Get a clearer look at the M87 black hole

The iconic image of the supermassive black hole at the center of M87 — sometimes referred to as the “orange fuzzy donut” — got its first official makeover with the help of machine learning. The new image further reveals a larger, darker central region, surrounded by bright accretion gas in the shape of a “skinny donut”. The team used data obtained through the Event Horizon Telescope (EHT) collaboration in 2017 and achieved for the first time the full resolution of the array.

In 2017, the EHT collaboration used a network of seven pre-existing telescopes around the world to gather data on M87, creating an “Earth-sized telescope.” However, because it is not possible to cover the entire Earth’s surface with telescopes, gaps appear in the data – like missing pieces in a jigsaw puzzle.

“With our new machine learning method, PRIMO, we have been able to achieve maximum accuracy for the existing matrix,” says lead author Lia Medeiros of the Institute for Advanced Study. “Because we cannot study black holes up close, the details of the image play an important role in our ability to understand their behavior. The width of the ring in the image is now about two times smaller, which will be a strong limitation for our theoretical models and gravity tests.”

PRIMO, which stands for Principal Component Interferometric Modeling, was developed by EHT members Lia Medeiros (Institute for Advanced Study), Dimitrios Psaltis (Georgia Tech), Tod Lauer (NOIRLab) and Feryal Özel (Georgia Tech). Their publication, “Image of M87 Black Hole Reconstructed Using PRIMO,” is now available at Astrophysical Journal Letters.

“PRIMO is a new approach to the difficult task of creating images from EHT observations,” said Lauer. “It provides a way to compensate for missing information about the object being monitored, which is required to generate the image that would have been seen with a single giant Earth-sized radio telescope.”

PRIMO is based on dictionary learning, a branch of machine learning that enables computers to generate rules based on large sets of training material. For example, if a computer is fed a series of different banana images – with enough training – it might be able to determine whether or not the unknown image is a banana. Beyond this simple case, the versatility of machine learning has been demonstrated in many ways: from creating Renaissance-style artworks to completing An unfinished work by Beethoven. So how can machines help scientists provide an image of a black hole? The research team has answered this very question.

Using PRIMO, the computers analyzed more than 30,000 high-resolution simulated images of gas accreting black holes. The suite of simulations covered a wide range of models of how a black hole accumulates matter, looking for common patterns in the structure of the images. The different patterns of structure were sorted by how commonly they occur in the simulations, and then blended to provide a highly accurate representation of the EHT observations, simultaneously providing a high-resolution estimation of the missing structure of the images. A paper related to the algorithm itself has been published in Astrophysical Journal On February 3, 2023.

“We’re using physics to fill in missing data areas in a way that hasn’t been done before with machine learning,” Medeiros added. “This could have important implications for interferometry, which plays a role in fields from exoplanets to medicine.”

The team confirmed that the newly provided image is consistent with the EHT data and with theoretical predictions, including the bright ring of emissions expected to result from hot gas falling into the black hole. Creating an image requires assuming the proper shape of the missing information, and PRIMO did that by building on the 2019 discovery that the M87 black hole in vast detail looked just as predicted.

“Almost four years after revealing the first horizontal-scale image of a black hole by EHT in 2019, we have set another milestone, producing an image that uses the full resolution of the matrix for the first time,” Psaltis stated. “The new machine learning techniques we have developed provide a golden opportunity for our collective work to understand the physics of black holes.”

The new image should lead to more precise determinations of the M87 black hole’s mass and the physical parameters that determine its current appearance. The data also provides an opportunity for researchers to place greater constraints on event horizon alternatives (based on lower dark central brightness) and perform more robust tests of gravity (based on narrower ring size). PRIMO can also be applied to additional EHT notes, including those for Sgr A*The central black hole in our Milky Way galaxy.

M87 is a relatively nearby massive galaxy in the Virgo group of galaxies. More than a century ago, a mysterious jet of hot plasma was observed emitting from its center. Starting in the 1950s, new radio astronomy technology showed that the galaxy had a bright compact radio source at its center. During the 1960s, it was suspected that M87 had a supermassive black hole at its center that was powering this activity. Measurements made from ground-based telescopes starting in the 1970s, and later by the Hubble Space Telescope starting in the 1990s, provided strong support that M87 was indeed harboring a black hole weighing several billion solar masses based on observations of the high velocities of stars and gas orbiting its center. The 2017 EHT observations of M87 were obtained over several days from several different radio telescopes linked together at the same time for the highest possible resolution. The now famous “orangecake” image of the M87 black hole, released in 2019, reflected the first attempt to produce an image from such observations.

“Photo 2019 was just the beginning,” Medeiros said. “If a picture is worth a thousand words, the data behind that image has many more stories to tell. PRIMO will continue to be a critical tool in extracting such insights.”