Rates of convergence in long time asymptotics of an alignment model with symmetry breaking
Alexandre Surin Ceremade (CNRS UMR n o 7534), PSL University, Université Paris-Dauphine, Place de Lattre de Tassigny, 75775 Paris 16, France Email: [email protected] · 2025
Read Rates of convergence in long time asymptotics of an alignment model with symmetry breaking by Alexandre Surin Ceremade (CNRS UMR n o 7534), PSL University, Université Paris-Dauphine, Place de Lattre de Tassigny, 75775 Paris 16, France Email: [email protected] (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-2509-06765.txt.
Introduction
Flocking models with noise settle into aligned motion at rates Alexandre Surin pins down in a solo 2025 analysis paper. The model is a nonlinear Fokker-Planck equation derived from the Cucker-Smale flocking system with noise added. Noise strength decides a phase transition: strong noise leaves one symmetric, unpolarized steady state, while weaker noise allows a whole continuum of polarized steady states.
Above the threshold, solutions converge to the unique radial steady state, even though that state is unstable in the symmetry-breaking range. There, polarized states attract all solutions whose entropy starts low enough. Surin proves convergence in a weighted norm happens at an exponential rate, and the average speed also converges exponentially to a unique limit. That limit selects a single polarized steady state, resolving which of the many possible aligned outcomes a given flock reaches.
Catalog record
| CREATOR | Alexandre Surin Ceremade (CNRS UMR n o 7534), PSL University, Université Paris-Dauphine, Place de Lattre de Tassigny, 75775 Paris 16, France Email: [email protected] |
|---|---|
| YEAR | 2025 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 62 KB |
| VERSION | 1.0 |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2509-06765
Browse the full library · About the Tower of Babel
Full text
The opening of the work, from the library's hosted text:
Rates of convergence in long time asymptotics of an alignment model with symmetry breaking Alexandre Surin Ceremade (CNRS UMR n o 7534), PSL University, Université Paris-Dauphine, Place de Lattre de Tassigny, 75775 Paris 16, France Email: [email protected] Abstract . We consider a nonlinear Fokker-Planck equation derived from a Cucker-Smale model for flocking with noise. There is a known phase transition depending on the noise between a regime with a unique stationary solution which is isotropic (symmetry) and a regime with a continuum of polarized stationary solutions (symmetry breaking). If the value of the noise is larger than the threshold value, the solution of the evolution equation converges to the unique radial stationary solution. This solution is linearly unstable in the symmetry-breaking range, while polarized stationary solutions attract all solutions with sufficiently low entropy. We prove that the convergence measured in a weighted L 2 L^{2} norm occurs with an exponential rate and that the average speed also converges with exponential rate to a unique limit which determines a single polarized stationary solution. Keywords: Collective behavior, flocking, non-linear Fokker–Planck equation, phase transition. AMS Subject Classification: 35B40; 35P15; 35Q92. 1. Introduction. In various models of mathematical biology and social sciences, the dynamics of a population of individuals without leader is often described by mean field interaction forces that depend on the parameters of the agents, for instance, their relative positions and velocities. These interaction rules are the basis for models of collective behavior and the description of emergent structures. There are two well-known models. The Cucker–Smale model [ 6 ] describes agents whose velocities tend to align with the average velocity of their neighbors, with an interaction strength that decreases with distance. The Vicsek model [ 14 ] , describes particles moving at a constant speed but the direction of the velocity is determined by the average orientation of the neighbors and by a stochastic component, typically a Brownian noise. This noise takes into account individual randomness and estimation errors. The model exhibits a noise-dependent phase transition, marking the emergence of global alignment as the noise intensity decreases. We study a variant of the Cucker–Smale model with Brownian noise and a self-propulsion force. This framework has been investigated in [ 2 , 11 , 12 ] and [ 13 ] . We consider the spatially homogeneous case, in which the evolution of the density f(t,v) of individuals is described by (1) \frac{\partial f}{\partial t}=D\,\Delta f+\nabla\cdot\big[\left(\nabla\psi_{\alpha}(v)+v-u_{f}\right)f\big]\quad\text{with}\quad\psi_{\alpha}(v)=\frac{\alpha}{4}\,|v|^{4}-\frac{\alpha}{2}\,|v|^{2}. Here t\geq 0 denotes the time variable, v\in\mathbb{R}^{d} is the velocity variable, D>0 is the intensity of the noise, \nabla\cdot and \Delta respectively denote the divergence and the Laplacian with respect to v , while \alpha>0 is a parameter of the external potential and u_{f} denotes the average velocity \displaystyle u_{f}:=\frac{\int_{\mathbb{R}^{d}}v\,f(t,v)\,dv}{\int_{\mathbb{R}^{d}}f(t,v)\,dv}. In