Classification of symmetric pairs with discretely decomposable restrictions of \((\mathfrak{g},K)\)-modules (Q2515819)

The subject of this article is discretely decomposable restrictions of irreducible representations with respect to symmetric pairs. Starting with a unitary irreducible representation \(\pi\) of a Lie group \(G\), one asks how the restriction \(\pi|_H\) decomposes, for \(H\subset G\) a subgroup. In the case where \((G,H)\) is a pair of real reductive groups, let \(K, H\cap K\) be maximal compact subgroups of \(G\) and \(H\) respectively. Then one asks how the underlying \((\mathfrak{g},K)\)-module \(X\) of \(\pi\) behaves as an \((\mathfrak{g},H\cap K)\)-module. It was proved by the first author that either (1) \(X\) is discretely decomposable as an \((\mathfrak{h},H\cap K)\)-module, or (2) \(\mathrm{Hom}_{\mathfrak{h}, H\cap K}(Y,X)=0\) for any irreducible \((\mathfrak{h},H\cap K)\)-module \(Y\). In the first case one can decompose both \(\pi|_H\) and \(X_{(\mathfrak{h},H\cap K)}\) as a Hilbert and algebraic direct sum respectively. For example, (1) always holds for pairs \((G,H)\) with \(H=K\) because K is compact, but never holds for the pair \((\mathrm{SL}(n,\mathbb{C}),\mathrm{SL}(n,\mathbb{R}))\) if \(\dim X=\infty\). The paper focuses on symmetric pairs \((G,G^\sigma)\) where \(G^\sigma\) is the fixed point group of an involution \(\sigma\) of \(G\). This setting contains tensor products of representations taking \(\mathrm{diag} G\) to be the fixed point group of \(G\times G\) under the automorphism \(\sigma(x,y)=(y,x)\). In particular the authors consider the following problems: Problem A: Classify all the reductive symmetric pairs \((G,G^\sigma)\) for which there exists at least one infinite-dimensional irreducible \((\mathfrak{g},K)\)-module \(X\) satisfying the property (1). This problem splits into two cases, first when \(\mathfrak{g}\) is simple and second when \((G,G^\sigma)\) is the group case \((G' \times G',\mathrm{diag}G')\) with \(\mathfrak{g}'\) simple. The authors provide a complete solution to this problem on the Lie algebra level. For simple \(\mathfrak{g}\) they get a large family of such symmetric pairs including the cases \(G^\sigma=K\), \((\mathfrak{g},\mathfrak{g}^\sigma)\) is of holomorphic type and \(\mathfrak{g}^\sigma\) is the real form of a complex \(\mathfrak{g}\). Problem B: If \((\mathfrak{g},\mathfrak{g}^\sigma)\) is a reductive symmetric pair, find a criterion for determining which infinite-dimensional irreducible \((\mathfrak{g},K)\)-modules \(X\) are discretely decomposable as \((\mathfrak{g}^\sigma,K^\sigma)\)-modules. It is proved in the paper that it is always possible to find such a unitarizable \((\mathfrak{g},K)\)-module (assuming it exists). Furthermore, they prove a usable criterion for discrete decomposability of a minimal representation \(X\). The authors finally use their results to prove that the tensor product of two infinite-dimensional irreducible \((\mathfrak{g},K)\)-modules is discretely decomposable if and only if \(G=K\) is a Hermitian symmetric space and these modules are simultaneously highest weight modules or they are simultaneously lowest weight modules. Very roughly, the results suggest that the ``smaller'' X is, the more likely the restriction \(X_{\mathfrak{h}, H\cap K}\) is discretely decomposable.