Machine learning assisted designing of organic compounds with high enthalpy of melting for batteries

The thermal stability of organic materials plays a crucial role in the safe and efficient operation of next-generation batteries. This study presents a machine learning (ML)-assisted framework for predicting and designing organic compounds with high enthalpy of melting (ΔH_m), a key property for enhancing thermal robustness. A dataset comprising over 4800 compounds was used to train and validate multiple ML models based on RDKit-derived molecular descriptors. Among the evaluated algorithms, the neural network model showed the best generalization performance with minimal overfitting. This model was then applied to virtually screen more than 50,000 compounds from the Harvard Organic Photovoltaic Database, successfully identifying 50 top candidates with high ΔH_m and favorable synthetic accessibility. The results highlight the utility of ML in accelerating the discovery of thermally stable organic compounds for energy storage applications.