Current status of room temperature magnetic compensation in impurity-doped Mn4N epitaxial thin films Authors: Tomohiro Yasuda, Takashi Suemasu Published: 2026 ABSTRACT ABSTRACTMn4N, a ferrimagnet that does not contain rare-earth elements, possesses attractive properties for spintronics applications. In particular, Mn4N epitaxial films exhibit a small spontaneous magnetization of about 100 kA m−1, a large spin polarization (p = 0.8), and large perpendicular magnetic anisotropy with a magnetic anisotropy constant of approximately 105 J m−3. More importantly, achieving magnetic compensation at room temperature is possible through impurity doping. This property is particularly significant for spin–torque-based spintronics applications. In the vicinity of the magnetization and/or angular momentum compensation points, magnetization dynamics can be substantially accelerated, enabling high-speed switching and domain wall motion. This article surveys the magnetic properties of impurity-doped Mn4N epitaxial films grown on SrTiO3(001) substrates, highlighting recent results obtained with Cu-, Ag-, Au-, and Pd-doped Mn4N films in comparison with Ni- and Cr-doped Mn4N films. Furthermore, with a view to device applications of Mn4N, we present the formation of ultrathin (~4 nm) Pt epitaxial films with a <100> orientation on MgO (001), which is essential for injecting spins into Mn4N-based overlayers by utilizing the spin Hall effect of heavy metals. Introduction In spintronics, one of the key areas of research is how efficiently spin precession and reversal can be induced using spin-transfer torque (STT) and spin-orbit torque (SOT) [1–3]. Under such circumstances, ferrimagnets (FIMs) have garnered increasing attention [4–22]. The main characteristic of ferrimagnets is that the magnetic moments of different sublattices arrange so as to partially compensate each other. Consequently, unlike antiferromagnets, ferrimagnets retain an imbalance between the magnetic moments of different sublattices, leading to a finite spontaneous magnetization that is much smaller than that of ferromagnets. This property offers a significant advantage in the development of application technologies based on STT and SOT, where the angular momentum carried by the incident spin currents acts on the magnetization, inducing precession or reversal of the magnetization. In the case of STT, the spin-polarized current is generated within the magnetic layer by exchange interactions, while in the case of SOT, a pure spin current can often be generated by the spin Hall effect (SHE) within a heavy-metal layer in contact with the magnetic layer. It is also generated by the Rashba-Edelstein effect at interfaces that break inversion symmetry or in topological insulators. In either case, the torque arises from the conservation of angular momentum. When a ferrimagnet is close to an angular momentum compensation (AC) and/or a magnetic compensation (MC) point, the angular momentum carried by spin currents readily affects the reduced magnetization, resulting in a high-speed magnetization dynamics [12,14,19]. Therefore, ferrimagnets are attractive materials for the development of devices based on STT and SOT, such as domain-wall memory, three-terminal magnetic random-access memory, and devices utilizing skyrmion motion [23–25]. The most commonly used approach to achieve MC and AC in spintronics is based on alloys combining rare earth elements and transition metals, such as Gd0.44Co0.56, Tb0.21Co0.79, and Gd3Fe5O12 [5,6,10,12,13,20]. When the magnetizations of the two species are antiparallel, there exists a certain concentration at which compensation occurs at room temperature (RT). One of such materials is rare-earth-free Mn4N [26–31], and thereby we have focused on this material [32,33]. As shown in Figure 1, Mn4N has an anti-perovskite structure. It consists of two sublattices: one containing Mn(I) at the corner sites and the other containing Mn(II) at the face-centered sites. Note that there are two sites, A and B, for Mn(II) in Mn4N epaxial films (crystallographic space group; P4/mmm) since they are subjected to tensile stress in the in-plane direction (c/a < 1) [34], differently from Mn4N bulk crystals (crystallographic space group; Pm3ˉm). Extensive research has been conducted on Mn4N-based mixed crystals, particularly on Mn3AN, a non-collinear antiferromagnet, and a wide variety of studies are summarized in Ref [35]. Detailed analyses of Mn4N bulk crystals revealed that the noncollinear frustrated moments of Mn(II) sites of the (111) kagome planes tilt about 20° out-of-plane and are easily influenced by the substitutions on Mn(I) sites [28]. This study also discusses the mechanism by which impurities induce MC. Since the observation of perpendicular magnetic anisotropy (PMA) in Mn4N films [36–38], interest in Mn4N-based epitaxial thin films has been growing together with the magnetic structures in Mn4N films, which is different from the triangular ferrimagnetism of the bulk [39,40]. Mn4N epitaxial thin films have attracted attention in the field of spintronics due to the following characteristics. Figure 1.Crystal structure of anti-perovskite Mn4N.The image shows the crystal structure of anti-perovskite MnN.4In. The structure consists of two types of manganese sites: Mn(I) located at the corner positions and Mn(II) at the face-centered positions. Mn(II) is further divided into two sites, labeled as Mn(II)A and Mn(II)B. The nitrogen atom is centrally positioned within the lattice. The arrangement forms a cubic structure with alternating positions occupied by manganese and nitrogen atoms. The Mn(I) atoms are depicted in green, while the Mn(II) atoms are shown in orange. The nitrogen atom is represented in purple. This structure is significant in spintronics research, particularly in the study of spin-transfer torque and spin-orbit torque effects.Crystal structure of anti-perovskite MnN.4In with Mn(I) and Mn(II) sites. A relatively large PMA and a small spontaneous magnetization (MS ~ 100 kA m−1) have been reported in Mn4N films deposited on various substrates, including glass, MgO (001), Si (001), LaAlO3[LAO] (001), (LaAlO3)0.3(Sr2TaAlO6)0.7(001), and SrTiO3[STO] (001) [34,36,37,41–44]. The uniaxial magnetic anisotropy constant (Ku) was reported to be of the order of 105 J m−3 for Mn4N thin films [32].The effective spin polarization derived from the domain wall (DW) velocity measurements in Mn4N strips is relatively high (p = 0.8) [45]. This enhances magnetic transport effects utilized in spintronics, such as the anomalous Hall effect (AHE) and the giant magnetoresistance, suggesting that Mn4N could be used as a magnetic soft layer in magnetic tunnel junction stacks.The magnetization reversal is caused by nucleation and the motion of Bloch DWs. The combination of a small MS and a relatively large P results in higher DW velocities due to the STT [19,45]. The current-induced DW velocity due to the adiabatic torque is expressed by(1)v≈11+α2gμB2eMSPJ∝PMS. Crystal structure of anti-perovskite Mn4N. A relatively large PMA and a small spontaneous magnetization (MS ~ 100 kA m−1) have been reported in Mn4N films deposited on various substrates, including glass, MgO (001), Si (001), LaAlO3[LAO] (001), (LaAlO3)0.3(Sr2TaAlO6)0.7(001), and SrTiO3[STO] (001) [34,36,37,41–44]. The uniaxial magnetic anisotropy constant (Ku) was reported to be of the order of 105 J m−3 for Mn4N thin films [32]. A relatively large PMA and a small spontaneous magnetization (MS ~ 100 kA m−1) have been reported in Mn4N films deposited on various substrates, including glass, MgO (001), Si (001), LaAlO3[LAO] (001), (LaAlO3)0.3(Sr2TaAlO6)0.7(001), and SrTiO3[STO] (001) [34,36,37,41–44]. The uniaxial magnetic anisotropy constant (Ku) was reported to be of the order of 105 J m−3 for Mn4N thin films [32]. The effective spin polarization derived from the domain wall (DW) velocity measurements in Mn4N strips is relatively high (p = 0.8) [45]. This enhances magnetic transport effects utilized in spintronics, such as the anomalous Hall effect (AHE) and the giant magnetoresistance, suggesting that Mn4N could be used as a magnetic soft layer in magnetic tunnel junction stacks. The effective spin polarization derived from the domain wall (DW) velocity measurements in Mn4N strips is