Chemical interactions between zirconium and free oxide in molten fluorides

Abstract

The chemical interactions between zirconium and free oxide in FLiNaK melts at different zirconium to oxide (n_Zr/n_O) molar ratios were studied by means of a carbothermal-reduction technique (LECO oxide analyzer) and Raman spectroscopy. ZrO₂ precipitates were formed with n_Zr/n_O ≤ 0.5, while ZrO₂ was converted to Zr₂OF_x^6−x complexes with n_Zr/n_O > 0.5. The maximum amount of Zr₂OF_x^6−x complexes in FLiNaK melts was found to be 0.020 mol kg⁻¹. With an initial oxide concentration lower than 0.020 mol kg⁻¹ and the required amount of ZrF₄, the free oxide in FLiNaK melts could be completely converted to Zr₂OF_x^6−x complexes, which would further prevent the formation of UO₂ precipitates.

---

1. Introduction

The molten salt reactor (MSR) is one of the six internationally recognized Generation IV reactor concepts,^1–3 which has been chosen as an ideal nuclear reactor to utilize uranium and thorium resources. Unlike other reactor types, MSR uses a ⁷LiF–BeF₂ melt as a carrier salt, in which the fissile material (uranium in the form of ²³³UF₄) and fertile material (thorium in the form of ²³²ThF₄) are dissolved.^4–6 UF₄ is prone to react with free oxide (such as H₂O and dissolved O²⁻), resulting UO₂ with much higher melting point and lower solubility. Despite the loss of dissolved uranium, this may cause criticality concerns, as the large accumulation of UO₂ would lead to superheated area. Therefore the oxide concentration should be strictly controlled.^7–9 Early studies at Oak Ridge National Laboratory (ORNL) by Rosenthal et al.¹⁰ showed that the solubility product (K_SP) of UO₂ in LiF–BeF₂ molten salt at 500 °C was 2.6 × 10⁻⁷ mol³ kg⁻³. Thus, the free oxide tolerance in the molten salt breeding reactor (MSBR) consisting of LiF–BeF₂–ThF₄–UF₄ (71.7 : 16 : 12 : 0.3 mol%) is probably no greater than 30 ppm.

ORNL proposed two ways to prevent UO₂ precipitation.¹¹ One is to eliminate the free oxide by bubbling HF–H₂ mixtures into the salt,^11,12 as O²⁻ ions react with an excess of HF according to the reaction O²⁻ + 2HF → H₂O↑ + 2F⁻. The other is to improve the oxide tolerance of uranium-based systems by adding some oxygen getters.^10,13,14 Due to its high solubility in the fluorides,^15,16 HF (g) retains stubbornly in the salts and causes the corrosion of structural material.^17,18 Furthermore, HF–H₂ purification is unable to confine the free oxide in fluorides to 30 ppm. Therefore, the preferable way is to introduce an oxygen getter to ensure the safe operation of MSR.

Five mole percent of ZrF₄ was added to the LiF–BeF₂–UF₄ salts to getter any oxide impurities through the reaction 2H₂O + ZrF₄ = 4HF + ZrO₂, as ZrF₄ was more favorable to react with moisture than with UF₄.⁸ Still, Korenko¹⁹ discovered that though ZrO₂ were found in KF–LiF–NaF–UF₄–ZrF₄ by the X-ray diffraction (XRD) phase analysis, the UO₂ could still be detected after the solidification of the molten system KF–LiF–NaF–UF₄ without addition of ZrF₄. Gibilaro²⁰ also found the precipitation of ZrO₂ and ZrO_1.3F_1.4 after CaO was added in the molten LiF–CaF₂–ZrF₄. These studies demonstrated that the zirconium-based systems reduced the free oxide by forming ZrO₂ to prevent the precipitation of UO₂. In this study, we found that the ZrO₂ precipitates formed at n_Zr/n_O ≤ 0.5, where n_Zr/n_O refers to the molar ratios of zirconium to oxide. While with n_Zr/n_O > 0.5, ZrO₂ would be converted to Zr₂OF_x^6−x complexes. Here, we primarily focused on the chemical interactions of zirconium and free oxide at different molar ratios and further explored the mechanism of ZrF₄ protection against UO₂ precipitation in molten FLiNaK.

