A groundbreaking achievement at the University of Warwick has been unveiled, where astronomers have successfully confirmed more than 100 exoplanets using an innovative artificial intelligence system. This remarkable accomplishment highlights the potential of AI in astronomy, specifically in the analysis of vast datasets provided by NASA’s Transiting Exoplanet Survey Satellite (TESS). TESS’s mission is to identify exoplanets by detecting slight dips in starlight that occur when these celestial bodies transit in front of their host stars.

This research, recently published in the Monthly Notices of the Royal Astronomical Society (MNRAS), is based on an extensive analysis encompassing observations from over 2.2 million stars collected during TESS’s initial four-year operations. The astronomers focused their efforts on identifying planets that orbit very close to their stars, completing full orbits in less than 16 days. Their systematic approach has yielded one of the most accurate measurements to date regarding the frequency of these short-period planets.

Dr. Marina Lafarga Magro, a postdoctoral researcher involved in the project, remarked, “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 which are entirely new to science. This represents one of the best characterized samples of close-in planets and will aid us in pinpointing the most promising systems for further exploration.” The implications of this enhanced understanding of exoplanets are considerable, potentially informing future studies aimed at investigating the atmospheres and compositions of these distant worlds.

The study has also identified several fascinating and atypical categories of planets that challenge existing theories. Among these are ultra-short-period planets, which complete their orbits in under 24 hours, and planets located within the so-called ‘Neptunian desert,’ a striking region where theoretical models suggest few such celestial bodies should exist. Furthermore, researchers have discovered tightly packed multi-planet systems, some containing previously unidentified pairs of planets orbiting a single star.

The RAVEN pipeline represents a significant advancement in the realm of planet detection technologies. Traditional planet-hunting missions frequently flag thousands of potential planets, yet distinguishing genuine signals from false ones poses a considerable challenge. Many false signals can appear reminiscent of exoplanets, such as those produced by eclipsing binary stars.

Dr. Andreas Hadjigeorghiou, who spearheaded the development of the RAVEN pipeline, explained the innovative features of this AI tool: “The challenge lies in determining whether a detected dimming is indeed caused by a planet orbiting a star or by alternative sources, such as eclipsing binary stars. RAVEN tackles this challenge head-on. Its strength derives from our carefully curated dataset of hundreds of thousands of realistically simulated planets and various astrophysical phenomena that could masquerade as planets. By training machine learning models to identify patterns in the data, we enable precise determination of the type of event being detected, a task for which AI models are uniquely well-suited.”

Moreover, RAVEN’s capability to manage the complete detection process is a crucial advantage. The pipeline integrates steps from signal detection through to machine learning validation and statistical verification, making it a comprehensive solution unlike many contemporary tools, which only focus on isolated segments of the workflow.

Dr. David Armstrong, an associate professor and senior co-author on the RAVEN studies, emphasized the efficiency of their approach: “RAVEN allows us to analyze extensive datasets consistently and objectively. The well-tested nature of the pipeline enhances our confidence in the reliability of the results obtained.” This study opens new avenues for research in the exoplanetary field, providing a framework that can potentially lead to the discovery of even more distant worlds with unique characteristics in the vast expanse of our universe.