betaFeedback

WorksMathematics

Polynomial growth and dynamic asymptotic dimension

Abstract

We show that principal ample groupoids with local polynomial growth of degree at most ss have dynamic asymptotic dimension at most ⌊s⌋\lfloor s\rfloor. As a consequence, σσ-compact principal ample groupoids with local polynomial growth have slow dynamic asymptotic dimension growth.

Introduction

The dynamic asymptotic dimension is a measure of the dynamical complexity of a topological dynamical system. It was introduced in [GWY17] for actions of discrete groups on locally compact spaces, and more generally for locally compact Hausdorff étale groupoids to produce an upper bound for the nuclear dimension of the associated groupoid C∗C^*-algebra (see also [10, 5] for extensions of this result). These upper bounds for the nuclear dimension are often far from optimal [9]. However, in [LLW26] the authors introduce a version of nuclear dimension for an inclusion D⊆AD \subseteq A of C∗C^*-algebras, called diagonal dimension, which seems to be very closely related to the dynamic asymptotic dimension of the groupoid model for the diagonal inclusion D⊆AD \subseteq A. In a different direction it was shown in [Bön+23] that the dynamic asymptotic dimension has a homological shadow: for a σ\sigma-compact principal ample groupoid GG, its homology vanishes in degrees above dad⁡(G)\operatorname{dad}(G).

The main contribution of the present article is a concrete sufficient criterion for finite dynamic asymptotic dimension (resp. control over the dynamic asymptotic dimension growth) in terms of the local polynomial growth. Roughly, local polynomial growth means that, for each compact open symmetric set DD of arrows, the number of arrows in DnD^n with any fixed source is bounded by a polynomial in nn, uniformly over the source. The polynomial may depend on DD. Degree at most ss means that its exponent can always be chosen to be ss.

Theorem A. Let GG be a principal ample groupoid with local polynomial growth of degree at most s≥0s \geq0. Then dad⁡(G)≤⌊s⌋\operatorname{dad}(G) \leq\lfloor s \rfloor.

The present work is inspired by [Pap23] where a similar statement is proved in the context of graphs and [8] who prove a similar statement in the Borel setting. In fact, via the coarse groupoid construction, Theorem A recovers Papasoglu’s bound asdim⁡(X)≤k\operatorname{asdim}(X) \leq k when the uniform growth of XX is O(rs)O(r^s) for some fixed s<k+1s < k+1, whereas [Pap23] requires only the weaker hypothesis o(rk+1)o(r^{k+1}). Moreover, the Borel asymptotic dimension result [8 Theorem 1.25] can be obtained by following our strategy as well, making minor modifications to the arguments presented below.

Finally, in [Wan+26], the authors introduce the concept of dynamic asymptotic dimension growth as a tool to study groupoids with infinite dynamic asymptotic dimension.

Corollary B. Let GG be a σ\sigma-compact, principal, ample groupoid with local polynomial growth and f:[0,∞)→[0,∞)f : [0,\infty) \to[0,\infty) an increasing unbounded function. Then there exists a continuous coarse length function ℓ\ell on GG such that dad⁡G,ℓ(t)≤f(t)\operatorname{dad}_{G,\ell}(t) \le f(t). In particular, ℓ\ell can be chosen so that (G,ℓ)(G,\ell) has slow dynamic asymptotic dimension growth.

In Section 4 we apply Theorem A to relevant examples. Beyond recovering known dynamic asymptotic dimension bounds in the case of coarse groupoids and AF groupoids, we also consider partial action groupoids and point set groupoids for point sets in locally compact groups. Specifically, we obtain the bound ⌊s⌋\lfloor s\rfloor for free partial actions of discrete groups with polynomial growth of degree at most ss on locally compact totally disconnected spaces. We also establish the same bound for point set groupoids, assuming separation, finite local complexity, trivial translation stabilizers throughout the non-empty hull, and polynomial volume growth of degree at most ss in the ambient group.

Acknowledgments. The first author wishes to thank Newcastle University for its hospitality during a research stay in February 2026, during which parts of this research were conducted. The first author was funded by the Research Council of Norway projects 324944 and 359602.

Groupoids and dynamic asymptotic dimension

Let GG denote a groupoid with unit space G(0)G^{(0)}, and let G(2)G^{(2)} denote its composable pairs. It comes equipped with range and source maps r,s:G→G(0)r,s : G \to G^{(0)}. The range and source fibres will be denoted by Gx=r−1(x)G^x = r^{-1}(x) and Gx=s−1(x)G_x = s^{-1}(x) for x∈G(0)x \in G^{(0)}. For a subset V⊆G(0)V \subseteq G^{(0)}, the restriction of GG to VV is G∣V:=s−1(V)∩r−1(V)G|_V := s^{-1}(V) \cap r^{-1}(V). We say that GG is principal if G∣{x}=s−1({x})∩r−1({x})={x}G|_{\{x\}} = s^{-1}(\{x\}) \cap r^{-1}(\{x\}) = \{x\} for all x∈G(0)x \in G^{(0)}. Furthermore, for subsets A,B⊆GA,B \subseteq G, we denote the set of all possible products of elements in AA and elements in BB by AB={gh∈G∣g∈A,h∈B,(g,h)∈G(2)}AB = \{gh \in G \mid g \in A, h \in B, (g,h) \in G^{(2)}\}. All groupoids considered in this article will be locally compact and Hausdorff. That is, we equip GG with a locally compact Hausdorff topology such that, when G(0)G^{(0)} is equipped with the relative topology from GG, and G(2)G^{(2)} is equipped with the relative topology from G×GG \times G, the multiplication and inversion maps on GG are continuous. By continuity of composition, if A,B⊆GA,B \subseteq G are compact, then ABAB is also compact. For A⊆GA \subseteq G we let ⟨A⟩⊆G\langle A\rangle\subseteq G be the algebraically generated subgroupoid generated by AA.

An open bisection is an open subset B⊆GB \subseteq G on which both ss and rr restrict to homeomorphisms onto open subsets of G(0)G^{(0)}. The groupoid GG is said to be étale if ss (equivalently rr) is a local homeomorphism, and is said to be ample if its topology has a basis consisting of compact open bisections. In an étale groupoid, products of open sets are open, and so in particular, finite products of compact open sets are again compact open. Since we exclusively work in Hausdorff topologies, such sets are also closed.

Definition 2.1. Let GG be an étale groupoid. We say that GG has local polynomial growth if for every compact open symmetric set D⊆GD \subseteq G, there exists an s≥0s \ge0 and a constant

CD≥1C_D \ge1 such that

sup⁡x∈G(0)∣Dn∩Gx∣≤CD(1+n)sfor all n∈N0.\sup_{x \in G^{(0)}} |D^n \cap G_x| \le C_D(1+n)^s \quad\text{for all } n \in\mathbb{N}_0.

If the same exponent s≥0s \ge0 can be chosen independently of DD, then we say that GG has local polynomial growth of degree at most $s.

