How thin does random interlacement have to be so that a random walk can see through it?
Nicolas Bouchot · 2025
Read How thin does random interlacement have to be so that a random walk can see through it? by Nicolas Bouchot (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-2505-15260.txt.
Introduction
Nicolas Bouchot studies how sparse a random tangle of walk paths must be before another walk can slip through it, in a 2025 paper. The tangle, called random interlacement, is a spray of independent walk trails across a grid, with a dial for density. Past work probed how the tangle meets large sets.
Bouchot finds the exact density window where the tangle's blocking power, measured as capacity, fades next to a plain blown-up set. Inside that window a far-off walk starts to see through the tangle. A matching result covers a single walk trapped in a large ball for a set time, with time swapped for density. The work sharpens limit theory for this model.
Catalog record
| CREATOR | Nicolas Bouchot |
|---|---|
| YEAR | 2025 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 81 KB |
| VERSION | 1.0 |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2505-15260
Browse the full library · About the Tower of Babel
Full text
The opening of the work, from the library's hosted text:
How thin does random interlacement have to be so that a random walk can see through it? Nicolas Bouchot Abstract The random interlacements ℐ ( u ) \mathscr{I}(u) at level u u has been introduced by Sznitman, as a Poissonian collection of independent simple random walk trajectories on ℤ d \mathbb{Z}^{d} , d ≥ 3 d\geq 3 , with intensity u > 0 u>0 . Since then, several works investigated the properties of the random interlacements intersected with large sets of ℤ d \mathbb{Z}^{d} . In this paper, we study the asymptotic behavior of the capacity of ℐ ( u ) ∩ D N \mathscr{I}(u)\cap D_{N} , where D N D_{N} is the blow up of a compact set D D , with typical size N N . We determine the correct window ( u N ) N ≥ 1 (u_{N})_{N\geq 1} of the intensity parameter for which the capacity cap ( ℐ ( u N ) ∩ D N ) \mathrm{cap}(\mathscr{I}(u_{N})\cap D_{N}) starts to become negligible compared to cap ( D N ) \mathrm{cap}(D_{N}) ; this roughly means that a random walk starting from far away starts to see through ℐ ( u N ) ∩ D N \mathscr{I}(u_{N})\cap D_{N} . In the same spirit, we investigate the capacity of the simple random walk conditioned to stay in a large Euclidean ball up to time t N t_{N} , and find similar asymptotics by taking t N = u N N d t_{N}=u_{N}N^{d} . Keywords: capacity, random interlacement, conditioned walk 2020 Mathematics subject classification: Primary 60G50, 60K35. 1 Introduction The random interlacement at level u>0 is a random subset \mathscr{I}(u)\subset\mathbb{Z}^{d} , d\geq 3 , which was introduced by Sznitman in [ 26 ] , in order to study the trace of the simple random walk on the torus \mathbb{T}_{n}\vcentcolon=\mathbb{Z}^{d}/n\mathbb{Z}^{d} . It can be thought as a Poissonian collection of bi-infinite simple random walk trajectories on \mathbb{Z}^{d} (that are transient since d\geq 3 ), with an intensity parameter u>0 . We refer to [ 15 ] for a detailed introduction and overview. In the following, we denote by \mathbf{P}_{x} the law of a simple random walk (S_{n})_{n\geq 0} on \mathbb{Z}^{d} , with starting