Efficient Computation of the Interaction Energies of Very Large Non-covalently Bound Complexes

Abstract

We present a new benchmark set consisting of 16 large non-covalently bound systems (LNCI16) ranging from 380 up to 1988 atoms and featuring diverse interaction motives. Gas-phase interaction energies are calculated with various composite DFT, semi-empirical quantum mechanical (SQM), and force field (FF) methods and are evaluated using accurate DFT reference values. Of the employed QM methods, PBEh-3c proves to be the most robust for large systems with a relative mean absolute deviation (relMAD) of 8.5% with respect to the reference interaction energies. r2SCAN-3c yields an even smaller relMAD, at least for the subset of complexes for which the calculation could be converged, but is less robust for systems with smaller HOMO–LUMO gaps. The inclusion of Fock-exchange is therefore important for the description of very large non-covalent interaction (NCI) complexes in the gas phase. GFN2-xTB was found to be the best performer of the SQM methods with an excellent result of only 11.1% deviation. From the assessed force fields, GFN-FF and GAFF achieve the best accuracy. Considering their low computational costs, both can be recommended for routine calculations of very large NCI complexes, with GFN-FF being clearly superior in terms of general applicability. Hence, GFN-FF may be routinely applied in supramolecular synthesis planning.1 Introduction2 The LNCI16 Benchmark Set3 Computational Details4 Generation of Reference Values5 Results and Discussion6 Conclusions

Reliable Dimerization Energies for Modeling of Supramolecular Junctions

“…This figure indicates that the ωB97X-3c/vDZP approach quite accurately reproduces relative differences in the Δ E values. As a consequence, an application of this method would be useful for ranking tentative structures in large supramolecular arrangements [ 53 ].…”