Remark 2.2. We note here that if GG admits a controlled length function ℓ:G→[0,∞)\ell: G \to[0,\infty) with polynomial growth of degree s≥0s \ge0 in the sense of [1], then GG has local polynomial growth of degree at most ss. Indeed, write Bℓ(t):={g∈G∣ℓ(g)≤t}B_\ell(t) := \{g \in G \mid\ell(g) \le t\} for the closed balls for ℓ\ell. Let C≥1C \ge1 be such that

sup⁡x∈G(0)∣Bℓ(t)∩Gx∣≤C(1+t)sfor all t≥0,\sup_{x \in G^{(0)}} |B_\ell(t) \cap G_x| \le C(1+t)^s \quad\text{for all } t \ge0,

and let DD be a compact open symmetric subset of GG. Define MD:=max⁡{1,sup⁡g∈Dℓ(g)}M_D := \max\{1,\sup_{g \in D}\ell(g)\}, which is finite as ℓ\ell is assumed controlled. Since ℓ\ell is a length function, we have Dn⊆Bℓ(nMD)D^n \subseteq B_\ell(nM_D). Then

sup⁡x∈G(0)∣Dn∩Gx∣≤sup⁡x∈G(0)∣Bℓ(nMD)∩Gx∣≤C(1+nMD)s≤CMDs(1+n)s.\sup_{x \in G^{(0)}} |D^n \cap G_x| \le\sup_{x \in G^{(0)}} |B_\ell(nM_D) \cap G_x| \le C(1+nM_D)^s \le CM_D^s(1+n)^s.

Lastly in this section we recall the definition of dynamic asymptotic dimension.

Definition 2.3. Let GG be an étale groupoid. We say GG has dynamic asymptotic dimension bounded by dd, d∈N0d \in\mathbb{N}_0, and write dad⁡(G)≤d\operatorname{dad}(G) \le d, if for every relatively compact open set E⊆GE \subseteq G, there are open subsets U0,…,Ud⊆G(0)U_0,\ldots,U_d \subseteq G^{(0)} covering s(E)∪r(E)s(E) \cup r(E) such that ⟨E∩G∣Ui⟩\langle E \cap G|_{U_i}\rangle is relatively compact in GG for each ii. The least such dd, or ∞\infty if none exists, is the dynamic asymptotic dimension dad⁡(G)\operatorname{dad}(G) of $G.

Our arguments in § 3 will be simplified in the case of a compact unit space. Given an étale groupoid GG we may pass to its Alexandrov groupoid [Ken+21, Definition 7.8]. Suppose G(0)G^{(0)} is noncompact. Let (G(0))+=G(0)⊔{∞}(G^{(0)})^+ = G^{(0)} \sqcup\{\infty\} be the one-point compactification of the unit space. The Alexandrov groupoid G+G^+ is the set G∪(G(0))+G \cup(G^{(0)})^+, with topology determined by requiring that the inclusions G,(G(0))+⊆G+G,(G^{(0)})^+ \subseteq G^+ are open, and the groupoid structure extends that of GG by making ∞\infty a unit. Since GG is locally compact Hausdorff étale, so is G+G^+. Moreover, if GG is ample, so is G+G^+. We remark that principality and local polynomial growth pass to G+G^+, with the same exponent whenever a uniform exponent is available.

Lemma 2.4 ([10 Proposition 3.13]). Suppose that GG is an étale groupoid. Then dad⁡(G)=dad⁡(G+)\operatorname{dad}(G) = \operatorname{dad}(G^+).

When GG is an étale groupoid with compact unit space and GG is equipped with a coarse continuous length function ℓ:G→[0,∞)\ell: G \to[0,\infty), then we follow [Wan+26] and define the dynamic asymptotic dimension function

dad⁡G,ℓ:[0,∞)→N0∪{∞}\operatorname{dad}_{G,\ell} : [0,\infty) \to\mathbb{N}_0 \cup\{\infty\}

by letting dad⁡G,ℓ(t)\operatorname{dad}_{G,\ell}(t) be the least d∈N0d \in\mathbb{N}_0 such that there are open subsets U0,…,Ud⊆G(0)U_0,\ldots,U_d \subseteq G^{(0)} covering s(Bℓ(t))s(B_\ell(t)) such that ⟨Bℓ(t)∩G∣Ui⟩\langle B_\ell(t)\cap G|_{U_i}\rangle is relatively compact in GG for each ii, and setting dad⁡G,ℓ(t)=∞\operatorname{dad}_{G,\ell}(t) = \infty if no such dd exists. The growth type of the function dad⁡G,ℓ\operatorname{dad}_{G,\ell} is called the dynamic asymptotic dimension growth of (G,ℓ)(G,\ell). For noncompact unit space, following [Wan+26, Definition 4.10] we define dad⁡G,ℓ:=dad⁡G+,ℓ+\operatorname{dad}_{G,\ell}:=\operatorname{dad}_{G^{+},\ell^{+}}, where ℓ+∣G=ℓ\ell^{+}|_{G}=\ell and ℓ+(∞)=0\ell^{+}(\infty)=0. This extension is again continuous and coarse.

Main Results

Before launching into the details of the proof of Theorem A, let us outline the main ideas and key steps in the proof. A key ingredient in the proof is to show that for every N∈NN\in\mathbb{N} and every compact open subset A⊆G(0)A\subseteq G^{(0)} there exists a compact open set Y⊆AY\subseteq A such that

⟨D∩G∣A∖Y⟩⊆D2N.(1)\langle D\cap G|_{A\setminus Y}\rangle\subseteq D^{2N}. \tag*{(1)}

A set YY with the properties above plays the role of the separating subsets in [Pap23], hence we will sometimes refer to YY as a separating set in AA. As outlined in [8 Section 3.3] and the references therein, such a separating subset can be produced via an explicit ball carving construction associated with suitable radius choice functions t:A→{0,…,N−1}t:A\to\{0,\ldots,N-1\}. If tt happens to be locally constant, the set YY obtained from the ball carving procedure is compact open and satisfies (3.1).

The strategy to prove Theorem A is to successively apply this procedure to obtain a decreasing sequence of compact open subsets G(0)=A0⊇A1⊇A2⊇⋯G^{(0)}=A_{0}\supseteq A_{1}\supseteq A_{2}\supseteq\cdots, each Ai+1A_{i+1} separating in AiA_i, i.e. ⟨D∩G∣Ai∖Ai+1⟩⊆D2N\langle D\cap G|_{A_i\setminus A_{i+1}}\rangle\subseteq D^{2N}. Thus, the goal is to use local polynomial growth to ensure that this process eventually halts in the sense that AiA_i is empty at some finite stage. This is where our argument significantly differs from both the minimality argument in [Pap23] and the probabilistic approach taken in [8]. We use the fact that local polynomial growth of GG implies that suitable balls in the orbit graph with respect to DD have small relative boundaries to produce for every ε>0\varepsilon>0 and all sufficiently large N∈NN\in\mathbb{N} a locally constant radius choice function t:G(0)→{0,…,N}t:G^{(0)}\to\{0,\ldots,N\} such that in addition to (3.1) the set YY produced by the ball carving construction also has a small growth function γY\gamma_Y (see (3.3) below for the definition) relative to that of AA, in the sense that

γY(R)≤εγA(R+2N).(2)\gamma_Y(R)\leq\varepsilon\gamma_A(R+2N). \tag*{(2)}

Applying this estimate iteratively to our decreasing sequence of subsets A0⊇A1⊇A2⊇⋯A_0\supseteq A_1\supseteq A_2\supseteq\cdots and using the polynomial growth assumption we conclude that γA⌊s⌋+1(0)<1\gamma_{A_{\lfloor s\rfloor+1}}(0)<1 and hence A⌊s⌋+1=∅A_{\lfloor s\rfloor+1}=\varnothing.

