On Fri, Oct 24, 2008 at 11:40 AM, Roy Stogner <roystgnr@...> wrote:
> On Fri, 24 Oct 2008, M. Adil Sbai wrote:
>> Just would like to know if LibMesh does have built-in support for spectral
>> finite elements ?
> Yes and no. The hierarchic FE type (on at least some geometric
> element types) can in theory be p-refined (or adaptively p-refined) to
> an arbitrary p... but with the current implementation, you're not
> going to get very far (less than p=10 in 3D, IIRC) before your
> solutions are swamped by floating point error (either from the shape
> function evaluations or the system ill-conditioning, I can't say
There's also the Berenstein polynomials (the SZABAB FE type) which go
up to arbitrary order. From what I've read I think they are better
about conditioning than Hierarchics.