Some problems arising from two sublattice modelling of ordered phases

Although two (or effectively two) sublattice modelling of ordered phases has been the norm in carrying out phase diagram assessments, it is argued that this kind of modelling should, in general, be discouraged for these phases. Two-sublattice modelling gives a Gibbs energy minimum (with reference states at the same temperature and with the same parent structure as the ordered phase) for every ordered phase, irrespective of its composition. In doing so, it fails to take into account that at most ordered phase compositions there are only kinks and not minima in the Gibbs energy-composition curve with respect to these reference states. Kinks can be obtained if the ordered phases are modelled with more than two sublattices. Because of the incorrect shape of the Gibbs energy-composition curve when two-sublattice modelling is used, there is an excessive but unnecessary demand on the sublattice L parameter terms in the excess Gibbs energy for the fitting of experimental results. This often leads to unreliable extrapolations of the Gibbs energy-composition curves.