Data Driven Spectral Models for APOGEE M dwarfs

The Cannon (Ness et al. 2015) is a flexible, data-driven spectral modeling and parameter inference framework, demonstrated on high-resolution Apache Point Galactic Evolution Experiment (APOGEE; λ/Δλ~22,500, 1.5-1.7µm) spectra of giant stars to estimate stellar labels (Teff, logg, [Fe/H], and chemical abundances) to precisions higher than the model-grid pipeline. The lack of reliable stellar parameters reported by the APOGEE pipeline for temperatures less than ~3550K (Schmidt et al. 2016), motivates the extension of this approach to M dwarf stars. Using a training set of 51 M dwarfs with spectral types ranging M0-M9 obtained from SDSS optical spectra, we demonstrate that The Cannon can infer spectral types to a precision of 0.6 types. We then use 30 M dwarfs ranging 3072 < Teff < 4131K, and -0.48 < [Fe/H] < 0.49 to train a two-parameter model precise to 44K and 0.05 dex respectively. Additionally we discuss the extension of a model to other labels, and the scientific objectives a data-driven pipeline could enable