relatively high (p = 0.8) [45]. This enhances magnetic transport effects utilized in spintronics, such as the anomalous Hall effect (AHE) and the giant magnetoresistance, suggesting that Mn4N could be used as a magnetic soft layer in magnetic tunnel junction stacks. The magnetization reversal is caused by nucleation and the motion of Bloch DWs. The combination of a small MS and a relatively large P results in higher DW velocities due to the STT [19,45]. The current-induced DW velocity due to the adiabatic torque is expressed by(1)v≈11+α2gμB2eMSPJ∝PMS. The magnetization reversal is caused by nucleation and the motion of Bloch DWs. The combination of a small MS and a relatively large P results in higher DW velocities due to the STT [19,45]. The current-induced DW velocity due to the adiabatic torque is expressed by(1)v≈11+α2gμB2eMSPJ∝PMS. Here, v is the DW velocity, α is the damping constant, g is the Landé factor, e is the electron charge, and J is the current density. The interest of the small magnetization is both for STTs through magnetic tunnel junction and for SOTs. iv) The combination of a high PMA and a moderate exchange stiffness results in a relatively sharp Bloch wall. A sharp DW can enhance the spin transfer effect through non-adiabatic contributions [46] and may contribute to the rapid motion of the DW at smaller currents [47]. The width (Δ) of the DW in a nanowire is given by Eq. (2) [48], (2)Δ=Astiff/Ku.Here, Astiff is the exchange stiffness. As Ku increases, Δ decreases, and the threshold current density (jth) required to move the DWs decreases [47,49]. In this regard, materials with PMA are advantageous for current-induced domain wall motion (CIDWM) devices. Furthermore, to scale down DW-based logic devices [50], three terminal memories [51], or racetrack memories [52] to the ultimate technology node, the DW width must be sufficiently reduced. This paper outlines our research group’s recent findings on impurity-doped Mn4N epitaxial films. Mainly focusing on experimental results from X-ray absorption spectroscopy (XAS), X-ray magnetic circular dichroism (XMCD), and AHE loop measurements, we investigate the presence or absence of MC at RT in Mn4N epitaxial films doped with various impurities (Ni, Cu, Ag, Au, and Pd). Finally, with a view to device applications of Mn4N on MgO (001) surfaces, we present the formation of approximately 4 nm thick <1 0 0>-oriented Pt epitaxial films on MgO (001) substrates. This is essential for injecting spin into the Mn4N-based film by utilizing the SHE of heavy metals. Methods Thin film growth Mn4N films have been grown on various substrates by conventional thin-film growth techniques like molecular beam epitaxy (MBE), sputtering, and others [35]. Instead of the MgO (001) substrates commonly used in spintronics, we employed STO (0 0 1) substrates, which exhibit minimal lattice mismatch with Mn4N. This was done to grow high-quality Mn4N epitaxial films and to clearly demonstrate the effects of impurity doping [32]. The lattice constant of MgO, STO, and Mn4N is 0.421 nm, 0.391 nm, and 0.387 nm, respectively. Before deposition, the STO substrates were etched for 40 s in buffered hydrofluoric acid. This was done to create a step-and-terrace surface structure, thereby ensuring a surface that was atomically smooth and terminated with Ti [53]. Approximately 10–25-nm-thick impurity doped Mn4N layers were grown on STO (0 0 1) substrates at a substrate temperature of around 450°C by MBE. The optimum substrate temperature during thin-film deposition may vary depending on the type of impurity. Metal elements including Mn were supplied by Knudsen cells, and nitrogen was supplied using a radio-frequency (RF) plasma source. To prevent oxidation, the sample surfaces were covered with sputter-deposited 2–3-nm thick capping layers like SiO2, Pt, and others. Characterizations of film properties The crystalline quality of the grown films was characterized by 20-kV reflection high-energy electron diffraction (RHEED) and X-ray diffraction (XRD; Rigaku SmartLab) with a Cu Kα radiation source. The composition ratio of Mn to doped impurity element was measured by energy dispersive X-ray spectroscopy (EDX, HITACHI SU7000) or just by their deposition rates. AHE loops were acquired with a physical properties measurement system (PPMS, Quantum Design, Inc.) using the Van der Pauw method. The transverse voltage (Vy) is expressed as Eq. (3) [54]: (3)Vy=RHBzt+ρAHEtIx=ρyxtIx. Here, RH, Bz, t, ρAHE, Ix, and ρyx are the normal Hall coefficient, the magnetic flux density perpendicular to the sample surface, the film thickness, the anomalous Hall resistivity, the longitudinal current, and the Hall resistivity, respectively. X-ray absorption spectroscopy (XAS) and X-ray magnetic circular dichroism (XMCD) measurements were conducted at beamline BL-16A of KEK-PF, Japan. During the measurement, circularly polarized soft X-rays were incident. XMCD spectra were acquired in a single scan, while switching between left and right polarizations at a frequency of 10 Hz using two undulators. The spectra at the Mn-L2,3 absorption edges were measured by the total electron yield (TEY) method. All measurements were performed at RT, either in the saturated magnetization state at 3‒5 T or in the remanent magnetization state at 0 T. To minimize systematic errors arising from the measurement system, the sign of the XMCD spectrum obtained at μ0H = −5 T was inverted, and the sum of this spectrum and the XMCD spectrum obtained at μ0H = +5 T was adopted as the spectrum obtained at 5 T. Similarly, the XMCD spectrum at 0 T consists of two spectra: one obtained by increasing the external magnetic field from μ0H = −5 T to −0 T, and the other obtained by decreasing the external magnetic field from μ0H = +5 T to + 0 T. The X-ray absorption near edge structure (XANES) measurements were performed in the fluorescence yield mode (TFY) at BL08W of the NanoTerasu facility, targeting the Pd and Ag-L3 edges. All measurements were performed at RT. The experimental data were processed to eliminate the backgrounds and normalize the spectra using Athena, an X-ray absorption spectroscopy (XAS) processing system [55]. The XANES simulations were conducted using the finite-difference method for near-edge structure (FDMNES) codes [56]. Results and discussion XAS and XMCD spectra of Mn4N epitaxial films Figure 2 shows typical examples of XAS and XMCD spectra measured for an Mn4N epitaxial film on STO (001) at the Mn-L2,3 edges under external magnetic fields of μ0H = 0 T and 5 T applied normal to the sample plane. About 5 T was sufficiently high to saturate the magnetization. The spectra at μ0H = 0 T and 5 T are nearly identical. The XMCD spectrum of the Mn4N film shows a sharp peak (peak α) and a broad peak (peak β) with the opposite sign. Figure 3 shows the theoretical density of states of Mn4N by the all-electron full-potential linearized augmented-plane-wave method [57]. Narrow and wide bandwidths are obtained for Mn(I) and Mn(II) atoms, respectively. This is because the Mn(I) atoms at the corner sites are farther from the nitrogen atoms in the body-centered lattice, making it more difficult for them to form hybridized orbitals compared to the Mn(II) atoms at the face-centered sites. As a result, the electrons of the Mn(I) atoms are relatively localized and are thought to contribute primarily to the sharp peak α. On the other hand, the broad peak β is primarily caused by the itinerant electrons of the Mn(II) atoms at the face-centered sites, due to hybridized orbitals formed with nitrogen [57]. Therefore, the fact that the signs of peak α and peak β are opposite reflects that the magnetic moments of the corner-site Mn(I) sublattice and the face-centered Mn(II) sublattice are oriented in opposite directions, and the magnetic moment of the Mn(I) sublattice is parallel to the magnetization. Using the Bohr magneton (μB), the calculated spin magnetic moment for the Mn(I) is 3.07 μB/atom, and those for the Mn(II) at inequivalent A and B sites are −2.29 μB/atom and 0.64 μB/atom, respectively. This means that the Mn(I) moment is aligned antiparallel to the magnetic filed, which is opposite to what was observed in the XMCD spectra (Figure 2). This suggests that even for the metallic Mn4N, more advanced modeling is required that