Abstract. We prove existence of reflexive sheaves on singular surfaces and threefolds with prescribed numerical invariants and study their moduli.
MotivationSheaves on singular varieties have become very popular recently because of their appearance in Physics, String Theory and Mirror Symmetry. In particular, many open questions about sheaves on singular varieties have come to light. The corresponding mathematical tools, however, are waiting to be developed. Our aim in this paper is to entice singularists to develop some basic techniques needed to approach such questions. It is extremely common for a physicist or string theorist to start up a lecture by giving a partition function for a theory, and now even algebraic geometers are quite often doing the same. It is not just a fashion, but the fact is that this is an extremely efficient way to present results. The general format of such partition functions is of an infinite sum whose terms contain integrals over moduli spaces. Here are some examples. We will not need details from these expressions, just the observation that they all contain integrals over moduli spaces.Example 1.1. (String Theory) The Nekrasov partition function for N = 2 supersymmetric SU (r) pure gauge theory on a complex surface X is given by an expression of the formwhere M r,d,n (X) is the moduli space of framed torsion-free sheaves or rank r, and Chern classes c 1 = d and c 2 = n. For the case of gauge theories with matter, one writes a similar expression but with more interesting integrands, see [GL]. Example 1.2. (Donaldson-Thomas Theory) For a Calabi-Yau threefold X, the partition function for Donaldson-Thomas theory is given by: where M g (X, β) is the moduli space of genus-g curves representing the class β ∈ H 2 (X, Z).There is a precise sense in which this partition function is equivalent to the one in Example 1.2, see [MNOP].These examples illustrate the appearance of integrals over moduli spaces of sheaves. Even in the case of moduli spaces of maps of Example 1.3 the theory is still related to a theory given by integration over moduli of sheaves. Observe that the definition of moduli spaces itself requires a choice of numerical invariants: in Example 1.1 the Chern classes and in Example 1.2 the Euler characteristic. So, we now agree that we are interested in moduli spaces of sheaves on surfaces and threefolds. Of course, the physics motivation is just a bonus, and we could have been interested in such moduli spaces for purely geometric reasons, as they are part of classical algebraic geometry. Now physics dictates that we should consider theories defined over singular varieties. In fact, some of the most popular categories considered currently by physicists and string theorists turn out empty in the absence of singularities; such is the case of the FukayaSeidel category and the Orlov category of singularities. Thus we arrive at the conclusion that we need to understand moduli of sheaves on singular varieties. Both the case of global moduli of sheaves on projective varieties and the ...