Hardy spaces for non-compactly causal symmetric spaces and the most continuous spectrum (Q1411993): Difference between revisions

Let \(G/H\) be a pseudo-Riemannian semisimple symmetric space. As is known, the quasiregular representation of \(G\) on \(L^2(G/H)\) decomposes into a finite sum of series of representations induced from parabolic subgroups of \(G\). The goal of the so-called Gelfand-Gindikin program (1977) was to give a geometric description of subspaces of \(L^2(G/H)\) where different summands of the above finite sum act. The main idea consisted in studying Hardy spaces on certain complex manifolds and in taking limit values of functions in these Hardy spaces. These complex manifolds were supposed to be submanifolds of complexifications of the space \(G/H\). The first step in this program was the realization of the \textit{discrete} part of the spectrum, more exactly, of the holomorphic discrete series for spaces \(G/H\) of Hermitian type, or compactly causal symmetric spaces (I. M. Gelfand, S. G. Gindikin, G. I. Olshanskii, R. J. Stanton, G. O'lafsson, J. Hilgert, B. Ørsted, V. F. Molchanov, B. Kroetz). In contrast to this first step, the paper under review is devoted to the opposite part of the spectrum, namely, to the \textit{most continuous part} of the spectrum, i.e., to the series of representations induced from a minimal parabolic subgroup. It is the first time when the most continuous part of the spectrum acquires a geometric realization -- in the framework of the above-mentioned program. This description is given for a subclass of the class of the non-compactly causal symmetric spaces, see below. A causal symmetric space is a semisimple symmetric space \(G/H\) (\(H\) is an open subgroup of the fixed point subgroup of an involution \(\sigma\)) equipped with an invariant causal structure, i.e. to any point \(x\) in \(G/H\) there is assigned a cone in the tangent space at \(x\) and this correspondence is \(G\)-invariant. Equivalently, in the decomposition \({\mathfrak g} = {\mathfrak h} + {\mathfrak q}\) of the Lie algebra \(\mathfrak g\) of \(G\) into \(\pm 1\)-eigensubspaces of \(\sigma\), the space \(\mathfrak q\) has an \(H\)-invariant cone (open, convex, containing no affine lines). The space \(G/H\) is called non-compactly causal, if it is irreducible and the cone consists of elements with real eigenvalues in the adjoint representation. Let \(K\) be a maximal compact subgroup of \(G\) invariant with respect to the involution \(\sigma\). Let \(\Xi\) be the \textit{complex crown} of the Riemannian symmetric space \(G/K\) (this notion was introduced by Akhiezer and Gindikin in 1990). The subclass of spaces \(G/H\) considered in the paper consists of such \(G/H\) that the crown \(\Xi\) is a Hermitian symmetric space \(S/U\) (so that \(G\subset S\) and \(K\subset U\)). In this case the Shilov boundary \(\partial_S\Xi\) of \(\Xi\) is just \(G/H\). All these spaces are classified (from the local point of view), the list has 8 items (7 classical and 1 exceptional). The Hardy space \({\mathcal H}^2(\Xi)\) differs from the Hardy spaces in the realizations of \(S/U\) as a bounded domain or as a tube domain. The group \(G\) acts on this space by translations. Denote this representation by \(L\). The fact that \(G/K\) is a totally real submanifold of \(S/U\) allows to construct the generalized Segal-Bargmann transform. This transform is a unitary \(G\)-isomorphism of \(L^2(G/K)\) onto \({\mathcal H}^2(\Xi)\). It gives that the representation \(L\) of \(G\) is a direct integral of principal series representations (with multiplicities one). On the other hand, the boundary map \(b\) which to any function in \({\mathcal H}^2(\Xi)\) assigns its limit values at \(G/H\) is a \(G\)-equivariant isometric embedding. Therefore, \(\text{ Im }b\) is a full subspace of the most continuous part \(L^2(G/H)_{\text{mc}}\) of \(L^2(G/H)\) -- but with multiplicity one. Since the space \(L^2(G/H)_{\text{mc}}\) is not multiplicity free, the problem to separate different multiplicities remains still open.