Skip to main content
NCBI home page
ACS Omega logoLink to ACS Omega
. 2025 Sep 5;10(36):41820–41827. doi: 10.1021/acsomega.5c05781

Ab Initio Molecular Dynamics Simulation Study on the Molten Structure of Cryolite to the Cryolitic Bath, Na3AlF6–AlF3–Al2O3–CaF2

Konstantinos Betsis †,*, Konstantinos Karalis ‡, Anthimos Xenidis †
PMCID: PMC12444678  PMID: 40978365

Abstract

This study investigates the structural properties of cryolitic melts using ab initio molecular dynamics (AIMD) simulations. Aluminum production relies on the electrolysis of alumina in a molten cryolite bath, which primarily consists of cryolite, aluminum fluoride, alumina, and calcium fluoride and operates at temperatures between 1213 and 1243 Κ. Despite significant advancements, the local structure and speciation within these melts remain incompletely understood. This research employs computational techniques to examine the atomic structure and charge distribution in cryolitic melts, with a particular focus on aluminum atom interactions and the role of bridging anions. AIMD simulations were performed by using the CP2K software package. The Perdew–Burke–Ernzerhof (PBE) approximation was applied for the exchange-correlation functional, and Goedecker–Teter–Hutter (GTH) pseudopotentials were used to model core electrons. The study investigated systems with varying amounts of AlF3, Al2O3, and CaF2 in molten cryolite, maintaining temperatures slightly above the liquidus point. Structural analysis was conducted using radial distribution functions (RDFs) to determine bond distances and coordination numbers, while electronic distribution was analyzed through Mulliken population analysis. Key findings include the dominance of the AlF5 2– complex in molten cryolite, which is in agreement with previous studies. The addition of alumina influences the formation of oxyfluoroaluminate species, with Al2OF6 2– and Al2O2F4 2– being prevalent at low and high alumina concentrations, respectively. Calcium fluoride impacts the melt’s structure by increasing the presence of AlF5 2– and altering molecular conformation due to the strong anionic nature of calcium. The electronic structure analysis revealed minor changes in the average charge of atoms but an overall increase in the anionic character of the melt with the addition of O2– and Ca2+. This study provides valuable insights into the atomic and electronic behavior of cryolitic melts, contributing to a deeper understanding of these complex molten systems and supporting the optimization of aluminum production processes.

graphic file with name ao5c05781_0009.jpg

graphic file with name ao5c05781_0007.jpg

1. Introduction

Aluminum production is based on the reduction of alumina to molten aluminum via electrolysis. The electrolytic bath, consisting primarily of cryolite, aluminum fluoride, alumina, and calcium fluoride, operates in a temperature range of 1213–1243 Κ, with the specific operational temperature influenced by the chemical composition of the melt. Physicochemical properties such as density, viscosity, and electrical conductivity are influenced by structural speciation within the melt, which is investigated using both experimental techniques (Raman spectroscopy and nuclear magnetic resonance (NMR)) and theoretical computational methods. Despite substantial progress, our understanding of the local structure and the role of the main components within the cryolitic bath remains incomplete.

Numerous investigations have aimed to elucidate speciation in the Hall–Héroult electrolyte. Using Raman spectroscopic measurements, Gilbert et al. − proposed the existence of three main aluminum complexes in the cryolite melt: AlF4 –, AlF5 2–, and AlF6 3–. Their findings indicate that AlF5 2– is the dominant complex in molten Na3AlF6. These results are consistent with NMR experiments conducted by Cikit et al., − which also highlight the predominance of specific aluminum-containing molecules in the melt. Dewing suggested the coexistence of AlF4 –, AlF5 2–, and AlF6 3–, as proposed by a thermodynamic data model. Ab initio molecular dynamics (AIMD) simulations have been used by several researchers − to gain deeper insight into the local structure of molten cryolite and to quantify aluminum fluoride complexes in the melt. These simulations show that Na+ ions remain uncoordinated and free in the melt. The results reaffirm that AlF5 2– is the dominant complex in the melt, with either AlF4 – or AlF6 3– being the second most abundant complex in the melt.

