I'm not sure, but computations with a dielectric resonance can be numerically unstable if your resolution isn't able to resolve the frequency at which it oscillates. I think Steven has some plans to make these calculations more stable. Looking at the fields for a small number of timesteps may give you some idea of what's going wrong. I haven't looked to see if you've got PML, but that can also lead to interesting situations involving NaNs, although I don't remember when exactly it'd be problematic.
David On Tue, Oct 31, 2006 at 05:58:34PM +0000, Dr Chris Stevens wrote: > Hi David, > I've not actually run one to completion in this form although in an > earlier test I did get NaN for the outputs when using a complex > dielectric. I was wondering if this might be an issue - any idea why it > would be happening ? > > > many thanks for your help > > > Chris > > David Roundy wrote: > > On Tue, Oct 31, 2006 at 05:38:53PM +0000, Dr Chris Stevens wrote: > > > >> Dear all, > >> am still playing with this complex dielectric issue. I am running > >> two ctl files listed below. The first, sea_fake.ctl runs in about 75s > >> having only real dielectrics ie water with zero conductiivty. The second > >> which simulates the same situation but with complex dielectric set by > >> using a seawater conductivitty of 4S/m takes over 200000s to run! > >> I'm using meep which responds to the meep -V command with: > >> > >> Meep 0.10, Copyright (C) 2005 Massachusetts Insitute of Technology. > >> Using libctl 3.0.2 and Guile 1.6.7. > >> > >> > >> Why is there this huge difference ? I've tried asking for complex fields > >> which doesn't make much difference in either case. I'm running for 30000 > >> as I'm using rather low frequeencies but comapred to their free space > >> wavelengths also rather small volumes. > >> > > > > I haven't actually looked at your inputs, but have you checked whether you > > are getting NaNs or infinities with the sea water? NaNs slow everything > > down dramatically. _______________________________________________ meep-discuss mailing list [email protected] http://ab-initio.mit.edu/cgi-bin/mailman/listinfo/meep-discuss
