Deep learning-assisted methods for accelerating the intelligent screening of novel 2D materials: New perspectives focusing on data collection and description

Abstract

Since the isolation of graphene, the interest in two-dimensional (2D) materials has been steadily growing thanks to their unique chemical and physical properties, as well as their potential for various applications. Deep learning (DL), currently one of the most sophisticated machine learning (ML) models, is emerging as a highly effective tool for intelligently investigating and screening 2D materials. The utilization of abundant data sources, appropriate descriptors, and neural networks enables the prediction of the structural and physicochemical properties of undiscovered 2D materials based on DL. Specifically, high-quality and well-described data plays a crucial role in effective model training, accurate predictions, and the discovery of new 2D materials. It also promotes reproducibility, collaboration, and continuous improvement within this field. This tutorial review is dedicated to an examination of the characterization, prediction, and discovery of 2D materials facilitated by various DL techniques. It focuses on the perspective of data collection and description, aiming to provide a clearer understanding of underlying principles and predicting outcomes. In addition, it also offers insights into future research prospects. The growing acceptance of DL is set to accelerate and transform the study of 2D materials.