Deformation theory of objects in homotopy and derived categories. I: General theory (Q838129): Difference between revisions

This article is concerned with the deformation theory for DG modules over a DG category. If \({\mathcal A}\) is a given DG category (over a fixed field \(k\)), fix a DG module \(E\) over \({\mathcal A}^{\text{op}}\), i.e., a DG functor from \({\mathcal A}^{\text{op}}\) to the DG category of complexes of \(k\)-vector spaces. Infinitesimal deformations are understood in this context as deformations along artinian DG algebras (finite-dimensional \(k\)-algebras \({\mathcal R}\) of which the maximal ideal \(m\) is a DG ideal such that \({\mathcal R}/m \cong k\)). For such an artinian \(\mathcal R\), considered as a DG category with a single object, one forms the DG category \({\mathcal A}\otimes{\mathcal R}\). Roughly speaking, a homotopy deformation of \(E\) is then a DG \(({\mathcal A}\otimes{\mathcal R})^{\text{op}}\)-module \(S\) isomorphic to \(E \otimes{\mathcal R}\) as a graded module together with a compatible isomorphism of \({\mathcal A}^{\text{op}}\)-modules between \(E\) and \(S\otimes_{\mathcal R}k\). This defines a pseudo-functor \( \text{Def}^h(E)\) from the category of artinian DG algebras to groupoids. There is also a dual notion of codeformation, but both are equivalent. In fact the authors prove that \( \text{Def}_{\mathcal R}^h(E)\) can be described as the Maurer-Cartan groupoid of the DG algebra \(\text{End}(E) \otimes{\mathcal R}\). This implies then that it depends only on the homotopy type of \(E\), whence the name. By passing further to the derived category of \({\mathcal A}^{\text{op}}\)-modules, a pseudo-functor \(\text{Def}(E)\) of derived deformations is finally introduced (as well as the dual version of derived codeformations). Unlike their homotopy counterparts, these two pseudo-functors are not always equivalent. This is explained by the fact that derived deformations of \(E\) do not agree in general with homotopy deformations of an \(h\)-projective resolution of \(E\). The last section of the article investigates certain conditions under which such a statement does hold: the resolution of \(E\) should be bounded and one should restrict the pseudo-functors to negative artinian DG algebras.