Approximating \(L^2\)-invariants by their classical counterparts (Q502141)

Suppose that \(G\) is a discrete group and \(G_0\supseteq G_1\supseteq\cdots\) is a descending chain of subgroups, where \(G_i\) is normal in \(G\), the index \([G:G_i]\) is finite, and the intersection taken over all subgroups is trivial. For a \(G\)-covering \(\overline{X}\rightarrow X\), let \(X[i]\) denote the quotient space \(G_i\backslash\overline{X}\), and \(X[i]\rightarrow X\) the induced covering map. For a topological invariant \(\alpha\), the \textit{\(L^2\)-approximation conjecture} states that the limit of the quotient \(\alpha(X[i])/[G:G_i]\) as \(i\rightarrow\infty\) is equal to the corresponding invariant \(\alpha^{(2)}(\overline{X})\) in \(L^2\)-(co)homology, where the latter has coefficients in the von Neumann group algebra of \(G\). The paper gives an extensive and in-depth survey of currently known results for which the \(L^2\)-approximation conjecture holds and for which it fails. Among the invariants considered are the Euler characteristic, Betti numbers, the manifold signature, the minimal number of generators of a homology group, the truncated Euler characteristic, the Fuglede-Kadison determinant, and \(L^2\)-torsion. Also addressed is the speed of convergence in the approximation conjecture.