Enhanced Separation of Neodymium and Dysprosium by Nonaqueous Solvent Extraction from a Polyethylene Glycol 200 Phase Using the Neutral Extractant Cyanex 923.

Neodymium and dysprosium can be efficiently separated by solvent extraction, using the neutral extractant Cyanex 923, if the conventional aqueous feed phase is largely replaced by the green polar organic solvent polyethylene glycol 200 (PEG 200). While pure aqueous and pure PEG 200 solutions in the presence of LiCl or HCl were not able to separate the two rare earth elements, high separation factors were observed when extraction was performed from PEG 200 chloride solutions with addition of small amounts of water. This addition of water bridges the gap between traditional hydrometallurgy and novel solvometallurgy and overcomes the challenges faced in both methods. The effect of different variables was investigated: water content, chloride concentration, type of chloride salt, Cyanex 923 concentration, scrubbing agent. A Job plot revealed the extraction stoichiometry is DyCl3·4L, where L is Cyanex 923. The McCabe-Thiele diagram for dysprosium extraction showed that complete extraction of this metal can be achieved by a 3-stage counter-current solvent extraction process, leaving neodymium behind in the raffinate. Finally, a conceptual flow sheet for the separation of neodymium and dysprosium including extraction, scrubbing, stripping, and regeneration steps was presented. The nonaqueous solvent extraction process presented in this paper can contribute to efficient recycling of rare earths from end-of-life neodymium-iron-boron (NdFeB) magnets.

Introduction

Strong neodymium-iron-boron (NdFeB) permanent magnets are key components in a number of important electronic appliances and green energy technologies, such as computer hard drives, voice coil motors, wind turbines, and (hybrid) electric vehicles.[1−7] It is expected that the demand for NdFeB magnets will continue to grow to the extent that concerns have arisen over the long-term availability of the REEs. The supply risk and strategic and economic importance have lead the European Commission to include them in the list of critical raw materials.[8] Similarly, the U.S. Department of Energy stated that, among others, neodymium and dysprosium have the highest supply risk within the near and foreseeable future.[9] Recycling of end-of-life magnets can contribute to the future supply of REEs, thanks to the high concentration of REEs present in these magnets.[5,10] Neodymium (Nd) is the main REE component, and accounts for about 30 wt % of the mass of a magnet. Dysprosium (Dy) is often added in concentrations to about 10 wt % to enhance temperature stability and coercivity, mainly in applications where heat is generated during operations, as in wind turbines and electric motors. Direct reuse of NdFeB magnets is possible for large magnets such as the ones used in wind turbines, and direct recycling via the hydrogen decrepitation process is attractive for the recycling of magnets in hard disk drives or electric motors. In other cases, direct magnet-to-magnet recycling is not feasible and recovery of the REEs is the most viable option.[11,12] Both pyrometallurgical and hydrometallurgical processes have been developed, with the hydrometallurgical ones being the most versatile, because they allow to separate mixtures of REEs in the individual elements. The most useful process for separation on a larger scale is solvent extraction.[13] In the case of NdFeB magnet recycling, especially the separation of neodymium from dysprosium is important. Different researchers have investigated the Nd/Dy separation by solvent extraction, with the aim of improving the separation efficiency.[14−23] Moreover, as the use of green solvents is attractive and most recommended, ionic liquids[24,25] and supercritical carbon dioxide[26,27] were investigated as well.

However, some issues hamper further improvement of the sustainability of the REE separation. One example is the impossibility of using neutral extractants for REE extraction from chloride feed solutions. Neutral extractants (e.g., Cyanex 923 or TOPO) are preferred over acidic extractants because the consumption of chemicals for stripping of metals from the loaded organic phase and for pH control can largely be avoided. Chloride salts are preferred over nitrate salts because they are cheaper and allow for easier wastewater treatment.[28] The inefficient extraction of REEs from chloride aqueous solutions by neutral extractants is largely caused by the strong hydration of REE ions and the poor coordinating ability of chloride ions toward REE ions.[29] The issue of strong hydration of REE ions by water molecules can be mitigated by replacing water by a polar organic solvent as is done in nonaqueous solvent extraction.[30−36] Nonaqueous solvent extraction is a unit operation in solvometallurgy, which is analogous to hydrometallurgy but with replacement of the aqueous phase by an organic solvent.[30]

It is evident that selection of the polar organic solvent in nonaqueous solvent extraction has to be done very carefully and attention must be paid not only to the extraction performance but to the sustainability of the solvent as well.[30,37,38] For this reason, green solvents must be selected. One attractive candidate is polyethylene glycol 200 (PEG 200). PEG 200 is a green, sustainable solvent that is readily soluble in water in all proportions but is insoluble in nonpolar organic solvents.[39] Its risks having been well studied, it is generally considered to have very low toxicity, flammability, and environmental risks while being commercially available at a low cost. Besides, it is nonvolatile, readily biodegradable, and chemically stable to acids and bases. PEG 200 has been applied for the extraction of metal ions in aqueous biphasic systems[40−43] as well as in triphasic systems.[44] Although PEG 200 itself is, inherently, less sustainable than water, its remarkable properties and coordinating abilities allow the separation of rare earths to take place in fewer stages, while reducing acid consumption, by being able to choose a neutral extractant rather than acidic extractants conventionally used in rare-earth solvent extraction. This makes PEG 200 a suitable solvent to replace water (partially) in rare-earth solvent extraction.

The objective of this paper is to improve separation factors for REE extraction from chloride media using a neutral extractant, Cyanex 923, using a solvometallurgical approach, while bridging the gap between hydrometallurgy and solvometallurgy by the incorporation of the minimum possible amount of aqueous solutions into the water-miscible more polar organic phase. We developed a sustainable process for the separation of Dy(III) and Nd(III) by nonaqueous solvent extraction using PEG 200 in the more polar (MP) phase and Cyanex 923 as extractant in the less polar (LP) phase. Cyanex 923 is a commercial mixture of trialkyl phosphine oxides, with C6 and C8 chains.[45] It has the advantage over trioctylphosphine oxide (TOPO) that it is a liquid at room temperature, and it is a stronger extractant than tri-n-butyl phosphate (TBP).

Experimental Section

Chemicals

Polyethylene glycol with a number-average molecular mass of 200 g mol–1 (PEG 200) was obtained from J&K Scientific (Zedelgem, Belgium). NdCl3·6H2O (99.9%) was purchased from Strem Chemicals (Newburyport, USA), Nd2O3 (99.9%) from Alfa-Aesar (Geel, Belgium), DyCl3·6H2O (99.9%),) from abcr GmbH (Karlsruhe, Germany), and Dy2O3 from Strem Chemicals (Newburyport, USA). Cyanex 923 was supplied by Solvay (Toulouse, France). Oxalic acid (>99%) and LiCl (99.9%) were ordered from Sigma-Aldrich (Diegem, Belgium). Hydrochloric acid (37%) was purchased from VWR Chemicals (Haasrode, Belgium). The aliphatic hydrocarbon diluent Shell GTL Solvent GS190 (GS190), composed of C10–C13n- and iso-alkanes with a boiling range of 187–218 °C, was supplied by Shell (Rotterdam, Netherlands). 1-Decanol (99%) was ordered from Acros Organics (Geel, Belgium). 1-Butanol (99%) was purchased from Thermo Fisher Scientific (Geel, Belgium). The silicone solution in isopropanol was supplied by SERVA Electrophoresis GmbH (Heidelberg, Germany). The gallium and scandium standard (1000 mg L–1 in 2–5% HNO3) were obtained from Chem-Lab NV (Zedelgem, Belgium). Triton X-100 was obtained from Merck KGaA (Darmstadt, Germany). Water was always of ultrapure quality, deionized to a conductivity of <0.055 μS cm–1 (298.15 K) with a Merck Millipore Milli-Q Reference A+ system. All chemicals were used as received without any further purification.

Feed Solutions

In a first series of experiments, the feed solutions were prepared using rare-earth chloride salts to optimize the feed conditions, i.e. the amount of water, concentration of acid, and concentration of salting-out agent. The concentrations of Dy(III) and Nd(III) in these feed solutions were 4 g L–1 for both REEs. In a next step, the optimized parameters were applied on simulated feed prepared by using rare-earth oxides, with concentrations mimicking those found in NdFeB magnet leachate, i.e., 12.2 g L–1 Nd(III) and 1.1 g L–1 Dy(III). The oxides were directly dissolved in mixed PEG 200/aqueous HCl solution to prepare the feed solution.

