No convergence for extension/plume moel

Dear all,
I am writing because I am trying to run a model where the lithosphere is simultaneously subjected to extension and plume impingement. In my intention, when the plumes reach the LAB, I should start the lithospheric extension and produce fast delamination.
However, I have noticed that, as soon as the extension starts, my model does not converge anymore, regardless of the solver I use for my simulation. In addition, the residual is often very large (sometimes higher than 10^18).
Here I enclose one of my input files (actually one of the most successful ones, where the residual is “only” 260). In this model, I apply extension on the sides of the lithosphere, whereas the remaining parts of the lateral boundaries are open.
original.prm (18.2 KB)
Any help would be very appreciated. Thank you very much! :smiley:

Dear Alessio,

Thanks for posting to the forum! Indeed, a residual of 260-10^18 is too high as you have noted :smile:

In addition to the PRM file, what would would also help to diagnose the issue is the following:

  • The log.txt file that accompanies the PRM file
  • Images from the model run that illiustrate the dynamics and any potential issues/features leading to the poor convergence behavior
  • A history of how you developed the PRM file and at what stage things stopped “working”. For example, you note that currently the solvers break when you apply extension on the boundaries. Was there a previously model that did show good convergence when you applied extensional BC?


Dear John,

thank you very much for your quick answer and for your help. Here I attach the log.txt file and a few images of one simulation, the time step before it stops converging.

Regarding the development history of the model, previous versions had a thickness of 400 km, activation volumes for mantle materials between 18e-6 and 25e-6 J/Pa, and lower densities for the lithosphere and the mantle (around 3300 kg/m3). I realized that the model stopped converging when I decreased the activation volumes and increased the density of both the lithosphere and the mantle.
Right now I am trying to run a successful model by changing each parameter one at a time, and then I will try to combine more parameters simultaneously, in order to understand what does not work.

Another aspect to mention, the lateral boundaries were not open in my previous simulations, but I implemented a box geometry with a lateral outflow to keep the volume of the model constant. In any case, when I re-apply my old boundary condition I still do not obtain convergence.

Once again, thanks for everything.

log.txt (5.5 MB)

Dear all,
I ran several different tests, and it seems that, as soon as introduce extension, the model stops converging. This happens both when I apply extension to the boundaries of the lithosphere and I open the mantle boundaries, and when I counterbalance the outflow of the lithosphere with a mantle inflow. I have tested models where there was no extension (“Zero velocity boundary indicators” on all boundaries") and the model was running without any problem. Right now I am using the “single Advection, iterated Stokes” nonlinear solver scheme, because I tested the other solvers and they slowed down my simulations without adding any benefit to the stability of my solution.
Is there any suggestion you can give me?
Thanks a lot!

Hi Alessio,

As with one other forum post, this fell off my radar and I forgot to follow up. Apologies for the long delay.

The images you posted certainly showed very sharp viscosity contrasts, from what appear to be the influence of the plume. I am not surprised the solver had issues, but what I am unsure of is why the solvers crash when extension (through kinematic BC) are turned on.

If you are still looking for a solution, here is my suggestion on how to proceed:

  1. For the purpose of simplification, create a simulation without a thermal anomaly (plume).
  2. Decrease the model depth to 400 km, and apply outflow along the entire length of the sides, which will be then balanced by inflow at the base.
  3. Try reverting to outflow along the lithosphere, followed by a transition (linear) to inflow in the asthenosphere.
  4. Finally, increase the model depth back to 800 km.

If an example model would help, here is a link to a rifting setup with kinematic bc on the sides(outflow transitioning to inflow) for a paper in review:

Apologies again for the delay, and if you are struggling with this issue let’s keep the discussion going.