AI System RAVEN Discovers Over 100 Hidden Exoplanets Including Rare 'Neptunian Desert' Worlds
University of Warwick researchers confirm 31 new planets and identify thousands more candidates using machine learning analysis of NASA TESS data.
Astronomers at the University of Warwick have achieved a major breakthrough in exoplanet discovery, using an advanced artificial intelligence system called RAVEN to confirm more than 100 exoplanets from NASA's Transiting Exoplanet Survey Satellite data. The sophisticated analysis, published in MNRAS, includes 31 completely new worlds and has identified over 2,000 additional high-quality planet candidates, nearly 1,000 of which were previously unknown.
The AI system analyzed observations from more than 2.2 million stars collected during TESS's first four years of operation, focusing specifically on planets that orbit very close to their host stars with periods shorter than 16 days. This targeted approach has produced one of the most precise measurements yet of how common these short-period planets are throughout our galaxy. The research represents a significant advancement in automated planet detection capabilities.
Particularly exciting among the discoveries are several rare and extreme planet types that challenge current astronomical theories. Some newly confirmed worlds are ultra-short-period planets that complete full orbits in less than 24 hours, while others belong to the mysterious "Neptunian desert" – a region where current models predict few planets should exist. The study also revealed tightly packed multi-planet systems, including previously unknown pairs of planets orbiting the same star.
Dr. Marina Lafarga Magro, the study's first author, emphasized the significance of the comprehensive dataset: "Using our newly developed RAVEN pipeline, we were able to validate 118 new planets, and over 2,000 high-quality planet candidates, nearly 1,000 of them entirely new. This represents one of the best characterized samples of close-in planets and will help us identify the most promising systems for future study."
The RAVEN system addresses a critical challenge in modern astronomy: distinguishing genuine planet signals from false positives created by eclipsing binary stars and other astrophysical phenomena. The machine learning models were trained on hundreds of thousands of realistically simulated planets and other events that can masquerade as planets. Unlike existing tools that focus on specific parts of the detection workflow, RAVEN handles the entire process from signal detection through statistical validation, giving it a significant advantage in processing enormous datasets consistently and objectively.
