[Mesa-users] Timestep and convergence problems for massive stars

[Francis Timmes]

hi jeremy,

ok. i'll give this one a whirl.

basically, yes. i've seen the same with public release 12778.
i found no magic bullet(s) in 12778 for, say, the solar metallicity 
100 msun zams model (others can weigh in here).

the issue is the behavior of the outer layers as helium depletes - large velocities 
and wild excursions in the hr diagram. these are driven, i think, from a large luminosity 
meeting the opacity bumps and the attendant density inversion in the radiation dominated outer layers. 

for my purposes, which may not be suitable for your purposes, i noted the 
opacity in the vast interior is dominated by simple electron scattering. 
following that line, i evolved a 100 msun model from the pre-ms to the ms as usual.
at the ms i then set use_simple_es_for_kap to true. the model then easily
went to core he-depletion and beyond (my purpose). i stress one should consider 
if this approximation is appropriate, and how it may impact, what one is modeling.

fxt







> On Nov 19, 2020, at 12:47 PM, Jeremy Scholtys via Mesa-users <mesa-users at lists.mesastar.org> wrote:
> 
> Hi,
> 
> My classmate and I are trying to study the effect of metallicity and rotation on the mass loss history of massive stars from the Pre-MS to the end of He burning in the center.
> Our parameter set includes 4 different masses (15, 25, 50 and 100Msun), 4 rotation omega values (0, 0.2, 0.4, 0.6 omega/omega_crit) and 4 different metallicities (Z=0.01*Zsun, Z=0.1*Zsun, Z=Zsun, Z=0.04). 
> We are running the mesa version 12778.
> 
> Error:
> The stars with M =15 or 25 Msun can run with our 4 metallicities values, i.e. for Z=0.000142 to Z=0.04.
> While the simulation also runs with M=50Msun at Z=0.0142, they crash at lower metallicities when they get to the end of the MS. The inlist and error message are attached to the email.
> I tried changing the mesh resolution by decreasing mesh_delta_coeff, choosing a lower value for the maximum timestep and a  lower varcontrol, but apart from taking more time to run, the error is the same when it crashes.
> It looks like the star can't converge when it reconfigures itself for the helium burning phase in the core.
> 
> Problem:
> In the case of our 100Msun run for Z=Zsun=0.0142, the timestep drops dramatically half-way through the helium burning phase in the center, and at that rate, the simulation would take at least a few days to run. 
> The limiting factor of the timestep is always said to be varcontrol. The inlist for the 100Msun is also attached to the email.
> I tried multiple things to get the simulation to finish faster, like slightly increasing the varcontrol value comparatively to the standard value.
> I also tried increasing the minimum timestep to a few years (8, 16, 32, 64, 128 years). While they run faster, the mass-loss history post-MS depends too sensitively on their chosen value, as is the mass of the carbon core.
> Are there control parameters that allow the code to evolve faster in that phase without short-cutting the relevant physics?
> 
> Cheers,
> 
> Jeremy Scholtys
> 
> 
> 
> <Error50MsolZ0.00142.txt><inlist_PreMS_to_HeExhaustion_50Msol><inlist_PreMS_to_HeExhaustion_100M>_______________________________________________
> mesa-users at lists.mesastar.org
> https://lists.mesastar.org/mailman/listinfo/mesa-users
>