Epitaxial Fe₁₆N₂ thin film on nonmagnetic seed layer

Fe₁₆N₂ is an interesting material in that it has both high saturation magnetization (M_s) and large magnetic anisotropy.^1–5 So, this compound is one of the potential candidates for permanent magnets.^6,7 Efforts have also been made to lower its anisotropy through doping and to use Fe₁₆N₂ as a write head material in hard disk drives,^8,9 because large M_s is the key to improving write-ability of magnetic writers, thus breaking the magnetic recording trilemma and further increasing the data storage density.¹⁰ However, accurate determination of the saturation magnetization of Fe₁₆N₂ and theoretical explanation of its large value are still challenging. An important part of the experimental difficulty is material synthesis. Fe₁₆N₂ can hardly be obtained in pure phase, whether the sample is thin film, fine particle,¹¹ or bulk.^12,13 The coexisting α-Fe and other iron nitrides affect the determination of M_s of Fe₁₆N₂. In thin film studies, a layer of Fe was mostly used as a seed layer to support the growth of Fe₁₆N₂ (e.g., Refs. 3 and 14). This layer of magnetically soft Fe not only interferes with the determination of M_s of Fe₁₆N₂ but also limits its application because it couples with the Fe₁₆N₂ layer.³ Although some effort was made to decouple the Fe and Fe₁₆N₂ layers by inserting a nonmagnetic Ag^15–17 or Cr¹⁸ layer between them, the Fe sublayer was still present in the heterostructures. The difficulty of a theoretical explanation lies in the fact that Density Functional Theory (DFT) calculations have failed to reproduce the very high M_s of Fe₁₆N₂ as in some experimental reports.¹⁹ This situation, together with the fact that giant M_s was mostly found in epitaxial films, inspired Ji et al. to propose that in-plane tensile strain is the cause of high M_s in Fe₁₆N₂.²⁰ In this letter, we demonstrated that a nonmagnetic Cr seed layer can epitaxially stabilize high quality Fe₁₆N₂ thin films without a Fe sublayer. Cr, like Fe, has a body-centered cubic (bcc) structure and a perfect lattice constant match with Fe₁₆N₂. The as-obtained strain-free Fe₁₆N₂ film has high phase purity. The order parameter, which quantifies the degree of interstitial N ordering in the mixture of the $α′$-Fe₈N martensite phase and $α″$-Fe₁₆N₂, is among the highest ever reported.

The Fe₁₆N₂ thin films were grown by a facing-target sputtering (FTS) technique.² A Cr seed layer was deposited at elevated temperature on a single crystal MgO (001) substrate. Fe₁₆N₂ was then deposited on this Cr layer by sputtering Fe targets in the N₂ and Ar gas mixture at room temperature. A Cr capping layer was then deposited on top of the Fe-N layer for protection from oxidation. The structure of samples used in this study is MgO/Cr (10.2)/Fe-N (26.7)/Cr (10.2) (thickness in nm is indicated in parentheses). The whole structure was deposited in a single chamber, which accommodates multiple pairs of targets, and then annealed in vacuum at 160 °C for 6–36 h. This annealing condition was optimized to give the highest N ordering.

X-ray diffraction (XRD) and X-ray reflectometry (XRR) were carried out on a PANalytical X'Pert diffractometer with Cu K-α radiation. To accurately calculate the order parameter, a Siemens D5005 diffractometer with Cu K-α radiation was used to obtain a higher signal at Fe₁₆N₂ (00L) reflections. Polarized neutron reflectometry (PNR) as a means to obtain the depth-resolved magnetization profile²¹ was carried out at the Magnetism Reflectometer at the Spallation Neutron Source (SNS), Oak Ridge National Lab (ORNL).²² A 10 kOe external field was applied to saturate Fe₁₆N₂ in the film plane. Room temperature M-H loops were measured on a Quantum Design PPMS vibrating sample magnetometer (VSM).

Figure 1(a) shows a typical out-of-plane XRD pattern of the Cr-seeded Fe₁₆N₂ thin film. Only (00L) reflections of Fe₁₆N₂ can be observed, indicating that Fe₁₆N₂ is well textured. Based on the integrated intensities of (002) and (004) reflections, the order parameter of Fe₁₆N₂ can be calculated as follows:

where $I002 exp . (I004 exp .)$ and $I002theo. (I004theo.)$ are the experimental and theoretical intensities of the (002) ((004)) peak, respectively. The theoretical intensity ratio of (002) and (004) reflections is calculated by considering structure factors and Lorentz-polarization factors and is equal to 0.074.²³ In calculating the integrated intensities, the peaks were fitted to pseudo-Voigt functions first. The samples in this study generally have an order parameter above 0.6, the highest being 0.75, suggesting that Cr serves as a good seed layer for Fe₁₆N₂ thin films. The annealing conditions and order parameters of 5 samples used in this study are summarized in Table I.

