A Study of the Utility of a Machine-Learning Approach Applied to the Prediction of Site Occupancy and New Members of the Half-Heusler Family

• Author / Creator

  Gzyl, Alexander

• Predicting the formation and structures of non-molecular inorganic compounds has long been a fundamental goal in solid state chemistry. In this thesis, machine learning approaches have been applied to confront this challenge, focusing in particular on the large family of half-Heusler compounds because they exhibit many useful materials properties but have not always been structurally well characterized. Two specific problems have been tackled: assigning the correct site distributions in existing half-Heusler compounds, and predicting the formation of new half-Heusler compounds.
  The site preference within the structures of half-Heusler compounds have been evaluated through a machine-learning approach. A support-vector machine algorithm was applied to develop a model which was trained on 179 experimentally reported structures and 23 descriptors based solely on the chemical composition. The model gave excellent performance with sensitivity of 93%, specificity of 96% and accuracy of 95%. As an illustration of data sanitization, two compounds (GdPtSb, HoPdBi) flagged by the model to have potentially incorrect site assignments were resynthesized and structurally characterized. The predictions of the correct site assignments from the machine-learning model were confirmed by single-crystal and powder X-ray diffraction analysis. These site assignments also correspond to the lowest total energy configurations as revealed from first-principles calculations.
  A machine-learning ensemble was used to predict new half-Heusler compounds. Compositions were selected for synthesis if they were also adopted by a full-Heusler compound counterpart. The model gave excellent performance with sensitivity of 90.0%, Specificity of 98.0%, and accuracy of 97.7%. Perturbations in site occupancy (e.g., vacancies, disorder) led to changes in crystal symmetry. Synthetic minority oversampling (SMOTE) and ensemble methods have been combined and applied for the first time to a materials science problem, and the performance of this approach has been evaluated

• Subjects / Keywords

  • Solving the coloring problem
  • half-Heusler
  • Machine-Learning Approach

• Graduation date

  Fall 2019

• Type of Item

  Thesis

• Degree

  Master of Science

• DOI

  https://doi.org/10.7939/r3-wc2c-hr02

• License

  Permission is hereby granted to the University of Alberta Libraries to reproduce single copies of this thesis and to lend or sell such copies for private, scholarly or scientific research purposes only. Where the thesis is converted to, or otherwise made available in digital form, the University of Alberta will advise potential users of the thesis of these terms. The author reserves all other publication and other rights in association with the copyright in the thesis and, except as herein before provided, neither the thesis nor any substantial portion thereof may be printed or otherwise reproduced in any material form whatsoever without the author's prior written permission.