Optimization of accurate SEM imaging by use of artificial images

Today the vast majority of the scanning electron microscopes (SEMs) are incapable of taking repeatable and accurate images at high magnifications. Geometric distortions are common, so are drift, vibration, and problems related to disturbing electro-magnetic fields, contamination. These issues tend to degrade the quality of the image. Hence, in many cases it is not the focusing ability of the electron optical column, but these factors that limit the achievable resolution, repeatability and accuracy. This is a significant issue for nanometer-scale measurements, because the errors are many times greater than the measured distances. However, there are new image acquisition techniques that could improve the accuracy and repeatability of such measurements. One of these techniques is being developed at National Institute of Standards and Technology (NIST). This technique is based on cross-correlation combined with frequency filtering. Because the power this technique strongly depends on many conditions, like shape and periodicity of the sample features, noise, frequencies of the vibrations, etc., it needs to be properly evaluated and limits of usability of the technique should be specified. For this, a statistically significant number of SEM images varying in all of these parameters is necessary and the parameters must be known. Unfortunately, this is impossible to achieve using an SEM, because most of these parameters are random and the instrument parameters are unknown. A possible solution is to use the artificial SEM images, which can simulate these effects repeatably and deterministically. It is not an issue to generate a large number of such images in a reasonable time. The artificial image generator was developed at NIST originally for the evaluation of resolution-calculation techniques, but it is also very usable for assessment of new imaging techniques too. The new version of the generator is being developed at NIST. It allows for modeling different types of samples and several new effects, which also allow for taking into account the different characteristics of different SEMs. This paper describes, how the artificial images are built and how they may be used to improve the new SEM imaging techniques.