Spatial Sobolev regularity for stochastic Burgers equations with additive trace class noise

Arnulf Jentzen, Felix Lindner, Primož PušnikNonlinear 2021 · DOI · arXiv

Abstract

This paper investigates the spatial Sobolev regularity of mild solutions to stochastic Burgers equations driven by additive trace class noise. By employing a combination of spectral Galerkin approximations and recursive bootstrap-type arguments, we demonstrate that these solutions exhibit significantly higher regularity than those driven by space-time white noise. Specifically, we prove that the solutions possess spatial regularity up to Sobolev order 2, providing a rigorous mathematical foundation for the smoothing effects of spatially correlated noise on non-linear partial differential equations.

Introduction

The Burgers equation is a fundamental non-linear partial differential equation (PDE) used extensively to model phenomena in fluid dynamics, gas flow, and traffic modeling. When augmented with a stochastic term, the resulting Stochastic PDE (SPDE) allows for the modeling of systems subject to external uncertainty or turbulence.

Historically, research in this area has focused heavily on systems driven by space-time white noise. While mathematically rich, white noise is spatially uncorrelated and extremely "rough," which limits the spatial regularity of the resulting solutions—typically to a Sobolev order of less than 1/2. This paper shifts the focus toward "additive trace class noise," which possesses inherent spatial structure and correlation. The central challenge addressed is determining the extent to which this smoother noise translates to smoother solutions, particularly in the presence of the Burgers equation's aggressive quadratic non-linearity ($v \cdot v'$), which naturally tends to produce shocks and steep gradients.

Methodology

The study utilizes the semigroup (mild solution) framework, treating the SPDE as an evolution equation within a Hilbert space. The proof of higher-order regularity rests on three methodological pillars:

1. Spectral Galerkin Approximations

To establish existence and uniqueness, the authors project the infinite-dimensional SPDE onto a finite-dimensional subspace. This transformation results in a system of random ordinary differential equations (ODEs). By proving the stability and convergence of these approximations, the authors provide a pathway to analyze the full infinite-dimensional system.

2. Bootstrap-type Arguments

The core of the regularity proof is a recursive "bootstrap" strategy. The authors begin with a known lower-order bound on the solution's regularity. By substituting this bound back into the equation and leveraging the smoothing properties of the Laplacian operator, they derive a higher-order bound. This process is repeated until the regularity is established up to order 2.

3. Detailed Non-linearity Analysis

The quadratic term $F(v) = -v'v$ is the primary obstacle to regularity. The authors perform a rigorous analysis of this operator across various interpolation spaces, establishing specific local Lipschitz conditions. This ensures that the smoothing effect of the diffusion term (the Laplacian) effectively counteracts the roughening effect of the non-linearity.

Key Insights and Results

The Regularity Threshold

The research identifies a precise relationship between the smoothness of the driving noise (denoted by $\beta$) and the resulting smoothness of the solution ($\gamma$). The findings show that if the noise resides in a Sobolev space of order $\beta$, the solution can achieve a regularity order $\gamma$ up to $\min\{1, 1/2 + \beta\}$. Under trace class assumptions, this allows the solution to reach Sobolev order 2.

Pathwise Solvability

A significant contribution of the paper is the demonstration of pathwise solvability. The authors show that for almost every realization of the random noise, the resulting equation behaves predictably. This pathwise approach simplifies the transition from finite-dimensional ODE approximations to the full SPDE solution.

Existence and Convergence

The paper culminates in Theorem 1.1 (and 5.10), which establishes the unique existence of mild solutions for the stochastic Burgers equation with additive trace class noise under zero Dirichlet boundary conditions on the unit interval $(0, 1)$. Furthermore, the proof provides a rigorous basis for the convergence of spatial spectral Galerkin approximations, which is essential for the development of reliable numerical simulations.

Conclusion and Future Directions

This work provides a substantial improvement over existing regularity results for the stochastic Burgers equation, moving from the rougher regime of white noise to the smoother regime of trace class noise. However, several avenues for future research remain:

  • Multiplicative Noise: This study focuses on additive noise. In many physical systems, noise is multiplicative (state-dependent), which introduces significantly higher analytical complexity.
  • Boundary Conditions: The current results are specific to zero Dirichlet boundary conditions; exploring periodic or Neumann boundaries may yield different regularity profiles.
  • Higher Dimensions: While this study is confined to the 1D unit interval, extending these high-regularity proofs to 2D or 3D models, such as the stochastic Navier-Stokes equations, remains a primary goal in the field of stochastic fluid dynamics.

Summary updated

BibTeX
@inproceedings{jentzen2021spatial,
  author = {Arnulf Jentzen and Felix Lindner and Primož Pušnik},
  title = {{Spatial Sobolev regularity for stochastic Burgers equations with additive trace class noise}},
  booktitle = {Nonlinear},
  year = {2021},
  doi = {10.1016/j.na.2021.112310},
}
Figure