III. RESULTS AND DISCUSSION

The selected area electron diffraction (SAED) patterns of the three main zone axes ([001], [011], and [111]) of Mn₃Zn_0.8Ni_0.2N are shown in Figs. 1(a), 1(c), and 1(e), respectively. In these patterns, there are some weaker spots around the basic diffraction spots corresponding to the antiperovskite crystal structure, and some of them are much brighter than the others. According to our TEM knowledge, these brighter spots indicate the presence of a second phase, which must be very minor according to the XRD spectrum, and these patterns can be produced on the principle of multiple diffraction. Based on some detailed analyses, we assumed that the sample particles are wrapped by a thin layer of a second phase. This model is supported by the corresponding morphology images shown in Figs. 1(b), 1(d), and 1(f). The near edge area is clearly different from the inner part, indicating that it should be the second phase.

FIG. 1.

SAED patterns along (a) [001], (c) [011], and (e) [111] directions and the corresponding morphology images (b), (d), (f) of Mn₃Zn_0.8Ni_0.2N.

To identify the second phase, we characterized its chemical composition with EDX and EELS measurements. We collected EDX and EELS spectra of the edge and inner area of many particles. We found that there are only Mn and O compositions in the edge area of some particles, and in some other cases, there is also little Zn, Ni, and/or N, whereas there are much Zn, Ni, and N in the inner area. A typical result is shown in Fig. 2. So, we reasonably concluded that the second phase is one of the manganese oxides. The SAED results shown in Fig. 1 suggest that the second phase should have a face-centered cubic crystal structure with a cell parameter a ≈ 4.4 Å. In the manganese oxides, only MnO can satisfy this restriction, which has a space group of Fm-3m with a = 4.436 Å (PDF No.: 1-1206). As a result, we concluded that the second phase is MnO.

FIG. 2.

(a) Demonstration of different areas for collecting EDX and EELS spectra. Related EDX and EELS spectra are shown in (b) and (c), respectively. The arrows indicate the position of the peaks of Zn and Ni. The EDX and EELS spectra are normalized to have equal intensity of the highest peak. The Mo and Co in EDX spectra are from the microgrid and the sample holder, respectively. The quantification results of the EDX spectra are shown in the inset table in (b). Due to the MnO layer, the Mn content is slightly higher than the nominal value.

The distribution of Zn and Ni elements has a sharp step as going further from the edge to the inner area as indicated by the EDX spectra shown in Fig. 2(b), this significant composition difference and thus resulted different crystal structures should be responsible for the appearance of the sharp border between the edge area and the inner part on the morphology images shown in Figs. 1 and 2.

On the EELS spectra collected from the middle and inner parts shown in Fig. 2(c), the obvious O K-edge should be contributed by the MnO surface.

We also measured the width of the edge area to estimate the thickness of the MnO layer and found that it ranges from about 3 nm to more than 40 nm, showing no dependence of zone axis direction.

When we found that the Mn₃Zn_0.8Ni_0.2N particle is wrapped in a thin film of MnO, we have a question naturally: is this a general feature? Or is there just something wrong with our sample? So we studied more manganese nitrides: Mn₃Zn_0.5Ni_0.5N and Mn₃ZnN. The SAED patterns along [001] direction of these compounds shown in Figs. 3(a) and 3(b) are almost the same as that of Mn₃Zn_0.8Ni_0.2N shown in Fig. 1(a), and the particles also have a clear edge area as shown in Figs. 3(c) and 3(d). EDX and EELS results also suggest that the edge area is mainly manganese oxide as shown in Fig. 4. In summary, we found the same phenomenon in these compounds as in Mn₃Zn_0.8Ni_0.2N, so it should be a general feature in this kind of compound.

FIG. 3.

SAED patterns along [001] direction of (a) Mn₃Zn_0.5Ni_0.5N and (b) Mn₃ZnN, respectively. The corresponding morphology images are shown in (c) and (d), respectively. EDX and EELS spectra of these two particles are shown in Fig. 4.

FIG. 4.

EDX and EELS spectra collected from different areas of the particles shown in Fig. 3: (a) and (c) for Mn₃Zn_0.5Ni_0.5N; (b) and (d) for Mn₃ZnN. The EDX and EELS spectra are normalized to have equal intensity of the highest peak. The Mo and Co (and/or Fe) in EDX spectra are from the microgrid and the sample holder, respectively. The quantification results of the EDX spectra are shown in the inset tables in (a) and (b). Due to the MnO layer, the Mn content is slightly higher than the nominal value.

To confirm our TEM conclusion, we carried out XPS measurements on Mn₃ZnN powder, which could reveal the information about the surface of the sample within a few nanometers macroscopically. A full-scan spectrum is shown in Fig. 5. The Mn and O peaks are very strong in the spectrum, whereas the N-related peak is negligible. This result also suggests that the surface of the Mn₃ZnN powder is covered by manganese oxide. There are also some Zn-related peaks on the spectrum, and in our study, Zn(Ni) was also detected at the edge area of some particles, although its relative content is lower than that in the inner part. This fact may mean that the transformation from manganese nitrides to MnO is incomplete in some cases. As a result, we get consistent result at both the macroscopic and microscopic scales.

