The Effects of Knife Milling and Ball Milling on Hydrogen Decrepitated Sm₂TM₁₇ Sintered Magnet Powder for Short-Loop Recycling

Abstract

Sm₂TM₁₇ sintered magnets (TM = Co, Fe, Cu, Zr) are utilised in high-temperature rotor applications due to their stable magnetic properties at elevated temperatures of 200–350 °C. However, Sm and Co are critical elements, and the reliance on virgin material supply chains must be reduced. Hydrogen decrepitation (HD) could facilitate magnet-to-magnet recycling of scrap material, but the milling characteristics of the powders generated by HD requires investigation. Sm₂TM₁₇ sintered magnets were exposed to 18 bar and 2 bar hydrogen pressure at 100 °C for 72 h and then knife-milled, roller ball-milled, and planetary ball-milled for varying milling times utilising a variety of surfactants. The particle size and morphology of the powders were investigated, and sintered magnets manufactured from chosen powders were characterised in terms of composition, density, microstructure, and magnetic properties. Knife milling for two minutes showed major particle size reductions of 70 and 82% in D50 for 18 bar and 2 bar samples respectively. Roller ball milling trials showed that a cyclohexane and oleic acid mixture was the most effective at reducing particle size, reducing D10, 50, and 90 by 92, 91, and 80% respectively. Knife milling HD powder for two minutes and then planetary ball milling this powder in a cyclohexane and 1 wt.% oleic acid mixture generated a particle size distribution of 1.3–6.8 µm. This powder formed a sintered compact with a density 0.08 g/cm³ lower than the as-received material. Sm losses due to oxidation and sublimation in addition to carbon impurities from surfactant usage caused the precipitation of an α-Fe/Co phase and formed ZrC phases respectively. Sm-hydride additions of 2–3 wt.% mitigated the formation of the α-Fe/Co phase, but ZrC phases remained and likely prevented cell structure formation and inhibited domain wall pinning in recycled magnets.

1. Introduction

Samarium Cobalt 2:17 type sintered magnets (Sm₂TM₁₇, where TM = Co, Fe, Cu, and Zr) are intermetallic materials known for their high coercivities (>2000 kA/m) and Curie temperatures (~850 °C) [1,2,3]. Therefore, these magnets are utilised in high-temperature applications (200–350 °C) such as actuators and permanent magnet motors used in the aerospace industry, as their magnetic properties remain stable at these temperatures [4].

Sm₂TM₁₇ sintered magnets exhibit a nanoscale cellular structure generating ferromagnetic behaviour through a domain wall pinning coercivity mechanism [2,4]. The cellular structure consists of a rhombohedral matrix phase (2:17R), which is enriched in Fe, a hexagonal cell boundary phase (1:5H), enriched in Cu, and a Zr-rich lamellar phase with a rhombohedral (1:3R) structure [2,4,5]. The large magnetocrystalline anisotropy difference between the cell interior and cell boundary gives rise to domain wall pinning behaviour [6].

This cellular structure is formed due to a series of complex heat treatments. A green compact of Sm₂TM₁₇ is usually sintered at 1190–1220 °C and subsequently homogenised at 1130–1175 °C, allowing for the formation of a 1:7H phase before quenching the material rapidly to preserve this structure [7,8,9,10]. The 1:7H phase is a precursor to the desired 2:17R phase; hence, forming it at this stage is crucially important. The sintered compact then undergoes heating to 750–850 °C to isothermally age the material and precipitate out the 2:17R phase and 1:3R phase that become enriched in Zr [11]. The sintered compact is then subject to a slow cooling process at 0.7–1 °C/min to 400 °C, with the Zr-rich lamellar structures acting as diffusion pathways for Cu to reach cell boundaries and form an ordered 1:5H phase [11]. The sintered compact can be further aged at 400 °C to further establish the cellular structure via Cu diffusion to the cell boundary phase before a final rapid quench stabilises the nanostructure. This heat treatment is usually conducted under vacuum and inert atmospheres to prevent oxidation [4,12] and reduce the sublimation of Sm [13] from the material.

Sm and Co are critical elements due to their high economic importance but also high likelihood of supply risk in the UK [14,15,16,17,18,19]. Therefore, there is a need to reduce the reliance on virgin supply chains of these materials, with one potential solution being to recycle Sm₂TM₁₇ sintered magnets from end-of-life (EoL) applications [20]. One method that may facilitate this is the hydrogen decrepitation (HD) process, which may offer low-energy ‘short-loop’ magnet-to-magnet recycling capabilities [21]. This process is being implemented in the recycling of NdFeB sintered magnets [22] and has previously been investigated at a lab scale for SmCo₅ magnets [23]. An investigation into the HD processing of Sm₂TM₁₇ material has recently been focused on cast alloy material [24,25,26], with little emphasis on the HD interactions of sintered magnets and recycling.

HD applies a hydrogen overpressure and, in most instances, elevated temperatures to Sm₂TM₁₇ sintered magnets, which causes the magnet to fragment and form a powdered product due to hydrogen uptake [21,27]. Hydrogen adsorbs to the surface of the material and then diffuses into the interstitial spaces within the crystal lattice, specifically the 2:17R phase, forming Sm₂TM₁₇H₅ [27]. The differential volume expansion between unreacted and reacted material results in intergranular and transgranular crack propagation and forms a powder [27]. The hydrogen must then be desorbed from the material at temperatures of 200–300 °C under vacuum conditions in order to restore the original magnetic properties of the material [28,29,30].

Previous work has been focused on the required pressures, temperatures, and processing times needed for the efficient and complete decrepitation of Sm₂TM₁₇ sintered magnets [29,30,31]. It has been shown that an applied temperature of 100–150 °C is required for thermal activation of the sample, with decrepitation being observed at low pressures of 2 bar in a 48–72 h period [30,31]. However, it should be noted that temperatures as low as 25 °C have facilitated HD at higher pressures of 10 bar and 18 bar in work by Zakotnik and Griffiths [29,30].

The powder generated from this process has the potential to be milled, magnetically aligned, pressed, sintered, and heat-treated to form a recycled magnet in a ‘short-loop’ recycling process [21]. This prevents the repetition of highly energy-intensive processing steps used at the beginning of the primary production process, such as recasting and coarse crushing [21]. However, if required, the powder generated could also be fed into ‘medium-loop’ remelting and recasting-based processes [32] or ‘long-loop’ acid leaching/chemical processes which precipitate and extract Sm and Co from the powdered material [33,34,35,36,37].

To date there has been no investigation into the processing steps needed after HD processing to facilitate the ‘short-loop’ recycling route, the first step of which is the milling of the HD powder to a suitable powder size for pressing into a green compact [12,21,38]. Ideally, the particle size distribution used for producing Sm₂TM₁₇ sintered magnets should be between 2 and 10 µm [4,12,38]. To produce a fully dense Sm₂TM₁₇ sintered magnet (8.3–8.5 g/cm³, depending on exact magnet composition) the average particle size, span of particle size distribution, and extent of oxidation are of utmost importance [38]. If the average particle size is too coarse or the particle size distribution is too wide, then porosity will be present in the magnet microstructure, which will have a detrimental effect on magnetic properties. The loss of the main ferromagnetic 2:17R phase will result in a loss of remanence, and pores will act as reverse domain nucleation sites, reducing the coercivity of the magnet.

In comparison to primary magnet manufacturing, the required energy needed for milling may be significantly lower than that of as-cast material due to it being a friable brittle hydride [30,39]. There are multiple methods that could be applied to mill powders to 2–10 µm, with two popular approaches previously applied to HD-processed NdFeB sintered material being knife milling and ball milling [40].

