Organic polymer energy conversion materials are increasingly vital in renewable energy and modern electronics. Leveraging machine learning to explore structure-property relationships and guide rational material design has become a key research focus. However, obtaining reliable computational and experimental data for polymers remains challenging. To address this, we curated the Organic Polymer Energy Conversion Materials (OPECM) Dataset, featuring 3225 polymers across three unified categories: organic semiconductors, photovoltaics, and dielectric materials. This dataset integrates structural and property data from over thousands of sources, with three key advances: First, it establishes experimental benchmarks for all three material classes alongside stable molecular structures and high chemical diversity. Second, it introduces polymer units as fundamental building blocks, providing a novel data organization framework where material functions are mapped to designable structural units (e.g., π-conjugated cores for semiconductors/photovoltaics, polar groups for dielectrics). Third, each entry is augmented with quantum-chemical descriptors from high-throughput DFT calculations, establishing OPECM as an integrated resource for QSPR modeling in energy conversion. Using this dataset, we developed polymer unit-guided regression models to predict key properties: electron/hole mobility (${\mu }_e/{\mu }_h$), power conversion efficiency (PCE), and dielectric constant ($\varepsilon$), while identifying the key polymer units that critically influence material performance. Furthermore, by applying the SISSO method, we derived interpretable symbolic regression models, which reveal critical molecular features and their impact on energy performance. This work aims to inspire further research into the targeted design of multi-property energy materials.

Close

PrevNext

View figure in article

• Organic electronics
• Photovoltaic absorbers
• Polymers
• Density functional theory
• Machine learning
• Regression analysis