[Mesa-users] MLT++ for massive MS stars?

[GM Mirouh]

Dear all,

I have already written to you a little while ago about this, and I am 
sorry to insist.
I am computing massive stars on the main sequence stopping once in a 
while to save a model (with mesa 12115). It works nicely for M<320Msun, 
but I am hitting all kinds of numerical difficulties for anything above 
that:

*1) *at some point during evolution, the timestep drops to almost 
nothing, and I think this is due to the luminosity hitting the Eddington 
limit (at Z=0.02, M=500, center_h1=0.45, my star has Teff=7600, logTcntr 
= 7.6 and logL=7.2 and evolves with log_dt=-0.3 ). It keeps running for 
a little while though, albeit very slowly, until...

*2)* my runs end up crashing with errors such as

" retry  193406
                      mlt_partials(mlt_cv_var,mlt_gradT) 
792                       NaN
  retry  193406
                      mlt_partials(mlt_cv_var,mlt_gradT) 
792                       NaN
backup  193406
                      mlt_partials(mlt_cv_var,mlt_gradT) 
774                       NaN
  retry  193405
                      mlt_partials(mlt_cv_var,mlt_gradT) 
774                       NaN"

Can anyone enlighten me about what is happening here? I am going out 
some interpolation table? is there something I can do?

*3)* I have tried playing with some numerical parameters, such as the 
starting timestep or the varcontrol value, and I have varied mixing 
parameters such as semiconvection or overshooting efficiencies, to no 
avail: the code as I have set it up seems to be digging a hole it cannot 
get out of.

*4) *Some people have mentioned MLT++ as a possible solution, while some 
others told me to avoid it like the plague. Recent events show people do 
not really avoid the plague, so I thought I could give MLT++ a try: I 
have turned "okay_to_reduce_gradT_excess = .true." but it does not seem 
to change much in terms of convergence (I have the exact same issues, 
only earlier on the MS). Do you know of a manual or do you have general 
advice for MLT++? Would it make sense to compute the MS as long as it 
works, save a model and then turn MLT++ on and proceed? Or maybe MLT++ 
is not the right approach at all?

At this point, I just want to reach hydrogen depletion for these stars, 
and I am ready to make reasonable concessions about the accuracy of my 
models as long as I know what is happening in there.
I am sorry I don't know what kind of files to attach here, but I can 
provide anything you think can help you to help me. Thank you for any 
advice you can give me.

Please take care,
Best.
-- 
Giovanni Mirouh

-------------- next part --------------
An HTML attachment was scrubbed...
URL: <https://lists.mesastar.org/pipermail/mesa-users/attachments/20201125/4a122c1e/attachment.htm>