Mullins-Sekerka as the Wasserstein flow of the perimeter

Antonin Chambolle, Tim LauxProceedings of the American Mathematical Society 2020 · DOI · arXiv

Abstract

We prove the convergence of an implicit time discretization for the one-phase Mullins-Sekerka equation, possibly with additional non-local repulsion, proposed in [F. Otto, Arch. Rational Mech. Anal. 141 (1998), pp. 63–103]. Our simple argument shows that the limit satisfies the equation in a distributional sense as well as an optimal energy-dissipation relation. The proof combines arguments from optimal transport, gradient flows & minimizing movements, and basic geometric measure theory.

Introduction

The Mullins-Sekerka equation is a fundamental mathematical model describing the evolution of phase boundaries, such as the interface of a fluid droplet. A classic physical manifestation is found in the Hele-Shaw cell, where a viscous fluid is compressed between two parallel plates. The boundary's motion is driven by a competition between surface tension—which seeks to minimize the perimeter—and non-local repulsion, which can lead to the formation of intricate, labyrinthine patterns.

In 1998, Felix Otto proposed a revolutionary perspective: this evolution could be interpreted as a gradient flow. In this framework, the fluid's shape moves along a path of "steepest descent" to minimize its total energy, with the "speed" of this descent measured by the Wasserstein distance ($W_2$) from the field of optimal transport. While this interpretation was conceptually powerful, a rigorous proof that discrete time-steps in this framework actually converge to the continuous physical equation remained elusive for over two decades. This paper provides that proof.

The Variational Framework: Minimizing Movements

To solve the Mullins-Sekerka equation, the authors utilize the Minimizing Movements scheme. Rather than tackling the complex partial differential equations (PDEs) directly, the evolution is discretized into small time steps ($h$). At each step, a new set $E$ (representing the fluid's shape) is found by solving a variational optimization problem:

$$\min_{E} \left\{ \text{Energy}(E) + \frac{1}{2h} W_2^2(E, E_{prev}) \right\}$$

Here, the Energy term accounts for both the perimeter and any non-local repulsion, while the Dissipation term (the squared Wasserstein distance) represents the physical cost of moving the fluid from its previous configuration.

Methodological Innovations

The proof's success hinges on several key technical advancements:

  1. De Giorgi’s Variational Interpolations: The authors use these interpolations to bridge the gap between discrete steps, allowing them to recover a continuous-time energy-dissipation relation.
  2. Geometric Measure Theory and Varifolds: To account for boundaries that may become highly irregular, split, or merge, the authors employ the language of varifolds and sets of finite perimeter. This allows the proof to remain valid even when the fluid boundary lacks classical smoothness.
  3. Non-linear Interpolation Inequality: A critical component of the proof is a new technical inequality (Lemma 2) derived by the authors. This inequality relates the $L^1$ difference between two sets to their perimeters and their Wasserstein distance, providing the necessary compactness to ensure the sequence of approximations converges.

Key Results and Insights

The primary contribution of this work is Theorem 1, which establishes that as the time-step $h$ approaches zero, the sequence of shapes generated by the optimization problem converges to a limit that satisfies the Mullins-Sekerka equation in a distributional sense.

Optimal Energy Dissipation

The limit solution is shown to obey an optimal energy-dissipation relation. This confirms that the resulting motion is not just a solution, but the most efficient path possible for the fluid to take according to the underlying physical laws. Furthermore, the authors successfully recover the balance-of-forces condition on the free boundary, where the pressure is equal to the sum of the curvature and the repulsive forces.

Simplicity and Robustness

One of the most striking aspects of this proof is its relative simplicity. Unlike previous attempts that relied on heavy regularity theory—assuming the surface remains smooth and well-behaved—this approach shows that the energy-dissipation structure itself is robust enough to guarantee convergence. This suggests that the Wasserstein distance is the "natural" metric for modeling energy dissipation in fluid interfaces.

Limitations and Future Directions

While the paper provides a rigorous foundation for Otto's scheme, it utilizes a "weak solution" concept. Because the proof relies on varifolds, it leaves open the theoretical possibility of "hidden boundaries"—mathematical artifacts that do not correspond to the actual fluid edge—though these are not expected to occur in physical applications.

Additionally, this work focuses on the "one-phase" version of the equation (fluid moving in a vacuum). Future research may extend these methods to the "two-phase" version, where two distinct fluids interact, and further investigate the conditions under which these weak solutions maintain geometric regularity (rectifiability).

Summary updated

Abstract

We prove the convergence of an implicit time discretization for the one-phase Mullins-Sekerka equation, possibly with additional non-local repulsion, proposed in [F. Otto, Arch. Rational Mech. Anal. 141 (1998), pp. 63–103]. Our simple argument shows that the limit satisfies the equation in a distributional sense as well as an optimal energy-dissipation relation. The proof combines arguments from optimal transport, gradient flows & minimizing movements, and basic geometric measure theory.
BibTeX
@inproceedings{chambolle2020mullins,
  author = {Antonin Chambolle and Tim Laux},
  title = {{Mullins-Sekerka as the Wasserstein flow of the perimeter}},
  booktitle = {Proceedings of the American Mathematical Society},
  year = {2020},
  doi = {10.1090/proc/15401},
}
Figure