Decoding the Complex Short-Range Orders in Semiconductor GeSn Alloys: Insights from Accurate and Efficient Machine Learning-Based Atomistic Simulations*

GeSn alloys have emerged as versatile materials with significant potential for electronic, photonic, and topological quantum applications, yet their structural properties remain elusive and challenging to determine. By employing extensive first-principles density functional theory (DFT) calculations and statistical sampling, our prior study unveiled the presence of short-range order (SRO) behaviors across the entire investigated composition range of GeSn alloys¹, and SRO was further predicted to substantially affect their electronic properties¹. Recent experiments using various characterization techniques have confirmed its existence^2,3, but important questions remain regarding the actual structures, spatial domain size and its distribution, and corresponding changes in the properties of SRO. To bridge the spatiotemporal scale gap and facilitate a direct comparison with advanced methods like atom probe tomography, we develop machine-learning interatomic potentials for these alloys. We show that the machine-learning interatomic potentials not only can accurately reproduce the results based on DFT calculations, but also enable an interesting discovery of subtle coexistence of two distinct types of SROs in GeSn alloys, similar to SiGeSn alloys⁴. Our results shed more light on the intricate SRO structural properties of Group IV alloys.

1. ACS Appl. Mater. Interfaces 12, 57245 (2020)

2. Small Methods 6, 2200029 (2022)

3. APL 122, 062103 (2023)

4. Communications Materials 3, 66 (2022)