The Asteroid Terrestrial-impact Last Alert System (ATLAS) carries out its primary planetary defense mission by surveying about 13000 deg^2 at least four times per night. The resulting data set is useful for the discovery of variable stars to a magnitude limit fainter than r~18, with amplitudes down to 0.01 mag for bright objects. Here we present a Data Release One catalog of variable stars based on analyzing 142 million stars measured at least 100 times in the first two years of ATLAS operations. Using a Lomb-Scargle periodogram and other variability metrics, we identify 4.7 million candidate variables which we analyze in detail. Through Space Telescope Science Institute, we publicly release lightcurves for all of them, together with a vector of 169 classification features for each star. We do this at the level of unconfirmed candidate variables in order to provide the community with a large set of homogeneously analyzed photometry and avoid pre-judging which types of objects others may find most interesting. We use machine learning to classify the candidates into fifteen different broad categories based on lightcurve morphology. About 10% (430,000 stars) pass extensive tests designed to screen out spurious variability detections: we label these as `probable variables. Of these, 230,000 receive specific classifications as eclipsing binaries, pulsating, Mira-type, or sinusoidal variables: these are the `classified variables. New discoveries among the probable variables number more than 300,000, while 150,000 of the classified variables are new, including about 10,000 pulsating variables, 2,000 Mira stars, and 70,000 eclipsing binaries.
تحميل البحث