Ab initio valence bond calculations are performed for the three lowest states of the oxygen molecule ('Xi, *Ag, and '2;). One objective of the present study was to make a contribution to previous valence bond discussions about the oxygen "double" bond. Further, we study the origin of a small barrier in the potential energy surface of the ground state. Two compact models are employed to maintain the clear picture that can be offered by the valence bond method. The first model has only the Rumer structures that are essential for bonding and a proper dissociation. The second model, in addition, has structures which represent excited atoms. These prove to be important for the dissociation energies. For both models, the orbitals are fully optimized. The spectroscopic data obtained are significantly better than are the (few) valence bond results on 0, that have been published and have the quality of multiconfiguration self-consistent field calculations in which the same valence space is used. The "hump" in the potential energy surface of the ground state is shown to arise from a spin recoupling. The free atoms correspond to a spin coupling that is incapable of describing the formation of bonds. Only at short distances, an alternative spin coupling provides bonding and the repulsive curve is converted into an attractive one. Our results on this subject support a valence bond explanation previously given by McWeeny [R. McWeeny, Int. J. Quantum Chem. Symp. 24, 733 (1990)l. 0 1996 John Wiley & Sons, Inc.
VB method failed even qualitatively to account for the experimental facts. Wheland (1937) was the first who showed that the VB method is capable of providing the correct answer [3]. However, his lengthy VB analysis did not receive much acclaim since the molecular orbital (MO) approach offered an easier explanation. It took a long time before quantitative VB studies appeared. Moss et al. in 1975 presented results obtained using the general-