Homogenization theory of random walks in degenerate random environment

Sebastian Andres · 2025

Read Homogenization theory of random walks in degenerate random environment by Sebastian Andres (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-2504-06690.txt.

Introduction

Sebastian Andres reviews recent progress on the Random Conductance Model, which describes a walker moving on a network whose links carry random strengths. The focus is homogenization: showing that, seen from far away, the messy random walk behaves like a clean, predictable diffusion.

The review stresses results that hold for almost every layout of the environment. These include central limit theorems describing the walk's spread, local versions that pin down probabilities at single points, and bounds on heat flow. The walk's link strengths may be unbounded, but they must meet certain average conditions. The piece surveys how far the theory has come for these degenerate environments, where some links can be arbitrarily weak.

Catalog record

CREATORSebastian Andres
YEAR2025
COLLECTIONPAPERS
FORMATTXT
FILE SIZE84 KB
VERSION1.0
LICENSEhttp://creativecommons.org/licenses/by/4.0/
RIGHTSCC BY
SOURCEarXiv

Source: https://arxiv.org/abs/2504-06690

Browse the full library · About the Tower of Babel

Full text

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

Homogenization theory of random walks in degenerate random environment Sebastian Andres Abstract . Recent progress on the understanding of the Random Conductance Model is reviewed. A particular emphasis is on homogenization results such as functional central limit theorems, local limit theorems and heat kernel estimates for almost every realization of the environment, established for random walks among stationary ergodic conductances that are possibly unbounded but satisfy certain moment conditions. 2000 Mathematics Subject Classification Primary 60K37, 60F17; secondary 80M40, 82B43 1. Introduction 1.1. Background One main motivation for the study of random walks in random environment is to explore the relationships between the microscopic properties of a disordered medium and its large scale conductivity. Examples include heat conduction in porous media or composite materials, electrical conducting properties in metals with impurities, to name just a few. One particular question of interest is whether homogenization occurs, in the sense that on macroscopic scales the irregular structures in the random medium are ‘smoothed out’ and only an averaged effect remains; the alternative being that an effect of the irregularities persists even at macroscopic scales leading to an anomalous behaviour. However, the relation between the microscopic properties of the medium and its effective ones is subtle and highly sensitive to the geometrical, spatial and statistical properties of the underlying medium. So it is natural to ask what kind of statistical properties of the random medium ensure a homogenization of the rescaled random walker on macroscopic scales. The mathematical rigorous analysis of random walks in random environments has been at the centre of interest in probability theory since the early 1980s, see e.g. the reviews [ 117 , 141 , 144 ] . During the last decades there has been a major research effort focused on the derivation of heat kernel estimates and functional central limit theorems (a.k.a. invariance principles), i.e. the convergence of the random walk towards Brownian motion under the diffusive rescaling of space and time. In this article we will review such homogenization results by means of the random conductance model, one the most prominent models for a random walk in random environment. On the other hand, depending on the properties of the environment, sub-diffusive behaviour may occur, which is typical for diffusion processes on fractal spaces, see e.g. [ 25 , 125 ] . One manifestation of such irregular behaviour is the presence of heat kernel fluctuations, appearing in many instances of stochastic processes on fractals and in random media. In particular, for processes in low dimensions and models at criticality, heat kernel fluctuations are known to occur, caused by local irregularities in the random medium, see [ 6 ] for a recent review on this topic. 1.2. The model The random conductance model (RCM) is a particular class of reversible random walks in random environments defined as follows. For d\geq 2 consider the Euclidean lattice (\mathbb{Z}^{d},E_{d}) , where E_{d}=\{e=\{x,y\}:x,y\in\mathbb{Z}^{d},|x-y|=1\} denotes the set of non-oriented nearest neighbour edges. We call two vertices x,y\in\mathbb{Z}^{d} adjacent if \{x,y\}\in E_{d} , and we then write x\sim y . Each edge \{x,y\}\in E_{d} is endowed with a non-negative weight \omega(\{x,y\}) , also called conductance. We write \omega(x,y)=\omega(\{x,y\})=\omega(y,x) , and \omega(x,y)=0 if \{x,y\}\not\in E_{d} . We are particularly interested in the situation when the collection \{\omega(e):e\in E_{d}\} is itself random and distributed according to a stationary and ergodic law \prob on the measurable space (\Omega,\mathcal{F})\mathrel{\mathop{:}}=([0,\infty)^{E_{d}},\mathcal{B}([0,\infty))^{\otimes