An approach by using the γ′-Fe4N precursor was proposed to prepare a bulk magnet with high α″-Fe16N2 content, and the feasibility for high N content α′-Fe(N)/α′-Fe8N was demonstrated in ribbon samples. γ′-Fe4N decomposed when it was heated to a high temperature, and then, a supersaturated γ-Fe(N) phase formed because of N atoms diffusing out from the γ′-Fe4N lattice. The short-lived supersaturated γ-Fe(N) had the face-centered cubic lattice, and its nitrogen content is beyond 3.0 wt. %. The supersaturated γ-Fe(N) acts as a precursor of the martensite transformation for a body-centered tetragonal phase. After the cryo-treatment, which was essential to complete the martensite transformation, a mixture of α′-Fe(N) and α′-Fe8N was obtained. The N content of more than 2.5 wt. % in the samples was achieved.

α″-Fe16N2 has been regarded as one of the most promising candidates for rare-earth-free permanent magnets.1 Its saturation magnetization can be up to 270 emu/g. Its reported magnetocrystalline anisotropy constant is between 1.0 × 107 and 1.8 × 107 erg/cm3.1 Bulk α″-Fe16N2, an ordered body-centered tetragonal (bct) phase, was prepared by tempering α′-Fe8N (bct) at a low temperature (<200 °C).2–6 N atoms in α′-Fe8N lattice randomly distribute in the interstitial sites, and then, the tempering (with/without stress) makes N atoms orderly arrange, resulting in the transformation from an N disordered α′-Fe8N to an N ordered α″-Fe16N2 phase.7,8 α′-Fe8N is a quenched product of a martensite transformation, from a face-centered cubic (fcc) γ-Fe(N) solid solution (austenite). According to the Lehrer diagram, the formation of γ-Fe(N) occurs in an optimal combination of nitrogen potential and high temperature.9 Jack used the nitrogenizing temperature of 750 °C to get γ-Fe(N), and then, γ-Fe(N) was immersed into a quenchant, as shown as route ① in Fig. 1.2–6 Martensite transformation occurred during the quenching, and γ-Fe(N) transferred into α′-Fe8N. After tempering, α″-Fe16N2 was obtained.

FIG. 1.

Fe–N phase diagram and different routes of martensite transformation. The first route as indicated in a yellow line is Jack’s approach,4 and the second route as indicated in a red line is our proposed quenching approach from γ′-Fe4N based ribbons following rapid heating from room temperature. The Fe–N phase diagram is reported by Jack in 1951. The N contents at three intersections (labeled as green spots) are 2.4, 2.8, and 4.5 wt. %, respectively.

FIG. 1.

Fe–N phase diagram and different routes of martensite transformation. The first route as indicated in a yellow line is Jack’s approach,4 and the second route as indicated in a red line is our proposed quenching approach from γ′-Fe4N based ribbons following rapid heating from room temperature. The Fe–N phase diagram is reported by Jack in 1951. The N contents at three intersections (labeled as green spots) are 2.4, 2.8, and 4.5 wt. %, respectively.

Close modal

However, the maximum solubility of N in γ-Fe(N) is 2.4 wt. %, which is smaller than that of α″-Fe16N2 (N content in α″-Fe16N2: 3.0 wt. %), and then, the α-Fe phase accompanies α″-Fe16N2 after the quenching and tempering processes even there is no loss of N atoms.5,6 As shown in Fig. 1, the dual-phase combination of α-Fe and α″-Fe16N2 is expected in the Fe–N phase diagram (1.3–2.6 wt. %); however, the α″-Fe16N2 content reported in the literature, thus far, is always less than 50% in the matrix.5,6 Both saturation magnetization and magneto-crystalline anisotropy of Fe are smaller than those of α″-Fe16N2. Fe significantly undermines the hard-magnetic properties of the synthesized samples. The degradation of hard-magnetic properties becomes even worse with the increase of the Fe content. Thus, it is critical to search for an approach that increases the N content in the γ-Fe(N) precursor before the quenching process.

As shown in Fig. 1, there are three single phases, γ′-Fe4N, ε-Fe2-3N, and ζ-Fe2N, which all have a much higher N content than 3 wt. %. If the decrease in the N content can be controlled, those phases will provide an exact 3 wt. % N content to form α′-Fe8N and then α″-Fe16N2. In this paper, we proposed and demonstrated a new approach (route ② in Fig. 1) to use γ′-Fe4N (N content is 5.9 wt. %) as a precursor, to synthesize α′-Fe8N. We systematically investigated the temperature-dependent quenching effect of the γ′-Fe4N based ribbons, and a α′-Fe(N)/α′-Fe8N phase was revealed after a cryo-treatment.

