[Mesa-users] split_mixing_choice parameter in MESAv15140

[Evan Bauer]

Hi Anthony,

Split mix has been removed from MESA. You're right that this change does not appear to be documented in the the changelog.

There was some discussion back in 2020 among the developers as to whether split mixing was useful for anybody, and the conclusion was reached that nobody felt it was useful enough to maintain. Here's what Bill Paxton had to say at the time:

split_burn allows us to do multi-substep solves indendently for each zone.  this is important at high temperatures where the simple backward euler can’t give decent results unless we force extremely small timesteps.  

split_mix doesn’t currently have any similar value.   it might be useful if it allowed the newton solver to work on only the structure variables and skip the possibly large number of composition variables.  but that only happens if we do split_burn and split_mix for all cells.  we currently use a logT_start cut so that we only pay the price of the split_burn for high temperature cells that need it.   so we cannot limit the newton to exclude the composition.  so there isn’t really much value to split_mix.

and for large nets and high temperatures, the total runtime seems to be dominated by the 1 zone burns rather than the matrix solve anyway.  so doing more cells using split_burn in order to make the matrix solve faster is counter productive.


Cheers,
Evan


> On Sep 29, 2021, at 10:11 AM, Anthony Noll <anoll at irap.omp.eu> wrote:
> 
> Dear MESA users,
> 
> I noticed that the split_mixing_choice parameter is not available anymore in MESA r15140. It used to be in MESA r12778, and I cannot find the reason of this change in the last changelog... Is there any reason for it?
> 
> Cheers,
> 
> Anthony.
> 
> 
> _______________________________________________
> 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/20210929/271ab9c4/attachment.htm>