Interior and atmosphere errors with low density, volatile rich sub-Neptunes

[IKisvardai]

Problem description

When running grids simulating 3 M_Earth, 2 R_Earth planet GJ 9827 d a number of simulations crashed and produced the errors in the attached log files.

A large percentage of simulations also take longer than 3 days to complete, while others finished within 3 days with either reached target time or total volatile escape.

Evidence

case_000001_proteus_00.log
case_000144_proteus_00.log
case_000149_proteus_00.log

My computer

Habrok cluster, Python 3.12.13, julia version 1.12.6

Relevant people

@timlichtenberg

[timlichtenberg]

Thanks for the detailed logs, Imre. The three crashes are all the same underlying problem: in these cases the volatile inventory meets or exceeds the total planet mass, so there is no mass left for a condensed interior. That configuration is unphysical, and it breaks three solvers in turn: the interior structure (Zalmoxis), the atmosphere (AGNI, where the envelope extends past the Hill radius), and the outgassing equilibrium (CALLIOPE).

I have pushed a fix to tl/fix-gj9827d-structure-coupling. PROTEUS now detects these cases and ends them cleanly with a clear log message instead of crashing or running for days. The shared message is Unphysical planetary regime, with a status code per cause:

• 20: non-positive dry mass (volatile envelope >= planet mass), interior side
• 22: atmosphere extends beyond the Hill radius and AGNI cannot converge
• 27: outgassing equilibrium did not converge (volatile-rich surface state)

To use it:

• Update your PROTEUS checkout to the latest tl/fix-gj9827d-structure-coupling.
• Point your Zalmoxis checkout at tl/fix-subneptune-zalmoxis-complex-R. This is the same ma/volatile-partitioning you are already on, plus a guard in the interior radius solver, so you keep everything and just gain the safety net.
• Exclude the affected cases from the grid by grepping the logs for Unphysical planetary regime, or by filtering on status 20, 22, or 27 in the status files.

This should also cut the runtime: the degenerate cells that were grinding through solver retries for days now fail in seconds. If you still see cases running past 3 days after this, let me know and we can check whether those are a separate bottleneck.

[nichollsh]

This has been identified as two separate issues:

• Oxidising conditions - missing O2 opacities (handled by new spectral files; New revised SOCRATES spectral files SOCRATES#8)
• Reducing conditions - atmosphere is unbound (handled by new ZEPHYRUS schemes; Add an escape regime dispatcher with per-species fractionation ZEPHYRUS#34)