The 3D extensional model with weak zone


I made 3D model of multi-components, like 3D sand-box orthogonal extension model with a small weak zone in the middle layer. However , the strain rate distribution showed the model always deformed from the base layer, not from weak zone. Using the same parameter for making 2D model can reach the target but can’t in 3D model. Which criteria should be noticed?


This is hard to answer given the information you provided. Do you prescribe deformation at the bottom boundary, or at the side boundaries? Where are the highest stresses in your model? What is the top boundary condition?


@cgw0814 - Another apology for not seeing your posts earlier! As @gassmoeller mentioned, it is a bit tough to diagnose the issue with the current information. Would you be willing to post a parameter file? Strain localization is a tricky issue going from 2D to 3D.


Hi Dr. @gassmoeller and Dr. @jbnaliboff

Thanks for reply, I’m sorry to respond so late because I was busy in Taiwanese Geoscience Association recently.

Few days ago, I was intersting in a “Continental Extension Cookbook” on the ASPECT-github

Result of 2D model shows the shear band convergs at seed or weak zone as manual describes. I thought I could refer to the values of parameters in that prm file and applied to 3D model. I reduced the model resolution because the limitation of proccessors (only 32 cores).

Continental Extension-3D model output and prm are put on

Apparently, the output shows shear band don’t converge at seed but model base

Another example I have focused on is in “ buiter_et_al_2008_jgr ” in benchmarks

I increased the velocity and let it extended in one side only.

Output and prm are put on

Outputs shows a clear half-graben forms and lower part of model flow upward to the surface. That what I expected.

So, I transfered figure_6_1e20.prm to 3D with reduction of size and resolution with respect to original exampme. Output and prm are put on

Output shows shear band converges from seed, I thank this example it possiblely works and it’s worthy to refer if temperature will not be considered. But it’s must be improved until reaches what I want as same as I saw in 2D.

I have studied some reference to figure out the key points, like Allken et al, 2011

Allken et al., 2012a

Allken et al., 2012b

It’s going to spend some time to figure out.




HI @cgw0814,

Thank you for posting all of that information! Strain-weakening will certainly help localize deformation and potentially improve the solver behavior in 3D. A bit hard to analyze the modified continental extension model, as the height was significantly decreased (very different dynamics).

I think you are on the right path going forward. As strain-localization in 3D is inherently a very tough problem, my suggestion would be to pick one the Allken models and try and reproduce the setup (and results) exactly. Reproducing the setup exactly will help you get familiar with all of the key parameters. After that step, you can add temperature to the models to make the problem closer to the continental extension cookbook.

The Buiter et al. benchmark is a good place to start for making the 3D Allken (or other similar) models.

Does this sound like a good plan? Happy to take a look at parameter files and offer suggestions along the way.

A few suggestions based on the continental extension cookbook 3D model.

  1. For testing purposes, try reducing the non-linear solver tolerance to 1e-3. Later on you should use a stricter value (5e-4, 1e-4, etc).
  2. Start with a small 3D model (e.g., not long in the y-direction) and then systematically increase the length.
  3. Try using harmonic material averaging in the Material model section: “set Material averaging = harmonic average”. For an example, please see “tests/point_value_01.prm”.

Let’s keep the discussion going and if you run into trouble, please do not hesitate to reach out here!

1 Like