Detailed chemical studies of Solar-type stars have long been routine in stellar astrophysics, making possible studies in Galactic chemodynamics and exoplanet demographics. However, a similar understanding of the chemistry of M and late-K dwarfs—the most common stars in the Galaxy and most likely to host planets—has been greatly hampered by the complex molecular chemistry of their cool atmospheres. Here we present a new implementation of the Cannon, a data-driven model widely used in stellar astrophysics, developed for low–medium resolution optical (400–700nm) cool dwarf spectra. Our novel four parameter implementation in Teff, log g, [Fe/H], and [Ti/H] models both label uncertainties and missing labels, and is trained on 121 cool dwarf benchmarks—21 of which have literature elemental abundances measured from a warmer binary companion. Under leave-one-out cross-validation, we recover Teff, [Fe/H], and [Ti/H] with precisions of 2%, ±0.12 dex, and ±0.09 dex respectively, a precision which allows insight into the giant planet–[Fe/H] correlation for our sample of 65 TESS candidate planet hosts. Our work highlights the importance of precisely known benchmark systems; the promise of data-driven models for chemical analysis of the rich, but challenging to model, optical spectra of cool stars; and the utility of both in studying exoplanet demographics in the era of TESS and Gaia.

{"references": ["Rains, A. D. et al. (2021), MNRAS, 504, 5788"]}