point x\in\mathbb{Z}^{d} . We also simply write \mathbf{P}=\mathbf{P}_{0} . We also denote K\Subset\mathbb{Z}^{d} to say that K is a finite subset of \mathbb{Z}^{d} . Then, a characterization of the law \mathbb{P} of the random interlacement (\mathscr{I}(u))_{u>0} is given by the following relation \forall K\Subset\mathbb{Z}^{d}\,,\quad\mathbb{P}\big(\mathscr{I}(u)\cap K=\varnothing\big)=\exp\big(-u\,\mathrm{cap}\left(K\right)\big)\,. (1.1) Here, \mathrm{cap}\left(K\right) denotes the capacity of the set K in \mathbb{Z}^{d} , defined as \mathrm{cap}\left(K\right)\vcentcolon=\sum_{x\in K}\mathbf{P}_{x}\big(H_{K}=+\infty\big)\,, (1.2) with H_{A}\vcentcolon=\min\{i\geq 1,S_{i}\in A\} the first return time of the set A ; by convention \min\varnothing=+\infty . The capacity is a non-decreasing function which characterizes the size of a subset seen by a (simple) random walk coming from far away. In this paper, we are interested in how the capacity of a set compares with the capacity of the trace of random walk(s) on this set. 1.1 Our main question We consider a bounded open set D\subset\mathbb{R}^{d} and we let D_{N}\vcentcolon=(N\cdot D)\cap\mathbb{Z}^{d} be its (discrete) blow up of size N . We are interested in \mathscr{I}(u)\cap D_{N} the trace of the random interlacement inside D_{N} , and more precisely in the ratio of their capacities \varsigma^{\mathrm{RI}}_{N,u}=\varsigma^{\mathrm{RI}}_{N,u}(D)\vcentcolon=\frac{\mathrm{cap}\left(\mathscr{I}(u)\cap D_{N}\right)}{\mathrm{cap}\left(D_{N}\right)}\,, (1.3) in the regime n\to\infty and u\downarrow 0 . Notice that since \mathscr{I}(u)\cap D_{N}\subseteq D_{N} , we have that \varsigma_{n,u}\in[0,1] . In particular, we are looking for a condition on the sequence (u_{N})_{n\geq 0} for the ratio \varsigma_{n} to converge either to 1 or 0 , in \mathbb{P} -probability. Let us define \Theta_{N}\vcentcolon=\begin{cases}N&\text{ if }d=3\,,\\ N^{2}/\log N&\text{ if }d=4\,,\\ N^{2}&\text{ if }d\geq 5\,.\end{cases} (1.4) The motivation behind this definition ( 1.4 ) of \Theta_{N} is so that \Theta_{N} and \mathbf{E}\big[\mathrm{cap}(\mathcal{R}_{N^{2}})\big] are of the same order (see Theorem 2.2 below). We then have the following criterion. Theorem 1.1 . Let d\geq 3 and let (u_{N})_{N\geq 1} be a sequence of positive numbers. For any compact regular set D\subset\mathbb{R}^{d} , we have the following convergences in \mathbb{P} -probability (and in L^{1} ): \varsigma^{\mathrm{RI}}_{N,u_{N}}=\varsigma_{N,u_{N}}^{\mathrm{RI}}(D)\vcentcolon=\frac{\mathrm{cap}\left(\mathscr{I}(u_{N})\cap D_{N}\right)}{\mathrm{cap}\left(D_{N}\right)}\xrightarrow[N\to+\infty]{}\begin{cases}0&\quad\text{ if }\ \lim_{N\to\infty}u_{N}\Theta_{N}=0\,,\\ 1&\quad\text{ if }\ \lim_{N\to\infty}u_{N}\Theta_{N}=+\infty\,.\end{cases} (1.5) Moreover, if 0<\liminf_{N\to\infty}u_{N}\Theta_{N}\leq\limsup_{N\to\infty}u_{N}\Theta_{N}<+\infty , then 0<\liminf_{N\to\infty}\mathbb{E}[\varsigma^{\mathrm{RI}}_{N,u_{N}}]\leq\limsup_{N\to\infty}\mathbb{E}[\varsigma^{\mathrm{RI}}_{N,u_{N}}]<1\,. (1.6) Remark 1.2 . In Theorem 1.1 , the assumption that D is regular means that \mathrm{cap}_{\mathbb{R}}\left(D\right)=\mathrm{cap}_{\mathbb{R}}\left(\mathrm{Int}(D)\right) , where \mathrm{cap}_{\mathbb{R}}\left(\,\cdot\,\right) denotes the Newtonian capacity on \mathbb{R}^{d} (see [ 24 ] for the necessary