Pyrochemical reprocessing of spent nuclear fuel (SNF) primarily intends to reprocess the advanced fuels and transmutation blankets which reach very high burn-ups and are expected to contain high level of Pu and minor actinides (MAs) [1e5]. Typically, the key step in the pyrochemical reprocessing is the molten salt electrorefining, in which most actinides (Ans) are recovered and decontaminated from the fission products (FPs) [2,6]. During the electrorefining process, the chemically active FPs exchange with the actinide chlorides in melt and afterwards the concentration for FPCl x gradually increases [7]. Consequently, to avoid lowering the actinide/FPs separation efficiency and be conducive for cyclic utilization of the electrolyte, the molten chlorides media must be regularly purified or regenerated when the FPCl x concentration exceeds roughly 10 wt% in LiCleKCl eutectic salt [8].
Praseodymium, which exists in molten salt in the form of trivalent ion Pr(III), is a typical fission element and difficult to be removed from actinides. To recover and separate praseodymium from actinides, it is of much significance to investigate the electrochemistry of Pr(III) for in-depth understanding the separation process of lanthanides (Lns) and Ans. Up to now, several groups explored the electrochemical properties of Pr(III) in molten chlorides [9e12]. Particularly, various kinds of active cathode materials have been carried out to extract praseodymium, such as Al [9,12e14], Zn [15,16], Bi [17e19], Cd [17,20], Mg [21] and Ni [10,22,23], allowing Pr(III) ions to be reduced at more anodic potentials via forming Pr alloys, which affords a mild and convenient approach for the electrolysis.
Investigations on the distribution of Ans and Lns between liquid Al metal/AleCu alloy and molten fluoride melts has demonstrated that the strong interaction between Ans and the liquid Al alloys might lead to the excellent recovery efficiency of Ans over Lns [24e26]. In recent years, our group has also employed an Al electrode to achieve the separation of Ans from Lns, and effective recovery of Ans has been attained with a high An/Ln separation factor [27e33]. However, at the temperature of the normal electrorefining process, the Al electrode is in solid form, and An or Ln elements will form an alloy on the cathode surface and continuously diffuse into the interior Al bulk. As such, the entire process is largely controlled by the diffusion of metal into the interior of the electrode. On the other hand, zinc, as one of the active cathode materials, has attracted considerable attention towards the recovery of Lns from molten salt [15,16,34e37], since the lower melting point (692.65 K) and boiling point (1180.15 K) for zinc enable the recovery process to be conducted at relatively lower temperatures, to some extent, which can lower the melting point of Al based on the AleZn binary phase diagram and compensate for the disadvantage of using Al as a cathode.
Hence, in this work, to recover praseodymium from molten LiCleKCl salt more efficiently, we propose an electrochemical extraction method for Pr assisted with Al and Zn together, and the present work focuses on the Pr electrochemistry at both the solid W and liquid AleZn electrodes as well as on making clear the formation process and related mechanisms regarding Pr-Al-Zn intermetallic compounds.
Anhydrous LiCl, KCl, BiCl 3 , AlCl 3 and ZnCl 2 (A.R. grade) were purchased from Sinopharm Chemical Reagent Co., Ltd, while praseodymium metal (>99.9%) came from Alfa Aesar. The experimental cell consists of a straight wall alumina crucible placed in a cylindrical quartz cell inside an electric furnace, and during the experiment the salt temperature was monitored with a nickelchromium thermocouple and kept to ±1 K. Moreover, storage and operations for all reagents were treated under inert argon atmosphere to avert exposure of oxygen and moisture. The concentrations of oxygen content and moisture levels were controlled to be less than 1 ppm.
All of the electrodes and thermocouple were positioned in molten salt using a custom-built quartz structure. Unless otherwise stated, the potentials were referred to this Ag/Ag þ couple, which was fabricated from a silver wire (Alfa Aesar, 99.99% purity, f ¼ 1 mm) dipped into the solution of AgCl (1 wt%) in LiCleKCl melts contained in an closed-end alumina tube, playing a role of a reference electrode (RE). As for the counter electrode (CE), a 6 mm diameter graphite rod was used. A W wire (Alfa Aesar, 99.9% purity) sheathed with an alumina tube of 1 mm in diameter served as the working electrode (WE). Prior to run, the working electrode was cleaned by galvanostatic anodic polarization and polished by fine sand paper in case of necessity. The active W electrode surface area (S ¼ 0.56 cm 2 ) was calculated after each experiment via measuring the length of the working electrode submerged into the melts. Moreover, the liquid AleZn electrode (S z 0.97 cm 2 ) with the eutectic composition (15 wt.% Al) was prepared by fusing pure individual metals (Al > 99.99% and Zn > 99.99%) in the glove box. To make the alloy more uniform, we stirred the liquid AleZn alloy during the fusing process.
The eutectic LiCleKCl (45:55 wt%) was dried under vacuum for more than 48 h at 473 K to get rid of residual water. Praseodymium metal block was used as the raw material of Pr(III) ions following the reaction according to Ref. [19], which elaborates praseodymium metal could be chlorided by BiCl 3 oxidant with praseodymium chloride yielded. This reaction could be expressed as:
BiCl 3 is a volatile salt that will rapidly volatilize at about 500 C and would be thus introduced into LiCleKCl eutectic directly. At the same time, Pr metal block connected with a molybdenum wire was also directly put into molten salt. In order to increase the contact area between the BiCl 3 and Pr metal block, we continuously shook the molybdenum wire connected with Pr metal. As for the amount of BiCl 3 added into the system, there was absolutely excess of Pr metal in the cell, so there will be no Bi(III) ions impurities in molten salt system, and the green color of the melt proved the generation of Pr(III) ions. The reaction was ended until the Bi element was not detected in molten LiCleKCl eutectic by inductively coupled plasma atomic emission spectrometer (ICP-AES Perkin Elemer NEXION 300D) and the Pr metal block should be lifted up from molten salt by a molybdenum wire. Actually, for convenience, higher concentration of PrCl 3 in LiCleKCl molten salt was prepared. In the light of different studied objects and purposes, the yielded LiCleKCl-PrCl 3 salts with higher concentration was added to the LiCleKCl eutectic and diluted to the desired concentration for experimental investigation.
All electrochemical measurements were executed with an Autolab PGSTAT 302 N electrochemical workstation controlled with the Nova 1.11 software package.
Specimens of Pr-Al-Zn alloys were prepared by electrolysis using molybdenum net (10 Â 15 Â 0.1 mm, 200 mesh) and/or liquid AleZn alloy as cathode, the graphite rod of 6 mm diameter as anode. The cathodic deposits were drawn out from the bath and washed ultrasonically with ethanol, which scarcely reacts with Pr-Al-Zn alloys, to remove solidified salt attached to their surfaces and afterwards stored in the glove box before subjected to further analyses. The cathodic deposits were characterized by X-ray diffraction (XRD, Bruker, D8 Advance). The composition and structure of Pr-Al-Zn samples were examined by utilizing scanning electron microscopy (SEM, Hitachi S-4800) coupled with energy dispersive spectrometry (EDS, GENESIS 2000).
Cyclic voltammogram (CV) measurements on the W electrode were carried out at 723 K. Fig. 1a elucidates a series of CV curves in LiCleKCl eutectic before (pink curve) and after (gray curve) the PrCl 3 (1.1 wt%) introduction. The cathodic signal A at appropriately À2.55 V and its corresponding striping signal A 0 , are due to the formation and depletion of lithium metal, respectively [16,35]. As can be seen, when PrCl 3 is present in LiCleKCl eutectic, one couple of cathodic/anodic signals, B/B', emerge in the voltammogram. The sharp cathodic wave B shows that the cathodic deposit at close to À2.00 V is an insoluble phase with respect to the formation of Pr metal.
