Inhibition effect of ZrF₄ on UO₂ precipitation in the LiF–BeF₂ molten salt

Abstract

The dissolution–precipitation behavior of zirconium dioxide (ZrO₂) in molten lithium fluoride–beryllium fluoride (LiF–BeF₂, (2 : 1 mol, FLiBe)) eutectic salt at 873 K was studied. The results of the dissolution experiment showed that the saturated solubility of ZrO₂ in the FLiBe melt was 3.84 × 10⁻³ mol kg⁻¹ with equilibrium time of 6 h, and its corresponding apparent solubility product (K′_sp) was 3.40 × 10⁻⁵ mol³ kg⁻³. The interaction between Zr(IV) and O²⁻ was studied by titrating lithium oxide (Li₂O) into the FLiBe melt containing zirconium tetrafluoride (ZrF₄), and the concentration of residual Zr(IV) in the melt gradually decreased due to precipitate formation. The precipitate corresponded to ZrO₂, as confirmed by the stoichiometric ratio and X-ray diffraction analysis. The K′_sp was 3.54 × 10⁻⁵ mol³ kg⁻³, which was highly consistent with that from the dissolution experiment. The obtained K′_sp of ZrO₂ was in the same order of magnitude as that of uranium dioxide (UO₂), indicating that a considerable amount of ZrF₄ could inhibit the UO₂ formation when oxide contamination occurred in the melt containing ZrF₄ and uranium tetrafluoride (UF₄). Further oxide titration in the LiF–BeF₂–ZrF₄ (5 mol%)–UF₄ (1.2 mol%) system showed that ZrO₂ was formed first with O²⁻ addition less than 1 mol kg⁻¹, and the precipitation of UO₂ began only after the O²⁻ addition reached 1 mol kg⁻¹ and the precipitation of ZrO₂ decreased the ZrF₄ concentration to 0.72 mol kg⁻¹ (3 mol%). Lastly, UO₂ and ZrO₂ coprecipitated with further O²⁻ addition of more than 1 mol kg⁻¹. The preferential formation of ZrO₂ effectively avoided the combination of UF₄ and O²⁻. This study provides a solution for the control of UO₂ precipitation in molten salt reactors.

---

1. Introduction

Molten salt reactors (MSRs) use molten fluoride mixtures, usually lithium fluoride–beryllium fluoride mixed melt (LiF–BeF₂ (2 : 1 mol, FLiBe)) as the carrier solvent of fuel, with uranium tetrafluoride (UF₄) or thorium tetrafluoride (ThF₄) dissolved as fuel salt.^1–3 However, these fluorides can absorb water due to their hygroscopic nature, and pyrohydrolysis commonly occurs upon melting, thus producing an active oxide impurity (O²⁻) (eqn (1)).^4,5

H₂O + 2F⁻ = 2HF + O²⁻ (1)

Unfortunately, UF₄ has a high sensitivity toward O²⁻ in fluoride melts and forms some insoluble compounds of oxides or oxyfluorides,⁶ e.g., uranium dioxide (UO₂) according to our previous investigations (eqn (2)).^7,8

UF₄ + 2O²⁻ = UO₂(s) + 4F⁻. (2)

The large accumulation of the fuel precipitate UO₂ would generate a superheated area in the reactor and cause a criticality risk.^7–11 Hence, the generation of UO₂ in the fuel salt should be strictly controlled to ensure the safe operation of MSR. Two methods have been proposed to protect uranium from UO₂ precipitate formation. One method is to remove the O²⁻ impurity in the fuel salt before loaded into the reactor, whereas the other approach is to develop feasible additives to collect O²⁻ or facilitate the solubility of UO₂ in molten fluorides during the MSR operation. The Oak Ridge National Laboratory (ORNL) has proposed that bubbling hydrogen fluoride and hydrogen gas (HF/H₂) mixtures into the salt is an effective means to remove O²⁻ (eqn (3)).^12–14

O²⁻ + 2HF(g) = H₂O(g) + 2F⁻. (3)

However, the HF/H₂ purification method has a limited capacity for oxide removal and cannot totally eliminate the O²⁻ due to the chemical equilibrium. Moreover, the active oxygen still can be produced through the unexpected nuclear reaction and brought by the cover gas and the external environment.^11,12,15 That means, the residual O²⁻ impurity contained in fluorides is inevitable and the formation of UO₂ during the MSR operation remains possible.

Therefore, the development of certain feasible additives to timely collect O²⁻ or the improvement of the solubility of UO₂ in the molten fluorides is more effective. Zirconium tetrafluoride (ZrF₄) may be a good selection, which has a strong affinity toward oxide and is a good oxide getter that can easily capture O²⁻ to form zirconium oxides (e.g., ZrO₂) or zirconium oxy-fluoride species (e.g., ZrOF₂ and Zr₂OF₁₀⁴⁻) in the molten fluorides.^16,17 In this manner, the combination of U(IV) and O²⁻ can be effectively avoided. Shen et al.¹⁶ have studied the interactions between Zr(IV) and O²⁻ in lithium fluoride–sodium fluoride–potassium fluoride (LiF–NaF–KF (46.5 : 11.5 : 42 mol%, FLiNaK)) molten salt and found that different products are formed at different molar ratios of nZr(IV)/nO²⁻. ZrO₂ is first formed with nZr(IV)/nO²⁻ ≤ 0.5, and the formed ZrO₂ redissolves into the melt as Zr₂OF₁₀⁴⁻ when nZr(IV)/nO²⁻ > 0.5. These results indicate that the free oxide ions (O²⁻) in the melt can be converted into oxyfluoride complex species (Zr₂OF₁₀⁴⁻) with enough ZrF₄ additive, which further prevents the UO₂ generation successfully. Besides, Gibilaro¹⁸ has found that the addition of calcium oxide (CaO) into a ZrF₄-containing lithium fluoride–calcium fluoride (LiF–CaF₂) salt leads to the formation of a close to equimolar mixture of solid ZrO₂ and ZrO_1.3F_1.4. Korenko et al.¹⁹ and our previous studies^7,8,16,20 have used X-ray diffraction (XRD) phase analysis and electrochemical techniques and discovered the ZrO₂ formation when oxide contamination occurs in the ZrF₄-containing systems (i.e., LiF–NaF–KF–ZrF₄ and LiF–NaF–KF–UF₄–ZrF₄), whereas only UO₂ is observed in the systems without ZrF₄ (i.e., LiF–NaF–KF–UF₄ and LiF–BeF₂–UF₄). These studies demonstrate that O²⁻ in zirconium-based systems can be reduced by forming Zr(IV) oxide or oxyfluorides, preventing the UO₂ precipitation.

ZrF₄ can also increase the solubility of oxides because of its strong affinity toward oxygen ions. Song et al.²¹ have studied the dissolution behavior of chromium sesquioxide (Cr₂O₃) in various molten fluorides and found that the ZrF₄ additive remarkably increases the solubility of Cr₂O₃ by 19 and 2 times in FLiNaK and FLiBe molten salt, respectively, most likely yielding the dissolution product of [ZrO_xF_y]^4−2x−y complex species. Peng et al.¹⁷ have studied the dissolution–precipitation behaviors of UO₂ in molten fluorides and found that ZrF₄ can also improve the solubility of UO₂. Their results show that the maximum solubility of UO₂ in FLiNaK melt increases by a factor of 5.76 (i.e., increase from 0.247 wt% to 1.422 wt%) when the added ZrF₄ concentration is up to 2.91 wt% and the oxide is dissolved as UOF₂ and ZrOF₂ species. These studies demonstrate that ZrF₄ can easily combine with oxygen ions and form soluble Zr(IV)–oxyfluorides to improve the solubility of oxides.

From the above, ZrF₄ is probably an ideal additive for controlling the precipitation of UO₂ in molten fluorides. Moreover, ZrF₄ has a low neutron-absorption cross-section, which allows its use in the fuel of MSR. Based on this finding, ORNL^10,11,22 adopts the LiF–BeF₂–ZrF₄–UF₄ (65 : 29.1 : 5 : 0.9 mol%) as the fuel salt of the molten salt reactor experiment, and the use of 5 mol% ZrF₄ successfully raises the oxide tolerance (the maximum allowable amount of O²⁻ added without the production of UO₂ in the molten salt) to above 100 ppm compared with that of 30 ppm in the molten salt breeder reactor by using a zirconium-free system, i.e., LiF–BeF₂–ThF₄–UF₄ (71.7 : 16 : 12 : 0.3 mol%). At present, the thorium molten salt reactor-liquid fuel (type 1) (TMSR-LF1) program of China is being launched in the Shanghai Institute of Applied Physics, Chinese Academy of Sciences (SINAP@CAS), which also uses a zirconium-containing system composed of LiF–BeF₂–ZrF₄–UF₄ (65.1 : 28.7 : 5 : 1.2 mol%) as the fuel salt.

In this paper, the dissolution–precipitation behaviors of ZrO₂ and the interactions between Zr(IV) and O²⁻ in the FLiBe melt are first studied. Then the apparent solubility product (K′_sp) of ZrO₂ is obtained and compared with that of UO₂ to predict the generation of oxide precipitations in the fuel salt containing both UF₄ and ZrF₄. The oxide titration experiment is performed in the FLiBe–ZrF₄ (5 mol%)–UF₄ (1.2 mol%) system, and the evolutions for the equilibrium concentrations of U(IV), Zr(IV), and O²⁻ with oxide additions are studied to further confirm above theoretical prediction. Finally, the inhibiting effect of ZrF₄ on UO₂ formation is further verified. This study provides an effective solution for controlling and monitoring the nuclear fuel precipitation (UO₂) in molten fluorides, which is of great importance for the safe operation of MSR.

2. Experimental

2.1 Molten salt

The highly purified FLiBe (99.9%) eutectic salt with eutectic temperature of 733 K was prepared and supplied by the SINAP@CAS. The received FLiBe eutectic salt was stored in a glovebox and further dehydrated before use by heating under vacuum from ambient temperature to its melting point for 72 h, to guarantee the high-purity of salt and ensure the reliability of the experiments. During this process, the moisture absorbed by the salt was effectively removed. In order to avoid residual HF remaining in FLiBe, the melt was further purified by sparging the ultra-pure H₂ for 8 h. The FLiBe–ZrF₄ and FLiBe–ZrF₄–UF₄ melts were thus prepared by artificially adding ZrF₄ (99.99%; Sigma-Aldrich) and UF₄ (99.99%; China North Nuclear Fuel Co., Ltd) to the prepared FLiBe salt.

2.2 Experimental method

During the dissolution experiments, the excess of powdery ZrO₂ (ca. 5 g, 99.99%; Sigma-Aldrich) was manufactured into sheets by tableting. The ZrO₂ sheet was dissolved in 50 g FLiBe melt contained in a vitreous carbon crucible, and the saturated solution of FLiBe–ZrO₂ was formed after equilibrium time. The ZrO₂ sheet was immersed in the salt bath with a graphite pellet holder, as shown in Fig. 1(a). The melt was sampled after the extraction of the pellet from the bath. Thus, no ZrO₂ compound would stick to the sampler, and all the zirconium in the sample was considered to be obtained from ZrO₂ dissolution and could reflect the solubility of ZrO₂.

Fig. 1 Schematics of the (a) dissolution and the (b) oxide titration experiments.

During the oxide titration experiments, the oxide ions (O²⁻) were introduced to the FLiBe–ZrF₄ and FLiBe–ZrF₄–UF₄ baths in the form of lithium oxide (Li₂O, 99.99%; Sigma-Aldrich), as shown in Fig. 1(b). The interaction between M(IV) (M = Zr or U) and O²⁻ in the melt would produce ZrO₂ (ρ = 5.89 g cm⁻³) or UO₂ (ρ = 10.96 g cm⁻³) precipitate and gradually deposit to the bottom of the bath (ρ ≈ 2 g cm⁻³). When the system reached equilibrium, the supernatant of the molten salt was sampled. The sampling process addressed the use of sintered nickel filter, which prevented the precipitates from sticking to the sampler as proposed by the ORNL,²³ to ensure the accuracy of the experimental results.

The schematic of the experimental apparatus used in present study was shown in Fig. 2, which was described in detail in our previous studies.^{7,8,16,17,20,21,24} The whole experiment was conducted in a glove box protected by dry argon atmosphere (99.999%), and the oxygen and water contents in the glove box were both strictly controlled below 0.5 ppm.

Fig. 2 Schematic of the experimental set-up.

2.3 Analytical techniques

The oxide concentration in molten fluorides was determined using the LECO oxygen detector (RO600, LECO Co., Ltd) through the carbothermal reduction technique, which was in accordance with the work of Mediaas^25,26 and our previous studies.^7,8,16,20 Herein, the result measured by LECO was the total oxide content including O²⁻ and oxide in oxygenated anions. The oxygenated anions in the melts were determined using ICE ion chromatography (IC, ICS-2100, Dionex Co., Ltd.) with a valve-switching technology.⁸

The concentration of Zr(IV) or U(IV) ions in molten fluorides was determined using inductively coupled plasma optical emission spectroscopy (ICP-OES, Arcos, Spectro Co., Ltd). The detailed testing process has been described previously.^7,8,17 The oxide precipitates formed in molten salt were analyzed using the XRD technique (DY3614, PANzlytical Co., Ltd).

3. Results

3.1 Concentration of free oxide ions (O²⁻) in molten fluorides

The oxygen in fluoride salt exists in two forms: (i) free oxide ions (O²⁻) and (ii) oxygenated anions (e.g., SO₄²⁻, NO₃⁻, and PO₄³⁻). However, only O²⁻ could form oxide precipitates (e.g., UO₂ and ZrO₂) due to its high sensitivity toward UF₄ and ZrF₄ in molten fluorides. Therefore, the O²⁻ concentration in molten fluorides should be measured to clarify the dissolution–precipitation behavior of the fuel salt for MSR. The practical O²⁻ concentration ([O²⁻]) in molten fluorides was determined by subtracting the oxide concentration in oxygenated anions ([O]_IC) from the total oxide concentration ([O]_LECO), as shown in eqn (4). Table 1 gives the oxide analysis results of two batches of FLiBe melts used in the present study. As shown in Table 1, the total oxygen content of 1# FLiBe was 145 ppm, and the oxygenated anions included NO₃⁻, SO₄²⁻, and PO₄³⁻ with content of 83, 55, and 6 ppm, respectively. According to the mass proportion of oxide in the oxygenated anion, the oxide concentrations in NO₃⁻, SO₄²⁻, and PO₄³⁻ were 64, 37, and 4 ppm, respectively. Thus, the sum of oxide concentration in these three oxygenated anions were up to 105 ppm, and the O²⁻ content was 40 ppm in accordance with eqn (4). Likewise, the O²⁻ content of 2# FLiBe was 1383 ppm.

[O²⁻] = [O]_LECO − [O]_IC. (4)

Table 1 Oxide analysis results of the FLiBe melts used in the present study

Batch number of FLiBe	Total oxide concentration ([O]LECO)/ppm	Oxygenated anion concentration/ppm		Oxide in oxygenated anions ([O]IC)/ppm	O^2− concentration ([O^2−])/ppm
1# (for oxide titration experiment)	145	NO3^−	83	105	40 (2.5 × 10^−3 mol kg^−1)
		SO4^2−	55
		PO4^3−	6
2# (for dissolution experiment)	1500	NO3^−	135	117	1383 (8.64 × 10^−2 mol kg^−1)
		SO4^2−	19
		PO4^3−	—

---

3.2 K′_sp of ZrO₂

3.2.1 Dissolution method: FLiBe–ZrO₂ system. The ZrO₂ pellet was immersed into the FLiBe melt to determine the solubility of ZrO₂ in this melt. The melt was sampled at different holding times at 873 K, and the Zr(IV) concentration in each sample was measured. Fig. 3 gives the plots of Zr(IV) concentration in the melt versus the dissolution time of ZrO₂. Apparently, the initial zirconium concentration in the FLiBe melt was below the detection limit of ICP-OES (<0.01 ppm or 1 × 10⁻⁷ mol kg⁻¹) and could not be detected. When the ZrO₂ was dissolved for the first 2 h, the Zr(IV) concentration remarkably increased to 3.37 × 10⁻³ mol kg⁻¹. Afterwards, the growth rate of Zr(IV) concentration slowed down, and its value gradually increased to 3.84 × 10⁻³ mol kg⁻¹ when the dissolution time further increased to 6 h. Then, no remarkable difference was observed in the results of samples collected from 6 h to 22 h, and the Zr(IV) concentration finally maintained at approximately 3.84 × 10⁻³ mol kg⁻¹, which corresponded to the solubility of ZrO₂ with an equilibrium time of 6 h.

Fig. 3 Zr(IV) concentration versus the dissolution time of ZrO₂ in the FLiBe melt at 873 K.

In the FLiBe–ZrO₂ system, when the dissolution equilibrium of ZrO₂ was achieved, the substances of ZrO₂ compound, Zr(IV), and O²⁻ ions, coexisted and equilibrated, as shown in eqn (5). Thus, the real solubility product (K_sp) of ZrO₂ could be obtained by multiplying the equilibrium activity of Zr(IV) (a_Zr(IV)) and O²⁻ (a_O²⁻), as shown in eqn (6). Actually, a certain deviation existed between the activity (a) and the apparent concentration, and this deviation could be quantitatively expressed using the activity coefficient (γ) with the value less than 1, as shown in eqn (7) and (8). In fact, the [Zr(IV)] and [O²⁻] values obtained by ICP-OES, LECO, and IC techniques were the results of apparent concentrations. Thus, the real solubility product (K_sp) of ZrO₂ could be calculated through eqn (6), where K′_sp is the apparent solubility product that can be obtained using experimental methods, as shown in eqn (9). However, the measurement of a and γ (γ₁ for Zr(IV) and γ₂ for O²⁻, which can be considered as constant in this case) in molten fluorides is a challenging task and leads to a difficult determination of the real K_sp. Nevertheless, the present study only considered the apparent concentration to deal with this issue, and the K′_sp calculation could also make sense in case of apparent concentration consideration (eqn (9)).

ZrO₂(s) ↔ Zr(IV) + 2O²⁻. (5)

K_sp = a_Zr(IV)·(a_O²⁻)² = γ₁γ₂²·[Zr(IV)][O²⁻]² = γ₁γ₂²·K′_sp. (6)

a_Zr(IV) = γ₁·[Zr(IV)]. (7)

a_O²⁻ = γ₂·[O²⁻]. (8)

K′_sp = [Zr(IV)][O²⁻]². (9)

[O²⁻] = [O²⁻]_ZrO2 + [O²⁻]_FLiBe. (10)

[O²⁻]_ZrO2 = 2[Zr(IV)]. (11)

Then, the equilibrium [Zr(IV)] and [O²⁻] were measured to determine the value of K′_sp. In accordance with the solubility of ZrO₂ in FLiBe, as shown in Fig. 3, the equilibrium [Zr(IV)] in the bath was 3.84 × 10⁻³ mol kg⁻¹, whereas the O²⁻ concentration caused by ZrO₂ dissolution ([O²⁻]_ZrO2) was twice of the [Zr(IV)], i.e., 7.68 × 10⁻³ mol kg⁻¹ (eqn (5) and (11)). Besides, the initial O²⁻ content in the FLiBe melt ([O²⁻]_FLiBe) was 8.64 × 10⁻² mol kg⁻¹ (1383 ppm, Table 1). Thus, the equilibrium [O²⁻] was 9.41 × 10⁻² mol kg⁻¹, which was obtained using the sum of [O²⁻]_ZrO2 and [O²⁻]_FLiBe (eqn (10)). Then, with the obtained values of [Zr(IV)] and [O²⁻], the K′_sp was calculated to be 3.40 × 10⁻⁵ mol³ kg⁻³ (eqn (9)).

3.2.2 Oxide titration method: FLiBe–ZrF₄–Li₂O system. Oxide ions in the form of Li₂O were gradually added to the FLiBe–ZrF₄ (5 mol%) melt to investigate the interaction between Zr(IV) and O²⁻ and its corresponding product. Then, the supernatant melt was sampled after the system reached equilibrium at 873 K, and the Zr(IV) concentration in each sample was analyzed. Table 2 shows the variation of Zr(IV) concentration with different oxide additions at 873 K. Apparently, the residual Zr(IV) concentration in the melt linearly decreased with increasing oxide addition, and the slope of this fitted straight line was 0.51, which approximately equaled to 0.5, as shown in Fig. 4. This result meant a zirconium-containing precipitate with the stoichiometry of 1Zr⁴⁺:2O²⁻, probably ZrO₂ or Zr_xO_2xF_y species, was formed. Moreover, a kind of white precipitate at the bottom of the melting bath was observed during oxide titration, as shown in Fig. 5(a). To identify the precipitated species, the bottom precipitate was sampled, distilled (to remove the residual matrix salt) and further transferred for XRD analysis. The result confirmed that this precipitate corresponded to ZrO₂, as shown in Fig. 5(b).

Table 2 Residual Zr(IV) and O²⁻ concentrations in the FLiBe–ZrF₄ (5 mol%) melt with different oxide additions, and the variation of calculated K′_sp (ZrO₂) value (equilibrium time: 10 h; temperature: 873 K)

Added O^2− (mol kg^−1)	Equilibrium [Zr(IV)] in melt (mol kg^−1)	Equilibrium [O^2−] in melt (mol kg^−1)	K′sp of ZrO2 (mol^3 kg^−3)
0	1.33	—	—
0.11	1.29	5.20 × 10^−3 (83 ppm)	3.49 × 10^−5
0.16	1.27	5.29 × 10^−3 (85 ppm)	3.55 × 10^−5
0.26	1.23	5.43 × 10^−3 (87 ppm)	3.63 × 10^−5
0.34	1.18	5.48 × 10^−3 (88 ppm)	3.54 × 10^−5
0.47	1.08	5.71 × 10^−3 (91 ppm)	3.52 × 10^−5

---

Fig. 4 Linear relationship between the residual [Zr(IV)] and the added [O²⁻] in the FLiBe–ZrF₄ (5 mol%) melt at 873 K.

Fig. 5 (a) The observed white precipitate formed at the bottom of the FLiBe–ZrF₄ (5 mol%) melt with O²⁻ addition at 873 K. (b) XRD pattern of this precipitate obtained after distillation treatment.

With the formation of ZrO₂, its K′_sp could be calculated by multiplying the equilibrium concentration of O²⁻ by Zr(IV) (eqn (9)). As shown in Table 2, when the added O²⁻ increased from 0 mol kg⁻¹ to 0.47 mol kg⁻¹, the remaining Zr(IV) in the melt gradually decreased from 1.33 mol kg⁻¹ to 1.08 mol kg⁻¹ whereas the equilibrium O²⁻ concentration increased from 5.20 × 10⁻³ mol kg⁻¹ to 5.71 × 10⁻³ mol kg⁻¹. Thus, a set of K′_sp values were obtained respectively during each oxide addition, and the results located within (3.49–3.63) × 10⁻⁵ mol³ kg⁻³. The slight variation in the K′_sp may be attributed to the testing error caused by the chemical analysis. The eqn (9) could be written as:

The linear relationship between the residual [Zr(IV)] and 1/[O²⁻]² was fitted (Fig. 6), and the slope of this straight line corresponded to the statistical result of K′_sp. Thus, the final result of K′_sp was determined to be 3.54 × 10⁻⁵ mol³ kg⁻³. This value was highly consistent with that obtained by the dissolution method in FLiBe–ZrO₂ system (3.40 × 10⁻⁵ mol³ kg⁻³), as stated in Sect. 3.2.1.

