Recently, an international research team led by Professor Ge Jian from the Shanghai Astronomical Observatory of the Chinese Academy of Sciences announced a significant discovery. Utilizing a novel deep learning algorithm, they successfully identified five ultra-short period planets with diameters smaller than Earth and orbital periods of less than one day in the photometric data released by the Kepler Space Telescope in 2017. This achievement not only represents a major breakthrough in the use of artificial intelligence for detecting and identifying celestial signals but also provides new crucial insights for the field of planetary science.
After five years of persistent efforts, the research team successfully developed a new algorithm named GPFC. This algorithm cleverly integrates GPU phase folding techniques with convolutional neural networks, significantly enhancing the speed, accuracy, and completeness of star transit signal searches. Compared to the traditional BLS method, the GPFC algorithm improves search speed by approximately 15 times and increases detection accuracy and completeness by about 7% each.
Using this advanced algorithm, the team discovered five new ultra-short period planets in the Kepler dataset: Kepler-158d, Kepler-963c, Kepler-879c, Kepler-1489c, and Kepler-2003b. Among these, Kepler-879c, Kepler-158d, Kepler-1489c, and Kepler-963c rank first, second, third, and fifth respectively among the smallest ultra-short period planets discovered to date. Additionally, Kepler-879c, Kepler-158d, Kepler-1489c, and Kepler-2003b are the smallest planets closest to their host stars, with orbital radii within five stellar radii.
These newly discovered ultra-short period planets are significant for studying the early evolution of planetary systems, interactions between planets, and the dynamical relationships between stars and planets. Their faint transit signals particularly demonstrate the GPFC algorithm's robust capability in detecting subtle signals within high-precision photometric data.
The related research findings have been published in the Monthly Notices of the Royal Astronomical Society (MNRAS). This accomplishment not only provides the astronomical community with a new research method—leveraging artificial intelligence to rapidly and efficiently search for transit signals in vast astronomical datasets—but also highlights the immense potential and prospects of AI in advancing astronomical research.
Professor Ge Jian stated that this achievement was made possible through the collective effort and continuous innovation of the team members. In the future, they will continue to deepen the application of artificial intelligence technologies in the field of astronomy, contributing further wisdom and strength to humanity's exploration of the universe's mysteries.
