Microstructure and magnetic properties of Mn-Al-C permanent magnets produced by various techniques

BulkMn52Al46C2 in t-phase was prepared by vacuum inductionmelting and used as precursor for the production bulk permanent magnets by suction casting and hot-extrusion. Part of the precursor alloy was mechanically milled into a t-phase powder and used as precursor for production of samples by electron beam melting, hot-compaction and high pressure torsion processes. The microstructure and magnetic properties of all samples were investigated and correlated. It was found that the mechanical deformation enhances coercivity, up to 0.58 T, while the absence of this strain is beneficial for magnetization. Among the observed techniques, hot extrusion and high pressure torsion have shown promising possibilities to further developMn-Al-C as permanent magnets. However, it should be taken into account the challenges related to design a proper processing window for hot extrusion and the limitation of HPT regarding the absence of texture.


Introduction
Mn-Al alloys belong to the well-known material systems, which have been intensively studied since the late 60's as perspective candidates for use as permanent magnets (PM) [1,2], but then faded into oblivion after the discoveryof the Nd-Fe-B magnets in the 80's [3]. The abrupt increase of the rare earth elements price in 2011, along with the growing demand on permanent magnets, triggered the renewed interest in this material system in the past years [3,4]. This combination of circumstances has motivated many scientists to not only search for novel materials, but also to reinvestigate already know materialsystems like MnAl and work further on the development of new fabrication techniques, involving novel routes of powder metallurgy [5][6][7], hot deformation [8][9][10][11][12] and additive manufacturing [13][14][15][16] which can provide hard magnetic properties for these non-rare-earth relatively cheap permanent magnets [4]. The MnAl material system is characterized by the use of non-critical elements, low cost and reasonable magnetic properties [2,3,17]. These characteristics are pointed as crucial to use and develop this material system as "gap magnets" [18,19]. This terminology is adopted to show that this material system has the intrinsic magnetic properties suitable to fill the magnetic performancegap (in the meaning of energy product À BH max ) between the cheap but low performance ferrites (BH max ≈ 40 kJ/m 3 ) and the expensive, rare-earth based high performance magnets (N45 grade Nd-Fe-B-BH max ≈ 450 kJ/m 3 ) [20]. The intrinsic properties of the Mn-Al compound show a high anisotropy constant (K 1 of 1.7 MJ/m 3 ), relatively high anisotropy field (H A between 4.0 and 5.5 T), moderate saturation magnetization (M S of 0.75 T) which yields in the theoretical value of BH max ≈ 95 kJ/m 3 [20,21].
The magnetic properties of Mn-Al are related to the solely ferromagnetic metastable phase in this material system, the tetragonal t-phase (L1 0 type structure). As a general rule, the processing window has to be design in a way to produce pure ferromagnetic phase in the Mn-Al system in order to maximize magnetization. This can be achieved by two common methods: (i) quenching from the melt or from high temperatures (around 1100°C) to stabilize the parent e-phase (hexagonal structure-HCP) followed by annealing at moderate temperatures (around 500°C); (ii) high temperature annealing followed by controlled cooling,whichresults in t-phase nucleation and growth during this last step. As mentioned, the ferromagnetic phase is metastable, which means that long annealing times or high temperatures can lead to decomposition of the t-phase and nucleation of the non-magnetic stable g 2 and b-phases [20,22,23]. The formation of the ordered L1 0 t-phase, from the parent chemically disordered HCP e-phase,is reported to happen through two main mechanisms: massive and displacive. The first one is associated with diffusional nucleation at the e-phase grain boundaries followed by growth via interphase boundary motion [24]. The latter is relatedtothe shear of HCP atomic planes with short-range diffusion [25]. It can also be the case that these two mechanisms happen simultaneously, as observed by insituexperimentsat temperatures around 500°C or above [26]. The growth of the t-phase develops lattice/microstructural defects which affects substantially the magnetic performance, as will be discussed later [24].
To increase the stability of the ferromagnetic phase, carbon is often added as interstitial dopant to the t-phase compositional range, Mn 50+x Al 50Àx (51 x 58), at the expenses of decreasing the Curie temperature (T C ) from 630 to 570 K [20,22]. But only phase purity is not a guarantee for optimized magnetic properties, since different manufacturing processes affect the microstructure and, consequently, the extrinsic magnetic properties.
Different studies have shown the complex relation between the several types of microstructural defects of t-MnAl phase and the magnetic properties, in terms of remanence, coercivity and, consequently, the BH max value [10,21,[27][28][29][30][31][32]. Among the possible defects in this material system, the existence of twin boundaries is often related to one of the major difficulties to obtainhighly textured samples and improved remanence in Mn-Al magnets [31]. This is linked to the high density of twin variants created during the t-phase formation, which will prevent the achievement high degree of texture along the easy magnetization axis since randomly multi-variant grains will be formed. Moreover, in addition to twin boundaries, other defects and metallurgical variables whichwere reported, namely: stacking faults, antiphase boundaries, dislocations, grain size and lattice strain; are related and affect coercivity, as reported in the literature [10,21,31,32]. The effect of each specific defect on the magnetic properties is still under investigation, as different characterization techniques are necessary to obtain qualitatively and/or quantitatively data about the densityand distribution of these defects. Furthermore, these microstructural features are very often reported to coexist, which increase the challenge to understandthe coercivity mechanism in t-Mn-Al-based compounds.
Based on these factors, a complete understand of the processing route on the phase stability, microstructure and magnetic properties are indispensableto optimize the Mn-Al material system to be suitable for permanent magnet applications. Therefore, this present work is focused on the use of five different processing techniques for production of Mn-Al-C bulk samples and the correlation between microstructure and magnetic properties for each of these processes.

