Integrating Experiments and Machine Learning for Measuring Topological and Quantum Materials

In recent years, there has been a surge in research on the application of machine learning in chemistry and materials sciences. Machine learning has led to the discovery of new pharmaceutical molecules and energy materials, marking a paradigm shift in both research and industry. However, quantum materials have faced significant challenges despite numerous reports on machine learning applications. These challenges arise due to the complex interplay between charge, spin, orbital, and lattice degrees of freedom, and the frequent occurrence of out-of-distribution (OOD) problems. In particular, for topological materials, identifying experimental signatures to reveal topology is critical since "topology" itself is not directly measurable.<br/> <br/>In this MRS seminar, we present our recent efforts to connect machine learning with various topological and quantum materials, particularly through experimental techniques. For band topology materials with weak correlation, we introduce our classifier that determines the topological class based on x-ray absorption (XAS) signals [1]. For topological quantum computation, we discuss our approach to applying machine learning to distinguish Majorana zero modes from other spurious signals in tunneling spectroscopy [2]. For quantum materials where phonons play a role, we showcase a graph neural network method that predicts phonon dispersion relations much faster than traditional machine learning potentials [3]. Finally, recognizing that discovered materials represent only a small fraction of all possibilities, we describe our efforts to generate new materials with constrained lattice types. We conclude by highlighting additional examples that demonstrate the increasingly important role of machine learning in topological and quantum materials, despite data scarcity and computational challenges.<br/> <br/>[1] NA, ML,“Machine learning spectral indicators of topology,” <i>Advanced Materials</i> (2022)<br/>[2] MC, ML, “Machine Learning Detection of Majorana Zero Modes from Zero Bias Peak Measurements,” <i>Matter</i> (2024). <br/>[3] RO, AC, ML, "Virtual Node Graph Neural Network for Full Phonon Prediction," <i>Nature Computational Science</i> (2024).