Assessment of standard force field models against high-quality Ab Initio potential curves for prototypes of π-π, CH/π, and SH/π interactions

Several popular force fields, namely, CHARMM, AMBER, OPLS-AA, and MM3, have been tested for their ability to reproduce highly accurate quantum mechanical potential energy curves for noncovalent interactions in the benzene dimer, the benzene-CH₄ complex, and the benzene-H₂S complex. All of the force fields are semi-quantitatively correct, but none of them, is consistently reliable quantitatively. Re-optimization of Lennard-Jones parameters and symmetry-adapted perturbation theory analysis for the benzene dimer suggests that better agreement cannot be expected unless more flexible functional forms (particularly for the electrostatic contributions) are employed for the empirical force fields.