Tree-like is not a transitive relation on paths

Jeremy Brazas, Gregory R. Conner, Paul Fabel, Curtis Kent · 2026

Read Tree-like is not a transitive relation on paths by Jeremy Brazas, Gregory R. Conner, Paul Fabel, Curtis Kent (2026) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2603-16774.txt.

Introduction

Brazas and co-authors revisit a question about tree-like paths in metric spaces. Hambly and Lyons showed that Lipschitz tree-like paths form an equivalence relation on paths of bounded variation. They asked whether the same holds without the Lipschitz condition. The authors show it does not. Once that condition is dropped, the tree-like relation stops being transitive, so it is no longer an equivalence relation. Their counterexample comes from an explicit fractal construction in the plane.

Catalog record

CREATORJeremy Brazas, Gregory R. Conner, Paul Fabel, Curtis Kent
YEAR2026
COLLECTIONPAPERS
FORMATTXT
FILE SIZE31611
LICENSEhttp://creativecommons.org/licenses/by/4.0/
RIGHTSCC BY
SOURCEarXiv

Source: https://arxiv.org/abs/2603.16774

Browse the full library · About the Tower of Babel

Full text

The opening of the work, from the library's hosted text:

Tree-like is not a transitive relation on paths Jeremy Brazas, Gregory R. Conner, Paul Fabel, Curtis Kent Abstract . The notions of tree-like loop and Lipschitz tree-like loop were introduced by Hambly and Lyons in their 2010 Annals of Mathematics paper. They showed that the Lipschitz tree-like property determines an equivalence relation on the set of paths of bounded variation in a given metric space and then asked if this notion could be extended to paths without the Lipschitz requirement. We show that after eliminating the Lipschitz requirement, the resulting relation is no longer transitive and thus is not an equivalence relation. The counterexample is obtained by analyzing an explicit fractal construction in the plane. 2010 Mathematics Subject Classification 54F50, 55R65, 57M10 1. Introduction In their 2010 Annals of Mathematics paper, Hambly and Lyons considered a notion of a path with bounded variation being tree-like [ 5 ] . Their definition encoded \mathbb{R} -trees using positive continuous functions to the real line, called height functions. Definition 1.1 ( [ 5 ] ) . A continuous function \alpha:[0,1]\to X in a metric space (X,d) is tree-like if there exists a positive real-valued continuous function h:[0,1]\to\mathbb{R} such that, for all s\leq t , d\big(\alpha(t),\alpha(s)\big)\leq h(t)+h(s)-2\inf\limits_{u\in[s,t]}h(u). The function h is a height function for \alpha . Two paths \alpha,\beta:[0,1]\to X are tree-like equivalent , if \alpha*\overline{\beta} is tree-like. We say \alpha is a Lipschitz tree-like path if \alpha has a height function with bounded variation. Two paths \alpha,\beta:[0,1]\to X are Lipschitz tree-like equivalent , if \alpha*\overline{\beta} is Lipschitz tree-like. Hambly and Lyons showed that Lipschitz tree-like is an equivalence relation on the set of paths with bounded variation and that each equivalence class has a unique (up to reparameterization) representative of shortest length. They then showed that, for paths of bounded variation in \mathbb{R}^{d} , the signature of a path is trivial if and only if the path is Lipschitz tree-like [ 5 , Theorem 4] . Hambly and Lyons asked how important was the finite length condition and posed the following two questions. Question 1.2 . [ 5 , Problem 1.8] Does tree-like equivalent (without requiring the height function to be Lipschitz) define an equivalence relation on the set of continuous paths? Question 1.3 . [ 5 , Problem 1.9] Is there a unique tree reduced path associated to any continuous path? The main result of this paper, Theorem 6.1 , gives a negative answer to both of these questions. To prove this theorem, we require an alternative definition of “tree-like” that was introduced by Boedihardjo, Geng, Lyons, and Yang in [ 1 ] and which applies to paths that do not have bounded variation. The same definition was considered independently in [ 2 ] where it was called “ \mathbb{R} -tree nullhomotopic.” To clarify the distinction for general paths, we will use this latter term (see Definition 1.4 ). Boedihardjo et al. were able to show that weakly geometric rough paths have trivial signature if and only if the path is \mathbb{R} -tree nullhomotopic. Definition 1.4 ( [ 1 , 2 ] ) . We say that \alpha:[0,1]\to X is \mathbb{R} -tree nullhomotopic , if \alpha factors through a loop in an \mathbb{R} -tree, that is, if there exists an \mathbb{R} -tree E , a loop \pi:[0,1]\to E and a map g:E\to X such that g\circ\pi=\alpha . We say that paths \alpha,\beta:[0,1]\to X are \mathbb{R} -tree homotopic , and we write \alpha\simeq_{\mathbb{R}}\beta , if \alpha*\overline{\beta} is \mathbb{R} -tree nullhomotopic. A principle result of [ 2 ] (Theorem 1.4) is that \simeq_{\mathbb{R}} is not an equivalence relation on the space of paths, specifically \simeq_{\mathbb{R}} does not satisfy transitivity. However, this in and of itself does not answer the questions of Hambly and Lyons, as it was not demonstrated that the paths causing non-transitivity admit a height function. While it is possible to work through the construction there to show that constructed paths admit height functions, here we will present an alternative geometric construction where it is immediate that the constructed loops admit height functions. This will also provide an independent proof of Theorem 1.4 of [ 2 ] . The rest of the paper is structured as follows. In Section 2 , we show that every tree-like loop is \mathbb{R} -tree nullhomotopic and that the converse holds when the factorization through the \mathbb{R} -tree E can be achieved with a Lipschitz map g:E\to X . In Section 3 , we show how to modify certain piecewise-linear paths that factor through a simplicial tree. In Section 4 , we inductively apply the construction from Section 3 to build two sequences of paths, g_{n}\circ\pi_{n} and \tilde{g}_{n}\circ\tilde{\pi}_{n} that factor through successively larger simplicial trees. In Section 5 , we show that these two sequences of paths limit on the same path \gamma , which is space-filling in the convex hull of a right triangle 1 1 1 After constructing \gamma independently, the authors recognized that \gamma is a parameterization of the Sierpsiński Curve and that our approximation of it is essentially Knopp’s representation [ 8 , Section 4.2] . We also identify general metric conditions ensuring that constructions such as that given in Section 4 result in \mathbb{R} -tree homotopic paths. In Section 6 , we give a negative answer to Hambly and Lyons’ first question by showing that \alpha=g_{1}\circ\pi_{1} is tree-like equivalent to \gamma , \beta=\tilde{g}_{1}\circ\tilde{\pi}_{1} is tree-like equivalent to \gamma , but that \alpha is not tree-like equivalent to \beta . Since the paths \alpha and \beta are rectifiable arcs which only intersect at their ends, they will be reduced and hence give a negative answer to Hambly and Lyons’ second question. 2. Relation between types of tree-like We will first give a short proof that tree-like implies \mathbb{R} -tree nullhomotopic. The interested reader can also see [ 3 , 4 , 6 ] for additional proofs. Recall that a topological space underlying a compact \mathbb{R} -tree is called a dendrite . Equivalently, a dendrite is a uniquely arcwise connected Peano continuum. Definition 2.1 (Dendrites from height functions) . For any continuous function h:[0,1]\to\mathbb{R} , we can define an equivalence relation \sim_{h} on [0,1] by s\sim_{h}t , if h(s)=h(t)=a and s,t are contained in a single component of h^{-1}\big([a,\infty)\big) . Note, for s<t , this is equivalent to a=h(s)=h(t)=\inf\limits_{u\in[s,t]}h(u) . Then, for any t\in[0,1] , the equivalence class of t is [t]_{h}=C\cap h^{-1}\big(h(t)\big) where C is the component of h^{-1}\big([h(t),\infty)\big) containing t . Let \pi:[0,1]\to E be the decomposition map where E=[0,1]/\hskip-2.5pt\sim_{h} . The following two properties are an immediate consequence of the definition. (*) A component C of h^{-1}\big([a,\infty)\big) is closed and \pi -saturated. Thus [0,1]\backslash C is open and \pi -saturated. (**) For every component C of h^{-1}\big([a,\infty)\big) , the components of C\backslash h^{-1}(a) are open and \pi -saturated. Lemma 2.2 . The equivalence relation \sim_{h} induces an upper semi-continuous decomposition of [0,1] such that the quotient \pi:[0,1]\to E=[0,1]/\hskip-2.5pt\sim_{h} is a dendrite. Proof. By Theorem 10.2 in [ 7 ] , we need only show that E is a continuum and that for any two points in E there is a third point separating them in E . If E is Hausdorff, then \pi is an upper semi-continuous decomposition and E

Read the complete text