The most detailed simulation of our Milky Way galaxy ever created on a supercomputer was created by combining AI with numerical models. Running 100 times faster than previous, equally detailed models, the new program gives astronomers the ability to map billions of years of the Milky Way's evolution in months instead of decades.
See also: On Saturn's Titan, water and oil mix
The new simulation includes 100 billion particles that correspond to stars – about as many as there are in the Milky Way itself. Previous high-resolution simulations could only handle a billion stars and were very slow. Modeling a million years of galactic evolution with the older methods took 315 hours, or about 13 days of real time; this meant that simulating a billion years would take almost 36 years of computing time.
Now, a team led by Keiya Hirashima from the RIKEN Center for Interdisciplinary Theoretical and Mathematical Sciences in Japan has developed a new supercomputer-powered simulation that is faster and offers much greater resolution of stars, thanks to a new methodology that allows for the incorporation of both short- and long-term evolutionary processes.
See also: Humanoid robot danced for Vladimir Putin
By comparison, previous simulations with the best resolution had just a billion “particles.” Each particle represented about 100 stars, but that “smoothed out” details — for example, the effect that a single supernova might have on the surrounding gaseous environment.
Thus, older models emphasized mainly long-term events and less so short-term phenomena associated with individual stars, even though these short-term phenomena are often crucial to the broader, long-term evolution of the galaxy.
See also: James Webb: Galaxies in the early universe were chaotic
The results of the new simulation were published in the context of an international supercomputing conference called SC '25.
