Enabling accurate chemical modeling of shocked energetic materials using a machine learning interatomic potential

Area of Science:

• Computational chemistry
• Materials science
• Chemical dynamics

Background:

• Dynamic compression of organic materials involves complex, multi-timescale reactions.
• Accurate modeling of energetic materials like 1,3,5-triamino-2,4,6-trinitrobenzene (TATB) under detonation is crucial for various applications.
• Existing methods face challenges in capturing the intricate chemistry under extreme conditions.

Purpose of the Study:

• To develop an efficient machine learning potential for simulating TATB under detonation.
• To establish a robust framework for modeling shocked organic energetic materials.
• To gain detailed insights into the chemical transformations of TATB during shock compression.

Main Methods:

• Utilized Chebyshev polynomials to construct a machine learning potential.
• Developed a strategy for generating diverse training data to capture complex TATB chemistry.
• Performed large-scale, multi-nanosecond simulations of shocked TATB.

Main Results:

• The machine learning potential demonstrated strong transferability across various thermodynamic conditions and other explosives.
• Simulations accurately reproduced experimental Hugoniot equation of state data for TATB.
• Observed the rapid formation of nitrogen-rich carbon clusters following shock compression.

Conclusions:

• The developed machine learning approach enables accurate and reliable chemical modeling of organic materials under extreme conditions.
• The study provides a robust framework for future investigations into shocked energetic materials.
• The findings offer detailed insights into the detonation chemistry of TATB and related compounds.