Predicting the Stability of an HEA: a First-Principles Analysis

In the quest of engineering new materials with superior strength and ductility, there has been an upsurge in the design and synthesis of multi-component alloys (known as high-entropy alloys) consisting of 4-5 elements in almost equi-atomic proportions. These alloys have shown significant solid-solution strengthening and excellent high-temperature properties, and have been found to exist primarily as a single phase, either in FCC or BCC structure. With the formation of a medley of innumerable configurations at the atomic scale, the complex interactions among the co-existing elements, the stability of HEA in such simple structural forms (FCC or BCC) is still enigmatic. Here, in this work, we use a combination of first-principles analysis, cluster expansion method and machine learning algorithms to understand the origin of the structural stability of a solid solution of 5 elements in a particular structure at the atomic scale, in terms of the relative energies of vast number of configurations. The energy of a configuration is then disintegrated into its elementary clusters, the number densities of which are regarded as the descriptors in our neural network. This network obtains the relative weights of each cluster in local stabilization or de-stabilization of an HEA.