Predicting the mechanical properties of borophene by artificial neural networks

Abstract

A material of great relevance in the current research context is borophene, a monolayer nanomaterial composed only of boron atoms with extraordinary electrical and mechanical properties. In the present work, a neural network was designed and trained in order to predict the mechanical properties of this material, such as Young’s modulus, fracture strength and fracture strain. The training data set was obtained through molecular dynamics simulations, with different parameter scenarios in order to analyze the effects of temperature, strain rate and strain direction. The trained machine learning model succeeded in predicting the material’s behavior with high accuracy. Its results reflect a decrease in yield stress with increasing temperature and a slight improvement in the fracture strain with increasing strain rates, as expected. Additionally, a web application with a graphical interface was developed, which uses the trained model, in order to make this tool available to any user. This interface also makes it possible to visualize the approximate stress-strain curve, drawn based on the resulting fracture stress and strain.