Migrating MESA to GitHub

Earlier today, I was pleased to announce that the MESA stellar evolution code is now publicly hosted on GitHub.

The MESA stellar evolution code began its life in 2007 as a Subversion repository hosted on SourceForge.

MESA is somewhat unusual in that a checkout of the code is multiple GB in size. This is because it versions its key binary assets (e.g., equation of state data files) along with the code. In addition, there are also numerous large text files (containing stellar models) that can frequently change, but whose diffs are not meaningful.

The SVN proved satisfactory for MESA’s early life, especially as its development was primarily due to a single person (Bill Paxton).

Around 2016-2017, the developer base had expanded to include several other regular contributors and SourceForge was becoming increasingly unreliable.

At that time, we considered moving to git, but solutions for versioning large binary files with git were in their infancy. Instead, we migrated the SVN to a paid hosting plan at Assembla, immediately resolving our performance problems.

Developers who wanted to interface via git (including myself), were able to do so using git-svn. This was an acceptable stop-gap measure, but the large binary files meant that the git repository constructed in this way was in excess of 50 GB and generally suffered from poor performance when doing complex operations.

With an understanding that a git migration remained desirable, sustained effort was applied to reduce the number of large files tracked in the repository. Some simple improvements were realized by removing various types of cruft. Others required more involved structural changes. For example, redesigning the testing infrastructure so that solutions could be compared between different machines via a central server removed the necessity to version-control reference stellar model files.

By 2020, the MESA developer base had continued to grow, GitLFS was mature, and there was a desire to more effectively engage with a new generation of users and developers. With the help of my colleagues, especially Rob Farmer and Bill Wolf, we put together a plan to migrate MESA to GitHub. In advance of the migration, documentation was written to provide a basic introduction to git/GitHub for those who were unfamiliar.

In the end, we chose not to include the complete MESA development history in the git repository. Instead, the git repository was initialized to the most recent release (r15140) and a modest number of patches originally committed to the SVN were applied to the git repository.

The MESA developers moved to git with minimal interruption in mid-December 2020. We took about 6 months to get comfortable and establish and document new workflows, and then finally made the repository public. This should pave the way for a new era of collaborative MESA development and interaction with the astrophysics community.

KITP Conference: White Dwarfs from Physics to Astrophysics

At the KITP conference White Dwarfs from Physics to Astrophysics, I gave a talk entitled Ashes to Ashes: White Dwarfs from Double White Dwarf Mergers.

UFRGS Astronomy Seminar

I am “visiting” Universidade Federal do Rio Grande do Sul to give the Astronomy Seminar. A recording of my talk is available on YouTube.

Evolutionary Models for the Remnant of the Merger of Two Carbon-Oxygen Core White Dwarfs

The Astrophysical Journal, Volume 906, Issue 1, id.53, 14 pp.

Josiah Schwab

We construct evolutionary models of the remnant of the merger of two carbon-oxygen (CO) core white dwarfs (WDs). With total masses in the range 1-2 M, these remnants may either leave behind a single massive WD or undergo a merger-induced collapse to a neutron star (NS). On the way to their final fate, these objects generally experience a ∼10 kyr luminous giant phase, which may be extended if sufficient helium remains to set up a stable shell-burning configuration. The uncertain, but likely significant, mass-loss rate during this phase influences the final remnant mass and fate (WD or NS). We find that the initial CO core composition of the WD is converted to oxygen-neon (ONe) in remnants with final masses ≳1.05 M. This implies that the CO core/ONe core transition in single WDs formed via mergers occurs at a similar mass as in WDs descended from single stars and thus that WD-WD mergers do not naturally provide a route to producing ultramassive CO-core WDs. As the remnant contracts toward a compact configuration, it experiences a "bottleneck" that sets the characteristic total angular momentum that can be retained. This limit predicts that single WDs formed from WD-WD mergers have rotational periods of ≍10-20 minutes on the WD cooling track. Similarly, it predicts remnants that collapse can form NSs with rotational periods ∼10 ms.

Northwestern CIERA Seminar

I am “visiting” Northwestern University to give the CIERA Astro Seminar. A recording of my talk is available on YouTube.