Moreover, the "crossover" phenomenon is observed in the cyclic voltammogram recorded at a lower scan rate of 20 mV/s (see Fig. 1a inset), manifesting that it was the result of surface area growth caused by metal dendrite formation or that nucleation was involved in the reduction process [38e40], similar phenomena have been also reported previously in the reduction process of Mg(II) [41] and U(III) [42].
On the other hand, the potential value of signal B gradually shifts towards more negative direction and that the associated peak current increases continuously along with the rise of scan rate, as depicted in Fig. 1a. The reason for this behavior may be the influence of ohmic drop which varies with the scan rate and also is a function of the cathodic current because of the electrolyte resistance [28,43]. In Fig. 1b (inset), the electrolyte resistance was measured to be 0.61 U via electrochemical impedance spectroscopy (EIS). Fig. 1b shows a group of CV curves obtained in the LiCleKCl-PrCl 3 molten salt at various scan rates after doing ohmic compensation, at 723 K. The cathodic peak potentials are almost independent on the scan rate and locate at around À2.01 V.
According to Ref. [44], Haarberg et al. demonstrated that the Berzin-Delahay equation would be invalid since the electrode was not covered by multiple monolayer products in the deposition process. Another alternative approach is to convolute the cyclic voltammetric data, afterwards, the semi-integral curve is very similar to the steady-state voltammetric curve, which provides convenience for further data analysis [45e48]. The semi-integral curve of the voltammetric data can be given by applying the where i(t) denotes the current of the CV, m(t) presents the semiintegral current.
The current of a representative semi-integral curve (Fig. 2 pink curve) obtained from the CV curve with the scan rate of 100 mV/s reaches a semi-integral limiting value, 0.0528, as shown in Fig. 2. For Pr(III)/Pr couple, this diffusion coefficient can be evaluated from m* per Eq. ( 2):
where m* represents semi-integral limiting current, n corresponds to the number of electrons, S is the electrode surface area, C 0 the concentration of Pr(III), D designates the diffusion coefficient. Hence, the diffusion coefficient of Pr(III) is determined to be 2.02 Â 10 À5 cm 2 /s by convolution voltammetry, which is higher than that published in Ref. [22], and the diffusion coefficients calculated from the convoluted curves for different scan rates are shown in Table 1. With these characteristics the forward and backward scans are not coincident and the hysteresis phenomenon takes place, the electrode reaction for Pr(III) on the W electrode might be not fully reversible [48,49].
Further, the electrochemical behaviour for praseodymium chloride in eutectic LiCleKCl molten salts at liquid AleZn electrode was studied. In Fig. 3, the blue curve (1) shows the results collected at the liquid AleZn electrode in the blank LiCleKCl melt, with a scan rate of 50 mV/s. The electrochemical window at liquid AleZn electrode (blue curve 1) is much narrower than that on inert W electrode (about 1.0 V). The cathodic limit for CV curve derives from the lithium reduction at liquid AleZn electrode (signal B, LieAl alloy or LieZn alloy), whereas the anodic one is associated with the anodic dissolution process of the liquid AleZn electrode (signal A 0 ). As for the pink curve (2), one new cathodic wave C and its corresponding striping wave C 0 were detected at around À1.29/-0.82 V after PrCl 3 was added into the LiCleKCl system, owing to the deposition and depletion of praseodymium metal at the liquid AleZn electrode, respectively. In addition, we can find that there is one more anodic signal (D') during the backward scan, and the similar phenomenon was detected during the Ln reduction process [50].
