Aurora supercomputer harnesses AI to stabilize fusion plasma

The Aurora supercomputer at the U.S. Department of Energy’s Argonne National Laboratory is being leveraged to tackle the complex challenge of fusion energy by simulating plasma behavior inside tokamaks. As one of the world’s fastest exascale machines, capable of performing a quintillion calculations per second, Aurora enables researchers to model the extreme physics of plasma confined by magnetic fields at temperatures exceeding those of the sun’s core. This modeling is critical because plasma instabilities, such as magnetic islands and disruptions, can abruptly halt fusion reactions or damage reactor components. Aurora’s advanced computational power allows scientists to solve multi-dimensional equations representing trillions of particles, improving understanding of plasma flow and edge physics relevant to ITER, the international fusion project aiming to achieve 150 million degrees Celsius. In addition to simulations, Aurora is instrumental in training artificial intelligence models that predict imminent tokamak disruptions. Using extensive historical experimental data from facilities like DIII-D and the Joint European Torus, AI systems developed on Aurora can assign disruption risk