Deep Learning Based Prediction of Perovskite Lattice Parameters from Hirshfeld Surface Fingerprints

Abstract

This Letter describes the use of deep learning methods on Hirshfeld surface representations of crystal structure, as an automated means of predicting lattice parameters in cubic inorganic perovskites. While Hirshfeld Surface Analysis is a well-established tool in organic crystallography, we also introduce modified computational protocols for Hirshfeld Surface Analysis tailored specifically to account for nuanced but important differences dealing with inorganic crystals. We demonstrate how two-dimensional Hirshfeld surface fingerprints can serve as a rich "database" of information encoding the complexity of relationships between chemical bonding and bond geometry characteristics of perovskites. Our results are compared with other studies on lattice parameter prediction involving both experimental and computationally derived data, and it is shown that our approach is an improvement over other reported methods. The paper concludes by discussing how this work opens new avenues for data-driven high throughput computational predictions of structure−property relationships involving complex crystal chemistries.

Mlatticeabc: Generic Lattice Constant Prediction of Crystal Materials Using Machine Learning

“…According to the input information of the prediction models, the approaches can be divided into composition (such as atomic properties of their constituent elements)-based lattice parameter prediction models 4 and structure-based prediction models. 9 While the majority of methods are based on composition information, the structure-based approaches can also bring interesting insights. 8 In this paper, 9 a deep learning method is proposed to predict lattice parameters in cubic inorganic perovskites based on Hirshfeld surface representations of crystal structures.…”

“… 9 While the majority of methods are based on composition information, the structure-based approaches can also bring interesting insights. 8 In this paper, 9 a deep learning method is proposed to predict lattice parameters in cubic inorganic perovskites based on Hirshfeld surface representations of crystal structures. They showed that two-dimensional Hirshfeld surface fingerprints contain rich information encoding the relationships between chemical bonding and bond geometry characteristics of perovskites.…”

“… 26 In terms of ionic radii, there are several ways to combine the values of the component elements such as the sum, differences, or ratios. In the structure-based lattice constant prediction, 9 both graph representations and two-dimensional (2D) fingerprints have been used by measuring (di, de) at each point on the surface and then binning them into a 2D histogram. In lattice constant prediction of binary body-centered cubic crystals, Takahashi et al 12 recommended seven descriptors for predicting lattice constant a including the atomic number of element A and B, density, atom orbital of elements A and B, difference in electronegativity between A and B, and atomic orbital B + difference in the electronegativity between A and B.…”

MLatticeABC: Generic Lattice Constant Prediction of Crystal Materials using Machine Learning

“…During the past 15 years, a series of prediction approaches have been proposed for lattice constant prediction, which can be categorized by their input information used, the descriptors or features, the machine learning model, and the chemical system or materials family they are trained for. According to the input information of the prediction models, the approaches can be divided into composition (such as atomic properties of their constituent elements) based lattice parameter prediction models [4] and structure based prediction models [9]. While the majority of methods are based on composition information, the structure based approaches can also bring interesting insights [8].…”

“…While the majority of methods are based on composition information, the structure based approaches can also bring interesting insights [8]. In [9], a deep learning method is proposed to predict lattice parameters in cubic inorganic perovskites based on Hirshfeld surface representations of crystal structures. They showed that two-dimensional Hirshfeld surface fingerprints contain rich information encoding the relationships between chemical bonding and bond geometry characteristics of perovskites.…”