Higher rank stable pairs and virtual localization (Q530649)

The article under review generalizes to higher rank the notion of stable pairs introduced by \textit{R. Pandharipande} and \textit{R. P. Thomas} [Invent. Math. 178, No. 2, 407--447 (2009; Zbl 1204.14026)]. Recall that on a smooth Calabi-Yau threefold \(X\) a pair \((F,s)\) consisting of a one-dimensional sheaf \(F\) and a section \(s\in H^0(X,F)\) is called stable if \(F\) is pure (that is, it has no non-zero subobjects of smaller dimension) and if \(\mathrm{Coker }s\) has zero-dimensional support when \(s\) is considered as a morphism \(s:\mathcal O_X\to F\). Stable pairs are roughly sheaf theoretic objects representing a system of curves on \(X\) and points on those curves, and they have sparked great interest since their introduction because they allow for the application of the robust machinery of the deformation theory of complexes in the derived category \(\mathrm{D}^b(X)\) to important enumerative problems for Calabi-Yau threefolds. For example, the proof of the MNOP conjecture for complete intersections [\textit{R. Pandharipande} and \textit{A. Pixton}, J. Am. Math. Soc. 30, No. 2, 389--449 (2017; Zbl 1360.14134)] shows that calculating the far easier stable pair invariants is equivalent to calculating Gromov-Witten invariants, which have proven to be much more difficult to actually calculate explicitly. In the present article, the author generalizes the setup by considering triples of type \((P_E,P_F)\) which are defined as triples \((E,F,\phi)\) where \(E\) and \(F\) are coherent sheaves with fixed Hilbert polynomials \(P_E\) and \(P_F\), respectively, \(F\) is a pure one-dimensional sheaf, and \(\phi\in\mathrm{Hom}(E,F)\). The paper mostly focuses on \textit{highly frozen triples} of type \((r,P_F)\) which are triples in which \(E\) is ``frozen'' such that \(E\cong\mathcal O_X(-n)^{\oplus r}\) via some morphism \(\psi: E\to \mathcal O_X(-n)^{\oplus r}\) that is part of the data. Forgetting the isomorphism \(\psi\) from the data gives the notion of a \textit{frozen triple}, and the stability condition in both cases is again that \(\phi\) has zero-dimensional cokernel. The author goes on to determine the deformation theory of these two types of triples and shows, in particular, that the deformation space of frozen triples is the same as that of the associated fixed-determinant 2-term complex \(\mathcal O_X(-n)^{\oplus r}\to F\) , at least for large enough \(n\), while the deformation space of highly frozen triples is different. Nevertheless, when it comes to defining a virtual fundamental class, the deformation-obstruction theory for the moduli stack of highly frozen triples is easier to study. The author considers the 4-term deformation-obstruction theory over the moduli stack of frozen triples and shows that its pull-back to the moduli stack of highly frozen triples, which is a \(\mathrm{GL}_r(\mathbb C)\)-torsor over the moduli of frozen triples, Zariski locally admits a suitable truncation giving a perfect 2-term deformation-obstruction theory. Upon assuming a technical assumption, the author can nevertheless get a \textit{globally} defined virtual fundamental class for the moduli stack of highly frozen triples. In the second part of the paper, the author studies those \(X\) admitting a torus action and shows that over the torus-fixed loci highly frozen triples are direct sums of \(r\) copies of torus-equivariant Pandharipande-Thomas (PT) stable pairs. While these threefolds are not compact (because of the torus action), the fixed loci are \(r\)-fold products of fixed loci in the PT moduli space, which they conjectured to be smooth and compact. This allows the author to apply Graber-Pandharipande virtual localization to calculate these higher rank invariants. The author calculates explicitly the 1-legged equivariant Calabi-Yau vertex for the local \(\mathbb P^1\) toric Calabi-Yau (that is, the total space of \(\mathcal O_{\mathbb P^1}(-1)\oplus\mathcal O_{\mathbb P^1}(-1)\to\mathbb P^1\)).