Electromagnetic Field Simulation of Dielectric Resonators

Simply put, it's a device where electromagnetic waves are confined and resonate inside a high-permittivity ceramic ($\varepsilon_r = 20 \sim 90$). Unlike a cavity resonator surrounded by metal walls, it uses the permittivity difference between the dielectric and air to cause "internal reflection" of the waves.

Yes. It's the same principle of total internal reflection as in optical fibers. The higher the $\varepsilon_r$, the shorter the wavelength becomes—since $\lambda = \lambda_0 / \sqrt{\varepsilon_r}$, the resonator can be made smaller for the same frequency. For example, with a ceramic of $\varepsilon_r = 36$, the size can be about 1/6 that of a metal cavity. They are used for frequency stabilization in filters and oscillators for 5G base stations.

The unloaded Q factor ($Q_u$) is extremely high, ranging from 10,000 to 100,000. Metal cavities are limited to Q factors of a few thousand due to conductor loss, but dielectrics can use low-loss materials with $\tan\delta$ below $10^{-4}$. That's why base station filters can sharply cut off out-of-band signals.

In metal cavities, pure TE and TM modes exist, but DRs have an open structure, so they become "hybrid modes" where both electric and magnetic field components remain in the axial direction. HE modes are hybrid modes where the magnetic field component is dominant, and EH modes are where the electric field component is dominant.

The most widely used is the TE_01δ mode. It is the lowest-order mode that has no axial (z-direction) electric field component $E_z$, and the electric field is distributed in the circumferential direction. The subscript $\delta$ indicates that the axial confinement is incomplete—meaning the electromagnetic field slightly leaks outside the dielectric. It has the highest Q factor and good frequency separation from other modes, making it the standard for filter design.

In dual-mode filters, the double degeneracy (two orthogonal polarizations) of the HE_11δ mode is sometimes used to achieve two resonances with a single DR. This allows the number of elements to be halved, which is valuable for input/output filters in satellite communications.

For cylindrical DRs, there is no exact solution, but Courtney's model or Kajfez's approximate formulas are often used. A typical approximate formula for the TE_01δ mode is this:

$a$ is the cylinder radius, and $h$ is the height. For example, for a DR with $\varepsilon_r = 36$, $a = 5\,\text{mm}$, $h = 4\,\text{mm}$, $f_0 \approx 4.8\,\text{GHz}$. In practice, we use this formula to estimate the initial value and then refine it precisely using FEM.

The Q factor is the ratio of "stored energy" to "energy lost per cycle." In formula form:

For DRs, losses are divided into three parts: dielectric loss $Q_d$, conductor loss $Q_c$ (if there is a metal enclosure), and radiation loss $Q_r$:

Dielectric loss is often dominant, and it can be approximated as $Q_d = 1/\tan\delta$. For a material with $\tan\delta = 10^{-4}$, $Q_d \approx 10{,}000$. When placed in a metal shielding enclosure, $Q_c$ is on the order of tens of thousands, and $Q_r$ can be considered virtually infinite. So often $Q_u \approx Q_d$.

The temperature coefficient $\tau_f$ (unit: ppm/K) evaluates the frequency shift due to temperature change:

Here, $\tau_\varepsilon$ is the temperature coefficient of permittivity, and $\alpha_L$ is the linear expansion coefficient. For practical DR materials, the composition is adjusted to achieve $\tau_f \approx 0\,\text{ppm/K}$. For example, the Ba-Zn-Ta (BZT) system achieves $\tau_f = 0 \pm 2\,\text{ppm/K}$, keeping frequency drift below 0.01% even in environments from $-40°$C to $+60°$C.

Yes. Using multiphysics tools like COMSOL for coupled thermal-electromagnetic field analysis allows evaluation of how local $\varepsilon_r$ variations due to temperature distribution within the enclosure affect the resonant frequency. Especially for high-power filters, the effect of self-heating cannot be ignored, making coupled analysis essential.