A Search for Exoplanet Candidates in TESS 2-min Light Curves using Joint Bayesian Detection

In this work, we apply an exploratory joint Bayesian transit detector (Taaki et al. 2020), previously evaluated using Kepler data, to the 2 min simple aperture photometry light curve data in the continuous viewing zone for the Transiting Exoplanet Survey Satellite (TESS) over three years of observation. The detector uses Bayesian priors, adaptively estimated, to model unknown systematic noise and stellar variability incorporated in a Neyman-Pearson likelihood ratio test for a candidate transit signal; a primary goal of the algorithm is to reduce overfitting. The detector was adapted to the TESS data and refined to improve outlier rejection and suppress false alarm detections in post-processing. The statistical performance of the detector was evaluated using transit injection tests, where the joint Bayesian detector achieves an 80.0 % detection rate and a 19.1% quasi-false alarm rate at a detection threshold {\tau} = 10; this is a marginal, although not statistically significant, improvement of 0.2% over a reference sequential detrending and detection algorithm. In addition, a full search of the input TESS data was performed to evaluate the recovery rate of known TESS objects of interest (TOI) and to perform an independent search for new exoplanet candidates. The joint detector has a 73% recall rate and a 63% detection rate for known TOI; the former considers a match against all detection statistics above threshold while the latter considers only the maximum detection statistic.