New software helps Webb spot faint planets by modeling stellar glare, revealing a planetary signal missed in earlier image processing.
Photographing faint planets around other stars requires exceptionally stable telescope optics. Even the James Webb Space Telescope’s optical components can shift by a few nanometers during observations.
A shift that small changes the glare around a star enough to swamp a planet’s light.
A team led from Northwestern University has now reduced the effects of those shifts with software instead of hardware.
Rodrigo Ferrer-Chavez, a doctoral student at Northwestern University and its astrophysics center, CIERA, led the work with his adviser, Jason Wang.
The team’s program builds a working model of Webb’s own optics and tunes it until the model’s starlight matches the real glare, wobbles included.
Run on Webb images taken two years ago, it picked out the giant planet 51 Eridani b at a wavelength where standard processing had found nothing.
Across four sets of observations, it made Webb two to 15 times more sensitive to faint planets.
“That is a very tricky problem because even the tiniest changes in the telescope can affect how that glare appears,” Ferrer-Chavez remarked.
“Slight shifts or tilts – even on the scale of just a few nanometers – can change the pattern of light.”.
Young giant planets shine at about one hundred-thousandth the brightness of their stars in infrared light.
Sources indicate that A mask inside the camera, called a coronagraph, blocks most of the starlight. A halo of leftover glare remains, and a planet has to be found inside it.
For about 20 years, astronomers have handled that halo by photographing a second, planet-free star and subtracting its glare from the first image.
The method works best when the telescope’s optics remain unchanged between observations.
The optics do not remain perfectly stable.
Pointing errors and heat drifts change the optics between one picture and the next, so the two halos don’t match.
That mismatch sets the limit on how faint a planet can be.
Recent findings suggest that wang said engineers try to make telescopes extremely stable and haven’t fully succeeded. “Small thermal changes could cause a mirror to expand or contract by a picometer,” he said.
“Telescopes like JWST have cryocoolers to keep the instrument super cold. Those coolers have little motors that can cause the mirrors to vibrate.”.
The team’s program, called FLIP, starts from the standard computer model of Webb’s optics maintained by the Space Telescope Science Institute.
It adds about a million adjustable settings for the mirrors, the position of the star behind the mask and the response of every pixel.
Then it nudges those settings until the simulated glare matches a real image.
The nudging uses the mathematical machinery that trains artificial intelligence, which can adjust millions of settings at once.
According to recent reports, “We used the same computational tools that researchers use to train AI models, but we did not actually use AI,” Wang said. “Because we already know the physics, we don’t need AI to understand what’s happening.”.
FLIP first tunes itself on a reference star, then fine-tunes on the image that might hold a planet.
To keep it from erasing a real planet along with the glare, the team plants fake planets in the image and stops the fitting when their signals stop improving.
A full run takes about 35 minutes on a single graphics card.
The team reprocessed published Webb images of three stars and compared FLIP with the standard method on each.
It has been revealed that around HIP 65426, the known planet’s signal was about 1.6 times stronger. Around 51 Eridani, the planet’s signal doubled at one wavelength.
At 4.3 micrometers, where the original analysis had found nothing, FLIP produced a detection.
The 4.3-micrometer band matters because carbon dioxide in the planet’s atmosphere absorbs light there, which is part of why the planet was so faint.
Earth’s own atmosphere blocks that light, so no telescope on the ground can make the measurement.
The third star, TYC 5899, has no known planets, and its reference images were taken four months before the science image, the kind of gap that defeats the standard method.
In a recent development, FLIP still recovered a fake planet planted in the data that the standard method missed. Close to the star, it was about 15 times more sensitive.
“We indicated that we can actually recover a planet that people previously couldn’t see,” Wang said. “This is real data in a real scenario.”.
In all four observations, residual glare still limited FLIP’s sensitivity, leaving room for further improvement.
The authors also wrote that their model can correct for the telescope’s small shifts but hasn’t yet been calibrated to measure them.
The detector model and the computing cost both need work, they wrote.
The team’s next step is to measure the planet’s brightness at 4.3 micrometers and test earlier hints that its atmosphere holds about five times the heavy elements of its star.
More than 100 Webb observations have been taken in surveys like the one that produced the TYC 5899 image. The team wrote that the discovery potential of reprocessing them is large.
The biggest test the authors named is NASA’s planned Habitable Worlds Observatory, designed to photograph Earth-like planets around Sun-like stars.
Current intends call for it to hold steady to about a picometer, 100 to 1,000 times steadier than Webb.
Whether software can ease that requirement hasn’t been tested, and the authors pointed to the Nancy Grace Roman Space Telescope as the place to find out.
The full study was published in the journal Nature.
Check us out on EarthSnap, a free app brought to you by Eric Ralls and Earth.com.
