A Data-driven Spectral Model of Main Sequence Stars in Gaia DR3

Abstract: Precise spectroscopic classification of planet hosts is an important tool of exoplanet research at both the population and individual system level. In the era of large-scale surveys, data-driven methods offer an efficient approach to spectroscopic classification that leverages the fact that a subset of stars in any given survey has stellar properties that are known with high fidelity. Here, we use The Cannon, a data-driven framework for modeling stellar spectra, to train a generative model of spectra from the Gaia Data Release 3 Radial Velocity Spectrometer. Our model derives stellar labels with precisions of 72 K in Teff , 0.09 dex in log g, 0.06 dex in [Fe/H], 0.05 dex in [{\alpha}/Fe] and 1.9 km/s in vbroad for main-sequence stars observed by Gaia DR3 by transferring GALAH labels, and is publicly available at https://github.com/isabelangelo/gaiaspec. We validate our model performance on planet hosts with available Gaia RVS spectra at SNR>50 by showing that our model is able to recover stellar parameters at {\geq}20% improved accuracy over the existing Gaia stellar parameter catalogs, measured by the agreement with high-fidelity labels from the Spectroscopic Observations of Cool Stars (SPOCS) survey. We also provide metrics to test for stellar activity, binarity, and reliability of our model outputs and provide instructions for interpreting these metrics. Finally, we publish updated stellar labels and metrics that flag suspected binaries and active stars for Kepler Input Catalog objects with published Gaia RVS spectra.