Solvent Extraction Procedure and Instrumentation

The more polar (MP) organic phase consisting of a PEG 200 solution of REEs and the less polar (LP) organic phase containing Cyanex 923 diluted with GS190 and modifier (10 vol % 1-decanol) were mixed using a wrist action shaker at ambient temperature (21 ± 1 °C) for 60 min, which is sufficient to attain equilibrium. The MP:LP phase ratio was 1:1, unless stated otherwise. After reaching equilibrium, the phase separation was accelerated by centrifugation (Heraeus Labofuge 200, Thermo Scientific, United States). After proper dilution with a 5 vol % Triton-X100 solution, the REE concentrations in the PEG 200 phase were analyzed by a S2 Picofox TXRF spectrometer (Bruker, Germany), equipped with a molybdenum X-ray source and operated at a voltage of 50 kV.[46,47] A certain amount of 1000 mg L–1 gallium standard solution was added as internal standard. Quartz glass carriers were pretreated with 30 μL of SERVA silicone solution in isopropanol to siliconize the carrier, in order to avoid the spreading of the sample, and subsequently dried at 60 °C for 5 min. A sample of 2 μL was placed on the pretreated quartz glass carrier and dried at 60 °C for 30 min. All samples were measured in duplicate.

The distribution ratio D is defined as the ratio of the total metal concentration cLP in the LP organic phase (Cyanex 923) to the total metal concentration cMP in the MP organic phase (PEG 200) at equilibrium and is represented as

The percentage extraction (%E) is the total amount of metal transferred to the LP organic phase divided by the total amount of metal in the MP organic feed, expressed as a percentage. With applying the definition of the distribution ratio and defining VMP and VLP as the volume of the MP and LP phase, respectively, %E is defined as

The separation factor α is the quotient of distribution ratio of a metal A to the distribution ratio of a metal B and is defined as

Stripping

Equal volumes of an aqueous oxalic acid solution and the loaded LP phase were contacted for 60 min at room temperature (21 ± 1 °C). The concentration of oxalic acid used in stripping was the exact stoichiometric amount needed for precipitation of all rare earths in the LP phase. For a LP phase containing 0.9 g L–1 Dy(III), this corresponds to 8.3 mmol L–1. After reaching equilibrium, the phase separation was accelerated by centrifugation (Heraeus Labofuge 200, Thermo Scientific, United States). The remaining Dy(III) concentration in the LP phase was determined via inductively coupled plasma-optical emission spectroscopy (ICP-OES), performed with an Avio 500 spectrometer (PerkinElmer, United States) equipped with a GemCone low-flow nebulizer, baffled cyclonic spray chamber, alumina injector, and PerkinElmer Hybrid XLT torch. The samples of the LP phase, the calibration solutions, the blank solution, and quality controls were diluted in 1-butanol, with 5 ppm of scandium added as internal standard. The line at 353.170 nm for Dy(III) was measured in axial viewing mode.

Viscosity Measurements

The viscosity of the feed solutions with varying water content was measured using a rolling-ball viscometer (Anton Paar LOVIS 2000 M/ME, Austria). Density and viscosity were measured simultaneously, which allowed the software to calculate the dynamic viscosity. The capillaries had a diameter of 1.8 mm for samples with a water content up to 30 vol % and 1.59 mm for higher water content. Gold-coated steel balls (7.88 g cm–3) were used. The operating temperature was 25 °C; the angle was 45°.

Results and Discussion

Bridging the gap between hydrometallurgy and solvometallurgy often requires addition of a fraction of water, albeit at a concentration lower than 50 vol %, as a solution having a higher water content is to be considered a dilute aqueous solution of organic solvent.[30] The addition of water in a nonaqueous system can have several advantages. First, rare-earth oxides are not soluble in appreciable amounts in most of the organic solvents. They need to be converted into rare-earth salts in order to dissolve them and to prepare the feed for solvent extraction studies. Therefore, a limited amount of aqueous acid is added to the MP organic phase, in this case PEG 200, enabling the dissolution of rare-earth oxides. Another possibility is to add the aqueous leachate containing a rare-earth chloride mix directly to the MP organic phase, thus facilitating integration into hydrometallurgical flow sheets. Besides, addition of water helps to lower the viscosity of the MP organic phase (Figure ).

