Evolutionary optimization of the multidimensional parameter space governing tungsten oxidation in fusion environments

Abstract Tungsten is a primary candidate for plasma-facing components in fusion devices, but its susceptibility to rapid oxidation during accidental air ingress poses significant safety risks. Existing models of tungsten oxidation often rely on individual parameters (such as oxygen diffusivity and adsorption energies) derived from first principles simulations, which frequently fall far from accurately explaining experimental oxidation rate trends. This study uses an improved multilayer interface tracking model that treats the oxide scale as a dynamic system of WO 3 , WO 2.9 , WO 2.72 , and WO 2 layers. To address the unknown nature of many layer-specific parameters, we employ a Differential Evolution genetic algorithm to optimize a 13-dimensional parameter space against a comprehensive database of experimental growth constants. The resulting optimized model accurately reproduces experimental oxidation kinetics and correctly predicts the temperature-dependent evolution of specific oxide phases, providing a more reliable tool for assessing the performance of fusion materials under extreme conditions. The model can be dynamically improved by incorporating experimental data as they become available, leading to parameters that reflect the state of the art in oxidation measurements and characterization.