Dominant-pair free energies predict phase selection in high-entropy alloys

Abstract

Phase selection in multicomponent alloys is governed by the competition between entropic stabilization of disordered solutions and enthalpic driving forces for chemical ordering. However, widely used parametric criteria reduce it to a single scalar, carrying no explicit free energy for any competing ordered phase. Herein, we develop a thermodynamic framework based on the semi-empirical macroscopic atom model and the Dinsdale lattice stability database to fill this gap. We show that a dominant-pair mechanism, in which the Al-transition-metal interaction family dominates the ordering enthalpy, enables the complex multicomponent B2-ordering problem to be reduced to an effective pseudo-binary system with an analytically evaluated Bragg-Williams free energy. Combined with a minimum-free-energy classifier, the framework predicts the lowest-energy phase as a function of composition and temperature. This provides continuous phase stability maps rather than the single-value predictions of conventional descriptors. Demonstrated on high-entropy alloys using a dataset of 269 experimentally characterized samples, the model outperforms widely used phase-selection criteria in the class-balanced macro-F1 metric and achieves 77.9% on the well-posed three-class task, outperforming the valence electron concentration criterion. The model is general by construction and computationally efficient for predicting phase stability in multicomponent alloys over a broad range of compositions and temperatures.