Abstract
Atomistic tight-binding based simulations are widely used to study transition metal alloys properties. However, they still require to be improved if one aims at modelling segregation and ordering phenomena in the case of magnetic materials, since they generally rely on local charge neutrality rules per site, valence orbital and element, but not per spin! We propose here a strategy to overcome this difficulty, that we illustrate in two magnetic systems of particular interest: CoPt and FeNi alloys.