Transferable Deep Learning Potential Reveals Intermediate-Range Ordering Effects in LiF–NaF–ZrF₄ Molten Salt

Abstract

LiF−NaF−ZrF 4 multicomponent molten salts are promising candidate coolants for advanced clean energy systems owing to their desirable thermophysical and transport properties. However, the complex structures enabling these properties, and their dependence on composition, is scarcely quantified due to limitations in simulating and interpreting experimental spectra of highly disordered, intermediate-ranged structures. Specifically, sizelimited ab initio simulations and accuracy-limited classical models used in the past are unable to capture a wide range of fluctuating motifs found in the extended heterogeneous structures of liquid salt. This greatly inhibits our ability to design tailored compositions and materials. Here, accurate, efficient, and transferable machine learning potentials are used to predict structures far beyond the first coordination shell in LiF−NaF−ZrF 4 . Neural networks trained at only eutectic compositions with 29% and 37% ZrF 4 are shown to accurately simulate a wide range of compositions (11−40% ZrF 4 ) with dramatically different coordination chemistries, while showing a remarkable agreement with theoretical and experimental Raman spectra. The theoretical Raman calculations further uncovered the previously unseen shift and flattening of bending band at ∼250 cm −1 which validated the simulated extended-range structures as observed in compositions with higher than 29% ZrF 4 content. In such cases, machine learning-based simulations capable of accessing larger time and length scales (beyond 17 Å) were critical for accurately predicting both structure and ionic diffusivities.

First-Principles Investigation of the Effects of UF4 and ThF4 Fuels on the Structural, Dynamic, and Thermodynamic Properties of LiF–NaF

“…The prepeak manifested consistently across all simulations featuring UF 4 and ThF 4 additives, but it remained absent in all simulations of the pure system. Medium-range ordering has also been reported in the literature for similar molten salt systems: Li et al 12 identified cluster formation in NaCl with the addition of UCl 3 and UCl 4 , while Chahal et al 29 observed the formation of networks in LiF−NaF upon the introduction of ZrF 4 .…”

“…A medium-range order has been investigated for decades in metallic melts, ionic liquids, and protein solutions. − These structures are linked to the emergence of a prepeak in the structure factor, − indicating spatial correlations beyond the length scale represented by the main structure factor peak. These extended correlations seem to have an impact on the thermophysical and transport properties of the liquid, though the exact relationship is not entirely clear. − In polymers, for instance, it is well-established that the structure and weight of the chain influence viscosity, , with viscosity increasing as molecular weight and chain length increase. In contrast, for molten salts and liquids in general, there is not a prevailing theory relating viscosity to their structure.…”

Tracing mechanistic pathways and reaction kinetics toward equilibrium in reactive molten salts

“…Investigating properties such as solubility, viscosity, diffusivity, and conductivity, and connecting them directly to ion exchange and speciation, is out of the scope of this AIMD study. However, in the future, through the development of accurate machine learning 30,57,58 interatomic interaction potentials trained on ab initio forces and energies, accessing both large length scales and long simulation times will be possible. This will help establish the property-speciation-ion exchange relations, providing a means to enhance the performance of the high-temperature energy storage and production technologies.…”

Compositional transferability of deep potential in molten LiF–BeF2 and LaF3 mixtures: prediction of density, viscosity, and local structure

“…36 At present, the DPMD simulations have been successfully performed on molten chlorides 37–39 and fluorides. 40,41…”

“…36 At present, the DPMD simulations have been successfully performed on molten chlorides [37][38][39] and fluorides. 40,41 In this work, LaF 3 as a representative lanthanide fluoride is added into molten Flibe eutectic salt to investigate the macroscopic properties and microscopic structure of molten mixtures. Firstly, diverse FPMD simulations of molten Flibe + LaF 3 are conducted at 823 K with small simulation cells under the complete consideration of the position of La.…”