Normal incident optical reflectance spectroscopy for thin-film thickness measurement with genetic algorithm

Abstract

Thin films are widely used in applications such as corrosion prevention, semiconductor insulation, and photoelectric conversion, where precise thickness measurement is critical for quality control and performance optimization. Reflectance spectroscopy is attracting interest as a non-contact, non-destructive, and efficient method for industrial thin film thickness measurements. This paper proposed a normal incident optical reflectance spectroscopy to enhance the lateral resolution, signal-to-noise ratio, and decrease the incident angle range, thereby capturing a higher proportion of reflected light. The system integrates a white light lamp, fiber Y-bundles, a self-designed reflectance probe, and a spectrometer. Genetic algorithm (GA) is employed to extract the optimal film thickness by minimizing the discrepancy between the theoretical and measured reflectance spectra. Reflectance spectra of the silica film samples on silicon substrates were acquired to match well with those from the angular resolution spectrometer. The thickness deviations between the proposed system and an elliptical polarimeter were less than 2 nm, with a spectral fitting error below ±0.5 %. The system achieves a measurement stability of 0.018 nm, and a processing time of 0.37 s per point. Furthermore, the thickness distribution of a patterned film sample was successfully mapped, validating the system’s capability for in-situ thin film thickness measurements.