摘要：This work reflects the worth of intelligent modeling in controlling the nanostructure morphology in manufacturing organic bulk heterojunction (BHJ) solar cells. It suggests the idea of screening the pool of material design possibilities inspired by machine learning. To fulfill this goal, a set of experimental data on a BHJ solar cell with a donor structure of diketopyrrolopyrrole (DDP) and backbone of benzothiadiazole (BT) are fed into a home-written artificial neural network (ANN)/genetic algorithm (GA) hybrid code to optimize film-casting parameters. The annealing temperature, spin coating spin rate, and donor/acceptor ratio taken from available literature are applied to give the machine chance of learning trends in the power conversion efficiency (PCE). DPP-BT structures virtually born in the mind of machine are then screened for resemblance survey to receive a tag of desired characteristic. The results enable device manufacturers to identify the sensitivity of designed molecules to specific film casting conditions, while homologous structures may result in similar responses against design variables.
关键词：Organic electronics;Bulk heterojunction solar cell;Machine learning;Artificial intelligence;Molecule tagging
