Hi, Thank you for the quick reply. This is a real problem we have at the moment, and the "coeff" doesn't (as far as we can see) update itself. No, you are right, the specific solver shouldn't have any bearing on this problem I guess.
Joakim -----Ursprungligt meddelande----- Från: [email protected] [mailto:[email protected]] För Guyer, Jonathan E. Dr. Skickat: den 7 april 2014 14:43 Till: FIPY Ämne: Re: Solving the Cahn-Hilliard equation On Apr 7, 2014, at 4:42 AM, Joakim Odqvist <[email protected]> wrote: > If the default solver (trilinos LinearGMRESSolver) is used we find that the > mobility/diffusion coefficient and the thermodynamic factor is only evaluated > at the beginning of the time-step which causes numerical instabilities if > these have a strong variation with composition. In order to have a fully > implicit integration the mobility and the thermodynamic factor need to be > evaluated also at the end of the time-step, taking the composition variation > of the mobility into account. Is this possible ? Do you know that this is a problem or do you worry that it is? I ask because the self.PHI = self._requires(PHI) line *should* cause coeff to update after any change to phi. I won't have any time to do diagnostics until tomorrow at the earliest. Also, I don't know why the particular solver you use would have any bearing on the behavior. _______________________________________________ fipy mailing list [email protected] http://www.ctcms.nist.gov/fipy [ NIST internal ONLY: https://email.nist.gov/mailman/listinfo/fipy ] _______________________________________________ fipy mailing list [email protected] http://www.ctcms.nist.gov/fipy [ NIST internal ONLY: https://email.nist.gov/mailman/listinfo/fipy ]
