AI reveals new details about the Milky Way’s monster black hole

A new series of images and measurements of the Milky Way’s central black hole – Sagittarius A* (Sgr A*) – a region of space where gravity is so strong that nothing, not even light, can escape. This analysis suggests that Sgr A* turns rapidly and faces us with a slight tilt.

The new view of Sgr A* comes from software that learns patterns in the telescope’s signals that were previously considered too complicated.

In May 2022, astronomers studied our black hole with the Event Horizon Telescope (EHT), a global network of radio observatories that together act as a single giant telescope.

These beginnings raised big questions about how the system behaves over minutes and hours, and how matter swirls and heats up near the point of no return.

Using AI to study Sgr A* spin

The study was led by Michael Janssen of Radboud University, who helped design a machine learning pipeline that handles complex data from the Event Horizon Telescope.

His team developed an approach called a Bayesian neural network, a type of artificial intelligence that combines data models with probability theory to estimate both outcomes and uncertainty.

In the new study, the lattice indicates a high spin for Sgr A*, on the order of 0.8 to 0.9 in dimensionless units, and a viewing angle close to 20 to 40 degrees.

The analysis also defines a preferred direction on the sky for the axis of rotation, in a band between 106 and 137 degrees east of north.

The same work looks forward to near-term upgrades to the global radio network.

The team reports that the addition of the Africa Millimeter Telescope, a new high-altitude observatory planned in Namibia, would reduce some error bars by about a factor of three for some models that exceed the standards.

How EHT and AI measure Sgr A* spin

The Event Horizon Telescope is not a single observatory, it is an array of parabolic antennas that record arriving waves almost in synchrony and then combine them to mimic a planet-sized mirror.

The technique is called very long baseline interferometry (VLBI), a method that connects widely separated telescopes to create an image with extremely high resolution.

This method is powerful, but it is sensitive to tiny timing errors and air conditions above each dish. The collaboration showed how tropospheric fluctuations, or changes in the lower atmosphere caused by water vapor and temperature, and site-by-site differences challenge the calibration of these data and why careful processing is essential.

The new project addresses these issues by directly modeling the measurement process, rather than working only with the final images. This choice allows the network to learn parts of the dataset that older pipelines had to discard.

How the neural network learned

To teach the system what to look for, the researchers generated millions of snapshots of simulated black holes using general relativistic magnetohydrodynamics, a complex physics framework that describes the behavior of magnetic fields and charged gases according to Einstein’s theory of gravity.

These synthetic scenes were passed through a virtual Event Horizon telescope to create fake observations that resemble the real thing.

The neural network architecture, named ZINGULARITY, was then trained to map these observables to physical properties such as spin, disk orientation, and the electron temperature parameter (ETP). The comprehensive framework documents the Bayesian setup and steps used to avoid overconfidence.

One of the main advantages of the Bayesian design is that it reports a range of likely answers, not a single precise number. This is important when the sky is variable and the array only captures sparse snapshots during short nighttime windows.

What the new image claims

For Sagittarius A*, the inference favors a rapidly rotating hole with a prograde accretion disk, meaning that the gas and dust orbit in the same direction as the black hole’s rotation.

The model prefers a low tilt to our line of sight, meaning we look primarily down toward the flow rather than edge-on.

These parameters help explain why the central ring appears bright on one side and why the polarization pattern, the orientation of light waves caused by magnetic fields, behaves as it does in other independent data sets.

The team also notes that the jet, if present, should be weak near the hole, consistent with the lack of strong jet detection so far in this source.

Results include a sky orientation for the rotation axis that overlaps with previous polarization-based indices. This consistency, across very different methods, provides assurance that the network is exploiting real data characteristics rather than artifacts.

Why some scientists are cautious

“It is very difficult to process data from the Event Horizon Telescope. A neural network is ideally suited to solve this problem,” Janssen said, highlighting the problem the method tries to solve.

Skeptics worry that an array could latch onto subtle biases if the training library leaves out real-world effects or if the telescope misses key baselines, the distances between paired antennas that define the resolution.

Proponents respond that Bayesian estimates, bootstrapped noise models, and cross-checks with traditional pipelines can detect many of these failure modes.

Black hole spin and Sgr A*

Rapid rotation, if confirmed, defines how energy can be extracted from the rotation of the hole and how particles are accelerated near the inner edge of the flow.

It also narrows down theories about the evolution of the Milky Way’s center, including possible past galactic mergers, events in which two galaxies collide and their central black holes combine.

Knowing the tilt tells us which parts of the supermassive black hole’s environment are visible and which are hidden.

This detail allows interpretation of flickers and reflections observed at other wavelengths and relates them to movements of the ring that radio interferometers can resolve.

The result also highlights the value of combining data-driven tools with solid physics. Networks that explain their confidence and are tested on synthetic observations bridge the gap between raw measurements and the parameters that shape our models.

What’s next for the telescope

The Event Horizon Telescope has already been observed with more stations and at more frequencies since the 2017 campaign used here.

Rerunning the pipeline on these newer datasets should test whether the same rotation and orientation disappears again under different observing conditions.

Hardware improvements evolve in parallel. The integration of new stations on other continents and the lengthening of certain baselines should strengthen the fidelity of the image and reduce the risk that atmospheric vagaries will direct the responses.

As the library of simulations grows, the training set can expand beyond the usual families of disks and magnetic states. This will give the network a broader menu of possibilities and make its findings more resilient.

The study is published in Astronomy and astrophysics.

Image credit: EHT/Janssen et al collaboration.

—–

Do you like what you read? Subscribe to our newsletter for engaging articles, exclusive content and the latest updates.

Check us out on EarthSnap, a free app brought to you by Eric Ralls and Earth.com.

—–