Geometricity for derived categories of algebraic stacks (Q730337)

The authors consider the derived category of an algebraic stack, traditionally studied in the context of triangulated categories. These have some short-commings; there are in general no tensor products, no concept of duals, and categories of triangulated categories have no obvious triangulated structure. To fulfil these conditions, one can work instead in differential graded categories (dg categories). In this work, the authors define a dg category \(D_{\mathrm{pf}}^{\mathrm{dg}}(X)\) which enhances \(D_{\mathrm{pf}}(X)\) in the sense that its homotopy category is equivalent to \(D_{\mathrm{pf}}(X)\). The dg categories simulate categories of sheaves on a hypothetical noncommutative space. It is well known that a variety cannot in general be recovered from its associative dg category, but several homological invariants can, including the algebraic \(K\)-theory spectrum and various variants of cyclic homology. In noncommuatative algebraic geometry, proper and smooth varieties translates to \textit{saturated} dg categories. The dg category \(D_{\mathrm{pf}}^{\mathrm{dg}}(X)\) associated to a variety \(X\) is saturated if and only of \(X\) is smooth and proper. On the other hand, the saturated dg categories are have an intrinsic characterization as the homotopy dualizable objects in the category of all dg categories with respect to a certain localization. These type of enhanced dg category can give answers to how far a noncommutative space is from being commutative. Smooth projective varieties have a dominant role in commutative geometry, and based on this, \textit{D. Orlov} [Adv. Math. 302, 59--105 (2016; Zbl 1368.14031)] recently introduced the concept of a geometric dg category. By definition, when \(X\) is a smooth, projective variety, \(D_{\mathrm{pf}}^{\mathrm{dg}}(X)\) is geometric, and it then follows that all its \textit{admissible} subcategories also are geometric. It is a known fact that every geometric dg category is saturated, and Orlov asks the question if all saturated dg categories are geometric. This question is open, and one of the main results in this article is that when \(X\) is a smooth, proper Deligne-Mumford stack over a field of characteristic zero, then the dg category \(D_{\mathrm{pf}}^{\mathrm{dg}}(X)\) is geometric, and in particular saturated. This result is extended to stacks over arbitrary fields. Because there in general are no resolutions of singularities over a field in positive characteristic, these extensions are restricted to \textit{projective} algebraic stacks, which in principle means that they are proper over the field \(k\). For a smooth, proper scheme \(X\) over a field of positive characteristic, it is not clear if \(D^{\mathrm{dg}}_{\mathrm{pf}}(X)\) is geometric if \(X\) is not projective. The authors consider tame algebraic stacks, which are more restrictive than Deligne-Mumford stacks. When the ground field is of characteristic zero, Deligne-Mumford stacks are equivalent to the class of separated tame algebraic stacks, but the last have better conditions in positive characteristic, verbatim: ``under mild finiteness assumptions, the tame algebraic stacks are scheme-like from a noncommutative perspective in the sense that their derived categories are generated by a single compact object, and the compact objects coincide with the perfect complexes''. This discussion ends with the second main result of the article, stating that the dg category \(D^{\mathrm{dg}}_{\mathrm{pf}}(X)\) is geometric, in paricular saturated, when \(X\) is a tame, smooth, projective algebraic stack over an arbitrary field. The author apply their results to invariant theory, proving that for \(U\) a smooth, quasi-projective variety over a field \(k\), \(G\) a linear algebraic group over \(k\) acting properly on \(U\) such that the action admits a geometric quotient \(U\rightarrow U/G\) such that \(U/G\) is projective over \(k\) with linearly reductive stabilizers, then the dg category enhancing the bounded derived category \(D^b(\text{Coh}^G(U))\) of \(G\)-equivariant coherent sheaves on \(U\) is geometric, and in particular saturated. If the base field \(k\) has characteristic zero, then the condition that the stabilizers are linearly reductive is not necessary. Also, if \(G\) is finite and \(U\) projective over \(k\), then the action is automatically proper and the existence of a projective geometric quotient \(U/G\) is guaranteed. The proof of the main results builds on the destackification theorem by Bergh and Rydh, allowing to compare a smooth, tame algebraic stack to a smooth algebraic space via a sequence of birational modifications, i.e. \textit{stacky blowups} via \textit{root stacks}, and the glueing theorem for geometric dg categories by Orlov. The article is self-contained, precise, and gives a very nice study of noncommutative algebraic geometry, pinpointing the challenges in doing this purely categorical, and giving reasonable results. The introduction is in it self a very good review of the article, and could have served as such.