Electrochemical Recovery of Rare Earth Elements from Magnets: Conversion of Rare Earth Based Metals into Rare Earth Fluorides in Molten Salts

Abstract

In the present work, selective extraction of rare earth (RE) metals from NdFeB magnets is investigated by studying the effects of various fluxes, viz. AlF₃, ZnF₂, FeF₃ and Na₃AlF₆ in the LiF-NdFeB system. The aim is to convert RE from RE magnet into the fluoride salt melt. The results show the complete selective separation of neodymium (also dysprosium) from the magnet and formation of rare earth fluoride, leaving iron and boron unreacted. The formed rare earth fluoride can subsequently be processed in the same reactor through an electrolysis route so that RE can be deposited as a cathode product. The results of XRD and EPMA analysis of the reacted samples indicate that AlF₃, ZnF₂ and FeF₃ can act as strong fluorinating agents for extraction of rare earth from NdFeB magnet, converting the RE to REF₃.

The results confirm the feasibility of the rare earth metals recovery from scrap NdFeB magnet as raw material. The fluoride conversion-electrolysis route suggested in the present work enables the extraction of rare earth metals in a single step using the above-mentioned fluxes.

1. Introduction

In view of the issues of supply and demand of the rare earth metals due to the uneven distribution of these metals, there is a strong demand for recycling waste products, as for example, rare earth magnets. Due to the lack of technologically and economically feasible recycling technologies, only less than 1% of the rare earth elements were recycled until 2011¹⁾. An effective process for the recycling of rare earth scrap is a major concern since most developed countries import their required rare earth elements²⁾.

Hydrometallurgical and pyro-metallurgical process routes have been the subject of investigation by a number of researchers^3–7). However, the drawbacks of hydrometallurgy processes are the low reaction rates and the serious environmental problems arising from the huge amounts of waste water generated in the process, which, in turn, need further processing. Alternatively, molten salt processing routes offer a more attractive flow sheet in view of the lower energy consumption, higher efficiency and higher purity of the deposits^8–10). The molten salt process is currently used in the production of reactive metals such as aluminium, magnesium, sodium, potassium and also in general for the production or separation of lanthanides from actinides¹¹⁾.

Based on the Salt Extraction Process^12,13) and for recovery of metal in the molten salt systems¹⁴⁾, the feasibility of the neodymium extraction from NdFeB magnet in chloride salt electrodeposition method has been investigated¹⁵⁾. A new approach involving molten fluoride bath has been proposed in a recent study at TU Delft for solving the problem of low solubility and oxyfluoride formation of rare earth oxides (REOs) in molten fluorides. In this approach, AlF₃, ZnF₂, FeF₃ and Na₃AlF₆ were used as fluorinating agents in the LiF-Nd₂O₃ system. The rare earth oxide was converted into rare earth fluoride, which can subsequently be processed in the electrolysis route in which rare earth metal is electrochemically extracted as the deposition on the cathode. Complete conversion of REO in molten fluorides can overcome the problem of the low solubility of REOs in the molten fluorides. The objective of the present work is to examine the applicability of this process concept for the extraction of neodymium and other rare earth elements from NdFeB magnets, using AlF₃, ZnF₂, FeF₃ and Na₃AlF₆ in the LiF-NdFeB system. It is interesting to examine the possibility of selective extraction of the rare earth metal from Nd magnet by fluorination. The REF₃ which is formed as the result of the chemical reaction between the additives and the RE-containing components of the magnet, will be subjected to electrolysis to extract the RE as the cathodic product. To overcome fluorocarbon formation on the anode a reactive anode was employed instead of the conventionally used graphite anode. This reactive anode is anodically dissolved to re-generate the fluorinating agent in-situ in the electrochemical reactor. This method provides a one-step process for direct recycling of rare earth metal from neodymium magnet scrap by molten fluoride salt extraction and electrolysis.

