🌀 nonlinear sciences

Lagrangian formulation and Eulerian closure in alignment dynamics

This paper establishes the global well-posedness and quantitative flocking of a continuum Lagrangian $p$ -alignment system, constructs a corresponding Eulerian closure that accounts for Reynolds stress and defect forces, and proves that these defect terms vanish asymptotically under heavy-tailed interactions to yield mono-kinetic limits, while also providing sharp critical thresholds and uniform mean-field convergence results for the linear case.

Original authors: José A. Carrillo, Young-Pil Choi, Eitan Tadmor

Published 2026-04-14

📖 6 min read🧠 Deep dive

CC BY 4.0

Original authors: José A. Carrillo, Young-Pil Choi, Eitan Tadmor

Original paper licensed under CC BY 4.0 (http://creativecommons.org/licenses/by/4.0/). ✨ This is an AI-generated explanation of the paper below. It is not written or endorsed by the authors. For technical accuracy, refer to the original paper. Read full disclaimer

Imagine a massive flock of birds, a school of fish, or even a crowd of people trying to move together. Each individual is constantly looking at their neighbors and adjusting their speed to match them. This is the phenomenon of alignment.

This paper is a mathematical deep dive into how we can predict the behavior of these huge groups. The authors, José A. Carrillo, Young-Pil Choi, and Eitan Tadmor, propose a new way of looking at the problem that solves several long-standing puzzles in physics and mathematics.

Here is the story of their discovery, explained simply.

1. The Two Ways to Watch a Crowd

To understand the paper, you first need to know there are two ways to watch a crowd move:

The Eulerian View (The "Weather Station" Approach): Imagine standing on a street corner with a camera. You don't care which specific person is walking past; you only care about the density of people and the average speed at that specific spot. This is how meteorologists watch clouds. It's great for big pictures, but it gets messy if people cross paths or if the crowd splits and merges in complex ways.
The Lagrangian View (The "Name Tag" Approach): Imagine every single person in the crowd has a unique name tag. You follow Person A from start to finish, then Person B, and so on. You track their exact journey. This is very precise, but if you have a million people, it's a lot of data to manage.

The Problem: For decades, mathematicians tried to jump from the "Name Tag" view to the "Weather Station" view. But when the crowd gets chaotic—when people cross paths or pile up—the "Weather Station" view often breaks down. It creates "ghosts" or "defects" in the math that don't make physical sense.

2. The Authors' Big Idea: The "Lagrangian First" Strategy

Instead of trying to force the "Weather Station" view to work, the authors decided to start with the "Name Tag" view (the Lagrangian approach) and build the "Weather Station" view from there.

Think of it like this:

Old Way: Try to draw a smooth map of traffic flow, but when cars crash, the map tears.
New Way: Follow every single car's GPS track perfectly. Then, look at where all those GPS tracks end up to draw the map.

By starting with the individual tracks, they proved that the system is always well-behaved (mathematically "well-posed"), even if the starting crowd is messy or the rules for how they interact are weird.

3. The "Reynolds Stress": The Hidden Chaos

Here is the most creative part of their discovery.

When you follow individual "Name Tags" (Lagrangian), sometimes two different people end up at the exact same spot at the same time, but they are moving at different speeds.

Example: A slow walker and a fast runner arrive at the same corner.
The "Weather Station" View: It sees a crowd at that corner. It calculates the average speed. But it loses the information that there is a slow person and a fast person.
The "Ghost": This missing information creates a hidden pressure or "stress" in the math. The authors call this Reynolds Stress.

In traditional physics, this stress is often treated as a mystery or a modeling error. The authors say: "No, it's not a mystery. It's a real, measurable thing caused by the crowd mixing!"

They created a new equation (the Euler-Reynolds-Alignment System) that includes this "stress" term explicitly. It's like upgrading a weather forecast to include not just the average wind speed, but also the turbulence caused by wind gusts hitting each other.

4. The Magic of "Flocking" (The Cleanup Crew)

The paper asks a crucial question: Does this hidden stress last forever?

The answer is no, thanks to the power of flocking.

The Metaphor: Imagine a chaotic dance floor. At first, people are moving in all directions (high stress). But as they start listening to each other and aligning their moves, they eventually all start dancing in perfect unison.
The Result: Once the group "flocks" (everyone moves at the same speed), the "Name Tags" no longer cross paths with different speeds. The "slow walker" and "fast runner" eventually match speeds.
The Conclusion: As the group aligns, the hidden "Reynolds Stress" vanishes. The messy, complex equation simplifies back into the clean, classic "Weather Station" equation.

