Advanced techniques for plasma radiation measurement in nuclear fusion reactors

In an era when the increasing demand for energy intersects with environmental consciousness, the exploration of clean and efficient energy sources becomes paramount. Nuclear fusion has emerged as a promising candidate to meet global energy needs, holding revolutionary potential in the landscape of energy resources. The core of nuclear fusion, a process that fuels the stars, represents an inexhaustible reservoir of energy, offering the prospect of power generation without the environmental impacts associated with traditional sources. However, building a functional fusion reactor is a challenging task, and the journey to achieve this goal has been marked by scientific, engineering, and financial challenges. To extract energy from the fusion of two nuclei is no straightforward task, primarily due to the strong e Coulomb repulsion between charged particles they contain. Accelerating particles to very high energies is necessary to achieve a reasonable probability of the fusion reaction occurring. This is achieved by heating a gas to enormous temperatures, transitioning it into a plasma state. However, for two particles to undergo fusion, they must be sufficiently close and remain so for an adequate interval of time, necessitating a method to confine the plasma. Various methods exist for plasma confinement, with magnetic confinement being one of the most extensively studied. This method is based on the principle that charged particles constituting the plasma are diverted by magnetic fields. The current leading configuration is the tokamak. Despite the construction of numerous experimental tokamaks over the past 70 years, achieving a controlled reaction operating in a continuous regime remains far in the future. A major challenge lies in maintaining plasma confinement, requiring extensive studies to enhance performance. To advance plasma physics understanding and optimize reaction confinement, measuring plasma parameters is crucial. Hence, diagnostics play a fundamental role in both real-time machine control and offline data analysis. Power emission emerges as a focal measure to determine plasma health since it represents a loss of energy, essential for establishing power balances within the reactor. Additionally, many proposed scenarios to improve confinement, such as detachment or seeding, are associated with radiative phenomena. Controlling these phenomena is crucial for future reactor development. Understanding where radiation is emitted from and its movement within the plasma is fundamental for studying transport phenomena. Determining the dynamics of ELMs, for instance, can significantly enhance access to advanced confinement regimes. Detecting anomalies in emission locations within the cross-section can contribute to the development of advanced strategies to avoid disruptions. However, understanding how radiation is distributed within the cross-section is not straightforward. This is due to the fact that radiation measurements are line-integrated, with detectors measuring the entire emitted power within their line of sight. Thus, to obtain local information, a tomographic inversion is necessary. In this thesis work, two new approaches have been developed for tomographic inversion to assess the spatial distribution of plasma emissivity within a reactor. These approaches serve for both real-time control and enhanced offline analysis. The developed approaches are based on the Maximum Likelihood Expectation Maximization algorithm, incorporating an interupdating operation. The first approach implements Gaussian smoothing, while the second the adoption of anisotropic diffusion with an adaptive algorithm. One significant advantage of utilizing Maximum Likelihood is its capability to estimate how measurement errors propagate onto the reconstruction. This knowledge is essential for improving the measurement quality. However, the direct demonstration of error propagation using an inter-updating approach had not been previously established. In this work, it has been demonstrated how to determine error propagation using both the inter-updating approach with Gaussian smoothing and anisotropic diffusion. Based on these new algorithms, a tool has been developed to test the design of new systems, enabling the selection of optimal geometries for line-of-sight measurements. In the first chapter of this thesis, nuclear fusion and the principles of magnetic confinement will be introduced. The second chapter will explain the role of radiation measurements, how they are acquired, and the main radiative phenomena occurring in the plasma. Methods for obtaining tomographic reconstruction from these measurements will then be explained. The third chapter will introduce Maximum Likelihood and the developments obtained for it. The fourth chapter, through the use of synthetic cases of major radiative phenomena, will verify the effectiveness of the developed methods for both accurately reconstructing and correctly propagating errors. The validity of error propagation will be demonstrated through a Monte Carlo approach within this chapter. In the fifth chapter, the application of these techniques to analyse experimental data from JET will be provided, analysing and illustrating different radiative cases. The sixth chapter will explain how this approach can improve the procedure for designing a tomographic diagnostic for fusion reactors.