Fusion, the nuclear reaction that powers the Sun and stars, has incredible potential as a safe, carbon-free and essentially limitless source of energy. But making it a practical and economic reality has tormented scientists since the 1950s. Today, fusion researchers are gearing up to harness the power of exascale computing to unravel the mysteries of what could be the source of ultimate renewable energy.
“To really understand what’s going on and what’s going to happen in the next experiment, you need big codes and big computers,” says Dr. Choongseok “CS” Chang, principal investigator of a multi-partnership center. American multidisciplinary institutional SciDAC for High-Fidelity Limit Plasma Simulation, headquartered at the Princeton Plasma Physics Laboratory, Princeton University.
Exascale supercomputers are exactly what current fusion research needs, says Chang. One of the biggest challenges today is to make accurate predictions about the processes that occur inside tokamak reactors, which use giant magnetic fields to confine the torus-shaped plasma fuel to achieve the necessary conditions to the merger. To advance this science, Chang’s team is preparing to use the Aurora Exascale supercomputer, the nation’s first Intel-based exascale HPC system to be deployed at the US Department of Energy’s Argonne National Laboratory ( DOE).
It is essential to be able to predict and control the disturbances which occur inside a tokomak and which could bring the ultra-hot plasma into contact with the wall of the reactor. Engineers have windows measured in milliseconds to control instabilities before the plasma bursts out of its magnetic confinement and potentially damages the reactor.
The walls of the reactor must be made of materials capable of withstanding the incredibly high heat and pressure of the plasma. Tungsten has the highest tensile strength of all pure metals, which is why it is incorporated into the tokamak of the International Thermonuclear Experimental Reactor (ITER), an international fusion research and engineering project in the South of France. Launched in 2013 and slated to have its first plasma in 2026 and begin full operation in 2035, ITER is the world’s largest tokamak and aims to prove that large-scale fusion power is possible.
Experiments at JET (Joint European Tokamak) have shown that the use of tungsten in the walls of the tokamak leads to weaker plasma confinement than expected. “It was totally unexpected,” Chang says, “so they’re really worried. If this were true, then ITER could struggle to produce 10 times more fusion energy than the input energy in its current design state.
But experiments have also shown that the injection of a very light material such as nitrogen or neon gas can restore containment levels. Nevertheless, even without injecting light impurity particles, researchers at the Joint European Torus (JET) fusion project in Oxfordshire, UK, were able to produce a breakthrough energy of 59 megajoules in a pulse of five seconds in December 2021, more than double the previous one. world record. “It was a historic event, enough to say that yes, the merger is actually practical,” Chang offers. However, the rate of energy efficiency was not yet at the level required by ITER because they did not inject light impurity particles into the edge of the tokamak.
Since JET uses the same wall material as ITER, Chang says “our first science on Aurora is to understand this tungsten wall experiment and how it will extrapolate to ITER. We need to do a high fidelity simulation based on first principles and fundamentally verify the physics. Their questions include why tungsten degrades fusion performance so much, why light impurity particles would bring performance back, and how best to incorporate them into reactor design.
The increased processing powers of exascale enable much more faithful scientific predictions and offer the possibility of training more specialized surrogate models that can be shared in real time with experimental facilities. “By using large-scale HPCs optimized for AI and ML, there can be daily communication and progress between exascale computers running large-scale simulations and large-scale experiments,” says Chang, in the comparing to the current process of trial and error which can take years. . “Aurora should have a maximum double accuracy of 2 exaFLOPs – that will be perfect.”
Chang’s team uses the XGC Gyrokinetic code, a modern particle-in-cell code designed and optimized for large-scale computers, especially GPU machines. It is highly scalable and open source for the US community. “It’s big code designed to take advantage of modern big HPC – when I see Aurora’s specs, I get excited,” he laughs. He hopes to have the code ready for Aurora in 2022 or early 2023.
Aurora will incorporate upcoming HPC and AI hardware and software innovations from Intel, including next-generation Intel Xeon Scalable processors (codenamed Sapphire Rapids HBM), and accelerated by future Intel Data Center GPUs (codenamed Ponte Vecchio) . It is based on the Slingshot 11 framework and the HPE Cray EX supercomputer platform. It will support ten petabytes of memory and will take advantage of Intel Distributed Asynchronous Object Storage (DAOS) technology, supported on Intel Optane Persistent Memory. “The oneAPI unified programming model will simplify development on various architectures,” Chang said.
Chang is confident that Aurora will shorten the time frame for commercial fusion power development, which he says has always seemed only a few decades away. In the meantime, he says, “there is still a lot of work to do.”
Julian Smith is an award-winning green tech, conservation, and travel writer based in Portland, Oregon, whose work has appeared in Wired, Smithsonian, New Scientist, and The Washington Post, among others.