Materials Screening for the Discovery of New Half-Heuslers: Machine Learning versus ab Initio Methods

Abstract

Machine learning (ML) is increasingly becoming a helpful tool in the search for novel functional compounds. Here we use classification via random forests to predict the stability of half-Heusler (HH) compounds, using only experimentally reported compounds as a training set. Cross-validation yields an excellent agreement between the fraction of compounds classified as stable and the actual fraction of truly stable compounds in the ICSD. The ML model is then employed to screen 71,178 different 1:1:1 compositions, yielding 481 likely stable candidates. The predicted stability of HH compounds from three previous high throughput ab initio studies is critically analyzed from the perspective of the alternative ML approach. The incomplete consistency among the three separate ab initio studies and between them and the ML predictions suggests that additional factors beyond those considered by ab initio phase stability calculations might be determinant to the stability of the compounds. Such factors can include configurational entropies and quasiharmonic contributions.

A Critical Review of Machine Learning of Energy Materials

“…Here, we will review ML works aimed at specifically identifying thermoelectrics through property predictions. There are several works that aim to identify new intermetallic structures (e.g., Heuslers being a major class) and their stability, but make no attempt to predict other critical thermoelectric properties . These works will not be reviewed here.…”

“…There are several works that aim to identify new intermetallic structures (e.g., Heuslers being a major class) and their stability, but make no attempt to predict other critical thermoelectric properties. [117,299,300] These works will not be reviewed here.…”

Machine learning in materials science

“…Many classification and regression algorithms can be applied to predict the chemical composition of a material from its structure . Perovskite is an important crystal structure in many fields .…”

“…This model realized a high true‐positive rate of .94 and successfully predicted 12 novel gallides, namely, MRu 2 Ga and RuM 2 Ga (M = Ti − Co), as Heusler compounds. Legrain et al trained a model using experimentally reported compounds to predict the stability of half‐Heusler compounds. The model, which was based on the random forest algorithm, retrieved 71 178 compositions and yielded 30 results, which mostly matched half‐Heusler compounds, for further exploration.…”

“…Many classification and regression algorithms can be applied to predict the chemical composition of a material from its structure. 28,40,41,129,130 Perovskite is an important crystal structure in many fields. 131,132 41 Pilania et al 121 developed a systematic feature engineering approach and an efficient machine learning model for predicting electronic bandgaps of double perovskites and for selecting stable perovskite candidates.…”