Formation of clusters and coarsening in weakly interacting diffusions
N. J. Gerber, R. S. Gvalani, M. Hairer, G. A. Pavliotis, A. Schlichting · 2025
Read Formation of clusters and coarsening in weakly interacting diffusions by N. J. Gerber, R. S. Gvalani, M. Hairer, G. A. Pavliotis, A. Schlichting (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-2510-17629.txt.
Introduction
Nicolai Gerber, with Rishabh S. Gvalani and three co-authors, studies clumping in weakly interacting particles in a 2025 arXiv paper. Their particles drift randomly on a ring while a short range attraction pulls nearby ones together. The group asks when the particles spread evenly and when they gather into clusters. The answer ties closely to sudden phase changes in a matching mean field equation, which describes the average density.
A new variant of a rearrangement inequality anchors their proofs. It shows the lowest energy states are either uniform or a single symmetric cluster. Particle clusters can merge and trade mass by diffusion. Mean field clusters cannot move, so they trade mass only. The authors model that trade and confirm it with careful simulations.
Catalog record
| CREATOR | N. J. Gerber, R. S. Gvalani, M. Hairer, G. A. Pavliotis, A. Schlichting |
|---|---|
| YEAR | 2025 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 91 KB |
| VERSION | 1.0 |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2510-17629
Browse the full library · About the Tower of Babel
Full text
The opening of the work, from the library's hosted text:
Formation of clusters and coarsening in weakly interacting diffusions Thanks: The research of RSG was partially supported by the Deutsche Forschungsgemeinschaft (DFG, German Research Foundation) under the Special Priority Programme 2410/1 “Hyperbolic Balance Laws in Fluid Mechanics: Complexity, Scales, Randomness”. GAP is partially supported by an ERC-EPSRC Frontier Research Guarantee through Grant No. EP/X038645, ERC Advanced Grant No. 247031 and a Leverhulme Trust Senior Research Fellowship, SRF \ \backslash R1 \ \backslash 241055. AS is supported by the Deutsche Forschungsgemeinschaft (DFG, German Research Foundation) under Germany’s Excellence Strategy EXC 2044 –390685587, Mathematics Münster: Dynamics–Geometry–Structure. N. J. Gerber, R. S. Gvalani, M. Hairer, G. A. Pavliotis, A. Schlichting Abstract . This paper studies the clustering behavior of weakly interacting diffusions under the influence of sufficiently localized attractive interaction potentials on the one-dimensional torus. We describe how this clustering behavior is closely related to the presence of discontinuous phase transitions in the mean-field PDE. For local attractive interactions, we employ a new variant of the strict Riesz rearrangement inequality to prove that all global minimizers of the free energy are either uniform or single-cluster states, in the sense that they are symmetrically decreasing. We analyze different timescales for the particle system and the mean-field (McKean–Vlasov) PDE, arguing that while the particle system can exhibit coarsening by both coalescence and diffusive mass exchange between clusters, the clusters in the mean-field PDE are unable to move and coarsening occurs via the mass exchange of clusters. By introducing a new model for this mass exchange, we argue that the PDE exhibits dynamical metastability. We conclude by presenting careful numerical experiments that demonstrate the validity of our model. 1. Introduction Interacting particle systems and their mean-field limits appear in many applications ranging from physics to mathematical models for swarming, flocking, opinion dynamics, and collective behavior. They are also used in the development of algorithms for sampling and optimization. Often, the emergence of collective behavior, e.g. consensus formation [ HK02 ] or synchronization [ Kur81 ] , can be interpreted as a (first- or second-order) phase transition from a disordered to an ordered state; see [ CGPS20 ] and the references therein, in particular [ CP10 ] . Noise-driven interacting particle systems can exhibit metastable states that persist over long time scales. As examples, we mention the formation of clusters in opinion models [ GPY17 ] as well as in mathematical models for transformers [ GLPR25 , GKPR24 , BPA24 ] ; see also [ SS25 , Section 6.1] . Noise-induced metastability in interacting particle systems is a manifestation of the presence of discontinuous phase transitions [ Pen+84 , GS20 ] . It is expected that, near a transition point, the mean-field description is no longer valid for arbitrarily long times [ DGPS23 ] and that the effect of noise, through the