Knife milling is a technique that fractures powder through repeated impacts with stainless-steel blades rotating at ~20,000 RPM under an inert atmosphere [40]. The repeated collisions between the blades and the brittle/friable powder continually fracture the material, with a reduction in D50 powder size of 66% achieved in previous work applied to NdFeB HD powders after five minutes of milling [40]. However, due to the large number of collisions between powder particles and the blades, this process is prone to generating heat, resulting in agglomerations. These agglomerations impact the particle size distribution; e.g., Nayebossadri reported an increase in D90 of 144 µm when comparing HD and knife-milled powders [40]. Hence, cooling periods must be incorporated into knife milling procedures to minimise this effect.

HD powder can also be milled via ball milling under an inert atmosphere whilst using an organic liquid as a surfactant (surface active agent) [38,41,42,43]. This can typically take the form of either roller ball milling or planetary ball milling [4,12,38]. Both techniques are based upon the continual fracture and cold welding of particles as a result of collisions between the milling balls and powder until a saturation point is reached in regard to particle size [41,42].

In the case of roller ball milling, magnet powder is loaded into a milling jar alongside milling balls and then rotated at low rotation speeds (~100 RPM). The repeated collisions between milling balls and powder gradually reduce the particle size; however, due to this being a low-energy technique, the required time to reach the ideal 2–10 µm can be in the order of ~10s hours [44].

Planetary ball milling (also known as high-energy ball milling) utilises high rotation speeds of up to 800 RPM, with each milling pot seated in a carousel rotated around a central point. The milling pots themselves then counter-rotate as the main carousel spins to generate high g-forces and improve milling efficiency greatly over roller ball milling (<1 h milling time), but this approach is less scalable in terms of the powder mass milled per run [45,46,47,48,49,50].

The milling efficiency of both techniques can be influenced by the ball-to-powder ratio and the addition of different surfactants. Generally, ball-to-powder ratios by mass are between 5:1 and 20:1, with a greater ratio potentially increasing milling efficiency at the risk of inducing crystal structure damage or generating a large proportion of ultrafine particles (<1 µm), which are highly susceptible to oxidation [41,42]. In the case of Sm₂TM₁₇ sintered magnets, these ultrafines will likely form Sm₂O₃ due to their high surface area-to-volume ratio, which will dilute the overall magnetic properties of the finished magnet [51]. The milled powders are also highly pyrophoric due to their large surface area initiating a rapid exothermic oxidation reaction.

The addition of surfactants improves the efficiency of ball milling, as they minimise cold welding and reduce the overall particle size attainable [41]. Organic molecules can be used as a means of modifying the surface of powder particles and preventing the direct metal-to-metal contact needed for cold welding to occur [41]. Typical additions of surfactants can range from 1 to 5 wt.%, with the effectiveness or potency of the surfactant effect increasing with the length of the organic molecule chain and being dependent on the available functional groups (e.g., –COOH groups in fatty acids like oleic acid anchor to the surface of metal particles) [41,52].

The primary processing of Sm₂TM₁₇ sintered magnets previously utilised ball milling in conjunction with surfactants, such as cyclohexane, as a means of generating powder for green compacts before the advent of jet milling [4,38]. More recent work on SmCo₅ compositions has investigated using fatty acids such as oleic and stearic acids to generate nanoflakes from the high-energy ball milling process [45,48,50]. Therefore, surfactant-assisted ball milling may also be a feasible way to mill recycled magnet powder in much the same way.

However, a significant disadvantage of this technique is the contamination of samples with carbon additions that may result in reduced magnetic properties of the recycled Sm₂TM₁₇ sintered magnet. Previous investigations have shown that carbon contamination > 0.1 wt.% can result in the formation of ZrC, reducing the volume fraction of the 1:3R Zr-rich lamellar phase and by consequence the abundance of Cu within the 1:5H phase [53]. This resulted in a reduction in the coercivity of the Sm₂TM₁₇ sintered magnet due to the domain wall pinning coercivity mechanism being disrupted [53]. Attempts have been made to remove surfactants from planetary ball-milled SmCo₅ powders through heat treatment in argon and in air, showing that a 96% reduction in residual surfactant can be achieved using a heat treatment at 400 °C [46]. At present, there is no recorded literature on applying the same process to Sm₂TM₁₇ sintered magnets that have been milled using a surfactant.

The application of different milling methods on the particle size distribution of Sm₂TM₁₇ sintered magnet HD powders and their applicability in producing recycled magnets has yet to be explored. Specifically, the most appropriate milling methods, which surfactants are required, potential sources of contamination, and the influence of this contamination on the density of a sintered Sm₂TM₁₇ magnet require investigation.

Therefore, the aim of this study was to assess the effect of knife milling, roller ball milling, and planetary ball milling on Sm₂TM₁₇ sintered magnet HD powder particle size and recycled magnet density. This was investigated by applying each milling method to HD powders, generated at 18 bar and/or 2 bar and at 100 °C for 72 h, and observing the resultant particle size distributions and particle morphologies. In the case of both ball milling techniques, a variety of surfactants (isopropanol, cyclohexane, heptane, oleic acid) were tested to investigate their impact on milling efficiency. The main objective of this study was to generate a particle size distribution of 2–10 µm comparable to that used in primary magnet production. Finally, the milled powders with the most similar particle size distribution to the target distribution were sintered, and the resultant compacts were characterised to assess their density, microstructure, composition, and magnetic properties.

2. Methods and Materials

2.1. Starting Materials Characterisation

Commercial sintered magnet production scrap was supplied in the form of blocks ranging from 0.5 to 5 kg in mass. To provide a fresh reaction surface for decrepitation trials, these blocks were initially cracked using a hydraulic press. For the purposes of this article, the production scrap material will be referred to as ‘as-received’, and the recycled material manufactured from milled powder will be referred to as ‘recycled sintered compacts/magnets’.

As-received scrap material, HD-processed powders, and recycled sintered magnets were analysed to assess their microstructures using a Hitachi 4000 + (Hitachi, Tokyo, Japan) benchtop scanning electron microscope (SEM) with 15 kV accelerating voltage on backscattered electron imaging mode (BSE). Energy dispersive X-ray spectroscopy (EDS) was also conducted on as-received and recycled sintered magnets using an Oxford Instruments EDS detector (Oxford Instruments, High Wycombe, UK). Grain size analysis was conducted using a Carl Zeiss (Jena, Germany) microscope with axioCam ICc1 attachment on samples polished to 0.25 µm and etched in 3% Nital.

Proto tabletop AXRD (Proto Manufacturing, LaSalle, ON, Canada) with a nickel kβ absorber (0.02 mm; Kβ = 1.392250 Å) producing Kα radiation (Kα1 = 1.540593 Å, Kα2 = 1.544426 Å) was used for phase analysis of the starting materials, which was then processed through CrystalDiffract 7.0 software.

The magnetic properties of the as-received and recycled sintered magnets were measured using a Magnet Physik EP5 Permagraph (Magnet-Physik, Cologne, Germany), with the forming being measured at temperatures of 20–200 °C. The density of the as-received and recycling magnets were calculated using the Archimedes principle with a densitometer.

The compositions of as-received and recycled magnets were investigated using inductively coupled plasma optical emission spectroscopy (ICP-OES) (Agilent, Santa Clara, CA, USA) via an Agilent 5110 with ~500 mg of sample dissolved in aqua regia.

2.2. Hydrogen Decrepitation Trials

Samples were placed inside a furnace tube comprised of Inconel or stainless steel and then sealed using a copper gasket or rubber O-ring. The furnace tube was then attached to the HD processing rig and evacuated to a 10⁻³ mbar atmosphere before the introduction of hydrogen. Flowing hydrogen and applied temperature were then applied to each sample. The processing temperature was 100 °C, with a hydrogen pressure of 2 bar or 18 bar selected for HD trials for 72 h. The powders generated were stored in an inert Mbraun glovebox (Mbraun, Garching, Germany) filled with nitrogen.

In addition to HD powders, Sm metal was also exposed to hydrogen to form an Sm-hydride that would act as an additive to recycled sintered compacts and magnets. Sm-hydride was formed by exposing pure Sm metal to 2 bar hydrogen at 330 °C based upon prior work [54].

