Interface Fe magnetic moment enhancement in MgO/Fe/MgO trilayers

In spintronic applications, magnetic tunnel junctions (MTJ) play an important role as read heads for ultra-high density hard disk drives¹ and for magnetic random access memory implementations.² MTJs containing Fe/MgO interfaces are promising candidates because of their high tunnelling magnetoresistance (TMR).³ In such devices, the interfaces play an important role for the electric properties.^4–7 For example, Negusse et al. have shown that improving the interfacial moment behavior by using Fe alloys is a key to further increase the TMR.⁸ The magnitude of TMR also depends on the possible oxidation of the Fe monolayer (ML) next to the MgO,^4,6,9–12 as well as on the presence of a perpendicular magnetic anisotropy for ultrathin Fe films.^7,9,13–16 Particularly, it has been shown that a high TMR can be achieved for interfaces without any Fe oxidation close to the MgO layers.^17,18 For such Fe/MgO interfaces, the hybridization between Fe and MgO has been theoretically predicted to result in a significant enhancement of the interfacial Fe magnetic moment similar to the one expected for freestanding Fe monolayers.^19,20 Experimentally these predictions have been corroborated by probing the thickness dependent total magnetic moment of 1–6 ML thick Fe films on MgO yielding indirect evidence for interfacial moment enhancement in actual device structures^16,21,22 or other substrates with similarly strong electronic hybridization.²³ 

There has been no experimental study probing the depth dependence of the Fe magnetic moment within a single MgO/Fe/MgO heterostructure. Consequently, it is unknown if the interface moment enhancement persists for this case. Some depth resolved information has been derived from Mössbauer spectroscopy and nuclear resonant scattering experiments on MgO/Fe/MgO trilayer samples with a Fe thickness below 10 Å (7 Fe ML) at low temperature.^13,14 Although those studies show the necessity of separately modelling interface and bulk Fe magnetic moments, they mainly provide information about the magnetic anisotropy and not about the size of the magnetic moments.^13,14

In this letter, we experimentally determine the room-temperature interfacial magnetic moment enhancement of a MgO/Fe(12 nm)/MgO(001) trilayer system by using soft x-ray resonant magnetic reflectivity (SXRMR). We find an increase of 24 ± 19% of the Fe magnetic moment within 3.8 Å (corresponding to 2.7 ML) of both MgO interfaces. We show that the magnetic moment in the center of the Fe layer is 2.1μ_B, the bulk value, whereas the interface moment is equal to 2.6 ± 0.4μ_B.

The Fe layer was deposited on a MgO substrate which was sequentially cleaned with acetone and methanol using ultrasonic bath for 10 min in each solvent and then annealed at 500 °C for 1 h at pressure <10⁻⁹Torr. The MgO substrate was next cooled to 200 °C and exposed to atomic oxygen for 10 min. The 15 nm Fe film was deposited at a substrate temperature of 50 °C from an e-beam source and then annealed at 350 °C for 1 h. In situ RHEED reveals the Fe film to be single crystalline, epitaxial, and with a smooth surface. The Fe layer was then capped with a 3 nm MgO layer deposited by reactive magnetron sputtering to prevent its oxidation during exposure to air. The samples were characterized chemically and magnetically using x-ray absorption spectroscopy (XAS) and x-ray magnetic circular dichroism (XMCD), respectively, at beamline 4.0.2 of the Advanced Light Source in Berkeley. X-ray absorption at the Fe L_3,2 absorption edges was detected by monitoring the total electron yield as function of the photon energy for opposite x-ray helicities. Sums and differences of the two curves provide XAS and XMCD spectra, respectively, which are also the imaginary parts of the optical constants used as reference for the SXRMR analysis below. XAS and XMCD spectra (Fig. 1) were found to be identical, after normalization, to the ones reported by Chen et al.²⁴ They also demonstrate the absence of any form of Fe oxidation.

