Assessing the performance of commonly used DFT functionals in studying the chemistry of frustrated Lewis pairs

Abstract

The benchmark study has assessed the performance of 18 density functional theory (DFT) functionals, including GGAs, hybrid-GGAs, meta-GGAs, and hybrid meta-GGAs, in predicting bonding strength, barrier height and structure for systems involving Lewis acids and bases. Three databases were built for the study, including 15 bonding enthalpies of dative bonds (DBH15), 10 reaction barriers (BH10) and 10 X-ray structures (XCS10). Wavefunction-based ab initio calculations were also carried out for comparisons. The benchmark data were computed at the CCSD(T)/BSI//MP2/BSI(BSI=aug-cc-pVTZ) level. The 6-311++G(d,p)(BSII) and 6-31G(d,p)(BSIII) basis sets were employed in DFT calculations. Generally, M06-2X/BSII and M05-2X/BSII outperform the other tested DFT methods. M05-2X/BSIII and M06-2X/BSIII are less accurate than M05-2X/BSII and M06-2X/BSII (or M05-2X/BSII//M05-2X/BSIII and M06-2X/BSII//M06-2X/BSIII), suggesting that large basis sets (e.g. BSII) are necessary to improve energetics. SCS-MP2 is less accurate than MP2, consistently overestimating bonding enthalpies and reaction barriers. Moreover, six DFT functionals (M05-2X, M06-2X, B3LYP, ωB97X-D, BMK and B97-D) were examined by comparing with the experimental bonding enthalpies of eighteen donor–acceptor complexes, which indicate that M05-2X and M06-2X are still better than others. Nevertheless, M05-2X and M06-2X significantly overestimate or underestimate the bonding enthalpies of F-substituted complexes, implying the necessity of improving the two functionals for describing fluorides. Using the six basis sets (BSI, cc-pVTZ, aug-cc-pVDZ, cc-pVDZ, TZVP and BSII) and DBH15 and BH10 databases, the influence of the basis sets on the performance of M06-2X functional was examined, which reveals that BSII is the most suitable basis set for the functional.