Supernova simulations using the OLCF’s Summit shed new light on earthly elements and their origins

Simulations performed on Oak Ridge National Laboratory’s Summit supercomputer provide a more detailed look at how stars die and could help unlock new insights into the origins of Earth’s heavy elements.

The study, conducted by scientists at ORNL and the University of Tennessee, Knoxville, modeled the collapse and explosion of two stars—one 9.6 times the mass of Earth’s sun, the other 10 times the mass of the sun. The computational power of Summit, the Oak Ridge Leadership Computing Facility’s flagship IBM AC922 system, enabled 3D simulations to track the evolution of each supernova, from the early stages of each explosion to the breakout of the surface-level shockwaves that scatter heavy-metal isotopes across the reaches of the universe.

“These simulations help us compare to observations and answer the questions that will solve the origin story of the elements around us,” said Michael Sandoval, an ORNL high-performance computing engineer and lead author of the study recently published in The Astrophysical Journal. “The calcium in your bones, the iron in your blood—they come from stars, ejected during explosions like this one. Models like this can bring us closer to unraveling how it all began.”

Scientists seldom get the chance to observe a nearby supernova in action. One of the few opportunities in the modern era came in early 1987, when telescopes in orbit and on the ground captured the death of a star on the outskirts of the Tarantula Nebula in the Large Magellanic Cloud, a satellite galaxy of the Milky Way. The explosion, known as SN1987A, erupted with the brightness of 100 million suns to produce the most visible supernova in four centuries and provide data that’s fueled astronomy and nuclear physics studies for more than three decades.

Scholars typically stop short of trying to model a full supernova. Sandoval and his team wanted to push those limits.

“Most simulations focus on the beginning of the supernova,” he said. “That’s the first couple seconds of the big boom. But the full force of the explosion doesn’t reach the surface of the star where we can see it until hours or days later. We wanted to follow that thread all the way to the end. So if you think of a supernova as the word ‘Boom,’ you can think of most studies as looking at the ‘B-,’ and ours as focused on the ‘-Oom.’ ”

The 3D simulations on Summit began with each explosion already in progress and focused on tracing the explosive evolutions in time and the paths of the heavy elements that erupt from a supernova like cosmic shrapnel at speeds as high as 4,500 kilometers, or 2,796 miles, per second.

3D simulation of a supernova using Summit

Concentrations of red (a lighter isotope of nickel) and green (a heavier isotope of nickel) reveal the early formations of nickel fingers or bullets, surrounded by a shell of silicon, that will be expelled from a star’s core during a 3D simulation of a supernova using Summit. Credit: Sandoval, et al./ORNL

“As the explosion progresses, the shock wave hits the various layers of the star, such as carbon and oxygen between helium or helium between hydrogen,” Sandoval said. “When that shock hits, these instabilities form fingers that spread throughout the star and punch through these layers. Helium acts as the coating around the star’s heavier elements, such as nickel, so we see these radioactive isotopes ejected like bullets. Because they’re fast and heavy, they’re not slowed down by the star’s mass.

“We wanted to model and track these bullets as far as we could. It’s like a domino effect—one thing sets off another—so we needed to capture all the rough edges and asymmetries that form early and seed the bullets later.”

Because researchers typically lack access to a computational powerhouse like Summit to run 3D simulations, the team incorporated a 2D approach as well. Most 2D models treat the supernova like an orange, simulating a single wedge and treating the other wedges as identical.

“The problem is that 2D simulations assume symmetry,” Sandoval said. “But the universe is not symmetrical, and things are not that simple. Because these kinds of models take so long to run and are so computationally expensive, there aren’t a lot of options to choose from. We tilted the most appropriate 2D model by 90 degrees and mapped it in 3D. It’s still not nearly realistic enough, but it’s closer than any purely 2D model.”

The power of Summit’s 200 petaflops—equal to 200 quadrillion calculations per second—equipped the team to simulate the process in unprecedented detail. The model required 100,000 node hours and 10,368 GPUs to track 160 species of isotopes, using FLASH, a multiphysics hydrodynamics code.

“Most simulations have tracked only about 15 nuclear species,” Sandoval said. “There are plenty more isotopes to track, but that takes more computational power. That’s why we love Summit. We tried it on Titan (Summit’s 27-petaflop predecessor that was retired in 2019) first, but we didn’t get as realistic a picture. These simulations require repeating the same calculations over and over, so Summit’s capabilities were a godsend for us.

“By tracking so many species, we get a much more realistic picture of what’s ejected at the end of the supernova. We can go higher up the periodic table, for example, and see a lot more of the neutron-rich material ejected from the star, such as these nickel bullets. It’s the difference between seeing two or three colors or a whole chunk of the rainbow—and this rainbow is very big.”

Sandoval expects further simulations on Summit and the next generation of supercomputers—such as Frontier, soon to be the nation’s first exascale system at speeds exceeding 1 quintillion calculations per second—could reveal even more of that rainbow. His next study calls for modeling the collapse of a star nearly 26 times the size of the sun, a project expected to require more than 400,000 node hours on Summit.

“We hope to build more detailed models, which means we’ll have to expand our network to track even more bullets with even heavier metal isotopes,” Sandoval said. “Moving to exascale will help with that. These simulations help show these metals are ejected from stars, and they show what we may be made of. Ultimately it’s about the origin story of us and what we’re made of. We’re getting there. We’re closer than ever, but we need just a little more computer power.”

Related Publication: M.A. Sandoval, W.R. Hix, O.E.B. Messer, E.J. Lentz, and J.A. Harris, “Three-Dimensional Core-Collapse Supernova Simulations with 160 Isotopic Species Evolved to
Shock Breakout,” The Astrophysical Journal 921, No. 1 (2021),

This research was supported by the DOE’s Office of Science. UT-Battelle LLC manages Oak Ridge National Laboratory for DOE’s Office of Science, the single largest supporter of basic research in the physical sciences in the United States. DOE’s Office of Science is working to address some of the most pressing challenges of our time. For more information, visit