On admissibility and temperedness of representations of real reductive groups (Q1271919)

In an earlier article [Acta Appl. Math. 37, 267-309 (1994; Zbl 0836.22020)], \textit{Z. Magyar} considered the representations of a broad class of reductive Lie groups, namely, those groups whose Lie algebras are reductive. For the `really reductive' groups, i.e. those groups \(G\) such that there exists a normal subgroup \(G_1\) of \(G\) such that \(G/G_1\) is finite and \(G_1=Z(G_1)G_0\), he addressed the problem of the Langlands classification. In this article, the author considers admissible and tempered representations for the class of Lie groups whose Lie algebras are reductive, thus allowing for highly disconnected groups. This class is larger than the `really reductive' groups, which is larger than the class studied by \textit{J. A. Wolf} [Mem. Am. Math. Soc. 138 (1974; Zbl 0288.22022)] and \textit{R. A. Herb} and \textit{J. A. Wolf} [Compos. Math. 57, 271-355 (1986; Zbl 0587.22005); 58, 73-110 (1986; Zbl 0587.22006)]. Wolf's class is itself a broadening of the groups originally studied by Harish-Chandra. An important consequence of this broadening of the class of groups allowed is that the author must also modify what is meant by `admissible' and `tempered' representations. Here, the definition for `tempered' is in terms of the leading exponents, with the estimates by spherical functions derived as consequences. Another consequence is that the natural generalization of the discrete series representations for Harish-Chandra-style reductive groups (the quasi-discrete series representations) need not have an infinitesimal character. The first section of the article develops the notions of weakly admissible and admissible \(({\mathfrak g},K)\)-modules. A \(({\mathfrak g},K)\)-module is weakly admissible if all of its elements are \(K\)-finite and \({\mathcal Z}\)-finite. Here, \(K\) is the fixed point group of a Cartan involution on \(G\); it need not be compact. Also, \({\mathcal Z}\) is the center of the universal enveloping algebra of the complexification of \([{\mathfrak g},{\mathfrak g}]\), the semi-simple part of \({\mathfrak g}\). A weakly admissible module is admissible if its \(K\)-structure is unitarizable. A finitely-generated weakly admissible \(({\mathfrak g},K)\)-module is tempered if the leading exponents of its \(K\)-finite matrix coefficients satisfy a certain negativity condition. An arbitrary weakly admissible module is tempered if its finitely generated parts are tempered. By the nature of the topic, the paper is necessarily fairly technical. However, the last two sections end with convenient summaries of the main results. Thus, Section 2 ends with a list of equivalent conditions for a module to be rapidly decreasing. Section 3 ends with a discussion of the weakly admissible spectrum and the weakly tempered spectrum of \(G\). The paper contains two explicit examples for the group \(G=SL(2,{\mathbf R})\). Both are `almost principal' modules. The first is actually a family of examples where the inducing representation of \(MAN\) has \(A\) acting by \(2\times 2\) unipotent matrices. The second example has \(A\) acting by \(3\times 3\) unipotent matrices. In both examples, the underlying module of \(K\)-finite vectors of the resulting representation of \(G\) is admissible and tempered in Magyar's sense. However, in the second example, there is at least one matrix coefficient \(f\) such that \((f(\exp tH))/(t^2\phi_0(t\exp tH))\) does not go to zero as \(t \rightarrow \infty\). The usual requirement for a matrix coefficient of a tempered representation is that \(| f(g)| < c_f | \phi_0(g)| \), where \(\phi_0\) is the `standard' spherical function.