2 9 A ug 2 00 1 Cohomology Ring of Crepant Resolutions of Orbifolds
Yongbin Ruan
2008-01-01
Abstract:Suppose that X is an orbifold. In general, KX is an orbifold vector bundle or a Q-divisor only. When the X is so called Gorenstein, KX is a bundle or a divisor. For Gorenstein orbifold, a resolution π : Y → X is called a crepant resolution if πKX = KY . Here, ”crepant” can be viewed as a minimality condition with respect to canonical bundle. Crepant resolution always exists when dimension is two or three. A nice way to construct it is to use Hilbert scheme of points. However, the crepant resolution in dimension three is not unique. Different crepant resolutions are connected by flops. When the dimension is bigger than four, the crepant resolution does not always exist. It is an extremely interesting problem in algebraic geometry to find out when it does exist. One of famous example is Hilbert scheme of points of algebraic surfaces, which is a crepant resolution of the symmetry product of an algebraic surface. It has long been an interesting problem to compute cohomology of Hilbert scheme of points of algebraic surfaces. A great deal of works has been done by algebraic geometors. As far as the author’s knowledge, everything has been computed except its ring structure. Among all the progress, a less known but crucial ingredient is the link to the orbifold cohomology of the symmetry product. Suppose that M is an algebraic surface. We use M [n] to denote the Hilbert scheme of points of length n of M . In 1994, in order to explain the modularity of the generating function ∑∞ n=1 χ(M [n])qn computed by Göttsche [G], Vafa-Witten [VW] computed ⊕nH (M/Sn,C). Motivated by orbifold conformal field theory, they directly wrote ⊕nH (M/Sn,C) as a ”Fock space” or a representation of Heisenberg algebra. Then, the generating function of Euler characteristic is interpreted as the correlation function of an elliptic curve. Therefore, it should be invariant under modular transformations of the elliptic curve. Orbifold string theory conjecture predicates that ⊕nH ∗(M [n],C) should also admit a representation of Heisenberg algebra. This conjecture was verified by a beautiful work of Nakajima [N] and others. One of theme of this short note is that the orbifold M/Sn will continue to play a crucial role to compute ring structure of H∗(M [n],C). During last two years, there was a surge of activities to study mathematics of the orbifold string theory, which author called the stringy orbifolds. A few curious physical concepts such as the orbifold Euler-Hodge numbers of global quotients found their places in a much broader and deeper theory. For example, a new cohomology (orbifold cohomology) was constructed [CR1]. The growth was so explosive that the author believes that there is an emerging new subject of mathematics. He learned in graduate school that the test of the relevance of a new theory has been the progress it made from old problems. Therefore, it is particularly significant to revisit the problem of computing the ring structure of M [n] and a crepant resolution in general. In [R2], the author proposed a conjecture in the case of a hyperkahler resolution.