New software squeezes more out of Webb Telescope data for exoplanet imaging
Researchers have developed FLIP, a method that uses machine-learning techniques to improve data from the James Webb Space Telescope. It let them spot one exoplanet in Webb data for the first time.
A team of researchers has developed a software method called FLIP that improves the quality of images from the James Webb Space Telescope (Webb), making it easier to see exoplanets, which are planets outside our Solar System. The results were published in the journal Nature in 2026. The method reduces noise in the images, so it may reveal fainter objects that previously went undetected.
How FLIP works
FLIP stands for Physics-based Learning for Imaging Planets, though it is not a true acronym. It relies on a physical model of NIRCam, Webb's near-infrared camera, which has 106 possible parameters. The software tunes these parameters using gradient descent, a mathematical technique widely used to train large neural networks. The process begins with a simulated point spread function and then optimizes it against data from a real star. On a current graphics processing unit (GPU), this takes a little over half an hour.
The results are then fine-tuned on the specific scientific image being analyzed. This step is critical because some variables differ from one observation to another.
Test results
The team tested the method on two stars with known exoplanets, HIP 65426 and 51 Eridani, and a third, TYC 5899, which has no known exoplanets but whose observations are known to vary considerably. These are young stars, and two of them host massive planets that are still warm from their gravitational collapse, which allowed them to be imaged before.
With FLIP, the signal-to-noise ratio improved by a factor of 1.5 to 2, depending on the observation. At a wavelength that can be used to detect carbon dioxide, one of the planets was seen in Webb data for the first time. For TYC 5899, sensitivity increased by a factor of 15.
Possible uses
The researchers note that the applications go beyond a better view of known planets. Lower noise could allow detection of dimmer, and therefore older or lower-mass, planets. They also think FLIP could be modified to show where it makes the biggest changes to the model, which might reveal useful information about the telescope's hardware.
