Change of polarization and Hodge numbers of moduli spaces of torsion free sheaves on surfaces (Q5961425): Difference between revisions

To define a reasonable moduli space for torsion-free sheaves over a smooth projective complex surface \(S\), one needs a concept of semi-stability, such as \(L\)-semi-stability in the sense of Gieseker and Maruyama, for \(L\) an ample line bundle on \(S\). As Qin showed, the resulting moduli space depends on the choice of the polarization \(L\). The aim of this paper is to analyze these changes in order to calculate the Hodge numbers of certain moduli spaces. Specifically, let \(M_L(c_1,c_2)\) be the moduli space of \(L\)-semi-stable torsion free sheaves on \(S\) of rank \(2\) and Chern classes \(c_1\in\text{Pic} (S)\) and \(c_2\in H^4(S,{\mathbb{Z}})\). There is a chamber structure on the ample cone of \(S\) such that \(M_L(c_1,c_2)\) depends only on the chamber of \(L\). The author refines (and corrects) Qin's results and calculates the change in the Hodge numbers of \(M_L(c_1,c_2)\) as \(L\) crosses a wall. The method is the following: The \(M_L(c_1,c_2)\)'s have a common open set (possibly empty) whose complement is a disjoint union of projective bundles over schemes of the type \[ \text{Pic} (S)\times \text{Hilb}^m (S)\times \text{Hilb}^n (S) \] for various \(m\) and \(n\). One then uses the fact that virtual Hodge polynomials \[ h(X;x,y)= \sum_{p,q} (-1)^{p+q}h^{p,q}(X) x^py^q \] behave nicely to reduce the calculation to that of \(h(\text{Hilb}^n (S);x,y)\), which was done by \textit{L. Göttsche} and \textit{W. Soergel} [Math. Ann. 296, No. 2, 235-245 (1993; Zbl 0789.14002)]. For \(S\) a \(K3\) or abelian surface, and \(c_1\) not divisible by \(2\), one finds that the Hodge numbers of \(M_L(c_1,c_2)\) are independent of \(L\), as long as \(L\) does not lie on a wall. When \(S\) is a ruled surface, the Hodge numbers of \(M_L(c_1,c_2)\) are determined by starting from a chamber for which this space is empty and summing up the wall-changing formulas obtained above. Since the paper under review was written, it was proved by \textit{G. Ellingsrud} and \textit{L. Göttsche} [J. Reine Angew. Math. 467, 1-49 (1995; Zbl 0834.14005)] that in certain cases, the change in the moduli space \(M_L(c_1,c_2)\) as \(L\) crosses a wall is a sequence of blow-ups and blow-downs.