[Dune] Re: from Sreejith (AG Bastian)

[Oliver Sander]

Hi Sreejith!
It is correct that I work with Dirichlet boundary conditions on a
regular basis.  There's not much support for it in Dune; I ended
up writing a few tools for myself.

I treat Dirichlet nodes by carrying around a BitField with an entry
for each degree of freedom.  Bits are set for Dirichlet dofs and
unset otherwise.  There is such a Bitfield for each grid level.

There was some talk about incorporating those bitfields into
the function spaces, but nothing happened so far.

For initializing the Dirichlet field I use two methods:
- read it from a file for the coarse grid and then propagate it
   to the higher levels.  I have written code for that.
- Describe the Dirichlet field by some analytical function
   and sample it.  It can be helpful here to determine the
   dofs on the grid boundary (for which I have code, too)
   but it's not strictly necessary.

This is pretty much all there is to it.  Once you know for each
dof on each level whether it's Dirichlet or not you can treat
them in your assemblers and solvers any way you please.

It seems like I'll be in HD next week.  If there's interest I'll
explain a bit more about Dirichlet handling and the more general
problem of treating data on the boundary.

Good luck,
Oliver

************************************************************************
* Oliver Sander                **                                      *
* Freie Universität Berlin     ** email: sander at math.fu-berlin.de      *
* Institut für Mathematik II   ** phone: + 49 (30) 838 75217           *
* Arnimallee 2-6               ** fax  : + 49 (30) 838 54977           *
* 14195 Berlin, Germany        ** URL  : www.math.fu-berlin.de/~sander *
************************************************************************

On Tue, 11 Jan 2005, Sreejith Pulloor Kuttanikkad wrote:

> Dear Oliver,
>
> I am forwarding a mail from Robert in which he recommended to contact
> you for further information  regarding the
> boundary conditions in DUNE. could you please share your ideas.
>
> thanks and regards,
> Sreejith
>
>
> ----- Forwarded message from Robert Kloefkorn <robertk at mathematik.uni-freiburg.de> -----
>
> Date: Tue, 11 Jan 2005 09:33:15 +0100
> From: Robert Kloefkorn <robertk at mathematik.uni-freiburg.de>
> To: Sreejith Pulloor Kuttanikkad <sreejith at hal.iwr.uni-heidelberg.de>
> Subject: Re: from Sreejith (AG Bastian)
>
> Sreejith Pulloor Kuttanikkad wrote:
>> Dear Robert,
>>
>> I am Sreejith working in the group of Prof. Peter Bastian. I am
>> contacing you to clear a doubt.
>> I am writing a small FEM program using the available models in the DUNE
>> (in order to get a grip of both FEM and DUNE). I am using quadrilateral
>> elements. Now I am facing a problem in incorporating boundary conditions
>> (say Dirichlet) and I think there are no libraries available in DUNE for
>> doing the same. But Peter told me that you had written a similar
>> piecewise linear FEM before and managed to incorporate BC's in someway.
>> could you please give me some tips on it.
>>
> Dear Sreejith,
>
> there are some things to work with boundary conditions in such a case.
> Most things done so far are more or less beta status. For Neumann
> Condtions the situation is a bit easier to handle, because in that case
> you can use the intersection iterator. The more compilcated case is
> Dirichlet conditions, because you need the information of boundary on
> verticies and ....
>
> The information of Dirichlet vertex or not should be stored on the
> corresponding functions space (if you are using this construct,
> otherwise were you like to store it). For this, when the function space
> is created, the information of dirichlet vertex or not should be
> generated by using the intersection iterator. Then, later in progress,
> you can get this information from the functions space via the map that
> was generated, the map maps for example global dof number to dirichlet
> or not, or something like that.
>
> If you're using the function spaces then this feature should be added.
>
> I hope I could explain a least a little bit of how it should work in
> Dune. Unfortunately there is not implemantion of this within dune, but
> Oli Sander work with that strategy and perhaps you can ask him for the
> code.
>
> Regards
>
> Robert
>
> -- 
>
>  Robert Kl?fkorn           <robertk at mathematik.uni-freiburg.de>
>
>  Mathematisches Institut              Tel: +49 (0) 761 203 5642
>  Abt. f?r Angewandte Mathematik       Fax: +49 (0) 761 203 5632
>  Universit?t Freiburg
>  Hermann-Herder-Str. 10
>  79104 Freiburg
>
>  http://www.mathematik.uni-freiburg.de/IAM/homepages/robertk
>
> ----- End forwarded message -----
>
> -- 
> Sreejith Pulloor Kuttanikkad
> Research Student (PhD)
> Interdisciplinary Centre for Scientific Computing
> Ruprecht-Karls-University of Heidelberg
> Im Neuenheimer Feld 368
> D-69120,Heidelberg,Germany
> ------------------------------------------------
> Ph:+49-06221-545689/544412(Off)
>  :+49-017624228904(Mob)
> http://hal.iwr.uni-heidelberg.de/~sreejith/
> ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
>