takes into account the core hole-3d and 3d–3d interactions. Another possible factor is that defects present in the thin films formed by the experiments may have an influence. This could be the reason why the experimentally obtained XMCD spectra differ from the predictions derived from first-principles calculations. However, as will be discussed later, when performing XAS and XMCD measurements on Mn4N thin films doped with impurity elements, we always use Mn4N epitaxial thin films as the reference sample. Consequently, the results shown in Figure 2 have been consistently obtained; we have never obtained results in which the signs of peaks α and β are reversed, as predicted by theoretical calculations, nor have we found any reports of such results. Anyway, now that we understand the characteristics of XAS and XMCD spectra of MBE-grown Mn4N epitaxial films, we will next discuss the results of impurity doped Mn4N epitaxial films. Figure 2.XAS and XMCD spectra of Mn4N epitaxial films measured at the Mn-L2,3 edges at μ0H = 0 T and 5T. The spectrum at 0 T is obtained by first applying an external magnetic field up to 5 T, then lowering the field to 0 T, and finally performing the measurement.Image A displays a line graph titled 'Mn L₂,₃ edges' and 'TEY, RT'. The x-axis, labeled 'Photon energy (eV)', spans 630 to 670 in increments of 10. The y-axis is labeled 'XAS (arb. unit)' without numeric ticks. Two curves, 'μ₀H = 0 T' and 'μ₀H = 5 T', rise sharply to a peak near 640, drop to a smaller peak near 653 and flatten towards 670. Labels 'L₃' and 'L₂' mark the peaks. Image B shows a similar graph with the same x-axis range and label. The y-axis is labeled 'XMCD (arb. unit)' with a dotted baseline at zero. Two curves, 'μ₀H = 0 T' and 'μ₀H = 5 T', feature a deep negative trough near 640 labeled 'alpha', a positive peak near 642 labeled 'beta', a smaller hump near 653 and approach zero towards 670.Two line graphs of X ray absorption spectroscopy and X ray magnetic circular dichroism versus photon energy. Figure 3.Total density of states of Mn (I), Mn(II)A, and Mn(II)B in Figure 1 [38].Three line graphs show density of states for Mn I, Mn IIB and Mn IIA. Mn I graph: m_spin = 3.07 μB, n_d = 5.47. Energy (eV) from -8 to 4, Density of states (states/eV/atom) from -5 to 5. Up curve: near 0 below -4, peaks between 0.5 and 2 at 3-4. Down curve: near 0 above 0, trough at -3 to -2.5 reaching -3 to -4. Mn IIB graph: m_spin = 0.64 μB, n_d = 5.32. Same axes. Up curve: peak near -1 at 2-3, structure between 0.5 and 2 at 2. Down curve: dips between -2 and 2, deepest near -1 at -3. Mn IIA graph: m_spin = -2.29 μB, n_d = 5.35. Same axes. Up curve: hump from -4 to 0, peak near -3 at 1-2, near 0 above 0. Down curve: near 0 below 0, strong negative above 0, trough near 1.5 to 2 at -3 to -4.Three line plots of density of states versus energy for Mn I, Mn IIA and Mn IIB, split by spin. XAS and XMCD spectra of Mn4N epitaxial films measured at the Mn-L2,3 edges at μ0H = 0 T and 5T. The spectrum at 0 T is obtained by first applying an external magnetic field up to 5 T, then lowering the field to 0 T, and finally performing the measurement. Total density of states of Mn (I), Mn(II)A, and Mn(II)B in Figure 1 [38]. Magnetic compensation in Ni-doped Mn4N epitaxial films Figure 4 shows the XAS and XMCD spectra of the Mn- and Ni-L2,3 edges of Ni-doped Mn4N epitaxial films (Mn4−xNixN) formed on STO (001) under conditions where an external magnetic field of μ0H = 3 T was applied perpendicularly normal to the sample plane [26]. The Ni composition x was determined from the deposition rates of Mn and Ni. In the XAS spectrum of Ni, as there is no shoulder peak appearing 2 eV higher than the main peak, which would be expected if Ni occupies an Mn(II) site. Therefore, it is thought that Ni substituted for an Mn(I) site, i.e., it is Ni(I). Since there was no significant difference in the XAS spectra of the Mn- and Ni-L2,3 edges regardless of the value of x, we focus here on the XMCD spectra. The XMCD spectrum of the Mn-L2,3 edges for the x = 0.10 sample, shown in Figure 4(a), is similar to that for the Mn4N film in Figure 2, meaning that the magnetic moment of the Mn(I) sublattice is aligned parallel to the magnetization at x = 0.10. In contrast, the sign of the XMCD signal of the Ni-L3 edge in Figure 4(b) is opposite to that of peak α in Figure 4(a). This result indicates that the magnetic moment of Ni(I) is oriented antiparallel to the magnetization. The element-specific magnetic moment can be calculated using the sum-rule analysis [58,59]. Table 1 summarizes the spin (mspin), orbital (morb), and total (mtot = mspin + morb) magnetic moments per Mn and Ni atom for samples at x = 0.10 and 0.25. For the number of holes in Mn and Ni atoms, 5.2 was adopted for Mn [60] and 1.4 for Ni [61], regardless of whether they were at I or II sites. For the x = 0.1 sample, as suggested by the sign reversal in Figure 4(b), the magnetic moment of Ni(I) is negative; therefore, the addition of Ni causes a decrease in magnetization. When the Ni composition was increased further to x = 0.25, the sign of the XMCD signals of Mn and Ni reversed compared to that at x = 0.10, and the magnetic moment of the Mn(II) sublattice is parallel to the magnetization. Furthermore, the magnitude of mtot for Ni remained almost unchanged between x = 0.1 and 0.25 in Table 1, suggesting that the preferred occupancy sites of the Ni atoms did not change. These results are consistent with the aforementioned analysis based on the XMCD spectrum. We therefore conclude that the MC composition (xMC) in Mn4−xNixN lies between 0.10 and 0.25. The Ni composition x = 0.10 is considered very close to xMC. This is because the mtot per unit cell was as small as approximately 0.009 μB at x = 0.10 and increased to approximately 0.39 μB at x = 0.25. These values correspond to approximately 1.5 kA m−1 and 63 kA m−1, respectively. Furthermore, the coercive force of the MOKE hysteresis loop was approximately four times that at x = 0.25. Similar sign reversal of the XMCD spectra were also observed at the Mn-L2,3 and Co-L2,3 edges in the Mn4-xCoxN epitaxial films [27,57], and the results can be explained by the view that Co first substitutes for Mn(I) sites and then for Mn(II) sites as the Co concentrations increases [27]. Figure 4.XAS (blue) and XMCD (black) spectra measured at (a)(c) Mn-L2,3 and (b)(d) Ni- L2,3 edges for (a)(b) Mn3.9Ni0.1N films and (c)(d) Mn3.75Ni0.25N films [26]. Reproduced by permission from [26], copyright [2020, AIP Publishing].Image A: Mn L2,3 edges (x=0.10) graph shows photon energy (630-670 eV) vs. intensity. XAS peaks at 639-640 (L3) and 652-654 (L2). XMCD, amplified 20x, shows a negative dip at 639-640 (alpha), a positive peak at 641-643 (beta) and smaller features at 652-655. Image B: Ni L2,3 edges (x=0.10) graph shows photon energy (830-900 eV) vs. intensity. XAS peaks at 851-852 (L3) and 869-871 (L2). XMCD, amplified 5x, shows a positive spike at 851-852 and a negative feature at 869-871. Image C: Mn L2,3 edges (x=0.25) graph shows photon energy (630-670 eV) vs. intensity. XAS peaks at 639-640 (L3) and 652-654 (L2). XMCD, amplified 20x, shows a positive peak at 638-639 (alpha), a negative dip at 640-642 (beta) and smaller features at 652-655. Image D: Ni L2,3 edges (x=0.25) graph shows photon energy (830-900 eV) vs. intensity. XAS peaks at 851-852 (L3) and 869-871 (L2). XMCD, amplified 8x, shows a strong negative dip at 851-852 and a positive peak at 869-871.Four line plots of X ray absorption spectroscopy and X ray magnetic circular dichroism spectra at Mn and Ni edges.Table 1.Spin (mspin), orbital (morb), and total (mtot) magnetic moments of Mn and Ni atoms in Mn4−xNixN deduced by sum-rule analysis.CompoundsAtomMagnetic moment [μB per atom]mspinmorbmtotMn3.9Ni0.1NMnNi0.024 ± 0.011−0.20 ± 0.03−0.016 ± 0.002−0.024 ± 0.0040.008 ± 0.014−0.22 ± 0.03Mn3.75Ni0.25NMnNi0.081 ± 0.0200.17 ± 0.020.011 ± 0.0050.010 ± 0.0020.092 ± 0.0250.18 ± 0.02 XAS (blue) and XMCD (black) spectra measured at (a)(c) Mn-L2,3 and (b)(d) Ni- L2,3 edges for (a)(b) Mn3.9Ni0.1N films and (c)(d) Mn3.75Ni0.25N films [26]. Reproduced by permission from [26], copyright [2020, AIP Publishing]. Table 1.Spin (mspin), orbital (morb), and total (mtot) magnetic moments of Mn and Ni atoms in Mn4−xNixN deduced