Phase transitions for interacting particle systems on random graphs

Benedetta Bertoli, Grigorios A. Pavliotis, Niccolò Zagli · 2025

Read Phase transitions for interacting particle systems on random graphs by Benedetta Bertoli, Grigorios A. Pavliotis, Niccolò Zagli (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-2504-02721.txt.

Introduction

Benedetta Bertoli, Grigorios Pavliotis, and Niccolò Zagli study swarms of interacting particles placed on random networks in a 2025 paper. Each particle drifts and feels pulls from the others, and the group as a whole can settle into more than one steady pattern.

The authors focus on the large-group limit, where a smooth equation takes over. They find the exact points where the steady pattern splits, and tie those points to traits of the network itself. A self-consistency check recovers the main split point and reveals further splits along uneven branches. They also show the group's energy can serve as a signal of the shift, and that past the shift the system can linger between states.

Catalog record

CREATORBenedetta Bertoli, Grigorios A. Pavliotis, Niccolò Zagli
YEAR2025
COLLECTIONPAPERS
FORMATTXT
FILE SIZE55 KB
VERSION1.0
LICENSEhttp://creativecommons.org/licenses/by/4.0/
RIGHTSCC BY
SOURCEarXiv

Source: https://arxiv.org/abs/2504-02721

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 interacting particle systems on random graphs Benedetta Bertoli, Grigorios A. Pavliotis, Niccolò Zagli Abstract In this paper, we study weakly interacting diffusion processes on random graphs. Our main focus is on the properties of the mean-field limit and, in particular, on the nonuniqueness of stationary states. By extending classical bifurcation analysis to include multichromatic interaction potentials and random graph structures, we explicitly identify bifurcation points and relate them to the eigenvalues of the graphon integral operator. Furthermore, we characterize the resulting McKean-Vlasov PDE as a gradient flow with respect to a suitable metric. We combine these theoretical results with the spectral analysis of the linearized McKean-Vlasov operator and extensive numerical simulations to gain insight into the stability and long-term behaviour of stationary solutions. In addition, we provide strong evidence that (minus) the interaction energy of the interacting particle system serves as a natural order parameter. In particular, beyond the transition point and for multichromatic interactions, we observe an energy cascade that is strongly linked to the dynamical metastability of the system. 1 Department of Mathematics, Imperial College London, London SW7 2AZ, UK, [email protected] and [email protected] 2 Nordita, Stockholm University and KTH Royal Institute of Technology - Hannes Alfvéns väg 12, SE-106 91 Stockholm, Sweden. [email protected] 1 Introduction 1.1 General introduction The study of stochastic interacting particle systems (SIPS) and their mean-field limit has been a topic of extensive research in recent decades due to their wide-ranging applications in physics [ 32 , 1 , 39 ] , biology [ 58 , 8 ] , and even the social sciences [ 55 , 68 , 38 , 14 , 42 ] . One of the main interesting aspects of such systems is the emergence of collective behavior at the macroscale, due to the interaction between the particles at the microscale. Examples include the emergence of consensus in models for opinion dynamics [ 38 , 70 ] , chemotaxis [ 65 ] , collective synchronization [ 39 ] , emergence of order in active matter [ 18 ] , self-organized alignment dynamics [ 25 ] , mean field games and macroeconomics [ 26 ] , synchronization dynamics in biological and technological systems [ 63 , 64 ] . In recent years, it has been recognized that the emergence of collective behavior at the macroscale can be interpreted as a disorder-order phase transition [ 16 ] . The area of SIPS and of their mean field limit has experienced enormous progress in recent years, and many extensions to, e.g. multi-species models [ 36 ] or moderately interacting diffusions [ 40 ] have been made. In many applications, it is important to consider interacting particle systems and, more generally, agent-based models, on graphs. Interacting multiagent systems on graphs have been used for modelling in many contexts, such as biology and the social sciences; examples include the dynamics of power grid networks [ 34 ] , opinion dynamics [ 29 , 2 , 61 , 57 ] , models of biological neurons [ 43 , 44 ] , social networks [ 59 ] , mean-field (stochastic) games [ 15 , 56 ] . In recent years, the rigorous analysis of interacting particle systems on graphs and of their mean field limit has seen significant development in the case where white noise is incorporated in the system. Stochastic interacting particle systems (SIPS) on random graphs have been studied in many recent works, including [ 5 , 12 , 27 , 11 , 50 , 21 , 22 , 37 , 24 ] . These studies have established, under appropriate assumptions on the interaction potential and on the graph structure, several results on propagation of chaos, the law of large numbers and central limit theorems for SIPS on graphs. A crucial element of the analysis presented in these (and many other recent) papers is the systematic use of the theory of graphons [ 48 , 54 ] ; see also recent developments on network dynamics on graphops [ 47 ] . Clearly, different network topologies can have a profound impact on the collective dynamics of the SIPS. This is a topic that has been studied extensively in recent years, in particular for Kuramoto-type models and their variants. Kuramoto-type models on graphs have been extensively studied in the deterministic setting in a series of papers by Medvedev and collaborators ( [ 41 ] , [ 19 ] , [ 54 ] , [ 45 ] , [ 20 ] ). In these important papers, the authors obtained several results on the long-term behavior and bifurcations of such systems; in particular, the influence of the network topology on the synchronization onset was studied. Both first-order (overdamped) and second-order (inertial) dynamics were considered. It is well known that, for fully coupled SIPS at equilibrium in the absence of an underlying network topology, the mean-field dynamics can exhibit several stationary states, corresponding to critical points of the system’s free energy. These dynamics are characterized by phase transitions, where changes in control parameters, such as temperature and/or interaction strength, lead to shifts in the global minimizer of the free energy. A standard example is the disorder-order phase transition from the uniform to the synchronized state in the noisy Kuramoto model [ 9 ] . A detailed study of bifurcations for the noisy Kuramoto model on random graphs is presented in [ 37 ] . The main goal of this paper is to systematically study phase transitions for SIPS on random graphs, with particular emphasis on how the interaction potential and the network topology influence the nature of these transitions. In particular, we extend the bifurcation analysis presented in the aforementioned works beyond the Kuramoto model, i.e. we consider the multichromatic interaction potentials studied in [ 10 ] and consider a variety of network topologies. There are several applications, e.g. in polymer dynamics [ 49 ] and biology [ 60 ] where it is necessary to consider interaction potentials (on the torus or the sphere) with several nonzero (negative) Fourier modes. As examples we mention the Onsager and Maiers-Saupe potentials [ 16 , 49 , 67 ] or the Hegselmann-Krause interaction potential [ 68 , 38 ] used in opinion dynamics models. In [ 10 ] the stability of multipeak solutions was studied in detail. In particular, it was shown that, in general, such states tend to be unstable and that the dynamics converges, in the long-time limit, to a single cluster. Understanding the stability of cluster/multipeak states is a very interesting problem related to the phenomenon of dynamical metastability [ 13 , 4 ] . One of the questions that we address in this paper is whether multipeak solutions are unstable for a variety of random graph topologies. In addition, we introduce an ”order parameter”, namely the negative of the interaction energy, that keeps track of the number of clusters of the SIPS. Our main contributions are as follows. • We identify the gradient flow structure of the McKean-Vlasov equation in the space of probability measures equipped with a modified Wasserstein metric. We derive the associated free energy functional and establish its connection to the stationary states of the PDE. • Motivated by Tamura’s approach [ 66 ] ,

Read the complete text