As shown in Fig. 1 labeled as route ②, γ′-Fe4N based ribbons were heated from room temperature to a high temperature (550–705 °C) and then quenched in the cold water. The experimental recipe has six steps, as shown in Fig. 2, including raw materials selection, ingots and ribbons preparation, microstructure buildup (MB), nitrogenizing, quenching, and cryo-treatment. The Fe thin foil (99.99%), Cu powder (99.5%), and Fe2B powder (99.9%) were used to prepare the ingots by using an arc melting system. Cu and B were doped into the Fe matrix to improve the microstructure (Ref. 8 and papers cited therein).8 B was used to control the grain size, grain boundary, and crystallinity of the alloy. Non-magnetic Cu precipitated in the grain boundary to decouple the ferromagnetic grains. The Cu and Fe2B powder were wrapped by the Fe foil and the vaporization of powders was then suppressed during the melting. The ingots were flipped several times until there was no color change. Then, the ingots were used to prepare Fe–Cu–B ribbons by using a melt spinner. Fe-3 wt. %Cu-2 wt. % B ribbons were synthesized with a thickness in the range of 30–50 μm. Figure 2-2(a) shows the morphology of the ingot, with the grain size no less than 10 µm. Large rod-like grains of the ingot are revealed. This is due to the relatively slow cooling rate of the arc melting system. The ribbon shows the branch-like structure with a relatively small size, and the composition inhomogeneity was greatly reduced in the ribbons [Fig. 2-2(b)].

FIG. 2.

Six-step experimental process of the formation of α′-Fe(N)/α′-Fe8N, and the SEM images of the samples prepared at each stage: 2-2(a) ingot prepared by arc melting, 2-2(b) ribbon prepared by melt spinner, 2-3(a) microstructure after build up, 2-3(b) zoom-in images of 2-3(a), 2-4(a) sample after gas nitrogenizing, 2-4(b) zoom-in image of 2-4(a), and 2-5(a) sample after the quenching related to the martensite transformation.

FIG. 2.

Six-step experimental process of the formation of α′-Fe(N)/α′-Fe8N, and the SEM images of the samples prepared at each stage: 2-2(a) ingot prepared by arc melting, 2-2(b) ribbon prepared by melt spinner, 2-3(a) microstructure after build up, 2-3(b) zoom-in images of 2-3(a), 2-4(a) sample after gas nitrogenizing, 2-4(b) zoom-in image of 2-4(a), and 2-5(a) sample after the quenching related to the martensite transformation.

Close modal

In the third step of a microstructure buildup process (MB), the ribbons were annealed at 705 °C for 15 min to precipitate Cu at grain boundaries and quenched into cold water to control the grain size. The Fe–Cu–B ribbons were then solutionized at 400 °C for 4 h to homogenize the microstructure. Figure 2-3(b) shows the microstructure after the buildup process, and a typical grain is circled by a dashed line. The grain size is about several hundred nanometers.

Then, a gas nitrogenizing was used to prepare face-centered cubic γ′-Fe4N based ribbons. Fe–Cu–B ribbons were polished with a sandpaper first to remove the severely oxidized layer generated in the third step and then put into a tube furnace with a tube diameter of 1 in. The ribbons were reduced at 550 °C with H2 flow of 120 SCCM. The hydrogen reduction helped to remove remained oxide phases on the surface of these ribbons. Fe–Cu–B ribbons were then nitrogenized for 36 h at 500 °C in a gas combination of NH3 and H2. After nitrogenizing, the ratio of NH3 to H2 was kept until the temperature returned to the room temperature. Then, γ′-Fe4N formed with Cu and B in the grain boundary.

The fifth step is the second quenching process. A box furnace was set at a temperature in the range of 550 and 705 °C. The ribbons were put in the furnace for 2–30 min. Afterward, the ribbons were rapidly put into the distilled water. After the step, we did a cryo-treatment by using a physical properties measurement system (PPMS) whose temperature could be as low as −268 °C with the temperature ramping rate of 12 °C/min.

The structural properties of the samples were measured by x-ray θ-2θ diffraction (XRD) and scanning electron microscopy (SEM). To characterize the content change of the γ′-Fe4N [or fcc-Fe(N), bct-Fe(N)], the intensity ratio of the XRD peaks of γ′-Fe4N (111) or γ-Fe(N) (111) to α-Fe (110) was used. The magnetic properties were measured by a vibrating sample magnetometer (VSM) and the physical properties measurement system (PPMS). The maximum field in the VSM measurements is 5 kOe. The temperature-dependent saturation magnetization was measured by using PPMS, and the applied field is 20 kOe.

Samples from steps 1–5 were prepared for SEM observation using a mechanical polish process up to a final step of 0.2 μm using diamond pastes. High resolution SEM images were collected using Hitachi SU8230 Field Emission Gun microscopy. The composition of the samples was measured by the energy dispersive x-ray spectroscopy (EDX) in the SEM.

Figure 3 shows the XRD θ-2θ scans of the samples from the as-prepared ribbons to the samples after the cryo-treatment. As shown in Fig. 3(a), the as-prepared ribbon is composed of α-Fe, as well as a tiny amount of γ-Fe whose peak was around 45.5°. Melt spinning was a rapid solidification process and then a small amount of fcc phase formed at a high temperature (∼1500 °C) was left. The γ-Fe disappeared after the microstructure buildup process, and only α-Fe left [Fig. 3(b)].

FIG. 3.

XRD θ-2θ scans of the experimental recipe of each step: (a) as-prepared ribbon by melt spinner, (b) after microstructure-buildup (M-B), (c) nitrogenizing in the mixed gases, (d) 665 °C (3 min) quenching in cold water, and (e) after cryo-treatment (Cryo-T) in low temperature. Four phases were labeled by the symbols. The iron oxide peaks located at 33.3° and 35.4° were not labeled in the plot.

FIG. 3.