by sum-rule analysis.CompoundsAtomMagnetic moment [μB per atom]mspinmorbmtotMn3.9Ni0.1NMnNi0.024 ± 0.011−0.20 ± 0.03−0.016 ± 0.002−0.024 ± 0.0040.008 ± 0.014−0.22 ± 0.03Mn3.75Ni0.25NMnNi0.081 ± 0.0200.17 ± 0.020.011 ± 0.0050.010 ± 0.0020.092 ± 0.0250.18 ± 0.02 Spin (mspin), orbital (morb), and total (mtot) magnetic moments of Mn and Ni atoms in Mn4−xNixN deduced by sum-rule analysis. Figure 5 shows the dependence of ρAHE on the perpendicular magnetic field for Mn4−xNixN epitaxial films (x = 0, 0.10, 0.25, and 0.5) on STO (001) at RT [62]. For the x = 0, 0.10, and 0.25 samples, the squareness ratio was almost 1. The sign of ρAHE reversed between x = 0.10 and 0.25. This shows that the relative orientation of the spin polarization and the magnetic moment changed between x = 0.10 and 0.25, confirming the crossover in the MC composition. Researchers generally use such a sign reversal to confirm the presence of MC in rare-earth ferrimagnets [12,13]. When the x was increased further to x = 0.5, a much greater magnetic field was required to saturate the magnetization, probably due to a degradation in crystal quality. Figure 5.AHE loops measured at RT for Mn4−xNixN (x = 0, 0.1, 0.25, 0.5) films on STO (001). A magnetic field was applied normal to the sample plane [62]. Reproduced by permission from [62], copyright [2019, AIP Publishing].A line graph with an inset schematic shows rho_AHE (µΩ·cm) vs. µ₀H (T). The µ₀H axis ranges from -3 to 3 and the rho_AHE axis from -4 to 4. Dotted lines mark rho_AHE = 0 and µ₀H = 0. Curve x=0: a hysteresis loop near rho_AHE +4 at negative field, switching steeply near µ₀H 0 to 0.5 to about -4 at positive field, returning near µ₀H 0 to -0.5. Curve x=0.1: a wider loop with plateaus near +3 at negative field and -3 at positive field, switching steeply near µ₀H 0. Curve x=0.25: a loop with plateaus near -2 at negative field and +2 at positive field, switching near µ₀H 0. Curve x=0.5: a slanted, weakly hysteretic trace from about -0.6 at µ₀H -3 to +0.6 at µ₀H +3, crossing rho_AHE = 0 near µ₀H 0. Inset schematic: a square sample with four corner contacts labeled V+, V-, I+, I-, circles marked V and I and a symbol indicating H applied out of plane.A line graph of anomalous Hall resistivity versus magnetic field for four Mn4 minus x Nix N films. AHE loops measured at RT for Mn4−xNixN (x = 0, 0.1, 0.25, 0.5) films on STO (001). A magnetic field was applied normal to the sample plane [62]. Reproduced by permission from [62], copyright [2019, AIP Publishing]. As mentioned above, Mn4N epitaxial films containing a small amount of Ni exhibit a decrease in magnetization, which approaches zero at RT near x = 0.1. In line with the reduction in magnetization of Mn4−xNixN, extremely fast current-induced domain wall (DW) motion driven purely by STT, exceeding 3,000 m s−1, was demonstrated in Mn4−xNixN microstrips at RT near the xMC [19]. Figure 6(a) shows the relationship between DW velocity and current density in Mn4−xNixN microstrips (x = 0, 0.1, 0.15, 0.2, 0.25) with different Ni concentrations on either side of the xMC. For comparison, the results for Mn4N microstrips are also shown in Figure 6(b) [45]. In Figure 6(a), for MC compositions of x ≤ 0, 0.1, 0.15, and 0.2, the DW moved in the direction of the electron flow, and its mobility increased with increasing Ni composition. However, when the value of x exceeded 0.25, the direction of DW motion reversed, and at a current density of j = 1.2 × 1012 A m−2, a large vDW of up to 3000 m s−1 relative to the current flow direction was obtained. This high mobility, obtained in Mn4−xNixN microstrips near the xMC, exceeded the highest velocity obtained by SOT driving in the absence of an external magnetic field in a ferromagnetic Pt/GdCo thin film at 250 K [14]. In these materials, the xMC and the angular momentum compensation composition (xAC) are different. In contrast, in Mn4N, the two magnetic sublattices consist of Mn, Mn(I), and Mn(II). Considering that the Ni composition in Mn4−xNixN thin films is only about 5 at% of Mn(I) atoms, it can be assumed that the gyromagnetic factors of Mn(I) and Mn(II) are the same. As a result, the xMC and the xAC coincide. Figure 6.(a) Domain wall (DW) speed against current density for Mn4−xNixN strips for various Ni concentrations on either side of the MC composition. MS = 71 kA m−1 corresponds to Mn4N (x = 0). The filled/empty symbols show the DW velocity below/above the MC composition: the direction of DW motion depends on whether the Ni composition is below or above the MC composition [19]. Reproduced by permission from [19], copyright [2021, American Chemical Society]. (b) DW velocity versus current density measured for Mn4N strips. The DW velocity reaches an average above 900 m s−1 at 1.3 × 1012 A/m2. The dotted line is a guide to the eye [45]. Reproduced by permission from [45], copyright [2019, American Chemical Society].Image A: Mn4-x Nix N strips scatter plot with error bars. Velocity (km/s) on y-axis from -3 to 3, x-axis from 0.0 to 1.2. Legend: +71, +26, +22, -20, -29 kA/m at 10 or 30 nm. Positive series rise from 0 at 0.1 to 0.9, 1.5, 2.0 at 1.2. Negative series fall from 0 at 0.1 to -0.7, -2.7 at 1.2. Image B: Mn4 N strips scatter plot with error bars and dashed diagonal guide. Current density (10^12 A/m^2) on x-axis from 0.0 to 1.5, Velocity (m/s) on y-axis from 0 to 1000. Two series: 1 µm-wide and 2 µm-wide. Values rise from 0 at 0.3 to 850-950 near 1.2-1.3, with 1 µm-wide series generally above 2 µm-wide series.Two scatter plots of domain wall velocity versus current density, showing rising and falling trends by sample. (a) Domain wall (DW) speed against current density for Mn4−xNixN strips for various Ni concentrations on either side of the MC composition. MS = 71 kA m−1 corresponds to Mn4N (x = 0). The filled/empty symbols show the DW velocity below/above the MC composition: the direction of DW motion depends on whether the Ni composition is below or above the MC composition [19]. Reproduced by permission from [19], copyright [2021, American Chemical Society]. (b) DW velocity versus current density measured for Mn4N strips. The DW velocity reaches an average above 900 m s−1 at 1.3 × 1012 A/m2. The dotted line is a guide to the eye [45]. Reproduced by permission from [45], copyright [2019, American Chemical Society]. Non-magnetic compensation in Cr-doped Mn4N epitaxial films Here, we present an example in which MC did not occur despite the introduction of transition metal impurities. This occurred in both Cr-doped Mn4N epitaxial films and Fe-doped Mn4N ones [63]. Figure 7 shows the XAS and XMCD spectra of Mn4−xCrxN films (x = 0.13, 0.37, and 0.66) at the Mn-L2,3 edges [63]. Both an external magnetic field of up to μ0H = ± 5 T and circularly polarized soft X-rays with right or left polarization were applied at the so-called “magic angle of 54.7◦ to the sample normal [64–66]. At x = 0.13, the XAS and XMCD spectra at the Mn-L2,3 edges in Figure 7(a) were not significantly different from those of Mn4N in Figure 2. With increasing the Cr composition x further, the intensities of both peaks α and β decrease in Figure 7(b,c). This suggests that both the Mn(I) and Mn(II) sites were replaced by Cr atoms. No sign reversal of the XMCD spectrum, such as that observed in Mn4−xNixN, was observed. Figure 8 shows the AHE loops measured for Mn4−xCrxN epitaxial films with various x values. Polarity reversal was not observed in the ρAHE loops within the measured Cr composition range. The difference in Ni and Cr atoms is the magnetic moment of these elements and their preference of site occupation. From an experimental perspective, the magnetic moment of the Mn(I) sublattice is parallel to the magnetization; consequently, its magnitude is greater than that of the Mn(II) sublattice. Therefore, it can be stated that an MC occurs when at least one of the following conditions is met. Figure 7.XAS and XMCD spectra measured measured at the Mn-L2,3 edges in Mn4−zCrzN epitaxial films for (a) z = 0.13, (b) z = 0.37, and (c) z = 0.66 [63]. Reproduced by permission from [63], copyright [2022, Elsevier].Image A: Two stacked line graphs labeled 'Mn L₂,₃ edges' and 'x = 0.13'. Top graph: X-axis 'Photon energy (eV)' from 630 to 670; tall 'L₃' peak near 640, smaller 'L₂' peak near 655. Bottom graph: Y-axis 'XMCD