摘要
基于对光谱测量数据中系统误差和随机误差的不同特性分析,提出了两种方法来分别减小其对薄膜光学参数反演测量精度的影响。方法一是利用薄膜光谱系数关于薄膜光学参数的一阶偏导数信息筛选数据点,选取偏导数符号相反的波段或入射角区域作为最佳测量数据点采集区域,以最小化系统误差引起的薄膜光学参数相对真实值的偏差大小。方法二是在实测光谱数据中注入独立同分布的随机噪声,利用基于偏差函数最小化方法的混合优化算法多次进行反演计算,将每次反演得到的薄膜光学参数估计值的统计均值作为薄膜真实光学参数的估计值,以减小甚至消除随机误差引起的薄膜光学参数不确定度。这两种方法不仅在提高薄膜光学参数测量精度上具有明显的实用价值,而且物理意义明确,可操作性强,不局限于特定波段和特定材料的薄膜光学参数测量,具有良好的普适性,可望在薄膜光学参数的准确测量技术中发挥重要作用。
Based on different characteristics analysis of the systematic and random errors found in the spectral measurement data, two methods are put forward to diminish their impacts on the reverse determination accuracy of thin films＇ respectively optical parameters. The first method uses spectral coefficient＇s first order partial derivatives of thin films with respect to layer optical parameters to select measurement data. The spectral region or incident angle region characterized with opposite operation symbol of partial derivatives is chosen as the optimal measurement data region to minimize the deviations from the real optical parameters of thin films caused by systematic measurement errors. The second method injects independent random noise of same distribution to the actual measurement data. The data with noise are utilized as the real spectral coefficients of the thin film being measured in the reverse determination for many times. The reverse determination algorithm used is a mixed optimization method based on the discrepancy function minimization technique. The statistics mean values of all fit optical parameters obtained each time by the mixed optimization method are chosen as estimates of the real optical parameters of the measured thin film to diminish or eliminate the uncertainty of optical parameters caused by random measurement errors. These two methods are of obviously practical value to improve the determination accuracy of thins films＇ optical parameters. And they also characterize with concise and definite physical meaning, expedient manipulation ability and good universality, not restricted to specific spectral regions or specific materials. They are of promising importance in the accurate reverse determination techniques of thins films＇ optical parameters.
出处
《中国激光》
EI
CAS
CSCD
北大核心
2009年第8期2171-2177,共7页
Chinese Journal of Lasers
关键词
薄膜光学
光学参数反演不确定度
偏导数筛选
噪声注入
光谱测量误差
遗传算法
thin films optics
reverse determination uncertainty of optical parameters
selection by partial derivatives analysis
noise injection
spectral measurement errors
genetic algorithm
