Outflow conditions with reduced.h

[Marc Vacher]

Dear everyone,

I am trying to simulate a jet issued from the bottom that impinges on a free-surface, like an underwater jet if you prefer. It works well inside of a tank (see https://basilisk.fr/sandbox/MCVacher/slosh_reg.c), it is even quite fast to simulate with classical gravity conditions.

However, I would like to study the phenomenon with no walls, so I need an "infinite ocean" condition so to say. I think reduced.h is the tool I need. However, fixing the right boundary conditions on pressure is now not so obvious for me. I fixed 0 pressure as a reference at the bottom. I tried here : https://basilisk.fr/sandbox/MCVacher/jet_flutter.c

I encounter several problems :

1) Without "

 f[left]=  y < h ? dirichlet(1.) : dirichlet(0.);
 f[right]= y < h ? dirichlet(1.) : dirichlet(0.); 

", the simulation is fast but the level is decreasing on average. I expected it. Interestingly, the behaviour and the rate of simulation is affected by whether I initiated the interface right on the edge of a cell or not. Maybe my problem is somewhat there ? Or related to ghost cells values ?

But except that, I don't have massive backflows whatsoever. 

2) With these conditions back in, the water level is on average constant, which is what I want. But the the simulation is very time consuming, plus I have important backflows.

I tried many things seen on the sandbox for this kind of problems that seems to be not so uncommon when one uses reduced.h. 

Neither the solution of coarsening the mesh at the boundaries to reduce the backflows : 

"event adapt (i++) {

  adapt_wavelet ({u}, (double[]){3e-2,3e-2}, 9, 4);
  unrefine (.......);
}" 

of JMF FAQ sandbox worked, nor various tries on boundary conditions involving face centered velocities uf.n and uf.t worked very well.

With all of these pieces of investigation, I hope someone will have a solution for this problem.

The code is also in attached document (same as sandbox).

Please don't hesitate to tell me if I misunderstoood something basic about reduced.h or something else.