Grazing incidence XRD 2θ–ω scan shows clearly Fe₁₆N₂ (400) and Cr (200) peaks [Fig. 1(b)], together with the substrate (220). Fe₁₆N₂ (220) and Cr (110) reflections were also observed along with MgO (200). Because the a lattice constant of Fe₁₆N₂ (bulk value 5.72 Å)²⁴ is very close to twice that of Cr (bulk value 2.87–2.89 Å), it is not possible to separate Fe₁₆N₂ (2H, 2H, 0) ((2H, 0, 0)) and Cr (HH0) ((H00)) reflections. We did 360°-ϕ scan with the rotation axis along the film normal at Fe₁₆N₂ (220) and Cr (110) reflections, and the result is shown in Fig. 1(c). The fourfold symmetry is clearly seen, and thus, the good crystallinity of the epitaxial Fe₁₆N₂ thin film is confirmed. In addition to the 4 main peaks 90° apart from each other, there is a peak 5.3° to the right of the second main peak, labeled “°” in Fig. 1(c). This additional peak is also found in the ϕ scan of substrate (200) reflection, also 5.3° to the right of the second main peak [Fig. 1(d)]. So, this effect is due to substrate defects and shows good epitaxy of the Fe₁₆N₂ film. Now, the epitaxial relation can be summarized as follows: $MgO (001)[110]∥Cr (001)[100]∥Fe16N2 (001)[100]$⁠. By a rotating crystal approach, the (202) and (103) reflections of Fe₁₆N₂ were observed, as shown in Figs. 1(e) and 1(f). (103) is a fingerprint peak of Fe₁₆N₂, which differentiates it from the N-disordered $α′$-Fe₈N martensite phase. Rocking curves [inset Figs. 1(e) and 1(f)] around (202) and (103) have a FWHM of 1.25° and 1.62°, respectively. The lattice constants of Fe₁₆N₂ were calculated by a least squares approach to minimize the deviation from d-spacings of (004), (400), (220), (202), and (103) reflections. The obtained values are a = 5.722 Å and c = 6.272 Å, very close to the well-accepted bulk values of a = 5.72 Å and c = 6.29 Å.²⁴ 

In the course of our experiments, the authors noticed a similar work.¹⁸ Conventional magnetron sputtering was used to grow Fe₁₆N₂ thin films on Cr seed layers, confirming that Cr is a reasonable choice to stabilize this metastable compound. However, a layer of Fe was still used in that work. In addition, Fe was deposited directly on MgO, with a 4% lattice mismatch. So, it became difficult to accurately account for the contribution of Fe in the total magnetic moment. Also in that work, a high order parameter, 0.6 as reported, can only be obtained for 10 nm-thick Fe₁₆N₂. For films thicker than 15 nm, the order parameter is below 0.3. In our samples, however, order parameters over 0.6 can be routinely obtained for 26.7 nm thick Fe₁₆N₂. This shows the superiority of FTS in fabricating Fe₁₆N₂.

XRR and PNR were performed to study the chemical and magnetic depth profiles of sample 5 as listed in Table I. PNR, for a not-so-thin film, is necessary due to the inability of magnetometry to account for nitrogen diffusion. The formation of Fe₁₆N₂ generally requires annealing at 125–180 °C, while at this temperature, N diffusion has become significant. For example, in the Fe/Fe₁₆N₂ structure, a 6.6–14.7 nm-thick interface was caused solely by diffusion when annealed at 152 °C for 6 h based on the diffusion coefficients measured by different methods.²⁵ Within the interface, the local environment of Fe atoms changes and the stoichiometry will deviate from Fe₁₆N₂. This diffusion length is not at all negligible when compared to the usual film thickness, i.e., 10–100 nm.

The GenX software was used to fit both XRR and PNR experimental curves.²⁶ First, the XRR curve was fitted to give the chemical depth profile. This was then used as the initial structure model to optimize the magnetization depth profile by fitting PNR curves. The experimental and fitted XRR curves are shown in Fig. 2(a). As can be seen, the fitting adequately describes the true structure. The corresponding x-ray electron scattering length density (ESLD) profile is shown in Fig. 2(b). N diffusion into both bottom and top Cr layers can be identified as dips in ESLD, as N has a lower ESLD than Cr and may cause lattice expansion. Neutron reflectivity for both spin up and spin down are plotted in Fig. 2(c), along with the fitted curves. The corresponding nuclear and magnetic scattering length density (SLD) profiles (NSLD and MSLD, respectively) are shown in Fig. 2(d). MSLD is directly proportional to magnetization (saturated in this experiment),²⁷ and M_s of Fe₁₆N₂ was calculated to be 1840 emu/cm³ (2.31 T). With 5% error in the fitting figure of merit (FOM), the M_s range of 1820 emu/cm³ (2.29 T) to 1870 emu/cm³ (2.35 T) was obtained.²⁸ A 3.2 nm-thick interface layer with reduced M_s at the top Fe₁₆N₂/Cr interface is likely due to roughnesss. The integrated MSLD (after converting to M_s) gives the total magnetic moment of the sample. This value is 4.1% larger than the VSM result. As the two techniques give consistent results, we believe that the M_s value determined from PNR reflects the intrinsic property of Fe₁₆N₂. This saturation magnetization value does not support the giant M_s results and may indicate that seed layers can affect the magnetization of Fe₁₆N₂. It has been shown by PNR depth-resolved magnetization that under tensile strain (from MgO epitaxy, with a very thin Fe buffer layer), Fe₁₆N₂ has a giant M_s of 3.1 T, while as the strain releases, the M_s decreases to 2.1 T.²⁰ In this work, the Cr seed layer is thick enough to overwrite the strain information of the MgO substrate and our sample contains a strain-free Fe₁₆N₂ layer, as can also be seen from XRD. This observation, together with the low M_s Fe₁₆N₂ particle results,^5,29 points to the possibility that strain is the reason for the observed giant M_s in Fe₁₆N₂ thin films.

