Discussion Forum

Hi All,

I am using the RF module to simulate SiC pillars supporting surface modes. I am exciting modes in the pillar utilising an electric point dipole source placed close to the vertex of the cylinder and wish to evaluate the outgoing Poynting vector on a surface enclosing the point dipole and the cylinder. My simulation domain is bounded with Floquet PB conditions on two sides and PMLs with scattering boundary conditions on the other two.
I am having a problem achieving reliable and consistent solutions. There is a huge sensitivity to mesh density in the SiC and in the region around the dipole. I've tried adapting the mesh manually in these regions and the electric field in the cylinder fluctuates wildly. I have also tried to use adaptive meshing to achieve convergence in the observables. This doesn't seem to work as expected. Initially I tried utilising the integrated power observable, as the mesh refinement proceeds this increases exponentially and does not converge before the solver fails. I have also tried to use the electric field evaluated at a point within the pillar. As mesh refinement proceeds the oscillations in the field amplitude blow up in line with the power and convergence is not achieved. I have also tried to refine the mesh utilising the L2 norm error estimation. This appears to succeed (the solver exits without failing) however when plotting the observables over the course of the refinement it is clear that the degree of error in the observables has not been altered by the refinement (oscillation is still of the same magnitude).

Does anyone have any idea how to achieve mesh convergence in this kind of problem? At the moment I have no trust in any of the results I am extracting.

Cheers,
Chris