aifollow.news
Back IT之家 科技新闻
IT之家 科技新闻· · Original publication time AI score54

Northwestern University team identifies known giant planet in earlier Webb data with new methodMachine translation

西北大学团队用新方法从韦布旧数据中识别出已知巨行星

AI-assisted summary

The Northwestern University team developed FLIP to model and subtract a star's glare using the telescope's physical properties, revealing the giant planet 51 Eridani b in Webb infrared data where earlier processing had missed it. Ground-based telescopes had already discovered the planet. In tests using data from three stars, FLIP improved detection sensitivity by about 2 to 15 times over traditional processing, but its computational cost needs further optimization and it needs validation on more data.

Article · AI translation

IT Home reported on November 21 that a research team at Northwestern University in the United States has developed a new exoplanet imaging technique and successfully identified a giant planet in data previously captured by the James Webb Space Telescope (JWST) that conventional image processing methods had failed to detect.

The researchers said the technique can improve the telescope’s sensitivity to faint planets by 2 to 15 times. The findings were published in Nature on November 4.

The planet, named 51 Eridani b, had previously been discovered by ground based telescopes, but interference from the intense light of its host star had prevented astronomers from identifying it in a set of infrared observations by the Webb telescope.

A major challenge in directly imaging exoplanets is the enormous difference in brightness between stars and planets. Even relatively bright, young giant planets may be about 100,000 times fainter than their host stars. After passing through a telescope’s optical system, starlight forms a bright halo in the image that obscures the faint light signals from nearby planets.

Astronomers have long used coronagraphs to block starlight so they can observe surrounding planets. However, some starlight still leaks around the instrument, creating a residual halo.

For about the past 20 years, researchers have typically observed another star to obtain a reference image, then compared it with and subtracted it from an image of the target star to remove unwanted starlight. But a telescope’s mirrors and other optical components cannot remain perfectly stable at all times. Tiny shifts, tilts and temperature changes can alter the distribution of light, creating differences between the reference image and the actual observation and leaving enough residual light to obscure a planet.

To address this problem, a research team led by Northwestern University doctoral student Rodrigo Ferrer-Chavez developed a computational method called Physics-based Learning for Imaging Planets (FLIP). IT Home noted that Jason J. Wang, an assistant professor of physics and astronomy at Northwestern University, is also a co-author of the paper.

Instead of relying entirely on actual images of reference stars, FLIP builds a digital twin of the telescope. By simulating how starlight passes through the telescope’s mirrors and optical components before reaching the detector, it reconstructs the stellar halo in the image.

The researchers use computational methods including automatic differentiation and gradient optimization to adjust millions of parameters in the model, bringing the simulation as close as possible to the telescope’s actual observations. The model also accounts for optical distortions, tiny movements of the instrument and other imperfections, allowing it to reconstruct starlight more accurately and subtract it from the original image.

The research team stressed that although FLIP uses computational optimization tools similar to those used to train artificial intelligence models, it does not itself use artificial intelligence to identify planets. Instead, it builds a model based on the known physics of the telescope.

Wang said the researchers already understand the physics of how light passes through the telescope, so they do not need artificial intelligence to understand what is happening in the image.

The research team tested FLIP on data previously captured by the Webb telescope for 3 stars: HIP 65426 and 51 Eridani, which have known planets, and TYC 5899, which previously had no known planetary companions.

The tests showed that FLIP outperformed traditional image processing methods across different sets of observations, improving detection sensitivity by about 2 to 15 times. Its most notable result was identifying the giant planet 51 Eridani b in the Webb telescope’s infrared observations of the 51 Eridani system.

Researchers had previously discovered the planet using ground based telescopes, but traditional processing methods had consistently failed to identify it in Webb telescope observations at a wavelength of 4.3 micrometers. This band is especially important for studying carbon dioxide in planetary atmospheres because carbon dioxide strongly absorbs infrared light at that wavelength, further weakening the planet’s already faint signal.

After reprocessing the same observations with FLIP, the researchers successfully isolated the light signal from 51 Eridani b. According to the paper, the new method achieved a signal-to-noise ratio of about 6.3 in observations made with the F430M filter, where traditional methods had failed to detect the planet, reaching a level of significant detection.

Wang said the result shows that some past telescope observations may already contain valuable information about planets, but sufficiently effective tools to extract it had been lacking.

The technique could also affect future space telescope missions. NASA is advancing plans for the Habitable Worlds Observatory (HWO), one of whose goals is to directly image Earth like planets orbiting Sun like stars.

Because Earth like planets are fainter than giant planets, future telescopes may need to maintain the stability of their optical systems at the picometer level, a scale of one trillionth of a meter. This requirement could mean that telescopes need to be about 100 to 1000 times more stable than the Webb telescope.

The research team believes that image processing techniques based on physical models, such as FLIP, might in the future compensate for some small instabilities in telescopes and reduce the need for extreme hardware stability. However, the method still needs further optimization of its computational costs and validation with more telescopes and observational data.

Ferrer-Chavez, the paper’s first author, said that humans have not yet found planetary systems around other stars similar to the planets of the solar system. But that does not mean such systems do not exist; current observational techniques may simply be unable to identify them. As detection techniques continue to improve, astronomers may find more types of planets and further study how planetary systems form and evolve.

Found an error? Send a correction