Figure 1 presents the electrochemical set up used in this work, composed of a vitreous carbon crucible (3) placed in a boron nitride secondary crucible (2). The two crucibles were placed in a cylindrical Inconel reactor (1) closed by a lid (4) cooled by air circulation. The experiments were carried out under an inert argon atmosphere (<0.1 ppm O₂ and <0.5 ppm H₂O). The cell was heated using a programmable furnace. The temperature was controlled by a platinum-rhodium thermocouple. This equipment was placed in a glovebox under dry argon atmosphere (<5 ppm O₂ and <2 ppm H₂O).

The electrolytic bath was composed of LiF–NaF (Sigma Aldrich 99.99%) eutectic mixture (60 - 40 mol%). This mixture was initially dehydrated by heating under vacuum (7 × 10⁻⁵ bar) from ambient temperature up to its melting point (650˚C) for 70 h. Zirconium ions were added to the bath in the form of zirconium (IV) fluoride ZrF₄ (Alfa Aesar 99.9%).

Zirconium, chromium, iron, niobium, tin, cobalt, nickel wires (1 mm diameter) and Zircaloy-4 rod (3 mm diameter) were used as working electrodes. The surface area of the working electrode was determined by measuring the immersion depth in the salt after each experiment. The counter electrode was a zirconium plate for electrodeposition investigation and a vitreous carbon (3000C) rod (3 mm diameter) for electrorefining process. The potentials were referred to a platinum wire (1 mm diameter) acting as a quasi-reference electrode Pt/PtO_x/O²⁻ [17] .

Transient electrochemical techniques used in this study are cyclic voltammetry, square wave voltammetry and chronoamperometry. Investigations were carried out with a VSP potentiostat/galvanostat controlled with EC-Lab V11.01 software.

The electrode surfaces and cross-sections were characterized by a Scanning Electron Microscope (Phenom XL) coupled with an Energy Dispersive Spectroscopy probe.

ICP-AES (Horiba Ultima 2) was used to analyze the impurities content in the bath after cooling.

In order to characterize Zr coatings after electrodeposition runs, it was particularly important to remove the residual salt eventually present at the surface of the electrode. As the solubility of fluoride salts in water is very low, it was decided to immerse the deposits into an ultrasonic cleaner full of pure water for 15 - 20 hours.

The efficiency of this ultrasonic cleaning methodology is shown in Figure 2 with a Zr deposit on a graphite substrate. All the Zr deposits obtained in this study have been cleaned before observation and characterization.

Electrorefining is an electrochemical method used for metal purification. It consists in oxidizing the material (metal, alloy, …) to be purified at the anode, while depositing a pure metal at the cathode. The main issues to be considered are the electrodeposition conditions at the cathode and the anodic behaviour of the material.

The electrorefining process selectivity depends on the dissolution potentials of the metal to be treated. Considering a material containing zirconium and impurities, if the impurity dissolution potential is higher the Zr one, the impurity is not anodically dissolved. The electrorefining selectivity is thus ensured by the anode potential. On the contrary, if the impurity dissolution potential is lower, the impurity is simultaneously dissolved with zirconium but is not reduced at the cathode. The process selectivity is governed by the cathode potential. Finally, if the dissolution potentials are similar, the impurity is dissolved together with Zr: its electrochemical reduction behaviour has to be studied to evaluate the possibility of its co-deposition with zirconium.

The electrochemical behaviour of Zr(IV) species in molten fluorides has been already established in previous studies in our laboratory in molten LiF-CaF₂ by Gibilaro et al. [10] and in LiF-NaF by Quaranta et al. [13] . The results showed that Zr(IV) is reduced into Zr metal in a diffusion controlled one-step process exchanging four electrons. Quaranta et al. [13] also studied the electrocristallization phenomenon and demonstrated that the nucleation mode of Zr electrodeposition process is progressive. To obtain a smooth coating structure, the creation of a high number of nuclei is essential, which is favoured by both high temperature and overvoltage. Nevertheless, for too high overvoltage, the electrolyte diffusion limitation can promote dendritic growth phenomena. Considering these conflicting effects, it was decided to investigate zirconium coatings in different operating conditions by varying Zr(IV) concentration and applied current density. To have the possibility to clearly highlight the influence of these operating conditions, two significantly different values were selected: 0.49 and 1.96 mol·kg⁻¹ for Zr(IV) content and an applied current density of −35 and −110 mA·cm⁻².

