Even the most powerful telescopes have blind spots. The James Webb Space Telescope (JWST) can peer billions of years into the past, but photographing a planet around a neighboring star is hindered by the blinding light of the star itself - its glare drowns out the planet's faint glow.
A group of astrophysicists led by Northwestern University (USA) has developed a computational method that makes it possible to "subtract" this glare and see hidden worlds. The study was published in the journal Nature.
The new method, called Physics-based Learning for Imaging Planets (FLIP), creates a "digital twin" of the telescope - a virtual copy that models how starlight passes through the mirrors and reaches the detectors. The researchers then tune millions of parameters of this model so that it matches the real image as closely as possible, and subtract the simulated glare from the image.
"Planets are a million times fainter than their stars," explained the lead author of the work, Rodrigo Ferrer-Chavez of Northwestern University. "To photograph an exoplanet, you need to remove this glare. This is a very difficult task, because even the smallest changes in the telescope affect the pattern of light. Shifts or tilts of just a few nanometers change the pattern." According to him, the method takes these tiny changes into account and makes it possible to remove the glare much more accurately.
Traditionally, astronomers combat the glare differently: they image another star, use it as a reference and subtract it from the original image. But the telescope cannot remain perfectly still: thermal fluctuations and vibrations from the motors of cooling devices change the scattered-light pattern, so the reference never matches the real image perfectly.
To test FLIP, the scientists reprocessed archival JWST observations of three stars: HIP 65426 and 51 Eridani, which have known planets, as well as TYC 5899, where no planets have yet been found. In all cases, the new method performed better than standard algorithms, increasing sensitivity to faint planets by 2-15 times.
The main result: in data from two years ago, they managed to discern the giant planet 51 Eridani b. Its atmosphere contains carbon dioxide, which absorbs infrared light at the wavelength where JWST observes, making the planet's signal even weaker. Conventional processing could not distinguish the planet from residual glare, but FLIP could.
"We showed that we can actually recover a planet that was not seen before," noted co-author Jason Wang. "This is real data in a real situation. The science was sitting on the table because we didn't have the right tools to analyze the data in an optimal way."
The authors emphasize that for now this concerns an already known planet that was seen in a new quality. But the method could help find smaller, colder and fainter planets without changing the telescopes themselves. In the future, such algorithms will be useful for future missions like NASA's proposed Habitable Worlds Observatory, which is to photograph Earth-like planets around distant Sun-like stars.
"We don't know planets around other stars that are similar to the planets of our Solar System," says Ferrer-Chavez. "Not because they don't exist, but because we don't see them. Every time we improve our methods, we discover new types of planets."