The irreducible components of moduli spaces of vector bundles on surfaces (Q1319244)

Let \(S\) be a smooth projective surface over an algebraically closed field \(k\) of characteristic zero. Fix an ample divisor \(H\) on \(S\), a divisor \(D\) on \(S\) and an integer \(c\). Denote by \({\mathcal M} (c) = {\mathcal M} (S,H;D,c)\) the moduli space of \(H\)-slope stable rank-two vector bundles on \(S\) with \(c_ 1 = {\mathcal O}_ S (D)\) and \(\deg (c_ 2) = c\). Similarly, \({\mathcal M}^ G (c) = {\mathcal M}^ G (S,H;D,c)\) denotes the moduli space of \(H\)-Gieseker semistable rank-two torsion free sheaves on \(S\) with \(c_ 1\) and \(c_ 2\) fixed as above. Then \({\mathcal M} (c)\) is an open subscheme of the projective scheme \({\mathcal M}^ G (c)\). The deep understanding of \({\mathcal M}^ G (c)\) for arbitrary \(S\) begins with Donaldson's generic smoothness result [\textit{S. K. Donaldson}, Topology 29, No. 3, 257--315 (1986; Zbl 0715.57007)]. Later, J. Li proved that \({\mathcal M}^ G (c)\) is normal and has local complete intersection singularities at stable sheaves provided \(c\) is large. He also showed that when \(S\) is a surface of general type satisfying some mild technical conditions, then \({\mathcal M}^ G (c)\) is of general type for \(c \gg 0\) [\textit{J. Li}, Invent. Math. 115, No. 1, 1--40 (1994; Zbl 0799.14015)]. The main theorem of the paper under review states that for fixed \(S,H\) and \(D\) the number of irreducible components of \({\mathcal M} (c)\) stabilizes for large \(c\). [Compare with the paper by \textit{D. Gieseker} and \textit{J. Li}, J. Differ. Geom. 40, No. 1, 23--104 (1994), where the stronger result is shown that \({\mathcal M} (c)\) is irreducible for large \(c\).] For the proof the author studies determinant line bundles on moduli spaces. Furthermore, he studies the restriction of semi-stable sheaves to a curve \(C\) in \(S\) and degenerations of elementary modification along \(C\) of \(H\)-slope stable bundles on \(S\). All this is put together to show that (for large \(c)\) each component of \({\mathcal M}^ G (c)\) meets the boundary (corresponding to non locally free sheaves) of the moduli space. Using this and counting parameters, the author proves for large \(c\) that \({\mathcal M} (c)\) is dense in \({\mathcal M}^ G (c)\). Hence, both spaces have the same number of irreducible components. Associating to a sufficient general boundary point \(F\) of \({\mathcal M}^ G (c)\) the double dual \(F^{\vee \vee}\), the author obtains a point in \({\mathcal M} (c - 1)\). On the level of irreducible components, this construction has an inverse: Let \(X\) be an irreducible component of \({\mathcal M}^ G (c - 1)\). If \(E\) represents a sufficiently general point of \(X\) and \(f : E \twoheadrightarrow k_ P\) is a surjection onto a skyscraper sheaf on \(S\), \(F : = \ker f\) is a smooth point of \({\mathcal M}^ G (c)\) and \(F^{\vee \vee} \cong E\). In this way the author obtains (for large \(c)\) a surjection of the set of irreducible components of \({\mathcal M}^ G (c - 1)\) onto the set of irreducible components of \({\mathcal M}^ G (c)\). This implies that the number of components of \({\mathcal M}^ G (c)\) stabilizes.