A new method pulls a known planet out of Webb’s glare

The same Webb exposures that once hid 51 Eridani b now show it at 4.3 micrometers. The difference is how researchers modeled the light left over from its star.

The planet was already in the image. What astronomers couldn’t see was its light at one particular infrared wavelength. By rebuilding the glare around the young star 51 Eridani in software, researchers recovered the planet from the same James Webb Space Telescope exposures that had previously yielded no detection in that band.

The study, published November 4 in Nature, reports the first direct image of 51 Eridani b at 4.3 micrometers. The planet itself has been known since its 2015 discovery with the Gemini Planet Imager. Webb took the exposures on October 18, 2023, and the original analysis already detected the giant in other bands. What changed was how the team removed starlight from the 4.3-micrometer exposures.

Why blocking a star still leaves a glare

Webb’s Near-Infrared Camera has a coronagraph: a set of masks that suppresses light from a star while allowing light from slightly farther away to reach the detector. Even with the star centered behind a mask, some of its light spreads into rings and speckles. Close to the star, a speckle can be as convincing as the faint point an astronomer is looking for.

A standard approach builds a picture of this leftover glare from reference exposures. Subtract the estimate from the target image and much of the glare goes away. But slight shifts in the star’s position behind the mask, or changes in the path light takes through the telescope, change the pattern. A reference image can be close enough to remove most of the starlight and still leave a bright patch exactly where the planet ought to be.

How the team separated the planet from the star

Rodrigo Ferrer-Chavez and colleagues instead simulated how light from a point source passes through Webb’s optics and lands on the detector. Their method, Physics-based Learning for Imaging Planets (FLIP), starts with measurements of Webb’s optics and fits the model to observations of a reference star. It then adjusts selected parameters on the target exposures, such as the alignment of the star behind the mask. The result is a glare model tailored to those exposures, which the team subtracts from the image.

The danger is that a flexible model could absorb some of the planet’s light as it fits the star. To measure that risk, the researchers added artificial planets of known brightness to the data and watched how much of their signal survived. Those tests helped them decide when to stop adjusting the model and estimate what either method could reliably recover.

In the comparison, the new result shows a point at the known planet’s expected position with a shape close to what Webb should record from a faint point source. The team reports a signal-to-noise ratio of 6.3 for 51 Eridani b in that filter. The authors also note a much fainter spot whose origin remains unclear.

What one more band can tell us

Carbon dioxide absorbs infrared light near 4.3 micrometers. Measuring how bright the planet is in that band, alongside other wavelengths, can help astronomers compare models of its atmosphere. A single image in the band does not identify the gas on its own. The authors plan to use the recovered signal to improve estimates of the planet’s brightness there and revisit its atmospheric composition.

In their tests, the detection threshold improved by about fourfold at the position of 51 Eridani b. The largest gain, roughly fifteenfold, came at a particular distance from a different star in a test with inserted planet signals. How much the method helps depends on the observation and how close the planet is to its star. The practical result is that some exposures may still hold measurements that earlier processing couldn’t recover.