Unsupervised clustering for identifying spatial inhomogeneity on local electronic structures

Spatial inhomogeneity on the electronic structure is one of the vital keys to provide a better understanding of the emergent quantum phenomenon. Given the recent developments on spatially resolved ARPES (ARPES: angle-resolved photoemission spectroscopy), the information on the spatial inhomogeneity on the local electronic structure is now accessible. However, the next challenge becomes apparent as the conventional analysis encounters difficulty handling a large volume of a spatial mapping dataset, typically generated in the spatially resolved ARPES experiments. Here, we propose a machine-learning-based approach using unsupervised clustering algorithms (K-means and fuzzy-c-means) to examine the spatial mapping dataset. Our analysis methods enable automated categorization of the spatial mapping dataset with a much-reduced human intervention and workload, thereby allowing quick identification and visualization of the spatial inhomogeneity on the local electronic structures.

Automated summary

Spatial inhomogeneity on the electronic structure is vital for understanding emergent quantum phenomena. Recent advancements in spatially resolved ARPES have made it possible to access information on local electronic structure. However, conventional analysis struggles with handling large spatial mapping datasets, leading to the proposal of a machine-learning-based approach using unsupervised clustering algorithms for automated categorization of spatial mapping datasets to quickly identify and visualize spatial inhomogeneity.