Organic photovoltaic (PV) cells are believed by some researchers to be the way forward for solar power, due to their use of inexpensive, lightweight materials and ease of production. However, it’s been difficult to find materials suitable for these cells with sufficiently high power-conversion efficiencies to rival the silicon-based commercial cells on the market today.
Researchers in Japan believe they may have found a way to solve this problem by using artificial intelligence (AI) to search for polymers that can perform well in organic PV cells, they said.
A team at Osaka University has used computational power to automate the search for well-matched solar materials, which could lead to vastly more efficient devices, they said in a recent press release from Osaka. “We have just constructed an AI-based predictor, and demonstrated how to use it for the design of a new polymer,” explained Akinori Saeki, one of the researchers on the work.
|The graphic shows the exploration of new polymers for polymer solar cells using materials informatics. Researchers at Osaka University in Japan developed an artificial-intelligence predictor for finding polymers that are well-suited to creating highly efficient organic solar cells, such as perovskite cells. (Image source: Osaka University)|
Organic PV cells depend on both the organic and the polymer layer to convert light into electricity. Traditionally, chemists have experimented with different combinations of these by trial-and-error, leading to a lot of wasted time and effort.
To expedite this process, the Osaka team gathered data on 1,200 organic PVs from about 500 studies. They then used Random Forest machine learning to build a model combining the band gap, molecular weight, and chemical structure of these previous OPVs—together with their power-conversion efficiency—to predict the efficiency of potential new devices. Random Forest is an ensemble learning method for classification, regression, and other tasks.
“A polymer for solar cell application is composed of a donor unit, an acceptor unit, a spacer, and alkyl chains,” Saeki explained to Design News. “Assuming 20 kinds of choices for each unit, the combinations exceed 1 million. It is hard to synthesize all of the combinations,” he added.
Even quantum chemical calculations cannot predict the solar-cell efficiency, because the efficiency is a result of complex factors such as film morphology, interface at p-type and n-type semiconductors, and solubility of the materials, he said.
To simplify combining the parameters and thus choosing materials, researchers manually collected about 1,000 experimental parameters—including efficiency, molecular weight, and electronic properties—from the literature and subjected them to machine learning with digitized chemical structures, Saeki said.
“Our AI allows for virtual screening of polymer structures and predicts the efficiency without experiments and simulation,” he explained. “Note that the accuracy is not perfect—around 20 percent to 50 percent. So, this AI can be used as a tool for rough screening.” The team published a paper on their work in The Journal of Physical Chemistry Letters.
From their research, the team developed a new, previously untested polymer, and is working with private companies and other academic institutions to use the AI-based predictor to explore new materials for use in organic PVs, Saeki said.
“These days, organic solar cell [research] is behind the emerging perovskite solar cell, which was reported in 2012 and has efficiency now comparable to those of commercialized inorganic solar cells,” he said to Design News. “So, I strongly hope to revive an organic solar cell by finding a new, high-efficiency material.”
Elizabeth Montalbano is a freelance writer who has written about technology and culture for 20 years. She has lived and worked as a professional journalist in Phoenix, San Francisco, and New York City. In her free time, she enjoys surfing, traveling, music, yoga, and cooking. She currently resides in a village on the southwest coast of Portugal.
|The Critical Power Education You Need Now. Join our in-depth conference program to learn about managing risk, reducing operational costs, maximizing microgrids, and much more. Critical Power Expo. Sept. 11-13, 2018, in Novi, MI. Register for the event, hosted by Design News’ parent company UBM.|