04/24/2024
By Danielle Fretwell
The Francis College of Engineering, Department of Energy Engineering - Nuclear, invites you to attend a Doctoral Dissertation Proposal defense by Shubhojit Banerjee on "Understanding structure and transport properties of molten salts from a time average and time-resolved perspective."
Candidate Name: Shubhojit Banerjee
Degree: Doctoral
Defense Date: April 29, 2024
Time: 1 to 3 p.m.
Location: Perry 215
Committee:
- Advisor: Stephen T Lam, Assistant Professor, Chemical Engineering, UML
- Fanglin Che, Assistant Professor, Chemical Engineering, UML
- Jerome Delhommelle, Associate Professor, Chemistry, UML
Brief Abstract:
Various advanced nuclear reactors have been proposed using molten salt coolants, including fluoride high-temperature salt-cooled (FHRs), molten salt reactors (MSRs), and fusion devices. Optimizing the performance of the molten salts for a particular application requires a substantial amount of knowledge of salts' thermophysical properties, which further depend on salt structures. However, due to the high cost, exploring the wide composition space of salt chemistry with standalone experiments is limited. Herein, predictive molecular dynamics plays a crucial role as it is known to predict the structure and transport properties of the molten salt accurately.
Firstly, the performance of the dispersion models, such as D2, D3, D3_BJ, and VDW-DF, for predicting the thermophysical and structural properties of G-I and G-II binary molten fluoride salts are tested. The performance of the dispersion methods was tested in terms of density error, and it is found that simulations with only PBE XC functional exhibited a higher error in density, whereas adding dispersion improves the prediction quality. Local structure was found to be well predicted across the tested dispersion models. Where pair distribution functions showed similar first peak distances (± 0.1 Å) and first shell coordination numbers, indicating accurate simulation of chemical structures and atomic distances. Diffusivity was also well predicted across different dispersion models ( ± 20%). As such, this rigorous benchmarking provides us with the best possible dispersion model for simulating individual binary fluoride salts. Validated DFT and AIMD will then be used to build a robust neural network interatomic potential to simulate the LiF-BeF2 system. Commonly used potential, such as the polarizable ion model, fails when the BeF2 molar concentration is higher. On the other hand, the inherited limitations of AIMD simulations limit the accurate prediction of transport properties such as diffusivity and viscosity. Therefore, we have proposed the development of a robust NNIP for simulating LiF-BeF2 system with the help of an active learning cycle. The proposed active learning method will shorten the potential development time by reducing the required number of DFT calculations. The developed NNIP will then be used to explore a wide range of composition spaces of LiF-BeF2 mixtures.
Lastly, after exploring time/ensemble-averaged properties from molecular dynamics simulations, I will focus on time-dependent ion dynamics. The ion dynamics that govern different properties of molten salts are poorly understood due to challenges in precisely quantifying the spatial and temporal fluctuations of specific ions in highly disordered systems. While the Van Hove correlation function (VHF) obtained from inelastic neutron scattering (INS) probes these dynamics directly, its interpretation is limited by the inherent species-averaging of experiments, which obscures analysis of key ion transport and solvation mechanisms. Here, we have used ab initio molecular dynamics to model the Van Hove correlation function (real space and time correlation function) to explore single as well as collective ion dynamics. This study improves our understanding of picosecond-scale ion dynamics, mechanisms, and emergent physical properties of molten salts.