A global-search method which applies the concept of genetic algorithm (GA) with gradient-based optimizer is proposed for the problem of experimental data analysis from spectroscopic ellipsometry on thin films. The method is applied to evaluate the data obtained for samples with different structure complexity, starting with transparent monolayers (SiO₂, HfO₂) on a substrate, through absorbing film (diamond-like carbon) and multilayer structures. We demonstrate that by using this method we are able to find material parameters even for limited a priori knowledge about the sample properties, where classical methods fail.

中文翻译：

---

混合GA梯度法在薄膜椭偏数据评估中的应用

针对薄膜的椭圆偏振光谱法对实验数据进行分析的问题，提出了一种基于梯度优化器的遗传算法概念的全局搜索方法。该方法适用于评估具有不同结构复杂性的样品的数据，该数据从基材上的透明单层（SiO ₂，HfO ₂）开始，通过吸收膜（类金刚石碳）和多层结构。我们证明了通过使用这种方法，即使对于传统方法失败的关于样品特性的有限先验知识，我们也能够找到材料参数。