8

Are there geometric characterizations of property (T) and the Haagerup property, i.e., such that can be stated in terms of a Cayley graph, in analogy to the Folner set characterization of amenability?

Vladimir
  • 1,210
  • 6
    In principle yes, because f.g. groups $G_1,G_2$ with the same Cayley graph satisfy: $G_1$ has T iff $G_2$ has T, and the same for Haagerup. But there is no known definition in these terms. And unlike amenability there is no coarse characterization, because these are not QI-invariants (for Haagerup it's a recent result of Carette). – YCor May 23 '14 at 17:25
  • Have you seen this question? http://mathoverflow.net/q/154431/1345 – Ian Agol May 24 '14 at 03:50
  • @Ian I don't see any close links about these questions. Vladimir's question is whether Property T can be defined purely in terms of the (unlabeled) Cayley graph. Ozawa's question is about algorithmic recognition of Property T from a presentation. That the set of presentations of Prop T groups is enumerable is not what I'd call a geometric characterization of Cayley graphs. – YCor May 24 '14 at 08:58
  • @YvesCornulier: I assumed that Vladimir meant the labeled Cayley graph. My understanding of the Folner condition is that it implicitly uses the group multiplication, so I assumed that he meant to incorporate the group action. There is a graph-theoretic intrinsic characterization due to Brooks, e.g. in terms of the Cheeger constant or smallest eigenvalue $=0$, but this is not mentioned. – Ian Agol May 24 '14 at 17:47
  • @Ian: the Følner condition can be made purely metric, since: if $F$ is the Følner set (that is $|SF-F|/|F|$ is small where $S$ is the generating set), then the 1-neighbourhood of $F^{-1}$ in the Cayley graph is small with respect to $|F|$. This condition (existence of Følner sets) is actually a QI-invariant for connected graphs of bounded valency. – YCor May 24 '14 at 18:20
  • Thanks everyone - this is a very interesting discussion. @Ian: I didn't know about this - thanks a lot! – Vladimir May 27 '14 at 02:35
  • @YvesCornulier: could you please direct me to the reference showing that property (T) is not QI-invariant? Thanks! – Vladimir May 27 '14 at 02:35
  • @Vladimir: see the answer to this question: http://mathoverflow.net/a/98727/1345 – Ian Agol May 27 '14 at 03:48
  • @Vladimir: the Bekka-Harpe-Valette book. The examples are one cocompact lattice in $G\times\mathbf{R}$ and one cocompact lattice in $\tilde{G}$, where $G$ is a simple lie group of rank $\ge 2$ and infinite fundamental group, e.g. $Sp_{2n\ge 4}(\mathbf{R})$ or $SO(2,n\ge 3)$. – YCor May 27 '14 at 08:43

0 Answers0