On perturbations that preserve the connectivity properties in tree percolations

arXiv:2607.01291 · math.PR, math-ph, math.CA, math.MG, math.MP · Submitted 2026-07-01 · Read on arXiv

Listen

Radio episode about this paper

Transcript

Introduction to the show: ident: Mathematics. The day in mathematics: new theorems, new proofs, and what they open.

Ada: We're looking at the paper: "On perturbations that preserve the connectivity properties in tree percolations".

Paper summary: Ada: So, to recap, this paper, "On perturbations that preserve the connectivity properties in tree percolations," is looking at how you modify edge probabilities in a percolation model on an infinite tree. The authors are investigating when you can apply these modifications without losing the fundamental property of whether the system has infinite clusters or not.

Emmy: They set up this scenario where edge retention probabilities p are replaced by something like the minimum of one and some scaled version of p, depending on how far that edge is from the root vertex. They explore two main scenarios based on whether your original model had infinite clusters to begin with.

Kolya: If you start with an infinite cluster, they study perturbations where the scaling factors are less than or equal to one. The goal there is to preserve that infinite connectivity property under these constrained changes in probability.

Ada: And if you start without an infinite cluster—meaning everything is finite—they look at perturbations where the scaling factors are greater than or equal to one. In this case, they want to see how far you can push the system while keeping those clusters finite.

Emmy: Their main result hinges on analyzing the cumulative effect of these perturbation factors along every infinite ray starting from the root vertex, which they denote by some product Q infinity n equals one q n. If this product converges to a positive and finite value, it implies that the perturbation is essentially just rescaling your original connectivity probabilities by a bounded factor.

Kolya: That convergence is what’s important because it means the qualitative behavior of the model won't change much; it suggests the system just gets rescaled in a predictable way. If that product doesn't converge nicely, they look at more complex stability conditions involving those boundary sets, which brings us into Theorem A.

Ada: Theorem A is where they get pretty technical, showing that even when the cumulative effect of those perturbations diverges in certain ways, the connectivity state of the system can still persist under mild necessary conditions.

Emmy: They prove this by using an energy functional called E T n q p in a truncated version of the tree and showing that under certain conditions, the probability of having an infinite cluster in the perturbed model is bounded by some multiple of your original probability.

Kolya: The proof for part one specifically addresses cases where you start with no infinite clusters, meaning almost surely there are no infinite clusters to begin with. They show that if a certain limit involving the perturbation factors and the original probability is zero, then the perturbed model also doesn't have an infinite cluster almost surely.

Ada: That sounds like they are providing a way to rule out new infinite clusters based on how much you change the probabilities along those rays, even when things get complicated.

Conclusion: Emmy: So, wrapping up the discussion on "On perturbations that preserve the connectivity properties in tree percolations," this paper is essentially about establishing robustness for percolation models on trees under specific types of edge probability modifications.

Ada: The authors show that if you start with a certain connectivity state, say infinite clusters, you can often keep it by using small enough perturbations where those scaling factors are less than or equal to one. Conversely, if the original system is finite, they find conditions on larger perturbations that still allow for stability.

Kolya: And the paper demonstrates that this stability isn't always obvious just by looking at a simple product along an infinite ray; it requires checking specific conditions on those products and sometimes using deeper tools like capacity characterization to show what happens.

Emmy: The title itself points to the idea that you are trying to find exactly what kind of perturbations are allowed—the ones that respect the underlying structure of the tree percolation. They aren't just any arbitrary changes to probabilities; they have specific constraints.

Ada: And their work has a direct application linking these ideas to problems in geometry, specifically relating them to how you can study Cantor sets using tree percolations and the Erdős similarity conjecture.

Kolya: So, what this means for the broader field is that we've found a way to rigorously prove that percolation models on trees are much more stable than we might have thought when you introduce these kinds of quantitative, non-trivial changes to their defining probabilities.

Emmy: It’s not a claim about everything being easy; it’s a statement about where the boundaries of stability lie for these specific models and how those boundaries are defined by these scaling factors.

MIRMUHSIN MAXMUD, VILLE SUOMALA

math.PR, math-ph, math.CA, math.MG, math.MP

Submitted: 2026-07-01

Updated: 2026-11-16

Comments: Accepted version, to appear in Electronic Communications in Probability

License: http://creativecommons.org/licenses/by/4.0/

Importance score: 89/100

The gist: On perturbations that preserve the connectivity properties in tree percolations we consider a general bond percolation on an infinite locally finite tree where edge retention probabilities are…

Key concepts

Tree Percolation
This is a mathematical model where edges on an infinite tree structure are retained or removed based on certain probabilities. It's used to study how clusters of connected nodes form within this branching network, often looking for the existence of infinitely large connected components.
Level-Dependent Perturbations (q|e|pe)
This refers to changing the edge retention probability based on its distance from the root in the tree. The perturbation factor $q$ depends on $|e|$, which is the distance from the root. This allows researchers to study how connectivity changes when different parts of the tree are perturbed differently.
Capacity Characterization
This is a key mathematical tool that relates a set's connectivity probability in percolation to its 'capacity'. Capacity measures how well a region can be connected or separated, providing a rigorous way to analyze the system's structure and stability.

