Global matrix factorizations (Q2434617): Difference between revisions

The notion of category of matrix factorizations goes back to ideas of Kontsevich in the context of mirror symmetry, and had precursors in work of Eisenbud and others on singularity theory. Rigorous definitions and proofs of important properties were obtained in the foundational paper of \textit{D. O. Orlov} [Proc. Steklov Inst. Math. 246, 227--248 (2004); translation from Tr. Mat. Inst. Steklova 246, 240--262 (2004; Zbl 1101.81093)]. Following Orlov, given a pair \((X, W)\) where \(X = \mathrm{Spec}(R)\) is an affine algebraic variety and \(W:X \rightarrow \mathbb{A}^1\) is a flat map, the category of matrix factorizations \(MF(X, W)\) is a differential \(\mathbb Z_2\)-graded triangulated category whose objects are pairs of projective \(R\)-modules: \[ u_1: P_1 \leftrightarrows P_2 : u_2 \] satisfying \(u_2 \circ u_1 = W\). In the paper under review, the authors extend the theory of matrix factorizations to the case of non-affine total space. They show that, dropping the affineness assumption on \(X\), it is still possible to attach to any flat map \(W: X \rightarrow \mathbb{A}^1\) is a differential \(\mathbb{Z}_2\)-graded triangulated category of matrix factorizations \(MF(X, W)\) which generalizes Orlov's definition (independently also \textit{D. Orlov} has obtained results in this direction [Math. Ann. 353, No. 1, 95--108 (2012; Zbl 1243.81178)]). This requires a careful study of the homological algebra of curved complexes over \(X\) and is done in Section 2 of the paper. In the following sections, the authors generalize to their set-up many results on matrix factorizations which were known in the affine case. This includes a computation of the Hochschild cohomology of \(MF(X, W)\) and a proof of the fact that \(MF(X, W)\) has good homological properties: that is, when the the critical locus of \(W\) is proper, it is a smooth and proper category (in fact, smoothness is always satisfied). This generalizes work of Dyckerhoff in the affine case, [\textit{T. Dyckerhoff}, Duke Math. J. 159, No. 2, 223--274 (2011; Zbl 1252.18026)]. In the last section of the paper the authors prove that if the total space \(X\) is Calabi-Yau, that is the canonical bundle of \(X\) is trivial, then \(MF(X, W)\) is a Calabi-Yau category.