Through-focus scanning optical microscopy (TSOM) is an economical and nondestructive method for measuring three-dimensional nanostructures. After obtaining a TSOM image, a library-matching method is typically used to interpret optical intensity information and determine the dimensions of a measurement target. To further improve dimensional measurement accuracy, this paper proposes a machine learning method that extracts texture information from TSOM images. The method extracts feature vectors of TSOM images in terms of the Gray-level Co-occurrence Matrix (GLCM), Local Binary Pattern (LBP), and Histogram of Oriented Gradient (HOG). We tested models trained with these vectors in isolation, in pairs, and a combination of all three to test seven possible feature vectors. Once normalized, these feature vectors were then used to train and test three machine-learning regression models: random forest, GBDT, and AdaBoost. Compared with the results of the library-matching method, the measurement accuracy of the machine learning method is considerably higher. When detecting dimensional features that fall into a wide range of sizes, the AdaBoost model used with the combined LBP and HOG feature vectors performs better than the others. For detecting dimensional features within a narrower range of sizes, the AdaBoost model combined with HOG feature extraction algorithm performs better.

Through-focus scanning optical microscopy (TSOM) is a high-efficient, low-costed, and nondestructive model-based optical nanoscale method with the capability of measuring semiconductor targets from nanometer to micrometer level. However, some instability issues resulted from lateral movement of the target and angular illuminating non-uniformity during the collection of through-focus (TF) images restrict TSOM’s potential applications so that considerable efforts are needed to align optical elements before the collection and correct the experimental TSOM image before differentiating the experimental TSOM image from simulated TSOM image. An improved corrected TSOM method using Fourier transform is herein presented in this paper. First, a series of experimental TF images are collected through scanning the objective of the optical microscopy, and the ideally simulated TF images are obtained by a full-vector formulation. Then, each experimental image is aligned to its corresponding simulated counterpart before constructing the TSOM image. Based on the analysis of precision and repeatability, this method demonstrates its capability to improve the performance of TSOM, and the promising possibilities in application of online and in-machine measurements.

Through-focus scanning optical microscopy (TSOM) is one of the recommended measurement methods in semiconductor manufacturing industry in recent years because of its rapid and nondestructive properties. As a computational imaging method, TSOM takes full advantage of the information from defocused images rather than only concentrating on focused images. In order to improve the accuracy of TSOM in nanoscale dimensional measurement, this paper proposes a two-input deep-learning TSOM method based on Convolutional Neural Network (CNN). The TSOM image and the focused image are taken as the two inputs of the network. The TSOM image is processed by three columns convolutional channels and the focused image is processed by a single convolution channel for feature extraction. Then, the features extracted from the two kinds of images are merged and mapped to the measuring parameters for output. Our method makes effective use of the image information collected by TSOM system, for which the measurement process is fast and convenient with high accuracy. The MSE of the method can reach 5.18 nm² in the measurement of gold lines with a linewidth range of 247–1010 nm and the measuring accuracy is much higher than other deep-learning TSOM methods.

Unlike the optical information taken from a single in-focus image of general optical microscopy, through-focus scanning optical microscopy (TSOM) involves scanning a target through the focus and capturing of a series of images. These images can be used to conduct three-dimensional inspection and metrology with nanometer-scale lateral and vertical sensitivity. The sensitivity of TSOM strongly depends on many mechanical and optical factors. In this study, how illumination polarization and target structure affect the sensitivity of TSOM is analyzed. Firstly, the complete imaging procedure of the polarized light is investigated. Secondly, through-focus scanning results of different targets with two illumination polarizations are simulated using the finite-difference time-domain method. Thirdly, a few experiments are performed to verify the influence of illumination polarization and target structures on the sensitivity of TSOM. Both the results of the simulation and experiments illustrate an apparent influence of polarization on the sensitivity of inspecting the targets with center asymmetric structures. For enhanced sensitivity, illumination polarization should be perpendicular to the target texture. This conclusion is meaningful to adjust illumination polarization purposefully for different structure characteristics and improve the sensitivity of metrology.

The measurement accuracy of a point diffraction interferometer is influenced by the quality of the diffraction wavefront generated by a nanometer-scale pinhole. This paper analyzes the diffraction process of a pinhole illuminated by deep ultraviolet converging light with a large numerical aperture (NA) based on vector diffraction theory. Key metrics, including the diffraction NA and the wavefront aberrations relative to an ideal spherical wavefront, are examined in relation to the polarization state, pinhole diameter, and Cr thickness. Additionally, the impact of alignment errors is investigated. The results indicate that for a converging beam with and , the optimal polarization state is linear polarization, and the diffraction wavefront most closely approximates an ideal spherical wavefront when the pinhole diameter is 200 nm and the Cr thickness is 100 nm. Alignment errors alter the magnitude of the original central-symmetric aberration components and introduce asymmetric components such as coma.