The iconic 2019 image of M87*, a black hole the size of our solar system located at the center of the Virgo galaxy cluster, was formed by synthesizing radio waves that traveled to us over 53 million light-years in the universe.
.jpg)
Compare the image of the M87* black hole before (left) and after (right) being enhanced using the PRIMO algorithm.
Now, scientists have employed machine learning to clean up the image, sharpening it to achieve the highest possible resolution and reveal a larger and darker central region surrounded by a glowing ring described by astronomers as a “thin bagel.” The updated image was published on April 13 in The Astrophysical Journal Letters.
“With our new machine learning technique, PRIMO, we could achieve the maximum resolution of the current array of telescopes,” said lead author Lia Medeiros, an astronomer at the Institute for Advanced Study in Princeton, New Jersey, in a statement. “Since we can’t study black holes up close, the details of the image play a crucial role in our ability to understand their behavior. The inner diameter of the ring in the current image is smaller than we thought, which will be a strong constraint for theoretical models and tests of gravity,” added astronomer Lia Medeiros.
The black hole Messier 87, which is as wide as our solar system and has a mass 6.5 billion times that of the Sun, was captured by the Event Horizon Telescope (EHT), an array of eight synchronized radio telescopes across the globe.
Black holes possess such strong gravitational pull that nothing (including light) can escape their grasp, but that doesn’t mean they can’t be observed. This is because active black holes are surrounded by accretion disks — massive rings of matter ripped away from gas clouds and stars orbiting around the black hole’s event horizon — which become heated to red-hot temperatures through friction, emitting faint yet detectable light.
It is from these faint radio emissions that astronomers were able to reconstruct the distant anomaly as a glowing ring-like shape. However, gaps in the data, resulting from missing puzzle pieces without any radio telescopes to capture them, caused the image to be blurry and unclear.
To enhance the image, researchers turned to a new AI technique called Principal Component Interferometric Imaging (PRIMO), analyzing over 30,000 high-fidelity simulated images of black hole gas accumulation to identify common patterns. These patterns were then sorted by their prevalence before being combined and applied to the original image to generate a sharper estimation.
By comparing the newly derived image with EHT data and theoretical predictions of what a black hole should look like, researchers confirmed that their image closely aligns with reality.
The research team further stated that this new image will allow for more detailed studies of extreme effects caused by cosmic black holes, where our theories of gravity and quantum mechanics break down and merge.