Probabilistic modeling of simulated plasma turbulence

• Promotors: Prof. Geert Verdoolaege, Prof. Peter Beyer (Aix-Marseille Université)
• Mentor: David Garrido González
• Study programs: Master of Science in Engineering Physics, Master of Science in Physics and Astronomy, Master of Science in Teaching in Science and Technology (Physics and Astronomy), European Master of Science in Nuclear Fusion and Engineering Physics
• Location: Technicum, at home and with (funded) visits to Aix-Marseille Université

One of the key challenges in the development of fusion energy based on magnetic confinement relates to understanding and controlling the transport of heat and particles, both in the core and edge region of the hot fusion plasma. This is because the transport processes are very complex, involving plasma turbulence that is both difficult to diagnose and model. Reduced order models (ROMs) allow approximating high-fidelity simulations at a manageable computational cost. Validation of the ROM approach includes matching probability distributions of fluctuating quantities like density and electrical potential. This requires a proper metric in spaces of probability distributions.

In this thesis project, you will use the methods of information geometry to assess the matching of ROM with higher-fidelity models, under varying conditions of the simulation. This will involve first defining a suitable probabilistic model, starting with univariate distributions, but later also covering multivariate models. Next, distributions will be compared on their Riemannian probabilistic manifold by calculating similarities based on geodesics between the distributions. Trends of distributional properties may be characterized by regression analysis on the manifold, while projections using distance-based methods could be used to visually explore structure in the simulated data. This will help to optimize the ROM in reproducing characteristics of drift-wave turbulence over a wide operational domain.