Discovery of Novel Energetic Materials via High-Throughput Computations and Machine Learning

Modern challenges demand novel multi-functional materials, such as energetic materials (EMs) with high-performance and thermal stability. However, difficulty in synthesis and a weak understanding of the features that drive macroscopic-level properties can lead to protracted cycles of trial-and-error. Atomistic simulations can accelerate this process, but even these techniques are not efficient enough to scan the vast chemical landscape. Advances in computer hardware and large language models have enabled the development of artificial intelligence (AI) models that can achieve high fidelity predictions of molecular properties from basic structural information. By training on high-quality simulation data, AI models can be employed to rapidly identify novel EMs.

In this work, we demonstrate our high-throughput workflow for density functional theory calculations which generates a rich database of properties related to the performance and thermal stability of EMs. This database is used to train a surrogate machine learning (ML) model capable of screening vast numbers of molecules to assess their potential as stable and performant EMs. In addition, the calculated properties are analyzed using interpretable ML models to explore factors that govern thermal stability of EMs.