[Mesa-users] 'bad negative xa' problem keeping timstep small

[Francis Timmes]

hi arman,

i looked at your run_star_extras.f90 but have not run your model, 
so at this point i can only offer some general debugging guidance.

does the model run as expected when your increased_mixing_diffusion_coefficient 
and diff_coeff_calc routines are removed?

does the model run as expected when the imposed timescale for mixing is extremely long?
if yes, explore what happens as the timescale is slowly shortened to the desired 600 sec.

does the model run as expected when the chosen mixing region is made very small?
if yes, explore what happens as the mixing region is slowly expanded to the desired envelope mass.

the timestep controls seem fairly generous.
you might explore what happens when much smaller timesteps are used.

fxt






> On Aug 18, 2021, at 8:23 AM, Arman Aryaeipour via Mesa-users <mesa-users at lists.mesastar.org> wrote:
> 
> Hello MESA users,
> I'm using mesa-r15140 to run nova models atop a carbon-oxygen white dwarf.
> However, my simulation experiences a problem where the timestep size is driven down to ~1 second or less due to “retry: some xa < min_xa_hard_limit” and kept there for many tens of thousands of timesteps. This retry message is printed at every 3-4 timesteps and appears to be the reason why my timetep size cannot get any bigger. Using the the control “report_bad_negative_xa = .true.”, I can see that the 'bad negative xa' are from the cells within the burning region of the WD-envelope interface, and specifically from the CNO cycles beta-minus decaying isotopes synthesized there (mostly N13).
> In my model, I use the other_D_mix hook routine to implement some mixing between the white dwarf and its accreted envelope when the maximum temperature reaches 10^8 K.
> This mixing phase lasts for about 600 seconds, and the mixing region extends into a portion of the WD substrate and out to the outer parts of the envelope (it does not include the surface, however). The reduced timestep due to the consistent “retry: some xa < min_xa_hard_limit” message arises after this mixing phase. This problem stops the model finishing the nova cycle.
> I understand that artificially mixing C12 from the WD into the hydrogen rich envelope using the other_D_mix hook routine will synthesize a much greater quantity of beta-minus decaying isotopes, as compared to no artificially implemented mixing. But why does the solver consistently get bad mass fraction values for the beta-minus decaying isotopes when the timestep size is 1s, which is much smaller than the half-life of the isotopes? Using a higher mesh resolution and smaller value of 'varcontrol_target' does not seem to help the problem.
> How can I get the simulation to have a larger timestep size after my implemented mixing has finished, and stop the solver from producing 'bad negative xa'?  
> This problem can be reproduced using the attached inlists, modified run_star_extras.f routine and the CO WD model.
> The inlist 'inlist_before_mix' evolves the WD model through the accretion phase and reduces the timstep to 60s once the maximum temperature in the model approaches 10^8K, this is to prepare the model for the implemented mixing phase. Once the model reaches 10^8K, the run terminates and outputs the model 'model_at_10^8K.mod'.
> The inlist 'inlist_do_mix' uses the 'model_at_10^8K.mod' model to carry out the mixing phase, and then continues the evolution of the model. It is after this point that I get the 'bad negative xa' problem which limits the timestep size.
> I would be grateful for any suggestions anyone may have regarding this issue! Thank you to anyone who took the time to read this message, and my apologies that it is on the longer side!
> Kindest regards,
> Arman 
> <run_star_extras.f90><inlist_before_mix><inlist_do_mix><wd2_CO_L--2.mod>_______________________________________________
> mesa-users at lists.mesastar.org
> https://lists.mesastar.org/mailman/listinfo/mesa-users