Discussion Closed This discussion was created more than 6 months ago and has been closed. To start a new discussion with a link back to this one, click here.
Iterative Solvers for 3D Electromagnetic Waves Problems with PMLs
Posted 2020年4月14日 GMT-4 08:44 Low-Frequency Electromagnetics, Wave Optics, Studies & Solvers Version 5.1 1 Reply
Please login with a confirmed email address before reporting spam
Dear all,
I am having a very hard time with trying to set an iterative solver that does converge for electromagnetic wave problems in 3D when using perfectly matched layers (PMLs). In particular I want to solve the emission of a point dipole into a photonic crystal cavity. I have done a precalculation of the eigenmodes (Eigenfrequency analysis) of my photonic structure, therefore I know at what frequencies my dipole should lie in the Frequency Domain simulation. The eigenfrequency analysis runs well with a direct solver (I think it is not possible to run it with an iterative solver, although that would be EXTREMELY HELPFUL)
The problems arise when I try to solve the Frequency Domain problem with an iterative solver. When run with a direct solver (for small geometric domains that allow for a direct solver in terms of memory requirements), I get no problems and recover a field that looks very much alike my eigenmode (as expected). However I seem to be unable to find any iterative solver that converges; I would say I am even unable to find any iterative solver that does not look more like diverging or oscillating over iterations. Actually this is true when the PML is activated. When it is not and I only have Scattering Boundary Conditions on the outter boundaries, the iterative solver set by default (BiCGStab with Multigrid preconditier) also does converge. So I guess either it becomes very hard for the iterative solver to converge with the PMLs activated, or I am doing something fundamentally wrong... I have to say that I tried solving for the emission of a dipole in air with PMLs and the same problems occur.
I have tried to read through Discussion Forums but have not found any specific threads on specific settings of iterative solvers in the presence of PMLs, so I decided to launch this new thread. Attached a file of the photonic crystal cavity with a dipole emitter in the center of the cavity. Perfect Magnetic or Electric Conductor boundary conditions are set in some boundaries to use the symmetries of the solution. If someone with more experience on using iterative solvers in Comsol for electromagnetism problems could have a look I would be very grateful.
Thanks in advance for the help,
Guillermo
Attachments:
Please login with a confirmed email address before reporting spam
Guillermo,
PMLs should have homogeneous material parameters which is not the case in your model. I tried to set up a second PML only for the silicon parts with user defined stretching only in x- and y-direction as z is probably irrelevant in this domain. It converges then with the default iterative solver. It converges also with a cartesian PML in the silicon, but slower.
I did these tests in the current version 5.5, because I don't have 5.1 installed anymore.
You should carefully check if those PMLs work as expected.
Cheers Edgar
-------------------Edgar J. Kaiser
emPhys Physical Technology
www.emphys.com
Attachments: