Hyperfinite Boundary Actions

• Hyperfinite Boundary Actions are defined as countable group actions on boundaries that induce orbit equivalence relations expressible as increasing unions of finite Borel relations.
• They employ geometric methods such as horofunction compactification, combinatorial sectors, and lexicographic coding to achieve finite symmetric differences and Borel definability.
• These actions extend beyond hyperbolic groups to relatively hyperbolic and acylindrically hyperbolic groups, trees, and cubical models, illuminating boundary dynamics and rigidity.

Hyperfinite boundary actions are boundary actions of countable groups whose orbit equivalence relations are hyperfinite in the sense of descriptive set theory. In the standard setting, a finitely generated hyperbolic group $G$ acts by homeomorphisms on its Gromov boundary $\partial G$, and the central theorem of the subject states that this boundary action induces a hyperfinite equivalence relation (Marquis et al., 2019). The area connects geometric group theory, countable Borel equivalence relations, and boundary dynamics, and it has expanded from free groups and cubulated hyperbolic groups to relatively hyperbolic groups, acylindrically hyperbolic groups, actions on trees, Roller boundaries of virtually special groups, and coned-off boundaries of graphical small cancellation groups (Huang et al., 2017).

1. Descriptive-set-theoretic framework

A Borel equivalence relation $E$ on a standard Borel space $Z$ is hyperfinite if it can be written as an increasing union of finite Borel equivalence relations,

$$E=\bigcup_{n=1}^\infty E_n, \qquad \text{each } E_n \text{ finite}.$$

In the field of countable Borel equivalence relations, hyperfiniteness is equivalent to being Borel reducible to the tail equivalence relation $E_0$ on $2^{\mathbb N}$ (Marquis et al., 2019).

For a Borel action $G\curvearrowright X$ of a countable group, the induced orbit equivalence relation is

$x\,E_G^X\,y \iff \exists g\in G:\ y=gx.$

A boundary action is hyperfinite when this orbit relation is hyperfinite. In the hyperbolic-group setting, the relevant space is the Gromov boundary; for trees, one works with the boundary of infinite geodesic rays modulo initial segments; for relatively hyperbolic groups, one also encounters the geodesic boundary of the relative Cayley graph and the Bowditch boundary (Karpinski, 2022).

A basic distinction runs between Borel hyperfiniteness and measure hyperfiniteness. Borel hyperfiniteness is a property of the equivalence relation as a Borel object. Measure hyperfiniteness requires hyperfiniteness only almost everywhere with respect to a Borel probability measure. All Borel hyperfinite relations are measure-hyperfinite, but the converse need not hold (Elayavalli et al., 2023). This distinction is essential in the literature on boundary actions.

2. From free groups to all hyperbolic groups

The classical point of departure is the free group boundary. For free groups acting on their boundaries, Connes–Feldman–Weiss, Vershik, and Adams established measurable hyperfiniteness, and Dougherty, Jackson, and Kechris proved hyperfiniteness in the pure Borel sense by using the tail equivalence relation on rays (Marquis et al., 2019).

A major intermediate generalization was obtained for cubulated hyperbolic groups. Huang, Sabok, and Shinko proved that if a hyperbolic group acts geometrically on a CAT(0) cube complex, then the induced action on the Gromov boundary is hyperfinite (Huang et al., 2017). Their proof relied on a geometric lemma asserting that for any boundary point $\gamma$ and any two vertices $\partial G$0, the sets of vertices lying on combinatorial geodesic rays from $\partial G$1 and $\partial G$2 toward $\partial G$3 differ by only finitely many points.

That finite-symmetric-difference phenomenon does not hold in complete generality. Touikan showed that the key geometric property used in the cubulated argument fails for some Cayley graphs of hyperbolic groups, so a general proof required new ideas (Marquis et al., 2019). Marquis and Sabok supplied those ideas and proved the full theorem: for every finitely generated hyperbolic group $\partial G$4, the action of $\partial G$5 on $\partial G$6 is hyperfinite (Marquis et al., 2019). This settled the general Borel problem for Gromov boundary actions of hyperbolic groups.