XRD θ-2θ scans of the experimental recipe of each step: (a) as-prepared ribbon by melt spinner, (b) after microstructure-buildup (M-B), (c) nitrogenizing in the mixed gases, (d) 665 °C (3 min) quenching in cold water, and (e) after cryo-treatment (Cryo-T) in low temperature. Four phases were labeled by the symbols. The iron oxide peaks located at 33.3° and 35.4° were not labeled in the plot.

Close modal

The nitrogenizing conditions follow the Lehrer diagram.9 In the current experiments, we adjusted the combination of NH3 and H2 to make the intensity ratio of γ′-Fe4N (111) (2θ = 41.2°) to α-Fe (110) (2θ = 44.6°) in the range of 1–5 [Fig. 3(c)]. Thus, γ′-Fe4N layers have the similar thickness. Fe (110) peak was used as a reference for the intensity change of the diffraction peaks due to phase transformation. After the second quenching, a strong Fe(N) peaks appear at 43.13° and 50.27°, as shown in Fig. 3(d). It is FeN0.09 (2.21 wt. % N). The intensity ratio of γ-Fe(N) (111) to γ-Fe(N) (200) is 2.2 in the powder diffraction, but the present intensity ratio is about 22.5. Thus, the ribbon was highly textured.

Table I summarizes the lattice parameters and XRD diffraction peaks of the γ-Fe (fcc), γ′-Fe4N (fcc), γ-Fe(N) (fcc), α′-Fe(N) (bct), α″-Fe16N2 (bct), and α-Fe (bcc). The lattice parameters vary with the N content, having a similar dependence like steel Fe(C) of lattice parameters on the interstitial atom content. A wide peak implies that there might be multiple Fe(N) phases with slightly different N content.

TABLE I.

Lattice parameters and 2θ of x-ray diffraction peaks of γ-Fe (fcc), γ′-Fe4N (fcc), γ-Fe(N) (fcc), α′-Fe(N) (bct), α″-Fe16N2 (bct), and α-Fe (bcc). The N content is in at. %. Also, the IDs of the PDF cards are listed.

MatterLatticeN (wt. %)a = b (nm)c (nm)Peak 1 (deg)Peak 2 (deg)Peak 3 (deg)PDF ID
Fe fcc 0.3652  42.86 49.90 73.25 98-001-9707 
Fe4fcc 5.90 0.3790  41.22 47.97 70.18 01-071-1294 
FeN0.095 fcc 2.33 0.3646  42.93 49.99 73.39 01-075-2136 
FeN0.076 1.88 0.3626  43.18 50.28 73.86 01-075-2131 
FeN0.059 1.46 0.3610  43.38 50.52 74.24 01-075-2130 
FeN0.032 0.796 0.3594  43.58 50.76 74.63 01-075-2127 
FeN0.090 bct 2.22 0.2854 0.3080 43.18 44.88 60.02 01-075-2140 
FeN0.076 1.88 0.2856 0.3047 43.39 44.84 60.74 01-075-2138 
FeN0.056 1.38 0.2859 0.3016 43.58 44.80 61.44 01-075-2137 
Fe bcc 0.2866  44.67 82.34 65.02 00-006-0696 
Fe16N2 bct 3.00 0.572 0.629 42.70 44.78 46.07 01-078-1865 
MatterLatticeN (wt. %)a = b (nm)c (nm)Peak 1 (deg)Peak 2 (deg)Peak 3 (deg)PDF ID
Fe fcc 0.3652  42.86 49.90 73.25 98-001-9707 
Fe4fcc 5.90 0.3790  41.22 47.97 70.18 01-071-1294 
FeN0.095 fcc 2.33 0.3646  42.93 49.99 73.39 01-075-2136 
FeN0.076 1.88 0.3626  43.18 50.28 73.86 01-075-2131 
FeN0.059 1.46 0.3610  43.38 50.52 74.24 01-075-2130 
FeN0.032 0.796 0.3594  43.58 50.76 74.63 01-075-2127 
FeN0.090 bct 2.22 0.2854 0.3080 43.18 44.88 60.02 01-075-2140 
FeN0.076 1.88 0.2856 0.3047 43.39 44.84 60.74 01-075-2138 
FeN0.056 1.38 0.2859 0.3016 43.58 44.80 61.44 01-075-2137 
Fe bcc 0.2866  44.67 82.34 65.02 00-006-0696 
Fe16N2 bct 3.00 0.572 0.629 42.70 44.78 46.07 01-078-1865 

Figure 3(e) shows the θ-2θ scan of the sample after the cryo-treatment. The intensities of both γ-Fe(N) peaks (at 43.13° and 50.27°) become very weak. Moreover, the peak around 43.2° becomes broader than that before the cryo-treatment. The results imply the phase transformation happened during the cryo-treatment.

Weak but obvious iron oxide peaks are seen in the θ-2θ scan of the samples after the quenching. This is due to the oxidation of the sample during the high-temperature duration and the followed water quenching.

