Machine learning bandgaps of double perovskites

Area of Science:

• Materials Science
• Computational Chemistry
• Data Science

Background:

• Accurate prediction of electronic bandgaps in double perovskites is vital for materials discovery.
• Traditional quantum mechanical methods are computationally expensive for high-throughput screening.
• Informatics and machine learning present a viable alternative for rapid bandgap prediction.

Purpose of the Study:

• To develop an efficient and accurate computational framework for predicting double perovskite bandgaps.
• To identify key material descriptors that govern the electronic bandgap properties.
• To establish a robust machine learning model for high-throughput bandgap prediction.

Main Methods:

• Systematic feature engineering and selection from over 1.2 million potential descriptors.
• Development and validation of a machine learning framework for regression analysis.
• Utilizing data science best practices for model training, testing, and validation.

Main Results:

• Identification of lowest occupied Kohn-Sham levels and elemental electronegativities as the most significant predictors.
• Demonstration of efficient and accurate bandgap predictions for double perovskites.
• Successful validation of the developed machine learning models.

Conclusions:

• Machine learning, particularly with engineered features, provides an efficient and accurate method for predicting double perovskite bandgaps.
• Lowest occupied Kohn-Sham levels and elemental electronegativities are critical features for bandgap prediction.
• The developed framework facilitates rapid screening of double perovskites for targeted applications.