A question regarding PML related divergence in FDTD simulations of a silicon waveguide.

    • and93

      Hi all!

      I am having some problems when simulating (FDTD) a silicon waveguide turn with some subwavelength structures at the side. I am using two ports. One for mode excitation and another to monitor the output. I am using a PML BC. All structures are extended through the PML region. The simulation appears to be running fine and the autoshutoff level drops down to around 1e-4, however it never reaches 1e-5 and at some point starts to rise unlimitedly. When observing the movie monitor, simulation appears physicaly correct and the light appears to be propagating as expected. Long after most of the light appears to be absorbed by the PML, the field around the input port starts to rise unlimitedly causing the simulation to diverge. I tried to change the PML profile to stable and in that case the simulation converges and the autoshutoff level of 1e-5 is reached. However in this case there apears to be a significant back reflection from the PML back into the waveguide that is rendering simulation results nonphysical. This reflection can also be seen on the movie monitor. The simulations take a significant amount of time and I am somewhat time limited. Since I am trying to avoid extensive convergence testing, I decided to ask here first whether anyone would have an insight into my problem. My question is what do you think I should change first to achieve a stable simulation? Any help would be gratly appreciated.


    • Greg Baethge
      Ansys Employee

      Thanks for posting your question and apologies for the delayed response. Diverging simulations can be tricky to troubleshoot. Assuming you already looked at this page, a couple of things you can try:
      As you already mentioned using the stabilized profile helps with the divergence, you can try and increase the number of PML layers to reduce the back reflection (downside is, it will make the simulation larger). Stabilized profile is typically less efficient, that's why y default, we already use a large number of layers (42) but it might not be enough.
      With dispersive material, change the bandwidth of the material fit in the material explorer. Sometimes, this is enough to help with the stability.
      I hope this will help!
Viewing 1 reply thread
  • You must be logged in to reply to this topic.