Maximal variation of curves on \(K3\) surfaces (Q2097277)

The paper studies the maximality of the variation of smooth curves in the linear system \(|L|\) with \(L\) an ample line bundle on a \(K3\) surface \(S\). Mapping a point \(s\in |L|\) corresponding to a smooth curve \(C\) to the corresponding point \([C]\in M_g\) in the Deligne-Mumford stack of smooth curves of genus \(g\), with \(g=\frac{L^2}2+1\), describe a rational map \[ \Phi:|L|\dashrightarrow M_g. \] The curves in \(|L|\) have maximal variation if \(\Phi\) is generically quasifinite, or equivalently the differential \[ d\Phi:H^0(C,L|_C)\rightarrow H^1(C,T_C) \] is injective. The main result of the paper states that for \(L=mH,m\geq 2\), with \(|H|\) base point free ample linear system such that the locus of nonreduced curves in \(|H|\) has codimension at least three, the curves in \(|L|\) have maximal variation. The strategy is to show that \(H^0(C,T_S|_C)=0\) for a generic smooth curve \(C\) in \(|L|\), since \(H^0(C,T_S|_C)\) is exactly the relative Zariski tangent space \(T_{\Phi}(C)\) of \(\Phi\) at the point \(C\in |L|\). If \(C\) is nonhyperelliptic, then \(T_S|_C\) is semistable for generic \(C\) in \(|H|\) by Hein's result and semistability of \(T_S\). If \(T_S|_C\) is stable, \(H^0(C,T_S|_C)=0\) follows immediately. If \(T_S|_C\) is strictly semistable or if \(C\) is hyperelliptic, then one still have \(H^0(C,T_S|_C)\leq 2\) and \(H^0(C,T_S|_C(-C))=0\), which can imply \(H^0(C',T_S|_{C'})=0\) for \(C'\) in \(|mH|\) with \(m\geq 2\). The assumption on the locus of nonreduced curves in \(|H|\) can be removed for \(m\geq 3\). The paper also provides another approach to the main result: using the comparison with the Hitchin system. More precisely, specialization from curves \(D\in |mH|\) on a \(K3\) surface to spectral curves \(D\twoheadrightarrow C\) with \(C\) a chosen smooth curve in \(|H|\) yields that \(\dim \Phi^{-1}([D])\leq g+1\). Combining this equality with a result of van Geemen and Voisin on Matsushita's conjecture, one sees directly the general quasifinity of \(\Phi.\) The paper is very well written, such that the readers should feel not difficult to understand all the details of the argument. The authors are considerate to add many remarks explaining the relation between their main result and other existing results in the literature.