SXRMR experiments were performed at beamline 10ID-2 of the Canadian Light Source. Measured scattering intensities depend on the depth profile of charge and magnetic contributions to the atomic scattering factors, including their x-ray energy and polarization dependence. Modeling SXRMR measurements allow us to determine magnetic characteristics, i.e., orientation and magnitude of the magnetic moments, with element, site, and depth sensitivity.^25,26 In this study, we measured angle dependent specular reflectivity curves, I_p and I_n, for opposite circular x-ray polarization with the sample in a saturating 1 T longitudinal field produced by a permanent magnet (Fig. 2). In this experimental configuration, SXRMR is sensitive to the in-plane and out-of-plane components of the magnetization within the scattering plane.²⁷ Because the XAS and XMCD provide no evidence that interfacial Fe has substantially different charge or magnetic spectra than the bulk Fe, our modeling assumes that all Fe is described by the same values of charge optical constant f_c,2 and we allow the magnetic optical constant f_m,2 to vary in magnitude to describe variations in interfacial magnetization.²⁵ This assumption is justified below.

Figure 2(a) shows the reflectivity at 680 eV far from the resonance L₃. The thickness fringes caused by the difference in optical properties for the MgO and Fe layers are clearly visible. These data are analyzed to determine the structural parameters of the system. Indeed, at 680 eV energy, there is a reduction of the amplitude of the real and imaginary correction terms of the resonant atomic scattering factor as well as a reduction of the sensitivity to the magnetic terms.²⁸ The structural parameters are derived from the best fit of the reflectivity (red line) using the Paratt formalism²⁹ and are shown in Table I.

The magnetic asymmetries (I_p – I_n)/(I_p + I_n) are displayed in Figs. 2(b)–2(d) for 3 different energies: two near the L₃ edge ((b) and (c)) and one near the L₂ edge (d). For all three photon energies, the asymmetry shows a strong dependence on the scattering angle and values of up to 80%. For the different photon energies, the scattering asymmetries differ in sign and details of the fine structure. Modeling the asymmetry allows us to determine the magnetic depth profile of Fe thin films with high accuracy^30,31 and is utilized to determine the magnetic moments of the MgO/Fe/MgO system. To fit the magnetic asymmetry, the structural parameters are fixed, and only the magnetic parameters are adjusted. There are three free parameters for each magnetic layer: a scaling factor weighting the bulk value of the bcc Fe magnetic moment (2.1μ_B (Ref. 24)) and two angles $φ$ and γ defining the orientation of the magnetic moment. Specifically, $φ$ corresponds to the out-of-plane angle of the magnetization, and γ describes the in-plane angle of the magnetization (Fig. 2). To fit the experimental asymmetry, the magnetic layer is divided into several magnetic sub-layers allowing for different magnetic properties throughout the Fe layer.³² 

Using the structural parameters derived from the 680 eV energy, the magnetic asymmetry is modeled as described in Ref. 33. In our case, the longitudinal saturating field orients all magnetic moments in the scattering plane (Fig. 2) so the in-sample-plane angle γ is fixed and not considered as a free parameter. Only the scaling factor and the out-of-sample-plane angle $φ$ are refined in the following. If only one magnetic layer is considered, the overall shape of the fit is given by the blue curves in Fig. 2. The derived model gives a magnetic moment of 2.4 ± 0.2μ_B. There is an apparent disagreement with the experimental asymmetry that grows systematically with the incident angle θ or scattering vector $q=4π sin(θ)/λ$⁠, implying that one or more magnetic sub-layers exist with different magnetic properties. Given the importance and even expectation of interface effects^{13,14,30–32} we divided the Fe layer into 3 sub-layers with one bulk and two interface values of the magnetic moments. From this model, we refined the thickness of the interface sub-layers as well as the magnetic moments and the angles $φ$ for the 3 magnetic sub-layers. A simultaneous fit of the 3 resonant energies gives a very good agreement (red curves). The magnetic model derived from this fit has out-of-plane angle $φ=0$ for all sub-layers, whereas there is an increase of 24% of the magnetic moments for the interface sub-layers with a thickness of 0.38 nm, i.e., 2.7ML. The magnetic moment as well as the thickness of the magnetic sub-layers are summarized in Table I.