The result was presented as completing the descriptive-set-theoretic classification of boundary actions for the whole class of hyperbolic groups, without any auxiliary hypothesis such as cubulation (Marquis et al., 2019). In that sense, hyperfiniteness became a general feature of hyperbolic-group boundary dynamics rather than a byproduct of special combinatorial models.

3. Marquis–Sabok constructions and the finite-symmetric-difference mechanism

The Marquis–Sabok proof replaces the earlier ray-bundle criterion by a refined geometric apparatus built from the horofunction compactification, combinatorial sectors, and special vertices (Marquis et al., 2019). For a boundary point $\partial G$7, the construction introduces a finite set $\partial G$8 of horoboundary points corresponding to combinatorial geodesic rays toward $\partial G$9. For $E$0, the combinatorial sector $E$1 consists of vertices lying on a combinatorial geodesic ray from $E$2 toward $E$3 converging to $E$4.

A vertex is declared $E$5-special when the intersection of all sectors $E$6, for $E$7, contains a combinatorial geodesic ray. These special vertices are used to define the refined sets

$E$8

where $E$9 is the finite set of $Z$0-special vertices $Z$1 closest to $Z$2 with sector $Z$3 (Marquis et al., 2019).

The decisive geometric statement is the finite symmetric difference theorem: for any two vertices $Z$4 in the hyperbolic Cayley graph and any boundary point $Z$5, the sets $Z$6 and $Z$7 are finite (Marquis et al., 2019). This is the replacement for the older bundle argument.

The proof is not only geometric. It also establishes Borel definability for the relevant assignments—special vertices, combinatorial sectors, and the sets $Z$8. That Borel control is indispensable because the theorem concerns pure Borel hyperfiniteness rather than measurable hyperfiniteness. The endgame is a reduction to a tail-type equivalence relation, together with Lusin–Novikov uniformization and a reduction to the hyperfinite relation $Z$9 (Marquis et al., 2019). The resulting architecture has become a model for later work on other boundaries.

4. Streamlined proofs and Borel asymptotic dimension

A later paper gave a shorter proof of the hyperbolic-group theorem and added a dimension-theoretic refinement (Naryshkin et al., 2023). The central device is a Borel embedding of $$E=\bigcup_{n=1}^\infty E_n, \qquad \text{each } E_n \text{ finite}.$$0 into a shift space $$E=\bigcup_{n=1}^\infty E_n, \qquad \text{each } E_n \text{ finite}.$$1, obtained by fixing an order on a finite generating set $$E=\bigcup_{n=1}^\infty E_n, \qquad \text{each } E_n \text{ finite}.$$2 and assigning to each boundary point the lexicographically minimal infinite geodesic ray from the identity. This yields a Borel injective map

$$E=\bigcup_{n=1}^\infty E_n, \qquad \text{each } E_n \text{ finite}.$$3

whose image can be compared with tail equivalence (Naryshkin et al., 2023).

Tail equivalence is the relation

$$E=\bigcup_{n=1}^\infty E_n, \qquad \text{each } E_n \text{ finite}.$$4

which is a standard hyperfinite countable Borel equivalence relation (Oyakawa, 2023). The short proof shows that each boundary orbit class maps into only finitely many tail-equivalence classes, which is enough to deduce hyperfiniteness (Naryshkin et al., 2023).

The same argument establishes that the boundary action has finite Borel asymptotic dimension. More precisely, if the Cayley graph is $$E=\bigcup_{n=1}^\infty E_n, \qquad \text{each } E_n \text{ finite}.$$5-hyperbolic and $$E=\bigcup_{n=1}^\infty E_n, \qquad \text{each } E_n \text{ finite}.$$6 denotes the maximal size of a ball of radius $$E=\bigcup_{n=1}^\infty E_n, \qquad \text{each } E_n \text{ finite}.$$7, then

$$E=\bigcup_{n=1}^\infty E_n, \qquad \text{each } E_n \text{ finite}.$$8

(Naryshkin et al., 2023). This links the subject to the general theory of Borel asymptotic dimension, where finite Borel asymptotic dimension implies hyperfiniteness of the finite-distance equivalence relation (Conley et al., 2020). The dimension-theoretic perspective recasts hyperfiniteness as a consequence of controlled large-scale Borel geometry.

