Autonomous closed-loop framework for reproducible perovskite solar cells.

Area of Science:

• Materials Science
• Chemical Engineering
• Renewable Energy

Background:

• Perovskite solar cell commercialization is hindered by inefficient, manual methods for material discovery and device fabrication.
• Current approaches rely heavily on human expertise, leading to slow progress and reproducibility issues.

Purpose of the Study:

• To create an autonomous closed-loop framework integrating machine learning (ML)-driven material discovery with automated manufacturing for perovskite solar cells.
• To overcome the limitations of trial-and-error methods and accelerate the development of high-performance, stable perovskite solar cells.

Main Methods:

• An integrated system combining active learning and quantum modeling for rapid identification of high-performance molecules.
• A closed-loop feedback system using Bayesian optimization and symbolic regression to refine the automated fabrication process.
• Utilized machine learning for autonomous material discovery and high-fidelity automated manufacturing.

Main Results:

• Discovery of a novel passivation molecule, 5-(aminomethyl)nicotinonitrile hydroiodide (5ANI).
• Achieved high power conversion efficiencies (PCEs): 27.22% for 0.05 cm² solar cells and 23.49% for 21.4 cm² mini-modules.
• Demonstrated excellent long-term operational stability, retaining 98.7% efficiency after 1,200 hours.
• Automated fabrication achieved nearly 5x higher efficiency reproducibility compared to manual methods.

Conclusions:

• The developed autonomous closed-loop system effectively synergizes ML-powered discovery with automated manufacturing.
• This integrated approach sets a new benchmark for autonomous discovery and manufacturing in photovoltaics and materials science.
• The system significantly enhances efficiency, stability, and reproducibility in perovskite solar cell production.