Linearized topologies and deformation theory (Q2634699)

The article builds on the notion of a deformation of a category given in [the author and \textit{M. van den Bergh}, Trans. Am. Math. Soc. 358, No. 12, 5441--5483 (2006; Zbl 1113.13009)], extending Gerstenhaber's deformation theory of associative algebras to abelian categories. Namely, let \({\mathcal C}\) be a Grothendieck category over a field \(k\). A \textit{deformation} of \({\mathcal C}\) over an Artin local \(k\)-algebra \(R\) is an \(R\)-linear Grothendieck abelian category \({\mathcal D}\) which reduces to \({\mathcal C}\) upon restricting to \(k\)-linear objects. Let us consider a representation \({\mathcal C}\cong\text{Sh}({\mathfrak a},{\mathcal T})\) of \({\mathcal C}\) as the category of sheaves on the \(k\)-linear site \(({\mathfrak a},{\mathcal T})\), associated to a functor \(\gamma:{\mathfrak a}\to{\mathcal C}\). The goal is to realize the deformation as \({\mathcal D}\cong\text{Sh}({\mathfrak b},{\mathcal S})\) for a linear site \(({\mathfrak b},{\mathcal S})\) such that (A)\quad \({\mathfrak b}\) is obtained as a linear (i.e. Gerstenhaber type) deformation of \({\mathfrak a}\) and (B)\quad \({\mathcal S}\) is a natural ``\(R\)-linear variant'' of \({\mathcal T}\). In general, both requirements may fail, due on the one hand to dependence of (A) on homological conditions (some Ext vanishing) and on the imprecision in the statement of (B). The article under review fixes both issues in the framework of map-graded categories. For a \(k\)-linear \({\mathcal U}\)-graded category \({\mathfrak a}\) over a small category \({\mathfrak a}\), the author describes how to linearize a Grothendieck topology \(\tau\) on \({\mathcal U}\) in order to obtain a linear topology \({\mathcal T}_{\tau}\) on \({\mathfrak a}\). In the framework of deformation theory, \({\mathcal U}\) will be the category associated to a poset. The guiding example is the classical case where the poset is the poset of open sets of an algebraic scheme \(X\) and the coverings in the Grothendieck topology are simply given by unions of open sets. Drawing on earlier work [\textit{W. Lowen}, Int. Math. Res. Not. 2008, Article ID rnn118, 32 p. (2008; Zbl 1160.13010)], a linear version of a construction of Grothendieck leads to a \({\mathcal U}\)-graded linear category \({\mathfrak o}\) with the same objects as \({\mathcal U}\) and \({\mathfrak o}(V,U)={\mathcal O}_X(V)\) for \(V\subset U\). In this setting, the author obtains for the linearized topology \({\mathcal T}_{\tau}\) a realization of the category \(\text{Sh}(X,{\mathcal O}_X)\) of sheaves of \({\mathcal O}_X\)-modules on the linearized site \(({\mathfrak o}, {\mathcal T}_{\tau})\). In order to realize (A), the author refines here to the full subcategory of affine open sets. One general question addressed in this article is the following: suppose \({\mathfrak a}\) is a \({\mathcal U}\)-graded category over \(({\mathcal U},\tau)\), \(\varphi:{\mathcal V}\to{\mathcal U}\) is a ``refinement'' functor, \(\sigma=\varphi^{-1}\tau\) is the induced topology on \({\mathcal V}\) and \(\phi:{\mathfrak a}^{\varphi}\to{\mathfrak a}\) is the naturally induced linear functor. What are natural conditions under which \(\phi\) is \textit{localizing}, i.e. under which we have an equivalence of categories \[ \text{Sh}({\mathfrak a},{\mathcal T}_{\tau})\cong \text{Sh}({\mathfrak a}^{\varphi},{\mathcal T}_{\sigma}) ? \] One main result of the article is that (Theorem 4.3) if \({\mathfrak a}\) is fibered over \({\mathcal U}\) and \(\varphi\) is localizing with respect to the topology \(\tau\) on \({\mathcal U}\), then \(\phi\) is localizing. The non-fibered case is also treated. In a last part, the article treats similar questions for a special topology, the \textit{tails topology}.