Let us now turn to the details of the proof. By Lemma 2.4, we may assume that the unit space is compact, and so the auxiliary lemmas in the sequel will be stated and proved under that assumption. The following remark tells us that in the definition of dynamical asymptotic dimension in Definition 2.3 it suffices to consider precompact sets which contain the compact unit space.

Remark 3.1. When G(0)G^{(0)} is compact, the definition of dynamical asymptotic dimension may be tested on symmetric relatively compact open sets containing G(0)G^{(0)}, see [Bön24, Definition 2.9] and the surrounding discussion. If GG is ample, it moreover suffices to restrict to compact open sets. Indeed, given a relatively compact open E⊆GE\subseteq G, choose compact open bisections B1,…,BnB_1,\ldots,B_n covering E‾\overline{E} and set

D=G(0)∪⋃j=1n(Bj∪Bj−1).D=G^{(0)}\cup\bigcup_{j=1}^{n}(B_j\cup B_j^{-1}).

Then DD is compact open, symmetric, and contains both G(0)G^{(0)} and EE. Any DAD cover U0,…,UdU_0,\ldots,U_d for DD works for EE, as ⟨E∩G∣Ui⟩⊆⟨D∩G∣Ui⟩\langle E\cap G|_{U_i}\rangle\subseteq\langle D\cap G|_{U_i}\rangle for all ii.

Thus to prove dad⁡(G)≤d\operatorname{dad}(G)\le d, it suffices to construct, for every such DD, an open cover U0,…,UdU_0,\ldots,U_d of G(0)G^{(0)} for which ⟨D∩G∣Ui⟩\langle D\cap G|_{U_i}\rangle is relatively compact for every 0≤i≤d0\le i\le d.

In light of Remark 3.1, we will focus on compact open symmetric sets D⊆GD\subseteq G with D⊇G(0)D\supseteq G^{(0)}. Indeed, fix such DD, set D0=G(0)D^0=G^{(0)}, and define

BD(x,n)=r(Dn∩Gx).B_D(x,n)=r(D^n\cap G_x).

We identify G(0)G^{(0)} with the vertices of a graph, where two units x,yx,y are connected by an edge exactly when there is g∈Dg\in D with s(g)=xs(g)=x and r(g)=yr(g)=y. We then naturally have an extended graph metric on G(0)G^{(0)}, and since G(0)⊆D=D−1G^{(0)}\subseteq D=D^{-1}, the set BD(x,n)B_D(x,n) is exactly the radius nn ball about x∈G(0)x\in G^{(0)} in this extended metric. The distance between different graph components is set to be infinite.

Moreover, for any A⊆G(0)A\subseteq G^{(0)} we define

γA(n)=sup⁡x∈G(0)∣A∩BD(x,n)∣,(3)\gamma_A(n)=\sup_{x\in G^{(0)}}|A\cap B_D(x,n)|, \tag*{(3)}

where we remark that the supremum is taken over G(0)G^{(0)}, not just AA. A DD-component of AA is a connected component of the induced graph on AA.

Note the trivial observation

γA(0)={1A≠∅,0A=∅.\gamma_A(0)= \begin{cases} 1 & A\ne\varnothing,\\ 0 & A=\varnothing. \end{cases}

If GG is principal, the range map is injective on GxG_x for every x∈G(0)x\in G^{(0)}, and so

∣BD(x,n)∣=∣Dn∩Gx∣.|B_D(x,n)|=|D^n\cap G_x|.

Lemma 3.2. Let GG be a principal ample groupoid with compact unit space G(0)G^{(0)}. If T⊆GT\subseteq G and W⊆G(0)W\subseteq G^{(0)} are compact open, then T∩r−1(W)T\cap r^{-1}(W) and r(T∩s−1(W))r(T\cap s^{-1}(W)) are compact open. Moreover, the function

x↦∣T∩Gx∣x\mapsto|T\cap G_x|

is bounded and locally constant. In particular, the map x↦∣W∩BD(x,n)∣x\mapsto|W\cap B_D(x,n)| is locally constant for each fixed nn.

Proof. Since rr and ss are local homeomorphisms, they are both continuous and open maps. The first claim is then immediate from TT and WW being compact open.

Using compactness of TT, cover TT by finitely many compact open bisections. Using the fact that GG is ample, find a finite disjoint refinement, i.e., find finitely many disjoint compact open bisections T1,…,TkT_1,\ldots,T_k. Then we may write

x↦∣T∩Gx∣=∑j=1k1s(Tj)(x)x\mapsto|T\cap G_x|=\sum_{j=1}^{k}1_{s(T_j)}(x)

which is locally constant. The last assertion then follows by setting T=Dn∩r−1(W)T=D^n\cap r^{-1}(W) and calculating

∣r(r−1(W)∩Dn∩Gx)∣=∣r(r−1(W)∩Dn∩Gx)∣=∣W∩BD(x,n)∣|r(r^{-1}(W)\cap D^n\cap G_x)|=|r(r^{-1}(W)\cap D^n\cap G_x)|=|W\cap B_D(x,n)|

where the first equality follows by injectivity of rr on each source fibre. □\square

Lemma 3.3. Let GG be a principal ample groupoid with compact unit space G(0)G^{(0)}, and let A⊆G(0)A \subseteq G^{(0)}. If every DD-component of AA is contained in some BD(u,N)B_D(u,N), where uu is allowed to depend on the component, then

⟨D∩G∣A⟩⊆D2N.\langle D \cap G|_A\rangle\subseteq D^{2N}.

Proof. If h∈⟨D∩G∣A⟩h \in\langle D \cap G|_A\rangle, then s(h)s(h) and r(h)r(h) are in the same DD-component of AA. This component is contained in BD(u,N)B_D(u,N) for some uu, so choose p,q∈DNp,q \in D^N with s(p)=s(q)=us(p)=s(q)=u and r(p)=s(h)r(p)=s(h) and r(q)=r(h)r(q)=r(h). Then s(qp−1)=s(h)s(qp^{-1})=s(h) and r(qp−1)=r(h)r(qp^{-1})=r(h), so principality yields h=qp−1∈D2Nh=qp^{-1}\in D^{2N}. □\square

Lemma 3.4. Let GG be a principal ample groupoid with compact unit space G(0)G^{(0)}. For every compact open symmetric set D⊆GD \subseteq G and every N≥1N \geq1, there exists a finite partition of compact open sets G(0)=C1⊔⋯⊔CkG^{(0)}=C_1\sqcup\cdots\sqcup C_k such that

BD(u,N)∩BD(v,N)=∅B_D(u,N)\cap B_D(v,N)=\varnothing

for all u≠v∈Ciu\ne v\in C_i and 1≤i≤k1\leq i\leq k.

Proof. As GG is étale, G(0)G^{(0)} is open in GG, and so D2N∖G(0)D^{2N}\setminus G^{(0)} is compact. Then the image

R=(r,s)(D2N∖G(0))⊆G(0)×G(0)\mathcal{R}=(r,s)(D^{2N}\setminus G^{(0)})\subseteq G^{(0)}\times G^{(0)}

is compact, and since G(0)×G(0)G^{(0)}\times G^{(0)} is Hausdorff, we deduce that R\mathcal{R} is closed. Furthermore, since GG is principal, the image misses the diagonal of G(0)×G(0)G^{(0)}\times G^{(0)}.

