Hi everyone,
I have some P,T,rho… profiles that I generated with BurnMan for hypothetical planets with different Core/Mantle ratios, I set the sampling of these profiles equally-spaced and put them in ASCII format just the same as the example_isentrope.txt. The problem is when I run the ASPECT cookbook ‘burnman’ with these profiles, I get and error of convergence. I read the manual and changed some values in the parameter section of Stokes solver parameters to increase the iterations but is still crashing (See the image). In that case, I don’t know what I’m doing wrong, even the input parameters or the profiles itself. As my models need to change the core size, I’m using the Planet class object from BurnMan to generate these profiles instead of the 1D adiabatic profiles example. I would appreciate it if you could help me to understand what’s happening with this error. I attached the error and the .prm I’m using in this case.
Also, as a second question: How can I set the dt constant? In all models that I’ve run so far, the dt changes in every timestep, even when I set the mesh to be non-adaptive. In my last model, the program achieved a certain timestep and started to show a dt~e-114. So, I had to stop it otherwise, it will take a lot of time, and I don’t know which parameter is the one that controls that. Thanks a lot for your help.
Best,
Mariana Villamil
cookbook-exoplanet.prm (4.1 KB)
Dear Mariana,
First, the convergence problem:
- Can you post the isentropic profile that causes this failure too? It’s not possible to bug-fix if we can’t reproduce the issue.
- I suspect the problem may be related to the smoothing of the adiabatic profiles - very fine spacing will result in very large compressibilities and thermal expansivities, and could conceivably cause convergence issues.
- You may be interested in an alternative way of modelling isochemical reactive flow in ASPECT: https://academic.oup.com/gji/article-abstract/231/3/1833/6652101
This may be more suitable for your purposes, especially if you are investigating large exoplanets.
Second, you can set a minimum timestep in the prm file:
However, before you do this, consider why ASPECT is making the timestep smaller - it often implies that there is some process that cannot be accurately captured with large timesteps.
1 Like
P.S. Another problem may be that the refinement level in your simulation is not capturing the latent heat / density change of reaction. You should make sure your simulation has at least four cells width for each transition if you aren’t usng the enthalpy/entropy method implemented in Dannberg et al. (2022).
1 Like
Dear Bob,
Thanks for your answer. I could set a minimum time step and increase the spacing, but still the convergence problem persists. I could use the entropy method proposed in the cookbooks, but I don’t know how can I change the core/mantle boundary in those files. The main important factor that I want to study is the parameter space of core/mantle ratios in convection. That’s why I decided to use the entire Planet profile from BurnMan, because I can set the size of the mantle relative to the planet radius and obtain a self-consistently profile of these exoplanets. If there’s a way in which I can set that parameter directly, I wouldn’t have any problem using that cookbook. Thanks a lot for your help!
BurnMan_profiles.txt (60.8 KB)
Dear Mariana,
I haven’t tested your model, but I wouldn’t be surprised if either/both of the two points above is the cause of your woes.
Other points:
-
Similar to the cookbook profile, your profile doesn’t include latent heats of reactions. That might be fine for you - I just wanted to make sure you knew.
-
I could use the entropy method proposed in the cookbooks, but I don’t know how can I change the core/mantle boundary in those files.
If that’s the only thing holding you back, perhaps you could ask for help with that problem in a separate question on this forum?
Best wishes,
Bob
1 Like
Dear Bob,
I understand the problem now. I will ask in a separate question for the Core/Mantle boundary to have adiabatic profiles self-consistent. Thanks a lot for your help!
Best
Dear Mariana,
For the Burnman-aspect coupling, I hope my answer above helps (you need to fix those two problems with your input profile).
I’ll reply to your other question now.
Best wishes,
Bob
1 Like