Study on the Adsorption Performance of Ionic Liquids Based on Molecular Dynamics and Interpretable Machine Learning

The stable adsorption behavior of ionic liquid lubricants at metal interfaces is a key mechanism for achieving their excellent friction-reducing and antiwear properties. This study employs a research strategy that combines high-throughput molecular dynamics simulations with interpretable machine learning to construct a data set of adsorption energy for 354 different alkyl chain structures of ammonium phosphate esters. By integrating statistical analysis with a feature recursive elimination algorithm, we effectively reduced the dimensionality of high-dimensional descriptors while fully preserving the physicochemical characteristic information. The reliability of the feature selection method was validated using four typical machine learning models. The quantitative structure-property relationship model established through symbolic regression indicates that, compared to branched alkanes, the increase in chain length of linear alkanes significantly enhances interfacial van der Waals interactions by promoting molecular conformational expansion and σ-electron delocalization effects. However, when exceeding a critical chain length, the dominant intramolecular forces lead to a gradual increase in adsorption energy. This research provides an important theoretical basis for the molecular design of high-performance ammonium phosphate ester ionic liquid lubricants.