Advanced industrial S/TEM automation and metrology: Boundary of precision

摘要

Developments in the semiconductor industry are driving the need for new methods to characterize smaller 3D devices in a productive and reproducible way. The automation of sample preparation, TEM imaging, and offline CD metrology is able to provide sample information in the form of both images and quantitative data. In this article, we evaluate the TEM imaging automation workflow in order to optimize the experimental configuration towards better measurement precision and higher throughput. It is found that the top contributor to CD precision is the signal-to-noise ratio of the STEM image, which is determined by the electron flux. We investigated the top 5 most important experimental factors (probe current, image size, dwell time, Drift Corrected Frame integration, and image Field of View) and their interactions for a secondary contributor to CD precision. And we found that the combination of those factors play very minor role as soon as they contribute to the same electron flux. This learning guides us to configure our experiment parameters to optimize the trade-off between measurement precision and throughput.