Control of spin orientation in antiferromagnetic NiO by epitaxial strain and spin–flop coupling

Antiferromagnets (AFMs) have recently attracted attention due to their potential application as the active element for spintronic devices.^1,2 A unique set of properties such as robustness to external magnetic fields, high packing density due to the lack of stray fields, and the potential terahertz operation speed³ makes AFMs serious candidates to replace ferromagnetic materials in future spintronic devices. Recent research concerning AFMs has identified an important issue in the ability to manipulate the direction of magnetic moments in AFM materials. In the case of ferromagnets (FMs), a moderate external magnetic field can be used to align the magnetic moments of ferromagnetic layers, but the magnetic field necessary to manipulate spin structures in AFMs is usually relatively large.¹ However, magnetic fields can be used to indirectly manipulate the spin directions in AFMs through interface coupling mechanisms in an FM/AFM bilayer. Coupling at FM/AFM interfaces can lead to either collinear^4–7 or non-collinear alignment of the magnetic moments of FMs and AFMs.^8,9 Besides interactions at the FM/AFM interface, manipulation of spin structures in AFMs can be realized using strain. Lattice distortion induced by strain can modify the magneto-crystalline anisotropy and reorient the direction of magnetic moments in the AFM layer.^10,11 Strain driven control of spin orientations was realized by growing AFM films on flexible,¹² piezoelectric,¹³ or lattice-mismatched substrates.¹⁴ Finally, electrical and optical control of AFMs has been recently proposed as effective methods to tune the magnetic properties of AFM materials.^1,2,15

Recent demonstrations of ultrafast magnetic dynamics,¹⁶ current-induced magnetic switching,¹⁷ and spin Hall magnetoresistance^18–22 in NiO have revealed the potential of this 3d metal oxide for antiferromagnet-based spintronics. NiO is an antiferromagnetic insulator that can be epitaxially grown on an MgO substrate. Bulk NiO crystallizes in a cubic NaCl structure. Below the Néel temperature (T_N = 523 K), the magnetic moments of Ni²⁺ ions in NiO align ferromagnetically within the {111} planes, while the adjacent {111} planes are coupled antiferromagnetically. In ultrathin layers, the NiO spin direction can be modified by strain induced by the substrate. Previous studies have shown that compressive strain imposes an in-plane NiO spin direction, while out-of-plane spin alignment is preferred for NiO grown under tensile strain.²³ In particular, an in-plane NiO spin direction was reported for layers grown on Ag(001) with a lattice constant smaller than the lattice parameter of NiO (a_Ag = 4.086 Å < a_NiO = 4.176 Å). An out-of-plane spin orientation was observed in NiO films grown on MgO(001)^24–27 with a larger lattice constant (a_NiO = 4.176 Å < a_MgO = 4.212 Å).

In this study, we used strain and interface coupling to manipulate the spin structure in NiO. First, x-ray magnetic linear dichroism (XMLD) was used to show that while an out-of-plane spin direction is observed for NiO layers grown on an MgO substrate in a Fe/NiO/MgO stack, the NiO spins are aligned in-plane when the layers are grown on a 20-nm Cr layer. Furthermore, a continuous strain-induced spin reorientation transition (SRT) from the out-of-plane to in-plane directions was observed when the wedge-shaped Cr buffer thickness increased from 0 nm to 3.5 nm. Finally, the analysis of in-plane magnetic structures for NiO in the Fe/NiO/Cr stack revealed a pronounced uniaxial anisotropy in the thin AFM layers with a spin direction perpendicular to that of Fe due to spin–flop interaction.

All layers were deposited under ultrahigh vacuum (UHV) conditions using molecular beam epitaxy on a polished MgO(001) single crystal. The MgO substrate was annealed at 775 K before deposition. A 6 nm homoepitaxial MgO buffer layer was evaporated at 725 K using electron beam evaporation (EBV). Next, half of the substrate was coated with a 20 nm Cr buffer layer deposited at 475 K and annealed at 725 K to improve its surface quality. The low-energy electron diffraction (LEED) pattern collected after annealing the Cr confirmed the epitaxial growth of the buffer layer with Cr[110]||MgO[100]. A wedge-shaped NiO with a thickness (d_NiO) between 5 Å and 40 Å over 10 mm was grown by reactive deposition of metallic Ni under an oxygen pressure of 1 × 10⁻⁶ mbar. The NiO wedge was simultaneously deposited on both the un-buffered and Cr-buffered regions of the substrate. Following NiO deposition, a 1 nm Fe layer was grown and capped with a protective MgO layer.

