Exotic ideals in the Fourier-Stieltjes algebra of a locally compact group (Q2236000)

Let \(G\) be a locally compact group, and for any ideal \(D\) (often assumed translation invariant) of \(L^\infty(G)\), we say that a unitary representation \(\pi\) of \(G\) on a Hilbert space \(H\) is a \(D\)-representation if there is a dense subspace \(H_0\) of \(H\) such that every coefficient \(\pi_{\xi,\eta}\) is a member of \(D\), for each \(\xi,\eta\in H_0\). This idea was used in [\textit{N. P. Brown} and \textit{E. P. Guentner}, Bull. Lond. Math. Soc. 45, No. 6, 1181--1193 (2013; Zbl 1337.46039)] to study new \(C^\ast\)-completions of \(L^1(G)\) which lie between the reduced and full group \(C^\ast\)-algebras, \(C^*_r(G)\) and \(C^*(G)\). There has subsequently been much study of this idea, especially in the case when \(D = L^p(G)\) for some \(p\). The paper studies the more general idea of looking at certain ideals in \(B(G)\), the Fourier--Stieltjes algebra, the dual of \(C^*(G)\). We write \(A_{L^p}(G)\) and \(B_{L^p}(G)\) for the norm-closed, respectively, weak\(^*\)-closed, linear span of \(P(G)\cap L^p(G)\), where \(P(G)\subseteq B(G)\) is the positive definite functions. A great many results are shown for various classes of groups, but a unifying theme is that these spans are ``large'' in various senses, and that they are distinct for different \(p>2\). For example, defining \(A_{L^{p+}}(G)\) to be the intersection of \(A_{L^{p+\epsilon}}(G)\) for \(\epsilon>0\), then it is shown that \(A_{L^{p+}}(G) / A_{L^{p}}(G)\) is non-separable, for each non-compact abelian \(G\), and for each non-compact almost connected [IN]-group. Some remarks about bounded approximate identities are made. It is tempting to believe that we obtain the same spaces by looking at \(B(G) \cap L^p(G)\), instead of the linear span of \(P(G)\cap L^p(G)\), but there appears to be no obvious why this is actually true (despite \(B(G)\) being the span of \(P(G)\)). The authors exhibit the first known example where the resulting \(B_{L^p}(G)\) spaces do agree, with \(G=SL(2,\mathbb R)\). It remains an open problem whether this might hold for any \(G\), and various remarks around this problem are made. The paper is well-written, and in particular the second section carefully reviews the literature and the various objects to be studied, and would form a good starting point for other researchers interested in this area.