Rotational Characterization of Tess Stars with Deep Learning

Abstract

Rotation is a fundamental property of stars. The Kepler mission revolutionized the field of stellar rotation, delivering periods of over 50,000 stars near the plane of the Milky Way. The distribution of periods revealed unexpected gaps, dips, and edges that cannot be described by current rotational evolution models, demanding new physical explanations. To sharpen the features in the distribution and to disentangle the effects of star formation history, more measurements of rotation are needed across the entire sky. The TESS mission has the potential to probe stellar rotation in millions of stars across the entire sky, but mission systematics—instrumental noise, observing gaps, and changes in detector sensitivity—have prevented recovery of rotation periods longer than 13.7 days. We used deep learning to see through TESS systematics and recover periods from year-long light curves. Our approach uses a training set of synthesized light curves from realistic star spot evolution simulations, with real light curve systematics from quiet TESS stars. Evaluating the network on real TESS data, we estimated reliable periods for 9,837 cool dwarfs. We recovered key features of the Kepler and K2 distributions, including periods up to 60 days. We reproduced the intermediate rotation period gap for the first time using TESS, as well as a dip in photometric activity surrounding it. Combining our TESS rotation periods with spectroscopic temperatures and abundances from APOGEE, we examined the detectability of rotation across fundamental stellar parameters, finding a strong dependence on temperature and age. Using gyrochronology, we inferred masses, ages, and other fundamental properties for the 6,632 TESS stars with APOGEE spectroscopy and corroborated evolution trends of Galactic chemistry and magnetic activity seen with Kepler. Finally with new measurements of spot filling factor from APOGEE, we investigated the spottedness of stars across the period distribution. We found that stars exhibit elevated spot fractions in the same regime where magnetic braking temporarily stalls in young open cluster stars, suggesting a common cause. Now with the ability to estimate rotation periods, including long periods, across the entire sky, we can characterize stars along many more lines of sight than before, enabling detailed study of the Galaxy's stellar populations.