CAREER: Microstructure Formation in Chemically-Modified Eutectics: Bridging Real-Time Imaging, Machine Learning, and Problem-Based Instruction

Regents Of The University Of Michigan - Ann Arbor, Ann Arbor MI

Investigators

Abstract

NON-TECHNICAL SUMMARY Patterns in nature typically form when a system changes from one phase of matter to another - for example, from a liquid phase to a geometrically patterned solid phase during solidification. The resulting patterns resemble a labyrinth of streets in a metropolitan area, although they are approximately one-hundred million times smaller and in three dimensions, called crystals. The structures of the as solidified crystals arranged in the pattern closely relate to the properties of the material, even after subsequent processing steps. To our advantage, the shapes and sizes of these crystals can be tailored to meet technological demands by manipulating the chemical composition of the parent liquid phase. For instance, trace metal impurities dissolved in the liquid are known to transform solid silicon (Si) from coarse, blocky particles into fine, web-like fibers. This results in a dramatic enhancement of the strength and ductility of the Si-based alloy and expands its potential for new applications, including space-frames and electric vehicles. The objective of this CAREER award is to understand how and why such transformations occur in the presence of trace metal impurities, by harnessing one of the brightest sources of hard X rays in the world at Argonne National Laboratory. The incident X-radiation can penetrate through an otherwise opaque metal, allowing one to capture the details of solidification in real time. Following the experiments, the PI and his team will extract information from the time lapse videos of solidification using state-of-the-art machine learning algorithms. It is anticipated that their new vision will advance the field of alloy solidification from metallurgical alchemy to predictive science. Ultimately, understanding the evolution of solid patterns during synthesis is the key to controlling the manufacture of advanced materials from the bottom-up. The new discoveries generated by this program will be integrated into problem-based learning units for underrepresented middle school students, in partnership with the Detroit Area Pre-College Engineering Program (DAPCEP). The PI will assess the impact of these activities using annual and multi-year surveys distributed through the DAPCEP organization. TECHNICAL SUMMARY In the past 50 years, there has been increasing interest - both fundamental and applied - in the process of solidification in multicomponent metal alloys. One longstanding problem concerns microstructure formation in eutectic alloy systems. While binary eutectics with non faceted interfaces are relatively well understood, comparatively little is known about the solidification behavior of eutectics with faceted interfaces and/or more than two components. Such is the case for Al-Si eutectics that are solidified in the presence of trace metal species (so called chemical modifiers). The higher degrees of freedom in the univariate reaction (i.e., three components and two solid phases) brings about morphological transitions not seen in nonvariant eutectics. For instance, the chemical modifiers transform the morphology of the Si phase from coarse flakes to fine fibers, thereby improving the mechanical properties of the cast alloy. Despite our technical experience with chemical modification, the underlying mechanisms by which the trace metal species influences the solidification pathway have not yet been satisfactorily explained. To this end, the PI and his team will develop a combined experimental computational research program that will focus on eutectic nucleation in chemically-modified alloys. Building on his track record in real-time imaging, he will harness the new capabilities in "fast" X-ray tomography to watch the solidification of Al-Si eutectics doped with various modifier species and concentrations. Given the volume of data to be collected in such experiments (tens of TB), it is impossible to process the data manually. Instead, the PI and his team will implement state-of-the-art machine learning algorithms (convolutional neural networks) to extract information from the large datasets, including details on the nucleation density, rate, and undercooling. The richness of these quantitative findings provides an unparalleled opportunity to verify kinetic theories with great precision. Lastly, the PI's team will conduct further correlative microscopy - covering over six orders-of-magnitude in length scale - to advance a comprehensive model of chemical modification in eutectic alloys. This award reflects NSF's statutory mission and has been deemed worthy of support through evaluation using the Foundation's intellectual merit and broader impacts review criteria.