Rare-earth-free ferrimagnetic Mn₄N sub-20 nm thin films as potential high-temperature spintronic material

As industry rapidly transitions to using big data in their operations and decision making, there is an urgent need to develop technologies that accommodate the increasing requirements of high-density data storage.¹ One promising candidate that has received increasing attention is using magnetic skyrmions as information carriers and manipulating them with current by spin–orbit torques (SOTs) for logic or memory operations^2,3 (e.g., racetrack memories). Magnetic skyrmions are swirling spin configurations of neighbor atoms in a magnetic material with topologic protection. The size of the skyrmion could be as small as a few nanometers,⁴ and the manipulation of skyrmions could be energy efficient^5,6 in high-density data storage applications.^7,8 To date, most RT skyrmions have been discovered in ferromagnetic (FM) multilayer stacks.^7–10 The use of a ferromagnets as skyrmions has certain shortcomings. For instance, ferromagnets suffer from large stray fields and large saturation magnetization (M_s), which makes the region of parameter space for small skyrmion less accessible, especially at room temperature (RT).¹¹ 

Unlike ferromagnets, the ferrimagnetic counterparts have small stray fields and low Ms, which makes them suitable for hosting small RT skyrmions.¹¹ This makes ferrimagnet a promising candidate material for high-density data storage applications. One prototypical example is the amorphous GdCo ferrimagnetic thin film. Small (10 nm–30 nm) room temperature skyrmions have been reported in Pt/GdCo (6 nm)/TaO_x.¹² While the amorphous GdCo ferrimagnet shows promising characteristics relative to the ferromagnets, it suffers from poor thermal stability. It has been shown that in amorphous GdCo, PMA is lost after annealing at 300 °C–400 °C,¹³ the temperature range applied in complementary metal–oxide–semiconductor (CMOS) fabrication.¹⁴ Since Néel skyrmions can only exist in a heterostructure with PMA, the poor thermal stability of RE-TM films will have a deleterious effect in both the fabrication and performance of amorphous RE-TM in devices that leverage skyrmions for storage technology. One of the potential solutions to this problem is to explore crystalline ferrimagnets with PMA synthesized by deposition at high-temperature near 400 °C, thus ensuring compatibility with conventional CMOS processing.

In addition to the intrinsic effects that we have discussed thus far, the film thickness is an important extrinsic effect that impacts the overall device performance. The importance of film thickness applies to both ferromagnets and ferrimagnets. It is now well-established that the interfacial Dzyaloshinskii–Moriya interaction (DMI), which stabilizes the magnetic skyrmions in multilayers and heterostructures, decays exponentially with increasing film thickness.^15–17 Further, the SOT scales inversely with the thickness.¹⁸ This necessitates the growth of thin film magnets with sub-20 nm thickness for the realization of small skyrmions in practical applications.^12,19,20

Anti-perovskite Mn₄N has been known as a crystalline, rare-earth-free ferrimagnetic material. In Mn₄N, the Mn-atoms have a face-centered cubic structure with one N-atom at the body center. A schematic of the unit cell is shown in Fig. 1. The Mn-atoms at the corner and the face center have inequivalent magnetic moments and are ferrimagnetically coupled.²¹ Although the easy axis of bulk Mn₄N is along the [111] direction, PMA is repeatedly and reproducibly observed in crystalline Mn₄N films.^22–28 As a result, ferrimagnetic Mn₄N thin films have also attracted increasing interest in spintronics applications. Compared to the amorphous RE-TM ferrimagnetic alloys, the Mn₄N system has better thermal stability for two main reasons. First, in most thin film studies, it has been shown that the anti-perovskite crystal structure is formed at 400 °C–450 °C. No loss of PMA is reported in thin films after annealing, which is an encouraging outcome. Second, there is no known evidence for any structural phase transformation on cooling to room temperature. Thus, we interpret that the anti-perovskite crystal structure is tolerant of high-temperature device fabrication processes (unlike the GdCo amorphous alloys). As noted earlier, the emergence of PMA is an important magnetic property for its use in spintronic applications. One of the plausible reasons for the PMA in Mn₄N thin films could be attributed to the deviation of the out-of-plane lattice constant (c) to the in-plane lattice constant (a) (c/a) ratio from 1^22–28 due to the in-plane epitaxial strain. A recent study also showed that the anisotropy energy is correlated with the c/a ratio.²⁷ We note that further studies are warranted to understand the interplay of magnetic and strain effects on the PMA and this is beyond the scope of this paper.

