Solar cells based on perovskite - a material that absorbs light well - have not yet reached the mass market: the technology is too capricious. The film is made from a liquid solution that dries and crystallizes, but the slightest fluctuations in temperature or humidity change its properties and reduce output. A group of scientists from Osaka University has proposed an automated platform that combines a robotic setup, optical imaging and machine learning. The results were published in the journal ACS Nano.
The platform applies the film and simultaneously records how its color changes and how droplets of antisolvent behave - the liquid that is added to accelerate crystallization. These processes proceed rapidly, so measurements must be very fast: the system captures the color change in fractions of a second, and the movement and impact of droplets in several milliseconds (thousandths of a second).
"We want to move away from trial and error by measuring what happens to the film directly during its formation," the press release quotes lead author Ryosuke Kado as saying. "Our system makes it possible to link these changes in real time to film quality and identify the most important factors."
Senior author Akinori Saeki adds: "The advantage of automation is that we collect far more data than can be obtained manually. Machine learning helps make sense of this multitude of variables and find connections that are hard to notice."
The researchers applied the knowledge gained to improve the application of the antisolvent. The analysis identified the optimal range of droplet behavior, and with its help a solar cell based on lead perovskite was fabricated. Under these conditions, the device showed a power conversion efficiency (PCE) of 24.4% - the share of sunlight that it converts into electricity. This is a demonstration that data collected during film growth can be used to improve a real cell.
The authors emphasize: the platform offers a more systematic approach to optimizing production, and the combination of automated experiments, real-time measurements and data analysis can reveal previously unknown connections between processing conditions and film quality. The scientists hope that their development will accelerate research in photovoltaics and help create more practical solar batteries. There is as yet no talk of industrial adoption or timelines for market entry.