RIKEN Researchers' AI Simulation Models 100 Billion Stars in Milky Way
- MM24 News Desk
- 3 days ago
- 2 min read
Scientists at Japan's RIKEN Center for Computational Science, led by Keiya Hirashima, have created the world's first Milky Way simulation to accurately model over 100 billion individual stars using a novel AI-powered approach.
This breakthrough, achieved on the Fugaku supercomputer using 7 million CPU cores, simulates galactic evolution 100 times faster than previous state-of-the-art models while representing 100 times more stars, opening new frontiers for understanding our galaxy's formation.
For decades, astrophysicists have struggled to create a simulation of our galaxy that captures both its immense scale and the intricate lives of individual stars. The challenge lies in the vastly different scales of space and time involved; a supernova explosion is a fleeting, local event, while galactic evolution unfolds over billions of years.
Conventional simulations hit a hard limit, forced to represent clusters of stars as single, massive particles. This meant the fate of individual stars was averaged out, and key phenomena driving galactic evolution were lost.
The research team's ingenious solution was to merge traditional physics-based simulations with a deep learning surrogate model. This AI was specifically trained to predict how gas expands in the 100,000 years following a supernova explosion.
"I believe that integrating AI with high-performance computing marks a fundamental shift in how we tackle multi-scale, multi-physics problems across the computational sciences," said lead researcher Keiya Hirashima. This AI component acted as a sophisticated shortcut, handling fine-scale events without draining computational resources from the main galaxy model.
The performance leap is staggering. According to the team's paper, published in the international supercomputing conference SC ’25, a conventional simulation attempting individual star resolution for the Milky Way would take an impractical 36 years to simulate just 1 billion years of galactic evolution.
The new AI-accelerated method slashes that time to a feasible 115 days. This means simulating 1 million years of galactic life now takes just 2.78 hours, a pace that was previously unimaginable.
Verification runs on RIKEN’s supercomputer Fugaku and The University of Tokyo’s Miyabi Supercomputer System confirmed the simulation's accuracy, producing a dynamic model that faithfully represents stellar evolution over 10 thousand years.
This achievement goes beyond pattern recognition; it establishes AI as a core tool for genuine scientific discovery. Hirashima emphasized this point, noting the simulation will help "trace how the elements that formed life itself emerged within our galaxy."
The implications of this methodology extend far beyond astrophysics. The same approach to bridging scale gaps can revolutionize simulations in climate science, weather forecasting, and oceanography, where linking small-scale turbulence to global systems remains a monumental computational challenge.
By successfully modeling the unfathomable complexity of 100 billion stars, the RIKEN team has not only recreated our galactic home in unprecedented detail but has also pioneered a new paradigm for computational science itself.
Comments