Machine learning-assisted development of organic photovoltaics via high-throughput in situ formulation

Abstract

The discovery of high-performance non-fullerene acceptors and ternary blend systems has resulted in a breakthrough in the efficiency of organic photovoltaics (OPVs) and has created new opportunities for commercialization. However, manufacturing technology has remained far behind expectations. Here we show a new research approach to develop OPVs via industrial roll-to-roll (R2R) slot die coating in conjunction with the in situ formulation technique and machine learning (ML) technology. The formulated PM6:Y6:IT-4F ternary blends deposited on continuously moving substrates resulted in the high-throughput fabrication of OPVs with various compositions. The system was used to produce training data for ML prediction. The composition/deposition parameters, referred to as deposition densities, and the efficiencies of 2218 devices were used to screen ML algorithms and to train an ML model based on a Random Forest regression algorithm. The generated model was used to predict high-performance formulations and the prediction was experimentally validated by fabricating 10.2% efficiency devices, the highest efficiency for R2R-processed OPVs so far.