Intelligent Selection Of Dielectric Function Models In Spectroscopic Ellipsometric Data Analysis

Spectroscopic ellipsometry is a kind of precision optical instrument that uses the principle of polarization optics to measure film samples.By analyzing the spectral ellipsometry data,useful film optical properties and film thickness can be obtained.In the traditional spectral ellipsometric data analysis process,the operator manually selects the dielectric function model based on his own optical experience,expertise requirements are high and the results of the analysis vary from person to person.This thesis builds on the problem of inter-distribution electric function model of spectral ellipsometric data analysis.Combined with artificial intelligence and related methods in machine learning,this paper proposes the research on the intelligent selection of the median electric function model for spectral ellipsometric data analysis.The main research contents of the thesis and The innovations are as follows:(1)The basic flow and principle of ellipsometric data analysis are systematically studied,and the optical characteristics modeling(including multi-layer film reflection/transmission characteristic model and dielectric function model)and the parameter extraction method based on nonlinear regression are described in the ellipsometric data analysis process.(2)The traditional ellipsometric data analysis has the following problems that the choice of dielectric function model depends heavily on experience,prior knowledge and manual operation are complex,and the analysis results vary from person to person.In view of the above problems,this paper proposes a dielectric based on support vector machine.The automatic recognition method of function model reduces the influence of human factors in modeling and ensures the consistency of analysis results.(3)An evaluation method of ellipsometric data analysis model based on information theory criterion is proposed.Considering the fitting effect and complexity of the model in ellipsometric data analysis,the ellipsometric data analysis model is more reasonablely judged and selected.(4)The ellipsometric data analysis mediation electric function model intelligent selection system is developed,which includes intelligent classification of dispersive oscillator,generation of dielectric function model set and model selection.The experimental verification is carried out for the developed system.The results show that the intelligent selection system of the dielectric function model can obtain a model with high fitting precision and relatively simple model structure.