Hi Steven, I have been trying to put together a source that emits only to one side by combining electric and magnetic sources. If the electric and magnetic sources are identical, then the fields on one side of the sources interfere to cancel out, while on the other side they reinforce each other (as is of course well known and easy to understand from symmetry considerations). However, in meep, it seems that the destructive interference is not quite complete.
Attached is a simple example in one dimension, a gaussian pulse in empty space emitted from the center to the right. It shows the total energy emitted to each side, as well as the flux (it is redundant, I know). As an extra, the .ctl file defines a parameter n, the delay of the magnetic source with respect to the electric source, in units of simulation time steps. Here is an example run: $ meep n=0 1d_electric_and_magnetic_source.ctl command-line param: n=0 ----------- Initializing structure... Working in 2D dimensions. time for set_epsilon = 3.24799e-08 s ----------- energy: 0.00591944535609377 energy: 0.892598771248037 energy (ideal): 0.886226925452758 run 0 finished at t = 20.0 (800 timesteps) flux1:, 1.0, -0.00154548650201008 flux1:, 1.0, 0.247960040697758 flux (ideal): 0.25 Elapsed run time = 0.75 s $ For n=0, the ratio flux-to-the-left:flux-to-the-right is 0.015:0.25, as you can see above. For n=1, the ratio turns out to be 9.6e-6:0.25. Clearly the cancellation is much better for n=1, suggesting that meep evaluates the magnetic source one time step ahead of the electric one. Of course the update steps for the magnetic and electric fields are half a time step apart in the Yee algorithm, and the sources need to be sampled half a time step apart. Could it be that there is a sign problem with this half time step in meep? So the sources are a full step apart instead of coinciding? Best regards, Mischa PS: As a further indication that the electric and magnetic sources are not synchronized for n=0: try increasing the Courant number from 0.5 to 0.99 (so the numerical dispersion is reduced). For n=0, the flux to the left *increases* (about 4x). For n=1, the emitted flux decreases (slightly), as expected for synchronous sources.
1d_electric_and_magnetic_source.ctl
Description: Binary data
_______________________________________________ meep-discuss mailing list [email protected] http://ab-initio.mit.edu/cgi-bin/mailman/listinfo/meep-discuss