Figure 1

Influence of water content on the viscosity of PEG 200 feed solutions. [Nd] = 12 g L–1, [Dy] = 1 g L–1, [HCl] = 2 mol L–1. The PEG 200 + 0 vol % water sample does not contain HCl. Temperature: 25 °C.

Being able to introduce a certain amount of water in the solvometallurgical system gives one an extra degree of freedom to optimize the process parameters. The effect of water was therefore studied by varying its concentration from 0 to 100 vol % in PEG 200, with a constant chloride concentration (Figure ). The chloride source was either LiCl or HCl, both at a concentration of 1 mol L–1. The feed concentration was 4 g L–1 for both Nd(III) and Dy(III), and the Cyanex 923 (C923) concentration was 1 mol L–1. The Nd/Dy separation by solvent extraction is clearly not possible from either pure PEG 200, which resulted in 100% extraction of both rare earths, or from pure water, in which case barely any rare earths were extracted. Addition of water to PEG 200 resulted in a decrease of the percentage extraction, with the decrease for Nd(III) being stronger than for Dy(III). This decrease can be explained by differences in solvation in PEG 200 and aqueous solutions. In aqueous solutions, rare-earth ions strongly coordinate to water molecules; in PEG 200 solutions also chloride ions can enter the first coordination sphere, replacing water.[35,48] For instance, according to Rogers et al., Nd(III) and Dy(III) species in tri(ethylene) glycol-chloride solutions appeared to exist as [MCl2(OH2)(EO3)]Cl2, with M = Nd, Dy and EO3 = tri(ethylene) glycol.[48] Since it is easier to form the neutral lanthanide chloride salt complex in nonaqueous PEG 200 solutions, extraction of these ions by Cyanex 923 is enhanced. Mixtures of PEG 200 and 30–40 vol % water allowed us to separate Nd(III) and Dy(III) well, with separation factors of 69 and 54 for the LiCl and HCl systems, respectively. The effect of addition of water to PEG 200–HCl was found to be larger than the effect on the PEG 200–LiCl system, over the complete range of concentrations. This observation can be explained by the competition of metal ion extraction and the efficient extraction of mineral acids, such as HCl, from aqueous solutions by Cyanex 923.[49]

Figure 2

Effect of water on the extraction of Nd(III) and Dy(III) from PEG 200 by Cyanex 923 at constant chloride concentration. Feed: [Nd] = [Dy] = 4 g L–1, [LiCl] or [HCl] = 1 mol L–1, [C923] = 1 mol L–1 in GS190 (+10 vol % 1-decanol). Temperature: 21 ± 1 °C.

Both for LiCl and HCl, the effect of the chloride concentration was studied at two different conditions: PEG 200 + 30 vol % water and 40 vol % water. The studied chloride concentrations for PEG 200 + 30 vol % water were 0.1–3 mol L–1, while for PEG 200 + 40 vol % water this was 0.1–4 mol L–1. Figure a shows that percentage extraction for both REEs increased up to 3 mol L–1 of LiCl, after which extraction slightly decreased. The same positive trend for Dy(III) extraction was observed with increasing HCl concentrations (Figure b). However, the Nd(III) extraction efficiency remained negligible, even at high HCl concentrations, which is in contrast to the effect of LiCl. As a result at 2 mol L–1 HCl and 30 vol % water, a separation factor of almost 260 was obtained. In comparison, at 0.5 mol L–1 LiCl and 30 vol % water, a separation factor of 38 was achieved. With regard to the water content, 30 vol % water was found to be the optimal concentration, a further increase in the water concentration led to a decrease in extraction efficiency and separation factor.

Figure 3

Effect of (a) LiCl or (b) HCl concentration on the extraction of Nd(III) and Dy(III) from PEG 200 + 30 and 40 vol % water. Feed: [Nd] = [Dy] = 4 g L–1, [LiCl] or [HCl] = 0.1–4 mol L–1, [C923] = 1 mol L–1 in GS190 (+10 vol % 1-decanol). Temperature: 21 ± 1 °C.

