Comparative Performance Analysis of Alternative Noble Gas Propellants for Hall Thrusters Using Machine Learning Prediction

Abstract

The growing demand for high-efficiency electric propulsion systems in modern satellite missions has intensified interest in alternative propellants for Hall-effect thrusters (HETs). Although xenon remains the standard propellant due to its favorable atomic mass and low ionization energy, its high cost and limited global supply motivate the exploration of viable substitutes such as krypton and argon. This study presents a comprehensive comparative analysis of noble gas propellants for Hall thrusters using a machine-learning-based predictive framework. An expanded dataset comprising more than 600 experimental data points collected from published literature was used to train and evaluate five regression models: Random Forest, Gradient Boosting, Extreme Gradient Boosting (XGBoost), Multi-Layer Perceptron neural networks, and Gaussian Process Regression. The models were developed to predict key thruster performance metrics including thrust, specific impulse, and overall efficiency under varying discharge voltages, magnetic field strengths, and propellant mass flow rates. Among the tested algorithms, the XGBoost model achieved the highest predictive accuracy with a coefficient of determination exceeding 0.97 for thrust prediction, demonstrating its capability to capture the nonlinear plasma–propellant interactions inherent in Hall thruster operation. Feature importance analysis reveals that thruster power and propellant atomic mass are the dominant factors governing thrust generation, while ionization energy significantly influences efficiency losses in lighter noble gases. The results confirm that krypton provides a competitive compromise between performance and cost, exhibiting only moderate efficiency degradation compared with xenon while significantly reducing propellant expenditure. Conversely, argon demonstrates substantially lower propellant utilization efficiency due to its higher ionization potential and increased beam divergence. A cost performance trade-off index is introduced to quantify mission-level implications of propellant selection. The findings provide a data-driven framework for evaluating alternative propellants and offer practical guidance for spacecraft designers seeking economically sustainable electric propulsion solutions for next-generation satellite constellations and deep-space missions.

Keywords: Hall-Effect Thruster; Electric Propulsion; Noble Gas Propellants; Xenon; Krypton; Machine Learning; Spacecraft Propulsion

Citation

Sultan M, Anik Chawdhuy NH, Akash NS. Comparative Performance Analysis of Alternative Noble Gas Propellants for Hall Thrusters Using Machine Learning Prediction. WebLog J Comp Sci Technol. wjcst.2026.d2901. https://doi.org/10.5281/zenodo.19951731