High Temperature Recovery of Rare Earth Ortho-Ferrites from Permanent Magnets

Abstract

There is a growing interest in Magnetoelectric (ME) materials in view of both fundamental understanding and novel desirable applications. The application potential for multiferroic materials with the possibility of reversing the magnetization by applying an electric field (or vice versa) is quite promising especially in data storage application. The rare earth ortho-ferrites (REFeO₃), such as NdFeO3 , have been studied extensively in recent study in the context of discovery of magneto-electric (ME)/multiferroic nature materials. In this paper a potential high temperature recovery of rare earth elements in the form of REFeO₃ from waste permanent magnets using direct oxidation in air was investigated. It was found that REFeO₃ (NdFeO₃) phase can be separated from the other oxide phases by thermal treatment of NdFeB magnet waste at 1100 °C in air atmosphere. The REFeO₃ also found to have different physical characteristic (density) from the other layers and can be separated physically. This paper will include a proposed flowsheet to recover Nd as NdFeO₃ through powder separation . The technique can be beneficial and favourable for RE recycling and recovery from waste permanent magnets .

Introduction

REFeO₃ have attracted special attention and have been extensively studied for many applications due to the great deal of various phase transitions and high-frequency phenomena [1]. One of the most common perovskite oxides, NdFeO₃ nanocrystal, is very well known due to having a wide variety of applications in, for instance, magnetic materials [2, 3], gas sensors for CO [4], arsenate adsorption [5] and H₂S detection [6]. REFeO₃ crystallize in the distorted perovskite structure (orthorhombic space group: Pnma) of general formula ABO₃, where B is usually a 3d transition metal surrounded by six oxygen atoms in octahedral coordination and A is normally a rare earth cation, 12-coordinate by oxygen atoms. The structure can be described as a three dimensional network of FeO₆ octahedra surrounding RE atoms, where Fe–O–Fe angle deviates from 180. This angle directly depends on RE size: the larger the RE is, the more the chains of octahedra stretch, and the Fe–O–Fe bond angle approaches 180, and the lattice tends to be cubic. This angle directly drives distance between oxygen 2p orbitals and iron 3d orbitals, which are mainly responsible of electrical and optical properties [7].

REFeO₃ possesses two sets of magnetic sublattices: one consists of 4f-electron of the R ions, and the other is made up of 3d-electrons of the iron ions (Fe). The R-sublattice only orders antiferromagnetically at much lower temperature . Meanwhile, due to the Dzyaloshinskii–Moriya (DM) interaction, the Fe spins in REFeO₃ are slightly canted at relatively higher temperatures and, as a result, a spontaneous weak magnetization is presented on a macro-scale [2]. This weak magnetization exhibits a combination of the static ferromagnetic properties with the dynamical antiferromagnetic properties, which consequently produces some unique magnetic phenomena, such as excellent magneto-optical coefficient and magnetic phase transition [8, 9].

The preparation of rare earth orthoferrites have been achieved by many methods, including hydrothermal, combustion, sol-gel, precipitation methods and sonication assisted precipitation [6]. However, it is difficult to obtain pure-phase REFeO₃ from the high-temperature synthesis methods [10]. The change in microstructure and oxidation mechanism of the NdFeB magnet in air at high temperature has been discussed by the author in previous publications [11, 12], the oxidation resulted in the formation of an external scale (external oxidation zone—EOZ) and development of subsurface layers containing complex mixtures of phases and solid solutions. It was observed that the oxidation in air at temperature range 1300–1500 K resulted in the development of a layer rich with distinct NdFeO₃ phase. The layer consists of elongated NdFeO₃ of around 40 µm interspersed with pockets of less than 20 µm Fe₂O₃. Based on the in situ XRD and thermodynamic prediction described in the publication [13], the equilibrium product should consist of 3 phases: Fe₂O₃, NdFeO₃, and NdBO₃. The proposed mechanism [11] predicted that separate layer of NdFeO₃, NdBO₃ and Fe₂O₃ could potentially be obtained under this temperature range (1300–1500 K), if the oxidation is allowed to occur indefinitely thus the system is allowed to reach equilibrium, as shown schematically in Fig. 1. In this case NdFeO₃ micron sized powders can be acquired through powder processing and separation . The information of the physical properties for the possible phases in the oxidised layers that is important for separation is listed in Table 1.