2. Thermodynamic Evaluations

Prior experiments by the present authors indicate that AlF₃, ZnF₂ and FeF₃ can be used for the extraction of neodymium from neodymium oxide. As thermodynamic calculations confirm the feasibility of the conversion of neodymium oxide to neodymium fluoride using AlF₃, ZnF₂ and FeF₃ as fluorinating agents, these fluorides were used in the present work for the rare earth extraction from neodymium magnet (NdFeB). Due to the high price of AlF₃ compared to cryolite, the use of cryolite as a substitute for AlF₃ was also studied. Hence the LiF-NdFeB-Na₃AlF₆ system was also investigated as part of the present work.

Thermodynamic calculations of the pure components show that AlF₃, ZnF_2, FeF₃ and Na₃AlF₆ act as the fluorinating agents leading to formation of REF₃ (RE = Nd, Dy), while Fe remains unreacted in the molten fluorides. The standard Gibbs free energies of the binary reactions of the four systems were calculated using FactSage (Tables 1–4).

Table 1 Gibbs free energies of possible binary reactions for the formation of fluoride in the LiF-Nd₂Fe₁₄B-AlF₃ system at T = 1223 K.

Reaction	ΔGo [kJ/mol]
Nd + AlF3 ⇄ NdF3 + Al	−176
Dy + AlF3 ⇄ DyF3 + Al	−199
Fe + AlF3 ⇄ FeF3 + Al	411
3Fe + 2AlF3 ⇄ 3FeF2 + 2Al	769
B + AlF3 ⇄ BF3 + Al	130

Table 2 Gibbs free energies of possible binary reactions for the formation of fluoride in the LiF-Nd₂Fe₁₄B-ZnF₂ system at T = 1223 K.

Reaction	ΔGo [kJ/mol]
2Nd + 3ZnF2 ⇄ 2NdF3 + 3Zn	−1076
2Dy + 3ZnF2 ⇄ 2DyF3 + 3Zn	−1121
2Fe + 3ZnF2 ⇄ 2FeF3 + 3Zn	99
Fe + ZnF2 ⇄ FeF2 + Zn	15
2B + 3ZnF2 ⇄ 2BF3 + 3Zn	−462

Table 3 Gibbs free energies of possible binary reactions for the formation of fluoride in the LiF-Nd₂Fe₁₄B-FeF₃ system at T = 1223 K.

Reaction	ΔGo [kJ/mol]
Nd + FeF3 ⇄ NdF3 + Fe	−588
Dy + FeF3 ⇄ DyF3 + Fe	−610
B + FeF3 ⇄ BF3 + Fe	−281

Table 4 Gibbs free energies of possible binary reactions for the formation of fluoride in the LiF-NdFeB-Na₃AlF₆ system at T = 1223 K.

Reaction	ΔGo [kJ/mol]
Nd + Na3AlF6 ⇄ NdF3 + 3NaF + Al	−67
Dy + Na3AlF6 ⇄ DyF3 + 3NaF + Al	−89
Fe + Na3AlF6 ⇄ FeF3 + 3NaF + Al	521
3Fe + 2Na3AlF6 ⇄ 3FeF2 + 6NaF + Al	989
B + Na3AlF6 ⇄ BF3 + 3NaF + Al	240

According to the standard Gibbs energies of the reactions in Tables 1–4, aluminium fluoride reacts only with the REs in the magnet. The formation of FeF₃ and BF₃ are not thermodynamically feasible in the standard state. However, ZnF₂ and FeF₃ can not only react with REs but also with B in the magnet (Table 2 and Table 3), showing the feasibility of BF₃ formation in both cases. Yet, the formation of FeF₃ in the LiF-Nd₂Fe₁₄B-ZnF₂ and LiF-Nd₂Fe₁₄B-Na₃AlF₆ systems at T = 1223 K is not favoured. In the case of the presence of Dy in the magnet, DyF₃ would be formed in all systems. It is important to note that these reactions refer to the case of the reacting species at their standard state as pure substances. In the real systems, the activity of the components would be lower than unity. Due to the absence of thermodynamic data for the fluoride systems, it is not possible to calculate the equilibrium potential of the fluorides in a similar way as it has been calculated for chlorides¹⁶⁾. Thus, the ideal Temkin model was used in order to calculate the activities of the different fluorides in the system, assuming that the magnet is dissolved in the salt, thereby forming fluorides which are completely ionized in the salt melt. Hence, the system is a mixture of cations and fluorine ions. Based on this model, the activity of a compound such as NdF₃ was calculated as:   

\[a_{NdF_3} = (X_{Nd^{3+}})(X_{F^-})^3\](1)

