The nuclear reactions that fuel the sun could soon be harnessed to generate electricity on Earth - with help from supercomputers at the Department of Energy's Oak Ridge National Laboratory.
Type One Energy Group , a Knoxville-based startup, expects to build the world's most advanced stellarator fusion device by 2030, with a pilot power plant to follow that would produce commercial fusion energy by the mid-2030s. The Type One Energy team discussed their pilot plant concept in a series of six papers recently published in the Journal of Plasma Physics .
"We have a target for the pilot plant within the decade, and we expect to have a functional prototype much sooner," said Walter Guttenfelder, principal scientist for Type One Energy. "The scientific understanding at this point is mature enough to know there are no obvious showstoppers, and we wouldn't be this far along without the leadership computing machines at ORNL."
The prototype, Infinity One, won't produce electricity but would demonstrate the company's design works and clear the path for the pilot plant, Infinity Two, which would generate an effective output of 350 megawatts for the electric grid. The detailed modeling offered by ORNL's Summit supercomputer, which has since ceased operations, shaved at least a year off the time from drawing board to reality, the company estimates - maybe more.
"These sorts of high-fidelity performance projections have never been used before to design a fusion power plant," said Noah Mandell, a Type One Energy computational scientist. "The scale of Summit was absolutely necessary for these calculations."
The promise of fusion power
Fusion occurs when two atoms' nuclei combine to form a single nucleus. The difference in mass between the two fusing nuclei and the one resulting nucleus converts to raw energy. That type of energy powers the sun and stars, and its promise fuels hopes for a potentially unlimited power source for the world.
But first scientists and engineers must figure out how to make fusion work - safely, reliably and consistently at a mass scale.
A stellarator uses an intricate set of superconducting electromagnetic coils to confine a plasma made up of the hydrogen isotopes deuterium and tritium at temperatures 10 times hotter than the core of the sun . That's an average temperature of 270 million degrees Fahrenheit, or 150 million degrees Celsius.
"The sun's a little too massive for us to have economical fusion on Earth, so we've got to go much hotter to make that core small enough to fit here on our planet," Guttenfelder said. "These machines have been built at the laboratory level, so we know the overall concept can work. The current models just aren't big enough to produce energy on a commercial scale."
The world's largest stellarator, the Wendelstein 7-X at the Max Planck Institute for Plasma Physics near Munich in Germany, has a radius of just 5.5 meters, or a little more than 18 feet. Type One Energy engineers expect they'll need a device about twice that radius to make commercial fusion viable.
"We know a lot about the right way to get there, but we want to confirm it," Guttenfelder said. "We need the greatest accuracy possible because we're projecting to a size that's never been built. It's easy to talk in the abstract about orders of magnitude, but in engineering we need to know everything we can at a much smaller tolerance for error so we can reduce risks."
The trouble with nuclear turbulence
The main risk to solve? Turbulence - the unstable, chaotic flow of heat and mass within a plasma. Too much plasma turbulence, and the stellarator core could leak energy and fail to reach the necessary temperatures for fusion.
"Turbulence occurs everywhere, from the flow around airplane wings and cars to stirring cold cream into hot coffee ," Guttenfelder said. "A great comparison for our purpose is between the turbulence in our devices and the atmospheric turbulence around Earth. As the sun heats the equator more than the poles, that reaction drives airflows and turbulence that dictate the weather patterns from day to day. The same process happens inside the stellarator. We're losing heat from the core to the edge, and that's holding us back."
That problem traditionally prompts two expensive attempts at a solution: build a bigger machine or generate a stronger magnetic field to contain the plasma.
The stellarator's unique flexibility offers a third solution: optimize the stellarator's shape to keep turbulence under control.
"What if we could squish or expand Earth's equator?" Guttenfelder said. "That would disrupt and change the turbulent flow patterns in Earth's atmosphere. That's the principle that we needed to explore for the stellarator using modeling and simulation: Can we find an optimized 3D shape to disrupt those turbulent flows inside the stellarator and minimize that leaking of energy so we can sustain a really hot, high-efficiency fusion plasma?"
Crafting a stellarator concept
The computing power required to simulate that turbulence and predict the stellarator's performance far surpassed the capabilities of any in-house computer at Type One Energy. The team, which included Type One Energy computational physicist Guillaume Le Bars, turned to the Oak Ridge Leadership Computing Facility , home to ORNL's Summit supercomputer at the time.
Summit's high-resolution modeling capabilities helped the Type One Energy team develop the knowledge and tools to make their solution a reality and to confirm an optimized stellarator as an economically viable concept for a fusion power plant.
Summit's speeds of 200 petaflops, or 200 quadrillion calculations per second, offered an ideal match for the GX code developed by Mandell with colleagues at the University of Maryland and Princeton Plasma Physics Laboratory. The code, tailored especially for GPUs, solves nonlinear 5D equations that track the behavior of magnetized plasmas.
"For this study, we ran two families of calculations," Mandell said. "We ran large ensembles of turbulence simulations, searching for optimal 3D shapes that can keep turbulence under control. Once we found a few shapes we liked, we also ran large, coupled turbulence calculations to make high-fidelity performance predictions, from densities and temperatures of the plasma to the total fusion energy output of the power plant. No one's ever been able to use turbulence simulations of this kind at these scales to design a fusion device. Summit allowed us to do that."
The team, granted 250,000 node hours of simulation on Summit, used the results to pinpoint the most promising design, detailed in the Journal of Plasma Physics study.
"We're laser-focused on building our prototype and the pilot plant," Guttenfelder said. "At the same time, we think there are some areas that deserve further study, so we want that extra level of confidence. Thanks to Summit, we've been able to run thousands of evaluations to make our design choice and uncover new areas of interest that deserve further exploration. Summit's results gave us the confidence to continue to advance and accelerate our chosen design toward the finish line."
The team hopes to further refine the design by using Frontier - Summit's faster, more powerful successor at speeds of roughly 2 exaflops, or 2 quintillion calculations per second.
Support for this research came from the DOE Office of Science Advanced Scientific Computing Research program. The OLCF is a DOE Office of Science user facility.
UT-Battelle manages ORNL for DOE's Office of Science, the single largest supporter of basic research in the physical sciences in the United States. The Office of Science is working to address some of the most pressing challenges of our time. For more information, please visit energy.gov/science .
