Realizing smart scanning transmission electron microscopy using high performance computing.

This article focuses on the integration of Scanning Transmission Electron Microscopy (STEM) combined with Electron Energy Loss Spectroscopy (EELS) and machine learning (ML) workflows through a Python-based server software that enables remote control and high-performance computing (HPC) interaction. The software acts as a wrapper over Gatan, Inc.'s DigitalMicrograph platform, widely used in STEM instruments, facilitating tasks such as image acquisition, beam positioning, and EELS data collection with demonstrated workflows including object detection, deep convolutional neural networks (DCNNs), and deep kernel learning (DKL). The integration addresses challenges related to data acquisition speed and computational demands, enabling more efficient and targeted data collection strategies like active learning, which can significantly reduce experimental time and resource use. The open-source code and example notebooks are available on GitHub, aiming to enhance the capabilities of approximately 70% of STEM microscopes globally that use DigitalMicrograph software.