Plate Nº 23 · recorded November 14, 2026
Space & AstronomyReported finding
Software Found a Hidden Planet in Two-Year-Old Webb Data
A digital twin of Webb's camera erased stellar glare and revealed planet 51 Eridani b hiding in two-year-old data — the first sighting at a key wavelength.
By Elena Vasquez4 min read791 words
In brief
- FLIP, a physics-based machine learning method, revealed planet 51 Eridani b in two-year-old Webb data, first spotted from the ground in 2015.
- The technique proved two to 15 times more sensitive to faint planets than standard glare-removal methods, roughly fourfold at the planet's position.
- The findings were published in the journal Nature; the team plans to adapt the method for the Nancy Grace Roman Space Telescope.
- NASA's proposed Habitable Worlds Observatory targets Earth-like planets some 10 billion times fainter than their stars.
A giant planet that sat unnoticed in James Webb Space Telescope data for two years popped into view after researchers built a virtual replica of the observatory's main camera and used it to erase stellar glare. The findings appear in the journal Nature, and the method behind them could change how astronomers hunt for planets around other stars.
The team calls the technique FLIP, short for Physics-based Learning for Imaging Planets. It works on images already collected, which means astronomers can now go back and mine Webb's archive for planets that were photographed but never noticed.
Why is photographing a planet so hard?
Most known exoplanets — planets orbiting stars other than the Sun — were detected indirectly, through their effects on their host stars. Imaging one directly is brutal, because the star overwhelms everything near it.
"Planets are a million times fainter than their stars," said Rodrigo Ferrer-Chavez of Northwestern University, who led the study. "So, to image exoplanets, we need to remove that glare."
That glare is fickle. Even the tiniest shifts in the telescope — tilts on the scale of a few nanometers, billionths of a meter — change the pattern of light the instruments record. Standard methods lean on reference photos to subtract the starlight. Small errors pile up.
FLIP takes a different route. Instead of comparing images, the researchers simulated how starlight physically travels through Webb: bouncing off its mirrors and landing on its detectors. This simulation is the "digital twin" — a virtual copy of the telescope's Near Infrared Camera, or NIRCam, Webb's main infrared imaging instrument.
Then the computer hunts. It searches for the exact combination of tiny flaws and fluctuations that would reproduce what the real telescope recorded. Nail the model, and you know precisely what the star's glare should look like. You can delete it cleanly.
"What's exciting is that we're showing how computation can help overcome limitations of the telescope itself," said co-author Katie Bouman of Caltech. "By combining a physical model of the instrument with optimization techniques from machine learning, we have been able to remove starlight much more precisely."
What did the digital twin actually find?
The team tested FLIP on existing Webb observations of two stars with known planets: HIP 65426 and 51 Eridani. Across those tests, the method proved two to 15 times more sensitive to faint planets than standard techniques.
The showcase result involves 51 Eridani b, a young Jupiter-like world first spotted in 2015 from ground-based telescopes. Webb had observed the system at a wavelength of 4.3 microns — an infrared band astronomers use to detect carbon dioxide in planetary atmospheres. Processed the conventional way, the planet simply wasn't there.
Run through FLIP, the same data revealed 51 Eridani b exactly where, and how, it should appear. It's the first time anyone has seen the planet at that wavelength. At the planet's position, FLIP improved sensitivity roughly fourfold.
It's not a new planet. That's the point: astronomers knew the answer, and FLIP found it where older methods failed.
"It was science left on the table because we didn't have the right tools," said Jason Wang, an assistant professor of physics and astronomy at Northwestern and a co-author on the study.
Could the model fake a planet?
There's a built-in risk. A model flexible enough to reproduce any glare pattern could also explain away a real planet as starlight. The team guarded against this by planting fake planet signals in the data and confirming FLIP recovered them. It did.
The project grew from failure. Ferrer-Chavez first tackled the problem as an undergraduate at Mexico's Autonomous University of Yucatán, in a thesis co-mentored by Wang and Bouman. None of his early approaches worked. Wang credits those dead ends with paving the way for the final method.
What comes next?
The plans are ambitious. Because the technique doesn't depend on Webb's specific design, the team intends to adapt it for NASA's upcoming Nancy Grace Roman Space Telescope and for ground-based observatories.
The biggest prize lies further out. NASA's proposed Habitable Worlds Observatory aims to directly image Earth-like planets around Sun-like stars — worlds roughly 10 billion times fainter than their hosts. If software can clean up optical instability after the fact, the engineers designing that telescope get breathing room on hardware precision.
There's a deeper motive too. Astronomers currently know of no planets around other stars that resemble the ones in our own solar system.
"It's not that they don't exist. It's that we cannot see them," Ferrer-Chavez said. FLIP, and the tools it inspires, may change that.
via zmescience.com (Source)
More from Elena Vasquez
Show full bio
Correspondent covering money, funding and market shifts around general science news at SciBeat.
422 articles
Nearby plates
- Webb Data Reveals Hidden Giant Planet After Software Clears Starlight
- Independent Method Confirms the First Black Hole Image of M87*
- Rubin Observatory Finds New Dwarf Planet Candidate Beyond Pluto
- ALMA captures first direct image of gas swirling around a forming planet
- James Webb Catches a Lopsided Bulge in an Exoplanet's Air