Supercritical percolation on nonamenable graphs: Isoperimetry, analyticity, and exponential decay of the cluster size distribution
Jonathan Hermon, Tom Hutchcroft
TL;DR
This paper proves exponential decay of cluster size distribution in supercritical percolation on nonamenable graphs, showing anchored expansion of infinite clusters and analyticity of key observables in the supercritical phase.
Contribution
It establishes exponential decay of cluster sizes and analyticity of observables for supercritical percolation on nonamenable graphs, answering longstanding questions.
Findings
Exponential decay of cluster size probabilities for p > p_c.
Infinite clusters exhibit anchored expansion almost surely.
Percolation probability and related functions are analytic in the supercritical phase.
Abstract
Let be a connected, locally finite, transitive graph, and consider Bernoulli bond percolation on . We prove that if is nonamenable and then there exists a positive constant such that \[\mathbf{P}_p(n \leq |K| < \infty) \leq e^{-c_p n}\] for every , where is the cluster of the origin. We deduce the following two corollaries: 1. Every infinite cluster in supercritical percolation on a transitive nonamenable graph has anchored expansion almost surely. This answers positively a question of Benjamini, Lyons, and Schramm (1997). 2. For transitive nonamenable graphs, various observables including the percolation probability, the truncated susceptibility, and the truncated two-point function are analytic functions of throughout the supercritical phase.
Click any figure to enlarge with its caption.
Figure 1
Figure 2
Figure 3Peer Reviews
No public reviews on file for this paper yet. If you reviewed it on a platform where reviews are public (OpenReview, ICLR, NeurIPS, ICML), you can paste yours below so the community can read it here.
Videos
No videos yet. Explain this paper in a talk, walkthrough, or lecture? Add one.
Supercritical percolation on nonamenable graphs: Isoperimetry, analyticity, and exponential decay of the cluster size distribution
Jonathan Hermon and Tom Hutchcroft
Abstract
Let be a connected, locally finite, transitive graph, and consider Bernoulli bond percolation on . We prove that if is nonamenable and then there exists a positive constant such that
[TABLE]
for every , where is the cluster of the origin. We deduce the following two corollaries:
Every infinite cluster in supercritical percolation on a transitive nonamenable graph has anchored expansion almost surely. This answers positively a question of Benjamini, Lyons, and Schramm (1997). 2. 2.
For transitive nonamenable graphs, various observables including the percolation probability, the truncated susceptibility, and the truncated two-point function are analytic functions of throughout the supercritical phase.
1 Introduction
Let be a connected, locally finite graph. In Bernoulli bond percolation, we choose to either delete or retain each edge of independently at random with retention probability to obtain a random subgraph of with law . The connected components of are referred to as clusters, and we denote the cluster of in by . Percolation theorists are particularly interested in the geometry of the open clusters, and how this geometry changes as the parameter is varied. It is natural to break this study up into several cases according to the relationship between and the critical probability
[TABLE]
which always satisfies when is transitive and has superlinear growth [21] (this result is easier and older if has polynomial growth [47, Corollary 7.19] or exponential growth [46, 34]).
Among the different regimes this leads one to consider, the subcritical phase is by far the easiest to understand. Indeed, the basic features of subcritical percolation have been well understood since the breakthrough 1986 works of Menshikov [49] and Aizenman and Barsky [1] which, together with the work of Aizenman and Newman [4], establish in particular that if is a connected, locally finite, transitive graph then
[TABLE]
for every , where we write for the set of edges that have at least one endpoint in the cluster . See also [23, 22, 39] for alternative proofs, and [16] for more refined results. Note that is well-defined for every and every connected, locally finite graph , and in particular that the definition does not depend on the choice of the vertex .
In contrast, the supercritical phase is rather more difficult to understand, and no good theory yet exists for supercritical percolation on general transitive graphs. A central role in the theory of this phase is played by the distribution of finite clusters. For with , most results concerning this distribution rely crucially on the important and technically challenging work of Grimmett and Marstrand [28], which allowed in particular for a complete proof of the asymptotics
[TABLE]
The upper bounds of (1.2) and (1.3) were proven by Chayes, Chayes, Grimmett, Kesten, and Schonmann [18] and by Kesten and Zhang [43] respectively, both conditional on the then-conjectural Grimmett-Marstrand theorem. All of these upper bounds rely essentially on renormalization, a technique that is unavailable outside the Euclidean setting. The lower bound of (1.2) is trivial, while the lower bound (1.3) was proven by Aizenman, Delyon, and Souillard [2]; see also [27, Sections 8.4-8.6]. The more refined properties of finite clusters in supercritical percolation on remain an area of active research, see e.g. [17, 15] and references therein. We note moreover that percolation in the slightly supercritical regime remains very poorly understood on even when is large, in which case critical () percolation is now well understood following in particular the work of Hara and Slade [32]; See [33] for an overview of progress and problems in this direction.
In this paper, we develop a theory of supercritical percolation on nonamenable transitive graphs, studying in particular the distribution of finite clusters, the geometry of the infinite clusters, and the regularity of the dependence of various observables on . Our main result, from which various corollaries will be derived, is the following theorem. An extension of the theorem to the quasi-transitive case is given in Section 4.
Theorem 1.1**.**
Let be a connected, locally finite, nonamenable, transitive graph. Then for every .
We stress that our methods are completely different to those used in the Euclidean context. We also note that we do not study the question of the (non)uniqueness of the infinite cluster, which remains completely open at this level of generality. We refer the interested reader to [40, 35] and references therein for an up-to-date account of what is known regarding this question and the closely related problem of understanding critical percolation on nonamenable transitive graphs.
Here, we recall that a locally finite graph is said to be nonamenable if its Cheeger constant
[TABLE]
is positive, where for each set we write for the set of edges with one endpoint in and the other outside of . For example, Euclidean lattices such as are amenable, while regular trees of degree at least three and transitive tessellations of hyperbolic spaces are nonamenable. Previously, Theorem 1.1 was known only for , where it follows by a simple counting argument [12, Theorem 2] and does not require transitivity. We believe that the full conclusions of Theorem 1.1 were only previously known for trees. Note that the theorem fails without the assumption of transitivity: For example, the graph obtained by attaching a -regular tree and a -regular tree by a single edge between their respective origins has and .