The influence of extractant concentration on the separation was studied from both LiCl and HCl media. Figure clearly shows that the separation of Nd(III) and Dy(III) was better for extraction from HCl media, as there was almost no coextraction of Nd(III), resulting in a separation factor larger than 6000 at 0.5 mol L–1 Cyanex 923. In LiCl media, the maximum attainable separation factor was 21. At higher Cyanex 923 concentrations, the separation factor decreases as the extraction of Dy(III) levels off, while Nd(III) extraction increases. At 0 mol L–1 Cyanex 923, the 1-decanol/GS190 mixture cannot extract the rare-earth ions.

Figure 4

Effect of Cyanex 923 concentration on the extraction of Nd and Dy from PEG 200 + 30 vol % water solution. Feed: [Nd] = [Dy] = 4 g L–1, [LiCl] = 0.5 mol L–1 or [HCl] = 2 mol L–1, [C923] = 0–1 mol L–1 in GS190 (+10 vol % 1-decanol). Temperature: 21 ± 1 °C.

To find out the extraction mechanism, the method of continuous variation was used, creating the Job plot displayed in Figure .[50] This plot is constructed by varying the feed concentration of Dy(III) with regard to the concentration of Cyanex 923 in the less polar phase, while keeping the total concentration constant, i.e., [Dy(III)] + [C923] = 0.5 mol L–1. As a result, the X-axis shows the variation of the mole fraction of Dy(III), calculated aswhere nDy and nC923 are the number of moles of Dy(III) and Cyanex 923, respectively. The Y-axis corresponds to the concentration of Dy(III) in the LP phase, and thus the concentration of Dy–C923 complexes in the LP phase. The experiment has been performed using different MP phases, containing 10, 30, and 50 vol % water. A least-squares linear fit through the first three data points (R2 = 0.955, 0.979, and 0.981 for 10, 30, and 50 vol % water, respectively) and the final 6 data points (R2 = 0.992, 0.992, and 0.995 for 10, 30, and 50 vol % water, respectively) has been used for construction of the plot. The maximum very close to XDy = 0.2 indicates that four molecules of Cyanex 923 are coordinating to Dy(III), which agrees with earlier observations for the extraction of Yb(III) from ethylene glycol chloride solutions by Cyanex 923.[31] The following extraction mechanism is thus proposedHere, the overbar indicates the species in the less polar phase, L represents Cyanex 923 and x, y the stoichiometric number of PEG and water molecules coordinating to Dy(III) in the more polar phase, respectively. Furthermore, the water content in the MP phase has little to no influence on the extraction mechanism, as the maxima of the three studied conditions coincide. In other words, the difference in extraction efficiency can only be ascribed to a difference in speciation of the rare-earth ion in the MP phase, as suggested earlier, rather than a difference in extraction mechanism.

Figure 5

Job plot for Dy(III) extraction. Conditions: [C923]+[Dy(III)] = 0.5 mol L–1, PEG 200 + 30 vol % water, [HCl] = 2 mol L–1, Cyanex 923 dissolved in GS190 (+ 10 vol % 1-decanol). Temperature: 21 ± 1 °C.

Extraction from Simulated Rare-Earth Oxide Feed

On the basis of the optimized extraction parameters, a new solvent extraction process was developed, based on a feed composition which resembles the concentration in real pregnant leach solutions from NdFeB magnet leaching. A synthetic feed solution with 12.2 g L–1 Nd(III) and 1 g L–1 Dy(III) was prepared starting from rare-earth oxides. The increased Nd/Dy molar ratio compared to the previous experiments resulted in a decrease of Dy(III) extraction efficiency and an increase of Nd(III) coextraction, as can be seen in Figure . At 0.5 mol L–1, 72% of Dy(III) was extracted, and a separation factor of 42 was attained.

Figure 6

Effect of Cyanex 923 concentration on the extraction of Nd(III) and Dy(III) from PEG 200 + 30 vol % water. Feed prepared from oxides: [Nd] = 12.2 g L–1, [Dy] = 1.1 g L–1, [HCl] = 2 mol L–1, [C923] = 0.1–1 mol L–1 in GS190 (+10 vol % 1-decanol). Temperature: 21 ± 1 °C.

