New research published in Advanced Functional Materials highlights the most recent advancement in photovoltaic technology developed by researchers at Osaka University. The researchers used machine learning to generate new polymers that could be utilized to enhance photovoltaic capabilities.
First author Kakaraparthi Kranthiraja explains that their machine learning algorithms allowed the Osaka University team to test combinations of the donor polymers and acceptor molecules that compose an organic solar cell at a much faster rate than would have been possible manually. "Basing the construction of our machine learning model on an experimental dataset drastically improved the prediction accuracy," Kranthiraja adds.
As a result of the machine learning algorithms, the team discovered 200,932 pairs of donor:acceptor from the original 382 donor molecules and 526 acceptor molecules. From there, they analyzed the pairs with a virtual screening to assess their energy conversion efficiency. Of all those screened, the team then chose one special pair to actually synthesize in order to test if it measured up to the algorithms’ predictions.
They found that, indeed, the pair that they synthesized did have the predicted properties, suggesting that the algorithms are accurate. The team says that they have high hopes for their technique and the implications it holds for the field of materials science.
"This project may contribute not only to the development of highly efficient organic solar cells but also can be adapted to material informatics of other functional materials," noted senior author Akinori Saeki.
Sources: Advanced Functional Materials, Eureka Alert
