Three-dimensional Structure Recognition of Circuit Patterns on Semiconductor Devices Using Multiple SEM Images Detected in Different Electron Scattering Angles

We propose a method for recognizing the three-dimensional structure of circuit patterns to achieve automatic pattern width measurement of semiconductor devices using scanning electron microscope (SEM). Pattern measurement requires pattern recognition technique to identify the measurement position from an SEM image. However, pattern width and brightness values on an image fluctuate according to patterning conditions, type of material, etc., so distinguishing between line and space through image matching based on brightness information is difficult. In view of this problem, we investigated a line and space discrimination method that focuses on fundamentally different three-dimensional structures (concavity and convexity) and considering that this difference between concavity and convexity appears as a difference in the scattering angles of secondary electrons, we captured two images corresponding to these two scattering angles. In this way, we were able to design image feature that portray the difference between line and space from these two images and perform high-speed and easy capture of concavity and convexity information. In experiments, the proposed method achieved a 100% accuracy rate in automatic critical-dimension measurements.