Seminar Speaker: Honggyu Kim, University of Florida

Unlocking Phase Complexity in Functional Materials Using 4D-STEM and Machine-Learning-Aided Digital Image Processing

Abstract

Four-dimensional scanning transmission electron microscopy (4D-STEM) has emerged as a powerful technique for characterizing crystal structures and uncovering previously inaccessible knowledge about novel and intricate material systems. This method creates a 4D dataset – a collection of two-dimensional (2D) diffraction patterns captured at every position of the electron probe scanned across a 2D region within the sample. It allows for mapping crucial material properties, such as crystal symmetry, orientation, strain, electric fields, and defects. A typical 4D dataset comprises a substantial number of 2D diffraction patterns, attainable in just a few tens of minutes using a fast electron detector system. This implies that a single microscopy session can yield millions of 2D diffraction patterns, posing a significant challenge in reliably and quantitatively analyzing structural properties embedded within extensive experimental data. As the field progresses swiftly in data quality and experimental efficiency, the methodology for data analysis becomes an integral part of the materials design and exploration process. This talk will initially introduce our recent technique for analyzing 4D-STEM data, employing digital image processing, dynamic electron diffraction simulation, and machine learning (ML) methods. We will then discuss our approach to investigating the complex microstructure of nanoscale fluorite-structured ferroelectric thin films, featuring the coexistence of multiple metastable phases with non-uniform grain size and random orientation. The results reveal the spatial distribution of multiple polymorphs with precise orientation information across a large field of view, enabling statistical analysis of microstructure characteristics and ultimately providing insights into how materials design impacts ferroelectric performance. The latter part of this talk will focus on demonstrating the clustering of diffraction patterns within a 4D dataset using an unsupervised ML method coupled with Cepstral analysis of electron diffraction patterns. This method is applied to quantify strain and microstructure in precipitation-hardened shape memory alloys. The results from these studies introduce a novel methodology for exploring the microscopic mechanisms through which nano and microscale phase competition controls material properties.

Biography

Honggyu Kim is an assistant professor in the Department of Materials Science and Engineering at the University of Florida (UF). He received his Ph.D. in the Department of Materials Science and Engineering at the University of Illinois at Urbana-Champaign, where he studied point defect and strain measurement in III-V semiconductor superlattices. Before joining the faculty at UF in Fall 2019, he was a postdoctoral researcher at the University of California, Santa Barbara, investigating vacancy characterization and phase evolution in functional oxides. His primary research focuses on the development and application of advanced transmission electron microscopy techniques to establish direct relationships between the structure and properties of materials on the atomic scale. His recent research topics include quantitative imaging of defect and domain structure in ferroelectric thin films, symmetry determination of quantum materials, and characterization of the phase transformation in novel metallic alloys, which are supported by NSF, DOE, DARPA, and UF. He is the recipient of a 2024 NSF CAREER award.