Methodological perspective demonstrates the necessity of modeling dynamic atomic environments in complex semiconductors, highlighting machine-learned potentials for accurate prediction.
Key Points
To outline computational strategies for incorporating realistic physical phenomena—such as thermal motion, disorder, and defects—into atomistic simulations of complex semiconductors.
Perspectival synthesis analyzing theoretical approaches that bridge idealized crystal models and realistic disordered semiconductor configurations.
Evaluation of machine-learned interatomic potentials for sampling structural configurations prior to electronic structure calculations.
Demonstrated that incorporating thermal motion, structural disorder, charge localization, defects, and interfaces substantially alters the physical interpretation of semiconductor electronic properties.
Identified machine-learned interatomic potentials as an effective tool to overcome sampling limitations and access realistic atomic configurations for subsequent electronic structure calculations.