search

geodynamics / aspect

A parallel, extensible finite element code to simulate convection in both 2D and 3D models.
https://aspect.geodynamics.org/
Other
227 stars 237 forks source link

"Shrinking" spherical shell #6114

Open Minerallo opened 3 weeks ago

Minerallo commented 3 weeks ago

Hi everyone I ran the cookbook spherical shell and enabled the free surface, the sphere starts to shrink as shown in the picture below. The mesh seems quite unstable with 1 level of global refinement in this example, as shown in the video: https://youtu.be/P2cjp4ToBVw This shrinking also happens for higher resolution models (~25km at the surface), but at a much slower rate ~50km over a period of 600 Myr. The mesh can be more stable by giving a constant level of refinement via the minimum and maximum refinement function over depth but the shrinking still occurs. Do you have any idea what could cause this "shrinkage" and how to fix it ? thanks in advance ! cheers Michael Pons image image test_shrinking_example1.prm also 2 levels of global refinement in this other example (https://youtu.be/b1IzhLPSi9w) which is equivalent to using the default parameters (for having tried).

tjhei commented 3 weeks ago

The computation seem to repeatedly coarsen and refine the mesh. Can you turn off adaptive refinement to see what happens then?

Minerallo commented 3 weeks ago

Hi Timo, thanks for your reply, the shrinking still happens when the global level of refinement is set to 0 (video : https://youtu.be/rXHt5qS6Vww). The model crashed after 20Myr.

image 
subsection Mesh refinement
  set Initial global refinement          = 0
end

If no parameters are set, the default parameters are used (which gives a global refinement of 2 : https://youtu.be/b1IzhLPSi9w).

tjhei commented 3 weeks ago

You misunderstood. Please set https://aspect-documentation.readthedocs.io/en/latest/parameters/Mesh_20refinement.html#parameter-name-time-steps-between-mesh-refinement to 0.

Minerallo commented 3 weeks ago

Oh yes, sorry, here is the initial model with global refinement 1 and time steps between mesh refinement 0 (https://youtu.be/xMQs2K0hdZ8).

image

Shrinkage still occurs, but is much less -60 km mean topography after 143 Myr following a progressive decrease. The negative temperature we get after 140 Myr might have some effect (?). I am not sure why this happens, but I could add an adabatic heating model or something else to prevent it ...

Minerallo commented 3 weeks ago

@MFraters advised me to switch to the normal projection, I tried the normal projection instead and it seems to work https://youtu.be/PfKEuJMkuV8. The temperature also matches the initial boundary condition. I will run more tests. Thanks!image.