Tempered representations of semisimple Lie groups
Representation theory of semisimple (or reductive) Lie groups is a rich and vast area. The area is, of course, dominated by the monumental work by Harish-Chandra in which he computed the Plancherel measure. In this excellent article the author explains in detail the classification of almost all tempered representations starting from the definition of a linear semisimple Lie group. The article is sprinkled with many examples which make the general theory clearer even to a first-time reader.\N\NSections 1 and 2 cover the preliminaries. In particular, it is assumed that \(G\) is a linear Lie group (closed subgroup of \(GL(n, \mathbb C)\)) which is closed under conjugate transposition and has finite center (which is taken as the definition of being semisimple). Examples include the orthogonal group \(O(n)\) (disconnected), the special orthogonal group \(SO(n),\) the unitary group \(U(n)\) (reductive) and \(SU(n), Sp(n)\) (all compact). In the non-compact case \(SL(n, \mathbb C), SO(n, \mathbb C), Sp(n, \mathbb C)\) are complex semisimple Lie groups and \(SL(n, \mathbb R), SO(p, q), SU(p, q)\) are real semisimple Lie groups. Their Lie algebras, Cartan subgroups, root space decomposition, restricted roots and nilpotent Lie algebras that come up, are explained. This provides a quick revision of what is needed to follow the rest of the article.\N\NSection 3 deals with tempered representations which is the heart of the matter. Let \(G\) be a locally compact topological group of type I. A unitary irreducible representation of \(G\) is called tempered if it lies in the support of the Plancherel measure of \(G.\) In other words a \(\pi \in \widehat{G}\) is tempered if and only if every neighbourhood of \(\pi\) in the Fell topology has positive Plancherel measure. For example, all unitary irreducible representations of a compact group are tempered (by Peter-Weyl theorem).\N\NNext, let \(G\) be a connected semisimple linear Lie group and \(K\) a fixed maximal compact subgroup. If \(\pi\) is a representation of \(G\) on a vector space \(V,\) a vector \(v \in V\) is called \(K\)-finite if its \(K\)-orbit under the \(\pi\) action spans a finite-dimensional space. If \(\pi\) is unitary, it is known that the subspace of \(K\)-finite vectors is dense in \(V.\)\N\NTheorem. Suppose that \(G\) is a connected, semisimple, linear Lie group. Then a representation \(\pi \in \widehat{G}\) is tempered if and only if the function \N\[\Ng \to \langle v, \pi(g)w \rangle\N\]\N is in \(L^{2+\varepsilon}(G)\) for all \(K\)-finite vectors \(v\) and \(w,\) and for all \(\varepsilon > 0.\) \N\NAfter stating the above theorem and discussing the example of \(SL(2, \mathbb R)\) the author dives into discrete series representations.\N\NProposition. Let \(G\) be a connected semisimple linear Lie group and \(\pi \in \widehat{G}.\) The following are equivalent:\N\begin{itemize}\N\item[1.] For all \(w, v \in V_\pi,\) the function \(g \to \langle v, \pi(g)w \rangle\) lies in \(L^2(G)\);\N\item[2.] The Plancherel measure of \([\pi] \in \widehat{G}\) is positive.\N\end{itemize}\NA unitary irreducible representation \(\pi\) is called a discrete series representation if it satisfies the equivalent conditions above. A famous theorem due to Harish-Chandra provides a necessary and sufficient condition for \(G\) to have discrete series.\N\NTheorem. [Harish-Chandra] The group \(G\) has discrete series representations if and only if \(G\) has a compact Cartan subgroup. \N\NThe author provides a table of classical groups and points out which ones have discrete series representations. Next, a deep theorem due to Harish-Chandra which gives a parametrization of discrete series representations is stated.\N\NTheorem. Suppose that \(G\) has a compact Cartan subgroup \(T.\)\N\N\begin{itemize}\N\item[1.] For regular \(\lambda \in i\mathfrak{t}^\ast\) such that \(\lambda + \rho_\lambda^G\) is analytically integral, there is a discrete series representation \(\pi_\lambda^G\) of \(G\) for which highest weights of all \(K\)-types are of the form \(\lambda + \rho_\lambda^G-2 \rho_\lambda^K +\sum_{\alpha \in R^+_\lambda(\mathfrak{k}, \mathfrak{t})}~n_\alpha \alpha, \quad n_\alpha \geq 0.\)\N\item[2.] Two discrete series representations \(\pi_\lambda\) and \(\pi_\mu\) are unitarily equivalent if and only if there exists a \(w \in W(K, T)\) such that \(\mu = w \cdot \lambda.\)\N\item[3.] Every discrete series representation occur as in part (\(1\))\N\end{itemize}\N\NThe above was a landmark achievement by Harish-Chandra, as this was necessary to construct all principal series representations (induced from parabolic subgroups). After discussing the example of \(SL(2, \mathbb R)\) (discussing an example right after a main theorem is really useful), the author moves on to parabolic induction. Parabolic induction refers to unitary representations that are induced from (cuspidal) parabolic subgroups of \(G.