Deep Learning-based Study Of The Work Function Of Boron-doped Graphene

Work function,as a fundamental property characterizing graphene’s electron exchange capability with the external environment,holds crucial significance.Particularly,in the realm of two-dimensional hybrid materials,the work function of boron-doped graphene assumes a pivotal role in determining its prospective applications.The introduction of boron atoms induces hole states within graphene,thereby exerting a significant influence on its electrical and optical characteristics,concomitantly altering its work function.Hence,investigating the work function of boron-doped graphene assumes paramount importance in comprehending its intrinsic properties and potential applications.Employing density functional theory,this study delves into the structural aspects of boron-doped graphene,subsequently establishing a deep learning model to capture the intricate non-linear relationship between the work function and doping concentration.By optimizing this model,a precise correlation between the work function of boron-doped graphene and the doping concentration is ascertained.This investigation offers valuable insights into the design and fabrication of doped graphene,thereby presenting a guiding framework with promising implications for diverse domains,including electronic devices and energy systems.Moreover,the devised descriptor and deep learning model present a novel approach for predicting the performance of two-dimensional materials.Grounded in density functional theory and deep learning methodologies,this paper conducts an in-depth exploration of the relationship between the work function of boron-doped graphene and the doping concentration.The following presents a concise summary of the principal contributions and focal points of this research endeavor:In Chapter 1,starting from the research background of boron-doped graphene,briefly introduces the research results of domestic and foreign researchers on the work function of boron-doped graphene in recent years,and expounds the main research content of this paper.In Chapter 2,this chapter introduces the relevant theories and calculation methods used in this paper in detail,including machine learning,deep learning methods,density functional theory and density functional theory calculation software.In Chapter 3,the model of boron-doped graphene at different concentrations was constructed,and the theoretically stable structure was obtained through structural optimization.Formation energy and work function data were then calculated using density functional theory and boron-doped graphene features were extracted.On this basis,more than 30 000 samples and more than 115 000 symmetrical samples were extracted,and the characteristics of the samples are mainly composed of 0-1 two-dimensional matrices representing atomic types and position information.Next,the formation energy was predicted using KNN,SVM,and MLP machine learning models to evaluate the stability of boron-doped graphene.The results showed that the MSE and R~2 of MLP reached 0.2394 and 0.9286,respectively,which proved that machine learning can evaluate the stability of boron-doped graphene.In Chapter 4,using the improved ACN and VCN convolutional networks,the work function of boron-doped graphene was successfully predicted using more than 115,000symmetrical samples.Among the two convolutional networks,the MSE and R~2 values of VCN reached 0.0013 and 0.9546,respectively.The experimental results showed that both ACN and VCN can effectively predict the work function.This research achievement provides valuable reference for the study of other two-dimensional material properties.The experimental results proved that both ACN and VCN can effectively predict the work function.This research result provides a valuable reference for the research on the properties of other two-dimensional materials.In Chapter 5,summary and prospect,summarizes the main work and achievements of this paper and introduces the future work prospect,and also analyzes the shortcomings of this work.