Sample synthesis and characterization
In the following subsections will be given experimental details of the different processing routes discussed in this work, according to the schematic presented in Figure 1. The master alloy composition, Mn 52 Al 46 C 2 , chosen to perform these processes was based on the low ratio between Mn and Al atoms, to avoid decreasein magnetization from the Mn-Mn antiferromagnetic interaction, and interstitial C doping to prevent decomposition of the metastable hard magnetic L1 0 t-phase.

Vacuum induction melting
Vacuum induction melting (VIM) followed by casting into graphite crucible is a well-known fabrication technique, especially for Mn-Al production [2]. Mn 52 Al 46 C 2 alloy was prepared by melting pure elements (purity above 99 wt. %) using a single chamber vacuum induction melting furnace (ConsarcCorp.). Before melting, the chamber purged with argon 5 times in order toreduce the oxygen content to a minimum level. The melting occurred under protective atmosphere of Ar (purity of 99.999%) and the alloy was kept in a molten state for 2 h for homogenization before casting into a graphite crucible. The so obtained bulk alloy was further characterized and used as a precursor in all subsequent processing routes, as shown in Figure 1, ensuring the same chemical composition for a direct comparison between the processes.

Suction casting
Part of the VIM Mn-Al-C bulk sample was subject to arc melting and, when in molten state, the alloy was rapidly "sucked" into a cylindrical cooled copper mold of 10 mm diameter. This is a method of casting that ensure high 95 cooling rates, in the order of 10 3 Ks À1 , and allows to obtain microstructural refinement and non-equilibrium phases. This technique has been reported for different intermetallic rare-earth based compounds [33][34][35]. However, there are no reports of the use of suction casting for Mn-Al, to the best of authors knowledge. The suction cast samples were subjected to the following heat treatment procedure: annealing for 72 h at 1100°C, followed by water quenching and subsequently annealing for 30 min at 550°C.

