Phase transitions for contact processes on one-dimensional networks
Benedikt Jahnel, Lukas Lüchtrath, Christian Mönch · 2025
Read Phase transitions for contact processes on one-dimensional networks by Benedikt Jahnel, Lukas Lüchtrath, Christian Mönch (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-2501-16858.txt.
Introduction
Benedikt Jahnel, Lukas Luchtrath, and Christian Monch study when an infection-like process can die out on a one-dimensional random network, in work from 2025. The network starts from the integer line and adds long-range links at random. Contact processes on such networks can either survive for good or fade away, with a sharp threshold between the two.
The authors give conditions that guarantee a fading zone exists, so the threshold is real and not trivial. Their results cover scale-free geometric random graphs with any well-behaved degree pattern, a wider class than past methods could handle. For one classic graph type, their condition is exact. For long-range percolation on the integers, they confirm a standing conjecture about when the threshold is nontrivial. The work widens the map of survival thresholds on random networks.
Catalog record
| CREATOR | Benedikt Jahnel, Lukas Lüchtrath, Christian Mönch |
|---|---|
| YEAR | 2025 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 67 KB |
| VERSION | 1.0 |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2501-16858
Browse the full library · About the Tower of Babel
Full text
The opening of the work, from the library's hosted text:
Phase transitions for contact processes on one-dimensional networks Benedikt Jahnel, Lukas Lüchtrath, Christian Mönch Abstract We study the survival/extinction phase transition for contact processes with quenched disorder. The disorder is given by a locally finite random graph with vertices indexed by ℤ \mathbb{Z} that is assumed to be invariant under index shifts and augments the nearest-neighbour lattice by additional long-range edges. We provide sufficient conditions that imply the existence of a subcritical phase and therefore the non-triviality of the phase transition. Our results apply to instances of scale-free random geometric graphs with any integrable degree distribution. The present work complements previously developed techniques to establish the existence of a subcritical phase on Poisson–Gilbert graphs and Poisson–Delaunay triangulations (Ménard et al., Ann. Sci. Éc. Norm. Supér. , 2016), on Galton–Watson trees (Bhamidi et al., Ann. Probab. , 2021) and on locally tree-like random graphs (Nam et al., Trans. Am. Math. Soc. , 2022), all of which require exponential decay of the degree distribution. Two applications of our approach are particularly noteworthy: First, for Gilbert graphs derived from stationary point processes on ℝ \mathbb{R} marked with i.i.d. random radii, our results are sharp. We show that there is a non-trivial phase transition if and only if the graph is locally finite. Second, for independent Bernoulli long-range percolation on ℤ \mathbb{Z} , with coupling constants J x , y ≍ | x − y | − δ J_{x,y}\asymp|x-y|^{-\delta} , we verify a conjecture of Can ( Electron. Commun. Probab. , 2015) stating the non-triviality of the phase transition whenever δ > 2 \delta>2 . Although our approach utilises the restrictive topology of the line, we believe that the results are indicative of the behaviour of contact processes on spatial random graphs also in dimensions d > 1 d>1 as long as the degree distribution of the underlying network has at least finite d d -th moment. We support this by proving that no phase transition exists if the d d -the moment is infinite. AMS-MSC 2020 : 60K35 (primary); 05C82, 91D30 (secondary) Key Words : contact process, Gilbert graph, long-range percolation, continuum percolation, random connection model, scale-free network, SIS epidemics 1 Background and motivation We investigate contact processes on spatial random graphs. Let us briefly introduce the contact process and the question of its global survival/extinction, both on fixed and on random graphs. 