X-ray absorption spectroscopy (XAS) measurements were performed at the XAS/photoemission electron microscopy (PEEM) beamline of the National Synchrotron Radiation Center SOLARIS.²⁸ The XAS spectra were collected under the total-electron-yield detection mode by measuring the sample current. To probe the NiO spin orientation, XAS spectra were collected using a linearly polarized x-ray beam with a photon energy corresponding to the Ni L_2,3 edges. The energy resolution was set to be better than 0.2 eV. The measurements were performed as a function of the incidence angle γ, which is defined as the angle between the propagation direction of the x rays and the sample surface normal [Fig. 1(a)]. With an increased γ, the footprint of the beam on the sample was limited by the slits used to illuminate the same sample area. The spin orientation of the Fe sublayer was studied using the x-ray magnetic circular dichroism (XMCD) technique by collecting the XAS spectra for the photon energy tuned to the Fe L_2,3 edge for two circular polarizations with opposite helicities. The magnetic hysteresis curves were collected with the longitudinal magneto-optic Kerr effect (LMOKE). A standard lock-in detection setup was used that consisted of an s-polarized laser source (λ = 635 nm) and a photo-elastic modulator with a modulation frequency of 50 kHz. The second harmonic (2f) signal measured by using the detector was proportional to the Kerr rotation and was taken as a measure of the magnetization.

The exemplary XAS spectra for the Ni²⁺ L₂ edge with a NiO thickness of 21 Å in the Fe/NiO/MgO (black curve) and Fe/NiO/Cr (blue curve) are shown in Figs. 1(b) and 1(c). The spectra were collected under normal (γ = 0°—dashed lines) and grazing (γ = 70°—solid lines) incident x-ray angles at 80 K and normalized using the intensity of the lower energy peak. For Fe/NiO/MgO, there was a greater intensity for the higher energy peak under grazing incidence [Fig. 1(b)], while the spectrum collected for Fe/NiO/Cr showed the opposite behavior [Fig. 1(c)]. Figure 1(d) shows the L₂ ratio, which is defined as the ratio of the intensities of the lower to the higher energy peaks as a function of the x-ray incident angle. R_L2 follows the theoretic angular dependence of the XAS spectrum, i.e., R(γ) = A cos² γ + B. In agreement with previous studies at γ = 0°, the maximum of the L₂ ratio for Fe/NiO/MgO [Fig. 1(d), black] proves that the NiO spins possessed a considerable out-of-plane component.^25,27 The minimum of R_L2 observed for the Fe/NiO/Cr [Fig. 1(d), blue] indicates an in-plane NiO spin direction. Analogous results with comparable amplitudes of R_L2 were obtained at 300 K (not shown), which suggests that the Néel temperature of the NiO layer with d_NiO = 21 Å is well above room temperature (RT).

Dichroism effects in x-ray linear dichroism (XLD) spectroscopy can be influenced not only by the exchange fields and magnetic moments but also by the local crystal field symmetry.²⁹ To determine whether the observed XLD effects are related purely to the magnetic order within the AFM layers, the R_L2 ratios were determined for γ = 0° and γ = 70° at different NiO thicknesses at 300 K and 390 K. As shown in Fig. 2(a), the L₂ ratio difference defined as |ΔR_L2| = |R_L2(γ = 0°) − R_L2(γ = 70°)| decreased with a reduced NiO thickness for Fe/NiO/MgO and Fe/NiO/Cr. Unfortunately, the technical restrictions did not allow measuring the XAS spectra at T = 523 K as this is the Néel temperature of bulk NiO. However, there was a decreased XLD magnitude as the temperature increased from 300 K to 390 K. This indicates that the observed XLD effects are associated with magnetic properties and not with the temperature-independent crystal-field dichroism. Another indication that the XLD effects originate from the antiferromagnetic order in NiO is a lack of distinct shifts for the spectra collected at γ = 0° and γ = 70°. As previously shown, a fingerprint of the crystal-field contribution to the XAS spectrum has a strong polarization dependence for the spectra accompanied by energy shifts for the L₃ line.²⁹ Figure 2(b) shows the γ angle dependence on the Ni L₃ spectra as collected for Fe/NiO/MgO (black) and Fe/NiO/Cr (blue). The spectra were measured for a thin NiO layer (d_NiO = 12 Å) for which any existing crystal-field dichroism effect should be strong. For Fe/NiO/Cr, the energy positions of the L₃ peak for γ = 0° and γ = 70° were the same. An energy shift of 0.1 eV was seen for Fe/NiO/MgO, which is within the resolution limit. These results indicate that the crystal-field effects do not significantly contribute to XLD in the considered samples.