In which the X-term refers to the ionic fraction of cations and anions:   

\[X_{Nd^{3+}} = \frac{n_{Nd^{3+}}}{\Sigma n_+} \quad {\rm and} \quad X_{F^-} = \frac{n_{F^-}}{\Sigma n_-}\](2)

n is the number of moles of ions. Using these activities, the relative partial molar Gibbs energies (chemical potentials) of the fluorides species in the LiF-NdFeB-AlF₃, LiF-NdFeB-ZnF₂, LiF-NdFeB-FeF₃ and LiF-NdFeB-Na₃AlF₆ systems were then calculated using the expression:   

\[\Delta \bar{G}_{NdF_3} = RT\ Ln\ a_{NdF_3}\](3)

The activities and relative partial molar Gibbs energies of the fluorides are listed in Table 5. The relative partial molar Gibbs energies of NdF₃ in all systems are lower than the corresponding values for FeF₃. Using a reactive anode in which FeF₃ is constantly generated in the system will generate sufficient amounts of FeF₃ that are instantly consumed in the fluorination of the magnet scrap. This should enable NdF₃ formation.

Table 5 Relative partial molar Gibbs energies and activities of the fluoride species in the systems with different fluorinating agents, assuming ideal Temkin behavior of the systems at T = 1223 K.

System	LiF-NdFeB-AlF3		LiF-NdFeB-ZnF2		LiF-NdFeB-FeF3		LiF-NdFeB-Na3AlF6
	a	Δ$\bar{G}$ [kJ/mol]	a	Δ$\bar{G}$ [kJ/mol]	a	Δ$\bar{G}$ [kJ/mo]	a	Δ$\bar{G}$ [kJ/mol]
LiF	0.26	−13.6	0.237	−14.6	0.260	−13.6	0.211	−15.8
NdF3	0.077	−25.9	0.071	−26.9	0.077	−25.9	0.063	−28
FeF3	0.544	−6.1	0.496	−7.1	0.622	−4.8	0.441	−8.3
BF3	0.039	−33	0.035	−33.9	0.039	−33	0.031	−35.1
ZnF2			0.160	−18.6
AlF3	0.077	−25.9
Na3AlF6							0.0004	−78.8

The REF₃ formed in the system will dissolve in the LiF melt which will further participate in the electrolysis process. Based on the calculated Gibbs free energies for the formation of the pure substances (Table 1 Table 4), the decomposition voltage of NdF₃ at 1223 K for pure substances is about 4.7 V, not considering the over-potentials which need to be determined experimentally. This enables the extraction of Nd from the fluoride bath. In case of BF₃ formation, B can be co-deposited on the cathode with the RE elements, since it has a lower decomposition voltage than NdF₃. As the boron concentration in the magnet is however quite low (1 mass%), this can be neglected. Using cryolite as the extracting agent, NaF will form in the molten salt, which has a lower decomposition voltage than NdF₃ (decomposition voltage of NaF at 1223 considering calculation with pure substances is ～3.6 V). The cathodic reactions will be (Reaction (5) will occur only in the case of using cryolite):   

\[{\rm RE}^{3+} + 3{\rm e}^- \rightleftarrows {\rm RE}\](4)

  

\[{\rm Na}^+ + {\rm e}^- \rightleftarrows {\rm Na}\](5)

The anodic reaction using a reactive anode (M) is:   

\[{\rm M}\ ({\rm anode}) + {\rm xF}^-\ ({\rm salt}) \rightleftarrows {\rm MF}_{\rm x}\ ({\rm Salt}) + {\rm xe}^-\](6)

A schematic diagram of the salt extraction process, including the fluorinating step and the electrochemical extraction step is shown in Fig. 1.

Fig. 1

Schematic diagram of the salt extraction process.

3. Experiments

Lithium fluoride (98.5%-Alfa Aesar) was mixed in a glove box with NdFeB magnets (supplied by Magneti Ljubljana), which were ground into fine powder. The magnet composition as provided by Magneti Ljubljana is shown in Table 6.

