Solving Maxwell's Equations with Mimetic Methods

1. Problem Statement

The paper addresses the numerical solution of Maxwell's equations for computational electromagnetics. While the standard Finite-Difference Time-Domain (FDTD) method (specifically the Yee scheme) is widely used due to its simplicity and explicit time-marching nature, it suffers from specific limitations:

• Accuracy: It is inherently limited to second-order spatial accuracy. Extending it to higher orders is difficult and often compromises the underlying physics.
• Boundary Issues: Standard finite-difference and Summation-by-Parts (SBP) operators typically suffer from reduced accuracy near domain boundaries.
• Conservation: While the Yee scheme preserves discrete divergence conditions initially, extending this to higher orders or complex coordinate systems (e.g., curvilinear) often requires ad-hoc divergence correction methods.

The goal is to develop a numerical framework that maintains high-order accuracy uniformly across the domain (including boundaries) while strictly preserving the fundamental vector calculus identities and conservation laws of the continuous Maxwell equations.

2. Methodology

The author proposes a Mimetic Finite-Difference approach using the Mimetic Operators Library Enhanced (MOLE). This method constructs discrete operators (gradient, divergence, curl, Laplacian) that satisfy the same algebraic identities as their continuous counterparts.

Core Mathematical Framework

• Mimetic Operators: The method defines discrete operators $G$ (gradient), $D$ (divergence), and $C$ (curl) that satisfy identities such as $D \cdot G = L$ (Laplacian) and $D \cdot C = 0$.
• Discrete Gauss Theorem: A key feature is the satisfaction of a discrete analogue of the extended Gauss divergence theorem:
  $$\langle D\nu, f \rangle_Q + \langle Gf, \nu \rangle_P = \langle B\nu, f \rangle$$
  This ensures that conservation properties (like charge conservation) are preserved at the discrete level.
• Staggered Grids: Similar to the Yee grid, the method uses staggered grids where scalar quantities (fields) are stored at cell centers/boundaries and vector components at cell edges.
• Uniform Accuracy: Unlike standard finite differences, the Corbino–Castillo (CC) operators used here achieve uniform order of accuracy (e.g., second-order) throughout the entire domain, including the boundaries, by solving Vandermonde systems.

Implementation Strategy

• 1D Formulation: Uses $E_x$ and $H_y$. The update equations are formulated as sparse matrix-vector products:
  $$E^{n+1} = E^n - \Delta t \frac{1}{\varepsilon_r} D H^{n+1/2}$$
  $$H^{n+3/2} = H^{n+1/2} - \Delta t \frac{1}{\mu_r} G E^{n+1}$$
  Material properties (permittivity $\varepsilon$, conductivity $\sigma$) are applied via element-wise scaling coefficients ($c_a, c_b$) rather than modifying the operator matrices.
• 2D Formulation (TMz): Uses $E_z, B_x, B_y$. 2D operators are constructed via Kronecker products of 1D operators.
• Boundary Conditions:
  • Lossy Media: Implemented via semi-implicit discretization of conduction currents, introducing damping coefficients.
  • Absorbing Boundaries: Uses Uniaxial Perfectly Matched Layers (UPML). The PML is implemented by adding artificial conductivity profiles ($\sigma_x, \sigma_y$) that ramp up smoothly. The damping is applied via Hadamard products with the field vectors, keeping the core mimetic operators unchanged.

3. Key Contributions

1. Operator-Based Formulation: Replaces complex index-level update loops with sparse matrix-vector multiplications. This simplifies the code structure, making it shorter and less error-prone, regardless of dimensionality or order of accuracy.
2. Uniform Boundary Accuracy: The use of Corbino–Castillo operators eliminates the accuracy degradation typically found at boundaries in standard FDTD or SBP schemes, preventing boundary errors from polluting the entire solution.
3. Discrete Conservation: By satisfying the discrete extended Gauss divergence theorem, the method guarantees the preservation of fundamental physical conservation laws, leading to superior long-term stability.
4. Modularity and Extensibility:
   • Changing the order of accuracy (e.g., from 2nd to 4th or 6th order) requires only changing a single parameter in the MOLE library.
   • Material properties and boundary conditions (like UPML) are decoupled from the operator construction, allowing for easy implementation of complex media without altering the core solver.

4. Results

The paper validates the methodology through two numerical examples:

• 1D Example: A sinusoidal wave (700 MHz) interacting with a lossy dielectric slab ($\varepsilon_r=4, \sigma=0.04$ S/m). The simulation successfully demonstrates the exponential attenuation of the wave within the lossy region and the correct phase velocity reduction ($\lambda \propto 1/\sqrt{\varepsilon_r}$).
• 2D Example: A Gaussian pulse radiating from the center of a $100 \times 100$ domain, absorbed by a UPML layer. The simulation shows the pulse propagating outward and being absorbed by the PML without significant reflection, demonstrating the effectiveness of the separable damping approach combined with mimetic operators.

5. Significance

This work provides a principled alternative to classical FDTD for computational electromagnetics. Its significance lies in:

• Physical Fidelity: It ensures that numerical solutions respect the topological and conservation properties of Maxwell's equations, which is critical for long-time simulations.
• Scalability: The framework is easily extensible to higher-order accuracy and complex geometries (curvilinear grids) without rewriting the core update logic.
• Practicality: The open-source MOLE library implementation demonstrates that these advanced mathematical concepts can be translated into efficient, compact, and robust code for engineering applications.

The paper concludes that while the presented 1D and 2D examples utilize gradient and divergence operators (as curl reduces to scalar derivatives in these specific polarizations), the framework is fully capable of handling 3D problems using the mimetic curl operator, paving the way for future work in 3D and curvilinear geometries.