Graph dynamical networks for unsupervised learning of atomic scale dynamics in materials.

Area of Science:

• Materials Science
• Computational Chemistry
• Artificial Intelligence

Background:

• Understanding atomic and molecular dynamics in condensed phases is crucial for designing advanced functional materials.
• Complex local environments in materials like electrolytes and membranes hinder the analysis of these dynamics.
• Molecular dynamics (MD) simulations generate vast datasets essential for materials design.

Purpose of the Study:

• To develop an unsupervised learning approach for analyzing atomic-scale dynamics in diverse material phases and environments.
• To extract complex dynamical information that is challenging to obtain through traditional methods.
• To provide an automated and broadly applicable tool for materials scientists.

Main Methods:

• Development of graph dynamical networks (GDNs), an unsupervised machine learning approach.
• Application of GDNs to analyze data from molecular dynamics simulations.
• Testing the approach on various multi-component amorphous material systems.

Main Results:

• GDNs successfully learned significant dynamical information from MD simulations.
• The approach proved effective for complex, multi-component amorphous materials.
• Key dynamical insights were extracted that are difficult to obtain otherwise.

Conclusions:

• Graph dynamical networks offer a powerful, automated method for understanding atomic-scale dynamics in materials.
• This AI-driven approach can accelerate the design of next-generation functional materials for energy and environmental applications.
• The method is broadly applicable across various material systems and simulation data.