What was done

A research group at Northwestern University has developed an image-processing method for direct observation of exoplanets, called Physics-based Learning for Imaging Planets (FLIP). The study was published on November 4 in Nature; the first author is doctoral student Rodrigo Ferrer-Chavez, and among the co-authors is assistant professor of physics and astronomy Jason J. Wang. According to the report, the technique increases the sensitivity of the James Webb Space Telescope (JWST) to fainter planets by about 2–15 times.

The most obvious result is the recovery of the gas giant 51 Eridani b, already discovered by ground-based telescopes but not identifiable in Webb's infrared observations with traditional methods, because of the glare of the host star.

Why it matters

Direct imaging of exoplanets has to contend with extreme contrast: a young gas giant can be about 100,000 times less luminous than its star. Coronagraphs block starlight, but some of it leaks through, forming a residual halo; for about twenty years, this has been corrected by subtracting a reference image of another star, an expedient that leaves residuals when the mirrors and instruments shift or change temperature.

FLIP takes a different approach: instead of relying on a reference star, it builds a digital twin of the telescope and simulates the path of light through mirrors, optics and detector, reconstructing the halo to be subtracted. The model's parameters—millions of them—are tuned using automatic differentiation and gradient-based optimization, including optical distortions and small instrumental shifts.

Tests and future prospects

The method was tested on Webb data for three stars: HIP 65426 and 51 Eridani, with known planets, and TYC 5899, with no known companions. In all cases it proved better than traditional methods. For 51 Eridani b, in the 4.3-micrometer band relevant to CO2, the new procedure achieved a signal-to-noise ratio of about 6.3 in the F430M filter, where the classical approach detected nothing.

The authors emphasize that FLIP is not an artificial intelligence system for recognizing planets: it uses optimization tools similar to those of machine learning, but is based on the telescope's known physical laws. Its limitations remain to be clarified: computational cost needs to be optimized, and validation must be extended to other telescopes and data. The method could be relevant to future missions such as NASA's Habitable Worlds Observatory, which aims to directly photograph Earth-like planets around Sun-like stars, an objective that would require optical stability 100–1,000 times greater than that of Webb.

Source and attribution: this original summary was prepared by AICronaca based on the report by IT之家 ↗The statements are attributed to the source; possible informational limitations: The news is based on a single IT之家 article that summarizes the Nature study; neither the original text of the paper nor independent comments are available. The sensitivity figures and signal-to-noise ratio come from the authors and have not been independently verified.