Supervisor/s: Francis Casson (UKAEA, lead), David Dickinson (University of York).
Plasma turbulence acts to degrade the confinement of particles and heat in magnetic confinement devices, such as tokamaks. Worse confinement leads to the need for larger, more expensive, reactors reducing the efficiency and economic viability of fusion produced electricity. Plasma turbulence is also a fascinating area of physics, rich in interesting and complex behaviour. As such, there is significant research activity around the theoretical, computational and experimental study of plasma turbulence.
The heat and particle transport in spherical tokamaks (STs), such as the UK’s MAST-U, is qualitatively different to that in conventional aspect ratio tokamaks such as JET (where several reduced complexity models have been developed which do a good job of predicting confinement). In STs, the electrostatic microturbulence is often suppressed by the geometry, allowing access to high pressure. This leads to electromagnetic microturbulence which is less well understood and for which accurate reduced models are not yet available. Reduced models are invaluable in predicting confinement since first principles approaches, such as high fidelity gyrokinetics, are generally too expensive to evaluate over confinement timescales.
This project has the potential to go through full cycle of model development and improvement through experimental validation. Nonlinear gyrokinetic simulations can be used to characterize the transport from the highest fidelity (but computationally expensive models). These simulations can be used to build reduced quasilinear models based on a combination of theory assumptions and machine learning. When the reduced models are integrated into a transport solver they can be used to compare global confinement predictions against MAST-U experiments. Any regions of parameter space where the integrated predictions do not agree with experiment can then be used to motivate subsequent cycles of model improvement. Comparison with experiment is vital to build confidence in such models and their extrapolations to future ST powerplants such as STEP.
The student will have opportunities to gain experience in high performance computing, the application of machine learning techniques to develop reduced models and experimental validation, as well as opportunities to develop skills in scientific communication.
The project will be mainly based in Culham Science Centre but may also involve travel to support collaborations. There are also opportunities for travel to conferences.
This project may be compatible with part time study, please contact the project supervisors if you are interested in exploring this.