Chemical Model Reduction for Negative Hydrogen Ion Density Predictions Using Global Sensitivity Analysis

Negative hydrogen ion sources (NHIS) are the preferred mode of plasma heating in tokamak devices for nuclear fusion. Accurate predictions of the H^- density are crucial for the design of such machines. This work exploits state-of-the-art statistical and data science tools developed for machine learning and uncertainty quantification to improve the predictive capabilities of the Global Model for NHIS by W. Yang et al. To achieve high fidelity in reproducing the important kinetics features, this model involves detailed pathways composed of hundreds of reactions. While increased complexity represents a benefit in terms of completeness, the resulting augmentation of model capacity requires parameter calibration through sufficient validation data, which are not always available. With this constraint in mind, this study relies on an efficient framework consisting of: i) global sensitivity analysis (GSA) enabled by polynomial chaos expansions to inform the construction of reduced-order models that omit the uninformative reactions, and ii) solution of an inverse problem via Bayesian inference to update the most influential chemical reactions from experimental data. The analysis suggests that reducing the NHIS mechanism from more than 1300 reactions to only 35 provides accuracy within ±10% for the prediction of H^- number density. Such a low-dimensional pathway is dominated by H2+e dissociation, associative detachment, wall recombination, and dissociative electron attachment.