机器学习增强椭偏仪测量

利用人工智能技术发明了一种新型的全自动椭偏分析系统。该系统首先利用仿真数据进行训练，以学习椭偏参数和（n, κ, d）的基本对应关系；其次，由于仿真数据与实验数据存在一定差别，导致系统模型精度不够。为此，他们提出了一种深度神经网络驱动的迭代学习框架，实现了椭偏逆问题的高精度全自动求解。我们在实验中利用金属、半导体和电介质等几十种材料薄膜，成功验证了该技术的可行性与可靠性，且精度明显高于传统技术。该方法可与现有的椭偏测量技术兼容，为实现材料特性的自动、快速、高通量椭偏表征铺平了道路。数据量1.72MB。A novel fully automated ellipsometry analysis system was developed using artificial intelligence technologies. This system first utilizes simulated data for training to learn the basic correspondence between ellipsometric parameters and (n, κ, d). Secondly, certain discrepancies exist between simulated and experimental data, which leads to insufficient accuracy of the system model. To address this problem, they propose a deep neural network-driven iterative learning framework that enables high-precision fully automated solving of the ellipsometric inverse problem. In experiments, we used dozens of thin film materials including metals, semiconductors and dielectrics to successfully verify the feasibility and reliability of this technology, with its accuracy significantly higher than that of conventional techniques. This method is compatible with existing ellipsometry measurement technologies, paving the way for automatic, rapid and high-throughput ellipsometric characterization of material properties. The dataset size is 1.72 MB.

搜集汇总

背景与挑战

背景概述

该数据集基于人工智能技术开发了一套全自动椭偏分析系统，通过深度神经网络驱动的迭代学习框架实现高精度椭偏逆问题求解，已在多种材料薄膜上验证其可行性和可靠性，精度优于传统方法。