Discovery of high-energy nitro-azolo-pyridine compounds with high-throughput computation and machine learning

Abstract

The nitro-azolo-pyridine compounds have planar molecular backbones with high nitrogen content, which are considered as a candidate system of insensitive high-energetic explosive (IHE). In this work, 30969 energetic molecules were generated based on 7 kinds of azolo-pyridine backbones; among them 2000 sample molecules were chosen to perform high-throughput calculations (HTC) of detonation velocity and pressure, then the machine learning (ML) models were constructed according to the computed molecular descriptors and detonation parameters; the detonation parameters for the other molecules besides the HTC samples were predicted by the ML models, and the molecular stability for the highest energy molecules were computed using quantum chemistry. Finally, 3 nitroazolo-pyridine compounds with high-energy and low-sensitivity were screened, and their synthetic accessibility were verified.