[mesa-users] Z=0.04 and MESA r9793

[Josiah Schwab]

Hi Aaron,

> I've done experiments like this in the past.  This situation seems to be in
> line with those.  The key difference in the EOS is the treatment of
> ionization.  HELM assumes full ionization and so will have a very different
> mean molecular weight than the OPAL tables in the outer layers of the
> model.  This is a metal-rich model evolving toward the RGB.  The attached
> shows two different evolutionary tracks, done with the same inlist and
> initial model supplied by Philip.  1000 timesteps in each.  The only
> difference is that I've only one EOS or the other in each.
> [image: Inline image 1]
> The difference is quite striking and becomes more pronounced as Teff
> decreases.  I'm not saying there is anything wrong with HELM, it does what
> it's designed to do. The only problem occurs when it is used out of
> context.

Indeed; that's the physical explanation.  (I'm still slightly puzzled by
the difference in behavior between r8845 and r9793.)

There is one subtlety of this whole EOS mess, which is that there are a
bunch of places where MESA attempts to mock up the effects of ionization
in HELM by blending between a version that includes the electrons and a
version that doesn't.  There are at least 3 such blends that I'm aware
of, all in different places in the code.  There's one embedded in the
MESA tables, one inside the HELM code, and then a user-controllable one
in the eos module.  The controllable one is

  http://mesa.sourceforge.net/star_job_defaults.html#set_HELM_ion_neutral_blends

but the default values are quite high for a hydrogen-helium dominated
composition.  (This blend is usually only hit in evolved stars.)

> I would suggest based on this that we should have a test_suite case
> that specifically checks the behavior at the Z=0.04.

Yes, that is a good idea.

Josiah