2025AJ....170..303I

The unprecedented photometric precision of the Kepler mission allows searches for Earth-like planets. However, it remains difficult to distinguish these low signal-to-noise-ratio planets from the false alarms originating from correlated and non-Gaussian noise. It reduces the resulting planetary catalog reliability and makes it hard to measure the occurrence rate of small long-period planets. We aim to obtain a more reliable catalog of small long-period planet candidates from Kepler data and use it to improve their occurrence rate estimates. This work develops an independent search pipeline for small (Kepler multiple-event statistic (MES) ≲ 12) long-period (50-500 days) planets. We design and implement a detection statistic that takes into account noise non-Gaussianity and physical priors. For every threshold-crossing event, we run permutation and injection procedures to calculate the probability of it being caused by a real planet. The provided detection statistic has a tailless background distribution with a rate of ∼1 false alarm per search for MES ∼ 7.8. We demonstrate an increase in detection efficiency for MESs of 7.5-9 and >4 transits due to the background distribution control. The pipeline was tested and found to be able to detect most of the faint confirmed Kepler planets. The pipeline was applied to the entirety of Kepler data and detected ∼50 candidate events with a high probability of originating from real planets, which will be presented in our future work.