Recovery of TESS Stellar Rotation Periods Using Deep Learning

Abstract

We used a convolutional neural network to infer stellar rotation periods from a set of synthetic light curves simulated with realistic spot-evolution patterns. We convolved these simulated light curves with real TESS light curves containing minimal intrinsic astrophysical variability to allow the network to learn TESS systematics and estimate rotation periods despite them. In addition to periods, we predict uncertainties via heteroskedastic regression to estimate the credibility of the period predictions. In the most credible half of the test data, we recover 10% accurate periods for 46% of the targets, and 20% accurate periods for 69% of the targets. Using our trained network, we successfully recover periods of real stars with literature rotation measurements, even past the 13.7 day limit generally encountered by TESS rotation searches using conventional period-finding techniques. Our method also demonstrates resistance to half-period aliases. We present the neural network and simulated training data, and introduce the software butterpy used to synthesize the light curves using realistic starspot evolution.

Rotation Distributions around the Kraft Break with TESS and Kepler: The Influences of Age, Metallicity, and Binarity

“…Strides are additionally being made in detecting long rotation periods from TESS light curves using more modern techniques such as machine learning (Lu et al 2020;Claytor et al 2022). An improved characterization of the TESS instrument as well as an improved understanding of scattered light across all TESS observing sectors will improve our ability to detect rotation signals beyond tens of days.…”

“…To correct and stitch together our light curves, we initially followed the methods performed in García et al (2011) on the Kepler sample. However, upon implementing this method on several simulated TESS light curves from Claytor et al (2022), we found that these methods preserved too many TESS-specific systematics for us to accurately measure rotation periods. In light of this, we devise our own method to correct for TESS systematics while preserving any possible rotation signals.…”

“…This detection limit is shown in Figure 9, where we compare the distribution of rotation periods detected in TESS to those detected in Kepler, where rotation periods up to tens of days were measured. Although other attempts have been made to measure rotation periods for stars that span multiple TESS observing sectors (e.g., Lu et al 2020;Martins et al 2020;Claytor et al 2022), up to this point, only studies using machine-learning-based rotation period extraction methods, i.e., neural networks, have successfully recovered rotation periods longer than 13.7 days (Lu et al 2020;Claytor et al 2022).…”

Activity and Rotation of Nearby Field M Dwarfs in the TESS Southern Continuous Viewing Zone

“…The resulting, nearly year-long light curves were processed to remove data affected by systematics. This processing, however, is still far from perfect, with TESS systematics not yet well enough understood for confident measurement of P rot longer than roughly half the length of a sector, or ≈15 days (Claytor et al 2022).…”

“…However, as was highlighted recently by Claytor et al (2022), there are fundamental issues with trying to measure longer P rot with TESS light curves produced by stitching together data from several 27 day sectors. In the resulting light curves, it is difficult to assess whether a signal is caused by true fluctuations in stellar brightness, or whether it is a result of systematics in the data either incorrectly removed or added in the processing of the light curves.…”