Learning Forum

[Ruzan Sokhoyan]

Hello,

 

I am simulating the transmittance of an array of high-Q resonators. In my simulations, I use auto mesh to reduce the simulation time.  The issue is that the simulation results are sensitive to geometrical parameters. If I find interesting geometrical conditions in simulations and then apply constant mesh override region and restimulate the structure, the results will look completely different. To recover the previously found interesting condition, I would have to tweak the geometrical parameters.  I think the difference stems from the fact that the geometry simulated by Lumerical is different, depending on the meshing.

 

My question is:

 

1.      If I use the refractive index monitor, does it accurately display the factual geometry simulated by the structure?

2.      If I use automesh, and start changing the pillar height by, say, 1.5 nm, the varying height is captured in optical simulations since my results change with height. If I use refractive index monitor, I indeed observe that the overall height of the pillar is roughly modified by 1.5 nm. At the same time, I also observe that the factual thickness from the refractive index monitor is significantly different as compared to the target thickness, which I prescribed. In some cases I observe that the software overlaps two materials in some regions, while I designed them to be placed sequentially (without the material overlap region). Is there a way to still use automesh and get an accurate representation of geometry?

3.      What would be optimal strategies for accurately representing geometry? Should I use a mesh override region over the geometrical structure and make sure that the top and bottom of the pillar are aligned with the mesh grid lines?

4.      When I run parameter sweeps with automesh,  I get results that look jugged. I presume this is because of geometry sampling. Is there a way to avoid this other than adding the mesh override region?

Thank you!

 

 

 

[Ruzan Sokhoyan]

I now see that there is “based on a structure” option available in the mesh override regions. But even if I use the mesh override snapped to the structure, the height value I extract from the refractive index monitor differs from the target value. Why is that the case?

[Kirill]

Hello Ruzan,

The issues you’ve described are related to meshing, a fundamental aspect of the FDTD algorithm. Both the representation of the object and the simulation results strongly depend on the mesh. When using auto-mesh, FDTD aims to maintain a specific number of cells per shortest wavelength in the simulation (within the material). The auto-generated mesh structure changes with geometry modifications, but not every change can be finely resolved by auto-mesh.

The resolution of the index monitor is determined by the mesh size used in the simulation. If you’re viewing the index preview result (before running the simulation), keep in mind that it doesn’t account for conformal meshing or Yee grid offsets. The index preview is optimized for speed and is useful for confirming the basic structure but doesn’t provide an exact representation of what the solver will use. When visualizing the index from the IndexMonitor in Analysis mode, the results depend on the spatial interpolation setting under the Advanced tab.

To achieve reliable results, consider increasing the mesh density and conducting convergence testing.

 Please check follwing references:
1.       Selecting the best mesh refinement option in the FDTD simulation object – Ansys Optics
2.       Understanding Mesh Refinement and Conformal Mesh in FDTD – Ansys Optics
3.       Refractive index monitor - Simulation object – Ansys Optics
4.       Convergence testing process for FDTD simulations – Ansys Optics

Best regards,
Kirill