Several groups have grown crystalline Mn₄N epitaxial films (30 nm–100 nm) on MgO, SrTiO₃, or LaAlO₃ substrates by magnetron sputtering,^22,23 molecular beam epitaxy (MBE),^24–27 and pulsed laser deposition (PLD),²⁸ and they reported similar c/a ratios of ∼0.99. The reported uniaxial magnetic anisotropy constant (K_u) and M_s of those Mn₄N films were about 0.5 Merg/cc–1 Merg/cc and 50 emu/cc–100 emu/cc, respectively,^22–28 comparable with the data observed for amorphous GdCo thin films (K_u ∼ 0.25 Merg/cc and M_s ∼ 50 emu/cc).¹² 

To date, however, only a few groups have grown sub-20 nm Mn₄N thin films with good magnetic hysteresis [M(H)] loops, which has the remanent magnetization M_r to M_s ratio larger than 0.5 (M_r/M_s > 0.5), by MBE.^26,27 For those reported Mn₄N films by sputtering, the thicknesses were 30 nm²² to 100 nm,²³ some of them even up to hundreds nanometers.^29,30 Similar results on sputter deposited sub-20 nm Mn₄N thin films have not been reported. Compared to MBE, sputter deposition is a more widely adopted method in CMOS technology. In this work, we have epitaxially grown sub-20 nm Mn₄N thin films on the MgO (001) substrate with PMA by reactive sputtering. The effect of MgO substrate morphology on the quality of Mn₄N films is studied. We then compare the magnetic properties of the Mn₄N films on the MgO substrate from experiment and first principles-based density functional theory (DFT) calculations. The main contribution of this paper lies in the demonstration of the growth of high magnetic quality sub-20 nm crystalline Mn₄N films on the MgO substrate using reactive sputtering that is more promising for scale-up production.

Mn₄N thin films with nominal thicknesses of ∼15 nm and 10 nm were deposited on MgO(001) 5 × 5 × 0.5 mm³ substrates by reactive rf-sputtering at 400 °C and 450 °C substrate temperatures and a base pressure of 7 × 10⁻⁸ Torr. The flow rates of Ar and N₂ gases were controlled by a mass flow meter, and we maintained a flow rate Ar:N₂ ratio of 93:7. A 3 nm Pt capping layer was deposited on the Mn₄N layer at room temperature to prevent oxidization. Before loading the MgO (001) substrate into the vacuum chamber, it was wet-cleaned with 2% diluted Hellmanex III alkaline detergent, acetone, and isopropanol and sealed in a vacuum tube with pressure 30 mTorr. The MgO substrate was baked at high temperature 1000 °C or 1100 °C for 4 h. Then, the substrates were annealed inside the chamber at 500 °C for 1 h to remove the surface contaminations. The Mn target was pre-sputtered with Ar gas for 20 min to remove the surface oxide. The surface roughness of the MgO substrate was measured using atomic force microscopy (AFM). The deposition rate was measured by x-ray reflectometry (XRR). The film compositions were determined with x-ray photoelectron spectroscopy (XPS). The film thickness was verified by XRR with an XRR simulation/calculation program (Rigaku, GXRR).³¹ 

The epitaxial growth of the Mn₄N crystal layer was demonstrated using x-ray diffraction (XRD) with Cu-Kα radiation. The magnetic properties of the samples were measured at room temperature with vibrating sample magnetometry (VSM). The diamagnetic component of the substrate was deduced from the slope of raw M-H curves at a large H region and subtracted from the raw data. K_u was calculated from effective anisotropy K with the following equations:

where M_easy and M_hard are the magnetization in the out-of-plane applied field and in-plane applied field, respectively.

Ab initio electronic structure calculations were carried out in the Density Functional Theory (DFT) framework using the plane-wave pseudopotential Quantum ESPRESSO code.^32,33 Core and valence electrons were treated using the ultrasoft pseudopotential method.^34,35 The exchange-correlation functionals were described using the Perdew–Burke–Ernzerhof parameterization of the generalized gradient approximation modified for solids (PBEsol).³⁶ The plane-wave cutoff energy was set to 60 Ry, and a Γ-centered Monkhorst–Pack k-point mesh of 12 × 12 × 12 was used to sample the Brillouin zone.³⁷ Lattice parameters were fixed to an experimentally measured value (c/a = 0.987), and M_s was calculated using the formula $Ms=μtotalV$⁠, where $μtotal$ is the absolute value of the total magnetization from DFT calculations given in Bohr magnetons and V is the unit cell volume.³⁸ Self-consistent spin-polarized calculations were performed for collinear spin structures with no spin–orbit coupling term in the Hamiltonian. The Mn-atoms (Mn I) located in the cell corners with coordinates (0, 0, 0) were ferromagnetically coupled to Mn-atoms (Mn IIa) located at (0.5, 0.5, 0). Both Mn I- and Mn IIa-atoms were anti-ferromagnetically coupled to the Mn-atoms (Mn IIb) in the face centers with coordinates (0, 0.5, 0.5) and (0.5, 0, 0.5). Since the total atomic magnetic moments at all three Mn-sites do not cancel each other out (Mn I = 3.47 µ_B, Mn IIa = 0.75 µ_B, and Mn IIb = −2.36 µ_B), the ground state is a ferrimagnet.

Table I lists the five samples (S1–S5) that were examined in this study. We begin by investigating different substrate annealing temperatures to understand the impact of the MgO surface on the epitaxial growth of Mn₄N. S1–S3 were deposited at 400 °C, and S4 and S5 were deposited at a higher temperature of 450 °C. The film thickness was determined by XRR measurement. In the case of S2–S5, the MgO substrates were pre-baked before loading into the sample deposition chamber.

