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

Area of Science:

• Materials Science
• Crystallography
• Computational Chemistry

Background:

• Low-dimensional organic-inorganic halide perovskites exhibit unique optoelectronic and magnetic properties.
• Predicting the dimensionality of these perovskites based on organic cation structure is challenging due to the lack of quantitative models.

Purpose of the Study:

• To develop a machine learning (ML)-assisted approach for predicting the dimensionality of lead iodide-based perovskites.
• To identify the key structural features of organic cations that dictate perovskite dimensionality.

Main Methods:

• A literature review identified 86 amines, classified as "2D"-forming or "non-2D"-forming.
• Machine learning models were trained using descriptor features of ammonium cations.
• Four key structural features were identified: steric effect index, eccentricity, largest ring size, and hydrogen-bond donor.

Main Results:

• A quantitative equation was derived to predict the probability of forming 2D perovskites based on amine features.
• The ML model accurately predicted the dimensionality of untested amines.
• Experimental verification through single crystal growth confirmed the model's predictions.

Conclusions:

• This work establishes a predictive model for low-dimensional hybrid halide perovskite structures.
• Machine learning serves as a powerful tool for understanding and predicting crystal structures in materials science.
• The identified features provide fundamental insights into structure-directing effects in perovskite formation.