Figure 4 shows the XRD θ-2θ scans of the samples with different duration times at the temperature of 550 °C. With the increase in the duration time, the intensity ratios of γ′-Fe4N peaks to Fe peak first increase at 3 min duration and then gradually decrease. The increased ratio is due to the transformation from ε-Fe2-3N to γ′-Fe4N phase during the heating.10 ε-Fe2-3N formed on the surface of the γ′-Fe4N layer, during the cooling from 500 °C nitrogenizing to room temperature under the same nitrogen potential according to the Lehrer diagram.9,11–13 The gradual intensity decrease of γ′-Fe4N is due to the decomposition of γ′-Fe4N. The diffraction peak of γ′-Fe4N is still visible even after 30 min staying at 550 °C. The result implies that the decomposition of γ′-Fe4N is not an instantaneous chemical reaction, but a process of nitrogen atoms diffusing off.

FIG. 4.

XRD θ-2θ scans of the samples with different duration times at 550 °C before the water quenching.

FIG. 4.

XRD θ-2θ scans of the samples with different duration times at 550 °C before the water quenching.

Close modal

Figure 5 shows the XRD θ-2θ scans of the samples with different duration times at the temperature of 665 °C. The intensity ratio of γ′-Fe4N peaks increases at 3 min duration and then decreases rapidly. The increased temperature fastened the diffusion of the nitrogen atoms, and then, there are no γ′-Fe4N diffraction peaks anymore when the duration is longer than 7 min. That is, γ′-Fe4N completely decomposes after 7 min annealing at 665 °C. Similar results occur in the case of 705 °C quenching, with an even quicker decomposition of γ′-Fe4N.

FIG. 5.

XRD θ-2θ scans of the samples with different duration times at 665 °C before the water quenching.

FIG. 5.

XRD θ-2θ scans of the samples with different duration times at 665 °C before the water quenching.

Close modal

The characteristics of phase transformation in quenching are summarized in Figs. 6(a) and 6(b). As shown in Fig. 6(a), the intensity ratio of γ′-Fe4N (111) to Fe (110) is about 60% for the nitrogenizing sample. With the increase of the duration time, a similar trend of the intensity ratio is indicated, increasing first and then decreasing. The decomposition of γ′-Fe4N depends on the temperature, and the decrease in the intensity ratio becomes quick and quick with the increase in the quenching temperature.

FIG. 6.

(a) The influence of quenching temperature and duration time on the intensity ratio of γ′-Fe4N (111) to Fe (110). The dashed line represents zero. (b) The influence of the quenching temperature and duration time on the intensity ratio of Fe(N) solution to Fe(110).

FIG. 6.

(a) The influence of quenching temperature and duration time on the intensity ratio of γ′-Fe4N (111) to Fe (110). The dashed line represents zero. (b) The influence of the quenching temperature and duration time on the intensity ratio of Fe(N) solution to Fe(110).

Close modal

Figure 6(b) shows the influence of the quenching temperatures and duration times on the intensity ratio of γ-Fe(N) to Fe (110). In these cases of 550 and 620 °C, the intensity ratio of γ-Fe(N) is very small. This tiny intensity ratio indicates that there is little austenite phase. Considering that there are strong γ′-Fe4N diffraction peaks, there is no martensite transformation by directly quenching γ′-Fe4N phase. However, apparent γ-Fe(N) was obtained in the 665 and 705 °C quenching. That is, there was another precursor before the quenching. It was γ-Fe(N), which is due to the decomposition of γ′-Fe4N. In the current composition, 665 °C quenching produces more γ-Fe(N) solid solution than that 705 °C quenching does.

Figure 7(a) shows the influence of the temperature and duration time on the saturation magnetization (MS) of the Fe–N samples. There is a big MS dip for the Fe–N ribbons quenched at 665 °C after 3 min duration. It is well known that γ-Fe(N) austenite is paramagnetic. Thus, the dip is due to the existence of the γ-Fe(N), which is about 40% in the sample. With the increase in the duration time, MS increases gradually in all the quenched samples, where there are less Fe–N phases but more Fe content. Figure 7(b) shows the influence of the temperature and duration time on the coercivities. When the duration is longer than 5 min, the coercivity slightly decreases with either the quenching temperature or the duration. When the duration is less than 5 min, the abnormal variation indicates the drastic change in the microstructure.

FIG. 7.

(a) Saturation magnetization (MS) of the samples quenched at different temperatures with different duration times. The maximum field for the measurements is 5 kOe. (b) Coercivity (HC) of the samples quenched at different temperatures with different duration times. The maximum field for the measurements is 5 kOe. (c) Temperature-dependent saturation magnetization (MS) of a quenched sample with high fcc-Fe(N) content (MS = 110 emu/g). The sample was measured by PPMS with an applied field of 20 kOe. The temperature ramp is 10 °C/min.

FIG. 7.

(a) Saturation magnetization (MS) of the samples quenched at different temperatures with different duration times. The maximum field for the measurements is 5 kOe. (b) Coercivity (HC) of the samples quenched at different temperatures with different duration times. The maximum field for the measurements is 5 kOe. (c) Temperature-dependent saturation magnetization (MS) of a quenched sample with high fcc-Fe(N) content (MS = 110 emu/g). The sample was measured by PPMS with an applied field of 20 kOe. The temperature ramp is 10 °C/min.