2. Experimental section

2.1 Chemicals

The 300 g of LiF–NaF–KF [FLiNaK] mixture with a mole percent of 46.5 : 11.5 : 42 was weighed. The FLiNaK powders were heat to 300 °C and held at this temperature for 24 h, and then gradually heated to 600 °C under argon atmosphere. The salt was then dehydrated and deoxygenated by bubbling a mixture of HF (g)–H₂ (g) (1 : 10 mole ratio). This process is consistent with that reported by ORNL-4616.¹¹ After treatment, the oxide in FLiNaK melts can be lowered from 0.134 mol kg⁻¹ to 0.008 mol kg⁻¹ (Table 1), as detected by carbothermal-reduction technique using LECO RO-600 (LECO Co., Ltd.). Finally, this treated FLiNaK salt was transferred to a nickel crucible with an internal diameter of 78 mm. A certain amount of Li₂O (Sigma-Aldrich, 99.99%) was added into FLiNaK melts to acquire a quantified dissolved free oxide concentration, and the zirconium ions were then gradually introduced into the melts in the form of zirconium tetrafluoride (ZrF₄) (Sigma-Aldrich, 99.99%) to prepare assorted melts with different n_Zr/n_O. The fluorides in a nickel crucible were placed in a stainless steel cell inside an electric furnace and melted at 600 °C. The temperature was measured by a nickel–chromium thermocouple positioned just outside the crucible. The whole experiment was conducted inside a glove box with dry argon atmosphere (99.99%), and the concentration of moisture and oxygen in the glove box were both under 2 ppm.

Table 1 The concentration of oxide in FLiNaK melts before and after dehydration

LiF–NaF–KF (46.5 : 11.5 : 42 mol%) [FLiNaK] mixture	Concentration of oxide analyzed by LECO (mol kg^−1)
Original material	0.134
After dehydration by HF–H2	0.008

---

2.2 Analytical techniques

During experiment, a portion of the melts was withdrawn from the crucible by a nickel sampler (Φ6 mm) for further analysis.

The oxide concentration of the extracted samples was determined by carbothermal reduction technique using LECO, according to the work done by Mediaas et al.^21,22 A 0.1–0.15 g sample of molten fluorides was sealed in a tin capsule. Then the tin capsule was heated in a graphite crucible to 2700 °C over a short period of time (up to 400 s). The oxides in the sample would completely react with carbon at this temperature to form CO (g) and CO₂ (g), which can be detected in an Infrared Radiation (IR) cell. To avoid any systematic error, five samples were chosen from different locations of the melting salts and analyzed respectively. A mean value as a final result was reported.

The zirconium in FLiNaK was determined by Inductively Coupled Plasma Optical Emission Spectroscopy (ICP-OES, Arcos, Spectro Co., Ltd.). The samples were dissolved in the 4% (v/v) HNO₃ (Jingrui) and then introduced in the spectrometer.

Raman analysis was performed to identify the structures of salt. The 532 nm line of an argon ion laser with 100 mW average power was used for exciting the sample. For analyzing and collecting the spectra, the HR 800 (Jobin Yvon) Raman system equipped with CCD detector was used in the triple configuration. The system was interfaced with a personal computer and the spectra were saved in digital form. We carried out in situ studies using a Linkam TS1000 environmental cell. The sample was put in a ceramic furnace with a thermocouple in intimate contact. Gas flowing over the sample was regulated by electronic thermal mass flow controllers and the laser was focused on the sample through a water-cooled silica window in the cover of the cell. The in situ studies detailed here used a ramp with rate of 5 °C min⁻¹ to the appropriate temperature, and the temperature was then held to be constant while Raman data were acquired.

3. Results and discussion

3.1 The calibration of LECO

The demonstration of accuracy for LECO was implemented in advance. After addition of different amounts of Li₂O into the FLiNaK melts, LECO was used to detect the increments of oxide,²³ which should be consistent with the theoretical values. Fig. 1 shows the relation of the theoretical increments vs. detected ones. The slope is approximately equal to 1, which indicates the LECO is reliable to detect the variation of oxide in FLiNaK melts.

Fig. 1 The relation between the theoretical and detected increment of oxide in FLiNaK after addition of different amounts of Li₂O.

