Invariant deformation theory of affine schemes with reductive group action (Q2349948)

Consider a reductive group \(G\) over an algebraically closed field \(k\) of characteristic zero. For an affine \(G\)-scheme \(W\) of finite type over \(k\) and a given function \(h: \mathrm{Irr}(G) \to \mathbb{Z}_{\geq 0}\), \textit{V.~Alexeev} and \textit{M.~Brion} [J.~Algebraic~Geom. 14, No. 1, 83--117 (2005; Zbl 1081.14005)] studied the invariant Hilbert scheme \(\mathcal{H} = \mathrm{Hilb}^G_h(W)\); morally, it classifies the closed \(G\)-subschemes (or more generally, flat families of such objects over a base scheme) \(\mathcal{Z}\) of \(W\) such that the coordinate algebra of \(\mathcal{Z}\) decomposes under \(G\) according to \(h\). In particular, there is a universal family \(\mathcal{U} \subset W \times \mathcal{H}\). The Hilbert-Chow morphism \(\gamma: \mathcal{H} \to W /\!/ G\) is a canonically defined morphism covered by \(\mathrm{pr}_1: \mathcal{U} \to W\). It is known to be a projective morphism; furthermore, there is a largest open subset \(U \subset W /\!/ G\) over which the GIT quotient morphism from \(W\) is flat. The main component \(\mathcal{H}_{\mathrm{main}}\) is defined to be the closure of \(\gamma^{-1}(U)\). Many geometric properties of the invariant Hilbert scheme remain inaccessible, such as the singularity, connectedness, etc. The main purpose of this paper is to give an algorithm for calculating the universal deformation at a point \([X]\) of \(\mathcal{H}\). An effective theory of invariant deformations is carefully set up under suitable hypotheses. As applications, the authors considered three cases: (1) \(\mathrm{GL}_3\) acts on \((k^3)^{\oplus n_1} \oplus (k^{3*})^{\oplus n_2}\), (2) \(\mathrm{SO}_3\) acts on \((k^3)^{\oplus 3}\), and (3) \(\mathrm{O}_3\) acts on \((k^3)^{\oplus n}\). In each case there is a \(\mathbb{G}_m \subset \mathrm{Aut}^G(W)\) with strictly positive weights on \(k[W]\) and on \((T_{[X]} \mathcal{H})^\vee\). Let \(h\) be the Hilbert function of the general fibers of \(W \to W/\!/G\). In the cases (1) and (2), it turns out that \(\mathcal{H}_{\mathrm{main}}\) is smooth, there is a decomposition \(\mathcal{H} = \mathcal{H}_{\mathrm{main}} \cup \mathcal{H}'\) into irreducible components, such that \(\mathcal{H}'\) is smooth (resp. singular) in case (1) (resp. case (2)). In the case (3), \(\mathcal{H}\) turns out to be connected and has at least two irreducible components. Grosso modo, the technique is to find a suitable algebraic subgroup \(H \subset \mathrm{Aut}^G(W)\), a Borel subgroup \(B \subset H\) and try to ``localize'' the geometric properties of \(\mathcal{H}\) at some \(B\)-fixed point; one looks then for \(\mathbb{G}_m \subset B\) satisfying the positivity property alluded to above.