A novel application of hematite precipitation for high effective separation of Fe from Nd-Fe-B scrap

Rare earths, e.g. neodymium (Nd), praseodymium (Pr) and dysprosium (Dy), are abundant in the rare earth sintered magnet scrap (Nd-Fe-B scrap), but their recycling is tedious and costly due to the high content of impurity Fe. Herein, a novel approach was developed to effectively recycle rare earths from the scrap via an integrated acid dissolution and hematite precipitation method. The scrap contained 63.4% Fe, 21.6% Nd, 8.1% Pr and 3.9% Dy. It was dissolved in nitric, hydrochloric and sulfuric acids, separately. Nearly all impurity Fe in the scrap was converted to Fe3+ in nitric acid but was converted to Fe2+ in hydrochloric and sulfuric acids. After hydrothermal treatment, the rare earths in the three acids were almost unchanged. From nitric acid, 77.6% of total Fe was removed, but total Fe was not from the hydrochloric and sulfuric acids. By adding glucose, the removal of total Fe was further increased to 99.7% in nitric acid, and 97% of rare earths remained. The major mechanism underlying total Fe removal in nitric acid was the hydrolysis of Fe3+ into hematite, which was promoted by the consumption of nitrate during glucose oxidation. This method effectively recycled rare metals from the waste Nd-Fe-B scrap and showed great potential for industrial application.

the formed Fe 3+ oxyhydroxide generated abundant hydrogen groups, in which rare earths could be coordinated, thereby resulting in low levels of dissolved rare earths in the solution.
When Fe 3+ oxyhydroxide was converted to the well-crystallised Fe oxides, the two adjacent Fe-OH bonds on Fe 3+ oxyhydroxide were dehydrated to form the Fe-O-Fe bond 18,19 , and the average number of coordination sites on Fe 3+ oxyhydroxide decreased 20,21 , thereby subsequently reducing the precipitation of rare earths. He et al. reported that 90.7% of Fe 3+ was eliminated as hematite when the Fe 3+ /Zn 2+ -bearing sulfuric acid solution was hydrothermally treated at 210 °C for 2 h with the addition of H 2 O 2 22 . Despite the effective removal of Fe 3+ , the Fe 3+ residual was still high (nearly 1,500 mg/L) 23 and needed to be removed before rare earth extraction.
In this study, an integrated acid dissolution and hematite precipitation method was developed for the effective removal of the impurity Fe from scrap. After the scrap's dissolution in nitric acid, 99.7% of total Fe was hydrothermally converted to hematite with the addition of glucose. Meanwhile, more than 97.1% of rare earths remained. This is the first report on the effective removal of impurity Fe from a rare earth-bearing solution with high rare earth retention.