The addition of alumina to molten cryolite influences the melt structure, depending on its concentration. Oxygen serves as a bridging atom between aluminum atoms for the formation of oxyfluoroaluminate species, with the presence of Al–O–Al bonds in the bath with low alumina content in the melt, less than 3.5%, or the presence of Al<OO>Al at high alumina concentrations. These findings have been proposed by Lacassagne et al., using high-temperature NMR techniques, and by Gilbert et al., , based on Raman spectroscopy, with the presence of Al2OF6 2– in low alumina content. Concurrently, Al2O2F4 2– is the predominant species at high alumina percentages. In the NaF–AlF3–Al2O3 system, DFT calculations by Picard et al. suggest the presence of AlOF2 –, Al2OF6 2–, and Al2O2F4 2–, while recent work by Machado et al. proposes the existence of Al2OF6 2–, Al2OF7 3–, Al2OF8 4–, Al3OF11 4–, Al3OF12 5–, Al3O2F8 3–, Al3O2F6 4–, and Al4O2F13 5–, with the presence of these complexes increasing accordingly with the content of alumina in the melt. Lin et al., using in situ Raman spectroscopy, suggest that the oxygen-containing species in the melt, for various molar ratios and alumina content, are Al2OF4, Al3O2F8 3–, Al2OF8 4–, and Al2O2F4 4–, while Zhang et al. indicate the presence of Al2OF6 2– with alumina content less than 3.5 wt % and Al2O2F4 2– for a higher percentage. Furthermore, Chen et al. investigate the behavior of molten cryolite with calcium fluoride using AIMD simulations. Their results show that the Ca2+ ions compete with Al3+ for F– and as a result, Ca2+ attracts [AlF x ](x−3)– complexes, with preferential cross-linking with AlF5 2–.

In this study, we examine the structural properties of cryolitic melts using AIMD simulations. Computational techniques are employed to investigate the atomic structure and charge distribution within these melts. By focusing on the interactions between aluminum atoms and their surrounding environment, as well as the role of bridging anions, we aim to gain a deeper understanding of the behavior of these complex molten systems at temperatures slightly above the liquidus point.

2. Methods

2.1. Ab Initio Molecular Dynamics Simulations

Ab initio molecular dynamics (AIMD) simulations were performed using the CP2K 2.6.2 software package employing the Quickstep method, which is based on the Gaussian and Plane Wave (GPW) method. This approach combines the efficiency of Gaussian-type orbintals for wave function representation with a plane-wave expansion of the electronic density. To describe electron exchange-correlation effects, Density Functional Theory (DFT) was used with the Perdew–Burke–Ernzerhof (PBE) functional within the Generalized Gradient Approximation (GGA). To account for long-range van der Waals interactions, Grimme’s DFT-D3 dispersion correction was applied.

Core electrons were treated using Goedecker–Teter–Hutter (GTH) norm-conserving pseudopotentials, and valence electrons were represented with the MOLOPT-SR Double-Zeta Valence Polarization (DZVP) basis sets. This combination of PBE, GTH pseudopotentials, and MOLOPT-SR basis sets ensures a balance between computational efficiency and accuracy. A plane-wave cutoff of 400 Ry was applied for the electronic charge density, and a relative cutoff of 60 Ry was used for the energy and force evaluations. A time step of 0.5 fs was used for the AIMD simulations. Each system was equilibrated for 15 ps, followed by a production run of 50 ps for sampling.

2.2. Supercells of Investigated Systems

The canonical ensemble (NVT) was employed for all simulations at temperatures slightly above the liquidus point of the respective cryolite melts, with liquidus temperatures calculated using the FactSage 7.0 software. The Nosé–Hoover thermostat was used to control temperatures to match the structural dynamics of each system, allowing for the observation of the effects of adding AlF3, Al2O3, and CaF2 to the molten cryolite, as shown in Tables and . Following industrial practice, the cryolitic bath consisted of cryolite with excess aluminum fluoride, alumina, and calcium fluoride. In this study, a representative cryolitic bath composition was selected: 10 wt % AlF3, 3 wt % Al2O3, and 4.5 wt % CaF2. For each system, one component was incrementally added to molten cryolite to reach the target chemical composition. The addition of new components altered the liquidus temperature and the kinetic energy of the melt. To ensure that the systems remained in the liquid state, all simulations were conducted at temperatures slightly above their respective liquidus points.

