Predictive Simulations May Point to the Best Design for Fusion Reactors

By Scott Gibson

Fusion energy, which harnesses a controlled thermonuclear fusion reaction, could one day transform the way we humans power our activities. This source will be clean, inexpensive, and nearly unlimited, with sea water supplying its basic fuel.

The largest fusion reactor under construction, the International Tokamak Experimental Reactor (ITER) in France, will attempt to produce ten times the power required to run the device. Magnetically confined fusion plasmas are being designed within ITER and other projects that will operate in physics regimes never achieved through experiment. Accordingly, modeling and simulation activities that require the power of exascale computers, or the next generation of supercomputers, are necessary to design and optimize these new facilities.

A US Department of Energy Exascale Computing Project (ECP) effort called WDMApp aims to produce a whole-device model of the fusion reactor as a means of supporting all the plasma processes that go on in the fusion device, from the core to the boundary. It will also provide predictive numerical simulations of the physics required for magnetically confined fusion plasmas to enable design optimization and fill in the gaps for ITER and future fusion devices.

In past years, and more recently, the WDMApp project research team has run each of its codes separately on petascale computers such as the Titan system at the Oak Ridge Leadership Computing Facility.

“This has been a very productive experience for each of these codes,” said WDMApp principal investigator Amitava Bhattacharjee of the Princeton Plasma Physics Laboratory. “We have used these codes to the utmost of their capabilities in petascale plasmas, but we have fallen short, because even with a machine as powerful as Titan, it is not possible to do the whole tokamak fusion reactor for a plasma the size of ITER within reasonable time constraints on the petascale computer. A pre-exascale computer such as Summit takes us much further, within a factor of 5 of the Aurora, which is the exascale computer that we expect to have at Argonne in a couple of years. And, already, the fruits of our labor are paying off.”

He explained that during a typical run of the project’s XGC code on Titan that would take between two weeks and a month, can be done within a couple of days on Summit and in a few hours on Aurora. The Aurora exascale machine is expected to arrive in 2021 at Argonne National Laboratory.

“When we have access to an exascale computer, the impact on our field of fusion simulation will be transformational,” Bhattacharjee said. “There are things that we could not even dream of attempting, even in the era of petascale, that now appear to be within reach. For example, predictive simulations may enable us to choose between several very expensive designs to build an optimum fusion reactor.”

Exascale will allow the addition of new capabilities to the whole-device model, including effects of the fusion boundary, effects of the fusion products such as alpha particles, the influence of sources of heating through radio frequency waves or through neutral beams, and the inclusion of the superimposed engineering structure that would make a fusion reactor operate as a unit.

“All of this now becomes within the realm of possibility within the next 10 years or so,” Bhattacharjee said.