CSIRO researchers independently validate Event Horizon Telescope M87 black hole image

Four researchers from CSIRO used a new imaging method to independently validate the Event Horizon Telescope’s landmark 2019 image of the M87* black hole shadow, publishing the results in Nature in 2026.

Dr Samuel Lai, the CSIRO astronomer and lead author, said the resulting image matched the historic Event Horizon Telescope picture in size and shape. He said the model had no prior knowledge of black hole appearance before it processed the M87 data. “It is like standing in your backyard and trying to take a photo of a doughnut on the moon,” Lai said, describing how hard it is to resolve an object so distant. The peer-reviewed paper is Samuel Lai et al., “Independent interferometric observable confirms black hole shadow in M87*,” published in Nature in 2026 with DOI 10.1038/s41586-026-11241-9.

Dr Laurent Loinard, director of the Event Horizon Telescope Collaboration, said the GenDIReCT method is entirely independent and recreates the M87* ring shape over multiple years, in close agreement with results the collaboration has published. He said such studies are central to science, giving valuable independent validation of Event Horizon Telescope results and enriching the toolkit for the whole community.

Instead of the usual pairs, CSIRO’s method uses groups of three telescopes to remove errors and biases from very long baseline interferometry, in which worldwide telescopes act as one planet-sized instrument. Dr Nithyanandan Thyagarajan, the CSIRO project leader, built the technique using experience from synchrotron experiments. He said imaging with very long baseline interferometry is tricky because it can introduce errors and biases that raise accuracy questions. Grouping telescopes in threes rather than pairs lets the method preserve structural information while canceling errors from the atmosphere or the telescopes.

To reduce bias from preconceptions about a black hole‘s appearance, the Australian team trained a machine-learning program on pictures that are not of black holes. The result closely resembled the original image released in 2019. By using advanced machine learning and artificial intelligence to remove atmospheric interference, the Nature-published approach reduces the effect of calibration mistakes and personal choices in image processing. The first M87 image changed understanding of extreme objects, but scientists have questioned how much processing methods for sparse telescope measurements influenced the final picture.

Researchers at Kobe University in Japan reported in 2022 that slight changes to assumptions, applied through the Event Horizon Telescope project’s usual data-analysis methods, removed the ring entirely from M87 and left an image of a radio-emission clump instead. Some researchers had thought the ring came from image-reconstruction methods rather than a real object. Thyagarajan said the small number of telescopes worldwide leaves large measurement gaps that imaging algorithms must fill, and short radio wavelengths allow the atmosphere to distort signals. Event Horizon Telescope data leaves more room for interpretation than many other astronomical observations, and processing choices can strongly shape the final picture.

The Event Horizon Telescope project said in The Astrophysical Journal Letters in 2019 that coordinated observations of the M87 black hole in the Virgo Cluster confirmed, for the first time, a ring-shaped black hole. The project connected radio dishes worldwide to work as one Earth-sized antenna, building the picture from April 2017 observations at eight telescopes in places such as Antarctica and the Andes. More than 20 institutions globally, including the Korea Astronomy and Space Science Institute, participated, and hundreds of scientists produced the image.

The black hole in M87 lies 55 million light-years from Earth, and its mass is 6.5 billion solar masses. A light-year is the distance light covers in a year, roughly 9.46 trillion km. A black hole’s gravity is so powerful that it traps everything, including light, inside. Researchers studied the dark outline using light from material moving at extreme speeds around the object.

The boundary is called the event horizon, and matter that crosses it cannot return.

In 1915, Einstein’s general relativity explained that a massive celestial object can collapse under its own gravity to form a black hole. The M87 image supported that century-old theory by showing how gravity curves spacetime and redirects light. Scientists first proposed the concept of a black hole in the 18th century, and it became a subject of scientific study through Einstein’s general relativity in 1915. If an object with Earth’s mass became a black hole, its diameter would shrink to one centimeter.

Material near the event horizon travels at close to light speed, and the disk moving toward Earth appears brighter than the part moving away, making one side of the M87 image look brighter than the other. The 2019 image displayed light from orbiting material instead of the black hole itself.

The researchers said the result confirms the Event Horizon Telescope’s finding that general relativity provides an accurate description of spacetime around a black hole. Lai called it an important step in the scientific process and the first independent validation of the Event Horizon Telescope image.

Thyagarajan said the method developed by CSIRO could also assist with medical imaging and earthquake mapping. He described it as a space-derived answer to practical problems on Earth. “we reconstructed the data from the 2019 paper using a newly developed independent imaging algorithm,” Thyagarajan said.

Wong said the work began as a chat among office colleagues and exceeded her initial expectations, with Samuel Lai’s ingenuity playing a key role. She said working with a diverse team of experts is rewarding and described the results as the cherries on top. “It has been incredibly rewarding to see how far this project has come,” Wong said.

Diakogiannis said Lai created a method in which custom-built telescope measurements teach the system to merge and improve AI-generated images, tailoring the learning to each observation. He called it a smart contribution and a key part of making the method work, and said it shows Lai’s creativity in solving a hard scientific problem.