A magnetic M-H loop of the Cr-seeded Fe₁₆N₂ thin film is shown in Fig. 3. An interesting feature of the loops, which is found in all Cr-seeded samples, is the clear switching in the out-of-plane loop. The field at which this switching happens and the magnitude of the switching as a fraction of M_s for the 5 samples are listed in Table I. The out-of-plane loop clearly has a nonzero remanence. A similar but less prominent phenomenon was previously found in Fe/Fe₁₆N₂ thin films, and this switching was attributed to the small amount of Fe₁₆N₂ that is perpendicularly magnetized.³ In those Fe/Fe₁₆N₂ samples, a soft Fe seed layer was used and N ordering was low in the iron nitride layer, i.e., the main phase is $α′$-Fe₈N martensite. A model was proposed that the Fe layer is magnetized in plane and couples with the Fe₁₆N₂ layer. So, the magnetization vector gradually goes to the film normal because of the perpendicular magnetic anisotropy (PMA) of Fe₁₆N₂.³ However, in our samples, there is no soft Fe layer and the only ferromagnetic layer is Fe₁₆N₂. Also, the degree of N disorder is small. Thus, a new model is needed for an explanation. One possibility is that some Fe₁₆N₂ grains are isolated from others so that there is no dominating shape anisotropy that favors in-plane magnetization. Magnetocrystalline anisotropy (MCA) determines the easy axis to be out-of-plane. In this case, the MCA calculated from the switching field, however, is only one third to half that of previous reports (∼1.0 × 10⁷ ergs/cc).^3–5 Another possible explanation is that some interface layer in the MgO/Cr/Fe₁₆N₂/Cr system has PMA, and its magnetic moment takes up about 2%–4% of the total moment (Table I). In this case, the MCA of Fe₁₆N₂ can be calculated (assuming coherent rotation) as $2πMs2−MsHK/2$⁠, where $2πMs2$ is the shape anisotropy and H_K is the anisotropy field labeled in Fig. 3 and is 13.4 kOe. The calculated MCA, 8.9 × 10⁶ ergs/cc, agrees well with literature values, and this indicates that the interface layer model is more likely to be true. Ongoing work on thickness and seed layer dependence of the magnetic behavior of Fe₁₆N₂ will help clarify this situation.

In conclusion, we grew Fe₁₆N₂ thin films on the MgO/Cr template using facing-target sputtering. The advantages of the Cr seed layer are as follows: (1) Cr has a bcc structure and a perfect lattice constant match with Fe₁₆N₂; (2) the only ferromagnetic material in the system is Fe₁₆N₂, making it easier to study its intrinsic magnetism; (3) the as-obtained Fe₁₆N₂ film has high phase purity, i.e., high order parameter. The XRD results show that our Fe₁₆N₂ film is textured in all three directions and has good crystallinity. By (H, K, L ≠ 0) reflections, Fe₁₆N₂ is identified from Cr, and by its fingerprint peaks (002) and (103), interstitial N ordering is confirmed. We believe that this work provides a platform for further study of Fe₁₆N₂. The intrinsic saturation magnetization of the strain-free Fe₁₆N₂ thin film was determined to be 2.31 T by PNR and confirmed with VSM. The out-of-plane M-H loop of the Cr-seeded Fe₁₆N₂ film shows an interesting easy-axis switching behavior, the magnitude being 2%–4% of the total magnetic moment. Further study is being carried out to find the origin of this perpendicular component.

Research at ORNL's Spallation Neutron Source was sponsored by the Scientific User Facilities Division and by the Laboratory Directed Research and Development Program of Oak Ridge National Laboratory, managed by UT-Battelle, LLC, for the U. S. Department of Energy. Parts of this work were carried out in the Characterization Facility, University of Minnesota, which receives partial support from NSF through the MRSEC Program.