The film thicknesses measured by the XRR technique indicated that the thicknesses of all five thin film samples were less than 20 nm (see Table I). The analysis of XRR results is shown in Fig. 1 of the supplementary material. S1–S4 had a thickness of 16.8 ± 0.5 nm, whereas S5 had a thickness of 11.5 ± 0.6 nm. Figure 2(a) shows the normalized M-H curves with the out-of-plane applied magnetic field for 16.8 nm Mn₄N films on different heat-treated MgO substrates (S1–S3). The M-H curves show a strong dependence on the annealing temperature. All out-of-plane hysteresis loops were open, indicating PMA. The substrate annealing temperature of S3 (1100 °C) is higher than that of S2 (1000 °C), whereas S1 is unannealed. The squareness of the M(H) hysteresis loop improves from S1–S3. Figure 2(b) shows the dependence of the substrate annealing temperature on M_r/M_s. The M_r/M_s ratio increased from 0.47 to 0.72 as the annealing temperature increased from RT to 1100 °C. The increment of M_r/M_s is attributed to the improved epitaxial growth of the thin film. High-temperature annealing not only decomposes the Mg(OH)₂ and MgCO₃ contaminants but also reconstructs the surface of the MgO substrate.^39,40 During this process, the substrate forms an atomically smooth surface,^39,40 which is beneficial for epitaxial growth.⁴¹ The difference between S1 and S3 is mainly attributed to the different substrate annealing temperatures. Thin films annealed at 1100 °C produced a smoother surface than 1000 °C. The surface morphology as characterized by the atomic force microscopy revealed an average root-mean-square (rms) roughness of 0.592 nm, 0.308 nm, and 0.278 nm for the as-received, 1000 °C annealed, and 1100 °C annealed thin films, respectively (see Fig. 2 of the supplementary material).

To explore further improvement, we increased the deposition temperature to 450 °C while retaining the substrate annealing temperature and time at 1100 °C and 4 h. It has been shown that the squareness of the Mn₄N film can be improved by increasing deposition temperature.²⁹We found that the M_r/M_s ratio of these samples, S4 and S5, increased further to near 1, as shown in Fig. 2(c). The difference between S4 and S5 is in the film thickness: The S4 and S5 thin films were 16.8 ± 0.5 nm and 11.5 ± 0.6 nm thick, respectively (see Table I). S5, the thinnest film, kept a high M_r/M_s ratio of ∼0.8, which is an encouraging outcome.

Figure 2(d) shows the out-of-plane and in-plane M-H curves for the S4 film. M_s and K_u for the S4 film are 43 ± 1 emu/cc and 0.70 Merg/cc, respectively. All five samples had similar M_s (40 emu/cc–60 emu/cc). Although M_s of these films are smaller than the reported M_s (∼100 emu/cc) in thicker films (30 nm–100 nm),^22–25,28 they are comparable with that of the MBE-grown sub-20 nm Mn₄N film on MgO, 50 emu/cc–80 emu/cc.^26,27 The lower M_s of sub-20 nm Mn₄N films may be due to surface oxidation.²⁵ 

Figure 3 shows the out-of-plane 2θ-θ XRD profile (a) and φ scan (b) of S3. Besides the MgO substrate peaks, only the Mn₄N (002) peak is observed in the 2θ-θ XRD profile, which indicates the Mn₄N (00l) orientation is parallel to MgO (00l). In the XRD 360° φ scan, both Mn₄N and MgO show four peaks with 90° interval, which correspond to (202), (022), (−202), and (0–22). The overlap of Mn₄N peaks and MgO peaks in φ-scan confirms their epitaxial relationship to be MgO (001)[100]//Mn₄N (001)[100]. The out-of-plane lattice constant c and in-plane lattice constant a deduced from the (002) and (202) diffraction peaks are 0.386 nm and 0.391 nm, respectively. Therefore, the c/a ratio was calculated to be 0.987, which is close to the value reported earlier.^22–28 

Based on the experimental lattice constant, DFT predicts an M_s value of 153 emu/cc, which is larger than our experiment results. However, DFT is a zero-temperature electronic structure calculation. Li et al. have shown that M_s of the bulk Mn₄N would decrease as a function of increasing measurement temperature.⁴² DFT also does not consider the possible effects of the mixing layer at the interface as well as defects. Future investigation will address the temperature dependence of saturation magnetization and magnetic anisotropy energy.

We have grown sub-20 nm ferrimagnetic Mn₄N epitaxial thin films with high M_r/M_s ratios at 400 °C–450 °C substrate temperatures on MgO substrates by reactive sputtering. The quality of the epitaxial growth was optimized by ex situ pre-annealing of MgO substrates at temperatures as high as 1100 °C. The annealing was found to reduce the surface roughness of the MgO substrates at the atomic level, which was found to improve the quality of the out-of-plane magnetization hysteresis loops. The present results established the Mn₄N thin film as a thermally stable ferrimagnet for further investigation as a promising skyrmion-based spintronic material.