A giant exoplanet that had vanished into the glare of its own star in images from NASA’s James Webb Space Telescope has been recovered by a new computational technique developed at Northwestern University. The method, called Physics-based Learning for Imaging Planets, or FLIP, improved the telescope’s sensitivity for detecting faint planets by two to fifteen times when applied to existing observations, and it revealed the planet 51 Eridani b at a wavelength where conventional image processing had failed to distinguish it from residual starlight. The study describing the technique is published on November 4 in the journal Nature.
The fundamental obstacle is one of contrast. Planets are a million times fainter than their stars, according to Rodrigo Ferrer-Chavez, a Ph.D. student in astronomy at Northwestern’s Weinberg College of Arts and Sciences who led the study. Even relatively bright young giant planets can be 100,000 times fainter than the stars they orbit. When a telescope points at a nearby star, that overwhelming brightness swamps the faint glow of any planet circling it, which is why photographing a planet outside our solar system has been compared to trying to spot a firefly hovering next to a blazing floodlight.
Telescopes such as JWST carry specialized instruments called coronagraphs that block out starlight so that faint objects can be detected. But blocking the star is not enough. Some starlight still seeps through, producing a hazy pattern that can obscure planets. Astronomers therefore turn to image-processing techniques to remove the remaining glare. For the past two decades, the standard approach has been to observe a different star and use its image as a reference, then subtract that reference image from the original observation to reveal faint planets that might otherwise be hidden, explained Jason Wang, an assistant professor of physics and astronomy at Weinberg and co-author of the study.
The trouble is that no telescope stays perfectly still. Nearly imperceptible changes in the telescope’s physical structure alter the patterns of scattered starlight, so the reference image never perfectly matches the actual observation. Even the slightest mismatch can leave behind just enough glare to hide a faint planet. Wang noted that there are known and unknown causes of this instability: small thermal changes could cause a mirror to expand or contract by a picometer, and the motors in the cryocoolers that keep JWST’s instruments super cold can cause the mirrors to vibrate. There are also issues that remain poorly understood as astronomers learn how to build these complex telescopes.
FLIP takes a different route. Instead of relying on reference photographs alone, the Northwestern team built a digital twin of the telescope, a virtual replica that simulates how starlight travels through the telescope’s mirrors and reaches its detectors. Using computational techniques similar to those used to train AI models, the researchers fine-tuned millions of parameters in the digital twin to account for tiny imperfections in the telescope, including subtle shifts in position and distortions in its optics. Crucially, the team did not use AI to identify planets. They combined their knowledge of the telescope’s physics with powerful optimization tools. Because the physics is already known, Wang said, there is no need for AI to understand what is happening.
By repeatedly adjusting the digital twin until it closely matched the telescope’s actual observation, FLIP recreated the unwanted starlight with remarkable precision. The researchers then subtracted that simulated glare from the image. The method models how light travels through the telescope and accounts for tiny changes in its structure, Ferrer-Chavez said. Slight shifts or tilts, even on the scale of just a few nanometers, can change the pattern of light, and modeling those changes allows the glare to be removed much more precisely than reference-based subtraction allows.
To test the approach, the team reanalyzed past JWST observations of three stars: HIP 65426 and 51 Eridani, both with known planets, and TYC 5899, which has no known planetary companions. Across the observations, FLIP consistently outperformed conventional image-processing techniques, improving sensitivity for detecting faint planets by two to fifteen times. The gain matters because it was achieved without changing the telescope itself. The advance could enable astronomers to discover smaller, colder and fainter planets without new hardware, and it could allow scientists to study known planets at different wavelengths to reveal new clues about their atmospheres.
The headline result came from 51 Eridani. Although astronomers had previously discovered the giant planet 51 Eridani b with ground-based telescopes, conventional image processing could not distinguish it from residual glare at the wavelength JWST uses to study carbon dioxide in planetary atmospheres. Carbon dioxide in the planet’s atmosphere absorbs light in that infrared band, making the planet’s signal even fainter. When Ferrer-Chavez applied FLIP to the same observations, the planet emerged. In a JWST/NIRCam observation of the star in the F430M filter, a standard glare-subtraction method called KLIP was unable to detect the planet, while FLIP produced a significant detection exceeding five sigma, at the planet’s expected location and with the expected signal morphology.
Wang emphasized that the demonstration used real data in a real scenario. It was science left on the table, he said, because researchers did not previously have the right tools to analyze the data in the most optimal way. The same property that makes a planet harder to detect can also make it more interesting to study. JWST can make sensitive observations at 4.3 microns, where carbon dioxide strongly absorbs infrared light, and detecting features at this wavelength potentially could give astronomers new clues into a planet’s atmosphere and composition.
Looking ahead, Ferrer-Chavez and Wang hope FLIP will help astronomers uncover planets previously hidden within existing observations. The technique could also shape the next generation of planet-hunting telescopes. NASA’s proposed Habitable Worlds Observatory aims to directly photograph Earth-like planets orbiting distant sun-like stars, and finding such small planets may require the telescope to maintain optical stability at the picometer scale, roughly 100 to 1,000 times more stable than JWST. FLIP offers a potentially different path forward: rather than relying on engineering solutions to keep a telescope perfectly stable, astronomers might be able to correct for its tiny imperfections. As Ferrer-Chavez put it, every time techniques improve, new types of planets are discovered, and finding more planets helps scientists understand how planetary systems form and evolve, and ultimately how typical our own solar system truly is.
Subject of Research: A physics-based glare-removal technique that improves exoplanet detection sensitivity in JWST images and reveals a previously hidden giant exoplanet.
Article Title: Hidden planet emerges from the glare
Article References: Hidden planet emerges from the glare. Original publication (Bibliographic details unavailable.)
Image Credits: AI Generated
DOI: Not provided
Keywords: exoplanets, JWST, starlight glare, image processing, digital twin, 51 Eridani b, coronagraphy, machine learning optimization, Northwestern University, Nature, carbon dioxide absorption, direct imaging
News Source: Grant Pearson. (November 4, 2026). New algorithm strips starlight from JWST images to reveal a hidden giant exoplanet. Scienmag.
