[Mesa-users] Binary evolution problem

[Pablo Marchant]

Hi Margaret,

On Mon, Aug 19, 2019, 19:28 Margaret Streamer via Mesa-users <
mesa-users at lists.mesastar.org> wrote:

> Hi all,
>
> I have recently upgraded from mesa -r8845 to mesa -r11701.
>
>
>
> I have successfully installed the new version on a Linux system and can
> compile and start my code running OK.  However, running some inlists and
> inlist-project that worked successfully using
>
> -r8845 now fail on -r11087.
>
>
>
> I am modelling Algol-type binaries with conservative mass transfer.
>
>
>
> With -r8845, I had investigated the effect of changing mixing length
> alpha and found that for m1, m2 values of 2 and 1.8, respectively, were the
> most optimal.  With -r11701, these values give code terminations as
> described below.  If I change mixing length alpha values for  m1, m2 to
> 1.6, 1.5,resp, then I am able to get some models to run as expected.
>
>
>
> Also I am now finding that small changes (0.05) in initial masses of the
> two stars or small changes in initial orbital period (0.1 or less) result
> in termination of the code with either:
>
>
>
> “termination code: Terminate because accretor (r-rl)/rl >
> accretor_overflow_terminate” or “terminated evolution: cannot find
> acceptable model; termination code: min_timestep_limit”.
>
>
>
> This was not happening with -r8845 where  I could investigate a wide range
> of initial conditions.
>

>
> I attach a sample inlists etc that worked for one condition.
>
>
>
> I had assumed backwards compatibility between the two MESA releases but
> maybe I’m missing something?
>
We cannot hope to comprehensively assure everything modelled on older
versions will still work on newer ones. As we are not an army of developers
we can't either support every single feature forever, parts can get
deprecated and removed. Please bear that in mind and consider that there
have been plenty of changes in the 3+ years since 8845. We operate as a
community, and depend on our users to keep things going steady.

These are a few things you could try to see if your models can work better.
- You appear to load some zams tables at the beginning. Have these been
recreated with the latest version of MESA or are you using the same old
ones? The EOS among other things has undergone various changes since 8845,
and that could be an issue.
- We have also adopted stringent tolerances for the equations. You can see
if this makes a difference by setting use_gold_tolerances=.false., but I
would not recommend just leaving that if it makes your models work.

Perhaps your models can work with gold tolerances if you use

use_dedt_form_of_energy_eqn = .true.
use_eps_mdot =.true.

See the latest mesa paper (paxton+ 2019) for details.
- If the above don't work, then you need to find at which point there was a
change in MESA that broke your models. For us to help you with this, you'd
need to go through the releases that have been made since 8845 and try to
pinpoint when the issues started.
- Are you sure your models with 8845 were converged? At some point since
8845 (don't know off the top of my head) the timestep controls in binary
where adjusted, leading in generall to smaller timesteps during mass
transfer. If you observe physical differences in the outcomes of your
models this could be related to it. You indicate for example that some
models terminate because they go into contact, that is a different physical
outcome, so if older models did something different you should verify the
new and old MESA results are converged.

Please let us know if you can't find a solution after trying these
suggestions.

Cheers!

>
>
> Regards,
>
> Margaret
>
>
>
>
> _______________________________________________
> 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/20190819/57d6be60/attachment.html>