High-pressure and temperature neural network reactive force field for energetic materials.

Area of Science:

• Computational materials science
• Chemical physics
• Materials chemistry

Background:

• Reactive force fields (ReaxFF) are vital for molecular dynamics simulations, offering computational efficiency for large-scale systems.
• Traditional force fields face limitations in accuracy due to fixed functional forms, hindering refinement.
• Accurate modeling of energetic materials under extreme conditions is crucial for safety and performance.

Purpose of the Study:

• To develop a neural network-based reactive interatomic potential for energetic materials.
• To predict mechanical, thermal, and chemical responses under ambient and shock-loading conditions.
• To overcome the accuracy limitations of traditional force fields.

Main Methods:

• Developed a neural network-based reactive interatomic potential.
• Utilized an automatic iterative approach to expand the training dataset.
• Included various CHNO materials and their reactions under diverse conditions.

Main Results:

• The new potential demonstrates improved accuracy over state-of-the-art force fields.
• Accurate prediction of detonation performance and decomposition product formation.
• Accurate prediction of vibrational spectra under ambient and shock-loading conditions.

Conclusions:

• Neural network potentials offer a significant advancement for reactive force fields.
• This approach enables more accurate simulations of energetic materials at extreme conditions.
• The developed potential provides a robust tool for studying material behavior under high-stress environments.