arXiv reader

On the uniform-in-time propagation of chaos for the Motsch-Tadmor model

Tamari Kldiashvili, Stefano Rossi and Lara Trussardi

arXiv
2610.21828v1 · PDF · 15 October 2026
Categories
math.AP
Licence
CC BY 4.0 The site shows the text.

Abstract

We consider the Motsch-Tadmor model, a modification of the Cucker-Smale model for flocking dynamics that incorporates asymmetric communication weights. In the case of uniformly positive interactions, we consider the validity problem for the limiting kinetic equation, proving a uniform-in-time propagation of chaos result in the moving frame given by the consensus velocities. The result is obtained by studying the reversibility of the communication weights and by implicitly characterizing the asymptotic consensus.

Items 8

Proposition

Proposition 2.1 (Velocity consensus)

O738

Consider the Motsch–Tadmor model (?) with initial data \{(X_i^N(0), V_i^N(0))\}_{i=1}^N and assume that the influence function \varphi satisfies the assumption in (?). Introducing the quantity

\langle \boldsymbol{V}^N(t)\rangle_{N}:= \sum_{i=1}^N \pi_i^N(t) V_i^N(t),

where \{\pi_i^N(t)\}_{i=1}^N is given in (?), there exists \mathbf{V}^N_\infty\in \mathbb{R}^d, characterized in (Proof), such that \langle \boldsymbol{V}^N(t)\rangle _{N} \to \mathbf{V}^N_\infty as t \uparrow \infty. Moreover, there exists C>0 such that

\max_{i\in \{1,\dots,N\}}\left|V_i^N(t)-\mathbf{V}^N_\infty\right|\le C \, \mathrm e^{-\eta t}, \quad \eta:= \frac mM.

Similarly, consider a regular solution f_t(x,v) of the Motsch–Tadmor model (?) with compactly supported initial datum f_{\mathrm{in}} \in C^1_{\mathrm c}(\mathbb{R}^d \times \mathbb{R}^d) and introduce the quantity

\langle v(t)\rangle_{f}:= \displaystyle \int_{\mathbb{R}^d\times \mathbb{R}^d}\pi^\rho(t,x) v\, \mathrm df_t(x,v),

where \pi^\rho(t,x) is given in (?). Then, there exists \mathbf{v}^f_\infty\in \mathbb{R}^d, characterized in (?), such that \langle v(t)\rangle _{f} \to \mathbf{v}^f_\infty as t \uparrow \infty. Moreover, there exists C>0, such that

\left|v-\mathbf{v}^f_\infty\right|\le C \, \mathrm e^{-\eta t}, \quad \eta:= \frac mM,

for any v \in \text{supp}f_t.

Used by Proof
Proof of Proposition 2.1

We consider the weighted average of the velocities introduced in (2.1) and we show that the limit in time of this quantity exists and corresponds to the consensus velocity of the model. Taking the time derivative

\dfrac{{\mathrm d}}{\mathrm dt} \langle \boldsymbol{V}^N(t)\rangle_{N}= \sum_{i=1}^N \dot \pi_i^N(t) V_i^N(t)+ \sum_{i=1}^N \pi_i^N(t) \dot V_i^N(t),

by a straightforward computation, the second term is identically equal to zero thanks to (?). Hence

\dfrac{{\mathrm d}}{\mathrm dt} \langle \boldsymbol{V}^N(t)\rangle_{N}= \sum_{i=1}^N \dot \pi_i^N(t) V_i^N(t)= \sum_{i=1}^N \dot \pi_i^N(t)\left[ V_i^N(t)- \langle \boldsymbol{V}^N(t)\rangle_{N}\right],

since \sum_i \dot \pi_i=0. By (?), it follows that

\left| \dfrac{{\mathrm d}}{\mathrm dt} \langle \boldsymbol{V}^N(t)\rangle_{N}\right| \le \sum_{i=1}^N \pi_i^N(t)\left|\dfrac{\dot \pi_i^N(t)}{\pi_i^N(t)}\right|\sum_{j=1}^N \pi_j^N(t)|V_i^N(t)-V_j^N(t)|\le \dfrac{2\text{Lip}(\varphi)}{m}[D^N_V(t)]^2.

By the asymptotic flocking (?), we conclude that

\left| \dfrac{{\mathrm d}}{\mathrm dt} \langle \boldsymbol{V}^N(t)\rangle_{N}\right| \le \dfrac{2 \text{Lip}(\varphi)}{m} [D^N_V(0)]^2 \mathrm{e}^{-2\eta t}.

