-2}(X, \mathbb Q)\) is surjective. Here \(Q_X \in H^4(X, \mathbb Q)\) is a cohomology class which is constructed using the Beauville-Bogomolov form. The first main result of the paper is a generalization of the previous result to any hyper-Kähler. More precisely the author proves that if \(X\) is an hyper-Kähler 2n-folds such that \(b_3(X) \neq 0\) then the map \(\bigwedge^{2} H^{3}(X, \mathbb Q) \to H^{6}(X,\mathbb Q) \xrightarrow{Q_X^{n-2}} H^{4n-2}(X, \mathbb Q)\) is surjective. Moreover if \(X\) is projective \(J^{3}(X)\) contains a simple component of the Kuga-Satake abelian variety \(H^{2}(X,\mathbb Q)_{tr}\). The author gives also a lower bound for the second Betti number, more precisely she proves that \(b_3(X) \geq 2^{k}\) where \(k=\frac{b_2(X)-2}{2}\) if \(b_2(X)\) is even, and \(k=\frac{b_2(X)-1}{2}\) if \(b_2(X)\) is odd. The author proves also that \(b_{2n-1}(X) \geq 2^k\), if \(H^{2n-1}(X,\mathbb Q) \neq 0\), \(H^{2n-3}(X,\mathbb Q) =0\), where k is as before. In particular if \(X\) is an hyper-Kähler manifold of dimension \(6\) and if \(b_{odd}(X) \neq 0\) then \(b_{odd}(X)\geq 2^k\), where \(k = b_2(X)-2\) if \(b_2(X)\) is even, \(k\ = b_2(X)-1\) if \(b_2(X)\) is odd. Recall that we have a precise bound for the Betti numbers of hyper-Kähler manifolds in dimension \(4\) in [\textit{D. Guan}, Math. Res. Lett. 8, No. 5--6, 663--669 (2001; Zbl 1011.53039)], and in higher dimension there are some partial results, see [\textit{J. Sawon}, Eur. J. Math. 8, No. 3, 1196--1212 (2022; Zbl 1503.53104), \textit{Y.-J. Kim} and \textit{R. Laza}, Bull. Soc. Math. Fr. 148, No. 3, 467--480 (2020; Zbl 1457.14090)], but the subject remains however wide open. A key point in both cases is the fact that the weight 3 or weight \(2n - 1\) Hodge structure one considers is of Hodge level 1, that is, they satisfy the property \(h^{p,q} =0\) for \(|p-q|\geq 1\). The second part of the paper concerns hyper-Kähler manifolds of generalized Kummer deformation type and provides a complement to Markman's paper. In this paper, Markman proves the Hodge conjecture for the Weil Hodge classes on the Weil abelian fourfolds that are the intermediate Jacobians \(J^3(X)\) of a projective hyper-Kähler manifold of generalized Kummer deformation type with \(\rho(X) = 1\) . The second main result of the paper states that for \(X\) a projective hyper-Kähler manifold of generalized Kummer deformation type with \(n \geq 2\), the Kuga-Satake correspondence between \(X\) and its Kuga-Satake variety \(KS(X)\) is algebraic.