Discovering Process Dynamics for Scalable Perovskite Solar Cell Manufacturing with Explainable AI

Area of Science:

• Materials Science
• Energy Science
• Photovoltaics

Background:

• Large-area processing of perovskite thin-films presents challenges in quality control, hindering commercialization of perovskite photovoltaics.
• Current scalable fabrication methods rely on inefficient trial-and-error approaches.
• In situ photoluminescence (PL) video data offers insights but is too complex for manual analysis.

Purpose of the Study:

• To apply deep learning (DL) and explainable artificial intelligence (XAI) to understand perovskite thin-film formation.
• To correlate sensor data during processing with solar cell performance.
• To translate complex data relationships into understandable and actionable recommendations for industrial-scale manufacturing.

Main Methods:

• Utilized deep learning (DL) algorithms to analyze high-dimensional in situ photoluminescence (PL) video data.
• Employed explainable artificial intelligence (XAI) techniques to interpret the relationships discovered by DL models.
• Integrated sensor data from the thin-film formation process with performance indicators of the final solar cells.

Main Results:

• Identified key relationships between sensor data acquired during perovskite thin-film processing and resulting solar cell performance.
• Successfully rendered complex data-performance correlations understandable through XAI.
• Demonstrated the potential for XAI to guide process optimization and improve manufacturing consistency.

Conclusions:

• Deep learning and XAI are powerful tools for accelerating materials science research in energy applications.
• Explainable AI is crucial for bridging the gap between complex data analysis and practical industrial implementation.
• This approach facilitates the development of robust and scalable manufacturing processes for perovskite solar cells.