Hot deformation/hot extrusion
The effect of hot extrusion was investigated by Matsushita Electrical Industrial company in 1977, it was reported values of H C = 0.30 T, (BH) max ≈ 56 kJ/m 3 for Mn-Al-C compound [36]. Texture along the hot extrusion direction was observed from the anisotropy behavior in the magnetic measurements. This result still remains as state of the art in terms of magnetic performance and difficulties on reproducing it were reported in literature (references  [37]. In thiscomprehensive study, it was highlighted the role of Ni-doping on the improving the plasticity of Mn-Al-C, which is imperative for the hot extrusion process. A press (Beckwood corp.), with tools heated up to 550°C, was used for the extrusion process. The diameter of the die used was 25 mm while the initial diameter of the samples was 50 mm, resulting in a ratio of the starting andfinal cross section area of 4. Prior extrusion the precursor alloy was heated up to 500°C and kept at this temperature for ca. 20 min. Afterwards the preheated alloy was placed into the press and pressure of up to 225 MPa was applied However, due to the rigid behaviour of the intermetallic phase at this temperature, the applied pressure was not sufficient to successfully complete theextrusion process.
Even though a small volume of the sample was extruded, a section of this volume was used for further characterization.

Synthesis of the powder precursor alloy
Various techniques have been previously applied to prepare Mn-Al basedpowder precursor for permanent magnets, including mechanical alloying, gas atomization and mechanical milling etc. [5][6][7]17,38,39]. Among this methods, mechanical milling (MM) seems to be the most costeffective and efficient approach for preparation of powder with enhanced coercivity. Therefore, powder was produced by mechanical milling of the VIM obtained Mn-Al-C alloy, by using a planetary ball mill Fritsch-Pulverisette 6. The milling was done in protective gas atmosphere, for 2 h at rotation speed of 250 RPM, using 10:1 ball to powder mass ratio, and 10 mm hardened steel balls.

Hot compaction
After milling, the randomly shaped powder and flake like particles, with D 90 100 mm was obtained. The powder was sieved and separated in fraction by particles sizes. The fraction with particle size below 100 mm was used for hot compaction. The hot compaction has been performed using a standard hydraulic hot press machine and closed die with inner diameter of 300 mm and height (after compaction) of 800 mm. The compaction was done at pressure of 150 MPa and die temperature of 450°C under protective argon atmosphere with holding time of 20 min.

High pressure torsion
The influence of plastic deformation as well as of crystal defects related to crystals plasticity on the formation magnetic properties of Mn-Al based alloys was intensively studied through recent years [30,31,40,41]. The high pressuretorsion (HPT À Walter Klements GmbH) process was used to compact and deform t-phase Mn-Al-C powder (particle size below 80 mm) into disc shaped samples of 10 mm diameter and 1 mm height. The pressure used was 4 GPa and 50 revolutions were applied with 1 RPM. The process was performed at room temperature and the tools were kept below 50°C during the process.
During the HPT process, the strain generated within the sample is proportional to the radius, which means that the edges of the disc will be higher deformed than the center, leading to an inhomogeneous microstructure and, in this case, a magnetic properties gradient [42]. For this reason, we evaluated three regions of the sample: center (R 0 ), half radius (R 0 . 5 ) and on the edge(R 1 ).

Additive manufacturing
An Arcam EBM A2 machine (Arcam EBM, Sweden) was used for the additive manufacturing process. The EBM system has reduced working volume, optimized for the experiments with small powder batches e.g., for testing newalloys [43]. Processing parameters used: layer thickness 100 mm; line offset 100 mm; maximum beam current (EB) 30 mA and average chamber temperature of 830°C. More details about the processing and optimization can be found in [15].

Characterisation techniques 2.3.1 SEM/Kerr microscopy
To investigate the present phases and microstructure, the samples were analyzed using scanning electron microscope (SEM ÀTescan VEGA 3 and FEG-SEM JEOL JSM-7600F) with backscattered electrons (BSE) detector. Energy-dispersive X-ray spectroscopy (EDS) measurements were performed to quantify chemical composition of the present phases.In addition, EDS measurements were taken in larger portions of the sample (area scans) to ensure the Mn/Al ratio was preserved after each processing technique. In all cases, the overall error/deviation from the initial aimed stoichiometry was around 1 at%, within the limits of the EDS detector. The carbon content was evaluated qualitatively by comparing the different samples, since the quantitative determination through EDS is challenging and can be influenced by different measurements artifacts (specimen surface or SEM chamber contamination).

