Sharpness of the phase transition for constrained-degree percolation

Ivailo Hartarsky, Roger W. C. Silva · 2025

Read Sharpness of the phase transition for constrained-degree percolation by Ivailo Hartarsky, Roger W. C. Silva (2025) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2509-16162.txt.

Introduction

In some spread models on a grid, a link can open only if both its ends still have few open links. Ivailo Hartarsky and Roger W. C. Silva study when such a process stays small and when it can take off. Standard tools fail here, since links depend on each other over long range and lack key order traits.

The pair prove that below the critical point, the chance of reaching far falls off at an exponential rate. Their proof adapts a randomized algorithm method and a trick that boosts pivotal links. The result settles an open problem, and the trick may help in other dynamic spread models.

Catalog record

CREATORIvailo Hartarsky, Roger W. C. Silva
YEAR2025
COLLECTIONPAPERS
FORMATTXT
FILE SIZE45 KB
VERSION1.0
LICENSEhttp://creativecommons.org/licenses/by/4.0/
RIGHTSCC BY
SOURCEarXiv

Source: https://arxiv.org/abs/2509-16162

Browse the full library · About the Tower of Babel

Full text

The opening of the work, from the library's hosted text:

Sharpness of the phase transition for constrained-degree percolation Ivailo Hartarsky, Roger W. C. Silva Abstract We consider constrained-degree percolation on the hypercubic lattice. Initially, all edges are closed, and each edge independently attempts to open at a uniformly distributed random time; the attempt succeeds if, at that instant, both end-vertices have degrees strictly less than a prescribed parameter. The absence of the FKG inequality and the finite energy property, as well as the infinite range of dependency, make the rigorous analysis of the model particularly challenging. In this work, we show that the one-arm probability exhibits exponential decay in its entire subcritical phase. The proof relies on the Duminil-Copin–Raoufi–Tassion randomized algorithm method and resolves a problem of dos Santos and the second author. At the heart of the argument lies an intricate combinatorial transformation of pivotality in the spirit of Aizenman–Grimmett essential enhancements, but with unbounded range. This technique may be of use in other dynamical settings. MSC2020: 60K35; 82B43 Keywords: sharp threshold; constrained-degree percolation; pivotality 1 Introduction Percolation models have been extensively studied for decades. Since their introduction in [ 3 ] , they have become central objects of research in probability and mathematical physics, with numerous variants of the original model proposed over time. Several models (see e.g. [ 17 , 19 , 22 , 10 , 23 , 11 , 12 , 16 , 18 ] ) purposefully feature degree constraints, often making their analysis intricate. Our model of interest is the constrained degree percolation (CDP) introduced in [ 24 ] . In CDP, each vertex may accept connections up to a fixed limit, after which no new edges can be formed, while all previously established connections remain intact. Specifically, it is a dependent continuous-time percolation process, defined by the following dynamics. At time zero, all edges are closed; each edge attempts to open at a uniformly distributed random time, and the attempt is successful if, at that moment, both of its end-vertices have degrees strictly smaller than a given parameter \kappa . A formal definition of the model is provided in Section 1.1 . Unlike classical Bernoulli percolation, the CDP exhibits dependencies of all orders at any fixed time. Furthermore, the absence of the Fortuin–Kasteleyn–Ginibre (FKG) inequality and the lack of the finite energy property make its analysis challenging. 1.1 Model and result Let us start by introducing the CDP on {\mathbb{Z}}^{d} , d\geq 2 . Let 2\leq\kappa\leq 2d be an integer, which is fixed throughout the paper and kept implicit. We denote by \mathcal{E} the set of nearest neighbour (non-oriented) edges of {\mathbb{Z}}^{d} . Let U=(U_{e})_{e\in\mathcal{E}} be a sequence of independent uniform random variables on [0,1] with corresponding product measure {\mathbb{P}} . We denote by \Omega the set of U\in[0,1]^{\mathcal{E}} with distinct coordinates and such that there is no infinite walk 1 1 1 A walk is a sequence of vertices such that consecutive ones are joined by edges. A path is a walk consisting of distinct vertices. x_{1},x_{2},\dots in {\mathbb{Z}}^{d} such that i\mapsto U_{x_{i}x_{i+1}} is strictly monotone. It is classical that {\mathbb{P}}(\Omega)=1 , so we systematically restrict our attention to U\in\Omega . For any p\in[0,1] , we define the function \omega_{p}:\Omega\to\{0,1\}^{\mathcal{E}} as