Fully kinetic simulations of magnetic reconnection and tearing mode instability in collisionless low-beta plasmas

Magnetic nuclear fusion (especially plasma physics) is one of the most leading-edge areas of nuclear energy research. Experiment and simulation are the mainstream methods in those research. On the one hand, the ITER, a new big science experiment, is approaching its final stages of construction and was expected to become operative in 2025. Meanwhile, smaller experiments are also being pursued to investigate specific issues and innovative approaches, as for example the VEST reactor in Korea that will be one of our focuses. On the other hand, simulation methods play a pivotal role in the plasma physics research, especially considering the high cost of conducting nuclear fusion experiments. Simulation methods are of course developed in synergy with experiments, which provide input for simulations, which in turn help their design and interpretation. 6D simulations provide a statistical description of the population of particles in this 6D space. The 6D approach is based on the probability distribution function that measures the probability of finding a particle with a given velocity at a given position. This model is also called kinetic and is based on the Boltzmann Equation. This description is considered first principle because it includes all known physics relevant to the evolution of the device. Now it is believed that the progress in supercomputing has reached a level where the full 6D description of fusion devices is becoming possible. One of the processes receiving the most concentrated attention is tokamak disruption. In a disruption, the plasma suddenly loses its confinement and releases its energy to the surrounding structures. In past tokamaks disruptions were a severe annoyance, resulting in experimental delays and sometimes requiring costly repairs. Consequently, effective measures must be taken to prevent this from happening or mitigate its influence. There are active controls that can be deployed when a disruption is about to start but the response system requires a predictive algorithm that can forecast with accuracy the possibility of a disruption happening with sufficient advance notice that countermeasures can be deployed. For the safe operation of ITER, this approach then requires that we understand disruptions better and that we have better algorithms to make a forecast of disruption onset. This is the goal of our project: we want to develop methods to study tokamak disruptions from first principle simulations and we want to develop machine learning (ML) approaches that using the insight from our simulations will be more effective in predicting the onset of a disruption.