Fig. 6 Determination of K′_sp (ZrO₂) in the FLiBe–ZrF₄ (5 mol%) melt with O²⁻ additions at 873 K: linear relationship between the residual [Zr(IV)] and 1/[O²⁻]².

3.3 Oxide titration to the molten FLiBe–ZrF₄–UF₄ system

Oxide ions in the form of Li₂O were gradually added to the FLiBe–ZrF₄ (5 mol%)–UF₄ (1.2 mol%) melt to investigate the inhibiting effect of ZrF₄ on UO₂ formation. Then, the evolutions of the equilibrium concentrations of Zr(IV), U(IV), and O²⁻ in the melt with oxide addition were determined, and the results could be shown in two stages, as shown in Fig. 7.

Fig. 7 Evolutions of the equilibrium [Zr(IV)], [U(IV)], and [O²⁻] in the FLiBe–ZrF₄ (5 mol%)–UF₄ (1.2 mol%) melt with oxide additions at 873 K.

3.3.1 Stage 1: ZrO₂ formation (O²⁻ addition <1 mol kg⁻¹). The initial Zr(IV) and U(IV) contents of the molten salt were 5 mol% (1.21 mol kg⁻¹) and 1.2 mol% (0.28 mol kg⁻¹), respectively. Before the added O²⁻ reached 1 mol kg⁻¹, the residual Zr(IV) concentration in melt linearly decreased to 0.72 mol kg⁻¹ and the U(IV) concentration almost remained invariable, whereas the equilibrium O²⁻ increased from 90 ppm to 110 ppm. This result indicated that only zirconium-containing precipitate was formed in this period. The slope of the linear relationship between the residual [Zr(IV)] and the added [O²⁻] (0–1 mol kg⁻¹) was fitted as 0.49, which approximately equaled to 0.5 (Fig. 7), presuming that insoluble ZrO₂ or Zr_xO_2xF_y species was formed in the melt. Moreover, a kind of white precipitate was observed at the bottom of the bath during experiment, and XRD analysis further proved that this precipitate corresponded to ZrO₂ (Fig. 8). Therefore, only ZrO₂ precipitate was formed with oxide addition less than 1 mol kg⁻¹.

Fig. 8 XRD patterns of the bottom samples obtained in the FLiBe–ZrF₄ (5 mol%)–UF₄ (1.2 mol%) melt with oxide additions after distillation treatment. Stage 1: oxide addition <1 mol kg⁻¹; stage 2: oxide addition >1 mol kg⁻¹.

3.3.2 Stage 2: ZrO₂ and UO₂ formations (O²⁻ addition ≥1 mol kg⁻¹). When the added O²⁻ reached 1 mol kg⁻¹, the U(IV) concentration drop began, and the equilibrium O²⁻ at this time was 110 ppm. When the added O²⁻ exceeded 1 mol kg⁻¹ and reached 2.72 mol kg⁻¹, the Zr(IV) and U(IV) in melt were simultaneously consumed and finally decreased to 0.07 and 0.09 mol kg⁻¹, respectively. Meanwhile the equilibrium O²⁻ further increase to 208 ppm. These results indicated that both zirconium-containing and uranium-containing precipitates were formed in this stage. In addition, the stoichiometry of the total reacted metal ions M(IV) (M = Zr or U) and the added O²⁻ was approximately 1 : 2 during this stage. After completion of the experiment, two kinds of precipitates were observed at the bottom of salt. The white one corresponded to ZrO₂ and the rufous one corresponded to UO₂, as confirmed by XRD (Fig. 8). Therefore, both ZrO₂ and UO₂ were formed with oxide addition more than 1 mol kg⁻¹.

4. Discussion

4.1 K′_sp: analysis of oxide precipitation behavior in the FLiBe melt

The K′_sp of ZrO₂ and UO₂ in FLiBe melt were compared to clarify the precipitation law of these two oxides. According to our previous investigation,⁸ the K′_sp (UO₂) in the FLiBe melt obtained using the dissolution and the oxide titration methods at 873 K were 1.67 × 10⁻⁵ and 1.33 × 10⁻⁵ mol³ kg⁻³, respectively. In this work, the K′_sp (ZrO₂) using the dissolution and the oxide titration methods at 873 K were 3.40 × 10⁻⁵ and 3.54 × 10⁻⁵ mol³ kg⁻³, respectively. As an equilibrium constant, the K_sp theoretically depended on temperature only, and the present results also agreed well with those reported by ORNL at the same temperature,²² as summarized in Table 3.

Table 3 Comparison between K′_sp (ZrO₂) and K′_sp (UO₂) and their ratio K′_sp (ZrO₂)/K′_sp (UO₂) (k) in the FLiBe melt at 873 K

	K′sp by dissolution method (mol^3 kg^−3)	K′sp by oxide titration method (mol^3 kg^−3)	Ksp reported by ORNL^22 (mol^3 kg^−3)
ZrO2	3.40 × 10^−5	3.54 × 10^−5	3.20 × 10^−5
UO2	1.67 × 10^−5 (ref. 8)	1.33 × 10^−5 (ref. 8)	1.20 × 10^−5
k	2.04	2.66	2.67

---

Evidently, K′_sp (ZrO₂) and K′_sp (UO₂) were in the same order of 10⁻⁵ mol³ kg⁻³, and K′_sp (ZrO₂) was 2–3 times higher than K′_sp (UO₂). Based on these results, we can predict the formation of ZrO₂ and UO₂ precipitates in the FLiBe melt. (i) When the FLiBe melt only contained UF₄, the reactive oxide resulted in UO₂ precipitation. (ii) When the FLiBe melt contained equimolar ZrF₄ and UF₄, the oxide contamination should first yield the UO₂ precipitate. (iii) In addition, when the FLiBe melt contained considerably more ZrF₄ than UF₄, the oxide contamination should first yield the ZrO₂ precipitate. The precipitation of UO₂ began only after the precipitation of ZrO₂ had dropped the ZrF₄ concentration to near 2–3 times as high as that of UF₄. With further oxide addition, UO₂ and ZrO₂ would precipitate simultaneously. Therefore, excessive ZrF₄ with more than 3 times higher than UF₄ can effectively avoid the formation of UO₂ precipitate in the FLiBe melt.

