The first positive position of a lattice random walk
Claude Godrèche, Jean-Marc Luck · 2025
Read The first positive position of a lattice random walk by Claude Godrèche, Jean-Marc Luck (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-2510-17329.txt.
Introduction
Where a wandering walker first steps into positive ground is a basic question in the study of random walks. Claude Godrèche and Jean-Marc Luck give a detailed study of that first positive position for walks on a line of equally spaced points. The walker starts at the origin, moves symmetrically, and each step lands on one of finitely many whole number offsets.
Godrèche and Luck work out the full distribution of that first positive landing spot in closed analytical form. Their self contained account ties the result to the wider study of extremes and records in one dimensional walks. Readers get exact formulas rather than simulations, which makes the paper a reference for checking approximations in related problems.
Catalog record
| CREATOR | Claude Godrèche, Jean-Marc Luck |
|---|---|
| YEAR | 2025 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 49 KB |
| VERSION | 1.0 |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2510-17329
Browse the full library · About the Tower of Babel
Full text
The opening of the work, from the library's hosted text:
The first positive position of a lattice random walk Claude Godrèche, Jean-Marc Luck Abstract The distribution of the first positive position reached by a random walker starting at the origin is central to the analysis of extremes and records in one-dimensional random walks. In this work, we present a detailed and self-contained analytical study of this distribution for symmetric finite-range lattice walks, whose steps are drawn from a distribution supported on finitely many integers. , 1 Introduction We consider a random walker on the one-dimensional lattice, whose integer position at discrete time n is given by x_{n}=x_{n-1}+\eta_{n}. (1.1) The walker is launched from the origin ( x_{0}=0 ). The steps \eta_{n} are independent, identically distributed (iid) integer random variables drawn from an even distribution supported on finitely many integers: \mathbb{P}(\eta_{n}=k)=\rho_{k}\qquad(k=-K,\dots,K), (1.2) with \rho_{k}=\rho_{-k} . The integer K\geq 1 is called the range of the step distribution. We assume that \rho_{1} is non-zero, i.e., that steps with unit length are allowed. This hypothesis will allow us to discard non-generic behavior. The probability \rho_{0} for the walker to stop at every time step may either be zero or not. The main subject of this paper is the distribution of the first positive position of the walker, denoted by H . This distribution is fundamental for understanding the statistics of successive record positions, or ‘ladder heights’, in one-dimensional random walks 1 1 1 See [ 1 ] for a review on records for random walks and Lévy flights. . The present work is a direct continuation of our recent comprehensive investigation of the same quantity in continuous symmetric random walks [ 2 ] . The latter provides a detailed account of the subject, including its history, and sets key observables in perspective with one another. Most of the bibliographic references cited therein will not be repeated here. In the present situation, namely symmetric finite-range lattice walks, the Wiener-Hopf equations governing the problem can be solved analytically, without having to resort to the general formalism. The essence of the method is to factorise the relevant analytic functions over their complex zeros and poles. This approach was put forward long ago by Wick [ 3 ] and Chandrasekhar [ 4 ] for Wiener-Hopf equations of the kind encountered in continuous random walks (see [ 5 ] for a review). In that setting, the method works whenever the step distribution is a finite superposition of decaying exponentials, so that its Laplace transform is a rational function. The possibility of applying a similar approach for finite-range lattice walks is mentioned in Feller’s book [ 6 , Ch. XII] . The method has been rediscovered and given the name ‘algebraic trick’ in a study of the crossings of a biased simple random walker with a ballistic obstacle moving at rational velocity [ 7 ] . In continuity with the above, the case of general symmetric finite-range lattice walks is investigated in the present work, with the following setup. Section 2 presents a self-contained derivation of explicit expressions for all fundamental quantities of interest, including the distribution of the first positive position of the walker (section 2.5 ) and