Obtaining physical insights for diffusion through machine learning for renewable energy storage applications

Lu, Grace M

Permalink

https://hdl.handle.net/2142/129933 Copy

Description

Title

Obtaining physical insights for diffusion through machine learning for renewable energy storage applications

Author(s)

Lu, Grace M

Issue Date

2025-07-14

Director of Research (if dissertation) or Advisor (if thesis)

Trinkle, Dallas

Doctoral Committee Chair(s)

Trinkle, Dallas

Committee Member(s)

• Bellon, Pascal
• Ertekin, Elif
• Schleife, Andre

Department of Study

Materials Science & Engineerng

Discipline

Materials Science & Engr

Degree Granting Institution

University of Illinois Urbana-Champaign

Degree Name

Ph.D.

Degree Level

Dissertation

Keyword(s)

• machine learning
• diffusion
• hydrogen storage
• perovskite
• pyrochlore
• li-ion battery

Abstract

Hydrogen storage, oxide fuel cells, and Li-ion batteries are three techniques that enable renewable energy storage and transport. To improve their performance, new materials need to be discovered with ideal transport properties. To explore the large material space efficiently, machine learning methods provide physical insights and enable the rapid screening of new materials. While analytic models use hand-selected features that have clear physical ties, they often lack accuracy when making quantitative predictions. Machine learning models are capable of making accurate predictions, but their inner workings are obscured, rendering it unclear which features are important. Additionally, machine learned interatomic potentials can allow near-DFT levels of accuracy on large systems for which DFT calculations would be prohibitively expensive. To develop interpretable machine learning models to predict the activation energies of hydrogen diffusion in metals and random binary alloys, we create a database and fit six machine learning models. Grouped feature importances, formed by combining the features via their correlations, reveal that the two groups containing the packing factor and electronic specific heat are particularly significant for predicting hydrogen diffusion in metals and random binary alloys. This framework allows us to interpret machine learning models and enables rapid screening of new materials with the desired rates of hydrogen diffusion. We then expand this framework by showing its applicability to predicting transport properties in more complex materials and as a feature down-selection method. For predicting oxygen diffusion in perovskites and pyrochlores, we build a database of experimental activation energies and use our grouping framework to reduce the number of material property features. These features are then used to fit seven different machine learning models. An ensemble consensus determines that the most important features for predicting the activation energy are the ionicity of the A-site bond and the partial pressure of oxygen for perovskites. For pyrochlores, the two most important features are the A-site $s$ valence electron count and the B-site electronegativity. The most important features are all constructed using the weighted averages of elemental metal properties, despite weighted averages of the constituent binary oxides being included in our feature set. This is surprising because the material properties of the constituent oxides are more similar to the experimentally measured properties of perovskites and pyrochlores than the features of the metals that are chosen. Inclusion of Ag at the electrolyte-anode interface has been shown to reduce dendrite growth, and enable the use of anode-free lithium-ion batteries through an alloying process. Using the pre-trained MACE-MP-0 potential, we explore interfacial structures between FCC Ag and Li, and conclude that the FCC Li phase is more energetically favorable. We demonstrate that there are negligible migration barriers for Li atoms to migrate across the interface into the Ag. However, larger migration barriers for diffusion of vacancies from the interface into the Li slab can impede the mixing process kinetically.

Graduation Semester

2025-08

Type of Resource

Thesis

Handle URL

https://hdl.handle.net/2142/129933

Copyright and License Information

Copyright 2025 Grace Lu

Owning Collections