Problem of formation of quotients and base change (Q706183)

In this article, a \(S\)-semistable curve \(X/S\) means a flat projective morphism \(X\rightarrow S\) of schemes (over \(k\)) such that each geometric fiber is a connected curve with at worst ordinary double points. Consider a faithful action of a finite group \(G\) on \(X\rightarrow S\). Then the quotient \(X/G\rightarrow S\) may not be semistable. To specialize this question, consider a normal subgroup \(H\vartriangleleft G\). Then there is an induced action of \(G/H\) on the quotient \(X/H\) which makes the quotient map equivariant with respect to \(G\rightarrow G/H.\) The authors wish to compare the \(G\)-equivariant deformation functor \(\text{Def}_G(X)\) of \(X\) with the corresponding \(G/H\)-equivariant deformation functor \(\text{Def}_{G/H}(X/H)\). The case \(G=H\) already gives information to the first question. Provided that for any base change \(T\rightarrow S\) the natural morphism \((X/G)\times_S T\rightarrow(X\times_S T)/G\) is an isomorphism (base change and quotients are commuting operations), there exists a morphism between the two deformation functors: \[ \text{Def}_H(X)\overset{\text{res}}\leftarrow\text{Def}_G\overset{\text{Ind}}\rightarrow\text{Def}_{G/H}(X/H). \] It is known that base change and quotients commute if \(G\) acts freely, and if the order of \(G\) is invertible in the structure sheaf of \(S\). One of the main results of the article is a new condition under which base change and quotients commute: If \(G\) is a finite group of \(S\)-automorphisms of \(X/S\) and if the action of \(G\) is free on an open dense set on any geometric quotient, then base change and quotients commutes, and moreover \(X/G\) is semistable. Suppose we have a semistable curve over a one dimensional regular base. Assume that the generic fibre is smooth, and that the curve is acted upon by a finite group \(G\). Then the group \(G\) will not act freely on on any open dense set of the special fiber. The hypothesis is restored by letting the group itself degenerate. This leads to a local version of the result above for a finite, flat \(S\)-group scheme \(\mathbf G\) acting on a smooth affine \(S\)-curve. This is the idea of the article, but the method for proving this leads to lots of beautiful mathematics: An approximation to the ``quotients'' \(A^{\mathbf G}\) is the Kleiman-Lønsted algebra \(\Sigma_R^{\mathbf G}(A)\) associated to a finite flat group scheme action. \(\Sigma_R^{\mathbf G}(A)\) and \(A^{\mathbf G}\) have the same maximal spectrum, and if the action of the constant group \(G\) is free, then they are equal. The questions above are solved for the Kleiman-Lønsted algebra, and then generalized to the original question. In particular this applies to the case of quotients of semistable curves, where it gives nice results. After proving that restriction and induction morphisms exists, it is commented that a local-global principle reduces the study of the deformation functor \(\text{Def}_G(X)\) to the study of the local deformation functor at singular points and wildly ramified points. This is an interesting tread to follow, and the authors have a paper in preparation. The final example in the article illustrates the local theory.