“Numerical exploration of the universe”
Astronomer Eiichiro Kokubo was also fascinated by planetary science, but it led him down a different path. Kokubo wanted to watch a solar system actually taking shape. For that, he turned to a supercomputer.
“It’s very difficult to investigate the formation of planets by telescope because the timescale is very long, and the planets are very small and far away,” he says. “But what we can do is a numerical experiment. We can see how a protoplanetary disk evolves with time, or how the planets form. We can do this only with numerical simulations because we cannot see them directly.”
Today, Kokubo is director of NAOJ’s Center for Computational Astrophysics (CfCA), which operates the Cray XC30 supercomputer ATERUI, the most powerful supercomputer in the world dedicated exclusively to astronomy. Alongside observational and theoretical astronomy, Kokubo calls simulation astronomy “the third way of astronomical research.”
“The solar system was formed 4.6 billion years ago. We cannot see it. But we can set up a virtual universe in the computer, using laws of physics,” he says. “We often say supercomputers are a kind of theoretical telescope to see unseen things.”
The center’s role as a branch within NAOJ gives it a unique position to work closely with observational and theoretical astronomers, Kokubo notes. Computing time is available not only to native and foreign astronomers in Japan, but also to Japanese astronomers working abroad.
CfCA completed a major upgrade of ATERUI last year, increasing performance from 502 teraflops to 1.058 petaflops, giving it the capability to perform 1,000 trillion calculations per second. The added speed will allow ATERUI to serve more researchers, as well as to take on even more complex simulations, Kokubo says.
An early achievement for CfCA was its work on a project that simulated the so-called “giant impact scenario,” showing how a massive collision with Earth created a disk of rock and debris that quickly formed the Moon. The process had been theorized in the 1970s as a model, but had never been tested. “From this simulation, we could understand how the Moon is formed from the impact-generated disk, and also why we have only one moon,” Kokubo says.
More recently, ATERUI was used to create a high-resolution simulation of a supernova. “This is the most inner part of a supernova explosion, which we cannot see from observation,” he says. The simulation was so complex, it required the use of the entire supercomputer at once, something CfCA schedules for large projects about once a month.
Kokubo says working with simulations has allowed him to fulfill not only his scientific ambitions, but also a childhood dream. “When I was an elementary student, I thought I wanted to be an explorer,” he says. “When I’m looking at the solar system, I can’t physically be an explorer, but I can do it with supercomputers. It’s a numerical exploration of the universe.”