Moduli of vector-bundles on surfaces

From MaRDI portal



Abstract: This is a survey paper: we discuss certain recent results, with some improvements. It will appear in the S. Cruz proceedings.


This article surveys the latest results about moduli spaces of vector bundles on an algebraic surface, providing mostly (sometimes partial) proofs. There is also a quite complete list of references, to which should be added the book by \textit{D. Huybrechts} and \textit{M. Lehn} [``The geometry of moduli spaces of sheaves (1997; Zbl 0872.14002)] which appeared after this article. Let \(S\) be a smooth projective complex algebraic surface and let \(H\) be an ample divisor on \(S\). There exists a projective moduli space \({\mathcal M}\) for \(H\)-semi-stable (in the sense of Gieseker-Maruyama) rank \(r\) torsion free sheaves on \(S\) with fixed Chern classes \(c_1\) and \(c_2\). Set \(\Delta=2rc_2-(r-1)c_1^2 \). It was proved by O'Grady and Gieseker-Li that for \(\Delta\) big enough, \({\mathcal M}\) is integral of the expected dimension, and the open set corresponding to stable vector bundles is dense. The proof of this result is sketched here and goes roughly as follows. Note that the choice of polarization is irrelevant, since moduli spaces for various polarizations are birational for \(\Delta\) big enough (Qin, Matsuki-Wentworth). We may fix the determinant; by standard deformation theory, \({\mathcal M}\) is smooth of the expected dimension at a point corresponding to a stable sheaf \(F\) if the space \(\Hom(F,F\otimes K_S)^0\) of traceless homomorphisms vanishes. The method consists in proving that the expected dimension of \({\mathcal M}\) is both bigger than the dimension of the space of non-stable sheaves, and than the dimension of the space \(W\) of stable sheaves with \(\Hom(F,F\otimes K_s)^0\neq 0\). The first point only requires a dimension count and is skipped here; the latter is the main result of the first section, where \(\dim(W)\) is shown to grow slower than the expected dimension of \({\mathcal M}\) with \(\Delta\). This is done by first proving that any subvariety of \({\mathcal M}\) with small codimension must have a non-locally free point. Now if \(W\) has small codimension, the double duals of its non-locally free points yield a subvariety \(W_1\) of small codimension of another moduli space \({\mathcal M}_1\) of sheaves with same rank and determinant but smaller \(c_2\); iterating the construction, it is shown that the numbers \(\dim(W_j)\) must increase except for a controlled number of \(j\)'s, until they get bigger than the dimension of the moduli space they live in (there is an upper bound on this dimension due to Simpson), giving a contradiction hence proving that the codimension of \(W\) in \({\mathcal M}\) is big. One proves irreducibility with an argument of Gieseker and Li: Fix a \(\Delta\) such that the corresponding moduli space is smooth of the expected dimension, with irreducible components \(X_1,\ldots,X_n\); for higher \(\Delta\), consider the irreducible subvariety \(Y_i\) of the moduli space \({\mathcal M}\) corresponding to sheaves whose double dual is in \(X_i\). It is proved that each component of \({\mathcal M}\) contains at least one of the \(Y_i\)'s and is smooth at its generic point; one then constructs an irreducible family of sheaves that contains all the \(Y_i\)'s to reach the conclusion. The second part is devoted to the construction of \(2\)-forms on a moduli space \({\mathcal M}\). Given a universal sheaf on \({\mathcal M}\times S\), one associates to its second Chern class, by the Mumford construction, a \(2\)-form on \({\mathcal M}\). This \(2\)-form is shown to be a multiple of the \(2\)-form introduced by Mukai and Tyurin, and to be generically non-degenerate if \({\mathcal M}\) has even dimension and \(S\) has a \(2\)-form whose zero-locus is either empty or a smooth irreducible curve of genus at least \(1\) (Mukai had treated in 1984 the case where \(\omega_S\) is trivial, in which case \({\mathcal M}\) is symplectic). The last statement actually depends on the validity of a conjecture which is known only for rank \(2\) and for certain degrees. Following an argument of Li, this implies that under these hypothesis, \({\mathcal M}\) is of general type (for large \(\Delta\)). For the entire collection see [Zbl 0882.00032].




Cited in
(31)








This page was built for publication: Moduli of vector-bundles on surfaces

Report a bug (only for logged in users!)Click here to report a bug for this page (MaRDI item Q4379980)