UC Berkeley

In recent years, the increased availability of computing resources has enabled the construction of public databases/datasets from automated, high-throughput calculations for materials structure and properties, often at the atomic level. Rising alongside the heightened popularity of data science and machine learning (ML) techniques, these datasets have greatly benefited the community by guiding materials design and the development of ML models such as universal force fields. Thus far, these computational workflows have primarily been developed for bulk crystals and molecules, and they are relatively less mature for interfacial properties. In the third chapter, I will present my work building a workflow for calculating antiphase boundary (APB) energies in Ni3Al, which is important for the high-temperature performance of Ni-based superalloys in jet turbines. This workflow combines statistical thermodynamics, first-principles calculations, and machine learning to produce a database and predictive models for the APB energy. In the fourth chapter, I will present a computational tool, the GRand canonical Interface Predictor (GRIP), that can be used for high-throughput, grand canonical optimization of grain boundary (GB) structures. We specifically demonstrate the application of GRIP on tilt GBs in hexagonal close-packed (HCP) titanium, as previous GB optimization studies focused solely on cubic systems. By allowing the density of atoms in the GB to vary, we discover novel GB phases in HCP Ti with distinct localized dislocation cores. We use high-temperature molecular dynamics to validate phase stability and study phase transformations through a novel point defect-induced dislocation pairing mechanism. In the fifth chapter, I present some complementary work where I use many of these same tools to enhance materials informatics education, as there exists a critical need to train materials scientists to harness the growing amounts of computing resources and scientific data. Our nation’s workforce must be proficient in these skills in order to accelerate materials development to solve grand challenges in areas such as sustainability and public health. Specifically, I use the Jupyter Book software to design interactive, digital texts that we have successfully used for a summer research curriculum and integrated into a materials characterization laboratory course. I will conclude with some ideas for how to build on the work in this dissertation to continue pushing the frontiers of research and teaching in materials science and engineering.