Data-assisted approach for optimal designing of small molecules for perovskite solar cells

Conventional computational methods have long history in designing the organic compounds, however, these approaches generally require significantly higher computational cost. To overcome these challenges, machine learning is applied as a powerful approach to screen and design high performance materials in a rapid and computationally cost-effective manner. Reorganization energy (Re) is predicted using machine learning. Mordred software is used to calculate molecular descriptors. Different algorithms such as random forest regressor, gradient boosting regressor, K-neighbors regressor, and extra tree regressor models are used to train the machine learning models. Random forest regressor model reveals higher predictive capability (R² = 0.73). Automatic method is used to design new compounds. 30 potential candidates are identified and their synthetic ability score are predicted. Clustering is used for similarity analysis. Interestingly, synthetic accessibility score reveals that these compounds can be synthesize with ease. The proposed approach holds immense potential for screening and designing high performance hole transport materials for perovskite solar cells in a cost-effective and rapid manner.