The structures and vibrational properties of trans-[V(N 2 ) 2 (PH 3 ) 4 ] - , trans-[Cr(N 2 ) 2 (PH 3 ) 4 ], [Mn(H)(N 2 )(PH 3 ) 4 ], [Fe(N 2 )(PH 3 ) 4 ], [Fe(H)(N 2 )(PH 3 ) 4 ] + and [FeCl(N 2 )(PH 3 ) 4 ] + have been computed using density functional theory. Good reproduction of metal ligand bond lengths and the trend in N-N stretching frequencies ν(N-N) is obtained showing that simple PH 3 is a good model for the more complicated phosphine ligands employed experimentally. Analysis of the theoretical M N binding energies shows a good correlation between increasing bond strength and decreasing ν(N-N). trans-[V(N 2 ) 2 (PH 3 ) 4 ] - has the lowest value of ν(N-N) (~1740 cm - 1 ) and the largest calculated M N 2 bond energy (223 kJ mol - 1 ) while [Fe(H)(N 2 )(PH 3 ) 4 ] + has the highest value of ν(N-N) (~2100 cm - 1 ) and the lowest computed M-N 2 bond energy (126 kJ mol - 1 ). The biggest discrepancy between theory and experiment is for trans-[V(N 2 ) 2 (PH 3 ) 4 ] - . The error is removed by explicitly modelling solvation effects and the ion-pair interactions with alkali metals which are vital for stabilising dinitrogenvanadates(-1). The strong V-N 2 bond is apparently at odds with the reported lability of dinitrogenvanadate(-1) complexes. However, this assumes that the lability is reversible. The modelling suggests that N 2 loss is accompanied by decomposition.