In Figure 2, even the best fits (red curves) are not perfect for θ angles exceeding 45°. This increasing discrepancy at the highest angles could result from a less-than-adequate model, or possibly could represent an experimental artifact from low signal levels. Adding more magnetic layers could improve the fit, although we were hesitant to increase the number of fit parameters. Since it is known that an out-of-plane (⁠$φ≠0°$⁠) interfacial moment increases the magnetic asymmetry at high angles,²⁷ we explored this further in the 3 layer model by carefully fitting the data from each energy separately. While non-zero $φ$ values are obtained for such separate fit, they vary in sign and size for the 3 energies considered, leading us to conclude that the simultaneous fit yielding $φ=0$ is the most consistent with the data. Furthermore, we note that at the highest θ angles the reflectivity signal is in the 10⁻⁷ range and even less at the interference minima. These values are near the dynamic range limit of the diode detector used, whose small dark current makes it increasingly difficult to accurately measure the reflectivity minima (Fig. 2(a) at θ angles of 48° and 62°). As the asymmetry is normalized to the total intensity at a particular angle, θ, the small intensity signals will artificially enhance small deviations in the signal, causing the asymmetry to appear larger than it actually is. This can lead to systematic artifacts in the asymmetry at the highest θ angles. Based on these considerations, we believe that the refinement using a 3 magnetic layer model to fit all data simultaneously provides the most reliable model for the magnetic structure of the Fe layer.

Our model results agree with findings in several previous studies. First, the absence of an out-of-plane magnetization component is expected for a Fe thickness of 12 nm.³⁴ Second, the increase of the magnetic moment in the 2–3 ML near both MgO interfaces confirms the theoretical predictions^19,20 and is in good agreement with experimental work published for samples of only a few Fe monolayer thicknesses.²¹ Indeed, theoretically, Li and Freeman demonstrated that an enhancement of 30% of the magnetic moment was expected for a 2ML Fe thick film on MgO.¹⁹ Experimentally, Miyokawa et al. showed by XMCD that the magnetic moment of 2 ML of Fe on MgO was 2.66 ± 0.1μ_B.²¹ Our model shows that the magnetic moments of 2.6 μ_B at both MgO/Fe and Fe/MgO interfaces are clearly enhanced outside the experimental uncertainty of 0.3 and 0.4 μ_B, respectively. Furthermore, owing to the very good depth resolution of the SRMR,³² our model is sensitive to the bottom interface buried at more than 12 nm. This reveals the symmetry of both interfaces which correlates favorably with the high TMR values observed in Fe/MgO based MTJs.³⁵ 

Using the magnetic depth profile discussed above, we calculated the energy dependence of the reflectivity and of the asymmetry for an angle θ = 15° (solid line in Fig. 3). The calculated and experimental data agree very well (Fig. 3), justifying our assumption that only the resonant magnetic optical properties vary appreciably near the interfaces. An explanation for the magnetic moment enhancement at MgO interfaces could be the change in layer spacing.

Our work emphasizes the need to model data from several energies systematically in the analysis of SXRMR data. The strength of a model derived from a fit comes from the ability to reproduce, with the same model, the experimental data for different experimental configurations. For our longitudinal configuration, using at least two resonant energies is the minimum to claim the uniqueness of the model. Data from the added energy near the L₃ inflection point further constrains the model. In particular, the sensitivity to the bottom interface layer is confirmed by the use of energy at inflection points. Furthermore, we find that it is important to model data collected at different energies simultaneously, as modeling different energy datasets individually led to inconsistent values for possible out-of-plane magnetic component.

In conclusion, this work further confirms the value of using SXRMR to measure the details of magnetic depth profiles across buried interfaces. We have revealed that even in a thick Fe layer sandwiched between MgO, the enhancement of the magnetic moment in the 2–3 ML near the interfaces persists, as predicted and shown for thin Fe layers. This inhomogeneity of the magnetic moment in depth is of great interest for the understanding of the high TMR effect measured in Fe/MgO tunnel junction and highlights the importance of the interface in such materials.

Work at SIMES was supported by the Department of Energy, Office of Science, Basic Energy Sciences, Materials Sciences and Engineering Division, under Contract No. DE-AC02-76SF00515. Work at the REIX beamline (10ID-2) at the Canadian Light Source was supported by the Canada Foundation for Innovation, Natural Sciences and Engineering Research Council of Canada, the University of Saskatchewan, the Government of Saskatchewan, Western Economic Diversification Canada, the National Research Council Canada, and the Canadian Institutes of Health Research. Work at the Advanced Light Source was supported by the Director, Office of Science, Office of Basic Energy Sciences of the U.S. Department of Energy under Contract No. DE-AC02-05CH11231.