Hi everyone,
I write this email to know whether any of you has managed to get
symmetries to work for 3D simulations.
After 6 years of working with Meep, we are still unable to get our 3D
simulations to converge using mirror symmetries.
We routinely use symmetries for 2D simulations but we are unable to get
them working for 3D simulations.
We are simulating slabs of PhC with mirror symmetries in x, y and z and
we have double- and triple-checked that both the structure and the
source are preserved while applying these symmetries.
But our simulations are still diverging whereas they work perfectly if
we don't use symmetry or if we just do 2D simulations in the XY or XZ
planes.
Here is the protocol we follow to check that our structure and source
are symmetric:
- set k-point to false,
- switch off symmetries,
- ensure that resolution and cell sizes along x,y,z are such that we end
up with an even number of grid points in any direction where we want to
apply a symmetry,
- output epsilon and hz (our source is hz only),
- load eps.h5 and hz.h5 in Scilab (or Matlab) and check that everything
is symmetric: eps(x,y,z)=eps(-x,y,z)=eps(x,-y,-z). We do the same with
hz (we have triple-checked the phase sign and it works in 2D).
So far, we have managed to get almost perfect symmetry (ie
max_{x,y,z}(abs(eps(x,y,z)-eps(-x,y,z)))~10^-16) and still our
simulations are diverging in 3D whereas they work fine if we restrict
ourselves to 2D simulations in XY or XZ planes.
Have we overlooked something?
Is there a magic trick that we should use?
Is there a bug or a flaw somewhere in Meep that causes this behaviour?
Thank you all in advance for your kind help,
Antoine
(feeling depressed)
_______________________________________________
meep-discuss mailing list
[email protected]
http://ab-initio.mit.edu/cgi-bin/mailman/listinfo/meep-discuss
