Independent study confirms Event Horizon Telescope’s 2019 M87 black hole ring

Independent researchers using new imaging techniques have confirmed the bright ring in the Event Horizon Telescope’s first picture of a black hole, revealed April 10, 2019, showing M87’s center about 55 million light-years from Earth; researchers completed the check Dec. 4, 2025, and published the result Nov. 15, 2026, in Nature.

The confirmation addresses a dispute, because independent researchers disagree over whether the bright circle in that iconic image is authentic light or a product of the imaging process: a 2022 study rejected the ring while a 2026 study in Nature confirmed it. The question matters because the Event Horizon Telescope Collaboration’s original publication drew on four separate imaging teams, and every team generated a ring-like structure while differing in fine-scale features. The collaboration traced those differences to human and algorithmic biases. Phys.org describes the new Nature research as enhancing confidence in a significant modern scientific result and restoring attention to the black-hole image.

The Event Horizon Telescope links radio telescope signals around the world so that many dishes act as one Earth-sized instrument, a technique called interferometric imaging that produces the sharpest pictures available from Earth’s surface. Resolving M87’s black hole is like photographing a doughnut on the moon as seen from Earth; reaching that sharpness would demand an instrument spanning the full planet. The array observes the black hole’s shadow, defined by light from hot gas swirling around the object.

A team led by Makoto Miyoshi of NAOJ, with Yoshiaki Kato of RIKEN and Junichiro Makino of Kobe University, reanalyzed the Event Horizon Telescope’s publicly available radio data from April 2017 in work dated June 30, 2022, with a preprint already available. The team aimed to publish it in the Astrophysical Journal. Many supermassive black holes emit astrophysical jets, and astronomers have known about M87’s jet for more than 100 years; M87 is an elliptical galaxy. Instead of a ring, the team reported a core, an astrophysical jet and knots at M87’s center, and it believes it resolved the base of that jet.

The team contended that the 40-micro-arcsecond ring seen in the image, equal to 1/25,000th of an arcsecond, probably came from inadequate data owing to the small number of telescopes involved. At that time fewer telescopes took part, leaving insufficient data to resolve structures at that scale, unlike structures at other sizes. The radio data and analysis methods for the center of M87 had been accessible to the public worldwide.

An independent study published in Nature re-examined the same data sets with new imaging techniques and validated the original ring image. Its central method, closure invariants, combines triplets of telescope measurements at once rather than pairs. Standard analysis combines pairs of measurements, so the three-way combination guards against atmospheric and instrumental errors. Every dish adds instrument and atmosphere errors that residual correction may not fully remove, and the team designed the method to strip out errors from instruments and the atmosphere.

The researchers also built a machine-learning imager called GenDIReCT. They trained it on ordinary photographs of cats, dogs and cars so it entered the task with no prior knowledge of black holes, then froze its training to block human bias. Before locking the procedure, they ran extensive validation exams on non-ring structures and on real data from other galaxies, including Centaurus A.

On Dec. 4, 2025, researchers ran the fully validated independent imager over the Event Horizon Telescope’s M87 observations and found the ring structure held up across every data set. The independent findings match the earlier pictures in diameter and general direction, while ring width and surface brightness vary around the circumference, and the agreement resolves discrepancies raised by the earlier challenge.

The team did not rest its case on machine learning alone. It also ran a separate traditional statistical imaging method, and that method also confirmed the ring structure using closure invariant data from the team. Phys.org’s reporting describes this separate method as evidence that machine learning was not the only reason the check succeeded. The same reporting suggests the technique might help fields such as medical imaging and earthquake mapping, which blend many sensor views into charts of unknown structures.

SpaceNews’s reporting expects new data reanalysis, fresh method examination and planned follow-up observations to yield more credible insight into M87’s center and the structure of the jet blasting from its galactic center. The outlet frames such independent re-checks as the normal scientific process behind credible results drawn from public observational data. The reporting also notes that the radio data and analysis methods had been accessible to the public worldwide.