[aspect-devel] boundary pb

[Wolfgang Bangerth]

> In order to fix ideas, we decided to start all over again, and ran a few
> timesteps of the convection-box.prm setup … and witnessed the same
> abnormal behaviour for three versions of ASPECT (one from two weeks ago, i.e.
> the virtual box made for the GFZ workshop, one from a few weeks back,
> and one from last december).
> As you know, the convection-box is a buoyancy-driven experiment where the
> temperature field is initially prescribed to be
>   T(x,z)=(1-z)-0.01*cos(pi*x)*sin(pi*x)
> so that one expects the velocity field to be symmetric at t=0  (in magnitude,
> not direction). However attached are some screen captures of what we obtain.
> Please keep in mind that we get the convection-box.prm from git (or svn for
> the older version), that we do not modify it at all and that the file
> prescribes free slip on all sides.
> So why does the velocity field display a kink on the right boundary ?

Do I interpret your pictures correctly that
- there is/should be an exact left-right symmetry in the velocity field
- the boundary layer exists on only one side?

Issues where a symmetry should exist but doesn't are often much simpler to 
debug because one has an invariant that should hold but doesn't. If you agree 
with the two points above, do you still see the issue on a much coarser mesh?

Best
  W.

-- 
------------------------------------------------------------------------
Wolfgang Bangerth               email:            bangerth at math.tamu.edu
                                 www: http://www.math.tamu.edu/~bangerth/