Back reflection from stabilized PML boundary


Hello Lumerical,

If I simulate a simple wave propagation through a Si waveguide in varFDTD with stabilized PML boundary I see a significant portion of light reflected back from the boundary. If I set the boundary to standard PML then there is no reflection. Please see the example file and look at the movie. I also have stability factor set to 0.6. The reason I need stabilized PML and 0.6 stab factor is to suppress instability in ring-resonator simulations. Is it possible to set the PML boundary so that it does not reflect the light back but stil provides good stabilization?
WG_reflection.lms (594.6 KB)

General tips to solve divergence problem in metal-insulator-metal (MIM) structures

Comment: I guess the light is actually reflected from the waveguide termination, not from the PML boundary. But still this means that the stabilized PML does not absorb the radiation and lets it propagate back into the simulation region,

Another comment: I guess one option around this problem is to extent the WG to a long distance so the light never comes back within the simulation time window


Hi Eugene,

The stabilized settings for the PML provide stability at the price of less absorption per layer. For that reason you need more PML layers; the default is 42 layers but you can increase this up to 256. In your simulation file I increased the number of layers to 100 and the reflection is drastically reduced: the movie monitor does not show the pulse coming back from the right side.

Regarding the dt stability factor of 0.6 I don’t think that is helping much because the divergence in your simulation seem to be due to the PML. To check if the problem is the PML you can set all the simulation boundaries to metal; if the simulation does not diverge then you know the problem is the PML. This is described in more detail here. I would try the simulation with the ring resonator setting the dt stability factor back to the default (0.99) and increasing the number of PML layers in the stabilized profile.

The waveguide as you have drawn it in the simulation file already extends beyond the simulation region and the PML boundaries. This means that effectively, the waveguide ends are not included in the simulation. The reflection that you see is because at the outer end of the PML the solver always imposes metal boundary conditions; therefore, if the PML has poor absorption, the portion of light reflected by that metal boundary and not absorbed by the PML will come back into the simulation region.


Thank you for your help!