Apr 8, 2025
5:00pm - 7:00pm
Summit, Level 2, Flex Hall C
Carter Francis1,2,Barnaby Levin1,Paul Voyles2
Direct Electron1,University of Wisconsin-Madison2
Carter Francis1,2,Barnaby Levin1,Paul Voyles2
Direct Electron1,University of Wisconsin-Madison2
Four-dimensional scanning transmission electron microscopy (4D STEM) uses a pixelated detector to acquire a 2D electron diffraction pattern at each point of a 2D scan of a specimen. The resulting information-rich 4D datasets allow multiple types of analyses of a specimen, including spatially resolved mapping of crystal grain structure or molecular orientation. Recently developed, fast direct detectors such as the Celeritas XS (Direct Electron LP, San Diego, CA) allow for the practical acquisition of
in situ 4D STEM (also known as 5D STEM) datasets, which allow nanoscale changes in the local structure of a material to be characterized over time.
One area where 5D STEM offers critical new insights is the study of nucleation in metallic supercooled liquids. This is important for understanding the mechanics of solidification, controlling crystal morphology/structure, and guiding the development of more thermally stable bulk metallic glasses with lower critical cooling rates.
Here, we have used 5D STEM to continually probe the structure of a metallic supercooled liquid above the glass transition and though crystal nucleation and growth. A Celeritas XS detector was used to continuously acquire a 4D STEM scan of 256 x 256 probe positions, containing multiple metallic glass nanodroplets, every 1.6 seconds over a total acquisition time of around 10 minutes prior to crystallization. After crystallization, the metallic droplets were heated above their melting point and then quenched inside the TEM to ensure reproducibility.
One significant challenge of 5D STEM is the analysis of very large five-dimensional datasets. Here, a clustering method in the pyxem python package was used to track strong diffraction and group it both spatially and in time. These clusters formed a method for identifying and tracking individual structures in both the supercooled liquid and eventually the crystal.
Our results reveal a two-step nucleation process. First, large (3-4 nm radius) structures form within the supercooled liquid. These structures are disordered in nature, lack higher order diffraction characteristics of nanocrystalline materials and fluctuate in size. Second, from within the disordered structures, ordered nanocrystals of similar lower order diffraction form with higher ordered diffraction and the crystal starts to grow. This suggests a strong correlation between the structure of the supercooled liquid structure and that of the resultant crystal suggesting that crystal-like structures within the supercooled liquid are particularly important to understanding the nucleation behavior.