We present the results of an LDA and LDA+U band structure study of the monoclinic and the corundum phases of V2O3 and argue that the most prominent (spin 1/2) models used to describe the semiconductor metal transition are not valid. Contrary to the generally accepted assumptions we find that the large on site Coulomb and exchange interactions result in a total local spin of 1 rather than 1/2 and especially an orbital occupation which removes the orbital degeneracies and the freedom for orbital ordering. The calculated exchange interaction parameters lead to a magnetic structure consistent with experiment again without the need of orbital ordering. While the low-temperature monoclinic distortion of the corundum crystal structure produces a very small effect on electronic structure of v2o3, the change of magnetic order leads to drastic differences in band widths and band gaps. The low temperature monoclinic phase clearly favors the experimentally observed magnetic structure, but calculations for corundum crystal structure gave two consistent sets of exchange interaction parameters with nearly degenerate total energies suggesting a kind of frustration in the paramagnetic phase. These results strongly suggest that the phase transitions in V2O3 which is so often quoted as the example of a S=1/2 Mott Hubbard system have a different origin. So back to the drawing board!
Download