Insights into the diffusion coefficient and adsorption energy of NH3 in MgCl2 from molecular simulation, experiments, and machine learning

Abstract

Molecular models were developed to evaluate the adsorption behavior between NH3 and MgCl2. These models were utilized to assess the diffusion coefficient and adsorption energy at different temperatures and pressures through the application of molecular dynamics (MD) simulations. Subsequently, a comprehensive dataset comprising the diffusion coefficient and adsorption energy was established. The analysis of the relative diffusion and adsorption mechanisms involved calculating the radius distribution function, coordination numbers, and energy values within the primary solvent layer. The molecular simulation results revealed that the highest values for the diffusion coefficient and adsorption energy of NH3 in MgCl2 were observed at a temperature of 348 K and a pressure of 0.2 MPa. Moreover, the experimental findings exhibited good agreement with the computational simulation conclusions. The preparation of Magnesium hydroxide (MH) under the aforementioned temperature and pressure conditions resulted in a concentrated particle size distribution, effective dispersion, and a complete hexagonal sheet morphology. Furthermore, machine learning predictions were performed using significant features (i.e., molarity(M), temperature(T), pressure(P), volume(V), density(D), and total energy(E)). The results demonstrated that the decision tree regression (DTR) model exhibited superior performance in predicting the diffusion coefficient, while the k-nearest Neighbors (KNN) regression model achieved the best performance in predicting the adsorption energy.