Fix x∈G(0)x\in G^{(0)}. Since (x,x)∉R(x,x)\notin\mathcal{R} and R\mathcal{R} is closed, there are open neighbourhoods Vx,WxV_x,W_x of xx such that (Vx×Wx)∩R=∅(V_x\times W_x)\cap\mathcal{R}=\varnothing. Using ampleness of GG, we find a compact open neighbourhood Ox⊆Vx∩WxO_x\subseteq V_x\cap W_x and observe

(Ox×Ox)∩R=∅.(O_x\times O_x)\cap\mathcal{R}=\varnothing.

Since G(0)G^{(0)} is compact, we obtain a finite subcover O1,…,OkO_1,\ldots,O_k. From this we may construct a disjoint compact open cover by setting

C1=O1,Ci=Oi∖⋃j<iOjfor 2≤i≤k,C_1=O_1,\qquad C_i=O_i\setminus\bigcup_{j<i}O_j\quad\text{for }2\leq i\leq k,

Note that since Ci⊆OiC_i\subseteq O_i, we have (Ci×Ci)∩R=∅(C_i\times C_i)\cap\mathcal{R}=\varnothing.

We claim that for u≠v∈Ciu\ne v\in C_i, 1≤i≤k1\leq i\leq k, we have BD(u,N)∩BD(v,N)=∅B_D(u,N)\cap B_D(v,N)=\varnothing, which will finish the proof. Suppose for the sake of contradiction that u≠v∈Ciu\ne v\in C_i and z∈BD(u,N)∩BD(v,N)z\in B_D(u,N)\cap B_D(v,N). By the definition of these balls, there are elements a,b∈DNa,b\in D^N with

s(a)=u,s(b)=v,r(a)=r(b)=z.s(a)=u,\qquad s(b)=v,\qquad r(a)=r(b)=z.

The product b−1ab^{-1}a then has source uu and range vv, and by symmetry of DD we have b−1a∈D2Nb^{-1}a\in D^{2N}. Since u≠vu\ne v, b−1a∈D2N∖G(0)b^{-1}a\in D^{2N}\setminus G^{(0)}. But then

(v,u)=(r(b−1a),s(b−1a))∈R∩(Ci×Ci),(v,u)=(r(b^{-1}a),s(b^{-1}a))\in\mathcal{R}\cap(C_i\times C_i),

which contradicts R∩(Ci×Ci)=∅\mathcal{R}\cap(C_i\times C_i)=\varnothing from above. Thus BD(u,N)∩BD(v,N)=∅B_D(u,N)\cap B_D(v,N)=\varnothing. □\square

The following is our variant of the ball carving construction relative to a family of radius choice functions. Note that the choice functions do not need to satisfy any compatibility condition. The flexibility of moving to a family of radius choice functions allows us to explicitly control the growth of the set AA relative to that of YY.

Lemma 3.5. Let GG be a principal ample groupoid with compact unit space. Let further N≥1N \ge1, let A⊆G(0)A \subseteq G^{(0)} be compact open, and suppose that for every compact open W⊆AW \subseteq A a locally constant function tW:W⟶{0,…,N−1}t_W: W \longrightarrow\{0,\ldots,N-1\} is given. Then there exists a compact open Y⊆AY \subseteq A such that every DD-component of A∖YA \setminus Y is contained in a ball of radius N−1N-1. In particular, ⟨D∩G∣A∖Y⟩⊆D2N\langle D \cap G|_{A\setminus Y}\rangle\subseteq D^{2N}.

Moreover, if

η:=sup⁡{∣W∩(BD(u,tW(u)+1)∖BD(u,tW(u)))∣∣W∩BD(u,tW(u))∣:W⊆A compact open, u∈W}\eta:= \sup\left\{ \frac{\left|W \cap\left(B_D(u,t_W(u)+1) \setminus B_D(u,t_W(u))\right)\right|}{\left|W \cap B_D(u,t_W(u))\right|} : W \subseteq A\text{ compact open},\ u \in W \right\}

is finite, then YY can be chosen to satisfy γY(R)≤ηγA(R+2N)\gamma_Y(R) \le\eta\gamma_A(R+2N).

Proof. If A=∅A=\varnothing, we use Y=∅Y=\varnothing and we are done. We may therefore assume A≠∅A\ne\varnothing. Using Lemma Lemma 3.4 we find a compact open partition G(0)=C1⊔⋯⊔CkG^{(0)}=C_1\sqcup\cdots\sqcup C_k such that for each ii, u≠v∈Ciu\ne v\in C_i implies BD(u,N)∩BD(v,N)=∅B_D(u,N)\cap B_D(v,N)=\varnothing.

Set W0=AW_0=A. For i≥1i\ge1, suppose that Wi−1W_{i-1} has been defined and is compact open. Set Si=Ci∩Wi−1S_i=C_i\cap W_{i-1}, and define

Pi,u:=Wi−1∩BD(u,tWi−1(u))P_{i,u}:=W_{i-1}\cap B_D(u,t_{W_{i-1}}(u))
Qi,u:=Wi−1∩(BD(u,tWi−1(u)+1)∖BD(u,tWi−1(u))).Q_{i,u}:=W_{i-1}\cap\left(B_D(u,t_{W_{i-1}}(u)+1)\setminus B_D(u,t_{W_{i-1}}(u))\right).

Then by construction

u∈Pi,u,∣Qi,u∣≤η∣Pi,u∣,Pi,u⊔Qi,u⊆BD(u,N).(4)u\in P_{i,u},\qquad|Q_{i,u}|\le\eta|P_{i,u}|,\qquad P_{i,u}\sqcup Q_{i,u}\subseteq B_D(u,N). \tag*{(4)}

For 0≤t<N0\le t<N, put Ei,t:={u∈Si∣tWi−1(u)=t}E_{i,t}:=\{u\in S_i\mid t_{W_{i-1}}(u)=t\}. Since tWi−1t_{W_{i-1}} is locally constant, the sets Ei,tE_{i,t} are compact open. For a general compact open set E⊆G(0)E\subseteq G^{(0)}, we write

Bn(E):=r(Dn∩s−1(E))=⋃u∈EBD(u,n),B_n(E):=r(D^n\cap s^{-1}(E))=\bigcup_{u\in E}B_D(u,n),

which is again compact open by Lemma Lemma 3.2. We deduce that the sets

Pi:=⋃u∈SiPi,u=Wi−1∩⋃t=0N−1Bt(Ei,t),andP_i:=\bigcup_{u\in S_i}P_{i,u}=W_{i-1}\cap\bigcup_{t=0}^{N-1}B_t(E_{i,t}),\qquad\text{and}
Ti:=⋃u∈Si(Pi,u⊔Qi,u)=Wi−1∩⋃t=0N−1Bt+1(Ei,t)T_i:=\bigcup_{u\in S_i}(P_{i,u}\sqcup Q_{i,u})=W_{i-1}\cap\bigcup_{t=0}^{N-1}B_{t+1}(E_{i,t})

are compact open. Now observe that for u≠v∈Si⊆Ciu\ne v\in S_i\subseteq C_i we must have Qi,u∩Pi,v=∅Q_{i,u}\cap P_{i,v}=\varnothing since Qi,u⊆BD(u,N)Q_{i,u}\subseteq B_D(u,N) and Pi,v⊆BD(v,N)P_{i,v}\subseteq B_D(v,N), and the latter sets are disjoint by Lemma Lemma 3.4. Together with Qi,u∩Pi,u=∅Q_{i,u}\cap P_{i,u}=\varnothing we deduce that

Qi:=⋃u∈SiQi,u=Ti∖Pi.Q_i:=\bigcup_{u\in S_i}Q_{i,u}=T_i\setminus P_i.

