Bi-Lipschitz geometry of quasiconformal trees (Q2152774)

There are two long-standing, and difficult (or impossible), questions: \begin{itemize} \item[(a)] classify metric spaces up to quasisymmetric mappings; \item[(b)] to determine which metric spaces bi-Lipschitz embed into Euclidean space. \end{itemize} In the present paper, Question (a) is cast as giving a catalogue of metric spaces up to bi-Lipschitz maps. The simplest setting, where both these questions are understood, are doubling and bounded turning arcs, i.e., quasi-arcs, see the work of Herron and Meyer and Rohde, which are cited in the paper. Given this background, it is very natural to take the next step, to study compact metric trees, which satisfy the analogous doubling and bounded turning conditions. Such trees are called quasiconformal trees. The authors of the present paper focus first on an aspect of Question (a) by giving a catalogue of combinatorial models so that every metric tree is bi-Lipschitz to one of the trees in this catalogue. For arcs Rohde as well as Herron and Meyer have studied this question. This is related to earlier work by Bonk and Meyer, who showed that any quasiconformal tree is quasisymmetric to a geodesic tree. Here, the focus is all on bi-Lipschitz geometry, which requires substantial additional care. Indeed, the idea is to use the work of Bonk and Meyer to iteratively subdivide the tree into pieces at vertices of degree two. These pieces are then glued and subdivided in order to ensure that each of them has roughly the same size. The second component of the paper concerns the bi-Lipschitz embeddability. The authors show that if the leaves of the tree bi-Lipschitz embed, then so does the entire tree. As pointed out in the paper, this does not fully resolve the question, but it does yield some useful initial insights. The methods for this embedding are the following. We already know that every arc bi-Lipschitz embeds. Further, if the leaves embed, so does their closure. We are just left to consider the complement of the closure of the leaves which is an open set. At Whitney scales of this open set, the tree is a finite union of arcs. Thus, at such scales, one has embeddings. By using a natural idea from Seo (which has independently been discovered by a number of authors such as Semmes), the authors then patch up an embedding for the full space. The paper is very clearly written. First, the authors discuss Question (a), and then later in the paper Question (b). It is worth pointing out that Question (b) was fully resolved by the reviewer and the two authors in a subsequent paper. In that work, this paper was a crucial motivator.