follows. We write \omega_{p}(U,e) for the value of \omega_{p}(U) at edge e\in\mathcal{E} and refer to e as p -open if \omega_{p}(U,e)=1 , and p -closed if \omega_{p}(U,e)=0 . Initially, only the (at most one) edge with U_{e}=0 is open, that is, \omega_{0}(U,e)=\mathbbm{1}_{U_{e}=0} . Each edge e\in\mathcal{E} attempts to open at time U_{e} and is successful only if its addition to the currently open edges would not create a vertex of degree larger than \kappa . In other words, p\mapsto\omega_{p}(U,xy) jumps from 0 to 1 at p=U_{xy} if |\{z\in{\mathbb{Z}}^{d}:xz\in\mathcal{E},\omega_{U_{xy}-}(U,xz)=1\}|<\kappa\text{ and }|\{z\in{\mathbb{Z}}^{d}:yz\in\mathcal{E},\omega_{U_{xy}-}(U,yz)=1\}|<\kappa, and stays equal to 0 for all p\in[0,1] otherwise. This definition is well-posed in \Omega , since for every e\in\mathcal{E} only a finite set of coordinates of U suffices to determine (\omega_{p}(U,e))_{p\in[0,1]} . We call a path p -open if all of its edges are p -open. The event that x\in{\mathbb{Z}}^{d} is connected to y\in{\mathbb{Z}}^{d} by a p -open path is denoted by x\xleftrightarrow{p}y . We write x\xleftrightarrow{p}\infty for the event that there exists an infinite p -open path starting at x . We consider the standard one-arm probability \displaystyle\theta_{n}(p) \displaystyle{}={\mathbb{P}}(\exists x\in{\mathbb{Z}}^{d},\|x\|_{1}=n,0\xleftrightarrow{p}x), \displaystyle\theta(p) \displaystyle{}={\mathbb{P}}(0\xleftrightarrow{p}\infty)=\lim_{n\to\infty}\theta_{n}(p). (1) Since \theta:[0,1]\to[0,1] is clearly non-decreasing, we naturally define p_{\mathrm{c}}=\sup\{p\in[0,1]:\theta(p)=0\}, with the convention \sup\varnothing=0 , which will never enter into effect. Our main result is the following. Theorem 1 (Sharp phase transition) . Consider the CDP on {\mathbb{Z}}^{d} , d\geq 2 , with \kappa\in\{2,\dots,2d\} . For p<p_{\mathrm{c}} , there exists \alpha_{p}>0 such that for all n\geq 1 , \theta_{n}(p)\leq e^{-n\alpha_{p}} . Remark 2 . Theorem 1 is stated and proved for {\mathbb{Z}}^{d} with its usual graph structure for concreteness. However, the proof applies mutatis mutandis for any transitive locally finite bipartite graph. For non-bipartite graphs, the structure of the switching path from Lemma 6 below becomes more complex, but is still sufficiently path-like for the argument to adapt to this case as well. 1.2 Background For \kappa=2d , the CDP reduces to the standard Bernoulli percolation model, for which Theorem 1 is a classical result [ 1 , 9 , 20 ] . We direct the reader to [ 13 ] for some background on this model. It is known that the CDP undergoes a nontrivial phase transition for all d\geq 2 and most values of \kappa . For example, [ 4 ] establishes this non triviality for d=2 and \kappa=3 , while also proving that percolation does not occur when d\geq\kappa=2 , even at p=1 . Further progress was made in [ 14 ] , which provides quantitative upper bounds on the critical time and characterizes the phase transition for all d\geq 3 and most choices of \kappa\geq 3 . Nonetheless, it remains open to determine whether the phase transition is trivial ( p_{\mathrm{c}}=1 ) for \kappa\in\{3,\dots,9\} for some dimensions, but it is conjectured that this is never the case. In view of this, let us emphasise that Theorem 1 applies regardless of whether the phase transition is trivial. It is straightforward that the CDP is stochastically dominated by standard Bernoulli percolation for every fixed p\in[0,1] . Consequently, Theorem 1 holds trivially whenever p lies below the Bernoulli percolation critical threshold, denoted by \overline{p_{\mathrm{c}}} . However, it is known that p_{\mathrm{c}}>\overline{p_{\mathrm{c}}} (see Theorem 1 in [ 4 ] for the case d=2 and \kappa=3 ; the argument extends for the general case), and therefore Theorem 1 remains nontrivial for all choices of d and \kappa . Let \widehat{p_{\mathrm{c}}}=\sup\{p\in[0,1]:\sum_{x\in{\mathbb{Z}}^{d}}{\mathbb{P}}(0\xleftrightarrow{p}x)<\infty\}\leq p_{\mathrm{c}} denote the suscptibility critical threshold for the CDP. The recent work [ 5 ] proves Theorem 1 for p<\widehat{p_{\mathrm{c}}} . Our result strengthens their findings: in particular, Theorem 1 implies p_{\mathrm{c}}=\widehat{p_{\mathrm{c}}} . This answers [ 5 ] *Question 1. in the setting of homogeneous (vertex-independent) constraints \kappa . 1.3 Outline of the proof A standard approach to proving sharpness in percolation models is the one of [ 8 ] using a randomised algorithm (see Section 2.4 ) and the OSSS inequality [ 21 ] (see Section 2.3 ). Applying this method directly fails. On the one hand, the measure \omega_{p}\circ{\mathbb{P}} is not monotonic and lacks other good properties.

Read the complete text