Automatic and statistically robust spatio-temporal detection and tracking of fusion plasma turbulent fronts

The controlled production of thermo-nuclear fusion energy is critical for providing an alternative, environmentally friendly, and renewable energy source on our planet. The technical challenge is to stabilize the dynamic turbulent flow of hot plasma in magnetic fields in a fusion energy reactor. More specifically, the issue is how to control fusion plasma instabilities, or turbulence–an analogy comparable to learning how to “Hold the Sun.” To address this issue, we create computational and mathematical methodology to discover and track–both in space and time–the intricate patterns of dynamic plasma turbulence during fusion energy production in a reactor simulated on a supercomputer. In doing so, we establish several aims. First, from extreme scale data, we automatically discover the structure of fronts, or patterns where the turbulence starts, and analyze their dynamic behaviour including speed and direction of front propagation. To do this, we create an algorithm for automatic turbulent front detection, track and quantify front propagation in space and time, and assess the predictability of the proposed methodology. Comprehensively, this process can potentially predict the structure, dynamics, and function of fusion plasma turbulence. It could also enable similar analyses required in other fields, such as astrophysics and combustion.