4.2 Oxide precipitation behavior of the MSR fuel salt system FLiBe–ZrF₄–UF₄

The oxide titration experiment in the FLiBe–ZrF₄–UF₄ melt with considerably more ZrF₄ than UF₄ was conducted to validate the above theoretical prediction and further clarify the influence of ZrF₄ on UO₂ generation, as shown in Fig. 7. At first, the ZrF₄ in the system was consumed by forming ZrO₂, thereby protecting uranium from being precipitated. Until the ZrF₄ concentration dropped to 0.72 mol kg⁻¹ (3 mol%), precipitation of UO₂ began to form. At this point, the Zr(IV) concentration was about 2.5 times as high as that of U(IV). This result was highly consistent with the theoretical prediction, as stated in Sect. 4.1.

With the measured equilibrium concentrations of M(IV) (M = Zr or U) and O²⁻, a set of K′_sp values for ZrO₂ and UO₂ could be respectively obtained after each oxide addition, as shown in Table 4. The K′_sp results were basically consistent with that obtained in the FLiBe melt (Table 3) but varied with O²⁻ addition. This was caused by the concentration change of M(IV) in the melt. The K′_sp (ZrO₂) showed no great difference when the added O²⁻ was less than 1.71 mol kg⁻¹, but decreased from 3.45 × 10⁻⁵ mol³ kg⁻³ to 1.19 × 10⁻⁵ mol³ kg⁻³ when the added O²⁻ further increased from 1.71 mol kg⁻¹ to 2.72 mol kg⁻¹. At this time, the concentration of Zr(IV) was below 0.39 mol kg⁻¹. Whereas the K′_sp (UO₂) had no remarkable change except some slight fluctuations. The variation of K′_sp (ZrO₂) probably arose from the deviation between the apparent concentration (c) and the activity (a) for Zr(IV) in the melt, as shown in eqn (7). The extent of this deviation (γ ≤ 1) was correlated to the concentration of ions. γ increased and decreased when the solute concentration decreased and increased, respectively. In this case, the initial Zr(IV) concentration was about four times greater than the initial U(IV) concentration, and the decrement in [Zr(IV)] varied in a wide range during oxide titration. Thus, when the residual Zr(IV) concentration continued decreasing to a low extent (the added O²⁻ was more than 1.71 mol kg⁻¹ and [Zr(IV)] was below 0.39 mol kg⁻¹, which was 1/3 of the initial one), the γ₁ for Zr(IV) would increase. Meanwhile, [O²⁻] changed slightly (only about 100 ppm change), indicating that γ₂ for O²⁻ remained virtually unchanged. According to eqn (6) and (9), the calculated K′_sp (ZrO₂) gradually decreased under low [Zr(IV)] when the added O²⁻ was more than 1.71 mol kg⁻¹ and [Zr(IV)] was below 0.39 mol kg⁻¹. With regard to U(IV), whose concentration still stayed at a relatively low level, and the decrement occurred on a smaller scale compared with that of Zr(IV). Herein, the γ_U(IV) might change slightly, so did the γ₂ for O²⁻. Thus, no remarkable variation in K′_sp (UO₂) was observed in the whole oxide titration procedure.

Table 4 Variation in K′_sp for ZrO₂ and UO₂ in the FLiBe–ZrF₄ (5 mol%)–UF₄ (1.2 mol%) melt with oxide addition at 873 K

Added [O^2−] (mol kg^−1)	[O^2−] (mol kg^−1)	[Zr(IV)] (mol kg^−1)	[U(IV)] (mol kg^−1)	K′sp (ZrO2) (mol^3 kg^−3)	K′sp (UO2) (mol^3 kg^−3)
0.15	5.59 × 10^−3 (89 ppm)	1.14	0.29	3.57 × 10^−5	— (no UO2 formed)
0.36	5.79 × 10^−3 (93 ppm)	1.06	0.28	3.56 × 10^−5	— (no UO2 formed)
0.54	6.01 × 10^−3 (96 ppm)	0.96	0.29	3.47 × 10^−5	— (no UO2 formed)
0.73	6.35 × 10^−3 (102 ppm)	0.85	0.29	3.43 × 10^−5	— (no UO2 formed)
0.84	6.53 × 10^−3 (105 ppm)	0.81	0.28	3.46 × 10^−5	— (no UO2 formed)
1.01	6.80 × 10^−3 (109 ppm)	0.72	0.28	3.33 × 10^−5	1.31 × 10^−5
1.14	7.03 × 10^−3 (113 ppm)	0.64	0.26	3.16 × 10^−5	1.30 × 10^−5
1.28	7.44 × 10^−3 (119 ppm)	0.55	0.25	3.07 × 10^−5	1.40 × 10^−5
1.39	8.10 × 10^−3 (130 ppm)	0.51	0.23	3.36 × 10^−5	1.50 × 10^−5
1.50	8.73 × 10^−3 (140 ppm)	0.46	0.22	3.53 × 10^−5	1.66 × 10^−5
1.71	9.40 × 10^−3 (150 ppm)	0.39	0.19	3.45 × 10^−5	1.68 × 10^−5
1.99	1.01 × 10^−2 (161 ppm)	0.28	0.16	2.81 × 10^−5	1.61 × 10^−5
2.23	1.08 × 10^−2 (173 ppm)	0.21	0.13	2.47 × 10^−5	1.56 × 10^−5
2.51	1.17 × 10^−2 (186 ppm)	0.15	0.10	1.99 × 10^−5	1.36 × 10^−5
2.72	1.30 × 10^−2 (208 ppm)	0.07	0.09	1.19 × 10^−5	1.50 × 10^−5

