Uniform Local Amenability (ULA)

• Uniform Local Amenability is a local geometric property ensuring every finite subset of a bounded geometry space contains a small-diameter subset with a relatively small boundary.
• ULA bridges classical amenability with coarse geometric notions like Property A, metric sparsification, and operator norm localization, providing key insights into space structure.
• Spaces such as expanders fail ULA, while amenable groups and Property A spaces illustrate its practical significance in index theory and the analysis of Roe algebras.

Uniform Local Amenability (ULA) is a large-scale geometric property introduced to capture a robust, local version of amenability for discrete metric spaces of bounded geometry. ULA links the existence of controlled Følner sets of bounded diameter inside arbitrary finite subsets or local measures to key notions such as Property A, operator norm localization, and metric sparsification. This property provides both an intuitive and an effective obstruction to expanderness and plays a central role in modern coarse geometry, index theory, and the theory of Roe algebras (Brodzki et al., 2012, Elek, 2019).

1. Formal Definitions and Variants

Let $X$ be a discrete metric space of bounded geometry, i.e., for each $R>0$, there exists $N_R<\infty$ such that $|B(x;R)|\leq N_R$ for all $x\in X$. Several versions of ULA are considered:

(a) Classical Amenability (Block–Weinberger):

$$X\ \text{is amenable} \iff \forall R,\epsilon>0\ \exists\ \text{finite } E\subset X: |\partial_R E|<\epsilon |E|$$

where $$\partial_R E = \{x \in E \mid d(x, X \setminus E) \leq R\}$$.

(b) Uniform Local Amenability (ULA) [Definition 2.3, (Brodzki et al., 2012)]:

$$\forall R,\epsilon>0\ \exists S>0\ \text{s.t. } \forall \text{ finite } F\subset X\ \exists\ E\subset X:\ \mathrm{diam}(E)\le S,\ |\partial_R E \cap F| < \epsilon |E \cap F|$$

(c) Measure-Theoretic Variant—$ULA_\mu$ [Definition 2.5, (Brodzki et al., 2012)]:

$$\forall R,\epsilon > 0\ \exists S > 0\ \text{s.t. } \forall \text{prob. measures } \mu\ \exists E \subset \mathrm{supp}(\mu):\ \mathrm{diam}(E) \le S,\ \mu(\partial_R E) < \epsilon \mu(E)$$

Equivalent formulations utilize finitely supported $\ell^1$ functions of controlled support and a variational inequality.

(d) Graph Theoretic Version (Elek, 2019):

For a family $\mathcal{G}$ in the class $\mathcal{H}_d$ of finite degree ≤ d graphs, $\mathcal{G}$ is uniformly locally amenable if for each $\varepsilon>0$ there is $K$ such that every $G\in\mathcal{G}$ contains a vertex set $E$,

$$|E|\leq K,\quad |\partial_G E| \leq \varepsilon |E|$$

where $\partial_G E$ is the vertex boundary in $G$.

2. Intuition and Conceptual Framework

ULA systematically strengthens classical amenability by imposing a local Følner condition that is uniform both in the scale of the boundary and the diameter of the subsets. Unlike global amenability, which is detected through asymptotically large Følner sets, ULA requires that every finite subset or measure on $X$ possesses a corner $E$ with uniformly bounded diameter and small boundary-to-volume ratio. Thus, ULA asserts that the space "looks amenable everywhere at small scales," precluding "large-scale expander-type" behavior at any location (Brodzki et al., 2012).

$ULA_\mu$ allows arbitrary probability measures, reinforcing this uniformity in a measure-theoretic context. The Reiter–Følner argument demonstrates the equivalence of the set-based and measure-based variants.

