(The Herald Post) – Researchers have successfully utilized machine learning to generate an updated image of the supermassive black hole at the center of the M87 galaxy, offering an unprecedented view of the celestial phenomenon. The groundbreaking technique, called PRIMO, enabled the scientists to achieve full resolution of the array, which previously resulted in an image commonly referred to as the “fuzzy, orange donut.”
PRIMO, or principal-component interferometric modeling, was developed by a team of Event Horizon Telescope (EHT) members, including Lia Medeiros, Dimitrios Psaltis, Tod Lauer, and Feryal Özel. The publication detailing the image reconstruction, titled “The Image of the M87 Black Hole Reconstructed with PRIMO,” is now available in The Astrophysical Journal Letters.
The technique employs dictionary learning, a branch of machine learning, to analyze over 30,000 high-fidelity simulated images of black holes accreting gas. The algorithm then identified common patterns in the structure of the images, which were sorted based on the frequency and blended to produce an accurate representation of the EHT observations. The breakthrough could have significant implications for interferometry across various fields, including exoplanets and medicine.
Researchers have confirmed that the new image aligns with both the EHT data and theoretical expectations, such as the bright ring of emission produced by hot gas falling into the black hole. PRIMO builds on the 2019 discovery that the M87 black hole appeared as predicted in broad detail.