Left and Right Boundary producing different stress values for Continental extension

Hi,
I’m running a continental extension model with top =free slip, and looking at left and right boundary it produce different stress component compared to each other. If we set up all the variable change only with depth, what could be the reason to have different stress values at boundary ? :slight_smile:
these are the stress_xx @10Ma (i can also see different values at 1st time step.
box_2d_right.0.txt (2.5 KB)
box_2d_left.0.txt (2.5 KB)

When I try to restart the model with these two different stress profile I ran into this error message





Best,
Chameera

Dear Chameera,

First, a couple of tips for asking questions on the forum:

  • You have introduced two unrelated problems - the first is a result that you don’t understand, and the second is a problem when restarting from a snapshot. Please open separate discussions for each issue; this will help ensure that you get answers to both parts of the question. Here I will only deal with the first problem (I do not have experience with restarting from snapshots)
  • Your prm file is complicated. If you want people to help with something, it is a good idea to spend a few hours finding a minimal example that demonstrates the problem at hand. In finding that minimal example, you might discover the solution to the problem.

Here are a couple of relevant observations from your output and prm:

  • The pressure differences are quite small - only a few megapascal.
  • You are using a non-linear rheology.

The second point is important because in non-linear rheologies, symmetric solutions to the governing equations are often metastable. Small perturbations (due to numerical noise) can lead to the development of asymmetry. You could test this explanation by running your simulation again with isoviscous layers and confirm that the solution is symmetric.

We see asymmetry in nature too: in extensional regions, small heterogeneities lead to asymmetric rifts, and even conjugate normal faults require asymmetric extension when they interact (https://doi.org/10.1306/8626BEF7-173B-11D7-8645000102C1865D). So although the origin of asymmetry in your model may be from numerical noise, it isn’t an inherently unnatural thing.

Best wishes,
Bob

1 Like

Thanks Bob.

Best,
Chameera