Payne4GAIN: NLTE Corrections for Red Giants in Milky Way Mapper using H-Band Neural Network Emulators

arXiv:2607.22499 · astro-ph.SR, astro-ph.GA · Submitted 2026-07-24 · Read on arXiv

Pierre Thibodeaux, Alexander P. Ji, Nicholas Storm, Maria Bergemann, Rana Ezzeddine, Yuan-sen Ting, Andrew R. Casey, Emily Griffith, José Fernández-Trincado, Guilherme Limberg, Szabolcs Mészáros, Amaya Sinha, Danny Horta, Andrew K. Saydjari, Joel Brownstein

astro-ph.SR, astro-ph.GA

Submitted: 2026-07-24

Comments: 20 pages, 11 figures, 4 tables. To be submitted to Open Journal of Astrophysics

Code: https://github.com/TSFitPy-developers/TSFitPy

License: http://creativecommons.org/licenses/by/4.0/

The gist: The majority of spectroscopic surveys assume local thermodynamic equilibrium (LTE) during the modeling of stellar spectra.

Terminology

Abstract

The majority of spectroscopic surveys assume local thermodynamic equilibrium (LTE) during the modeling of stellar spectra. This assumption begins to break down for luminous stars, like the red giants targeted by SDSS-V's Milky Way Mapper Survey in its Galactic Genesis program. In this work, we present non-LTE (NLTE) abundances for 360,000 red giant stars in Milky Way Mapper DR19, from infrared APOGEE spectra. We generate NLTE spectra using precomputed departure coefficient grids for Na, Mg, Si, Al, Ca, Ti, Mn, and Ni. To fit APOGEE spectra at scale, we train neural network emulators (NNEs) to synthesize LTE and NLTE H-band spectra. After verifying that the NNEs are accurate, we fit the APOGEE spectra with ASPCAP results that fall within the same parameter range as the training data. We find strong NLTE effects on the order of 0.1,dex for Al, Mn, and Ti, and smaller effects for Si and Ni. We provide a catalog of the results of our LTE and NLTE fits, as well as NLTE-corrected ASPCAP abundances using a polynomial fit correction.

Sources

Related papers