Predicting Solvation Free Energies of Molecules and Ions via First-Principles and Machine-Learning Molecular Dynamics

Area of Science:

• Computational Chemistry
• Physical Chemistry
• Molecular Dynamics

Background:

• Solvation free energy (SFE) is crucial for understanding molecular behavior and solubility.
• Alchemical methods for SFE computation face end-point singularities, causing numerical instability, especially in advanced simulations.

Purpose of the Study:

• Introduce a novel 'bubble method' for calculating SFEs from first-principles.
• Address the end-state problem in ab initio and machine-learning molecular dynamics.
• Provide a versatile method applicable to various molecular and ionic species.

Main Methods:

• Developed the bubble method for first-principles SFE calculations.
• Applied the method to classical, ab initio, and machine-learning molecular dynamics.
• Incorporated corrections for periodic boundary conditions and interface potentials in DFT calculations for ions.

Main Results:

• Successfully computed SFEs for methane, methanol, water, piperazine, sodium, and potassium ions.
• Demonstrated the method's applicability across different simulation types.
• Validated the avoidance of end-state problems and numerical instabilities.

Conclusions:

• The bubble method offers a robust, data-independent approach to SFE calculation.
• Enables reliable simulations of solvation under extreme conditions (high pressure, temperature, nanoconfinement).
• Presents a significant advancement over traditional methods, particularly where experimental data or reliable force fields are unavailable.