Abstract
Designing compounds for organic solar cells is a hot topic. In the present study, a new approach is introduced to design acceptor materials for organic solar cells. Building blocks are mined from the chemical database. New libraries of building blocks are also enumerated. New acceptors are designed using searched building blocks. Machine learning is used to predict the power conversion efficiency of accepters. Best molecular descriptors (features) are selected with the help of statistical methods. Multiple machine learning models are trained using best descriptors. The bagging regressor and random forest regressor are the best models. Energy levels of designed acceptors are calculated using density functional calculations. Electronic properties and electrostatic potential are also calculated. Synthetic accessibility of designed acceptors is predicted. The synthetic accessibility score of most acceptors is higher than the experimentally reported acceptor named QIP-4F. Our proposed framework has the potential to easily select efficient materials for organic solar cells in a short time. Fast designing and performance prediction can speed up the goal of commercialization.