1. Investigated Systems with Calculated Values of Density and % wt Percentage.

System Name DensityFactsage (g/cm3) DensityKvade H (g/cm3) Chemical composition
1 Cryolite 2.095 2.099 100 wt % Na3AlF6
2 AlF3 2.054 2.048 90.09 wt % Na3AlF6
9.91 wt % AlF3
3 Alumina 2.034 2.036 87.03 wt % Na3AlF6
9.95 wt % AlF3
3.02 wt % Al2O3
4 Bath 2.078 2.070 82.5 wt % Na3AlF6
9.90 wt % AlF3
3.00 wt % Al2O3
4.6 wt % CaF2
Open in a new tab

2. Compositions and Simulation Conditions of the Investigated Molten Cryolite Systems.

Name Al atoms F atoms Na atoms O atoms Ca atoms Volume (Å3) Temperature (K)
Cryolite 40 240 120 0 0 6670.15 1285
AlF3 46 267 129 0 0 7489.27 1270
Alumina 60 288 126 9 0 8266.91 1255
Bath 49 261 120 9 6 7384.12 1240
Open in a new tab

The initial atomic configurations were generated by randomly placing atoms within the supercell using the Packmol software. The volume of each system was determined based on estimated densities derived from thermochemical data, calculated using FactSage 7.0 software. These values were found to be consistent with empirical density estmates obtained from Kvande’s equation. For structural and electronic distribution analyses, radial distribution functions (RDFs) were computed, and visualizations were performed using VMD and Ovito software packages.

2.3. Structure Analysis

Structural analysis was conducted using RDFs, which describe the probability of finding a given atom at a distance of r from a reference atom. The first peak of the RDF provides information about the bond distance between the two atoms. For cryolite melts containing more than 3% w/w alumina, Al x OF y species are more prevalent than Al x O2F y . When alumina is added to a cryolitic melt with a cryolite ratio (CR) below 3, the focus is on the Al–F bond distance across the full CR range. However, in systems with higher alumina content, such as samples 3 and 4, the key bond distances of interest are O–Al and O–F. Bond distances involving Al or alkaline elements in the cryolitic bath give insight into the melt’s structure. Elements such as sodium and calcium are distributed in the volume of the melt and act as bridges for electrical conductivity and electron movement.

The peak of the normalized RDF function g(r) indicates the most probable interatomic distance and reflects both chemical bonding and coordination environments. An atom is considered coordinated to a central atom if it lies within a cutoff distance defined by the first minimum following the RDF’s first peak. The coordination analysis is, therefore, crucial to understanding the preferred molecular structures in the cryolite melt. All systems were analyzed after a 15 ps equilibration step and were based on the calculated Mean Squared Displacement (MSD) analysis, which estimated the average distance squared by the current position from the initial position. System equilibration was confirmed when the log­(MSD) versus log­(time) reached a slope of approximately 1, as shown in Figure .

1.

1

Open in a new tab

Mean Squared Displacements (MSD) of Al for each system examined (see Table ).

2.4. Electronic Distribution

The total electron density derived from the overlap of atomic orbitals allows for the calculation of partial atomic charges and the assessment of charge transfer phenomena.

The electronic structure and charge distribution in the investigated systems were analyzed by using Mulliken population analysis. This method highlights differences in electronic structure and charge distribution across the molten cryolite systems, including components such as Na3AlF6, AlF3, Al2O3, and CaF2. The insights gained from this analysis help clarify properties such as electrical conductivity and coordination numbers, with a particular focus on the evolution of clusters in the melt and the structural effects of adding or omitting specific substances in the cryolitic bath.