It follows that both QiQ_i and Wi:=Wi−1∖TiW_i:=W_{i-1}\setminus T_i are compact open.

The sets under consideration are pairwise disjoint across different stages of the induction, because the later stages take place in the remainder left after removing earlier pieces. Moreover, Wi∩Ci=(Wi−1∖Ti)∩Ci=(Wi−1∩Ci)∖Ti=Si∖Ti=∅W_i\cap C_i=(W_{i-1}\setminus T_i)\cap C_i=(W_{i-1}\cap C_i)\setminus T_i=S_i\setminus T_i=\varnothing because Si⊆Pi⊆TiS_i \subseteq P_i \subseteq T_i. The remainders decrease and the CiC_i cover G(0)G^{(0)}, so we deduce Wk=∅W_k = \varnothing. Hence

Y:=⋃i=1kQisatisfiesA∖Y=⋃i=1k⋃u∈SiPi,u.(5)Y := \bigcup_{i=1}^{k} Q_i \qquad\text{satisfies} \qquad A \setminus Y = \bigcup_{i=1}^{k}\bigcup_{u\in S_i} P_{i,u}. \tag*{(5)}

and both YY and A∖YA \setminus Y are compact open.

We show that distinct pieces in the decomposition of A∖YA \setminus Y can not be joined by an element of DD. First, suppose a∈Pi,ua \in P_{i,u} and b∈Pi,vb \in P_{i,v} are joined by an element of DD, and that u≠vu \ne v. Then b∈BD(u,tWi−1(u)+1)⊆BD(u,N)b \in B_D(u,t_{W_{i-1}}(u)+1) \subseteq B_D(u,N) while also b∈BD(v,N)b \in B_D(v,N), which contradicts the disjointness of the balls provided by Lemma 3.4. If instead a∈Pi,ua \in P_{i,u} and b∈Pj,vb \in P_{j,v} are joined by an element of DD while i<ji < j, then b∈Wj−1⊆Wib \in W_{j-1} \subseteq W_i, but also

b∈Wi−1∩BD(u,tWi−1(u)+1)=Pi,u⊔Qi,u⊆Ti,b \in W_{i-1} \cap B_D(u,t_{W_{i-1}}(u)+1) = P_{i,u} \sqcup Q_{i,u} \subseteq T_i,

which contradicts Wi∩Ti=∅W_i \cap T_i = \varnothing. We conclude that every DD-component of A∖YA \setminus Y is contained in some piece Pi,u⊆BD(u,N−1)P_{i,u} \subseteq B_D(u,N-1), which verifies the first claim of the lemma. The statement ⟨D∩G∣A∖Y⟩⊆D2N\langle D \cap G|_{A\setminus Y}\rangle\subseteq D^{2N} follows from Lemma 3.3.

We end by verifying the final claim. Fix x∈G(0)x \in G^{(0)} and R∈N0R \in\mathbb{N}_0, and set

J:={(i,u)∣1≤i≤k, u∈Si, Qi,u∩BD(x,R)≠∅}.J := \{(i,u) \mid1 \le i \le k,\ u \in S_i,\ Q_{i,u} \cap B_D(x,R) \ne\varnothing\}.

The Qi,uQ_{i,u} are pairwise disjoint, and BD(x,R)B_D(x,R) is finite, so we deduce that JJ is finite. If (i,u)∈J(i,u) \in J, choose some z∈Qi,u∩BD(x,R)z \in Q_{i,u} \cap B_D(x,R). Then every point of Pi,uP_{i,u} can be reached from xx through a path via zz and uu, which by construction then has length at most R+(tWi−1(u)+1)+tWi−1(u)≤R+2N−1R + (t_{W_{i-1}}(u) + 1) + t_{W_{i-1}}(u) \le R + 2N - 1. Then

⋃(i,u)∈JPi,u⊆A∩BD(x,R+2N).\bigcup_{(i,u)\in J} P_{i,u} \subseteq A \cap B_D(x,R+2N).

By using (4) we then calculate

∣Y∩BD(x,R)∣=∑(i,u)∈J∣Qi,u∩BD(x,R)∣≤∑(i,u)∈J∣Qi,u∣≤η∑(i,u)∈J∣Pi,u∣≤η∣A∩BD(x,R+2N)∣.\begin{aligned} |Y \cap B_D(x,R)| &= \sum_{(i,u)\in J} |Q_{i,u} \cap B_D(x,R)| \le\sum_{(i,u)\in J} |Q_{i,u}| \\ &\le\eta\sum_{(i,u)\in J} |P_{i,u}| \le\eta|A \cap B_D(x,R+2N)|. \end{aligned}

This holds for each x∈G(0)x \in G^{(0)}, and so taking the supremum over all x∈G(0)x \in G^{(0)} finishes the proof. □

Our goal is now to produce a family of functions (tW)W(t_W)_W as in the hypothesis of the previous lemma such that η\eta is small. The form of η\eta of course immediately leads us to consider conditions under which the balls themselves form Følner sets. Hence the following subexponential growth type criterion should not come as a surprise:

Lemma 3.6. Let N∈NN \in\mathbb{N} and ε>0\varepsilon> 0 such that

γG(0)(N)<(1+ε)N.(6)\gamma_{G^{(0)}}(N) < (1+\varepsilon)^N. \tag*{(6)}

Then for every compact open W⊆G(0)W \subseteq G^{(0)} the function tW:W→{0,…,N−1}t_W : W \to\{0,\ldots,N-1\} given by

tW(u)=min⁡{t:∣W∩(BD(u,t+1)∖BD(u,t))∣≤ε∣W∩BD(u,t)∣}t_W(u) = \min\{t : |W \cap(B_D(u,t+1) \setminus B_D(u,t))| \le\varepsilon|W \cap B_D(u,t)|\}

is well defined and continuous.

Proof. Set f(t)=∣W∩BD(u,t)∣f(t)=|W\cap B_D(u,t)|, and observe that f(0)=1f(0)=1 and f(N)≤γG(0)(N)f(N)\le\gamma_{G^{(0)}}(N). Note that for any W⊆G(0)W\subseteq G^{(0)} and any u∈Wu\in W, there exists some 0≤t<N0\le t<N which satisfies

∣W∩(BD(u,t+1)∖BD(u,t))∣≤ε∣W∩BD(u,t)∣.(7)|W\cap(B_D(u,t+1)\setminus B_D(u,t))|\le\varepsilon|W\cap B_D(u,t)|. \tag*{(7)}

Indeed, if the inequality (7) failed for each 0≤t<N0\le t<N, then we would immediately get a contradiction to (6). The claim that the smallest such tt is locally constant then follows from Lemma 3.2. □

We are now ready to prove our main result:

Proof of Theorem A. We will prove the result under the additional assumption that G(0)G^{(0)} is compact. As explained in § 2, the general case follows by passing to the Aleksandrov groupoid G+G^{+}.

Fix a compact open and symmetric set G(0)⊆D⊆GG^{(0)}\subseteq D\subseteq G. Since GG has local polynomial growth of degree at most s≥0s\ge0 we can find a constant CD≥1C_D\ge1 such that

sup⁡x∈G(0)∣Dn∩Gx∣≤CD(1+n)sfor all n∈N0\sup_{x\in G^{(0)}}|D^n\cap G_x|\le C_D(1+n)^s\qquad\text{for all }n\in\mathbb{N}_0

Set m=⌊s⌋+1m=\lfloor s\rfloor+1, and for n≥2n\ge2 set

an=log⁡(2CD(1+n)s)n,andεn=ean−1.a_n=\frac{\log(2C_D(1+n)^s)}{n},\qquad\text{and}\qquad\varepsilon_n=e^{a_n}-1.

