Chemical bonding is perhaps the central concept to all of chemistry, and is certainly of great interest to theorists. Homolytic bond cleavage separates electron pairs, leading to strong electron correlation that is challenging to model. Chemical bond breaking thus is often an excellent regime to identify limitations of electronic structure approximations. I have often explored this aspect, using response properties as sensitive probes for errors in underlying electronic structure.
These investigations led me to a fuzzier but more fundamental question: When can we say a chemical bond is broken? After all, energy is generally a monotonically increasing function with increasing distance, with no well defined cutoffs. Lots of popular metrics are either not observables or only tangentially connected to electrons, which are central to bonds. So we proposed the static polarizability as the central quantity for defining bond breaking. Qualitatively, as a bond is stretched from equilibrium, the bonding electrons initially fill up the additional accessible volume, resulting in a more diffuse (and thus more polarizable) electron density. Post bond cleavage however, the electrons should start localizing on individual fragments, leading to a decrease in polarizability. Therefore the polarizability maximum along the dissociation coordinate represents the point of bond rupture.
The above argument can be placed on a firmer footing using the electron position spread tensor and excitation energies, which we describe in a recent work. We also empirically show the presence of perceptible polarizability maxima along the dissociation curves of a wide range of bonds. Further investigation would naturally still be needed for increasingly more exotic interactions. We do nonetheless hope this proposal stimulates interest, and recommend perusal of our work on the topic.
Our work on the limitations of various electronic structure methods in describing bond dissociations led to some other interesting discoveries, such as spurious inversion of bond polarities with non self-consistent double hybrid methods. These works are listed in the response properties section.
Diptarka Hait 