Etude GYrocinétique des Plasmas Turbulents – EGYPT

Turbulence in fusion plasmas is among the issues of first principle physics that are questioned by the ITER project. Turbulence is not only a major challenge to theoretical physics, it is the main performance limitation of the confinement capability of the device. Specific physics of fusion plasmas still increases the complexity of turbulence transport in this media. On the one hand, electromagnetic fields, including those generated by the plasma motion itself, drive the turbulence. Such a system is thus prompt to instabilities and requires a self-consistent analysis of the plasma and the electromagnetic fields. On the other hand, binary collisions are negligible owing to the weak density of such plasmas so that a kinetic description must be considered. In the so-called gyrokinetic framework, the computation of the distribution function of each species must still be performed in a reduced 5 dimension phase space. The numerical simulation of plasma turbulence then requires advanced numerical schemes that must be thoroughly optimised to tackle the characteristic parameter space of ITER plasmas. The United States are leaders in this research activity taking advantage of their capability to gather the required teams and making available the most powerful computing facilities. Given this situation, the Département de Recherche sur la Fusion Contrôlée at CEA has developed the code GYSELA based on a completely novel scheme. In the last year this code has proven to be among the most powerful simulation tool of plasma turbulence. This effort stems from a large scale collaboration of French research institutions bridging numerical skills, applied mathematics and physics of kinetic turbulence in fusion plasmas. In Europe, they are only two other codes addressing these physics. The present goal is to take part in the vigorous debate taking place on the issue of numerical errors as well as extending our understanding of plasma transport in the operating regimes foreseen in ITER. Present simulations aim at computing transport on some 4.29 1010 grid points using more than 4000 processors with massively parallel computing. In such a demanding framework, the code must take benefit of all properties of the computing platform that governs therefore the way of addressing the physics. Code optimisation will then lead to recast the physical problem as well as to work on the numerical scheme itself. Such issues can only be addressed by combining various skills and fields of expertise. Regarding the physics of fusion plasmas, the main goal is to recover the experimental scaling laws of the turbulent transport by taking into account the heat sources, collisions and the interplay between electron and ion physics. In such a procedure, one must also identify the mechanisms that govern a reduction of turbulent transport and therefore allow one to achieve improved confinement regimes. This introduces the issue of control, in particular that of turbulent eddies and that of phase space mixing. The theoretical starting point to investigate such aspects of control is the dimension reduction of the phase space performed in the Hamiltonian framework. This step provides the means to investigate the distribution functions in the gyrokinetic framework while preserving the Hamiltonian structure of the system. This provides the correct setting to address problems such as that of Hamiltonian control or that of the choice of variables. Of particular interest is the separation between slow and fast variables as well as defining the most appropriate flow pattern for setting the time splitting procedure of the code. Furthermore, this way of casting the evolution of the distribution function must be compatible with solving the quasi-neutrality constraint that relates particle density in real space to the electromagnetic fields. Such an ambitious programme must also analyse novel means of casting the problem such as the multifluid approach (also called waterbag) obtained by an alternative projection of the distribution function on discrete functions or by considering the solution of the kinetic equations as the output of as specific stochastic process. The financial support that is requested for the present project should cover the expense of extending the computing power of a platform dedicated to the code development as well as hiring post-docs who will take part in the collaborative work between the various teams of physicists, DRFC Cadarache, CPT Marseille, LPMIA in Nancy, and the applied mathematics teams located at IRMA in Strasbourg. Such a support would favour the effort committed to the development of the GYSELA code and the physics addressed by the simulations.