Machine learning-driven atomistic analysis of mechanical behavior in silicon nanowires

Abstract

This study investigates the modulus of elasticity of silicon nanowires using a combination of molecular dynamics simulations and machine learning techniques. The research presents a substantial dataset with over 3000 data points obtained from molecular dynamics simulations, which reveals detailed insights into the mechanical properties of silicon nanowires and underscores the importance of accurate model calibration. Machine learning surrogate models are employed to predict the elasticity of silicon nanowires, focusing on the influence of surface state and crystal orientation. By analyzing partial dependencies and using inverse pole figures, the study demonstrates that the modulus of elasticity exhibits significant orientation dependence. This work bridges computational and experimental approaches, offering a refined understanding of the mechanical behavior of silicon nanowires. The findings highlight the potential of integrating machine learning with atomistic simulations to improve the predictive accuracy of material properties, building the framework for advancements in nanoelectromechanical applications.