Statistical variability in mechanical properties of amorphous silica predicted by molecular dynamics

Abstract

This paper examines the statistical variation in the mechanical behavior of amorphous silica using molecular dynamics (MD) simulations. The variability arises from random initial atomic arrangements in replicate samples. Reactive molecular dynamics (R-MD) with the ReaxFF–SiO force field was used to simulate 100 samples, each prepared via a melt-quench process to ensure unique atomic configurations, but identical radial distribution functions. Uniaxial tension simulations on uncracked and pre-cracked samples were carried out to predict properties such as elastic modulus, strength, strain to failure, and fracture energy. Results show that initial atomic configurations significantly influence mechanical properties, introducing statistical variation that follows a normal distribution, consistent with the central limit theorem. The normal distribution of strength suggests that amorphous silica exhibits ductile-like failure at the nanoscale, despite brittle behavior macroscopically. Identifying the mean and standard deviation of these distributions enables quantification of variability in MD predictions, enhancing the reliability and understanding of such simulations.