🔢 mathematics

Optimal regularity at the free boundary in one-dimensional first-order mean field games

This paper establishes sharp regularity results for the value function, pressure, and free boundary in one-dimensional first-order mean field games by transforming the problem into Lagrangian coordinates and utilizing a singular change of variables to recast boundary degeneracy as a removable radial axis in an effective higher dimension.

Original authors: Sebastian Munoz

Published 2026-05-11

📖 5 min read🧠 Deep dive

CC BY 4.0

Original authors: Sebastian Munoz

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 crowded room where thousands of people are trying to move from one side to another, but they don't want to bump into each other. In the world of mathematics, this is modeled by something called a Mean Field Game. Instead of tracking every single person, we look at the "density" of the crowd—where it's thick and where it's empty.

This paper by Sebastian Muñoz tackles a specific, tricky version of this problem: What happens at the very edge of the crowd?

The Setup: A Crowd with a Hard Edge

In many real-world scenarios, the crowd doesn't just fade away gradually into nothingness; it has a sharp, defined boundary. Inside this boundary, there are people (density > 0). Outside, there is empty space (density = 0). This boundary is called a free boundary because it moves and changes shape as the game progresses.

The paper asks: How smooth is this edge? Is it jagged and rough, or is it a perfectly polished line? And how smooth are the "pressure" (how much the crowd pushes against itself) and the "value function" (the optimal strategy for each person) right up to that edge?

The Problem: The "Rough" Edge

Previous research had shown that near this edge, things get messy. The math gets "degenerate," meaning the usual rules of smoothness break down. It was known that the pressure (the push of the crowd) is at best "Lipschitz continuous"—think of a line that is continuous but has a sharp corner, like the roof of a house. It's not perfectly smooth.

The big question was: Is this roughness unavoidable? Or is it just a limitation of the math tools we were using? Could the edge actually be smoother if we looked at it the right way?

The Solution: A Magic Lens (Lagrangian Coordinates)

The author's breakthrough is changing the point of view. Instead of watching the crowd from a fixed spot on the street (Eulerian coordinates), he imagines riding along with the people at the very edge of the crowd (Lagrangian coordinates).

He uses a clever mathematical trick, like putting on a special pair of glasses or using a magic lens:

The Square-Root Trick: He transforms the distance from the edge using a square root. Imagine the edge is a wall. If you walk toward it, the distance shrinks. By taking the square root of that distance, he stretches out the space near the wall.
The "Extra Dimensions" Analogy: This transformation reveals a hidden secret. The messy, degenerate math at the edge looks exactly like the math for a radial axis in a higher-dimensional space (specifically, a space with about 4 to 6 dimensions, depending on the crowd's behavior).

Think of it like this: A shadow cast by a 3D object on a 2D wall looks distorted and hard to analyze. But if you step back and look at the 3D object itself, the geometry becomes clear and smooth. Muñoz's method lifts the problem out of the "distorted shadow" and into the "3D object" view.

The Results: Smoothness Revealed

Once he looked through this "magic lens," the results were surprisingly clean:

The Edge is Smooth in Time: The boundary of the crowd doesn't jitter or wiggle. It moves smoothly over time, like a well-choreographed dance.
The Pressure is "Lipschitz" (Sharp but Continuous): The pressure (the push) is indeed as smooth as a sharp corner (Lipschitz). It cannot be smoother than that, no matter how smooth the starting crowd was. This confirms that the "roughness" is a fundamental feature of the physics, not a math error.
The Strategy is "C1,1/2": The optimal strategy for the people (the value function) is smoother than the pressure. It's like a curve that is smooth but has a specific type of "bend" at the edge.
If You Start Smooth, You Stay Smooth: If the crowd starts out perfectly smooth, the pressure and strategy remain smooth right up to the edge, as long as you stay inside the crowd.

The "Why" Behind the Magic

The paper explains why this works using an analogy of a removable axis.
In the transformed math, the edge of the crowd becomes the center of a circle (the axis). Usually, math breaks down at the center of a circle. But because of the specific way the crowd density behaves (it vanishes linearly), the math treats this center as "removable." It's as if the singularity (the breakdown point) is an optical illusion. Once you realize the axis is removable, you can apply powerful, existing mathematical tools (Schauder estimates) to prove the solution is smooth.

Summary

In simple terms, this paper proves that for a specific type of crowd movement game:

The edge of the crowd moves perfectly smoothly over time.
The "push" of the crowd is as smooth as physically possible (it has a sharp corner, but nothing worse).
The "smartest path" for the crowd members is even smoother than the push.
The author achieved this by changing the coordinate system to reveal a hidden, higher-dimensional smoothness that was previously hidden by the complexity of the edge.

The paper settles a long-standing question: The "roughness" at the edge is real and optimal, but the movement of that edge and the strategies inside are as smooth as the math allows.

