Perovskite solar cells are gaining significant momentum in the search for cheaper and more efficient solar energy. However, the material degrades over time, limiting its widespread commercial use.
To overcome one of the biggest barriers to commercializing perovskite solar cells, Marina Leite, professor of materials science and engineering at the University of California, Davis, and an interdisciplinary team of researchers are leveraging AI.
In a paper published in Advanced Materials, researchers demonstrated how AI can dramatically accelerate research progress. Rather than relying solely on trial-and-error experiments, the team used AI to learn from thousands of automated experiments to accurately predict how new material compositions would respond to heat, a key environmental stressor, and more quickly identify the most promising materials.
Search for stable perovskites
Compared to traditional silicon solar cells, perovskites are lighter, more flexible, cheaper to manufacture, and more efficient. However, it is unstable to environmental stressors such as heat, moisture, and light, which limits its scalability.
“Our scientific community is very interested in understanding the chemical and physical processes that cause the stability, or lack thereof, of these materials,” Leite said.
It is nearly impossible to test all possible perovskite material compositions under all environmental conditions. Instead, Leite’s team asked whether AI could be taught to recognize patterns and predict behavior from a carefully selected subset of experiments.
The team tested 10 different perovskite compositions by subjecting them to repeated temperature cycles, resulting in 137,000 unique measurements. The measurements were used to train a machine learning model that can accurately predict how previously untested material compositions will respond to repeated heating.
From these predictions, the researchers determined which compositions are more thermally stable. Compositions with low levels of cesium, a highly reactive alkali metal, generally recover with repeated heating, whereas compositions with high cesium content are more likely to permanently decompose.
AI: Research partner
The discovery provides researchers with a roadmap for developing more durable perovskite solar cells. Instead of experimentally testing thousands of possible recipes, you can now focus your efforts on the candidates that are most likely to withstand real-world operating conditions.
“AI does not replace experimentation in our research,” Leite said. “Rather, it can be used to increase the efficiency of scientific discovery.”
Leite and his collaborators used AI to learn relationships from a limited number of high-throughput experiments. Using appropriate algorithms, AI was able to accurately predict the stress response behavior of halide perovskites under previously unmeasured conditions.
Ultimately, Leite hopes the same approach can be used to predict how perovskite materials will perform in different climates around the world, allowing researchers to design solar cells tailored to real-world operating conditions and paving the way to more efficient solar energy.
“Successfully implementing a machine learning model to analyze experiments is an important demonstration of how AI can truly help.”
Co-authors of this paper include Abigail Herring and Dr. Mansha Dubey. student in materials science and engineering at the University of California, Davis. Meghna Srivastava, a graduate student in materials science and engineering; PhD student Elaheh Hosseini and Professor of Electrical and Computer Engineering Homan Homayun from the University of California, Davis. Yu An and Juan-Pablo Correa-Baena of Georgia Tech;
The project received funding from several organizations, including the Defense Advanced Research Projects Agency, the National Science Foundation, and the Department of Energy.
