Using graph neural network and symbolic regression to model disordered systems.

Area of Science:

• Computational physics
• Materials science
• Chemical physics

Background:

• Accurate simulation of atomic trajectories is crucial for modeling disordered systems using molecular dynamics (MD).
• The precision of MD simulations hinges on the interatomic potential function, which governs atomic movement calculations.
• Traditional methods for deriving interatomic potentials are knowledge-intensive and computationally expensive.

Purpose of the Study:

• To introduce a novel approach integrating machine learning with molecular dynamics for disordered systems.
• To provide precise interatomic potential energy calculations, enhancing simulation accuracy.
• To address the limitations of traditional interatomic potential derivation methods.

Main Methods:

• Integration of machine learning algorithms with molecular dynamics simulations.
• Development of a novel computational framework for potential energy calculations.
• Application to disordered systems requiring accurate atomic trajectory simulation.

Main Results:

• Achieved precise interatomic potential energy calculations for disordered systems.
• Demonstrated the efficacy of the machine learning-integrated approach.
• Overcame the need for extensive prior physical knowledge and reduced computational cost.

Conclusions:

• The novel machine learning-assisted method offers a more efficient and accurate way to model disordered systems.
• This approach advances the capabilities of molecular dynamics simulations.
• It paves the way for more sophisticated modeling of complex materials and phenomena.