FLIP method raises JWST NIRCam coronagraph sensitivity 2–15 times; 51 Eridani b detected at 4.3 μm
Authors report a physics-based starlight subtraction method tested on real JWST NIRCam data. It reduces dependence on the stability of reference images.

Astronomers led by researchers at Northwestern University have developed FLIP, a physics-based method of starlight subtraction for direct exoplanet imaging, according to a paper on nature.com nature.com
nature.comnature.comOpen source ↗. The method was tested on observations from the NIRCam coronagraph of the James Webb Space Telescope (JWST). The authors report sensitivity 2–15 times higher than standard techniques, including the first detection of the planet 51 Eridani b at 4.3 μm
nature.com
nature.comnature.comOpen source ↗. To their knowledge, this is the first time such an approach has substantially improved sensitivity on real on-sky coronagraphic images
nature.com
nature.comnature.comOpen source ↗.
Why starlight subtraction is hard
Even the brightest young giant planets are about 10^5 times fainter than their host stars in the near-infrared nature.com
nature.comnature.comOpen source ↗. A coronagraph blocks part of the starlight, but the stellar point spread function (PSF) still has to be modelled and removed from the image
nature.com
nature.comnature.comOpen source ↗. Standard algorithms build this model from a library of reference observations taken at other times or wavelengths, or of other stars
nature.com
nature.comnature.comOpen source ↗.
Such empirical approaches are limited by instabilities. In space telescopes these are mostly pointing errors and thermal drifts, which make the reference PSF differ from the science PSF and reduce sensitivity nature.com
nature.comnature.comOpen source ↗. Model-based PSF subtraction had been considered before, but the complexity of optical systems made accurate modelling difficult. With JWST NIRCam it had never given sensitivity better than standard techniques
nature.com
nature.comnature.comOpen source ↗.
How FLIP works
FLIP first calibrates a model of the instrument on reference images so that it reproduces the reference PSF. It then solves for small changes in the telescope aberrations that better explain the science image nature.com
nature.comnature.comOpen source ↗. The method combines a multistep optimization with a flexible physical model
nature.com
nature.comnature.comOpen source ↗. Automatic differentiation is used to optimize millions of model parameters, and GPU-accelerated differentiable models make this computationally feasible
nature.com
nature.comnature.comOpen source ↗.
The authors state that the framework can be extended to other ground-based and space-based observatories, including the Nancy Grace Roman Space Telescope and the Habitable Worlds Observatory (HWO) nature.com
nature.comnature.comOpen source ↗. HWO aims to image Earth analogues around Sun-like stars, a brightness contrast of 10^10. This requires stability at the picometre level, 100 to 1,000 times better than JWST
nature.com
nature.comnature.comOpen source ↗. The authors note that the approach could potentially relax these stability requirements
nature.com
nature.comnature.comOpen source ↗.