The XMLD measurements reveal a change in the NiO spin orientation from out-of-plane in Fe/NiO/MgO to in-plane in Fe/NiO/Cr. A dedicated sample was prepared to identify the origin of the SRT. An Fe(10 Å)/NiO(21 Å) bilayer was grown on a wedge-shaped Cr buffer with a thickness that continuously changed in the range of 0 nm–8 nm at 1 nm/mm. After Cr deposition, LEED patterns were collected systematically across the wedge. Figure 3 shows the in-plane lattice spacing along the Cr[110] (MgO[100]) direction as determined from the LEED patterns as a function of Cr thickness d_Cr. For d_Cr < 1.5 nm, a_Cr[110] = (4.2 ± 0.04) Å, which proves pseudomorphic growth for the ultrathin Cr on MgO. For d_Cr > 1.5 nm, a relaxation of the in-plane Cr lattice constant occurred up to d_Cr = 3 nm for which the lattice parameter reached a bulk value of 4.07 Å. Changes in the Cr lattice parameter were accompanied by the evolution of the spin direction in the NiO layer. The ΔR_L2 dependence on d_Cr reveals a monotonic decrease up to a d_Cr of 3.5 nm (Fig. 3, black curve). For Cr layers thicker than 3.5 nm, for which we noted the relaxed Cr surface in the LEED ΔR_L2 remained nearly constant. These results suggest that the SRT in the system was induced by epitaxial strain exerted on the NiO by the underlying Cr layer. For ultrathin buffer layers for which pseudomorphic growth of Cr was observed, NiO experienced tensile stress (a_Cr = 4.2 Å > a_NiO = 4.176 Å) and its spins were aligned out-of-plane, which is similar to the NiO grown on MgO. Together with the increased Cr thickness, a relaxation in the lattice parameter occurred, and NiO became compressed (a_Cr = 4.07 Å < a_NiO = 4.176 Å). Therefore, the AFM spins rotate toward the in-plane direction. Moreover, the results show that the continuous nature of the reported transition allows a proper choice of Cr thickness to precisely tune the out-of-plane component of the antiferromagnetic NiO spins.

To determine the in-plane orientation of NiO spins in Fe/NiO/(Cr)MgO, XMLD measurements were collected as functions of the azimuthal angle φ, while the γ angle was set to 45° [Fig. 4(a)]. The external magnetic field was applied along the easy Fe[100] (NiO[110]) direction before the XMLD measurements. The XAS spectra recorded at the Fe L_2,3 edge for two circular polarizations confirmed the strong polarization dependence of the spectra for φ = 0° (Fe[100], NiO[110]), while there was no XMCD for φ = 90° [Figs. 4(b) and 4(c)]. Figure 4(d) shows R_L2(φ) expressed in polar coordinates. For Fe/NiO/Cr, the measurements were performed for two representative NiO thicknesses: d_NiO = 14 Å (a relatively thin AFM layer with a pronounced XLD signal at 300 K) and the thickest available NiO of d_NiO = 37 Å.

Based on previous studies, if the R_L2 reaches an extremum when the polarization vector is parallel to the [110] NiO direction, the AFM spin alignment is determined from the maximum of the L₂ ratio.^30,31 For Fe/NiO(14 Å)/Cr, well-pronounced extrema in the R_L2(φ) dependence were observed for the NiO[110] and [1–10] directions. There was a maximum for the x-ray polarization parallel to the [1–10] NiO crystal axis (φ = 90°) [Fig. 4(d), black curve]. Hence, Ni²⁺ spins in Fe/NiO(14 Å)/Cr are aligned parallel to the NiO[1–10] axis and perpendicular to the Fe spins. After switching the magnetization of Fe by 90°, the NiO spins rotated by 90° (not shown). The in-plane anisotropy of NiO weakened with an increase in NiO thickness. For Fe/NiO(37 Å)/Cr, there was nearly no change in the R_L2(φ) dependence [Fig. 4(d), blue curve].

