(350e) Molecular Models with Accurate Descriptions in Electrostatic Potential for Molecular Simulations of Adsorption

Nanoporous materials, such as zeolites or metal-organic frameworks, have shown great potential as adsorbents for the removal or separation of gas mixtures in energy or environment-related applications. However, while such a large database of nanoporous materials provide great opportunities, it has also made the discovery of optimum materials unprecedentedly challenging for a given application. To this end, computational materials discovery by employing classical molecular simulations can play a key role in facilitating materials discovery. To ensure the reliability of the predictions, accurate descriptions of the framework-molecule and molecule-molecule interactions are critically significant. For this reason, a molecular model that is able to accurately describe the molecule of interest is of utmost importance. Thus, we herein present a collection of molecular models of a variety of gas molecules, which their separation or purification is critically essential in energy- and environment-related processes. These gas molecules include carbon monoxide (CO), sulfur dioxide (SO₂), carbon dioxide (CO₂), nitrogen (N₂), hydrogen sulfide (H₂S), carbonyl sulfide (COS), nitrous oxide (N₂O), and ammonia (NH₃). For these models, the number of massless sites, the location of massless sites, and the point charges are first fully optimized to reproduce the electrostatic potential (ESP) surrounding the molecule determined by density functional theory (DFT) calculations, while their van der Waals (vdWs) interaction parameters are subsequently fitted by reproducing the experimental vapor liquid equilibrium (VLE). Thanks to the optimization of key model parameters as noted above in reproducing the ESP, a variety of properties in both the adsorbed and the bulk liquid phase can be well predicted. We also identified that different vdWs interaction parameters with a given number and location of massless sites and the corresponding point charges are able to resemble the experimental VLE curve equally well. However, these variations in the vdWs parameters can still yield consistent predictions in various properties, such as binding energy and geometry and bulk liquid density, emphasizing again the importance of reproducing ESP. On the other hand, we found that radial distribution function in the bulk liquid phase is sensitive to these variations in the vdWs parameters. This suggests that a relative orientation between the molecules can serve as a good secondary reference in addition to the VLE curve in the optimization of the vdWs interaction parameters. Overall, we have developed a set of molecular models with accurate description of the electrostatic potential for small molecules that are highly relevant to a variety of energy- and environment-related separation applications, which can be readily implemented to facilitate the computational discovery of optimum adsorbents.