Multiplicity one theorem for \((\text{GL}_{n+1}({\mathbb{R}}), \text{GL}_{n} ({ \mathbb{R}}))\) (Q1047911)

Let \(F\) be an archimedean local field, i.e., \(F\) is either the field of the real or the complex numbers. In the paper under review the authors prove a result on representations of \(\text{GL}_n(F)\) which appear as restrictions of representations of the group \(\text{GL}_{n+1}(F)\) with \(\text{GL}_n(F)\) embedded in the upper left. To formulate this result, called Theorem B in the paper, consider irreducible admissible smooth Fréchet representations \(\pi: \text{GL}_{n+1}(F)\to \text{GL}(V)\) and \(\tau : \text{GL}_n(F) \to \text{GL}(W)\). Then Theorem B asserts that the vector space of continuous \(\text{GL}_n(F)\)-equivariant linear maps \(\varphi : V \to W\) has dimension at most 1. This strengthens the results of an earlier paper of the same authors (jointly with \textit{E. Sayag}) [Compos. Math. 144, No. 6, 1504--1524 (2008; Zbl 1157.22004)], where \(\tau\) is the trivial representation of \(\text{GL}_n(F)\). Moreover, an analogous result for non-archimedean local fields of characteristic zero is contained in a joint paper of the two authors with S. Rallis and G. Schiffmann, which will appear in Ann. of Math., see [arxiv:0709.4215]. Theorem B is deduced from a result about distributions on \(\text{GL}_{n+1}(F)\), named Theorem A in the paper. It says that every \(\text{GL}_n(F)\)-invariant distribution on \(\text{GL}_{n+1}(F)\) is invariant with respect to the transposition of matrices. Similar results are known for other pairs of a group \(G\) and a subgroup \(H\), such as symmetric groups, orthogonal groups and unitary groups. The way to deduce the representation theoretic result from the corresponding result on distributions is fairly standard and goes back to the classical theory of \textit{I. M. Gelfand} and \textit{D. A. Kazhdan} [Lie Groups Represent., Proc. Summer School Bolyai János math. Soc., Budapest 1971, 95--118 (1975; Zbl 0348.22011)]. The proof of Theorem A however is quite involved and consumes most of the 24 pages of the paper. It uses Fourier transform, Harish-Chandra descent and some new ingredients, such as D-modules. The interested reader is directed to a paper of Binyong Sun and Chen-Bo Zhu [arXiv:0903.1413], where the main results of the paper under review have been obtained in a different way and are presented in a slightly more readable manner (in the reviewer's opinion).