On the commuting variety of a reductive Lie algebra and other related varieties (Q290450)

Consider a reductive Lie algebra \(\mathfrak{g}\) of finite dimension \(\ell + 2n\), \(\ell = \mathrm{rg}(\mathfrak{g})\), over an algebraically closed field \(\Bbbk\) of characteristic zero. Let \(G\) denote its adjoint group. The main geometric objects under consideration in this paper are following subvarieties of \(\mathfrak{g}^k\) \[ \mathcal{B}^{(k)} = \left\{ (x_1, \ldots, x_k) \in \mathfrak{g}^k : \text{in the same Borel} \right\}, \] \[ \mathcal{N}^{(k)} = \left\{ (x_1, \ldots, x_k) \in \mathfrak{g}^k : \text{nilpotent} \right\}, \] \[ \mathcal{C}^{(k)} = \overline{\left\{ (x_1, \ldots, x_k) \in \mathfrak{g}^k : \text{in the same Cartan} \right\}}. \] Fix a Cartan subalgebra \(\mathfrak{h}\) inside a Borel \(\mathfrak{b}\) and define \(\mathcal{X}\) to be the variety of \((x,y) \in \mathfrak{g} \times \mathfrak{h}\) such that \(y \in \overline{G x}\). Define \(\mathcal{B}^{(k)}_x\) to be the subvariety of \(\mathcal{X}^k\) consisting of tuples \((gx_i, \overline{x}_i)_{i=1}^k\) with \(g \in G\), \(x_i \in \mathfrak{b}\) and \(x \mapsto \bar{x}\) being the quotient map \(\mathfrak{b} \to \mathfrak{h}\). The main Theorem 1.1 states the equivalences in each of the following cases: (i) normality of \(\mathcal{N}^{(k)}\) and that of \(\mathcal{B}_x^{(k)}\), (ii) same for Cohen--Macaulay property; (iii) rational singularity and Cohen--Macaulay property of \(\mathcal{N}_x^{(k)}\), (iv) same for \(\mathcal{B}_x^{(k)}\). Furthermore, (v) \(\Bbbk[\mathcal{B}_x^{(k)}]\) is a free extension of its \(G\)-invariant subalgebra, which identifies with \(\text{Sym}(\mathfrak{h}^k)\), (vi) \(\Bbbk[\mathcal{B}^{(k)}]^G\) identifies with \(\mathrm{Sym}(\mathfrak{h}^k)^{W(\mathcal{R})}\) with respect to the diagonal action of the Weyl group \(W(\mathcal{R})\) on \(\mathfrak{h}^k\); (vii) the ideal of \(\Bbbk[\mathcal{B}^{(k)}]\) generated by \(\Bbbk[\mathcal{B}^{(k)}]^G_+\) is strictly contained in the ideal defining \(\mathcal{N}^{(k)}\). The second main Theorem 1.2 states that by denoting \(X'\) the set of centralizers of regular elements of \(\mathfrak{b}\) with at most subregular semisimple components, we have (i) all element of \(X\) is a commutative algebraic subalgebra of \(\mathfrak{g}\); (ii) for all \(x \in \mathfrak{g}\) the set of elements of \(G . X\) containing \(x\) has dimension \(\leq \dim \mathfrak{g}^x - \ell\); (iii) the sets \(X \smallsetminus B . \mathfrak{h}\) (resp. \(G . X \smallsetminus G . \mathfrak{h}\)) are equidimensional of dimension \(n-1\) (resp. \(2n-1\)); (iv) the sets \(X'\) (resp. \(G . X'\)) are smooth open subsets of \(X\) (resp. \(G . X\)) of codimension \(\geq 2\). Roughly speaking, the final main Theorem 1.3 relates the desingularizations of \(X\) to those of \(\mathcal{C}_x^{(k)} = \eta^{-1}(\mathcal{C}^{(k)})\), where \(\eta: \mathcal{B}^{(k)}_x \to \mathcal{B}^{(k)}\) is the canonical projection. This work fits into a project by the authors that studies the geometric objects derived from the diagonal action of \(G\) on \(\mathfrak{g}^k\), which partly generalizes the classical theory of Springer resolutions, etc.