The authors proved that alignment itself is the cleanup crew. You don't need to force the crowd to be smooth; if they just keep aligning, the math naturally becomes smooth over time.

5. Special Cases: The One-Dimensional "Traffic Jam"

The authors also looked at a 1D line (like cars on a single-lane road).

The Rule: If the people in the front are faster than the people in the back, they will eventually crash into each other (collide).
The Insight: They found a sharp "tipping point." If the initial speeds are ordered correctly, the crowd flows smoothly forever. If not, they crash.
The Surprise: Even if they crash, the math doesn't break. It just switches to the "Reynolds" version (the one with the stress term) to describe the pile-up. This gives a complete picture of what happens when a traffic jam forms.

6. From Birds to Math: The "Mean-Field" Limit

Finally, they connected this back to the real world. They showed that if you have a huge number of individual agents (like $N$ birds), and you let $N$ go to infinity, the behavior of the birds converges perfectly to their new "Lagrangian-first" model.

They proved that this convergence happens uniformly in time. This means their model is accurate not just for a few seconds, but for the entire lifetime of the flock.

Summary: Why This Matters

This paper is a bridge.

It starts with the individual (the Lagrangian view), which is always reliable.
It builds a macroscopic view (the Eulerian view) that admits when things get messy (Reynolds stress).
It proves that nature fixes itself: as the group aligns, the messiness disappears, and the system becomes simple and predictable again.

In a nutshell: They found a way to mathematically track a chaotic crowd from start to finish, showing that even when the crowd gets messy, the underlying rules are simple, and eventually, everyone just moves together in perfect harmony.

1. Problem Statement

The paper addresses the mathematical challenge of connecting microscopic particle models of collective alignment (flocking) to macroscopic continuum descriptions. Specifically, it investigates the Cucker–Smale (CS) model with nonlinear velocity coupling ( $p \ge 2$ ), where the interaction force is proportional to $|v_j - v_i|^{p-2}(v_j - v_i)$ .

The core difficulties in this field include:

Deriving Hydrodynamic Limits: Rigorously deriving the Euler-alignment system (a macroscopic PDE) from the particle system or kinetic equations, particularly when the initial data is weak (e.g., measures rather than smooth functions).
Handling Non-Injectivity: In continuum limits, Lagrangian trajectories may cross (collide), causing the mapping from initial labels to current positions to lose injectivity. This leads to "microscopic" velocity fluctuations at a single Eulerian point, breaking the standard "mono-kinetic" assumption (where density $\rho$ and velocity $u$ are uniquely defined).
Closure Issues: Standard hydrodynamic limits often fail to close because they cannot account for the stress tensor generated by these velocity fluctuations (Reynolds stress) or nonlinear defect forces arising from $p \neq 2$ .

2. Methodology

The authors adopt a Lagrangian-first approach, reversing the traditional Eulerian-centric methodology.

Lagrangian Foundation: They treat the continuum Lagrangian system as the primary object. Instead of assuming a smooth Eulerian velocity field, they define the dynamics via a nonlocal mean-field ODE for a continuum of agents labeled by a reference measure $\rho_0$ .
$\partial_t \eta_t = v_t, \quad \partial_t v_t = \kappa \int \phi(\eta_t(x) - \eta_t(y)) G_p(v_t(y) - v_t(x)) \rho_0(dy)$
This formulation is well-posed even for discontinuous initial data and arbitrary probability measures $\rho_0$ .
Measure-Theoretic Correspondence: To recover Eulerian variables, they use pushforward and disintegration (via the disintegration theorem).
- Density $\rho_t$ and momentum $m_t$ are pushforwards of $\rho_0$ .
- The barycentric velocity $u_t$ is the Radon-Nikodym derivative.
- Crucially, they define the Reynolds stress $\tau_t$ as the fibre variance of the velocity distribution within the pre-image (fibre) of the flow map $\eta_t$ .
Kinetic Lifting: They lift the Lagrangian flow to phase space to form a kinetic measure $\mu_t = (\eta_t, v_t)_\# \rho_0$ , which satisfies a Vlasov-type equation. This allows the use of compactness arguments for long-time limits.
Stability Analysis: For the mean-field limit, they employ modulated Wasserstein distances to compare the $N$ -particle system directly with the limiting Lagrangian flow, establishing uniform-in-time stability estimates.

