Magnetic Fusion

Supervisor:            Dr Yasmin Andrew

Type:                        Experimental, theoretical and computational blend    

Funding:                 Eligible for FOSTER industry partnership funding

This proposed PhD project aims to further the fundamental understanding of confinement regime transitions in tokamaks, by applying statistical methods and machine learning to experimental data. Specifically, this PhD will build off established research at Imperial into correlations between confinement transitions and the probability density functions of key variables related to plasma turbulence and zonal flows. This includes fluctuations in electron density and edge plasma turbulence velocity. The fundamental methods used to characterise the shape and evolution of these probability density functions falls under the field of information geometry, a mathematical field that is highly novel in fusion. Building upon this work, this proposed PhD shall combine this analysis of experimental probability distributions into a machine learning model, developed under the supervision of digiLab. The aim of this project is for these models to predict the onset of a transition, which could be L-H, H-L, or between different types of H-mode (dithering ELM, ELM free etc). The predictions of a gaussian process regression model created by digiLab to predict turbulence characteristics for spherical tokamak power-plants. The machine learning models developed for this project should be inherently probabilistic in nature, such that we can build an understanding of model uncertainty and how to improve it.     

Supervisor:  Dr Yasmin Andrew

Title:             The role of edge and SOL plasma profiles in the H-mode density limit

Type:             Theory and experiment blend

Funding:       UKAEA studentship agreement

Future burning plasma devices need to operate at high density to increase the fusion power and improve power handling capability. However, the high confinement plasma (H-mode) cannot always be sustained at high pedestal density. This so-called H-mode density limit (HDL) limits high fusion performance in future devices. By using pellet fuelling to elevate the central density while keeping the edge density low, it is possible to exceed the HDL. Experiments show that the divertor configuration and wall material have a direct impact on the HDL, indicating that the HDL is sensitive to the physics of the boundary plasma (edge, scrape-off layer (SOL) and divertor regions).

This project aims to investigate the synergistic roles of edge plasma self-regulating dynamics and SOL broadening in determining the HDL, focusing on the physics of collisional boundary plasmas. Previous experiments on AUG and JET have shown that the SOL profiles broaden as the density approaches the HDL. Under certain condition, the SOL region may develop a flat density profile, known as a 'density shoulder'. The plasma can remain in H-mode for a while following the density shoulder formation, but once the temperature profile in the SOL broadens, it soon transitions back to L-mode. The power requirements and turbulence dynamics of the erosion or collapse of the external transport barrier in the lead-up to or at the HDL has been shown to be sensitive to the magnetic equilibrium and divertor configuration.

The project will utilize an extensive database from previous JET campaigns, complemented by new experiments on the MAST-U device. By leveraging the broad parameter ranges on JET, and advanced diagnostics on MAST-U, the student will investigate the conditions and mechanisms leading to SOL broadening and its impact on the HDL, aiming to develop current theoretical understanding and optimizing the performance of next-generation fusion experiments.