2.3. Oxygen/Nitrogen/Carbon Content Analysis

To investigate nitrogen and oxygen content within the as-received and recycled magnets, an Elementrac ONH-p Analyser (Eltra, Haan, Germany) was loaded with ~50 mg samples of each material. The content of each element was determined by inert gas fusion in an impulse furnace. Samples were also loaded into an Elementrac CS-i Analyser (~500 mg) (Eltra, Haan, Germany) for carbon analysis. Carbon content was characterised by melting under a pure oxygen atmosphere.

2.4. Milling of HD Powders

Multiple milling techniques were used in this work to investigate the effectiveness of each method on powder size reduction.

• Knife Milling—A total of 200–300 g of HD powder was loaded into a CGoldenWall electric grain grinder (CGoldenwall, Gongyi city, China), using 304 stainless-steel blades rotating at 20,000 RPM, and was milled for 1–4 min. After each minute of milling, the knife mill chamber was allowed to cool for 45 min to prevent cold welding of the sample to the chamber walls. All knife milling was completed inside a nitrogen-filled Mbraun glovebox to prevent sample oxidation.
• Roller Ball Milling—A total of 35 g of HD powder or knife-milled powder was loaded into a stainless-steel milling pot alongside 250 g of stainless-steel milling balls (10 mm diameter) for a ball-to-powder ratio of approximately 7:1. Surfactants of isopropanol, cyclohexane and cyclohexane/heptane and oleic acid mixtures were added to the milling pot, with the former three being added as a 60 wt.% addition and the oleic acid being added at 2.5 wt.% of the HD powder mass. The milling pot was rotated at a speed of 100 RPM. Milled powders were dried in a vacuum chamber to remove volatile organic surfactants.
• Planetary Ball Milling—A total of 80 g of knife-milled powder was loaded into a tungsten carbide milling pot alongside 1200 g of tungsten carbide milling balls (5 mm diameter) for a ball-to-powder ratio of approximately 15:1. A surfactant mix of 60 wt.% cyclohexane and 0.5–2.5 wt.% oleic acid were added to the milling pot across the milling trials. The milling pot was rotated at a speed of 400 RPM for a total milling time of 10 min, with pauses in the cycle added to allow the pots to cool. Milled powders were dried in a vacuum chamber to remove volatile organic surfactants. For magnets manufactured with Sm-hydride additions, the hydride was crushed with a mortar and pestle and added to the planetary ball milling pot to be milled and mixed with HD powder.

All milling pots were loaded inside a nitrogen-filled Mbraun glovebox, and the sample pots were sealed under this atmosphere using a rubber O-ring prior to milling to prevent sample oxidation. Contamination from abrasion of the ball milling pots or milling balls was not investigated in this study due to the high abrasion resistance of the milling equipment materials utilised.

2.5. Particle Size Analysis

To analyse the particle size of all HD and milled powders, a Malvern Mastersizer 3000 (Malvern, UK) was used under a nitrogen atmosphere with a gas pressure of 4 bar.

2.6. Vacuum Degassing and Surfactant Removal for Milled Powders

Planetary ball-milled samples (10 g) were loaded into a vacuum tube furnace with a Pirani gauge attachment. The system was then evacuated to a vacuum pressure of <10⁻³ mbar. The samples were then heated to 350 °C at 5 °C/min whilst under dynamic vacuum, and the change in vacuum pressure during gas release was monitored using a Pirani gauge (Edwards, Burgess Hill, UK). During degassing, the furnace was held at 200 °C and 350 °C for 45 min to desorb hydrogen and residual surfactant, respectively.

2.7. Recycled Magnet Manufacture

Degassed planetary ball-milled powders were loaded into rubber isostatic press bags and sealed under a nitrogen atmosphere before being magnetically aligned using an 8T pulse magnetising field. The material was then isostatically pressed at a pressure of 30 tonnes to form cylindrical green compacts with a 10 mm diameter and 20 mm height. These compacts were then inertly loaded into a tube furnace with a ceramic furnace tube. The tube was then evacuated to an atmosphere of 10⁻⁶ mbar before being subject to the heat treatment profile shown in Figure 1.

Samples sintered to test for densification (sintered compacts) were sintered in a standard tube furnace and only underwent section (A) of the heat treatment profile shown in Figure 1A. Samples that underwent the full heat treatment cycle to form recycled magnets were loaded into a vertical quench furnace, where the sample was suspended in the furnace hot zone. When heat treatment profiles in Figure 1A,B part (A) were completed, the sample was dropped from the hot zone into a water-cooled area of the furnace to quench the material.

The first stage of the sintering and heat treatment profile (a) consisted of a 10 °C/min heat to sinter, sintering for 90 min at 1215 °C, cooling to homogenisation temperature at 2 °C/min, homogenising at 1160 °C for 480 min, and then quenching the sample to room temperature. The second stage of heat treatment (a) consisted of 10 °C/min heating to isothermal ageing temperature, isothermal ageing for 540 min at 860 °C, slow cooling to the secondary ageing temperature at 0.7 °C/min, secondary ageing at 400 °C for 240 min, and then quenching the sample to room temperature.

3. Results

3.1. As-Received Material Characterisation

Table 1. ICP-OES, oxygen, nitrogen, and carbon analysis of the as-received material.

Sample	Sm (wt.%)	Co (wt.%)	Fe (wt.%)	Cu (wt.%)	Zr (wt.%)	O (ppm)	N (ppm)	C (ppm)
As-received	23.9 ± 0.16	49.3 ± 0.28	19.3 ± 0.10	4.8 ± 0.09	2.8 ± 0.12	1600 ± 86	1100 ± 84	850 ± 20

shows the composition of the Sm₂TM₁₇ sintered magnet sample used as a recycling feedstock for this study. Overall, the composition contained all the typical alloying elements found within Sm₂TM₁₇ sintered magnets, with the only anomalies being a relatively low Sm content and a high carbon impurity content.

Figure 2A shows an optical micrograph of the grain structure of the as-received material; the calculated grain size of this material was 76 ± 6.5 µm, with a density of 8.40 ± 0.002 g/cm³. Figure 2B shows an SEM micrograph of the as-received material. As shown in Figure 2B, the grey-coloured phase denotes the Sm₂TM₁₇ matrix phase, the white phase shows an Sm-rich region, and the darkest phase shows a Zr-rich impurity phase. An additional EDS analysis comparing the composition of each phase is shown in

Table 2. EDS analysis showing the compositions of three respective phases present in the Sm₂TM₁₇ as-received material.

EDS Site	Sm (wt.%)	Co (wt.%)	Fe (wt.%)	Cu (wt.%)	Zr (wt.%)
A1	24.9 ± 0.41	48.8 ± 0.54	18.8 ± 0.31	5.30 ± 0.78	2.21 ± 0.17
P1	56.4 ± 0.58	28.8 ± 0.41	10.6 ± 0.30	2.82 ± 0.71	1.44 ± 0.19
P2	8.86 ± 0.36	15.2 ± 0.29	6.09 ± 0.22	1.24 ± 0.52	68.6 ± 0.54

.

Figure 3 shows the XRD phase analysis of the Sm₂TM₁₇ as-received material. Three main phases were identified: the main ferromagnetic rhombohedral (2:17R) phase, the hexagonal (1:5H) cell boundary phase, and the rhombohedral Zr-rich lamellar phase (1:3R). The 2:17R and 1:5H phases are particularly evidenced by the large peak intensities at the 42–43° 2θ angles, with the 1:3R exhibiting smaller intensities at 2θ angles such as 62–63°.

Figure 4 shows the magnetic properties of the Sm₂TM₁₇ as-received sample across a range of operating temperatures. The traces for both the 25 °C and 50 °C measurements did not allow for the magnet to fully demagnetise, as evidenced by the demagnetisation curve not crossing the x-axis, and hence a coercivity value could not be determined, as shown in

Table 3. Collated magnetic properties of Sm₂TM₁₇ as-received material at a variety of operating temperatures of 25–200 °C.

