Tensor Products of Analytic Continuations of Holomorphic Discrete Series (Q4390898)

One of the major problems in the representation theory of type I groups is the decomposition of tensor products of representations into direct integrals of irreducible representations. Such decompositions have already been established for some semi-simple Lie groups. In the present paper the authors treat \(SU(2,2)\). Let \(D\) be the bounded symmetric domain of contractive complex \(2\times 2\)-matrices, and for \(\nu>3\) let \(H_\nu\) denote the Hilbert space of holomorphic functions \(f\) on \(D\) with the norm \(c_\nu \int_D \bigl | f(Z) \bigr|^2 \text{det} (1-Z^*Z)^{\nu-4} dZ<\infty\) \((c_\nu\) a suitable constant). \(G\) acts unitarily on \(H_\nu\) by \[ \pi_\nu \left( \begin{matrix} A & B \\ C & D \end{matrix} \right) f(z)= \text{det}\bigl( (CZ+D) \bigr)^{-\nu} f\left( \left( \begin{matrix} A & B \\ C & D\end{matrix} \right)Z \right). \] Then \((H_\nu, \pi_\nu)\) has an analytic continuation in \(\nu\), and this yields a unitary representation for every \(\nu\geq 1\). The objective is to find the explicit direct integral decomposition of \(\pi_\nu \otimes \overline \pi_{\nu}\), respectively \(H_\nu \otimes \overline H_\nu\). It turns out that, partly due to different orthogonality relations, one has to distinguish the cases \(\nu=1\), \(1<\nu <{3\over 2}\) and \(\nu\geq {3\over 2}\). For simplicity we only outline the result in the latter case. Let \({\mathfrak g}\) be the Lie algebra of \(G\) and take as a maximal abelian subspace corresponding to a Cartan decomposition \[ {\mathfrak a}= \left\{a= \left( \begin{matrix} 0 & \text{diag} (s,t)\\ \text{diag} (s,t) & 0 \end{matrix} \right): s,t\in\mathbb{R} \right\}. \] Define linear functionals on \({\mathfrak a}\) by \(f_1(a)=s\) and \(f_2(a)=t\). Write elements of \({\mathfrak a}^*_\mathbb{C}\), the complexification of the dual of \({\mathfrak a}\), as \(\underline \lambda= \lambda_1 f_1+ \lambda_2 f_2\). Now, for \(\underline\lambda\in {\mathfrak a}^*_\mathbb{C}\), the \(G\)-translates of the associated spherical function generate a Hilbert space \({\mathcal H} (\underline \lambda)\), and the representation of \(G\) on \({\mathcal H} (\underline \lambda)\) is equivalent to the induced representation \(\text{ind}^G_P (1\otimes i\underline \lambda \otimes 1)\), where \(P\) is the minimal parabolic subgroup corresponding to \({\mathfrak a}\). Then \[ H_\nu \otimes \overline H_\nu =\int^\oplus_{\lambda_1 >\lambda_2 >0} {\mathcal H} (\underline \lambda) B_\nu (\underline \lambda) | C (\underline\lambda) |^{-2} d\underline \lambda. \] (Theorem 2.13). Here \(C(\underline \lambda)\) denotes the Harish Chandra \(C\)-function of \(G\) and \(B_\nu (\underline \lambda)\) is the symbol of the Berezin transform on the domain \(D\). The results of the paper are very interesting and appear to be difficult, and the authors indicate that some of these extend to more general situations.