Molecular Modeling and Simulation of Hydrogen Bonding Pure Fluids and Mixtures

Hydrogen bonding is a strong, directional intermolecular interaction that drives association and network formation and thereby causes unusual thermodynamic properties such as high critical temperatures and large enthalpies of vaporization, and it also plays a key role in biological structure. Accurately modeling hydrogen-bonding fluids is challenging because both energetic and structural effects must be represented consistently. This work evaluates a simple, computationally efficient molecular-modeling strategy for predicting thermodynamic and structural properties of hydrogen-bonding fluids in pure substances and mixtures. For efficiency, dispersion, polarity, and hydrogen bonding are captured using only two potential types: anisotropic united-atom Lennard–Jones interactions plus point charges. Based on this framework, high-accuracy models are developed for methanol, ethanol, mono- and dimethylamine, methanethiol, formic acid, and water, with geometries largely taken from quantum-chemical calculations (Hartree–Fock/6-31G) and remaining parameters fitted mainly to saturated liquid densities and vapor pressures over nearly the full vapor–liquid coexistence range using an NpT + test-particle approach. Typical deviations are below ~1% for saturated densities and below ~5% for vapor pressures, with vaporization enthalpies predicted around ~5% for most substances; strongly associating formic acid and especially water reveal remaining limitations of the simple point-charge description, motivating a focused water pilot study comparing TIP4P/TIP5P-inspired variants and an all-atom refinement. For mixtures, the work analyzes how unlike Lennard–Jones parameters affect VLE and shows that standard combining rules are generally insufficient for quantitative vapor-pressure prediction; instead, using the Lorentz rule for unlike size and fitting a single binary interaction parameter for unlike energy can yield consistent predictions across multiple properties. This strategy is applied to 32 binary mixtures containing a hydrogen-bonding component, demonstrating that one fitted, state-independent binary parameter can reproduce vapor pressures, compositions, and Henry constants over ranges of conditions. Structural fidelity is assessed primarily for methanol by comparing simulated radial distribution functions with neutron-diffraction data, showing good agreement despite the reduced united-atom representation. To probe hydrogen bonding in mixtures where diffraction data are scarce, new NMR measurements of methanol in near- and supercritical CO₂ are performed and compared with simulation-based hydrogen-bond statistics using an established geometric bonding criterion. By combining simulated donor/acceptor population statistics with only two state-independent parameters and the experimentally inferred monomer reference shift, the relative chemical shifts are predicted with excellent agreement, confirming that the approach can capture both thermodynamic behavior and hydrogen-bonding structure with high accuracy while remaining computationally efficient.