Propagation of chaos for the homogeneous Boltzmann equation with moderately soft potentials
Nicolas Fournier and Stéphane Mischler · 2025
Read Propagation of chaos for the homogeneous Boltzmann equation with moderately soft potentials by Nicolas Fournier and Stéphane Mischler (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-2512-24065.txt.
Introduction
Nicolas Fournier and Stéphane Mischler prove a basic result for the Boltzmann equation. That equation governs how gas particles collide and spread. The claim is called propagation of chaos. It says a large system of colliding particles matches the equation's smooth solution. In the limit, the particles act as if independent.
The pair prove this for moderately soft potentials. In that regime, slow particles collide very often. The math becomes singular and delicate. They adapt recent work by Imbert, Silvestre, and Villani. One estimate does the key work. Fisher information never grows along solutions of the particle system's master equation. That bound tames the singular interaction. The result closes a gap between particle-level randomness and the continuum picture of gases.
Catalog record
| CREATOR | Nicolas Fournier and Stéphane Mischler |
|---|---|
| YEAR | 2025 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 79 KB |
| VERSION | 1.0 |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2512-24065
Browse the full library · About the Tower of Babel
Full text
The opening of the work, from the library's hosted text:
Propagation of chaos for the homogeneous Boltzmann equation with moderately soft potentials Nicolas Fournier and Stéphane Mischler Abstract . We show that the Kac particle system converges, as the number of particles tends to infinity, to the solution of the homogeneous Boltzmann equation, in the regime of moderately soft potentials, γ ∈ ( − 2 , 0 ) \gamma\in(-2,0) with the common notation. This proves the propagation of chaos. We adapt the recent work [ 24 ] of Imbert, Silvestre and Villani, to show that the Fisher information is nonincreasing in time along solutions to the Kac master equation. This estimate allows us to control the singularity of the interaction. 2020 Mathematics Subject Classification 82C40, 60K35, 65C05 1. Introduction and main results 1.1. The homogeneous Boltzmann equation The homogeneous Boltzmann equation describes the evolution of the density f=f(t,v)=f_{t}(v)\geq 0 of particles with velocity v\in{\mathbb{R}}^{3} at time t\geq 0 in a spatially homogeneous dilute gas. It writes (1.1) \partial_{t}f=Q(f,f)\;\;\;\text{in}\;\;\;(0,\infty)\times{\mathbb{R}}^{3},\quad f(0,\cdot)=f_{0}\;\;\;\text{in}\;\;\;{\mathbb{R}}^{3}, where the collision operator is given by (1.2) \displaystyle Q(f,f)(v):=\int_{{\mathbb{R}}^{3}}{\rm d}v_{*}\int_{{\mathbb{S}}_{2}}{\rm d}\sigma B(v-v_{*},\sigma)\big[f(v^{\prime})f(v^{\prime}_{*})-f(v)f(v_{*})\big], and where in the above formula the pre-collisional velocities are given by (1.3) v^{\prime}=v^{\prime}(v,v_{*},\sigma)=\frac{v+v_{*}}{2}+\frac{|v-v_{*}|}{2}\sigma,\quad v^{\prime}_{*}=v^{\prime}_{*}(v,v_{*},\sigma)=\frac{v+v_{*}}{2}-\frac{|v-v_{*}|}{2}\sigma. The cross section takes the form (1.4) B(z,\sigma)\,\sin\theta=\alpha(|z|)\beta(\theta) for some \alpha:{\mathbb{R}}_{+}\to{\mathbb{R}}_{+} and some \beta:(0,\pi]\to{\mathbb{R}}_{+} , where \theta=\theta(z,\sigma) is defined by \cos\theta=\frac{z}{|z|}\cdot\sigma . We will always assume that this cross section corresponds to the situation where particles repel each other according to a force proportional to 1/r^{s} , with s\in(2,\infty) . In that case, we have (1.5) \alpha(r)=r^{\gamma}\;\;\text{and}\;\;c_{1}\theta^{-1-\nu}\leq\beta(\theta)\leq c_{2}\theta^{-1-\nu},\;\;\hbox{with}\;\;\gamma=\frac{s-5}{s-1}\in(-3,1)\;\;\text{and}\;\;\nu=\frac{2}{s-1} for some constants c_{2}>c_{1}>0 . One speaks of hard spheres when \gamma=1 , hard potentials when \gamma\in(0,1) , Maxwell molecules when \gamma=0 and soft potentials when \gamma\in(-3,0) . The present paper is concerned with moderately soft potentials, i.e. \gamma\in(-2,0) . The main physical features of the Boltzmann equation is that any solution conserves mass, momentum and kinetic energy and that its entropy is decaying. From Galilean invariance, we may (and will) always normalize the initial datum, and we thus have (1.6) \int_{{\mathbb{R}}^{3}}f_{t}(v){\rm d}v=1,\quad\int_{{\mathbb{R}}^{3}}f_{t}(v)v{\rm d}v=0,\quad\int_{{\mathbb{R}}^{3}}f_{t}(v)|v|^{2}{\rm d}v=3,\quad\forall\,t\geq 0, as well as the entropy bound H(f_{t}):=\int_{{\mathbb{R}}^{3}}f_{t}(v)\log f_{t}(v){\rm d}v\leq H(f_{0}),\quad\forall\,t\geq 0. In a recent paper [ 21 ] , Guillen and Silvestre have established that the Fisher information (1.7) I_{1}(f):=\int_{{\mathbb{R}}^{3}}\frac{|\nabla f(v)|^{2}}{f(v)}{\rm d}v is nonincreasing in time along the solutions of the Landau equation. This was extended by Imbert, Silvestre and Villani in [ 24 ] to the Boltzmann equation ( 1.1 ), for any physical kernels ( 1.4 )-( 1.5 ). Although such a property result was known for Maxwell molecules, see McKean [ 27 ] , Toscani [ 38 ] and Villani [ 39 ] , it was rather unexpected for other kernels. This remarkable result and the possibility of extending it to the Kac particle system are the starting point and a cornerstone argument in the present work. 1.2. The Kac particle systems Kac [ 26 ] introduced a stochastic particle system, expected to approximate the Boltzmann equation. This system is composed of a large number N of particles characterized by their velocities v_{1},\dots,v_{N} and such that each couple of particles with velocities (v_{i},v_{j}) randomly collides, for each \sigma\in{\mathbb{S}}_{2} , at rate \frac{B(v_{i}-v_{j},\sigma)}{2(N-1)} and are then replaced by particles with velocities v^{\prime}(v_{i},v_{j},\sigma) and v^{\prime}_{*}(v_{i},v_{j},\sigma) . Equivalently, the Kac particle system is a {({\mathbb{R}}^{3})^{N}} -valued Markov process (\mathbf{V}^{N}_{t})_{t\geq 0} with generator \mathcal{L}_{N} defined (at least informally) by \mathcal{L}_{N}\phi(v)=\frac{1}{2(N-1)}\sum_{1\leq i\neq j\leq N}\int_{{\mathbb{S}}_{2}}[\phi(v^{\prime}_{ij})-\phi(v)]B(v_{i}-v_{j},\sigma){\rm d}\sigma, for any v=(v_{1},\dots,v_{N})\in{({\mathbb{R}}^{3})^{N}} and any \phi:{({\mathbb{R}}^{3})^{N}}\mapsto{\mathbb{R}} sufficiently regular. Here, we have set (1.8) v^{\prime}_{ij}=v^{\prime}_{ij}(v,\sigma)=v+(v^{\prime}(v_{i},v_{j},\sigma)-v_{i})\mathbf{e}_{i}+(v^{\prime}_{*}(v_{i},v_{j},\sigma)-v_{j})\mathbf{e}_{j}\in{({\mathbb{R}}^{3})^{N}}, where for h\in{\mathbb{R}}^{3} , h\mathbf{e}_{i}=(0,\dots,0,h,0,\dots,0) with h at the i -th place. It turns out that the law F^{N}_{t} of \mathbf{V}^{N}_{t} satisfies the N -particle linear Boltzmann equation, also called Kac master equation, (1.9) \partial_{t}F^{N}_{t}=\mathcal{L}_{N}F^{N}_{t}\;\;\;\text{in}\;\;\;(0,\infty)\times({\mathbb{R}}^{3})^{N},\quad F^{N}(0,\cdot)=F^{N}_{0}\;\;\;\text{in}\;\;\;({\mathbb{R}}^{3})^{N}. That equation has the same main physical properties as the Boltzmann equation since its solutions also conserve mass, momentum and kinetic energy and have a decaying entropy. As alluded to above, we will establish that the N -particle Fisher information (1.10) I_{N}(F)=\frac{1}{N}\int_{{({\mathbb{R}}^{3})^{N}}}\frac{|\nabla F(v)|^{2}}{F(v)}{\rm d}v is nonincreasing along solutions of ( 1.9 ). 