Kuga-Satake construction and cohomology of hyperkähler manifolds (Q2001576): Difference between revisions

The Kuga-Satake construction associates to any polarized \(K3\) surface an abelian variety. More generally, it defines an injective map between all Hodge structures of \(K3\) type (i.e., integral Hodge structures of weight 2 with possibly indefinite polarization and Hodge diamond of \(H^2(X,\mathbb{C})\) for a \(K3\) surface \(X\)) and all Hodge structures of weight 1. Hence many aspects of Hodge structures of \(K3\) type are reduced to certain Hodge structures of weight 1. The crucial input is the even Clifford algebra \(\mathrm{Cl}^+(H_{\mathbb{R}},q)\) of a Hodge structure \(H\) of \(K3\) type together with the quadratic form \(q\) defined by the (possibly indefinite) polarization where \(H_{\mathbb{R}}=H\otimes_{\mathbb{Z}} \mathbb{R}\). On the real torus \[ \mathrm{Cl}^+(H_{\mathbb{R}},q)/\mathrm{Cl}^+(H,q) \] there is a natural complex structure yielding a complex torus \(\mathrm{KS}(H)\). The first cohomology of \(\mathrm{KS}(H)\) gives the desired Hodge structure of weight 1. Moreover, there is a natural embedding \[ H\hookrightarrow H^2(\mathrm{KS}(H)) \tag{\(\ast\)} \] of Hodge structures of weight 2. The present article generalizes the Kuga-Satake construction from \(K3\) surfaces to any compact simply connected hyper-Kähler manifold \(M\), equivalently irreducible holomorphic symplectic manifold, via \(k\)-symplectic structures. A \(k\)-symplectic structure on a vector space \(V\) is a subspace \(\Omega\subset \Lambda^2 V^*\) of dimension \(k\) with the following properties: (1) If \(w\in \Omega\) is non-zero, then the rank of \(w\) is either maximal or \(\tfrac{1}{2}\dim V\). In particular, the set of degenerate forms in \(\Omega\) is a quadric. (2) The quadric inside \(\mathbb{P}(\Omega)\) of degenerate forms is non-degenerate as a quadric. For \(k=3\) we recover hypersymplectic structures, in particular the previously considered case. To generalize \((*)\), the authors start with the Beauville-Bogomolov-Fujiki (BBF) form \(q_{\mathrm{BBF}}\), a non-degenerate primitive integral quadratic form on \(H:=H^2(M,\mathbb{Z})\). Let \(\mathfrak{g}_{\mathrm{tot}}(M)\) be the Lie algebra generated by all Lefschetz \(\mathfrak{sl}(2)\)-triples on the cohomology algebra of \(M\). These triples are generalizations of the \(\mathfrak{sl}(2)\)-triples determined by the Lefschetz operator of a compact Kähler manifold. The Lie algebra \(\mathfrak{g}_{\mathrm{tot}}(M)\) has the beautiful property that its action on \(H^\bullet(M,\mathbb{C})\) recovers all Hodge structures on \(H^\bullet(M,\mathbb{C})\) of hyper-Kähler type, i.e., all which are determined by the complex structures of the hyper-Kähler manifold \(M\). Given this setup, the authors construct a \(\mathrm{Cl}(H_{\mathbb{R}},q_{\mathrm{BBF}})\)-module \(V\) (for some \(k\)) such that \(H\hookrightarrow \Lambda^2 V^*\) is a \(k\)-symplectic structure. Let \(T\) be the real torus with \(H^2(T,\mathbb{R})=\Lambda^2 V^*\). Then \(\mathfrak{g}_{\mathrm{tot}}(M)\) acts on both \(H^\bullet(M,\mathbb{R})\) and \(H^\bullet(T,\mathbb{R})\). The author's main result is the construction of an embedding \[ \Psi\colon H^\bullet(M,\mathbb{R})\hookrightarrow H^{\bullet + l }(T,\mathbb{R}) \] of \(\mathfrak{g}_{\mathrm{tot}}(M)\)-modules. Here \(l\geq 0\) is an integer which takes into account the difference of the dimensions of \(M\) and \(T\). Now \(\Psi\) is the aforementioned generalization of \((*)\). Moreover, they show that for any Hodge structure of hyper-Kähler type on \(H^\bullet(M,\mathbb{C})\), there is a complex structure on \(T\) such that \(\Psi\) becomes a morphism of Hodge structures. This follows from the fact that the \(\mathfrak{g}_{\mathrm{tot}}(M)\)-action encodes all Hodge structures of hyper-Kähler type on \(H^\bullet(M,\mathbb{C})\). Finally, the authors show that \(T\) is in fact an abelian variety if \(M\) is projective. In this sense, their construction beautifully generalizes the original Kuga-Satake construction to compact simply-connected hyper-Kähler manifolds.