Defocus correction for angle-resolved scatterometry using waveguide-based surface plasmon resonance.

• A method to correct the defocus effect of ARS is proposed. • A polymer planar waveguide is employed to calibrate the incident angle at different defocus positions. • A defocus compensation function was derived based on the bare silicon wafer. Angle-Resolved Scatterometry (ARS) enables the acquisition of images at the back focal plane (BFP) of the objective lens. These images contain target geometry information, which is a function of the incident angle and azimuthal angle. This technique has recently exhibited significant potential in film measurement, owing to its advantages of being fast, accurate, low-cost, and non-contact for critical dimensions (CD) measurement. However, during on-line CD metrology, environmental vibrations or sample stage displacement errors can cause the measurement plane to deviate from the objective lens's focal plane. Consequently, the recorded defocused BFP images lack crucial large-angle information, exacerbating the discrepancy between experimental and simulated data. In this study, we propose a method to address the defocus effect in ARS. To calibrate the incident angle under varying defocus positions, a polymer planar waveguide is employed. By illuminating the waveguide with polarized light, surface plasmon resonances are excited at a specific incident angle. At this resonance angle, the reflectance becomes exceptionally low, resulting in the formation of a distinct dark ring at the BFP. Subsequently, we capture BFP images of the bare silicon under different defocus positions, which serve as a reference light source for reflectance normalization. This enables effective compensation for the corresponding spectral errors associated with defocus. The experimental results demonstrate that for the SiO 2 film with a reference thickness of 30.5 nm, the maximum measurement error reduces significantly (at −25 µm defocus position) from 8.4 nm to 2.4 nm. Across the entire defocus range, the measurement error is consistently controlled within 3.4 nm. [ABSTRACT FROM AUTHOR]