Abstract
The structural, thermodynamic and transport properties of the CaF2-MgF2 molten salt system were investigated with ab initio molecular dynamics (AIMD), system-specific neural network interatomic potentials (NNIPs) and universal PreFerred Potentials (PFP). We trained an NNIP model using AIMD data as input and used this potential to efficiently simulate the interactions within a large supercell in a temperature range of 1273–1773 K. The Large-scale Atomic/Molecular Massively Parallel Simulator (LAMMPS) code was employed to validate our trained NNIP model. The Matlantis software with universal PFP is also presented to prove its feasibility for MD calculations and can be considered as a useful alternative simulation tool for higher-order systems where existing potentials are not readily available. We calculated structural and thermodynamic properties including radial distribution function (RDF), angular distribution function (ADF), specific heat capacity, ionic self-diffusivity, and viscosity. Our results indicate that the system exhibited a high degree of structural disorder, with the Ca, Mg, and F ions forming a liquid solution. Using PFP, the positions of the first peak in RDFs for Ca-F and Mg-F pairs are only slightly left-shifted (<0.05 Å), and the estimated viscosity of the melt decreases from 4.613 mPa·s to 1.846 mPa·s with an increase in temperature from 1273 K to 1773 K, in agreement with the NNIP trained specifically for CaF2-MgF2. Our results provide valuable insights into the properties of the CaF2-MgF2 system at high temperatures and serve as predictive models for the development of new electrolytes that could be used for silicon epitaxy by adding silica.
Original language | English |
---|---|
Article number | 113294 |
Journal | Computational Materials Science |
Volume | 245 |
DOIs | |
State | Published - Oct 2024 |
Funding
This research at Worcester Polytechnic Institute, Boston University, and University of Massachusetts-Lowell was supported by the US Department of Energy (DOE) NEUP program Award DE-NE0008937 and DE-NE0009204, Solar Energy Technologies Offices Award DE-EE0008988, and the US National Science Foundation (NSF) Award No. CMMI-1937818 and CMMI 1937829. In addition, we would like to acknowledge the computing resources from the Advanced Cyberinfrastructure Coordination Ecosystem: Services & Support (ACCESS) program Award: TG-DMR190004 and the technical support from the Academic & Research Computing team at Worcester Polytechnic Institute, especially James Leonard Kingsley and Siamak Mohammed Z. Najafi.
Keywords
- Angular distribution function
- CaF-MgF molten salt
- Matlantis
- Neural networks interatomic potentials
- Radial distribution function
- ab initio molecular dynamics