EFFICIENT MATERIALS INFORMATICS BETWEEN ROCKETS AND ELECTRONS

• A Dissertation in Materials Science and Engineering by Adam M. Krajewski
• Submitted in Partial Fulfillment of the Requirements for the Degree of Doctor of Philosophy in August 2024
• 352 total pages or 326 body pages
• 109 figures
• Final Version PDF

Abstract

The true power of computational research typically can lay in either what it accomplishes or what it enables others to accomplish. In this work, both avenues are simultaneously embraced across several distinct efforts existing at three general scales of abstractions of what a material is - atomistic, physical, and design. At each, an efficient materials informatics infrastructure is being built from the ground up based on (1) the fundamental understanding of the underlying prior knowledge, including the data, (2) deployment routes that take advantage of it, and (3) pathways to extend it in an autonomous or semi-autonomous fashion, while heavily relying on artificial intelligence (AI) to guide well-established DFT-based ab initio and CALPHAD-based thermodynamic methods.

The resulting multi-level discovery infrastructure is highly generalizable as it focuses on encoding problems to solve them easily rather than looking for an existing solution. To showcase it, this dissertation discusses the design of multi-alloy functionally graded materials (FGMs) incorporating ultra-high temperature refractory high entropy alloys (RHEAs) towards gas turbine and jet engine efficiency increase reducing $CO_2$ emissions, as well as hypersonic vehicles. It leverages a new graph representation of underlying mathematical space using a newly developed algorithm based on combinatorics, not subject to many problems troubling the community. Underneath, property models and phase relations are learned from optimized samplings of the largest and highest quality dataset of HEA in the world, called ULTERA. At the atomistic level, a data ecosystem optimized for machine learning (ML) from over 4.5 million relaxed structures, called MPDD, is used to inform experimental observations and improve thermodynamic models by providing stability data enabled by a new efficient featurization framework.

List of Chapters

• Chapter 1 - Introduction
• Chapter 2 - Extensible Structure-Informed Prediction of Formation Energy with Improved Accuracy and Usability employing Neural Networks
• Chapter 3 - Efficient Structure-Informed Featurization and Property Prediction of Ordered, Dilute, and Random Atomic Structures
• Chapter 4 - Handling Millions of Atomic Structures
• Chapter 5 - Ergodic Ensemble Approach to the Material Discovery
• Chapter 6 - Creating an Efficient Database Infrastructure for Discovery of Real Materials Exemplified with High Entropy Alloys
• Chapter 7 - Detecting Abnormalities in Materials Data
• Chapter 8 - Optimization of Compositional Dataset Domain towards Reliable Machine Learning Training and Deployment
• Chapter 9 - Inverse Design of Compositionally Complex Alloys
• Chapter 10 - Efficient generation of grids and traversal graphs in compositional spaces towards exploration and path planning exemplified in materials
• Chapter 11 - Infeasibility Gliding in Compositional Spaces
• Chapter 12 - Path Planing In Compositional Spaces through Graph Traversals
• Appendix Chapter A - Supplementary Discussions
• Appendix Chapter B - Additional Developed Software
• Appendix Chapter C - Nimplex Workshop No.1 - Quick Start Guide to using Nimplex through Python and Command Line Interface
• Appendix Chapter D - Nimplex Workshop No.2 - Additive Manufacturing Path Planning Made Effortless
• Appendix Chapter E - MatSE580 Guest Lecture 1 - Quick Guide to Manipulating Materials With pymatgen, Setting up MongoDB, and Getting Started with pySIPFENN
• Appendix Chapter F - MatSE580 Guest Lecture 2 - Running ML Models in pySIPFENN and Guiding Limited DFT Calculations Based on KS2022 Embedding Cluster Exploration
• Appendix Chapter G - nimCSO Basic Tutorial on Selecting Elements for High Entropy Alloy Modeling

Graphical Abstract

Vita

Adam Krajewski was born in Europe, where he spent his childhood and received pre-college education at a school nationally recognized for its university-level chemistry curriculum. He first came to the United States in 2013 and moved completely in 2015 to join the Materials Science Department at Case Western Reserve University. Within the first two months, Adam began research in Prof. Welsch's group. After just one year, he enrolled in graduate courses and also joined Prof. Willard's group, progressively moving from experiments towards theory, modeling, and simulations. In the Fall of 2017, he enrolled in graduate courses in Artificial Intelligence, starting to specialize in applying AI techniques, including Machine Learning, to his research which became focused hidden process modeling, materials data processing, and data-driven design of magnetocaloric metallic glasses.

After earning his B.S.E. degree in 2019, Adam moved directly to pursue PhD under world-renowned thermodynamics expert Prof. Zi-Kui Liu at Penn State. He had the pleasure of working on implementing various computational techniques, ranging from atomistic machine learning through materials data curation, to purely theoretical considerations, while having the support of colleagues who are specialists in applied ab-initio modeling, thermodynamic calculations, and materials discovery. Since 2022, he has also extensively collaborated with Lawrence Livermore National Lab, where he spent two summers on-site.

As of May 2024, Adam has published several computational tools and scientific publications listed under his ORCID record 0000-0002-2266-0099 and Google Scholar (id:3tvHo8kAAAAJ including 4 first-author publications and 9 co-author publications. Furthermore, eight first-author papers are under preparation.