Simulating the NaK Eutectic Alloy with Monte Carlo and Machine Learning.

Area of Science:

• Materials Science
• Computational Chemistry
• Statistical Mechanics

Background:

• Materials discovery can be accelerated by integrating atomistic simulations with machine learning.
• Predicting phase transitions and material behaviors is crucial for developing new alloys.

Purpose of the Study:

• To apply a combined atomistic simulation and machine learning approach for predicting the melting transition and amorphous-solid behavior of NaK alloy at eutectic concentration.
• To demonstrate the efficacy of machine learning methods trained on topological local structural properties for efficient property prediction.

Main Methods:

• Utilized Monte Carlo annealing to generate configurations of the NaK eutectic alloy.
• Analyzed atomic configurations using topological attributes derived from Voronoi tessellation.
• Employed expectation-maximization clustering and Random Forest classification to categorize atomic configurations.

Main Results:

• Voronoi topological fingerprints accurately and rapidly predicted the alloy's thermal behavior.
• Atomic configurations were successfully catalogued into three distinct phases: liquid, amorphous solid, and crystalline solid.
• Melting transition identified at 230 K, with a sharp distinction between crystalline solid and liquid phases.
• An arrest-motion temperature determined at 130-140 K for amorphous solid configurations.

Conclusions:

• The statistical learning paradigm effectively predicts thermal behavior and phase transitions in NaK alloy.
• Topological attributes offer significant utility beyond this specific application, extending to other materials and thermodynamic studies.
• This methodology harnesses the discovery of novel material properties through advanced computational techniques.