High-Frequency FEM

Category: Electromagnetic Field Analysis | Consolidated Edition 2026-04-06
CAE visualization for fem high frequency theory - technical simulation diagram
High-Frequency FEM

High-Frequency FEM: Theoretical Foundations

High-Frequency Electromagnetic Field FEM

🧑‍🎓

Professor, what is the governing equation for solving high-frequency electromagnetic fields with FEM?


🎓

The Helmholtz equation (frequency domain):


$$ \nabla \times \left(\frac{1}{\mu_r}\nabla \times \mathbf{E}\right) - k_0^2 \varepsilon_r \mathbf{E} = 0 $$

$k_0 = \omega/c = 2\pi f/c$: free-space wavenumber. Discretized using edge elements (Nédélec elements), solving a complex symmetric matrix.


🧑‍🎓

That's the method used by HFSS, right?


🎓

Yes. Ansys HFSS is the leading example of frequency-domain FEM. It can analyze complex geometries with high accuracy using tetrahedral meshes + 2nd-order edge elements. Its adaptive meshing feature automatically iterates until S-parameter convergence.


Summary

🎓
  • Helmholtz Equation — Wave equation in the frequency domain
  • Edge Elements — Natural discretization for vector fields
  • Complex Symmetric Matrix — Solved by direct or iterative methods

Coffee Break Yomoyama Talk

The "Numerical Dispersion" Problem in High-Frequency FEM—Errors Accumulate as Wavelengths Shorten

High-frequency FEM has an inherent problem called "numerical dispersion." It's a phenomenon where the phase velocity of electromagnetic waves propagating within the FEM approximation deviates from the true speed of light. This error becomes larger when the mesh is coarser and the frequency is higher (phase error accumulates proportionally to propagation distance). Countermeasures are: ① Use a finer mesh (element edge length below 1/6 to 1/10 of the wavelength), ② Use higher-order elements (2nd/3rd order polynomials), ③ Adopt methods with less numerical dispersion, such as DG-FEM (Discontinuous Galerkin Finite Element Method). In the millimeter-wave band (30–300 GHz), this problem directly impacts design accuracy.

Computational Methods for High-Frequency FEM

FEM Implementation

🧑‍🎓

Please tell me the implementation considerations for high-frequency FEM.


🎓

1. Mesh: 5–10 elements per wavelength $\lambda$ (for 2nd-order elements)

2. Port Settings: Set mode patterns for waveguide ports and extract S-parameters

3. Absorbing Boundary: Model open space with PML (Perfectly Matched Layer)

4. Matrix Solver: Iterative methods (GMRES+preconditioner) for large-scale problems. HFSS uses Domain Decomposition Method (DDM) for domain-decomposed parallelization


🧑‍🎓

How about frequency sweeps?


🎓

HFSS's Fast Frequency Sweep uses rational function approximation to interpolate S-parameters across the entire band from a few direct solutions. It's very efficient because it doesn't need to solve again at each frequency.


Summary

🎓
  • $\lambda/5$–$\lambda/10$ Mesh — For 2nd-order elements
  • DDM — Domain-decomposed parallelization for large-scale problems
  • Fast Frequency Sweep — High-speed via rational function interpolation

Coffee Break Yomoyama Talk

Edge Elements (Nedelec Elements)—Why They Are Essential for High-Frequency FEM

If you use regular node elements (Lagrange elements) for high-frequency electromagnetic field FEM, "spurious modes"—fake eigenmodes that don't physically exist—appear. Edge elements (also called Nedelec elements or Whitney elements) were developed to prevent this. They assign degrees of freedom not to nodes but to element edges, automatically satisfying the ∇·E=0 (zero divergence of electric field) condition. Almost all major high-frequency FEM tools, including HFSS, are based on edge elements. If asked "Why does HFSS use Nedelec elements?", the simplest answer is "to eliminate spurious solutions."

Related Simulators

Experience the theory firsthand with the interactive simulator for this field

All Simulators

Related fields

Coupled AnalysisStructural AnalysisThermal Analysis
Rate this article
Thank you for your feedback!
Helpful
More details
Report error
Helpful
0
More details
0
Report error
0
Written by NovaSolver Contributors
Anonymous Engineers & AI — Sitemap
About the Authors