A Digital Twin Pulled 51 Eridani b From Webb’s Glare
A physics-based model of Webb’s NIRCam did more than clean a picture. It recovered 51 Eridani b at 4.3 microns—where earlier pipelines saw nothing.
The Discovery
Direct images of giant planets are usually fights against starlight. A Nature paper published 4 November 2026—also available as arXiv:2611.13257—reports a new way through that glare: FLIP, short for Physics-based Learning for Imaging Planets.
The team built a differentiable physical model of Webb’s NIRCam coronagraph, tuned millions of parameters with automatic differentiation, and subtracted a reconstructed stellar point-spread function from real on-sky frames. In tests against the standard KLIP pipeline on published Webb data, FLIP improved detection sensitivity by roughly 2 to 15 times, depending on the field and separation.
The headline result is concrete: FLIP produced the first image of 51 Eridani b at 4.3 micrometres, recovering the planet in the F430M filter at a reported SNR of 6.3 after an earlier non-detection with standard methods.
What Makes It Strange
Most starlight-subtraction recipes lean on libraries of reference images. Tiny pointing errors and thermal drifts make those references imperfect copies of the science frame, so leftover glare can bury a planet. FLIP instead pre-calibrates an instrument model on a reference, then solves for small aberration changes that better match the science observation.
- Observed: improved recoveries on HIP 65426 and 51 Eridani, plus injected-planet tests on TYC 5899.
- New wavelength: 4.3 μm sits in a carbon-dioxide absorption region that ground telescopes struggle to reach cleanly through Earth’s atmosphere.
- Not claimed: a newly discovered fifth planet, an Earth twin, or proof of life.
51 Eridani b was already a known giant planet. What changed is the reach of the archive. A wavelength that once returned a non-detection now yields a usable detection after the model peels the star away.
The Implications
The authors argue that physics-based, autodiff-enabled subtraction can push past limits set by telescope stability—an issue that only gets harder for future missions such as the Nancy Grace Roman Space Telescope and a proposed Habitable Worlds Observatory. That is a methodological claim, not a guarantee that Earth-like worlds are about to appear in old folders.
For 51 Eridani b specifically, the new 4.3-μm detection opens a path to photometry and atmospheric work that can test earlier hints of elevated metallicity. Those follow-ups are future research, not results already in hand.
The planet was already in the light. The star was still writing over it.
The accompanying image is a NASA photograph of Webb’s gold hexagonal primary-mirror segments in a clean facility. It is hardware context for the telescope that took the data, not a picture of 51 Eridani b or a FLIP residual frame.