Close modal

Figure 7(c) shows the temperature-dependent MS of a quenched sample. When the temperature decreases from 25 to −30 °C, the MS slightly increases, just like a typical behavior of temperature-dependent MS of a magnetic material. However, when the temperature is below −30 °C, MS dramatically increases. Martensite transformation occurred, and γ-Fe(N) transferred to α′-Fe(N)/α′-Fe8N, resulting in the increase of the saturation magnetization.7,8 The martensite transformation temperature of this specific sample is around −30 °C. As the temperature decreases below −75 °C, the MS ramp becomes slow. When the temperature is below −180 °C, the temperature-dependent MS returns to the normal tendency. It indicates the phase transformation completes. When the temperature increases from −263 °C to room temperature, the saturation magnetization undergoes a different trend. The sample had different Fe–N phases—α′-Fe(N)/α′-Fe8N, after the cryo-treatment.

As shown as route ② in Fig. 1, decomposition of γ′-Fe4N is used to prepare α′-Fe(N)/α′-Fe8N mixture and then α″-Fe16N2. Compared to route ①, route ② also follows the martensite transformation but by heating γ′-Fe4N to get γ-Fe(N) precursor at a high temperature. The difference between the two methods is whether the precursor is in an equilibrium status. In route ①, the samples were nitrogenized in a gas atmosphere with ammonia and hydrogen for a long time to attain equilibrium before the quenching. That is, the precursor is in an equilibrium status, which can be identified on the Fe–N phase diagram. However, the precursor in route ② is in a transient or unstable γ-Fe(N) phase(s), which cannot be clearly labeled in the phase diagram. This also means that the N content in the precursor can be increased, and then, high α″-Fe16N2 content may be realized.

We would like to point out that the Fe–N phase diagram differs in an important feature from the common metallurgical phase diagram of a binary alloy.14 The metallurgical diagram is determined by the composition and temperature at a constant atmosphere, but the Fe–N phase diagram is a projection onto a temperature-concentration plane. The wide-used Fe–N phase diagram is determined by the controlled nitrogenizing methods, and the nitrogen potential that is defined by the ratio of NH3 to H2 is crucial. In general, each point in the Fe–N phase diagram was obtained by the following procedure:14 

  1. Fixing a temperature.

  2. Setting a nitrogen potential by varying the ratio of NH3 to H2.

  3. Nitrogenizing the Fe powders.

  4. Quenching the nitrogenized sample as fast as possible to avoid the phase transformation.

  5. Measuring the lattice structure and the composition.

  6. Plotting the phase diagram by many isothermal series of controlled nitrogenizing experiments.

This means that both the temperature and the nitrogen potential determine the composition of the sample. The same composition can be adjusted if both the temperature and the nitrogen potential change accordingly. Different compositions will be reached if the nitrogen potential keeps the same but the temperature changes or the temperature keeps the same but the nitrogen potential changes. If a Fe–N sample is heated in the air, then the Fe–N phase diagram cannot be used to predict the phase transformation because there is no environment with an exact nitrogen potential to attain equilibrium. Moreover, nitrogen atoms will escape during the heating. When γ′-Fe4N is heated in air, its decomposition starts when the temperature is higher than 400 °C and greatly enhances when the temperature is higher than 600 °C.10,15–18 Therefore, the red dash line schematically labels the composition change during the heating.

The nitrogenizing process in route ① requires that the dissociation of ammonia be constant throughout an isothermal series of experiments and that equilibration be attained between the gas and solid phases. However, in route ②, the nitrogenized samples are heated in the air. That is, the precursor is mutative with a variable N content as N atoms escape from the γ′-Fe4N lattice. Since both γ′-Fe4N and γ-Fe(N) are of the fcc lattice with different N contents, it is expected the escape of N atoms in γ′-Fe4N results in an fcc lattice with an N content somehow in the range of 2.4–6.0 wt. %. Thus, the proposed precursor in route ② is an N deficient γ′-Fe4N or N supersaturated γ-Fe(N) at a high temperature.

The quenching results show that there was no martensite transformation related to γ′-Fe4N, but martensite transformation occurred in γ-Fe(N). Clearly, partial γ′-Fe4N converted into γ-Fe(N). When the samples were quenched from 665 °C (Fig. 5), there are no γ′-Fe4N diffraction peaks after 7 min duration. The absence of γ′-Fe4N indicates that it completely decomposed under the current conditions. On the contrary, the γ-Fe(N) appears in Fig. 5. If the quenching does not change the sample composition and lattice,14 γ-Fe(N) is the phase at 665 °C before the quenching. That is, γ-Fe(N) forms when γ′-Fe4N decomposed at a high temperature. This phase transformation was also reported by an in situ powder neutron diffraction.10 

As far as we know, the minimum y value in γ′-Fe4Ny is 0.94 in the XRD diffraction PDF card (01-071-1294), which corresponds to the N content of 5.6 wt. %. However, y = 0.86 (5.1 wt. %) was reported in the in situ neutron diffraction.10 In Table I, the N content in γ-Fe(N) is less than 2.4 wt. %. Thus, the phase with 5.1 wt. % is not in the Fe–N phase diagram and it may be recognized as an N deficient γ′-Fe4N or N supersaturated γ-Fe(N). Moreover, the phase is metastable, which only exists for a short time at the high temperature and decomposes at room temperature. Therefore, N supersaturated γ-Fe(N) with a high N content may be obtained by heating γ′-Fe4N. Then, it is used as the precursor to the martensite transformation to realize a high N content in the quenched sample.

