An alternative method to the classical fit of semi-empirical, statistical, or artificial intelligence-based models to retention data is proposed to predict surface excess adsorption and retention factors in liquid chromatography. The approach is based on a fundamental, microscopic description of the liquid-to-solid adsorption of analytes taking place at the interface between a bulk liquid phase and a solid surface. Molecular dynamics (MD) simulations are performed at T=300 K in a 100 Å wide slit-pore model (β-cristobalite-C18 surface in contact with an acetonitrile/water mobile phase) to quantify a priori the retention factors of small molecules expected in reversed phase liquid chromatography (RPLC). Uracil is chosen as the reference “non-retained” marker, whereas benzyl alcohol, acetophenone, benzene, and ethylbenzene are four selected retained, neutral compounds. The MD simulations allow to determine the pore-level density profiles of these five compounds, i.e., the variation of the analyte concentration as a function of distance from the silica surface. The retention factors of the retained analytes are expressed using their respective calculated surface excess adsorption relative to uracil. By definition, the retention factors are proportional to the surface excess adsorbed and the proportionality constant is directly scaled to the retention time of the “non-retained” marker. Experimentally, a 4.6 mm × 150 mm RPLC-C18 column packed with 5 μm 100 Å High Strength Silica (HSS)-C18 particles is used and the retention times of these five compounds are measured. The volume fraction of acetonitrile in water increases from 20 to 90% generating a wide range of retention factors from 0.15 to 183 at T=300 K. The results demonstrate very good agreement between the MD-predicted surface excess adsorption data and measured retention factors (R2> 0.985). A systematic error is observed as the proportionality constant is not exactly scaled to the retention time of uracil. This is most likely caused by the differences between the chemical and morphological features of the slit-pore model adopted in the MD simulations and those of the actual HSS-C18 particles: the average surface coverage with C18 chains, the geometry of the mesopores, and the pore size distribution. Specifically, the impact on RPLC retention of slight, local variations in surface chemistry (e.g., functional group density and uniformity) and how this aspect is affected by the pore space morphology (e.g., pore curvature and size) is worth investigating by future MD simulations.
Read full abstract