Singular flows with time-varying weights

Immanuel Ben-Porat, José A. Carrillo, Pierre-Emmanuel Jabin · 2025

Read Singular flows with time-varying weights by Immanuel Ben-Porat, José A. Carrillo, Pierre-Emmanuel Jabin (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-2503-02276.txt.

Introduction

Immanuel Ben-Porat, José Carrillo, and Pierre-Emmanuel Jabin extend mean-field theory to swarms of particles whose interactions are singular and whose weights change over time. Earlier landmark work handled Coulomb-like repulsion with fixed weights. Time-varying weights break parts of that machinery, so new estimates are needed.

The authors combine the known commutator estimates with a new functional inequality to control the error between the particle system and its limiting equation. They also prove that the limiting partial differential equation and its trajectory system are well posed. The results apply to models where each particle's importance or mass evolves, such as opinion dynamics and biological aggregation with growth.

Catalog record

CREATORImmanuel Ben-Porat, José A. Carrillo, Pierre-Emmanuel Jabin
YEAR2025
COLLECTIONPAPERS
FORMATTXT
FILE SIZE151 KB
VERSION1.0
LICENSEhttp://creativecommons.org/licenses/by/4.0/
RIGHTSCC BY
SOURCEarXiv

Source: https://arxiv.org/abs/2503-02276

Browse the full library · About the Tower of Babel

Full text

The opening of the work, from the library's hosted text:

Singular flows with time-varying weights Immanuel Ben-Porat, José A. Carrillo, Pierre-Emmanuel Jabin Abstract We study the mean field limit for singular dynamics with time evolving weights. Our results are an extension of the work of Serfaty [ 12 ] and Bresch-Jabin-Wang [ 8 ] , which consider singular Coulomb flows with weights which are constant time. The inclusion of time dependent weights necessitates the commutator estimates of [ 12 , 8 ] , as well as a new functional inequality. The well-posedness of the mean field PDE and the associated system of trajectories is also proved. 1 Introduction We aim to study the well posedness and the mean field limit of the following doubly non-local transport PDE \partial_{t}\mu-\mathrm{div}(\mu\,\mathbf{a}\star\mu)=h\left[\mu\right],\ \mu(0,x)=\mu_{0}\,, (1.1) where \mathbf{a}:\mathbb{R}^{d}\rightarrow\mathbb{R}^{d} is a given vector field and h\left[\mu\right] is the signed measure given by h\left[\mu\right]\coloneqq\mu(x)\int_{\mathbb{R}^{d}}S(x-y)\mu(y)dy\,, (1.2) for some function S:\mathbb{R}^{d}\rightarrow\mathbb{R} . We use \mu(x)dx as the notation for integrating against the measure \mu independently of being absolutely continuous or not with respect to the Lebesgue measure. Moreover, we shall denote by \mu(x) its density with respect to the Lebesgue measure in case it is absolutely continuous. Here, both \mathbf{a} and S satisfy some appropriate structural hypothesis specified in Section 2. The initial data \mu_{0} belongs to \mathcal{P}(\mathbb{R}^{d}) , where \mathcal{P}(\mathbb{R}^{d}) is the set of probability measures on \mathbb{R}^{d} . The solutions \mu(t,\cdot)\in\mathcal{P}(\mathbb{R}^{d}) sought after are curves of probability measures, and thus the source term h\left[\mu\right] has to have zero average. The existence and uniqueness theory of PDEs of the type ( 1.1 ) or variants thereof has been handled in several works: in [ 21 ] the existence and uniqueness of measure valued solutions has been established for non-negative source terms h\left[\mu\right] which satisfy a boundeness and Lipschitz condition with respect to the Wasserstein distance. In [ 22 ] , a variant of the Wasserstein distance for signed measures has been introduced and applied in order to remove the positivity assumption on h . The non-positivity assumption is important if one wishes to study models in which h\left[\mu\right] has zero mean. Both of these results do not cover the important case where h\left[\mu\right] is given by the formula ( 1.2 ), since such source terms typically satisfy the boundeness and Lipschitz conditions required only locally. This obstacle has been successfully overcome in [ 14 ] , in which the derivation of the PDE ( 1.1 ) as a mean field limit is also studied. Let us denote by \mathbb{R}^{dN}\setminus\triangle_{N} the set of non-overlapping configurations, that is, the set of \mathbf{x}_{N}:=(x_{1},\dots,x_{N})\in\mathbb{R}^{dN} such that x_{i}\neq x_{j} for all i\neq j , and let us denote by \mathbb{M}^{N} the set of admissible weights, that is, the set of \mathbf{m}_{N}:=(m_{1},\dots,m_{N}) such that m_{i}\geq 0 , with \sum_{i=1}^{N}m_{i}=N . We can define the empirical measure associated to this configuration as \displaystyle\mu_{N}\coloneqq\frac{1}{N}\mathrel{{\mathop{\sum}\limits}}m_{i}\delta_{x_{i}}. (1.3) The following system of ODEs governs the particle dynamics: \left\{\begin{array}[]{lc}\dot{x}_{i}^{N}(t)=-\frac{1}{N}\mathrel{{\mathop{\sum}\limits}}m_{j}^{N}(t)\mathbf{a}(x_{i}^{N}(t)-x_{j}^{N}(t)),\ x_{i}^{N}(0)=x_{i}^{0,N}\\ \dot{m}_{i}^{N}(t)=\frac{1}{N}\mathrel{{\mathop{\sum}\limits}}m_{i}^{N}(t)m_{j}^{N}(t)S(x_{i}^{N}(t)-x_{j}^{N}(t)),\ m_{i}^{N}(0)=m_{i}^{0,N}.\end{array}\right. (1.4) Subject to several technical assumptions on the functions \mathbf{a} and S , the ODE system ( 1.4 ) has a globally well defined flow (\mathbf{x}_{N}(t),\mathbf{m}_{N}(t)) , see [ 2 , Theorem 3] , where \mathbf{x}_{N}(t):=(x_{1}^{N}(t),\dots,x_{N}^{N}(t))\in\mathbb{R}^{dN} and \mathbf{m}_{N}(t):=(m_{1}^{N}(t),\dots,m_{N}^{N}(t))\in\mathbb{M}^{N} . We will denote by \mu_{N}(t) the empirical measure associated to the system ( 1.4 ), with a slight abuse of notation, as the empirical measure given by ( 1.3 ) associated to the configuration (\mathbf{x}_{N}(t),\mathbf{m}_{N}(t)) . The empirical measure is readily checked to be a distributional solution of the PDE ( 1.1 ). We say that ( 1.1 ) is the mean field limit of the interacting particle system ( 1.4 ) if W_{1}(\mu_{N}(0),\mu_{0})\underset{N\rightarrow\infty}{\rightarrow}0\Longrightarrow W_{1}(\mu_{N}(t),\mu(t))\underset{N\rightarrow\infty}{\rightarrow}0,\ t\in[0,T], where \mu(t) is the solution to the PDE ( 1.1 ) with initial data \mu_{0} at time t and W_{1} is the Wasserstein (or Monge-Kantorovich) distance. Before we proceed with the mathematical discussion, let us say a few words about the real life phenomena that the above models idealizes. Given N agents interacting pairwise via a given interaction \mathbf{a} with time evolving opinions (x_{1}^{N}(t),\cdots,x_{N}^{N}(t)) and weights of influence (m_{1}^{N}(t),\cdots,m_{N}^{N}(t)) the system ( 1.4 ) describes the evolution both of the opinions and the weights of influence in time. As it can be seen from ( 1.4 ), the i -th opinion and the i -th weight evolve according to a law which takes into consideration all the other involved opinions and influences. A notable example of a model of this type is the Kuramoto model and its variants [ 1 , 10 ] . Similar models appear also in crowd dynamics and have received mathematical treatment in [ 18 ] and [ 22 ] . The scientific literature seems to be so far mostly limited to treat continuous interaction kernels, although Hegselmann-Krause type models [ 9 ] includes also jump discontinuities. To the best of our knowledge, the blow up singularities which are considered in the present work have not yet been studied when the weights evolve in time. Viewed as a system of opinion dynamics, the mean field limit can be interpreted as a large population limit. This limit has been proven in [ 14 ] , with an approach which resembles the celebrated Dobrushin theorem [ 11 ] . In [ 2 ] a weak version of the mean field limit has been recovered through a graph limit approach. By “weak” we mean that the result is limited to a special choice of initial configurations (at the microscopic level) and a special choice of initial data for the Cauchy problem ( 1.1 ). Among other works which consider the graph limit and its link with the mean field limit we mention [ 7 , 20 ] as well as [ 3 ] for a general overview. The case where the weights form a non-symmetric matrix, a scenario which arises for instance in Neuron dynamics, is also very interesting and has been analyzed in [ 17 ] . All of these results considered measure valued solutions, and typically one has to require that the functions \mathbf{a} and S at least satisfy some Lipschitz continuity. However, the case where the function \mathbf{a} admits discontinuities is also relevant and requires a different mathematical treatment. In the recent work [ 5 ] , the well posedness and mean field limit has been established for the one dimensional attractive Coulomb kernel, which corresponds to the choice \mathbf{a}(x)=\mathrm{sgn}(x) in one dimension. The argument proposed there relies on recasting the limit PDE as a Burgers type equation only valid in one dimension, and invoking Kruzkhov theory of conservation laws in order to prove stability estimates for the resulting Burgers equation. It should be mentioned that singularities emerging from S are also relevant and have been considered in [ 6 , 18 ] . The purpose of this work is twofold: to study the well posedness and mean field limit for ( 1.1 ) for arbitrary dimension d\geq 3 in the case where \mathbf{a} exhibits a Coulomb type singularity, thereby addressing a question left open in [ 5 ] . More precisely, we assume the following hypotheses ( H1 ) on the functions \mathbf{a} and S in ( 1.1 ): • ( H1 -i): \mathbf{a}=\mathbb{J}\nabla V , with V(x) the d- dimensional repulsive Coulomb interaction ( d\geq 3 ), V(x)=-\frac{|x|^{2-d}}{2-d}

Read the complete text