Table 6 Composition of the neodymium magnet supplied by Magneti Ljubljana analyzed with XRF.

Element	Nd	Dy	Al	Fe	B	Co	Pr	Cu	Ga	Total
mass%	28.9	2.72	0.14	63.4	-	2.91	0.67	0.19	0.07	99.0

Either aluminium fluoride (99+%-Alfa Aesar), zinc fluoride (99%-Alfa Aesar), iron fluoride (97%-Alfa Aesar) or cryolite (99%-Alfa Aesar) powder was added to the lithium fluoride and magnet mixture. This mixture was then charged into a graphite crucible. The fluorinating agents were added in the stoichiometrical amount that is needed for fluorinating the rare earth in the magnet. The mixtures were heated up to 1223 K for 3 hours under argon atmosphere. The Ar gas was purified by passing through KOH flakes for CO₂ removal and also through silica gel and P₂O₅ to remove moisture. The gas was further purified by passing through a tube furnace containing Ti sponge which was heated to 1123 K in order to remove the traces of O₂. After 3 hours treatment, the samples were quenched in liquid nitrogen.

The samples were analysed by X-Ray Diffractometry (XRD) as well as by Electron Probe Micro Analysis (EPMA) in order to determine the phases that are formed during the experiments and their compositions. For the EPMA analyses, WDS point analyses and mappings were performed at either 15 kV-15 nA or at 5 kV-15 nA. For the quantification, standards were used.

4. Results

4.1 AlF₃ in LiF-NdFeB system

From the XRD pattern of the LiF-NdFeB-AlF₃ sample after fluorination treatment (Fig. 2), the peaks of NdF₃ phase are identified along with LiF, α-Fe and the Fe₃Al intermetallic phase. Because of the low concentration of B in the magnet, B-containing phases were not detected by the XRD analysis.

Fig. 2

XRD pattern with phase identification of the LiF-Nd₂Fe₁₄B-AlF₃ sample after fluorination treatment.

Figure 3 shows the EPMA mapping of the LiF-NdFeB-AlF₃ sample after fluorination treatment. Nd and Fe are separately distributed in the sample. This proves that Nd is selectively extracted from the Nd magnet, whereas Fe remains in the magnet matrix. F is detected in the Nd-containing regions and Al is distributed in the Fe-containing regions. B could not be detected in the EPMA mapping due to its low atomic number and low concentration in the NdFeB magnet.

Fig. 3

EPMA mapping of the LiF-NdFeB-AlF₃ sample after fluorination treatment showing F, Nd, Fe and Al distribution.

The sample was also analyzed using EPMA point analysis. A BSE image and the measured composition of the RE-fluoride, the FeAl and the FeAlB phase are shown in Fig. 4. The results clearly show the formation of a FeAl intermetallic phase (89 at% iron and 9 at% aluminum). Boron was quantified in the FeAlB phase with concentrations of 80 at% iron, 15 at% aluminum and 1.1 at% boron. The quantification data show a fluorine concentration of 72 at% and Nd concentration of 25 at% in the REF phase. The presence of 2.5 at% dysprosium along with 25 at% neodymium was detected in the rare earth fluoride phase.

Fig. 4

BSE image and composition of the phases (in at%) of the LiF-NdFeB-AlF₃ sample after fluorination treatment showing the presence of REF, FeAl and FeAlB phases.

4.2 ZnF₂ in LiF-NdFeB system

The XRD result of the experiment of the LiF-NdFeB-ZnF₂ system after fluorination shows the formation of NdF₃ in the sample (Fig. 5). LiF, Fe-Zn, Li₂ZnF₄ and Fe₂B are the other identified phases. Nd was thus selectively extracted by ZnF₂ as extracting agent and Fe remains unreacted. B also stays unreacted in the Fe₂B phase. Part of the Zn has formed Fe₉Zn with Fe and the rest is dissolved in the salt by forming Li₂ZnF₄.

Fig. 5

XRD pattern with phase identification of the LiF-NdFeB-ZnF₂ sample after fluorination treatment.