3.2 Chemical interactions between zirconium and free oxide in molten FLiNaK

3.2.1 n_Zr/n_O ≤ 0.5. To calculate n_Zr/n_O, n_O is a constant of 0.067 mol kg⁻¹ as derived from the LECO data of the melts used for calibration, while n_Zr represents the sum of ZrF₄ that has been added. The concentration of oxide in FLiNaK melts varying with n_Zr/n_O was analyzed at 600 °C and the results were plotted in Fig. 2. Interestingly, by increasing n_Zr/n_O from 0 to 0.5, the oxide concentration decreases from 0.067 to 0.010 mol kg⁻¹. This is probably caused by the formation of ZrO₂ precipitates, as shown in eqn (1).

2O²⁻ + ZrF₄ = ZrO₂↓ + 4F⁻ (1)

Fig. 2 The concentration of oxide in FLiNaK melts versus n_Zr/n_O at 600 °C.

When n_Zr/n_O equals to be 0.5, the whole cylinder block is taken out of the nickel crucible after the molten salt was solidified. A relatively thin layer of precipitates is observed at the bottom (see Fig. 3(a)), which presents a different appearance from that of the rest of the cylinder (Fig. 3(b) shows the upper surface as reference). Then ca. 0.1 g precipitates were sampled and transferred to a sealed high-purity quartz container filled with argon gas for Raman analysis. Fig. 4 shows that the Raman spectrum of the bottom sample is in good agreement with that of the monoclinic ZrO₂ (Sigma-Aldrich, 99.99%). Therefore, we can conclude that the precipitates formed with n_Zr/n_O ≤ 0.5 in FLiNaK melts correspond to ZrO₂ (see eqn (1)). However, the oxide concentration began to increase when it furthered to add ZrF₄, which can be concluded that a solubility equilibrium of ZrO₂ may be established. This will be discussed in the next section.

Fig. 3 The solidified FLiNaK melts prepared with initial oxide concentration of 0.067 mol kg⁻¹ and n_Zr/n_O = 0.5; (a) bottom surface; (b) upper surface.

Fig. 4 The Raman spectra of monoclinic ZrO₂ and the bottom precipitates sampled from the solidified FLiNaK melts at initial oxide concentration of 0.067 mol kg⁻¹ and n_Zr/n_O = 0.5.

3.2.2 n_Zr/n_O > 0.5. By continuously adding ZrF₄ to increase the n_Zr/n_O, the oxide concentration then rises up and reaches a plateau, as shown in Fig. 2. Specifically, the oxide concentration rises from 0.009 to 0.020 mol kg⁻¹ by increasing n_Zr/n_O from 0.7 to 2.3. This increment might suggest a process that the precipitated ZrO₂ is capable of reacting with the excessive ZrF₄ to form Zr₂OF_x^6−x complexes, which re-dissolves in the FLiNaK melts.

After n_Zr/n_O = 3.9 is attained, a nickel rod was quickly inserted into the melts vertically without touching the bottom precipitates and then it was rapidly extracted. During this process, the melts adhered to the rod and experienced a quenching process. Due to its inhomogeneity, two Raman spectra of the sample were measured at 25 °C, with two peaks at 250 and 548 cm⁻¹ in Fig. 5(b) and five peaks at ca. 223, 300, 395, 520 and 548 cm⁻¹ in Fig. 5(d).

Fig. 5 The Raman spectra of (a) the calculated Raman spectrum of ZrF₆²⁻ complex; (c) the calculated Raman spectrum of Zr₂OF₁₀⁴⁻; (b) and (d) the upper melts extracted from FLiNaK at initial oxide = 0.067 mol kg⁻¹ and n_Zr/n_O = 3.9 measured at 25 °C; (e) measured at 600 °C.

The spectrum of the calculated octahedral [ZrF₆]²⁻ structure was shown in Fig. 5(a). It also appeared two peaks at 231 cm⁻¹ and 548 cm⁻¹, corresponding to the bending and symmetrical stretching vibration of Zr–F, respectively, which was consistent with that of Fig. 5(b). Besides, the predominance of octahedral [ZrF₆]²⁻ was also detected in KF–ZrF₄ melts as reported by Dracopoulos²⁴ and Pauvert.^25,26 Thus, the as-detected [ZrF₆]²⁻ structure in the melting FLiNaK prepared with n_Zr/n_O = 3.9 is probably formed via the following reaction [eqn (2)].