Temperature (°C)	Remanence (T)	Coercivity (kA/m)	(BH)max (kJ/m3)
25	1.21	>2055	>258
50	1.14	>1464	>232
100	1.13	1267	221
200	1.06	818	191

. The reason this occurred was because the electromagnet applying the magnetic field to these samples saturated before the sample could be fully demagnetised.

The shape of both the 25 and 50 °C demagnetisation curves exhibited multiple undulations and changes in the curve gradient during demagnetisation, whereas the measurements for 100 and 200 °C did not show these features. As evidenced in Table 3, with increasing operating temperature, there was a reduction in remanence across the measurements of 12.4%, with coercivity also decreasing from >2055 to 818 kA/m when comparing 25 and 200 °C measurements.

3.2. Particle Size Analysis of HD-Processed and Milled Powders

3.2.1. HD Powder

Figure 5 shows the particle size analysis results for samples processed at 18 bar and 2 bar at 100 °C for a 72 h processing period.

As shown by Figure 5, the particle size of HD powders generated at 18 bar was always finer than those generated at 2 bar. For example, the D50 of the 18 bar powder was 148 µm lower than that of the 2 bar powder. The D10, 50, and 90 of the 18 bar powder were all lower than the 2 bar powders, reflecting a shift in the entire particle size distribution to a finer particle size. Figure 6 shows SEM micrographs of the Sm₂TM₁₇ as-received material processed at 18 bar and 2 bar at 100 °C for 72 h. Both powders exhibited intergranular and transgranular fracturing throughout the powder particles.

Larger powder particles typically had smaller powder flakes attached to their surface, with 18 bar powders generally being smaller than those generated at 2 bar. Despite the variations between the 18 bar and 2 bar powders, the extent of cracking present in both powders indicated that the material was likely high-friable and could easily be broken apart upon mechanical agitation.

3.2.2. Knife-Milled Powder

Figure 7 shows the particle size analysis for knife-milled HD powders that were processed at 18 bar and 2 bar and at 100 °C for 72 h. Figure 7 shows that the majority of the particle size reduction occurred for both powders within the first minute of knife milling. For example, the powders reduced in D90 by 120 µm and 314 µm for 18 bar and 2 bar HD-processed material, respectively, in one minute, relative to 147 and 367 µm decreases across four minutes of milling. This effect also occurred in the D50 and D10 values, showing that the entire particle size distribution shifted significantly after one minute of milling; see Figure A1 for particle size distribution traces. There was a fixed difference between the final particle size achieved via knife milling for the 18 bar and 2 bar HD powders, with the 18 bar powders reaching a final particle size lower than that of the 2 bar powders across D10, 50, and 90.

Figure 8 shows the particle morphology of HD powder generated at 2 bar that was knife-milled for four minutes. Image A shows that the knife-milled particles were below 100 µm in size, and there was a tendency for these powders to agglomerate together into clusters. The powder morphology was non-spherical and random in nature, with little visible cracking within the particles. Extremely small particles such as those highlighted in Figure 8B adhered to the surface of larger powder particles, forming fine powder agglomerations.

3.2.3. Roller Ball-Milled Powder

Figure 9 shows the particle size analysis for HD powders generated at 18 bar that were ball-milled in a variety of surfactants for 1–7 h.

It can be seen from Figure 9 that each surfactant/surfactant combination had varying levels of effectiveness at improving the milling efficiency and reducing the particle size. Isopropanol-based milling trials showed a reduction in D90, D50, and D10 after 1 h and 4 h of processing, but then the overall particle size increased again after 7 h of processing. For example, D50 decreased from 102.3 to 48.4 µm after 4 h but then increased to 60 µm after 7 h of processing.

Similarly, roller ball milling with cyclohexane resulted in an initial particle size reduction after 1 h of milling, but, in the case of D50 and D90 percentiles, the particle size increased again after 4 h. After 7 h of milling, the D50 and D90 decreased to a size similar to what was observed after 1 h of milling; for example, the D50 after 7 h was 52 µm, compared to 64.7 µm after 1 h. Interestingly, the D10 value continued to decrease as milling time was increased, showing a 92% reduction in particle size.

Surfactant mixes utilising oleic acid in conjunction with heptane or cyclohexane showed an overall much greater milling efficiency than samples milled without oleic acid. Both samples showed a decreasing trend in particle size across D10, D50, and D90 with increased milling time. A combination of heptane and 2.5 wt.% oleic acid showed a gradual decrease in particle size, with D10, D50, and D90 all decreasing by 65, 59, and 64%, respectively, after 7 h of milling. However, a combination of cyclohexane and 2.5 wt.% oleic acid showed a much steeper reduction in powder coarseness, with D10, D50, and D90 all decreasing by 92, 91, and 80%, respectively, after 7 h of milling. The particle size against volume density traces shown in Figure A2 present additional information on the effect of milling time on particle size distribution for each surfactant.

Figure 10 shows the particle morphology for roller ball-milled powders generated after 7 h of milling in cyclohexane (Figure 10(1A,1B)) and a cyclohexane + 2.5 wt.% oleic acid mix (Figure 10(2A,2B)). Overall, it was clear that the powders milled using only cyclohexane as a surfactant were far coarser than those with an oleic acid addition. Both powders showed a high degree of agglomeration, as seen in images 1A and 2A in Figure 10, with the particle morphology being non-spherical and with a prevalence of flake-like structures present. The powder milled in cyclohexane had some particles > 100 µm present after 7 h of milling, onto which smaller, more finely milled particles had adhered; see image 1B in Figure 10. In contrast, there were very few particles > 100 µm found in the powders milled with the cyclohexane and 2.5 wt.% oleic acid mix. In this sample, agglomerates of thin flake-like powder particles were observed; see image 2B in Figure 10.

Following the initial roller ball milling trials shown in Figure 11, further milling trials took place using the cyclohexane and 2.5 wt.% oleic acid combination. This utilised a longer milling time of 14 h and 18 bar HD powder feedstock that was knife-milled for 2 min as a basis for the roller ball milling. For comparison purposes, the particle size volume density traces for the aforementioned trial have been presented alongside the trace of powder milled in a cyclohexane 2.5 wt.% oleic acid for 7 h using 18 bar HD powder feedstock.

It is clearly shown in Figure 11 that the combination of increased milling time and a reduced feedstock particle size allowed for a much finer powder to be generated. The D10, D50, and D90 values for the trial presented in Figure 11 were 1.09, 3.38, and 6.32 µm, respectively. Both of the milled powders shown in Figure 11 exhibited two peaks, showing that there were two distinct powder size populations.

3.2.4. Planetary Ball-Milled Powder

Figure 12 shows the particle size analysis for planetary ball-milled powder generated using cyclohexane and a varying oleic acid surfactant (0.2–2.5 wt.%) with HD powder generated at 2 bar that was knife-milled for two mins as a feedstock.

The D10 and D50 values after planetary ball milling did not change greatly with differing oleic acid surfactants, with all D10 values being between 1.2 and 2.1 µm and all D50 values being between 3.0 and 4.8 µm. However, there was a sharp decrease in the D90 particle size when the oleic acid addition was ≥1 wt.%. For example, at 0.5 wt.% oleic acid addition, the D90 was 44.9 µm, and at 1 wt.% addition the D90 decreased to 6.82 µm, with little variance in the particle size occurring with further additions of oleic acid. Full particle size and volume density traces are shown in Figure A3, illustrating the effects of differing oleic acid additions on the particle size distribution generated through planetary ball milling knife-milled magnet powder.

Figure 13 shows the particle morphology of planetary ball-milled powder generated using 1.0, 1.5, and 2.0 wt.% oleic acid content. The particles across all three of the powder samples were recorded to be <10 µm in size, with most particles being observed as a component of larger agglomerations. The particle morphology was varied, with some flake- or plate-like structures observed among concave disc-like particles.

