Machine Learning Spectral Indicators of Topology

Abstract

AbstractTopological materials discovery has emerged as an important frontier in condensed matter physics. While theoretical classification frameworks have been used to identify thousands of candidate topological materials, experimental determination of materials’ topology often poses significant technical challenges. X-ray absorption spectroscopy (XAS) is a widely-used materials characterization technique sensitive to atoms’ local symmetry and chemical environment, which are intimately linked to band topology by the theory of topological quantum chemistry. Moreover, as a local structural probe, XAS is known to have high quantitative agreement between experiment and calculation, suggesting that insights from computational spectra can effectively inform experiments. In this chapter, we show that XAS can potentially uncover materials’ topology when augmented by machine learning. Using the computed X-ray absorption near-edge structure (XANES) spectra of more than 10,000 inorganic materials, we train a neural network classifier that predicts topological class directly from XANES signatures with F1 scores of 82% and 87% for topological and trivial classes, respectively, and achieves F1 scores above 90% for materials containing certain elements. Given the simplicity of the XAS setup and its compatibility with multimodal sample environments, the proposed machine learning-empowered XAS topological indicator has the potential to discover broader categories of topological materials, such as non-cleavable compounds and amorphous materials, and may further inform a variety of field-driven phenomena in situ, such as magnetic field-driven topological phase transitions.KeywordsX-ray absorption spectroscopyX-ray absorption near-edge structureTopological materialsTopological quantum chemistryPrincipal component analysisConvolutional neural network