Multiscale Prediction of Thermodynamic Properties of Supercritical Aqueous Sodium Sulfide Using Synchrotron Experiments, Atomistic Simulations, and Artificial Neural Networks

Abstract The increase of fossil fuel use is responsible for increased greenhouse gas emissions, which have a great impact on the environment, leading to global warming. Consequently, substantial efforts are devoted to decreasing our dependence on fossil fuels and there is a growing interest in harnessing the chemical energy from other carbon-neutral sources like biomass. Supercritical water gasification (SCWG), which is a catalyst-mediated process, provides a promising method to transform organic streams and wastes into fuels due to the utilization of water as a nontoxic green solvent. However, the process is limited by sulfur impurities that deactivate the catalysts required for efficient conversion. The optimization of SCWG remains challenging because the solvent properties of aqueous mixtures in the supercritical regime are unknown, and they have significant differences while crossing the Widom line, which separates liquid- and gas-like states. In this study, we propose a framework to determine accurate thermodynamic properties for complex systems under supercritical conditions, where experimental data are scarce. We demonstrate that a methodology combining structural characterization experiments, atomistic simulations, and machine learning enables the development of reliable equations of state for systems with previously unknown properties. This approach characterizes changes in the thermophysical properties of the aqueous solution across the Widom line, with potential implications for ion solvation and sulfur speciation in SCWG. By providing thermophysical property predictions that may inform process models aimed at improving sulfur removal and mitigating catalyst poisoning, our framework advances the SCWG method as a clean energy technology. More broadly, the methodology is generalizable and can be applied to a wide range of complex fluid systems with poorly understood thermodynamic behavior.