As part of a joint project, the EPFL Center for Imaging (IMAGING), the Swiss Plasa Center (SPC) and the Atomic Energy Center (CEA) in France are looking for a motivated candidate to perform her/his Ph.D. in a stimulating and dynamic atmosphere. The project takes place at the forefront of scientific research in computerised tomography for fusion plasma diagnostics in modern and next-generation tokamaks such as TCV, WEST and ITER. The thesis will be hosted by IMAGING and SPC at the Swiss Federal Institute of Technology in Lausanne (EPFL), a world-renowned science and education centre that offers students an ideal environment to continue their scientific carrier, together with an excellent connection with industry. Regular trips and/or prolonged visits to the CEA and ITER in France for testing/deploying algorithmic solutions are expected.
Thermonuclear fusion promises a clean and almost unlimited source of energy. Fusion reactors currently plan to employ tokamak configurations to magnetically confine a very hot fusion plasma in a toroidal volume. The performance and stability of the fusion reaction can be assessed by means of computerised tomography (CT), a common plasma diagnostic tool.
CT notably monitors the plasma confinement and temperature, to prevent damage to the vacuum vessel walls interfacing with the plasma, and to detect the impurities radiated by the plasma upon the facing components of the tokamak chamber.
Tomographs used in tokamak are often composed of up to a few hundred detectors, yielding somewhat sparse tomographic measurements. This results in poor reconstructions with strong “bleeding” artefacts along the detector line-of-sights. Current artefact-mitigation strategies lack robustness and do not allow for uncertainty quantification. This complicates comparison and interoperability between diagnostics, and reduces the accuracy of broadband multichannel tomographic reconstructions and, consequently, physics understanding.
The goal of this PhD thesis is to develop a suite of novel tomographic reconstruction algorithms for broadband multichannel plasma probing of fusion reactors. These algorithms will take advantage of, and develop upon, recent advances in the fields of variational inverse problems, convex/non-convex optimisation, signal processing, sparse/deep/Bayesian learning, together with numerical and high-performance computing, with the goal of achieving unprecedented reconstruction accuracy, speed, uncertainty quantification and robustness. Together with designing and testing these algorithms, the PhD student will also be expected to implement, package, validate and deploy them within the production tomographic pipelines specifically for TCV, WEST and ITER tokamaks. European research in this field is managed by the EuroFusion organisation where this research may be expected to propagate to other fusionresearch devices.
The successful candidate is expected to hold a MSc degree Applied Mathematics, Physics, Computational Science Engineering, Electrical Engineering, or a closely related discipline. In addition, the following qualifications are desired:
• Solid background in mathematics, optimisation, machine learning and/or signal processing. A previous experience with inverse problems, computational imaging or image processing is a bonus.
• Solid programming skills in Python, and good knowledge of the Python scientific data stack (Numpy, Scipy).
• Good background in physics and an interest in the field. Previous experience with plasma/nuclear fusion and/or experimental physics would be a plus.
• Practical skills (experimental validation, automating measurement setups).
• Familiarity with various open environments (eg: Jupyter notebooks, Github).
• Excellent communications skills (written and oral) and a team player attitude.
• Strong self-direction.
The mathematical emphasis of these requirements reflects a strong algebraic and numerical need but the goal is highly practical using vast amounts of data with a real goal of providing statistically meaningful and physics rich results.
The selected PhD student would need to enrol in the Physics program of the EPFL doctoral school. After one year of successful probation, the initial contract will be extended up to a total of four years. Doctoral school information and employment conditions at EPFL are described in:
We offer a stimulating, collaborative, cross-disciplinary research environment in collaboration with world-class research institutions at the forefront of imaging and nuclear science. The PhD candidate will engage, on a daily basis, with a dynamic and creative community of scientists,
eager to share their expertise and knowledge, will be exposed to a large variety of state-of-the-art computational imaging methods, and given the opportunity to deploy her/his algorithms in the tomography diagnostic pipelines of operational tokamaks. Finally, the EPFL is equipped with advanced imaging and strong experimental nuclear facilities that provide access to the latest cutting-edge technologies and a continual source of challenging data.
For inquiries and applications, please contact Dr. Matthieu Simeoni (email: matthieu.simeon[email protected]), principal scientist and head of the EPFL Advanced Image Reconstruction Hub.
Applications should include CV, grade transcript and names of 3 persons who may be contacted for reference letters.