its moments (section 2.6 ). In section 3 we examine the simplest examples of walks whose range is either 1 or 2. The case of a discrete uniform step distribution with an arbitrary range K is then studied in section 4 . A special emphasis is put on the regime where the range becomes large. The asymptotic expressions of various quantities are compared with their counterparts for continuous random walks with a symmetric uniform step length distribution. Section 5 contains a discussion in which analogies and differences between continuous and discrete walks are highlighted, and the findings of the present paper are set in perspective with those of earlier works [ 2 , 8 ] . Finally, two appendices are devoted respectively to the detailed derivations of selected key results and to integer renewal processes. 2 Fundamental observables In this section we present a self-contained derivation of general results concerning the fundamental quantities, valid for arbitrary step distributions of the form ( 1.2 ). 2.1 Preliminaries The generating series of the step distribution, \hat{\rho}(z)={\langle z^{\eta}\rangle}=\sum_{k}\rho_{k}z^{k}=\frac{\rho_{K}}{z^{K}}+\cdots+\rho_{K}z^{K}, (2.1) is a self-reciprocal (or palindromic 2 2 2 The term palindromic refers to the fact that the sequence of coefficients reads the same forwards and backwards, just like a palindrome word such as radar . ) Laurent polynomial in z , obeying \hat{\rho}(1/z)=\hat{\rho}(z) . The factorisation of the difference \phi(z)=1-\hat{\rho}(z) (2.2) over its complex zeros plays a central role hereafter. We have \phi(z)=-D(z-1)^{2}+\cdots (2.3) as z\to 1 , where D=\frac{{\langle\eta^{2}\rangle}}{2}=\frac{1}{2}\sum_{k}k^{2}\rho_{k} (2.4) is the diffusion coefficient of the walk. Besides the above double zero at z=1 , \phi(z) has no other zero on the unit circle. Setting z=\rme^{{\rm i}\alpha} , we have indeed \phi(\rme^{{\rm i}\alpha})=\sum_{k}\rho_{k}(1-\cos k\alpha)\geq 2\rho_{1}(1-\cos\alpha). (2.5) The factorisation of the Laurent polynomial \phi(z) therefore reads \displaystyle\phi(z) \displaystyle= \displaystyle-\frac{\rho_{K}}{z^{K}}(z-1)^{2}\prod_{a}(z+z_{a})(z+1/z_{a}) (2.6) \displaystyle= \displaystyle\rho_{K}(1-z)(1-1/z)\prod_{a}\frac{(z_{a}+z)(z_{a}+1/z)}{z_{a}}. (2.7) The second form ( 2.7 ) makes the invariance under the change of z into 1/z manifest. Here and throughout the following, the label a runs over 1,\dots,K-1 . The K-1 complex numbers z_{a} are inside the unit circle, whereas their inverses 1/z_{a} are outside the unit circle. The zeros of \phi(z) are therefore 1 (double), as well as -z_{a} and -1/z_{a} . Henceforth, for the sake of brevity, we simply refer to the z_{a} as being ‘the zeros’. They come in complex conjugate pairs. For K even, there is also a real zero such that 0<z_{a}<1<1/z_{a} . In generic circumstances, as the range K is increased, the zeros come closer to the unit circle and become more evenly spread. Plots of these zeros will be shown in section 4 (see figure 3 ) for discrete uniform step distributions. The leitmotiv of the present approach is that all quantities of interest can be expressed as symmetric functions of the complex zeros z_{a} . A first example, D=\rho_{K}\prod_{a}(1+z_{a})(1+1/z_{a}), (2.8) is obtained by comparing ( 2.3 ) and ( 2.7 ). 2.2 Enhancement factor The enhancement factor of the walk is defined by {\cal E}=\exp\biggl(\frac{1}{2}\sum_{n\geq 1}\frac{1}{n}\,\mathbb{P}(x_{n}=0)\biggr). (2.9) In the probabilistic literature, and especially in the classic books [ 6 , 9 ] , 1/{\cal E}^{2} is denoted by c , whereas D is denoted by \sigma^{2}/2 . For a random walk with a continuous step distribution, \mathbb{P}(x_{n}=0) vanishes, and so {\cal E}=1 . If the step distribution is discrete, like e.g. for the lattice walks considered here, \mathbb{P}(x_{n}=0) may be non-zero, at least for some n , and so {\cal E}\geq 1 . The enhancement factor {\cal E} therefore measures to what extent a given lattice walk is different from the class of continuous ones. This quantity enters the expressions of many observables. To take a characteristic example [ 10 , 11 ] , we recall that the mean maximum of a long random walk of n steps takes the universal asymptotic form {\langle M_{n}\rangle}\approx 2\sqrt{\frac{Dn}{\pi}}, (2.10) depending only on the diffusion coefficient of the walk, whereas the mean number of records, or ‘ladder heights’, of the same walk grows as {\langle R_{n}\rangle}\approx\frac{2}{{\cal E}}\sqrt{\frac{n}{\pi}}. (2.11) We have furthermore [ 10 , 11 ] M_{n}\approx{\langle H\rangle}R_{n}. (2.12) The ratio between the estimates ( 2.10