This implies that the limit in time of \langle \boldsymbol{V}^N(t) \rangle_{N} exists and is given by

\mathbf{V}^N_\infty:= \langle \boldsymbol{V}^N(0) \rangle_{N}+\int_0^\infty \sum_{i=1}^N \dot \pi_i^N(t) V_i^N(t)\, {\mathrm d}t.

Moreover

\begin{aligned} |V_i^N(t)- \mathbf{V}_\infty^N| &\le |V_i^N(t)- \langle \boldsymbol{V}^N(t) \rangle_{N}| + |\langle \boldsymbol{V}^N(t) \rangle_{N}- \mathbf{V}_\infty^N|\\ &\le \sum_{j=1}^N \pi_j^N(t) |V_i^N(t)- V_j^N(t)| + |\langle \boldsymbol{V}^N(t) \rangle_{N}- \mathbf{V}_\infty^N|\le C{\mathrm e}^{-\eta t}, \end{aligned}

with C:=D^N_V(0)+\text{Lip}(\varphi) [D^N_V(0)]^2/ \eta m.

Similarly, at the kinetic level we can consider the average velocity introduced in (2.1). Taking the time derivative and integrating by parts, we get

\begin{aligned} \dfrac{{\mathrm d}}{\mathrm dt} \langle v(t)\rangle_{f}&=\int \partial_t \pi^\rho(t,x) v\, \mathrm df_t(x,v)+ \int \pi^\rho(t,x) v\, \mathrm \partial_tf_t(x,v) \mathrm dx \mathrm dv\\ &=\int \left(\partial_t \pi^\rho(t,x) + v \cdot \nabla_x \pi^\rho(t,x)\right) v \, \mathrm d f_t(x,v) \\ &\quad+ \int \pi^\rho(t,x) \left[\,\int a^\rho(t,x,y)(w-v)\, \mathrm df_t(y,w)\right] \mathrm df_t(x,v), \end{aligned}

by a straightforward computation, the last term is identically equal to zero thanks to (?). Hence

\begin{aligned} \dfrac{{\mathrm d}}{\mathrm dt} \langle v(t)\rangle_{f}=\int \left(\partial_t \pi^\rho(t,x) + v \cdot \nabla_x \pi^\rho(t,x)\right) [v-\langle v(t) \rangle_{f}] \, \mathrm d f_t(x,v) \end{aligned}

since \partial_t \int \pi^\rho(t,x) \mathrm d\rho_t(x)=0. By (?), it follows that

\begin{aligned} \left| \dfrac{{\mathrm d}}{\mathrm dt} \langle v(t)\rangle_{f}\right| &\le \int \, \mathrm df_t(x,v)\pi^\rho(t,x)\left|\dfrac{\partial_t\pi^\rho(t)+v\cdot \nabla_x \pi^\rho(t)}{\pi^\rho(t)}\right|\int \pi^\rho(t,y)|v-w|\, \mathrm df_t(y,w)\\ &\le \dfrac{2\text{Lip}(\varphi)}{m}[D^f_V(t)]^2. \end{aligned}

By the asymptotic flocking (?), we conclude that

\left| \dfrac{{\mathrm d}}{\mathrm dt} \langle v(t)\rangle_{f}\right| \le \dfrac{2 \text{Lip}(\varphi)}{m} [D^f_V(0)]^2 \mathrm{e}^{-2\eta t}.

This implies that the limit in time of \langle v(t) \rangle_{f} exists and is given by

\mathbf{v}^f_\infty:= \langle v(0) \rangle_{f}+\int_0^\infty \int(\partial_t \pi^\rho(t)+v\cdot \nabla_x \pi^\rho(t)) v \, \mathrm df_t(x,v)\, {\mathrm d}t.

Moreover, for any v \in \text{supp} f_t

|v- \mathbf{v}_\infty^f| \le |v- \langle v(t) \rangle_{f}| + |\langle v(t) \rangle_{f}- \mathbf{v}_\infty^f|\le C{\mathrm e}^{-\eta t},

with C:=D^f_V(0)+\text{Lip}(\varphi) [D^f_V(0)]^2/ \eta m.