---

Fig. 9 illustrates the evolution of K′_sp (ZrO₂)/K′_sp (UO₂) ratio (k) in the oxide titration duration. The k value was also equal to the ratio of [Zr(IV)]/[U(IV)] because the O²⁻ was simultaneously equilibrated by ZrO₂ and UO₂, as shown in eqn (13). Apparently, K′_sp (ZrO₂) was 2.54–1.46 times greater than K′_sp (UO₂) when the added O²⁻ was in the range of 1.01–2.51 mol kg⁻¹. However, the difference between them narrowed (decreasing k) with oxide addition, and K′_sp (ZrO₂) < K′_sp (UO₂) happened when the oxide addition reached 2.72 mol kg⁻¹ (Fig. 9, bottom X-axis). Meanwhile, the decrease in [Zr(IV)]/[U(IV)] ratio with decreasing uranium concentration was also found (Fig. 9, top X-axis). These results reveal two important information: (i) proposing a solution for monitoring the generation of UO₂ by measuring the k value in the melt with known [U(IV)]. For example, in this case the FLiBe–ZrF₄ (5 mol%)–UF₄ (1.2 mol%) melt, as long as k ＞ 2.5, namely the Zr(IV) concentration remains more than 2.5 times that of U(IV), the oxide contamination will not cause UO₂ precipitation; (ii) when [U(IV)] is different, the required [Zr(IV)] (actually, the ratio of [Zr(IV)]/[U(IV)], k) which can prevent UO₂ precipitation also changes. The lower the [U(IV)] is, the lower the required k value will be. For example, when [U(IV)] is 0.29 mol kg⁻¹, k ≥ 2.5 is needed to prevent UO₂ precipitation. However, when [U(IV)] is 0.09 mol kg⁻¹, it only requires k ≥ 0.8. That means the required [Zr(IV)] decreased with decreasing [U(IV)]. These results provide reference data for fuel salt design of MSR with different uranium content.

Fig. 9 K′_sp ratio of ZrO₂ to UO₂ (k) during the oxide titration of FLiBe–ZrF₄ (5 mol%)–UF₄ (1.2 mol%) melt at 873 K. k = 2.66 obtained by oxide titration method; k = 2.04 obtained by dissolution method; k = 1 means K′_sp (ZrO₂) = K′_sp (UO₂).

5. Conclusion

The formation of the UO₂ precipitate would lead to a superheated area in MSR and cause a critical risk. Developing appropriate additives for oxide ion capture is an effective way to protect uranium from forming the UO₂ precipitate in the fuel of MSR. This study focused on the dissolution–precipitation behavior of ZrO₂ in the FLiBe melt, and the inhibition of UO₂ generation by the ZrF₄ additive was also clarified.

First, the results of the dissolution experiment showed that the saturated solubility of ZrO₂ in the FLiBe melt at 873 K was 3.84 mol kg⁻¹ with the equilibrium time of 6 h. Then, the interactions between Zr(IV) and oxide ions was studied by adding O²⁻ into the FLiBe–ZrF₄ melt. The Zr(IV) concentration in the melt linearly decreased with O²⁻ addition because of ZrO₂ precipitation, which was confirmed by the stoichiometric ratio (chemical analysis) and XRD. Moreover, the K′_sp of ZrO₂ were determined to be 3.40 × 10⁻⁵ and 3.54 × 10⁻⁵ mol³ kg⁻³ in these two processes, which showed good conformity and were highly consistent with that reported by ORNL.

The obtained K′_sp of ZrO₂ was in the same order of 10⁻⁵ mol³ kg⁻³ and 2–3 times higher than that of UO₂. Based on this result, the formation of ZrO₂ and UO₂ precipitates in the FLiBe melt can be predicted. If the FLiBe melt contained considerably more ZrF₄ than UF₄, the oxide contamination first yielded the ZrO₂ precipitate, and the precipitation of UO₂ began only when the precipitation of ZrO₂ had dropped the ZrF₄ concentration to near 2–3 times as high as that of UF₄. The oxide titration experiment of the FLiBe–ZrF₄ (5 mol%)–UF₄ (1.2 mol%) melt was further carried out, showing that only ZrO₂ was formed when the O²⁻ addition was less than 1 mol kg⁻¹. When the added O²⁻ reached 1 mol kg⁻¹ and the [Zr(IV)] reduced to 0.72 mol kg⁻¹ (3 mol%), UO₂ began to form and the [Zr(IV)]/[U(IV)] ratio was 2.5 at this point. Afterwards, ZrO₂ and UO₂ coprecipitated with further O²⁻ addition. Experimental results agreed well with the theoretical prediction from the obtained K′_sp (ZrO₂) and K′_sp (UO₂).

The results of present work clarified the inhibiting effect of ZrF₄ on the UO₂ generation in the FLiBe melt and further affirmed the approach of ZrF₄ additive for the control of nuclear fuel precipitation (UO₂) in MSR. When oxide contamination occurred, the considerable amount of ZrF₄ first formed ZrO₂, which prevented the combination between UF₄ and O²⁻. Thus, the precipitation of UO₂ could be effectively avoided. The critical ratio of [Zr(IV)]/[U(IV)] to prevent the precipitation of UO₂ was more than 2.5 in the FLiBe–ZrF₄ (5 mol%)–UF₄ (1.2 mol%) melt, and it decreased with decreasing initial uranium concentration. The precipitation mechanism of ZrO₂ and UO₂ in the FLiBe molten salt can provide the following inspirations: (i) a solution for monitoring the generation of UO₂ precipitation in the MSR fuel salt can be proposed by measuring the [Zr(IV)]; (ii) the required [Zr(IV)]/[U(IV)] ratio which can prevent UO₂ precipitation changes with different uranium content. All of these will be helpful for the safe operation and fuel salt design of MSR.

