RIS models have two main objectives: (1) to describe the reaction of a system submitted to unusual driving forces, such as point defect concentration gradients; and (2) to reproduce the atomic diffusion enhance­ment induced by an increase of the local point defect concentration. In comparison with the thermal aging situation, gradients of point defect chemical potential are nonnegligible. The L-coefficients are considered as variables that vary with nonequilibrium point defect concentrations. With L-coefficients varying in time, such models do not satisfy the TIP hypothe­sis. Instead, the authors of the first publica- tions11,30,102 considered new quantities, the so-called

 

[25]

 

where ZVa is defined in terms of the local equilibrium vacancy concentration (see eqn [14]). Flux of B is deduced from the flux of A by exchanging the letters A and B. In a multicomponent alloy, equivalent kinetic equations are provided by Perks’s model.10 In this model, point defect concentrations are assumed to be independent of chemical concentra­tions: in other words, parameters ZVa and ZVB are set to zero. Most of the RIS models are derived from Perks’ model although they neglect the cross-coeffi — cients.5 Flux of species i is assumed to be independent of the concentration gradients of the other species. In doing so, not only the kinetic couplings, but also some of the thermodynamic couplings are ignored. Indeed, as shown in eqn [9], a chemical potential gradient is a function of all the concentration gradients. An atomic flux results from a balance between the so-called IK effect, atomic fluxes induced by point defect concentration gradients (first term of the RHS of eqn [24]), and the so-called Kirkendall (K) effect