High-throughput full-field thickness mapping of semiconductor wafers via hyperspectral imaging with a field-modulated...

Thin-film thickness is a key parameter governing optical, mechanical, and electronic properties. Due to the limitations of discrete sampling, traditional point-scanning methods fail to meet the requirements for real-time characterization of full-size spatial uniformity in mass industrial production. In this paper, we propose a high-throughput, full-field online thickness monitoring method based on hyperspectral imaging (HSI). A high-uniformity linear illumination system was designed, tailored to the imaging characteristics of the hyperspectral camera, to achieve stable detection in a large field of view. We found that in large-field detection, the measured reflectance spectrum exhibits a "spectral drift" phenomenon at the interference minima. This distortion is primarily caused by substrate-induced diffuse scattering, as the dielectric thin film is highly transparent and its surface scattering is negligible. To overcome this physical bottleneck, we investigated the diffuse reflection mechanism specifically at the film-substrate interface and established a physical model coupling specular reflection and field-modulated diffuse scattering. By introducing an effective interface roughness coefficient, the background noise caused by substrate scattering was successfully decoupled. Combined with the least-squares nonlinear fitting algorithm (LSNFA), the system achieves accurate extraction of film thickness by introducing an effective roughness parameter for spectral baseline correction. Experimental results show that the HSI system can complete the full-size thickness mapping of a 4-inch wafer (containing more than 120,000 sampling points) within 60 seconds, with an absolute error of less than 1 nm compared with standard ellipsometry. This work resolves the signal shift problem in HSI-based film thickness measurement and provides a precise metrology solution with both physical depth and detection efficiency for large-area film uniformity characterization in industrial environments.