Predictive Design Model for Low-Dimensional Organic-Inorganic Halide Perovskites Assisted by Machine Learning.

Low-dimensional organic-inorganic halide perovskites have attracted interest for their properties in exciton dynamics, broad-band emission, magnetic spin selectivity. However, there is no quantitative model for predicting the structure-directing effect of organic cations on the dimensionality of these low-dimensional perovskites. Here, we report a machine learning (ML)-assisted approach to predict the dimensionality of lead iodide-based perovskites. A literature review reveals 86 reported amines that are classified into "2D"-forming and "non-2D"-forming based on the dimensionality of their perovskites. Machining learning models were trained and tested based on the classification and descriptor features of these ammonium cations. Four structural features, including steric effect index, eccentricity, largest ring size, and hydrogen-bond donor, have been identified as the key controlling factors. On the basis of these features, a quantified equation is created to calculate the probability of forming 2D perovskite for a selected amine. To further illustrate its predicting capability, the built model is applied to several untested amines, and the predicted dimensionality is verified by growing single crystals of perovskites from these amines. This work represents a step toward predicting the crystal structures of low dimensional hybrid halide perovskites using ML as a tool.