Identification and reconstruction of diffraction structures in optical scatterometry using support vector machine method

Abstract

A library search is a widely used method for the reconstruction of diffraction structures in optical scatterometry. In a library search, if the actual geometrical model of a measured signature is different from the model used in the establishment of a library, the search result will be meaningless. Therefore, the identification of the geometrical profile for a measured signature is critical. In addition, fast searching of the library is essential to find a best-matched signature even though the library may have huge amounts of data. The authors propose a support vector machine (SVM)-based method to deal with these issues. First, an SVM classifier is trained to identify the geometrical profile of a diffraction structure from its measured signature, and then another set of several SVM classifiers are trained to map the measured signature into a sublibrary to accelerate the search process. Simulations and experiments have demonstrated that the SVM classifier can identify the geometrical profile of one-dimensional trapezoidal gratings accurately, and the SVM-based library search strategy can achieve a fast and robust extraction of parameters for diffraction structures. © The Authors. Published by SPIE under a Creative Commons Attribution 3.0 Unported License. Distribution or reproduction of this work ih whole or in part requires full attribution ot the original publication, including its DOI.

Development of a broadband Mueller matrix ellipsometer as a powerful tool for nanostructure metrology

“…A weighted least-squares regression analysis (Levenberg-Marquardt algorithm) [10,11] or library search [12][13][14] is then performed, during which the measurands are varied until the calculated and measured data match as much close as possible. The solution of measurandsx can be achieved by minimizing a weighted mean-square error function χ r 2 defined by…”

“…The reconstruction of the structural profile from the measured signature is a typical inverse diffraction problem with an objective of finding a profile whose theoretical signature can best match the measured one. The solution of the inverse problem usually employs two kinds of methods [9], namely the nonlinear regression method [10,11] and the library search method [12][13][14]. Both of these two approaches involve the establishment of a theoretical diffraction model that relates the optical signatures and the structural profiles associated with these signatures.…”

Dependence-Analysis-Based Data-Refinement in Optical Scatterometry for Fast Nanostructure Reconstruction

“…There are many reliable forward-modeling techniques such as the rigorous coupled-wave analysis (RCWA), the finite element method (FEM), the boundary element method (BEM), or the finite-difference time-domain (FDTD) method. The profile 2 of 12 reconstruction process conducted under a fixed measurement configuration is an inverse problem with the objective of optimizing a set of floating profile parameters (e.g., CD, sidewall angle, and height) whose theoretical signatures can best match the measured ones through regression analysis or library search [12][13][14]. The measurement configuration is defined as a combination of specially selected wavelengths and incident and azimuthal angles.…”

Automatic inspection of a residual resist layer by means of self-organizing map

“…In 2008, we demonstrated the possibility of identifying profile geometry by using a classical MLP from the ellipsometric signature 15 used in the characterization process. Other works 16 have employed another classification method, such as support vector machines, combined with the library search to solve the inverse problem. In terms of results, this classification method shows performances similar or complementary to the ANN method.…”