High-throughput Density Functional Perturbation Theory and Machine Learning Predictions of Infrared, Piezoelectric and Dielectric Responses
Kamal Choudhary, Kevin F. Garrity, Vinit Sharma, Adam J. Biacchi, Francesca M. Tavazza, Angela R. Hight Walker
In this work, combining high-throughput (HT) density functional perturbation theory and supervised machine learning approaches, we explored the territory of compounds with interesting infrared, piezoelectric and dielectric properties. We have computed Γ-point phonons, infrared intensities, Born-effective charges (BEC), piezoelectric (PZ) tensors and dielectric (DL) tensors for 5015 inorganic materials in the JARVIS-DFT database. Analyzing this data, we find 3230 and 1943 materials with at least one far-IR and mid-IR peak, respectively. Similarly, from our computed PZ tensor data, we identify 577 high-PZ materials, using a threshold of 0.5 C/m2 as an absolute maximum value. Using a threshold of 20, we find 593 potential high-dielectric materials. Importantly, we analyze the chemistry, symmetry, dimensionality (0D/1D/2D/3D), geometry of materials and several property-correlations to find trends. Finally, we train regression models for the highest IR frequency and maximum BEC, and classification models for maximum PZ and average DL tensors to systematically accelerate materials screening.
, Garrity, K.
, Sharma, V.
, Biacchi, A.
, Tavazza, F.
and Hight, A.
High-throughput Density Functional Perturbation Theory and Machine Learning Predictions of Infrared, Piezoelectric and Dielectric Responses, npj Computational Materials, [online], https://doi.org/10.1038/s41524-020-0337-2
(Accessed June 14, 2021)