3. Results and Discussion

3.1. Radial Distribution Function

In this study, aluminum or oxygen atoms were selected as reference atoms for RDF analysis depending on the system investigated. For molten cryolite and the molten Na3AlF6–AlF3 mixture, the aluminum atom is the most critical factor for further investigation of these two systems, using the g(r) values to understand the basic structure and estimate the melt’s speciation, Figure . For cryolite, Table , there is a small peak, which indicates cross-linking between the Al-coordinated species and the intense peak of this bond at 1.75 Å, referring to the distance in the liquid state. The excess of AlF3 has a minor effect on the preferred bond distance, as it increases by 0.1 Å for all of the bonds except for F–F, which is located at 2.25 Å, as a result of the increasing number of fluoride atoms in the melt. Although there is minor alteration in the melt’s structure, the bond intensity is more significant than in molten cryolite. Thus, the added AlF3 affects strongly the covalent and ionic interactions, mainly the coordination between aluminum atoms and fluoride atoms, and due to the deficient value of the radial distribution factor, it indicates that fluoride ions must cross a significant barrier to exit the first solvation sphere of aluminum ions, making the Na3AlF6–AlF3 mixture more electronegative than pure cryolite. When alumina dissolves into the mixture, it significantly influences the distance between the Al–Al pair, reducing it to 3.13 Å from 5.75 and 5.86 Å in cryolite and AlF3, respectively. The minimization of the Al–Al distance takes place as a result of alumina molecules in the first part and since the creation of oxyfluoroaluminate anions, with low-concentration alumina, less than 3.5%, where oxygen is located at the center of the molecule, creating two bonds with two atoms of aluminum, which are bonded with three atoms of fluoride for each aluminum atom. In the molten mixture of Na3AlF6–AlF3–Al2O3–CaF2, all the preferred distances of the atoms were not affected by the presence of calcium fluoride. However, the resulting molecular structures exhibited different orientations due to the difference in ionic radius between Ca2+ and Na+ ions, as well as their ability to interact not only with fluoride ions but also with oxygen and aluminum centers. These interactions promote the formation of larger molecular structures, which is in agreement with experimental results showing that the presence of CaF2 increased the viscosity of the melt.

2.

2

Open in a new tab

Radial Distribution Function (RDF) of cryolite (a), Na3AlF6–AlF3 (b), Na3AlF6–AlF3–Al2O3 (c), and cryolitic bath (d) (Na3AlF6–AlF3–Al2O3–CaF2).

3. First Peak Distance for Each Pair of Elements.

  First Peak distance (Å)
Pair Cryolite Na3AlF6–AlF3 Na3AlF6–AlF3–Al2O3 Na3AlF6–AlF3–Al2O3–CaF2
Al–Al 5.75 5.86 3.13 3.205
Al–F 1.75 1.78 1.76 1.78
Al–Na 3.38 3.49 3.62 3.48
F–F 2.65 2.60 2.60 2.61
F–Na 2.27 2.24 2.30 2.27
O–Al     1.73 1.74
O–F     2.81 2.81
O–Na     2.42 2.41
O–O     2.95 2.92
Open in a new tab

3.2. Coordination Analysis

All of the trajectories are analyzed for further structural analysis based on the molecules in the melt and the estimation of the coordination number of molecules added for each system, from cryolite to the mixture of the cryolitic bath.

There are free sodium atoms in molten cryolite, Figure , while fluoride atoms create complex molecules with aluminum. The AlF5 2– complex is the predominant species in the melt, followed by four-coordinated aluminum species, as shown in Table . An excess of aluminum fluoride is added to the molten cryolite, Figure , although AlF5 2– is retained as the molecule with the highest percentage in the melt, Table , and AlF6 3– has a greater presence in the melt. Simultaneously, the AlF3 and AlF4 – rate decreased.

3.

3

Open in a new tab

Illustrative representation of molten cryolite.

4. Percentages of AlF3 and Complex Anions in Molten Cryolite.

Molecule Percentage (%)
AlF3 13.99
AlF4 – 19.11
AlF5 2– 46.27
AlF6 3– 13.77
Open in a new tab

4.

4

Open in a new tab

Illustrative representation of molten Na3AlF6–AlF3.

5. Percentages of AlF3 and Complex Anions in Molten Na3AlF6–AlF3 .

Molecule Percentage (%)
AlF3 11.59
AlF4 – 17.32
AlF5 2– 46.56
AlF6 3– 19.50
Open in a new tab

Numerous investigations , have shown that if the percentage of alumina is more than 4%, the preferred complex of the system is molecules with two atoms of oxygen in the oxyfluoroaluminate molecules. On the other hand, one oxygen atom creates bonds with aluminum and fluoride with less than 4% of alumina content in the melt. This is in agreement with previous spectroscopic and computational studies, − which demonstrate that the structure of the melt evolves depending on alumina addition.

