No subduction when running "Mantle convection with continents in an annulus" cookbook

ASPECT
1

Hi everyone (maybe especially Cedric),

I ran into issue when trying to run the prm file come with the “Mantle convection with continents in an annulus” cookbook. The slabs in my model cannot be subducted and all end up diffusing away. It looks very different from Fig. 149 in the manual.

result_1
result_11552×1118 176 KB

We are using ASPECT 2.6.0-pre with deal.II 9.6.0. We appreciate any thoughts or suggestions about what may cause this issue.

Thanks in advance,
Lucy

2

Hi,
interesting. Did you change the resolution?
I have an MSc student starting this week, I’ll ask him to run this cookbook tomorrow.
More soon.

3

Hi Cedric,

Thanks for the response! We tried the original AMR setting in the cookbook and the uniform gird with a global refinement level of 8. No subduction occurred in either case. Very looking forward to what your student model results will look like.

Lucy

4

Hi Lucy (and Cedric),

Looking at the commit history for the cookbook prm, here are the key changes that have occurred:

  1. Switching from fields to particles for tracking materials
  2. Switching from the AMG to GMG solver and turning on averaging (harmonic) over the entire element.
  3. Minor updates to the particle section of the PRM file.

My guess is the change in behavior occurred in the commit that added the first two changes.

Lucy, can you try the following?:

  1. Switching back to AMG (use set Stokes solver type = block AMG)
  2. Turning off averaging across the element (use set Material averaging = none)
  3. Switch back to using fields instead of particles (remove all lines related to particles).

Cheers,
John

5

hi,
we just ran the prm, and agree with your conclusions. Long story short, Erik van der Wiel, who built the prm & wb files actually had a few additional cold blocks in the mantle to kickstart things. During the last hackathon, I had to rewrite his wb from scratch (new wb version) and omitted the blocks. I now suspect that these are necessary (even more than the prescribed slabs?) to get the model running…

Screenshot from 2024-10-02 14-23-11
Screenshot from 2024-10-02 14-23-111920×1080 467 KB

6

John,

Thanks for the suggestions! We tried to rerun our model following you first two suggestions and the result does not change. The slabs still diffused instead of getting subducted.

Cedric,

Thank you so much for being willing to share these insights with us! We are trying to increase the length of the initial slabs to see if we can kick off subduction without having the cold blocks. We will reply to this thread about our progresses.

Best,
Lucy