On the unitary globalization of cohomologically induced modules (Q2444477): Difference between revisions

The orbit method suggests a close connection between irreducible representations of a Lie group \(G_\mathbb R\) and the co-adjoint orbits. When \(G_\mathbb R\) is a real reductive group, attached to elliptic co-adjoint orbits is a family of irreducible unitarizable \((\mathfrak g,K)\)-modules, \(A_{\mathfrak q}(\lambda)\). These are called cohomologically induced modules. The purpose of this paper is to give a geometric description of the corresponding \(G_\mathbb R\) unitary representations. Under some positivity assumptions on \(\lambda\), \(A_{\mathfrak q}(\lambda)\) can be realized as the space of \(K\)-finite solutions of an elliptic differential operator \(\mathcal D\), acting on sections of a bundle over \(G_\mathbb R/K_\mathbb R\). The geometric construction of Dolbeault cohomology is related to the solution space of \(\mathcal D\) via the so-called Real Penrose transform. Moreover, the Real Penrose transform determines an isomorphism between the cohomology realization and \(\text{Ker}\,\mathcal D\). In this paper, the authors determine (a) the Hermitian dual \((\text{Ker}\,\mathcal D)^h\), (b) the space of continuous \(G_\mathbb R\)-intertwining maps from \((\text{Ker}\,\mathcal D)^h\) to \(\text{Ker}\,\mathcal D\). Such intertwining maps are also of ``kernel type''. The authors describe the kernels in terms of generalized spherical functions, \(F\). The function \(F\) determines the unique (up to a scalar) invariant Hermitian form on \((\text{Ker}\,\mathcal D)\) and depends solely on the minimal \(K\)-type of \(A_{\mathfrak q}(\lambda)\). This is consistent with the fact that \(A_{\mathfrak q}(\lambda)\) is unitary if and only if the Hermitian form is definite on its bottom layer. In particular, when \(A_{\mathfrak q}(\lambda)\) is the Harish-Chandra module of a representation in the discrete series, \(F\) is the function defined by Flensted-Jensen. The cohomology class \([\omega]\) that corresponds to \(F\) via the Penrose transform is completely determined by \(F\).