Curves on surfaces with trivial canonical bundle (Q1749766)

In the present survey, the author focuses on the so-called Severi varieties and variation in moduli of curves lying in \(K3\) or abelian surfaces. For a smooth irreducible projective surface \(S\) and a polarization \(L \in \mathrm{Pic}(S)\) we consider curves in the linear system \(| L|\). Let \(\delta \in\mathbb Z_{\geq 0}\) be a fixed integer, the Severi variety of \(\delta\)-nodal curves in \(| L|\) is the locally closed subscheme \[ | L|_\delta:= \{C \in| L|:C \text{ is integral and has }\delta\text{ nodes as its only singularities}\}. \] The expected dimension of \(| L|_\delta\) is equal to \(\dim| L|-\delta\), and we say that \(| L|_\delta\) is regular if it is smooth and of the expected dimension. In the case when \(\omega_S=\mathcal O_S\) one can show that \(| L|_\delta\) is regular as soon as it is non-empty. After presenting several interesting results on \(| L|_\delta\) in the case of polarized \(K3\) surfaces and abelian surfaces the author formulates some open problems, one of the most interesting (at least from my very subjective point of view) is the following. Problem 1. Given a general abelian surface \(S\) with polarization \(L\) of type \((n_1,n_2)\) with \(n_1>1\), when is the Severi variety \(| L|_\delta\) non-empty? Next, the author studies variation in moduli of curves lying on \(K3\) or abelian surfaces. The key problem is to determine the dimension of the family of \(K3\) surfaces containing the same curve as the (canonical) hyperplane section, and this problem is strictly connected with the Gaussian map of the curve itself. Let us recall that the moduli space of polarized \(K3\) surfaces of genus \(p\) is irreducible of dimension \(19\). We denote by \(\mathcal P_p\) the parameter space for pairs \(((S,L),C)\) where \((S,L) \in \mathcal F_p\) and \(C \in| L|\) is smooth and irreducible, along with the natural forgetful morphisms \[ f_p:\mathcal P_p\longrightarrow \mathcal M_p, \] \[ q_p:\mathcal P_p\longrightarrow \mathcal F_p. \] The map \(q_p\) realizes \(\mathcal P_p\) as an open subset of a \(\mathbb P^p\)-bundle over \(\mathcal F_p\), while \(f_p\) sends \(((S,L),C)\) to the class of \(C\) in the moduli space \(\mathcal M_p\) of genus \(p\) curves. It can be observed that \(\dim\mathcal P_p=19+p\geq\dim\mathcal M_p=3p-3\) if and only if \(p \leq 11\), so the last inequality is a necessary condition for a general curve of genus \(p\) to lie in a \(K3\) surface. It turns out that this condition is also sufficient by a result due to \textit{S. Mori} and \textit{S. Mukai} if only \(p \leq 11\) and \(p \neq 10\) -- in this case \(f_p\) is dominant [Lect. Notes Math. 1016, 334--353 (1983; Zbl 0557.14015)]. Moreover, \textit{C. Ciliberto} et al. showed that \(f_{p}\) is birational onto its image for \(p\geq 13\) and \(p=11\) [Invent. Math. 114, No. 3, 641--667 (1993; Zbl 0807.14028)]. Another interesting part of the survey is devoted to Mukai's program [Lect. Notes Pure Appl. Math. 179, 189--197 (1996; Zbl 0884.14010)] for curves in \(K3\) surfaces. Let \(C\) be a general curve of genus \(11\) and consider the moduli space \(M_C(2,K_C)\) of semistable rank two vector bundles on \(C\) with canonical determinant. The Brill-Noether locus \(M_C(2,K_C,5)\) of vector bundles in \(M_C(2,K_C)\) with a space of global sections of dimension at least \(5\) turns out to be a Fourier-Mukai transform of the unique \(K3\) surface in which \(C\) lies. Mukai also suggested that a similar procedure should make it possible to reconstruct a general polarized \(K3\) surfaces of any genus \(p \equiv 3\mathrm{mod }4\) starting from a general hyperplane section of it. This path was explored recently by \textit{E. Arbarello} et al. if only \(p\geq 15\) [Algebr. Geom. 1, No. 5, 532--557 (2014; Zbl 1322.14062)] and it turned out that Mukai's prediction was correct. These considerations lead to the following question/problem. Problem 2. Find a generalization of Mukai's program for all genera \(p\geq 13\) and \(p=11\).