Unable to reproduce meaningful results in Lumerical


parameters fixed source on si substrate no oxide monitor at 2.81 above source.fsp (607.2 KB)
parameters fixed source on si substrate oxide monitor at 2.81 above source.fsp (526.4 KB)

Hi Everyone,

We have been using Lumericals in the past year to simulate photonic crystal devices, but recently we notices that we are unable to reproduce meaningful results. As an example I have attached two simulation files:
FSP File1: Source oriented in x y z are placed on top of a silicon substrate, source emit at 1.13 micron, monitor above the source records the power spectrum in the z direction
FSP File2: Same as file 1 except a SiO2 layer of quarter wavelength lamda/4n is inserted on top of silicon, designed to have a Bragg reflection at the boundary and constructive interference at the design wavelength of 1.13 micron.

The images have the monitor’s power measurement in Visualizer 1 and the source power in visualiser 2. We notice the following observations that do not make sense:

  1. monitor transmission produces a sloped line, We know from an earlier post that this is because there is a normalization with the source power, but why does the slop change sign when we add a silicon dioxide layer?

  2. Looking at lamda=1.13 on visualizer 1 in both images, The design with bragg reflection (source on SiO2/Si) shows T=0.15 whereas without bragg reflection (Source on Silicon directly) T=0.185! Lumerical seem to tell us that the Bragg reflection does not work!

I am sure there’s something wrong with our simulation parameters but we can’t figure out what that might be, we have tried a lot of different simulations with SiO2/Si that does not seem to make sense at all.

Please help us figure out what the problem is,

Many thanks



Hi Yameng,

If you use a quarter-wavelength layer I think it will act as an anti-reflection coating; therefore, the reflection with the SiO2 will be smaller than without it, just as you found. You can confirm this reduction analitically using a plane wave at normal incidence with the stackrt script command, as I did in this script file: example_quarterlambda.lsf (611 Bytes). The script compares the reflection with and without the quarter-wavelength layer confirming that the destructive interference reduces the reflection. If you increase the SiO2 layer thickness to half the wavelength then you get constructive interference and the reflection is the same as without the SiO2 layer (at the design wavelength 1.13um):

For the simulation with the dipoles I would suggest increasing the size of the power monitor to collect as much reflected light as possible. Also, you can reduce the simulation span (unless you plan to have some additional structures). Finally, you can use a nonuniform mesh and use an override mesh to make sure the SiO2 layer is meshed appropriately. I have made this changes in this file: parameters fixed source on si substrate oxide monitor at 2.81 above source_modFG.fsp (252.8 KB)

Hope this helps!