Computational and Informatics Guided Design of Magnetic Materials

Author: ORCID icon
Hartnett, Timothy, Materials Science - School of Engineering and Applied Science, University of Virginia
Balachandran, Prasanna, University of Virginia

Theory-based computational simulations have a long history of modeling the underlying physics of materials systems. In principle, they can also be used to guide the exploration and optimization of new materials for use in targeted applications. Unfortunately, these computational tools can be extremely expensive when considering the vast space of unexplored materials and their complexities. Recently, scientists have leveraged the boom in data science to develop new and interesting techniques for the rapid exploration of vast compositional, structural, and processing spaces. However, the “black box” nature of machine learning models make them difficult to interpret the result of these explorations. This dissertation illustrates how first-principles calculations and data science methods can be leveraged to establish hitherto unknown structure-property relationships in magnetic materials. Three promising classes of magnetic materials are considered: (1) Bulk ferromagnetic (FM) MTX alloys that exhibit a giant magnetocaloric effect (MCE) with potential in solid-state cooling, (2) Heterostructures of ferrimagnetic (FiM) Mn4N, which shows promise as a nearly compensated ferrimagnet for hosting magnetic skyrmions, and (3) Bulk FM, Fe-rich FeGe that poses intrinsic Dzyaloshinskii-Moriya interactions (DMI) and high FM Curie temperature (TC ) also as a potential skyrmion host.

In the case of MTX alloys, using density functional theory (DFT), we explore the impact of site-specified chemistry on the relevant materials properties in the system. Using this information, we rigorously train an ensemble of machine learning models to rapidly predict the martensitic transition temperature (Tt) of the system and experimentally validate the predictions. Using post hoc model interpretability techniques, we also identify trends that give insight into the underlying mechanism of phase stability in the system. Based on the prediction of the machine learning, we explore the impact of high entropic, single-site substitution using DFT and find a substantial reduction in total energy between the relevant phases with a minimal impact on the magnetism in the system providing a potential path towards achieving room temperature magnetostructural phase transitions while maintaining high saturation magnetization.

Bulk Mn4N is a centrosymmetric, FiM compound with high Néel temperature (TN) (740 K). Its relative ease of processing makes it a viable material for hosting skyrmions in magnetic heterostructures. Using DFT we calculate the bulk properties of Mn4N then proceed to calculate the interfacial Dzyaloshinskii-Moriya interactions (i-DMI) for different substrates. Using these results and models developed for FM//heavy metal interfaces, we explore the possible electronic transitions responsible for the magnitude and sign of the i-DMI. Existing models also provide a prediction of skyrmion radius based on DFT results which we can be compared to experimental findings. Due to the labor-intensive nature of i-DMI calculation we also test a novel feature for describing the total and atom-projected density of states (DOS). The Kullback-Leibler divergence metric (KLD) between substrate and magnetic layers shows a strong correlation with i-DMI strength for a given substrate. Combining the KLD with the substrate total energy difference based on inclusion or exclusion of SOC in the DFT calculation and using a generative algorithm we arrive at a simple analytic expression for i-DMI that may be used to rapidly predict i-DMI in magnetic heterostructures.

FM FeGe is one of the well-studied skyrmion hosts but its TC is below 300 K. Using nonequilibrium techniques our colleagues at The Ohio State University have grown Fe-rich FeGe which has elevated TC and skyrmion stability. Using DFT we investigate the location and electronic structure of point defects in this system and provide a simulated Fe K-edge XAS, which we hope will motivate further experimental analysis to provide conclusive evidence of the dominant point defect in the system. Based on different defect structures we also calculate magnetic properties of the system and investigate the electronic structure for potential origins of bulk DMI in Fe-rich FeGe

PHD (Doctor of Philosophy)
Magnetism, Materials Informatics
All rights reserved (no additional license for public reuse)
Issued Date: