[ad_1]
Researchers in Japan and Spain have created the first simulation of the Milky Way. It can track over 100 billion stars over a period of 10,000 years. It combines deep learning with high-resolution physics.
The researchers are led by Keiya Hirashima. The study was conducted at the RIKEN Center for Interdisciplinary Theoretical and Mathematical Sciences (iTHEMS) in Japan. According to ANI, they worked with partners from the University of Tokyo and the Universitat de Barcelona in Spain.
Scientists have long struggled to create a model of a galaxy as large as the Milky Way with enough detail to track single stars. Current simulations can handle systems equivalent to about a billion Suns. Still, that’s far less than the galaxy’s more than 100 billion stars.
The new AI learned how gas behaves. This model is now hundreds of times faster than older methods. The simulations used 100 times more stars than previous work.
The breakthrough, demonstrated at the SC ’25 supercomputing conference, marks a major step forward for astrophysics, high-performance computing and AI-supported modeling. The same method could also help in larger Earth system studies, including climate and weather research.
Each tiny “particle” in those models often represents about 100 stars. This hides individual behavior and weakens accuracy. The problem arises from the small time steps required to capture fast events such as supernovae.
Small steps require large computing power. Complete star-by-star galaxy simulations will take decades. It also adds more supercomputer cores which is neither efficient nor practical.
How researchers dealt with the limitations of galactic modeling
Hirashima and his team tackled the limitations of galactic modeling by combining deep learning with traditional physics. Their method used a trained surrogate model that learned gas behavior from detailed supernova simulations.
This surrogate predicted how the gas would expand for about a million years without slowing the main flow after each eruption. The hybrid approach kept the galaxy’s broad structure accurate. It still captures smaller events like individual supernova details.
The team found close agreement when testing the results against larger runs on the Fugaku and Miyabi system. Similar methods could transform weather, ocean, and climate studies.
“I believe that integrating AI with high-performance computing marks a fundamental shift in the way multi-scale, multi-physics problems are tackled in computational science,” ANI quoted Hirashima as saying.
“This achievement also shows that AI-accelerated simulations can move beyond pattern recognition and become a real tool for scientific discovery – helping us discover how the elements that create life emerged within our galaxy,” Hirashima said.
[ad_2]
