Special components of Noether-Lefschetz loci (Q2047462)

For any \(d \ge 4\), the \textit{Noether-Lefschetz locus} \(\mathrm{NL}_d\) parametrizes smooth degree \(d\) surfaces in \(\mathbb{P}^3\) with Picard number at least \(2\). Denote by \(U_d\) the space parameterizing smooth degree \(d\) surfaces in \(\mathbb{P}^3\). It is known that \(\mathrm{NL}_d\) is the union of countably many irreducible subvarieties of \(U_d\). Moreover, the codimension of any irreducible component of \(\mathrm{NL}_d\) is bounded above by \(\binom{d-1}{3}\). If an irreducible component of \(\mathrm{NL}_d\) is not of codimension \(\binom{d-1}{3}\), it is called \textit{special}. Harris conjectured that \(\mathrm{NL}_d\) has only finitely many special components. In this article, the author gives a conjectural description of an infinite number of reduced special components of \(\mathrm{NL}_d\) for \(d=8\). The idea is to choose a smooth degree \(d\) surface \(X_0\) in \(\mathbb{P}^3\) and two curves \(C_1, C_2\) in \(X_0\) such that their cohomology classes are linearly independent in the primitive cohomology. For any \(r \in \mathbb{Q}\), denote by \(\mathrm{NL}([C_1]+r[C_2])\) the Hodge locus associated to the Hodge class \([C_1]+r[C_2]\). The author introduces the notion of \textit{special pencil of Noether-Lefschetz loci}. This is a family of Hodge loci of the form \(\mathrm{NL}([C_1]+r[C_2])\) satisfying certain conditions, as \(r\) varies over \(\mathbb{Q}\). The key criterion necessary for the family of Hodge loci to be a special pencil is that for all but finitely many \(r \in \mathbb{Q}\), \(\mathrm{NL}([C_1]+r[C_2])\) is smooth. One observes that if a special pencil of Noether-Lefschetz loci exists then it will give rise to an infinite number of reduced special components of \(\mathrm{NL}_d\), thereby giving a counterexample to the Harris conjecture. The author focuses on the case when \(X_0\) is a Fermat surface and \(d=8\). He gives criterion on \(C_1, C_2\) and \(r\) under which \(\mathrm{NL}([C_1]+r[C_2])\) is singular, hence cannot give rise to a special pencil of Noether-Lefschetz loci. In the positive direction, the author gives an example of \(C_1, C_2\) such that for \(r=r_1/r_2\) with \(0 \le |r_1| \le 10\) and \(1 \le r_2 \le 10\), we have \(\mathrm{NL}([C_1]+r[C_2])\) is smooth upto \(5\)-th order i.e., the \(5\)-th order infinitesimal deformation of \(\mathrm{NL}([C_1]+r[C_2])\) coincides with the \(5\)-th order infinitesimal deformation of a smooth variety. One hopes that such a family of Hodge loci will give rise to a special pencil of Noether-Lefschetz loci.