2025A&A...704A.317O

Context. Stellar variability and transient events provide critical insights into many areas of astrophysics. Progress in these fields has been accelerated by high-precision space-based photometry missions such as CoRoT, Kepler, and K2. NASA's ongoing Transiting Exoplanet Survey Satellite (TESS) represents another significant milestone, offering a unique combination of long observational baseline, high cadence, and nearly all-sky coverage. However, extracting high-quality light curves from TESS full-frame images (FFIs) remains challenging due to contamination from scattered light, primarily from Earth or the Moon, and source blending in crowded fields.
Aims. In this study, we processed TESS FFIs to produce a comprehensive catalog of light curves for variable point sources observed during the satellite's prime mission. The resulting database is named TESS quick-look and light curve analysis (TEQUILA) and is intended to support diverse scientific investigations, enable large-scale statistical studies of stellar variability and transient phenomena, and relieve researchers of the need to process TESS FFIs from raw pixel data.
Methods. We applied the difference image analysis technique, constructing high signal-to-noise photometric reference images via the median combination of quality-filtered FFIs for each charge-coupled device and camera across TESS sectors 1-26. An iterative subtraction method was applied to mitigate instrumental systematics and other variable background features. Light curves were created using simple aperture photometry with a fixed 3-pixel radius centered on sources whose brightness was found to vary significantly in one of the residual images.
Results. Our pipeline yields over six million light curves of variable point sources from the first two years of TESS data. These include stellar variables, transient events, instrumental systematics, and moving objects. Approximately 6 × 10⁵ light curves span multiple sectors, with around 10³ originating from the continuous viewing zones. In the median normalized light curves, we achieve a median point-to-point differential variability noise level ranging from 10^–3 to 10⁰ for sources between 5.0 T_mag and 16.0 T_mag, while the typical photometric root mean square variability ranges from 10^–2 to 10¹. To identify light curves whose creation was prompted by instrumental systematic noise, we employed a convolutional neural network trained in a supervised learning framework. A score was assigned to each classification, reflecting the network's confidence in the predicted class. To avoid confusion between astrophysical variables and Solar System objects (SSOs), we also include in the catalog a flag that identifies light curves whose creation was prompted by known SSOs.
Conclusions. All extracted light curves are publicly accessible as a high-level science product through the Mikulski Archive for Space Telescopes (MAST). The new catalog can be used as a discovery tool for previously unknown variable point sources, such as astrophysical transients and moving SSOs. In future works, we aim to refine our methods, mitigate remaining systematics, classify the light curves by their phenomenological characteristics, analyze some of the newfound variables, and extend the catalog to include observations from the TESS extended mission.