Application of First-Principles-Based Artificial Neural Network Potentials to Multiscale-Shock Dynamics Simulations...

Area of Science:

• Computational materials science
• Chemical physics
• Solid mechanics

Background:

• Artificial neural network molecular dynamics (ANN-MD) offers first-principles accuracy for large-scale simulations.
• Current ANN-MD applications are primarily limited to near-equilibrium processes.
• Simulating far-from-equilibrium phenomena like shock waves requires computationally intensive methods.

Purpose of the Study:

• To extend the applicability of ANN-MD to far-from-equilibrium shock phenomena.
• To integrate ANN-MD with multiscale shock theory (MSST) for enhanced simulation capabilities.
• To achieve first-principles accuracy in modeling shock-wave propagation with reduced computational cost.

Main Methods:

• Development of a combined artificial neural network molecular dynamics and multiscale shock theory (ANN-MSST-MD) approach.
• Training ANN potentials using first-principles calculations for accurate interatomic interactions.
• Application of the ANN-MSST-MD method to simulate shock-wave propagation in solids.

Main Results:

• The ANN-MSST-MD approach successfully describes far-from-equilibrium shock phenomena with first-principles accuracy.
• Simulations using ANN-MSST-MD were approximately 5000 times faster than traditional first-principles MD.
• The method resolved fine, long-time elastic deformation at low shock speeds, a feat not possible with conventional first-principles MD due to computational limitations.

Conclusions:

• The novel ANN-MSST-MD framework significantly advances the simulation of dynamic processes in materials.
• This work overcomes the computational barriers associated with studying far-from-equilibrium phenomena.
• The developed simulation methodology provides a foundation for exploring a broad spectrum of non-equilibrium material behaviors.