It was observed in [8] that the argument of [2] can be generalized to prove that for every whenever is a Cayley graph of a finitely presented amenable group. In Corollary 3.5 we prove via a different argument that in fact for every amenable transitive graph and every . This gives a converse to Theorem 1.1, so that, combining both results, we obtain the following appealing percolation-theoretic characterization of nonamenability for transitive graphs:
Corollary 1.2**.**
Let be a connected, locally finite, transitive graph. The following are equivalent:
* is nonamenable.* 2. 2.
* and for every .* 3. 3.
* and for some .*
We note in particular that Theorems 1.1 and 1.2 resolve several questions raised by Bandyopadhyay, Steif, and Timár [8, Questions 1 and 2 and Conjecture 1].
1.1 Analyticity
One nice corollary of Theorem 1.1 is that many quantities including the percolation probability , the truncated susceptibility , the free energy , and the truncated two-point function all depend analytically on throughout the entire supercritical phase. The regularity properties of these functions have historically been a subject of great importance in percolation, motivated in part by the still uncompleted project to rigorize the heuristic computation of for various planar lattices by Sykes and Essam [55]. See e.g. [41, 27, 25] for further background.
Let us now state our general analyticity result. Let denote the set of finite connected subgraphs of containing , and say that a function has subexponential growth if . It is an observation essentially due to Kesten [41, 42], and a consequence of Morera’s Theorem and the Weierstrass -test, that if has subexponential growth then the function
[TABLE]
is analytic in on the set , which always contains when is transitive as discussed above. More precisely, this means that for every with there exists and a complex analytic function on the complex ball of radius around whose restriction to agrees with the function under consideration. See Proposition 3.1 for details. Thus, Theorem 1.1 has the following corollary.
Corollary 1.3**.**
Let be a connected, locally finite, nonamenable, transitive graph, let and let have subexponential growth. Then is an analytic function of on .
Previously, it was not even known that was differentiable on under the same hypotheses. On the other hand, it is an immediate consequence of the mean-field lower bound on [23, Theorem 1.1] that the percolation probability is always non-differentiable at when is transitive, so that Corollary 1.3 is best-possible in this case. (Note that analyticity of follows from Corollary 1.3 by taking .) It remains open to prove that the free energy (a.k.a. open-clusters-per-vertex) is not analytic at .
We remark that in amenable transitive graphs, subexponential decay of the cluster volume distribution makes analyticity in the supercritical phase a much more delicate issue. Indeed, while infinite differentiability of , , , and in the supercritical phase of is an easy consequence of superpolynomial decay [27, Theorem 8.92], the analyticity of these functions was settled only in the very recent work of Georgakopoulos and Panagiotis [25, 26].
1.2 Isoperimetry and random walk
In this section, we discuss applications of Theorem 1.1 to the isoperimetry of infinite percolation clusters on nonamenable transitive graphs. It is easily seen that percolation clusters cannot be nonamenable when , since infinite clusters will always contain arbitrarily large isoperimetrically ‘bad regions’, such as long paths. Nevertheless, one has the intuition that supercritical percolation clusters on nonamenable transitive graphs ought to be ‘essentially nonamenable’ in some sense. With the aim of making this intuition precise, Benjamini, Lyons, and Schramm [10] defined the anchored Cheeger constant of a connected, locally finite graph to be
[TABLE]
where is a vertex of whose choice does not affect the value obtained, and said that has anchored expansion if . They asked [10, Question 6.5] whether every infinite cluster in supercritical Bernoulli bond percolation on has anchored expansion whenever is transitive111In fact they stated the question without the assumption of transitivity. An example showing that the question has a negative answer without this assumption is given in Remark 3.6. and nonamenable. (See also [47, Question 6.49].) Partial progress on this question was made by Chen, Peres, and Pete [19], who proved in particular that every infinite cluster has anchored expansion a.s. for sufficiently close to . Their result does not require transitivity. Their paper also treats some other related models, most notably establishing anchored expansion for supercritical Galton-Watson trees. Anchored expansion has subsequently been established for various other random graph models, see [20, 7, 11].
Applying Theorem 1.1 together with the argument of [19, Theorem A.1], we are able to give a complete solution to [10, Question 6.5].
Corollary 1.4**.**
Let be a connected, locally finite, nonamenable, transitive graph, and let . Then every infinite cluster in Bernoulli- bond percolation on has anchored expansion almost surely.
Corollary 1.4 also allows us to analyze the behaviour of the simple random walk on the infinite clusters of Bernoulli percolation under the same hypotheses. Indeed, it is a theorem of Virág [57, Theorem 1.2] that if is a bounded degree graph with anchored expansion then there exists a positive constant such that the simple random walk return probabilities on satisfy the inequality for every vertex of , where is a -dependent constant. In Section 3.3 we establish that a matching lower bound also holds in our setting, thereby deducing the following corollary. We write for the -step return probability of simple random walk from on the percolation configuration .
Corollary 1.5**.**
Let be a connected, locally finite, nonamenable, transitive graph, let , and let be a vertex of . Then
[TABLE]
almost surely on the event that the cluster of is infinite.
Remark 1.6*.*
Virág [57, Theorem 1.1] also established that anchored expansion implies positive speed of the random walk on bounded degree graphs. For supercritical percolation on unimodular transitive nonamenable graphs, positive speed of the random walk was already proven to hold by Benjamini, Lyons, and Schramm [10, Theorem 1.3]. Combining Corollary 1.4 with the aforementioned result of Virág allows us to extend this result to the nonunimodular case. The resulting fact that every cluster is transient in supercritical percolation on nonunimodular transitive graphs was first proven by Tang [56].
The same return probability asymptotics given by Corollary 1.5 also appear in many other random graph models of nonamenable flavour [53, 7]. For results on isoperimetry and random walks in supercritical percolation on , see e.g. [51, 9] and references therein.
About the proof and organization.
The proof of the main theorem, Theorem 1.1, is given in Section 2. The starting point of the proof is to use Russo’s formula to express the -derivative of the truncated exponential moment as the sum of a positive term, which corresponds to the cluster growing while remaining finite, and a negative term, which corresponds to the finite cluster becoming infinite. (To do this rigorously we instead truncate at a large finite volume.) The proof then hinges on two key ideas, each of which allows us to bound the absolute value of one of these two terms. The bounds on the negative term work by re-purposing ideas from the Burton-Keane proof that there is at most one infinite cluster in percolation on amenable transitive graphs [14]. We use here the fact that supercritical percolation on a nonamenable transitive graph always stochastically dominates an invariant percolation process with trifurcation points (Lemma 2.5), which is a sort of weak converse to Burton-Keane and is due in the unimodular case to Benjamini, Lyons, and Schramm [10]. Meanwhile, for the positive term, we first bound the derivative in terms of certain ‘skinny trees of bridges’, and then bound the resulting expression using an inductive argument. This is the most technical part of the paper. Finally, the derivative itself, which is the difference of these two terms, can be bounded using martingale methods. Once these three bounds are in hand, the finiteness of follows easily.
