Curve counting and DT/PT correspondence for Calabi-Yau 4-folds (Q2213772)

Donaldson-Thomas and stable pair (or Pandharipande-Thomas) invariants are curve counting invariants defined as intersection numbers on the Hilbert scheme of 1-dimensional subschemes or the moduli space of stable pairs. For a Calabi-Yau threefold, they are the degree of a zero-dimensional virtual fundamental class (in the sense of Behrend-Fantechi). Stable pair invariants were introduced by Pandharipande-Thomas to give a geometric explanation of the MNOP conjectures, relating Gromov-Witten invariants and Donaldson-Thomas invariants. The relation between the two invariants can be summarized by the slogan \[ DT=DT_0\cdot PT,\] where \(DT_0\) is the generating series of 0-dimensional Donaldson-Thomas invariants. The DT/PT correspondence in its original form was proven by Toda (using wall-crossing arguments) and by Bridgeland (using Hall algebra techniques). In this paper, the authors formulate various ``DT/PT correspondences'' for Calabi-Yau 4-folds. Two types of DT (or PT) invariants are mainly considered here. If \(X\) is a projective Calabi-Yau 4-fold, by the seminal work of Cao-Leung and Borisov-Joyce there exist a virtual fundamental class (which does not come from a perfect obstruction theory) and invariants are defined by integrating primary insertions against it. In this setting, the conjectured DT/PT correspondence can be read as \[ DT=PT.\] In fact, differently from the 3-fold theory, the DT/PT correspondence does not see the correction of ``floating points'', since the Hilbert scheme of points has in general positive virtual dimension. The question becomes more interesting for a toric Calabi-Yau 4-fold (which is never projective). Despite invariants cannot be defined directly (as to the moduli space is in general not proper) the authors, whenever the torus fixed locus of the moduli space is proper, define invariants by a virtual localization formula. To do this, they carefully describe the torus fixed locus of the Hilbert scheme of curves (which is always 0-dimensional and reduced) and of the moduli space of stable pairs (which is 0-dimensional and reduced only under suitable assumptions). Both fixed loci are described in terms of (possibly infinite) solid partitions obeying prescribed rules. The ``localized'' DT invariants, for a general toric Calabi-Yau 4-fold, are reduced to the case of \(\mathbb{C}^4\). In this case, the authors develop a vertex formalism that computes DT and PT invariants purely in terms of the combinatorics of solid partitions. The equivariant DT (and similarly the PT) 4-fold vertex is defined as \[ V^{DT}_{\lambda \mu \nu \rho}(q)=\sum_{\pi}(-1)^{o^{DT}_\pi}\sqrt{(-1)^{a_\pi}e_T(-V_\pi^{DT})}q^{|\pi|}\in \mathbb{Q}(\lambda_1, \lambda_2, \lambda_3)((q)) \] where \( V_\pi^{DT}\) is the vertex term and \((-1)^{o^{DT}_\pi}\) is a sign assigned to each solid partition. The toric DT/PT conjecture predicts that there exists precisely a unique choice of signs \((-1)^{o^{DT}_\pi},(-1)^{o^{PT}_\pi} \) such that \[ V^{DT}_{\lambda \mu \nu \rho}(q)=V^{DT}_{\emptyset\emptyset\emptyset\emptyset}(q)\cdot V^{PT}_{\lambda \mu \nu \rho}(q) \] in complete analogy with the 3-fold theory. This sign should be really seen as depending on the orientation of the moduli space and no formulae are known for them. Both type of conjectures are supported by several examples ( of geometric nature in the case of projective Calabi-Yau 4-folds and of combinatorial nature in the toric case) and several computations of DT and PT invariants. Finally, the authors prove a virtual localization formula in some special cases.