Now note that εn=O(log⁡(n)/n)\varepsilon_n=O(\log(n)/n), and so

εnmγG(0)(2mn)≤CDεnm(1+2mn)s=O(ns−m(log⁡(n))m)⟶0.\varepsilon_n^m\gamma_{G^{(0)}}(2mn)\le C_D\varepsilon_n^m(1+2mn)^s=O\left(n^{s-m}(\log(n))^m\right)\longrightarrow0.

We may therefore choose an N≥1N\ge1 so large that

εNmγG(0)(2mN)<1.(8)\varepsilon_N^m\gamma_{G^{(0)}}(2mN)<1. \tag*{(8)}

By construction, (1+εN)N=2CD(1+N)s>γG(0)(N)(1+\varepsilon_N)^N=2C_D(1+N)^s>\gamma_{G^{(0)}}(N), and so Lemma 3.6 provides a family of locally constant functions tW:W→{0,…,N−1}t_W:W\to\{0,\ldots,N-1\} such that

∣W∩(BD(u,tW(u)+1)∖BD(u,tW(u)))∣∣W∩BD(u,tW(u))∣≤εN\frac{|W\cap(B_D(u,t_W(u)+1)\setminus B_D(u,t_W(u)))|}{|W\cap B_D(u,t_W(u))|}\le\varepsilon_N

for every compact open W⊆G(0)W\subseteq G^{(0)} and u∈Wu\in W. Now set A0=G(0)A_0=G^{(0)} and inductively define a sequence of compact open sets A0⊇A1⊇A2⊇⋯A_0\supseteq A_1\supseteq A_2\supseteq\cdots by applying Lemma 3.5 to AjA_j to find Aj+1⊆AjA_{j+1}\subseteq A_j such that

  1. ⟨D∩G∣Aj∖Aj+1⟩⊆D2N\langle D\cap G|_{A_j\setminus A_{j+1}}\rangle\subseteq D^{2N}, and

  2. γAj+1(R)≤εN⋅γAj(R+2N)\gamma_{A_{j+1}}(R)\le\varepsilon_N\cdot\gamma_{A_j}(R+2N).

We claim that Am=∅A_m=\varnothing so that setting Uj:=Aj∖Aj+1U_j:=A_j\setminus A_{j+1} for j=0,…,m−1j=0,\ldots,m-1 yields the desired open cover of G(0)G^{(0)} required by the definition of dynamic asymptotic dimension by item (1). To see that AmA_m is indeed empty, repeatedly apply the estimate (2) to obtain

γAm(0)≤εNm⋅γG(0)(2mN)<1.(9)\gamma_{A_m}(0)\le\varepsilon_N^m\cdot\gamma_{G^{(0)}}(2mN)<1. \tag*{(9)}

Since γAm\gamma_{A_m} only takes non-negative integer values we conclude that Am=∅A_m=\varnothing as desired. □

Finally, we turn to the consequences concerning the dynamic asymptotic dimension growth. As the statement is slightly more technical, we repeat it for the convenience of the reader:

Corollary 3.7. Let GG be a σ\sigma-compact, principal, ample groupoid with local polynomial growth and f:[0,∞)→[0,∞)f : [0,\infty) \to[0,\infty) an increasing unbounded function. Then there exists a continuous coarse length function ℓ\ell on GG such that dad⁡G,ℓ(t)≤f(t)\operatorname{dad}_{G,\ell}(t) \le f(t). In particular, ℓ\ell can be chosen so that (G,ℓ)(G,\ell) has slow dynamic asymptotic dimension growth.

Proof. We may assume that G(0)G^{(0)} is compact by passing to G+G^{+}, as this preserves σ\sigma-compactness, principality, ampleness, and local polynomial growth. A continuous coarse length on G+G^{+} restricts to one on GG, and the dimension functions agree by our convention above. Since GG is σ\sigma-compact we can find an increasing sequence G(0)⊆K1⊆K2⊆⋯G^{(0)} \subseteq K_{1} \subseteq K_{2} \subseteq\cdots of compact open symmetric subsets exhausting all of GG. Next, use that GG has local polynomial growth to find corresponding sequences of constants Cn≥1C_{n} \ge1 and sn≥0s_{n} \ge0 such that for every k∈Nk \in\mathbb{N}:

sup⁡x∈G(0)∣Knk∩Gx∣≤Cn(1+k)sn.\sup_{x \in G^{(0)}} \left|K_{n}^{k} \cap G_{x}\right| \le C_{n}(1+k)^{s_{n}}.

Let GnG_{n} denote the subgroupoid of GG generated by KnK_{n}. Then GnG_{n} has local polynomial growth of degree at most sns_{n}. Indeed, given a compact open symmetric set D⊆GnD \subseteq G_{n}, there exists q∈Nq \in\mathbb{N} such that D⊆KnqD \subseteq K_{n}^{q}. It follows that

sup⁡x∈G(0)∣Dk∩(Gn)x∣≤sup⁡x∈G(0)∣Knqk∩Gx∣≤Cn(1+qk)sn≤Cnqsn(1+k)sn.\sup_{x \in G^{(0)}} \left|D^{k} \cap(G_{n})_{x}\right| \le\sup_{x \in G^{(0)}} \left|K_{n}^{qk} \cap G_{x}\right| \le C_{n}(1+qk)^{s_{n}} \le C_{n}q^{s_{n}}(1+k)^{s_{n}}.

It follows from Theorem A that dad⁡(Gn)≤⌊sn⌋\operatorname{dad}(G_{n}) \le\lfloor s_{n} \rfloor. Since ff is non-decreasing and unbounded, we can find a strictly increasing sequence (an)n(a_{n})_{n} of positive integers such that f(an)≥⌊sn⌋f(a_{n}) \ge\lfloor s_{n} \rfloor. Define a function ℓ:G→[0,∞)\ell: G \to[0,\infty) by

ℓ(g)=min⁡{∑j=1manj ∣ m≥1, g=g1⋯gm, gj∈Knj}\ell(g) = \min\left\{ \left. \sum_{j=1}^{m} a_{n_{j}} \ \right|\ m \ge1,\ g = g_{1} \cdots g_{m},\ g_{j} \in K_{n_{j}} \right\}

for all g∈G∖G(0)g \in G \setminus G^{(0)} and ℓ(g)=0\ell(g) = 0 for g∈G(0)g \in G^{(0)}. It is routine to check that ℓ\ell is a continuous coarse length function on GG.

