[mesa-users] mesa release 5232

[Bill Paxton]

Hi,

I've been busy working on nuclear reaction nets.  
More about that later.  Here are some of the other things in this release.



Roni Waldman has contributed code to turn on velocities prior to core collapse caused by gamma1 < 4/3.
The control is called gamma1_integral_for_v_flag and lives in star_job.defaults
         
      gamma1_integral_for_v_flag = 0d0
         ! automatically turn on velocities if center_gamma1_integral drops below this limit.
         ! this is useful for evolution leading up to pair instability core collapse.
         ! integral is sum over all cells of (gamma1-4d0/3d0) weighted by dm*P/rho



Rich Townsend has continued to improve GYRE so that it is now ready for use with star/astero.
Check astero/defaults/astero_search_controls.defaults for details.



As part of the changes to improve reliability of late stage burning, I've added these new controls  for timestep selection.

            delta_Ye_highT_limit = 1d-5 ! limit for magnitude of max change in Ye in high T cells
            delta_Ye_highT_hard_limit = -1
            minT_for_highT_Ye_limit = 7d9 ! only consider delta_Ye_highT_limit if cell T >= this limit



Also related to late stage burning, there are some new controls to "soften" the extreme stiffness of certain reactions at very high T's.  These apply to the reactions that toss around neutrons and protons as the temperatures reach NSE levels.  Dave Arnett suggested that we should be able to scale down these rates to make the equations easier to solve while still keeping the rates large enough to give a good approximation to equilibrium abundances.   It is like with convective mixing --- we can often get good mixing without using the actual large diffusion coefficients from MLT that in some case would cause ill-conditioning of the matrices.  Enough really is enough in some cases -- and more than enough can cause unnecessary problems.  The hope is that we can similarly limit some of the troublesome reaction rates at high temperatures to reduce numerical problems without sacrificing adequate accuracy.  This will of course need testing, lots of it.    But the initial results are looking good enough to be encouraging.  And it is nice that we haven't had to limit  pure alpha capture/emit rates -- just the rates using p's and n's, and just at very high temperatures.

Of course if you want to opt out of this, you can do so by setting turn_off_T_limits = .true.

Otherwise, you can experiment with size of the decreases in rates and the T ranges where the decreases are used.  The relevant controls are in star_job.defaults -- look for "rate limit factors for high temperatures".



Finally, the approx family of nets has had a population explosion!   As you probably know, these nets are for advanced burning and are intended to get energy generation "right" without worrying about exact composition details.
The foundation of the family is approx20.net -- it has these isos:

         neut h1 he3 he4 c12 n14 o16 ne20 mg24 si28 s32 ar36 ca40 ti44 cr48 fe52 fe54 fe56 ni56 prot

BTW: approx20 is actually the approx19 net from Frank Timmes with fe56 added; it has an ancient lineage!  The basic idea is to limit the net (from oxygen up) to even-even isotopes (i.e., even numbers of both protons and neutrons), and link them by compound reactions that assume equilibrium abundances for intermediates not included in the net.

The approx20 net does a good job with the energy, but if the Ye profile in the core matters, we've suggested using approx21.net.  It adds cr56 and a compound multi electron capture reaction from fe56 to cr56 using a fake rate based on ni56 electron capture.  That was useful, but if you wanted a more accurate Ye profile, the next option was a "soft wired" net with a large set of isotopes; e.g., Dave uses a net with 177 isos for this sort of thing.  I wanted to see if we could extend the "hard wired" approach of approx21 to get a relatively small net that would give sufficient accuracy for some (many?) uses without resorting to fake reaction rates.   

I've ended up with a series of new nets, ranging from 25 isos up to 50.  The approx25 net is now used in many test_suite cases. I'm hopeful that it will be the new workhorse to replace approx21 for massive stars (25 is the new 21).  The larger approx nets may be useful in special situations and provide a relatively quick check of how the small approx nets are doing.  Of course a real verification will require running test cases with one of the very large soft nets to compare.  As usual, I'll leave that to interested users!  Any volunteers?

Here are some new approx nets and the isos each adds (see above for the isos in approx20 and 21).

approx25 -- ti50 cr54 fe58 fe60
approx27 --  o18 ne22
approx31 -- ni58 ni60 ni62 ni64
approx36 -- ti46 ti48 ti52 cr50 cr52
approx40 -- ca42 ca44 ca46 ca48
approx41 -- ni66
approx42 -- fe62
approx46 -- zn64 zn66 zn68 zn70
approx50 -- ge70 ge72 ge74 ge76

Here's the full set for approx50 with alpha chains in columns.
The approx25 isos are in black and the rest are in red.

      h1 he3 he4 neut prot c12 n14
       o16  o18
      ne20 ne22
      mg24
      si28
       s32
      ar36
      ca40 ca42 ca44 ca46 ca48
      ti44 ti46 ti48 ti50 ti52
      cr48 cr50 cr52 cr54 cr56
      fe52 fe54 fe56 fe58 fe60 fe62
      ni56 ni58 ni60 ni62 ni64 ni66
                zn64 zn66 zn68 zn70
                     ge70 ge72 ge74 ge76

So far I've been able to resist the temptation to add more --
but it cries out for mg26, si30, s34, s36, ar38, and ar40;
approx56.net anyone?


Cheers,
Bill






-------------- next part --------------
An HTML attachment was scrubbed...
URL: <https://lists.mesastar.org/pipermail/mesa-users/attachments/20130718/e8ad0ebc/attachment.html>