Molecular Dynamics Study on Deep Learning Potential of the (LiF-YF₃)_eut.-Y₂O₃ Molten Salt System

Area of Science:

• Materials Science
• Computational Chemistry
• Chemical Engineering

Background:

• Molten salts are crucial in various industrial applications.
• Understanding the structural evolution of molten salt systems is essential for optimizing their performance.
• The (LiF-YF3)eut.-Y2O3 system presents unique challenges due to complex ion interactions.

Purpose of the Study:

• To develop a reliable molecular dynamics potential function for the (LiF-YF3)eut.-Y2O3 molten salt system using machine learning.
• To investigate the dynamic evolution of cluster structures and ion coordination in the molten salt system.
• To analyze the structural changes upon the dissolution of Y2O3 in the molten salt.

Main Methods:

• A machine learning workflow, Deep Potential Generator (DPGEN), was employed to train a molecular dynamics potential function.
• Radial distribution functions (RDFs) and angular distribution functions (ADFs) were calculated to analyze cluster structures.
• System density was computed and compared with experimental values for validation.

Main Results:

• The trained potential function achieved a high accuracy, with a maximum deviation of 3.0% in system density compared to experimental values.
• In the (LiF-YF3)eut. system, dominant clusters were [YF6]3- and [YF7]4-, with coordination numbers of 4-5 for Li-F and 6-7 for Y-F pairs.
• Upon Y2O3 dissolution, O2- ions replaced F- ions, forming new [YOFx]n- clusters, primarily [YOF4]3- and [YOF5]4-, which constituted 75% of the total clusters.

Conclusions:

• The machine learning approach provides an accurate method for simulating complex molten salt systems.
• The study identified key short-range ordered clusters in the (LiF-YF3)eut. and Y2O3-doped systems.
• The findings offer insights into the structural behavior of molten salts, crucial for their application in areas like nuclear energy and materials processing.