Data-driven designing of conjugated organic chromophores: Chemical space generation and property prediction

This study presents a machine learning-assisted approach for the designing of conjugated organic chromophores. 10 machine learning models are trained to predict exciton binding energy, random forest has appeared as best model (R-squared = 0.723). A database of new chromophores is generated and exciton binding energy of chromophores is predicted. 30 organic chromophores with low exciton binding energy values are identified. Clustering and chemical similarity analyses, based on chemical fingerprints, are conducted on the selected chromophores. Additionally, the synthetic accessibility scores of the newly designed chromophores are evaluated. This approach enables rapid screening of organic chromophores for use in organic solar cells. The proposed framework provides a strategic and efficient pathway for discovering optimal materials for organic solar cell applications.