To complete the proof we show that dad⁡G,ℓ(t)≤f(t)\operatorname{dad}_{G,\ell}(t) \le f(t). If 0≤t<a10 \le t < a_{1}, then Bℓ(t)=G(0)B_{\ell}(t) = G^{(0)} and dad⁡G,ℓ(t)=0≤f(t)\operatorname{dad}_{G,\ell}(t) = 0 \le f(t). For t≥a1t \ge a_{1}, there is a unique n∈Nn \in\mathbb{N} such that an≤t<an+1a_{n} \le t < a_{n+1}. Suppose g∈Bℓ(t)g \in B_{\ell}(t) is a non-unit element. Let g=g1⋯gmg = g_{1} \cdots g_{m} be such that ∑anj\sum a_{n_{j}} realises ℓ(g)\ell(g). Then ∑anj=ℓ(g)≤t<an+1\sum a_{n_{j}} = \ell(g) \le t < a_{n+1}. Thus, since (an)n(a_{n})_{n} is strictly increasing we must have nj≤nn_{j} \le n and hence gj∈Knj⊆Kng_{j} \in K_{n_{j}} \subseteq K_{n} for all j=1,…,mj = 1,\ldots,m. It follows that Bℓ(t)⊆⟨Kn⟩=GnB_{\ell}(t) \subseteq\langle K_{n} \rangle= G_{n}. Since a cover witnessing finite dynamic asymptotic dimension at scale Bℓ(t)B_{\ell}(t) in GnG_{n} also witnesses finite dynamic asymptotic dimension at scale Bℓ(t)B_{\ell}(t) in GG we get

dad⁡G,ℓ(t)≤dad⁡(Gn)≤⌊sn⌋≤f(an)≤f(t),\operatorname{dad}_{G,\ell}(t) \le\operatorname{dad}(G_{n}) \le\lfloor s_{n} \rfloor\le f(a_{n}) \le f(t),

as desired. □

Examples

Coarse groupoids and AF groupoids

Let (X,d)(X,d) be a uniformly locally finite extended metric space satisfying ∣Bd(x,R)∣≤C(1+R)s\left|B_{d}(x,R)\right| \le C(1+R)^{s} uniformly in x∈Xx \in X. Its coarse groupoid G(X,d)G(X,d) is principal and ample with unit space βX\beta X, where βX\beta X is the Stone–Cech compactification of XX (see [STY02] for more details on this construction). Moreover dad⁡(G(X,d))=asdim⁡(X,d)\operatorname{dad}(G(X,d)) = \operatorname{asdim}(X,d) by [GWY17, Theorem 6.4]. Given a compact open symmetric D⊆G(X,d)D \subseteq G(X,d), choose RR such that DD lies in the closure of the controlled set ER={(y,x)∣d(x,y)≤R}E_R = \{(y,x) \mid d(x,y) \le R\}. For x∈Xx \in X we then have ∣Dn∩G(X,d)x∣≤C(1+nR)s|D^n \cap G(X,d)_x| \le C(1+nR)^s. By Lemma 3.2 the function ω↦∣Dn∩G(X,d)ω∣\omega\mapsto|D^n \cap G(X,d)_\omega| is locally constant on βX\beta X, so density extends this bound to all ω∈βX\omega\in\beta X. Thus G(X,d)G(X,d) has local polynomial growth of degree at most ss, and Theorem A gives asdim⁡(X,d)≤⌊s⌋\operatorname{asdim}(X,d) \le\lfloor s \rfloor. For graphs with their extended graph metrics, this recovers the polynomial exponent bound in [Pap23].

AF groupoids have local polynomial growth of degree zero. Indeed, by writing G=⋃jKjG = \bigcup_j K_j as an increasing union of principal clopen subgroupoids with Kj(0)=G(0)K_j^{(0)} = G^{(0)} and Kj∖G(0)K_j \setminus G^{(0)} compact, every compact DD lies in some KjK_j, see e.g. [AOP25, Section 3]. Consequently Dn⊆KjD^n \subseteq K_j and sup⁡x∈G(0)∣Dn∩Gx∣≤sup⁡x∈G(0)∣Kj∩Gx∣<∞\sup_{x \in G^{(0)}} |D^n \cap G_x| \le\sup_{x \in G^{(0)}} |K_j \cap G_x| < \infty uniformly in nn. Theorem A therefore recovers dad⁡(G)=0\operatorname{dad}(G) = 0.

Partial action groupoids

We refer the reader to [2] and [1 Section 4.1] for material on partial action groupoids.

Let θ=(Xg,θg)g∈Γ\theta= (X_g,\theta_g)_{g \in\Gamma} be a partial action of a discrete group Γ\Gamma on a locally compact totally disconnected Hausdorff space XX with open domains. The partial action θ\theta is free if θg(x)=x\theta_g(x) = x implies g=eg = e whenever defined.

Write X⋊θΓX \rtimes_\theta\Gamma for the associated partial transformation groupoid, whose elements are (θg(x),g,x)(\theta_g(x),g,x) with x∈Xg−1x \in X_{g^{-1}}, see [1 Section 4.1] and [2]. There is a natural identification of (X⋊θΓ)(0)=X(X \rtimes_\theta\Gamma)^{(0)} = X.

For a discrete group, local polynomial growth of degree at most ss means that for every finite symmetric set F⊆ΓF \subseteq\Gamma there is a constant CF≥1C_F \ge1 such that ∣Fn∣≤CF(1+n)s|F^n| \le C_F(1+n)^s for all n∈N0n \in\mathbb{N}_0.

Corollary 4.1. Suppose Γ\Gamma is a discrete group with local polynomial growth of degree at most s≥0s \ge0 and that θ=(Xg,θg)g∈Γ\theta= (X_g,\theta_g)_{g \in\Gamma} is a partial action on a locally compact totally disconnected Hausdorff space XX with open domains. If θ\theta is free, then

dad⁡(X⋊θΓ)≤⌊s⌋.\operatorname{dad}(X \rtimes_\theta\Gamma) \le\lfloor s \rfloor.

Proof. The groupoid X⋊θΓX \rtimes_\theta\Gamma is ample because XX is locally compact and totally disconnected, and it is principal because θ\theta is free. The map

c ⁣:X⋊θΓ→Γ,c(θg(x),g,x)=g,c\colon X \rtimes_\theta\Gamma\to\Gamma,\qquad c(\theta_g(x),g,x)=g,

is a continuous groupoid homomorphism which is injective on each source fibre. If D⊆X⋊θΓD \subseteq X \rtimes_\theta\Gamma is compact open and symmetric, F=c(D)∪{e}F = c(D) \cup\{e\} is finite and symmetric and c(Dn)⊆Fnc(D^n) \subseteq F^n. Consequently,

∣Dn∩(X⋊θΓ)x∣≤∣Fn∣≤CF(1+n)s|D^n \cap(X \rtimes_\theta\Gamma)_x| \le|F^n| \le C_F(1+n)^s

for x∈Xx \in X. Theorem A now applies. □

For global actions of finitely generated groups the result above can also be obtained in a bit of a roundabout way by a combination of existing results in the literature: if a finitely generated group Γ\Gamma has polynomial growth of degree at most dd, then by Gromov’s Theorem it is virtually nilpotent. Combining [4 Corollary 1.10] and [GWY17, Theorem 4.7] we conclude that the dynamic asymptotic dimension of any free action of Γ\Gamma on a compact, metrizable, totally disconnected Hausdorff space is finite. Thus, [Pil25, Corollary 3.4] shows dad⁡(X×Γ)=asdim⁡(Γ)\operatorname{dad}(X \times\Gamma) = \operatorname{asdim}(\Gamma) and the latter is at most dd by [Pap23]. Thus, even though our result is not new in this situation, we believe it does provide a much more direct route.

Point set groupoids

In the interest of brevity, we refer the reader to [AOP25, Section 5.2] or [1 Section 4.2] for material on point set groupoids.

