Advanced High-Throughput Rational Design of Porphyrin-Sensitized Solar Cells Using Interpretable Machine Learning.

Area of Science:

• Materials Science
• Renewable Energy
• Computational Chemistry

Background:

• Predicting power conversion efficiency (PCE) is vital for designing dye-sensitized solar cells (DSSCs).
• High-throughput screening of novel dye sensitizers requires accurate predictive models.
• Zn-porphyrin dyes are a key focus for solar cell development.

Purpose of the Study:

• Develop precise, predictive, and interpretable machine learning (ML) models for Zn-porphyrin-sensitized solar cells.
• Utilize theoretically computable molecular descriptors (MDs) for model development.
• Facilitate rational design and efficient screening of new dye sensitizers.

Main Methods:

• Employed machine learning (ML) algorithms for PCE prediction.
• Leveraged molecular descriptors (MDs) that are computable, effective, and reusable.
• Validated models using a "blind test" on 17 newly designed cells.
• Utilized SHAP analysis to identify key molecular descriptors.

Main Results:

• Achieved excellent predictive performance with a mean absolute error (MAE) of 1.02% on the blind test.
• Successfully predicted PCE for 10 dyes within a 1% error margin.
• Identified crucial molecular descriptors through SHAP analysis, aligning with experimental findings.
• Discovered promising Zn-porphyrin-based dyes with high PCE.

Conclusions:

• The developed ML models are accurate, predictive, and interpretable for Zn-porphyrin DSSCs.
• The models enable efficient in silico screening, reducing experimental analysis time.
• The findings provide valuable chemical guidelines for rational dye design in DSSCs.
• A publicly accessible prediction tool is available to aid researchers.