Convergence Failure with velocity conditions to produce lithospheric thickening

ASPECT
1

Hello Everyone,

Hope you’re doing well!
I am currently working with ASPECT and trying to model a scenario of lithospheric thickening. My goal is to thicken a lithosphere by putting velocity conditions at the boundary of my model. I must say that I am pretty new to ASPECT.

My model consists of a 2D box of 800 km in depth and 2400 km in width. I have 2 compositions: crust and mantle. Crust is top 40km, while the rest of the model is mantle. I also have a thermal anomaly at the bottom boundary to produce plumes. Finally, I am trying to impose an inflow condition at the top 200 km and an equally balanced outflow condition below it. I would Ideally like to start the convergence after the lithosphere has been affected by multiple plumes, but currently, only for debugging, I am starting the convergence from 5 million years.

The problem I face is that the stokes solver never converges. I have tried multiple velocity conditions, but the same problem persists. The velocity conditions, along with the implementations, are:

  1. Inflow of 1 cm/yr at both left and right boundary at the top 200 km and an equal outflow below it.

    linear_velocity
    linear_velocity905×273 37 KB

    This model runs for 7.6 million years

  2. Inflow of 1 cm/yr at both left and right boundary at the top 200 km(but with a sin function) and an equal outflow below it.

    sin_velocity
    sin_velocity901×260 34.2 KB

    This model runs for 5.08 million years

  3. Inflow of 1 cm/yr at both left and right boundary at the top 200 km(but with a tanh function) and an equal outflow below it.
    This model runs for 5.18 million years

I have also tried removing the crust and keeping only one mantle composition to make it simpler. But that did not solve the issue.

I have attached the output error message of the linear velocity(1) boundary condition and also a picture of the first timestep of the model when the convergence starts, along with the log.txt file.

output.e12043736.txt (6.1 KB)
log.txt (73.7 KB)

velocity_ss
velocity_ss1184×923 110 KB

Really appreciate any help with this issue.

Thanks,
Arijit

2

Hi @arijitchkc,

Welcome and thanks for posting to the forum!

The way I have seen most groups implement this type of boundary conditions on the side walls is to have an intermediate section where there is a linear transition from inflow to outflow.

For example, you could have inflow in the upper 300 km, a linear transition to outflow over the next 200 km (300-500 km depth) and then outflow from 500-800 km depth.

I think this will produce a much smoother flow field than what you currently have (something looks off in the images).

My suggestion would be to try this approach first in a simplified setup with a constant viscosity and no initial thermal perturbations. If the flow field at time 0 looks correct and the solver converges in a stable fashion, then start re-introducing complexity back into the model.

This will be farther down the line in your workflows, but I also suggest having the imposed flow patterns at 5 Myr gradually ramp up over some period of time. I’ve found many instances where instantaneous changes in the boundary velocity produce solver convergence issues.

I hope this helps and let’s keep the discussion going as you test various scenarios.

Cheers,
John

3

Here are my notes with regards to the boundary conditions that John describes above.
I hope this will help.

image
image889×898 57.6 KB

image
image901×954 54.2 KB

4

Hello John,

Thanks for your help. I could solve the issue, with the suggestions. Also thanks to @cedrict , those notes are really helpful. I have used those formulas in the models and I have attached a picture below.

I initially tried with no thermal perturbation in the model and gradually introduced complexity. The total duration of the models I am currently running are 500 Mya.

Then I tried varying the onset and duration of convergence, and this works quite well now. I also varied the size of the box models a bit, to make sure everything works for a larger setup.

Here’s an example of the model where I started convergence from 140 Mya and continued till 240 Mya, and they ran without any convergence issue. I have attached the velocity(X direction) and Temperature plot for the starting time step and the one just before it.

T_138
T_1381168×903 116 KB

T_140
T_1401151×839 115 KB

Regards,
Arijit

5

@arijitchkc - Thanks for the update and glad the models are working now!

Cheers,
John