Machine learning quantitatively characterizes the deformation and destruction of explosive molecules

Area of Science:

• Materials Science
• Computational Chemistry
• Chemical Engineering

Background:

• Explosives are vital in various industries, but their reaction mechanisms remain incompletely understood.
• This knowledge gap limits optimal energy utilization and safety protocols for explosives.
• Understanding molecular-level changes is crucial for advancing explosive science.

Purpose of the Study:

• To develop quantitative and qualitative models for explosive molecular structure deformation and destruction.
• To investigate the shock-loading response of ε-CL-20 at a molecular level.
• To provide new insights into the rapid chemical changes occurring during explosive reactions.

Main Methods:

• Employed molecular dynamics simulations to analyze shock-loaded ε-CL-20.
• Utilized machine learning algorithms, including Delaunay triangulation, clustering, and gradient descent.
• Developed quantitative models linking molecular volume, position, and distance changes.

Main Results:

• Established quantitative relationships between molecular structural changes under shock.
• Observed inward shrinking of peripheral structures, initially stabilizing the cage structure.
• Identified expansion and destruction of the cage structure after significant compression, alongside hydrogen atom transfer.

Conclusions:

• The study elucidates the intricate structural transformations and chemical reactions in explosive molecules under shock.
• The developed machine learning-based quantitative characterization method offers a novel approach for analyzing microscopic reaction mechanisms.
• Findings contribute to a deeper understanding of detonation processes and can inform the design of safer, more effective energetic materials.