[Mesa-users] Convergence issues with massive stars

Wed Aug 4 17:19:04 UTC 2021

Hi Frank,

I went through the files in $MESA_DIR/eos/private and compared with what I
did in run_star_extras and I think my mods indeed use the consistency
conditions you mentioned.

Essentially, what I am doing is adding a new boson to the EOS by performing
Bose-Einstein integrals and fitting the results as a function of density
and temperature.

My mods then add this to the EOS by first calling eosDT_get() and then
updating the entries in res(:) and the partial derivatives. For quantities
like pressure, entropy, C_V and derivatives at constant T or rho, these are
just additive. For other quantities, there are some rescalings to account
for the fact that the old pressure used in logarithmic derivatives needs to
have the boson added to it, and for others, I use thermodynamic relations
to compute them from scratch from the newly updated variables described
above.

Looking ta e.g. eos/private/eoselm_eval.f90, it looks like I am using the
same relations as the default EOS to calculate these changes, so the
consistency relations should be satisfied.

The version I sent that won't converge corresponds to a boson with some
specific mass. For other masses, the models converge as I would expect, and
even for this mass higher and lower mass models converge as well.

This is why I suspect it may be something that I can solve by changing one
of the controls but I'm happy to look at my EOS mods again too if you see a
problem with what I have done.

