Study of line edge roughness using continuous wavelet transform for 65-nm node

This paper introduces the continuous wavelet transform (CWT) techniques to characterize spatial frequencies of LER. A 890 nm length of line pattern was dissected with 448 measured-points along line-edge from the image of scanning electron microscope (SEM), and the dissection of measurement points is around 2 nm. The measured data of line-edge roughness (LER) were transformed to spatial power spectrum with commercial software packages of wavelet transform, and the characterization of spatial frequency correlated to lithographic process parameters, such as the soft-bake (SB) temperature, the numerical aperture (NA), the temperature of post-exposure baking (PEB), and the molecular weight of resist (MW) were investigated. The higher NA and lower SB give a significant improvement from low spatial frequency (long range LER) to higher one (short range LER). However, both the higher temperature of PEB and lower MW improve edge roughness only on long range order roughness (lower spatial frequency).