Thin film characterization by learning-assisted multi-angle polarized microscopy.

ABSTRACT

Thin film characterization is a necessary step in the semiconductor industry and nanodevice fabrication. In this work, we report a learning-assisted method to conduct the measurement based on a multi-angle polarized microscopy. By illuminating the film with a tightly focused vectorial beam with space-polarization nonseparability, the angle-dependent reflection coefficients are encoded into the reflected intensity distribution. The measurement is then transformed into an optimization problem aiming at minimizing the discrepancy between measured and simulated image features. The proposed approach is validated by numerical simulation and experimental measurements. As the method can be easily implemented with a conventional microscope, it provides a low cost solution to measure film parameters with a high spatial resolution and time efficiency.