../images/468727_1_En_64_Chapter/468727_1_En_64_Fig1_HTML.gif — Fig. 1
A schematic diagram showing the process of morphological change as the NdFeB oxidised in air with time at temperature range 1300–1500 from Firdaus et al. [11]. The figure is not to scale

Table 1

Physical properties of the possible phase in the oxidised layers

	Phase	Density (g/cm³)	Magnetic Properties @25 °C		Hardness (Mohs scale)
	Phase	Density (g/cm³)	χ_m (cm³/mol)	Type	Hardness (Mohs scale)
EOZ	Fe₂O₃	5.24	+3,586 × 10⁻⁶	Paramagnetic	5.5–6.5
Layer 1	NdFeO₃	6.98	N/A	Weak ferromagnetic	–
Layer 2 and 3	NdBO₃	5.31	–	–	–
	Fe₂B	7.15	N/A	Ferromagnetic	–
	Nd₂O₃	7.24	+10,200 × 10⁻⁶	Paramagnetic	–
	α-Fe	7.87	N/A	Ferromagnetic	4

Experimental Methods

In this study, the samples were prepared from new and clean magnet samples. The sintered N45-NdFeB magnets were supplied by Alpha Magnetics Ltd. The magnet was thermally demagnetised at 573 K for 30 min in a Nabertherm TR 60 oven furnace and bulk samples (10 × 5 × 5 mm) were cut from the magnets by Struers Secotom-15/-50 high performance cutting machine. The coating was removed manually from each sample by peeling or grinding the surface. To quantify the chemical elements and composition of the sample, an inductively coupled plasma—atomic emission spectroscopy (ICP-AES) was used. The ICP-AES analysis was conducted at Spectrometer Services Pty. Ltd. Coburg, Melbourne. The sample was dissolved in a mixture of three parts HCl to one part HNO₃. The resultant solution was analysed using a Varian 730-ES Optical Emission Spectrometer. The power used was 1.20 kW with a Plasma Flow of 15 L min⁻¹ Argon. Calibration curves were generated for each element analysed using primary stock solutions. The bulk composition of the prepared magnet samples, determined using inductively coupled plasma (ICP-AES) is shown in Table 2.

Table 2

Chemical composition of clean REPM sample (mass%)

	Nd	Pr/Dy	Fe	B
Typical commercial range	23–31	0–7	65–70	0.9–1.2
Sample used in this study (by ICP-AES)	22.4	8.5 (Pr)	68	1

More than 30 g of clean magnet samples were oxidised in air atmosphere at 1373 K for 27–30 h using a Nabertherm LT 15/13/P330 muffle furnace under air atmosphere. Some of the oxidised samples then crushed using a hammer to separate the oxidised layers.

High resolution EPMA (a Field Emission Gun (FEG) EPMA-JEOL 8500F Hyperprobe) was used to evaluate the topography and elemental distribution within the samples. The standards used to calibrate the EPMA WD spectrometers prior to mapping were Fe metal (Fe), synthetic neodymium phosphate (NdP₅O₁₄), praseodymium silicide (PrSi₂), lanthanum hexaboride (LaB₆) and hematite (Fe₂O₃). Elements that were not directly measured by WD detectors were measured using two additional energy -dispersive (ED) detectors operating in parallel. After mapping, the element distribution data were processed using the software package CHIMAGE [14] to construct the appropriate metals and possible compounds (e.g. oxides, borides) map.

X–ray diffraction was used for phase analysis of the separated oxidised layers. All the separated samples were crushed in a ring mill and sieved to less than 100 mesh size, and subsequently packed into a sample holder for analysis. The Bruker AXS–D8 diffractometer utilising Cu Kα radiation (1.54178 Å) operating at 40 kV and 30 mA was used to scan the sample from 20 to 90° (2θ) at a rate of 0.02° per 1.5 s time step. Phase identification was performed using Bruker DIFFRAC.EVA© software (V4.1).

Results and Discussion

The oxidised sample were found to have about 30% increase in mass from the original and estimated to have around 80% conversion based on the kinetics described by Firdaus et al. [13]. The microstructural change of the oxidation layer were found to be similar to that described in Firdaus et al. [11]. Figure 2 show the sectioned oxidised sample showing the different layers developed after partial oxidation .

../images/468727_1_En_64_Chapter/468727_1_En_64_Fig2_HTML.gif — Fig. 2
Section of oxidised waste NdFeB magnet in air at 1373 K for 27 h showing the formation of different oxide layers

Three different layers can be recognized visually with the middle layer around 1.28 mm thick found to contain high amount of elongated NdFeO₃ phase as shown in the EPMA coloured map in Fig. 3. Image analysis of the layer using ImageJ show that the NdFeO₃ occupies about 62.02% of the layer followed by 22.23% of Fe₂O₃ and the rest (15.75%) are pores. It can be assumed based on the calculated area and the densities in Table 1 that about 78.80 mass% of the layer is NdFeO₃ and the rest (21.20 mass%) is Fe₂O₃.

