Effect of molecular dynamics water models on flux, diffusivity, and ion dynamics for polyamide membrane simulations

While molecular dynamics (MD) models for reverse osmosis (RO) and nanofiltration (NF) membranes are useful for studying water filtration processes, the specific water model and force-field parameters have a profound impact on key measures such as flux and diffusivity. Moreover, ensemble related parameters such as temperature and how it is controlled via a thermostat also affect simulation results. To quantify these effects, we investigate the performance of three commonly used water models, SPC/E, TIP3P, and TIP4P, in simulating an NF membrane system at varying temperatures and thermostat settings. All three water models successfully capture the Arrhenius relationship between water diffusivity and temperature as well as the linear relationship between water flux through the membrane and water diffusivity in the membrane. However, SPC/E underpredicts, TIP3P overpredicts, and TIP4P most accurately reproduces the water dynamics, especially water self-diffusivity, compared to experiments. We also investigate the effect of the water model on common salt solutions at various concentrations. TIP4P generally outperforms the other two water models when describing ion diffusivities and water solvation shells around ions. These findings indicate that TIP4P is an appropriate water model to use in polyamide membrane MD simulations.