Artin's criteria for algebraicity revisited (Q2418586)

In algebraic geometry, moduli spaces are constructed by forming suitable quotients, the most common of these being projective methods. Grothendieck generalized this to the functor of points, and the main problem considered concerns representable functors, where the scheme representing the functor is the moduli space for families of geometric points. To prove representability, Grothendieck (together with Serre, MacLane et. al.) developed deformation theory and formal geometry as main tools for proving representability, but left this for the projective methods in proving the existence of Hilbert and Picard schemes. Later on, Artin generalized Grothendieck's existence results, and proved that Hilbert and Picard schemes exists as algebraic spaces. The authors claim that the correct setting is that of algebraic spaces and stacks (not schemes). Artin gave precise criteria for algebraicity of functors and stacks. These were clarified by Conrad and De Jong using Néron-Popescu desingularization, by \textit{H. Flenner} [Math. Z. 178, 449--473 (1981; Zbl 0453.14002)] using Exal, and by the first author [J. Reine Angew. Math. 722, 137--182 (2017; Zbl 1362.14012)] using coherent functors. This article uses the ideas of Flenner and the first author to give a new criterion for algebraicity of functors and stacks, giving a supplement to Artin's criteria. To give a wider application of the criteria, and to simplify the proofs of Artin and Flenner, several new concepts are introduced. Also, a place where Artin's proof for algebraicity of stacks doesn't work in positive characteristic is circumvented by these new techniques. The main result is stated for \(S\) an excellent scheme. Then a category \(X\) fibred in groupoids over \(\operatorname{Sch}/S,\) is an algebraic stack, locally of finite presentation over \(S\), if and only if it satisfies: \begin{itemize} \item[(1)] \(X\) is a stack over \(\operatorname{Sch}/S\) with the fppf topology, \item[(2)] \(X\) preserves limits, \item[(3)] \(X\) is weakly effective, \item[(4)] \(X\) is \(\mathbf{Art}^{\text{triv}}\)-homogeneous, \item[(5a)] \(X\) has bounded aumorphisms and deformations, \item[(5b)] \(X\) has constructible automorphisms and deformations, \item[(5c)] \(X\) has Zariski local automorphisms and deformations, \item[(6b)] \(X\) has constructible obstructions, \item[(6c)] \(X\) has Zariski local obstructions. \end{itemize} In addition, the main result contains special cases where some of the criteria are superfluous or can be replaced by simpler ones. The main content of the paper is to give the necessary definitions of the nine concepts (criteria) above, and to prove the result. The authors point out the homogeneity condition (4) above as the most striking difference between their condition and Artin's condition as this condition only involves local artinian schemes and that there is no conditions on étale localization of deformations and obstruction theories. When \(S\) is Jacobson, e.g. of finite type over a field, no compatibility with zariski localization is needed, nor conditions on the compatibility with completions for automorphisms and deformations. A detailed comparison between the results of this article and the results of Artin is given. The authors single out the following steps in algebraicity proofs up to now: \begin{itemize} \item[(i)] Existence of formally versal deformations, \item[(ii)] algebraization of formally versal deformations, \item[(iii)] openess of formal versality, \item[(iv)] formal versality implies formal smoothness. \end{itemize} Notice that these steps gives an explicit explanation of the link between deformation theory and the theory of algebraic stacks. The last two steps are the ones where the treatment of Artin, Flenner, Starr and Hall differs from the present. This yields both the criteria themselves and the techniques, and this treatment fills up the main body of the paper. Step (iv), formal versality implies formal smoothness: This criterion is weaker than Artin's two criteria, except that in positive characteristic, \(X\) needs to be a stack in the fppf topology or criteria (4) must be strengthened. This is similar to Artin's criteria where the functor is assumed to be an fppf-sheaf. This is crucial for his proof that formally universal deformations are formally étale, settling step (iv) for functors. This proof also depends on the existence of universal deformations and so does not extend to stacks with infinite or nonreduced stabilizers. Using a different approach, the authors extend this result to fppf stacks. Artin only assumes that the groupoid is an étale stack. The authors express that they do not understand Artin's proof of step (iv) when \(S\) is not of finite type over \(\mathbb Z\) or a perfect field. This special case is solved by the techniques given in the present paper. Step (iii), openness of formal versality: In the treatment of the present exposition, localization is only required when passing to nonclosed points of finite type. These points only exist when \(S\) is not Jacobson, that is if \(S=\operatorname{Spec} D\) of a DVR. The validation of the proof boils down to a matter of simple algebra, the criterion for the openness of the vanishing locus of half-exact functors that follows from the Ogus-Bergman Nakayama Lemma for half-exact functors. From step (ii) and (iv) it is proved that conditions (1)--(4) and (5a) at fields gives homogeneity for arbitrary integral morphisms, and so that \(\operatorname{Aut}_{X/S}(T,-),\;\operatorname{Def}_{X/S}(T,-)\) and \(\operatorname{Obs}_{X/S}(T,-)\) are additive functors. The distinct advantage of the criterion in this paper is the dramatic weakening of the homogeneity. The ideas of the paper have lead to a criterion for a half-exact functor to be coherent, and this does not follow from any algebraicity criterion. The authors recall the notion of homogeneity, limit preservation and extensions. They introduce homogeneity involving only artinian rings and show that residue field extensions are invariant for stacks in the fppf topology. They relate formal versality, formal smoothness and the vanishing of Exal, which is also defined. Then additive functions are considered, and their vanishing loci. This is applied to Exal, which assure that the locus of formal versality is open. Conditions are given on \(\operatorname{Def}\), \(\operatorname{Aut}\), \(\operatorname{Obs}\) that imply the corresponding conditions on Exal. The general theory is applied to introduce \(n\)-step obstruction theories. These are formulated without using linear obstruction theories of Artin, and are applied to the study of effectivity. This is a very far-reaching article, concatenating and giving a universal understanding of the correspondence and differences between algebraic stacks and obstruction theory. As there are very few authors that are experts in both fields, this article is a valuable contribution, with a lot of very important applications and comparisons.