High performance computing in magnetic fusion energy research

Obtaining energy from nuclear fusion requires heating the fuel to about 100 times the temperature of the sun to produce a plasma, and holding the hot plasma long enough for fusion reactions to produce a net energy gain. In a magnetic fusion device, this plasma is maintained in a largely self-organized state that far from equilibrium the mathematical description of which is characterized by high dimensionality, nonlinearity, extreme range of time and space scales, and sensitivity to geometric details. High-performance computing plays an essential role in fusion research not just to understand the theory and make quantitative comparison to experiments, but also to provide direct support to the experiments by interpreting measurements and designing experiments. The fusion community has begun a major new phase of research with ITER, an international prototype fusion energy reactor. This talk will review recent advances in the computational modeling of fusion plasmas and describe the challenges of simulating a reactor scale experiment such as ITER.