National Center for Computational Sciences

Nuclear fusion—the process that powers the sun—promises in the future to provide a significant portion of our own energy as well. First, however, researchers must determine how to use a magnetic field to control an ionized gas far hotter than the sun. A team led by Jeff Candy of General Atomics will use the computing power of ORNL’s petascale Jaguar supercomputer to model the turbulence of this plasma inside a doughnut-shaped tokamak reactor. Using a new simulation code called TGYRO, the team will calculate from first principles the temperature and density profiles to be expected in reactor plasmas and look for plasma instabilities. Candy is the lead developer of GYRO, TGYRO’s predecessor, a global gyrokinetic software code that last year successfully modeled 674 million coordinate points to simulate the transport of plasma inside a tokamak.

In the current simulation, the team will use 2 million to 4 million processor hours—and 40 times as many processing cores as GYRO—to solve Maxwell’s equations, which will describe the interaction between the turbulently fluctuating electromagnetic fields and the distribution of plasma ions and electrons. On Jaguar, the new TGYRO “supercode” will manage multiple instances of the massively parallel GYRO code to obtain global, steady-state plasma temperatures and densities. The result will give researchers realistic performance predictions for ITER, the international fusion reactor being built at Cadarache, in the south of France.