A McCabe–Thiele diagram was constructed to determine the number of theoretical stages required to separate Dy(III) from Nd(III) with the optimized system. It was estimated that at least three stages would be needed for the extraction of Dy(III) using a MP:LP phase ratio of 1:1 (Figure ).

Figure 7

McCabe-Thiele diagram for Dy(III) extraction isotherm. Feed prepared from oxides: [Dy] = 1.1 g L–1 ([Nd] = 12.2 g L–1), [HCl] = 2 mol L–1, [C923] = 0.5 mol L–1 in GS190 (+10 vol % 1-decanol). MP:LP phase ratio = 7:1–1:7. Temperature: 21 ± 1 °C.

To further purify the extracted Dy(III), the coextracted Nd(III) needs to be removed by a scrubbing step. A loaded LP phase was created by contacting a PEG 200 feed solution with 0.5 mol L–1 of Cyanex 923, at identical conditions as mentioned in Figure . Around 6% of Nd(III) was coextracted (0.7 g L–1), while 72% of Dy(III) was extracted (0.9 g L–1). Several PEG 200-HCl solutions were investigated for the removal of coextracted Nd(III) (Table ). While the PEG 200–HCl scrub feed without addition of Dy(III) could remove the coextracted Nd(III) quantitatively, there was a considerable loss of Dy(III). A scrub feed containing either 0.9 or 1.8 g L–1 Dy(III) removed the coextracted Nd(III) as well, while avoiding Dy(III) loss. Changing the phase ratio to MP:LP = 1:2 decreased somewhat the percentage scrubbing for Nd(III) but offered a possibility to increase the concentration of Nd(III) in the scrub raffinate. Ideally, the concentration of HCl should be 2 mol L–1, so that the scrub raffinate can be added to the initial feed solution. A small number of scrubbing stages will eventually remove Nd(III) quantitatively, using the optimized parameters: MP:LP = 1:2, [HCl]scrub = 2.0 mol L–1, [Dy(III)]scrub = 0.9 g L–1.

Table 1

Scrubbing Studies for Removal of Co-Extracted Nd(III)a

Phase ratio (MP:LP)	[HCl]scrub, mol L–1	[Dy(III)]scrub, g L–1	%Scrubbing, Nd(III)	%Stripping, Dy(III)
1:1	0.5	0.0	99.3	65.1
	1.0	0.0	100	46.6
	2.0	0.0	96.4	24.4
	1.0	0.9	99.0	b
	2.0	1.9	99.5	b
1:2	1.0	0.0	77.3	23.5
	1.0	0.9	81.0	14.8
	1.0	1.9	79.7	b
	2.0	0.9	81.1	b
	2.0	1.9	81.5	b

The variables within the PEG 200 scrub solution were: HCl concentration, Dy(III) concentration, and MP:LP phase ratio. The loaded LP phase contained [Nd] = 0.7 g L–1 and [Dy] = 0.9 g L–1. Temperature: 21 ± 1 °C.

Mass balance calculations based on scrub raffinate concentrations indicated that there was net extraction of Dy(III) from the scrub feed, hence no stripping percentage can be mentioned.

Eventually, the Dy(III) in the scrubbed LP phase can be precipitated quantitatively by use of an aqueous oxalic acid solution. A LP phase containing 0.9 g L–1 Dy(III) was contacted with the exact stoichiometric amount of oxalic acid aqueous solution (8.3 mmol L–1) at a phase ratio of MP:LP = 1:1, for 1 h at room temperature (21 ± 1 °C). After being stripped, the LP phase contained just 0.13 mg L–1 (RSD 4.8%) of Dy(III), corresponding to a stripping efficiency of >99.99%. The oxalate precipitate can be filtered off, dried, and calcined to produce pure Dy2O3 for reuse in magnets.

