Machine Learning Accelerated Design of Self-Assembled Monolayers for High-Performance Perovskite Solar Cells.

Area of Science:

• Materials Science
• Computational Chemistry
• Renewable Energy

Background:

• Self-assembled monolayers (SAMs) are advanced hole transport materials (HTMs) for perovskite solar cells (PSCs).
• Current SAM discovery relies on inefficient empirical methods, leading to long development cycles and high costs.

Purpose of the Study:

• To develop a machine learning (ML) platform for accelerated SAM discovery and design.
• To improve the efficiency and stability of perovskite solar cells through novel SAMs.

Main Methods:

• Constructed a feature space using RDKit molecular descriptors and Morgan fingerprints.
• Evaluated multiple ML algorithms, with XGBoost showing optimal performance.
• Utilized SHapley Additive exPlanations (SHAP) for structure-property relationship analysis.

Main Results:

• The RDKit-based XGBoost model achieved high accuracy (RMSE: 1.862, R²: 0.5058, r: 0.8161).
• SHAP analysis identified key molecular features influencing photovoltaic performance.
• External validation showed prediction errors within 10% for reported SAMs.
• Designed three new SAM molecules with predicted PCEs up to 27%.

Conclusions:

• The ML platform offers an efficient digital solution for SAM development.
• This approach significantly accelerates the discovery of next-generation photovoltaic materials.
• Provides valuable guidance for designing high-performance SAMs for PSCs.