\) Such a subgroup \(P\) has a decomposition \(P= MAN\) where \(M\) is reductive (may be disconnected), \(A\) is abelian and \(N\) is nilpotent. When \(P\) is the minimal parabolic subgroup, \(M\) is compact. The subgroup \(MA\) is called the Levi factor of \(P.\) Cuspidal means that the factor \(M\) has a compact Cartan subgroup and hence \(M\) has discrete series representations. However, \(M\) might be disconnected and so Harish-Chandra needed a result which is more general than the parametrization of the discrete series above (see Theorem 2.4.14 in this article). There are only finitely many cuspidal parabolic subgroups \(P\) (up to conjugation) and one has \(G = KP\) always. At this stage several examples are discussed. Let \(\sigma\) be a discrete series representation of \(M,\) \(\nu\) an irreducible representation of \(A\) (unitary character) and \(1_N\) the trivial representation of \(N.\) Consider \(\sigma \otimes \nu \otimes 1_N,\) which is a unitary representation of \(P=MAN.\) Normalized induction \(\text{Ind}_P^G(\sigma \otimes \nu \otimes 1_N)\) produces a unitary representation of \(G\) (note that \(G/P\) does not carry a \(G\)-invariant measure and so slight modification of the usual induction procedure is needed). Next comes another major result:\N\NTheorem. An irreducible unitary representation \(\pi\) of \(G\) is tempered if and only if there exists a cuspidal parabolic subgroup \(P = MAN,\) a discrete series representation \(\sigma \in \widehat{M}\) and \(\nu \in \widehat{A}\) such that \(\pi\) is unitarily equivalent to a subrepresentation of \(\mathrm{Ind}_P^G (\sigma \otimes \nu \otimes 1_N).\)\N\NNote that the above includes the discrete series representations (if they exist) of \(G\) (these are the ones obtained when \(P = G\)).\N\NThree pictures of parabolic induction are described in detail including the compact picture where the representation is realised on \(L^2(K)\) and the non compact picture where the representation is realised on \(L^2(\overline{N})\) (here \(\overline{N} = \theta (N)\)). The latter picture brings in the Knapp-Stein intertwining operators \(\mathcal{A}_P(w, \sigma, \nu)\) which intertwine the representations \(\text{Ind}_P^G(\sigma \otimes \nu \otimes 1_N)\) and \(\text{Ind}_P^G(w \cdot \sigma \otimes w \cdot \nu \otimes 1_N).\) These operators turn out to be singular integral operators on nilpotent groups. Completeness theorem due to Harish-Chandra can now be formulated in terms of these intertwining operators. If \(\sigma\) belongs to the discrete series of \(M,\) the algebra of intertwining operators of \(\text{Ind}_P^G(\sigma \otimes \nu \otimes 1_N)\) is spanned by \(\{\sigma(w) \mathcal{A}_P(w, \sigma, \nu):~w \in W_{\sigma, \nu} \}\) where \(W_{\sigma, \nu}\) is a certain explicit subgroup of \(W.\) It follows from Schur's lemma that \(\text{Ind}_P^G(\sigma \otimes \nu \otimes 1_N)\) is irreducible provided \(W_{\sigma, \nu}\) is trivial. This is the case if \(\nu\) is regular (which means that no element in \(W\) fixes \(\nu\)). Thus, for an open dense subset of \(\widehat{A}\) we have irreducibility. Using Langlands' disjointness theorem along with the above results leads to the following classification of almost all tempered representations:\N\NTheorem. Let \(H_1, H_2, \cdots H_n\) be \(\theta\)-stable representatives of conjugacy classes of Cartan subgroups of \(G.\) Let \(P_j =M_jA_jN_j\) be the cuspidal parabolic subgroup associated to \(H_j\) and a choice of positive restricted roots. The representations \(\mathrm{Ind}_{P_j}^G(\sigma_j \otimes \nu_j \otimes 1_{N_j})\) where \(\sigma_j \in \widehat{M_j}\) is a discrete series representation and \(\nu_j \in \widehat{A_j}\) is regular are irreducible and tempered. Almost every irreducible tempered representation is of this form. Two such representations \(\mathrm{Ind}_{P_j}^G(\sigma_j \otimes \nu_j \otimes 1_{N_j})\) and \(\mathrm{Ind}_{P_{j^{'}}}^G(\sigma^{'}_{j^{'}} \otimes \nu^{'}_{j^{'}} \otimes 1_{N_j^{'}})\) are equivalent if and only if \(j = j^{'}\) and there exists \(w \in W_j = N_K(\mathfrak{a}_j)/(K\cap M_j )\) such that \(\sigma_j^{'} \cong w \cdot \sigma_j\) and \(\nu_j^{'} = w \cdot \nu_j.\)\N\N``Almost every in the above means that we obtain all tempered representations in a dense open subset of the parameter space associated with each conjugacy class of Cartan subgroups, which is enough to compute the Plancherel measure. Classification of all tempered representations by \textit{A. W. Knapp} and \textit{G. J. Zuckerman} [Ann. Math. (2) 116, 389--455 (1982; Zbl 0516.22011)] is also discussed in some detail. In the final section all these results are used to compute the \(K\)-groups (\(K_0(G)\) and \(K_1(G)\)) of the reduced \(C^\ast\)-algebra of \(G.\)\N\NFor the entire collection see [Zbl 1555.19001].
This page was built for publication: Tempered representations of semisimple Lie groups
Report a bug (only for logged in users!)Click here to report a bug for this page (MaRDI item Q6918770)
