Hi everyone,
I’m running a mantle convection model with viscoplastic rheology but encountering unrealistically high velocities (~40 m/year), which is also drastically reducing my dt = ~150 years. Given that I need to run the model for about 4 Gya, the small time step increment makes it nearly impossible for the solve to reach completion. I’ve set a lower viscosity limit of 1e20 Pas. Has anyone experienced a similar issue or found a workaround?
Thank you in advance for any suggestions or advice!
log.txt (1.7 MB)
original.prm (5.7 KB)
Some additional context: I’m running the whole mantle convection model using spherical geometry. The goal is to prevent a stagnant lid by implementing temperature-dependent visco-plasticity. The model initially runs stably, but when yielding begins in the lithosphere, velocities spike to around 40 m/year as material subducts into the mantle.
@arnie94 You seem to have a complicated set up that leads to these results. Have you output the viscosity, for example, to see why the velocities are so large? If you can see areas of unreasonably low viscosity, you need to ask why that is so.
In the end, as mentioned in another post this morning, the way to debug these problems is to come up with the simplest version of the model that illustrates the issue, i.e., to find which single factor is causing things to fail.
Best
W.
@arnie94 - Yes, 40 m/yr is quite high.
As noted in this paper by Magali Billen and Margarete Jadamec, mantle velocities on the order of meters per year is not necessarily unphysical, but that tended to occur for lower viscosities (10^18 - 10^19 Pa s).
I’ve seen similarly high (meters per year) viscosities in both instantaneous regional subduction and global convection models with very low minimum viscosities, but again that is hard to reconcile with your minimum viscosity of 1e20 Pa s.
Does the viscosity always go to 1e20 Pa s in the yielding portions of the lithosphere?
One thing you could try is adding a plastic damper to the rheology model (manual section) with a higher viscosity (1e21 or 1e22 Pa s) than the minimum viscosity.
This should slow down convergence rates in subduction zone, but is really just a tuning parameter. I did not see anything in the PRM file that looked out of place. Perhaps try also making the linear and nonlinear solver tolerances a bit stricter just to make sure that is not the issue?
As @bangerth noted, any additional images or simplified models you can provide would be helpful.
Cheers,
John
Hi,
I just checked your model. I think there are such questions:
- You use the dislocation creep for the lower mantle. This may not be so suitable. You can add a phase transition at 660 km, the diffusion-dislocation-plastic for the upper mantle, and the diffusion(-plastic) for the lower mantle.
- The friction angle, 0.1 degree, looks too small. I think you could increase the friction angle.
- The timestep is controlled by the maximum velocity in the energy equation. Therefore, to avoid the too small timestep, you can modify the code and add a maximum limitation when solving the energy equation and particle advection.
Xinyu Li
