Gradient- and flux-driven global gyrokinetic simulations of ITG and TEM turbulence with an improved hybrid kinetic electron model

ORB5 is a global gyrokinetic code being developed by many scientists over the last 20 years. It allows performing so-called gradient-driven simulations, in which initial profiles such as density and temperature are specified and an adaptative source is used to relax the zonal component of the perturbed distribution function so that the total flux-surface-averaged distribution function (f) over bar does not deviate too much from the initial one, (f) over bar (0). Thus, f(0) can be used as a control variate to reduce the sampling noise. Even though the gradient-driven model is useful due to its relatively low numerical cost (there is no need to perform simulations over transport time scales) it does not reflect the experimental reality where fluxes are prescribed by the actual heat sources. Furthermore, comparison between experimental and gradient-driven numerical predictions of the heat fluxes is a challenging task because of the profile stiffness: a small change, within experimental error bars, in the imposed temperature gradients leads to a large variation of the computed heat fluxes. To address the stiffness problem, the ORB5 code has been adapted to be able to run flux-driven simulations. In this model, fluxes are prescribed by a radially localized heat source and sink and the background profiles slowly evolve toward a quasi-steady state. As a first step toward complete flux-driven simulations, a mixed approach is used, in which the run is started in the gradient-driven mode providing estimates for the heat sources/sinks, which are then used for pursuing the simulation in flux-driven mode. This allows us to keep the numerical cost relatively low as a quasi-steady state is more quickly reached. In this work, we present the heat source implemented in ORB5 to perform the flux-driven simulations as well as the numerical technique used to ensure that no other moment is injected by the source up to machine precision. Furthermore, simulation results illustrating the mode switching between gradient- and flux-driven will be shown as well as preliminary results of TEM dominated cases done using an upgraded hybrid electron model.