[Mesa-users] New public release r11701 (recommended upgrade)
jwschwab at ucsc.edu
Fri May 3 19:24:52 EDT 2019
I'm pleased to announce the release of MESA r11701. The MESA developers
strongly recommend that you upgrade to this version if you are using
r11532 or r11554.
This release corrects two equation of state related issues present in
these versions (described in more detail below). We thank those (Conny
Aerts, Joey Mombarg, Warrick Ball) who provided problem reports.
Other changes are minor. The install process has been visually
streamlined. Archive files are now compressed using xz (instead of
bzip2) which results in a slightly smaller download size.
When upgrading, we also suggest that users upgrade to the recently
released 20190503 SDKs.
The website will be updated shortly.
Summary of EOS issues
The EOS issues are related to DT2 and ELM (see MESA V, Appendix A).
As of r11532, these EOS options were turned on by default, i.e.
use_eosDT2 = .true. ! replaces eosDT
use_eosELM = .true. ! replaces HELM
These EOSes rely on the same underlying microphysical data as previous
MESA versions, but interpolate differently.
eosDT2 Issue: X Interpolation
eosDT2 has been corrected to default to the same X (hydrogen mass
fraction) interpolation scheme as eosDT.
MESA r11532 and r11554 defaulted to linear interpolation in X. This was
a change in behavior from the old eosDT, which uses a higher order
scheme. One manifestation of this was mean molecular weight profiles
that are piecewise-linear (i.e. continuous but not smooth). This can
become particularly apparent in the Brunt.
eosELM Issue: Composition coverage
eosELM has been corrected to better match HELM and to automatically fall
back to HELM outside its range of applicability.
ELM (which is a tabulated version of HELM) could return significantly
different answers than HELM at the same conditions. Because the ELM
tables are only compiled at particular values of (abar, zbar), MESA
interpolates. The details of this interpolation, plus some issues with
table placement, meant that there were places where the values returned
by ELM were more than 1% different than the HELM value.
More information about the Mesa-users