Cheers,

Jeremy

On Fri, Jul 30, 2021 at 10:38 AM Jeremy Sakstein <sakstein at hawaii.edu>
wrote:
>
> Hi Frank,
>
> I just sent my inlists and run_star_extras so that might answer some of
your questions.
>
> "i have no idea what eos modifications you've made, but if you've
modified the eos
> over any appreciable region in (rho,T) space, then its not helm, but a
helm with
> modifications on top, yes?"
>
> This is correct. I am calling eosdt_get and then updating the entries in
res(:) to include contributions from a new particle so it is HELM+DM. I
think the quantities e.g. gradad are computed using the relations you
mentioned below.
>
> For your second point, I am updating all of the partials as well as the
EOS variables so I think everything should be consistent. In fact, in an
earlier version I got failures until I updated the partials to high
precision.
>
> Cheers,
>
> Jeremy
>
> On Fri, Jul 30, 2021 at 9:46 AM Francis Timmes <fxt44 at mac.com> wrote:
>>
>>
>>
>> > On Jul 30, 2021, at 12:22 PM, Jeremy Sakstein <sakstein at hawaii.edu>
wrote:
>> >
>> > Thanks Frank,
>> >
>> > For (b), I am starting from helium cores, so as far as I can tell I am
always in the regime of the HELM EOS. I can get this to converge for lower
mass models (M<28 or so) and higher mass (M>40) no problem but it is just
this small range that gives the convergence issue.
>>
>> i have no idea what eos modifications you've made, but if you've
modified the eos
>> over any appreciable region in (rho,T) space, then its not helm, but a
helm with
>> modifications on top, yes?
>>
>>
>> > For (a), I am using fitting functions that have all quantities to 8
decimal places and tested that what MESA is spitting out matches what I put
in, but I am not sure how high a precision I need. Is there some benchmark
I should be shooting for?
>>
>> apologies for not being clear. starting from a helmholtz free energy F,
>> the pressure, entropy, and energy are related to the partial derivatives
of F.
>> in addition, the partials of P, S, and E - which are related to things
like
>> grad_ad, specific heats, and so on - have to obey consistency relations.
>> finally, the derivatives of grad_ad, c_v and so on should also have
continuous
>> and smooth behaviors. so its not necessarily the precision of a given
quantity,
>> although that helps, but how the entire thermodynamic framework hangs
together numerically.
>> for example, if an eos modification changes P but not its derivatives or
S or E and their
>> derivatives, then troubles usually ensue.
>>
>> again, i'm only raising the possibility that the behavior your seeing is
>> related to the eos modifications made. that the model runs without the
>> eos modification seems to be hint though.
>>
>> fxt
>>
>>
>>
>>
>>
>> >
>> > Cheers,
>> >
>> > Jeremy
>> >
>> > On Fri, Jul 30, 2021 at 8:53 AM Francis Timmes <fxt44 at mac.com> wrote:
>> > a comment from the cheap seats. changing the eos is fraught
>> > with danger due to (a) the myriad of thermodynamic relationships
>> > that must be satisfied numerically to high precision and
>> > (b) the mesa eos in 15140 and previous versions being a composite
>> > patchwork of different eos's with different treatments for how
>> > and where they merge with a neighboring eos. its possible that
>> > your modified eos is running into difficulties with (a) or (b)
>> > at the elevated temperatures associated with a ~30 msun zams model.
>> >
>> > fxt
>> >
>> >
>> >
>> >
>> > > On Jul 30, 2021, at 10:20 AM, Jeremy Sakstein <sakstein at hawaii.edu>
wrote:
>> > >
>> > > Thanks Rob!
>> > >
>> > > I am already using  so I turned off conv_vel to test that.
>> > >
>> > > This did speed up the runs but only in the sense that it hit the
phase where the time-step goes to -4.49 faster, so I don't think it
addressed the underlying issue.
>> > >
>> > > I am attaching some terminal output in case you have any more
suggestions.
>> > >
>> > > In the meantime, I am going to make a minimal working example that
encounters this problem so I can send my inlists and run_star_extras.
>> > >
>> > > Cheers,
>> > >
>> > > Jeremy
>> > >
>> > > On Thu, Jul 29, 2021 at 4:06 AM Rob Farmer <
robert.j.farmer37 at gmail.com> wrote:
>> > > Hi,
>> > > I've found lowering the starting core temperature helps low-Z models:
>> > >
>> > > pre_ms_T_c =2.8e5
>> > >
>> > > You may also want to turn off conv_vel until after core helium
burning.
>> > >
>> > > Rob
>> > >
>> > >
>> > > On Mon, 26 Jul 2021 at 19:21, Jeremy Sakstein <sakstein at hawaii.edu>
wrote:
>> > > Hi everyone,
>> > >
>> > > I'm trying to evolve Z= He cores from ZAHB to core collapse. I've
updated the EOS to include a new particle that softens the EOS and
everything runs fine for lower mass objects but for stars with ZAHB mass
30M I'm running into convergence issues.
>> > >
>> > > The time-step falls to -4.49 and stays there indefinitely.
>> > >
>> > > I've tried to vary parameters like the overshooting controls (both
location and f and f0 parameters) and turning on MLT++ but to no avail. The
same thing happens every time.
>> > >
>> > > I'm attaching some
>> > > Kippenhahn plots. "no_DM" (x-axis goes to model number 15,000) is
the default case with no modifications and DM (model number goes to 50,000)
includes my mods.
>> > >
>> > > I was hoping someone would be able to suggest other controls for me
to vary to try to get through to CC.
>> > >
>> > > Happy to send more info if this is not sufficient.
>> > >
>> > > Cheers,
>> > >
>> > > Jeremy
>> > >
>> > > <DM_2.png>
>> > >
>> > > <DM.png><noDM_2.png><noDM.png>
>> > >
>> > >
>> > > --
>> > > Jeremy Sakstein (He/Him/His)
>> > > Assistant Professor of Theoretical Physics
>> > > Department of Physics & Astronomy
>> > > University of Hawaiʻi at Mānoa
>> > > Watanabe Hall
>> > > 2505 Correa Road
>> > > Honolulu, HI, 96822, USA
>> > > http://www.jeremysakstein.com
>> > > _______________________________________________
>> > > mesa-users at lists.mesastar.org
>> > > https://lists.mesastar.org/mailman/listinfo/mesa-users
>> > >
>> > >
>> > >
>> > > --
>> > > Jeremy Sakstein (He/Him/His)
>> > > Assistant Professor of Theoretical Physics
>> > > Department of Physics & Astronomy
>> > > University of Hawaiʻi at Mānoa
>> > > Watanabe Hall
>> > > 2505 Correa Road
>> > > Honolulu, HI, 96822, USA
>> > > http://www.jeremysakstein.com
>> > > <DM_100_CV2.out>_______________________________________________
>> > > mesa-users at lists.mesastar.org
>> > > https://lists.mesastar.org/mailman/listinfo/mesa-users
>> > >
>> >
>> >
>> >
>> > --
>> > Jeremy Sakstein (He/Him/His)
>> > Assistant Professor of Theoretical Physics
>> > Department of Physics & Astronomy
>> > University of Hawaiʻi at Mānoa
>> > Watanabe Hall
>> > 2505 Correa Road
>> > Honolulu, HI, 96822, USA
>> > http://www.jeremysakstein.com
>>
>
>
> --
> Jeremy Sakstein (He/Him/His)
> Assistant Professor of Theoretical Physics
> Department of Physics & Astronomy
> University of Hawaiʻi at Mānoa
> Watanabe Hall
> 2505 Correa Road
> Honolulu, HI, 96822, USA
> http://www.jeremysakstein.com

--
Jeremy Sakstein (He/Him/His)
Assistant Professor of Theoretical Physics
Department of Physics & Astronomy
University of Hawaiʻi at Mānoa
Watanabe Hall
2505 Correa Road
Honolulu, HI, 96822, USA
http://www.jeremysakstein.com
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <https://lists.mesastar.org/pipermail/mesa-users/attachments/20210804/260bf37e/attachment.htm>