Researchers at the RIKEN Center for Interdisciplinary Theoretical and Mathematical Sciences (iTHEMS) in Japan have developed an unprecedented computer simulation that accurately models the Milky Way galaxy, capturing more than 100 billion stars over a span of 10,000 years. In collaboration with colleagues from the University of Tokyo and the Universitat de Barcelona, this groundbreaking achievement represents a significant leap forward in the fields of astrophysics and supercomputing.
The simulation stands out not only for its scale but also for its remarkable speed, completing calculations at a rate 100 times faster than previous models. By harnessing the power of 7 million CPU cores, machine learning algorithms, and advanced numerical simulations, the team has created a hyper-realistic representation of our galaxy. Their findings were detailed in a paper titled “The First Star-by-star N-body/Hydrodynamics Simulation of Our Galaxy Coupling with a Surrogate Model,” published in the *Proceedings of the International Conference for High Performance Computing, Networking, Storage and Analysis* (SC ’25).
Advancing Astrophysical Research
Simulating the Milky Way with such granularity allows astronomers to test theories regarding galactic formation, structure, and evolution. Researchers have faced challenges in developing increasingly complex simulations due to the multitude of forces involved, including gravity, fluid dynamics, and the effects of supernovae and supermassive black holes. Until now, the computational limitations meant that models could only represent about 1 billion solar masses, which is less than 1% of the Milky Way’s total stellar count.
Existing supercomputing systems would take approximately 315 hours (over 13 days) to simulate just one million years of galactic evolution. By contrast, the new model allows for the simulation of 1 million years in just 2.78 hours. This breakthrough means that astronomers can analyze 1 billion years of galactic history in under 115 days.
The research team, led by scientist Hirashima, utilized an innovative approach by incorporating a machine learning surrogate model. This AI shortcut predicted the impact of supernova explosions on surrounding gas and dust without consuming the resources required for the main simulation. By training the model on high-resolution supernova simulations, they achieved a balance of efficiency and detail, allowing them to track both large-scale galactic dynamics and small-scale stellar phenomena simultaneously.
Implications for Future Research
The successful verification of this model was conducted through extensive testing on the Fugaku and Miyabi Supercomputer Systems, demonstrating its ability to simulate galactic dynamics with unprecedented detail. As Hirashima noted in a RIKEN press release, this method not only enhances the study of galactic evolution but also showcases the potential of integrating AI models into complex simulations.
Beyond the realm of astrophysics, the techniques developed in this study could be applied to other fields requiring intricate simulations, such as meteorology, ocean dynamics, and climate science. The research marks a pivotal moment in our understanding of the universe and paves the way for future explorations of cosmic phenomena, proving that the intersection of supercomputing and artificial intelligence holds immense promise for scientific advancement.
This innovative simulation reflects a major milestone in computational astrophysics and exemplifies how collaborative efforts across institutions can lead to significant breakthroughs, ultimately expanding our comprehension of the cosmos.