An article published in “The Astrophysical Journal Letters” reports the results of using a machine learning system to obtain a sharper and more detailed version of the image showing the area around the supermassive black hole at the center of the galaxy M87. Lia Medeiros (Institute for Advanced Study), Dimitrios Psaltis (Georgia Tech), Tod Lauer (NOIRLab), and Feryal Özel (Georgia Tech), all members of the Event Horizon Telescope (EHT) Collaboration that obtained the image that has become famous, developed the PRIMO system to reprocess the data collected during the original observation campaign in 2017. Their aim is to obtain the maximum possible resolution with all the details present in the data collected by the various radio telescopes participating in the EHT Collaboration. The image, to which the four researchers hold the rights, shows the comparison between the famous result published in 2019 and that of the first test with PRIMO.
The interferometry technique makes it possible to combine the detections of different radio telescopes. In the case of the EHT Collaboration, the instruments were scattered all over the world achieving results never seen before precisely because they were enabled to work as if they were a single colossal instrument. However, the fact that the antennas that were used for the detections are scattered here and there around the world prevents us from obtaining the same results in terms of collected data as a single antenna with the Earth’s diameter. Again, very advanced computer technologies were used to obtain otherwise impossible results to compensate for that lack of data.
Machine learning systems have been used in many fields for years but the general public has only recently started getting to know them thanks to the common use of chatbots. These are systems often defined as artificial intelligence in a decidedly optimistic way, but that’s another topic. In an article published in February 2023 in “The Astrophysical Journal”, the four researchers of the EHT Collaboration who have now published the first results, described the PRIMO (Principal-component Interferometric Modeling) system.
In very simple terms, PRIMO was trained using over 30,000 very precise simulations of black holes surrounded by gas. This allowed PRIMO to reprocess the data of the galaxy M87 collected by the EHT Collaboration in the 2017 observation campaign, including a very precise estimate of the missing structures.
The result is visibly different from the original processing of the area around the supermassive black hole at the center of M87 published in 2019. The PRIMO-generated image shows the bright ring surrounding the black hole’s shadow much thinner and much more detailed. At the same time, the shadow of the black hole in the center of the image is larger in size.
A more precise reconstruction of the image is important because it allows more rigorous testing of black hole models and a more precise estimation of the characteristics of the supermassive black hole at the center of the galaxy M87.
For the researchers, this is just the beginning and the PRIMO system will also be applied to the data collected in subsequent observation campaigns of the EHT Collaboration such as the one that led to the publication last year of an image of the area around Sagittarius A*, the supermassive black hole in the center of the Milky Way. Gradually, these studies are allowing scientists to learn more about these truly extreme objects.