Results and Discussion
After the scrap was dissolved in the nitric, hydrochloric and sulfuric acids, the generated acidic solutions were designated as Nitric-A, Chloric-A and Sulfuric-A, respectively. The concentrations of rare earths and total Fe (including Fe 2+ and Fe 3+ ) were similar in the three acids, as shown in Fig. 1(a,b). However, in Nitric-A, Fe 2+ was only 54.9 mg/L, whereas Fe 3+ was about 10,038 mg/L, as shown in Fig. 1(b), thereby indicating that Fe 3+ predominated in the total Fe in Nitric-A. In comparison with Nitric-A, Fe 2+ was approximately 10,000 mg/L in both Chloric-A and Sulfuric-A, as shown in Fig. 1(b), thereby suggesting that Fe 2+ was rich in Chloric-A and Sulfuric-A due to the lack of oxidising agent (e.g. nitrate).
After hydrothermal treatment, the concentrations of rare earths were almost unchanged in the three acids, as shown in Fig. 2(a). However, in Nitric-A, the total Fe concentration decreased from 10,093 mg/L to 2,257 mg/L, corresponding to 77.6% of the total Fe removal rate, as shown in Fig. 2(b). Meanwhile, the solution pH slightly decreased from 0.38 to 0.19, as shown in Fig. 2(c), due to the generation of H + from the hydrolysis of Fe 3+ . The hydrolysed Fe 3+ was in irregular form with the uniform distribution of element Fe and sparse distributions of Nd, Pr and Dy (Fig. 3), demonstrating that element Fe was dominant in the generated particles. Moreover, only indicative peaks of hematite (JSCPDS 33-0664) were observed in the curve of the generated particles (Fig. 4), indicating that Fe 3+ was hydrolysed in the form of well crystallised hematite. Compared with Nitric-A, the total Fe concentrations in Chloric-A and Sulfuric-A were constant, as shown in Fig. 2(b), suggesting that the oxidation and hydrolysis of Fe 2+ did not occur.
To further remove the total Fe from Nitric-A, glucose was introduced. Glucose's efficiency to remove Fe is shown in Fig. 5. After hydrothermal treatment, the retention rates of rare earths were 98.4% for Nd, 97.5% for Pr and 97.1% for Dy, as shown in Fig. 5(a), and these rates were similar to those obtained without glucose ( Fig. 2(a)). However, the removal rate of total Fe increased to 99.6% ( Fig. 5(b)), much higher than that without glucose ( Fig. 2(b)), indicating that glucose was important factor for total Fe removal without losing rare earths. With glucose, the total Fe was removed as hematite particles ( Fig. 6(a)), similar to the product generated without glucose (Fig. 4), but with the average diameter of 80-100 nm, as shown in Fig. 7(a). During the process, the pH increased slightly from 0.24 to 0.71 (Fig. 5(c)), whereas total organic carbon dramatically decreased from 3,465 mg/L to 39.2 mg/L ( Fig. 5(d)). Moreover, the nitrate concentration considerably decreased from 80.9 g/L to 8.08 g/L (Fig. 5(e)). These findings demonstrated that redox reaction occurred between glucose and nitrate, in which abundant glucose was oxidised by nitrate to CO 2 and H 2 O with the consumption of H + .
In comparison with Nitric-A, no apparent change in total Fe concentration was observed in Chloric-A and Sulfuric-A after hydrothermal treatment with glucose, and only a few spherical particles with average diameters of 3-5 μm ( Fig. 7(b,c)) were precipitated. The spherical particles showed an extremely broad XRD peak at 2θ = 23.2° ( Fig. 6(b,c)) that probably belonged to the carbon sphere generated from the dehydration and polymerisation of glucose, similar to the hydrothermal product of glucose reported by Mi et al. 24 . www.nature.com/scientificreports www.nature.com/scientificreports/ This approach exhibited a high removal rate (99.6%) of total Fe in the recycling of rare metals from the rare earth-bearing scrap. This rate was higher than that obtained through other reported processes, such as the complex leaching and electrolysis process with addition of H 2 SO 4 and MnO 2 25 , the extraction processwithtri-n-butyl     26 , the mechano-chemical treatment with HCl and (COOH) 2 27 and selective leaching with nitric acid 2 , sulfuric acid 28 and ascorbic acid 29 (Table 1). When the scrap was dissolved in hydrochloric and sulfuric acids separately, Fe 2+ -bearing solutions were generated. Fe 2+ was stable in the two acids and not hydrolysed during the hydrothermal process, thereby resulting in low removal rates of total Fe. However, with nitric acid dissolution, Fe 3+ was generated from the oxidation of impurity Fe in scrap by nitrate. The generated Fe 3+ was further hydrothermally hydrolysed to hematite with  www.nature.com/scientificreports www.nature.com/scientificreports/ generation of nitric acid when the temperature increased to 160 °C (Eq. (1)) 30,31 . Hematite was well crystallised and had a protonated surface at pH <1 32 in which the net surface charge on its surface was positive and blocked the adsorption of metal ions, such as Nd, Pr and Dy.
As the reaction continued, Fe 3+ was hydrolysed to produce a large amount of nitric acid. An increase in nitric acid concentration decreased the solution pH from 0.38 to 0.19 (Fig. 2(c)) and shifted the hydrolysis equilibrium to the left (Eq. (1)) 30 , resulting in a decrease of Fe 3+ hydrolysis. Therefore, residual total Fe at a concentration of 2,257 mg/L was left in the solution (Fig. 2(b)).
Glucose was hydrothermally oxidised by nitrate to generate levulinic acid and 5-hydroxymethylfurfural, which were further oxidised to CO 2 and H 2 O via Eq. (2). With the oxidisation of glucose, nitrate was hydrothermally reduced to N 2 , and its concentration apparently decreased from 80.9 g/L to 8.08 g/L (Fig. 5(e)), thereby promoting Fe 3+ hydrolysis. Moreover, H + was also involved in the redox reaction between glucose and nitrate (Eq. (2)). Thus, the solution pH increased from 0.19 to 0.71 (Fig. 5(c)), which further accelerated the formation of hematite 33 . During glucose oxidisation, the generated intermediates levulinic acid and 5-hydroxymethyl furfural were electrostatic adsorbed on the positively charged surface of hematite particles, thereby inhibiting the aggregation and crystal growth of hematite particles 34 and resulting in smaller hematite particle sizes than without glucose.

Materials and Methods
Nd-Fe-B scrap. Nd-Fe-B scrap was acquired from the calcinator of a local alumina refinery in Jilin, China.

Separation of Fe from Nd-Fe-B scrap. Dissolution of Nd-Fe-B scrap.
The scrap was dissolved in acids as follows. Scrap (5 g) was dispersed in 250 mL of 3 M nitric acid under constantly stirring at 150 rpm overnight. A yellowish solution was generated and denoted as Nitric-A. The control experiments were also performed using 3 M hydrochloric and sulfuric acids, and the generated acidic solutions were designated as Chloric-A and Sulfuric-A, respectively.
Separation of Fe. Impurity Fe was separated from the acids via a one-step hydrothermal method. Nitric-A at 20 mL was transferred to a 50 mL Teflon kettle, hydrothermally treated at 160 °C for 10 h and cooled down to room temperature. The reddish particles generated at the bottom of kettle, were collected and vacuum-dried at 50 °C for 20 h before characterisation. The Chloric-A and Sulfuric-A were also hydrothermally treated, respectively, but no deposit was generated.
To further remove total Fe in Nitric-A, glucose was added at the glucose/total Fe molar ratio of 0.7. After agitating at 150 rpm for 5 min, a brownish suspension was generated. The suspension was hydrothermally treated according to the procedure described above. The obtained particles were collected for characterisation. Glucose was added to Chloric-A and Sulfuric-A with the same treatment as that used on Nitric-A, and the generated particles were collected and characterised separately.
Each experiment was performed thrice, and the averaged date was reported.  Table 1. Comparison of the removal rate of total Fe and the retention rate of rare earths.