Northwestern-led team finds giant exoplanet in two-year-old JWST data with FLIP

A Northwestern-led team revealed a giant exoplanet in two-year-old JWST archival data on Nov. 4, 2026, using a new computational technique called FLIP, with the study appearing in Nature that same day and Rodrigo Ferrer-Chavez leading the research.

The central obstacle is contrast. Planets are a million times fainter than their stars, according to Rodrigo Ferrer-Chavez, and even bright young giant planets can be 100,000 times fainter than their host stars. JWST uses coronagraphs to block starlight, but some starlight seeps through and creates a hazy pattern that can obscure planets, so astronomers use image processing to remove remaining glare. Coronagraphs act like artificial eclipses inside the telescope but cannot eliminate all starlight, leaving residual patterns that require further processing.

Carbon dioxide in 51 Eridani b’s atmosphere absorbs light at the infrared wavelength JWST observes, at 4.3 microns, making the planet’s signal even fainter. Rodrigo Ferrer-Chavez described a coronagraph as something like an artificial eclipse, explaining the mask inside JWST that blocks most of the star’s light. Yahoo Tech’s report on the Nature study says even the largest planets can be more than 10,000 times fainter than their stars, with their light disappearing into the glare.

Jason Wang said that for two decades, astronomers have observed a different reference star and subtracted its image from the target observation to remove glare. He also said small thermal changes could cause a JWST mirror to expand or contract by a picometer, a trillionth of a meter, and cryocooler motors can vibrate the mirrors, with other causes of instability still unknown. No telescope stays perfectly still, so nearly imperceptible structural changes alter scattered starlight patterns and leave a reference image that never perfectly matches the observation, hiding faint planets behind residual glare. JWST’s 6.5-meter mirror flexes, stars can sit slightly off-center, and temperatures shift between observations, according to Yahoo Tech.

FLIP builds a digital twin of the telescope, a virtual replica simulating how starlight travels through JWST’s mirrors to its detectors. The method fine-tunes millions of parameters for imperfections, including position shifts and optic distortions. FLIP repeatedly adjusts the digital twin until it matches the actual telescope observation, then subtracts the simulated glare from the image. The team did not use AI to identify planets; they combined known telescope physics with computational tools like AI training and optimization, Jason Wang said.

The team ran FLIP on archived JWST views of three stars: HIP 65426 and 51 Eridani, each hosting a known planet, and TYC 5899, which has no known companions. In those same observations, FLIP beat conventional image processing, lifting faint-planet sensitivity by two to 15 times. The sensitivity improvement applied across all three observed systems rather than only to the successful planet recovery.

The technique revealed 51 Eridani b, a known giant planet, hidden in two-year-old JWST data where conventional processing could not distinguish it from residual glare. Astronomers had discovered 51 Eridani b with ground-based telescopes. In a JWST/NIRCam observation of 51 Eridani in the F430M filter, the standard KLIP method could not detect the planet, while FLIP produced a detection exceeding five sigma at the expected location. The comparison used the same JWST/NIRCam observation and filter, showing the new method succeeded where the older KLIP algorithm failed.

Jason Wang emphasized that the demonstration used real data in a real scenario and called it science left on the table because researchers lacked optimal analysis tools. The researchers stressed that the discovery came from existing data rather than a new observation campaign. Because FLIP reprocesses existing images rather than requiring fresh observations, archival JWST data can yield newly detectable planets without telescope time.

The method may let astronomers find smaller, colder, and fainter planets without altering telescopes or adding hardware. FLIP gives JWST a better chance of spotting planets closer to their stars, and the sensitivity gains are biggest close to stars, where planets like Earth tend to orbit. FLIP may also let researchers examine known planets at other wavelengths, uncovering new details about their atmospheres and composition. Model-based glare removal had been discussed for more than a decade but could not beat reference-star methods until now, Dimitri Mawet said.

NASA’s proposed Habitable Worlds Observatory aims to directly photograph Earth-like planets orbiting distant sun-like stars and is planned to launch in the 2040s. The Habitable Worlds Observatory is a future NASA mission concept intended to image Earth analogs around sun-like stars. Finding Earth-like planets may require HWO to maintain optical stability at the picometer scale, equivalent to trillionths of a meter and roughly 100 to 1,000 times more stable than JWST. HWO’s optics must remain steady within a trillionth of a meter to photograph Earth-like planets roughly 10 billion times fainter than their stars.

FLIP offers a path to ease HWO’s stability error budget by correcting tiny imperfections computationally rather than relying solely on engineering stability, according to the researchers and Dimitri Mawet, meaning the hardware can be allowed to wobble a bit more. Software-based correction could reduce the precision hardware requirements for the next-generation observatory. Dimitri Mawet, a senior research scientist at NASA’s Jet Propulsion Laboratory not involved in the study, called the result significant. Rodrigo Ferrer-Chavez said if the FLIP approach works, building NASA’s Habitable Worlds Observatory may be really hard but not really, really, really hard.

Currently JWST can detect worlds roughly the size of Saturn orbiting farther out than Pluto orbits the sun. Rodrigo Ferrer-Chavez said Earth-sized worlds would require a telescope between 100 and 1,000 times more stable than JWST, not yet achieved even in lab conditions. He added that shifts or tilts of just a few nanometers alter the light pattern, and modeling those changes permits more precise glare removal. After tuning their simulation to match real observations, researchers used FLIP to remove glare and see planets smaller than the telescope’s hardware limits normally allow, according to Yahoo Tech.

Rodrigo Ferrer-Chavez is a Ph.D. student in astronomy at Northwestern’s Weinberg College and a member of Jason Wang’s research group and CIERA. Jason Wang is a Weinberg assistant professor of physics and astronomy, study co-author, and CIERA member.

The study, titled Imaging exoplanets below the telescope stability limit, was published on Nov. 4, 2026, in the journal Nature, with DOI 10.1038/s41586-026-11213-z. The research was supported by the Space Telescope Science Institute under award JWST-AR-07942, the Sloan Foundation under award FG-2025-24976, and the National Science Foundation under award PHY-2525568. The Northwestern news release was embargoed until 11 A.M. EST on Wednesday, Nov. 4, 2026.

Every time techniques improve, new types of planets are discovered, helping scientists understand how planetary systems form and evolve and how typical our solar system is, Rodrigo Ferrer-Chavez said. The original KLIP analysis of a JWST/NIRCam F430M observation of 51 Eridani was not able to detect the planet 51 Eri b, and its expected location was marked with a dashed circle, according to the Northwestern news release figure caption. The news release describes a three-panel visualization of the 51 Eridani observation: the left panel shows the KLIP result, the middle panel shows the FLIP detection exceeding five sigma, and the right panel shows a simulated image of the planet with expected morphology.