Data Analytics Predicts New Stable Perovskites

Article: New tolerance factor to predict the stability of perovskite oxides and halides

Guest Author: Dr. Michael Yeung, Northwestern University

Because of their relatively simple structure and their compositional breadth, perovskites are intensely studied as solar absorbers, ferroelectric materials, and superconductors. One hot example today is the field of halide perovskites, which are easy to deposit and have remarkable solar conversion efficiencies. While there has been some debate as to what makes a perovskite a perovskite, the purist definition defines the perovskite structure as corner sharing octahedra, typically of ABX3 composition. This structure was first solved by the mineralogist Victor Goldschmidt in 1926, who deduced by geometry that only ions of a certain size will make a perovskite structure. His discovery, known as the Goldschmidt Tolerance Factor, has been used by solid state chemists since 1926 to understand the stability of their perovskites. A pocket calculator and knowing the radius of A, B, and X is all that is needed.

Figure 1: The breadth of elements that perovskites can be composed of. The new tolerance factor discovered by Bartel et al can more accurately predict stable perovskites composed of these elements.

Since Goldschmidt’s discovery, there has been a massive uptick in the number of known perovskite compounds, including the rapidly growing field of halide perovskites. Unfortunately, the accuracy of Goldschmidt’s model for these new compounds is off. When comparing predicted perovskites of ABX3 composition with measured, synthesized compounds chlorides (51% accuracy), bromides (56%), and iodides (33%)] is far lower than oxides (83%) and fluorides (83%). Last month, Bartel et al (Science Advances, 5, eaav0693 (2019)) suggests a new tolerance factor by which to judge stable perovskites. This new tolerance factor is more accurate for oxides (92% accuracy), fluorides (92%), chlorides (90%), bromides (93%), and iodides (91%); more importantly, this new tolerance factor only needs a pocket calculator, just like Goldschmidt’s formulation. These new formulas were determined with a machine learning like, novel data analytics approach. Furthermore, they are able to predict ~23,000 new compounds that should be stable perovskites with their formula. This simple approach will help experimentalist and theorists quickly test the stability of their envisioned perovskites, and serve as a quick screen for viability in topics across materials chemistry.


Leave a Reply