Generative Chemical Language Models for Energetic Materials Discovery.

The discovery of new energetic materials remains a pressing challenge, hindered by the limited availability of high-quality data. To address this, we have developed generative chemical language models that have been pretrained on extensive molecular data and then fine-tuned with curated energetic material data. This transfer learning strategy extends the chemical language model capabilities beyond the pharmacological space in which they have been predominantly developed, offering a framework that is applicable to other data-spare discovery problems. After fine-tuning, we observe modest distributional shifts in predicted detonation performance as benchmarked by density functional theory as well as the presence of chemical motifs associated with energetic materials. Furthermore, we discuss the benefits of fragment-based molecular encodings for chemical language models, in particular, in constructing synthetically accessible structures. Together, these advances provide a foundation for accelerating the design of next-generation energetic materials.