| Abstract Scope |
Electronic structure simulations, like Density Functional Theory (DFT), accurately model materials properties, but are often computationally prohibitive for glasses due to their lack of long-range order. However, since DFT scales poorly with system size, simulating many small, ordered systems can be “cheaper” than one large enough to capture the disorder. Prior work showed that appropriate averages over these “crystalline ensembles” accurately reproduced a variety of experimentally measured properties. This approach used random structure sampling to fully capture configuration space. Here, we explore strategies to downselect from sampled structures, ultimately using DFT for a small subset of them while retaining accuracy. This efficiency could allow for using crystalline ensembles with even higher levels of electronic structure theory, enabling more accurate first-principles calculations of glassy materials.
While in this work we focus on structural characteristics, future research will explore the generality of downselection techniques across different materials properties. |