NaCl Dissociation Explored Through Predictive Power Path Sampling Analysis

Utilizing the replica exchange transition interface sampling (RETIS) technique, we simulated the dynamics of sodium chloride dissociation in water. Subsequently, the resulting trajectories were analyzed using predictive power analysis (PPA), enabling the identification and quantification of collective variables (CVs) capable of forecasting the reaction occurrence. We improved the robustness of the PPA method by incorporating the Savitzky-Golay (SG) filter on integrated histograms, effectively avoiding the limitations associated with binning. Applying this adapted PPA method, the previously designed solvent parameters and distances from the index invariant distance matrix were assessed. This revealed that the sixth closest oxygen to sodium serves as an equally effective predictor as the best complex solvent parameter. The latter, however, required more knowledge and human intuition as an input for its design, while the former provided such intuition purely as an output. Through a comparable analysis, the chloride solvation shell appears to contain less predictive information. Employing a linear combination of several CVs can further enhance predictability, albeit at the expense of a reduced human interpretability.