Interfacial mixing effect in a promising skyrmionic material: Ferrimagnetic Mn₄N

The interface plays an important role in the properties of thin-film heterostructures. For magnetic materials, many well-studied and important phenomena arise from interfacial effects, such as interfacial magnetic anisotropy,^1–4 Ruderman–Kittel–Kasuya–Yosida (RKKY) coupling in a synthetic antiferromagnet through a metal spacer,^5–8 and the interfacial Dzyaloshinskii–Moriya interaction (I-DMI).^9–14 Furthermore, it has been reported that in many of these heterostructures, interfacial mixing and dead layers exist at the interfaces of magnetic and non-magnetic layers. Such interfacial compositional heterogeneities have led to varying magnetic properties, such as thickness-dependent anisotropy and magnetization.^15–21 These present challenges in fabricating and developing devices with magnetic thin films, such as magnetic tunnel junctions.²² Therefore, investigating the effects due to interfacial mixing in heterostructures is a crucial step in designing new materials for spintronics applications.

Mn₄N thin films have recently garnered much attention in the spintronic community.²³ Mn₄N forms an anti-perovskite crystal structure.²⁴ It is a ferrimagnetic metal with a Curie temperature of 710 K.²⁵ Experiments have shown that Mn₄N exhibits perpendicular magnetic anisotropy (PMA)^26–32 that is desirable for use as free and fixed layers in magnetic devices. Furthermore, these properties are resilient to high temperatures, at least up to 450 °C. More recently, tunable magnetic skyrmions have been reported to be present in MgO/Mn₄N/Pt_xCu_1−x thin films,³³ wherein Pt_xCu_1−x is designed to provide a variable I-DMI that controls the size of the skyrmion. As such, Mn₄N can potentially be used as a building block in skyrmion-based memory devices. Nonetheless, in view of elemental mixing that tends to occur at the interfaces of thin-film heterostructures, the question remains as to what extent high-temperature deposition Mn₄N^26–32 would adversely affect the functionalities of the Mn₄N based heterostructures. For example, the interface is known to play a significant role in the effectiveness of spin currents and spin–orbit torque between magnetic and non-magnetic layers.^34–37 In a recent study, Mn₄N has been reported to have a 2 nm mixing layer from transmission electron microscopy.³⁰ Thus, further quantitative studies of the interfacial properties of Mn₄N based heterostructure films is pivotal in exploring the feasibility of using Mn₄N in the future devices.

In this study, we investigated the interfaces of epitaxially grown MgO/Mn₄N (17 nm)/Pt films, which were deposited at the same time, using x-ray photoelectron spectroscopy (XPS) and polarized neutron reflectometry (PNR). From XPS, mixing layers were revealed at both MgO/Mn₄N and Mn₄N/Pt interfaces. At the MgO/Mn₄N interface, an oxidation layer of 5 nm thick MgMnNO mixing layer was revealed. At the Mn₄N/Pt interface, a 5 nm thick MnPtNO mixing layer was found. Analysis of the PNR data revealed that both top and bottom mixing layers exhibit essentially zero magnetization. The 5 nm thickness of the mixed layer at the MgO/Mn₄N interface is comparable to other room temperature depositions,^19,20 despite Mn₄N being deposited at 450 °C. Equally important, micromagnetic simulations presented herein show that even with the presence of mixing layers, magnetic skyrmions, which arise from I-DMI, are found to be stable in MgO/Mn₄N/Pt_xCu_1−x thin films. This finding is consistent with recent experimental results. All these findings indicate that Mn₄N is a promising material for designing skyrmion-based memory.

We deposited 17 nm thick Mn₄N films on MgO substrates using a single Mn target by reactive radio frequency (rf) sputtering. The MgO substrates were wet-cleaned and heat-treated ex situ.³² The base pressure was 9 Pa. Sputter deposition was carried out at 450 °C and under an Ar:N₂ gas flow ratio of 93:7. A Pt capping layer was deposited at room temperature to prevent oxidation. Our previous work showed epitaxially growth MgO/Mn₄N/Pt through x-ray diffraction (XRD).³² 

We performed x-ray photoelectron spectroscopy (XPS) measurements to obtain the compositional depth profile using the PHI VersaProbe III XPS instrument (Note that certain commercial equipment is identified in this paper to foster understanding. Such identification does not imply recommendation or endorsement by NIST, nor does it imply that the materials or equipment identified are necessarily the best available for the purpose). XPS data were collected after sputtering off a few layers from the surface. Each sputtering lasted about 15 s, and the total sputtering time was 10 min. We analyzed the XPS results based on the following model with three assumptions. First, the composition from an XPS measurement is the averaged composition of a 3 nm layer from the surface, and the measurement sensitivity is the same within the 3 nm layer. Second, the distribution of Mn₄N near the Mn₄N/Pt and Mn₄N/MgO surfaces follows the cumulative distribution function (CDF) of a Gaussian as below:

where z is the distance from the MgO substrate, s is the interface boundary, and σ is the thickness of the mixing.

