Bringing Fusion Power Closer with High-Fidelity Tokamak Plasma Predictions
PI: Amitava Bhattacharjee
Princeton Plasma Physics Laboratory
In 2016, the US Department of Energy’s (DOE’s) Exascale Computing Project (ECP) set out to develop advanced software for the arrival of exascale-class supercomputers capable of a quintillion (10¹⁸) or more calculations per second. That meant rethinking, reinventing, and optimizing dozens of scientific applications and software tools to leverage exascale’s thousand-fold increase in computing power. That time has arrived as the first DOE exascale supercomputer—the Oak Ridge Leadership Computing Facility’s (OLCF’s) Frontier—opens to users around the world. “Exascale’s New Frontier” explores the applications and software technology for driving scientific discoveries in the exascale era.
The Science Challenge
Fusion power—generated when two nuclei combine to form a new nucleus, thereby releasing energy—holds many potential advantages as a source of electricity: plentiful fuels, zero carbon emissions, high reliability and efficiency, and none of the risks associated with fission power, such as meltdowns or long-lived radioactive waste. So, what’s holding us back from using fusion reactors now? Scientists are still learning how to create sustainable, controlled thermonuclear fusion reactions. One of the primary research facilities working to solve this problem is ITER, which is destined to be the world’s largest fusion reactor once its construction is completed in southern France in 2025.
The key to producing successful fusion reactions at ITER is the design of its tokamak, a machine that uses massive magnetic coils around a donut-shaped chamber to shape and control charged plasma particles formed from hydrogen fuel. To obtain sustainable energy confinement, which enables the performance of a fusion reactor, researchers need to predict the kinetic turbulence of the plasma in their tokamak designs.
To help understand how to run ITER for optimal efficiency and safety, the ECP’s WDMApp (Whole Device Model Application) effort will provide the most complete models so far of gyrokinetic turbulence within tokamaks. For the first time ever, WDMApp combines two advanced gyrokinetic codes: XGC (a particle-in-cell code optimized for modeling the edge plasma) with either GENE (a continuum code) or GEM (a particle-in-cell code), both of which are optimized for the core plasma. Based on first principles—the fundamental kinetic equations that govern plasma turbulence—WDMApp’s simulations should also be the most accurate of any so far.
“What is important about our project is that, for the first time, ITER researchers will have predictions that couple the core and the edge rather than treating these two regions of the plasma as separate units, each with its own prediction,” said Amitava Bhattacharjee, leader of the WDMApp project and a professor of Astrophysical Sciences at Princeton University. “To do that correctly requires the power of exascale computers and clever software engineering. It is the exascale that enables these calculations to be done now in a matter of days when it took months and years before. Without exascale, we cannot have a simulation of high-enough fidelity and of large-enough size to be predictive for ITER.”
WDMApp’s goal for coupled-code performance was to be at least 50× faster than an analogous code run on the Oak Ridge National Laboratory’s (ORNL’s) Titan, which was the fastest supercomputer in the world when the project started in 2016. WDMApp’s runs on ORNL’s Frontier—the current fastest supercomputer in the world—significantly exceeded that goal with a coupled-code performance of 301× faster than Titan (with XGC-GEM coupling).
The WDMApp team is planning its hero run on Frontier to make its first plasma predictions for the ITER tokamak.
“I think our result on the ITER plasma will be a first-of-its-kind. And beyond the lifetime of the project, which concludes in 2023, I expect there will be even further applications that will build on the foundation of WDMApp to produce results of even greater scientific impact,” Bhattacharjee said.
Furthermore, he foresees WDMApp becoming an important tool in the coming fusion-power industry, guiding designers to decide which reactors to build based on WDMApp’s predictions.
“I don’t think that industrial design of reactors can be done without knowing how the plasma, which sits at the heart of the machine, really performs. And these designs must be tested with predictive first-principles simulations of the kind that we are carrying out,” Bhattacharjee said.
Support for this research came from the ECP, a collaborative effort of the DOE Office of Science and the National Nuclear Security Administration, and from the DOE Office of Science’s Advanced Scientific Computing Research program. The OLCF is a DOE Office of Science user facility.
UT-Battelle LLC manages ORNL 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 https://energy.gov/science.