On the discrete Poincaré inequality for B-schemes of 1D Fokker-Planck equations in full space

Lei Li, Jian-Guo Liu, Zhen WangarXiv 2025 · DOI · arXiv

Abstract

This paper establishes a rigorous mathematical foundation for the long-term stability of numerical approximations to the Fokker-Planck equation (FPE). By proving the discrete Poincaré inequality for a broad class of finite volume methods known as B-schemes—including the widely used Scharfetter-Gummel scheme—in an unbounded 1D domain, the authors demonstrate that these numerical solutions converge to their equilibrium state exponentially fast. This result bridges a critical gap between the continuous physical theory of statistical mechanics and the discrete methods used in computational science.

Introduction

The Fokker-Planck equation is a cornerstone of statistical physics, describing the time evolution of probability density functions under the combined effects of deterministic drift and stochastic diffusion. A fundamental characteristic of these systems is their tendency to settle into a steady-state equilibrium. In continuous analysis, the rate of this convergence is typically proven to be exponential using the Poincaré inequality, provided the underlying potential is sufficiently well-behaved.

However, real-world applications in biology, finance, and physics require solving these equations on computers through discretization. A major challenge in numerical analysis is ensuring that the discrete version of the equation inherits the stability and convergence properties of the continuous model. While this has been established for bounded domains, proving such properties for "full space" (infinite domains) is significantly more difficult due to the potential for probability mass to "leak" toward infinity. This research addresses this difficulty for B-schemes in a one-dimensional setting.

Methodology

The authors analyze the discretized Fokker-Planck equation by treating it as a continuous-time Markov chain on a discrete grid. To establish the discrete Poincaré inequality, they propose two distinct mathematical frameworks:

1. Gamma-calculus (Bakry-Émery Criterion)

This approach adapts tools from differential geometry to the discrete setting. It focuses on a "curvature" condition: if the physical potential is sufficiently convex (resembling a bowl-shaped energy landscape), the numerical scheme will naturally drive the system toward equilibrium. The authors prove that for B-schemes, a discrete version of this curvature condition holds, ensuring stability for any grid size $h$.

2. Lyapunov Function Method

For cases where the potential might not be globally convex, the authors employ a "local-to-global" strategy. They first prove that the Poincaré inequality holds on a small, local patch of the grid. They then utilize a Lyapunov function—a mathematical tool that tracks the "distance" from equilibrium—to demonstrate that the stability on that local patch effectively controls the behavior of the entire infinite grid. This method is particularly robust for handling the complexities of an unbounded domain.

Key Insights and Results

The paper provides several critical advancements in the understanding of numerical FPE solvers:

  • Discrete Consistency and Detailed Balance: The authors show that B-schemes are structured to preserve the "detailed balance" of the original physical system. This symmetry is essential for applying Markov chain theory to numerical analysis.
  • Theorem 1.2 (Global Convexity): The research proves that if the physical potential is strongly convex and the numerical flux function ($B(s)$) is convex, the discrete Poincaré inequality holds for all grid sizes. This provides a universal guarantee of stability for these configurations.
  • Theorem 1.5 (General Potentials): The authors extend their findings to potentials that are not perfectly convex everywhere. As long as the potential behaves correctly at infinity, the inequality holds for sufficiently fine grids (small $h$).
  • Validation of the Scharfetter-Gummel Scheme: These results provide a formal mathematical justification for the Scharfetter-Gummel scheme, confirming that its widespread use in long-term simulations is supported by rigorous convergence guarantees.

Conclusion and Future Directions

By proving that discrete B-schemes exhibit exponential convergence to equilibrium in full space, this work ensures that numerical simulations remain faithful to the underlying physics of the Fokker-Planck equation. However, several avenues for future research remain:

  1. Higher Dimensions: The current proofs are limited to 1D. Extending these results to 2D or 3D is necessary for complex fluid dynamics and plasma physics applications.
  2. Non-Uniform Grids: The study assumes a uniform grid spacing. Proving similar inequalities for adaptive grids, which are common in engineering, remains an open challenge.
  3. Potential Smoothness: The current framework requires relatively smooth potentials; future work may address systems with sharp or discontinuous energy landscapes.

In summary, this paper provides the theoretical "stamp of approval" for a class of numerical methods, ensuring that the long-term behavior observed in simulations is a true reflection of physical reality rather than a numerical artifact.

Summary updated

BibTeX
@misc{li2025discrete,
  author = {Lei Li and Jian-Guo Liu and Zhen Wang},
  title = {{On the discrete Poincaré inequality for B-schemes of 1D Fokker-Planck equations in full space}},
  year = {2025},
  eprint = {2507.03941},
  archivePrefix = {arXiv},
  url = {https://arxiv.org/abs/2507.03941},
}
Figure