5. Extensions beyond hyperbolic groups

The hyperbolic-group theorem has several boundary-action analogues. For relatively hyperbolic groups, if $$E=\bigcup_{n=1}^\infty E_n, \qquad \text{each } E_n \text{ finite}.$$9 is finitely generated and hyperbolic relative to a finite collection of subgroups $E_0$0, then the natural action on the geodesic boundary of the associated relative Cayley graph induces a hyperfinite orbit equivalence relation, and consequently the action on the Bowditch boundary $E_0$1 is also hyperfinite (Karpinski, 2022). This strengthens a result of Ozawa that had required amenability or exactness assumptions on the peripheral subgroups.

For countable acylindrically hyperbolic groups, there exists a generating set $E_0$2 such that the Cayley graph $E_0$3 is hyperbolic, $E_0$4, the natural action on $E_0$5 is acylindrical, and the induced action on $E_0$6 is hyperfinite (Oyakawa, 2023). The proof again uses a lexicographically least geodesic coding and a finite-ambiguity reduction to tail equivalence. A corollary is that such actions are topologically amenable (Oyakawa, 2023).

Tree boundaries exhibit both positive criteria and genuine failures. If $E_0$7 is an action on a countable tree such that for every infinite geodesic $E_0$8 there is a finite initial segment $E_0$9 with $2^{\mathbb N}$0, then the induced action on $2^{\mathbb N}$1 is Borel hyperfinite (Elayavalli et al., 2023). Acylindrical actions satisfy this condition. A weaker hypothesis, replacing equality by uniform coamenability, yields Borel $2^{\mathbb N}$2-amenability and hence measure hyperfiniteness (Elayavalli et al., 2023). By contrast, there are modular actions of non-amenable groups whose boundary action is not measure-hyperfinite, and for antitreeable groups the modular boundary action is not measure-treeable (Elayavalli et al., 2023). These examples show that hyperfinite boundary actions are not automatic even for actions on trees.

6. Other boundary models and conceptual significance

The phenomenon is not limited to Gromov boundaries. For triangle buildings of order $2^{\mathbb N}$3, certain natural actions on the boundary are hyperfinite of type $2^{\mathbb N}$4 (Ramagge et al., 2013). In cubical geometry, if a countable group acts virtually specially on a CAT(0) cube complex, then the orbit equivalence relation induced by its action on the Roller boundary is hyperfinite (Oyakawa, 4 Sep 2025). This generalizes the cubulated hyperbolic case from Gromov boundaries to Roller boundaries and from hyperbolic groups to virtually special groups.

Small-cancellation theory provides another class of examples. A class of graphical small cancellation groups, including infinitely presented classical small cancellation groups, admit hyperfinite boundary actions on the Gromov boundaries of their coned-off Cayley graphs (Karpinski et al., 6 Sep 2025). The proof again proceeds by a Borel injection into a shift space and a finite-index comparison with tail equivalence, but the controlling hypothesis is the extreme fineness of the presentation graph.

Across these settings, hyperfiniteness functions as a strong regularity property of orbit equivalence relations. Hyperfinite relations are amenable, but hyperfiniteness is strictly stronger than amenability for actions on standard Borel spaces (Karpinski, 2022). In the hyperbolic-group context, the result has been described as an important signature for orbit equivalence and rigidity questions, and the methods based on horoboundaries, combinatorial sectors, Borel codings, and Borel asymptotic dimension suggest a general program for studying negatively curved or nonpositively curved boundary dynamics (Marquis et al., 2019). A plausible implication is that the decisive issue is not non-amenability of the acting group, but whether the boundary model admits sufficiently finite combinatorial control of geodesic representatives and stabilizers.