Accelerating the discovery of high-performance donor/acceptor pairs in photovoltaic materials via machine learning and density functional theory

Abstract

The gigantically unexplored chemical space for the combinations of donor and acceptor poses a mammoth challenge in enhancing power conversion efficiency. Herein, a strategy integrating machine learning and density functional theory was proposed to assist the rational screening of prominent donor/acceptor pairs. Notably, 10 promising donor/acceptor pairs were ultimately sieved out with power conversion efficiency lager than 18.22%. Furthermore, the density functional theory computation expressed that the new donor/acceptor pair of D18/BTP-eC9 possessed stronger absorption intensity, while the KCS/KCR (ratio of charge separation rate and charge recombination rate) and μe (electron mobility) were larger than those of the best known D18/Y6 with the increments of 201.31% and 125.98% respectively. The SHapley Additive exPlanations analysis revealed that the two most important features of A_nR08 and D_D/Dtr12 were positively related to the improvement of power conversion efficiency. It was found that A_nR08 was positively associated with π-conjugation strength, and high D_D/Dtr12 suggested the highly fused chemical structure, indicating the practical clues in discovering top-performing donor/acceptor pairs.