The EPMA mapping of the LiF-NdFeB-ZnF₂ sample after fluorination treatment is shown in Fig. 6. The formation of NdF₃ is confirmed by Fig. 6 (a). The areas with higher concentration of fluoride correspond to the LiF phase. Fe and Zn are distributed in the same areas and are separated from Nd and F. In order to detect B, an EPMA mapping was performed at 5 kV (Fig. 6 (b)). B is present in the areas with the highest Fe concentration. According to the XRD results, this can be the Fe₂B phase. In the areas where both Fe and Zn are present, B is present in a lower concentration. This is the FeZnB phase. The areas were both Nd and F are detected correspond to the NdF₃ phase.

Fig. 6

EPMA mapping of the LiF-NdFeB-ZnF₂ sample after fluorination treatment (a) at 15 kV and (b) at 5 kV.

In Fig. 7, the BSE image of the LiF-NdFeB-ZnF₂ sample after fluorination is shown. The light areas in this image are REF₃ phase (RE being 23 at% neodymium and 2 at% dysprosium) with 74 at% fluorine and the dark matrix is FeZnB intermetallic phase consisting of 91 at% iron, 8 at% zinc and 1 at% boron.

In Fig. 7, the EPMA quantification results show that the dark gray grains correspond to iron boron phase with high concentration of boron (25 at% boron and 73 at% iron).

For more detailed detection of the element distribution and especially for a better detection of B, a WDS mapping was also performed at 5 kV (Fig. 6(b)). This mapping agrees with Fig. 6(a), viz. the white phase is the NdF₃ phase dispersed in the grey matrix of Fe-Zn-B phase and the darker grains on the sample are Fe borides. A BSE image of the same area in the LiF-NdFeB-ZnF₂ sample after fluorination is shown in Fig. 7 together with the measured phase compositions.

Fig. 7

BSE image of the LiF-NdFeB-ZnF₂ sample in at% after fluorination treatment showing the presence of NdF₃, Fe boride and FeZnB.

4.3 FeF₃ in LiF-NdFeB system

The XRD result of the LiF-NdFeB-FeF₃ sample after fluorination treatment is shown in Fig. 8. NdF₃ has formed from the reaction between FeF₃ and the Nd in the magnet. LiF and α-Fe are the two other identified phases. For the same reasons mentioned earlier, B-containing phases were not detected with the XRD analysis.

Fig. 8

XRD pattern with phase identification of the LiF-NdFeB-FeF₃ sample after fluorination treatment.

The elemental mappings of F, Nd, B, Fe and Co elements are shown in Fig. 9. Nd is extracted from the Nd magnet and forms NdF₃ (the white appearing phase). Fe and Co are distributed in the same areas indicating that no FeF₃ or CoF₃ have formed during fluorination treatment. The slightly darker grains in the FeCoB matrix consist of Fe and B, indicating the presence of Fe borides. The concentration of Co in these grains is lower than in the FeCoB matrix. The EPMA mapping images show the presence of NdF₃, FeCo and FeB phases.

Fig. 9

EPMA mapping of the LiF-NdFeB-FeF₃ sample after fluorination treatment.

4.4 Na₃AlF₆ in LiF-NdFeB system

The XRD result of the LiF-NdFeB-Na₃AlF₆ sample after fluorination treatment is shown in Fig. 10. The formation of Na_1.5Nd_1.5F₆ shows that Nd in the magnet has substituted Al in the cryolite, hence Nd is extracted from the magnet. The small peaks related to cryolite show that a low concentration of this phase has remained in the sample. The XRD pattern also shows peaks related to α-Fe and LiF.

Fig. 10

XRD pattern with phase identification of the LiF-NdFeB-Na₃AlF₆ sample after fluorination treatment.

