A new numerical method for div-curl systems with low regularity assumptions

Shuhao Cao, Chunmei Wang, Junping Wang

TL;DR

This paper introduces a Primal-Dual Weak Galerkin (PDWG) method to solve div-curl systems, which are foundational to electromagnetics and fluid dynamics. The method stands out by providing accurate solutions even when the underlying physics is "rough" (low regularity) and remains robust regardless of the domain's topological complexity (e.g., holes or handles).

What Problem Does This Paper Address?

Div-curl systems—equations involving the divergence and curl of a vector field—are ubiquitous in science, appearing in Maxwell’s equations for electromagnetism and the velocity-pressure relations in fluid flow.

Two major hurdles often break standard numerical solvers:

  1. Low Regularity: In many real-world scenarios, such as domains with sharp corners or heterogeneous materials, the solution isn't "smooth." Most traditional finite element methods require high levels of regularity (smoothness) to guarantee convergence.
  2. Topological Complexity: If a domain has holes (like a torus), the solution to a div-curl system is not unique; it is only unique up to a "normal harmonic vector field." Accurately capturing these harmonic fields usually requires complex, specialized basis functions that account for the domain's cohomology.

The authors aim to create a unified numerical framework that handles these "rough" solutions and complex shapes without needing to manually "patch" the math for different topologies.

Methods

The authors employ the Primal-Dual Weak Galerkin (PDWG) finite element technique.

Key components include:

  • Weak Operators: Instead of standard derivatives, the method uses "weak gradient" and "weak curl" operators. These are defined using a distribution-based approach on a combination of polynomials inside mesh elements and on their boundaries.
  • Primal-Dual Formulation: The system is treated as a constrained optimization problem. The "primal" variable is the vector field $\mathbf{u}$, while "dual" variables (Lagrange multipliers) are introduced to enforce the div and curl constraints.
  • Stabilization: The algorithm includes stabilizer terms ($s_1, s_2$) that penalize discontinuities across element boundaries, ensuring the numerical scheme remains stable and well-posed.
  • Extended Formulation: They derive an "extended" weak formulation that involves no partial derivatives on the vector field $\mathbf{u}$ itself, which is the secret sauce for handling low-regularity data.

Datasets

As a theoretical and computational mathematics paper, the "datasets" consist of seven rigorous numerical test cases designed to stress-test the algorithm:

  • Example 1: A smooth solution on a unit cube.
  • Examples 2-4: Singular solutions on non-convex domains and domains with cavities (testing low regularity).
  • Examples 5-6: Toroidal domains with one and two holes (testing topological robustness).
  • Example 7: Direct computation of discrete normal harmonic vector fields.

Key Insights and Ideas

  • Topology-Agnosticism: The most striking insight is that the PDWG method naturally approximates the space of normal harmonic vector fields. You don't need to know the Betti numbers (the number of holes) of your domain beforehand; the method "discovers" the necessary harmonic components through its auxiliary variables.
  • Derivative-Free Primal Variable: By shifting the derivative burden to the test functions in the weak formulation, the method can solve for vector fields $\mathbf{u}$ that are barely integrable (specifically, $H^\alpha$ for $\alpha > 0$), where traditional methods would fail to converge.
  • The "Harmonic Difference": The authors show that the difference between the $L^2$ projection of the true solution and the numerical solution ($\mathbf{Q}_h\mathbf{u} - \mathbf{u}_h$) is itself an effective approximation of a harmonic field.

Key Results

  • Optimal Convergence: Across all tests, including those with singular solutions (low regularity), the method achieved optimal convergence rates. For the lowest-order element ($k=0$), the error decreased linearly with the mesh size ($O(h)$).
  • Singularity Handling: In Example 3 (a domain with a re-entrant corner), the method maintained a convergence rate of approximately $h^{2/3}$, matching the theoretical maximum possible for that specific level of "roughness."
  • Multi-Hole Robustness: In Example 6 (a torus with two holes), the method successfully resolved the complex field patterns where the solution is non-unique, demonstrating its ability to handle high-genus topologies.

Limitations and Future Work

  • Boundary Conditions: This specific paper focuses on normal boundary conditions ($\varepsilon \mathbf{u} \cdot \mathbf{n} = \phi_1$). While the authors have touched on tangential conditions in previous work, a unified implementation for mixed boundary conditions remains a practical next step.
  • Computational Overhead: The primal-dual approach introduces several auxiliary variables ($\lambda_h, q_h, s_h$). While this makes the method robust, it increases the size of the algebraic system compared to simpler, less stable methods.
  • Parameter Sensitivity: The stabilizers rely on parameters ($\rho_i$). While the authors suggest these are at the "user's discretion," further study on automated parameter selection for highly irregular meshes would be beneficial.

Summary updated

Abstract

This paper presents a numerical method for div-curl systems with normal boundary conditions by using a finite element technique known as primal-dual weak Galerkin (PDWG). The PDWG finite element scheme for the div-curl system has two prominent features in that it offers not only an accurate and reliable numerical solution to the div-curl system under the low $H^α$-regularity ($α>0$) assumption for the true solution, but also an effective approximation of normal harmonic vector fields regardless the topology of the domain. Results of seven numerical experiments are presented to demonstrate the performance of the PDWG algorithm, including one example on the computation of discrete normal harmonic vector fields.
BibTeX
@article{cao2022new,
  author = {Shuhao Cao and Chunmei Wang and Junping Wang},
  title = {{A new numerical method for div-curl systems with low regularity assumptions}},
  journal = {Comput. Math. Appl},
  year = {2022},
  doi = {10.1016/j.camwa.2022.03.015},
}

Loading related content...

Figure