Phase transitions for contact processes on sparse random graphs via metastability and local limits
Benedikt Jahnel, Lukas Lüchtrath, Christian Mönch · 2025
Read Phase transitions for contact processes on sparse random graphs via metastability and local limits 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-2505-22471.txt.
Introduction
Benedikt Jahnel, Lukas Lüchtrath, and Christian Mönch offer a new view of phase transitions in the contact process on sparse random networks. This process models spread, like infection, where active sites can recover. On sparse graphs, the key question is when spread dies out fast versus lingering for a very long time.
The authors tie the boundary between regimes to the presence of a stable middle density at which the process can hover. That framing suits tools based on local views of large graphs. They give general conditions for when this boundary matches the survival threshold of the limiting tree-like graph. They also argue that exponential time is the right scale for the split.
Catalog record
| CREATOR | Benedikt Jahnel, Lukas Lüchtrath, Christian Mönch |
|---|---|
| YEAR | 2025 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 58 KB |
| VERSION | 1.0 |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2505-22471
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 sparse random graphs via metastability and local limits Benedikt Jahnel, Lukas Lüchtrath, Christian Mönch Abstract We propose a new perspective on the asymptotic regimes of fast and slow extinction in the contact process on locally converging sequences of sparse finite graphs. We characterise the phase boundary by the existence of a metastable density, which makes the study of the phase transition particularly amenable to local-convergence techniques. We use this approach to derive general conditions for the coincidence of the critical threshold with the survival/extinction threshold in the local limit. We further argue that the correct time scale to separate fast extinction from slow extinction in sparse graphs is, in general, the exponential scale, by showing that fast extinction may occur on stretched exponential time scales in sparse scale-free spatial networks. Together with recent results by Nam, Nguyen and Sly (Trans. Am. Math. Soc. 375, 2022), our methods can be applied to deduce that the fast/slow threshold in sparse configuration models coincides with the survival/extinction threshold on the limiting Galton-Watson tree. AMS-MSC 2020 : 60K35 (primary); 05C82, 91D30 (secondary) Key Words : Local weak convergence, random network, SIS model, sparse random graph, locality of phase transition 1 Introduction and main results The contact process. Let G=(V,E) denote a locally finite connected graph with a dedicated root o\in V and we write (G,o) for the rooted graph. The contact process with infection rate \lambda>0 on G is the family of set-valued continuous-time Markov processes \big\{\xi^{A}=(\xi^{A}_{t})_{t\geq 0}\colon A\subset V\text{ finite }\big\} . The law of \xi^{A} is determined by setting \xi^{A}_{0}=A and the transition dynamics, for every v\in V , \displaystyle\xi^{A}_{t}\to\xi^{A}_{t}\setminus\{v\}\text{ at rate }1, (1) \displaystyle\xi^{A}_{t}\to\xi^{A}_{t}\cup\{v\}\text{ at rate }\lambda\sum_{w\colon v\sim w}\mathbbm{1}\big\{w\in\xi^{A}_{t}\big\}. Here and throughout we write v\sim w for \{v,w\}\in E . For ease of notation, we set \xi^{v}_{t}=\xi^{\{v\}}_{t} for v\in V and analogously drop the set notation for singletons in similar instances. We write \mathbf{P}_{G}^{\lambda} for the law of the contact process based on the graph G with infection rate \lambda . Three fundamental and well-known properties of the contact process are monotonicity: \xi^{A}_{t} is stochastically increasing in \lambda and G , attractivity: if A\subset B , then \xi^{B} stochastically dominates \xi^{A} , and (self-)duality: \mathbf{P}_{G}^{\lambda}(\xi_{t}^{A}\cap B=\emptyset)=\mathbf{P}_{G}^{\lambda}(\xi_{t}^{B}\cap A=\emptyset) for A,B\subset V finite. Duality immediately guarantees that \xi^{A} is well-defined for any A\subset V. The extinction time \tau_{\emptyset}(A) of \xi^{A} denotes the time at which (\xi^{A}_{t})_{t\geq 0} first gets absorbed into the state \emptyset . In this article, one central quantity of interest for the contact process on infinite graphs is the survival/extinction threshold of (G,o) , given by \lambda_{\textsf{1}}(G)=\sup\big\{\lambda>0\colon\mathbf{P}_{G}^{\lambda}(\tau_{\emptyset}(o)=\infty)=0\big\}, which is guaranteed to be well-defined by monotonicity. Note that the definition of the critical rate is independent of the choice of the root, since we assume G to be connected. For finite graphs, the critical rate is 0 and contains no information about the dynamics of \xi , and we discuss more appropriate thresholds for this setting below. Invoking monotonicity again, one sees that, for any infinite connected graph G , the value \lambda_{\textsf{1}}(G) is at most the extinction/survival threshold for the contact process on \mathbb{N} with nearest