[Mesa-users] inaccuracies from large mesh_delta_coeff and convergence helpers

[Thomas Steindl]

Hello Simon,

There are multiple reasons why such models may experience convergence issues. I assume you are using T-Tau boundary conditions since you are interested in pulsations. I have seen that setting the error tolerance of the atmospheric integration can help (atm_T_tau_errtol = 1d-9, for example). If the convergence issues happen very early in the evolution, setting scale_max_correction = 0.1 can help the solver to not overshoot the „correct“ solution in the convergence process.

That said, it is usually very hard to help without having the inlists that you are using. I am happy to take a look at them if your issues persist. 

If you change mesh_delta_coeff later in the run (you can do that in run_star_extras for example), you will obtain a better resolution in the model going forward. However, the error that has been made in the beginning of the run will propagate through to some degree. 

Best,
Thomas

__________________________________________________________________

Thomas Steindl, MSc
University of Innsbruck | Institute for Astro- and Particle Physics
Tel: +43 512507 52031
Mail: thomas.steindl at uibk.ac.at
Homepage: www.thomassteindl.com

> Am 03.09.2021 um 08:53 schrieb Simon Murphy via Mesa-users <mesa-users at lists.mesastar.org>:
> 
> Hi folks,
> 
> I'm supervising a student project to investigate systematic modelling uncertainties on pulsation frequencies of 1.5--2.0 Msun stars.
> 
> This involves evolving some models from very young ages (create_pre_main_sequence_model = .true.) to the late main sequence. A major difficulty is having models converge, because the solver eventually gives up with "hydro_failed" a lot of the time.
> 
> There are of course things that help. Having "Pextra_factor > 1", having "okay_to_reduce_gradT_excess = .true.", and, it seems, having larger "mesh_delta_coeff". I'm not sure why the latter helps, but it seems that it does. Would love to know why.
> 
> My question is this: if at some point in the evolution the run uses, say, Pextra_factor = 2 and mesh_delta_coeff = 1.0, is there any point in later using Pextra_factor = 1 and mesh_delta_coeff = 0.1? Or is that accuracy permanently lost?
> 
> The student or I would check ourselves, but we can't get anything to converge with Pextra_factor = 1 and mesh_delta_coeff = 0.1 all the way through.
> 
> Thanks,
> Simon
> 
> -----------------------------------------
> Dr Simon J. Murphy
> Incoming ARC Future Fellow
> School of Physics
> University of Sydney
> NSW 2006, Australia
> tel: (+61) 0290365104
> -----------------------------------------
> _______________________________________________
> mesa-users at lists.mesastar.org <mailto:mesa-users at lists.mesastar.org>
> https://lists.mesastar.org/mailman/listinfo/mesa-users <https://lists.mesastar.org/mailman/listinfo/mesa-users>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <https://lists.mesastar.org/pipermail/mesa-users/attachments/20210903/d5823566/attachment.htm>