[aspect-devel] cylindrical rigid body rotation

[Ian Rose]

Hi Cedric and all,

I have also run into this and related issues.  It has to do with the fact
that rigid body modes are in the nullspace of the Stokes operator, and over
many timesteps the solutions can pick up power in those modes without any
effect on the residuals.  You would also see rigid translations in a box
with free slip BCs and periodic boundary conditions.

Note that not only will it mess up the Vrms values, but it can really make
your Courant number plummet unnecessarily.

Citcom has the same phenomenon, and they just subtract off rigid rotations
at every timestep.  I do have some code that addresses this which is not in
the mainline.  I am meeting with Timo in a couple of weeks, perhaps we can
add this to a todo list.

Best,
Ian




On Thu, Feb 27, 2014 at 6:14 AM, cedric thieulot <geogarfield at gmail.com>wrote:

> Hi,
>
> I am running a few simple simulations in cylindrical geometry (constant
> viscosity, incompressible flow, low Ra number) for a benchmark article.
> When I run the simulation with no-slip bc, the solution looks fine and the
> measurements correspond to values obtained with other codes.
> As soon as I use free-slip on the inner and outer bc, I observe a rigid
> body rotation (see attached movie). Note that the measurements in general
> are not very much off,
> but the root mean square velocity measurements will undoubtedly suffer. Is
> there an algorithm implemented in the ASPECT to remove it ?
>
> Cheers,
>
> Cedric.
>
>
>
>
>
>
>
>
>
> _______________________________________________
> Aspect-devel mailing list
> Aspect-devel at geodynamics.org
> http://geodynamics.org/cgi-bin/mailman/listinfo/aspect-devel
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://geodynamics.org/pipermail/aspect-devel/attachments/20140227/1cdb1ec0/attachment.html>