I would like to investigate dynamic topography at the CMB in spherical geometry. This is typically done with a free-slip boundary, where topography is calculated based on stresses. Yet, it would be interesting to use a “free base”, similar to the models which study surface topography.
When I had a look at the manual (section Free surface), it sounds like you could apply that to any boundary of your domain. However, all set-ups that I have tried ran into a solver convergence error sooner or later, no matter what choice of “CFL number” and “free surface stabilization theta” I used.
I then tried the cookbook “free-surface.prm” and switched the free surface to the bottom, but again I end up with the convergence error after 3 time steps with a huge residual (~1e20).
Has anyone ever tested using the free surface at any other boundary than the top of the domain? And is this convergence error related to the fact that you cannot specify any material properties (e.g. density, viscosity) for the fluid outside your domain? I guess ASPECT intrinsically uses the values for air, which might cause problems for the CMB.
Thanks for your help