The EPMA elemental mappings of the LiF-NdFeB-Na₃AlF₆ sample after fluorination are shown in Fig. 11. The different phases in the sample were identified and are marked on Fig. 12. Fe and Co are present in the same areas. Al is distributed in the same areas as Nd and F, but its concentration is low compared to Nd. The light appearing grey grains contain Fe and Co and are clearly separated from the dark appearing phase consisting of Nd, Dy, F and Al. The very bright appearing phases in the middle of the Fe-rich grains are the phases in which the intensity of Nd and F are very high and Al is not present. This can be related to the Na_1.5Nd_1.5F₆ phase. According to the XRD results, both Na_1.5Nd_1.5F₆ and Na₃AlF₆ phases are present in the sample. The small black spots on the sample have high concentration of F in which Nd and Al are absent. This can be LiF phase.

Fig. 11

EPMA mapping analysis of LiF-NdFeB-Na₃AlF₆ sample after fluorination treatment.

Fig. 12

BSE image of the LiF-NdFeB-Na₃AlF₆ sample after fluorination treatment.

5. Discussion

In all the experiments, Nd (and Dy in the case of the fluorinated samples after treatment with cryolite) are completely extracted from the NdFeB magnet and have formed rare earth fluoride. No (intermetallic) phases consisting of both rare earth elements and Fe were found by XRD and EPMA. This proves the complete extraction of the rare earths from the magnet. Extracted Nd had been completely converted to NdF₃ after reaction with AlF₃, ZnF₂ and FeF₃. In the case of using cryolite as the fluorinating agent, Na_1.5Nd_1.5F₆ is formed instead of NdF₃. Dissolution of Na_1.5Nd_1.5F₆ in molten LiF during the electrochemical extraction process will lead to reduction of both Nd and Na on the cathode, due to the lower stability of NaF compared to NdF₃:   

\[{\rm NaF} \rightleftarrows {\rm Na} + 1/2{\rm F}_2 \quad \Delta G^{\rm o}(T = 1223\,{\rm K}) = 571\,{\rm kJ/mol}\](7)

  

\[{\rm NdF}_3 \rightleftarrows {\rm Nd} + 3/2{\rm F}_2 \quad \Delta G^{\rm o}(T = 1223\,{\rm K}) = 1665\,{\rm kJ/mol}\](8)

The decomposition voltages of NaF and NdF₃, which were calculated from the standard Gibbs energy of these reactions at their standard states, are −4.6 and −4.7 V for NaF and NdF₃, respectively. Na will therefore deposit on the cathode at lower voltages compared to Nd and co-deposition of Nd and Na will occur.

Both the XRD and EPMA results of the experiments with AlF₃ show that Al replaced Nd in the magnet, forming FeAl and FeAlB phases and no Al in the form of fluoride is present in the system. The same results can be seen using FeF₃ as fluorinating agent. All FeF₃ has reacted with Nd in the magnet and Fe has replaced the rare earths in the magnet matrix, thereby forming FeCo and Fe boride phases.

In the experiment with ZnF₂, XRD and EPMA results show that different Zn compounds have formed. The XRD result indicates the presence of the Li₂ZnF₄ phase. Besides this phase formation, EPMA analysis indicates that Zn has replaced Nd in the matrix, thereby forming FeZnB phase.

The results from the experiments with Na₃AlF₆ show that not all the cryolite has reacted with Nd, hence dissolution of the excess cryolite in LiF during the electrolysis process will lead to co-deposition of not only Na with Nd, but also Al.

Supposing that FeAl, FeB, FeZn and FeCo compounds (with varying stoichiometry) would not dissolve in the molten salt, it is expected that only rare earth deposition will occur during the electrolysis process. However, as was discussed earlier, in the case of cryolite treatment, co-deposition of Na and Al along with the rare earths can be expected.

Based on the thermodynamic calculations, the formation of boron fluoride appeared feasible using FeF₃ as the fluorinating agent. However, the experimental results analysed by EPMA mappings show that B and Fe are distributed in the same area, indicating the presence of Fe borides. The reason that boron had not reacted with iron fluoride can be the fact that stoichiometric amount of FeF₃ was used in the experiment and since REF₃ has higher stability than BF₃, REF₃ is formed and no excess iron fluoride is present for reaction with boron. The other reason might be related to the kinetics of the reactions.