Magnetisation measurements
Isothermal magnetization measurements were performed using PPMS-VSM(Quantum Design PPMS-14), at room temperature, under applied magnetic field up to 3 T. No corrections regarding demagnetizing factor were made.

Transimission electron microscopy-TEM
Prior to TEM analysis, lamellas were prepared using plasma FIB (TescanS9000X) from samples in the VIM, powder and HPT states. The TEM TitanThemis G2 60-300 (FEI/Thermo Fisher) with aberration correction and rapid camera with 4K resolution was used to analyze the lamellas.

Vacuum induction melting (VIM)
The sample after VIM shows ferromagnetic behavior, as presented in Figure 2b with relatively high magnetization at 3 T applied field (M 3T ) of 100 A · m 2 /kg, in agreement with the microstructure shown in Figure 2a, in which only t-phase is observed. It is important to notice the small value of coercivity (H C < 0.03 T), which is directly related with the coarse microstructure with high density of twin boundaries and the absence of defects which pin the domain wall motion, as similarly reported by [9,31]. The resultant phase/microstructure arises from the casting process, where the cooling is not fast enough to stabilize the high temperature e-phase and not slow to promote the stable g 2 -and b-phases. Since an intermediate cooling rate is achieved during VIM, the nucleation and growth of the t-phase was achieved, leading to a single-phase sample.

Powder preparation À ball milling
Mn 52 Al 46 C 2 powder was prepared from the VIM bulk precursor, through ball milling, for further processing routes. As can be seen from the SEM-BSE image on Figure 3a, the powder consists of particles smaller than 100 mm of different shapes, including spherical and flake like morphology. The correspondingmagnetization measurement is shown in Figure 3b, in which can be seen a M 3T of 90 A · m 2 /kg and H C of 0.12 T. This shows a substantial increase in coercivity from the VIM bulk precursor, H C < 0.03 T, and a slight decrease in the M 3T . This behavior is commonly attributed to the plastic deformation and strain of the produced powder during the mechanical milling, which is more evident inlonger milling procedure, as reported by [7]. It worth mentioning that longer and more intensive milling can be used to produce powder that shows even higher coercivity, reaching values around 0.50 T, but this comes with a decrease in magnetization and even a decomposition of the t-phase, as reported by Law et al. [39].

Suction casting
Through suction casting technique is expected to have a higher cooling rate compared to VIM method. Indeed, as shown in Figure 4a, the microstructure differs from the VIM sample where is observed the presence of three distinct phases: t-, g 2 -and e-phases. The e-phase arises from the higher cooling compared to VIM, which is also in agreement with the refinement of the t-phase. The high fraction of g 2 -phase present in the sample indicate that further annealing needs to be done to further homogenize and maximize the t-phase fraction. After annealing at 1100°C, temperature region of e-phase stability, for 72 h with subsequently quenching into water, leads to a change in the microstructure, ascan be observed in Figure 4b. In the outer shell of the cylindric sample is observed the  presence of e-phase because of the higher quenching rate on the edges of the sample, while in the interior we observe the t-phase. To further maximize the ferromagnetic phase fraction, an additional annealing step at 550°C for 30 min was adopted, which led to a microstructure of pure t-phase, as can be seen in the Figure 4c and from the magnetization values in Figure 4d. Similar to the VIM sample, features like twin boundaries and micro twins are observed in the microstructure. Accordingly, the magnetization measurements shown in Figure 4d agrees well with the microstructure observation. In the as suction casted state, small fraction of t-phase, the low magnetization represents the nonmagnetic g 2 -and e-phases, which are in large fraction in this state. By annealing at 1100°C, there is an increasing the t-phase fraction and the change on the curve shape to a ferromagnetic behavior with coercivity around 0.04 T and M 3T of 70 A · m 2 /kg. After annealing at 550°C, we promote the e ! t transformation, leading to a higher ferromagnetic phase fraction resulting in a higher magnetization (M 3T of 90 A · m 2 /kg) with similar coercivity. It is worth to mention that only a slight increase in the coercivity value was observed from the suction casted sample when compared to the VIM sample, from below 0.03 T to around 0.04 T.

