[aspect-devel] Checkerboard at large scale

[Wolfgang Bangerth]

On 04/07/2018 02:09 PM, Juliane Dannberg wrote:
> 
> If that is not the case, and all boundaries are closed, you will not get a 
> reasonable solution (or the model will just not converge). John, Wolfgang, 
> please correct me if you think that works differently.

Correct. For incompressible models, there must be a zero net in/outflux or 
either the solution you get is likely wrong, or the solver will simply not 
converge to high accuracy.


> If, on the other hand, your model is _compressible_, this would work 
> differently (so it would be good if you could tell us if it is).

Indeed. If the model is compressible, then a net influx will simply increase 
the pressure to the point where the density increases (due to the higher 
pressure) to accommodate the net influx.

Best
  W.


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