For Fe/NiO/MgO, there was a pronounced in-plane anisotropy [Fig. 4(d), red curve], which indicates that the NiO spins are not aligned perfectly perpendicular to the surface but possess a small in-plane component.

The orthogonal (spin–flop) coupling between Fe and NiO was previously reported for Fe(12 Å)/NiO(21 Å–62 Å)/Ag(001).²⁷ Kim et al. showed that spin–flop coupling is accompanied with an enhanced magnetic coercivity (H_C) in the hysteresis loops collected for Fe/NiO/Ag(001). To elucidate how the orthogonal FM–AFM coupling affects the FM magnetization reversal, LMOKE was performed as a function of the NiO thickness for Fe/NiO/Cr. The LMOKE hysteresis loops were recorded at 80 K after the sample was cooled in an external magnetic field of 5 kOe applied along the easy Fe[100] direction. The loops collected for Fe/NiO(d_NiO = 20 Å)/Cr [Fig. 5(a), blue curve] exhibit more than twice the coercivity than the H_C for the loop registered for the un-buffered sample [Fig. 5(a), black curve]. This confirms the enhanced coercivity for NiO layers with more pronounced changes in the R_L2(φ) dependence, as seen in Fig. 4(d). As the NiO spins in Fe/NiO/MgO possess strong out-of-plane components, their influence on the in-plane magnetic anisotropy of the Fe layer is much weaker than for Fe/NiO/Cr, which has an in-plane NiO spin direction.

Based on theoretical studies for perfectly flat FM/AFM interfaces, spin–flop coupling contributes to a uniaxial anisotropy and should not lead to exchange bias, which manifests in a horizontal shift in the hysteresis loop.³² Figure 5(b) presents the exchange bias field (H_EB) dependence on the NiO thickness for Fe/NiO/Cr as determined from the LMOKE hysteresis loops collected using an external magnetic field applied along the easy Fe direction. When below d_NiO = 30 Å, there was a negligible H_EB; however, there was a gradual increase in H_EB for d_NiO > 30 Å, which reached H_EB = 175 Oe at d_NiO = 40 Å. The absence of an exchange bias for NiO layers thinner than 30 Å is understood when considering the XMLD results. The existence of spin–flop coupling between Fe and NiO [Fig. 4(d), black curve] was proven for ultrathin NiO. A consequence of the orthogonal FM/AFM coupling is a lack of exchange bias noted for thin NiO films grown on Cr. When the NiO thickness increases, the in-plane anisotropy of the AFM is lost [Fig. 4(d), blue curve], which is accompanied by the appearance of H_EB for d_NiO > 30 Å [Fig. 5(b)].

In summary, strain and interaction with the ferromagnetic layer were shown to modulate the NiO spin direction in Fe/NiO/MgO(Cr). First, comparative XMLD studies revealed the existence of a continuous spin reorientation transition from the nearly out-of-plane direction in Fe/NiO/MgO to the in-plane direction in Fe/NiO/Cr. Analysis of the XMLD spectra recorded for different azimuthal angles showed that spins for an ultrathin NiO in Fe/NiO/Cr are aligned perpendicularly to the Fe magnetization due to spin–flop interaction. Orthogonal coupling between NiO and Fe in Fe/NiO/Cr is reflected in the LMOKE measurements as an increase in the coercive field and a lack of an exchange bias, which occurred along with a loss of the in-plane magnetic anisotropy for thicker NiO films in Fe/NiO/Cr. The results demonstrate the feasibility of using strain and coupling with ferromagnets to manipulate the spin structure of NiO.

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

This work was supported by the “Antiferromagnetic proximity effect and development of epitaxial bimetallic antiferromagnets – two routes towards next-generation spintronics” project, which is carried out within the Homing programme of the Foundation for Polish Science co-financed by the European Union under the European Regional Development Fund.