Determining the ultimate resolution of scanning electron microscope-based unbiased roughness measurements. I. Simulating noise
| Autor: | Chris A. Mack, Gian Francesco Lorusso |
|---|---|
| Rok vydání: | 2019 |
| Předmět: |
Materials science
Pixel Scanning electron microscope business.industry Process Chemistry and Technology Surface finish Real image Quantitative Biology::Other Signal Grayscale Computer Science::Computers and Society Surfaces Coatings and Films Electronic Optical and Magnetic Materials Metrology Noise Optics Computer Science::Computer Vision and Pattern Recognition Materials Chemistry Electrical and Electronic Engineering business Instrumentation |
| Zdroj: | Journal of Vacuum Science & Technology B. 37:062903 |
| ISSN: | 2166-2754 2166-2746 |
| DOI: | 10.1116/1.5122758 |
| Popis: | Measuring line-edge roughness in a top-down scanning electron microscope (SEM) is complicated by noise in the SEM image, which biases the measured roughness. When either the roughness is small or the noise is large, it can become very difficult to separate noise from roughness to produce an unbiased estimate of the feature roughness. Synthetic SEM images with known roughness and noise properties can be used to explore the ultimate limits of SEM-based roughness metrology, but only if the noise in the synthetic images mimics the noise behavior of real images. By carefully analyzing the properties of experimental SEM images as a function of the number of frames of averaging (which directly modulates SEM noise), a noise model is developed. This model uses a Gamma distribution for the grayscale noise and then scales the image so that no more than 0.3% of the pixels are pegged at the maximum grayscale value of 255. The resulting synthetic SEM images mimic experimental SEM images in both signal and noise and will serve as a valuable tool for studying roughness metrology.Measuring line-edge roughness in a top-down scanning electron microscope (SEM) is complicated by noise in the SEM image, which biases the measured roughness. When either the roughness is small or the noise is large, it can become very difficult to separate noise from roughness to produce an unbiased estimate of the feature roughness. Synthetic SEM images with known roughness and noise properties can be used to explore the ultimate limits of SEM-based roughness metrology, but only if the noise in the synthetic images mimics the noise behavior of real images. By carefully analyzing the properties of experimental SEM images as a function of the number of frames of averaging (which directly modulates SEM noise), a noise model is developed. This model uses a Gamma distribution for the grayscale noise and then scales the image so that no more than 0.3% of the pixels are pegged at the maximum grayscale value of 255. The resulting synthetic SEM images mimic experimental SEM images in both signal and noise and wil... |
| Databáze: | OpenAIRE |
| Externí odkaz: |
|