Interior pointwise $C^{1,α}$ estimates for Stokes systems in divergence form

Rong Dong, Dongsheng LiarXiv 2023 · DOI · arXiv

TL;DR

This paper establishes interior pointwise $C^{1,\alpha}$ (Hölder) estimates for the velocity of fluids modeled by Stokes systems in divergence form. Crucially, the authors prove these regularity estimates without assuming the system's coefficients or source terms are continuous in time, revealing a mathematically surprising result: while the fluid velocity itself isn't Hölder continuous in time, its curl (vorticity) is.

What Problem Does This Paper Address?

The Stokes system of equations describes the flow of highly viscous fluids (like honey or glycerin) where advective inertial forces are negligible compared to viscous forces. Mathematically, it is a type of parabolic partial differential equation (PDE), but it features a "degeneration" due to the pressure term and the incompressibility constraint (the fluid's volume doesn't change).

While the smoothness (regularity) of solutions to Stokes systems with constant or highly smooth coefficients is well-studied, real-world materials and complex fluids often require models with rougher, variable coefficients. Specifically, the authors tackle a challenging scenario: what happens if the system's coefficients and source data are continuous in space, but completely lack continuity in the time variable?

Because of a famous mathematical boundary (Serrin's counterexample), it is known that under these conditions, the fluid velocity cannot be guaranteed to be continuous in time. The authors set out to find exactly what can be proven about the pointwise smoothness of the solution under these extremely weak time-regularity assumptions.

Methods

To prove their estimates, the authors avoid trying to bound the derivatives directly. Instead, they use an iterative approximation technique based on Campanato's characterization of Hölder spaces.

In simple terms, Campanato's theory characterizes the smoothness of a function by looking at how well it can be approximated by polynomials within shrinking local neighborhoods (balls or cylinders). The authors approximate the fluid velocity at a $(1+\alpha)$-th order scale using a spatial first-order polynomial, where the coefficients of that polynomial are allowed to depend on time.

The primary technical novelty is how they handle the iteration. To make the math work, the velocity, its gradient, and its curl must be estimated simultaneously. They use a clever decoupling strategy:

  1. They estimate the curl of the velocity using techniques standard for parabolic equations.
  2. They estimate the gradient of the velocity using the curl and the velocity itself via techniques standard for elliptic equations.

Datasets

Not discussed in the paper. (This is a theoretical mathematics paper focused on PDE proofs; no empirical datasets are used).

Key Insights and Ideas

  • Directional Regularity: The findings highlight a phenomenon similar to "directional regularity" in PDEs, where a solution is smoother in certain directions (space) than in others (time).
  • The "Aha!" Moment on Curl: The most astonishing finding is that while the velocity field $u$ is only Hölder continuous in the spatial variables (and explicitly not in time), the curl of the velocity ($\nabla \times u$) is Hölder continuous in both space AND time.
  • Physical Intuition: While the paper is purely mathematical, this makes physical sense. In fluid dynamics, the curl of the velocity is the vorticity. Because the gradient of the pressure term vanishes when you take the curl ($\nabla \times \nabla p = 0$), the vorticity evolves smoothly in time even if the pressure and velocity are experiencing rough, discontinuous jumps in time. The math rigorously proves this behavior holds even when the underlying fluid matrix coefficients are rough in time.

Key Results

  • Pointwise Estimates (Theorem 1.1): The authors rigorously prove that if the coefficient matrix and source terms have small oscillations in space (but arbitrary behavior in time), the velocity can be tightly bounded and approximated by a linear spatial polynomial $a(t) + b(t)x$.
  • Classical Estimates (Corollary 1.2): They formally establish that the spatial gradient of the velocity ($\nabla u$) belongs to the Hölder space $C^\alpha_x$ (continuous in space only), while the curl ($\nabla \times u$) belongs to $C^{\alpha, \alpha/2}$ (continuous in both space and time).
  • Proof of Sharpness: The authors confirm that their results are perfectly sharp. Because of Serrin's counterexample, it is impossible to prove time-continuity for the velocity gradient under these conditions, making this the strongest possible regularity result for this specific setup.

Limitations and Future Work

  • Limitations: The paper strictly addresses interior estimates (what happens deep inside the fluid domain). As the authors note, because Stokes systems are parabolic with degeneration, global regularity heavily depends on boundary data. The behavior of the fluid near the boundaries of the domain is a completely different mathematical beast and is not covered here.
  • Future Work: While not explicitly detailed in a dedicated section, natural extensions of this work would involve exploring boundary estimates under similarly weak time-continuity assumptions, or applying these directional Campanato iteration techniques to other degenerate parabolic systems (like Navier-Stokes or non-Newtonian fluid models).

Summary updated

Abstract

Interior pointwise $C^{1,α}$ estimates are established for Stokes systems in divergence form where no continuity in time variable is assumed for the coefficients and the given data. The estimates are attained by iteration and are presented by Campanato's characterization. The sharpness of the conclusions can be seen from Serrin's counterexample.
BibTeX
@misc{dong2023interior,
  author = {Rong Dong and Dongsheng Li},
  title = {{Interior pointwise $C^\{1,α\}$ estimates for Stokes systems in divergence form}},
  year = {2023},
  eprint = {2304.03529},
  archivePrefix = {arXiv},
  url = {https://arxiv.org/abs/2304.03529},
}
Figure