Third, the distribution of the MnO is a Gaussian function (G) as below:

where w is the weight coefficient and l is the location of the MnO layers.

Table I summarizes the distribution of each layer, where A, B, C, D, and E are normalization coefficients. Using this model, we calculated the averaged compositions f_avr(z) at location z by integrating the distribution from z − 3 nm to z and fitted these f_avr(z) with the composition obtained from the XPS measurement by tuning parameters in CDF and G.

Polarized neutron reflectometry (PNR) measurements were performed at room temperature in a 3 T in-plane field using the polarized beam reflectometer (PBR) instrument at the NIST Center for Neutron Research. The reflectivities were measured as a function of the wavevector Q along the normal direction to the sample surface (i.e., specular scattering). Data were reduced and model-fitted with the reductus³⁸ and Refl1D³⁹ software packages, respectively. The modeling allowed us to deduce the depth profiles of the film’s nuclear scattering length density (ρ_N, indicative of the nuclear composition), and the in-plane magnetization.⁴⁰ 

Micromagnetic simulations of Mn₄N were carried out using the Object-Oriented Micromagnetic Framework.⁴¹ Details of the simulations can be found in a previous publication.³³ The simulation dimensions are 300 × 300 × 15 nm³, and each cell is 5 × 5 × 5 nm³. The exchange stiffness constant between each cell is 1.5 × 10⁻¹¹ J/m. The external field is 0.01 T. The anisotropy (K_u) is (0.5 – 1.2) × 10⁵ J/m³. We added a 5 nm thick magnetically dead layer at the Mn₄N/Pt interface, which is assumed to provide a DMI according to the fractional composition of Pt in the layer but no magnetic interaction. Skyrmion size is defined as the average diameter of a boundary with zero magnetizations.

Figure 1 shows the hysteresis loops of Mn₄N at 300 K. The out-of-plane and in-plane M(H) loops confirm perpendicular magnetic anisotropy (PMA). The coercivity is found to be about 0.55 T, and saturation magnetization is 40× kA/m. From the in-plane loop, the anisotropy field is determined to be over 3 T. The in-plane magnetization remains unsaturated at 3 T, which is the limit of the equipment. Using these results, the anisotropy (K_u) is estimated to exceed 70 kJ/m³, comparable to other studies.^26–32 Figure 2 shows (a) the compositional depth profile, and (b) the first derivative of the compositional depth profile, as a function of distance from the MgO substrate as determined from XPS. The surface of the MgO substrate is ideally set at z = 0 nm. In view of the presence of an interfacial layer, z = 0 nm is set at the position where the relative composition of Mg and O is 1:1. As such, z = 0 nm is the beginning of the MgO/Mn₄N interface region. The bulk of the Mn₄N layer is seen to extend to around 20 nm. Above approximately z = 20 nm is the Mn₄N/Pt interface region. From Fig. 2(a), elemental mixings exist at both MgO/Mn₄N and Mn₄N/Pt interfaces. At the MgO/Mn₄N interface region (z = 0 nm), both Mg and O atoms from the MgO substrate diffuse into Mn₄N. Since Mn reacts easily with O, but not Mg, O diffuses further into Mn₄N. Overall, the nominal average composition of this interface region is Mg₂₀O₃₇Mn₃₇N₆. From Fig. 2(b), the first derivative of the fitted compositional depth profile verifies that O diffuses further into the Mn₄N layer, as indicated by the peak of O occurring at higher z than that of Mg. Furthermore, the thickness of this interface region can be estimated using the full width at half maximum of the derivative. The mixing region with Mn₄N and MgO is about 5 nm thick with O diffuses about 3 nm further into Mn₄N. The presence of this interfacial layer at the MgO/Mn₄N interface is possibly due to rough MgO/Mn₄N interfaces. From Shen et al.,⁴² misfit dislocations were observed by transmission electron microscopy, while such dislocations were not found at the SrTiO₃/Mn₄N interface.²³ 

