Predicting the stability of the perovskite structure remains a longstanding challenge for the discovery of new functional materials for photovoltaics, fuel cells, and many other applications. Using a novel data analytics approach based on SISSO (sure independence screening and sparsifying operator), an accurate, physically interpretable, and one-dimensional tolerance factor, Ï, is developed that correctly classifies 92% of compounds as perovskite or nonperovskite for an experimental dataset containing 576 ABX3 materials (X = O2-, F-, Cl-, Br-, I-). In comparison, the widely used Goldschmidt tolerance factor, t, achieves a maximum accuracy of only 74% for the same set of materials. The probability of forming stable perovskites is mapped continuously as a function of the sizes of the A, B, and X ions revealing physical insights into how these relative sizes yield stable and unstable perovskite structures. Additionally, the new tolerance factor is shown to compare well with DFT-calculated decomposition enthalpies of single and double perovskite oxides and chalcogenides. Ï is applied to identify more than a thousand inorganic (Cs2BBâCl6) and hybrid organic-inorganic (MA2BBâBr6) double perovskites that are predicted to be stable.