definitions). It is required to compare \mathrm{cap}_{\mathbb{R}}\left(D\right) with \mathrm{cap}\left(D_{N}\right) (see Proposition 2.3 below). Remark 1.3 . In the case where \lim_{N\to\infty}u_{N}\Theta_{N}=a\in(0,+\infty) , we expect that the ratio \varsigma_{N,u_{N}} converges in distribution to a random variable \varsigma_{a}\in(0,1) . A first step towards this would be to prove the following tightness-type result: for any \varepsilon>0 there is some \eta>0 such that \mathbb{P}\big(\eta<\varsigma^{\mathrm{RI}}_{N,u_{N}}<1-\eta\big)\geq 1-\varepsilon\,. (1.7) The bounds ( 1.6 ) only shows that for sufficiently small \eta>0 , the probability \mathbb{P}\big(\eta<\varsigma_{N}<1-\eta\big) remains bounded away from 0 . Remark 1.4 . Let us notice that the parameter u is related to the density (with respect to the volume) of random interlacements in \mathbb{Z}^{d} . Indeed, for any fixed u , we have that \lim_{N\to\infty}\frac{|\mathscr{I}(u)\cap D_{N}|}{|D_{N}|}=1-e^{-u/g_{0}}=\vcentcolon\vartheta(u)\,, (1.8) with g_{0}\vcentcolon=\sum_{n=0}^{\infty}\mathbf{P}_{0}(S_{n}=0) . In particular, when u_{N}\downarrow 0 , the random interlacement \mathscr{I}(u_{N}) is very sparse and has large holes, but the capacity \mathrm{cap}(\mathscr{I}(u_{N})\cap D_{N}) may remains equivalent to \mathrm{cap}(D_{N}) , meaning that the holes are not “visible” for a random walk arriving from far away. In that sense, Theorem 2.2 gives a threshold on (u_{N})_{n\geq 1} for these holes to become “visible”. About the capacity of a random walk conditioned to stay in a ball Random interlacements can be interpreted as an infinite volume measure for the simple random walk on the torus \mathbb{T}^{d}_{N}=(\mathbb{Z}/N\mathbb{Z})^{d} . The link between these two objects has been a powerful tool to tackle problems involving this random walk on the torus (see [ 29 , 4 , 25 ] ) as well as large-deviation events for the simple random walk on \mathbb{Z}^{d} (see [ 21 , 10 ] ). The coupling between the two objects presented in [ 8 ] (and previous works) relies on a heuristical argument, which roughly explains that given a subset K of the torus, the random walk may visit K and then stay in \mathbb{T}^{d}_{N}\setminus K a time larger than its mixing time. The eventual return to K would then be independent from the last times the random walk was in K , and the number of such returns is well-approximated by a Poisson random variable. Consider the discrete Euclidean ball B_{N} with radius N\geq 1 and a sequence (t_{N})_{N\geq 1} , and write \mathcal{R}_{t_{N}}=\left\{S_{0},\,\dots,S_{t_{N}}\right\} the range up to time t_{N} of the simple random walk on \mathbb{Z}^{d} . Under \mathbf{P}_{0}(\cdot\,|\,\mathcal{R}_{t_{N}}\subseteq B_{N}) , the random walk behaves in a way such that the heuristical argument still has some truth (see [ 7 ] for a more rigorous statement). By comparison to random interlacements, the range under \mathbf{P}_{0}(\cdot\,|\,\mathcal{R}_{t_{N}}\subseteq B_{N}) can be expected to resemble a random interlacement with parameter u_{N}\sim ct_{N}N^{-d} . Therefore, we can conjecture an analogue to Theorem 1.1 for \mathcal{R}_{t_{N}} under \mathbf{P}_{0}(\cdot\,|\,\mathcal{R}_{t_{N}}\subseteq B_{N}) , by plugging u_{N}=t_{N}N^{-d} . We prove this result in Section 3 , although our method does not rely on any coupling. Theorem