“Machine Learning Sharpens Image of M87* Black Hole to Reveal New Details”

SPACE — In 2019, scientists made a breakthrough by releasing the first photo of a black hole. This iconic photo of the black hole known as M87* was produced by the Event Horizon Telescope (EHT) project.

The EHT is a series of eight globally synchronized radio telescopes. M87* or Messier 87 is a solar system-sized black hole at the center of the Virgo galaxy cluster. The image of M87*’s black hole was made by collecting radio light that had traveled 53 million light years toward Earth.

Now, artificial intelligence (AI) has succeeded in sharpening the iconic photos of the M87*. Astronomers use machine learning to sharpen the first direct image of the M87* black hole.

Scroll to read

Scroll to read

Machine learning managed to clean up the image, sharpen the M87* image to achieve the maximum possible resolution. The results reveal a larger, darker central region of the black hole surrounded by glowing gas. The researchers published the images updated April 13 at The Astrophysical Journal Letters.

“With a new machine learning technique, PRIMO, we were able to achieve the maximum resolution of the current (telescope) array,” said Lia Medeiros, astronomer at the Institute for Advanced Study in Princeton, New Jersey, quoted from Live Science.

Studying black holes up close is impossible. As a result, the detailed images obtained with this technology play an important role in understanding black hole behavior.

Messier 87’s black hole is as wide as our solar system. Its mass is about 6.5 billion times that of the sun.

Black holes have a very strong gravitational pull. Theoretically, nothing can escape it, including light. However, that does not mean that black holes cannot be seen. This is because an active black hole is surrounded by an accretion disk, which heats up enough to produce a faint but detectable light.

It was from this faint radio emission that astronomers were able to reconstruct the distant singularity as a donut hole surrounded by a halo of light. However, the presence of gaps arising from the missing bits of light that no radio telescope could receive makes the image blurry and indistinct.

To sharpen the images, the researchers turned to a new AI technique called PRIMO. This technique analyzes more than 30,000 high-fidelity simulated images of black hole gas accretion to find general patterns.

These patterns are then sorted by how often they occur before being combined together and applied to the original image to produce a sharper estimate.

By examining the newly rendered image with the EHT data and theories about what a black hole should look like, the researchers confirmed that their image was a very close approximation of the actual object.