We then apply Theorem 1.1 to deduce Corollaries 1.2, 1.3, 1.4 and 1.5 in Section 3, generalize Theorem 1.1 to the quasi-transitive case in Section 4, and give closing remarks and open problems in Section 5.
Remark 1.7*.*
An interesting feature of our proof is that, although we rely on an equality (2.3) obtained by expressing the derivative of in two different ways, we do not use the fact that this equality is related to the derivative. In particular, we do not integrate any differential inequality nor use any sprinkling, in contrast with the classical results mentioned above. Speculatively, this may mean that some of the techniques we develop here have applications to critical percolation on amenable transitive graphs under the (presumably contradictory) assumption that there is an infinite cluster at .
2 Proof of the main theorem
In this section we prove Theorem 1.1. Let be a connected, locally finite graph, and let be a vertex of . For each we write for the cluster of in , and write for the number of edges touching . We will usually take be an instance of Bernoulli- bond percolation on , in which case we write and when there is no risk of confusion. Recall that denotes the set of all finite connected subgraphs of containing . Given a function , we write
[TABLE]
for every and . To prove Theorem 1.1, it suffices to prove that if is transitive and nonamenable then for every there exists such that .
Given and , the truncated expectation is a polynomial in and is therefore differentiable. We begin our analysis by expressing the -derivative of in two different ways. The first way to express the derivative is in terms of the fluctuation in the number of open edges in the cluster of , which we now introduce. For each finite subgraph of , we define to be the set of edges of with at least one endpoint in the vertex set of , define to be the edges of , and define to be . (Note that is not always the same thing as the edge boundary of the vertex set of , and it is possible for edges of to have both endpoints in the vertex set of .) For each we also define the fluctuation
[TABLE]
Then for every , , and we may compute that
[TABLE]
In many situations, it is fruitful to bound the absolute value of this expression by viewing as the final value of a certain martingale (indeed, a stopped random walk) that arises when exploring the cluster of one step at a time. We will apply this strategy to bound the absolute value of the derivative in Section 2.2.
Next, we apply Russo’s formula to give an alternative expression for the derivative in terms of pivotal edges. For each and , let and let . (Here and elsewhere we abuse notation to identify elements of with subsets of whenever it is convenient to do so. This should not cause any confusion.) Russo’s formula [27, Theorem 2.32] states that if depends on at most finitely many edges, then
[TABLE]
Applying this formula to of the form for , we deduce that
[TABLE]
We denote the two terms appearing on the right hand side of this expression by
[TABLE]
so that
[TABLE]
for every , every , and every . Note that , , and all depend linearly on for fixed and , that is nonnegative if is nonnegative and that is nonnegative if is increasing.
Our strategy will be to use the equality (2.3) to obtain bounds on moments of certain random variables. In the remainder of this section we carry this out conditional on three supporting propositions, each of which gives control of one of the three quantities , , or , and which will be proved in the following three subsections. We begin by stating the following proposition, which is proven in Section 2.1.
Proposition 2.1**.**
Let be a connected, locally finite, nonamenable, transitive graph. Then for every there exists a constant such that
[TABLE]
for every nonnegative function , every , and every .
We remark that the complementary inequality always holds trivially on every connected, locally finite graph.
Applying Proposition 2.1 to the function , we obtain that for every there exists a positive constant such that
[TABLE]
for every , and . We want to show that if is sufficiently small then the expectation on the left is bounded uniformly in . To do this, we bound each term on the right hand side by the sum of a term that can be absorbed into the left hand side and a term that we will show is bounded as for sufficiently small values of . For the second term, this is quite straightforward to carry out using the bound
[TABLE]
The term on the second line can readily be shown to be bounded as for sufficiently small via a martingale analysis, as summarized by the following proposition which is proven in Section 2.2.
Proposition 2.2**.**
Let . Then
[TABLE]
for every locally finite graph , every , every , and every .
Bounding is rather more complicated. Let be a finite connected graph. For each edge of , let denote the subgraph of spanned by all edges of other than . Given a vertex and a set of vertices in , we write for the set of edges such that there exists such that is not connected to in , and define
[TABLE]
We will use these quantities to bound . To do this, we first write
[TABLE]
for each . Let have and for each edge let be an endpoint of in , and let . Then we have that
[TABLE]
Since this inequality holds for every set with , we deduce that
[TABLE]
for every and . Summing over , we obtain that
[TABLE]
for every . Similarly to (2.5), we can bound this expression by
[TABLE]
The second term is dealt with by the following proposition, which is proven in Section 2.3.
Proposition 2.3**.**
Let and . Then there exists such that
[TABLE]
for every locally finite graph , every and every .
Let us now see how the proof of Theorem 1.1 can be concluded given Propositions 2.1, 2.2 and 2.3.
Proof of Theorem 1.1.
Let be a connected, locally finite, transitive, nonamenable graph, and let . Letting be the constant from Proposition 2.1, we deduce from (2.4), (2.5), and (2.8) that
[TABLE]
for every and . Taking limits as we obtain that
[TABLE]
for every . Propositions 2.2 and 2.3 imply that there exists such that the right hand side is finite for every , completing the proof. ∎
It now remains to prove Propositions 2.1, 2.2 and 2.3.
2.1 Bounding the negative term
In this section we prove Proposition 2.1. Note that the proof of this proposition is the only place where the assumption of nonamenability and transitivity are used in the proof of Theorem 1.1. The proof is also ineffective in the sense that it does not yield any explicit lower bound on the constant , and is in fact the only ineffective step in the proof of Theorem 1.1.
Let be a locally finite graph, let , and let be a finite subset of . We define the quantity
[TABLE]
By Menger’s Theorem [31], this quantity is equal to the maximum size of a set of edge-disjoint paths from to in the subgraph of spanned by . Note in particular that is an increasing function of . Proposition 2.1 will be deduced from the following proposition.
Proposition 2.4**.**
Let be a connected, locally finite, nonamenable, transitive graph. Then for every there exists a positive constant such that
[TABLE]
for every finite.
