On the complexity of the quasi-isometry and virtual isomorphism problems for finitely generated groups (Q931867)

Let \(S\) and~\(T\) be finite sets of generators of groups \(G\) and~\(H\) with word metrics \(d_S\) and~\(d_T\), respectively. The groups \(G\) and~\(H\) are said to be quasi-isometric, written \(G\approx_{\text{QI}}H\), if there exist \(\lambda\geq1\), \(C\geq0\), and a~mapping \(\varphi:G\to H\) such that \(\frac1\lambda d_S(x,y)-C\leq d_T(\varphi(x),\varphi(y))\leq \lambda d_S(x,y)+C\) for all \(x,y\in G\) and \(d_T(z,\varphi[G])\leq C\) for all \(z\in H\). Two finitely generated groups \(G_1\) and~\(G_2\) are said to be virtually isomorphic up to finite kernels, written \(G_1\approx_{\text{VI}}G_2\), if there exist subgroups \(N_i\leq H_i\leq G_i\) for \(i=1\),~\(2\) such that \([G_i:H_i]<\infty\), \(N_i\)~is a~finite normal subgroup of~\(H_i\) for \(i=1\),~\(2\) and \(H_1/N_1\cong H_2/N_2\). It is known that virtually isomorphic finitely generated groups are quasi-isometric. The converse does not hold and it is natural to conjecture that the quasi-isometry relation is strictly more complex than the virtual isomorphism relation in the sense of Borel reducibility. A~known theorem of the author of the paper that the quasi-isometry relation~\(\approx_{\text{QI}}\) is strictly more complex than the isomorphism relation~\(\cong\) on the space of finitely generated groups can serve as a~justification of this conjecture. In the paper under the review the author studies the Borel complexity of the relations \(\cong\), \(\approx_{\text{QI}}\), and \(\approx_{\text{VI}}\) for the class of finitely generated groups. He proves that they are \(K_\sigma\)-equivalence relations. The quasi-isometry relation makes sense for arbitrary metric spaces, and the author proves that the quasi-isometry relation on the space of connected \(4\)-regular graphs is a~universal \(K_\sigma\)-equivalence relation. The conjecture that \(\approx_{\text{QI}}\) is a~universal \(K_\sigma\)-equivalence relation remains open. But \(\approx_{\text{VI}}\) for finitely generated groups is shown to be a~non-universal \(K_\sigma\)-equivalence relation.