In surprise, within the electrochemical window range, only the reduction signal arising from Al(III) ions is detected, and the reduction signal of Zn(II) ions at the liquid AleZn electrode vanishes. The CV measurement was thus carried out in molten LiCleKCl eutectic at the liquid Zn electrode, and the results were displayed in Fig. 3 (curve 3). Comparing curve 3 with curve 1, an anodic peak E' at À0.37 V corresponding to the anodic oxidation of liquid zinc electrode, related to a well-defined cathodic one E is observed. The reason why the reduction peak for Zn vanished at the liquid AleZn electrode may be that the reduction peak of Al(III) is too large, and thus the reduction peak of Zn is overlapped. Moreover, it is found that the reduction signal B is caused by the LieZn alloy [51].
One of the most sensible answers with respect to the reversibility for a redox reaction, will be generally found via the application of peak-voltammetry theories [52]. As shown in Fig. 4a, with scan rate increasing, the cathodic and anodic signals shift slightly to more cathodic and anodic values, respectively. Fig. 4a inset presents the variation of the peak potentials for C/C 0 couple versus logarithm of the scan rate. The peak potential alterations with regard to signals C/C' are insignificant at scan rates below 50 mV/s, manifesting the redox reaction of Pr(III) at the liquid AleZn electrode may incline towards being reversible; whereas for the scan rate up to 250 mV/s, DE c and DE a will be significantly increased and two linear relationships of the redox peak potentials as a function of the logarithm of the scan rate illustrate, which suggests that the overall electrode reaction of Pr(III) ions to metal Pr should be considered to be a quasi-reversible process. Fig. 4b shows a group of CV curves obtained in the LiCleKCl-PrCl 3 molten salt at various scan rates after doing ohmic compensation, at 723 K. It can be observed that the cathodic peak B potentials of the CV curves consistently shift to slightly more negative values, manifesting the possible presence of electron transfer rate control component during the overall redox reaction [40,53]. Because of the large background current at the liquid AleZn electrode, the reduction peak current of signal C in Fig. 4 actually is not distinguishable, from which it is of considerable difficulty to obtain the diffusion coefficient of Pr(III) directly due to the possible errors about peak current determination. In consequence, the CV curve is further processed by semi-differential technique, and related equation is described as follows:
where i(t) donates the current of the CV, e(t) designates the semidifferential current. Fig. 5a shows a representative semi-differential curve derived from CV curve recorded in LiCleKCl-PrCl 3 melt at the liquid AleZn pool electrode. It is observed that the reduction peak of the semidifferential curve is more clear and obvious, and has higher accuracy on confirming its peak current as well. Therefore, according to the semi-differential theories, the diffusion coefficient can be estimated using Eqs. ( 4) and ( 5):
where F, R and T have the usual meanings, W p and e p represent peak width and peak height, respectively. The peak width (W p ) concerning cathodic signal C is measured to be 0.343. With subsequent the conversion calculation, it is also discovered that the kinetic parameter 'an' value is equal to 0.58 for reduction of Pr(III). The peak height, e p decreases in proportion to the scan rate which is shown in Fig. Based on the slope of the fitted curve, the calculated value of diffusion coefficient here is determined to be 4.53 Â 10 À5 cm 2 /s, and for the scan rates of 50, 150, 200 and 250 mV/s, the diffusion coefficients for Pr(III) ions are given in Table 1, which are generally higher than these obtained on the W electrode, indicating that the electrode material has a some influence on the diffusion coefficient of Pr(III) ions. Furthermore, this slight discrepancy with respect to the diffusion coefficients between the W and liquid AleZn electrodes might be due to the systematic error in the surface area determination of the liquid AleZn electrode as well.
