Quasi-isometric rigidity of subgroups and filtered ends (Q2106586): Difference between revisions

Suppose \(X\) and \(Y\) are metric spaces with collections of subspaces \(\mathcal{A}\) and \(\mathcal{B}\) respectively. A quasi-isometry \(q\colon X \to Y\) is a \textit{quasi-isometry of pairs} \(q\colon (X,\mathcal{A}) \to (Y,\mathcal{B})\) if there exists \(M > 0\) such that each subspace in \(\mathcal{A}\) is sent to within Hausdorff distance \(M\) of a subspace in \(\mathcal{B}\), and conversely every subspace in \(\mathcal{B}\) is within Hausdorff distance \(M\) of the image of some subspace in \(\mathcal{A}\). The authors define what it means for a set of subspaces of a metric space \(X\) (for example a set of cosets of a collection of subgroups of a finitely generated group \(G\) equipped with a word metric) to be \textit{quasi-isometrically characteristic.} A first guess at the definition would be to suppose that \(\mathcal{A}\) is quasi-isometrically characteristic if every quasi-isometry \(X \to X\) is automatically a quasi-isometry of pairs \((X,\mathcal{A}) \to (X, \mathcal{A})\) with constant \(M\) depending only on the quality of the quasi-isometry. This turns out to be too permissive a definition to be useful. Instead the authors add the following two conditions. \begin{itemize} \item[1.] Suppose \(B \subset X\) is bounded. Then \(B\) meets only finitely many subspaces in \(\mathcal{A}\) up to the relation of finite Hausdorff distance. \item[2.] For each subspace \(A\) in \(\mathcal{A}\), there is a bound (depending on \(A\)) to the Hausdorff distance between \(A\) and any \(A' \in \mathcal{A}\) if it is finite. \end{itemize} For example, if \(X\) is a \(\delta\)-hyperbolic space and \(\mathcal{A}\) is the collection of all quasi-geodesics in \(X\) at bounded Hausdorff distance from a given bi-infinite geodesic, then condition (1) is satisfied, but in general if \(\mathcal{A}\) is the collection of \textit{all} quasi-geodesics in \(X\), condition (1) will be violated, even though this latter collection is stable under quasi-isometries. The second condition (2) rules out, for example, the case where \(\mathcal{A}\) is a collection of singletons which is coarsely dense in \(X\), like the integer points in \(\mathbb{R}\). So in some sense these conditions ensure that \(\mathcal{A}\) is not ``too large''. For a finitely generated group \(G\) equipped with the word metric and \(\mathcal{A}\) a collection of cosets of subgroups \(P \le G\), these extra conditions turn out to force the number of subgroups \(P\) to be finite and for each \(P\) to have finite index in its commensurator in \(G\). The main result of the paper is that if \(X\) is a metric space with a quasi-isometrically characteristic collection of subspaces \(\mathcal{A}\), then any finitely generated group \(G\) quasi-isometric to \(X\) possesses a finite collection of subgroups whose cosets form a quasi-isometrically characteristic collection of subspaces of \(G\), and in fact these subgroups are a well-chosen finite collection of ``quasi-stabilizers'' of subspaces in \(\mathcal{A}\) under the induced quasi-action of \(G\) on \(X\). As an application, the authors prove that if \(P\) is a quasi-isometrically characteristic subgroup of \(G\) and \(H\) is quasi-isometric to \(G\), then \(H\) contains a subgroup \(Q\) such that the number of \textit{filtered ends} (a notion introduced by \textit{B. H. Bowditch} [Trans. Am. Math. Soc. 354, No. 3, 1049--1078 (2002; Zbl 0988.20027)] as \textit{coends}) of \(G\) with respect to \(P\) is equal to the number of filtered ends of \(H\) with respect to \(Q\). The paper is clear and offers a useful common abstraction of arguments appearing elsewhere in the literature. Particularly nice is the careful setting of the notion of \textit{ends}, which is primarily topological, in the category of \textit{metric spaces}.