3.3. Degassing and Surfactant Removal from Planetary Ball-Milled Powder

Figure 14 shows an example set of vacuum degassing traces for the planetary ball-milled powder utilising a cyclohexane and 2.5 wt.% oleic acid surfactant mix.

Figure 14 shows two clear desorption events, the first desorption reached its maximum at ~130 °C and showed the removal of hydrogen from the milled powder. The second desorption peak, which reached maximum desorption at ~260 °C, was related to the removal of oleic acid from the powder. Overall, the traces show that proportionally more hydrogen gas was released than oleic acid.

Each sample of planetary ball-milled powder was analysed for the remaining carbon content as shown in Figure 15. Figure 15 clearly shows that, for the smallest oleic acid addition of 0.5 wt.%, the residual carbon content was far lower than that of the other samples, with 0.187 wt.% C compared to ≥0.679 wt.% C for all other samples. At additions of ≥1 wt.% oleic acid, the residual carbon content within the powders dramatically increased and reached a saturation point, with a minor decrease in carbon content seen for the 2.5 wt.% oleic acid milling trials.

3.4. Sintering Trials for Density (Planetary Ball Milling)

Planetary ball-milled powders utilising a cyclohexane and oleic acid surfactant mix (1.0, 1.5, 2.0 wt.%) were degassed and then subsequently aligned, pressed, sintered, and homogenised to form compacts. Further heat treatment was not performed at this stage, as these trials were conducted as initial experiments to investigate the densification of compacts manufactured using planetary ball-milled powders. The compositional information for each of these compacts, alongside the original magnet composition, are summarised in

Table 4. Compositional analysis of sintered and homogenised compacts manufactured from planetary ball-milled powders using ICP-OES and separate oxygen, nitrogen, and carbon analysis.

Sample	Sm (wt.%)	Co (wt.%)	Fe (wt.%)	Cu (wt.%)	Zr (wt.%)	O (ppm)	N (ppm)	C (ppm)
As-received	23.9	49.3	19.3	4.8	2.8	1600	1100	850
1.0 wt.%	22.6	49.9	18.7	4.8	2.6	3400	2300	8200
1.5 wt.%	19.4	50.2	19.4	4.4	3.0	4100	2700	7500
2.0 wt.%	22.1	49.1	18.8	4.6	2.9	3500	2100	7200

.

For each of the compacts formed, there were three key compositional differences compared to the as-received material. Firstly, the Sm content for each of the sintered compacts was significantly lower than that of the as-received material. Respectively, the planetary ball-milled powders utilising 1.0, 1.5, and 2.0 wt.% oleic acid had 5.4, 18.8, and 7.5% less Sm than the as-received material.

Secondly, the oxygen content of the sintered compacts was far greater than the as-received material, with the sintered compacts having at least 1800 ppm greater oxygen content. Finally, the carbon content of all sintered compacts was dramatically greater than the as-received material, showing an increase of 6350–7650 ppm depending on the sintered compact examined. Outside of these compositional differences, there was also an increase in the nitrogen content, though that was likely due to the powders being stored in a nitrogen-filled glovebox.

The densities of each sintered and homogenised compact are shown in

Table 5. Density measurements of sintered and homogenised compacts manufactured from planetary ball-milled powders.

Sample	Density (g/cm3)
As-received	8.40 ± 0.002
1.0 wt.%	8.32 ± 0.006
1.5 wt.%	8.30 ± 0.004
2.0 wt.%	8.30 ± 0.01

in comparison to the amount of oleic acid that was added during planetary ball milling. The densities of the sintered and homogenised compacts were very consistent despite the varying surfactant content used during the milling stage. Overall, the densities of the compacts were lower than that of the as-received material by ~0.1 g/cm³; however, it should be noted that the as-received material was fully heat-treated and therefore may have further densified because of this.

Figure 16 shows a compilation of micrographs for the sintered and homogenised compacts formed from planetary ball-milled trials using 1.0, 1.5, and 2.0 wt.% oleic acid additions. Four separate phases were identified in the microstructures, with a supporting EDS analysis for each of those phases provided in

Table 6. EDS analysis showing the compositions of the four respective phases present in the sintered and homogenised compact microstructures.

EDS Site	Sm (wt.%)	Co (wt.%)	Fe (wt.%)	Cu (wt.%)	Zr (wt.%)
1	24.4 ± 0.40	50.0 ± 0.53	18.6 ± 0.31	7.1 ± 0.77	-
2	8.20 ± 0.30	59.0 ± 0.58	31.8 ± 0.38	1.1 ± 0.80	-
3	15.5 ± 0.54	46.4 ± 0.55	22.3 ± 0.43	-	15.8 ± 0.38
4	70.4 ± 0.35	29.6 ± 0.35	-	-	-
1.0 wt.% Oleic Acid
1	23.8 ± 0.40	49.1 ± 0.54	18.3 ± 0.31	7.5 ± 0.79	1.4 ± 0.16
2	2.11 ± 0.26	60.2 ± 0.65	34.6 ± 0.43	3.1 ± 0.91	-
3	15.6 ± 0.37	29.3 ± 0.37	9.7 ± 0.25	4.2 ± 0.64	41.3 ± 0.43
4	70.8 ± 0.35	29.2 ± 0.35	-	-	-
1.5 wt.% Oleic Acid
1	27.0 ± 0.45	47.2 ± 0.58	17.1 ± 0.32	6.3 ± 0.91	2.3 ± 0.17
2	12.2 ± 0.51	47.6 ± 0.49	23.5 ± 0.38	-	16.4
3	4.1 ± 0.37	4.8 ± 0.24	2.0 ± 0.20	0.2 ± 0.47	88.9 ± 0.61
4	68.1 ± 0.39	31.8 ± 0.37	-	-	-
2.0 wt.% Oleic Acid

.

The dark grey phase within each of the sintered and homogenised compacts was identified as the Sm₂TM₁₇ matrix phase, with EDS analysis showing Sm, Co, Fe, and Cu all present within this phase. Zr was detected in this phase for the sample manufactured using 1.5 wt.% and 2.0 wt.% oleic acid but not detected in the 1.0 wt.% oleic acid sample. The darkest phase was determined to be a Co- and Fe-rich phase, with a high proportion of Co (~60 wt.%) and Fe (~30–35 wt.%) being detected, as shown in Table 6’s EDS site 2 measurements. The exception to this was the sintered and homogenised sample manufactured using powder milled with 2.0 wt.% oleic acid. EDS of the black phase in this sample showed Sm at 12.2 wt.% but also contained 16.4 wt.% Zr with far less Co and Fe enrichment, which was not observed in the other samples.

The darkest phase was often situated next to a lighter-coloured Sm-rich phase, annotated as EDS site 4, where only Sm and Co were detected. The proportion of Sm was >68 wt.% in these areas across all three samples. Finally, EDS site 3, which examined a light-grey-coloured phase, was linked to a Zr-rich phase where the specific wt.% of Zr varied widely from sample to sample, e.g., 15.8–88.9 wt.%. It should be noted that the sintered and homogenised powder formed from powder milled with the 2.0 wt.% oleic acid addition had far smaller precipitates of the lightest and darkest phases.

3.5. Sm Additions and Full Heat Treatment

Planetary ball-milled powder using cyclohexane and 1 wt.% oleic acid was used to manufacture sintered and fully heat-treated Sm₂TM₁₇ sintered magnets with additional Sm-hydride. Two fully heat-treated recycled magnets were manufactured with 2 and 3 wt.% Sm additions, respectively. The compositional information for each of those samples compared to the as-received material is presented in

Table 7. Compositional analysis of fully heat-treated Sm-compensated magnets using ICP-OES and separate oxygen, nitrogen, and carbon analysis.

