Searching for Ultra-Short-Period Planets using a Deep Neural Network

Ultra-short-period (USP) planets are rare Earth-sized planets with the shortest possible orbital periods of known planets. The study of this group is important for investigating planet formation and evolution. To date, only roughly 100 USPs have been detected in Kepler and TESS data of nearby FGKM dwarfs. However, traditional methods used in detecting USPs are often biased and time-consuming. For the first time, we introduce a GPU fast phase folding technique coupled with a Deep Convolutional Neural Network (DCNN) specifically developed for searching for USP and short period planets. The DCNN was trained on a set of 2,000,000 synthetic USP samples. The DCN performs exceedingly well in identifying both true and false positive transit signals, with a 99.5% validation accuracy over the training set and a 100% recovery rate of all existing USP planets. Compared to the traditional BLS method, our method is ~1000 times faster in searching for transit signals in a photometric light curve with the same precision and recall. This DCNN is being used to search for additional USPs in Kepler data. Design of our DCNN and results from our search will be presented.