Cyclic homology of categories of matrix factorizations
Alexander I. Efimov
TL;DR
This work establishes a precise link between noncommutative invariants of matrix factorization categories and classical vanishing-cycle cohomology. It develops a framework of curved DG categories and mixed complexes with $u$-connections to transfer information between periodic cyclic homology $HP_{ullet}(MF_{coh}(X,W))$ and vanishing cohomology, while identifying Hochschild homology with twisted de Rham hypercohomology $H^{\bullet}(X, (\Omega_X^{\bullet}, dW))$. The main achievement is proving, first in the affine case and then globally, that $HP_{ullet}(MF_{coh}(X,W))$ and $HH_{ullet}(MF_{coh}(X,W))$ correspond to twisted de Rham data and vanishing cohomology under a Riemann–Hilbert-type correspondence, including a monodromy twist by sign. These results yield a robust bridge between noncommutative algebro-geometric invariants and the theory of vanishing cycles, with implications for the categorical Chern character and conjectural Hodge-theoretic surjectivity onto vanishing-cycle Hodge classes. The work thus generalizes classical HKR-type identifications to the setting of matrix factorizations and curved DG categories, providing tools for studying degenerations, Thom–Sebastiani decompositions, and Hodge-theoretic conjectures in a derived categorical context.
Abstract
In this paper, we will show that for a smooth quasi-projective variety over $\C,$ and a regular function $W:X\to \C,$ the periodic cyclic homology of the DG category of matrix factorizations $MF(X,W)$ is identified (unde Riemann-Hilbert correspondence) with vanishing cohomology $H^{\bullet}(X^{an},φ_W\C_X),$ with monodromy twisted by sign. Also, Hochschild homology is identified respectively with the hypercohomology of $(Ω_X^{\bullet},dW\wedge).$ One can show that the image of the Chern character is contained in the subspace of Hodge classes. One can formulate the Hodge conjecture stating that it is surjective ($\otimes\Q$) onto Hodge classes. For W=0 and $X$ smooth projective this is precisely the classical Hodge conjecture.
