A Broad Structural Search of Binary Precipitates via Active Learning

Understanding the structure of thermodynamically stable precipitates is of great interest in material science as they can affect the electrical conductivity and mechanical properties of the matrix to a great degree. In this work, we use a relaxation-on-the-fly active learning algorithm in order to scan all possible binary candidates, for different types and concentrations of alloy elements (mainly Cu, Si, and Ni). Quantum-mechanical calculations are performed on a small number of candidates to train and improve the machine-learned potential. The model is then used to predict the enthalpy of formation of all candidates. The stability of binary precipitates, based on predicting the convex hull, is further assessed by the phonon density of states analysis.