The added alumina in the cryolite mixture with an excess of AlF3, Figure , leads to the formation of both single and double oxygen aluminum complexes. From the melt’s structure analysis, species such as Al2OF6 2– and Al2OF7 – are identified, while most of the complex ions in the melt consist of AlF5 2–, Table . As the percentage of alumina in the melt is 3.5 wt %, there are a few molecules of Al2O2F5 3–, suggesting the onset of a structural transition from low alumina to high alumina concentration in the cryolitic melt. The presence of such bridging oxygen species has been associated in the literature with enhanced alumina solubility and improved ion transport, − which are critical for sustaining steady electrolysis in industrial Hall–Héroult cells.

5.

5

Open in a new tab

Illustrative representation of molten Na3AlF6–AlF3–Al2O3.

6. Percentages of AlF3 and Complex Anions in Molten Na3AlF6–AlF3–Al2O3 .

Molecule Percentage (%)
AlF3 20.37
AlF4 – 20.20
AlF5 2– 27.25
AlF6 3– 9.37
Al2OF6 2– 1.72
Al2OF7 3– 1.36
Al2OF5 – 1.16
Al2O2F5 3– 0.79
Open in a new tab

In the cryolitic path, a mixture consisting of Na3AlF6–AlF3–Al2O3–CaF2, with calcium fluoride present at 4.5% w/w, affects the conformation of the atoms in the melt, Figure . The strong anionic nature of calcium also affects the conformation of molecules in the melt, resulting in an increasing percentage of AlF5 2–. This molecule can surround the calcium atoms in the melt. The percentage of oxyfluoroaluminate ions is increased, with Al2OF6 2– retained as the complex with the greatest presence in the melt, Table .

6.

6

Open in a new tab

Illustrative representation of molten Na3AlF6–AlF3–Al2O3–CaF2.

7. Percentages of Molecules in the Molten Na3AlF6–AlF3–Al2O3–CaF2 .

Molecule Percentage (%)
AlF3 14.32
AlF4 – 14.08
AlF5 2– 26.56
AlF6 3– 14.32
Al2OF6 2– 2.01
Al2OF7 3– 1.59
Al2OF5 – 1.35
Al2O2F5 3– 0.93
Open in a new tab

3.3. Mulliken Analysis

Mulliken population analysis was employed to elucidate the electronic structure and charge distribution in (a) molten cryolite, (b) a mixture of Na3AlF6–AlF3, (c) Na3AlF6–AlF3–Al2O3, and (d) a cryolitic bath, providing insight into electronic interactions. The results reveal that the addition, respectively, of AlF3, Al2O3, and CaF2 to molten cryolite affects the charge distribution in the melt, as shown in Table .

8. Mulliken Analysis Population for the Investigated Systems.

  Average charge distribution (Std)
Atom Cryolite Na3AlF6–AlF3 Na3AlF6–AlF3–Al2O3 Na3AlF6–AlF3–Al2O3–CaF2
Al 1.17 (0.05) 1.18 (0.06) 1.15 (0.07) 1.14 (0.08)
F –0.60 (0.08) –0.64 (0.09) –0.56 (0.06) –0.58 (0.07)
Na 0.79 (0.02) 0.78 (0.02) 0.79 (0.02) 0.79 (0.02)
O     –0.69 (0.02) –0.69 (0.04)
Ca       1.21 (0.05)
Open in a new tab

Based on the results, the average charge of Al atoms slightly increases after the addition of AlF3. However, the presence of alumina and calcium fluoride affects the aluminum atoms’ reversible charge due to the added aluminum atoms from alumina and calcium atoms with more cationic conduction. Fluoride atoms tend to maximize their charge in the presence of an excess of AlF3. However, the added alumina minimizes the negative charge, and calcium fluoride affects the charge of fluoride. However, in the cryolite bath, the average charge is less than that in the pure cryolite melt. The average charge of sodium atoms is stabilized at 0.79, except in the mixture of the cryolite bath with aluminum fluoride, in which the average charge of sodium decreases to 0.78. The average charge of oxygen shows a minor increase due to the presence of calcium in the bath, which has the most considerable value among all the atoms in the investigated systems.

