Learn-and-Match Molecular Cations for Perovskites.

Area of Science:

• Materials Science
• Computational Chemistry
• Solid-State Physics

Background:

• Predicting structural stability in hybrid organic/inorganic compounds is complex due to vast composition spaces and undefined organic molecule references.
• Perovskite structures, particularly ABC3 chalcogenides and halides, are crucial but challenging to analyze for stability.
• Traditional methods for materials discovery often rely on trial-and-error, which is inefficient for complex systems.

Purpose of the Study:

• To develop and apply machine-learning algorithms for systematically predicting the likelihood of cations forming stable perovskite structures.
• To identify key features that govern the phase stability of perovskite compounds.
• To provide an efficient computational strategy for materials design in hybrid organic/inorganic systems.

Main Methods:

• Utilized state-of-the-art density functional theory (DFT) data to train various machine-learning algorithms.
• Conducted a systematic analysis of cation properties influencing perovskite phase stability.
• Focused on both chalcogenide (I-V-VI3) and halide (I-II-VII3) perovskite structures.

Main Results:

• Identified effective atomic radius and the number of lone pairs on the A-site cation as sufficient descriptors for perovskite phase stability.
• Developed a machine-learning approach capable of efficiently mapping the phase stability of a wide range of compounds.
• Demonstrated the model's applicability to systems with mixed cations replacing a single A-site cation.

Conclusions:

• Machine learning, combined with advanced electronic structure theory, offers a powerful and efficient alternative to traditional trial-and-error in materials design.
• The identified key features provide a simplified yet effective way to predict perovskite stability.
• This work paves the way for accelerated discovery of novel hybrid organic/inorganic materials with desired structural properties.