Matrix factorizations and cohomological field theories (Q278873): Difference between revisions

Cohomological Field Theory (CohFT) was introduced by \textit{M. Kontsevich} and \textit{Yu. Manin} [Commun. Math. Phys. 164, No. 3, 525--562 (1994; Zbl 0853.14020)] to axiomatize the structure of Gromov-Witten invariants and certain aspects of mirror symmetry. The latter can be viewed as an isomorphism of CohFTs coming from the so-called A and B models in string theory. The work of Saito and Givental led to the construction of CohFTs corresponding to the B-side of the Landau-Ginzburg (LG) model related to a quasihomogeneous isolated singularity \(w\). A CohFT corresponding to the A-side was recently constructed by Fan, Jarvis and Ruan using moduli spaces of solutions to certain PDE corresponding to linear perturbatrions of potential \(w\). They proved it to be isomorphic to the B-side CohFT for the simple ADE singularities. In this paper, the authors present a purely algebraic construction of (conjecturally) a generalization of the Fan-Jarvis-Ruan (FJR) CohFT. The main tools are matrix factorizations, generalizations of differential complexes obtained by replacing \(d^2=0\) by \(d^2=w\), and the state space is built from the Hochschild homology of the dg-categories of equivariant matrix factorizations associated with the singularity \(w\) and its finite group of symmetries \(G\). The coefficients are in the representation ring of \(G\), and the specialization to \(\mathbb{C}\) conjecturally gives the FJR theory. This conjecture is proved for simple singularities only, the general case is obstructed by difficulties with verifying the dimension axiom. The ``\(w\)-curves'' of FJR, orbicurves with marked orbipoints and a collection of line bundles coming from the monomials of \(w\), are reinterpreted as attributes of \(G\) rather than \(w\). The notion of \(w\)-structure is reformulated using principal bundles, and leads to the moduli stack of spin curves which replace the \(w\)-curves. This removes a requirement of FJR that the group of the diagonal symmetries of \(w\) be finite. The authors also obtain a weak categorification of their CohFT by constructing a collection of functors inducing the CohFT maps (up to rescaling) after passing to Hochschild homology. The functors are given by kernels, termed fundamental matrix factorizations, living on the product of moduli spaces of spin curves with affine spaces. The factorizations play a role similar to the virtual fundamental class of the Gromov-Witten theory. The factorization axiom of FJR holds at the categorical level after passing to the appropriate finite covers of the moduli. It is expected that a version of quantum \(K\)-theory can also be developed in this context.