GIFT Professor Yixin Zhao's Team Publishes in Science: AI-Guided Design of Efficient and Stable Perovskite Solar Cells

On May 15, 2026, a team from the Future Photovoltaics Research Center at the Global Institute of Future Technology, Shanghai Jiao Tong University, published a paper in Science in collaboration with several partners. The paper titled “AI-guided design of efficient perovskite solar cells operationally stable at 100 ˚C” introduces a four-agent collaborative artificial intelligence (AI) for perovskite solar cell development, enabling holistic design of device architecture, material composition, and interfaces to achieve both high efficiency and exceptional operational stability.

The co-first authors of the paper are Jiahao Guo, a doctoral degree student from the School of Environmental Science and Engineering, SJTU; Bowei Li, Assistant Research Professor from GIFT Future Photovoltaics Research Center; Zeyu Zhang, Postdoctoral Fellow from GIFT Future Photovoltaics Center; and Fang Liu, Postdoctoral Fellow from the School of Environmental Science and Engineering, SJTU. Assistant Professor Congyi Li and Assistant Research Professor Yao Wang are the co-authors. Associate Professor Yanming Wang, Associate Research Professor Yanfeng Miao, and Professor Yixin Zhao serve as the co-corresponding authors.

Figure 1: The multi-agent AI workflow

Perovskite solar cells are promising for future photovoltaics due to their high efficiency and low cost. However, challenges remain in stability and degradation under operational conditions such as high temperature, bias, and illumination. Traditional trial‑and‑error approaches struggle with the multi‑variable optimization required. The AI platform developed by the researchers integrates fragmented literature, experimental data, and model analyses into viable design recommendations. The resulting devices exhibit outstanding stability under high‑temperature operation, demonstrating the potential of AI‑driven methods for advanced energy materials research.

Figure 2: Perovskite composition and transport layer design, and device stability

The platform comprises several specialized agents: a data agent extracts structured information from literature and experiments; a central agent manages task planning, knowledge search, workflow coordination, and device architecture suggestions; a composition agent optimizes the perovskite absorber layer and analyzes stability; and an interface agent screens and designs transport layers and interfacial materials. Through multi‑agent collaboration, the platform converts distributed knowledge into executable research recommendations, significantly accelerating the development of material systems.

Figure 3: Perovskite multi‑agent AI design platform

Guided by the agents, the researchers proposed a highly stable device structure with a double‑layer alumina protective layer, and further optimized the formamidinium‑cesium (FA‑Cs) perovskite composition and hole‑transport material. The composition agent indicated that moderate Cs incorporation helps regulate defects and enhance thermal stability. The researchers also designed a biphenyl‑based hole‑transport molecule (MeO‑DPPACz) with high carbon‑nitrogen bond energy that improves ultraviolet-light stability. The resulting perovskite solar cells achieved a power conversion efficiency of 25% and retained 97% of initial efficiency after 1000 hours of continuous operation at 100 ˚C.

This study demonstrates the value of AI in designing complex material systems. By combining multi‑agent collaboration, knowledge search, and data‑driven analysis, it greatly facilitates the development of highly stable perovskite solar cells.

The study was supported by SJTU's high-performance computing infrastructure along with a locally deployed DeepSeek‑V4 large language model. The AI platform is currently in internal trials and is expected to be launched soon. In the near future, it will be extended to cover more perovskite photovoltaic R&D scenarios, providing a general and reusable technical framework for intelligent development of energy science.

Author Profiles

Bowei Li

Assistant Professor at GIFT Future Photovoltaics Research Center. Prof. Li received his Ph.D. degree from the University of Surrey, UK. His research focuses on inverted perovskite solar cells, device stability, and degradation mechanisms. He has published over 30 papers in journals such as Science, Nature, Nat. Photonics, J. Am. Chem. Soc., and Adv. Mater.

Zeyu Zhang

Postdoctoral Fellow at GIFT Future Photovoltaics Research Center. Dr. Zhang received his Ph.D. degree from New Jersey Institute of Technology, USA. He has published more than 20 papers in journals including Science, Acc. Chem. Res., ACS Nano, Angew. Chem. Int. Ed., and Adv. Mater.

Yanming Wang

Associate Professor at GIFT Future Photovoltaics Research Center. Prof. Wang received his Ph.D. degree from Stanford University, USA. His research focuses on multi‑scale modeling and AI for advanced energy materials. He has published over 80 papers in journals such as Science, Nature, Nat. Commun., and Sci. Adv., holds 1 US patent and 7 software copyrights, and leads more than 10 projects funded by NSFC, MOST, and the Shanghai Municipal Commission of Science and Technology. He serves as a young editorial board member for the Journal of Energy Chemistry.

Yixin Zhao

Chair Professor and Director of the Future Photovoltaics Research Center. Prof. Zhao received his B.S./M.S. from Shanghai Jiao Tong University (2002/2005) and Ph.D. from Case Western Reserve University (2010), with postdoctoral experience at Penn State University and National Renewable Energy Laboratory. His honors include the National Young Talents Program, Fok Ying-Tong Education Fund of China, Shanghai Dawn Scholar Program, Shanghai Outstanding Young Academic Leader, National Science Fund for Distinguished Young Scholar, the first place in the Shanghai Natural Science Award, and Qingshan Science and Technology Award. He has published over 200 papers in journals, including Science, Nature, Nat. Mater., Nat. Sustain., Natl. Sci. Rev., JACS, Angew. Chem. Int. Ed., and Adv. Mater., with over 25,000 citations. He has been listed as a Clarivate Highly Cited Researcher for eight consecutive years (2018-2025), and one of his papers was selected as one of “China's 100 Most Influential International Academic Papers of 2019”.