Discovering High-Entropy Oxides with a Machine-Learning Interatomic Potential

TL;DR

This paper introduces a machine-learning interatomic potential to efficiently explore and identify stable high-entropy oxides, significantly accelerating materials discovery in complex chemistries.

Contribution

It presents a novel integrated computational-experimental approach using machine learning to map high-entropy oxide stability across composition space.

Findings

Accurately predicts stability of known high-entropy oxides.

Identifies dozens of new potential stable compositions.

Uses bond length distribution and mixing enthalpy as descriptors.

Abstract

High-entropy materials shift the traditional materials discovery paradigm to one that leverages disorder, enabling access to unique chemistries unreachable through enthalpy alone. We present a self-consistent approach integrating computation and experiment to understand and explore single-phase rock salt high-entropy oxides. By leveraging a machine-learning interatomic potential, we rapidly and accurately map high-entropy composition space using our two descriptors: bond length distribution and mixing enthalpy. The single-phase stabilities for all experimentally stabilized rock salt compositions are correctly resolved, with dozens more compositions awaiting discovery.