Descriptors for dielectric constants of perovskite-type oxides by materials informatics with first-principles density functional theory

Abstract

Dielectric materials that can realize downsizing and higher performance in electric devices are in demand. Perovskite-type materials of the form ABO 3 are potential candidates. However, because of the numerous conceivable compositions of perovskite-type oxides, finding the best composition is technically difficult. To obtain a reasonable guideline for material design, we aim to clarify the relationship between the dielectric constants and other physical and chemical properties of perovskite-type oxides using first-principles density functional theory (DFT) and partial least-squares regression analysis. The more important factors affecting the dielectric constants are predicted based on variable importance in projection (VIP) scores. The dielectric constant strongly correlates with the ionicity of the B cations and the density of states of the conduction bands of the B cations. ARTICLE HISTORY

Modeling the dielectric constants of crystals using machine learning

“…[31][32][33] The modern calculation of dielectric constants is typically done using density functional perturbation theory (DFPT). 4,[34][35][36][37][38] Despite being a perturbation theory, DFPT can be calculated with a relatively low cost, and the results are less dependent on the exchange-correlation functional used, compared to properties such as bandgaps. 4 Therefore, many studies have involved analysis of datasets of dielectric constants aiming to understand trends or predict values for new materials.…”

“…4,34 Some studies employ machine learning (ML) methods where a statistical model is trained. 35,36 For example, Umeda et al trained different ML models on a dataset of 3382 compounds. 35 They obtained good agreement with DFPT calculations; however, the reasons for this agreement are not discussed.…”

Machine learning models for predicting the dielectric constants of oxides based on high-throughput first-principles calculations

“…Pilania et al [24] constructed prediction models of various material properties, including the dielectric constants, of one-dimensional chain structures. Noda et al [29] reported a model for perovskite-type oxides using partial least-squares regression. Related to these studies, Umeda et al [30] corrected the DFT calculation errors of the dielectric constants using ML, while Choudhary et al [31] proposed a classification model for exploration of materials with dielectric constants higher than 10.…”

Entropy-based active learning of graph neural network surrogate models for materials properties

“…Active learning is particularly appealing in scenarios where data is scarce and expensive to obtain. For energy materials design properties such as high-quality electronic structure [40], dielectric constants [22,[41][42][43], effective masses [44] and defect properties [45,46] are some high-profile examples where data-driven approaches are possible, but obtaining large sets of labelled data can be prohibitive to applying deep learning approaches. In these scenarios the availability of a method for reducing the number of samples required to train accurate, and reliable models is rather critical to the application of deep learning approaches in materials' science.…”