An iconic first-ever view of a supermassive black hole has a dramatic new look thanks to machine learning.
The photo that captivated the world in 2019 showed a donut of bright, blurry lights. but, Astrophysics Journal Letter Sharpen the view to a narrow ring against an inky background on April 13th. The new images lay the foundation for future advances in understanding black holes, scientists say.
Carnegie Mellon University astrophysicist Tiziana Di Matteo said: I was working and was not involved in any new research. “It’s a beautiful example of how things can get better and see farther, you can literally see sharper,” she says.
Galaxy M87 lies about 54 million light-years from Earth. At its center is a black hole with about 6.5 billion times the mass of the Sun. The behemoth is one of his two primary targets for the Event Horizon Telescope (EHT), a coalition of radio observatories located around the world. Combining data from these sources, scientists have essentially built an Earth-sized telescope. This is powerful enough to capture details in the bright matter swirling around the black hole.
But EHT has a fundamental problem. The data was spotty, with light streaming through only a few patches, much like a scene viewed through a dirty window. The 2019 image and its new companion are based on data collected from just a handful of places on Earth, leaving a big gap in how scientists see black holes.
That’s where machine learning comes in. Behind his 2019 original and current enhanced view of M87’s black hole is imaging technology that uses machine learning to act as a sort of “mathematical detective,” says the Massachusetts Institute of Technology. says Kazunori Akiyama, an astrophysicist at Haystack Observatory, a member of the Event Horizon Telescope Collaboration, but not participating in the new research.
When scientists created the first images, they relied on popular machine learning systems to fill in the gaps. (For example, such a system might decide that two adjacent pixels are more likely to be of approximately the same brightness rather than being significantly different.) The characteristic ring-shaped image results from that process. appeared, it helped convince scientists that it was really what they were seeing. in a black hole. However, the blurred rings made it difficult to learn more about the black hole.
“Our thinking was, of course, that this was the first time we’d seen a black hole and we didn’t want to make any assumptions about it,” said Lia Medeiros, an astrophysicist at the Institute for Advanced Study. increase. Princeton, N.J., and the authors of the new study, which also contributed to the creation of the 2019 image.
Convinced that EHT’s first artificial intelligence augmentation method worked well on the 2019 images, Medeiros and her colleagues decided to use a slightly different and arguably more sophisticated alternative. (Primo).
PRIMO runs on rules derived from how scientists predict what a black hole would look like. This rule was gleaned from training on a large number of simulated black holes with different properties (different masses, different spins, etc.). The result is a more professional algorithm.
“This is a completely new method,” says Akiyama. “They use different assumptions about what kinds of images are likely.”
Medeiros and her colleagues then applied PRIMO to the same initial EHT data. A more physics-friendly rule creates a sharper image that depicts a narrower ring surrounding a truly black center. A sharper field of view could also change scientists’ understanding of massive objects, as scientists believe that properties such as the width of the rings reflect fundamental features of black holes. . However, the new research does not delve deeply into these potential effects. A paper for that is still in the works, she says, Medeiros.
Like the iconic 2019 image, the new PRIMO image isn’t the last portrait of M87’s black hole. Akiyama hopes the PRIMO algorithm will be tested more thoroughly, and Di Matteo stresses that the approach will become more powerful as scientists continue to improve their understanding of the physics governing black holes. I’m here.
Jiri Yunshi, an astrophysicist at University College London, who is a member of the Event Horizon Telescope Collaboration but was not involved in the new study, agrees. “Obviously, we have a lot more work to do to explore this algorithm and do more tests, but the results could be very exciting,” he says.
