Abstract
This paper describes a new approach for the detecting pattern edge positions and roughness of LSI ultra-fine patterns. As LSI pattern lines become narrower, exact estimation of line edge roughness and line width roughness as well as edge positions, becomes indispensable. The proposed approach comprises the function approximation of cross-sectional signals at each edge in a SEM image and the automatic determination of the optimum edge detection region used in the function approximation. The approach enables us to detect edge positions and roughness with high accuracy and high reproducibility.
