Weak Formulation of Electromagnetic Field Problems

In the field of computational electromagnetics (CEM), solving partial differential equations (PDEs) in their strong form—the direct differential representation of physical laws—presents significant mathematical and practical hurdles. While Maxwell’s equations provide a precise pointwise description of electromagnetic phenomena, they demand a high degree of smoothness (regularity) from the solution. Specifically, for a second-order PDE, the solution must be twice differentiable throughout the entire domain.

In real-world engineering applications, this requirement is rarely met. Material properties such as relative permittivity ($\epsilon_r$) and permeability ($\mu_r$) often undergo abrupt changes at interfaces between different media. These discontinuities cause the derivatives of the electromagnetic fields to be non-smooth or even undefined at the boundaries. Consequently, the strong form fails to hold at these critical points. To overcome this, we transition to the weak form (also known as the variational or integral form), which provides the mathematical foundation necessary for robust numerical techniques like the Finite Element Method (FEM).

From Strong Form to Weak Form: The Conceptual Shift

The fundamental difference between the strong and weak forms lies in how the equation is satisfied. While the strong form requires the PDE to hold at every single point in the domain, the weak form requires the equation to hold in an integral sense against a set of "test functions."

By introducing test functions and utilizing integration, we effectively "spread" the requirement of the equation over a local area. This process yields two transformative advantages:

  1. Reduced Regularity Requirements: Integration by parts allows us to shift one level of differentiation from the unknown field (the solution) onto the test function. This enables us to seek solutions in broader function spaces, such as Sobolev spaces, which can accommodate the discontinuous derivatives found in multi-material problems.
  2. Natural Handling of Boundaries: The integration process naturally generates boundary integrals, providing a mathematically elegant way to incorporate physical boundary conditions.

The Standard Derivation Procedure

The transformation from a strong-form PDE to a weak-form integral equation generally follows a rigorous four-step workflow:

  1. Weighting: Multiply the strong-form equation by an arbitrary test function $\mathbf{v}$ (or $\mathbf{E}_t$) from a predefined functional space.
  2. Integration: Integrate the weighted equation over the entire computational domain $\Omega$.
  3. Integration by Parts: Apply vector identities or Green’s Theorems to reduce the order of the derivatives. This step is the "engine" of the weak form, as it converts high-order differential terms into lower-order terms and surface integrals.
  4. Boundary Condition Integration: Incorporate the physical constraints of the problem into the resulting boundary integrals.

Mathematical Derivation: The Time-Harmonic Vector Wave Equation

To illustrate this procedure, let us consider the time-harmonic vector wave equation for the electric field $\mathbf{E}$ in a source-free region. The strong form is expressed as:

$$\nabla \times (\mu^{-1} \nabla \times \mathbf{E}) - \omega^2 \epsilon \mathbf{E} = 0$$

where $\mu$ is the permeability, $\epsilon$ is the permittivity, and $\omega$ is the angular frequency.

Step 1: Weighting and Volume Integration

We introduce a test function $\mathbf{E}_t$ and integrate the product of the equation and the test function over the domain $\Omega$:

$$\int_{\Omega} \left[ \nabla \times (\mu^{-1} \nabla \times \mathbf{E}) \right] \cdot \mathbf{E}t , d\Omega - \int{\Omega} \omega^2 \epsilon \mathbf{E} \cdot \mathbf{E}_t , d\Omega = 0$$

Step 2: Integration by Parts

To lower the order of the curl operator ($\nabla \times$), we employ the vector integration-by-parts identity:
$$\int_{\Omega} (\nabla \times \mathbf{A}) \cdot \mathbf{B} , d\Omega = \int_{\Omega} \mathbf{A} \cdot (\nabla \times \mathbf{B}) , d\Omega + \int_{\partial \Omega} (\mathbf{A} \times \mathbf{n}) \cdot \mathbf{B} , d\Gamma$$
where $\mathbf{n}$ is the outward unit normal vector on the boundary $\partial \Omega$.

By setting $\mathbf{A} = \mu^{-1} \nabla \times \mathbf{E}$ and $\mathbf{B} = \mathbf{E}_t$, the first term of our integral becomes:

$$\int_{\Omega} (\mu^{-1} \nabla \times \mathbf{E}) \cdot (\nabla \times \mathbf{E}t) , d\Omega + \int{\partial \Omega} (\mu^{-1} \nabla \times \mathbf{E} \times \mathbf{n}) \cdot \mathbf{E}_t , d\Gamma$$

Recognizing that $\mu^{-1} \nabla \times \mathbf{E}$ is related to the magnetic field $\mathbf{H}$ (via Faraday's Law, $\nabla \times \mathbf{E} = -j\omega\mu\mathbf{H}$), the boundary term can be expressed in terms of the tangential magnetic field.

Step 3: The Final Weak Form

Combining the terms, we arrive at the weak formulation. We seek a field $\mathbf{E} \in \mathcal{V}$ such that for all test functions $\mathbf{E}_t \in \mathcal{V}_0$:

$$\int_{\Omega} (\mu^{-1} \nabla \times \mathbf{E}) \cdot (\nabla \times \mathbf{E}t) , d\Omega - \int{\Omega} \omega^2 \epsilon \mathbf{E} \cdot \mathbf{E}t , d\Omega = - \int{\partial \Omega} (\mathbf{n} \times \mathbf{H}) \cdot \mathbf{E}_t , d\Gamma$$

In this structure, the left-hand side represents the bilinear form $a(\mathbf{E}, \mathbf{E}_t)$, and the right-hand side represents the linear functional $L(\mathbf{E}_t)$.

Classification of Boundary Conditions

In the weak form framework, boundary conditions are categorized into two distinct types based on how they enter the equation:

  • Essential (Dirichlet) Boundary Conditions: These are imposed directly on the solution space. For example, at a Perfect Electric Conductor (PEC) boundary, we require $\mathbf{n} \times \mathbf{E} = 0$. In the derivation, these are handled by restricting the space of trial and test functions $\mathcal{V}$ and $\mathcal{V}_0$ such that they satisfy the condition at the boundary.
  • Natural (Neumann) Boundary Conditions: These emerge "naturally" from the integration by parts process. For instance, if the tangential magnetic field $\mathbf{n} \times \mathbf{H}$ is specified on a boundary, this value is substituted directly into the boundary integral on the right-hand side of the equation.

Conclusion

The transition from the strong form to the weak form is not merely a mathematical exercise; it is a fundamental requirement for modern electromagnetic simulation. By shifting the problem from a pointwise differential requirement to an integral variational requirement, we gain the ability to:

  • Handle Material Discontinuities: Effectively model interfaces where field derivatives are discontinuous.
  • Accommodate Complex Geometries: Utilize integral formulations that are inherently compatible with the unstructured meshes used in FEM.
  • Unify Physical Constraints: Integrate various types of boundary conditions into a single, consistent mathematical framework.

Understanding the derivation and logic of the weak form is therefore indispensable for anyone working in high-fidelity electromagnetic modeling and numerical analysis.