Automorphisms and periods of cubic fourfolds (Q2069635)

Automorphism groups of \(K3\) surfaces were studied by Nikulin and by Mukai via lattice theory, and later in a more systematic way by Kondo, and the ideas have been extended to hyperkähler manifolds of higher dimension by Mongardi and others. Making use of this and other recent work by Höhn and Mason on fixed sublattices of the Leech lattice [\textit{G. Höhn} and \textit{G. Mason}, J. Algebra 448, 618--637 (2016; Zbl 1360.11078)], this paper gives a complete classification of groups of symplectic automorphisms of cubic fourfolds, including in most cases the equations of the cubics in question. \par The link between the cubic fourfolds and the hyperkähler manifolds is of course the Fano variety \(F(X)\) of lines on a cubic fourfold \(X\), which is a hyperkähler fourfold. An automorphism of \(X\) is said to be symplectic if the automorphism of \(F(X)\) that it induces is symplectic: that is, it preserves the essentially unique \(2\)-form on \(F(X)\). Equivalently, a symplectic automorphism of \(X\) is one that acts trivially on \(H^{3,1}(X)\). \par The main result in this paper classifies almost completely pairs \((X,G)\) where \(X\) is a smooth cubic fourfold and \(G\) is its group of symplectic automorphisms. The groups \(G\) are completely classified (there are 34 of them) along with the associated covariant lattice \(S\), the orthogonal complement of the \(G\)-invariant part of \(H^4(X,\mathbb{Z})\). It turns out that \(S\) is a primitive sublattice of the Leech lattice \(\mathbb{L}\) and is determined by \(G\) up to isometry. Inequivalent embeddings of \(S\) in \(\mathbb{L}\) correspond to different irreducible components of the corresponding moduli space. In most cases it is also possible to describe the equation of \(X\), but for a few this is still an open question. \par The maximal cases, where \(S\) is of rank 20, are particularly interesting. There are six groups, and the corresponding moduli space of cubic fourfolds is of dimension 0: for four of the groups \(X\) is unique, but in two cases (the alternating group \(A_7\) and the Mathieu group \(M_{10}\)) there is a second way to embed \(S\) in \(\mathbb{L}\) and hence a second, mysterious fourfold. \par The proofs rely on the Torelli theorem, to reduce the problem to one of lattice theory. Then one applies the idea that one should study the lattices by embedding them in the lattice \(2U\oplus 3E_8\) of signature \((2,26)\), here called the Borcherds lattice. This, which the authors call Borcherds polarisation, has been used in related contexts by several people but perhaps has not been clearly articulated before now in this generality. Note that \(3E_8\) can be replaced by any (negative-definite) Niemeier lattice, in particular the Leech lattice. Thus one should expect that the groups \(G\) will be related to the automorphism group of the Leech lattice, the Conway group \({\mathop{\mathrm {Co}}}_0\). The lattice theory part relies on the classification of Höhn and Mason and its relation to geometry is effected using Mongardi's notion of Leech pair. \par Additionally, the authors supply more detail in some cases and investigate some relations with \(K3\) surfaces of degrees 2 and 6. They also add some remarks about the full (not just symplectic) automorphism groups, which in particular can distinguish between the two fourfolds whose symplectic automorphism group is \(A_7\).