LEarning Gyro-fluid clOsure (LEGO)

Nuclear Fusion represents a virtually clean and unlimited energy source of the future, where net energy is to be produced by fusing charged particles. In order for this to happen, these charged particles are heated to temperatures that exceed those of the Sun and are confined by extremely strong magnetic fields that are 1000 times greater than that the Earths magnetic field. These magnetic fields force the particles to follow circular paths along the magnetic fields. An outstanding problem of nuclear fusion and the design of future fusion reactors is the accurate prediction and reduction of the heat flow to the adjacent reactor walls. This heat flow poses extreme challenges for wall materials, even exceeding that of rocket launches. The current state of the art mathematical models utilize the gyro-motion of the particles for an efficient description (gyro-kinetics) of the turbulent dynamics . However, these models reach their limits when predicting the heat flow in the most demanding calculations on the world`s fastest high performance computing machines. In the LEGO project (Learning gyro-fluid closure), exact relations (closure) of the heat flow are to be found, which are the key to a so called fluid description (gyro-fluid fluid). For example, the heat flow can be linked to temperature differences when particle collisions are dominant. Here, these relations are to be derived for gyro-fluid theories in collision-free and collision-dominated conditions, but in particular also for conditions that cannot be assigned to either regime. For this purpose, the most modern methods in analytical theory, numerical mathematics and simulations as well as machine learning are used. This should enable the design of nuclear fusion reactors and help to understand the ongoing dynamics therein using the much less complex gyro-fluid models.