Azimuth-guided image fusion method for sub-λ/18 defect inspection using a laser scattering darkfield imaging system

The continuous scaling of advanced process nodes imposes challenges on semiconductor manufacturing, while driving the demand for the development of high-sensitivity and high-throughput defect inspection methods. Although darkfield imaging technology offers high efficiency, systems operating in the visible spectrum struggle to achieve nanoscale sensitivity due to signal pollution caused by surface roughness and electronic noises. In this study, a rigorous physical model of defect imaging is constructed based on vectorial electromagnetic theory, systematically elucidating the mechanisms affecting defect detectability. Based on the strong dependence of defect scattering characteristics on illumination azimuth, an azimuth-guided image fusion strategy is proposed to enhance the detection capability of randomly oriented subwavelength defects. By incorporating the fusion method into an optimally configured laser scattering darkfield imaging system, we achieved the effective identification of sub-λ/18 wide defects on large-area wafers. This work overcomes the sensitivity limitation of darkfield imaging in the visible spectrum, providing an efficient solution for nanoscale defect inspection at advanced nodes.