Rapid evaluation method for anisotropic growth of WS2 monolayers by combining machine learning algorithms and kinetic Monte Carlo simulation data

Abstract

The growth of two-dimensional (2D) materials such as transition metal dichalcogenide (TMDC) monolayers is challenging and their growth mechanism has not been fully understood. It requires a large number of growth experiments and simulation calculations to reveal the growth mechanism. Also, a computational framework for combining the simulation results and the experimental results is essential. To speed up this time-consuming process for understanding the growth mechanism, a preliminary framework based on machine learning is herein proposed to analyze the simulation data and predict the growth trend. Specifically, the framework can be used for studying the anisotropic growth of tungsten disulfide (WS2) monolayers based on the data of kinetic Monte Carlo (kMC) simulations. Both an isotropic model with 91% accuracy and an anisotropic model with 90% accuracy are obtained. Moreover, both an undergrowth model with 98% accuracy and an overgrowth model with 99% accuracy are also obtained. This proposed framework can take both the experimental data and the simulation data as a single input data set, much speeding up the study of the growth mechanism.