Disruptions in tokamaks: from prediction and prevention to the assessment of loss of vacuum accident consequences

The continuous growth of the human population and the development of technologies lead to a persistent increase of the energy demand, which is only partially fulfilled by the renewables and CO2-free sources of energy, with a consequent rise of the CO2 and other pollutant concentrations in the atmosphere. The consequences of pollution are quite known: climate change, ocean acidification, effects on human health, etc. Developing sustainable and clean ways to produce energy is a key issue to build a greener and healthier world. In this field, nuclear fusion may play a key role. A nuclear fusion reactor is aimed at producing energy by fusing two light atoms, deuterium and tritium, into a heavier one, helium. The mass defect between the products and the reactants is directly converted in energy. The most studied and promising approach to realise nuclear fusion is the tokamak, a solution based on confining the plasma by a magnetic field. Despite the fact that the design is physically possible, nuclear fusion requires a lot of further studies and investigations to render nuclear fusion economically valuable on earth. In chapter 1 of this dissertation, a brief introduction to nuclear fusion reactors is given, describing the physical principles of the reaction, the magnetic confinement of the plasma and the main components of the tokamak. One of the events which threaten nuclear fusion feasibility is the disruption. Disruptions are events triggered by plasma instabilities which terminate rapidly the plasma, involving huge heat and electromechanical loads on the plasma-facing components and on the structure of the reactor. These loads, in future tokamaks, will be so large that even few disruptions may affect the life of the reactor. Techniques to mitigate, prevent or avoid disruptions have been developed but all of them require a real-time prediction of the disruption. In chapter 2, an algorithm for disruption prediction is shown and discussed. The algorithm has been tested on an ASDEX-Upgrade database and on a JET database. If a disruption happens, the heat loads and the electro-mechanical forces may critically damage the walls and accidents may occur. One of these accidents is the loss of vacuum accident, an event where the isolation barrier between the vacuum vessel of the tokamak and the external environment is lost. Thus, the external air flows inside the vessel, pressurising it. After the pressurisation, the walls of the reactor, still hot, warm up the air in the vessel, which increases in temperature and pressure, leading the internal air to flows outside. This air may drag outside some of the dangerous materials inside the vessel, such as tritium, Riccardo Rossi Abstract 11 causing a radiological event. In chapter 3, a scaling law to calculate the pressurisation time has been developed. The pressurisation time is important to understand how much time the operators have to take countermeasures in the case of loss of vacuum accident. During the normal operation of the tokamak, plasma-material interactions erode the plasma-facing components, producing a certain amount of dust. During a loss of vacuum accident, this dust may be resuspended, and it may be dragged outside. The dust is a risk for the environment and the human health since it may be radioactive and toxic. In chapter 4, experimental measurements of dust resuspension in case of LOVA have been performed in a scaled facility, STARDUST-Upgrade, developed by ENEA and upgraded by the University of Rome “Tor Vergata”, in order to investigate the dynamics of dust resuspension and mobilisation. Scaling the experimental results from a vessel like STARDUST-U to a tokamak-like device is experimentally impossible without additional measurements in tokamaks. On the other hand, making these experiments in situ would be unpractical and uneconomical. The scaling of the results may be performed by a numerical approach. In chapter 5, a preliminary multiphase computational fluid-dynamic has been tested and compared with the experimental results obtained in chapter 4.