ABYSS. IV. Identifying Signatures of Stellar Youth in APOGEE Spectra

Artikel i vetenskaplig tidskrift, 2026

We develop a convolutional neural network classifier that performs spectroscopic identification of stellar youth (<40 Myr) in Apache Point Observatory Galactic Evolution Experiment (APOGEE) spectra. This classifier is sensitive to several youth-related features, including rotational broadening, which is common in younger stars, and to several discrete lines that appear to be indicative of a very spotted photosphere. The model is successful at identifying youth across a wide range of stars, achieving its strongest performance at discriminating young M and K dwarfs, while maintaining useful discriminatory power for hotter stars as well. This work enables more robust separation of pre-main-sequence stars from more evolved sources in the field even when they have comparable Teff and (Formula presented) logg, providing a reliable means to substantially reduce contamination in photometrically selected young-stellar-object candidates. In addition to constructing a classifier for APOGEE spectra, we also perform classification of young stars in optical Baryon Oscillation Spectroscopic Survey spectra.