6. Conclusion

This study shows that AlF₃, ZnF₂, FeF₃ and Na₃AlF₆ can act as strong fluorinating agents for extraction of Nd and Dy in the NdFeB magnet. In all the treatments, Nd was separated from the Fe and B in the magnet and was converted to rare earth fluoride. Al, Zn and Fe from AlF₃ (Na₃AlF₆), ZnF₂ and FeF₃, respectively, replaced the rare earth in the magnet matrix, thereby forming compounds with Fe. In the case of cryolite, co-deposition of Al and Na along with the rare earth elements, makes cryolite as fluorinating agent unsuitable for the recovery of rare earths from NdFeB magnet.

Once NdF₃ is formed, it can subsequently be electrolysed and extracted on the cathode by an electrolysis process in the same batch. Hence, conversion of the rare earth in the magnet to rare earth fluorides provide the possibility of a one-step recycling process of the rare earths from magnets. The findings of our study on the electrolysis process will be reported in the future.

Acknowledgment

This research has received funding from the European Community's Seventh Framework Programme ([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 authors` view, exempting the Community from any liability. Authors would also like to thank Ruud Hendrikx and Richard Huizenga for the XRD analyses.

REFERENCES

• 1)  K. Binnemans, P.T. Jones, B. Blanpain, T. Van Gerven, Y. Yang, A. Walton and M. Buchert: J. Clean. Prod. 51 (2013) 1–22. 10.1016/j.jclepro.2012.12.037
• 2)  O. Takeda, T.H. Okabe and Y. Umetsu: J. Alloy. Compd. 408–412 (2006) 387–390. 10.1016/j.jallcom.2005.04.094
• 3)  Y. Bian, S. Guo, L. Jiang, J. Liu, K. Tang and W. Ding: ACS Sustainable Chemistry & Engineering 4 (2016) 810–818. 10.1021/acssuschemeng.5b00852
• 4)  T. Nohira, S. Kobayashi, K. Kobayashi, R. Hagiwara, T. Oishi and H. Konishi: ECS Trans. 33 (2010) 205–212. 10.1149/1.3484777
• 5)  A.M. Martinez, O. Kjos, E. Skybakmoen, A.r. Solheim and G.M. Haarberg: ECS Trans. 50 (2013) 453–461. 10.1149/05011.0453ecst
• 6)  S. Riaño and K. Binnemans: Green Chem. 17 (2015) 2931–2942. 10.1039/C5GC00230C
• 7)  H.D. Bandara, K.D. Field and M.H. Emmert: Green Chem. 18 (2016) 753–759. 10.1039/C5GC01255D
• 8)  B. Mishra and D.L. Olson: J. Phys. Chem. Solids 66 (2005) 396–401. 10.1016/j.jpcs.2004.06.049
• 9)  W. Han, Q. Chen, Y. Sun, T. Jiang and M. Zhang: Metall. Mater. Trans., B 42 (2011) 1367–1375. 10.1007/s11663-011-9567-5
• 10)  Y. Castrillejo, M.R. Bermejo, P. Díaz Arocas, A.M. Martínez and E. Barrado: J. Electroanal. Chem. 575 (2005) 61–74. 10.1016/j.jelechem.2004.08.020
• 11)  P. Taxil, L. Massot, C. Nourry, M. Gibilaro, P. Chamelot and L. Cassayre: J. Fluor. Chem. 130 (2009) 94–101. 10.1016/j.jfluchem.2008.07.004
• 12)  S. Seetharaman, O. Grinder: US patent application nr. 12/991128, ref. no.: 12057 (2010).
• 13)  A. Abbasalizadeh, S. Seetharaman, L. Teng, S. Sridhar, O. Grinder, Y. Izumi and M. Barati: Journal Of Materials 65 (2013) 1552–1558. 10.1007/s11837-013-0752-7
• 14)  T.H. Okabe, S. Shirayama: Method and apparatus for recovery of rare earth element, US8323592 B2 (2012).
• 15)  A. Abbasalizadeh, L. Teng, S. Sridhar and S. Seetharaman: Mineral Processing and Extractive Metallurgy 124 (2015) 191–198. 10.1179/1743285514Y.0000000062
• 16)  T. Uda: Mater. Trans. 43 (2002) 55–62. 10.2320/matertrans.43.55