Least popular vibrational entropy model provides the best accuracy and robustness

Vibrational contributions into free energies usually amount to several kcal/mol and can significantly affect computational predictions. However, they are generally estimated incorrectly for chemical systems in solutions because the usual (employed in ∼99% of cases) models of vibrational entropies are extremely sensitive to errors in low-lying frequencies (below 300 cm-1), and these low-lying frequencies involve solvent molecules that are usually neglected (computed implicitly) in quantum chemical calculations. We find that only one vibrational entropy approximation-the one proposed by Truhlar in 2011-which is used in only ∼2% of cases, is stable in the low-lying frequency region and does not exhibit this problem. Accordingly, this approximation shows the best accuracy and robustness on a diverse set of experimental complexation energies and can be somewhat improved even further.