The proof of Proposition 2.4 uses elements of the Burton-Keane [14] argument, which is usually used to establish uniqueness of the infinite cluster in percolation on amenable transitive graphs. This argument in fact establishes that the inequality (2.9) holds when there are infinitely many infinite clusters for percolation with parameter . (In the amenable case one can then prove by contradiction that such do not exist, since the left hand side of (2.9) is at most .) To apply this argument in our setting, since we are not assuming that there are infinitely many infinite clusters in Bernoulli- percolation, we will first need to find an appropriate automorphism-invariant percolation process that has infinitely many infinite connected components each of which is infinitely ended and which is stochastically dominated by Bernoulli -percolation. We do this by a case analysis according to whether or not is unimodular, applying the results of [10] in the unimodular case and of [40] in the nonunimodular case.
We will borrow in particular from the presentation of the Burton-Keane argument given in [47, Theorem 7.6]. Let be a graph and let . We say that a vertex is a furcation of if closing all edges incident to would split the component of in into at least three distinct infinite connected components.
Lemma 2.5**.**
Let be a connected, locally finite, nonamenable, transitive graph, and let . Then there exists an automorphism-invariant percolation process on such that the following hold:
The origin is a furcation of with positive probability. 2. 2.
The process is stochastically dominated by Bernoulli-* bond percolation on .*
Here, we recall that if is a connected, locally finite, transitive graph, a random variable taking values in is an automorphism-invariant percolation process if its law is invariant under the automorphisms of . See e.g. [47, 10] for background on general automorphism-invariant percolation processes.
Before beginning the proof of this lemma, let us briefly introduce the notion of unimodularity. We refer the reader to [47, Chapter 8] for further background. Let be a connected, locally finite, transitive graph with automorphism group . The modular function of is defined to be , where is the stabilizer of in and is the orbit of under . We say that is unimodular if and that is nonunimodular otherwise. Most graphs occurring in examples are unimodular, including all Cayley graphs of finitely generated groups and all transitive amenable graphs [54].
Proof of Lemma 2.5.
Recall the definition of the uniqueness threshold has a unique infinite cluster -a.s.. The usual Burton-Keane argument implies that if then the origin is a furcation for Bernoulli- percolation with positive probability; See in particular the proof of [47, Theorem 7.6]. Thus, if we may take to be Bernoulli- percolation for some . The main result of [40] states that if is nonunimodular (or more generally if has a nonunimodular transitive subgroup of automorphisms), which completes the proof in this case. Now suppose that is unimodular. A result of Benjamini, Lyons, and Schramm [10, Lemma 3.8 and Theorem 3.10] states that for every , Bernoulli- bond percolation on stochastically dominates an automorphism-invariant percolation process that is almost surely a forest all of whose components are infinitely-ended (see also [5, Theorem 8.13]). More precisely, [10, Theorem 3.10] implies that this is the case222Strictly speaking, [10, Theorem 3.10] states that can be taken to be a forest with nonamenable components. This is stronger than the statement we give here since every nonamenable tree is infinitely-ended. Similarly, the statement of [10, Lemma 3.8] requires that the Bernoulli percolation process has an infinitely-ended component rather than non-uniqueness per se, but it is well-known that these two properties are equivalent for Bernoulli percolation by insertion and deletion tolerance. when there is uniqueness at , while [10, Lemma 3.8] implies that this is the case when there is non-uniqueness at . (Here, we recall that a locally finite tree is infinitely-ended if it contains infinitely many disjoint infinite simple paths.) Any such automorphism-invariant process clearly has furcations, and so meets the conditions required by the lemma. ∎
The rest of the proof of Proposition 2.4 is very similar to the Burton-Keane argument as presented in [47, Section 7.3]; we include the details for completeness. We will use the following elementary combinatorial lemma.
Lemma 2.6**.**
Let be a tree in which every vertex has degree at least two. If is a finite set of vertices in , then there exists a set of edge-disjoint paths from to in .
Proof.
We may assume that is non-empty, the claim being trivial otherwise. Let be the smallest connected subtree of containing , so that is the union of all finite simple paths beginning and ending in . Let be the set of edges that are incident to a vertex of but do not belong to , and observe that the maximum number of edge-disjoint paths from to infinity is at least . (In fact this is an equality.) Indeed, for each element of we can choose an infinite simple path from to in with first edge that does not revisit , and any choice of such a path for each will result in a collection of paths that are mutually edge-disjoint. Let be the set of vertices that belong to . The definition of ensures that every vertex has degree at least in . Writing and for degrees in and and for the edge set of , we deduce that
[TABLE]
and hence that
[TABLE]
completing the proof. ∎
Proof of Proposition 2.4.
Let be a connected, locally finite, transitive, nonamenable graph, and let be a fixed root vertex of . Let , let be Bernoulli- bond percolation, and let be as in Lemma 2.5. Let be the set of furcation points of . Since stochastically dominates we have that for every finite, and so it suffices to prove that
[TABLE]
for every finite. Since is automorphism-invariant and is transitive, it suffices to prove the deterministic statement
[TABLE]
since (2.10) follows from (2.11) by taking expectations.
To prove (2.11), pick a spanning tree of each infinite cluster of and let be the union of all of these spanning trees. For each connected component of , let be the set of furcations of , and note that is contained in the disjoint union of the sets . For each component of , let be obtained from by iteratively deleting all vertices of degree at most , and note that the set of furcations of coincides with the set of furcations of . Since every vertex in has degree at least in , it follows from Lemma 2.6 that
[TABLE]
Since the components of are disjoint, we deduce that
[TABLE]
as claimed, where the sums are over the connected components of . ∎
Proof of Proposition 2.1.
Fix a vertex of . Let be independent copies of Bernoulli- bond percolation on , and let be defined by
[TABLE]
Since we can explore without revealing the status of any edge in , the configuration is also distributed as Bernoulli- bond percolation on . If is finite then any path from to in must pass through the set , and we deduce that
[TABLE]
when is finite. Taking expectations on both sides and using that and are independent, we deduce from Proposition 2.4 that
[TABLE]
where is the constant from Proposition 2.4. Using the standard fact that whenever and are bounded random variables and is a -algebra such that is -measurable, we deduce immediately that if is nonnegative then
[TABLE]
and the claim follows by taking expectations over . ∎
2.2 Bounding the total derivative
In this section we prove Proposition 2.2. This is the easiest of the three propositions used in the proof of Theorem 1.1; The methods used are completely standard and go back to some of the earliest work on percolation, see e.g. [44, 24] and [3, Section 3].
Proof of Proposition 2.2.