contrast with the classical Vicsek model, we do not require that |v| is constant. For any u\in\mathbb{R}^{d} , we define a Gibbs state as (2) \displaystyle G_{u}(v):=\mathcal{Z}(u)^{-1}{e^{-\frac{1}{D}\left(\frac{1}{2}|v-u|^{2}+\psi_{\alpha}(v)\right)}} where \mathcal{Z}(u) is a renormalization constant. A stationary state is a Gibbs state with a specific value of the speed |u| . In [ 2 ] it was proved that a phase transition occurs and the following refined result is taken from [ 11 ] . Proposition 1.1 ( [ 11 , Theorem 1.1] ) . There exists D_{*}>0 such that: • If D\geq D_{*} , then G_{0} is the unique stationary state. It is isotropic and stable. • If D<D_{*} , then there exists r(D)>0 such that G_{u} is a stationary state for any u\in\mathcal{S}\cup\{0\} where (3) \displaystyle\mathcal{S}:=r(D)\mathbb{S}^{d-1} and there are no other stationary states. Moreover, G_{0} is unstable and the set of polarized states \{G_{u}\}_{u\in\mathcal{S}} is stable. For two nonnegative functions f and g of unit mass we define the relative entropy as \displaystyle\mathcal{H}[f|g]:=\int_{\mathbb{R}^{d}}f\log\left(\frac{f}{g}\right)dv and the free energy associated to Equation 1 as (4) \displaystyle\mathcal{F}[f]=D\int_{\mathbb{R}^{d}}f\log f\,dv+\int_{\mathbb{R}^{d}}f\,\psi_{\alpha}\,dv+\frac{1}{2}\int_{\mathbb{R}^{d}}|v-u_{f}|^{2}\,f\,dv. Our first main result deals with the large time asymptotics of the solution and the identification of a limiting velocity. Theorem 1.1 . Let D<D_{*} . Let f be a nonnegative solution to Equation 1 with initial datum f_{\mathrm{\mathrm{ini}}}\in L^{1}_{+}(\mathbb{R}^{d}) such that \mathcal{F}[f_{\mathrm{\mathrm{ini}}}]<\mathcal{F}[G_{0}] and \|f_{\mathrm{ini}}\|_{L^{1}(\mathbb{R}^{d})}=1 . Then the average speed u_{f} converges as t\rightarrow\infty to a unique limit u_{\infty}\in\mathcal{S} and for any \beta>0 , \displaystyle\lim_{t\rightarrow\infty}t^{\beta}\left(\mathcal{H}\big[f(t,\cdot)\mid G_{u_{\infty}}\big]+|u_{f}(t)-u_{\infty}|^{2}\right)=0 It was known from [ 11 ] that \lim_{t\rightarrow\infty}\inf_{u\in\mathcal{S}}{\mathcal{H}\big[f(t,.)|G_{u}\big]}=0 but the existence of a unique limiting Gibbs state and the convergence of u_{f}(t) as t\rightarrow\infty are new, thus answering a long standing open question. Under additional conditions on the initial datum, we obtain an exponential rate of convergence. Theorem 1.2 . There is a positive constant \beta depending only on \alpha and D such that under the assumptions of Theorem 1.1 , if f_{\mathrm{ini}}\in\bigcap_{u\in\mathcal{S}}L^{2}(\mathbb{R}^{d},G_{u}^{-1}dv) , then \displaystyle\mathcal{H}\big[f(t,.)|G_{u_{\infty}}\big]+|u_{f}-u_{\infty}|^{2}=O(e^{-\beta t})\text{ as }t\rightarrow+\infty. The additional assumption is slightly stronger than asking f_{\mathrm{ini}} to belong to L^{2}(\mathbb{R}^{d},G_{u}^{-1}dv) for only one u . Indeed, the function v\mapsto\sqrt{G_{u}(v)/\left(1+|v|^{d+1}\right)} belongs to L^{2}(\mathbb{R}^{d},G_{u}^{-1}\,\mathrm{d}v) but not to L^{2}(\mathbb{R}^{d},G_{w}^{-1}\,\mathrm{d}v) if w\neq u . The proofs of Theorem 1.1 and Theorem 1.2 are based on a detailed analysis of the free energy. The exponential convergence is obtained by considering the projection of u_{f}(t) onto \mathcal{S} . We refer to [ 2 ] for an introduction to ( 1 ). The Vicsek model [ 14 ] is the limit of ( 1 ) as \alpha\rightarrow\infty according to [ 4 ] . Models with a large class of interaction potentials but a strong assumption of uniform convexity were investigated in [ 5 ] . Their results do not apply to Equation 1 . The first large time asymptotic result in the non-critical isotropic case D>D_{*} was established in [ 11 ] . In this paper, Li proved the convergence to the unique stationary state G_{0} with exponential decay rate. We give a different proof which includes the study of the critical case D=D_{*} and the more general initial datum of Theorem 1.1 . We adapt techniques from [ 7 , 8 ] for the Vicsek model. This paper is organized as follows. Definitions and preliminary results are collected in Section 2 . In Section 3 , we prove the convergence of the free energy ( Proposition 3.1 ). Theorem 1.1 and Theorem 1.2 are proved respectively in Section 4 and Section 5 . 2. Preliminary results. 2.1. Existence and uniqueness We provide here some already known results and references. Since Equation 1 has a gradient flow structure, a natural approach is to consider a discrete gradient descent scheme in the Wasserstein space, as introduced by Jordan, Kinderlehrer, and Otto in [ 10 ] . The drawback of this method is that it does not give uniqueness since the free energy is not convex. The existence is detailed in Chapter 8 of [ 15 ] . The proof only requires an initial datum in L^{1}(\mathbb{R}^{d},\,dx) with finite entropy and second moment. In [ 3 ] , Bolley, Cañizo and Carrillo proved the existence and uniqueness for a large class of kinetic PDE’s which covers ( 1 ). However, their method requires some additional integrability conditions on the initial datum, see [ 3 ] Theorem 1.2. In [ 3 , 15 ] , the solutions are measure-valued in the sense of distributions at t=0 , and admit a density with respect to the Lebesgue measure for any t>0 .