Terminology

Summary

On perturbations that preserve the connectivity properties in tree percolations we consider a general bond percolation on an infinite locally finite tree where edge retention probabilities are replaced by min(1, qepe) where qn is a sequence of positive perturbation factors and e denotes the distance between the edge e and the root If the original percolation model admits infinite clusters it is of interest to investigate under which perturbations 0 < qn <= 1 this connectivity property is preserved conversely if the original percolation does not admit infinite clusters we are led to study the stability of such a property under perturbations satisfying qn ≥ 1 In both cases under minimal assumptions on the original model we show that the percolative behaviour is stable against certain quantitative non-trivial perturbations

Stability of Connectivity under Perturbations

The work investigates the stability of infinite clusters in percolation models on a locally finite rooted tree under level-dependent perturbations q = min(qepe, 1) where e is the graph distance from the root The analysis distinguishes between two cases based on whether the original model admits infinite clusters or not.

The perturbation sequence qn is restricted based on the initial state of the system; specifically, if Pp(I) = 1 we consider perturbations 0 < qn ≤ 1 to preserve infinite clusters If Pp(I) = 0 we look at perturbing sequences with qn ≥ 1 to keep clusters finite while increasing their size. p. 1

Where the paper says it

This monotonicity allows us to simplify our analysis by restricting the range of the perturbation sequence based on the initial state of the system: • If Pp(I) = 1, we are interested in perturbations {qn}n that decrease the edge retention probabilities while preserving the existence of infinite clusters. Thus, we consider perturbations 0 < qn ≤ 1.

Heuristically the cumulative effect of the perturbation along each infinite ray is governed by the product Q∞ n=1 qn If this product converges to a positive and finite limit then the perturbation merely rescales connectivity probabilities by a bounded factor and is not expected to change the qualitative behaviour of the model. p. 1
Where the paper says it

Heuristically, the cumulative effect of the perturbation along each infinite ray is governed by the product Q∞ n=1 qn. If this product converges to a positive and finite limit, then the perturbation merely rescales connectivity probabilities by a bounded factor (see (5)) and is, therefore, not expected to change the qualitative behaviour of the model.

Formal Proof of Theorem A

Theorem A establishes that under mild necessary conditions the connectivity state of the system persists even when the cumulative effect of the perturbation diverges Part (i) addresses the case where there are no infinite clusters Pp almost surely and qn ≥ 1 satisfies lim inf n→∞ Yn i=1 qi · Pp[o ←→ Vn] = 0 then the perturbed model does not have an infinite cluster almost surely Part (ii) addresses the case where there is an uncountable boundary cluster with positive probability Pp[o u ←−→ ∂T] > 0 and shows that there exists a perturbation q = min(qepe, 1) under which the almost-sure existence of an uncountable cluster is preserved and limn→∞ qn = 1

The proof of part (i) utilizes the energy functional E Tn qp in the truncation Tn and shows that Pqp[o ←→ Vn] ≤ 2QnPp[o ←→ Vn] where Qk:= Qk i=1 qi This leads to the conclusion that Pqp[o ←→ ∂T] = inf n∈N Pqp[o ←→ Vn].

Geometric Applications and Erdős Similarity

The paper presents a geometric application of Theorem A using the dyadic expansion of real numbers to encode each Cantor set C ⊂ [0, 1] as a subtree T Q C An application of Theorem A (ii) yields a sequence 0 0 This observation suggests a potential avenue to study the Erd˝os similarity problem for Cantor sets via tree percolations.

Corollary 4. p. 9

Where the paper says it

The claim is now an immediate consequence of (2). □ Corollary 4.

2 states that given a Cantor set C there is a set B of zero Lebesgue measure such that B ∩ (λC + t) ≠ ∅ for almost all t, λ This result follows from the previous proposition and the second Borel-Cantelli lemma. p. 9
Where the paper says it

□ Corollary 4. Given a Cantor set C, then there is a set B of zero Lebesgue measure, such that B ∩ (λC + t) ̸= ∅ for almost all t, λ.

Connection to Capacity Characterization

The proof of Theorem A relies heavily on Lyons’ celebrated capacity characterisation of percolation on arbitrary locally finite trees This characterization establishes that for any Borel set E ⊆ ∂T its connectivity probability is governed by its capacity capp(E) The capacity is defined using the energy functional Ep(µ) which relates to the gauge function Ψp

The proof of part (ii) involves a Ψp-level decomposition of the boundary ∂T into R∞ and Rn Case I shows that if Pp[o u ←−→ R∞] > 0 then there exists a perturbation q = min(qepe, 1) under which the almost-sure existence of an uncountable cluster is preserved Case II demonstrates that if Pp[o u ←−→ R∞] = 0 but Pp[o u ←−→ Rn] > 0 then a similar construction of a perturbation q˜ = min(qepe, 1) can be made such that Pqp˜ [o ←→ Rn] > 0

