stemOrchestrator: Enabling Seamless Hardware Control and High-Throughput Workflows on Electron Microscopes

Abstract

Scanning Transmission Electron Microscopy (STEM) is one of the most powerful tools for materials characterization, providing access to atomic-scale structure via direct imaging, chemical composition, orbital populations, vibrational properties and quasiparticles via spectral methods, and crystallographic information through diffraction. However, these diverse functionalities are often supported by hardware components from different manufacturers, creating challenges in seamless operation and integration of detectors, holders, and cameras on a single STEM column. As the field moves toward machine learning (ML) enabled experiments and autonomous discovery, the need for combined control across these hardware interfaces becomes critical. This paper develops stemOrchestrator, a software framework which combines detector- and camera specific Application Programming Interfaces (APIs) in a cohesive platform for controlling various STEM hardware modules and developing sophisticated automated workflows. We illustrate the performance of stemOrchestrator using several model workflows including high-throughput particle characterization, hardware tuning using Bayesian Optimization (BO), cross correlation-based drift correction with informative logging of hardware status. Importantly, these workflows are presented not as isolated technical advances, but to highlight how stemOrchestrator renders their implementation almost trivial by abstracting hardware heterogeneity and execution logic. This framework also enables seamless integration with LLM (Large language model) agents to suggest and run complex automated workflows and opens pathway for orchestration of self-driving labs and geographically distributed instrumentation networks. The codes are available at this link for trying and contributing: https://github.com/pycroscopy/pyAutoMic/tree/main/TEM/stemOrchestrator