FIG. 5.

The full-scan XPS spectrum of Mn₃ZnN powder. The C is from organic contaminants which are hard to avoid in XPS measurements.

The presence of MnO is also supported by some other experiments. In the high-pressure x-ray powder diffraction measurements on Mn₃Zn_0.8Ni_0.2N performed by Hu et al., ^{Reference Hu, Sun, Cong, Zhao, Zhang, Zhang, Li, Liu, Jin and Yu14} there is an obvious peak of MnO (111). And there is also a small but obvious peak which can be determined as MnO (111) in the high-pressure angle dispersive x-ray powder diffraction patterns of Mn₃GaC reported by Meenakshi et al. ^{Reference Meenakshi, Vijayakumar, Lausi and Busetto15} However, this peak is almost invisible in the ordinary x-ray powder diffraction patterns.

Our work suggests that there is no or little M on the surface of the sample particles of manganese nitrides Mn₃MN (M = Zn, Ni, Cu…). As the shape of our sample particles is manifold, and there is no clue (Only MnO was detected as a secondary phase in XRD patterns as well as TEM mircroanalysis of the composition of the particles) to the presence of pure Mn or M, it is reasonable to conclude that this phenomenon is caused by the preferred loss of M on the surface in the process of our TEM sample preparation, rather than by Mn segregation during sintering. A reasonable explanation of the preferred loss of M is that M has a weaker bonding environment than Mn. On the surface of the sample particle, the activation energy of M may be much lowered due to the surface effect, thus the loss of M would occur under some conditions. Due to the high diffusibility of N, the loss of N is ordinary in Mn₃MN, especially on the surface of sample particles. In the presence of O, the oxidation of Mn may occur after the loss of N and M, or they occur at the same time, as a result, the particle is wrapped in a MnO layer with a few nanometers. In a word, antiperovskite manganese nitrides Mn₃MN (M = Zn, Ni, Cu…) have a strong surface activity.

Since our samples were sintered in vacuum, the MnO layer should not emerge during sintering. It most likely forms in the process of our TEM sample preparation, i.e., while crushing the samples. Due to the friction heat, some local areas could reach a temperature high enough for the activation of M and N and the diffusion of O, i.e., the formation of MnO surface layer during crushing. Since a higher energy is needed to activate M and N and to diffuse O in the inner area of a particle, a higher temperature should be needed for the formation of a thicker MnO layer. While at a certain temperature (or other equivalent conditions), the thickness of the MnO layer should reach a maximum if the condition lasts long enough.

During our TEM work, we have found that the MnO layer could develop due to the electron beam irradiation. As shown in Figs. 6(a) and 6(b), the border between the edge and middle areas is sharp and clear before the irradiation, and their chemical compositions are very different. After some time of irradiation, the middle area loses its M element (Zn and Ni) and the border becomes indistinct, as shown in Figs. 6(c) and 6(d). This phenomenon gives direct evidence to our former conclusion that the loss of M is prior to the loss of Mn in manganese nitrides. The related theoretical studies are eager to be carried out.

FIG. 6.

The morphology images and corresponding EDX results of a sample particle of Mn₃Zn_0.8Ni_0.2N before [(a), (b)] and after [(c), (d)] some time of electron beam irradiation. The EDX spectra are normalized to have equal intensity of the highest peak. The Mo is from the microgrid.

Hamada and Takenaka ^{Reference Hamada and Takenaka9} represented the different results about Mn₃ZnN of different researchers ^{Reference Fruchart, Madar, Barberon, Fruchart and Lorthioir7,Reference Kim, Chi, Kim, Hur, Lee and Choi8} by different heat treatments. Wang et al. ^{Reference Wang, Chu, Yao, Sun, Wu, Ding, Yan, Na, Tang, Li, Huang and Lynn16} found same behaviors in the samples with different Zn content, the samples with 1% and 4% Zn vacancy exhibit two phase transitions, whereas the sample with 7% Zn vacancy exhibits only one phase transition, similar to Hamada's samples sintered at 700 and 900 °C, respectively. This coincidence could be understood based on our results: M is easier to be lost in Mn₃MN (M = Zn, Ni, Cu…), and so, higher sintering temperature would result in more M(Zn) vacancies and a small change in the amount of M(Zn) vacancies could greatly alter the properties of the sample. This should be the major reason why different researchers got different, even opposite results.

Besides, when measuring the property of the samples with small size, the relative amount of MnO may be large and its influence cannot be neglected. So, the influence of MnO may be another reason for the difference among different researchers' results. ^{Reference Fruchart, Madar, Barberon, Fruchart and Lorthioir7–Reference Sun, Wang, Wen, Zhu and Zhao10}

To understand the physics under the interesting phenomena in manganese nitrides Mn₃MN (M = Zn, Ni, Cu…), a real and accurate description of their properties is necessary, so the influence of M vacancies and/or MnO should be carefully avoided to get an accurate result in the future study.