Let be a connected, locally finite graph, let , and let . Let be a sequence of i.i.d. mean-zero random variables with and , and let . Exploring the cluster of one edge at a time leads to a coupling of percolation on with the sequence and a stopping time such that and if then ; See e.g. [38, Section 3] for details. Using this coupling, we can write
[TABLE]
for each . Since for each , we also have that
[TABLE]
for each . Finally, since for every , we deduce from Azuma’s inequality that
[TABLE]
for every and , so that the right hand side of (2.12) is finite whenever . ∎
2.3 Bounding the positive term
In this section we prove Proposition 2.3. This is the most difficult of the three propositions going into the proof of Theorem 1.1. The basic idea is to control the probability by induction on . The proof will (implicitly) establish the explicit moment estimate
[TABLE]
which holds for every countable, locally finite graph , , , , and . It will be important for us to work on arbitrary graphs in this section to facilitate the induction. Indeed, working on arbitrary graphs (or arbitrary subgraphs of a given graph) in this way is a useful trick to avoid non-monotonicity problems in inductive analyses of percolation, which we believe was first used by Kozma and Nachmias in [45] following a suggestion of Peres.
We begin with the base case . Note that is bounded from above by the intrinsic radius of , i.e., the maximal graph distance in of a vertex from . It will therefore be relevant to bound the probability of having a large skinny cluster, whose intrinsic radius is of the same order as its volume; think of below as being at least in the following lemma.
Lemma 2.7**.**
Let be a locally finite graph and let be a vertex of . Then the bound
[TABLE]
holds for every and .
Proof.
Consider exploring the cluster of as follows: at stage , expose the value of those edges that touch , the set of vertices with intrinsic distance exactly from , and have not yet been exposed. Stop when . If and , there must exist at least stages where the sum of degrees in of the vertices in is at most . At each such stage, the conditional probability that given everything that has happened up to stage is at most . We deduce that
[TABLE]
as claimed, where the inequality has been used in the second inequality. ∎
Remark 2.8*.*
Using Lemma 2.7, we can already conclude the proof of the weaker statement that the intrinsic radius of a finite cluster has an exponential tail in supercritical percolation on a nonamenable transitive graph. This can be done via the same strategy used for the volume, but using Lemma 2.7 instead of Proposition 2.3 to bound the term , which is easily seen to be at most .
We now introduce the notation that we will use to set up our induction. Let be a finite connected graph. Recall that two vertices of are said to be (edge) -connected to each other if there exist two edge-disjoint paths in from to . Being -connected is an equivalence relation, so that can be decomposed into a collection of -connected components. An edge of is said to be a bridge if its endpoints are in different 2-connected components of , or equivalently if the subgraph of spanned by all edges other than is disconnected. We define to be the finite graph whose vertices are the 2-connected components of and where two 2-connected components and of are connected in if there is an edge of with one endpoint in and the other in . It follows readily from the definitions that the edges of are naturally in bijection with the bridges of and that is a tree (hence the notation).
Now let be a finite connected graph, let be a vertex of , and let . Write for the 2-connected component of in . We define to be the maximum number of edges in a subgraph of spanned by the union of the geodesics between and exactly leaves of the tree . (By a leaf we mean a degree one vertex distinct from the root .) If has fewer than leaves we set . Note that
[TABLE]
For each and we define the quantity
[TABLE]
Note that is trivially equal to zero when or but that can be positive. We will prove Proposition 2.3 by analysis of the following inductive inequality.
Lemma 2.9**.**
The inequality
[TABLE]
holds for every , , and .
Proof of Lemma 2.9.
Let be countable and locally finite and let . Let be i.i.d. Uniform random variables. We say that an edge is -open if , and otherwise that it is -closed, so that the subgraph of spanned by the -open edges is distributed as Bernoulli- bond percolation on . We write for the cluster of in , write for the number of edges of that touches, and write . Write for the law of the collection of random variables .
Fix , , and , and consider the event , . It suffices to prove that
[TABLE]
We may assume that has degree at least , since the claim is trivial otherwise. Suppose that the event holds, and let be a collection of leaves of attaining the maximum in the definition of . (Note that this collection is not unique, but that the choice will not matter. In particular, we can and do choose to be a measurable function of the cluster .) Let be the subtree of spanned by the union of the geodesics connecting and the root . We say that an edge of is a last-branching edge if deleting it from results in two connected components , where contains the root and the following conditions hold:
- •
is either a path or an isolated vertex;
- •
The endpoint of that belongs to is either equal to or has degree at least three in .
Observe that every last-branching edge of is naturally associated both to a leaf of and to a bridge of , which we call a last-branching bridge. In particular, we may enumerate the last-branching bridges of by in such a way that for each the geodesic from to in passes through the edge corresponding to but does not pass through the edge of corresponding to for any . Write .
As the form of the recursion (2.15) suggests, we will perform surgery to a random edge in . Write , and consider the event
[TABLE]
Since and there are at most -open edges in on the event , we may compute that
[TABLE]
where we have used that for every in the final inequality, and hence that
[TABLE]
Write for the subgraph of spanned by those edges of that do not touch . To bound we now argue that
[TABLE]
where is the event that the following conditions all hold:
- (i)
and .
- (ii)
There is exactly one -open edge touched by that is not -open, and this edge lies in the set . In particular, this edge has an endpoint that does not lie in .
- (iii)
The -cluster of in the graph touches edges of .
- (iv)
Every -open edge in the -cluster is also -open.
- (v)
.
It may seem that the unions in (2.17) should begin at , rather than , . What is written is correct, however: is positive by the assumption that has degree at least in , while would mean by (i) that has fewer than leaves, which is incompatible with the conditions that (ii) holds and that .
The only other part of the claim (2.17) that merits explanation is the implicit claim that if (ii) holds then
[TABLE]
Without loss of generality assume that . Consider the subgraph of spanned by the union of the geodesics between and the root. Since is the last-branching bridge associated to , the tree contains one of the endpoints of the edge corresponding to and we therefore observe that
[TABLE]
where the term corresponds to the edge itself. On the other hand, suppose that are leaves of attaining the maximum in the definition of and let be the distance in from to the subgraph of spanned by the union of the geodesics between the leaves and the root. Then considering the subgraph of spanned by the union of the geodesics between the leaves , the leaf , and the root yields that
[TABLE]
(again, the term corresponds to the edge itself) and hence that
[TABLE]
as claimed.
It remains to estimate the probability of the event . Let be the simpler event that the following conditions hold:
- (ib)
and .
