autoshutt doesn't reach and showing wrong R and T spectra.

vishal_vashisthavishal_vashistha Member Posts: 1

Why this simple simulation doesn't converge even for a large simulation time of 3000 fs? I notice that it never reaches to auto shut level of 1e-5. Because of this, it shows me R and T both value close to 1 which is non-sense. R plus T must be equal to one here.

Even I try to add extra mesh but it never reached to 1e-5 shut level.

Also earlier, there was a time monitor which shows how the signal decays over time on a femtosecond scale, now if I look through the time monitor, it shows the E-fields over wavelength, how does it make sense to understand the signal convergence?

PS: This new page is even strange, I can't even upload the Lumerical file in its original format, I wish I could go back to previous KX support which was excellent.



  • kjohnsonkjohnson Posts: 230Ansys Employee

    Hello @vishal_vashistha ,

    For some simulations (particularly simulations involving resonant structures), 3000 fs may not be a long enough time for the autoshutoff threshold to be reached. It's possible you may have to simply increase the simulation time for this simulation. It would be helpful if I could see the structure you are simulating, but Ansys staff are not permitted to download files from the ALF (as discussed in this post: Can you please include some screenshots of your simulation and structure?

    Regarding the time monitor results, the electric field as a function of time is given as the "E" result:

    In your screenshot it looks like you're viewing the "spectrum" result, which will give the Fourier transform of the time domain field results.

  • vishal_vashisthavishal_vashistha Posts: 2Member

    @kjohnson , Thanks for reply.

    I think you are right. This is a strong resonance structure having an air hole in a 1-micron Si substrate (see xy view). When I see the E-field spectrum, it has a long tail even for 5000 fs but it does solve to give correct T and R values. But, this takes too much time, I wonder if we can do it faster because I am doing a sweeping analysis for period and hole.

    I suppose we need some better PML to absorb the energy quickly so that the field decay soon.

    I try to manipulate the PML as step PML and stabilized PML but the simulation time is more or less the same with a long heading tail of E-field.

  • kjohnsonkjohnson Posts: 230Ansys Employee

    Hello @vishal_vashistha ,

    This is a symmetric structure and source, so you could use symmetry boundary conditions to speed up this simulation. With this source polarization you could use anti-symmetric BCs in the x direction and symmetric BCs in the y direction (assuming the substrate is in the XY plane, I can't tell from your screenshot). Make sure you enable "allow symmetry on all boundaries" and set the symmetry BCs on both sides so the simulation is also periodic:

    There is more information on symmetry BCs on this page:

    Also, for a periodic simulation, you generally should use the "steep angle" PML profile, to absorb light in higher grating orders.

    I am a bit surprised that the T and R results are both close to one, however. I would expect that they should be reasonable values, even if you end the simulation early. Can you please share a screenshot showing the placement of your monitors in the simulation?

Sign In or Register to comment.