1.3. Propagation of chaos We recall some more or less classical notions of chaos which have been introduced in Kac’s paper [ 26 ] and later in Sznitman [ 35 ] , Carlen et al. [ 2 ] and in [ 22 ] , to which we refer for more details. For a Polish space E , we denote by \mathcal{P}(E) the set of probability measures on E . We say that G^{N}\in\mathcal{P}(E^{N}) is symmetric (and write G^{N}\in\mathcal{P}_{\!s}(E^{N}) ) if \langle G^{N},\varphi_{\sigma}\rangle=\langle G^{N},\varphi\rangle for any \varphi\in C_{b}(E^{N}) and any permutation \sigma of \{1,\dots,N\} , where \varphi_{\sigma}(Z):=\varphi(Z_{\sigma(1)},\dots,Z_{\sigma(N)}) for any Z\in E^{N} . For 1\leq\ell\leq N , we define the \ell -th marginal G^{N}_{\ell}\in\mathcal{P}(E^{\ell}) by \langle G^{N}_{\ell},\varphi\rangle:=\langle G^{N},\varphi\otimes{\bf 1}^{N-\ell}\rangle , for any \varphi\in C_{b}(E^{\ell}) . For g\in\mathcal{P}(E) , we say that a sequence (G^{N})_{N\geq 1} of \mathcal{P}(E^{N}) is (weakly) g -chaotic if G^{N}_{\ell}\rightharpoonup g^{\otimes\ell}\ \hbox{ weakly in }\ \mathcal{P}(E^{\ell})\ \hbox{ as }\ N\to\infty, for any fixed \ell\geq 1 (or equivalently for at least one fixed \ell\geq 2 ). When E={\mathbb{R}}^{3} , we define the N -particle entropy of G^{N}\in\mathcal{P}_{\!s}(({\mathbb{R}}^{3})^{N}) such that m_{2}(G^{N}):=\langle G^{N},|v_{1}|^{2}\rangle<\infty by H_{N}(G^{N}):=\frac{1}{N}\int_{{\mathbb{R}}^{3N}}G^{N}(v)\log G^{N}(v){\rm d}v. Consider now a sequence of symmetric probability measures (G^{N})_{N\geq 1} with \sup_{N\geq 1}m_{2}(G^{N})<\infty . We say that (G^{N})_{N\geq 1} is entropically g -chaotic if G^{N}_{1}\rightharpoonup g\hbox{ weakly in } \mathcal{P}({\mathbb{R}}^{3})\hbox{ and }H_{N}(G_{N})\to H(g). A E^{N} -valued random vector {\mathbf{Z}}^{N} is said to be exchangeable if its law is symmetric. For any Z=(z_{1},...,z_{N})\in E^{N} , we define the empirical measure (1.11) \mu^{N}_{Z}:=\frac{1}{N}\sum_{i=1}^{N}\delta_{z_{i}}\in\mathcal{P}(E). Consider a family ({\mathbf{Z}}^{N})_{N\geq 1} of exchangeable random vectors, with {\mathbf{Z}}^{N} valued in E^{N} . For g\in\mathcal{P}(E) , we say that ({\mathbf{Z}}^{N})_{N\geq 1} is g -chaotic if its sequence of laws is g -chaotic, or equivalently, see [ 35 , Proposition 2.2] , if the associated \mathcal{P}(E) -valued random variable \mu^{N}_{{\mathbf{Z}}^{N}} converges to the deterministic random variable g : (1.12) \mu^{N}_{{\mathbf{Z}}^{N}}\,\Rightarrow\,g\quad\hbox{in probability as}\quad N\to\infty. Here \Rightarrow refers to the weak convergence of measures (tested against C_{b}(E) -functions). Similarly, when E={\mathbb{R}}^{3} , we say that ({\mathbf{Z}}^{N})_{N\geq 1} is entropically g -chaotic if the associated sequence of laws is entropically g -chaotic. 1.4. Main result We are now in position to formulate the main results of the paper. Theorem 1.1 . Consider a kernel B satisfying ( 1.4 )-( 1.5 ) with \gamma\in(-2,0) and a nonnegative initial condition f_{0} satisfying the normalization \int_{{\mathbb{R}}^{3}}f_{0}(v){\rm d}v=1,\quad\int_{{\mathbb{R}}^{3}}f_{0}(v)v{\rm d}v=0,\quad\int_{{\mathbb{R}}^{3}}f_{0}(v)|v|^{2}{\rm d}v=3, and with finite Fisher information. Denote by f\in C([0,\infty);L^{1}({\mathbb{R}}^{3})) the unique weak solution to the Boltzmann equation ( 1.1 ) with nonincreasing Fisher information and which which satisfies ( 1.6 ). For any N\geq 2 , there exists a {({\mathbb{R}}^{3})^{N}} -valued Kac particle system (\mathbf{V}^{N}_{t})_{t\geq 0} with initial law f_{0}^{\otimes N} and for all t\geq 0 , (\mathbf{V}^{N}_{t})_{N\geq 2} is entropically f_{t} -chaotic. Precise definitions of the objects involved in this theorem as well as more accurate or more general statements will be given in the next sections. In particular, we prove propagation of chaos in the sense of trajectories, see Theorem 6.1 . The main novelty is that we are able to deal with moderately soft potentials. Some propagation of chaos results of Kac’s system to the Boltzmann equation have been obtained for Maxwell molecules by Kac [ 26 ] , Graham-Méléard [ 19 ] , Desvillettes-Graham-Méléard [ 8 ] , see also