Autocorrelation-based beam shape optimization in scanning electron microscopy
Many automatically focusing algorithms for scanning electron microscopes exist. Typically, these optimize a single derived metric (such as image sharpness) without considering the intrinsic beam shape. We describe an algorithm to automatically focus and stigmate scanning electron microscopes based on the autocorrelation of captured images. In the first part of this work, we develop the theoretical basis of using the autocorrelation to approximate the current beam shape. In the second part, we use this to condition the beam shape on various surfaces, using Bayesian optimization to tune both working distance and, in particular, stigmation. Through a comparison of the full width at half maximum of the autocorrelation arrays before and after optimization and a visual comparison of images taken of the sample surface before and after optimization, we show that the algorithm is able to successfully optimize the scanning parameters. Overall, our algorithm can effectively focus on a variety of topographies at different magnifications.