Sample	Sm (wt.%)	Co (wt.%)	Fe (wt.%)	Cu (wt.%)	Zr (wt.%)	C (ppm)
As-received	23.9 ± 0.16	49.3 ± 0.28	19.3 ± 0.10	4.8 ± 0.09	2.8 ± 0.12	850 ± 20
2 wt.% Sm	23.9 ± 0.11	48.7 ± 0.21	18.8 ± 0.14	4.5 ± 0.003	3.3 ± 0.008	8000 ± 19
3 wt.% Sm	24.6 ± 0.02	48.5 ± 0.06	18.8 ± 0.009	4.5 ± 0.001	3.3 ± 0.02	7000 ± 130

.

Table 7 shows that the sample with the 2 wt.% Sm addition had an identical Sm content to the as-received material, and the sample with the 3 wt.% Sm addition had a 0.7 wt.% higher Sm than the as-received material. All other major alloying elements were present in similar quantities across the as-received and recycled magnet samples. The major outlier compositionally was related to the carbon content, where the level of impurity was an order of magnitude greater in the recycled magnets when compared to the as-received material.

Table 8. Density measurements of fully heat-treated 2–3 wt.% Sm-compensated sintered magnets.

Sample	Density (g/cm3)
As-received	8.40 ± 0.002
2.0 wt.% Sm	8.23 ± 0.017
3.0 wt.% Sm	8.37 ± 0.025

shows the densities obtained for the two recycled magnet samples in comparison to the as-received material. The 2 wt.% Sm-compensated sample had the lowest density, with the 3 wt.% Sm sample reaching a density just 0.03 g/cm³ lower than that of the as-received material.

Figure 17 shows micrographs of the 2 and 3 wt.% Sm-compensated recycled magnet microstructures; phases of interest have been annotated and analysed via EDS, as shown in

Table 9. EDS analysis showing the compositions of different phases present in the fully heat-treated Sm-compensated sintered magnets.

EDS Site	Sm (wt.%)	Co (wt.%)	Fe (wt.%)	Cu (wt.%)	Zr (wt.%)
1	24.7 ± 0.38	50.8 ± 0.51	18.1 ± 0.28	6.5 ± 0.74	-
2	1.51 ± 0.25	61.0 ± 0.31	37.5 ± 0.28	-	-
3	7.5 ± 0.31	5.1 ± 0.19	1.1 ± 0.16	-	86.3 ± 0.36
4	61.9 ± 0.55	32.2 ± 0.40	-	0.7 ± 0.65	5.2 ± 0.23
2.0 wt.% Sm
1	24.7 ± 0.39	49.5 ± 0.53	19.3 ± 0.30	6.5 ± 0.80	-
2	-	61.6 ± 0.27	38.4 ± 0.27	-	-
3	8.45 ± 0.32	14.8 ± 0.25	1.34 ± 0.5	-	69.3 ± 0.5
4	69.9 ± 0.36	27.5 ± 0.31	2.6 ± 0.26	-	-
3.0 wt.% Sm

.

In 2 wt.% and 3 wt.% Sm-compensated sample, four key phases were observed. The grey phase was attributed to the Sm₂TM₁₇ matrix phase, though in both samples no Zr was detected in either EDS scan; see Table 9’s EDS site 1 measurements. The darkest contrast phase was identified as a Co- and Fe-rich phase, with the majority of this phase being Co (≥61 wt.%), as shown by the EDS site 2 scans. This phase was present in a greater quantity in the 2 wt.% Sm-compensated sample when compared to the 3 wt.% Sm sample, as shown in Figure 17(2.0A,3.0A). Also identified in both samples were white-coloured Sm-rich regions (≥61.9 wt.% Sm), where the other main constituent element was Co, measured at 27.5–32.2 wt.%. These white phases were preferentially formed at the grain boundaries and triple-point junctions in both recycled Sm-compensated magnet microstructures, as shown in Figure 17. Finally, a Zr-rich phase was detected in both recycled magnets; this was denoted by a light grey phase decorating the grain boundaries of both sample microstructures. The proportion of Zr was very high, with EDS scans (site 3 shown in Table 9) indicating a 69.3 wt.% in the area scanned in the 3 wt.% Sm recycled magnet and an 86.3 wt.% in the 2 wt.% recycled magnet. The Zr-rich phase was a distinctive square shape and, much like the Sm-rich phase, was also identified in great concentrations at triple-point junctions in the microstructure.

Table 10. Magnetic properties of as-received material in comparison to recycled sintered Sm₂TM₁₇ magnets with 2 and 3 wt.% Sm addition, respectively.

Sample	Remanence (T)	Coercivity (kA/m)	(BH)max (kJ/m3)
As-received	1.21	>2055	>258
2 wt.% Sm	0.26	9.1	0.7
3 wt.% Sm	0.16	6.2	0.09

shows a comparison of key magnetic properties between the as-received material and the 2 wt.% and 3 wt.% Sm-compensated recycled magnets. All the recorded magnetic properties of remanence, coercivity, and (BH)_max were far lower in the recycled magnets than in the as-received material. As a result of HD processing and the subsequent post-processing techniques applied, the magnetic properties of the recycled material were diminished, with the 2 wt.% Sm-compensated recycled magnet having slightly better properties than the 3 wt.% Sm-compensated sample.

4. Discussion

Hydrogen decrepitation (HD) processing has the potential to be used in the recycling of Sm₂TM₁₇ sintered magnets from end-of-life (EoL) applications. Specifically, this technique could facilitate ‘short-loop’ magnet-to-magnet recycling which would omit the energy-intensive steps, such as casting and coarse crushing, used in primary sintered magnet production. This work set out to investigate the required post-processing, with emphasis on the fine milling stage, needed to generate fully dense sintered magnets using this ‘short-loop’ pathway.

The composition of the as-received feedstock material used in this work, as shown in Table 1, contained all the nominal elements expected of a Sm₂TM₁₇ sintered magnet [4]. The material was noted to be low in Sm content at just 23.9 wt.% (typically at least 24 wt.% Sm is utilised in commercial grade samples, [30,39]) and high in carbon impurity (850 ppm). This was likely due to this being a scrap material feedstock that was originally rejected from commercial production at the quality control phase. Figure 2 and Figure 3 showed that this feedstock consisted of the typical phases associated with this Sm₂TM₁₇ sintered magnet, those being the 2:17R, 1:5H, and 1:3R phases [1,4]. However, the magnetic properties of the material were not optimised, as shown in Figure 4 and Table 3, with various undulations being present in the demagnetisation loops, likely due to carbon impurities forming ZrC phases. This was shown at EDS site P2, as shown in Table 2, resulting in sub-optimal cell structure formation and irregular demagnetisation behaviour. HD trials taking place at 18 bar and 2 bar, and at 100 °C for 72 h, showed that 18 bar powders had a finer particle size than 2 bar powders; see Figure 5. Both the 18 bar and 2 bar powders showed areas of intergranular and transgranular fracture, as shown in Figure 6.

Knife milling trials of both the 18 bar and 2 bar HD powder showed that the vast majority of powder size decrease occurred during the first minute of milling, as shown in Figure 7. For example, the D90 value of the 18 bar and 2 bar HD-processed powders decreased by 120 and 314 µm, respectively, within the first minute of milling relative to an overall drop of 147 and 367 µm, respectively, after four minutes. Unlike the previous work investigating this technique for use on NdFeB, very little powder agglomeration was seen, and the D90 value did not increase relative to the starting HD powder size [40]. In their work, Nayebossadri saw an increase of 144 µm in powder size relative to HD powder after knife milling, which was not observed here. It should be noted that the D50 and D10 values of the knife-milled powder in Nayebossadri’s work were far lower than that shown in Figure 7, reaching 6.45 and 1.45 µm for D50 and D10, respectively [40]. Potentially, this small powder size led to increased agglomerations, leading to the very high D90 values, as metallic powders < 20 µm in size are often electrostatically attracted to one another. Agglomerations were present in knife-milled Sm₂TM₁₇ sintered magnet HD powder, but, as shown in Figure 8, this consisted of small particles (<10 µm) agglomerating on the surface of larger ~100 µm size particles.