int. (arb. unit)'; deep negative 'alpha' trough near 640, positive 'beta' peak near 642, smaller features near 655-660. Text: 'μ₀H = 5 T, TEY'. Image B: Same layout, 'x = 0.37'. Top: Peaks lower than in A. Bottom: Reduced 'alpha' trough and 'beta' peak, small bumps near 655-660. Text: 'μ₀H = 5 T, TEY'. Image C: Same layout, 'x = 0.66'. Top: Further reduced peaks. Bottom: Smallest 'alpha' trough and 'beta' peak among A-C, small features near 655-660. Text: 'μ₀H = 5 T, TEY'.Three line graphs of XAS and XMCD spectra showing decreasing peak intensity with increasing x. Figure 8.AHE loops of Mn4−zCrzN epitaxial films [63]. Reproduced by permission from [63], copyright [2022, Elsevier].A line graph of anomalous Hall resistivity versus magnetic field for five x values, with hysteresis. Substituting Mn(I) sites to reduce the magnetic moment of the Mn(I) sublattice.Substituting Mn(II) with magnetic moments increases the magnetic moment of the Mn(II) sublattice. XAS and XMCD spectra measured measured at the Mn-L2,3 edges in Mn4−zCrzN epitaxial films for (a) z = 0.13, (b) z = 0.37, and (c) z = 0.66 [63]. Reproduced by permission from [63], copyright [2022, Elsevier]. AHE loops of Mn4−zCrzN epitaxial films [63]. Reproduced by permission from [63], copyright [2022, Elsevier]. Substituting Mn(I) sites to reduce the magnetic moment of the Mn(I) sublattice. Substituting Mn(I) sites to reduce the magnetic moment of the Mn(I) sublattice. Substituting Mn(II) with magnetic moments increases the magnetic moment of the Mn(II) sublattice. Substituting Mn(II) with magnetic moments increases the magnetic moment of the Mn(II) sublattice. In the case of Mn4−xNixN films, MC occurred because the Mn(I) site was substituted with Ni, which has a magnetic moment antiparallel to that of Mn(I) atoms. In the case of Mn4−xCrxN films, however, no preferred substitution site has been identified. Therefore, it falls into neither of the two categories described above. Magnetic compensation in Mn4N epitaxial films using non-magnetic impurities (Cu, Ag, Au) We next move on to Mn4N epitaxial films doped with non-magnetic elements and compare the results of doping three kinds of impurity atoms (Cu, Ag, and Au) in almost equal amounts. Figure 9(a–c) shows reflection high-energy electron diffraction (RHEED) patterns for 10-nm-thick Mn4−xCuxN, Mn4−yAgyN, and Mn4−zAuzN films, respectively, taken along the [100] and [1 1 0] azimuths of STO (001). The composition of impurity element was varied in the range 0.05–0.30. When the impurity composition was small, the streaky RHEED patterns with Kikuchi lines were observed regardless of impurity element, meaning that smooth epitaxial films were grown for all the Mn4N-based films. However, when the impurity composition was further increased to 0.30, differences began to appear on the type of impurity. Kikuchi lines disappeared except in Mn4−xCuxN films, and the RHEED streaks lost their sharpness and became broad in Mn4−yAgyN, and Mn4−zAuzN films. Figure 9.RHEED patterns measured on (a) Mn4−xCuxN films, (b) Mn4−yAgyN films, and (c) Mn4−zAuzN films, taken along the [100] and [1 1 0] azimuths of STO.Images A, B and C display reflection high-energy electron diffraction patterns for Mn4-xCuxN, Mn4-yAgyN and Mn4-zAuzN, respectively. Each image lacks numeric axes or units, organized by azimuths 100 and 110 and dopant levels x, y, z ranging from 0.05 to 0.30. In Image A, vertical streaks rise from a bright band across all x values, with faint diagonal lines. Image B shows distinct vertical streaks for y values 0.05 to 0.25, becoming diffuse at y=0.30. Image C displays vertical streaks for z values 0.05 to 0.25, broadening at z=0.30. At dopant level 0.30, streak sharpness decreases for Ag and Au, while Cu maintains prominent streaks.Composite reflection high-energy electron diffraction patterns for Mn4N films doped with Cu, Ag and Au. RHEED patterns measured on (a) Mn4−xCuxN films, (b) Mn4−yAgyN films, and (c) Mn4−zAuzN films, taken along the [100] and [1 1 0] azimuths of STO. Figure 10(a–c) show the out-of-plane XRD patterns of Mn4−xCuxN, Mn4−yAgyN, and Mn4−zAuzN films, respectively. The 002 and 004 reflections of these films were observed as shown by close triangles. For Ag- and Au-doped Mn4N films, the peak position shifted toward lower 2θ angles as the doping concentration increases; at y, z = 0.30, it nearly overlapped with the substrate peak. These results indicate that <100>-oriented epitaxial films were grown for these samples. Figure 11(a–c) shows the in-plane XRD patterns of Mn4−xCuxN, Mn4−yAgyN, and Mn4−zAuzN films, respectively, with scattering vectors set along the [100] direction of the STO. In this case, almost no reflection from the epitaxial thin film was visible. This is likely because it overlapped with the peaks from the substrate. Figure 10.Out-of-plane XRD patterns for (a) Mn4−xCuxN films, (b) Mn4−yAgyN films, and (c) Mn4−zAuzN films.The images A, B and C display stacked line plots for Mn4-xCuxN, Mn4-yAgyN and Mn4-zAuzN, respectively. Each plot has an x-axis labeled 2 theta (deg) ranging from 20 to 120 and a y-axis labeled Log intensity (arb. unit). For each image, six traces are shown with varying values: x, y, z equals 0.05, 0.10, 0.15, 0.20, 0.25, 0.30. All traces exhibit sharp peaks near 2 theta at approximately 50 and 105, labeled 002 and 004, respectively.Three stacked line plots of out-of-plane X-ray diffraction patterns versus doping level. Figure 11.In-plane XRD patterns for (a) Mn4−xCuxN films, (b) Mn4−yAgyN films, and (c) Mn4−zAuzN films.Image A displays Mn4-xCuxN X-ray diffraction patterns as stacked line plots. The x-axis is 2 theta (degrees) from 20 to 120, with major ticks every 20. The y-axis is Log intensity (arb. unit). Six traces are labeled x = 0.05, 0.10, 0.15, 0.20, 0.25, 0.30, each showing peaks near 45, 50, 75 and 105. Image B shows Mn4-yAgyN patterns with identical axes and trace labels y = 0.05, 0.10, 0.15, 0.20, 0.25, 0.30, displaying similar peaks. Image C presents Mn4-zAuzN patterns, also with the same axes and trace labels z = 0.05, 0.10, 0.15, 0.20, 0.25, 0.30, showing peaks at the same positions.Three stacked line plots of in-plane X-ray diffraction patterns across increasing dopant concentration. Out-of-plane XRD patterns for (a) Mn4−xCuxN films, (b) Mn4−yAgyN films, and (c) Mn4−zAuzN films. In-plane XRD patterns for (a) Mn4−xCuxN films, (b) Mn4−yAgyN films, and (c) Mn4−zAuzN films. Figure 12(a–c) show the AHE loops of Mn4−xCuxN, Mn4−yAgyN, and Mn4−zAuzN epitaxial films, respectively. For all samples, the coercive field increased with increasing impurity concentration and then decreased. Furthermore, at impurity concentrations where the coercive force was high, the sign of ρyx reversed. Based on the sign reversal in the AHE loops in the Mn4−xNixN thin films in Fig. 5 and the corresponding sign reversal in the XMCD spectra in Figure 4, we conclude that the MC occurs between x, y = 0.10 and 0.15 in the Mn4−xCuxN [29] and Mn4−yAgyN epitaxial films [30], and between z = 0.15 and 0.20 in the Mn4−zAuzN epitaxial films. Figure 12.AHE loops of (a) Mn4−xCuxN films, (b) Mn4−yAgyN films, and (c) Mn4−zAuzN films.Image A: Mn4-xCuxN displays six rho_yx vs. mu_0H graphs. Y-axis: rho_yx (μΩ·cm), range -4 to 4. X-axis: mu_0H (T), range -1.5 to 1.5. Hysteresis loops saturate near ±4. For x=0.05, 0.10, loops are positive at negative fields, switching to negative near small positive fields. For x=0.15, 0.20, 0.25, 0.30, polarity reverses, switching from negative to positive fields. Image B: Mn4-yAgyN shows similar graphs. Y-axis: rho_yx (μΩ·cm), range -4 to 4. X-axis: mu_0H (T). For y=0.05, 0.15, 0.20, 0.25, 0.30, x-range is -1.5 to 1.5. For y=0.10, x-range is -9 to 9, with mu_0H_max=9 T. Hysteresis loops saturate near ±4, switching around 0 to 1. Image C: Mn4-zAuzN also has six graphs. Y-axis: rho_yx (μΩ·cm), range -4 to 4. X-axis: mu_0H (T). For z=0.05, 0.10, 0.20, 0.25, 0.30, x-range is -1.5 to 1.5. For z=0.15, x-range is -9 to 9, with mu_0H_max=9 T. Loops saturate near ±4; z=0.15 shows a broad, slanted loop spanning -9 to 9.Three sets of AHE loop line graphs showing hysteresis and sign reversal versus impurity concentration. AHE loops of (a) Mn4−xCuxN films, (b) Mn4−yAgyN films, and (c) Mn4−zAuzN films. Figure 13(a–c) show the XAS and XMCD spectra of