The ERC Advanced Project, "BALANCE," is fundamental research into intermolecular forces that are ubiquitous in chemistry and physics. They have been difficult to calculate accurately, and experimental methods had been limited to small molecular systems where the effects were also small. We proposed to investigate the intermolecular forces in molecules large enough that they would be representative for catalysts, reagents, and materials. We designed gas-phase "molecular torsion balances," which are molecules whose structures or conformational equilibria measure one force (poorly characterized) against another (well characterized), just as a balance measures an unknown weight against a set of standard weights in the macroscopic world. The readout of the balance is done by gas-phase laser spectroscopy using a technique called Cryogenic Ion Vibrational Predissociation (CIVP).
We found in the course of the project several indications that the current implementations of dispersion-corrected DFT overestimated non-covalent interactions significantly. Both in measurements of absolute binding energies, as well as in relative binding energies calibrated against other interactions--the molecular torsion balances--we found systematic overbinding which resulted in incorrect conformational predictions. Having solved a number of spectroscopic problems associated with CIVP--systems with Fermi resonances and perturbations due to the tag--and having added ion mobility measurements as yet another, structure-sensitive, experimental readout, we could ascertain that the systematic overbinding is very widespread. Expressing the result in an (over)generalization, one could say that our experiments show that, for systems simple enough that a skilled chemist can predict structure by chemical intuition, the dispersion-corrected DFT calculations can also predict a correct structure. As one systematically steps up the structural complexity of the system, one reaches a point where chemical intuition no longer suffices for a reliable prediction, but it is at this same point that dispersion-corrected DFT fails as well, calling into question the utility of the commonly used methods. Lastly, based upon some non-ERC funded results, the BALANCE project included a new experiment which showed that, in contrast to the previous results, non-covalent interactions between pi systems oriented face-to-face are, in fact, corrected modeled by dispersion-corrected DFT. An analysis of the reasons why some systems are corrected modeled, and others not, indicates that the dispersion corrections fail (sometimes) because the interaction is anisotropic in reality, but isotropic in the model. This calls for a more sophisticated approach from theory.
BALANCE provides the missing experimental data to test treatments of noncovalent interactions which are ubiquitous in organic and organometallic chemistry. As the dispersion-corrected DFT methods are used to design catalysts, materials, and even pharmaceutical active ingredients, the project contributes to the method development in societally relevant areas.
