Reaction Network of Ammonium Perchlorate (AP) Decomposition: The Missing Piece from Atomic Simulations

Abstract

The decomposition network of ammonium perchlorate (AP) is essential for combustion performance and safety of solid propellants, while the detailed reaction pathway during thermolysis is far from clear due to the ultrafast and complex reactions involved. Herein, we present direct atomic simulations of AP thermal decomposition and propose a detailed decomposition network to fill the missing piece in the kinetic models by using a neural network model derived from ab initio calculations. The proton transfer is the dominant channel (NH 4 + ClO 4 → NH 3 + HClO 4 ), which is also observed in previous mass spectra experiments. In addition, gas products from decomposition play a critical role in promoting the decomposition of solid AP. For example, the H abstraction reaction by OH is found to be a critical pathway for AP decomposition. These simulations provide atomic insights into the complex reaction dynamics of AP and can be extended to investigate the reaction mechanism of novel energetic materials.

The thermal decomposition mechanism of RDX/AP composites: ab initio neural network MD simulations

“…The evidence hereby further supports the great numerical stability of the NNP model in MD simulation and highlights the issues in the design of the ReaxFF model. In previous NNP-MD simulations, 17,22 a timestep of 0.2 fs is a default choice to maintain good numerical stability. The findings in Fig.…”

“…Among previous studies, proton transfer is regarded as the most favorable route of AP decomposition. 8,22 Regarding RDX/AP composites, Liu et al 23 studied the thermal analysis characteristics and gas-phase decomposition products via the DSC-TG-FTIR method. The results indicated a significant chemical coupling effect between AP and RDX, exerting a mutually enhancing effect on their thermal decomposition processes.…”

“…33 In recent years, a series of machine learning-based neural network potentials (NNPs) 34 have been applied in MD simulation with the development of machine learning algorithms. 35 Chu et al 17,22 successfully developed NNP models for RDX and AP to investigate the thermal decomposition mechanisms of RDX and AP. The results indicate that the NNP model can accurately and rapidly describe the thermal decomposition process of RDX and AP, achieving accuracy comparable to DFT methods while enhancing computational efficiency.…”

Research progress on the catalytic and thermal decomposition of ammonium dinitramide (ADN)

“…5,6 However, large amounts of hydrogen chloride (HCl) gas could be produced during a space shuttle launch due to the high halogen content, which can react with moisture in the air to form hydrochloric acid that can cause irreversible environmental damage. 7–9 Instead, ammonium nitrate (AN) is considered an alternative to AP due to its low cost, easy preparation, and halogen-free features, yet the relatively low specific impulse and burning rate greatly limit its application. 10,11 As the most widely used liquid monopropellant, hydrazine has been applied in various strategic missiles, rockets, and satellite launching systems.…”

Determining the mechanical and decomposition properties of high energetic materials (α-RDX, β-HMX, and ε-CL-20) using a neural network potential

“…However, recent research highlights certain limitations of the ReaxFF model, such as its inability to accurately capture complex chemical reactions during HEM decomposition, relatively poor energy conservation, and relatively low computational efficiency compared to neural network potentials (especially on large systems, i.e., 410 nm). [37][38][39][40][41][42][43][44][45][46] In recent works, machine learning-based neural network potential (NNP) provides new opportunities in predicting the physicochemical and mechanical properties of energetic materials with high accuracy and computational efficiency. 39,41,44,[47][48][49][50][51][52][53][54] This approach aims to establish a functional relationship between inputs (chemical descriptors) and outputs (properties), without the need to reconstruct the relationship from physics.…”