2F⁻ + ZrF₄ = ZrF₆²⁻ (2)

The cluster model of F–Zr–O was computed by the quantum chemical ab initio calculations based on the Gaussian 09 software and it was further optimized by the Restricted Hartree–Fock (RHF) method and connection of SDD. The calculated Raman spectrum of [Zr₂OF₁₀]⁴⁻ structure was plotted in Fig. 5(c) and it basically agreed with that of Fig. 5(d). But the peak at 223 cm⁻¹ in Fig. 5(d) was due to the overlapping of the two peaks 230 and 250 cm⁻¹ in Fig. 5(c), corresponding to the F–Zr–O bending vibration and F–Zr–O ω-rocking vibration, respectively. The peaks at 300, 520 and 548 cm⁻¹ were antisymmetric stretching vibration of Zr–F and the peak at 395 cm⁻¹ corresponded to the symmetrical stretching vibration of Zr–F. As the test temperature increased to 600 °C, the sample melted to liquid and became homogenous. The Raman spectrum was shown in Fig. 5(e). The expected trends was observed; i.e. red frequency shifting at high temperature.²⁴ Besides, another peak for F–Zr–O symmetrical stretching vibration was observed at 410 cm⁻¹. This peak was probably attributed to [ZrOF₆]⁴⁻ produced by the decomposition of [Zr₂OF₁₀]⁴⁻ at 600 °C.

The detected [Zr₂OF₁₀]⁴⁻ in the melting FLiNaK was probably formed via eqn (3), and the increment of detected oxide concentration by increasing n_Zr/n_O from 0.7–2.3 (Fig. 2) was also linked to the formation of Zr₂OF_x^6−x complexes.

ZrO₂ + 3ZrF₄ + 8F⁻ = 2[Zr₂OF₁₀]⁴⁻ (3)

The oxyfluorides in molten fluorides have already been addressed in the field of rare earth^27–34 and aluminum production.^35,36 Stefanidaki³² found that the solubility of Nd₂O₃ at 900 °C increased to 0.15, 0.22 and 0.38 mol% when NdF₃ in LiF melt was increased to 15, 23 and 30 mol%, respectively. They proved that dissolution of Nd₂O₃ in NdF₃–LiF is attributed to a reaction of Nd₂O₃ + NdF₃ + 3(x − 1)F⁻ = 3NdOF_x^(x−1)−. Haas³⁷ also suggested that the solubility of UO₂ in LiF–BaF₂–UF₄ or LiF–CaF₂–UF₄ is strongly influenced by the concentration of UF₄ in the melt, which is connected to the possible existence of complex UO_xF_y. Additionally, the solubility of La₂O₃ is lower in LiF melts²⁷ than that in LaF₃–LiF melts due to the formation of LaO_xF_y complexes.^28,29 In this study, the precipitated ZrO₂ also reacted with the excessive ZrF₄ and F⁻ at n_Zr/n_O > 0.5 to form Zr₂OF_x^6−x complexes as indicated in eqn (3), which obviously increased the solubility of ZrO₂ in the FLiNaK melts. This dissolution mechanism is similar to those reported by Stefanidaki,³² Haas,³⁷ M. Ambrová,²⁷ and Rollet.^28,29

3.3 The solubility equilibrium of ZrO₂ in FLiNaK–ZrF₄ melts

Under the condition of n_Zr/n_O = 3.9, the initial oxide concentration of 0.067 mol kg⁻¹ in FLiNaK melts was finally converted to that of in ZrO₂ precipitates and in Zr₂OF_x^6−x complexes. And the Zr₂OF_x^6−x complexes finally stabilized at 0.020 mol kg⁻¹. With an initial oxide concentration of 0.020 mol kg⁻¹ and addition enough amount of ZrF₄, one could assume that the free oxide in FLiNaK melts can be completely converted to that in the complexes without precipitating ZrO₂. As shown in Fig. 6, the oxide concentration drops sharply from 0.020 to 0.009 mol kg⁻¹ when n_Zr/n_O is increased from 0 to 0.5. With continual addition of ZrF₄, the oxide concentration reaches its lowest value 0.009 mol kg⁻¹ and then begins to go up. The curve in Fig. 6 reveals a similar trend as observed in Fig. 2. These observations indicate that the solubility equilibrium of ZrO₂ was established.

