Systematic assessment of generic force fields for CO2 adsorption in metal-organic frameworks

To ensure that computational screening of porous materials, such as MOFs, for carbon capture yields accurate predictions, it is essential to carefully, and thoroughly, validate and test the underlying molecular models. Yet, such validation studies are extremely scarce in the literature, and the vast majority of comparisons between simulated and experimental adsorption isotherms do not realistically consider the inherent uncertainty in either or both methods. In this paper, we conduct a systematic assessment of simulation force fields by comparing them against ‘consensus’ experimental isotherms derived from a curated dataset of carbon dioxide adsorption measurements. Our estimate for the average uncertainty in experimental adsorption isotherms is ∼15 %, while the average uncertainty arising from the choice of framework force field is ∼10 %; these uncertainties are quite significant and should be considered explicitly when comparing simulations to experiments. Remarkably, we observed that generic force fields taken ‘off the shelf’ only yielded good predictions of experimental data for one out of five MOFs studied here – IRMOF-1. The observed discrepancies for Cu-BTC and Co-MOF-74 can be explained by the inability of standard force fields to accurately describe the specific interactions of CO₂ with open metal sites. In contrast, the differences for UiO-66 can be rationalised by the presence of extensive defects in the MOF structure. However, the disagreement observed for MIL-47 has not been unequivocally explained, raising the need for more extensive experimental and simulation studies of this material. Based on these results, we provide concrete recommendations for future computational modelling studies of adsorption in MOFs.