On Aug 9, 2008, at 3:02 PM, Ryan Hao wrote: > 1. about the duplication command in MEEP > How to infinite duplicated the computational cell in one > direction in MEEP?
If I understand your question correctly, the right way to "infinitely" duplicate the computational cell in one direction in Meep is to use periodic boundary conditions. (There is a way to set the boundaries to be periodic in only one direction, if that is what you are asking. But you almost never need this capability in my experience: if you have PML layers in the other "non-periodic" directions, it doesn't matter what their boundary conditions are. Remember that PML is a layer of "material" adjacent to the boundary, and is totally independent of the boundary condition.) > T= S1/S1' (S1 and S1' has the same detector. S1 is hole > structure, S2 is without hole structure) > But in my opinion, the transmission spectrum should define as: > T= S1/S2 (S1 and S2 have different detector. Si is detector > at output, S2 is detector at input) No, you don't want S2 to be the power from a detector at the input. There are two problems with this: First, you *must* do a second "control" simulation without the scatterer(s). Otherwise, even if you put a flux plane at the input, it will include the input power minus the reflected power. Second, with a typical source, you will couple not only to the waveguide mode (or whatever your input channel is), but you will also couple into modes you don't want (e.g. radiating modes). Because of this, you want to put your normalization flux S2 in your "control" simulation far enough away from your source that the radiating/leaky modes will have decayed away. You can see how this worked in the example: the control is the waveguide without any holes, because in that case we want the transmission relative to the power the source couples into the uniform waveguide. But the, just like the field example shown in the tutorial, the source will couple power into radiating modes as well as into the waveguide mode, so your detector needs to be far enough away from the source. (For the same reason, your scattering structure needs to be far enough away from the source that the coupling from radiating modes doesn't play a significant role. If you are wondering how far is "far enough," the way to check is the same way one checks all other convergence issues in numerical methods: double the distance from your source to the scatterer/detector, and make sure the answer doesn't change to within your desired error tolerance.) Now, it is true that your control simulation can probably get away with a smaller computational cell than your simulation with the scatterers. However, usually it's more convenient just to use the same computational cell in both cases. Steven _______________________________________________ meep-discuss mailing list [email protected] http://ab-initio.mit.edu/cgi-bin/mailman/listinfo/meep-discuss
