A Systematic Study of DFT Performance for Geometry Optimizations of Ionic Liquid Clusters

Abstract

Clusters of two ion pairs of imidazolium-based ionic liquids were optimized with 43 different levels of theory, including DFT functionals and MP2-based methods combined with varying Dunning's basis sets, and added dispersion corrections. Better preforming DFT functionals were then applied to clusters consisting of four ion pairs. Excellent performance of some DFT functionals for the two ion pair clusters did not always match that of the four ion-paired clusters despite interionic distances remaining constant between the optimized two and four ion-paired clusters of the same ionic liquid. Combinations of DFT functional and basis set such as ωB97X-D/cc-pVDZ, M06-2X/aug-cc-pVDZ, B3LYP-D3/cc-pVTZ, and TPSS-D3/cc-pVTZ gave excellent results for geometry optimization of two ion-paired clusters of imidazolium ionic liquids but gave larger deviations when applied to the four ion-paired clusters of varying ionic liquids. Empirical dispersion corrections were seen to be crucial in correctly capturing correlation effects in the studied ionic liquid clusters, becoming more important in larger clusters. Dunning's double-ζ basis set, cc-pVDZ, is associated with the smallest root mean squared deviations for geometries; however, it also produces the largest deviations in total electronic energies. ωB97X-D and M06-2X produced the best performance with the augmented version of this basis set. The triple-ζ basis set, cc-pVTZ, leads to the best performance of most of the DFT functionals (especially the dispersion-corrected ones) used, whereas its augmented version, aug-cc-pVTZ, was not seen to improve results. The combinations of functional and basis set that gave the best geometry and energetics in both two and four ion-paired clusters were PBE-D3/cc-pVTZ, ωB97X-D/aug-cc-pVDZ, and BLYP-D3/cc-pVTZ. All three combinations are recommended for geometry optimizations of larger clusters of ionic liquids. PBE-D3/cc-pVTZ performed the best with an average deviation of 2.3 kJ mol −1 and a standard deviation of 3.4 kJ mol −1 for total electronic energy when applied to four ion-paired clusters. Geometries optimized with FMO2-SRS-MP2/cc-pVTZ produced total energy within 2.0 kJ mol −1 off the benchmark in two ion-paired clusters, with the cc-pVDZ basis set performing unsurprisingly poorly with the same method. The error increased to 4.8 kJ mol −1 on average in four ionpaired clusters, with the smallest RMSD deviations in geometries when compared to the benchmark ones. This study is the first report that investigated the performance of DFT functionals for two and four ion-paired clusters of a wide range of ionic liquids consisting of commonly used cations such as pyrrolidinium, imidazolium, pyridinium, and ammonium. It also identified the importance of assessing the performance of quantum chemical methods for ionic liquids on a variety of cation-anion combinations.

Carbon steel corrosion inhibition in H2SO4 0.5 M medium by thiazole-based molecules: Weight loss, electrochemical, XPS and molecular modeling approaches

“…For the spectrum of Fe 2p (Fig. 11) in the presence of the product P1, there is the presence of a doublet: 712 eV (Fe 2p1/2) and 725 eV (Fe 2p3/2), which confirms the oxidation of the surface of carbon steel [70][71]. Also, two little peaks, the first one at 707 eV characteristic of Fe 0 and the second one appears at 719 eV (Fe(III)) [66].…”

Topology Automated Force-Field Interactions (TAFFI): A Framework for Developing Transferable Force-Fields

“…Regarding the first source of error, the dispersion-corrected range-separated functional used here is among the highest performing in benchmarks of conformational energetics and cluster interactions for organic species. 62,[68][69][70][71][72] Nevertheless, even modern functionals have documented deficiencies for aqueous solutions and reaction barriers that would require higher fidelity training data for models of water or reactive force-fields, which are beyond the present scope. Thus, for the current study, we acknowledge this potential source of error but consider it negligible in comparison with the representability errors associated with the simple functional form of the force-field.…”

Topology Automated Force-Field Interactions (TAFFI): A Framework for Developing Transferable Force Fields

“…Thus, the errors in TAFFI-gen predictions can be decomposed into errors associated with the underlying DFT parametrization data and representability errors associated with the limited functional form of the force field. Regarding the first source of error, the dispersion-corrected range-separated functional used here is among the highest performing in benchmarks of conformational energetics and cluster interactions for organic species. ,− Nevertheless, even modern functionals have documented deficiencies for aqueous solutions and reaction barriers that would require higher fidelity training data for models of water or reactive force fields, which are beyond the present scope. Thus, for the current study, we acknowledge this potential source of error but consider it negligible in comparison with the representability errors associated with the simple functional form of the force field.…”