Machine learning assisted designing of polymers with lower reorganization energies for the possible use as donor and acceptors for organic solar cells

• Published in: Solar energy 2025-01, Vol.286, p.113169, Article 113169
• Main Authors: Hussain, Ejaz, Soliman, Mohamed Mohamed, El-Bahy, Salah M., Naeem, Sumaira, Khan, Numan
• Format: Article
• Language: English
• Subjects: Machine learning; Molecular descriptors; Organic solar cells; Reorganization energy
• ISSN: 0038-092X
• DOI: 10.1016/j.solener.2024.113169

[Display omitted] •Machine learning model is used to predict the reorganization energies for electron and hole transfers.•Shapley Additive Explanations are also performed to interpret and understand the contributions of individual features.•10,000 polymers are generated by using the Breaking Retrosynthetically Interesting Chemical Substructure Methodology.•The t-SNE plots are used to illustrate the generated the chemical space. This study presents a novel method for accurate prediction of reorganization energy descriptors which are critical for optimizing the performance of organic solar cells by employing machine learning (ML) based techniques. Traditional methodologies have drawbacks in accurately estimating these descriptors, which affects the dependability and efficiency of organic photovoltaic devices. The engaged AI-based methodology offers a quantitative understanding of these descriptors, which greatly improves capacity to predict and optimize the efficiency of organic solar cells. Novel polymers were systematically generated using the Breaking Retrosynthetically Interesting Chemical Substructures (BRICS) approach. Electron and hole rearrangement energies were predicted using AI-driven predictive models, revealing common ranges and distribution patterns. This study has demonstrated how AI-driven approaches have the potential to transform high-performance organic photovoltaic design and development, leading to breakthroughs in renewable energy technologies.