- (iib)
There is exactly one -open edge touched by that is not -open, and this edge has an endpoint that does not lie in .
- (iiib)
The -cluster of in the graph touches edges of .
- (vb)
.
By definition, the probability that the condition (ib) holds is at most . On the other hand, the conditional probability that (iiib) and (vb) hold given both that (ib) and (iib) hold and given the cluster and the edge is equal to
[TABLE]
which is at most since was defined by taking a supremum over all graphs. Thus, we have that
[TABLE]
for every and . Since was arbitrary, the claim now follows from (2.16) and (2.17). ∎
We now apply Lemmas 2.7 and 2.9 to prove Proposition 2.3 via a generating function analysis.
Proof of Proposition 2.3.
In order to analyze the inductive inequality of Lemma 2.9, we introduce for each and the function given by
[TABLE]
which is a sort of multivariate generating function. Lemma 2.9 implies the inductive inequality
[TABLE]
and using the change of variables and yields that
[TABLE]
for every , , and . (Note that both sides of this inequality could be equal to , but this will not cause us any problems.) On the other hand, Lemma 2.7 implies that
[TABLE]
for every and , and since the right hand side of (2.19) is increasing in it follows that
[TABLE]
for every and . If , , and then separate consideration of the two cases and yields that
[TABLE]
for every , and hence that
[TABLE]
for every , , and , where we used that for every . Substituting (2.20) into (2.18) and inducting over yields that
[TABLE]
for every , , , and .
We now apply (2.21) to conclude the proof. Let be a countable, locally finite graph and let . Note that
[TABLE]
for every , , and . Since
[TABLE]
for each and , it follows that
[TABLE]
for every , , , and , where we used (2.21) in the final inequality. Using that , we deduce that
[TABLE]
for every , , , and . It follows in particular that
[TABLE]
for every , , , and . (A similar computation yields (2.13).) Taking , it follows that for each and there exists
[TABLE]
such that
[TABLE]
for every and . This clearly implies the claim. ∎
3 Proofs of corollaries
In this section we apply Theorem 1.1 to deduce Corollaries 1.2, 1.3, 1.4 and 1.5.
3.1 Analyticity
In this section, we apply the following proposition to deduce Corollary 1.3 from Theorem 1.1.
Proposition 3.1**.**
Let be a connected, locally finite graph, let and let have subexponential growth. Then is an analytic function of on .
Although the statement is more general, the proof of this proposition is essentially identical to the original argument of Kesten [42, Pages 250-251]. Many further results on analyticity can be found in the works of Georgakopoulos and Panagiotis [25, 26]. (Indeed, Corollary 4.14 of [25] can be seen to imply Proposition 3.1, but we prefer to give a direct proof since that result is rather technical.)
Proof.
Let have subexponential growth. We wish to show that for each there exists and an analytic function such that
[TABLE]
for every . By Morera’s Theorem, it suffices to prove that for every there exists such that the series converges uniformly in . To prove this, it suffices by the Weierstrass -test to prove that for each there exists such that
[TABLE]
We stress that the supremum in this expression is taken over complex with . To prove eq. 3.1, we observe that
[TABLE]
for every and , and hence that
[TABLE]
for every and . The right hand side is finite if , concluding the proof. ∎
Proof of Corollary 1.3.
This is immediate from Theorem 1.1 and Proposition 3.1. ∎
Remark 3.2*.*
One can strengthen the conclusions of Proposition 3.1 to give analyticity at and using the methods of [25].
3.2 Anchored expansion
The following proposition, which allows us to deduce Corollary 1.4 from Theorem 1.1, is implicit in the proof of [19, Theorem A.1]. We give a proof both for completeness and to stress that the argument does not require any isoperimetric assumptions on the ambient graph .
Proposition 3.3**.**
Let be a connected, locally finite graph. Let and suppose that . Then every infinite cluster of Bernoulli- bond percolation on has anchored expansion with anchored Cheeger constant
[TABLE]
almost surely.
Before beginning the proof, let us note that if is a connected locally finite graph then
[TABLE]
for each vertex of . Indeed, if is any finite set of vertices in and is the subgraph of induced by then and , so that .
Proof.
Fix a vertex of . For each , let be the set of connected subgraphs of containing that touch exactly edges. Given , for each , let be the set of open edges in , where we recall that is the set of edges that touch but do not belong to . For each , we define to be the event that there exists a subgraph such that (equivalently, every edge of is open in ) and . By (3.2), the anchored Cheeger constant satisfies on the event . Thus, to prove the claim it suffices to prove that if then
[TABLE]
for every , since Markov’s inequality will then imply that for every as desired. To prove (3.3), first observe that if and then
[TABLE]
Summing over the possible choices of with and , we deduce that if then
[TABLE]
and hence that
[TABLE]
To conclude, simply note that if then is increasing in , so that
[TABLE]
If then by definition of , and it follows that the right hand side of (3.4) is finite for as claimed. ∎
Remark 3.4*.*
When is transitive, it is a consequence of indistinguishability [48, 29] that for each there exists a deterministic constant such that every infinite cluster has anchored Cheeger constant equal to almost surely.
We now turn to Corollary 1.2. It is a result of Häggström, Schonmann, and Steif [30] (see also [47, Corollary 8.38]) that if is an amenable transitive graph and is an automorphism-invariant percolation process on , then every cluster of has almost surely. Thus, as discussed in the introduction, Proposition 3.3 has the following corollary.
Corollary 3.5**.**
Let be a connected, locally finite, transitive graph. If is amenable then for every .
Proof of Corollary 1.2.
This is immediate from Theorem 1.1 and Corollary 3.5. ∎
Remark 3.6*.*
Consider the graph formed by attaching a binary tree to each vertex of . This graph is nonamenable, has bounded degrees, and satisfies . Moreover, if , then Bernoulli- percolation on has a unique infinite cluster almost surely, which is distributed as the graph obtained by taking the unique infinite cluster of percolation on and attaching an independent subcritical Galton-Watson tree to each vertex. This graph clearly has subexponential growth, and consequently does not have anchored expansion. This example shows that Theorem 1.1 cannot be extended to arbitrary connected, bounded degree, nonamenable graphs, and therefore gives a negative answer to [10, Question 6.5] as originally stated.
3.3 Random walk analysis
Proof of Corollary 1.5.
