Research On Anisotropic Growth Of Two-dimensional Materials Based On Machine Learning And Monte Carlo Method

Two dimensional transition metal dichalcogenide(TMDCs)have received widespread attention in recent years due to their unique properties,such as the transition from indirect bandgap to direct bandgap when diluted into a single layer,and valley-dependent photoluminescence.In addition,as a semiconductor with considerable mobility,these materials have also found suitable applications in next-generation electronic and optoelectronic devices that depend on atomic layer thickness.In addition,as a semiconductor with considerable mobility,these materials have also found suitable applications in next-generation electronic and optoelectronic devices that depend on atomic thickness.However,it is a huge challenge to produce a large area of 2D TMDCs with controllable direction and localized growth.At present,most synthetic methods are based on optimization of experimental parameters,and the growth mechanism behind it is not clear.With the rise of kinetic Monte Carlo(k MC)and the development of machine learning,materials science increasingly uses computational methods to deal with various complex problems in its technical field,which not only provides new methods for two-dimensional materials research,but also puts forward more and higher technical requirements for two-dimensional material research.Therefore,it is of great significance to combine k MC and machine learning to explore the mechanism behind the growth of two-dimensional materials.In this paper,we will explore the growth mechanism of WS₂ from chemical vapor deposition(CVD)preparation,k MC simulation and machine learning model.The growth preparation mainly uses sulfur powder and tungsten trioxide to react in the reaction furnace by CVD,and the effective single-layer film can be prepared,which is consistent with the previously reported single-layer WS₂characteristics.In the k MC simulation part,the model is designed according to the related parameters obtained in the experiment and the kinetic parameters are calculated by first principles.The key factors such as substrate adsorption and surface diffusion are also included in the model.By comparing and analyzing the simulation results of different growth conditions,it is preliminarily concluded that the ratio of metal atoms to sulfur atoms is the main reason affecting anisotropic growth and growth area polarization.In the part of machine learning,the learning task is divided into classification task and regression task.The classification task predicts the types of anisotropic growth and isotropic growth,under growth and over growth by analyzing the results of k MC simulation.The main task of regression is to complete the regression calculation of anisotropic growth degree.After training the model,certain progress has been made,including anisotropic prediction model with 91%accuracy and isotropic prediction model with accuracy of 90%,undergrowth prediction model with accuracy of 98%and overgrowth prediction model with accuracy of 99%.In addition,based on the preliminary growth mechanism,the growth trend surface is simulated by the least square method,and the anisotropic growth surface equation is further analyzed and calculated.This may be the key to the synthesis of large area or extremely anisotropic single-layer WS₂ films.This paper successfully designed a basic framework combining k MC and machine learning to study the anisotropic growth mechanism of WS₂.The feature of this paper is to use k MC simulation to accumulate statistics on the atomic level of the dynamic process,which helps quantify these different experimental conditions to better analyze the mechanism.At the same time,the machine learning model is used to predict its growth,reduce unnecessary simulations to reduce the total simulation time,and similarly can reduce the cost and time required for the experiment.In addition,this framework is also applicable to the research of other TMDCs materials and even two-dimensional materials,so it has great significance for the research of two-dimensional materials.