Theorem

Theorem 2.1 (Uniform-in-time propagation of chaos)

GMTL

Assume (?), and suppose f_{\mathrm{in}}\in C_{\mathrm c}^{1}(\mathbb{R}^{d}\times \mathbb{R}^d). Let f_t be the regular solution of the kinetic Motsch–Tadmor model (?) with initial datum f_{\mathrm{in}} and let F_{t}^N be the Liouville solution to (?) with initial datum F^N_{\mathrm{in}}=f_{\mathrm{in}}^{\otimes N}.

Then there exists a constant C>0, depending only on d, m, M, \|\varphi\|_{\mathrm{Lip}}, and the diameter of the velocity support of f_{\mathrm{in}}, such that

\sup_{t\geq0}\, \mathcal{W}_2\left( \widetilde F_{k,t}^N, \widetilde f_t^{\otimes k} \right) \leq C\sqrt{\frac{k}{N}},

where \widetilde f_t:= (\mathcal{G}_t)_\# f_t and, for 1\leq k\leq N, \widetilde F_{k,t}^N is the k-particle marginal of (\mathcal{G}_t^N)_{\#}F^N_{t} where \mathcal G_t(x,v) := \left( x-t\mathbf v^f_\infty,\, v-\mathbf v^f_\infty \right) and

\mathcal{G}^N_{t}(X_N,V_N):=(x_1-t \mathbf V_\infty^N,\dots, x_N -t \mathbf V_\infty^N, v_1- \mathbf V^N_\infty, \dots, v_N-\mathbf V_\infty^N).
Corollary

Corollary 2.1 (Asymptotic flocking in the moving frame)

1NCN

Assume (?), and let f_{\mathrm in}\in \mathcal C^1_{\mathrm c}(\mathbb{R}^d\times\mathbb{R}^d) be compactly supported. Let f_t be the solution to (?), and let \widetilde f_t=(\mathcal G_t)_\# f_t. Then there exist \rho_\infty\in\mathcal P(\mathbb{R}^d) and a constant C>0 such that

\mathcal W_2\left( \widetilde f_t,\, \rho_\infty\otimes\delta_0 \right) \leq C\,\mathrm e^{-\eta t}. \qquad \eta=\frac mM,

Similarly, let F_t^N=(\Phi_t^N)_\#F_{\rm in}^N be a solution of the Liouville equation (?), and let \widetilde F_{k,t}^N be the k-th marginal of (\mathcal G_t^N)_\#F_t^N where \mathcal G_t^N(X_N,V_N) is defined in (?). Then there exist \rho_{k,\infty}^N \in\mathcal P(\mathbb{R}^{dk}) and a constant C>0 such that

\mathcal W_2\left( \widetilde F_{k,t}^N,\, \rho_{k,\infty}^N\otimes\delta_0^{\otimes k} \right) \leq C\,\sqrt k\,\mathrm e^{-\eta t}.
Proof of Corollary 2.1

We give a proof for the discrete case. By Proposition 2.1, for any i=1,\ldots,N there exists a constant C>0 such that

\left| V_i^N(t)-\mathbf V_\infty^N \right| \leq C\,\mathrm e^{-\eta t}.

Since

\frac{\mathrm d}{\mathrm dt} \left( X_i^N(t)-t\mathbf V_\infty^N \right) = V_i^N(t)-\mathbf V_\infty^N,

we deduce that

X_{i,\infty}^N := X_i^N(0) + \int_0^\infty \left( V_i^N(s)-\mathbf V_\infty^N \right)\,\mathrm ds

exists and

\left| X_i^N(t) -t\mathbf V_\infty^N -X_{i,\infty}^N \right| \leq \frac C \eta\,\mathrm e^{-\eta t}.

This yields to

\left| \mathcal G^N_t\bigl(\boldsymbol{X}^N(t),\boldsymbol{V}^N(t)\bigr) - \bigl(X^N_\infty,0\bigr) \right|^2 \leq C^2\, N \left(1 +\frac{1}{\eta^2}\right)\mathrm e^{-2\eta t},

where X_{\infty}^N := \bigl( X_{1,\infty}^N,\ldots,X_{N,\infty}^N \bigr). Setting \rho_\infty^N:=(X_{\infty}^N)_\#F_{\rm in}^N, the definition of Wasserstein distance implies that

\mathcal W_2^2 \left( \widetilde F_t^N,\, \rho_\infty^N\otimes\delta_0^{\otimes N} \right) \leq \int_{(\mathbb{R}^d \times \mathbb{R}^d)^N} \left| \mathcal G^N_t\bigl(\boldsymbol{X}^N(t),\boldsymbol{V}^N(t)\bigr) - \bigl(X^N_\infty,0\bigr) \right|^2\mathrm d f_{\mathrm{in}}^{\otimes N} \leq C\,N\, \mathrm e^{-2\eta t},

which concludes the proof.