4. Conclusions

The structure of molten cryolite and its mixtures with alumina has been extensively investigated in the literature through both experimental techniques and molecular dynamics simulations. It is well established that the AlF5 2– complex is dominant in molten cryolite, followed by AlF4 – as the second most prevalent, as confirmed by multiple studies and thermodynamic modeling including data from FactSage 7.0. Furthermore, a decrease in the cryolite ratio (CR = NaF/AlF3) results in an increased proportion of AlF4 –, while AlF5 2– remains the primary complex. In this study, based on these established findings, further exploration of cryolite mixtures using AIMD simulations was investigated, with the presence of calcium fluoride and the Mulliken population analysis providing significant information about the structural properties of the cryolite bath.

The addition of alumina at approximately 4 wt % based on industrial practice promotes the formation of oxyfluoroaluminate complexes such as Al2OF6 2– and Al2O2F5 3–. While the presence of AlF5 2– is well documented in the literature, Al2O2F5 3– appears as a minor complex and becomes more prominent at an alumina content around 3.5 wt % in the melt. In the molten bath, Ca2+ ions, due to their higher charges, increase the concentration of AlF5 2– complexes by reducing the free fluoride anions. In parallel, the percentage of oxyfluoroaluminate complexes increases.

The electronic distribution in these mixtures can emerge using Mulliken population analysis. From a microscopic view of the investigation, it is clear that there are minor changes in the average charge of the atoms, but the total anionic character of the melt increases as the addition of O2– and Ca2+ boosts the total charge that has to be equilibrated. The average charge distribution of Al3+ ions is influenced by the addition of excess aluminum atoms from alumina. This increase in the anionic character of the melt correlates well with empirical models of electrical conductivity and aligns with descriptions of the atomic structure.

Acknowledgments

This research has been implemented within the framework of the National Recovery and Resilience Plan ‘Greece 2.0’, with funding from the European Union - NextGenerationEU through the European Recovery and Resilience Fund (project code: ΤΑΕΔΚ-06198).

The open access publishing of this article is financially supported by HEAL-Link.

The authors declare no competing financial interest.

