On 6/5/19 10:24 AM, Tom Mathew wrote: > Is this a foolish question or this group don't entertain discussions ?
Tom -- it's not a foolish question. As J-P already mentioned, it's often a matter of everyone waiting for someone else more qualified to answer. The questions that tend to linger are the ones where nobody seems to have the requisite knowledge because it's a question a bit outside of what we all do. J-P already gave you one answer. Let me try a different perspective that augments his: > I am a novice of dealii, but would like to work on it, if there is > something which I am looking for. I would like to implement Optimized > Restricted Additive Schwarz iterative solver or preconditioner for my > problem, may be by inheriting the dealii linear algebra solver or > iterative solver classes. Will it be possible to access the interface > between the subdomains (interface boundary), so that I can impose > optimized transmission (Robin) conditions for the iterative method. If > not, any possible alternative directions or suggestions are welcomed. I am > working on indefinite Maxwell's equations in complex algebra, so I guess I > will be working with block matrices in dealii. Or if there is an example > or project where the local subdomain matrices has been modified or any > Schwarz methods has been tried can also be wonderful place to start for > me. If you're interested in non-overlapping Schwarz methods (I suspect that's what you are after, since you're interested in "transmission conditions"), then that sort of implies that you are doing domain decomposition (DD) methods in which you decompose the mesh into different partitions. One of the issues you will encounter with deal.II if you do this is that non-overlapping DD methods require you to duplicate degrees of freedom that lie on domain boundaries: each side of the boundary has its own unique degree of freedom, even though both sides' DoFs are located at the same physical point. deal.II doesn't do this well. You can't do this easily on one Triangulation object because that means that your finite element space is discontinuous along the domain boundary. That is, unless you are willing to use discontinuous Galerkin methods to begin with. The only way you can really achieve that the space is discontinuous only along *some* faces, but not *all*, is to use the hp framework and using FE_Nothing. step-46 shows up to this; in that case, we actually solve for different physics on different sub-domains (with the equivalent of "transmission boundary conditions", but there is really nothing that prevents you from solving for the same physics on all sub-domains. Either way, I hope this is pointing you in the right direction. Feel free to ask more questions if you run into trouble. As always, we'd be happy to take any program that implements methods not yet covered in the tutorial, either for the tutorial itself, or for the code gallery. Best W. -- ------------------------------------------------------------------------ Wolfgang Bangerth email: [email protected] www: http://www.math.colostate.edu/~bangerth/ -- The deal.II project is located at http://www.dealii.org/ For mailing list/forum options, see https://groups.google.com/d/forum/dealii?hl=en --- You received this message because you are subscribed to the Google Groups "deal.II User Group" group. To unsubscribe from this group and stop receiving emails from it, send an email to [email protected]. To view this discussion on the web visit https://groups.google.com/d/msgid/dealii/c9900831-07f6-36a0-9009-ce032a876368%40colostate.edu. For more options, visit https://groups.google.com/d/optout.