../images/468727_1_En_64_Chapter/468727_1_En_64_Fig3_HTML.gif — Fig. 3
EPMA coloured map showing the phases formed in the NdFeO₃ rich layer

The layers were found to have different hardness, with the EOZ far more brittle under constant hammering than the others and inner core being the toughest. The particle of each layer can be differentiated by the difference in size with EOZ particles comes as the smallest (less than 150 µm). Figure 4 shows the crushed particles which were then categorized via visual inspection and sieving. The mass distribution of the separated EOZ, NdFeO₃ rich and the inner core layer particles were 28.20 mass%, 44.90 mass%, and 26.90 mass% with an average apparent density, determined by pycnometer analysis, of 2.17 g/cm³, 4.08 g/cm³, and 2.5 g/cm³ respectively. These suggest that the particles from the NdFeO₃ rich layer can be separated by the others through size and density separation .

../images/468727_1_En_64_Chapter/468727_1_En_64_Fig4_HTML.gif — Fig. 4
Crushed particles of the separated layers using hammer showing a the inner core of the magnet Layer 2 and Layer 3; b the NdFeO₃ rich—Layer 1; and c the External oxidation layer—EOZ

The XRD results in Fig. 5 indicate the major phases in the separated particles. The data for the EOZ contain peaks of mainly of Fe₂O₃, with minor NdFeO₃ . The presence of minor NdFeO₃ in the analysis mainly due to some of the NdFeO₃ rich layer particles reports to the undersize of 150 µm sieve. As predicted the NdFeO₃ rich layer particles contain peaks of mainly of NdFeO₃, with minor Fe₂O₃. The peaks of Fe₂O₃ indicate the presence of the phase in the inclusions, implying further size reduction is required to liberate the Fe₂O₃. The XRD data for the inner core layer particles contain peaks indicative of α-Fe with traces of NdBO_3, Fe₂O₃ and NdFeO₃. There was indication of a small amount of neodymium oxide (Nd₂O₃) and praseodymium oxide (Pr₂O₃) however the spectra might overlap with spectra of NdBO₃ and NdFeO₃ .

../images/468727_1_En_64_Chapter/468727_1_En_64_Fig5_HTML.gif — Fig. 5
XRD spectra collected for the visually separated oxidised sample. Significant peaks for the oxidisation phases are labelled in the plots of different layer

The results show that the NdFeO₃ phases are concentrated in the NdFeO₃ rich layer and in the EOZ, consequently the NdFeO₃ powder can be recovered by removing firstly the inner core particles. Figure 6 presents the proposed flowsheet to recover Nd in the permanent magnet as NdFeO₃. The oxidised magnet is firstly crushed in a hammer mill (crusher) to break up the samples into smaller particles and separate the layers. The crushed particles then screened through a 250 µm sieve and due to the different hardness most of the inner core particles will be removed at this stage by reporting to the oversize stream before then sent to the dense medium separator. Various dense liquid such as bromoform, tetrabromoethane (TBE) or Methylene iodide can be used but non-toxic liquid such as sodium polytungstate (SPT), lithium metatungstate (LMT) or lithium heteroplytungstate (LST) is much preferable to achieve 2.8 g/cm³ density cut point. Most of the NdFeO₃ rich powder in the separator will report to the underflow (sink) stream, these will then be ground before recycled back to the 250 µm sieve screen. The undersize stream will mostly be comprised of particles from the EOZ and NdFeO₃ rich layer, consisting of NdFeO₃ and Fe₂O_3. The particles from this stream will then be screened using the 150 µm sieve and the overflow will be sent to the second dense medium with cut point of 2.5 g/cm³. The underflow from this separator will consist mainly NdFeO₃ and combined with the underflow from the 150 µm screen, the stream then will go through the magnetic separator to separate the paramagnetic Fe₂O₃ to the ferromagnetic NdFeO₃. In this flowsheet , it is also proposed that the oxidation is carried out in air (in an open type furnace ) using a concentrated solar energy . This will avoid the need to generate heat for high temperature oxidation by combustion of fuel. It will also allow the process to be carried out in multiple days.

../images/468727_1_En_64_Chapter/468727_1_En_64_Fig6_HTML.gif — Fig. 6
Proposed flowsheet for recovering the Nd as NdFeO₃ through powder processing and separation

Conclusion

Permanent magnet oxidised in air at 1100 °C for 27–30 h show formation of separated layer rich of NdFeO₃ . This layer was found have different hardness and density from the other layers, and that it can be separated physically from the other layer. The mass distribution of the separated EOZ, NdFeO₃ rich and the inner core layer particles were found to be 28.20 mass%, 44.90 mass%, and 26.90 mass% with the average density determined by pycnometer analysis were 2.17 g/cm3, 4.08 g/cm3, and 2.5 g/cm3 respectively. The estimated NdFeO₃ in the NdFeO₃ rich layer was 78.80 mass%. A flowsheet to recover Nd as NdFeO₃ using series of density and magnetic separation after oxidation in air using concentrated solar energy was developed and the estimated NdFeO₃ powder recovered based on mass balance was 35.38 mass%. This study is a preliminary work for the development of a more comprehensive approach in recovering Nd in the form of NdFeO₃. It is hoped that this work will stimulate others to refine the process.

References