2D subduction model ASPECT 2.5.0 vs. 3.0.0 + convergence issues

Hi everyone.

I have a problem with a 2D Visco-plastic subduction model with all sides free slip. The model works well when using ASPECT 2.5.0., but exactly the same model fails on the Stokes solver first time step when using ASPECT 3.0.0. I tried different nonlinear solver schemes from simple advection and stokes to single Advection, iterated defect correction Stokes etc., but got the same result. I also tried different solver types AMG, GMG, but no luck. The setup is rather simple, an initial slab from WB and an overriding plate. Any ideeas how to solve this?
subduction_2D.prm (4.4 KB)

Thank you
Vlad

Hi @vladycm,

Are you using the same version of WB to generate the initial conditions.

My initial suggestion would be to use the nonlinear solver scheme no Advection, no Stokes (+ End time = 0), which will just setup the initial conditions and run the post processors. Using that output, then confirm is the initial model setup is the same between the two versions.

Cheers,
John

Hi John.

No, for ASPECT 2.5.0 I use WB 0.5 and for 3.0.0 I use WB 1.0.
The WB file is the same, for both. I am going to try this to see if the IC are the same.

Best.
Vlad

Hi John.

Just to confirm that the two initial model setups are identical.

Vlad

Hi @vladycm,

Some ideas for what to try text next:

  • Run the model for just one time step (set End time = 0) with one nonlinear iteration (set Maximum number of nonlinear iterations = 1).
  • Compare the outputs to see where things diverge.
  • Be sure to run the models in debug so that you can provide the full error messages if any model crashes.
  • You can further simplify the above procedure and outputs by setting all of the boundaries to be free slip, and then setting the Minimum strain rate and Reference strain rate equal to the same value.

Cheers,
John