Also of note was the fixed difference in final particle size between the two different HD powder feedstocks after knife milling. For each measurement taken, the 18 bar HD powder was always finer than the 2 bar HD powder after knife milling; see Figure 7. This may be linked to the initial powder size being smaller in the case of the 18 bar powder, e.g., a D90 of 202 µm compared to 452 µm in the 2 bar powder, see Figure 5, or a greater extent of cracking already present in the 18 bar powders after HD, leading to a more friable powder.

Roller ball milling trials of HD powder generated at 18 bar showed that different surfactants had varying effects on the milling efficiency over a 1–7 h milling time, as shown in Figure 9. Isopropanol and cyclohexane were not very effective at milling HD powder to a finer size, with both surfactants allowing milled powder to agglomerate over time during milling. For example, the powder milled in isopropanol saw an increase in D90 and D50 after 4 h of milling compared to 1 h, reaching sizes comparable to the starting material, as shown in Figure 9. This was likely due to the surfactant not preventing metal-to-metal surface contact areas due to their short carbon chain length and lack of polar functional groups to act as anchors to the surface of powder particles [41]. However, when oleic acid was utilised, the longer carbon chain and polar functional groups (-COOH) of this molecule seemed to greatly increase milling efficiency. It is likely that the cyclohexane/heptane base allowed the oleic acid to be distributed across the HD powder particles, with the long carbon chains forming layers on the surface of the particles and preventing cold welding in a similar fashion to previous work on SmCo₅ powder ball milling [41,45,50]. This greater milling efficiency was shown by the large reduction in particle size over time for the milling trials utilising the 2.5 wt.% oleic acid addition. The cyclohexane and oleic acid mix achieved a far greater D90 particle size reduction in 7 h of milling compared to the heptane and oleic acid mix, showing reductions of 80% and 64%, respectively. This suggests that cyclohexane may be a better dispersant for oleic acid than heptane due to the increased milling efficiency observed. The particle morphologies shown in Figure 10 illustrated the impact of oleic acid on powder size and shape, with far more flake-like particles being visible for a cyclohexane and oleic acid mixture when compared to only using cyclohexane as a surfactant.

Additionally, a longer milling time of 14 h and utilising HD powder knife-milled for 2 min as a feedstock for ball milling allowed for a very small particle size and very narrow particle size distribution to be achieved, as shown in Figure 11. Specifically, a distribution of 1.1–6.3 µm from D10 to D90 was achieved, which is suitable for generating sintered magnets, as a large proportion of this powder falls within the ideal 2–10 µm range required for Sm₂TM₁₇ sintered magnet production [4,12,38]. This result suggests that the cyclohexane and 2.5 wt.% oleic acid combination can reach the desired particle size for processing recycled magnets before reaching a particle size reduction plateau [41,42].

When applying the planetary ball milling process to knife-milled feedstock using cyclohexane and a variable amount of oleic acid, it was shown that a particle size distribution of D10 to D90 between 2 and 10 µm was achievable with ≥1 wt.% oleic acid addition; see Figure 12. For example, with a 1 wt.% oleic acid addition, D10, D50, and D90 values of 1.3, 3.02, and 6.82 µm were achieved, respectively, as shown in Figure 12. However, for a 0.5 wt.% addition, the D90 value recorded increased to 44.9 µm, suggesting that this addition for the 80 g powder sample was insufficient to coat the surface of the powder particles and prevent metal-to-metal contact, leading to cold welding after fracture, and hence agglomerations [41,42]. The morphology of the planetary ball-milled powders was much finer than those of the roller ball-milled powders, as shown in Figure 13. Some particles were very thin flakes, whereas others were more concave in shape, likely due to the high-energy nature of planetary ball milling imparting large stresses onto the particles.

To reduce carbon contamination within the powders, which may detrimentally impact the magnetic performance of the recycled Sm₂TM₁₇ sintered magnets, a degassing processing was utilised, as shown in Figure 14. The first larger peak showed the removal of hydrogen from the powder. As the powder was more friable in the hydride state [29,30,31,39], the removal of hydrogen was only completed after milling. Hydrogen removal was required before recycled magnet manufacturing, as it has been shown in the previous literature that the presence of hydrogen within interstitial sites during the 2:17R phase can reduce magnetic properties such as (BH)_max [31]. Additionally, the temperature at which the maximum hydrogen desorption occurred for planetary ball-milled powder, as shown in Figure 14, was 130 °C, compared to values of 180–260 °C for HD powders [28,29,30]. This could be attributed to the reduced particle size of the planetary ball-milled powder (<10 µm) compared to HD powder, which is often >100 µm. A smaller particle size may have led to an increased surface area for the desorption reaction to take place over and a reduced diffusion distance for hydrogen to leave the powder particles, hence lowering the required activation energy for desorption.

The second desorption peak occurring at ~260 °C, shown in Figure 14, was attributed to the removal of oleic acid. This method for surfactant removal was similar to that utilised by Leontsev, who used heat treatment in an Ar/air atmosphere to remove valeric acid from SmCo₅ planetary ball-milled powders [46]. The key difference between this investigation and the work by Leontsev was that the removal was completed under a vacuum atmosphere in this study. This atmosphere was chosen to reduce the vapour pressure of the oleic acid and potentially allow for the removal of the organic material occuring at lower temperatures. Despite the success seen by Leontsev in the removal of valeric acid, with a 96% removal when heat-treated in Ar at 400 °C, the residual carbon analysis shown in Figure 15 showed that a significant proportion of carbon remained within samples with ≥1 wt.% oleic acid addition. For oleic acid additions of ≥1 wt.%, the carbon content seemingly saturated at 0.74–0.78 wt.%, with a slight reduction to 0.68 wt.% for the 2.5 wt.% oleic acid addition. Only the 0.5 wt.% oleic acid addition showed a lower carbon content of 0.19 wt.%. In the case of the ≥1 wt.% oleic acid milling trials, the carbon content remaining was very high, likely due to the oleic acid being a very long chain molecule which is difficult to evaporate from the powder surface [41,42]. Additionally, the high-energy ball milling procedure likely broke down the molecule due to high-energy collisions and extreme local heating depositing smaller chain carbon molecules throughout the sample [41,42]. The 0.5 wt.% oleic acid trial did result in the lowest residual carbon content after the heat treatment for surfactant removal, but, as indicated earlier, this trial had a very large D90, which was likely because there was insufficient oleic acid to coat all powder particle surfaces. Therefore, this work shows that an ideal particle size distribution can be achieved using a cyclohexane and oleic acid surfactant mixture with ≥1 wt.% oleic acid, but this comes at the cost of a very large carbon impurity content, even after attempted surfactant removal. For this method of milling, there is a clear exchange between the lower oleic acid content acting to reduce the carbon content but also resulting in a far wider particle size distribution.

Sintering planetary ball-milled powders that utilised cyclohexane and 1.0, 1.5, and 2.0 wt.% oleic acid as surfactants to investigate their densification resulted in compacts with a similar density to the as-received material, 8.3–8.32 g/cm³ compared to 8.4 g/cm³, as shown in Table 5. Though this was within the ideal range for a sintered Sm₂TM₁₇ magnet (8.3–8.5 g/cm³), the composition of the sintered compacts was not ideal, as shown in Table 4. Compared to the as-received material, the Sm content of the powders milled using 1.0, 1.5, and 2.0 wt.% oleic acid were 5.4, 18.8, and 7.5% lower. Additionally, the oxygen content was far higher, by a minimum of 1800 ppm, and the carbon content was an order of magnitude greater in the recycled sintered compacts. The carbon content was high in each of the sintered compacts due to the use of oleic acid during milling; the Sm content dropped likely due to Sm sublimation during sintering and heat treatment [13]. The oxygen content increase was attributed to both oxygen absorption during HD processing/powder handling and the decomposition of the oleic acid used for milling, which itself contained oxygen that may have reacted with the powder [41,42]. Additionally, planetary ball milling inherently generates ultrafine particles which are included in the powder used to form sintered compacts. These ultrafine sub-micron particles are exceptionally susceptible to oxidation and may have contributed to the high oxygen content in the sintered and homogenised compacts [38].

