Multiscale Collisional Gyrokinetic Turbulence with GYSELA

First-principle predictive capability in fusion research is based on gyrokinetic simulations of plasma turbulence. Ideally, all scales must be self-consistently treated, from the small turbulent ones up to the large scales of e.g. axisymmetric flows. However, since gyrokinetic 5D simulations are numerically severely demanding, it is of wide use to maintain two scale separations. On the one hand, flux-tube simulations are radially localised and assume a small scale fluctuation-induced transport. On the other hand, the so-called δf approach assumes a fixed background Maxwellian distribution function and concentrates on the dynamics of the fluctuations only. The 5D full torus (global) and full–f GYSELA code [1] overcomes these limitations and allows to self-consistently describe large-scale fast transport events and complex non-Maxwellian distribution functions, relevant in weakly collisional plasmas. It models the electrostatic toroidal branch of the Ion Temperature Gradient (ITG) instability with adiabatic electrons. In global collisionless simulations, the choice of the initial state is important. Two situations with two strongly differing initialisations are investigated: the initial distribution function is canonical, i.e. is an equilibrium solution of the gyrokinetic equation or is non canonical, and a representative choice for such distribution functions is the classical Maxwellian. We address: (i) how the latter non equi