# JWST’s FLIP Method Reveals 51 Eridani b in Existing Data

> MKNIRALO — Science — Astronomy · Exoplanets

- Published: 2026-11-07
- Estimated reading time: 6 minutes
- Byline: Sora — Space Enthusiast
- Byline disclosure: Sora is a house editorial byline maintained by MKNIRALO, not a claim of an independently verified individual. It does not imply personal product testing or professional credentials.
- Author profile: https://mkniralo.com/authors/sora/
- Tags: Astronomy, Exoplanets, James Webb Space Telescope, JWST, Astrophysics, Space Science
- Cover image: https://mkniralo.com/images/jwst-flip-technique-reveals-51-eri-b-2026.svg

A physics-based image-processing method has recovered a known giant planet from JWST observations that standard starlight subtraction left unresolved.

A star can overwhelm the faint world orbiting beside it, even when a telescope is powerful enough to collect the planet’s light. The challenge is to separate the two signals without also erasing the object astronomers want to study. A new analysis of archived James Webb Space Telescope observations shows how a more accurate model of the telescope itself can change what survives that subtraction.

## A planet recovered, not newly discovered

The result appeared in Nature on November 4 in a paper led by Northwestern University astronomer Rodrigo Ferrer-Chavez. The team introduced FLIP, short for Physics-based Learning for Imaging Planets, and used it to recover 51 Eridani b in JWST data that conventional processing had left unresolved. The finding is a new detection in this particular set of Webb images, not the discovery of a previously unknown planet. The planet itself has been known for more than a decade.

51 Eridani b was first identified with the Gemini Planet Imager, using ground-based observations made in 2014; the discovery paper appeared in Science in 2015. NASA describes the system as about 96 light-years from Earth. The new work revisits JWST/NIRCam observations of the same system that were already roughly two years old, showing that valuable information can remain in an archive after an initial analysis has failed to reveal it.

Sources:
- [Imaging exoplanets below the telescope stability limit](https://www.nature.com/articles/s41586-026-11213-z) — Nature
- [Hidden planet emerges from the glare](https://news.northwestern.edu/stories/2026/11/hidden-planet-emerges-from-the-glare) — Northwestern University
- [Giant planet emerges from two-year-old JWST data thanks to glare-removal technique](https://phys.org/news/2026-11-giant-planet-emerges-year-jwst.html) — Northwestern University (republished by Phys.org)
- [Discovery and spectroscopy of the young jovian planet 51 Eri b with the Gemini Planet Imager](https://pubmed.ncbi.nlm.nih.gov/26272904/) — Science (PubMed record)
- [Exoplanet 51 Eridani b (Gemini)](https://science.nasa.gov/asset/webb/exoplanet-51-eridani-b-gemini/) — NASA Science

## Why a nearby star hides its planet

Direct imaging is difficult because a star can outshine a planet by an enormous margin. Northwestern’s account puts the contrast at roughly a million to one for the problem the team is addressing. JWST uses a coronagraph to block most of the central star’s light, much as an eclipse hides the Sun, but the mask cannot remove every photon. Residual starlight spreads across the detector as a structured pattern, or point-spread function, that can bury a much fainter companion.

Conventional processing often builds a reference image from another star and subtracts it from the target observation. The method used for comparison in the Nature paper is Karhunen–Loève Image Processing, or KLIP. Its weakness is that the telescope and instrument do not remain perfectly identical between exposures: tiny changes in pointing, optical path and detector response alter the glare pattern. A reference that is almost right can still leave residuals bright enough to conceal a planet close to its host star.

Sources:
- [Imaging exoplanets below the telescope stability limit](https://www.nature.com/articles/s41586-026-11213-z) — Nature
- [Hidden planet emerges from the glare](https://news.northwestern.edu/stories/2026/11/hidden-planet-emerges-from-the-glare) — Northwestern University

## FLIP builds a digital twin of the telescope

FLIP takes a model-based route. It creates a virtual representation of how light passes through JWST’s optics and coronagraph and reaches the NIRCam detector. The model combines an astrophysical scene with optical and detector descriptions, then uses wavefront-sensing information and the instrument’s nominal calibration as a starting point. Rather than assuming a reference-star image is a perfect match, the method calculates what the star’s light should look like under the conditions of the science observation.

The researchers then tune the model to match the observed stellar pattern. The paper describes automatic differentiation, gradient descent and a multi-stage optimization process, implemented in PyTorch; the optimization adjusts around a million free parameters. It is important not to mistake the word “Learning” in FLIP’s name for an AI system that spots planets. Northwestern co-author Jason Wang said the team used computational tools also found in AI training, but did not use AI to identify the planet: the method is driven by a physical model of the instrument. Artificial planets injected into the images helped the team track recovered signal and stop before optimization began fitting away real companions.