Recycling Studies

Reuse of the LP phase (0.5 mol L–1 Cyanex 923, 10 vol % 1-decanol in GS190 solvent) was investigated for six extraction-stripping cycles. The LP phase was contacted with a feed containing 12.2 g L–1 Nd(III) and 1.1 g L–1 Dy(III) in a separatory funnel for 1 h, after which it was stripped with oxalic acid as described in a previous section. As HCl from the MP phase is likely partially extracted by Cyanex 923, 2 contacts with pure water at a phase ratio MP:LP = 1:1 were performed after every stripping cycle, with each washing step taking 15 min.[49,51] The percentage Dy(III) and Nd(III) extracted during the first cycle was 71% (RSD: 0.3%) and 5% (RSD: 0.9%), respectively. The percentages Dy(III) and Nd(III) extracted were decreased slightly for the second cycle and were maintained constant over the next four cycles at a value of 68% (RSD: 5.1%) and 4% (RSD: 2.0%), respectively. Stripping percentage exceeded 99.9% in all stripping cycles. As expected, pH of the aqueous wash solutions decreased slightly as HCl was stripped back, with 5.9 for the first washing step, while for the second washing step, the pH was about neutral. These observations prove the recycling capacity and thus the stability of the extraction system.

Conceptual Flow Sheet

A conceptual flow sheet for the nonaqueous solvent extraction process based on above optimizations is presented in Figure . A mixture of Nd(III) and Dy(III) can be added as a rare-earth oxide, by dissolving it in PEG 200 + HCl, or can be added as an aqueous rare-earth chloride solution. Concentrations of HCl and water must be adjusted to the optimized values as mentioned above. Three stages of extraction are required for complete extraction of Dy(III). The coextracted Nd(III) can be removed easily using few stages at a phase ratio of MP:LP = 1:2 and a Dy(III) concentration of 0.9 g L–1. The extracted Dy(III) is stripped by a limited amount of oxalic acid from the loaded LP phase, generating pure Dy(III) oxalate precipitate. Nd(III) remaining in the raffinate cannot be precipitated directly by oxalic acid, as rare-earth oxalates are soluble in PEG 200. Instead, it can be extracted by Cyanex 923 in one stage, after (partial) removal of water, bringing back the water concentration between 0 vol % and 10 vol % (Figure ), followed by precipitation stripping by oxalic acid. Excess water can be removed by traditional distillation or by pervaporation using polymeric membranes,[52−55] ceramic membranes,[56] or zeolites.[57] Pervaporation is a green alternative for conventional distillation that has been suggested to be used in dehydration of ethylene glycol–polyethylene glycol mixtures after their synthesis from ethylene oxide, and for the dewatering of di(ethylene) glycol and tri(ethylene) glycol, used in natural gas dehydration. Dewatering is needed for a second reason: preparing a new feed solution requires the addition of the rare earths mixture as aqueous chloride solution or as rare-earth oxides, in which case fresh aqueous HCl solution needs to be added. In both cases, water is added in the process, requiring a partial removal of water prior to reuse of the PEG 200. The LP phase is regenerated after removing the extracted HCl by contacting it with an aqueous phase and can be reused in subsequent extraction cycles as well.[51] In summary, this process does not eliminate water but limits its use during extraction while providing better separation in less stages. Moreover, the possibility of working with neutral extractants limits the base consumption for pH control and excessive acid consumption during stripping.

Figure 8

Conceptual flow sheet for separation and recovery of Nd(III) and Dy(III) from a mixed-oxide feed.

Conclusion

While it is not possible to separate Nd(III) and Dy(III) from aqueous chloride solution by solvent extraction with Cyanex 923, present work has shown that high separation factors can be obtained when these rare-earth ions are extracted from the PEG 200 chloride solution as solvent in the more polar phase when small amounts of water are added. The separation factor increases even further when HCl is used as a chloride source, instead of LiCl. A solvometallurgical process was developed for feed solutions mimicking the real NdFeB magnet composition, which can be integrated in existing hydrometallurgical flow sheets, either through the dissolution of a mixed-oxide or the addition of an amount of concentrated aqueous rare-earth solution. The process comprises three extraction stages for the recovery of Dy(III), followed by at least one stage of scrubbing and one of stripping with oxalic acid, to recover pure Dy(III) oxalate. Eventually, this paper proposes an efficient way of separating Dy(III) and Nd(III) followed by recovery as rare-earth oxalate, which in turn can be calcined to obtain the much needed rare-earth oxides that can be used directly in the production of NdFeB magnet alloy.