#49 Artefact from ghost zones with boundery bcz='set'

[Anonymous]

Originally created by: rplas...@gmail.com
Originally owned by: rplas...@gmail.com

I noticed a problem when using the 'set' boundary condition for bcz.

If I start a run by imposing these boundaries:
  lperi=T,T,F
  bcz='s','s','a','s', 's','s','s','s','set'
  fbcz1=0.,0.,0.,0., 0.,0.,0.,0.,0.
  fbcz2=0.,0.,0.,0., 0.,0.,0.,0.,1.,
(the purpose is to impose a concentration 1 of a chemical compound on a top
layer, and of 0 on the bottom layer, in order to impose a matter flux)
Then the profile of concentration on the top layer for this compound
corresponds to the picture bcz_set.png and bcz_set2.png: at the interface
between the processors block, there is a spike of lower concentration.
This effect disappear if the 'set' boundary is put on the x axis instead of
z (see bcx_set.png).

As Axel suggested, this comes probably from the order in which the ghost
zones are initialized... I don't really know how this should be solved,
besides selecting the "good" axis...

[Anonymous]

Originally posted by: tobias.h...@gmail.com

Do these pictures show the state of the system after it has evolved for a while? Or is this the initial state?

[Anonymous]

Originally posted by: rplas...@gmail.com

This is showing the state shortly after the launching of the run (50 time steps), but
this persists all the time. I am sending a picture of the concentration profile at
the top layer after a long run (3000 time steps).

The chemical reactions implies non homogeneous concentration, even at the top
(variations of ≈1%) that are about 10× more important than the boundary artifact.
However, this artifact stays there and implies a systematic error. This may not be a
very important bug (instead of having a clean constant boundary, we are probably only
imposing a slightly tiled profile). But in the cases these variations would
noticeably influence the runs, this can be a problem as this "tiling" is dependent on
how the system is sliced to the processors.