Insights into cation ordering of double perovskite oxides from machine learning and causal relations

Abstract

This work investigates origins of cation ordering of double perovskites using first-principles theory computations combined with machine learning (ML). We have considered various possible oxidation states of A, A , B and B from the family of transition metal ions to construct datasets. A conventional framework employing traditional ML classification algorithm such as Random Forest (RF) combined with appropriate features including geometry-driven and key structural modes leads to highly accurate prediction (∼98%) of A-site cation ordering. This study evaluates accuracy of ML models by entailing analyses of decision paths, assignments of probabilistic confidence bound, and finally introducing a direct non-Gaussian acyclic structural equation model to investigate causality. A comparative study of utilization of first-principles driven features versus those computed with lower fidelity is added to further elucidate the role of structural modes for accurate prediction of cation ordering in double perovskites.