1.1 Survival and extinction of contact processes The contact process (sometimes also referred to as the SIS epidemic model) is a Markovian interacting particle system, which was first introduced to the mathematical literature by Harris [ 25 ] and has been studied extensively on hypercubic lattices throughout the 1970s, 1980s and early 1990s. An introduction to this classical theory as well as an overview of its main results is provided in the monographs [ 39 , 38 ] . In the 2000s renewed interest in the contact process emerged in the context of complex networks, and in particular researchers began to investigate the behaviour of the process on large but finite random graphs as a toy model for the spread of disease or information in a variety of mesoscopic systems, see, for example, [ 14 ] for an account of rigorous results in this area. Together with percolation and oriented percolation, the contact process is one of the simplest random interacting systems that may display a non-trivial phase transition, the existence of which is the subject of the present paper. More precisely, let G be a locally finite connected graph with vertex set V(G) and edge set E(G)\subset V(G)^{[2]}:=\{\{x,y\}\colon x\in V(G),y\in V(G),x\neq y\}. In the contact process on G , infected vertices in V(G) infect their susceptible neighbours at exponential rate \lambda>0 and turn back into susceptible vertices at rate 1 . Formally, we view this process as a Markov process with values in the set of subsets of V(G) and generator L given by Lf(A)=\lambda\sum_{\{u,v\}\in E(G)}\mathbbm{1}\{u\in A,v\notin A\}\big(f(A\cup\{v\})-f(A)\big)+\sum_{v\in V(G)}\mathbbm{1}\{v\in A\}\big(f(A\setminus\{v\})-f(A)\big), where \lambda>0 is called the infection rate and f is any non-negative bounded function on V(G) . Starting at time 0 with an initially infected set A_{0}\subset V(G) , we denote by \xi^{A_{0}}_{t}\subset V(G) the set of infected vertices at time t under the dynamics prescribed by L . The law of \xi is denoted by \mathbb{P}_{\lambda} . For the purpose of this paper, we always consider rooted graphs (G,o) , i.e., we distinguish a specific vertex o\in V(G) as the root of G . We usually set A_{0}=\{o\} and omit the initially infected set from the notation in this case. We say that the contact process (with infection rate \lambda ) survives on G if \mathbb{P}_{\lambda}(\xi_{t}\neq\emptyset\text{ for all }t>0)>0 and define the corresponding critical infection rate by \lambda_{\mathsf{c}}(G):=\inf\big\{\lambda>0\colon\mathbb{P}_{\lambda}(\xi_{t}\neq\emptyset\text{ for all }t>0)>0\big\}. If \lambda_{\mathsf{c}}(G)\in(0,\infty) , then the survival/extinction phase transition for \xi on G is called non-trivial . Since \emptyset is an absorbing state of the dynamics, it follows immediately that \lambda_{\mathsf{c}}(G)=\infty for any finite graph G . Conversely, if G is assumed to be infinite we have \lambda_{\mathsf{c}}(G)\leq\lambda_{\mathsf{c}}(\mathbb{Z}_{\textup{nn}})<\infty, (1) where \mathbb{Z}_{\textup{nn}} denotes the nearest-neighbour graph on \mathbb{Z} , which we always consider as rooted at 0 . The second inequality in ( 1 ) is due to Harris’ foundational work [ 25 ] and the first inequality follows from the fact that \lambda_{\mathsf{c}}(\cdot) is monotone with respect to the canonical partial order on rooted graphs. The non-triviality of the survival/extinction phase transition thus reduces to the question whether \lambda_{\mathsf{c}}(G)>0 , in which case we say that the contact process on G has a subcritical phase . A simple comparison with a branching process shows that \lambda_{\mathsf{c}}(G)>0 whenever the vertices of G have a uniformly bounded degree. At the other end of the spectrum, it is not difficult to construct graphs with unbounded degrees on which \xi survives for any \lambda>0 , for instance by considering a sequence of sufficiently rapidly growing star-graphs whose centre vertices are placed on a copy of the line graph \mathbb{Z}_{\textup{nn}} . Moreover, if we interpret the breadth-first exploration of a Bernoulli bond percolation cluster as a contact-type process in which every vertex can only be infected once, then an elementary coupling argument shows that any graph G with bond percolation threshold p_{\mathsf{c}}(G)=0 also satisfies \lambda_{\mathsf{c}}(G)=0 . 1.2 Contact processes on random graphs Let us now consider a random rooted graph (\mathscr{G},o) with distribution \mathbf{P} . We mostly consider cases in which \mathscr{G} is infinite but locally finite \mathbf{P} -almost surely 1 1 1 If \mathscr{G} is not locally finite, the contact process can still be defined in the way presented until it infects a vertex of infinite degree. If \mathscr{G} is also connected, this explosion event always has positive probability and we consequently have \lambda_{\mathsf{c}}(\mathscr{G})=0 . . Furthermore, we assume \mathbf{P} to be invariant and ergodic with respect to shifts of the root, e.g., see [ 3 , 33 ] for discussions of the concept. Note that these assumptions are naturally satisfied in many cases of interest, for instance whenever \mathscr{G} arises as a local limit of a sequence of finite graphs in which the degree of the root is uniformly integrable [ 2 ] , or if \mathscr{G} is the infinite cluster of some translation-invariant percolation model on the hypercubic lattice. We