Harnessing Automated SEM-EDS and Machine Learning to Unlock High-Throughput Compositional Characterization of Powder Materials

<title>Abstract</title> Compositional characterization is essential for understanding and optimizing material performance. For powder-based materials, ubiquitous across diverse technologies, accurately resolving the composition of individual phases at high throughput remains an unsolved challenge. This bottleneck slows material research and is particularly acute for autonomous laboratories, where large numbers of samples are synthesized and AI decision-making algorithms demand low-uncertainty data. Here, we integrate measurement automation, data filtering, and unsupervised machine learning to deliver a fully automated framework for high-throughput, phase-resolved compositional analysis of as-deposited powders using scanning electron microscopy with energy-dispersive X-ray spectroscopy (SEM-EDS). Our approach, implemented in the Python-based software AutoEMXSp, overcomes geometry-induced artifacts from irregular particle morphologies and enables accurate compositional quantification using only one experimental standard per element—including light elements conventionally difficult to quantify in powders. Validated across 74 powder samples with elements spanning nitrogen to bismuth, AutoEMXSp consistently estimates compositions within 5–10% of their expected values and resolves minor or intermixed phases with high confidence. By delivering a robust, operator-free, high-throughput framework, this work paves the way for integration of powder compositional characterization into autonomous laboratory platforms, accelerating the pace of materials discovery and development.