Fig. 6 The concentration of oxide in the FLiNaK melts versus n_Zr/n_O at 600 °C.

The descent part of the curve (n_Zr/n_O ≤ 0.5) in Fig. 6 reveals that a reaction occurred between the initial free oxide and the added ZrF₄, and ZrO₂ precipitates as well as its solubility equilibrium were formed. With the further addition of ZrF₄, the oxide concentration reached its minimum value which was 0.009 mol kg⁻¹ and the maximum amount of ZrO₂ precipitates was also achieved. Since the solubility equilibrium of ZrO₂ was preserved as long as the ZrO₂ precipitates existed, it can be concluded that the minimum oxide concentration of 0.009 mol kg⁻¹ was produced by the dissolved ZrO₂. According to the stoichiometric relation [eqn (1)], the solubility of ZrO₂ in FLiNaK melts equaled to be 0.0045 mol kg⁻¹ (0.019 mol%).

The ascending part of the curve (n_Zr/n_O > 0.5) in Fig. 6 shows that the ZrO₂ started to react with the excessive ZrF₄ to form Zr₂OF_x^6−x complexes according to eqn (3), and the solubility of ZrO₂ in FLiNaK–ZrF₄ melts gradually increased as well. Finally, the maximum dissolved oxide of 0.020 mol kg⁻¹ was attained in FLiNaK–ZrF₄ melts, indicating that the maximum solubility of ZrO₂ was 0.020 mol kg⁻¹ (0.080 mol%), which was much lower than that of Nd₂O₃ in LiF–NdF₃ (ref. 32) and UO₂ in CaF₂–LiF–UF₄ and BaF₂–LiF–UF₄.³⁷ It is worth mentioning that with an initial oxide concentration less than 0.020 mol kg⁻¹, the free oxide can be fully converted to [Zr₂OF_x]^6−x complexes without any ZrO₂ precipitates, given that enough ZrF₄ was added.

3.4 The mechanism of ZrF₄ protection against the UO₂ precipitation

Further validation for the conversion of free oxide to Zr₂OF_x^6−x complexes was carried out. Two sets of FLiNaK melts were prepared by adding Li₂O. They both had the same initial concentration of free oxide 0.020 mol kg⁻¹. One set was added with ZrF₄ to achieve n_Zr/n_O = 7.0, while the other set was not (n_Zr/n_O = 0). Varied amounts of UF₄ (China North Nuclear Fuel Co., Ltd, 99.99%) were then introduced into the two prepared melts, respectively.

In the melts prepared with n_Zr/n_O = 0, the oxide concentration detected by LECO was decreased from 0.020 mol kg⁻¹ to 0.016 mol kg⁻¹, while the uranium concentration analyzed by ICP-OES was reduced from 0.007 mol kg⁻¹ to 0.005 mol kg⁻¹ in the first 3 h, as shown in Fig. 7(a). From 3 to 60 h, the decrements of both oxide and uranium gradually diminished and their concentrations reached plateaus at 0.013 mol kg⁻¹ and 0.004 mol kg⁻¹, respectively. Since the decrement ratio of oxide and uranium approximately equals to 2, it can be concluded that the UO₂ precipitates were formed from the reaction between UF₄ and free oxide, as shown in eqn (4). In addition, a layer of precipitates with rufous color was observed on the bottom surface after solidification, as shown in Fig. 8(a). The Raman spectrum of this precipitates is consistent with that of UO₂ reported by Stefaniak et al.,³⁸ as shown in Fig. 9.

2O²⁻ + UF₄ = UO₂↓ + 4F⁻ (4)

Fig. 7 The concentration of oxide and uranium varied with time at 600 °C (a) in FLiNaK melts with initial oxide at 0.020 mol kg⁻¹ and n_Zr/n_O = 0; (b) in FLiNaK melts with initial oxide at 0.020 mol kg⁻¹ and n_Zr/n_O = 7.0.

Fig. 8 The solidified FLiNaK melts with initial oxide at 0.020 mol kg⁻¹ and different n_Zr/n_O; (a) n_Zr/n_O = 0; (b) n_Zr/n_O = 7.0.

Fig. 9 The Raman spectra of UO₂ (ref. 38) ore and the bottom precipitates taken from the solidified FLiNaK–UF₄ (0.007 mol kg⁻¹) melts at initial oxide = 0.020 mol kg⁻¹ and n_Zr/n_O = 0.

