Prediction of stable structures of two-dimensional (2D) materials that have not been experimentally realized is an important pre-requisite step for the development of these materials for various technological uses without the extensive trial-error experiments. Traditional methods such as density functional theory (DFT) can be used to find the energies of structures computationally, however, calculating the energies of the total number of structures possible would be daunting timewise as well. We propose using machine learning methods to reduce the search time for 2D materials’ geometric structures. Our case study for this process consists of hexagonal graphene-like boron–carbon–nitrogen (h-BCN). Our dataset consists of 300 randomly generated h-BCN structures optimized using DFT based calculations. A pattern recognition scheme based on nearest neighbors was developed to characterize the atomic environment of the atoms. We will be exploring the use of k-means clustering models to accurately predict the total energy of an h-BCN structure in order to find the lowest energy geometric structure.
Rahman, Talat S.
Bachelor of Science (B.S.)
College of Sciences
Length of Campus-only Access
Joshi, Sonali, "Predicting Structures of 2D Materials Enabled by Machine Learning" (2020). Honors Undergraduate Theses. 722.