Toward Predicting Efficiency of Organic Solar Cells via Machine Learning and Improved Descriptors

Organic solar cells (OSCs) have been dominated by donor:acceptor blends based on fullerene acceptors for over two decades. This situation has changed recently, with non-fullerene (NF) OSCs developing very quickly. The power conversion efficiencies of NF OSCs have now reached a value of over 13%, which is higher than the best fullerene-based OSCs. NF acceptors show great tunability in absorption spectra and electron energy levels, providing a wide range of new opportunities. The coexistence of low voltage losses and high current generation indicates that new regimes of device physics and photophysics are reached in these systems. This Review highlights these opportunities made possible by NF acceptors, and also discuss the challenges facing the development of NF OSCs for practical applications.

Fossil fuel alternatives, such as solar energy, are moving to the forefront in a variety of research fields. Polymer-based organic photovoltaic systems hold the promise for a cost-effective, lightweight solar energy conversion platform, which could benefit from simple solution processing of the active layer. The function of such excitonic solar cells is based on photoinduced electron transfer from a donor to an acceptor. Fullerenes have become the ubiquitous acceptors because of their high electron affinity and ability to transport charge effectively. The most effective solar cells have been made from bicontinuous polymer-fullerene composites, or so-called bulk heterojunctions. The best solar cells currently achieve an efficiency of about 5%, thus significant advances in the fundamental understanding of the complex interplay between the active layer morphology and electronic properties are required if this technology is to find viable application.