The GBS code for the self-consistent simulation of plasma turbulence and kinetic neutral dynamics in the tokamak boundary

Physics-informed machine learning techniques for edge plasma turbulence modelling in computational theory and experiment

“…Model approximations are thus inevitable, but once approximations are introduced into edge turbulence models, their effects on nonlinear dynamics need to be quantitatively evaluated-against higher fidelity simulations and/or experiment, if possiblefor model predictions to be meaningful. An oft-applied set of equations to study the plasma edge is drift-reduced Braginskii fluid theory, which has been studied for decades starting from Braginskii's original formulation [14], and remains a highly active research area today [47,207,85,200,236,211,66,30]. But the underlying model approximations tend to only be marginally satisfied in fusion plasmas and there is a scarcity of clear-cut validation tests of the theory against higher-fidelity simulations or experimental measurements.…”

“…But going forward, there are several extensions possible in translating these calculations towards empirical testing in magnetic confinement fusion devices to uncover new physics. For example, once flows and geometric effects arising from the poloidal magnetic field [116] are inserted into the deep learning framework (and validated using modern 3-dimensional codes [235,155,66]), the predictions from drift-reduced Braginskii theory can be directly compared to available experimental poloidal electric field measurements [143]. Such experimental information can then even be used to invert the computational technique to potentially begin learning missing or misrepresented physical terms (e.g.…”

“…In Chapter 4, a path was embarked on to start translating these techniques to the highest fidelity plasma of all: experiment. For this task, an entirely new optimization scheme based upon a multi-network physics-integrated framework was used to convert brightness measurements of HeI line radiation into local plasma and neutral fluctuations via the merging of transport physics and collisional radiative modelling for the 3 3 𝐷−2 as in [66,30,234] could enable this work to use the existing GPI data for comparison of turbulent electric fields with available probe measurements. These increasingly sophisticated full device simulations include self-consistent kinetic neutrals (i.e.…”

Turbulent transport regimes in the tokamak boundary and operational limits

“…The physical model considered here is based on the drift-reduced Braginskii model 22 implemented in GBS. 21 For simplicity, the coupling to the neutral dynamics is neglected, although implemented in GBS. 23 The validity of a drift-reduced fluid model is limited to the regime of electron mean free path shorter than the parallel connection length, λ e L 2πqR, and perpendicular scale lengths of the dominant modes larger than the ion Larmor radius, k ⊥ ρ i 1.…”

“…19 by leveraging a set of three-dimensional, flux-driven, two-fluid electromagnetic turbulence simulations, carried out with the GBS code. 21 With respect to the simulations in Ref. 19, we consider here simulations that include electromagnetic effects and avoid the use of the Boussinesq approximation.…”