How to discover exoplanets obscured by starlight? Internationally developed new high-sensitivity detection method

The international academic journal Nature published an astronomy paper online in the early morning of November 5, Beijing time, stating that researchers have developed a new method for detecting exoplanets, using a machine learning algorithm to remove starlight from telescope observations. Its sensitivity can be up to 15 times that of traditional exoplanet imaging methods, and may assist observatories in identifying Earth-like exoplanets.

According to the paper, direct imaging is a key strategy for observing exoplanets, but the brightness of their host stars often masks their presence. Even the brightest young giant planets are only 100,000 times as bright as their host stars. One way to solve this problem is to subtract the host star's light from the exoplanet observations. However, previous attempts have been difficult to succeed due to the physical limitations of the observatory or the computational limitations of simulating starlight.

In this study, the first author and corresponding author of the paper, Rodrigo Ferrer-Chavez of Northwestern University, worked with colleagues and collaborators to propose an image processing framework called "Physics-Driven Learning for Planetary Imaging" (FLIP), which can generate simulated glare of specific stars. They rationalized subtracting this simulated light from direct observations of a star system, allowing all obscured exoplanets to be exposed.

Subsequently, the author of the paper tested the FLIP framework using the near-infrared camera observation data of the Webb Telescope, and successfully obtained the first image of 51 Eridani b, which could not be detected using traditional methods. In other tests, FLIP was also 2 to 15 times more sensitive than traditional solar glare removal technology.

The authors of the paper concluded that FLIP can be used in conjunction with observations from the Webb Telescope to discover giant planets with lower temperatures, smaller masses, and older ages. In the future, this method may be adapted for use in ground-based observatories, and may also be used to image Earth-like objects in planned missions such as the Habitable World Observatory. However, they also cautioned that further research should focus on how to reduce computational costs and test with larger amounts of data.