Prediction of ABX₃ Perovskite Formation Energy Using Machine Learning.

Area of Science:

• Materials Science
• Computational Materials Science
• Solid-State Chemistry

Background:

• Perovskite materials are crucial for advanced devices like solar cells and sensors.
• Structural instability hinders the application of many perovskite compositions.
• Traditional prediction methods, such as the tolerance factor, have limitations in accuracy due to overlooking atomic interactions.

Purpose of the Study:

• To develop a robust machine learning model for predicting the formation energy of ABX3 perovskites.
• To overcome the limitations of existing analytical methods in predicting perovskite stability.
• To utilize formation energy as a key parameter reflecting atomic interactions for accurate material property prediction.

Main Methods:

• Application of machine learning algorithms for pattern recognition in large datasets.
• Development of a predictive model targeting formation energy of ABX3 perovskite structures.
• Validation of the machine learning model using first-principles computations.

Main Results:

• Achieved a high R-squared value of 0.928, indicating strong model performance.
• Obtained a root mean square error of 0.301 eV/atom, demonstrating prediction accuracy.
• Successfully predicted 75% of values within a low error margin of 0.06 eV/atom.

Conclusions:

• The developed machine learning model accurately predicts perovskite formation energy.
• This approach enhances the understanding and prediction of perovskite structural stability.
• The findings can accelerate research into solving perovskite instability for device applications.