Webb data yields a known planet through better optical modelling
· BehindLines newsroom
A planet that astronomers already know about can still be lost in an image they have recorded. A Northwestern-led team proposes a better way to use such observations with FLIP, a physics-based imaging method presented in Nature on 4 November. Accounts from Northwestern and Caltech describe how a detailed model of starlight helps reveal the weak signal of a neighbouring planet.
Recreating the light that gets in the way
A coronagraph blocks some of a star's glare. Residual light can nevertheless cover a planet's signal. Tiny changes within the telescope alter that pattern, so a straightforward comparison with an observation of another star will not always remove it effectively.
FLIP creates a digital representation of the instrument. Researchers tune its parameters to reproduce the passage of light and the optical imperfections present during an observation. They then remove the calculated stellar contribution from the recorded image. The computational tools resemble those used in training machine-learning models, while the calculation itself follows the instrument's physics.
Recovering a recorded signal
The researchers revisited observations made by the James Webb Space Telescope. Among their results was the recovery of 51 Eridani b, a giant planet previously identified using ground-based instruments. Conventional processing had failed to separate its signal in these Webb data. The advance concerns that recovery, rather than the original discovery of the planet.
The institutions report sensitivity gains ranging from a factor of 2 to 15 across the observations they compared. The upper end is not a benefit promised for every image. These figures describe the reported tests and do not guarantee that the method will reveal any planet researchers look for.
The immediate opportunity is to examine archived observations with improved tools. Extending the approach to other instruments and future observatories remains a research prospect. The result illustrates how computation can complement telescope engineering while working with light that an instrument has actually recorded.