Learning Forum

[yuqizhu]

I noticed that the auto shutoff level in job manager is oscillating towards the end of the run, e.g. between 1.7e-5 and 1.2e-5. In my experience with other simulations, this number just decreases monotonically to the set auto shutoff trigger level1e-5. I wander if this is indicative of some sort of numerical error. 

Is this a problem? if so, any advice on how to fix this? thanks in advance

[Ethan Keeler]

Hello, thanks for your question!

This doesn't necessarily imply that there is a numerical error in your simulation. Often when we see an oscillation in the auto-shutoff value, it might mean that there are some resonances arising in your simulation and it can take a longer time for the energy to sufficiently decay to the auto-shutoff level.

There are a few things you can consider:

1. You can check your boundary conditions. If they are PML, then you will want to double check that you are not getting any non-physical reflections from the simulation boundary.
2. You could slightly increase the auto-shutoff from its default value. The oscillations are very near the default level so I don't expect it will have a huge impact on the results, but if the simulation is terminated too early you may see some non-physical ripples in the spectrum.

Ethan

[yuqizhu]

Hi Ethan,

thanks for your reply. I do have mutiple PMLs. I tried to put a time monitor near a boundary and monitor the poynting vector. I suppose non physical reflections should be seen here? what level of oscillation would be considered as due to PML reflections? 

Yuqi

[yuqizhu]

I seem to have trouble uploading images directly... but attached are some time monitors near PMLs in my google drive. In this simulation, auto shuoff was not triggered

time monitor near PML z-

time monitor near PML z+