Developing Highly Efficient First-Principles Neuroevolution Potentials to Understand Short-Range Order and Properties in Complex Alloy Systems—From Group-IV Alloys to High-Entropy Materials

Alloying plays a critical role in enhancing the properties and functionalities of various materials, such as Group-IV alloys, III-V alloys, medium-entropy materials, and high-entropy materials (HEMs). Group-IV alloys, known for their silicon compatibility, hold promise for electronic, photonic, and topological quantum applications. Similarly, III-nitride semiconductors are essential in optoelectronics and electronics, incorporating new functionalities such as ferroelectricity, ferromagnetism, and superconductivity to advance next-generation semiconductor and quantum technologies. HEMs, including high-entropy and complex concentrated alloys, high-entropy oxides/nitrides, and high-entropy metallic glasses, offer diverse applications due to their customizable properties. However, understanding the complex behavior of these alloy systems, particularly their short-range order (SRO), poses significant challenges. Accurate modeling with density functional theory (DFT) calculations is often limited by spatiotemporal constraints. To address this limitation, we develop highly accurate and efficient machine-learning interatomic potentials based on a neuroevolution potential approach, incorporating new insights into high-quality DFT dataset preparation. Our first-principles neuroevolution potentials (NEPs) help deepen the understanding of SRO and properties in complex alloy systems, from Group-IV alloys to high-entropy materials. The developed NEPs demonstrate superior accuracy and efficiency, elucidating intricate SRO behavior and their impacts on material properties. By bridging the gap between atomistic modeling and advanced characterization techniques such as APT and 4D-STEM, our approach offers promising pathways for advancing the fundamental understanding of complex alloy systems, marking a significant step forward in materials science.<br/><br/>*This work was supported as part of the μ-ATOMS, an Energy Frontier Research Center funded by the U.S. Department of Energy, Office of Science, Basic Energy Sciences under award DE-SC0023412.