Variety of power sums and divisors in the moduli space of cubic fourfolds (Q526150)

The authors consider two different hyperkähler fourfolds associated to a general cubic fourfold \(F\subset\mathbb P(W)\), where \(\dim(W)=6\). The first fourfold is the Fano variety \(L(F)\) of lines in \(F\). The second is the variety of sum of powers \(\mathrm{VSP}(F)\) defined by \(F\) in the Hilbert scheme of sets of length \(10\) in \(\mathbb P(W)\). Indeed, a general cubic fourfold has Waring rank equal to \(10\) and \(\mathrm{VSP}(F)\) is the (closure of the) set of points \(\ell_1,\dots,\ell_{10}\) giving a decomposition \(F=\ell_1^3+\dots +\ell_{10}^3\). The two hyperkähler manifolds have Hodge structures which are related to the geometry of \(F\), and it is natural to ask for direct relations between them. The answer is surprisingly negative: the authors show that, for a general \(F\), all morphisms between \(H^4(F,\mathbb Q)_{prim}\) (which is isomorphic to \(H^2(L_F,\mathbb Q)\)) and \(H^2(\mathrm{VSP}(F),\mathbb Q)\) are trivial. The result is obtained by analyzing the relations between \(\mathrm{VSP}(F)\) and the variety of decompositions \(\mathrm{VSP}(C)\) of a general sestic plane curve \(C\), which is a \(K3\) surface. Namely, if one identifies \(W\) with \(S^2(V)\), \(V\) being a space of dimension \(3\), then the dual of the multiplication \(S^3(S^2(V^*))\to S^6(V^*)\) determines a map \(s:S^6(V)\to S^3(W)\). The image of this map is a divisor \(D\) in the space of cubic fourfolds. \(F\) belongs to \(D\) precisely when the derivative associated to \(F\) kills all the cubic forms in the ideal of a Veronese surface in \(\mathbb P^5\). For these cubics, apolar to a Veronese surface, the variety \(\mathrm{VSP}(F)\) can be analized closely. It turns out that it naturally contains \(\mathrm{VSP}(C)\) as its singular locus. As a result of the analysis, the authors prove that \(D\) induces a Noether Lefschetz divisor for the family of varieties of sum of powers over an open subset of \(\mathbb P(S^3(W))\). On the other hand, the authors prove that \(D\) is not a Noether Lefschetz divisor in the moduli space of cubic fourfolds. In their study on the decomposition of cubics in \(6\) variables, the authors also prove a result of independent interest (Lemma 3.25): a general cubic of rank \(9\) has two different minimal Waring decompositions of length \(9\).