Quasi-representations of groups and two-homology (Q2147980)

Two unitaries whose commutator has small norm give a quasi-representation of the group \(\mathbb{Z}^2\). It may happen that these unitaries are not close to exactly commuting unitaries. \textit{R. Exel} and \textit{T. Loring} [Proc. Am. Math. Soc. 106, No. 4, 913--915 (1989; Zbl 0677.15003)] showed that the nonvanishing of a certain winding number obstructs this. The group \(\mathbb{Z}^2\) is the fundamental group of a torus. The author [J. Topol. Anal. 4, No. 3, 297--319 (2012; Zbl 1258.46029)] has extended the Exel-Loring formula to quasi-representations of the fundamental group \(\Gamma_g\) of an oriented surface of genus \(g\ge1\). The first result in this article generalises this result further to a discrete group together with a class \(x\) in its second integral homology. Via the assembly map, \(x\) gives rise to a class in the \(K\)-theory of the \(\ell^1\)-Banach algebra of the group, and a quasi-representation to a finite matrix algebra attaches an integer to this. The first main theorem identifies this integer with the winding number of a certain loop. An analogous theorem holds for quasi-representations into a unital \(C^\ast\)-algebra with a finite trace. The setup above is linked to surface groups as follows: there are \(g\ge1\) and a group homomorphism \(f\colon \Gamma_g \to \Gamma\) so that the given homology class \(x\) is the \(f_*\)-image of a canonical element in \(H^2(\Gamma_g,\mathbb{Z})\). This is a key idea for the proof. The second main result in the article shows under some assumptions that there is a quasi-homomorphism for which the resulting integer is not zero. Of course, this only makes sense if \(x\) is not torsion. In addition, it is assumed that the group admits a \(\gamma\)-element and is isomorphic to a subgroup of the unitary group of a quasidiagonal \(C^\ast\)-algebra. Both assumptions hold, for instance, if the group in question is amenable.