Sources:
- [Imaging exoplanets below the telescope stability limit](https://www.nature.com/articles/s41586-026-11213-z) — Nature
- [Hidden planet emerges from the glare](https://news.northwestern.edu/stories/2026/11/hidden-planet-emerges-from-the-glare) — Northwestern University

## What the tests actually show

The team tested FLIP on published JWST/NIRCam observations of two systems with known planets, HIP 65426 and 51 Eridani, and on TYC 5899, a survey target with no known companion. It compared the results with KLIP reductions of the same data. Across these different observations, the authors report sensitivity improvements ranging from about two to 15 times. The largest gains depend on the target and separation; they are not a blanket multiplication of JWST’s hardware sensitivity for every observation. The TYC 5899 example shows the range is target- and separation-dependent: sensitivity at 0.5 arcseconds improved about 15-fold.

For 51 Eridani b, the key result is in the F430M filter, centred near 4.3 microns. The standard reduction did not detect the planet there, while FLIP produced a signal-to-noise ratio of 6.3. The paper estimates that sensitivity at the planet’s location improved by about a factor of four. In the F410M data, the team measured roughly 2.1 times the signal-to-noise ratio achieved with KLIP. These are results for this planet and these datasets, not a claim that every faint object is now detectable.

Sources:
- [Imaging exoplanets below the telescope stability limit](https://www.nature.com/articles/s41586-026-11213-z) — Nature
- [Hidden planet emerges from the glare](https://news.northwestern.edu/stories/2026/11/hidden-planet-emerges-from-the-glare) — Northwestern University
- [Giant planet emerges from two-year-old JWST data thanks to glare-removal technique](https://phys.org/news/2026-11-giant-planet-emerges-year-jwst.html) — Northwestern University (republished by Phys.org)

## Why the 4.3-micron image matters

The wavelength is scientifically useful because carbon dioxide absorbs strongly around 4.3 microns. That absorption makes the planet’s signal harder to see, but it also creates an opportunity: observations in this band can help researchers investigate the atmosphere rather than simply record a bright point of light. The Nature paper calls the F430M result the first image of 51 Eridani b at 4.3 microns, a region that is difficult to access from the ground because Earth’s atmosphere absorbs strongly there.

The new image is not, by itself, a complete atmospheric analysis. The authors say future work can use the recovered detection to measure the planet’s 4.3-micron photometry and carry out further atmospheric characterization, including tests of earlier evidence about its metallicity. That work remains ahead. For now, FLIP has supplied a signal that was not measurable in the earlier standard reduction; it has not established a new atmospheric composition, discovered life-related chemistry or revised the planet’s physical properties.

Sources:
- [Imaging exoplanets below the telescope stability limit](https://www.nature.com/articles/s41586-026-11213-z) — Nature
- [Hidden planet emerges from the glare](https://news.northwestern.edu/stories/2026/11/hidden-planet-emerges-from-the-glare) — Northwestern University

## A more capable analysis, not a new telescope

The wider implication is that existing observatory archives may contain planets that standard processing has missed. Reanalysing old observations could therefore expand the sample available for follow-up and help astronomers select better targets for new measurements. But the experiment remains a demonstration on a small set of observations. In the TYC 5899 test, the team used injected artificial planets to measure sensitivity; it did not announce the discovery of a real planet around that star. The distinction between a recovered known planet and a simulated signal is central to judging the result.

FLIP could also inform the design and analysis of future observatories, including NASA’s proposed Habitable Worlds Observatory, which is intended to image Earth-like planets around Sun-like stars. The paper presents that as a possibility, not a solved engineering problem. Its authors note that none of the tested observations reached the fundamental photon-noise limit and identify further work on detector models, parameter uncertainties, computation and larger datasets. The advance does not change JWST, guarantee Earth-like discoveries or replace careful telescope engineering. It shows that better physical modelling can make existing observations more revealing. The proposed observatory may require picometre-scale optical stability, but whether FLIP can reduce that demand remains untested.

Sources:
- [Imaging exoplanets below the telescope stability limit](https://www.nature.com/articles/s41586-026-11213-z) — Nature
- [Hidden planet emerges from the glare](https://news.northwestern.edu/stories/2026/11/hidden-planet-emerges-from-the-glare) — Northwestern University

## About this article

This article reflects the public sources available on the publication and update dates shown above. For developing stories, follow the linked source pages for subsequent statements or decisions.

## Sources and further reading

- [Imaging exoplanets below the telescope stability limit](https://www.nature.com/articles/s41586-026-11213-z) — Nature (published 2026-11-04) (accessed 2026-11-07)
- [Hidden planet emerges from the glare](https://news.northwestern.edu/stories/2026/11/hidden-planet-emerges-from-the-glare) — Northwestern University (published 2026-11-04) (accessed 2026-11-07)
- [Giant planet emerges from two-year-old JWST data thanks to glare-removal technique](https://phys.org/news/2026-11-giant-planet-emerges-year-jwst.html) — Northwestern University (republished by Phys.org) (published 2026-11-04) (accessed 2026-11-07)
- [Discovery and spectroscopy of the young jovian planet 51 Eri b with the Gemini Planet Imager](https://pubmed.ncbi.nlm.nih.gov/26272904/) — Science (PubMed record) (published 2015-10-02) (accessed 2026-11-07)
- [Exoplanet 51 Eridani b (Gemini)](https://science.nasa.gov/asset/webb/exoplanet-51-eridani-b-gemini/) — NASA Science (published 2021-04-21) (accessed 2026-11-07)