Data-mining and machine learning based search for optimal materials for perovskite and organic solar cells

• Published in: Solar energy 2025-02, Vol.287, p.113223, Article 113223
• Main Authors: Ahmad, Nafees, Ibrahim, Mahmoud A.A., Sayed, Shaban R.M., Ahmad Shah, Syed Shoaib, Tahir, Mudassir Hussain, Zou, Yingping
• Format: Article
• Language: English
• Subjects: Hole mobility; Machine learning; Perovskite solar cells; Regression analysis; Reorganization energy; Small molecules
• ISSN: 0038-092X
• DOI: 10.1016/j.solener.2024.113223

•Innovative data mining and machine learning approach for organic semiconductor discovery.•Efficient screening of photovoltaic materials using energy level predictions.•Selection of optimal materials for perovskite and organic solar cells.•Robust machine learning model selection: Random Forest outperforms 40 + models.•Synthesis-friendly organic semiconductors identified for practical application. A data mining-based approach is introduced to search the organic compounds for Photovoltaics applications. Organic semiconductors are search from a database of organic compounds having lower reorganization energy for hole transfer. Three polymer donors are selected as a standard structure to search similar materials from database. Energy levels are predicted using machine learning as a screening criterion for the selection of best materials for photovoltaics applications. Fingerprints are used for training the machine learning models. More than 40 machine learning models are tried, random forest has appeared as a best model (r-squared of 0.800 and 0.609 for training and test set, respectively). This machine learning model is used to predict the energy levels of new materials. Synthetic accessibility of selected organic semi-conductors is also predicted, all these semi-conductors are straight-forward to synthesize. Their Synthetic accessibility (SA) score is less than 6.