To evaluate the impact of Zr(IV) content in the solution on the coating morphology, Zr electrodepositions were done on a graphite substrate at −35 mA·cm⁻² and 750˚C for different Zr(IV) molalities. Figure 3 shows typical photographs of deposits done at 0.49 and 1.96 mol·kg⁻¹. The surface morphology of the deposits is clearly influenced by Zr(IV) concentration: the roughness of coating realized at high Zr(IV) content is lower than at low Zr(IV) concentration. This observation agrees with the progressive nucleation mode established in our previous study [13] : higher solute concentration leads to a higher germination rate, so a greater number of generated nuclei, improving the coating compacity.

The influence of the applied current density on the Zr deposits morphology can be observed in Figure 4 , for two electrolyses performed at −35 (a) and −100 (b) mA·cm⁻² on graphite electrode at 750˚C and Zr(IV) molality of 1.96 mol·kg⁻¹. This figure clearly highlights that at high current density, coatings become dendritic and thus have a high roughness. This result has already been observed in previous studies concerning silicon, tantalum and niobium electrodeposition in molten fluorides [18] - [20] and is also in agreement with a progressive nucleation mode. It can be noted that the applied current density has a greater impact on the morphology than the Zr(IV) concentration. The dendritic aspect of Zr deposits should not be problematic since they are adherent enough to the substrate to prevent any loss of Zr metal into the melt.

These results conclude that Zr recovery by electrodeposition can be efficient during long durations electrolyses with a applied current density of around −0.1 A·cm⁻² on graphite electrode.

As mentioned in the introduction part, Zircaloy-4 is a Zr-based alloy mainly containing Cr, Fe and Sn (cf. Table 1 ). In order to confirm the alloy homogeneity, a SEM micrograph of a cross-section is presented in Figure 5 : Zircaloy-4 is a single-phase material with homogeneous composition.

The anodic behaviour of Zircaloy-4 was compared to those of Zr and several metals considered as impurities in LiF-NaF at 750˚C by linear sweep voltammetry at 10 mV·s⁻¹.

The plots for Zircaloy-4, Zr, Cr, Nb, Fe, Sn, Co and Ni are shown in Figure 6 . As it can be observed, the anodic behaviour of Zircaloy-4 and Zr is similar, with the same dissolution potential and an oxidation rate slightly lowered for Zircaloy-4. Finally, Figure 6 also highlights that Cr, Fe, Nb, Sn, Co and Ni are anodically dissolved at a higher potential than Zr: the selective dissolution of Zr from Zircaloy-4 is thus possible.

Galvanostatic electrolyses were performed on Zircaloy-4 anodes in LiF-NaF at 750˚C. Cross-sections of the electrodes after electrolysis were characterized by SEM-EDS and a typical SEM micrograph obtained after electrolysis at −0.21 A·cm⁻² during 1.1 h is presented in Figure 7 . This micrograph evidences two zones: 1) a homogeneous area corresponding to the core of the anode which was not dissolved, and 2) a non-homogeneous degradation layer located at the surface of the anodes.

The composition of the degradation layer was characterized by EDS analyses at different depths, from the surface to the core of the anode, as presented in Figure 8 . This figure clearly evidences the two aeras observed in Figure 7 . The composition of non-degraded Zircaloy-4 is homogeneous and identical to fresh material. In the degradation layer, Zr content is significantly lower than in the core of the anode while the content of other elements remains constant.

