[Mesa-users] Differences between MESA versions

[Josiah Schwab]

Hi James,

Is it prudent to trust newer versions of the code over older ones, in this
> case?
>

This is a difficult question to answer.  We certainly strive to improve
MESA with time, so I tend to favor more recent versions, but one can't make
a blanket statement that models from newer MESA versions are more
trustworthy.

Ultimately, confidence in the models gets built by:

   Understanding the option choices (including both options you explicitly
set and key defaults) and their importance in the evolution of the model.
   Confirming that the models are converged (i.e., checking that the
results do not significantly depend on the time and spatial resolution)
   Comparing to other solutions (analytics, previous papers, different
codes, etc.)

Kudos to you for hitting that last point by doing careful comparisons to
past work.  The first point can be extra tricky when doing cross-version
comparisons.  Defaults change, internals are revised, etc.

Without seeing inlists or looking at internal structures, I'd say the
differences between r8845 and r15140 look like they're at the level where
they might plausibly be explained by differences in input physics (mostly
eos & kap) between those versions, though I am not confident in that
explanation.

As you note, the r12115 results look wrong.  My guess is that there's
something weird in the inlists.  Feel free to send them along and I can
take a quick look and see if anything jumps out.

Josiah
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <https://lists.mesastar.org/pipermail/mesa-users/attachments/20210225/56ece22f/attachment.htm>