Conflicts of interest

There are no conflicts to declare.

Acknowledgements

This work was supported by the National Natural Science Foundation of China (Grant No. 22006150), Shanghai Sailing Program (Grant No. 19YF1458200), “Strategic Priority Research Program” and the “Frontier Science Key Program” of the Chinese Academy of Sciences (Grant No. XDA02030000 and QYZDYSSW-JSC016).

References

1. M. Rosenthal, P. Kasten and R. Briggs, Molten-salt reactorsd history, status, and potential, Nucl. Appl. Technol., 1970, 8, 107–117.
2. W. R. Grimes, F. F. Blankenship, G. W. Keilholtz, H. F. Poppendiek and M. T. Robinson, Chemical aspects of molten fluoride reactors, in Progress in Nuclear Energy, Series IV, Pergamon Press, London, 1960, vol. 2.
3. A. Nuttin, D. Heuer, A. Billebaud, R. Brissot, C. Le Brun, E. Liatard, J.-M. Loiseaux, L. Mathieu, O. Meplan and E. Merle-Lucotte, Potential of thorium molten salt reactors: detailed calculations and concept evolution with a view to large scale energy production, Prog. Nucl. Energy, 2005, 46, 77–99.
4. F. Lantelme and H. Groult, Molten Salts Chemistry: from Lab to Applications, Elsevier, 2013.
5. S. Delpech, C. Cabet, C. Slim and G. S. Picard, Molten fluorides for nuclear applications, Mater. Today, 2010, 13, 34–41.
6. P. A. Haas and P. Cooper, Solubility of uranium oxides in fluoride salts at 1200 °C, J. Chem. Eng. Data, 1993, 38, 26–30.
7. H. Peng, M. Shen, Y. Zuo, X. X. Tang, R. Tang and L. D. Xie, Electrochemical technique for detecting the formation of uranium-containing precipitates in molten fluorides, Electrochim. Acta, 2016, 222, 1528–1537.
8. H. Peng, W. Huang, L. D. Xie and Q. N. Li, Solubility and precipitation investigations of UO₂ in LiF–BeF₂ molten salt, J. Nucl. Mater., 2020, 531, 152004.
9. ORNL-3812, Molten-salt Reactor Program: Semiannual Progress Report for Period Ending, Oak Ridge National Laboratory, 1965.
10. ORNL-3936, Molten-salt Reactor Program: Semiannual Progress Report for Period Ending, Oak Ridge National Laboratory, 1966.
11. ORNL-4812, Molten-salt Reactor Program: the Development Status of Molten-Salt Breeder Reactors, Oak Ridge National Laboratory, 1972.
12. ORNL-3708, Molten-salt Reactor Program: Semiannual Progress Report for Period Ending, Oak Ridge National Laboratory, 1964.
13. ORNL-4616, Reactor Chemistry Division: Preparation and Handling of Salt Mixtures for the Molten Salt Reactor experiment, Oak Ridge National Laboratory, 1971.
14. A. L. Mathews and C. F. Baes Jr., Oxide chemistry and thermodynamics of molten lithium fluoride-beryllium fluoride solutions, Inorg. Chem., 1968, 7, 373–382.
15. ORNL-TM-0728, report, MSRE Design and Operations Report: Part I. Description of Reactor Design, Oak Ridge National Laboratory, 1965.
16. M. Shen, H. Peng, M. Ge, C. Y. Wang, Y. Zuo and L. D. Xie, Chemical interactions between zirconium and free oxide in molten fluorides, RSC Adv., 2015, 5, 40708–40713.
17. H. Peng, M. Shen, Y. Zuo, H. Y. Fu and L. D. Xie, Chemical and electrochemical studies on the solubility of UO₂ in molten FLINAK with ZrF₄ additive, J. Nucl. Mater., 2018, 510, 256–264.
18. M. Gibilaro, L. Massot, P. Chamelot, L. Cassayre and P. Taxil, Investigation of Zr(IV) in LiF–CaF₂: stability with oxide ions and electroreduction pathway on inert and reactive electrodes, Electrochim. Acta, 2013, 95, 185–191.
19. M. Korenko, M. Straka, J. Uhlíř, L. Szatmáry, M. Ambrová and M. Šimurda, Phase analysis of the solidified KF–(LiF–NaF–UF₄)–ZrF₄ molten electrolytes for the electrowinning of uranium, J. Radioanal. Nucl. Chem., 2014, 302, 549–554.
20. M. Shen, H. Peng, M. Ge, Y. Zuo and L. D. Xie, Use of square wave voltammeter for online monitoring of O²⁻ concentration in molten fluorides at 600 °C, J. Electroanal. Chem., 2015, 748, 34–39.
21. Y. L. Song, M. Shen, H. Peng, C. Y. Wang, S. F. Zhao, Y. Zuo and L. D. Xie, Solubility of Cr₂O₃ in Molten Fluorides with Different ZrF₄ Contents and Fluoroacidities, J. Electrochem. Soc., 2020, 167, 023501.
22. ORNL-3419, Molten-salt Reactor Program: Semiannual Progress Report for Period Ending, Oak Ridge National Laboratory, 1963.
23. ORNL-1771, Aircraft Nuclear Propulsion Project: Quarterly Progress Report for Period Ending, Oak Ridge National Laboratory, 1954.
24. H. Peng, M. Shen, C. Y. Wang, T. Su, Y. Zuo and L. D. Xie, Electrochemical investigation of the stable chromium species in molten FLINAK, RSC Adv., 2015, 5, 76689–76695.
25. H. Mediaas, J. Vinstad and T. Østvold, Solubility of MgO in MgCl₂–NaCl–NaF Melts, Minerals, Metals and Materials Society, Warrendale, PA (United States), 1996.
26. 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.

---