The experiments were repeated for different durations, corresponding to a variation of anodic charge up to 2500 C. Similar morphologies of anodes were observed and the thickness of the degradation layer was plotted versus the anodic charge of the polarization (see Figure 9 ). A linear relationship was observed indicating that 1) only Zr was dissolved and 2) the anodic dissolution was not limited by intermetallic diffusion of Zr into Zircaloy-4.

As Cr, Fe, Nb, Sn, Co and Ni are anodically dissolved at a higher potential than Zr, these results prove the selective dissolution of Zr from Zircaloy-4 anode. The strategy to selectively dissolve Zr at the anode and recover pure Zr at the cathode during electrorefining runs in this electrorefining is thus to control the oxidation potential.

Five electrolyses corresponding to different electrical charges were performed at 750˚C in molten LiF-NaF-ZrF₄ (1.96 mol·kg⁻¹) with Zircaloy-4 as anode and graphite as cathode, and both anode and cathode potentials were recorded. A typical electrolysis record, plotting anode and cathode potentials and anode-cathode difference of potentials (ΔE) versus time is presented in Figure 10 for galvanostatic electrolysis at −0.15 A cm⁻² during 3.2 hours. This figure highlights first that for duration lower than 4000 s, anode and cathode potentials similarly slightly increase while ΔE remains constant at around 0.15 V. This observation can be attributed to a small variation of the reference electrode potential. Then, in the 4000 - 8000 s duration range, all potentials remained stable. Finally, for t > 8000 s, cathode potential did not vary while anode potential sharply increased up to values higher than 0.5 V vs (F_2(g)/F⁻). At this time, the electrolysis was immediately stopped to prevent impurities dissolution in the bath.

The dissolution progress can be calculated by weighting the anode before and after electrolysis (after salt cleaning) and the mass difference (Δm) is compared to the mass of Zr initially present in Zircaloy-4 anode immersed in the bath (m_{i Zr}) as follows:

$\text{Dissolution}\text{progress}=\frac{\Delta m}{m_{i\text{Zr}}}$ (8)

For electrolysis of Figure 10 , the maximum dissolution progress of the anode without dissolving impurities was calculated at the end of electrolysis to be 74% in the case of the electrolysis presented in Figure 10 .

The other key-points of the process are the purity of electrodeposited Zr, the anodic and cathodic faradic efficiencies.

At the end of the runs, both electrodes were cleaned and weighted. Photograph of the Zr deposits obtained after electrolysis runs are presented in Figure 11 . In this figure, the growth of the coating clearly leads to big dendrites formation for a high quantity of Zr, confirming the conclusions provided in paragraph 3.2. If the dendrites are too brittle, they can be broken during the cleaning process, leading to a decrease in the recovery efficiency. However, it can be noted that no lack of dendrites can be observed on the deposits in Figure 11 , meaning removing the salt from the deposit does not impact the metal recovery efficiency.

The deposits composition was analysed by SEM-EDS and the Zr, Sn, Fr and Cr content was measured (see Table 2 ). Only zirconium was detected by EDS, involving purity of the deposit higher than 99.9%. The content of other elements is thus lower than the limit of detection of EDS, estimated at 0.1%. In order to confirm that Sn, Fe and Cr were not dissolved during electrolyses, samples of the bath were analysed, after cooling, by ICP-AES after several runs. The results show that Sn, Fe and Cr are effectively not detected in the bath, with a limit of quantification of 0.001 mass%. These results prove the high selectivity of Zr dissolution from Zircaloy-4.

The mass variations of the anode and the cathode before and after electrolysis were measured and compared to the theoretical ones calculated from Faraday’s law, considering only Zr dissolution and electrodeposition, allowing to calculate anodic and cathodic faradic efficiencies. All the results are summarized in Table 3 .

These results highlight that both anodic and cathodic faradic efficiencies are closed to 100%, in agreement with cathodic and anodic investigations. The obtained values higher than 100% can be attributed to the residual salt present in the deposit after the cleaning process. Similar efficiencies were obtained in our laboratory concerning Si electrorefining from Si-Fe anodes [4] . All these results validate the electrorefining strategy developed during this study.