Extended-Scale Tokamak Pedestal Turbulence
Max Planck Institute for Plasma Physics, Garching (Germany)
Local Project ID:
HPC Platform used:
SuperMUC and SuperMUC-NG of LRZ
Turbulence is a phenomenon deeply rooted in our daily lives, and yet we rarely stop to think about it, unless we are sitting in a plane that gets rattled by the wind. Even to physicists, turbulence is often a complication that is better avoided if possible.
In fusion research, turbulence is one of the main factors that has challenged scientists for many decades. The highly irregular motion of particles in a turbulent environment is ultimately the reason why an energy-producing fusion reactor cannot be as small as shown in some popular movies, but instead can only be realized on the scale of a multi-national, multi-billion Euro endeavor that requires decades of planning. The flagship project of fusion research, ITER, is such an experiment, and is being constructed in the South of France overseen by an international consortium.
In the project "Extended-Scale Tokamak Pedestal Turbulence", Dr. Daniel Told from the Max-Planck Institute for Plasma Physics led a group of researchers that worked towards unraveling the properties of turbulence in a so-called "tokamak" (from Russian "ring-shaped chamber with a magnetic field") reactor, such as ITER or ASDEX Upgrade (shown in Fig. 1)
In a tokamak, researchers heat an ionized gas (plasma) to very high temperature, about 100 million degrees, to achieve the conditions for nuclear fusion reactions between the plasma particles. Using the hydrogen isotopes deuterium and tritium, this reaction releases energy that can then be used for electricity production. Unfortunately, the large difference in pressure between the center of plasma and the tokamak wall creates strong turbulence, which in turn leads to a fast loss of heat and particles from the plasma center, making it hard to maintain the conditions necessary for fusion reactions.
This is where supercomputing comes into play: Due to the complexities in plasma physics, it is not sufficient to simulate plasma turbulence in three-dimensional space. Not only the particle positions, but also their velocities need to be kept, so that ultimately a five-dimensional simulation is needed, using a computational grid consisting of hundreds of billion points. Such simulations have only become possible using the latest developments both on the hardware and on the software side.
Within this project, the team surrounding Dr. Told applied high-performance computing resources provided by the SuperMUC computer at the Leibniz Rechenzentrum Garching in order to achieve realistic simulations of plasma turbulence in the edge region of the tokamak plasma. Understanding this region is particularly critical, as it can develop a "transport barrier", in which the detrimental turbulence is suppressed. The wider a barrier is achieved, the better the overall confinement quality of the plasma, and the more energy production can be achieved in the fusion reactor
Leveraging the resources of SuperMUC, the researchers were able to push the simulations of this region to a new level. Improvements to their simulation code GENE [1,2] that involved the use of newly developed block-structured grids  allowed them to perform simulations that were previously not feasible.
In particular, the researchers were able to perform global simulations that encompassed the whole transport barrier region for several different experimental data sets, and obtained good agreement with the transport levels found in the experiments. Such modeling is an important cornerstone on the path towards predictive modeling that can then inform the design and optimization of future experiments.
The temperature of the plasma can change by a factor of 5 or more across the width of the transport barrier region (which is only a few millimeters). Accordingly, the character of the turbulence also changes, and sharp layers of sheared plasma flows can develop that serve to break up the turbulent structures, leading to its suppression. At the same time, this places strong resolution requirements on the simulation: The seemingly narrow simulation domain shown in the figure actually was resolved by more than 1,000 grid points across. Despite the use of block-structured grids, together with the remaining dimensions that also need to be kept, these simulations still used several ten billion grid points, requiring top-tier supercomputers as provided by the Gauss Centre for Supercomputing.
 F. Jenko et al., Physics of Plasmas 7, 1904 (2000).
 T. Görler et al., Journal of Computational Physics 230, 7053 (2011).
 D. Jarema et al., Computer Physics Communications 215, 49, (2017).
Dr. Daniel Told
Division Tokamak Theory
Max Planck Institute for Plasma Physics
Boltzmannstr. 2, D-85748 Garching (Germany)
e-mail: daniel.told [@] ipp.mpg.de
LRZ project ID: pr27fe