On Aug 12, 2008, at 7:06 AM, Juntao Xi wrote:
> In the meep, there is always a transition layer between two different
> epsilon structure.
> For example, there is a slab (epsilon=3) in the free-space. You will
> see a
> transition layer whose epsilon is changing gradually from 1 to 3.
>
> Is that possible to turn this feature offer? It is expected there is
> no
> transition layer.
If you set eps-averaging? to false, it will internally do no smoothing
and the dielectric constant will change abruptly. However, when you
output the epsilon.h5 file, even then you may see some "transition"
pixels. The reason is that, when you output to HDF5, it has to
interpolate from the Yee grid onto a regular grid, and this
interpolation leads to some intermediate epsilon values in the output.
However, I'm concerned that you want this for the wrong reasons.
Remember that Meep is simulating Maxwell's equations on a discrete
(finite-difference) grid, and the goal is to have the results of the
discrete grid converge as quickly as possible with increasing
resolution to the results of the exact Maxwell equations. Since the
"transition layer" you are worried about is only one pixel thick, it
disappears in the limit as you increase the resolution.
In fact, by doing some subpixel smoothing (eps-averaging? = true),
Meep generally *improves* the accuracy. That is, the structure with a
"transition layer" actually gives results *closer* to the exact
results of the discontinuous dielectric than a structure where you
change the discretized epsilon abruptly.
For the technical details of this, see the paper referenced on the
Meep web page (in particular, it depends critically on exactly how the
subpixel smoothing is done):
http://ab-initio.mit.edu/wiki/index.php/Citing_Meep
Steven
_______________________________________________
meep-discuss mailing list
[email protected]
http://ab-initio.mit.edu/cgi-bin/mailman/listinfo/meep-discuss
