Reliable prediction of association (free) energies of supramolecular complexes with heavy main group elements – the HS13L benchmark set

Abstract

We introduce a set of 13 supramolecular complexes featuring diverse non-covalent interactions with heavy main group elements (Zn, As, Se, Te, Br, I), high charges (−2 up to +4), and large systems with up to 266 atoms (HS13L). The experimental Gibbs free energies of association cover the typical range (−1.9 to −9.2 kcal mol⁻¹). An efficient automated multilevel theoretical workflow is applied for the determination of the respective minimum structures in solution by conformer ensemble generation with the CREST program at the semiempirical GFN2-xTB level. Subsequent refinement is performed with the r²SCAN-3c composite DFT method including thermostatistical corrections at the GFN2-xTB level and solvation contributions by COSMO-RS using the CENSO free energy ranking algorithm. Various density functional approximations in combination with three London dispersion correction schemes are assessed against “back-corrected” experimental association energies as well as accurate local coupled cluster reference values. Our protocol predicts association free energies with a mean absolute deviation of only 2 kcal mol⁻¹ from the measured values. Thus, it is well suited to generate reference association free energies for assessing theoretical methods on realistically sized supramolecular complexes or to support experimental chemists. For specifically evaluating methods for calculating gas-phase association energies, we recommend using the provided accurate coupled cluster reference values. We propose to use this set as an extension of the S30L benchmark set [Sure et al., J. Chem. Theory Comput., 2015, 11, 3785–3801] with a special focus on the challenging computation of non-covalent interactions of heavy main group elements.