A scientific journal article published by scientists from Osaka University’s school of applied sciences considers the use of artificial intelligence in designing new material combinations for organic photovoltaic (OPV) cells.
OPVs have struggled to reach commercially competitive power conversion efficiencies, even though several companies worldwide have invested large sums into R&D related to the technology. Organic photovoltaics are made from inexpensive, lightweight materials are safe to handle and benefit from easy production. However their power conversion efficiencies (PCEs) are still too low for full-scale commercialization. PCE depends on the organic and polymer layers and chemists have experimented with different combinations without finding a sufficient improvement in efficiency to date.
Big data informatics can help scientists navigate large, complex datasets, identifying potentially important statistical trends. The Japanese scientists mentioned in the article fed data from 1,200 OPVs sourced from around 500 studies into an artificial intelligence package, to navigate the otherwise time consuming trial-and-error phase. By analysing the performances of each material used, and considering its interplay with other compounds, the machine learning software can draw conclusions about possible combinations of different organic molecules and polymers faster than humans could. The AI makes performance predictions of the new material constellations it generates.
According to the Japanese university’s scientists, the prediction for the first tested material constellation was wrong, as the material under-performed. They stress, however, additional data added to the system will improve the process.
“Machine learning could hugely accelerate solar cell development since it instantaneously predicts results that would take months in the lab,” says Akinori Saeki, co-author of the article published in the Journal of Physical Chemistry Letters. “It’s not a straightforward replacement for the human factor – but it could provide crucial support when molecular designers have to choose which pathways to explore.”