Points in algebraic geometry (Q2348130): Difference between revisions

The paper under review gives a very pretty description of points in toposes of algebro-geometric origin. Recall that in sheaf theory one would like to check that a morphism of sheaves on a site is an isomorphism by checking it is one in all the stalks, as in the topological case. Because there exist examples of non-empty toposes with no points (i.e. fibre functors) this does not work in general. Nevertheless, under suitable finiteness hypotheses which are valid in all of the settings discussed below, Deligne proved that one can indeed check isomorphisms of sheaves using the fibre functors. In the case of a Grothendieck topology \(\tau\) on the category of finite type separated \(S\)-schemes \(\text{Sch}/S\) where \(S\) itself is separated noetherian, the authors give {\parindent=6mm \begin{itemize} \item[1.] an equivalence between fibre functors and \(\tau\)-local affine \(S\)-schemes, for topologies refining the Zariski topology and for which every covering family is refinable by a finite covering family; \item [2.] a description of \(\tau\)-local affine schemes for 12 Grothendieck topologies appearing in the literature. \end{itemize}} The first result is based on a classical translation of fibre functors into \(\tau\)-local pro-objects, which can be interpreted by classical limit arguments as \(\tau\)-local affine \(S\)-schemes if the topology is finer than the Zariski topology. The second result is an amalgam of classical, unpublished and new results. The authors give explicit definitions for each of the 12 topologies, describe the poset of refinements which results in a very nice picture in equation (6), and describe what the ``local rings'' are. Observe that not all of these topologies satisfy the conditions necessary to apply the first result, so it is not a description of the fibre functors in general. The fppf topology is omitted here, because the description in this case turns out to be problematic. A detailed discussion in this case (albeit without noetherian hypotheses) can be found in [\textit{S.\ Schröer}, ``Points in the fppf topology'', preprint, \url{arXiv:1407.5446}]. The paper ends with a description of which topologies have exact direct images for closed immersions. For some of the topologies used in this paper one needs to restrict to noetherian schemes, because various definitions which are equivalent in the noetherian setting become inequivalent for more difficult schemes.