Low-rank optimization of strongly magnetized fusion plasma

Nuclear fusion is a leading contender for producing clean and abundant energy. It holds the promise to combine many of the advantages of traditional technologies (large power density, no intermittency, scalability, suitability for baseload power, etc.) with that of renewable but often intermittent technologies (no greenhouse gas emission, virtually unlimited fuel supply, safety). The most advanced designs to make fusion viable use strong magnetic fields to enclose a plasma (as exemplified by the International Thermonuclear Experimental Reactor, ITER, currently under construction in southern France). If fusion is going to be a viable source of power, the design and operation of reactors will have to be optimized and new insights will have to be gained using computer simulation, similar to how this is now routinely done in building aircraft, designing industrial plants, etc. However, this is extremely difficult at present as the most accurate (i.e. kinetic) models for strongly magnetized plasma systems are extremely expensive. This makes them difficult to solve even on the world`s largest supercomputer. In this project we are developing low-rank based complexity reduction and optimization algorithms that drastically reduce the computational cost required. This will, for example, enable rapid exploration of different designs of fusion reactors and thus has the potential to drastically accelerate progress in the field.