Classification of Periodic Variable Stars from TESS

TL;DR

This paper classifies over 72,000 periodic variable stars from TESS data into 12 sub-types using a random forest approach, significantly expanding known catalogs and aiding astrophysical research.

Contribution

The study introduces a new classification method for periodic variable stars using TESS data and 19 parameters, identifying 63,106 newly classified stars and improving catalog purity.

Findings

Classified 72,505 variable stars from TESS data.

Achieved 94.2% to 99.4% purity for eclipsing binaries and pulsating stars.

Discovered 63,106 new variable stars, enriching existing catalogs.

Abstract

The number of known periodic variable stars has increased rapidly in recent years. As an all-sky transit survey, the Transiting Exoplanet Survey Satellite (TESS) plays an important role in detecting low-amplitude variable stars. Using 2-minute cadence data from the first 67 sectors of TESS, we find 72,505 periodic variable stars. We used 19 parameters including period, physical parameters, and light curve (LC) parameters to classify periodic variable stars into 12 sub-types using random forest method. Pulsating variable stars and eclipsing binaries are distinguished mainly by period, LC parameters and physical parameters. GCAS, ROT, UV, YSO are distinguished mainly by period and physical parameters. Compared to previously published catalogs, 63,106 periodic variable stars (87.0$\%$) are newly classified, including 13 Cepheids, 27 RR Lyrae stars, $\sim$4,600 $\delta$ Scuti variable stars, $\sim$1,600 eclipsing binaries, $\sim$34,000 rotational variable stars, and about 23,000 other types of variable stars. The purity of eclipsing binaries and pulsation variable stars ranges from 94.2$\%$ to 99.4$\%$ when compared to variable star catalogs of Gaia DR3 and ZTF DR2. The purity of ROT is relatively low at 83.3$\%$. The increasing number of variables stars is helpful to investigate the structure of the Milky Way, stellar physics, and chromospheric activity.