\(L^2\)-Betti numbers and Plancherel measure (Q2253133)

Given a locally compact unimodular group (with fixed Haar measure), its \(L^2\)-Betti numbers are the von Neumann dimensions of the (continuous) group cohomology of \(G\) with coefficients in \(L^2(G)\), where one realizes that these groups are modules (in the algebraic sense) over the group von Neumann algebra. This group von Neumann algebra has a semifinite trace induced by the unimodular Haar measure, and there is an associated von Neumann dimension for arbitrary modules (scaling with the Haar measure). In the paper under review, for postliminal such groups the \(L^2\)-Betti numbers are computed as an integral over the unitary dual with its Plancherel measure of the ordinary dimensions of the (reduced) cohomology with coefficients in irreducible unitary representations. In particular, the authors give explicit formulas for: the \(L^2\)-Betti numbers of semisimple Lie groups \(G\) with finite center (they vanish always, except if the rank of \(G\) is the rank of a maximal compact subgroup \(K\) and the degree is \(\dim(G/K)/2\), in which case the \(L^2\)-Betti number is strictly positive); simple algebraic groups over local fields with residue field of large enough order (they vanish if the degree is not the dimension of the associated Bruhat-Tits building, and are given in terms of the Steinberg module in the remaining case); automorphism groups of locally finite trees which act transitively on the boundary (they vanish if the degree is different from \(1\) and the first \(L^2\)-Betti number is given by the vertex degrees of the tree); postliminal locally compact unimodular semisimple groups with non-compact center (all \(L^2\)-Betti numbers vanish. It might be remarked that most of the explicit calculations are obtained without relying on the abstract main theorem. In the proof of Corollary 11, the authors state that \(L^2\)-cohomology is quasi-isometry invariant. Here, one has to be very careful with the context, otherwise this doesn't make sense (or becomes wrong). The equalities used in the proof, however, definitely hold.