Let HH be a second-countable locally compact group. For a nonempty separated point set Λ⊆H\Lambda\subseteq H, let Ω(Λ)\Omega(\Lambda) be its Chabauty–Fell hull and put Ω0(Λ)={P∈Ω(Λ)∣e∈P}\Omega_0(\Lambda) = \{P \in\Omega(\Lambda) \mid e \in P\}. As in [AOP25, Section 5.2], its point set groupoid is

G(Λ)={(h,P)∈H×Ω0(Λ):h−1∈P},G(\Lambda) = \{(h, P) \in H \times\Omega_0(\Lambda) : h^{-1} \in P\},
s(h,P)=P,r(h,P)=hP,s(h, P) = P,\qquad r(h, P) = hP,

with the topology and operations inherited from the translation groupoid. This is an étale groupoid. We say that Λ\Lambda has finite local complexity if for every compact K⊆HK \subseteq H the collection {(x−1Λ)∩K∣x∈Λ}\{(x^{-1}\Lambda) \cap K \mid x \in\Lambda\} is finite. We say P∈Ω(Λ)P \in\Omega(\Lambda) has trivial translation stabilizer if hP=PhP = P implies h=eh = e.

Corollary 4.2. Let HH be a second-countable locally compact group with left Haar measure mm and local polynomial growth of degree at most ss in the sense that for every compact subset K⊆HK \subseteq H, there is a constant CK>0C_K > 0 such that

m(Kn)≤CK(1+n)sfor all n∈N.m(K^n) \leq C_K(1+n)^s \quad\text{for all } n \in\mathbb{N}.

Suppose Λ⊆H\Lambda\subseteq H has finite local complexity and every nonempty P∈Ω(Λ)P \in\Omega(\Lambda) has trivial translation stabilizer. Then G(Λ)G(\Lambda) is principal and ample, and

dad⁡(G(Λ))≤⌊s⌋.\operatorname{dad}(G(\Lambda)) \leq\lfloor s \rfloor.

Proof. The groupoid is Hausdorff and étale with compact unit space Ω0(Λ)\Omega_0(\Lambda); see [AOP25, Section 5.2]. Finite local complexity makes this unit space zero-dimensional by the first part of the proof of [1 Proposition 4.22], which uses no relative density. Thus G(Λ)G(\Lambda) is ample. The stabilizer hypothesis makes it principal. Finally, we show that G(Λ)G(\Lambda) has local polynomial growth of degree at most ss. Let D⊆G(Λ)D \subseteq G(\Lambda) be a compact open symmetric subset and let K⊆HK \subseteq H be the image of DD under the canonical projection map. Then KK is compact. Choose a relatively compact open identity neighbourhood VV such that the translates pVpV, p∈Pp \in P, are pairwise disjoint for every P∈Ω(Λ)P \in\Omega(\Lambda). Let M⊆HM \subseteq H be a compact set such that K∪V⊆MK \cup V \subseteq M and e∈Me \in M. Since the disjoint translates pVpV for p∈P∩Knp \in P \cap K^n are contained in KnV⊆Mn+1K^nV \subseteq M^{n+1}, we have

∣P∩Kn∣m(V)=∑p∈P∩Knm(pV)=m(⋃p∈P∩KnpV)≤m(Mn+1).|P \cap K^n|m(V) = \sum_{p \in P \cap K^n}m(pV) = m\left(\bigcup_{p \in P \cap K^n}pV\right) \leq m(M^{n+1}).

Using that HH has local polynomial growth we can find CM>0C_M > 0 such that m(Mn)≤CM(1+n)sm(M^n) \leq C_M(1+n)^s for all n∈Nn \in\mathbb{N}. Thus, for every P∈Ω0(Λ)P \in\Omega_0(\Lambda) we have

∣Dn∩G(Λ)P∣≤∣P∩Kn∣≤m(Mn+1)m(V)≤CMm(V)(2+n)s≤CM2sm(V)(1+n)s.|D^n \cap G(\Lambda)_P| \leq|P \cap K^n| \leq\frac{m(M^{n+1})}{m(V)} \leq\frac{C_M}{m(V)}(2+n)^s \leq\frac{C_M2^s}{m(V)}(1+n)^s.

Thus, G(Λ)G(\Lambda) has local polynomial growth of degree at most ss and Theorem A applies. □

Remark 4.3. Note in particular that unlike in the comparison results of [1], we are not requiring Λ\Lambda be relatively dense.

Use of AI assistance: This project was initiated in February 2026 during a visit of the first author to Newcastle University. During this visit the authors developed a potential proof strategy for Theorem A and established some preliminary results. At a later stage, the authors had conversations with ChatGPT-6.0 Astra to flesh out the details of the strategy developed in February. The technical details in the present proof are based on these conversations. As the arguments produced by AI were difficult to read and obscured many of the underlying ideas, the authors completely reorganised and rewrote the proof. Finally, AI tools were used for proofreading the manuscript.

References

Are Austad, Department of Mathematics, University of Oslo, P.O. Box 1053 Blindern, 0316 Oslo, Norway

Email address: [email protected]

Christian Bönicke, School of Mathematics, Statistics and Physics, Newcastle University, Newcastle upon Tyne NE1 7RU, UK

Email address: [email protected]

References

  1. [1]Are Austad and Christian Bönicke. Growth Conditions for Étale Groupoids: Amenability and Comparison. 2026. eprint: arXiv:2605.16013v3.arxiv.org/abs/2605.16013
  2. [2]Fernando Abadie. “On partial actions and groupoids”. In: Proc. Amer. Math. Soc. 132.4 (2004), pp. 1037–1047. ISSN: 0002-9939,1088-6826. DOI: 10.1090/S0002-9939-03-07300-3.DOI
  3. [4]A. Bartels. “Coarse flow spaces for relatively hyperbolic groups”. English. In: Compos. Math. 153.4 (2017), pp. 745–779. ISSN: 0010-437X. DOI: 10.1112/S0010437X16008216.DOI
  4. [5]Christian Bönicke and Kang Li. “Nuclear dimension of subhomogeneous twisted groupoid C*-algebras and dynamic asymptotic dimension”. English. In: Int. Math. Res. Not. 2024.16 (2024), pp. 11597–11610. ISSN: 1073-7928. DOI: 10.1093/imrn/rnae133.DOI
  5. [8]Anton Bernshteyn and Jing Yu. “Large-scale geometry of Borel graphs of polynomial growth”. English. In: Adv. Math. 473 (2025). Id/No 110290, p. 51. ISSN: 0001-8708. DOI: 10.1016/j.aim.2025.110290.DOI
  6. [9]Jorge Castillejos, Samuel Evington, Aaron Tikuisis, Stuart White, and Wilhelm Winter. “Nuclear dimension of simple C*-algebras”. English. In: Invent. Math. 224.1 (2021), pp. 245–290. ISSN: 0020-9910. DOI: 10.1007/s00222-020-01013-1.DOI
  7. [10]Kristin Courtney, Anna Duwenig, Magdalena C. Georgescu, Astrid an Huef, and Maria Grazia Viola. “Alexandrov groupoids and the nuclear dimension of twisted groupoid C*-algebras”. In: J. Funct. Anal. 286.9 (2024), Paper No. 110372, 49. ISSN: 0022-1236,1096-0783.DOI

Paper details

vlnc:W_bxa4ejka
Version
v1 · latest
Published
Fresh — added to the Hub within the last 30 days
Added
License
CC BY 4.0
Usage

views · 0

downloads · 0

A view counts once per browser per day, a download each time the PDF is requested.

  • Mathematics

Not recorded for this paper

View statement
Looking for related papers…
Contents