For the γ-Fe(N) with a high N content, the fcc lattice keeps after the quenching because its martensite temperature is below the temperature of the cold water. As shown in Fig. 8, the martensite temperature decreases with the increase in the N content.19 When the N content is more than 2.3 wt. %, the martensite transformation temperature is below 0 °C, and then, the martensite transformation cannot occur in cold water. Thus, a cryo-treatment is required, as shown in the current experiments. The martensite transformation temperature of −30 °C for the sample shown in Fig. 7(c) is plotted as the dashed–dotted line in Fig. 8. The intersection of the two curves shows the N content of 2.5 wt. %, which is the high end of the range shown in Fig. 1. Figure 7(c) shows that a small amount of γ-Fe(N) has an even lower martensite temperature; it means that their N contents are high, for example, the N content is 2.95 wt. % when the martensite temperature is −100 °C. Therefore, the proposed route ② results in high N content in the sample, which benefits the preparation of α″-Fe16N2.

FIG. 8.

The dependence of martensite transformation temperature on the N content, according to the data in Ref. 19. The curves were measured by two techniques. The thermal arrest technique was used to measure the martensite transformation temperature beyond 100 °C, and the resistance technique was used to measure the martensite transformation temperature below room temperature. The dashed–dotted line is the martensite temperature of the sample shown in Fig. 7(c).

FIG. 8.

The dependence of martensite transformation temperature on the N content, according to the data in Ref. 19. The curves were measured by two techniques. The thermal arrest technique was used to measure the martensite transformation temperature beyond 100 °C, and the resistance technique was used to measure the martensite transformation temperature below room temperature. The dashed–dotted line is the martensite temperature of the sample shown in Fig. 7(c).

Close modal

Here, we proposed a dynamic model based on nitrogen diffusion/loss to explain the decomposition of γ′-Fe4N in our results. At an elevated temperature, N atoms in the γ′-Fe4N lattice absorb kinetic energy and escape from the fcc lattice through diffusion. The diffusion coefficient is exponentially proportional to the temperature, and the following formula was used to calculate the diffusion coefficient of nitrogen in γ′-Fe4N:20 

lnDγN=91.4×103RT21.7,
(1)

where R is the gas constant, T is the temperature, and DγN is the diffusion coefficient of N atom in γ′ lattice. The lattice parameters of γ-Fe(N) are proportional to the N content, following the formula below:21,22

aN=a0+k*CN,
(2)

where CN is the content of nitrogen in Fe lattice, a0 is the lattice parameter of fcc-Fe, and aN is the lattice parameter of γ-Fe(N). Here, k is a constant, representing the linear change of lattice parameter with the N content. Both γ′-Fe4N and γ-Fe(N) are of the fcc lattice, but with different N contents and lattice parameters. Thus, the difference between γ′-Fe4N and γ-Fe(N) is mainly due to the N content.

There is a big gap of the N content between γ′-Fe4N and γ-Fe(N). According to the XRD data, the minimum N content of γ′-Fe4N is 5.7 wt. %, and the maximum solubility of N in γ-Fe(N) is 2.4 wt. % (Table I). This implies that there might have γ-Fe(N) with the N content from 2.4 to 5.7 wt. %, but they are not stable so that the XRD technique cannot reveal their lattice. The recent work on neutron diffraction showed that the decomposition of γ′-Fe4N is governed by nitrogen loss, and there is transient austenite iron nitride with a composition of 5.1 wt. % formed during the heat treatment. Thus, we might utilize the transient γ-Fe(N) as the precursor of the martensite transformation to obtain a higher N content. To understand the underline mechanism during the γ′-Fe4N decomposition, we carried out the simulation of the decomposition process by applying self-diffusion of nitrogen atoms.

The assumption here is that the decomposition rate of γ′-Fe4N or supersaturated γ-Fe(N) is primarily controlled by the self-diffusion of nitrogen atoms in the matrix. The recombination of nitrogen atoms that occurred at the surface to form nitrogen gas will be considered as a minor factor. The premise of this simulation is to consider a homogeneous nitrogen distribution along the depth of the iron nitride ribbon. We also assume that the mobile nitrogen atoms will diffuse along the depth in both directions.

Figure 9(a) shows the simulated nitrogen profiles of the samples after 665 °C annealing for 3 min. We used a bulk γ′-Fe4N ribbon for the simulation, as N atoms diffuse out both sides of the surface instead to the middle of the ribbon. There is no nitrogen loss in the middle area, and it is still γ′-Fe4N. This is due to the homogeneous nitrogen environment as there are equal N atoms diffusing in and out. Figure 9(b) shows that the N loss occurs near the surface (∼2.5 μm). The slope of the nitrogen profile becomes more and more steep when the location is close to the surface. The nitrogen loss is faster at the surface than that in the middle of the ribbon. Also, the nitrogen loss is affected by the N content of the original Fe(N) phase. It will be much faster and deeper in the case of low N content. For Fe-3.3 wt. % N, the transition region is more than 10 μm from the surface. Thus, when nitrogen loss occurs in γ′-Fe4N, it will become fast with the decrease of the nitrogen content in the ribbons.

FIG. 9.

Simulation of the nitrogen profile based on a nitrogen diffusion model: (a) a 30-μm-thick ribbon with different N contents and (b) zoom-in plot of (a). The gray region is the air. (c) A 1-μm-thick ribbon with different heating durations. The temperature of the heating treatment is 665 °C.