Necessary Conditions for Homogeneous Percolation

When the percolation model is homogeneous, Theorem A provides a necessary condition (1) which is also sufficient provided the underlying tree is spherically symmetric and retention probabilities are homogeneous or more generally level-dependent This condition involves an expression Eqp(µ) = 1 + X∞ n=0 [1 − pqn+1]p−n−1Q−1 n+1 1/En+1 The derivation shows that for any µ in dom Eqp, ∞ > Eqp(µ) − 1 ≥ X∞ n=0 [1 − pqn+1]p−n−1Q−1 n+1 1/En+1

The final part of the paper explores how the stability of percolation clusters under perturbations qe of the probability weights may also be considered a stability of percolation type problem now with perturbations induced by scaling and translating the set C This involves replacing the tree TC with TλC+t where λ and t vary continuously. The validity of β) in the setting of Proposition 4. p. 3

Where the paper says it

) in the boundary ∂T, we define ξ ∧ η := ( xk, if ξ ̸= η, where k = sup{n ≥ 0 : xi = yi for all 0 ≤ i ≤ n}; ξ, if ξ = η. We can then endow ∂T with a metric d by setting d(ξ, η) = 0 if ξ = η, and d(ξ, η) = e −|ξ∧η| otherwise.

1 may also be considered a stability of percolation type problem now with perturbations induced by scaling and translating the set C

Acknowledgements and References

The authors thank Jeff Steif for many valuable suggestions on an early draft of the manuscript The references cited include works by Lyons, Peres, Jung, Lai, Mooroogen, Kersting, Kolountzakis, and Shmerkin and Yavicoli The paper is organized with Section 2 recalling Lyons’ capacity characterisation of percolation and Section 3 dedicated to the proof of Theorem A Section 4 discusses a connection between the main result and the Erd˝os similarity problem for Cantor sets The acknowledgements section thanks Jeff Steif for many valuable suggestions on an early draft of the manuscript. The references list includes works from Charalambos Aliprantis and Kim Border, Grimmett, Hara and Slade, Kersting, Kolountzakis, Lyons and Peres, Pemantle and Peres, Shmerkin and Suomala The email addresses are Mirmukhsin.Makhmudov@oulu.fi and Ville. p. 0

Where the paper says it

ON PERTURBATIONS THAT PRESERVE THE CONNECTIVITY PROPERTIES IN TREE PERCOLATIONS MIRMUHSIN MAXMUD AND VILLE SUOMALA Abstract. We consider a general bond percolation on an infinite locally finite tree, where the edge retention probabilities pe are replaced by min{1, q|e|pe}, where {qn}n≥1 is a sequence of positive perturbation factors and |e| denotes the distance between the edge e and the root.

Suomala@oulu. p. 0
Where the paper says it

ON PERTURBATIONS THAT PRESERVE THE CONNECTIVITY PROPERTIES IN TREE PERCOLATIONS MIRMUHSIN MAXMUD AND VILLE SUOMALA Abstract. We consider a general bond percolation on an infinite locally finite tree, where the edge retention probabilities pe are replaced by min{1, q|e|pe}, where {qn}n≥1 is a sequence of positive perturbation factors and |e| denotes the distance between the edge e and the root.

fi. The arXiv identifier is 2607. p. 0
Where the paper says it

Both authors were supported by the Research Council of Finland via the project “Fractals and randomness”, grant no. 1 arXiv:2607.

01291v1 [math. p. 0
Where the paper says it

1 arXiv:2607. 01291v1 [math.

PR] 1 Jul 2026. p. 0
Where the paper says it

01291v1 [math. PR] 1 Jul 2026

Green — the paper says it, and the page is printed. Red — no passage carries it, so treat that sentence as unverified. The same check runs on your own text: check a summary against its source, $5.

Improvements for AI systems

  1. Improved stability analysis for percolation models on trees by analyzing level-dependent perturbations qp = min(qepe, 1). The improved system can determine if an infinite cluster exists almost surely under these specific quantitative non-trivial perturbations based on whether Pp(I) is 1 or 0.

  2. Enhanced understanding of divergent perturbation effects by distinguishing between the cumulative effect of a perturbation Q infinity n=1 qn and the genuinely interesting divergent cases like qn = 0 or qn = infinity. This allows for more precise predictions about qualitative behavioral changes in percolation models.

  3. Application to the Erd˝os similarity conjecture for Cantor sets by showing that a set B of zero Lebesgue measure exists such that B intersected with scaled and translated copies of C is non-empty almost surely. The improved system can verify this property using the capacity characterization of percolation on tree subgraphs defined by dyadic expansions.

Abstract

We consider a general bond percolation on an infinite locally finite tree, where the edge retention probabilities p e are replaced by 1,q ep e, where q n n at least 1 is a sequence of positive perturbation factors and e denotes the distance between the edge e and the root. If the original percolation model admits infinite clusters, it is of interest to investigate under which perturbations 0<q n at most 1 this connectivity property is preserved. Conversely, if the original percolation does not admit infinite clusters, we are led to study the stability of such a property under perturbations satisfying q n at least 1. In both cases, under minimal assumptions on the original model, we show that the percolative behaviour is stable against certain quantitative non-trivial perturbations. We also discuss an application of our results to the Erdős similarity conjecture for Cantor sets.

Sources

Related papers