Both the Mn and N distributions shown in the composition derivative plot are separated by about 17 nm, which is the nominal thickness of the Mn₄N layer. It is worth noting that a small amount of O exists in the region. This is possibly due to the presence of O during the deposition. Even though the base pressure is below 9 Pa, and only Ar:N₂ was used, very tiny amount of O can enter the system through the Ar or N₂ gas feed lines. Since Mn reacts easily with O, O enters Mn₄N. Overall, the composition of this region is Mn_77.7N_19.6O_2.7. This means that Mn:N maintains 4:1 in this sample despite the presence of O. Above Mn₄N (z > 20 nm) lies the region of the Mn₄N/Pt interface. Both Pt and O atoms are found to diffuse into Mn₄N. The nominal average composition of this interface region is Mn₃₇N₆O₂₀Pt₃₇. The increase of the O content at the Mn₄N/Pt interface is possibly due to the extended time required to cool down before room temperature deposition of Pt with trace amounts of O in the system. Using the full width at half maximum of the Pt composition derivative in Fig. 2(b), this region is about 5 nm thick. It is worth noting that the mixing regions at both interfaces are about 5 nm thick, even though one was deposited at 450 °C and Pt was deposited at room temperature.

To understand the magnetic consequences of the compositional inhomogeneity, we characterized the magnetization depth profile with PNR. Figure 3(a) shows the model-fitted spin-dependent reflectivities (++ and −−) as functions of wavevector transfer Q. The data show clear oscillations, demonstrating sensitivity to the scattering length density depth profile. However, the scattering is dominated by nuclear contributions, as the difference between ++ and −− is essentially too small to detect in (a). In Fig. 3(b), the same fitted data are plotted as spin-asymmetry (SA, difference in ++ and −− divided by the sum), which highlights magnetic contributions to the scattering. Shown in this way, a spin-dependent signal is evident, demonstrating sensitivity to a non-zero magnetic depth profile. Figure 3(c) shows the magnetic and nuclear profiles corresponding to the best fit shown in panels (b) and (c). In this model, the compositional profile of the nominal Mn₄N layer determined from XPS is approximated in terms of a three sublayer structure smeared out by error function roughness⁴³ accounting for both conventional interlayer roughness and compositional gradients: MgO/Mg₂₀O₃₇Mn₃₇N₆/Mn₇₈N₁₉O₃/Mn₃₇N₆O₂₀Pt₃₇/Pt. All nuclear scattering length densities were calculated based on the expected composition and were treated as fixed parameters in the fitting. Mg₂₀O₃₇Mn₃₇N₆ and Mn₃₇N₆O₂₀Pt₃₇ sublayer thicknesses were held fixed at 2 nm, and the Pt cap layer thickness was fixed at 3 nm. Interlayer roughness was assumed to be equal for all interfaces, and it was found to be 1 nm. The Mn₇₈N₁₉O₃ layer thickness was a free parameter, which is found to be 17 nm, confirming XPS results. The magnetizations of all three sublayers of the nominal Mn₄N layer were treated as free parameters. The profile shows a center Mn₇₈N₁₉O₃ layer magnetization of 56 kA/m, with a sharp decline near both interfaces. While the magnetization does not go completely to zero in the interfacial sublayer regions, this is consistent with layers of zero magnetization smeared out by the apparent roughness.⁴⁴ The integrated magnetization for the entirety of the nominal Mn₄N film is 45 kA/m, consistent with the magnetometry data shown in Fig. 1. Despite the rigid constraints, Figs. 3(a) and 3(b) show that this model gives an excellent fit to the data, and we note that models with a uniform magnetization profile in the nominal Mn₄N layer result in a significantly worse fit. Thus, the PNR results confirm the compositional profile determined from XPS and show that the interfaces of the nominal Mn₄N layers contribute no net magnetization. This finding is crucial in designing spintronics devices with Mn₄N. With non-magnetic interfacial mixings, these regions can be treated as a non-magnetic layer for the effectiveness of spin currents and spin–orbit torque in Mn₄N heterostructures. Thus, Mn₄N remains a viable option for future devices.

