A geometric realisation of tempered representations restricted to maximal compact subgroups (Q2200761)

Let \(G\) be a connected, linear, real reductive Lie group with compact centre and \(K \subset G\) a maximal compact subgroup. For a tempered irreducible representation \(\pi\) of G, the authors realise the restriction \(\pi_{\vert K}\) as the \(K\)-equivariant index of a Dirac operator on a homogeneous space \(G/H\), for a Cartan subgroup \(H \subset G\). The result in fact applies to every standard representation. To begin with, the authors draw on the result of Knapp-Zuckerman which shows \(\pi\) is equivalent to a basic representation \(\operatorname{Ind}_{MAN}^G(\sigma \otimes exp(i\nu)\otimes 1)\). Here, \(\sigma\) is a limit for discrete series for \(M\) and \(\nu \in \mathfrak a^\star\). Next, they fix a \(\theta\)-stable Cartan subgroup \(H\) for \(G\) so that its split part is \(A\). These gives rise a almost complex structure \(J\) on the tangent bundle to \(T(G/H)\). They consider \(\Lambda_J (T(G/H))\otimes L_\pi \rightarrow G/H\), here \(L_\pi \rightarrow G/H\) is a line bundle attached to \(\pi\) and \(\Lambda_J (T(G/H))\) is the complex exterior algebra bundle of \(T(G/H)\) with respect to \(J \). This bundle has a natural Clifford action and a connection, hence, a Dirac operator \[ D: \Gamma^\infty( \Lambda_J (T(G/H))\otimes L_\pi) \rightarrow \Gamma^\infty( \Lambda_J (T(G/H))\otimes L_\pi). \] For a suitable \(\psi \in \mathfrak g^\star\), let \(\Phi : G/H \rightarrow \mathfrak k\sim \mathfrak k^\star\) be \(\Phi(gH)=(\operatorname{Ad}(g)^\star \psi)_{\vert \mathfrak k}\), the map \(\Phi\) induces a vector field \(v^\Phi\) via the infinitesimal action by \(\mathfrak k\) on \(G/H\). A Theorem shown by Braverman gives that the multiplicities \(m_\delta^\pm\) of irreducible representation \(\delta\) of \(K\) in \(\operatorname{ker}(D -if c(v^\Phi)) \cap L^2(\Lambda_J^\pm T(G/H) \otimes L\pi )\) are finite, for function \( f \in \Gamma^\infty(G/H)^K\) with suitable growth behaviour. The difference \(m_\delta^+-m_\delta^-\) is independent of \( f\) and of the specific Dirac operator \(D\) used. The main Theorem is: \[ \pi_{\vert K} =\pm \sum_{\delta \in \hat K}(m_\delta^+-m_\delta^-)\delta.\]