neighbour edges. The latter is well-known to be finite [ 17 ] and thus \lambda_{\textsf{1}}(G)<\infty . The survival/extinction threshold is also known as the lower critical value for the contact process on G , the upper critical value being the threshold at which the infection returns infinitely often to the root. The contact process and related interacting particle systems were first systematically studied in this form in the 1970s, see e.g. [ 17 , 14 ] . Classical choices for G are homogeneous lattices (usually the hypercubic lattice \mathbb{Z}^{d} ) and their finite subgraphs [ 15 , 16 , 10 , 9 , 11 , 12 ] . Using the fact that the contact process is a form of oriented percolation, Bezuidenhout and Grimmett famously proved that the critical contact process on \mathbb{Z}^{d} dies out [ 5 ] and established exponential decay of the volume of the infected set in [ 6 ] . A comprehensive overview of the classical theory is provided in the monographs [ 25 , 24 ] . The fast/slow threshold of the contact process on random graphs. With the advent of network science in the early 2000’s, interest in infection models on finite sparse random graphs arose, as they serve as models for the spread of diseases or information in inhomogeneous populations. Since the contact process dies out on any finite graph, the question of extinction vs. survival becomes a question of fast extinction vs. slow extinction. Here, slow extinction means that the infection survives for a time that scales exponentially in the size of the graph. Let (G_{n})_{n\in\mathbb{N}} denote a sequence of finite graphs with G_{n}=(V_{n},E_{n}) . Assume that |V_{n}|\to\infty , then the fast/slow threshold for (G_{n})_{n\in\mathbb{N}} is given by \lambda_{+}\big((G_{n})_{n\in\mathbb{N}}\big)=\sup\big\{\lambda>0\colon\lim_{n\to\infty}\mathbf{P}_{G_{n}}^{\lambda}\big(\tau_{\emptyset}(V_{n})>{\rm e}^{c|V_{n}|}\big)=0\text{ for all }c>0\big\}. The main purpose of this article is to investigate the question when \lambda_{+}\big((G_{n})_{n\in\mathbb{N}}\big) and \lambda_{\textsf{1}}(G) coincide, provided that the finite graphs (G_{n})_{n\in\mathbb{N}} approximate (G,o) in a suitable sense. To formalise this, we use the framework of [ 1 , 18 , 19 ] . Let {{\mathbb{G}}}_{\ast} denote the space of equivalence classes of connected locally finite rooted graphs modulo rooted isomorphisms, equipped with the local metric \mathsf{d}_{\ast} and let \mathcal{P}({{\mathbb{G}}}_{\ast}) denote the space of Borel probability measures on {{\mathbb{G}}}_{\ast}. We do not usually distinguish between rooted graphs and their equivalence classes in the same vein as one commonly speaks of ‘a random variable X\in L_{1} ’. Let us note that any locally finite rooted graph can be viewed as an element of {{\mathbb{G}}}_{\ast} if it is identified with the connected component of its root. More background on {{\mathbb{G}}}_{\ast} as a metric space can be found in [ 19 ] and the references therein. Let (\mathscr{G}_{n})_{n\in\mathbb{N}} denote a sequence of finite connected random graphs \mathscr{G}_{n}=(\mathscr{V}_{n},\mathscr{E}_{n}) with |\mathscr{V}_{n}|\overset{n\to\infty}{\longrightarrow}\infty in probability. We say (\mathscr{G}_{n})_{n\in\mathbb{N}} converges locally in probability to some random rooted graph (\mathscr{G},o) with distribution \mathsf{Q}\in\mathcal{P}({{\mathbb{G}}}_{\ast}) if, for any \varepsilon>0 , \lim_{n\to\infty}\mathbb{P}\big(\mathsf{d}_{\mathcal{P}}(Q_{n},\mathsf{Q})>\varepsilon\big)=0, (2) where Q_{n}(\,\cdot\,)=|\mathscr{V}_{n}|^{-1}\sum_{v\in\mathscr{V}_{n}}\mathbbm{1}\{(\mathscr{G}_{n},v)\in\,\cdot\,\}\in\mathcal{P}({{\mathbb{G}}}_{\ast}) denotes the empirical distribution on {{\mathbb{G}}}_{\ast} associated with \mathscr{G}_{n} if a root is chosen uniformly at random, and \mathsf{d}_{\mathcal{P}} denotes the Lévy–Prokhorov metric on \mathcal{P}({{\mathbb{G}}}_{\ast}) . We write \mathscr{G}_{n}\underset{n\to\infty}{\overset{\mathbb{P}}{\rightharpoonup}}(\mathscr{G},o) for local convergence in probability. Note that, if o_{n}\in\mathscr{V}_{n} is chosen uniformly at random and the distribution of (\mathscr{G}_{n},o_{n}) is denoted by \mathsf{Q}_{n} , then ( 2 ) implies that \mathsf{Q}_{n}\to\mathsf{Q} weakly 1 1 1 Let us mention that the convergence in ( 2 ) still makes sense if \mathsf{Q} is random, see [ 19 , Remarks 2.12 and 2.13] . In this case, the induced weak limit is the expectation of \mathsf{Q} . However, since we are mostly interested in a law-of-large-numbers-type convergence of the contact process observables, we will always assume \mathsf{Q} to be deterministic. Our results also hold for random \mathsf{Q} in an almost-sure sense, but we do not present them in this way to spare the reader another layer of randomness that has no bearing on the technical core of our results. . This corresponds to local weak convergence of (\mathscr{G}_{n})_{n\in\mathbb{N}} to (\mathscr{G},o) with distribution \mathsf{Q} . In particular, the above definitions apply to sequences (G_{n})_{n\in\mathbb{N}} of deterministic finite graphs. It