Partial mesh refinement triggering EB coarsening - question

[AlexGlock]

Hi WarpX team,
I have a general question related to the partial mesh refinements in WarpX. I am running electrostatic simulations with kind of complex STL embedded boundaries. So far I am using a uniform 3D cartesian grid for this and now I am trying to include a (static) refined region into my model domain, which does not intersect with any EmbeddedBoundary and whose spatial resolution should be two times FINER than the base grid.

When setting this up as a refined region, the Multigrid solver for some reason starts to coarsen my EmbeddedBoundaries and errors out on that because the EBs are borderline resolvable with my level 0 grid discretization. I am now wondering why the Solver is trying to coarsen my base grid in the first place, when I am actually trying to add a finer layer? Can I somehow avoid/deactivate this and if not: can I perhaps go the inverse route of defining my base grid with the finer resolution and then partially coarsen regions within it?

Cheers,
Alex

[RemiLehe]

Thanks for bringing this up! This is indeed does not look like the expected behavior.

• Would you be able to attach an input script, so that we can try to reproduce the error?
• Could you clarify whether the WarpX version that you used included this change: Store and use per-mesh-refinement-level EB IndexSpace #7034

[AlexGlock]

I was unable to reproduce the described error with the newest WarpX release, including the change mentioned above.

Mesh refinement now works as expected. Unfortunately, I am unable to specify the exact version in which the error occurred, but it seems to have been fixed.