Beam-based powder bed additive manufacturing
The part of the current work devoted on the Additive Manufacturing of MnAl-C magnets by Beam-based Powder Bed technology is a continuation of the work published previously [14] and [15], were ball milled binary Mn-Al alloy was used as a precursor. The samples were printed using a modified ArcamA2 EBM machine and ballmilled Mn-Al-C alloy with particles size of 50-60 mm (see Fig. 1). As already discussed above, the use of C doped powder was expected to improve the phase stability and enhance the magnetic properties of the printed magnets. The experiment has shown that contrary to the induction melting process, after electron beam melting the magnetic t-phase is not predominant. This can be due to the use of not optimal process parameter settings like scanning rate, hatching distance, process temperature, etc. Therefore, further investigation of the influence of process parameter settings on the microstructure forming aspects responsible for formation of proper magnetic properties should be done. On another hand, the amount of t-phase in the printed samples can be increased by post processing. Magnetization measurements confirmed (Fig. 5) that the properties of the printed magnets can be fully recovered by two step annealing at 1100 and 500°C. However, the properties cannot be further improved. Figure 6a and b presents the microstructure and magnetic properties of thehot deformed (slightly extruded) Mn-Al-C sample, respectively. It can be observed that the hot deformation has caused dynamic recrystallization, which explains the microstructure refinement while preserving the t-phase which can be also observed by the high M 3T value of 97 A · m 2 /kg. The coercivity value has substantially increase when compared to the previous processing routes, going from 0.03 to 0.12 T. This shows that the combination of grain refinement and the induced defects during plastic deformation, such as dislocation, can improve coercivity without affecting the magnetization, which is in agreement to results reported by Thielsch et al. [9] and Feng et al. [37].

Hot extrusion/deformation and hot pressing
As it was mentioned above, because of the brittle nature of this compound, the sample has been deformed and only slightly extruded at the beginning of the extrusion die. This show that the extrusion temperature of 500°C was not enough to induce plasticity for deformation process, but, as mentioned previously, higher temperatures could lead to decomposition of the metastable ferromagnetic phase. For this reason, a further detailed study onthe influence of chemical composition, process temperature, phase stabilization and magnetic properties would give an insight of the processing window for this material system, but this is beyond the scope of the present work.
Alternatively, with the aim to overcome the technical difficulties presented during the hot extrusion route and produce a refined t-phase microstructure,the hot compaction of the produced ball milled powder has been performed. The obtained hot-compacted samples had the microstructure as shown in Figure 7a. As can be seen, the metastable phase partially was decomposed during the process, but the remaining t-phase shows a refined microstructure when compared toVIM and suction casting processes. It is important to notice that was observedporosity in this sample, which led to a sample with density of 4,3 g/cm 3 , around 83% relative to the theoretical density, measure by Archimedes' principle.
The magnetization curve of the hot-pressed sample, Figure 7b, shows a magnetization M 3T of 50 A · m 2 /kg, related to a decrease of the ferromagnetic phase fraction. On the other hand, the coercivity has improved, reaching a value of 0.21 T, showing again, as in the case of hot extrusion, that grain refinement and defects induced during the process contribute to improve this property.

