Enveloping algebras of Slodowy slices and Goldie rank (Q649081)

Let \(U(\mathfrak{g},e)\) be a finite \(W\)-algebra associated with a nilpotent element \(e\) in a complex simple Lie algebra \(\mathfrak{g}=\text{Lie}(G)\). It is defined as follows. Let \(\{e,f,h\}\) be a \(\mathfrak{sl}_2\)-triple in \(\mathfrak{g}\). Let \((,)\) be an invariant bilinear form on \(\mathfrak{g}\), such that \((e,f)=1\). Let \(\chi\in \mathfrak{g}^*\) be a functional, such that \(\chi(x)=(e,x)\). Then \(U(\mathfrak{g},e)\) is the enveloping algebra of the Slodowy slice \(e+\text{Ker ad}\, f\) to the adjoint orbit \(\mathcal{O}=(\text{Ad}\, G)e\). One has \(U(\mathfrak{g},e)=(\text{End}_{\mathfrak{g}} Q_e)^{op}\). Here \(Q_e\) is a generalized Gelfand-Graev \(\mathfrak{g}\)-module. It is an induced module from a one-dimensional module \(\mathbb{C}_{\chi}\) of a nilpotent subalgebra \(\mathfrak{m}\) of \(\mathfrak{g}\) whose dimension equals \(d(e)=\frac{1}{2}\dim\mathcal{O}\). The action of \(\mathfrak{m}\) on \(\mathbb{C}_{\chi}\) is given by \(x(1_{\chi})=\chi(x)1_{\chi}\), where \(x\in\mathfrak{m}\), \(1_{\chi}\in \mathbb{C}_{\chi}\). It is know that if \(V\) is an irreducible finite-dimensional \(U(\mathfrak{g},e)\)-module then \(Q_e\otimes_{U(\mathfrak{g},e)} V\) is an irreducible \(U(\mathfrak{g})\)-module. One can associate with this \(U(\mathfrak{g})\)-module a primitive ideal \(I_V=\text{Ann}(Q_e\otimes_{U(\mathfrak{g},e)} V)\subset U(\mathfrak{g})\). This is in fact a one-to-one correspondence between primitive ideals in \(U(\mathfrak{g})\) and \(U(\mathfrak{g},e)\)-modules (proved independently by \textit{I. Losev} [J. Am. Math. Soc. 23, No. 1, 35--59 (2010; Zbl 1246.17015)] and the author [Adv. Math. 225, No. 1, 269--306 (2010; Zbl 1241.17015)]). Note that to every primitive ideal in \(I\subset U(\mathfrak{g})\) there corresponds a subvariety in \(\mathfrak{g}\), which, as proved by Borho-Brylinsky, is closure of a nilpotent orbit in \(\mathfrak{g}\). Put \(G_x=\{g\in G: (\text{Ad}\,g)x=x\}\) and let \(\Gamma(e)\) be the component group of \(G_e\). This group acts on \(U(\mathfrak{g},e)\) as automorphisms. As a corollary it acts by twisting on \(U(\mathfrak{g},e)\)-modules. In [\textit{I. Losev}, Duke Math. J. 159, No. 1, 99--143 (2011; Zbl 1235.17007)] it is proved that all modules that belong to a \(\Gamma(e)\) orbit correspond exactly to ideals in \(U(\mathfrak{g})\) that define to the same nilpotent orbit in \(\mathfrak{g}\). In particular all these modules have the same dimension. In [\textit{I. Losev}, J. Am. Math. Soc. 23, No. 1, 35--59 (2010; Zbl 1246.17015)] it was proved that \(\text{rk}(U(\mathfrak{g})/I_V)\leq \dim V\), where \(\text{rk}(U(\mathfrak{g})/I_V)\) is the Goldie rank. The first main result of the paper is the theorem stating that the Goldie rank of \(U(\mathfrak{g})/I_V\) divides \(\dim V\). The second main result is the theorem stating that if the Goldie field of \(U(\mathfrak{g})/I_V\) is isomorphic to the skew field of fractions of the Weyl algebra \(A_{d(e)}(\mathbb{C})\), that the Goldie rank of \(U(\mathfrak{g})/I_V\) equals \(\dim V\) for any finite dimensional irreducible \(U(\mathfrak{g},e)\)-module. In particular one gets that the Goldie rank of \(U(\mathfrak{g})/I_V\) equals \(\dim V\) in the case \(\mathfrak{g}=\mathfrak{sl}_n\). As an application of the obtained results the author disproves the Joseph's version of the Gelfand-Kirillov conjecture. It asserts that the Goldie field of a primitive quotient of \(U(\mathfrak{g})\) is always isomorphic to a Weyl skew-field and it fails in fact for \(\mathfrak{g}=\mathfrak{sp}_{2n}\).