Silicon Carbide Epitaxial Layer Thickness Measurement Based on Infrared Interferometry Method and Simulated Annealing-Least Squares Hybrid Optimization Algorithm
Download as PDF
DOI: 10.25236/iwmecs.2025.018
Author(s)
Pengyu Zhu, Zidong Yi
Corresponding Author
Pengyu Zhu
Abstract
As a third-generation semiconductor material, silicon carbide (SiC) epitaxial layer thickness is a key parameter to measure material quality and device performance. In order to achieve accurate and non-destructive measurement of its thickness, a systematic measurement and optimization model was constructed based on infrared interferometry. In the first step, a simplified dual-beam interferometric physical model is established. The model considers the single reflection and transmission of light waves at the interface between the epitaxial layer and the substrate, and inverts the thickness of the epitaxial layer by analyzing the interference fringes in the reflection spectrum. In order to accurately describe the optical properties of materials, the Sellmeier equation and the Drude model are introduced to describe the dispersion effect of silicon carbide refractive index with wavelength and carrier concentration, and the theoretical calculation formula of epitaxial layer thickness is derived by combining Snell's law and interference conditions. In the second step, based on the model of the first step, a thickness calculation and optimization algorithm combining least squares method and simulated annealing algorithm is designed. The algorithm optimizes the thickness value by minimizing the error function between the theoretical and measured reflectances, and the simulated annealing algorithm is used to avoid the local optimal solution to ensure the global nature of the solution. In order to deal with the randomness of s-polarized and P-polarized light in actual measurements, the average reflection coefficient of the two is used in the calculation. The algorithm was applied to process the experimental data, and the calculated average thickness of the bright pattern was 9.93 μm and 9.79 μm, the average thickness of the dark pattern was 11.29 μm and 10.91 μm, and the thickness difference was 12.75% and 10.84%, respectively, when the incidence angle was 10° and 15°, the results were in good agreement, which verified the effectiveness and accuracy of the model.
Keywords
Infrared Interferometry; Silicon Carbide; Thickness of Epitaxial Layer; Dual-Beam Interference; Least Squares Method; Simulated Annealing Algorithm