3. Key Examples: Spaces Satisfying or Failing ULA

Type	Satisfies ULA	Reason/Reference/Implication
Property A Spaces	Yes	Implies $ULA_\mu$ [(Brodzki et al., 2012), 3.1]
Box Spaces (of amenable groups)	Yes	All five properties equivalent [(Brodzki et al., 2012), 4.6]
Expanders	No	Contradicts ULA due to large Cheeger constant [(Brodzki et al., 2012), 4.2]
Large-girth graphs	No	Fail ULA via the large-girth trick
Coarse Embeddable (CE)	Yes (if MSP holds)	$ULA_\mu \iff MSP \implies CE$ (Brodzki et al., 2012)

Expanders and large-girth graphs of uniformly bounded degree provide canonical counterexamples: any bounded-diameter set inside such graphs always has boundary proportional to its size, violating the ULA Følner condition. In contrast, spaces with Property A, including box spaces from amenable groups, do exhibit ULA and all related properties become equivalent.

4. Relationships to Coarse Geometric Properties

The interplay between ULA and other coarse geometric or analytic properties forms the main structural insight:

• Property A $\implies ULA_\mu \iff$ Metric Sparsification Property (MSP) $\implies$ Operator Norm Localization (ONL) $\implies$ ULA (Brodzki et al., 2012):
  • Property A leads, via the Higson–Roe criterion and averaging, to ULA$_\mu$ (Proposition 3.1).
  • ULA$_\mu$ is equivalent to MSP (Theorem 3.4), via decompositions yielding small-boundary pieces.
  • MSP in turn implies ONL as established in prior work (Chen–Tessera–Wang–Yu).
  • ONL ensures ULA by leveraging Laplacian operator estimates.
• Equivalence for Bounded Geometry:

For bounded geometry spaces, Sako showed the equivalence of these properties, closing the implication cycle.

• Counterexamples:

Spaces failing ULA (e.g., expanders) necessarily lack Property A and ONL, providing positive detection for "bad" spaces.

5. ULA and Property A: The Main Theorem

Elek's resolution in (Elek, 2019) establishes that ULA implies Property A for bounded degree graphs. The proof hinges on translating ULA into the framework of local hyperfiniteness and then to strong hyperfiniteness, before culminating in Property A as weighted hyperfiniteness.

The argument proceeds via:

• Step A: ULA ⇒ local hyperfiniteness: finite subgraphs can be partitioned into small components after removing a small proportion of vertices.
• Step B: Local hyperfiniteness ⇒ local strong hyperfiniteness: probabilistic separator distributions ensure fractional coverage properties.
• Step C: Local strong hyperfiniteness ⇒ weighted hyperfiniteness (Property A): convex geometry duality links separator distributions and weighted vertex removals.

Hence, ULA at bounded degree ensures all finite subgraphs admit appropriate partitions, yielding Property A.

6. Quantitative Generalizations and Connections to Amenability

The framework extends to quantitative regimes via control functions. A weak metric sparsification property $WMSP(c, f)$ encapsulates a quantitative form of ULA, controlling the diameter and boundary profile in terms of a function $f$. This leads, in coarsely-geodesic, amenable spaces, to sharp bounds on the isodiametric profile $A_X(n)$. If $X$ further possesses finite asymptotic dimension with control $\tau$, then

$A_X(n) \preceq \tau(n)$

recovering and generalizing quantitative relationships between amenability and asymptotic dimension obtained by Nowak [(Brodzki et al., 2012), Section 5].

7. Operator Norm Localization and Concrete Applications

ULA, through its connections with ONL, allows effective operator-norm estimates for finite-propagation operators on $X$. Specifically, ONL guarantees: for any positive, finite-propagation operator $T$ (with $\|T\|=1$), there exists a unit vector $\psi$ supported in a ball of uniformly bounded radius such that

$\langle \psi, T\psi \rangle \geq c$

for a suitable constant $c>0$. This yields the estimate: $$\|T\| = \sup \{ \|T\psi\| : \|\psi\|=1,\, \mathrm{diam}(\mathrm{supp} \psi) \leq S\}$$ (up to a factor $c^{-1}$). Thus, operator norms are effectively computable by localization to small supports (Brodzki et al., 2012).

---

References:

• "Uniform Local Amenability" (Brodzki et al., 2012)
• "Uniform Local Amenability implies Property A" (Elek, 2019)