Molecular Simulations and Data-Driven Predictive Modeling of Mercury Adsorption in Special Solvents
Mercury contamination presents critical challenges to the oil and gas industry due to its toxicity and severe environmental and health repercussions. Traditional mercury removal methods often suffer from inefficiency, limited selectivity, and suboptimal predictive capabilities for adsorption performance. This study addresses these challenges by integrating molecular simulations and data-driven predictive modeling to evaluate and optimize mercury adsorption in specialized solvents, with a focus on ionic liquids. Leveraging the exceptional thermal stability and selectivity of ionic liquids, Turbomole and COSMO-RS were employed to calculate sigma surfaces and profiles, providing insights into molecular properties relevant to mercury adsorption. Key predictors, including viscosity, density, activity coefficient, and Henry's constant, were derived from COSMO-RS outputs and incorporated into regression models using MATLAB's Curve Fitting Toolbox. Multiple Linear Regression (MLR) and Multiple Non-Linear Regression (MNLR) were applied to establish relationships between these predictors and the mercury adsorption capacity and efficiency. Sigma surface and profile analyses revealed that molecules with higher electronegativity exhibited superior mercury adsorption. MNLR also effectively captured nonlinear interactions, yielding more accurate adsorption capacity and efficiency predictions than MLR with an adjusted R2 of 1 and 0.7105. Among the predictors, viscosity and activity coefficient emerged as the most reliable for adsorption capacity and efficiency predictions, registering an RMSE of 1.129 and 1.590, respectively. This study highlights the potential of combining computational chemistry with advanced data analytics to overcome limitations of current methods and enhance mercury removal strategies. Despite dataset limitations, the findings demonstrate the promise of integrating molecular simulations and data analytics for designing efficient mercury mitigation strategies, laying a foundation for future improvements in predictive accuracy and industrial applications.