Tempered representations for classical \(p\)-adic groups (Q470882)

Let \(F\) be a \(p\)-adic field and let \(G\) be a classical group over \(F\) by which we mean that \(G\) is either a symplectic, odd special orthogonal, even orthogonal or unitary group. The aim of the paper under review is to extend the classical work of \textit{C. Moeglin} and \textit{M. Tadić} [J. Am. Math. Soc. 15, No. 3, 715--786 (2002; Zbl 0992.22015)] classifying the discrete series of \(G\) to cover the case of tempered representations. Briefly, the Mœglin-Tadić classification gives a purely representation-theoretic way of parameterizing the square-integrable representations of \(G\) by associating to every discrete series \(\pi\) of \(G\) a certain triple \((\mathrm{Jord}_\pi,\varepsilon_\pi,\sigma_\pi)\). Here \(\mathrm{Jord}_\pi\) is a finite set of pairs \((\rho,a)\) where \(\rho\) is a self-dual (or conjugate-dual in the unitary case) supercuspidal representation of a general linear group and \(a\) is a positive integer whose parity is prescribed by \(\rho\). It is a substitute for the Langlands parameter of \(\pi\) and is defined in terms of the reducibility of certain parabolic inductions. The second part of the triple, \(\varepsilon_\pi\), is a ``partially'' defined function \(\mathrm{Jord}_\pi\to \{\pm 1\}\) (we refer the reader to section 1.3 of the paper or to the original article [Mœglin and Tadić, loc. cit.] for precision about the word ``partially'' here). Finally \(\sigma_\pi\), which is probably the most interesting part of the triple, is a supercuspidal representation of a group of the same type as \(G\). It is the partial cuspidal support of \(\pi\) meaning that \(\pi\) embeds in a representation of the form \(\mathrm{Ind}_P^G(\rho\otimes \sigma_\pi)\) for \(\rho\) a representation of a general linear group (this property uniquely characterizes \(\sigma_\pi\)). In [loc. cit.], Mœglin and Tadić define a set of admissible triples and they show that the map \(\pi\mapsto (\mathrm{Jord}_\pi,\varepsilon_\pi,\sigma_\pi)\) induces a bijection between discrete series of \(G\) onto the set of admissible triples. The author's extension of the Mœglin-Tadić classification involves the definition of some set of admissible quadruples \((\mathrm{Jord},\varepsilon,\sigma,m)\). Here Jord, \(\epsilon\) and \(\sigma\) are almost as before except that now more general pairs \((\rho,a)\) are allowed in Jord (\(\rho\) can now be any unitary cuspidal representation of a general linear group and there is no condition on the parity of \(a\)). The new ingredient is a multiplicity function \(m:\mathrm{Jord}\to \mathbb{N}\) which reflects the fact that the Langlands parameter of a tempered representation of \(G\) is no longer discrete in general. The main results of the paper, which are summarized in Subsection 1.4, is that there is a natural bijection between the set of tempered representations of \(G\) and the set of admissible quadruples. Moreover, as in the work of Mœglin-Tadić, the tempered representation corresponding to an admissible quadruple \((\mathrm{Jord},\varepsilon,\sigma,m)\) is entirely characterized by mean of parabolically induced representations in which it embeds. As noted by the author in the introduction, such a characterization is particularly well-suited for applying Jacquet module methods and this does not only apply to tempered representations but to the whole admissible dual of \(G\) thanks to the Langlands classification. The author explains that he expects to apply such a method to the study of the duality operator, which is defined in [\textit{A.-M. Aubert}, Trans. Am. Math. Soc. 347, No. 6, 2179--2189 (1995; Zbl 0827.22005)] or [\textit{P. Schneider} and \textit{U. Stuhler}, Publ. Math., Inst. Hautes Étud. Sci. 85, 97--191 (1997; Zbl 0892.22012)], on the categories of admissible representations of classical groups. The proof of the above-mentioned classification is given in Section 2. In the third section, the author gives additional properties of this classification and uses it to describe the \(R\)-group action on a parabolic induction \(\mathrm{Ind}_P^G(\lambda)\) where \(\lambda\) is a generic tempered representation of some Levi subgroup of \(G\). This last part appeals to results of Goldberg regarding the structure of the \(R\)-group [\textit{D. Goldberg}, Am. J. Math. 116, No. 5, 1101--1151 (1994; Zbl 0851.22021); Can. J. Math. 47, No. 2, 344--363 (1995; Zbl 0835.22015); Proc. Am. Math. Soc. 123, No.4, 1267-1276 (1995; Zbl 0856.22023)] and as a consequence the author has to assume the group to be split (quasi-split in the unitary case).