Thorough Descriptor Search to Machine Learn the Lattice Thermal Conductivity of Half-Heusler Compounds

Abstract

Predicting the lattice thermal conductivity (κL) of compounds prior to synthesis is an extremely challenging task because of complexity associated with determining the phonon scattering lifetimes for underlying normal and Umklapp processes. An accurate ab initio prediction is computationally very expensive, and hence one seeks for data-driven alternatives. We perform machine learning (ML) on theoretically computed κL of half-Heusler (HH) compounds. An exhaustive descriptor list comprising elemental and compound descriptors is used to build several ML models. We find that ML models built with compound descriptors can reach high accuracy with a fewer number of descriptors, while a set of a large number of elemental descriptors may be used to tune the performance of the model as accurately. Thereby, using only the elemental descriptors, we build a model with exceptionally high accuracy (with an R 2 score of ∼0.98/0.97 for the train/test set) using one of the compressed sensing techniques. This work, while unfolding the complex interplay of the descriptors in different dimensions, reveals the competence of the readily available elemental descriptors in building a robust model for predicting κL.

Experimentally validated machine learning predictions of ultralow thermal conductivity for SnSe materials

“…However, most models developed have not been used to validate the results with new experimental data (Table 1). [25][26][27][28][29] Moreover, these models do not implement the presence of elemental proportions as feature vectors -an elemental vector matrix within a dataset that contains the amount of constituent elements derived from their respective chemical compositionin the final training set. The significance of the vector matrix is vital when using dopants in the training dataset, as high performing thermoelectric materials almost always include various dopants.…”

Machine learning the vibrational free energy of perovskites

“…Naturally, enormous computational effort has been spent to build online repositories containing different physical properties of compounds, which may be used to build meaningful machine-learning (ML) models. [32][33][34][35][36][37][38][39][40][41][42] Many works have been carried out in this regard. ML models are being built to predict the inter-atomic potentials to circumvent the problems associated with the use of DFT in predicting various properties of materials, which may be computationally expensive.…”