On Tue, Apr 10, 2012 at 3:39 PM, Daniel Wheeler <[email protected]> wrote: > On Mon, Apr 9, 2012 at 1:02 PM, James Snyder <[email protected]> > wrote: >> >> I'm interested in looking at some problems that involve time harmonic >> versions of Maxwell's equations in looking around at how these have >> been approached in finite element models in the past, I've run into a >> few issues: > > > FiPy is not currently set up to do this effectively. The next release of > FiPy will allow the implementation of hyperbolic vector equations of the > form, > > \partial_t \phi^i + \partial_j \left( A^{ij}_k \phi^k \right) = 0 > > where the $A^{ij}_k$ is a coefficient matrix that might be dependent on > $\phi$. An effective solution to that equation is what is required to solve > Maxwell's correctly. The next version of FiPy will allow the problem to be > posed, but will solve it only in a naive way. > > I am currently working on a less naive approach to solving general coupled > hyperbolic equations using Riemann flux updates which will enable an > effective solution of Maxwell's equations in FiPy. Most likely that will be > part of version 3.1 (and we haven't released 3.0 yet). I'm planning on > including a Maxwell's equations example as part of this although I haven't > done so yet. The branch is > > <http://matforge.org/fipy/browser/branches/riemann> > > You can take a look at progress but I can't provide any support right now.
Are things in the branch functional or stable enough that it might be worth attempting something like this at the moment or should I hold off? Also, is the documentation or are examples updated for this functionality where I might be able to at least experiment with some things? Would you be interested in bug reports if it's something that's ready to be experimented with? Otherwise, I'm glad to hear that the functionality is coming. > > >> The first of which is how to represent a curl operator. Presumably it >> might be best to add the functionality, perhaps based on existing >> gradient functionality? I suppose that there are workarounds using >> the existing API, but if they're part of an equation I suspect that >> might get more complex than desirable. > > > You'll have to wait for vector equations to do this correctly. Anything else > will be a hack that probably won't work properly. > > Cheers > > -- > Daniel Wheeler > > _______________________________________________ > fipy mailing list > [email protected] > http://www.ctcms.nist.gov/fipy > [ NIST internal ONLY: https://email.nist.gov/mailman/listinfo/fipy ] > -- James Snyder Biomedical Engineering Northwestern University http://fanplastic.org/key.txt ph: (847) 448-0386 _______________________________________________ fipy mailing list [email protected] http://www.ctcms.nist.gov/fipy [ NIST internal ONLY: https://email.nist.gov/mailman/listinfo/fipy ]