Mn4−xCuxN, Mn4−yAgyN, and Mn4−zAuzN epitaxial films, respectively, at x, y, z = 0.2. In all the samples, an external magnetic field of 0 T or 5 T and circularly polarized soft X-rays with right or left polarization were applied normal to the sample plane. It is noted that the spectra at 0 T were obtained by applying a magnetic field up to μ0H = ±5 T and then reducing it to μ0H = ±0 T. Based on the results in Figure 12, these impurity compositions are believed to exceed the xMC. At μ0H = 0 T, the XMCD spectra of the Mn4−xCuxN, Mn4−yAgyN, and Mn4−zAuzN epitaxial films were reversed compared to that of Mn4N epitaxial films. Therefore, it is concluded that the MC occurs in these three types of impurity-doped Mn4N epitaxial films. On the other hand, the XMCD spectra at μ0H = 5 T shown by the red dotted lines differ from those at μ0H = 0 T. Figure 14 shows the differences in the XMCD spectra at the Mn-L2,3 edges for (a) Mn4N, (b) Mn3.8Cu0.2N, (c) Mn3.8Ag0.2N, and (d) Mn3.8Au0.2N epitaxial films, measured at μ0H = 0 T and 5 T. Similar difference was also observed in the XMCD spectra for the Mn4N epitaxial films shown in Figure 14(a). Since such an unusual phenomenon did not appear in the electrical characteristics (AHE loops) shown in Figure 13, we suspect that it may be due to the diffusion of Mn into the SiO2 capping layers formed by sputtering. Therefore, to verify and discuss the sign reversal in the XMCD spectrum, it is better to saturate the magnetization in a high magnetic field and then lower the field to 0 T to conduct XAS and XMCD measurements, rather than comparing spectra under high magnetic fields. Figure 13.XAS and XMCD spectra at the Mn-L2,3 edges measured at 0 T and 5 T for (a)(a’) Mn3.8Cu0.2N, (b)(b’) Mn3.8Ag0.2N, and (c)(c’) Mn3.8Au0.2N. The spectrum at 0 T is obtained by first applying an external magnetic field up to 5 T, then lowering the field to 0 T, and finally performing the measurement.A) Mn3.8Cu0.2N: Mn L2,3 edges, TEY, RT. XAS Intensity from 0.0 to 1.0, photon energy 630-670 eV. Spectra at mu0H 0 T and 5 T. L3 peak at 640 eV reaches 1.0; drops to 0.2 by 645 eV. L2 feature at 652 eV peaks at 0.3; tail 0.1-0.2 at 670 eV. XMCD Intensity -0.8 to 0.4, trough at 641 eV -0.7 to -0.8; bump at 646-648 eV 0.1; approaches 0 by 670 eV. B) Mn3.8Ag0.2N: Similar axes as A. Spectra at mu0H 0 T and 5 T. L3 peak at 640 eV reaches 1.0; shoulder 642-644 eV near 0.5. L2 peak at 652 eV reaches 0.3; tail 0.1-0.2 at 670 eV. XMCD trough at 641 eV -0.7 to -0.8; positive region 646-650 eV 0.1-0.2; near 0 by 670 eV. C) Mn3.8Au0.2N: Similar axes as A. Spectra at mu0H 0 T and 5 T. L3 peak at 640 eV reaches 1.0; dip to 0.2 by 645 eV. L2 peak at 652 eV reaches 0.3; tail 0.1-0.2 at 670 eV. XMCD trough at 641 eV -0.7 to -0.8; bump 646-649 eV 0.1; near 0 by 670 eV.Six line plots comparing XAS and XMCD spectra at Mn L2,3 edges for Mn3.8M0.2N at 0 T and 5 T. Figure 14.Differences in the XMCD spectra at the Mn-L2,3 edges for (a) Mn4N, (b) Mn3.8Cu0.2N, (c) Mn3.8Ag0.2N, and (d) Mn3.8Au0.2N epitaxial films, measured at 0 T and 5 T. The spectrum at 0 T is obtained by first applying an external magnetic field up to 5 T, then lowering the field to 0 T, and finally performing the measurement.The images A, B, C and D depict line graphs titled Mn4N, Mn3.8Cu0.2N, Mn3.8Ag0.2N and Mn3.8Au0.2N, respectively. Each graph has an x-axis labeled Energy (electron volt) ranging from 630 to 670 and a y-axis labeled Difference (percent) ranging from -0.8 to 0.4. All curves represent XMCD(5 T) minus XMCD(0 T). In image A, the curve stays near 0.0 from 630 to 638, dips to -0.75 near 640, then rises to 0.0 by 642, with small positive bumps peaking at 0.1 from 645 to 655, ending near 0.0. Image B's curve is similar, dipping to -0.75 near 640, rebounding slightly above 0.0 by 642, with small oscillations peaking at 0.05 from 650 to 655, ending near 0.0. Image C shows a dip to -0.75 near 640, rising to 0.0 by 642, with peaks around 645 to 655 reaching 0.1, finishing near 0.05. Image D's curve dips to -0.75 near 640, returns to 0.0 by 642, with undulations peaking at 0.05 to 0.1 from 650 to 655, ending near 0.0.Four line graphs of XMCD difference spectra for Mn4N and Mn3.8Cu0.2N, Mn3.8Ag0.2N, Mn3.8Au0.2N. XAS and XMCD spectra at the Mn-L2,3 edges measured at 0 T and 5 T for (a)(a’) Mn3.8Cu0.2N, (b)(b’) Mn3.8Ag0.2N, and (c)(c’) Mn3.8Au0.2N. The spectrum at 0 T is obtained by first applying an external magnetic field up to 5 T, then lowering the field to 0 T, and finally performing the measurement. Differences in the XMCD spectra at the Mn-L2,3 edges for (a) Mn4N, (b) Mn3.8Cu0.2N, (c) Mn3.8Ag0.2N, and (d) Mn3.8Au0.2N epitaxial films, measured at 0 T and 5 T. The spectrum at 0 T is obtained by first applying an external magnetic field up to 5 T, then lowering the field to 0 T, and finally performing the measurement. Discussion on substitution sites Regarding the substitution sites of impurities within the Mn4N film, XANES measurements were performed for Pd and Ag [30,31], and based on these results, it has been determined that the impurities occupy corner Mn(I) sites. Here, we present the results for Pd in Mn4–ₓPdₓN epitaxial films. Figure 15(a) shows the measured XANES spectra at the Pd L3 absorption edge for Mn4–ₓPdₓN (x = 0.10, 0.20, and 0.8). Even when the composition of Pd is changed, no change is observed in the peak positions of the spectrum [31]. This result suggests that the Pd substitutes for Mn at the same site in the three Pd compositions. Figure 15.(a) XANES spectra of the Pd-L3 edge measured on Mn4−xPdxN epitaxial films at x = 0.10 (orange), 0.20 (green), and 0.8 (purple). (b), (c), and (d) simulation and measurement results for the Pd L3-edge XANES spectra at x = 0.10, 0.20, and 0.8, respectively. The black lines are the measured spectra by the TFY method at RT [31]. The red and blue lines correspond to the simulation results for clusters, where the Pd atom are located at the face-centered and corner sites of the antiperovskite structures, respectively. Reproduced by permission from [31], copyright [2025, American Physical Society].The image A showing a line graph labeled Measured at Pd L subscript 3. The x-axis is labeled Relative photon energy (electron volt), from minus 20 to 40. The y-axis is labeled XANES (arbitrary units), from 0.0 to 1.6. Three measured spectra are shown for x equals 0.10, x equals 0.20 and x equals 0.8. All three rise from near 0 at minus 20 to a sharp peak near 0 electron volt at about 1.4 to 1.5, drop to about 0.9 near 3 electron volt, form a broad peak near 10 electron volt at about 1.2 to 1.3, then approach about 1.0 by 40 electron volt. The image B showing a line graph labeled x equals 0.10 at Pd L subscript 3. The x-axis is labeled Relative photon energy (electron volt), from minus 20 to 40. The y-axis is labeled XANES (arbitrary units), from 0.0 to 1.6. Three curves are labeled Face-centered, Corner and Measured. All rise steeply to a main peak near 0 electron volt. Face-centered peaks highest at about 1.6 near 2 electron volt, then drops to about 0.8 by about 8 electron volt. Corner peaks near 1.4 around 0 to 1 electron volt, dips to about 0.7 near 4 electron volt, then peaks again near 1.2 around 12 electron volt. Measured peaks near 1.3 around 1 electron volt, then shows a broad hump near 12 to 15 electron volt around 1.2 and approaches about 1.0 by 40 electron volt. The image C showing a line graph labeled x equals 0.20 at Pd L subscript 3. The x-axis is labeled Relative photon energy (electron volt), from minus 20 to 40. The y-axis is labeled XANES (arbitrary units), from 0.0 to 1.6. Curves are Face-centered, Corner and Measured. Face-centered shows two close peaks near 0 to 3 electron volt reaching about 1.5 to 1.6, then falls to about 0.8 by about 8 electron volt. Corner peaks near 1.4 around 0 to 1 electron volt, dips near 4 electron volt to about 0.7, then rises to about 1.2 near 12 electron volt. Measured peaks near 1.3 around 1 electron volt, then forms a broad peak near 12 to 15 electron volt around 1.2, trending toward about 1.0 by 40 electron volt. The image D showing a line graph labeled x equals 0.8 at Pd L subscript 3. The x-axis is labeled Relative photon energy (electron volt), from minus 20 to 40. The y-axis is labeled XANES (arbitrary units), from 0.0 to 1.6. Curves are