Splitting multiple bonds: a comparison of methodologies on the accuracy of bond dissociation energies

A benchmarking of different quantum chemical methodologies for the splitting of multiply bonded systems is presented, with an emphasis on quantitative reproduction of experimentally determined dissociation energies. New benchmark full configuration interaction (FCI) calculations are presented for nitrogen and acetylene, and comparisons are made between various methods with both the FCI results and with experiment in an effort to understand qualitatively and quantitatively how well these different methods cope with the bond-breaking process. It is found that the multireference methods fare well, but are sometimes slowly convergent with respect to the size of the basis set, and in particular the angular momenta of the polarization functions. Single-reference (SR) wavefunction-based methods perform poorly against full CI, while CCSD(T) is reasonably accurate for comparison with experiment. Density functional theory with hybrid functionals have very quick basis-set convergence, and offer reliable estimates of bond dissociation energies. Spin-restricted SR wavefunctions are found to be poor for such systems, while unrestricted wavefunctions give reasonably good energies, but with severe spin contamination.