Langlands classification for real Lie groups with reductive Lie algebra (Q1891671)

Let \(G\) be a Lie group with reductive Lie algebra \(g\) (not necessarily connected). In the paper under review the author establishes a ``Langlands classification'' for such \(G\). The group \(G\) is called really reductive if there exists an open normal subgroup \(G_1\) of \(G\) such that \(G/G_1\) is finite and \(G_1 = Z(G_1) G_0\). The author shows that any such group is CCR and that for such groups his Langlands classification includes in particular all the irreducible unitary representations. We explain roughly the main results of the paper. One starting point is that any \(G\) as above still has a Cartan decomposition \(G = K \text{exp }p\) and an Iwasawa decomposition \(G = KA_p N_p\). Let \(a_p \subseteq p\) be a maximal abelian subspace, \(\Sigma\) the system of restricted roots, and \(\Pi\) a basis of \(\Sigma\) defining the positive system \(\Sigma^+\). To each subset \(S \subseteq \Pi\) one associates a certain parabolic subgroup \(MAN\) of \(G\). A \((g,K)\)-module \(W\) is called admissible if the finite dimensional \(K\)-submodules of \(W\) are unitarizable and each simple \(K_0\)-type has finite multiplicity in \(W\). A continuous representation of \(G\) on a Banach space is said to be of Harish-Chandra type if the \((g,K)\)-module of analytic \(K\)-finite vectors is admissible. It is shown in Theorem 2.8 that continuous irreducible unitary representations of really reductive groups are of Harish-Chandra type. The crucial notion that makes it possible to overcome the possible non-compactness of the identity component \(K_0\) of \(K\) is the following. Let \(z_k\) denote the center of \(k\) and \(n = \dim z_k\). A continuous Banach representation \(U\) of \(G\) is called permissible if there exists a subgroup \(\Gamma \cong \mathbb{Z}^n \subseteq \mathbb{R}^n \cong z_k\) such that \(U(\text{exp } \Gamma)\) commutes with \(U(G_0)\) and the representation \(U |\text{exp }\Gamma\) is semisimple with finitely many simple types, each of which is one- dimensional. Let \(P\), resp., \(P^+ \subseteq a^*_p\) denote the non-negative, resp., positive combinations of elements of \(\Pi\) and \(\rho = {1\over 2} \text{trad}_{n_p}\). As in the classical case one associates to a permissible representation \(U\) of Harish-Chandra type its leading exponents of \(K\)-finite matrix coefficients which are certain linear functionals on the complexification of \(a_p\). The representation \(U\) is called tempered if the real parts of these leading exponents are contained in the set \(-\rho - P\). If, in addition, \(U\) is irreducible and the real parts of the leading exponents are contained in \(-\rho - P^+\), then the corresponding infinitesimal equivalence class is called the quasi-discrete series. Let \(a^+ := \{X \in a: (\forall \alpha \in \Pi \setminus S)\) \(\alpha(X) > 0\}\) be the Weyl chamber associated to \(S \subseteq \Pi\), where \(a = S^\perp \subseteq a_p\). If \(P = MAN\) is the parabolic subgroup associated to \(S\), then \((M, V, \nu)\) is called a Langlands data if \(\nu \in a^\mathbb{C}\) with \({\mathfrak R} \nu \in a^+\), \(M = Z_K (\nu)M_0\), and \(V\) is an irreducible tempered unitary representation of \(M\). The main results of the paper are that any non-tempered irreducible Harish-Chandra type representation of \(G\) is infinitesimally equivalent to the Langlands quotient obtained from an essentially unique triple \((M, V, \nu)\) of Langlands data; while tempered irreducible Harish-Chandra type representations are infinitesimal subrepresentations of some induced representation \(U^{V,\nu}\) with imaginary \(\nu\) and \(V\) a quasi-discrete series representation of a suitable \(M\).