Dear Meep Community,
I am performing 3D FDFD simulations with meep-mpi in order to get T and R
spectra of a photonic structure.
Here is a sketch of my simulation domain:
/ ------ / |
/ Trans/| |
/ / | /
/ / /
/ Refl / /
/-- ---- / / sz
/____ /| /
| | | /
sy | | /
Adiabatic absorbers are put at the two z-extremities and PBC at the 4
other faces. My control file outputs complex frequency dependent fields in
reflection and transmission planes. Therefore their surface are sx x sy
sizes. In my case, sx = 1 um and sy = 1.73205 um. Finally, resolution is
set to 40. The fields that I get on these planes have 42X72 dimensions. I
do not understand why.
Indeed, each field is output at the center of a voxel.
Along x, if voxels are all entirely comprised in the simulation domain, 40
fields should be output for a given y. If voxels are shifted by half the
spatial step size delta (= 1/resolution) along x, for a given y, then 41
fields should be output because 2 fields would lie on the boundaries ( 1
at the left side and 1 at the right side).
Along y, 69.282 voxels should be generated, which is not possible.
Therefore, it should be 70 voxels but not all entirely comprised within
the boundaries because sy < 70*delta. The same reasoning as before yields
either 70 or 71 fields for a given x.
Hence, there are extra-fields that are output in both cases.
Moreover, fields take the same value at some voxels close to each boundary
(of course, field symmetry is perfectly normal):
_ Along x, for a given y: [E1 E1 E2 E3 . E3 E2 E1 E1]
_ Along y, for a given x: [E1 E2 E1 E3 . E3 E1 E2 E1]
Why is that so ? Is it linked to the non-integer sy, to the Bloch PBC or
to the chunk structure ?
Which fields should I remove to perform Poynting vector integration on
these surfaces, the most extreme E1 fields ?
meep-discuss mailing list