Due to these sources of impurity, sub-optimal microstructures were generated in the sintered and homogenised compacts shown in Figure 16, supported by the EDS analysis of four distinct phases identified in Table 6. The Zr content in the grey matrix phase of the recycled sintered compacts was low/undetectable, likely because of the abundance of the Zr-rich impurity phase, shown as a dark grey phase. This Zr-rich phase was likely ZrC, an impurity compound identified in the previous literature [51,53], with the Zr content of these impurities varying widely from 15.8 to 88.9 wt.% across the recycled sintered compacts. The lack of Sm content within the recycled sintered compacts likely led to the precipitation of the Fe/Co-rich phase, denoted as dark spots throughout the microstructure. By lowering the Sm content, it is likely that, in some areas, the local composition of the microstructure entered the α-Fe/Co phase field [9]. This may explain why the dark phases had 30–35 wt.% Fe, ~60 wt.% Co, and a minor amount of Sm remaining. The high oxygen content indicated in the compositional analysis in Table 4, in addition to the precipitation of the Fe/Co-rich phase, also likely led to the formation of the white Sm-rich phase observed in all recycled sintered compacts. These phases had ~70 wt.% Sm, and often these Sm-rich phases are oxides due to the high affinity of Sm to react with oxygen, forming Sm₂O₃ [51].

To address the Sm losses observed in the sintered compacts, the fully heat-treated recycled Sm₂TM₁₇ sintered magnets utilised planetary ball-milled powder manufactured with a cyclohexane and 1.0 wt.% oleic acid surfactant mix, in addition to 2 wt.% and 3 wt.% Sm-hydride additions. As shown by the compositional data for the 2 and 3 wt.% Sm addition recycled magnets, the Sm content of these samples was far more comparable with the as-received material. The 2 wt.% addition matched the as-received material Sm content precisely, whereas the 3 wt.% addition was 0.7 wt.% greater in Sm content. The result of this was a reduction in the prevalence of the dark α-Fe/Co phase, especially in the 3 wt.% Sm addition sample, as shown in Figure 17. This was linked to the increased Sm content pushing the composition of the material away from the α-Fe/Co phase field [9]. The density of the recycled magnets was also very high, with the 3 wt.% Sm addition having a final density just 0.03 g/cm³ lower than the as-received material, as seen in Table 8.

Crucially, the magnetic properties of the recycled magnets were much lower than the as-received material; see Table 10. One reason for these poor properties was likely the abundance of the Sm-rich phase across the microstructure, which this research suggested was the Sm₂O₃ phase due to the high oxygen content of the sintered compacts, as shown in Table 4. The reason for this high oxygen content has already been discussed, but it may have contributed to lowered magnetic properties, as Sm₂O₃ formation potentially diluted the content of the main ferromagnetic Sm₂TM₁₇ matrix phase [51]. Therefore, with less of the desired matrix phase present, magnetic properties suffered as a result. It is important to note that the Fe/Co-rich phase was randomly distributed throughout the microstructure. Due to precipitation of this phase, the volume fraction of the Sm₂TM₁₇ matrix phase was also likely reduced. This may also have had a detrimental effect of the magnetic properties of the recycled material, as seen in Table 10, by reducing the fraction of the main ferromagnetic phases present.

The poor magnetic properties of the recycled sintered magnets were also linked to the very high carbon content in the recycled magnets, as shown in the compositional data in Table 7. The recycled magnets exhibited 7150 and 6150 ppm greater carbon contents for the 2 and 3 wt.% Sm addition samples, respectively. This led to the formation of the ZrC impurity phase, primarily occurring at the grain boundaries of the recycled sintered magnets, as shown in Figure 17 and supported by the EDS analysis shown in Table 9. Table 9 shows that Zr was not detectable within the matrix phase EDS scans taken, but was found to be in abundance in the dark grey phases at the grain boundaries, with Zr contents ranging from 69.3 to 86.3 wt.% for the areas examined. As a result of ZrC formation, the Zr-rich lamellar required to enrich the 1:5H cell boundary in Cu likely did not form, as has been shown in previous work by Tian [51,53]. Therefore, the required magnetocrystalline anisotropy difference for domain wall pinning to occur at the cell boundaries was not present, and thus the desired coercivity mechanism was not achieved.

Overall, this work has shown that, with a combination of knife milling and surfactant-assisted planetary ball milling utilising a cyclohexane and oleic acid mixture, a particle size distribution within 2–10 µm could be achieved. High density values within the 8.3–8.5 g/cm³ target range were also achieved, with Sm-hydride additions acting to compensate for Sm losses due to oxidation and sublimation during powder handling and sintering/heat treatment, respectively. However, the use of organic surfactants greatly increased the carbon content of the recycled magnets by an order of magnitude compared to the as-received material, even after heat treating the powders to remove the organic material. Reducing the oleic acid content to a <1 wt.% addition did reduce the carbon content, but at the expense of the particle size distribution. The D90 value of the milled powder increased to 44.9 µm for a 0.5 wt.% oleic acid addition, compared to 6.82 µm for a 1 wt.% addition. This article suggests that the increased carbon content caused an abundance of the ZrC impurity phase to form, stripping the matrix of Zr. As a result, the Zr-rich lamellar, needed for providing diffusion pathways for Cu to reach the 1:5H cell boundary phase during heat treatment, was not formed. Therefore, the domain wall pinning coercivity mechanism required for Sm₂TM₁₇ sintered magnets to retain their magnetic properties was not formed, and the magnetic properties of the recycled material were far lower than the as-received material.

Future work in this area will focus on improving the process for removing organic surfactants from powders after milling or employing milling techniques that use very little/zero surfactant addition, such as inert gas jet milling, for powder size reduction.

5. Conclusions

This work investigated the powder processing required to manufacture recycled Sm₂TM₁₇ sintered magnets from HD powder feedstock in a ‘short-loop’ magnet-to-magnet recycling process. A specific emphasis was placed on the fine milling processes required to achieve a 2–10 µm particle size distribution and reach sintered magnet densities of 8.3–8.5 g/cm³.

HD powders generated at 18 bar or 2 bar and at 100 °C for 72 h were subjected to knife milling, roller ball milling, and planetary ball milling trials, with the effect of different milling times and surfactants on the final particle size distribution being investigated.

Knife milling 18 bar and 2 bar powders showed that 70% and 82% reductions in D50 occurred after just two minutes of milling. Roller ball milling trials showed that a mixture of cyclohexane and oleic acid allowed for the greatest particle size reduction in D10, D50, and D90 (80–92%) over 7 h of milling compared to all other surfactants tests. The long carbon chain and the polar (-COOH) functional group present within oleic acid likely prevented metal-to-metal powder contact and reduced agglomerations.

Knife milling HD for two minutes and then planetary ball milling this powder in cyclohexane and 1.0 wt.% oleic acid resulted in a particle size distribution of 1.3–6.8 µm, generating a sintered compact with just a 0.08 g/cm³ lower density than the as-received material.

However, Sm losses of 5.4–18.8% were detected across various milling trials using 1.0, 1.5, or 2.0 wt.% oleic acid surfactant additions. These losses were a result of oxidation and sublimation during processing, which led to extensive Sm₂O₃ formation and the segregation of α-Fe/Co from the matrix phase. Sm-hydride additions at the milling stage of 2 or 3 wt.% mitigated the formation of the α-Fe/Co phase. Carbon content in the densest recycled magnet was 6150 ppm greater than the as-received material. This formed ZrC regions where EDS analysis indicated a Zr content of up to 86.3 wt.%. This phase likely inhibited the formation of the magnet cellular structure and prevented domain wall pinning, substantially reducing magnetic properties.

Future work will investigate removal techniques for organic surfactants from milled powders or trial processes such as inert gas jet milling for achieving a particle size distribution between 2 and 10 µm.