Moduli spaces of bundles over nonprojective \(K3\) surfaces (Q522113)

It is a well-known result due to Mukai that the moduli space \(M^\mu_v(S,H)\) of \(\mu_H\)-stable sheaves of Mukai vector \(v=(r,c_1(L),s)\) with respect to an ample divisor \(H\) on a \(K3\) surface \(S\) carries a holomorphic symplectic form [\textit{S. Mukai}, Invent. Math. 77, 101--116 (1984; Zbl 0565.14002)]. Moreover, if \(H\) is generic and \(\gcd(r,L)=1\), then there are no strictly \(\mu_H\)-semistable sheaves, so \(M^\mu_v(S,H)\) is in fact a projective homomorphically symplectic manifold. These moduli spaces are also known to be irreducible hyper-Käher manifold deformation equivalent to a Hilbert scheme of points on \(S\) [\textit{K. G. O'Grady}, J. Algebr. Geom. 6, No. 4, 599--644 (1997; Zbl 0916.14018)]. If the \(K3\) surface \(S\) is nonprojective, and thus does not carry an ample divisor, one can still define the notion of slope-stability with respect to a Käher class \(\omega\) and obtains a similar moduli space \(M^\mu_v(S,\omega)\) which is a smooth complex manifold carrying a holomorphic symplectic form. Moreover, if \(\omega\) is generic (in an appropriate sense) and \(\gcd(r,c_1(L))=1\), then \(M^\mu_v(S,\omega)\) is compact, in analogy with the projective case. It is thus natural to ask whether the analogy continues with regard to \(M^\mu_v(S,\omega)\) being irreducible symplectic, the subject considered in the paper under review. In their main theorem, the authors' prove that for \(\omega\) generic with respect to \(v=(r,c_1(L),s)\in H^*(S,\mathbb Z)\), which satisfies \(\gcd(r,c_1(L))=1\) and \(v^2\geq 0\), the moduli space \(M^\mu_v(S,\omega)\) is a compact, connected homomorphically symplectic complex manifold of dimension \(v^2+2\) deformation equivalent to the Hilbert scheme of points on a projective \(K3\) surface. They show, moreover, that in the general case \(H^2(M^\mu_v(S,\omega),\mathbb Z)\) continues to carry a nondegenerate quadratic form making it isometric to \(v^{\perp}\) if \(v^2>0\) (resp., \(v^{\perp}/\mathbb Z v\) if \(v^2=0\)). A consequence of their result is the second main result of the paper, stating that under the same assumptions the moduli space \(M^\mu_v(S,\omega)\) is projective if and only if the underlying \(K3\) surface \(S\) is. The authors' argument entails dealing with two cases. In the first case, \(S\) is assumed to be projective while \(\omega\) is only Kähler, and the authors show that \(\omega\) lies in the same chamber as ample divisor \(H\) so that \(M^\mu_v(S,\omega)=M^\mu_v(S,H)\), where all of the claims of the theorem are already known. The second, and more novel, case assumes \(S\) to be nonprojective. To establish the result in this case the authors deform \(M^\mu_v(S,\omega)\) along the twistor family \(\mathcal X\to\mathbb P^1\) associated to the pair \((S,\omega)\). Although the sheaves parametrized by \(M^\mu_v(S,\omega)\) do not necessarily deform in this family, by generalizing Yoshioka's construction of a relative moduli space of twisted sheaves [\textit{K. Yoshioka}, Adv. Stud. Pure Math. 45, 1--30 (2006; Zbl 1118.14013)] the authors connect \(M^\mu_v(S,\omega)\) to \(M^\mu_v(S',\omega')\) for generic polarization \(\omega'\) on a projective \(K3\) surface \(S'\). An upshot of these twistor and twisted sheaf techniques is that the main theorem remains true for moduli of twisted sheaves as well. Although the authors' techniques can be used to provide another proof that the locally free locus \(M^{\mu-lf}_v(S,\omega)\) admits a Kähler metric, it is still an open problem to determine if \(M^\mu_v(S,\omega)\) is Kähler, and thus a compact irreducible hyperKähler manifold, when \(S\) is nonprojective. Nevertheless, the authors show that for any positive even dimension \(d\) there exists a \(K3\) surface \(S\) and Mukai vector \(v\) such that \(M^\mu_v(S,\omega)=M^{\mu-lf}_v(S,\omega)\) is a compact irreducible hyperKähler manifold of dimension \(d\).