Statistical and Information Analysis of Plasma Diagnostics

Abstract

Model comparison using Bayesian evidence, an analysis technique not previously used in plasma physics, is demonstrated on data from two experimental fusion devices. In the first device, a Polywell at the University of Sydney, Langmuir probe data is analysed to compare models describing its electron velocity distribution function (EVDF). The best performing model of the EVDF across all datasets is found, and its posterior distribution is used to give statistical distributions of plasma parameters, such as the plasma potential and density. Compared across the different data sets, these parameters indicate the successful formation of an electric potential well, crucial to the operation of the Polywell. Intensity profiles collected by an infrared camera viewing the divertor strike plate of MAST, previously run at the Culham Centre for Fusion Energy, are analysed to compare models describing the scrape-off layer (SOL). One of the previously existing models in the literature, the Eich function, performs best overall according to the Bayesian evidence, but one of the new convection-diffusion models matches its performance in several of the frames, indicating that improvement is possible. A second analysis technique, based on the Fisher information, is developed to aid the design of discharge plasma experiments with spectroscopic diagnostics. The mathematical basis of this technique is completely novel, having been derived specifically for use in this problem. The analysis technique quantifies the sensitivity of a spectral line to changes in the electron energy distribution function (EEDF). This technique is recommended for those designing plasma discharge experiments where a particular deviation in the EEDF is to be determined from spectral data. It allows all lines in a potential spectrum to be compared and a small subset chosen that will still strongly indicate the deviation. A demonstration is given using a collisional-radiative model of an argon plasma.