Predictive methods for the heat of decomposition of reactive chemicals: From CHETAH, QSPR, and quantum chemical calculations to deep learning

The heat of decomposition of reactive chemicals is a critical parameter for characterizing their thermodynamic properties and has broad application across various essential fields, including hazard identification, chemical classification, and safety risk assessment of chemical reactions. Currently, the primary method for determining the heat of decomposition of compounds relies on experimental techniques, which are not only complex but also inherently associated with significant potential hazards. The existing predictive methodologies are still in the developmental stage. This paper first analyzes and summarizes the current research progress on the prediction of heat of decomposition, then provides an in-depth exploration of the three most prominent predictive methods: the Chemical Thermodynamic and Energy Release Computer Program (CHETAH) program, quantum chemistry calculation, and Quantitative Structure-Property Relationship (QSPR) methods. It focuses on their underlying prediction mechanisms, accuracy, advantages, and limitations. Among these, quantum chemical computation methods and the QSPR approach have demonstrated superior predictive performance. The study specifically emphasizes that integrating deep learning techniques can overcome existing bottlenecks: transfer learning can mitigate the challenge of limited samples in QSPR modeling, while large language models (LLMs) in chemistry can address the prediction difficulties of decomposition reaction equations. These innovative directions are expected to significantly enhance predictive accuracy and provide crucial technical pathways for future research.