A proof of the Donaldson-Thomas crepant resolution conjecture (Q2152482)

Donaldson-Thomas (DT) invariants are deformation-invariants defined to virtually enumerate stable objects in the derived category of coherent sheaves on a smooth projective threefold. Motivated by analogues phenomena in Gromov-Witten Theory, \textit{J. Bryan} et al. [Adv. Math. 229, No. 1, 531--595 (2012; Zbl 1250.14027)] conjectured a series of relations between curve-counting invariants of a Calabi-Yau 3-orbifold and the crepant resolution of its coarse moduli space. Notice that, for a Calabi-Yau 3-orbifold \(\mathcal{X}\) (that is, a 3-dimensional smooth Deligne-Mumford stack with trivial canonical bundle and vanishing first cohomology), there always exists a distinguished crepant resolution \(Y\to X\) of its coarse moduli space \(X\), given étale locally by Nakamura's \(G\)-Hilbert scheme. Moreover, \(Y\) and \(\mathcal{X}\) are derived equivalent, by the \textit{derived McKay equivalence} of \textit{T. Bridgeland} et al. [J. Am. Math. Soc. 14, No. 3, 535--554 (2001; Zbl 0966.14028)]. Given a Calabi-Yau 3-orbifold \(\mathcal{X}\), there are two main moduli spaces of interest for this paper: the \textit{Hilbert scheme of points and curves} on \(\mathcal{X}\), which parametrizes quotients \(\mathcal{O}_{\mathcal{X}}\to \mathcal{O}_Z\), where \(Z\) is a substack with fixed class \([\mathcal{O}_Z]\), such that its support is at most 1-dimensional, and the \textit{moduli space of stable pairs}, which parametrizes complexes \([\mathcal{O}_{\mathcal{X}}\to F]\), where \(F\) is a pure 1-dimensional sheaf and the section has 0-dimensional cokernel. In both cases, one defines DT and Pandharipande-Thomas (PT) invariants by the use of \textit{K. Behrend}'s function [Ann. Math. (2) 170, No. 3, 1307--1338 (2009; Zbl 1191.14050)], and defines the generating series \(DT(\mathcal{X})_\beta, DT(\mathcal{X})_0, PT(\mathcal{X})_\beta\), where \(\beta\) is a curve class. The authors prove a series of remarkable results on the generating series of DT and PT invariants. The first result is that the generating series of PT invariants is a \textit{rational function} -- for a suitable notion of Laurant expansion of a rational function -- whose possible poles are classified. Moreover, a symmetry property is proved for \(PT(\mathcal{X})_\beta\). Secondly, they prove an orbifold version of the classical DT/PT correspondence proved by Bridgeland, which reads \[ PT(\mathcal{X})_\beta=\frac{DT(\mathcal{X})_\beta}{DT(\mathcal{X})_0}. \] Finally, they prove the \textit{Donaldson-Thomas crepant resolution conjecture}, formulated in terms of Bryan-Steinberg relative stable pair invariants \[ PT(\mathcal{X})_\beta=BS(Y/X)_\beta, \] where \(Y\to X\) is the crepant resolution of the coarse moduli space. In the original formulation of Bryan-Cadman-Young, the latter was formulated in terms of DT and PT invariants. The main tool exploited to prove all these results is Joyce's wall-crossing machinery, with all the adaptations needed to take care of the stackiness of the problem.