Accelerating the search of CHONF-containing highly energetic materials by combinatorial library design and high-throughput screening

Simple rule for linking atoms to construct high energy isomers

“…5 Nowadays, the big data-driven molecular design becomes increasingly popular with the further enhancement of efficiency, and numerous fuel molecules and energetic molecules with excellent properties have been designed and synthesized. 6–9…”

“…5 Nowadays, the big data-driven molecular design becomes increasingly popular with the further enhancement of efficiency, and numerous fuel molecules and energetic molecules with excellent properties have been designed and synthesized. [6][7][8][9] Still, less attention has been paid to some basic issues in constructing new molecules. Among them, the rule for the efficient construction of high energy molecules is still absent from the molecular design of fuels and energetic materials.…”

Impact sensitivities of energetic materials derived from easy-to-compute ab initio rate constants

“…The design of energetic materials must consider their resilience to external loads, such as mechanical impacts, shock waves, friction, sparks and others. [1][2][3] To ensure their safe handling, impact sensitivity is of special significance. The experimental determination of this property involves dropping a hammer from a height h onto a confined sample and detecting any decomposition event.…”

Correlated RNN Framework to Quickly Generate Molecules with Desired Properties for Energetic Materials in the Low Data Regime

“…Alternatively, based on the size of the energetic data and the advantage of RNN, we explore a molecular designer via augmenting data and correlating multiple RNN models by the transfer learning strategy to generate new energetic molecules with desired properties. Herein, we focus on the most widely used CHNO-containing energetic molecules, and the detonation velocity is selected as a desired property as it is one of the most important properties characterizing the detonation performance of the energetic molecules and a main determinant for the target of experimental synthesis. ,− Our DL-based generator consists of a pretrained RNN model, a generative RNN model focusing on the energetic molecules, and a predictive RNN model. The pretrained RNN is designed to correlate with the predictive model and the generative model by a transfer learning way, through which sufficient structure knowledge can be introduced into the predictive model and the generative model to improve their performance.…”