Dean–Kawasaki SPDE [ Kaw73 , Dea96 ] , must be taken into account to capture metastable states accurately [ Weh+25 ] . The primary objective of this work is to provide a rigorous mathematical understanding of the emergence of clusters in weakly interacting diffusions exhibiting first-order phase transitions and to derive a model for the dynamics of these clusters. In [ PNT90 ] , the authors justify that the mean-field description for Hamiltonian systems is inadequate to describe all regimes in the Newtonian dynamics of particles interacting via a pairwise attractive Gaussian potential. They observed that for sufficiently strong interactions, the initially diffused particles form clusters that eventually collapse to form a single cluster (or, as the authors of [ PNT90 ] refer to it, the condensed phase). Although the mean-field Vlasov equation describes the final cluster state, which is a stationary state of the mean-field PDE, it is unable to explain the transient dynamics that leads to the formation of the condensed phase. We quote this main conclusion of theirs below: We find that the condensed phase is a stationary solution of the Vlasov equation, but the Vlasov dynamics cannot describe the collapse. The goal of this paper is to study a similar phenomenology in the setting of weakly interacting diffusions with localised attractive interaction potentials. The thermodynamic limit in our setting is the so-called McKean–Vlasov equation, and our main conclusion reads as follows: Although the single-cluster state is a stationary solution of the mean-field McKean-Vlasov PDE, coalescence of clusters is not captured by the mean-field dynamics. It turns out that the weakly interacting diffusion model studied in this paper exhibits very rich and interesting dynamics. While the interacting particle model in [ PNT90 ] is deterministic and the entropy in the corresponding mean-field Vlasov equation is conserved, the model of interacting diffusions considered in this paper implies that the corresponding mean-field equation is a diffusion-aggregation equation that exhibits coarsening through mass exchange between clusters. This mass exchange leads to dynamic metastability of the mean-field dynamics, which is related to the notion of slow motion of gradient flows studied in [ OR07 ] . We comment on some more general relations to similar effects observed in related models. First, we note that our identification of the two time scales for tunneling and coalescence is similar to the classical theory of Ostwald ripening : a configuration of near-equilibrium liquid droplets sitting on a precursor film that wets the entire substrate can coarsen in time by two different mechanisms: collapse or collision of droplets [ GORS09 ] . Here, the phenomenological model is a gradient flow of Cahn–Hilliard type, where the dynamics is entirely deterministic. The authors provide a reduced model for the position and volume of single droplets in [ GORS09 , Section 4] , which also allows them to study the crossover between the timescales of migration (leading to collapse) and ripening (through coalescence). The relevant time scales can be identified using suitable energies, in this case related to surface tension, and it is possible to derive a differential inequality for them. This approach was pioneered in [ KO02 ] and later refined in [ CNO06 , ORS06 , BOS11 , OSS13 ] . In the physical literature, a study of diffusive aggregating particles can be found in [ PTA14 ] , where the observed behavior is studied on the PDE level. There, Ostwald ripening is related to volume modes induced by the escape of particles from clusters due to random fluctuations. However, a crucial difference from our study is that the translational modes causing movement and joining of clusters are again deterministically caused by long-range aggregation forces between the clusters, whereas in our model, the interaction potential has finite interaction range and clusters can move only due to the Brownian motions. We also comment on the link between the problem studied in our paper and Kramers’ problem and its generalizations. Indeed, the mass leakage between clusters that occurs in the model that we derive in this paper is a generalization of Kramers’ problem pioneered in [ Kra40 ] , where a Brownian particle is confined in a double-well potential and the timescale is related to the energy barrier of the system. Our generalization requires considering potentials with multiple timescales, which in addition change dynamically as a result of the mass exchange. In recent years, several different techniques for studying metastability in time-independent potentials