Corollary

Corollary 2.2

0FF2

Under the assumptions and notation of Theorem 2.1, there exists a constant C>0, depending only on d, \eta, \mathrm{Lip}(\varphi), and the diameter of the velocity support of f_{\mathrm{in}}, such that

\mathcal{W}_2\left( \widetilde F_{k,t}^N, (\rho_\infty\otimes \delta_0)^{\otimes k} \right) \leq C\sqrt{k}\left({\frac{1}{\sqrt{N}}}+ \mathrm e^{-\eta t}\right),

for fixed k \leq N-1.

Lemma

Lemma 3.1

BK03

Consider the Motsch–Tadmor dynamics in (?) and (?) and let \{\pi_i^N(t)\}_{i=1}^N and \pi^\rho(t,x) as in (?). Under the assumption (?), the following results hold:

  1. the weights \{a_{ij}^N\} defined in (?) are reversible w.r.t. \{\pi_i^N(t)\}_{i=1}^N, i.e. \sum_{i=1}^N \pi^N_i(t)=1 and

    \pi^N_i(t) a_{ij}^N(t)=\pi^N_j(t) a_{ji}^N(t).

    Similarly, a^{\rho}(t) defined in (?) is reversible w.r.t. \pi^\rho(t), i.e. \int \pi^\rho(t,x)\mathrm d\rho_t(x)=1 and

    \pi^\rho(t,x) a^{\rho}(t,x,y)=\pi^\rho(t,y) a^{\rho}(t,y,x).
  2. It holds that

    \max_{1\le i\le N}\left|\dfrac{\dot \pi_i^N(t)}{\pi_i^N(t)}\right|\le \dfrac{2 \text{Lip}(\varphi)}{m}\, D_V^N(t),\quad \left|\dfrac{\partial_t \pi^\rho+v\cdot \nabla_x \pi^\rho}{\pi^\rho}\right| \le \frac{2 \text{Lip}(\varphi)}{m} D_V^f(t)

    where D^N_V(t) is introduced in (?), while D^f_V(t) in (?).

  3. For any i \in \{1,\dots,N\}, it holds that

    \eta\le\min \left\{N\pi^N_i(t), \pi^\rho(t,x)\right\} \le \max \left\{N\pi^N_i(t), \pi^\rho(t,x)\right\}\le \dfrac{1}{\eta}, \quad \eta=\dfrac mM.
Used by Proof
Proof of Lemma 3.1

We prove the result at the discrete level, a similar argument works for the kinetic quantities. If \varphi_{ij}:=\varphi(|X_i^N-X_j^N|), we get

\pi^N_i a^N_{ij}=\dfrac{\sum_\ell \varphi_{i\ell}}{\sum_{\ell h}\varphi_{\ell h}} \dfrac{\varphi_{i j}}{\sum_{\ell} \varphi_{i \ell}}= \dfrac{\sum_\ell \varphi_{j \ell}}{\sum_{\ell h}\varphi_{\ell h}} \dfrac{\varphi_{j i}}{\sum_{\ell} \varphi_{j \ell}}= \pi^N_j\, a^N_{j i}\,

since \varphi_{ij}=\varphi_{ji}, this proves the first item. Concerning (?), we have

\begin{aligned} \log \pi_i^N(t)=\log\left(\sum_\ell \varphi(|X_i^N(t)-X_\ell^N(t)|)\right)- \log\left( \sum_{\ell,h} \varphi(|X_\ell^N(t)-X_h^N(t)|\right). \end{aligned}

Taking time derivatives, using (?) and the definition of D_V^N(t), we get (?). Finally, last item follows by the uniform bounds in (?).

Source and licence

The text on this page is from arXiv:2610.21828v1 by Tamari Kldiashvili, Stefano Rossi and Lara Trussardi, under the licence CC BY 4.0. The corpus made these changes: it read the items out of the LaTeX source, it expanded the macros of the paper, it put each proof after its statement, and it converted the text to HTML. The formulas are the TeX of the paper, and KaTeX renders them. A formula or a reference can look different from the PDF. The page is published under CC BY 4.0.

The record of each item is in analysis/objects/2610/2610.21828.jsonl.