Predicting Band Gaps of 2D Phosphorene-Isoelectronic materials Using Machine Learning

Abstract

It also opened new possibilities of being used in electronic and optoelectronic ways in future since the two dimensional (2D) materials have been discovered. Among these materials, the phosphorene and their isoelectric counterparts are those materials that attracted the attention of people due to their adjustable electronic properties and high quality bandgap properties. However, the conventional methods of calculation such as Density Functional Theory (DFT) physically determining the gap is slow and expensive, particularly with large material libraries being screened. The framework is a machine-learning model that makes accurate and efficient prediction which bandgaps of 2D phosphorene-isoelectronic materials (PINs) are possible. A set of structural, electronic, and physicochemical properties of various 2D materials was used to develop and test various machine learning techniques, especially Random Forest Regression, Support Vector Regression (SVR), and Gradient Boosting algorithms. To determine the most important material properties that affect the bandgap values, feature engineering and feature selection techniques were implemented. The accuracy of the machine learning models developed was assessed based on commonly used regression tests, such as Mean Absolute Error (MAE), Root Mean Squared Error (RMSE), and coefficient of determination (R²). A set of statistical measures used to evaluate the prediction accuracy, reliability and generalizability of the models. The results obtained in the present work show the machine learning methods are very effective and can predict the materials bandgap properties with a very good accuracy. Moreover, the proposed models are able to substantially decrease the computational effort and time needed for traditional first-principles calculations, enabling practical applications in materials screening and discovery for larger systems. The work in this study demonstrates the power of data-driven strategies to unlock the rapid identification and design of 2D semiconductors with engineered electronic properties. The primary areas of interest include: 2D Materials, Phosphorene-Isoelectronic Materials, Bandgap Prediction, Machine Learning, Density Functional Theory (DFT), Materials Informatics and Semiconductor Design.