Accelerating Materials Discovery with Computation: Human-in-the-loop, Quantum Machine Learning, and First Principles Computation

Abstract

The central challenge of materials science is discovering and developing new materials that solve engineering challenges and enable new technologies. This thesis comprises three projects exploring the application of different computational techniques to materials discovery. The first project applied human-machine learning collaboration to autonomous experimentation. By presenting a human expert with a machine learning model's predictions and allowing them to provide their feedback, the performance and interpretability of the autonomous experiment was improved. The second project applied quantum kernel machine learning to x-ray diffraction pattern analysis. Comparing x-ray diffraction patterns using a quantum computer reduced the amount of experimental data needed to train machine learning models for some specific machine learning tasks similar to those in autonomous materials science. The third project applied high-throughput density functional theory calculations to the identification of novel candidate phase change materials. Using a first-principles thermodynamic approach, several promising materials systems were identified. These projects represent a sample of the future of computational materials discovery.