E_{d}}) equipped with the Borel- \sigma -algebra. For a fixed \omega\in\Omega set \displaystyle\mu^{\omega}(x)\mathrel{\mathop{:}}=\sum_{y\sim x}\omega(x,y),\qquad p^{\omega}(x,y)\mathrel{\mathop{:}}=\frac{\omega(x,y)}{\mu^{\omega}(x)}. We endow (\Omega,\mathcal{F}) with the d -parameter group of space shift (\tau_{x})_{x\in\mathbb{Z}^{d}} acting on \Omega as \displaystyle(\tau_{z}\,\omega)(\{x,y\})\;\mathrel{\mathop{:}}=\;\omega(\{x+z,y+z\}),\qquad\forall\,\{x,y\}\in E_{d}. We will study continuous time random walks on \mathbb{Z}^{d} which jump according to the probabilities p^{\omega}(x,y) . There are two natural choices of this. The first one is called variable speed random walk (VSRW) X=(X_{t}\!:t\geq 0) , which waits at x an exponential time with mean 1/\mu^{\omega}(x) , with generator \big(\mathcal{L}_{X}^{\omega}f)(x)=\sum_{y\sim x}\omega(x,y)\,\left(f(y)-f(x)\right). It is a reversible Markov chain with symmetrising measure given by the counting measure. We denote by \Prob_{x}^{\omega} the law of the process starting at the vertex x\in\mathbb{Z}^{d} . The corresponding expectation will be denoted by \Mean_{x}^{\omega} . The second walk, called the constant speed random walk (CSRW), denoted by Y=(Y_{t}\!:t\geq 0) waits at x for an exponential time with mean 1 , with generator \displaystyle\big(\mathcal{L}_{Y}^{\omega}\,f\big)(x)\;=\;\frac{1}{\mu^{\omega}(x)}\sum_{y\sim x}\omega(x,y)\,\big(f(y)-f(x)\big). If \mu^{\omega}(x)=0 we write \mathcal{L}_{X}f(x)=\mathcal{L}_{Y}f(x)=0 . The CSRW is reversible w.r.t. the measure \mu^{\omega} . Note that X and Y are time-changes of each other. The CSRW is mainly of probabilistic interest due to its resemblence to the corresponding random walk in discrete time. On the other hand, VSRWs and their generators are objects of study in stochastic homogenization in PDE theory, and they are also relevant due to their appearance in statistical mechanics models. We denote by p^{\omega}(t,x,y) and q^{\omega}(t,x,y) , x,y\in\mathbb{Z}^{d} , t\geq 0 , the transition density with respect to the reversible measure (or heat kernels ) of the VSRW and CSRW, respectively, i.e. \displaystyle p^{\omega}(t,x,y)\mathrel{\mathop{:}}=\Prob^{\omega}_{x}(X_{t}=y),\qquad q^{\omega}(t,x,y)\mathrel{\mathop{:}}=\frac{\Prob_{x}^{\omega}\big(Y_{t}=y\big)}{\mu^{\omega}(y)}. As an important example for an environment, consider the case when the conductances are independent and identically distributed (i.i.d.). Note that X or Y never jump across e whenever \omega(e)=0 . So if \prob(\omega(e)>0) is less than p_{c}\in(0,1) , the critical threshold for Bernoulli bond percolation on \mathbb{Z}^{d} , then both X and Y are \prob -a.s. confined to a finite set. Thus it is very natural to assume that \prob(\omega(e)>0)>p_{c} . Then, it is well known that there exists a \prob -a.s. unique infinite cluster \mathcal{C}_{\infty} of connected open edges. We denote by \prob_{0}[\,\cdot\,]\mathrel{\mathop{:}}=\prob\bigl[\,\cdot\,|\,0\in\mathcal{C}_{\infty}\bigr] the conditional distribution given the event \{0\in\mathcal{C}_{\infty}\} . 1.3. Questions and aims In the late 1980s, the pioneering works [ 87 ] and [ 120 ] already proved that a functional central limit theorem holds under the average , or annealed , law (i.e. the joint law of the environment and the process). The results also imply an ‘invariance principle in probability’. On the other hand, much effort in the past 20 years has been invested in the derivation of the following quenched statements for the random walks and their heat kernel, holding for almost all realisations of the environment (for simplicity only stated formally for the VSRW here). • Quenched functional central limit theorem (QFCLT). For almost all realisations of the environment, the rescaled random walk (n^{-1}X_{n^{2}t})_{t\geq 0} converges in law towards a Brownian motion on \mathbb{R}^{d} with deterministic covariance matrix \Sigma^{2} . • Quenched local central limit theorem. For almost all realisations of the environment, the transition kernel of the random walk converges under suitable rescaling towards the Gaussian transition kernel of Brownian motion with covariance matrix \Sigma^{2} , i.e. n^{d}p_{n^{2}t}^{\omega}(0,\lfloor nx\rfloor)\underset{n\to\infty}{\longrightarrow}\bar{p}^{\Sigma}(t,0,x), uniformly over compact sets, where (1.1) \displaystyle\bar{p}^{\Sigma}(t,x,y)\;\mathrel{\mathop{:}}=\;\frac{1}{\sqrt{(2\pi t)^{d}\det\Sigma^{2}}}\exp\!\left(-\frac{(x-y)\cdot(\Sigma^{2})^{-1}(x-y)}{2t}\right), and for any x\in\mathbb{R}^{d} we write \lfloor nx\rfloor for a closest point to nx in the underlying graph w.r.t. the |\cdot|_{\infty} -norm. • Gaussian heat kernel estimates. For almost all realisations of the environment, we have p^{\omega}(t,x,y)\leq c\,t^{-d/2}\exp\big(-c|x-y|^{2}/t\big) , and a similar lower bound, for sufficiently large t and |x-y| . Since the heat kernel is a fundamental solution of the heat equation \partial_{t}u-\mathcal{L}_{X}^{\omega}u=0 , a closely related, purely analytic statement is the parabolic Harnack inequality. In fact, since the 1960s Harnack inequalities have had a big influence on PDE theory and

Read the complete text