As discussed in the introduction, it is a theorem of Virág [57, Theorem 1.2] that if is a bounded degree graph with anchored expansion then there exists a positive constant such that the simple random walk return probabilities on satisfy the inequality for every vertex of , where is a -dependent constant. This theorem together with Corollary 1.4 yield the desired upper bound. Thus, it suffices to prove that for each we have that
[TABLE]
almost surely on the event that the cluster of is infinite. We call a path in a pipe if all of its internal vertices have degree two in . The estimate (3.5) will be deduced as a consequence of the following geometric claim:
[TABLE]
We now prove this claim. Fix and . Write for the intrinsic ball of radius around in , and for the set of vertices with intrinsic distance exactly from . Since has anchored expansion a.s. on the event that it is infinite by Corollary 1.4, it must trivially also have exponential growth a.s. on the event that it is infinite. Indeed, it follows from Theorem 1.1 and Proposition 3.3 that there exists a constant and a random variable that is almost surely finite on the event that is infinite such that for every . On the other hand, since the cardinality of the intrinsic -ball of any vertex is deterministically at most , where is the maximum degree of , it follows that
[TABLE]
for every and almost surely on the event that is infinite. It follows in particular that there exist constants and such that
[TABLE]
for every almost surely on the event that the cluster of is infinite. Let be the set of vertices in such that there exists a not necessarily open path of length in starting at that does not visit any vertex of other than at its starting point, so that
[TABLE]
for every almost surely on the event that the cluster of is infinite. Moreover, choosing points greedily shows that for every and there exists a subset of with cardinality at least such that any two distinct points in have distance at least in . Furthermore, we can and do choose in such a way that it is a measurable function of . (This property is crucial to the argument of the next paragraph. It is very important here that the definitions of and do not require the path of length to be open; the argument would not work otherwise.) Putting the above facts together, we deduce that there exist constants and such that
[TABLE]
for every almost surely on the event that the cluster of is infinite.
Let and let be the event that contains a pipe of length . Let . For each element of , the conditional probability given that contains a pipe of length starting at is at least , since we can take a path of the required length starting at that is disjoint from other than at , find all of its edges to be open, and find to be closed all the edges that touch a vertex of the path other than but are not included in the path. On the other hand, the separation between the points of makes all of these events independent from each other, and we deduce that
[TABLE]
for every and . Choosing
[TABLE]
and using the inequality yields that
[TABLE]
Thus, it follows by Borel-Cantelli that there exists an almost surely finite such that the event occurs for every almost surely. Together with (3.7), this shows that the event holds for every almost surely on the event that is infinite. This completes the proof of (3.6).
It remains to deduce (3.5) from (3.6). This argument is well-known and appears in e.g. [57, Example 6.1], so we will keep the discussion brief. Let and be random variables such that contains a pipe of length for every . Pick one such pipe, call it the -pipe, and let be the vertex of the -pipe at minimal intrinsic distance from . In time , the random walk on can return to using the following strategy: Walk to in exactly steps, spend the following steps performing an excursion from to itself inside the -pipe, then return to in the final steps. Each of these three stages has probability at least to occur for an appropriate choice of constant , concluding the proof. ∎
4 Extension to quasi-transitive graphs
Recall that a graph is said to be quasi-transitive if the action of its automorphism group on its vertex set has at most finitely many orbits. Theorems concerning percolation on transitive graphs can almost always be generalized to the quasi-transitive case, and ours are no exception.
Theorem 4.1**.**
Let be a connected, locally finite, nonamenable, quasi-transitive graph. Then for every .
Once this theorem is established, extensions of Corollaries 1.3, 1.4 and 1.5 to the quasi-transitive case follow by essentially the same proofs as in the transitive case.
Let be a connected, locally finite, nonamenable, quasi-transitive graph, and let be a vertex of . The only place in the proof of Theorem 1.1 where transitivity is used is in the proof of Proposition 2.1. Thus, to prove Theorem 4.1, it suffices to prove that for every there exist positive constants and such that
[TABLE]
for every and . Indeed, once this is established the proof may be concluded in a very similar way to the transitive case.
Let be the orbits of the action of on . The proof of Lemma 2.5 generalizes to show that for every , there exists an automorphism-invariant percolation process such that has furcations almost surely and is stochastically dominated by Bernoulli- bond percolation on . Thus, the proof of Proposition 2.4 yields that there exists and a positive constant such that
[TABLE]
for every finite set . The proof of Proposition 2.1 then yields that
[TABLE]
for every and . In order to deduce an inequality of the form (4.1) from (4.3), it suffices to prove the following lemma.
Lemma 4.2**.**
Let be a connected, locally finite, quasi-transitive graph with maximum degree , let , and let be the orbits of the action of on . Then for every there exist positive constants and such that
[TABLE]
for every and .
Proof of Lemma 4.2.
It suffices to prove the claim for . Fix . Since is connected, we may reorder so that is adjacent to for every . Since acts transitively on for each , every vertex in must be adjacent to at least one vertex in for each . Let be the maximum degree of , and write for each . We first claim that there exists a positive constant such that for each we have that
[TABLE]
for every . The first inequality is trivial. For the second, consider exploring one edge at a time. On the event under consideration, we must query some number edges with one endpoint in and the other in and find that at most of these edges are open. Since , the claimed inequality (4.5) follows by standard large deviation estimates for Binomial random variables.
To conclude, we note that we trivially have , and apply (4.5) and a union bound to deduce that
[TABLE]
for every . This immediately implies the claim. ∎
Proof of Theorem 4.1.
Let and be as in Lemma 4.2, and note that (4.3) implies that
[TABLE]
for every and . The second term on the right is finite by Lemma 4.2, so that (4.6) is of the form required by (4.1). Thus, the proof of Theorem 4.1 may be concluded in a very similar manner to the proof of Theorem 4.1 as discussed above. ∎
5 Closing remarks and open problems
5.1 Analyticity at the uniqueness threshold
The analyticity of and throughout the entire supercritical phase established by Corollary 1.3 is in stark contrast to the behaviour of the untruncated two-point function , which can be discontinuous on the same interval under the same hypotheses.
Indeed, it is known that there exists a unimodular transitive graph for which and for which there are infinitely many infinite clusters in Bernoulli- percolation almost surely. The most easily understood example with these properties is probably the product of two three-regular trees, which is treated by the results of [36] and [50]. We claim that for such a graph , there must exist vertices and such that is discontinuous at . Indeed, it is a theorem of Lyons and Schramm [48, Theorem 4.1] that if is transitive and unimodular and is such that there are infinitely many infinite clusters -a.s. then (see also [56] for the nonunimodular case). Thus, it follows by our assumptions that there exist with . On the other hand, the Harris-FKG inequality implies that both in the infinite cluster for every , so that must have a jump discontinuity at as claimed. (The fact that the qualitative change in behaviour at is not necessarily reflected in any failure of regularity of at was also remarked on in [25].)
