Learning Forum

[sagar]

I am simulating with high-index material (~ 3.5) for 3 um wavelength. What is the typical mesh size required for this wavelength where the typical size of the element is ~ 1 um? I used a minimum mesh size of 5 nm and a maximum of 50 nm (custom non-uniform conformal meshing). With this, I observed higher transmission, but when I reduced the maximum mesh size to 25 nm, I observed a significant change in transmission and even with auto non-uniform meshing with a minimum mesh step of 5 nm, I observed a significant change in transmission with absorption (1- R- T) in some wavelength, even though there is no absorbing material presence. So, I can not trust the value generated with a smaller mesh element, but the simulation should be accurate as we are going towards dinner mesh. Another problem with finer meshing is the significant increase in the simulation time as the wavelength is in the 3 um range, effectively increasing the simulation volume. Please suggest some solutions and explain how meshing depends on the wavelength. 

[Guilin Sun]

I guess it is related to PML performance. When mesh size is changed, to have better result, the PML thickness (number of layers multiplied by mesh size in front of PML) should be approximately constant. That said you may need to check the PML thickness (you can use "ruler" in the left tool panel in GUI). 

For a quick simulation, usually mesh accuracy 2 , minimum wavelength inside material divided by 10 is recommended. this is about 3/3.5/10~ 85nm. so for your device 50nm min mesh should be ok.  Of course sometime the result is very senstiive to mesh size. If this is the case, you will need to do a convergence testing:

Convergence testing process for FDTD simulations

Again, when mesh size is changed, please change the number of PML layers to maintain about the same PML thickness. 

Sometimes the conformal mesh may create issues so please use index monitor to check if there is anyhting abnormal.

I strongly suggest that you use the auto-non uniform mesh, and may use override mesh if necessary: /forum/forums/topic/ansys-insight-about-override-mesh-in-fdtd-its-use-and-settings/

Please also reduce the override mesh size when mesh accuracy increases to have a better converging test. 

It is normal that the simulation will be slow when finer mesh is used, since the simulation time will be 16 times longer when the mesh is reduced to half of the previous mesh size.