基于机器学习的一维材料功函数预测研究  

Prediction of Work Function for One-Dimensional Materials Based on Machine Learning

作　　者：胡泽宇 卢峰[1] Hu Zeyu;Lu Feng(College of Electronic Information and Optical Engineering,Nankai University,Tianjin 300350,China)

机构地区：[1]南开大学电子信息与光学工程学院,天津300350

出　　处：《南开大学学报（自然科学版）》2024年第2期79-87,共9页Acta Scientiarum Naturalium Universitatis Nankaiensis

基　　金：国家自然科学基金(51871121)。

摘　　要：以三元一维原子线的高通量计算为基础,基于机器学习方法,建立了一套可以有效预测一维原子线电子结构性质的程序.以功函数预测为例,基于晶格常数、键长、元素共价半径及其电负性等特征参数,结合特征工程和机器学习的方法,建立了有效的回归模型.结果表明,此模型程序可以实现对一维原子线材料电子结构性质的合理预测,从而加快新材料设计和开发.Based on the high-throughput calculation of ternary one-dimensional AWs,a set of programs has been developed that can effectively predict the electronic structure properties of one-dimensional AWs based on machine learning methods.Taking the work function as an example,the parameters of lattice constant,bond length,atomic covalent radius and electronegativity could be used as the input features.Meanwhile,the feature engineering and machine learning methods could be used to build regression models.The results show that the prediction of the electronic structure properties of one-dimensional AW materials can be realized through these programs proposed.Thus,the process of new material design and development can be accelerated.

关 键 词：机器学习 功函数 回归拟合 低维材料 一维原子线 

分 类 号：O469[理学—凝聚态物理]