The Many Uses of Interpretable Data-Driven Spectral Models: Optical Cool Dwarf [X/Fe] and Flux Recovery Across Spectral Resolution

Le résumé fourni par la source

Detailed chemical studies of F/G/K stars have long been routine in stellar astrophysics, enabling studies in both Galactic chemodynamics, and exoplanet demographics. However, similar understanding of the chemistry of M and late-K dwarfs—the most common stars in the Galaxy—has been greatly hampered both observationally and theoretically by the complex molecular chemistry of their atmospheres. This complexity results in extensive line blending and a marked sensitivity to incomplete molecular data for dominant opacity sources when modelling, limiting the viability of traditional spectroscopic methods like equivalent width measurement or spectral synthesis—especially in the optical. Data-driven spectroscopic models using reliable stellar benchmark training samples offer a way around this problem while at the same time acting as an empirical bridge between observations and theory. Such models have been shown to effectively reproduce K/M dwarf [X/Fe] when trained on gold standard chemical benchmark F/G/K–K/M binary systems, and their generative nature allows empirically testing the accuracy of physical models. These comparisons can highlight problem spectral regions to either optimise analysis of current observational data, or guide future improvements in modelling or line lists. Here we present a comparison between low (R~7000) and high (R~46,000) resolution optical implementations of the data-driven Cannon model for cool dwarf Teff, logg, [Fe/H], and [X/Fe]. Both models show excellent recovery of stellar parameters (≲ 1.5% in Teff; ≲ 0.11 dex in [Fe/H] and [X/Fe]) and stellar fluxes, allowing direct comparison with current generation physical model spectra, as well as exoplanet demographic studies.

Ce résumé expose les affirmations des auteurs. BNTIC ne l’interprète pas comme une validation indépendante des résultats.