Face-centered, Corner and Measured. All rise to a main peak near 0 electron volt. Corner reaches about 1.6 near 0 to 1 electron volt, then dips to about 0.7 near 4 electron volt and rises to about 1.2 near 12 electron volt. Face-centered peaks near 1.4 around 1 electron volt, then drops to about 0.8 by about 8 electron volt. Measured peaks near 1.3 around 1 electron volt, shows a broad hump near 12 to 15 electron volt around 1.2 and approaches about 1.0 by 40 electron volt.Four line graphs of XANES spectra versus relative photon energy, comparing measured and simulated curves. (a) XANES spectra of the Pd-L3 edge measured on Mn4−xPdxN epitaxial films at x = 0.10 (orange), 0.20 (green), and 0.8 (purple). (b), (c), and (d) simulation and measurement results for the Pd L3-edge XANES spectra at x = 0.10, 0.20, and 0.8, respectively. The black lines are the measured spectra by the TFY method at RT [31]. The red and blue lines correspond to the simulation results for clusters, where the Pd atom are located at the face-centered and corner sites of the antiperovskite structures, respectively. Reproduced by permission from [31], copyright [2025, American Physical Society]. To compare these XANES spectra with the calculation results, XANES spectra were calculated for Mn4−xPdxN using the FDMNES package. The calculations were performed for two structures: Mn4N-type cluster and Mn3PdN-type cluster. Both clusters included a center Pd as an absorbing atom. The difference between the two types is whether the central Pd atom is surrounded by a Mn4N crystal or a Mn3PdN crystal. Furthermore, we considered two substitutional sites for Pd: the corner and face-centered sites in the anti-perovskite structure in each case. To compare with experimental spectra, we adopted the spectra of the Mn4N-type cluster for a low-composition case (x = 0.1 and 0.2) and that of the Mn3PdN-type cluster for a high-composition case (x = 0.8). Prior to the comparison, we convoluted the simulated spectra using a Lorentzian function with an energy-dependent scale parameter, Γtot. This can be expressed by the following equation.(4)ΓtotE=Γhole+ΓelectronE=Γhole+Γmax12+1πarctanπ3ΓmaxELarge−1e2.(5)e=E−EFermiEcent. Here, E is the photon energy, Γhole is the core-hole width, ΓelectronE is the photo-electron width, Γmax is the maximum width of the final state, ELarg is the width of the arctan function, Ecent is the center position of the arctan function, and EFermi is the Fermi energy. Three parameters, EFermi, Γhole, and Γmax, were optimized using an FDMNES code so that the experimental spectra matched the calculated spectra for the cluster with a Pd atom that substituted Mn at corner sites. In contrast, we also tried to fit the calculated spectra for face-centered Pd, the fitting code did not settle. Therefore, the same parameters were applied to the convolution. Table 2 summarizes the optimized parameters for the convolution. The optimization was conducted with the default parameters of Ecent= 30 eV, and ELarg= 30 eV.Table 2.Optimized parameters for the convolution of simulated spectra for FDMNES.Pd composition xEFermi (eV)Γhole (eV)Γmax (eV)0.10.175502.1787515.422210.20.026062.3348912.982110.80.050172.6033612.99420 Table 2.Optimized parameters for the convolution of simulated spectra for FDMNES.Pd composition xEFermi (eV)Γhole (eV)Γmax (eV)0.10.175502.1787515.422210.20.026062.3348912.982110.80.050172.6033612.99420 Optimized parameters for the convolution of simulated spectra for FDMNES. Figure 15(b–d) shows a comparison of the experimental results at x = 0.10, 0.20, and 0.8 with the XANES spectra predicted under the assumption that Pd has substituted for the corner Mn(I) sites or the face-centered Mn(II) sites. The convolution parameters of the calculated spectra were optimized with reference to the experimental data. The energy on the horizontal axis is normalized to 3,173 eV, which is the Pd L3 absorption edge. When Pd substitutes for a face-centered Mn(II) site, the simulated spectrum exhibits split peaks in the energy range from 0 to 7 eV. On the other hand, when Pd substitutes for a corner Mn(I) site, only a single peak is observed in the same energy range. Furthermore, in the energy region above 7 eV, the simulation corresponding to Pd substitution at the corner sites shows a high degree of agreement with the experimental data for the three simulated Pd compositions. The agreement between the experimental and simulated spectra was evaluated using the D1 factor, a measure commonly used to assess spectral similarity. A smaller D1 coefficient indicates a higher degree of agreement between the experimental and simulated spectra. The D1 factor is given by Eqs. (6) and (7):(6)D1=12∫1csimIsimE−1cexpIexpEdE.(7)csim=∫EminEmaxIsimEdE,cexp=∫EminEmaxIexpEdE. Here, csim and cexp represent the integral of the simulation and experimental spectra, respectively, over the energy range between the minimum energy (Emin) and maximum energy (Emax) under consideration. IsimE and IexpE represent the simulated and experimental spectra, respectively, at a given energy E. For the three Pd compositions examined, the D1 values obtained assuming that Pd substituted for corner Mn(I) sites were smaller than those obtained assuming that Pd substituted for face-centered Mn(II) sites as shown in Table 3. We therefore conclude that the Pd atoms preferentially substitute for Mn(I) sites in Mn4−xPdxN epitaxial films in the range of x = 0.10–0.8. Similar results were obtained for Ga and Zn doped in Mn4N epitaxial films. In impurity-doped Mn4N epitaxial films, the impurities that cause MC at RT are currently Ni, Co, Cu, Ag, Au, and Pd, while Zn and Ga are under investigation, it can be stated that they preferentially substitute for the corner Mn(I) sites. A common feature of these impurity elements is that MC occurs at RT at doping levels of approximately 5 at%. Figure 16 shows the longitudinal electrical resistivity (ρxx) of Mn4N epitaxial films doped with Cr, Co, Ni, Pd, Cu, Ag, Au, In, Ge, and Sn, measured at RT. Regardless of the type of impurity, as doped impurities are increased, the electrical resistivity gradually increases compared to that of Mn4N, and then reaches a plateau. Looking ahead to device applications, it is not possible to determine which impurity element is preferable based on Mn4N layers alone. When using SOT, for example, it is necessary to form a steep heterostructure with heavy metals (Pt, W). In such a structure, it is essential to identify a fabrication process that suppresses the diffusion of constituent elements, including impurity elements. Figure 16.Longitudinal resistivity of impurity-doped Mn4N epitaxial films.The scatter plot displays resistivity (rho_xx in micro ohm cm) from 0 to 300 on the vertical axis and Mn_4-xA_xN composition (x) from 0.0 to 1.5 on the horizontal axis. Ticks are marked at intervals on both axes. The legend titled 'Group' lists elements by period: 4th (Cr, Co, Ni, Cu, Ge), 5th (Pd, Ag, In, Sn) and 6th (Au). A shape key indicates 'Closed' symbols on SrTiO_3(001) and 'Open' on MgO(001). Data points range from x=0.0 to 1.5 and resistivity from 160 to 295. Near x=0.0-0.2, points cluster between 165-225. From x=0.2-0.6, series rise to 240-290, peaking at 275-295. Between x=0.6-1.1, points spread from 190-270, with concentrations at 230-255 near x=1.0. From x=1.1-1.5, points lie between 185-200, especially near x=1.3-1.5 at 185-190.A scatter plot showing resistivity for Mn subscript 4 minus x A subscript x N composition x from 0.0 to 1.5.Table 3.D1 factors calculated from experimental and simulated spectra.Nominal Pd composition (x)D1 for corner simulation spectraD1 for face-centered simulation spectra0.103.2811515.6229760.202.7261355.3149040.82.8323844.792633 Longitudinal resistivity of impurity-doped Mn4N epitaxial films. Table 3.D1 factors calculated from experimental and simulated spectra.Nominal Pd composition (x)D1 for corner simulation spectraD1 for face-centered simulation spectra0.103.2811515.6229760.202.7261355.3149040.82.8323844.792633 D1 factors calculated from experimental and simulated spectra. Toward Mn4N on <100>-oriented Pt Starting with a Ni-doped Mn4N film [26], we have grown Mn4N epitaxial thin films doped with various impurity elements and investigated whether they exhibit MC at RT. When