3. Key Contributions and Results

A. Global Well-Posedness of Lagrangian Dynamics (Theorem 1.1)

Result: The authors prove global existence and uniqueness of solutions to the Lagrangian $p$ -alignment system for bounded initial velocities ( $u_0 \in L^\infty$ ).
Flocking: Under a "heavy-tail" condition on the communication kernel ( $\int^\infty \phi(r) dr = \infty$ ), they establish quantitative flocking estimates: the velocity diameter $d_v(t)$ decays to zero as $t \to \infty$ .
Significance: This provides a robust framework that does not require a priori smoothness of the Eulerian velocity, allowing for the study of "blobs" of particles with discontinuous initial configurations.

B. The Euler–Reynolds–Alignment (ERA) System (Theorem 1.2)

Result: By pushing forward the Lagrangian flow, they derive a closed macroscopic system:
$\partial_t \rho + \nabla \cdot (\rho u) = 0$
$\partial_t (\rho u) + \nabla \cdot (\rho u \otimes u + \tau) = \rho (A_p[\rho, u] + R_p[\rho, u])$
- Reynolds Stress ( $\tau$ ): A non-negative tensor representing velocity dispersion within Lagrangian fibres. It is not an ad-hoc closure but is uniquely determined by the Lagrangian flow.
- Nonlinear Defect Force ( $R_p$ ): A new term appearing only for $p > 2$ , arising from the nonlinearity of the interaction kernel $G_p$ and fibre fluctuations.
Significance: This system is the exact Eulerian description of the Lagrangian dynamics. It resolves the "closure problem" by explicitly identifying the defect terms rather than assuming they vanish.

C. Asymptotic Mono-Kinetic Closure (Theorem 1.3)

Result: The authors prove that velocity flocking alone is sufficient to suppress the defect terms. As $t \to \infty$ , the Reynolds stress $\tau_t$ and the nonlinear defect force $R_p$ vanish asymptotically.
Mechanism: The decay of the velocity diameter forces the fibre distributions to collapse to Dirac masses. Consequently, the ERA system asymptotically reduces to the standard mono-kinetic Euler-alignment system.
Significance: This demonstrates that the complex Reynolds stress is a transient phenomenon driven by initial misalignment, which disappears naturally due to the alignment mechanism itself, without requiring spatial confinement or injectivity of the flow.

D. Global Weak Solutions in 1D and High Dimensions (Theorems 1.4 & 1.5)

1D Case ( $p=2$ ): They establish a sharp critical threshold based on the monotonicity of an "effective velocity."
- If the effective velocity is non-decreasing, the flow remains injective, $\tau \equiv 0$ , and global mono-kinetic solutions exist.
- If not, collisions occur, and the solution is described by the ERA system with non-trivial Reynolds stress.
High Dimensions ( $p=2$ ): Under a large coupling strength condition, they prove global injectivity of the Lagrangian flow. This ensures $\tau \equiv 0$ for all time, yielding global weak solutions to the standard Euler-alignment system.
Significance: These results provide the first global existence theory for Euler-alignment with regular (non-singular) kernels in high dimensions without relying on singular dissipation.

E. Uniform-in-Time Mean-Field Limits (Theorems 1.6 & 1.7)

Result: For the linear case ( $p=2$ ), they prove a uniform-in-time stability estimate between the $N$ -particle Cucker–Smale system and the limiting Lagrangian flow.
Convergence: If the initial data is prepared such that the Lagrangian flow remains injective (or the disintegration is Dirac), the particle system converges uniformly in time to the mono-kinetic Euler-alignment system in the Wasserstein metric.
Significance: Unlike previous results that often yield finite-time convergence or require strong regularity, this provides a uniform-in-time convergence rate, validating the hydrodynamic limit for all time scales.

4. Significance and Impact

Unification of Scales: The paper bridges the gap between particle, kinetic, and hydrodynamic descriptions of alignment dynamics through a rigorous Lagrangian framework.
Resolution of Closure Problems: It introduces a mathematically precise definition of Reynolds stress and defect forces derived directly from the Lagrangian flow, removing the need for ad-hoc modeling assumptions in the Eulerian limit.
Robustness to Weak Data: By starting with the Lagrangian formulation, the theory accommodates initial data that are merely bounded measures, a significant generalization over classical PDE approaches requiring $C^1$ or Sobolev regularity.
Asymptotic Justification: It proves that the standard mono-kinetic Euler-alignment system is the correct long-time limit of the full dynamics, provided flocking occurs, even if the flow map is not globally injective.
Critical Thresholds: The sharp characterization of global existence in 1D and the large-coupling condition in higher dimensions offer new insights into the structural stability of alignment models.

In summary, this work provides a comprehensive theoretical foundation for alignment dynamics, showing that the complex micro-macro interactions can be rigorously tracked via Lagrangian methods, leading to a clear understanding of when and how macroscopic flocking emerges.