Mapping Structural Heterogeneity at the Nanoscale with Scanning Nano-structure Electron Microscopy (SNEM)
Yevgeny Rakita,James L. Hart,Partha Pratim Das,Daniel L. Foley,Stavros Nicolopoulos,Sina Shahrezaei,Suveen Nigel Mathaudhu,Mitra L. Taheri,Simon J. L. Billinge
DOI: https://doi.org/10.1016/j.actamat.2022.118426
2022-08-26
Abstract:Here we explore the use of scanning electron diffraction coupled with electron atomic pair distribution function analysis (ePDF) to understand the local order as a function of position in a complex multicomponent system, a hot rolled, Ni-encapsulated, Zr$_{65}$Cu$_{17.5}$Ni$_{10}$Al$_{7.5}$ bulk metallic glass (BMG), with a spatial resolution of 3 nm. We show that it is possible to gain insight into the chemistry and chemical clustering/ordering tendency in different regions of the sample, including in the vicinity of nano-scale crystallites that are identified from virtual dark field images and in heavily deformed regions at the edge of the BMG. In addition to simpler analysis, unsupervised machine learning was used to extract partial PDFs from the material, modeled as a quasi-binary alloy, and map them in space. These maps allowed key insights not only into the local average composition, as validated by EELS, but also a unique insight into chemical short-range ordering tendencies in different regions of the sample during formation. The experiments are straightforward and rapid and, unlike spectroscopic measurements, don't require energy filters on the instrument. We spatially map different quantities of interest (QoI's), defined as scalars that can be computed directly from positions and widths of ePDF peaks or parameters refined from fits to the patterns. We developed a flexible and rapid data reduction and analysis software framework that allows experimenters to rapidly explore images of the sample on the basis of different QoI's. The power and flexibility of this approach are explored and described in detail. Because of the fact that we are getting spatially resolved images of the nanoscale structure obtained from ePDFs we call this approach scanning nano-structure electron microscopy (SNEM), and we believe that it will be powerful and useful extension of current 4D-STEM methods.
Materials Science,Disordered Systems and Neural Networks