Technical Summary: Optimal Regularity at the Free Boundary in One-Dimensional First-Order Mean Field Games

Problem Statement
This paper addresses the regularity of solutions to one-dimensional first-order Mean Field Games (MFG) with compactly supported density and power coupling. The system is governed by the Hamilton–Jacobi and continuity equations:
$\begin{cases} -u_t + \frac{1}{2}u_x^2 = m^\theta & \text{in } \mathbb{R} \times (0, T), \\ m_t - (m u_x)_x = 0 & \text{in } \mathbb{R} \times (0, T), \end{cases}$
where $\theta > 0$ , $m$ is the agent density, and $u$ is the value function. The study focuses on the regime where the density $m$ has compact support, creating a free boundary between the positive phase ( $m>0$ ) and the zero phase. The central question is determining the sharp regularity of the value function $u$ , the pressure $p = m^\theta$ , and the free boundary curves themselves, particularly at the interface where the density vanishes.

Previous work established that solutions have finite speed of propagation and that the free boundary is $C^{1,1}$ . However, the precise regularity of the pressure and value function across the free boundary remained open. The paper notes that for the explicit self-similar Barenblatt-type solution, the pressure is Lipschitz but not $C^1$ across the boundary, and the value function is $C^{1,1/2}$ but not smoother. The goal is to prove that these bounds are sharp and generally attained under standard nondegeneracy assumptions on the initial data.

Methodology
The proof proceeds in Lagrangian coordinates, utilizing a singular change of variables to transform the free boundary problem into an interior regularity problem for a degenerate elliptic equation.

Lagrangian Formulation: The system is reformulated using the characteristic map $\gamma(y,t)$ , where $y$ is the initial coordinate. The free boundary corresponds to the fixed endpoints $y=0$ and $y=b$ in Lagrangian space. The flow $\gamma$ satisfies a space-time degenerate elliptic equation:
$\gamma_{tt} + \theta p_0 \gamma_y^{-(2+\theta)} \gamma_{yy} = p_0' \gamma_y^{-(1+\theta)}.$
Here, the degeneracy arises from the linear vanishing of the initial pressure $p_0(y) \sim y$ at the endpoints.
Singular Change of Variables: To handle the degeneracy, the author introduces a square-root distance coordinate $y = r^2/4$ . This transformation converts the linear vanishing of $p_0$ into a quadratic behavior in $r$ . Consequently, the degeneracy is recast as a radial weight $W(r) \sim r^{N-1}$ in an effective dimension $N = 4 + 2/\theta$ . The equation becomes a divergence-form equation with a weight that mimics the radial Jacobian in dimension $N$ .
Removability of the Axis: A crucial step is establishing that the "radial axis" ( $r=0$ ) is removable in the weighted Sobolev space because the effective dimension $N > 2$ . This allows the solution to be extended by even reflection across the axis without imposing additional boundary conditions. The problem is thus reduced to studying the interior regularity of an even solution to a weighted elliptic equation.
Regularity Bootstrap:
- Tangential Regularity: Using recent weighted Schauder estimates for even solutions (Sire–Terracini–Vita), the author establishes that the trace of the solution at the axis is smooth in time ( $C^\infty$ ). This implies the free boundary curves $\gamma_L(t)$ and $\gamma_R(t)$ are smooth.
- Normal Regularity: To recover regularity in the spatial variable, the author employs a "regular-singular descent." By analyzing the equation in the original Lagrangian variable $y$ , the problem is reduced to a regular-singular ordinary differential equation (ODE) structure. The non-integrability of the singular homogeneous solution ensures that the solution inherits the regularity of the source term, allowing for the derivation of higher-order normal derivatives.

Key Contributions and Results
The paper establishes the following optimal regularity results (Theorem 1.1) under standard nondegeneracy assumptions on the initial pressure $p_0$ (specifically, $p_0$ vanishes linearly at the boundary and is $C^{1,\sigma}$ ):

Free Boundary Smoothness: The free boundary curves $\gamma_L(t)$ and $\gamma_R(t)$ are $C^\infty$ in time.
Pressure Regularity: The pressure $p = m^\theta$ is globally Lipschitz continuous ( $W^{1,\infty}$ ) across the free boundary. Inside the positive phase, if the initial pressure $p_0$ is $C^{\ell,\sigma}$ , then $p$ is $C^{\ell,\sigma}$ up to the free boundary.
Value Function Regularity: The value function $u$ is $C^{1,1/2}$ globally. Inside the positive phase, if $p_0$ is $C^{\ell,\sigma}$ , then $u$ is $C^{\ell+1,\sigma}$ up to the free boundary.
Sharpness: These results confirm that the Lipschitz bound for $p$ and the $C^{1,1/2}$ bound for $u$ are sharp and cannot be improved, matching the behavior of the explicit self-similar solutions.

Significance
The paper settles open questions regarding the regularity of first-order MFGs with compact support, previously left unresolved in [4]. It demonstrates a deep structural analogy between these MFGs and the Porous Medium Equation (PME), where the linear vanishing of pressure and nondegeneracy assumptions similarly drive free boundary regularity. However, the paper highlights that the MFG context replaces the parabolic dissipation of the PME with a Hamilton–Jacobi/continuity structure.

The primary methodological contribution is the successful application of weighted Schauder estimates for even solutions to a degenerate elliptic problem arising from MFGs. By interpreting the free boundary degeneracy as a removable radial axis in a higher effective dimension, the author bridges the gap between free boundary problems and interior regularity theory for degenerate equations. This provides a rigorous framework for establishing optimal regularity in systems where the density vanishes at the boundary.