LER distribution measurement using HHCF-based noise correction

Line edge roughness (LER), which characterizes the sidewall roughness of line patterns, is used as a performance metric in lithography-related technologies. In current LER measurement practices, power spectral density (PSD) analysis is often used to evaluate roughness information based on the self-affine fractal model. However, due to the fundamental characteristics of PSD-based noise correction methods, only the mean LER over multiple patterns can be obtained. In this study, we propose a noise correction method based on the height-height correlation function (HHCF), thereby enabling the evaluation of roughness for individual line patterns and facilitating the measurement of the LER distribution. The proposed method was applied to silicon line/space patterns, where roughness parameters based on self-affine fractal model—namely, the standard deviation (σ), correlation length (ξ), and roughness exponent (α)—were evaluated for 1000 line edge profiles. Histogram results representing the distribution of these parameters were then obtained. The distributional information reflects the variations in the shapes of the fabricated patterns and constitutes critical data for process evaluation. The precise evaluation of the LER distribution, as demonstrated in this study, would be expected as a novel evaluation method for advanced process and yield control. Additionally, by utilizing this LER distribution information, it becomes feasible to estimate the required sample size to achieve a target mean σ within a specified confidence interval and confidence level, based on statistical analysis. This approach thereby provides high reliability and efficiency in LER measurement.