Computational challenges in magnetic-confinement fusion physics

Magnetic-fusion plasmas are complex self-organized systems with an extremely wide range of spatial and temporal scales, from the electron-orbit scales (similar to 10(-11) s, similar to 10(-5) m) to the diffusion time of electrical current through the plasma (similar to 10(2) s) and the distance along the magnetic field between two solid surfaces in the region that determines the plasma-wall interactions (similar to 100 m). The description of the individual phenomena and of the nonlinear coupling between them involves a hierarchy of models, which, when applied to realistic configurations, require the most advanced numerical techniques and algorithms and the use of state-of-the-art high-performance computers. The common thread of such models resides in the fact that the plasma components are at the same time sources of electromagnetic fields, via the charge and current densities that they generate, and subject to the action of electromagnetic fields. This leads to a wide variety of plasma modes of oscillations that resonate with the particle or fluid motion and makes the plasma dynamics much richer than that of conventional, neutral fluids.