Diradicals, transition metal complexes, and reactions which break chemical bonds are all very challenging (sometimes even impossible) for current quantum chemical methods to model. Why?
Standard electronic structure models are built around the Hartree-Fock molecular orbital approximation, which assumes that the electronic wavefunction is well-described by an antisymmetrized product of orbitals (i.e., a Slater determinant). This approximation generally works well if there is only one important electron configuration for the system of interest. However, imagine that a sigma bond is broken between two fragments, A and B. At the dissociation limit when A and B are infinitely far apart, the bonding sigma orbital becomes degenerate with the antibonding sigma* orbital. That means that two electron configurations, ...(sigma)2, and ...(sigma*)2, become degenerate. Since Hartree-Fock theory has only one of these two configurations, it only has half of the wavefunction. This can cause serious problems ... such as MP2 energies which go to negative infinity! Clearly, the reliability of quantum chemical models needs to be assessed for such problems, and new methods must be developed when they are found lacking.
Situations requiring more than one important electron configuration, such as the bond breaking example above, are technically described as having important nondynamical correlation. Diradicals, which can be considered as molecules with a ``permanently broken bond,'' are certainly problem cases. Transition metals, particularly in the first transition row, are also prone to having two or more nearly degenerate electron configurations. For any of these chemical systems, the usual quantum chemical models may fail utterly. Even worse, a quantum chemistry program may make predictions for such systems without any warnings at all that anything is amiss.
Our research is centered upon developing a better fundamental understanding of how and when current approaches go wrong for bond breaking and other nondynamical correlation problems, and upon creating new approaches to describe these challenging problems more reliably or more efficiently.