With n_Zr/n_O = 7.0, the concentration of oxide was merely varied in the range of 0.019 mol kg⁻¹ to 0.021 mol kg⁻¹ in 60 h, while the concentration of uranium approximately maintained at 0.005 mol kg⁻¹. Besides, the solidified melt consisted of a small amount of green colored bulk and no precipitates were observed at the bottom surface, as shown in Fig. 8(b). All these indicate that no chemical reaction between oxide and uranium was occurred. Therefore, with an initial oxide concentration less than 0.020 mol kg⁻¹, the free oxide can be fully converted to [Zr₂OF_x]^6−x complexes by introducing enough ZrF₄ into FLiNaK melts. The conversion of free oxide to Zr₂OF_x^6−x complexes prevented the formation of UO₂ precipitates. This method will provide a very efficient way to control the uranium species against the oxygenation in MSR.

4. Conclusions

This work focused on the interactions between zirconium and free oxide in FLiNaK system. It was found that the initial concentration of oxide which was 0.067 mol kg⁻¹ in FLiNaK melts was finally converted to that of in ZrO₂ precipitates at n_Zr/n_O ≤ 0.5 and in Zr₂OF_x^6−x complexes at n_Zr/n_O > 0.5 analyzed by LECO and Raman spectroscopy. The maximum Zr₂OF_x^6−x complexes of FLiNaK melts was found to be 0.020 mol kg⁻¹. When the initial concentration of oxide was less than 0.020 mol kg⁻¹, the free oxide in FLiNaK melts would be completely converted to Zr₂OF_x^6−x complexes without any ZrO₂ precipitates, given that enough ZrF₄ was added.

UF₄ was introduced to FLiNaK melts for further validating the conversion of free oxide to Zr₂OF_x^6−x complexes. The results showed that the UO₂ precipitates had been formed in the FLiNaK system at n_Zr/n_O = 0, while no chemical reaction between oxide and uranium was occurred at n_Zr/n_O = 7.0. The conversion of free oxide to Zr₂OF_x^6−x complexes prevented the formation of UO₂ precipitates. This method will provide a very efficient way to control the uranium species against the oxygenation in MSR.

Acknowledgements

This work was financially supported by the Strategic Priority Research Program of the Chinese Academy of Sciences (Grant no. XDA02020400) and the Young Scientists Fund of the National Natural Science Foundation of China (Grant no. 51404237).

