Gyrokinetics has been shown to be an appropriate model to simulate microturbulence in fusion plasma. The dimensionality of the phase-space grids, however, makes gyrokinetics computationally expensive already for the local flux-tube plasma simulations, where the radial direction range is small compared to the machine size and the plasma temperature radial variation can be neglected. In global simulations, nevertheless, the radial range has to be extended to the full machine size in order to capture global effects. The corresponding computational domain spans now over areas with significantly different temperatures. This leads to different plasma properties in different radial positions and, thus, other computational grids in the velocity directions are required. The temperature variations are reflected in the background distribution function, which appears in the numerically-solved gyrokinetic equations. We develop computational grids adjusted to the background distribution function, thus enabling faster and more stable gyrokinetic plasma simulations on massively parallel machines.     «

Gyrokinetics has been shown to be an appropriate model to simulate microturbulence in fusion plasma. The dimensionality of the phase-space grids, however, makes gyrokinetics computationally expensive already for the local flux-tube plasma simulations, where the radial direction range is small compared to the machine size and the plasma temperature radial variation can be neglected. In global simulations, nevertheless, the radial range has to be extended to the full machine size in order to captu...     »