周天宇, 杨开勇, 吴素勇. 多层高反膜光学常数和厚度的反演[J]. 应用光学, 2011, 32(1): 128-132.
引用本文: 周天宇, 杨开勇, 吴素勇. 多层高反膜光学常数和厚度的反演[J]. 应用光学, 2011, 32(1): 128-132.
ZHOU Tian-yu, YANG Kai-yong, WU Su-yong. ADetermination of optical constants and thicknesses of high-reflection multilayer system[J]. Journal of Applied Optics, 2011, 32(1): 128-132.
Citation: ZHOU Tian-yu, YANG Kai-yong, WU Su-yong. ADetermination of optical constants and thicknesses of high-reflection multilayer system[J]. Journal of Applied Optics, 2011, 32(1): 128-132.

多层高反膜光学常数和厚度的反演

ADetermination of optical constants and thicknesses of high-reflection multilayer system

  • 摘要: 基于精英保留策略的遗传算法具有良好的全局寻优能力和很强的鲁棒性,基于Levenberg-Marquardt算法的非线性最小二乘法具有快速、高效的局部收敛能力。研究了将这两种算法结合起来求解多层膜光学常数和厚度的可行性和有效性。结果表明:基于两种算法的组合反演算法综合了这两种算法的优点,能以较大的概率、较高的精度找到在实际约束条件下多层膜的光学常数和厚度,特别是在寻找全局最优解方面表现出色。根据全光谱透射拟合法基本原理,应用组合反演算法分别在考虑测量误差与不考虑测量误差两种情况下反演15层规整高反膜的光学常数和厚度,得到了较为精确的薄膜参数。模拟实验表明:此组合反演算法可有效地求解多层膜反演问题,在反演多层膜光学常数和厚度方面具有实际的应用价值。

     

    Abstract: Genetic algorithm based on elite selection strategy has good ability of global optimization and robustness. The non-linear least square method with Levenberg-Marquardt algorithm has fast and efficient local convergence features. The feasibility and effectiveness of reverse engineering of optical constants and thicknesses by the combination of both algorithms were studied. The results indicate that the combined fitting method can be used to calculate the optical constants and thicknesses of multilayer with great feasibility and efficiency under the actual restrictions, especially in finding the global optimum solution. At the same time, based on the theory of the whole spectrum fitting, the combined fitting method was applied to calculate the optical constants and thicknesses of 15-layer high-reflection multilayer and accurate thin films' parameters were obtained. The fitting experiments show that the combined fitting method is practical in reverse engineering of optical constants and thicknesses of multilayer, and it can be used in optical reverse engineering.

     

/

返回文章
返回