Highly accurate machine learning prediction of crystal point groups for ternary materials from chemical formula

Abstract

One of the most challenging problems in condensed matter physics is to predict crystal structure just from the chemical formula of the material. In this work, we present a robust machine learning (ML) predictor for the crystal point group of ternary materials (A$$_l$$ l B$$_m$$ m C$$_n$$ n ) - as first step to predict the structure - with very small set of ionic and positional fundamental features. From ML perspective, the problem is strenuous due to multi-labelity, multi-class, and data imbalance. The resulted prediction is very reliable as high balanced accuracies are obtained by different ML methods. Many similarity-based approaches resulted in a balanced accuracy above 95% indicating that the physics is well captured by the reduced set of features; namely, stoichiometry, ionic radii, ionization energies, and oxidation states for each of the three elements in the ternary compound. The accuracy is not limited by the approach; but rather by the limited data points and we should expect higher accuracy prediction by having more reliable data.

Resampling Techniques for Materials Informatics: Limitations in Crystal Point Groups Classification

“…Therefore, from the 32 classes, only 20 will be resampled. For the high-count classes, as found in ref , their performance is already good without the need for resampling.…”

“…Only the chemical formula and point group are of interest from the acquired 6,792,794 observations data. Sensibly, the charge acquired by an atom in a compound, or the oxidation state, which manifests itself in the Coulombic interactions governing the formation of compounds, is an essential feature to consider for predicting the crystal structure. , Since the oxidation state can differ for an element depending on the compound, the whole possible material space for ternary compounds needs to be constructed and matched with the extracted data. By imposing the condition of charge neutrality for each compound, the known oxidation states for each element allow for around 605 million possible permutations.…”

“…This paper focuses on examining and elevating the learnability of skewed data distribution within a multilabel multiclass classification setting through resampling methods. The effect of resampling techniques and dependencies on its variables is explored for the problem of imbalanced classification of ternary (containing three unique elements) crystal point groups, which has been investigated in a previous work . Knowledge of the point group aids in knowing the arrangement of the atoms in a crystal, since it represents the rotational symmetry of the crystal structure.…”

“…The effect of resampling techniques and dependencies on its variables is explored for the problem of imbalanced classification of ternary (containing three unique elements) crystal point groups, which has been investigated in a previous work. 33 Knowledge of the point group aids in knowing the arrangement of the atoms in a crystal, since it represents the rotational symmetry of the crystal structure. Rather than aiming to equate the classes count, the significance of sweeping for the ideal balancing degree is demonstrated.…”

A review about COVID-19 in the MENA region: environmental concerns and machine learning applications

“…In addition to detecting them, it may improve or change their actions over time. Efficiency and accuracy improve as data amount increases (Vabalas et al 2019 ; Guezzaz et al 2021 ; Alsaui et al 2022 ; Mirbolouki et al 2022 ). Better choices and predictions are made by the algorithm that learns from the data.…”