Designing Efficient Tandem Organic Solar Cells with Machine Learning and Genetic Algorithms

Abstract

Tandem organic solar cells can potentially drastically improve the PCE over single-junction devices. However, there is limited research on device development and often only ca. 1% improvement over single-junction devices. Because of the complex nature of organic material compatibility and properties, such as energy-level alignment and maximizing absorption spectra, and the vastness of chemical space, computational guidance is vital. The first part of this work uses a new data set of 1,225 donor/non-fullerene acceptor (NFA) pairs containing 1,001 unique pairs, one of the largest to date, to train an ensemble machine learning model to predict device efficiency (RMSE =1.60 +/- 0.14%). Next, a series of genetic algorithms (GA) are used to discover high-performing NFAs and polymer donors, and then combinations of them for potential high-efficiency tandem cells. Interesting design motifs show up in high-performing NFAs, such as diphenylamine substituents on the core and 3D terminal groups. The donor polymers from the GA reveal that it may be beneficial to arrange the monomers as a small-block copolymer instead of the common alternating copolymer. The GAs for selection of tandem cell materials successfully find material combinations, that when in a device together, have strong absorption across the entire visible-near-IR spectrum. Computational guidance is critical for the selection of tandem OSC materials, with genetic algorithms proving a highly successful technique.

Best Practices for Using Genetic Algorithms in Molecular Discovery

“…These aforementioned methods are all classified as fitness proportionate methods, meaning that higher-scored individuals have a higher chance of being selected, and the entire population is accessible for selection. A different and much simpler method is "random" selection, where 2 parents are randomly selected either from the entire population or from some top percentage of individuals 3,8,12,14 . By selecting from only the top candidates, this selection can allow for faster convergence.…”

“…The individuals in each population need to be scored to assess how well they optimize a given property. The fitness function is a model that scores each individual and can range widely in complexity, from relatively simple tasks such as maximizing molecular weight to requiring ML models for tasks such as predicting the power conversion efficiency of organic solar cells 12 . In most GAs, the fitness function is the time-limiting step, so care must be taken to use cost-effective calculations.…”