Micromagnetic simulations are employed to study the effect of interfacial mixings on skyrmion formation in Mn₄N. Figure 4 shows the DMI-K_u phase diagram of 15 nm thick Mn₄N with a 5 nm thick magnetic dead layer at 300 K. Only the Mn₄N/Pt mixing region is considered here. In the MgO/Mn₄N/Pt heterostructure, interfacial-DMI is dominated by Pt, so a magnetic dead layer at Mn₄N/Pt can potentially influence skyrmion formation. From XPS, this region has an average composition of Mn₃₇N₆O₂₀Pt₃₇. From density functional theory (DFT) calculations, interfacial-DMI at Mn₄N/Pt is near |7| mJ/m². Using a simple fractional composition model, this region is estimated to have a DMI of 2.6 mJ/m². On the other hand, little or no effect is expected for the MgO/Mn₄N as the interfacial-DMI is negligible between MgO and Mn.³³ Thus, this region is treated as a non-interacting empty space. Using experimental K_u of 0.7 × 10⁵ J/m³, robust skyrmions are observed in interfacial-DMI above 2.5 mJ/m². Below this DMI, only ferrimagnetic states are observed. With interfacial-DMI of 3.5 mJ/m², near 100 nm skyrmions are found in Mn₄N. As interfacial-DMI increases to 5.5 mJ/m², skyrmion sizes increase to 200 nm. At interfacial-DMI close to 7.0 mJ/m², 250 nm skyrmions are observed in the simulations. With larger interfacial-DMI, skyrmions start to evolve into stripes. However, this phenomenon could be due to the limit of the simulation space of 300 nm in the in-plane direction.

From Fig. 4, using experimental K_u of 0.7 mJ/m³ and calculated interfacial-DMI at Mn₄N/Pt of |7| mJ/m², skyrmions are just over 250 nm in size. This is comparable to the observed skyrmions in MgO/Mn₄N/Pt, which range from 250 to 350 nm in size. Furthermore, from the experiment, skyrmions in MgO/Mn₄N/Pt_0.1Cu_0.9 are around 100 nm. From DFT calculation, the interfacial-DMI of Cu/Mn₄N is 2.6 mJ/m².³³ Assuming a simple solution model, the interfacial-DMI of Mn₄N/Pt_0.1Cu_0.9 is estimated to be 3.0 mJ/m². From Fig. 4, with K_u of 0.7 mJ/m³ and interfacial-DMI of 3.0 mJ/m², skyrmions are about 60 nm in diameter. This is smaller than the measured skyrmions in MgO/Mn₄N/Pt_0.1Cu_0.9, which are 80–120 nm in size. This discrepancy may be due to the crude model to estimate the DMI in a mixture. Figure 5 shows the color mapping of out-of-plane magnetization at various K_u (0.5, 0.7, and 0.9 mJ/m³) and interfacial-DMI (1.5, 3.0, 4.5, and 6.0 mJ/m²), with blue regions corresponding to up magnetic moments and red regions corresponding to down magnetic moments. Robust shapes of skyrmions are found in these Mn₄N heterostructures throughout various K_u and interfacial-DMI, even with the presence of interfacial mixings. This is critical for any potential future for using Mn₄N in spintronics applications. Overall, these results verify the reliability of Mn₄N to host skyrmions, despite the existence of magnetic dead layers, and confirm the potential of using Mn₄N in skyrmion-based devices.

Ferrimagnetic Mn₄N thin films are recently recognized as promising skyrmionic materials due to thermal stability and the ability to host small skyrmions. These material attributes prompted us to study the interfacial mixing effect in MgO/Mn₄N/(Pt–Cu) heterostructures. From x-ray photoelectron spectroscopy, elemental mixings were revealed at both MgO/Mn₄N and Mn₄N/Pt interfaces. These mixing regions are about 5 nm thick, comparable to room temperature deposition. Using polarized neutron reflectometry, we found that these interface regions exhibit essentially zero magnetization. To address the effect of these inhomogeneities on the potential of using Mn₄N in spintronic devices, micromagnetic simulations were employed to investigate skyrmions in Mn₄N. Simulated skyrmions were found to be robust at room temperature despite the existence of magnetic dead layers. These results confirm that Mn₄N is a promising material for next-generation memory and logic devices.

This work was partially supported by the DARPA Topological Excitations in Electronics (TEE) program (Grant No. D18AP00009). The content of the information does not necessarily reflect the position or the policy of the Government, and no official endorsement should be inferred. Approved for public release; distribution is unlimited. The Phi VersaProbe III XPS used for acquiring the data was provided through the NSF-MRI under Award No. 1626201.

The authors have no conflicts to disclose.

Chung T. Ma: Data curation (equal); Formal analysis (equal); Investigation (equal); Methodology (equal); Writing – original draft (equal). Wei Zhou: Data curation (equal); Formal analysis (equal); Investigation (equal); Methodology (equal); Writing – review & editing (equal). Brian J. Kirby: Data curation (equal); Formal analysis (equal); Investigation (equal); Methodology (equal); Writing – original draft (equal). S. Joseph Poon: Funding acquisition (lead); Supervision (lead); Writing – review & editing (equal).

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