Evolution of oxide composition, liquidus temperature, and viscosity during lunar molten regolith electrolysis

Abstract

The design and implementation of systems for molten lunar regolith electrolysis (MRE) benefit significantly from detailed modeling of the composition trajectory and evolution of composition-dependent thermophysical properties of regolith-derived oxides during this process. Two properties of interest are the liquidus temperature and viscosity of the molten oxide. The liquidus temperature evolution governs both the minimum reactor temperature and the extent of viable electrolysis at reasonably achievable reactor temperatures, while the viscosity is essential to the design of reactor outlet flow paths. This paper presents a strategy for modeling the oxide composition trajectory in MRE using the integrated thermochemical modeling software FactSage and calculating liquidus and viscosity evolution curves from the same software. This strategy was employed to model MRE of a lunar highlands regolith simulant composition. In contrast to previously employed modeling strategies, composition trajectory modeling in FactSage allows consideration of parallel species reduction during electrolysis, supporting detailed predictions of the compositions of reduced metallic products as a function of electrolysis reaction progress, which is quantified by the oxygen extraction yield parameter ${\lambda }_{O_2}$. For the regolith simulant composition modeled, it is suggested that a means of isolating ferrosilicon produced in the early stages of MRE from aluminum produced in more advanced stages should be implemented near ${\lambda }_{O_2}$ = 0.30 to maintain the usability of each metallic product. The calculated liquidus temperature evolution from FactSage suggests a minimum reactor temperature requirement of 1675 °C, which allows electrolysis to progress to ${\lambda }_{O_2}$ = 0.37, beyond which the liquidus rises above practical operating limits. Preliminary differential scanning calorimetry experiments indicate FactSage is more accurate than existing regression models at predicting liquidus evolution. The viscosity evolution predicted by FactSage is compared to the prediction of five models from the literature. Spindle viscometry experiments on select compositions indicate that FactSage outperforms all the literature models at least for ${\lambda }_{O_2}$ ≤ 0.165, demonstrating agreement to within 28 % between 1300 and 1600 °C for this composition range, which is attributed to its incorporation of a quasichemical model for bridging oxygen concentration in silicate melts. The advantages of using this model diminish for compositions beyond ${\lambda }_{O_2}$ = 0.30, which only contain small concentrations of SiO₂. The accuracy of viscosity models for these late-stage compositions may depend primarily on the distribution of compositions within the SiO₂-CaO-Al₂O₃ ternary system sampled to calculate empirical parameters.