Astronomers were able to create the highest resolution video ever achieved, depicting a jet emanating from a supermassive black hole. To achieve this result, the team combined nearly three decades of data and applied a new machine learning model capable of reconstructing material movement and enhancing images.
The observed object is located at the center of the galaxy 3C 345, known as a blazar—an extremely active type of galaxy whose core emits a powerful jet of matter, including gas and plasma, at speeds close to the speed of light. The source of this jet is a supermassive black hole that potentially orbits a second black hole. The variable behavior of this jet has attracted astronomers' attention for many decades.
To create the video, researchers used 116 images taken between 1995 and 2022. These images were then processed by the Kine algorithm, which is based on neural networks and designed to improve such observations. Some team members also participated in the Event Horizon Telescope project, responsible for obtaining the first image of a supermassive black hole.
The video demonstrates three aspects of the jets: overall intensity, polarization (which serves as an indicator of the magnetic field), and optical flux, representing projected velocity. This result surprised scientists because there was a hypothesis that the brightest regions are shock fronts and should therefore move significantly faster than the surrounding material. However, the obtained data did not confirm this.
The study's authors state that they found no evidence that the moving bright components are regions heavily affected by shocks, as previously hypothesized. The interpretation based on shock fronts also weakened due to the lack of correlation between bright structures and polarization peaks. Although this does not completely rule out the shock front hypothesis, the observations raise new questions about what is actually happening in these mysterious relativistic jets.
One of the key elements of the work was the use of a neural network to reconstruct the jet's movement. Instead of analyzing each image individually, the algorithm allowed researchers to visualize the movement of plasma along the structure. This method provides a resolution approximately four times higher than traditional approaches, and a contrast 140 times higher. According to the researchers, this machine learning algorithm could likely be applied when it becomes possible to create the first video of a black hole's event horizon. The study was published in the journal Nature.