Stellar Magnetism, Activity, and Rotation with Time Series (SMARTS)

Conventional methods of detecting stellar rotation from TESS light curves have struggled to obtain periods longer than 13.7 days due to complicated systematics related to the telescope's orbit. Machine learning has been shown to see beyond TESS's systematics and obtain long periods, but it requires large training sets with known rotation periods. SMARTS DR1 consists a training set of synthetic light curves and binned wavelet transforms designed to mimic the full-frame image light curves of the TESS continuous viewing zones. The light curves were generated using physically realistic "butterpy" spot evolution models and include rotation, varying activity levels, magnetic cycles, spot emergence and decay, and latitudinal differential rotation. They are combined with real TESS galaxy light curves and stitched sector-to-sector to emulate TESS's systematics and noise. This HLSP contains 1 million simulations spanning rotation periods of 0.1—180 days. In SMARTS DR2, the authors present a new training set based on the Kepler Bonus Background light curves. This training set was used to infer new rotation periods from de-blending Kepler sources using machine learning. Like DR1, the light curves use butterpy simulations to create the rotational modulation, but use flattened red clump light curves as the noise templates.