FIG. 9.

Simulation of the nitrogen profile based on a nitrogen diffusion model: (a) a 30-μm-thick ribbon with different N contents and (b) zoom-in plot of (a). The gray region is the air. (c) A 1-μm-thick ribbon with different heating durations. The temperature of the heating treatment is 665 °C.

Close modal

It should be noted that the N content at the surface should be half of the original N content, as N atoms combine to molecular and escape. However, the current theoretical model cannot include this case. Also, the surface benefits the nitrogen loss in the ribbon as there is no reverse diffusion from the surface to the inside of the ribbon.

Figure 9(a) shows the inhomogeneous nitrogen profile. This causes great challenges under the quenching conditions, as the martensite transformation temperature depends on N content. In order to improve the quenching effect, a thin γ′-Fe4N ribbon (such as 1 μm thick) is a better precursor than the current ribbon. Figure 9(c) shows nitrogen profiles at different heating durations. The nitrogen should loss much quicker even within 10 s at 665 °C. It can be seen from the above figure that the center nitrogen content even drops from 5.9 to 5.6 wt. % while holding the ribbon at 665 °C for 10 s. After annealing at 665 °C for a minute, the center nitrogen content dropped below than 3.3 wt. % and the interface of the ribbon/ambient environment contained lower than 2.7 wt. %. Once the annealing time prolongs to 2 min, the thorough nitrogen content falls below 2.7 wt. %. Moreover, the variation of the N content is smaller than that of a thick ribbon.

Even though this simulation is not accurate enough to represent the physical reaction, it presents a feasibility that a γ-Fe(N) phase with a high nitrogen content can be obtained by the decomposition of γ′-Fe4N. This simulation could provide some general guidance to discuss the process parameters.

In summary, a new strategy using γ′-Fe4N as the precursor was proposed to prepare a bulk magnet with high α″-Fe16N2 content, and the feasibility of achieving high N content α′-Fe(N)/Fe8N was demonstrated. The new recipe separates the nitrogenizing process and the quenching process. First, the ribbons were nitrogenized under a controlled ammonia/hydrogen mixture gas atmosphere. Various ratios of the ammonia/hydrogen were adjusted based on the nitrogenizing, which is guided by the Lehrer diagram. Then, the nitrogenized samples were heated to a set temperature and then quenched in cold water.

γ′-Fe4N cannot be directly used as the precursor of the martensite transformation. When the temperature increased up to 665 °C, however, significant diffraction peaks of γ-Fe(N) were observed. Meanwhile, γ′-Fe4N disappeared after 7 min duration for 665 °C and 3 min duration for 705 °C, respectively. Then, the appearance of γ-Fe(N) after quenching indicated that γ-Fe(N) phases were obtained from the decomposition of γ′-Fe4N, with the same lattice but different nitrogen contents. This transformation can be explained by the nitrogen loss from the fcc lattice through atoms diffusion. Supersaturated γ-Fe(N) is a precursor with a high N content for the martensite transformation in the new approach.

The difference between our method and Jack’s method is whether the precursor is in an equilibrium status. In Jack’s work, the samples were nitrogenized in a gas atmosphere with ammonia and hydrogen for a long time to attain equilibrium before the quenching. That is, the precursor is in an equilibrium status, which can be identified on the Fe–N phase diagram. However, the precursor in our method is in a transient or unstable phase(s), which cannot be clearly labeled in the phase diagram. This also means that the N content in the precursor and the quenching product can be increased.

Martensite transformation temperature decreases with the N content, and then, the cryo-treatment is required for the high N content fcc-Fe(N). After the cryo-treatment, α′-Fe(N)/Fe8N appears, which was confirmed by an increase in the saturation magnetization of ribbons. The martensite transformation temperature of the main phase is −30 °C, deriving from the temperature-dependent saturation magnetization of quenched ribbons. The martensite temperature corresponds to the N content of more than 2.5 wt. %.

The work was partially sponsored by Niron Magnetics, Inc. We are grateful to Dr. Yiming Wu, Dr. John Larson, Dr. Frank Johnson, and technicians for their assistance in arranging nitrogenizing some samples using Niron facilities based on our pretreated Fe–Cu–B ribbons as input precursors and our proposed specific processing recipes.

Part of this work was carried out by the help of Dr. Javier Garcia Barriocanal, in the Characterization Facility, the University of Minnesota, which receives partial support from NSF through the MRSEC program. We also thank the partial support in VSM measurements from the Institute of Rock Magnetism.

Dr. Jian-Ping Wang has equity and royalty interests in and serves on the Board of Directors for Niron Magnetics, Inc., a company involved in the commercialization of FeN magnet. The University of Minnesota also has equity and royalty interest in Niron Magnetics, Inc. These interests have been reviewed and managed by the University of Minnesota in accordance with its Conflict of Interest policies.

The data that support the findings of this study are available within the article and from the corresponding author upon reasonable request.