5.2 Intrinsic geodesics in the hyperbolic plane
Theorem 4.1 also has consequences for the geometry of intrinsic geodesics in percolation in the hyperbolic plane. Indeed, let be a locally finite, quasi-transitive, simply connected, nonamenable planar map with locally finite dual , which is also quasi-transitive and nonamenable. Every edge of has a corresponding dual edge . (See e.g. [6, Section 2.1] for detailed definitions.) If is Bernoulli- bond percolation on then the configuration defined by is distributed as Bernoulli- bond percolation on , and it follows from the results of [13] that , so that -percolation on is dual to -percolation on .
Suppose that is an edge of with endpoints and , and let and be the two faces incident to . Let be Bernoulli- bond percolation on and let and be the clusters of and in . If is closed and is connected to in then at least one of or must be finite. Moreover, if is connected to and is finite then the collection of edges other than whose duals are in the boundary of must contain an open path from to , so that
[TABLE]
Thus, it follows from sharpness of the phase transition (1.1) and Theorem 4.1 that for every there exists a constant such that
[TABLE]
for every pair of adjacent vertices and of and every . (With a little more work one can also obtain similar bounds for non-neighbouring pairs of vertices.) This contrasts the behaviour at , where it is proven in [37, Theorem 6.1] that, under the same assumptions, \mathbf{P}_{p_{u}}\bigl{[}d_{\mathrm{int}}(x,y)\geq n\mid x\leftrightarrow y\bigr{]}\asymp n^{-1} for some neighbouring pairs of vertices.
5.3 Conjectures for amenable graphs
We end with some conjectures concerning supercritical percolation on general transitive graphs that could potentially unify our results with those from the Euclidean case [18, 43, 2, 51]. We expect that these conjectures have been around for some time as folklore. We recall that if is a connected, locally finite graph, the isoperimetric profile of is defined to be
[TABLE]
Conjecture 5.1**.**
Let be an infinite, connected, locally finite, transitive graph. Then for every there exist positive constants and such that
[TABLE]
Note that the nonamenable case of Conjecture 5.1 is exactly Theorem 1.1, and that the case is covered by (1.3). In the case that is a Cayley graph of a one-ended finitely presented group, the lower bound of Conjecture 5.1 is implicit in the proof of [8, Theorem 3]. For the same class of graphs and for sufficiently close to one can also establish the upper bound of Conjecture 5.1 via a Peierls argument.
Let us also draw attention to the following much weaker question, which also remains open.
Conjecture 5.2**.**
Let be an infinite, connected, locally finite, transitive graph. Then the truncated susceptibility is finite for every .
In contrast to the volume, we expect that the radius of a finite supercritical cluster has an exponential tail on every transitive graph. Similar statements should also hold for the intrinsic radius. The nonamenable case of this conjecture is implied by Theorem 1.1, while the case was proven by Chayes, Chayes, Grimmett, Kesten, and Schonmann [18].
Conjecture 5.3**.**
Let be an infinite, connected, locally finite, transitive graph. Then for every there exists a positive constant such that
[TABLE]
Finally, we conjecture that the infinite clusters in supercritical percolation on always inherit an anchored version of any isoperimetric inequality satisfied by .
Conjecture 5.4**.**
Let be an infinite, connected, locally finite, transitive graph, and let . Then there exists a positive constant such that
[TABLE]
almost surely on the event that is infinite.
The nonamenable case of this conjecture is implied by Corollary 1.4, while the case was established by Pete [51]. Similarly to above, for Cayley graphs of one-ended, finitely presented groups and close to the conjecture may be established via a Peierls argument, see [51, Theorem 1.5]. We remark that Conjectures 5.1 and 5.4 are also closely related to Pete’s exponential cluster repulsion conjecture [52, Conjecture 12.32].
5.4 Other models
It would be very interesting (and seemingly non-trivial) to extend our analysis to dependent percolation models such as the random cluster model. One can also wonder whether our main theorems hold for an arbitrary automorphism-invariant percolation process that is subjected to Bernoulli noise.
Question 5.5**.**
Let be an infinite, connected, locally finite, nonamenable transitive graph. Let be an ergodic, automorphism-invariant percolation process on , let be an independent Bernoulli process in which each edge is included independently at random with probability , and suppose that the symmetric difference has infinite clusters almost surely.
Is the origin exponentially unlikely to be in a large finite cluster of ? 2. 2.
Does every infinite cluster of have anchored expansion almost surely?
Note that most of our analysis extends straightforwardly to such a model; it seems that the only part of our proof that breaks down badly is in the proof of Proposition 2.1, where we use that conditioning on does not affect the distribution of the restriction of the process to the complement of .
Acknowledgments
We thank the anonymous referees for their close reading and helpful suggestions.
The reference list from the paper itself. Each links out to its DOI / PubMed record.
- 1[1] M. Aizenman and D. J. Barsky. Sharpness of the phase transition in percolation models. Comm. Math. Phys. , 108(3):489–526, 1987.
- 2[2] M. Aizenman, F. Delyon, and B. Souillard. Lower bounds on the cluster size distribution. J. Statist. Phys. , 23(3):267–280, 1980.
- 3[3] M. Aizenman, H. Kesten, and C. M. Newman. Uniqueness of the infinite cluster and continuity of connectivity functions for short and long range percolation. Comm. Math. Phys. , 111(4):505–531, 1987.
- 4[4] M. Aizenman and C. M. Newman. Tree graph inequalities and critical behavior in percolation models. J. Statist. Phys. , 36(1-2):107–143, 1984.
- 5[5] D. Aldous and R. Lyons. Processes on unimodular random networks. Electron. J. Probab. , 12:no. 54, 1454–1508, 2007.
- 6[6] O. Angel, T. Hutchcroft, A. Nachmias, and G. Ray. Hyperbolic and parabolic unimodular random maps. Geom. Funct. Anal. , 28(4):879–942, 2018.
- 7[7] O. Angel, A. Nachmias, and G. Ray. Random walks on stochastic hyperbolic half planar triangulations. Random Structures Algorithms , 49(2):213–234, 2016.
- 8[8] A. Bandyopadhyay, J. Steif, and A. Timár. On the cluster size distribution for percolation on some general graphs. Rev. Mat. Iberoam. , 26(2):529–550, 2010.