On Tue, 1 Jan 2008, Simon J. Bale wrote: > I'm interested in using meep to model the resonant frequencies and field > patterns inside a high Q cylindrical cavity resonator in 3 dimensions. > I'm totally new to meep and I've had a read through the tutorial and > documentation and I'm concerned that I may not be able to accurately > model the metal end wall and side wall losses. Is it possible to specify > a lossy metal over a wide bandwidth in meep and if so are there any > examples?
You can specify, in principle, just about any loss spectrum by fitting to a series of Lorentzian functions in the form described in the Meep manual. Of course, if you have a resonant cavity, then only the absorption losses at the resonant frequencies matter, so I'm not sure why you're concerned about losses over a wide bandwidth. A more efficient alternative, if the losses are small (which should be the case if you have a resonant cavity with a reasonable Q), is to model the cavity using perfect metals, get the field pattern of the resonant mode, and then compute the absorption losses a posteriori using perturbation theory. (I seem to recall that Jackson includes an example of how to compute absorption losses in metallic cavities/waveguides using perturbation theory.) This should be essentially exact if the absorption losses are small (Q > 100 or so), and has several advantages -- you can model different metals with no additional calculation, and you may not need as high a resolution (since you do not need to numerically model the finite skin depth of the field in the metal). [In general, I highly recommend perturbative methods as a supplement to brute-force numerics whenever you are trying to compute small effects (such as absorption loss in a low-loss cavity), as for such problems they can be far more powerful and revealing than numerics. However, they require a bit more care, so it is important to do a simple test case or two to be sure you are applying them correctly.] > I'm also interested in modelling the quality factor of some very high > Q distributed Bragg structures but again metal loss is important. Will > meep be a suitable fit for this type of problem? Yes, probably. See above. (Although if by "Bragg structure" you mean a multilayer film, then transfer-matrix methods are far more efficient for that specific geometry.) Regards, Steven G. Johnson _______________________________________________ meep-discuss mailing list [email protected] http://ab-initio.mit.edu/cgi-bin/mailman/listinfo/meep-discuss