For the purpose of getting more details in terms of the intermetallic compound formation, the co-reduction process for Pr(III), Al(III) and Zn(II) ions on the W electrode were investigated as well. Fig. 6a exhibits the comparison of typical CV curves recorded in the LiCleKCl eutectic containing ZnCl 2 (0.2 wt%), PrCl 3 (1.1 wt%) and ZnCl 2 (0.2 wt%)-PrCl 3 (1.1 wt%), respectively. Two redox couples A/ A 0 and C/C' (see Fig. 6a), associated with the deposition and its subsequent dissolution reaction of lithium and zinc metals, respectively, can be detected in LiCleKCleZnCl 2 system. When PrCl 3 is present in molten LiCleKCleZnCl 2 melts, except for the A/A 0 and C/C 0 redox systems described previously, five various waves labeled with D/D 0 , E/E 0 , F/F 0 , G/G 0 and H/H 0 , are discovered at À1.33/-1.21 V, À1.76/-1.55 V, À1.85/-1.70 V, À1.88/-1.74 V and À1.97/-1.82 V, respectively. Here, the five different kinds of redox signals are located within the electrochemical window limited by the reduction of Pr(III) (signal B) and the reoxidation of metallic Zn to Zn(II) ions (signal C'), and should be ascribed to the deposition/ dissolution of PreZn intermetallic compounds.
In the case of the gray curve in Fig. 6b, the CV test was performed in LiCleKCl eutectic with a solution of both of ZnCl 2 and AlCl 3 on the W electrode at 723 K. Three redox peaks A/A 0 , C/C 0 and I/I 0 are correlated to the formation and dissolution of metallic lithium, zinc and aluminum, respectively. Unexpectedly, between peaks C/C 0 and I/I 0 , one sharp anodic peak X 0 appears, probably due to depletion of the unstable AleZn alloy. The cyclic voltammogram consists of five new redox couples after PrCl 3 addition into the molten LiCleKCleAlCl 3 eZnCl 2 melts, as shown in Fig. 6b red curve, arising from the Pr-Al-Zn alloys. Compared with LiCleKCl-PrCl 3 -ZnCl 2 system (i.e. the comparison of the blue curve in Fig. 6a with Fig. 6b red curve), it is reasonable to suggest that the redox couples D/D 0 , F/ F 0 and H/H 0 in Fig. 6b are interpreted as the formation and dissolution for three kinds of PreZn intermetallic compounds, however, what substances peaks J/J 0 and K/K' belong to is still a question. Keeping this in mind, some voltammetric data were also employed to summary the redox couple potentials of PreAl intermetallic compounds.
The black curve of Fig. 6c illustrates an example of the cyclic voltammogram recorded on the W electrode in the LiCleKCleAlCl 3 molten salt. In consequence, a cathodic wave I is observed at À1.05 V and its corresponding sharp anodic wave I 0 at À0.90 V are attributed to the formation of aluminum metal and dissolution reaction. When PrCl 3 was introduced into the LiCleKCleAlCl 3 melts, five pairs of new cathodic signals and their corresponding anodic signals occurred (the purple curve, Fig. 6c). On the basis of Refs. [12,14], Tang et al. concluded that the redox signals (K/K 0 , L/L 0 and M/M 0 ) could be caused by the reduction and reoxidation of PreAl alloys in the LiCleKCl-PrCl 3 -AlCl 3 system. Howover, we obtained one more PreAl intermetallic compound (peak N') in the same melts than Tang [14]. This phenomenon might be due to the higher concentration of Pr(III) ions in ours, and there is enough Pr(III) ions to further generate Pr-rich intermetallic compound during the co-reduction process.
In general, comparing two curves with respect to molten LiCleKCl eutectic containing PrCl 3 -AlCl 3 and PrCl 3 -AlCl 3 -ZnCl 2 salt systems, it is very likely that redox couple K/K 0 in Fig. 6b is formed by the reduction and depletion of PreAl intermetallic compound, and another pair of couple J/J' maybe be possibly caused by the under-potentially deposition of praseodymium on the aluminum and zinc alloy already coated onto the W electrode to form ternary Pr-Al-Zn intermetallic compound.
Interestingly, it is worth to mention that there is the only one redox signal for the PreAl intermetallic compound found between the redox couples B/B 0 and C/C' in the LiCleKCl-PrCl 3 -AlCl 3 -ZnCl 2 system, while three signals corresponding to EreZn intermetallic compounds are observed. The electronegativity difference, adopted to predict the possibility of the formation of the intermetallic compound between two elements, could be the possible explanation for this phenomenon [54]. The greater the difference in electronegativity between two elements, the easier the formation for intermetallic compound. Hence, the PreZn intermetallic compounds will be formed easily because the electronegativily difference between praseodymium and zinc is larger than that between praseodymium and aluminum, and the results that the introduction of AlCl 3 into LiCleKCl-PrCl 3 -ZnCl 2 system has an insignificant influence on electrochemical properties of praseodymium are also obtained. 2, and the results obtained in OCPs are well reproduced with those attained from CVs in Fig. 6.
To confirm the cathodic products stemming from the coreduction process of Pr(III), Al(III) and Zn(III) ions on inert W electrode and the under potential deposition of Pr(III) ions at liquid AleZn electrode, potentiostatic and galvanostatic electrolyses were conducted to prepare alloy specimens with different electrode materials. After electrolysis process, the obtained specimens were washed using ethylene glycol and then stored inside the glove box for further analyses.
Potentiostatic electrolysis experiment was implemented with the W electrode in the LiCleKCl-PrCl 3 -AlCl 3 -ZnCl 2 molten melts at À1.5 V for 2 h at 723 K at first. There is only a small amount of Pr-Al-Zn alloys adhering on the W electrode although the experiment was repeated for several times. One reasonable explanation could be that all of intermetallic compounds deposited at this potential are based on W electrode coated with Zn metal, while at this temperature metal Zn was in the liquid state and might be easily drop from the W electrode during the electrolysis process.
Afterwards, the molybdenum net was adopted as the cathode, and the Pr-Al-Zn alloys were deposited by potentiostatic electrolysis at À1.5 V for 2 h. Fig. 8a shows the XRD spectrum of the cathodic deposits collected in LiCleKCl-PrCl 3 -AlCl 3 -ZnCl 2 system at 723 K, and also verifies that the components of the deposition product are Al metal as well as the intermetallic compounds Pr 2 Zn 17 and Al 2 PrZn 2 with tetragonal crystal system (I4/mmm). From Fig. 8b andc, it turns out that the morphology of Al 2 PrZn 2 intermetallic compound is lumpy shape. These lumps are formed by plenty of layers, and these layered structures are assembled tightly in a certain direction. The relative EDS analysis is shown in Fig. 8d, which confirms the existing of Pr, Al and Zn elements.
Moreover, galvanostatic (À6 mA) electrolysis experiment was carried out for 8 h at the liquid AleZn electrode in LiCleKCl-PrCl 3 (2.02 wt%) melts at 783 K. Fig. 9 displays the XRD results of the surface of the cathodic product and SEM images. Besides the monophase of Al and Zn metals, the Al 0.71 Zn 0.29 and Pr 2 Zn 17 intermetallic compounds were identified as well.
The co-reduction behaviors of Pr(III) ions with both Al(III) and Zn(II) ions as well as the under potential deposition of Pr(III) ions at the liquid AleZn binary electrode were studied systematically in LiCleKCl eutectic. In both cases, the reduction potential of Pr(III) ions occurs at more anodic potentials because of the decreasing of Pr activity in the metal phase. Meanwhile, the diffusion coefficients of Pr(III) ions in melt in the case of W electrode and liquid AleZn electrode were 2.02 Â 10 À5 cm 2 /s and 4.53 Â 10 À5 cm 2 /s, respectively, at 100 mV/s. The results show that more PreZn intermetallics could be formed than PreAl intermetallics. In addition, at the liquid AleZn electrode, Pr-Al-Zn alloys could be generated.
The authors declared that they have no conflicts of interest to this work.
Y.-l.Liu et al. / Electrochimica Acta 326 (2019) 134971