Quality Estimation of SEM Wafer Images

This paper describes a practical method for image quality estimation. In particular, this method applies to SEM images of VLSI wafers. Wafers are extremely sensitive to electron charging, since they may contain nonconducting materials. This forces utilization of low energy, low-current electron beams and short viewing times. The images generated under these SEM settings are noisy. An automatic inspection machine must use these noisy images as input to a machine vision algorithm in order to navigate on the wafer, locate the line to be measured, and measure it's width. During algorithm development, we encountered the need for an objective measure which can be used to evaluate SEM images. The development and application such a measure for SEM image quality is described herein. This measure is simple, and facilitates the objective estimation of SEM image quality. It is utilized for evaluation of SEM operating parameters, in order to establish optimal conditions to produce sufficient image quality without damaging the specimen. Experimental results presented show that this measure reflects the same changes SNR (signal to noise ratio) is expected to show following a change in SEM operational parameters.