With the massive size of chemical space, computational guidance is vital for inverse design problems such as high-efficiency organic photovoltaic (OPV) materials. Current experimental power conversion efficiencies (PCE) have only reached approximately half of their theoretical limit, suggesting possibilities for improvement. Although many device parameters contribute to the PCE, a key factor is the solar absorption and compatibility of materials in the active layer. In organic solar cells (OSC), the p-type and n-type active layer materials need to have a precise energy alignment and complementary solar absorption. In tandem cells, the opportunity for additional active-layer materials allows us to reach higher PCEs but at the cost of more complex material selection.<br/> <br/>We have created a pipeline combining machine learning and a genetic algorithm to efficiently screen chemical space and select ideal materials for tandem OSCs. In the first stage, a genetic algorithm discovers new electron acceptors and donors that have optimal electronic structure alignment and complementary absorption. The PCE of these devices is predicted using an ensemble machine learning model trained on 1000 experimental OPV devices. In the second stage, the top three performing devices from the first stage are selected as target systems for one layer of a tandem cell. A genetic algorithm is used to find optimal materials for the second cell that maximize absorption of different regions of the solar spectrum while minimizing overlap. Using this computational-guided approach, we offer candidates for highly efficient tandem OSCs and a method for screening potential new materials.

Selection of optimal materials for tandem organic solar cells using machine learning and genetic algorithms

Fall 2022 abstracts, posters, and presentations will launch in mid-September.

ACS Fall 2022

The official archive and repository for all ACS Meetings & Expositions abstracts and select posters or presentations.