1.
J.-P.
Wang
, “Environment-friendly bulk Fe16N2 permanent magnet: Review and prospective,”
J. Magn. Magn. Mater.
497
,
165962
(
2020
).
2.
K. H.
Jack
, “The occurrence and the crystal structure of α″-iron nitride; a new type of interstitial alloy formed during the tempering of nitrogen-martensite,”
Proc. R. Soc. London, Ser. A
208
(
1093
),
216
224
(
1951
).
3.
K. H.
Jack
, “The iron-nitrogen system: The preparation and the crystal structures of nitrogen-austenite(γ) and nitrogen-martensite(α′),”
Proc. R. Soc. London, Ser. A
208
(
1093
),
200
215
(
1951
).
4.
K. H.
Jack
, “The iron–nitrogen system: The crystal structures of ε-phase iron nitrides,”
Acta Crystallogr.
5
(
4
),
404
411
(
1952
).
5.
K. H.
Jack
, “The synthesis, structure, and characterization of α″-Fe16N2 (invited),”
J. Appl. Phys.
76
(
10
),
6620
6625
(
1994
).
6.
K. H.
Jack
, “The synthesis and characterization of bulk α″-Fe16N2,”
J. Alloys Compd.
222
,
160
(
1995
).
7.
Y.
Jiang
,
V.
Dabade
,
L. F.
Allard
,
E.
Lara-Curzio
,
R.
James
, and
J. P.
Wang
, “Synthesis of α″-Fe16N2 compound anisotropic magnet by the strained-wire method,”
Phys. Rev. Appl.
6
(
2
),
024013
(
2016
).
8.
M.
Mehedi
,
Y.
Jiang
,
B.
Ma
, and
J. P.
Wang
, “Nitriding and martensitic phase transformation of the copper and boron doped iron nitride magnet,”
Acta Mater.
167
,
80
(
2019
).
9.
E. J.
Mittemeijer
,
M.
Planck
, and
I.
Systems
,
Fundamentals of Nitriding and Nitrocarburizing, Steel Heat Treating Fundamentals and Processes
4
,
619
646
(
2018
).
10.
M.
Widenmeyer
,
T. C.
Hansen
,
E.
Meissner
, and
R.
Niewa
, “Formation and decomposition of iron nitrides observed by in situ powder neutron diffraction and thermal analysis,”
Z. Anorg. Allg. Chem.
640
(
7
),
1265
1274
(
2014
).
11.
Y.
Inokuti
,
N.
Nishida
, and
N.
Ohashi
, “Formation of Fe3N, Fe4N and Fe16N2 on the surface of iron,”
Metall. Trans. A
6
(
3
),
773
784
(
1975
).
12.
J.
Ratajski
,
J.
Tacikowski
, and
M. A. J.
Somers
, “Development of compound layer of iron (carbo)nitrides during nitriding of steel,”
Surf. Eng.
19
(
4
),
285
291
(
2003
).
13.
J.
Liu
,
G.
Guo
,
X.
Zhang
,
F.
Zhang
,
B.
Ma
, and
J.-P.
Wang
, “Synthesis of α″-Fe16N2 foils with an ultralow temperature coefficient of coercivity for rare-earth-free magnets,”
Acta Mater.
184
,
143
150
(
2020
).
14.
V. G.
Paranjpe
,
M.
Cohen
,
M. B.
Bever
, and
C. F.
Floe
, “The iron-nitrogen system,”
JOM
2
(
2
),
261
267
(
1950
).
15.
X. L.
Wu
,
W.
Zhong
,
H. Y.
Jiang
,
N. J.
Tang
,
W. Q.
Zou
, and
Y. W.
Du
, “Magnetic properties and thermal stability of γ′-Fe4N nanoparticles prepared by a combined method of reduction and nitriding,”
J. Magn. Magn. Mater.
281
(
1
),
77
81
(
2004
).
16.
H.
Xiang
,
F. Y.
Shi
,
M. S.
Rzchowski
,
P. M.
Voyles
, and
Y. A.
Chang
, “Epitaxial growth and thermal stability of Fe4N film on TiN buffered Si(001) substrate,”
J. Appl. Phys.
109
(
7
),
07E126
(
2011
).
17.
L. L.
Wang
 et al, “Structural and magnetic properties of nanocrystalline Fe–N thin films and their thermal stability,”
J. Alloys Compd.
443
(
1–2
),
43
47
(
2007
).
18.
I.
Dirba
 et al, “Growth, structure, and magnetic properties of γ′-Fe4N thin films,”
J. Magn. Magn. Mater.
379
,
151
155
(
2015
).
19.
T.
Bell
, “Martensite transformation start temperature in iron-nitrogen alloys,”
J. Iron Steel Inst.
206
,
1017
1021
(
1968
).
20.
M. A. J.
Somers
and
E. J.
Mittemeijer
, “Layer-growth kinetics on gaseous nitriding of pure iron: Evaluation of diffusion coefficients for nitrogen in iron nitrides,”
Metall. Mater. Trans. A
26
(
1
),
57
74
(
1995
).
21.
T.
Christiansen
,
M. A. J.
Somers
, and
A. S.
Preparation
, “Controlled dissolution of colossal quantities of nitrogen in stainless steel,”
Metall. Mater. Trans. A
37
,
675
682
(
2006
).
22.
T. S.
Hummelshøj
,
T. L.
Christiansen
, and
M. A. J.
Somers
, “Lattice expansion of carbon-stabilized expanded austenite,”
Scr. Mater.
63
(
7
),
761
763
(
2010
).