Monte Carlo Simulations of Au₃₈(SCH₃)₂₄ Nanocluster Using Distance-Based Machine Learning Methods

Area of Science:

• Computational chemistry and materials science
• Application of machine learning in atomistic simulations
• Nanoparticle research

Background:

• Thiolate (SR) protected gold nanoclusters are crucial in nanotechnology.
• Accurate simulation of their thermal dynamics requires reliable interaction potentials.
• Existing methods may struggle with the complexity of these nanostructures.

Purpose of the Study:

• To implement distance-based machine learning (ML) methods for creating accurate atomistic interaction potentials.
• To enable efficient Monte Carlo simulations of thermal dynamics for gold nanoclusters.
• To facilitate the study of thermal-dependent electronic and optical properties.

Main Methods:

• Training an ML potential using density functional theory (DFT) based molecular dynamics (MD) data.
• Utilizing data from two experimentally characterized structural isomers of Au38(SR)24.
• Validating the ML potential against independent DFT MD simulations.

Main Results:

• A realistic atomistic interaction potential was successfully implemented using ML.
• The ML potential accurately models the thermal dynamics of Au38(SR)24.
• The method demonstrates generalization and accuracy control for complex nanostructures.

Conclusions:

• This ML approach provides an efficient pathway for probing the configuration space of gold nanoclusters.
• It enables further investigations into thermal-dependent properties.
• The strategy is applicable to complex nanostructures with multiple elements and varying interaction strengths.