Polluted Modified Bootstrap Percolation
Janko Gravner, Alexander Holroyd, Sangchul Lee, David Sivakoff · 2025
Read Polluted Modified Bootstrap Percolation by Janko Gravner, Alexander Holroyd, Sangchul Lee, David Sivakoff (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-2503-15746.txt.
Introduction
Janko Gravner, Alexander Holroyd, Sangchul Lee, and David Sivakoff study a spread model on a grid where some sites are blocked forever, in work from 2025. In bootstrap percolation, empty sites fill when enough neighbors are filled. The modified rule here needs an occupied neighbor along each axis, and closed sites never fill.
The team asks what final share of sites ends up filled when both the start rate and the block rate are small. They find a sharp threshold with a logarithmic twist. Above it the final density fades to zero, and below it the density rises to one. The result settles a conjecture from 1997 about this correction.
Catalog record
| CREATOR | Janko Gravner, Alexander Holroyd, Sangchul Lee, David Sivakoff |
|---|---|
| YEAR | 2025 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 19 KB |
| VERSION | 1.0 |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2503-15746
Browse the full library · About the Tower of Babel
Full text
The opening of the work, from the library's hosted text:
Polluted Modified Bootstrap Percolation Janko Gravner, Alexander Holroyd, Sangchul Lee, David Sivakoff Abstract In the polluted modified bootstrap percolation model, sites in the square lattice are independently initially occupied with probability p p or closed with probability q q . A site becomes occupied at a subsequent step if it is not closed and has at least one occupied nearest neighbor in each of the two coordinates. We study the final density of occupied sites when p p and q q are both small. We show that this density approaches 0 0 if q ≥ C p 2 / log p − 1 q\geq Cp^{2}/\log p^{-1} and 1 1 if q ≤ p 2 / ( log p − 1 ) 1 + o ( 1 ) q\leq p^{2}/(\log p^{-1})^{1+o(1)} . Thus we establish a logarithmic correction in the critical scaling, which is known not to be present in the standard model , settling a conjecture of Gravner and McDonald from 1997 . 1 Introduction Bootstrap percolation has been extensively studied as a model for nucleation and metastability. In the version considered in this paper, each site in the lattice {\mathbb{Z}}^{2} is in one of the three states: occupied (also called infected), closed (immune), or open (susceptible). Initially, the state of each vertex is chosen independently at random: occupied with probability p and closed with probability q , where p+q\leq 1 . We will assume throughout that the probability parameters p and q are small. After the initial configuration is chosen, it evolves according to a simple deterministic rule in which occupied and closed sites never change their states, but an open site may become occupied. In the standard bootstrap percolation , an open site at time t becomes occupied at time t+1 if and only if at least 2 of its 4 nearest neighbors are occupied at time t . In the modified version, this transition is further restricted: at each step, every open site that has at least one occupied nearest neighbor to the east or west and at least one occupied nearest neighbor to the north or south becomes occupied. The standard “unpolluted” case with q=0 , a variant of which was introduced in [ CLR ] , has played a central role in the study of spatial dynamics in random settings. The first rigorous result, proved in [ vE ] , established that, as soon as p>0 , every site of {\mathbb{Z}}^{2} is eventually occupied with probability 1 . To probe the metastability properties of the model, a natural setting is to restrict the lattice to a finite square of large side length n . This lead to increased understanding of the mechanism for nucleation and the exponential scaling function is now known to a remarkable precision [AL, Hol, GHM, HM, HT] . Many variations were also extensively studied, e.g. [ BBDM , BDMS , BBMS1 , BBMS2 ] . A brief survey cannot do justice to the variety and depth of the results in this area, so we refer to the review [ Mor ] for further background. The “polluted” standard case with q>0 was introduced in [ GM ] . Here, the main focus is the final density of occupied sites. The main result of [ GM ] is that there exist constants c,C>0 so that when (p,q)\to(0,0) , (1.1) \mathbb{P}\left(\text{the origin is occupied}\right)\to\begin{cases}1&\text{if }q<cp^{2}\\ 0&\text{if }q>Cp^{2}\\ \end{cases} Rigorous extensions of this result have been relatively scarce, but higher-dimensional cases [ GraH , GHS , DYZ ] and general two-dimensional models [ Gho ] have been addressed recently. Figure 1: The dynamics with p=0.1 , q=0.01 attempts to fill in a square. Black sites are initially occupied, including all sites on the boundary, grey sites are eventually occupied, and red sites are closed. For the modified model, the argument in [ GM ] , which proves ( 1.1 ) for the standard rule, allows the possibility of a logarithmic correction in the scaling (see also open problem (2) in [ GHS ] ). In this paper, we confirm that, indeed, p^{2} in ( 1.1 ) needs to be replaced by p^{2}/\log p^{-1} , although we have to allow a possible correction power of \log\log p^{-1} in the lower bound. It is unclear whether such a power can be removed, and we leave this as a natural open question. While the modified model is in many senses similar to the standard one, notably in the dominant exponential term in the nucleation scaling [ Hol ] , there are important differences. The recent result on the scaling correction [ Har ] is one example, and the present paper provides another. The cause is illustrated in Figure 1 , where narrow “chimneys” — one site wide segments of open sites, with closed sites at their ends — block the inward progress of occupation. Unsurprisingly, the chimney phenomenon is closely reflected in our proofs. In the standard rule, occupation can easily invade such chimneys. We now formally state our results. Theorem 1.1 . There exists a constant C>0 so that if q\geq Cp^{2}/\log(1/p) , then (1.2) \lim_{q\to 0}\mathbb{P}_{p,q}\bigl((0,0)\text{ is eventually occupied}\,\bigr)=0. Theorem 1.2 . If q\leq\frac{p^{2}}{\log(1/p)(\log\log(1/p))^{4}}, then (1.3) \lim_{p\to 0}\mathbb{P}_{p,q}\bigl((0,0)\text{ is eventually occupied}\,\bigr)=1. We prove these two theorems in the following two sections respectively. 2 Stopping the infection Fix an odd integer m\geq 1 and an integer k\geq 2 . Fix also \epsilon,\delta>0 . Let M=\lfloor\delta p^{-1}\log p^{-1}\rfloor and N=2m\lceil\epsilon/(mp)\rceil . We partition the lattice into rectangles of width M and height N , which we call blocks : for z=(z_{1},z_{2})\in{\mathbb{Z}}^{2} , we let R_{z}=(Mz_{1},Nz_{2})+([0,M-1]\times[0,N-1]) . A block R_{z} is called safe if there exists an x in the upper half of R_{z} so that: (S1) x is closed; (S2) there is a rectangle of width m and height k\lceil\epsilon p^{-1}\rceil , whose top edge is included in the top edge of R_{z} , contains x within its middle vertical line and contains no initially occupied sites; and (S3) there is a rectangle of width k\cdot\lfloor\delta p^{-1}\log p^{-1}\rfloor and height m , whose left edge is contained within the bottom half of the left edge of R_{z} , and contains no initially occupied sites. \boldsymbol{\times} m m \frac{k\delta\log p^{-1}}{p} \frac{2\epsilon}{p} \frac{\delta\log p^{-1}}{p} \frac{k\epsilon}{p} Figure 2: A safe block (outlined in blue) has a closed site marked by \times and no occupied sites in the red protective region. Its core is outlined by thicker red lines. Dimensions of rectangles are given with integer parts omitted. If R_{z} is safe , we call the union of the two rectangles from conditions (S2) and (S3) its protective region ; the horizontal and vertical line segments in the middle of the two rectangles together form its core ; see Figure 2 . Lemma 2.1 . Fix m and k , and any \eta>0 . Then \epsilon , \delta and a constant C can be chosen so that, for any fixed z , \mathbb{P}\left(R_{z}\text{ is {\color[rgb]{0,0,0}safe}}\right)\geq 1-\eta, provided that q\geq C\frac{p^{2}}{\log p^{-1}} and p is small enough. Proof. Note first that the probability that a fixed rectangle of dimensions in condition (S2) has an occupied site is at most {\color[rgb]{0,0,0}p\cdot m\cdot k\cdot\frac{2\epsilon}{p}=2mk\epsilon.} Therefore, the conditional probability that (S2) happens