[Mesa-users] Time step error

[Pablo Marchant]

Hi Steven, I've been playing a bit with your input files and can give you a
few pointers. First, a couple of things
- in your last email you said you had issues with a 46 Msun model, but your
input files are for a 47 Msun model. I tried the 47 Msun model and found
the issue you describe
- Do you consistently run into these problems, or is this just one failing
model among many others? It is not uncommon when running a large set of
models for some of them to fail, if you really want an all-encompassing
solution that runs flawlessly that will require a significantly larger time
investment than you might have allotted.

Now, after running into timestep issues with large numbers of retries and
backups, its always useful to set on the option
report_hydro_solver_progress = .true.
in the controls section of your inlists. This will give you an idea as to
what equation in particular is causing trouble. By doing that I noticed
that the main issues come from the element abundance equations. During late
stages of burning the steep temperature dependence of these equations can
make them very challenging to solve, leading to these problems.There are
two potential "solutions" you might want to try:

- reduce the tolerance of the equations being solved. MESA uses what we
call "gold" tolerances by default, but it switches to lower tolerances in a
way controlled by the parameter
maxT_for_gold_tolerances = 5d9
if you were to lower this value, the solver might have an easier time going
through the pulses. But you should be careful about this, when not using
gold tolerances MESA can settle for a solution once the newton iterations
have small per-step variations, even if the equations themselves could have
significant residuals. It's up to you to evaluate the trustworthiness of
the results on whatever way you find appropriate.

- MESA also has a split burning option, which separates the burning
equations from the rest. This might help you prevent the numerical issues
you have without sacrificing the tolerance for the equations. You can
activate it by using
    ! split burning
    op_split_burn = .true.
    op_split_burn_min_T = 3d9

    op_split_burn_eps = 1d-7
    op_split_burn_odescal = 1d-8
in your controls, where I set the minimum temperature to a value
representative of the onset of pulses. But beware, this will make the code
much slower!

So you get to choose your poison. As usual, there's likely a better
solution, but finding it might not be simple.

Let me know how it goes.

Cheers

On Fri, Aug 2, 2019 at 11:00 AM Wampler, Steven via Mesa-users <
mesa-users at lists.mesastar.org> wrote:

> Hi Frank/everyone,
>
> *Here's a little more information about what I'm doing. I'm using the
> inlists and run_star_extras* downloaded from the Marchant paper
> repository on PPISN and using the current revision of MESA. The paper used
> r11123 ( I think). We expected to run sim on a 46 M_sun star and have it go
> through several pulsations then reach core-collapse. However after the
> first pulse the time step tanks and I'm not sure why or what steps to take
> to try and fix it.
>
> Thanks,
> Cole Wampler
>
>
>
> On Fri, Aug 2, 2019 at 6:11 AM Francis Timmes <fxt44 at mac.com> wrote:
>
>> hi steven,
>>
>> usually its advantageous to show what efforts have been made
>> to isolate or resolve an apparent issue before asking for help.
>>
>> for example, what physical changes might be going on that can cause the
>> timesteps to drop?
>> how does the model respond when adjusting parameters relevant to those
>> physical changes?
>>
>> fxt
>>
>>
>>
>>
>> > On Aug 1, 2019, at 11:16 AM, Wampler, Steven via Mesa-users <
>> mesa-users at lists.mesastar.org> wrote:
>> >
>> > Hey All,
>> >  slurm-23431158.out
>> >
>> > I'm having an issue where my MESA sim doesn't run to core implosion it
>> stops before. Attached is my insists and a shot of the slurm output. It
>> keeps ending on a maximum number of retrys and the timestep at the end is
>> like 10^-9 years.
>> > <image.png>
>> >
>> <inlist_ppisn><inlist_after_first_pulse><inlist_hydro_on><inlist_extra><inlist><inlist_hydro_off><history_columns.list><profile_columns.list>_______________________________________________
>> > mesa-users at lists.mesastar.org
>> > https://lists.mesastar.org/mailman/listinfo/mesa-users
>> >
>>
>> _______________________________________________
> mesa-users at lists.mesastar.org
> https://lists.mesastar.org/mailman/listinfo/mesa-users
>
>

-- 
Pablo Marchant Campos
M.Sc on Astrophysics, Universidad Católica de Chile
PhD on Astrophysics, Argelander-Institut für Astronomie, Universität Bonn
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <https://lists.mesastar.org/pipermail/mesa-users/attachments/20190805/cfd336e5/attachment.html>