Constructing a new predictive scaling formula for ITER's divertor heat-load width informed by a simulation-anchored machine learning

Abstract

Understanding and predicting divertor heat-load width λq is a critically important problem for an easier and more robust operation of ITER with high fusion gain. Previous predictive simulation data using the extreme-scale edge gyrokinetic code XGC1 in the electrostatic limit for λq under attached divertor plasma conditions in three major US tokamaks [C.S. Chang et al., Nucl. Fusion 57, 116023 (2017)] reproduced the Eich and Goldston attached-divertor formula results [T. Eich et al., Phys. Rev. Lett. 107, 215001 (2011); R.J. Goldston, Nucl. Fusion 52, 013009 (2012)], and furthermore predicted over six times wider λq than the maximal Eich and Goldston formula predictions on a full-power scenario ITER plasma. After adding data from further predictive simulations on a highest current JET and highest-current Alcator C-Mod, a machine learning program is used to identify a new scaling formula for λq as a simple modification to the Eich formula #14, which reproduces the Eich scaling formula for the present tokamaks and which embraces the wide λq XGC for the full-current Q = 10 ITER plasma. The new formula is then successfully tested on three more ITER plasmas: two corresponding to long burning scenarios with Q = 5 and one at low plasma current to be explored in the initial phases of ITER operation. The new physics that gives rise to the wider λq XGC is identified to be the weakly-collisional trapped-electron-mode turbulence across the magnetic separatrix, which is known to be an efficient transporter of the electron heat and mass. Electromagnetic turbulence and highcollisionality effects on the new formula are the next study topics for XGC1.

Turbulent broadening of electron heat-flux width in electromagnetic gyrokinetic simulations of a helical scrape-off layer model

“…The validity of extrapolating this empirical scaling to future devices like ITER is an important issue that can be addressed by first-principles modeling. For example, XGC1 electrostatic gyrokinetic simulations reproduce the empirical scaling for existing tokamaks, but predict a λ q that is about 6 times wider when extrapolated to ITER, with the widening due to an increase in trapped-electron turbulence 6,7 .…”

2022 Review of Data-Driven Plasma Science

“…Again, these discovered models (discussed more below in section D) can be either largely empirical [393], or additionally constrained to be physically consistent with theoretical considerations [394] or simulation-based considerations (e.g. high fidelity model predictions) [389,395]. It is generally thought that as the amount of both experimental and simulation data increases, data-driven physics models may become increasingly important for being able to predict and model experimental behaviors while simultaneously connecting the gained insights from these systems to traditional and first principle ways of understanding the plasma physics.…”

“…• 3D MHD equilibrium calculations for stellarator optimization applications [439] • MHD instability calculations, as part of a disruption predictor stack [440] • Surrogate formula for divertor heat-load width built from combined experimental and gyrokinetic simulation data [395]. Further extension of these techniques to incorporate all components of the MCF multi-physics simulation stack provides a pathway towards fast and accurate interpretation of present-day experiments, scenario design and optimization (including inter-shot), and control-oriented modelling.…”

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

“…The boundary region is critical in determining a fusion reactor's overall viability since edge plasma conditions strongly influence a myriad of reactor operations [51,125,23]. Validating edge turbulence models is accordingly a crucially important endeavour since gathering sufficient information to effectively test reduced turbulent transport models is vital towards developing predictive capability for future devices.…”