The review addresses one of the key challenges in applying artificial intelligence to microstructure characterization: the availability of reliable ground-truth data. While AI has demonstrated considerable potential for automating microstructure analysis, its performance ultimately depends on the quality of the datasets used for training. For complex microstructures, generating these datasets remains difficult because manual annotation is both time-consuming and subject to expert interpretation.

The article explores how correlative microscopy can overcome these limitations by combining complementary characterization techniques, including optical microscopy, scanning electron microscopy (SEM), and electron backscatter diffraction (EBSD). Through several case studies, the review demonstrates how these methods can be integrated to create robust and reliable ground-truth datasets for AI applications.

A central concept of this work is to leverage correlative microscopy as a one-time investment for generating high-quality training data. Once an AI model has been trained on reliable ground truth, routine microstructure characterization can be performed using simpler and faster microscopy techniques, significantly improving efficiency without compromising reliability.

This review reflects our ongoing research interests in microstructure characterization, correlative microscopy, and the integration of artificial intelligence into materials science workflows. It also highlights the importance of combining advanced characterization methods with data-driven approaches to enable the next generation of automated materials analysis.

You can find the article here: https://link.springer.com/article/10.1007/s13632-026-01379-3.

For decades, materials science education has largely followed a bottom-up approach: starting with crystal structures, phase diagrams, and dislocation theory before eventually arriving at engineering applications.

But what if we reversed the process?

This is exactly the idea behind MatGlobe EDU, the educational materials database currently being developed by Matplus GmbH. Inspired by Michael Ashby’s top-down design philosophy, students begin with a real engineering challenge and then identify, evaluate, and model suitable materials to solve it. This approach connects theory directly with practical decision-making from the very beginning.

I’m excited that we are among the early adopters working with MatGlobe EDU to develop new teaching formats and case studies.

Last week, we had the opportunity to participate in a workshop with other early adopters and the Matplus team. Together, we discussed the current state of the database and its accompanying wiki, shared experiences, and contributed to shaping the roadmap for its integration into university teaching.

As part of the visit, Martin Müller delivered a seminar entitled “Advancing AI-Driven Microstructure Analysis through Correlative Microscopy Approaches,” highlighting the role of correlative microscopy in enabling machine learning-based microstructure analysis.

The scientific exchange was highly engaging and inspiring, with many stimulating discussions that opened up exciting perspectives for future collaboration. We look forward to the next steps and would like to sincerely thank Prof. Raabe and his team for their warm hospitality.

This internal funding scheme supports promising research projects that lay the foundation for future third-party funding applications. We are grateful for the opportunity to develop an exciting new research direction: “Correlation Between Multiphase Material Microstructures and Local Micromechanical Properties Using Artificial Intelligence.”

Our goal is to better understand how complex microstructures influence local mechanical behavior—and to leverage artificial intelligence to uncover relationships that are difficult to identify using conventional analysis alone.

The project officially kicks off in September 2026, and I’m looking forward to sharing our progress, challenges, and key findings over the coming months.