References

  1. Haupin W.. The influence of additives on Hall-Héroult bath properties. JOM. 1991;43(11):28–34. doi: 10.1007/BF03222717. [DOI] [Google Scholar]
  2. Kline J., Tangstad M., Tranell G.. A Raman Spectroscopic Study of the Structural Modifications Associated with the Addition of Calcium Oxide and Boron Oxide to Silica. Metall. Mater. Trans. B. 2015;46(1):62–73. doi: 10.1007/s11663-014-0194-9. [DOI] [Google Scholar]
  3. Gilbert B., Materne T.. Reinvestigation of Molten Fluoroaluminate Raman Spectra: The Question of the Existence of AlF 5 2– Ions. Appl. Spectrosc. 1990;44(2):299–305. doi: 10.1366/0003702904085525. [DOI] [Google Scholar]
  4. Robert E., Lacassagne V., Bessada C., Massiot D., Gilbert B., Coutures J. P.. Study of NaF-AlF3 melts by high-temperature27Al NMR spectroscopy: Comparison with results from raman spectroscopy. Inorg. Chem. 1999;38(2):214–217. doi: 10.1021/ic980677b. [DOI] [Google Scholar]
  5. Gilbert B., Mamantov G., Begun G. M.. Raman spectrum of the AlF4- ion in molten fluorides. Inorg. Nucl. Chem. Lett. 1974;10(12):1123–1129. doi: 10.1016/0020-1650(74)80177-7. [DOI] [Google Scholar]
  6. Cikit S., Akdeniz Z., Madden P. A.. Structure and Raman spectra in cryolitic melts: Simulations with an ab initio interaction potential. J. Phys. Chem. B. 2014;118(4):1064–1070. doi: 10.1021/jp4080459. [DOI] [PubMed] [Google Scholar]
  7. Akdeniz Z., Madden P. A.. Raman spectra of ionic liquids: A simulation study of AlF3 and its mixtures with NaF. J. Phys. Chem. B. 2006;110(13):6683–6691. doi: 10.1021/jp060064h. [DOI] [PubMed] [Google Scholar]
  8. Akdeniz Z., Çiçek Z., Pastore G., Tosi M. P.. Ionic Clusters in Aluminium–Sodium Fluoride Melts. Mod. Phys. Lett. B. 1998;12(23):995–1002. doi: 10.1142/S0217984998001165. [DOI] [Google Scholar]
  9. Akdeniz Z., Çiçek Z., Tosi M. P.. Theoretical evidence for the stability of the (AlF5)­2– complex anion. Chem. Phys. Lett. 1999;308(5–6):479–485. doi: 10.1016/S0009-2614(99)00634-X. [DOI] [Google Scholar]
  10. Dewing E. W.. Thermodynamics of the system NaF-AIF3: Part VII. non-stoichiometric solid cryolite. Metall. Mater. Trans. B. 1997;28(6):1095–1097. doi: 10.1007/s11663-997-0064-9. [DOI] [Google Scholar]
  11. Lv X., Han Z., Chen J., Jiang L., Xu Z., Liu Q.. First-principles molecular dynamics study of ionic structure and transport properties of LiF-NaF-AlF3 molten salt. Chem. Phys. Lett. 2018;706:237–242. doi: 10.1016/j.cplett.2018.06.005. [DOI] [Google Scholar]
  12. Tan M., Li T., Shang B., Cui H.. Quantum chemical prediction of the spectroscopic properties and ionic composition of the molten NaF-AlF3 salts. J. Mol. Liq. 2020;317:113937. doi: 10.1016/j.molliq.2020.113937. [DOI] [Google Scholar]
  13. Lv X., Xu Z., Li J., Chen J., Liu Q.. Theoretical investigation on local structure and transport properties of NaFAlF3 molten salts under electric field environment. J. Mol. Struct. 2016;1117:105–112. doi: 10.1016/j.molstruc.2016.03.076. [DOI] [Google Scholar]
  14. Bučko T., Šimko F.. Effect of alkaline metal cations on the ionic structure of cryolite melts: Ab-initio NpT MD study. J. Chem. Phys. 2018;148(6):064501. doi: 10.1063/1.5017106. [DOI] [PubMed] [Google Scholar]
  15. Bučko T., Šimko F.. On the structure of crystalline and molten cryolite: Insights from the ab initio molecular dynamics in NpT ensemble. J. Chem. Phys. 2016;144(6):064502. doi: 10.1063/1.4941333. [DOI] [PubMed] [Google Scholar]
  16. Machado K., Zanghi D., Salanne M., Stabrowski V., Bessada C.. Anionic Structure in Molten Cryolite-Alumina Systems. J. Phys. Chem. C. 2018;122(38):21807–21816. doi: 10.1021/acs.jpcc.8b06905. [DOI] [Google Scholar]
  17. Lacassagne V., Bessada C., Florian P.. et al. Structure of high-temperature NaF-AlF3-Al2O3 melts: A multinuclear NMR study. J. Phys. Chem. B. 2002;106(8):1862–1868. doi: 10.1021/jp013114l. [DOI] [Google Scholar]
  18. Gilbert B., Mamantov G., Begun G. M.. Raman spectra of Al2O3 solutions in molten cryolite and other aluminum fluoride containing melts. Inorg. Nucl. Chem. Lett. 1976;12(5):415–424. doi: 10.1016/0020-1650(76)80052-9. [DOI] [Google Scholar]