High pressure torsion À severe plastic deformation
Another deformation processing route that has been investigated in this work was high pressure torsion (severe plastic deformation), but different from the 295 previous ones, the deformation was carried out at room temperature. As can be seen from the magnetization measurements, Figure 8c, there is a gradient on the coercivity and magnetization values across the diameter. The increase in  the coercivity towards the edge of the sample (R 1 ), from 0.22 to 0.58 T, can be associated to the higher strain that this area is subjected during the HPT process and, consequently, higher defect density. The high values of coercivity obtained by HPT are similar to the ones reported by other authors and also similar for powders produced by high energy ball milling, in which the powder is also subjected to high a deformation degree.
On the other hand, the magnetization value decreases because of higher Mn-Mn antiferromagnetic interaction caused by internal stresses, which has been shown also for other processing routes [8]. In addition to the decrease in the absolute magnetization value, it is possible to notice a step-like decrease in the magnetization value. This behavior is related to the appearance of secondary phase caused by the stress-driven decomposition of the metastable tinto b-phase, as confirmed by SEM-EDS (see Fig. 8b). The nucleated phase also contributes to the decrease in the absolute magnetization value since it is nonmagnetic.
To correlate the obtained magnetic results with microstructure, SEM analysis were made in the center (R 0 - Fig. 8a) and at the edges (R 1 -Fig. 8b), revealinga gradient on the microstructure morphology. Kerr microscopy was also used but it was not possible to distinguish features, like grain size or twins, as shown for the previous processing routes. For this reason, TEM analysis was carried in the as cast (VIM), powder precursor and in the HPT (R 1 region) states, as shown in Figure 9. In the as cast state (VIM), as shown previously in Figure 2a and also in Figure 9a, the microstructure is composed from coarse grains (micrometer range) with well-defined twins and twin boundaries. As for the powder, Figure 9b, it is possible to notice a difference in the morphology and size of the grains along the presence with a higher density of defects.  The sample after HPT, Figure 9c and d, shows even further microstructural refinement and change in morphology, being difficult to distinguish individual grains. But it is noticeable the increase of defect density, including polytwinned microstructure with high density of dislocations. The high density of defects, especially dislocations, can hinder the domain wall motion leading to a higher coercivity, in accordance with the magnetic results presented in Figure 8c, and as previous reported by [30,31,41].
The results indicate that the coercivity is strongly related to the defects, and the density of such defects, that can pin the domain wall motion in the magnetization reversal process. Even though plastic deformation/severe plastic deformation is beneficial for this figure of merit, the magnetization value decreases due to internal stresses, as seen and explained previously. For this reason, a compromise between these two properties can be established by using HPT or even with the addition post annealing processing step. These possibilities can be further studied and explored to achieve a balanced magnet in terms of magnetization and coercivity, but the absence of texture coming from the HPT still remains a challenge for increasing remanence and, consequently, the energyproduct BH max . Table 1 summarizes the magnetic characteristics of Mn-Al-C permanent magnets produced by various techniques that were found in literature. The results of the experimental findings presented in this work were also added to the Table 1 for comparison.

Conclusion
To investigate the influence of different production methods of bulk Mn-Al-C based permanent magnets on their magnetic properties, a batch of Mn 52 Al 46 C 2 precursor alloy was prepared by vacuum induction melting for further processing. MnAl based permanent magnets were produced by applying different degree of mechanical deformation and heat treatment. The resulting permanent magnets were studied with respect of the correlation between microstruture and magnetic properties.
It was clearly shown that the phase purity of the samples is important, but not the ultimate factor determining the magnetic properties of the MnAlC samples. Beside the amount of t-phase, the density and the type of microstructural defects can significantly affect the extrinsic magnet properties of this material system. The samples that were subject of plastic deformation during the manufacturing process (hot extrusion, high pressure torsion and milling) have higher defects density. By having these higher defect density, especially dislocations,high coercivity values were obtained because these defects can act as pinning center than can hinder the reversal domains movement. On the other hand, the strain related to these processes can reduce magnetization due to Mn-Mn antiferromagnetic interaction and, in some cases, can also leads to a decomposition of the ferromagnetic metastable t-phase.
Techniques related to melting with slow cooling (VIM) or with post annealing processing (suction casting) have  shown the highest magnetization values. Based on the obtained results, it can be concluded that introducing texture along the easy magnetization axis can (without negative effect on the magnetization) maximize the remanence and consequently the energy product BHmax of the magnets. From all investigated methods, as previously reported, only samples produced by hot extrusion has shown a significant degree texturing. Due to the experimental constraints (temperature and pressure), this could not be fully reproduced and therefore only isotropic samples are reported in this work. Among the explored techniques, hot extrusion and high pressure torsion have shown promising possibilities to further develop Mn-Al-C as permanent magnets. However, it should be taken into account the challenges related to design a proper processing window for hot extrusion and the limitation of HPT regarding the absence of texture.