On the nature of the oxide ion in nitrate melts