The knowledge of the phenomenological coefficients L, including their dependence on the chemical compo­sition, allows the prediction of RIS phenomena. Unfor­tunately, in practice, it is very difficult to get such information from experimental measurements, espe­cially for concentrated and multicomponent alloys, and for the diffusion by interstitials. As we have seen, it is also quite difficult to establish the exact relation­ship between the phenomenological coefficients and the atomic jump frequencies because of the compli­cated way in which they depend on the local atomic configurations and because correlation effects are very difficult to be fully taken into account in diffusion theories. An alternative approach to analytical diffusion equations, then, is to integrate point defect jump mechanisms, with a realistic description ofthe frequen­cies in the complex energetic landscape of the alloy, in atomistic-scale simulations such as mean-field equa­tions, or Monte Carlo simulations (molecular dynam­ics methods are much too slow — by several orders of magnitude — for microstructure evolution governed by thermally activated migration of point defects).

Atomic-scale methods are appropriate techniques to simulate nanoscale phenomena like RIS. They are all based on an atomic jump frequency model. From this point of view, the difficulties are the same as for the modeling of other diffusive phase transformations (such as precipitation or ordering during thermal aging), complicated by the point defect formation and annihilation mechanisms and by the self-interstitial jump mechanisms, which are usually more complex than the vacancy ones.76