Milky Way’s Central Black Hole Spins Near Light Speed, New Study Reveals
Sagittarius A*, the supermassive black hole at the center of the Milky Way, has been studied extensively, but recent research using artificial intelligence (AI) and extensive simulations has uncovered new, surprising details about its spin.
An international team of astronomers, supported by computing resources at the University of Wisconsin-Madison, trained a Bayesian neural network with nearly one million synthetic black hole images. These images were generated through general relativistic magnetohydrodynamics (GRMHD) simulations that model matter behavior under extreme gravity.
Revealing Hidden Patterns with AI
The Event Horizon Telescope (EHT) first imaged a black hole in 2019 (M87 galaxy) and then Sagittarius A* in 2022. However, the raw data held deeper insights that standard analysis couldn't extract. By pairing each simulated image with variables such as black hole spin, magnetic fields, and disk temperature, the AI learned to identify subtle patterns in the observational data.
Unlike typical AI models, the Bayesian neural network also estimates uncertainty, providing confidence levels for its conclusions. This approach revealed that Sagittarius A* spins near its theoretical maximum speed, with its spin axis oriented almost directly toward Earth. This challenges earlier assumptions about black hole orientation and magnetic field structure.
Computing at Scale Enables Breakthroughs
The breakthrough relied heavily on high-throughput computing (HTC), which distributes millions of small computational tasks across networks of computers. The Center for High Throughput Computing (CHTC) facilitated running over 12 million simulation jobs over three years, a scale unattainable with traditional supercomputers.
Software automation handled data scheduling and processing, streamlining the immense workload. The Open Science Pool contributed additional resources from more than 80 institutions nationwide, coordinated by the PATh project.
Improving Telescope Data Quality
To ensure reliable results, the team enhanced the original EHT data through improved calibration pipelines and signal stabilization software. They combined data across different frequency bands and polarizations, increasing signal clarity and reducing noise.
By forward modeling—predicting telescope observations based on theoretical models—and comparing hundreds of thousands of simulations with real data, they eliminated unrealistic scenarios. This produced a library of 932,000 synthetic datasets simulating effects like telescope errors and atmospheric turbulence, closely matching actual EHT observations.
Key Findings and Implications
- Spin Speed: Sagittarius A* spins at nearly the speed of light, influencing how it accretes matter and emits energy.
- Accretion Disk Dynamics: The bright ring observed is primarily due to hot electrons in the accretion disk, not jets as previously thought.
- Magnetic Fields: The magnetic field behavior near the black hole differs from earlier theoretical models.
The scale of simulation data—millions instead of dozens—was critical for the neural network’s success. This required collaboration between astronomers, computer scientists, and software engineers.
Looking ahead, future EHT upgrades will include more telescopes, higher frequencies, and improved resolution. This will allow AI models to train on even richer datasets, leading to more precise insights into black hole physics.
For professionals interested in AI applications in research, this study exemplifies how large-scale simulation data combined with Bayesian neural networks can deepen scientific analysis. To explore related AI training and courses, visit Complete AI Training.
References:
Your membership also unlocks:
