Fluoride Molten Salts

Fluoride Molten Salts


2MB Sizes 1 Downloads 16 Views




Aida Abbasalizadeh1, Lidong Teng2, Seshadri Seetharaman2, Jilt Sietsma1, Yongxiang Yang1 Department of Materials Science and Engineering, Delft University of Technology, Delft, The Netherlands1; Department of Materials Science and Engineering, Royal Institute of Technology, Stockholm, Sweden2

1. INTRODUCTION International concern has been raised regarding the supply shortage of rare earth elements (REEs) since China, the largest producer of rare earths, reduced the export of these elements. At the same time, global demand has increased over the past years. Among the REEs, neodymium (Nd) is extensively applied for the production of magnets. These magnets are used in different applications such as computer hard disk drives, voice coil motors, and magnetic resonance imaging sources because of their superior magnetic properties. Dysprosium (Dy) is often used as an additive element in Nd magnets to retain magnetic properties at high temperatures. An effective recovery method for Dy and Nd is needed because almost all magnets are disposed after being used (Kurachi et al., 2012). Electrochemical deposition is one method to recover REEs. However, electrochemical deposition in aqueous solutions is not a feasible method because REEs have a highly negative electrode potential and react with water and oxygen. Therefore, molten salt electrolytes are selected for the electrowinning of rare earths (Lodermeyer et al., 2006). In this work, we used molten salt chloride to electroreduce rare earth elements from Nd magnets. The other important issue that has been discussed in recent years is the recovery of rare earths from rare earth oxides (REOs). Different methods were used to remove oxygen from REOs: direct electrochemical deoxidation (Hirota et al., 1999), solid state electrotransport (Fort et al., 1987, 1995; Jordan et al., 1975), oxyhalide formation (Corbett et al., 1986), and calcium-halide deoxidation (Okabe et al., 1998), which was further combined with electrolysis to balance CaO activity in the molten salt (Hirota et al., 1999). Yet, the strong affinity of rare earth metals to oxygen (Carlson et al., 1974) has made it difficult to industrialize any of these methods except molten salt electrolysis. Currently, the main technology to produce pure rare earth metals and rare earth master alloys is REO electrolysis in Rare Earths Industry. http://dx.doi.org/10.1016/B978-0-12-802328-0.00024-3 Copyright © 2016 Elsevier Inc. All rights reserved.




fluoride molten salts (Siming et al., 2011). In general, because of the higher efficiency, lower energy consumption, no limitation resulting from H2 evolution, and higher purity of the deposits, a number of active metals such as aluminum, magnesium, sodium, and potassium are produced by molten salt reduction or electrolysis (Mishra and Olson, 2005). Chloride melts are mostly used for different applications because, compared with fluorides, they are less expensive and less corrosive and have lower melting temperatures (Han et al., 2011). However, fluorides have higher stability and higher conductivity compared with chloride salt. In this chapter we discuss the possibility of using aluminum chloride (AlCl3) as the chlorinating agent in an LiCl-KCl-NaCl ternary electrolyte for the reduction of Nd and Dy from NdFeB magnets containing Dy. Use of the eutectic composition of an LiCl-KCl-NaCl electrolyte (eutectic point at 354  C) made it possible to perform the experiments at lower temperatures compared with fluoride salts. However, owing to the low stability of the chloride salts, we decided to use the fluoride salts for the electrochemical reduction of rare earth metals from REOs. Hence, aluminum fluoride (AlF3) was used as a strong fluxing agent in the molten fluorides to react with REOs and form rare earth fluoride, which can be further subjected to electrolysis under the applied voltage.

2. THERMODYNAMIC CONSIDERATIONS 2.1 ELECTROCHEMICAL REDUCTION OF NdFeB MAGNETS CONTAINING Dy Efficient dissolution of metal in the molten salts depends on the choice of additives. Earlier work on the extraction of iron, chromium, and Nd from industrial electric arc furnace slag, chromite ore (Ge et al., 2010), and spent Nd magnets (Abbasalizadeh et al., 2013) proved that AlCl3 can act as a powerful chlorinating agent. The reaction between the magnets and AlCl3 leads to metal chloride formation. The formed metal chloride is subjected to electrolysis and reduced on the cathode. The standard Gibbs energy for different metal chloride formation (for metals present in the magnets), using AlCl3 as the chlorinating agent, was calculated using FactSage software (FactSage 6.3). The results showed that Nd and Dy trichlorides are more stable than AlCl3, whereas the formation of FeCl3, FeCl2, and BCl3 is not favored. The Gibbs energies of the corresponding reactions are listed in Table 1. From the Gibbs energy formations calculated in Table 1, it can be concluded that Nd and Dy react with AlCl3; as a result, NdCl3 and DyCl3 are formed in the salt bath. These can be further Table 1 Gibbs Energy Values of the Reaction of AlCl3 with Different Metals in the System

Chlorination Reactions AlCl3 AlCl3 AlCl3 AlCl3 AlCl3

(salt) þ Dy(s) ¼ DyCl3 þ Al (liquid) (salt) þ Nd(s) ¼ NdCl3 þ Al (liquid) (salt) þ Fe(s) ¼ FeCl3 þ Al (liquid) (salt) þ 1.5Fe(s) ¼ 1.5FeCl2 þ Al (liquid) (salt) þ B(s) ¼ BCl3 þ Al (liquid)

DGo (kJ/mol) at 1073 K

DHo (kJ/mol) at 1073 K

DGo (kJ/mol) at 298 K

DHo (kJ/mol) at 298 K

206.2 247.8 296.7 207 180.2

353.5 368 337.7 165.1 195.3

291.2 331.4 296.4 176.8 242.4

293.9 330.2 306.6 193.2 303



electrolyzed under the applied voltage. Iron (II,III) chloride as well as barium chloride will not form in the system because of their positive Gibbs energy value. Selective chlorination of the REEs by aluminum chloride is important because it separates the rare earths from iron and boron in the magnet. These calculations are based on the pure substances in their standard state at 1073 K, whereas the activity of Nd and Dy in the magnet as well as the activity of the chlorides would change after dissolution in the salt bath. The formed rare earth chloride will be decomposed according to the reaction under external voltage in the molten electrolytes (1): RECl3 ¼ RE3þ þ 3Cl


The decomposition voltage of the different metal chlorides and alkali chlorides was calculated using FactSage software. Results are presented in Figure 1. An overpotential of 0.8 V was suggested for the decomposition of NdCl3 in the earlier work (Abbasalizadeh et al., 2013). Considering that Nd and Dy possess similar properties, and based on the decomposition voltage of DyCl3 shown in Figure 1 (2.6 V at 800  C), a voltage of 3.4 V was applied for the electrodecomposition of DyCl3.

2.2 ELECTROCHEMICAL REDUCTION OF REOs For the electrochemical reduction of REOs, the first step is the electrolyte selection. The reduction potential of the electrolyte should be more negative than the reduction potential of the REOs, meaning that the molten salt electrolyte has to be more stable than the rare earth compounds in the system. To compare the stability of different chloride and fluoride electrolyte systems, thermodynamic studies were performed on these salt components as well as on the oxides. The decomposition voltages of Nd oxide (Nd2O3), Nd fluoride (NdF3), and the most common molten salts, calculated using FactSage,

FIGURE 1 Calculated standard decomposition voltage as a function of temperature for the chlorides formed in an LiCl-KCl-NaCl-AlCl3 molten salt bath.



FIGURE 2 Comparison of decomposition voltages of different salts and oxides at different temperatures.

are compared in Figure 2. Because of the similarity of the chemical behavior of the rare earths, thermodynamic calculation was done only for Nd compounds as the representative for other rare earths. Figure 2 shows that fluorides have higher decomposition voltages than chlorides; therefore, they are expected to be more stable. Among chlorides, KCl is the most stable and NaCl has the lowest stability. However, none of these chlorides (KCl, LiCl, CaCl2, and NaCl) can be used for the electrochemical reduction of rare earth fluorides owing to their low stability. From the thermodynamic results, it can also be seen that among CaF2, LiF, KF, and NaF at different temperatures, the most stable fluoride is CaF2. Hence, the relative stability of these metal fluorides is CaF2 > LiF > NaF > KF. Comparing the decomposition voltage of rare earth fluorides and the alkali fluorides and chlorides, only CaF2 and LiF are stabler than NdF3. In other words, in the case of using NaF, NaCl, KF, KCl, CaCl2, and LiCl, we can expect that Na, K, Ca, and Li, respectively, will be reduced on the cathode before the reduction of rare earths. Therefore, for the electrochemical deposition of REOs, the most suitable electrolytes among fluorides are CaF2 and LiF. Considering the high melting point of calcium fluoride (1418  C), the eutectic composition of LiF-CaF2 (79–21 mol%) (Figure 3) is a suitable option to be used as the molten salt electrolyte for the electrochemical reduction of the REOs. The experimental results of Hamel et al. (2004) support the thermodynamic results in the current study. Those authors measured the standard potential of different fluoride electrolytes to find the most suitable electrolyte for the reduction of Nd. Their results show that in LiF-NaF and LiF-KF systems, no electrochemical reduction of Nd is observed. In China since the 1990s fluorides have been substituted for chloride salts in rare earth metal production. LiF-REF3-REO (RE ¼ La, Nd, Dy, Ce, Pr, and rare earth master alloys) is the main electrolyte system in the rare earth electrochemical production industry (Siming et al., 2011). In the current work, AlF3 is suggested for the electrochemical reduction of Nd from Nd2O3, and AlCl3 is used for Nd and Dy extraction from Nd magnets. In a comparison of the relative advantages of AlF3 with AlCl3, trials with AlCl3 indicated the loss of some of the aluminum chloride added to the








FIGURE 3 LiF–CaF2 phase diagram from FactSage software package (Bale et al., 2012).

vapor phase from the molten chloride bath before being dissolved in the molten salt as a result of the high vapor pressure (Abbasalizadeh et al., 2013). Aluminum fluoride, on the other hand, was found in the current work to be a suitable reducing agent for the electrolysis of REOs in molten fluorides. From the negative Gibbs energy value of reaction (2), Nd2 O3 þ 2AlF3 ¼ Al2 O3 þ 2NdF3

DG ðT ¼ 850  CÞ ¼ 177:7

kJ mol


it is seen that AlF3 can react with REOs forming rare earth fluorides, which can be further reduced at the cathode. In situ formation of REF3 is important because the solubility of REOs is low in molten fluorides. Moreover, REOs form rare earth oxyfluoride in molten fluorides according to the reaction (Stefanidaki et al., 2002): þ 3ðx  1ÞLiþ Nd2 O3 þ NdF3 þ 3ðx  1ÞLiF ¼ NdOFðx1Þ x


Preliminary experiments in our lab show the formation of oxyfluoride in an LiF-NdF3-Nd2O3 system. Figure 4 shows the energy dispersive spectroscopy (EDS) line scan of a sample cross-section after 3 h at 900  C. The changes in oxygen and fluorine exhibit the same trend, showing the formation of oxyfluoride compounds in this system. It was not possible to determine Li changes using an EDS line scan because lithium is a light element. Stefanidaki et al. (2002) studied the oxide solubility and Raman spectra of Nd2O3 in alkali fluorides. They reported that the NdF6 3 anion is the dominant complex in the eutectic NdF3-LiF melt system, and when Nd2O3 is added, an NdOF5 4 complex might form in the melt. They found that the solubility of Nd2O3 varies from 0.15 to 0.38 mol% when the NdF3 concentration changes from



FIGURE 4 Line scan graph of the quenched LiF-NdF3-Nd2O3 sample after 3 h at 900  C indicate the same changes of oxygen and fluorine, indicating formation of the oxyfluoride compound.

15 to 30 mol% at 900  C. This is in support of earlier results showing that REO solubility in electrolytes is enhanced by the presence of a rare earth fluoride salt (Stefanidaki et al., 2002). Contradictory results have been reported for the electrochemical reduction of rare earth oxyfluorides in molten salts, which show that further investigation is needed. According to Taxil et al. (2009), Ln fluorides in the presence of metal oxides will form rare earth oxyfluoride, which is an insoluble product. Stefanidaki et al. (2001) showed that Nd oxyfluoride is not reduced to Nd metal. In the voltammetric characterization of an LiF-NdF3-Nd2O3 system, they observed the same voltammogram as that for an LiF-NdF3 system. They concluded that Nd is reduced on the tungsten cathode by electroreduction of NdF3 (present in the form of [NdF6]3–), whereas oxygen is generated on the glassy carbon anode by oxidation of Nd oxyfluorides (present in the form of [NdOF5]4–), producing CO and CO2 gases. They believed that electrochemical production of Nd in an oxyfluoride melt is possible at low-voltage electrolysis, in which fluorocarbon compounds are not formed. Thudum et al. (2010) showed that Nd in LiF-CaF2-NdF3-Nd2O3 and LiF-CaF2-LaF3-Nd2O3 systems can be reduced from both Nd oxyfluorides and NdF3 ions, depending on the molar ratio of Nd oxyfluorides to NdF3 ions (OF/F). At low OF/F ratios, [NdF6]3– is reduced to Nd. Meanwhile, above a critical Nd2O3 concentration, [NdOF5]4– are cathodically active ions and are reduced on the cathode. Kaneko et al. (1993) suggested that in the oxyfluoride system, oxygen is generated on the anode, whereas fluorine can be produced at the anode at a higher cell voltage. Based on the Gibbs energy value of reaction (2), rare earth fluoride formation is possible at 850  C. One important advantage of in situ formation of REF3 as the result of the reaction of REO with AlF3 is that the formation of rare earth oxyfluoride might be avoided. Based on reaction (2), aluminum oxide is also formed in the system. We should consider whether aluminum oxide participates in electrochemical reactions. There would be two scenarios after aluminum oxide formation in the system: either it is dissolved in the molten fluorides or it remains undissolved in the molten salt. The density of aluminum oxide is 3.95 g/cm3. Based on the density of the LiF containing different contents of NdF3 (Hu et al., 2010) shown in Table 2, an estimation shows that the density of LiF containing 5 mol% NdF3 is about 3.98 g/cm3. This means that the critical composition in which the aluminum oxide starts to float is LiF–5 mol% NdF3. Neodymium fluoride content used in molten salt is more than 5 mol% for



Table 2 Density of LiF at Different NdF3 Contents at 950  C (Hu et al., 2010) NdF3 Contents/mol %

Density/(g/cm3) at 950  C

0 25 30 35 40 45 50

1.541 4.451 4.618 4.763 4.856 4.952 5.031

the higher solubility of Nd2O3. Moreover, based on reaction (2), more NdF3 is formed in the salt as the result of the reaction between AlF3 and Nd2O3. Thus, we expect that the undissolved aluminum oxide would float on top of the salt. In this case, Al2O3 can be removed from the salt. If the formed aluminum oxide is not dissolved in the molten fluoride salt, another alternative is to add cryolite (Na3AlF6) to the system. Cryolite is used in the aluminum production industry as the solvent for aluminum oxide. Electrochemical decomposition voltages of aluminum oxide, Nd2O3, and NdF3 are calculated and compared in Figure 5. Figure 5 shows that AlF3 that is used as an additive to the LiF-CaF2-NdF3 electrolyte system, and Al2O3, which is formed in the system as a result of the chemical reaction between the AlF3 and Nd2O3, has a lower decomposition voltage than NdF3 and Nd2O3. This means that when the voltage for electrolysis of NdF3 is applied to the system, the formed Al2O3 and the remaining AlF3 in the system

FIGURE 5 Decomposition voltages of different components in the system.



will go through electrolysis as well. Hence, the co-deposition of aluminum along with the rare earth will occur. In this case, anodic and cathodic reactions at a cell voltage higher than 4.8 V (NdF3 decomposition voltage at 900  C) are: Cathodic reactions: Nd3þ þ 3e ¼ Nd


Al3þ þ 3e ¼ Al


Anodic reactions (on graphite anode): O2 ðSaltÞ þ C ðanodeÞ ¼ COðgÞ þ 2e


2O2 ðSaltÞ þ C ðanodeÞ ¼ CO2 ðgÞ þ 4e


According to the Al–Nd phase diagram shown in Figure 6, the formation of six different intermetallics is possible on the cathode. It is possible to control alloy formation by adjusting the voltage of the system. More investigation into the electrochemical behavior of LiF-CaF2-NdF3-Al2O3 is necessary. A cyclic voltammetry analysis of the system can give us better insight in this regard. Neodymium–aluminum co-deposition is beneficial because it causes the potential needed for Nd ion reduction on the cathode to move to more positive values, based on the depolarization effect

FIGURE 6 Nd–Al phase diagram (Okamoto, 2000).



(Nourry et al., 2009). The depolarization effect is expected in the case of binary systems that form intermetallic compounds. In the Nd-Al binary system, Nd reduces at a lower potential because the activity of Nd is decreased in the alloy. This phenomenon would increase the extraction efficiency of Nd. The Nd-Al alloy, which is formed on the cathode, can be used as the master alloy for the NdFeAl bulk amorphous alloys, which are attractive because of their glass-forming ability and their ferromagnetic properties at room temperature (Inoue et al., 1996). However, the use of Nd-Al alloy depends on which intermetallic compound is produced as the cathodic product. On the other hand, if the alloy formed is not marketable, an extra step is needed after the electrolysis process to separate the alloying element from the rare earth. In this case, the difference in the melting temperature of Al and Nd can be used to separate the two metals. The melting temperature of Nd is 1024  C and the melting temperature of Al is 660  C. Molten Al, which has much lower density than Nd, will float on the top and can be further removed (the density of Al is 2.7 g/cm3 and that of Nd is 7.01 g/cm3). As was discussed earlier, in situ formation of NdF3 might reduce the problem of low solubility of REOs in molten fluorides. Moreover, the alloy formation would increase the rare earth extraction efficiency because of the lower activity of the metal in the alloy, which decreases the reduction potential of the rare earth. A proper Al-Nd alloy with industrial applications should be explored as the target cathodic product.

3. EXPERIMENTS AND RESULTS 3.1 EXPERIMENTS FOR ELECTROCHEMICAL REDUCTION OF Dy AND Nd FROM NdFeB MAGNETS CONTAINING Dy IN CHLORIDE SALTS The ternary eutectic composition of LiCl-KCl-NaCl salt (55 mol% LiCl, 35 mol% KCl, and 10 mol% NaCl) was dried at 473 K for at least 24 h. High-purity AlCl3 (99%) was added as the chlorinating agent. Neodymium magnets containing approximately 6 wt% Dy were crushed into small particles (about 1 mm). According to results from scanning electron microscopy (SEM) equipped with an EDS probe analysis, the chemical composition of the magnet was indicated as Fe14 Nd1.4 Dy0.6 (because boron is a light element it cannot be detected in EDS analysis). Experiments on the reduction of Nd and Dy from scrap magnets in molten chloride were performed at 1073 K for 6 h. The flux/Nd and flux/salt ratios were chosen as 2 (molar fraction) and 20 (wt%), respectively. Table 3 shows the concentrations of different components used in the experiments. The salt mixture, flux, and magnets were heated to 1073 K in an alumina crucible in a vertical furnace. An inert atmosphere containing argon gas was used and was dehydrated by passing through silica gel. Graphite rods were chosen as the anode and cathode in view of their additional advantage as

Table 3 Amount of Different Components Used in the Electrolysis Experiment (wt%) Composition














FIGURE 7 (a) Schematic diagram of the setup. (b) Image taken from graphite cathode after electrolysis at 1073 K for 6 h in an argon atmosphere.

oxygen getters. The mixture was kept at 1073 K for 3 h, so the Nd chloride was formed and dissolved in the molten fluoride electrolyte. Then, the electrolysis was started by dipping the electrodes into the salt bath, and a constant voltage of 3.4 V was applied between anode and cathode by a DC power supply (HP, Hewlett, 6632A) based on the decomposition voltage of the DyCl3 at 1073 K. Figure 7(a) shows the schematic diagram of the setup. After 6 h of electrolysis, the crucible was cooled down under the argon gas. The deposited layer on the graphite cathode (Figure 7(b)) was separated and washed with distilled water to dissolve the salts. After removing the salts, the deposited powder was dried and prepared for analysis. To investigate the morphology and composition of the deposited product, SEM/EDS analysis was carried out.

3.2 RESULTS The microstructures of the cathode samples were analyzed by SEM and are presented in Figure 8. The phase with bright contrast in this image, indicated by A, was confirmed to be metallic, consisting mostly of Dy and Nd. The phase with dark contrast indicated by B is aluminum oxide and remaining salt. The composition of the metallic phase on the cathode samples, analyzed by EDS, is presented in Figure 9. Data from EDS point analysis show the presence of Nd and Dy in the deposited product to be dominant compared with other elements. It should be noted that oxygen detection by EDS is not reliable. The formation of a Dy-Nd metallic phase on the cathode was also investigated by EDS mapping analysis on the same region of the sample, shown in Figure 10. The intensity of the color in the image related to Nd is low but it can be seen that Dy and Nd are distributed in the same areas, confirming the formation of a metallic phase, because these regions are poor in oxygen. It can also be seen that aluminum and oxygen are distributed in the same areas, which shows the presence of an aluminum oxide phase. Dysprosium deposition and Nd were confirmed using the molten salt electrodeposition method.



FIGURE 8 Scanning electron microscopic image of cathode deposition after electrolysis of magnet, using AlCl3 as flux, at V ¼ 3.4 V and T ¼ 1073 K for 6 h. A is the metallic phase (Dy-Nd) and B is the oxide phase.

cps/eV 14

12 10

8 6

Cl Dy O








2 0 2






FIGURE 9 Energy-dispersive spectroscopic pattern of Dy-Nd deposit on graphite cathode in LiCl-KCl-NaCl molten salt at 1073 K.



FIGURE 10 Mapping images of Dy-Nd-Al-O deposit on graphite electrode in LiCl-KCl-NaCl molten salt at 1073 K.

It was not possible to detect boron in the EDS analysis because it is a light element. In earlier studies on the electrochemical reduction of Nd from Nd magnet scraps (Abbasalizadeh et al., 2014), results from wavelength dispersive spectroscopy showed that boron remains in the bulk salt bath, and the intensity of the boron peak in the cathode sample was lower than the detection limit, as shown in Figure 11.

FIGURE 11 Intensity scan over the boron peak position in the salt bath sample and cathode sample from the earlier study (Abbasalizadeh et al., 2014).



3.3 DISCUSSION The eutectic composition of an LiCl-NaCl-KCl ternary mixture was used as an electrolyte for the electrochemical decomposition of NdFeB magnets containing Dy. Aluminum chloride was used as the chlorinating agent to react selectively with rare earth metals in the Nd magnet. Lichum and Osteryoung (1981) showed that aluminum chlorides exist in the form of AlCl4 ionic species in the alkali chloride solvents, which results in the formation of a pseudo-binary solution. Results from SEM/EDS show the presence of an Nd-Dy metallic phase in the deposited material. This is explained by the proximity of their electrode potentials. It can be concluded that Nd and Dy have been dissolved in the alkali chloride melt, forming RECl3. In fact, results from Raman spectrometry confirmed that Nd(III) exists as an NdCl6 3 complex with octahedral symmetry in the molten alkali chlorides (Barbanel et al., 1990). Thus, in the chloride melt, REEs dissolve, according to (Castrillejo et al., 2003): RE3þ þ 6Cl ¼ RECl63


Considering the cathodic and anodic electrochemical reactions, these can be represented as: Anodic reaction: 3 3Cl ¼ Cl2 þ 3e 2


Cathodic reactions: Nd3þ þ 3e ¼ Nd Dy3þ þ 3e ¼ Dy


RE3þ is the stablest state of the REEs in the molten salts (Zhu, 2014). It has been reported that most REEs (La, Ce, Pr, Y) have a single decomposition signal in the molten chlorides; however, Nd reduction occurs in two steps (Castrillejo et al., 2003). The formation of divalent rare earth metal ions in the chloride melt is most likely one reason for decreasing the current efficiency, which would be caused by a two-step reduction in the rare earth metals: RE3þ þ e ¼ RE2þ


RE2þ þ 2e ¼ RE


Contradictory results have been reported on the electrochemical mechanism of Nd and Dy reduction in molten salts. It has been reported that the reduction process of NdCl3 to Nd metal in LiF-CaCl2 melts (Hamel et al., 2004), LiF (Stefanidaki et al., 2001), LiF-CaF2 (Nourry et al., 2009), and LiCl-KCl (Serp et al., 2005), is a one-step mechanism. However, De Co´rdoba et al. (2008) and Masset et al. (2005) confirmed that the reduction of NdCl3 takes place in two steps. Electrode material is one factor that can influence the electrochemical reduction behavior of rare earths in molten salts. Castrillejo et al. (2005) observed that their cyclic voltammogram results exhibit different behaviors of Dy on tungsten (W) and Al wire electrodes. They observed that on the W



FIGURE 12 Schematic diagram of the salt extraction process.

electrode, which is used as an inert cathode, Dy is reduced in two steps: Dy(III) / Dy(II) / Dy(0), whereas on the Al electrode, the electrochemical reaction would be (Stefanidaki et al., 2001): Dy3þ þ 3Al þ 3e ¼ DyAl3


in which Dy is reduced at more positive potentials compared with those on a W inert cathode, owing to the lower activity of Dy in the formed DyAl3 intermetallic compound. Study of the electrochemical reduction of Nd and Dy from Nd magnets containing Dy is based on a new process line for the electrolytic recovery of rare earths (Seetharaman and Grinder, 2010). In the molten salt process the rare earth compound, which can be magnet scrap or rare earth ore, is mixed with molten salts and heated to a specific temperature. The working temperature depends on the salt system and the state in which the rare earth is reduced on the cathode. The mixture is kept at this temperature for a certain time (about 3 h), so the magnet or REO is dissolved in the electrolyte and the metal ions are formed in the salt. After that, based on the rare earth metal that will be deposited on the cathode, a specific voltage is applied to the cell, and the pure RE metal or the alloy is reduced on the cathode. The process steps are shown in Figure 12.

4. CONCLUSION The feasibility of Nd and Dy extraction from Nd2Fe14B magnets containing 6% Dy was investigated using a combined method of molten salt extraction and electrolysis. The LiCl-NaCl-KCl ternary composition was used as electrolyte. Aluminum chloride proved to be a strong chlorinating agent that was reacted with Nd and Dy in the magnet; as the result, NdCl3 and DyCl3 were formed. However, the formation of iron chloride and barium chloride was not feasible. The formed rare earth chlorides were subsequently subjected to electrolysis; hence, Nd and Dy were reduced on the cathode. In the current



approach, it was shown that Nd and Dy recovery from magnetic scrap enables the direct separation of these metals from iron, eliminating the oxide or halide conversion steps. The simplicity of this method as the result of single-step recovery of the rare earth metals from the magnet scrap makes this process attractive from an industrial point of view. This process has the advantage of being environmentally friendly because the salt bath can be reused without contaminating the environment. Furthermore, based on thermodynamic calculations, it was shown that the strong tendency for AlF3 toward a reaction with Nd2O3 and Dy2O3 would enhance the in situ formation of rare earth fluorides in the salt bath and thereby increase the solubility of REOs in the molten fluorides. Hence, AlF3 can act as an efficient flux agent in the fluoride melt for the recovery of rare earth metals from their oxides. The use of AlF3 will be experimentally explored in forthcoming studies.

ACKNOWLEDGMENTS Part of this work related to rare earth extraction from scrap magnet was carried out at the Royal Institute of Technology (KTH) in Sweden and was supported by the Swedish Foundation for Strategic Environmental Research (MISTRA) through the Swedish Steel Producers Association (Jernkontoret). Research on the reduction of rare earth from REOs, carried out at TU Delft, received funding from the European Community’s Seventh Framework Program ([FP7/2007-2013]) under grant agreement n 607411 (MC-ITN EREAN: European Rare Earth Magnet Recycling Network, Project Website: www.erean.eu). This publication reflects only the author’s view, exempting the Community from any liability.

REFERENCES Abbasalizadeh, A., Seetharaman, S., Teng, L., Sridhar, S., Grinder, O., Izumi, Y., Barati, M., 2013. Highlights of the salt extraction process. Journal of Materials 65 (11), 1552–1558. Abbasalizadeh, A., Teng, L., Sridhar, S., Seetharaman, S., 2014. Neodymium extraction using salt extraction process. Mineral Processing and Extractive Metallurgy. http://dx.doi.org/10.1179/1743285514Y.0000000062. Bale, C., Pelton, A., Thompson, W., Eriksson, G., Hack, K., Chartrand, P., Decterov, S., Jung, I., Melanc¸on, J., Peterson, S., 2012. FactSageÔ 6.3. 1. Thermfact and GTT-Technologies. CRCT, Montreal, Canada. Barbanel, Y., Kolin, V., Kotlin, V., Lumpov, A., 1990. Coordination chemistry of actinides in molten salts. Journal of Radioanalytical and Nuclear Chemistry 143 (1), 167–179. Carlson, O.N., Lichtenberg, R.R., Warner, J.C., 1974. Solid solubilities of oxygen, carbon and nitrogen in yttrium. Journal of the Less Common Metals 35 (2), 275–284. Castrillejo, Y., Bermejo, M.R., Barrado, A.I., Pardo, R., Barrado, E., Martinez, A.M., 2005. Electrochemical behaviour of dysprosium in the eutectic LiCl-KCl at W and Al electrodes. Electrochimica Acta 50 (10), 2047–2057. Castrillejo, Y., Bermejo, M.R., Barrado, E., Martinez, A.M., Diaz Arocas, P., 2003. Solubilization of rare earth oxides in the eutectic LiCl-KCl mixture at 450  C and in the equimolar CaCl2-NaCl melt at 550  C. Journal of Electroanalytical Chemistry 545, 141–157. Corbett, J.D., Smith, J.D., Garcia, E., 1986. The selective removal of oxygen from rare earth and thorium metals through oxyhalide formation. Journal of the Less Common Metals 115 (2), 343–355. De Co´rdoba, G., Laplace, A., Conocar, O., Lacquement, J., Caravaca, C., 2008. Determination of the activity coefficient of neodymium in liquid aluminium by potentiometric methods. Electrochimica Acta 54 (2), 280–288. Fort, D., Beaudry, B.J., Gschneidner Jr., K.A., 1987. The ultrapurification of rare earth metals: gadolinium and neodymium. Journal of the Less Common Metals 134 (1), 27–44.



Fort, D., Pecharsky, V.K., Gschneidner Jr., K.A., 1995. Solid state electrotransport purification of dysprosium. Journal of Alloys and Compounds 226 (1–2), 190–196. Ge, X.L., Grinder, O., Seetharaman, S., 2010. The salt extraction process: a novel route for metal extraction part I – Cr, Fe recovery from EAF slags and low grade chromite ores. Mineral Processing and Extractive Metallurgy 119 (1), 27–32. Hamel, C., Chamelot, P., Taxil, P., 2004. Neodymium(III) cathodic processes in molten fluorides. Electrochimica Acta 49 (25), 4467–4476. Han, W., Chen, Q., Sun, Y., Jiang, T., Zhang, M., 2011. Electrodeposition of Mg-Li-Al-La alloys on inert cathode in molten LiCl-KCl eutectic salt. Metallurgical and Materials Transactions B 42 (6), 1367–1375. Hirota, K., Okabe, T.H., Saito, F., Waseda, Y., Jacob, K.T., 1999. Electrochemical deoxidation of RE–O (RE¼Gd, Tb, Dy, Er) solid solutions. Journal of Alloys and Compounds 282 (1–2), 101–108. Hu, X., Wang, Z., Gao, B., Shi, Z., Liu, F., Cao, X., 2010. Density and ionic structure of NdF3-LiF melts. Journal of Rare Earths 28 (4), 587–590. Inoue, A., Zhang, T., Takeuchi, A., 1996. Bulk Nd-Fe-Al amorphous alloys with hard magnetic properties. Materials Transactions 37 (2), 99–108. Jordan, R.G., Jones, D.W., Hems, V.J., 1975. The purification of the rare earth metals: II. Solid state electrotransport processing of terbium. Journal of the Less Common Metals 42 (1), 101–110. Kaneko, A., Yamamoto, Y., Okada, C., 1993. Electrochemistry of rare earth fluoride molten salts. Journal of Alloys and Compounds 193, 44–46. Kurachi, A., Matsumiya, M., Tsunashima, K., Kodama, S., 2012. Electrochemical behavior and electrodeposition of dysprosium in ionic liquids based on phosphonium cations. Journal of Applied Electrochemistry 42 (11), 961–968. Lichum, H., Osteryoung, R.A., 1981. Chemical and electrochemical studies in room temperature aluminium halide containing melts in Ionic Liquids. In: Inman, D., Lovering, D.G. (Eds.). New York, Plenum Press. Lodermeyer, J., Multerer, M., Zistler, M., Jordan, S., Gores, H.J., Kipferl, W., Diaconu, E., Sperl, M., Bayreuther, G., 2006. Electroplating of dysprosium, electrochemical investigations, and study of magnetic properties. Journal of the Electrochemical Society 153 (4), 242–248. Masset, P., Konings, R.J.M., Malmbeck, R., Serp, J., Glatz, J.-P., 2005. Thermochemical properties of lanthanides (Ln¼La, Nd) and actinides (An¼U, Np, Pu, Am) in the molten LiCl–KCl eutectic. Journal of Nuclear Materials 344 (1–3), 173–179. Mishra, B., Olson, D.L., 2005. Molten salt applications in materials processing. Journal of Physics and Chemistry of Solids 66 (2–4), 396–401. Nourry, C., Massot, L., Chamelot, P., Taxil, P., 2009. Electrochemical reduction of Gd(III) and Nd(III) on reactive cathode material in molten fluoride media. Journal of Applied Electrochemistry 39 (6), 927–933. Okabe, T.H., Hirota, K., Kasai, E., Saito, F., Waseda, Y., Jacob, K.T., 1998. Thermodynamic properties of oxygen in RE–O (RE¼Gd, Tb, Dy, Er) solid solutions. Journal of Alloys and Compounds 279 (2), 184–191. Okamoto, H., 2000. Desk Handbook: Phase Diagrams for Binary Alloys. ASM International, Materials Park, OH, USA. Seetharaman, S., Grinder, O., 2010. US Patent Application No. 12/991128, ref. no.: 12057(2010). Serp, J., Allibert, M., Terrier, A.L., Malmbeck, R., Ougier, M., Rebizant, J., Glatz, J.-P., 2005. Electroseparation of actinides from lanthanides on solid aluminum electrode in LiCl-KCl eutectic melts. Journal of the Electrochemical Society 152 (3), C167–C172. Siming, P., Shihong, Y., Zongan, L., Dehon, C., Liha, X., Bin, Z., 2011. Development on molten salt electrolytic methods and technology for preparing rare earth metals and alloys in China. Chinese Journal of Rare Metals 35 (3), 440–450. Stefanidaki, E., Hasiotis, C., Kontoyannis, C., 2001. Electrodeposition of neodymium from LiF-NdF3-Nd2O3 melts. Electrochimica Acta 46 (17), 2665–2670.



˜ stvold, T., 2002. Oxide solubility and raman Stefanidaki, E., Photiadis, G.M., Kontoyannis, C.G., Vik, A.F., A spectra of NdF3-LiF-KF-MgF2-Nd2O3 melts. Journal of the Chemical Society, Dalton Transactions 11, 2302–2307. Taxil, P., Massot, L., Nourry, C., Gibilaro, M., Chamelot, P., Cassayre, L., 2009. Lanthanides extraction processes in molten fluoride media: application to nuclear spent fuel reprocessing. Journal of Fluorine Chemistry 130 (1), 94–101. Thudum, R., Srivastava, A., Nandi, S., Nagaraj, A., Shekhar, R., 2010. Molten salt electrolysis of neodymium: electrolyte selection and deposition mechanism. Mineral Processing and Extractive Metallurgy 119 (2), 88–92. Zhu, H., 2014. Rare earth metal production by molten salt electrolysis. In: Kreysa, G., Ota, K., Savinell, R. (Eds.), Encyclopedia of Applied Electrochemistry. Springer, New York, pp. 1765–1772.