Spectral Operator Representations

Machine learning in atomistic materials science has grown to become a powerful tool, with most approaches focusing on atomic arrangements, typically decomposed into local atomic environments. This approach, while well-suited for machine-learned interatomic potentials, is conceptually at odds with learning complex intrinsic properties of materials, often driven by spectral properties commonly represented in reciprocal space (e.g., band gaps or mobilities) which cannot be readily atomically partitioned. For such applications, methods which represent the electronic rather than the atomic structure could be more promising. In this work, we present a general framework focused on electronic-structure descriptors which take advantage of the natural symmetries and inherent interpretability of physical models. Using this framework, we formulate two such representations and apply them respectively to measuring the similarity of carbon nanotubes and barium titanate polymorphs, and to the discovery of novel transparent conducting materials (TCMs) in the Materials Cloud 3D database (MC3D). A random forest classifier trained on 1% of the materials in the MC3D is able to correctly label 76% of entries in database which meet common screening criteria for promising TCMs.