  19. Malherbe C., Eppe G., Gilbert B.. Determination of the Al2O3 content in NaF-AlF 3-CaF2-Al2O3 melts at 950 °c by raman spectroscopy. Anal. Chem. 2014;86(16):8073–8081. doi: 10.1021/ac500923g. [DOI] [PubMed] [Google Scholar]
  20. Picard G. S., Bouyer F. C., Leroy M., Bertaud Y., Bouvet S.. Structures of oxyfluoroaluminates in molten cryolite-alumina mixtures investigated by DFT-based calculations. J. Mol. Struct.: THEOCHEM. 1996;368(1–3):67–80. doi: 10.1016/S0166-1280(96)04611-8. [DOI] [Google Scholar]
  21. Lin M., Hu X., Yu J., Yang Y., Shi Z., Wang Z.. Morphology of alumina in NaF-AlF3 systems determined by Raman spectroscopy and quantum mechanical calculations. J. Mol. Liq. 2020;315:113747. doi: 10.1016/j.molliq.2020.113747. [DOI] [Google Scholar]
  22. Zhang F., You J., Canizarès A.. et al. Quantitative studies on the microstructure evolution and its impact on the viscosity of a molten Al2O3-Na3AlF6 system by Raman spectroscopy and theoretical simulations. J. Mol. Liq. 2024;396:123732. doi: 10.1016/j.molliq.2023.123732. [DOI] [Google Scholar]
  23. Chen C., Li J., Luo R., Zhang H.. Ab initio molecular dynamics study on microstructure and diffusion properties of CaF2–2.2NaF-AlF3 system. Chem. Phys. 2021;550:111294. doi: 10.1016/j.chemphys.2021.111294. [DOI] [Google Scholar]
  24. Kühne T. D., Iannuzzi M., Del Ben M.. et al. CP2K: An electronic structure and molecular dynamics software package -Quickstep: Efficient and accurate electronic structure calculations. J. Chem. Phys. 2020;152(19):194103. doi: 10.1063/5.0007045. [DOI] [PubMed] [Google Scholar]
  25. Bale C. W., Bélisle E., Chartrand P.. et al. FactSage thermochemical software and databases - recent developments. CALPHAD. 2009;33(2):295–311. doi: 10.1016/j.calphad.2008.09.009. [DOI] [Google Scholar]
  26. Martinez L., Andrade R., Birgin E. G., Martínez J. M.. PACKMOL: A package for building initial configurations for molecular dynamics simulations. J. Comput. Chem. 2009;30(13):2157–2164. doi: 10.1002/jcc.21224. [DOI] [PubMed] [Google Scholar]
  27. Chakraborty B., Kidwai S., Ramaniah L. M.. Trapping of Li+ Ions by [ThFn]­4-n Clusters Leading to Oscillating Maxwell-Stefan Diffusivity in the Molten Salt LiF-ThF4. J. Phys. Chem. B. 2016;120(32):7926–7936. doi: 10.1021/acs.jpcb.6b05173. [DOI] [PubMed] [Google Scholar]
  28. Humphrey, W. ; Dalke, A. ; Schulten, K. . VMD: Visual Molecular Dynamics; VMD, 1996. [DOI] [PubMed] [Google Scholar]
  29. Stukowski A.. Visualization and analysis of atomistic simulation data with OVITO–the Open Visualization Tool. Modell. Simul. Mater. Sci. Eng. 2010;18(1):015012. doi: 10.1088/0965-0393/18/1/015012. [DOI] [Google Scholar]
  30. Machado K., Zanghi D., Salanne M., Bessada B. C.. Structural, Dynamic, and Thermodynamic Study of KF–AlF 3 Melts by Combining High-Temperature NMR and Molecular Dynamics Simulations. J. Phys. Chem. C. 2019;123(4):2147–2156. doi: 10.1021/acs.jpcc.8b11907. [DOI] [Google Scholar]
  31. Picard G., Bouyer F., Leroy M., Bertaud Y., Bouvet S.. Structures of oxyfluoroaluminates in molten cryolite-alumina mixtures investigated by DFT-based calculations. J. Mol. Struct.: THEOCHEM. 1996;368:67–80. doi: 10.1016/S0166-1280(96)90539-4. [DOI] [Google Scholar]
  32. Ratkje S. K.. OXY-FLUORO ALUMINATE COMPLEXES IN MOLTEN CRYOLITE MELTS. Electrochim. Acta. 1976;21(7):515–517. doi: 10.1002/chin.197641017. [DOI] [Google Scholar]
  33. Arkhipov P., Tkacheva O.. The electrical conductivity of molten oxide-fluoride cryolite mixtures. Materials. 2021;14(23):7419. doi: 10.3390/ma14237419. [DOI] [PMC free article] [PubMed] [Google Scholar]
  34. Gheribi A. E., Serva A., Salanne M.. et al. Study of the Partial Charge Transport Properties in the Molten Alumina via Molecular Dynamics. ACS Omega. 2019;4(5):8022–8030. doi: 10.1021/acsomega.9b01110. [DOI] [PMC free article] [PubMed] [Google Scholar]
  35. Votava I., Matiasovsky K.. Measurement of viscosity of fused salts. II. Viscosity of molten binary mixtures on the cryolite basis. Chem. Zvesti. 1973;27(5):582–587. [Google Scholar]
  36. Kvande H., Rorvik H.. Influence of bath density in aluminum electrolysis. J. Met. 1984;36(12):62. 420 COMMONWEALTH DR, WARRENDALE, PA 15086: MINERALS METALS MATERIALS SOC. [Google Scholar]

Articles from ACS Omega are provided here courtesy of American Chemical Society

RESOURCES