ON THE INNOCENCE OF BIPYRIDINE LIGANDS: HOW WELL DO DFT FUNCTIONALS FARE FOR THESE CHALLENGING SPIN SYSTEMS?

The electronic structures of a number of iron, cobalt, vanadium, and titanium complexes with the 2,2′-bipyridine (bpy) ligand were considered using the multireference CASSCF and NEVPT2 methods. Many of these systems have been studied in the past using B3LYP and were found to contain the bpy ligand as a radical anion. For many of the cases, this contradicts both our chemical intuition and our multireference calculations. While there are instances where the ligand is indeed a radical anion resulting from electron transfer from the metal centre, in many cases it remains neutral and is involved in π-backbonding from the metal centre.

Since there are many cases where CASSCF is too costly, a number of DFT functionals, including the newer double-hybrid functionals, were evaluated against the CASSCF data. It was found that nonhybrid functionals, especially those containing the kinetic energy density (i.e., meta-GGA functionals), were the best at predicting the electronic nature of the complexes. The τ-HCTH and HCTH functionals were the top performers, correctly predicting all eleven out of eleven test cases and with the lowest mean unsigned errors (MUE, 7.6 and 7.8 kcal/mol, respectively); the M06-L, N12, BLYP, PBE, and TPSS functionals also did well, while B3LYP had significant shortcomings.