The measurement of line-edge roughness (LER) has recently become a major topic of concern in the semiconductor industry. This paper proposed a methodology method to measure LER using atomic force microscopy (AFM). Pay attention to the 3-D imaging of AFM, an image analysis algorithm detecting the line edge is presented. The code has been developed using MATLAB, which is able to calculate the amplitude parameters of LER above from measured data. We used this method to deal with the experiment data and analyzed the dependence of the amplitude of LER. After then, a same sample is measured by ordinary probe, ultrasharp probe and carbon nanotube probe. Analysis and comparison of measurement results using established algorithm were made. Then, as the characterization of LER is not only a simple geometry feature, but also is a wide-band including the spatial complexity of the edge, the spatial frequency analysis of the detected edges using the power spectral density function is necessary. For the self-affinity edge roughness, a characterization of LER based on the fractal theory is briefly described. The analysis of experiment data using nanotube probe demonstrated this method can completely characterize LER. Finally, the problem in the study is thoroughly investigated with interesting conclusions.
The line-edges of the sample scanned by AFM is detected using cellular neural networks. Line-width, Line-width roughness and line edge roughness of line-structure are calculated respectively based on the analysis of the detected edge character. Since cellular neural network is characterized by high-speed parallel computation and easy to be implemented in hardware, it has more potential, comparing with other software technique, in the quick line-structure-parameters detection.