[Mesa-users] Resolution studies with binary module

[Wolf, William M]

Hi Roseanne,

That result (increased spatial resolution allowing larger time steps) is not all that unexpected. Those simulations were tuned to execute quickly on a underpowered laptops due to the constraints of the summer school. With low resolution comes worse residuals and other artifacts that require shorter time steps, which might seem to make the runs last longer. Indeed, that can happen, and it’s a great discovery when your realize that adding resolution actually saves compute time!

But often the lowered spatial resolution makes individual time steps take less compute time, while a higher spatial resolution can take longer to run even though it uses fewer time steps. I can’t remember the details of how we tuned those models to run under the constraints of the summer school, but I seem to remember there were some pretty large corners we cut to ensure they would run, perhaps even just to avoid overflowing the RAM on weaker laptops.

What settings you should use is something you will have to determine by conducting a resolution study, which it seems you are already doing. So long as the results have stopped appreciably changing beyond a certain point, you likely don’t need to keep adding more resolution. You should do something similar by enforcing shorter time steps as well to see if that makes any difference.

Regards,

Bill Wolf

On Jan 12, 2023, at 3:28 PM, Cheng, Roseanne Marie via Mesa-users <mesa-users at lists.mesastar.org> wrote:

Hi everyone,

Thank you for all of the resources and tests available.  We have been working on massive star and stellar black hole binary simulations in MESA using both the star_plus_point_mass test suite and the inlists from Pablo Marchant's 2017 Summer School slides on "Stellar rotation in binary systems" or "Numerical convergence in simulations."  The presentation slides are great -- I wish I was there!

We have been finding that in decreasing max_dq (or increasing the number of zones), the number of time steps decreases for the runs we have set up.  I've attached the inlists.  Naively, I would have expected the opposite -- the number of time steps increases (or stays the same) with decreasing max_dq.  Would we need to change the wind/rotation settings along with max_dq?  Are there more resources about these settings?  Do you know the best diagnostics we could plot to check if these settings are appropriate for our model such as the controls for the implicit wind?

Here are a few more details:

For a setup based on star_plus_point_mass, we use the same initial zams.mod model to run with max_dq = 1.0e-2 (default), 1.0e-3, 1.0e-4.  This results in the same number time steps for all three of the models resulting in an increasing number of zones.

For the new set up (attached), we use the same initial zams.mod model.  The difference between this set up and star_plus_point_mass is the rotation settings from Marchant 2017.



Thanks,
Roseanne

---
T-3: Fluid Dynamics and Solid Mechanics
Theoretical Division
Los Alamos National Laboratory
+1 505 667 1607

Pronouns: She/Her/Hers

This email is a safe space


<inlist><inlist_project><inlist1><inlist2>_______________________________________________
mesa-users at lists.mesastar.org<mailto:mesa-users at lists.mesastar.org>
https://lists.mesastar.org/mailman/listinfo/mesa-users

-------------- next part --------------
An HTML attachment was scrubbed...
URL: <https://lists.mesastar.org/pipermail/mesa-users/attachments/20230112/66413c79/attachment.htm>