Treffer: Designing of symmetric and asymmetric small molecule acceptors for organic solar cells: A farmwork based on Machine learning, virtual screening and structural analysis.

[Display omitted] • Machine learning (ML) models are made to predict efficient solar cells. • The small molecule acceptors (SMAs) have been designed through ML. • Molecular descriptors are used for model training. • Efficient solar cell acceptors have been designed based on data-driven framework. Organic solar cells (OSCs) have drawn a lot of interests because of their distinctive qualities, including flexibility and tunability. In present study, a detailed data-driven framework is introduced to design the small molecule acceptors (SMAs). Machine learning (ML) models are constructed to predict the energy levels of SMAs. Molecular descriptors are used for the model training. The influence of different descriptors on the output of best ML model is explained through SHapley Additive exPlanations (SHAP). Positive, negative and mean SHAP values are calculated. 5000 SMAs are generated using Breaking Retrosynthetically Interesting Chemical Substructures (BRICS) method. Energy levels of SMAs are predicted through best ML model. The SMAs are shortlisted on basis of band gap, SMAs with lower band gap are retained. The synthetic accessibility of selected SMAs is tested through Python software Synthetic Bayesian classifier (SYBA). The majority of SMAs are easy to synthesize. The structural similarity between selected SMAs is also studied. That is indicating that selected SMAs are disperse in nature. By effectively identifying and optimising new SMAs, the developed approaches can increase the chances to find the better materials. [ABSTRACT FROM AUTHOR]