applying Mn4N to devices as a next step, we aim to utilize the SOT that arises from a heterojunction with a heavy metal, and realize the layered structure of Mn4N/Pt/MgO (001). The reason for using Pt as an underlayer rather than a capping layer is to enable the formation of device structures, such as magnetic tunnel junctions (MTJ), on top of the Mn4N film. We also selected MgO rather than STO since MgO is commonly used for MTJs. The lattice constant of face-centered cubic Pt is 0.392 nm, which is close to that of Mn4N (0.387 nm). Therefore, we believe that if a Pt film with a <100> orientation can be formed on MgO, a Mn4N film can be epitaxially grown on top of it. When Pt is selected as the heavy metal [67–70], the key step is to form a <100>-oriented ultrathin Pt epitaxial film on MgO with a (001) surface. <100>-oriented Pt films are preferred compared to <111>-oriented films because they exhibit higher SOT efficiency via SHE [71]. The thickness of heavy-metal thin films used as spin sources is typically less than 10 nm. This is because the spin current generated by the SHE originates primarily from the region within the spin diffusion length (typically 10 nm or less) near the interface [72–76]. When the film is thicker than this length scale, the charge current flowing far from the interface contributes little to the generation of spin current; therefore, as the film thickness increases, the spin torque efficiency per unit current decreases [77,78]. However, reports of ⟨100⟩-oriented epitaxial Pt films that satisfy both continuity and smoothness have been limited to films thicker than 10 nm. This is because Pt films with (111) surfaces are the most stable [79–81]. We have addressed these challenges using a three-step growth method and have recently demonstrated the epitaxial growth of <100>-oriented Pt layers with a thickness of approximately 4 nm on MgO (001) substrates by radio-frequency sputtering [82]. Figure 17 shows a diagram of the three-step growth process of the Pt layers. The deposition rate of Pt was 0.23 nm min−1. Upon post-annealing, samples were heated to the target temperature at a rate of approximately 10°C min−1, then held at that temperature for 30 min under high vacuum. Samples were then cooled down to RT in the vacuum chamber. In the first step, Pt crystals oriented in the <100> direction were formed for 1.5 min at a substrate temperature of T1st = 870°C, followed by the Pt deposition for 13.5 min at a substrate temperature of T2nd = 250°C in the second step. In the third step, post annealing was performed at a substrate temperature of T3rd = 450°C. The Pt deposited in the second step inherited the crystal orientation and roughness from the seed crystals formed in the first step as shown in Figure 18(a,b). Normally, when Pt is deposited on an MgO (001) substrate at 250°C, <111>-oriented Pt is formed [82]; however, it is important to note that because <100>-oriented Pt seed crystals are formed in the first step, even when Pt is deposited at 250°C in the second step, the resulting structure retains the <100> orientation. After the third step, RHEED streaks appeared, indicating the improved surface flatness. The 1/5-period RHEED streaks taken along the [100] direction originate from a Pt (001) − (5 × n) surface reconstruction [83–85]. AFM images in Figure 18(c) show that the surface roughness after the third step was significantly smaller and the formation of the <100>-oriented ultrathin continuous Pt films was confirmed. The longitudinal electrical resistance of the Pt layer was 24 µΩ cm at RT. The development of such ultra-thin Pt films described in this paper will facilitate the realization of highly energy-efficient Mn4N-based SOT devices on MgO with a (001) surface. Figure 17.Schematic diagram of the 3-step growth process for high-quality <100>-oriented Pt films. The first step is the formation of seed crystals at T1st = 870°C to establish <100> orientation, the second step is the Pt deposition at T2nd = 250°C to ensure film continuity, and the third step is post-annealing at T3rd = 450°C to improve surface flatness.A diagram illustrating a three-step growth process for platinum films. The top graph shows temperature on the vertical axis and time on the horizontal axis. It depicts three steps: the first step at temperature T subscript 1st, the second step at T subscript 2nd and the third step at T subscript 3rd. Below the graph, three stages are shown: the first step is 'Growth of seed' with Pt seed on MgO(001) at T subscript 1st equals 870 degrees Celsius, the second step is 'Low-temp. deposition' with Epitaxial Pt on MgO(001) at T subscript 2nd equals 250 degrees Celsius and the third step is 'Post annealing' with Epitaxial Pt on MgO(001) at T subscript 3rd equals 450 degrees Celsius.Diagram of 3-step Pt film growth process with temperature and time. Figure 18.(a) Out-of-plane XRD patterns for Pt formed by the three-step growth process. The patterns are measured after the first step (growth of seed), the second step (low-temperature deposition), and the third step (post annealing). (b) RHEED patterns taken along the [100] azimuth. (c) AFM images of samples after each step.Image A displays four stacked line plots of out-of-plane X-ray diffraction intensity. The x-axis is labeled Cu minus K alpha 1 2 theta (deg), ranging from 20 to 120. The y-axis is Out-of-plane XRD intensity (arb. unit). Top plot: After 1st step; peaks near 40 and 95, smaller feature near 45 labeled Pt 002. Second plot: After 2nd step; peaks near 40 and 95, feature near 105 labeled Pt 004. Third plot: After 3rd step; peaks near 40 and 95, feature near 105. Bottom plot: MgO(001) sub.; peaks labeled MgO 002 near 40 and MgO 004 near 95. Image B shows three stacked RHEED images labeled After 1st, 2nd and 3rd steps. Top image: spotty diffraction; middle: mixed spots and streaks; bottom: stronger streaks. Image C shows three atomic force microscopy height maps labeled After 1st, 2nd and 3rd steps. Height scale: 0.00 nanometer at bottom; maxima are 0.95 nm (first), 1.06 nm (second), 2.08 nm (third). Scale bar under first map: 0.5 micrometer.Three plots show platinum film growth: sharper diffraction, smoother topography. Schematic diagram of the 3-step growth process for high-quality <100>-oriented Pt films. The first step is the formation of seed crystals at T1st = 870°C to establish <100> orientation, the second step is the Pt deposition at T2nd = 250°C to ensure film continuity, and the third step is post-annealing at T3rd = 450°C to improve surface flatness. (a) Out-of-plane XRD patterns for Pt formed by the three-step growth process. The patterns are measured after the first step (growth of seed), the second step (low-temperature deposition), and the third step (post annealing). (b) RHEED patterns taken along the [100] azimuth. (c) AFM images of samples after each step. Conclusions This paper summarizes experimental results regarding the potential of Mn4N thin films – ferrimagnetic materials containing no rare earth elements – with a particular focus on the presence or absence of MC at RT due to impurity doping. The presence of MCs at RT has been confirmed in Mn4N epitaxial films doped with Ni, Co, Cu, Ag, Au, and Pd. In Ni-doped Mn4N near the MC composition, current-induced domain wall motion driven solely by STT is realized at speeds exceeding 3000 m/s at RT, suggesting that the introduction of impurity elements that surpass Ni is expected. To further advance applications based on STT and SOT using Mn4N, it is necessary to study several physical properties in greater detail, such as spin current polarization and decay parameters. Furthermore, to utilize and enhance spin-orbit effects such as SOT and the Dzyaloshinskii Moriya interaction, it is essential to deposit ultrathin films of extremely high-quality Mn4N-based ferrimagnetic/heavy-metal structures. We have also demonstrated the formation of ultra-thin <100>-oriented Pt epitaxial film (approximately 4 nm in thickness) that serves as the foundation for such structures. This material development will help demonstrate the potential of Mn4N, extending beyond DW manipulation via pure STT to applications ranging from SOT to skyrmions and THz switching. We therefore hope that many researchers will take an interest in utilizing this material.