A generalized Harish-Chandra isomorphism (Q616879)

The authors generalize the Harish-Chandra isomorphism theorem for complex semisimple Lie algebras to the tensor product \(U(\mathfrak g)\otimes V\), where \(\mathfrak g\) is reductive complex Lie algebra and \(V\) a locally finite \(\mathfrak g\)-module. In section 1 they set up the notation as follows. Let \(\mathfrak g\) be a reductive complex Lie algebra of rank \(r\) and fix a decomposition \(\mathfrak g=\mathfrak n_-\oplus\mathfrak{t}\oplus\mathfrak n_+\), where \(\mathfrak{t}\) is a Cartan subalgebra, and \(\mathfrak n_-,\mathfrak n_+\) are the nilpotent nilradicals of two opposite Borel subalgebras containing \(\mathfrak{t}\). Let \(\Delta\subset\mathfrak t^{\ast}\) be the root system of \(\mathfrak g\), and \(\alpha_1,\dots,\alpha_r\) be simple roots with coroots \(H_i=\alpha_i^{\vee}\in\mathfrak t\). Let \(E_i,H_i,F_i\) be the corresponding \(\mathfrak{sl}_2\)-triple for each \(i=1,\dots,r\) and \(\mathfrak g_i\) be the \(\mathfrak{sl}_2\)-subalgebra generated by this triple. Let \(W\) be the Weyl group of \(\Delta\) and \(s_1,\dots,s_r\) the simple reflections. Let \(G\) be a connected reductive complex algebraic group with \(\text{Lie}(G)=\mathfrak g\), and \(V\) a locally finite \(G\)-module. Set \(M=U(\mathfrak g)\otimes V\). Considering the appropriate actions, \(M\) decomposes into a direct sum of \(U(\mathfrak t)\)-bimodules \(M=U(\mathfrak t)\otimes V\oplus(\mathfrak n_-M+M\mathfrak n_+)\). Taking \(Z=M/(\mathfrak n_-M+M\mathfrak n_+)\), the previous decomposition implies that \(Z\) is identified with \(U(\mathfrak t)\otimes V (\ast)\). Consider the set \(\mathcal{S}=\{H_\alpha+k/\alpha\in\Delta,k\in\mathbb{Z}\}\). If \(\overline{U(\mathfrak g)}, \overline{U(\mathfrak t)}\) are the fields of fractions of \(U(\mathfrak g), U(\mathfrak t) \) w.r.t. \(\mathcal{S}\), set \(\overline{M}=\overline{U(\mathfrak g)}\otimes V\). A similar decomposition to \((\ast)\) holds also for \(\overline{Z}:=\overline{M}/(\mathfrak n_-\overline{M}+\overline{M}\mathfrak n_+)\), namely \(\overline{Z}\) is identified with \(\overline{U(\mathfrak t)}\otimes V\). In section 2 the authors discuss the following operators on \(M\). For \(i=1,\dots,r\), let \(\eta_i: M\to \overline{M}\) be the linear map defined as \[ \eta_i(m)=\sum_{k=0}^\infty (k!H_i(H_i-1)\cdots(H_i-k+1))^{-1}E_i^k(\text{ad}\, F_i)^k\tilde{s_i}(m). \] Here \(\tilde{s_i}\in \text{Norm}(T)\cap G_i\) stands for a representative of the simple reflection \(s_i\) through the isomorphism \(\text{Norm}(T)/T\overset{\simeq}{\to}W\) between the quotient of the normalizer of the maximal torus \(T\) of \(G\), with \(\text{Lie}(T)=\mathfrak t\), and the Weyl algebra \(W\). The maps \(\eta_i\) induce linear maps \(\xi_i: Z\to \overline{Z}\) called the \textit{Zhelobenko operators}. Section 3 contains the main result of the article (Theorem 1) stating the following. If \(Z^0\subset Z\hookrightarrow \overline{Z} \) is the zero weight space w.r.t. the adjoint \(\mathfrak t\)-action, let \(Q:=\{z\in Z^0\mid\xi_i(z)=z,\,i=1,\dots,r\}\) and consider \(\gamma: M^G\to Z\), the restriction of the projection \(\pi: M\to Z\). Then the map \(\gamma\) is injective and \(\gamma(M^G)=Q\). In section 4 the authors give an alternative description of the space \(Q\) by means of the hypergeometric function \(F\) and the decomposition of \(V\) into a sum of irreducible \(\mathfrak g_i\)-submodules. In particular, for each \(j=0,1,\dots\) let \(V_{ij}\) be the sum of \((2j+1)\)-dimensional irreducible \(\mathfrak g_i\)-submodules of \(V\). Let also \(\Psi_{ij}:=(H_i+2)\cdots (H_i+j+1)\). Then (Proposition 4) \(Z^0\mathfrak ni z\) is in \(Q\) iff for each \(i=1,\dots,r\), \(z\) can be written as a finite sum of terms of the form \(\Theta\Psi_{ij}\upsilon\) where \(\upsilon\in V^0\cap V_{ij},\Theta\in U(\mathfrak t)\) and \(s_i(\Theta)=\Theta\) meaning the extension to \(U(\mathfrak t)\) of shifted action of \(W\) on \(\mathfrak t^{\ast}\). Theorem 1 has a graded commutative analogue proved in section 5. Namely, \[ \mathrm{gr}\, Z=\mathrm{gr}\,M/((\mathfrak n_-+\mathfrak n_+)\mathrm{gr}\,M)=(S(\mathfrak g)/(\mathfrak n_-+\mathfrak n_+)S(\mathfrak g))\otimes V \] is an \(S(\mathfrak t)\)-module and correspondingly to the noncommutative case, \(\mathrm{gr}\, Z\) is identified with \(S(\mathfrak t)\otimes V\) and \(\mathrm{gr}\, Z^0\) with \(S(\mathfrak t)\otimes V^0\). Then the map \(\gamma\) induces a map \(\mathrm{gr}\, \gamma: (\mathrm{gr} \,M)^G\to \mathrm{gr}\, Z\). The graded analogue of Theorem 1 states that (Theorem 2) the map \(\mathrm{gr}\, \gamma\) is injective and \(F\in(S(\mathfrak t)\otimes V^0)^W\) belongs in \(\mathrm{gr}\,\gamma\) \(((\mathrm{gr}\,M)^G)\) if for each \(i=1,\dots,r\) and \(j=0,1,\dots\) the projection of \(F\) to \(S(\mathfrak t)\otimes (V^0\cap V_{ij})\) is divisible by \(H_i^j\) (the highest order term of \(\Psi_{ij}\)) in the first tensor factor.