References

1. GIF-002-00, http://gif.inel.gov, December 2002.
2. C. W. Forsberg, The Americas Nuclear Energy Symposium (ANES), American Nuclear Society, Miami, Florida, 2002.
3. C. W. Forsberg, P. F. Peterson and L. Ott, Proceedings of the 2004 Americas Nuclear Energy Symposium, Miami Beach, Florida, October 2004.
4. M. Rosenthal, P. Kasten and R. Briggs, Nucl. Appl. Technol., 1970, 8, 107–117.
5. C. Forsberg, Prog. Nucl. Energy, 2005, 47, 32–43.
6. A. Nuttin, D. Heuer, A. Billebaud, R. Brissot, C. Le Brun, E. Liatard, J.-M. Loiseaux, L. Mathieu, O. Meplan and E. Merle-Lucotte, Prog. Nucl. Energy, 2005, 46, 77–99.
7. W. Grimes, Nucl. Technol., 1970, 8, 137–155.
8. L. Toth, U. Gat, G. Del Cul, S. Dai and D. Williams, Review of ORNL's MSR technology and status, Oak Ridge National Lab., TN, United States, Funding organisation: USDOE, Washington, DC (United States), 1996.
9. R. Ross, C. Bamberger and C. Baes, J. Inorg. Nucl. Chem., 1973, 35, 433–449.
10. Ed. M. Rosenthal, P. Haubenreich and R. Briggs, ORNL-4812, 1972.
11. J. H. Shaffer, in ORNL-4616, 1971.
12. A. Mathews and C. F. Baes Jr, Inorg. Chem., 1968, 7, 373–382.
13. R. B. Briggs, in ORNL-3122, 1960.
14. E. G. B. W. R. Grimes, H. F. McDuffie, G. M. Watson, F. F. Blankenship and C. H. Secoy, in ORNL-4076, 1966.
15. J. Shaffer, W. Grimes and G. Watson, J. Phys. Chem., 1959, 63, 1999–2002.
16. P. E. Field and J. H. Shaffer, J. Phys. Chem., 1967, 71, 3218–3222.
17. D. A. Petti, G. Smolik, M. F. Simpson, J. P. Sharpe, R. Anderl, S. Fukada, Y. Hatano, M. Hara, Y. Oya and T. Terai, Fusion Eng. Des., 2006, 81, 1439–1449.
18. P. Calderoni, P. Sharpe, H. Nishimura and T. Terai, J. Nucl. Mater., 2009, 386, 1102–1106.
19. M. Korenko, M. Straka, J. Uhlíř, L. Szatmáry, M. Ambrová and M. Šimurda, J. Radioanal. Nucl. Chem., 2014, 302, 549–554.
20. M. Gibilaro, L. Massot, P. Chamelot, L. Cassayre and P. Taxil, Electrochim. Acta, 2013, 95, 185–191.
21. H. Mediaas, J. Vinstad and T. Østvold, Solubility of MgO in MgCl {sub 2}–NaCl–NaF melts, Minerals, Metals and Materials Society, Warrendale, PA, United States, 1996.
22. G. Kipouros, H. Mediaas, J. Vindstad, T. Oestvold and O. Tkatcheva, Oxide solubilities and phase relations in the system Mg–Nd–O–Cl, Minerals, Metals and Materials Society, Warrendale, PA, United States, 1996.
23. P. Chamelot, P. Palau, L. Massot, A. Savall and P. Taxil, Electrochim. Acta, 2002, 47, 3423–3429.
24. V. Dracopoulos, J. Vagelatos and G. Papatheodorou, J. Chem. Soc., Dalton Trans., 2001, 1117–1122.
25. O. Pauvert, D. Zanghi, M. Salanne, C. Simon, A. Rakhmatullin, H. Matsuura, Y. Okamoto, F. Vivet and C. Bessada, J. Phys. Chem. B, 2010, 114, 6472–6479.
26. O. Pauvert, M. Salanne, D. Zanghi, C. Simon, S. Reguer, D. Thiaudière, Y. Okamoto, H. Matsuura and C. Bessada, J. Phys. Chem. B, 2011, 115, 9160–9167.
27. M. Ambrová, J. Jurišová, V. Danielik and J. Gabčová, J. Therm. Anal. Calorim., 2008, 91, 569–573.
28. A. L. Rollet, H. Matsuura and C. Bessada, Dalton Trans., 2015, 522–529.
29. A. L. Rollet, E. Veron and C. Bessada, J. Nucl. Mater., 2012, 429, 40–47.
30. M. Gibilaro, L. Cassayre, O. Lemoine, L. Massot, O. Dugne, R. Malmbeck and P. Chamelot, J. Nucl. Mater., 2011, 414, 169–173.
31. M. Gibilaro, J. Pivato, L. Cassayre, L. Massot, P. Chamelot and P. Taxil, Electrochim. Acta, 2011, 56, 5410–5415.
32. E. Stefanidaki, G. M. Photiadis, C. G. Kontoyannis, A. F. Vik and T. Østvold, J. Chem. Soc., Dalton Trans., 2002, 2302–2307.
33. A. F. Vik, V. Dracopoulos, G. N. Papatheodorou and T. Østvold, J. Alloys Compd., 2001, 321, 284–299.
34. J. Von Barner, E. Christensen, N. Bjerrum and B. Gilbert, Inorg. Chem., 1991, 30, 561–566.
35. V. Lacassagne, C. Bessada, P. Florian, S. Bouvet, B. Ollivier, J.-P. Coutures and D. Massiot, J. Phys. Chem. B, 2002, 106, 1862–1868.
36. C. Malherbe, G. Eppe and B. Gilbert, Anal. Chem., 2014, 86, 8073–8081.
37. P. A. Haas and S. P. Cooper, J. Chem. Eng. Data, 1993, 38, 26–30.
38. E. A. Stefaniak, A. Alsecz, I. E. Sajó, A. Worobiec, Z. Máthé, S. Török and R. Van Grieken, J. Nucl. Mater., 2008, 381, 278–283.

---