Microscopic origin of the reduced magnetocrystalline anisotropy with increasing oxide content in Co₈₀Pt₂₀ : oxide thin films

A well-known dilemma in state-of-the-art composite CoPt : oxide thin films for perpendicular magnetic recording is that while the desirable intergrain magnetic decoupling is improved with increasing oxide volume fraction the magnetocrystalline anisotropy, K₁, is adversely reduced [1]. Our previous experimental study of CoPt : oxide thin films has also shown that the magnetic anisotropy of the magnetic grains (K_1g) itself drops when the oxide volume fraction is increased, causing the reduction in K₁ [2, 3]. Revealing the underlying physical mechanism of such a drop in K₁ is not only of great scientific importance, but could also provide technological benefits, in deciding the optimum oxide volume fraction in the manufacturing process for high-density storage applications [4, 5]. Several mechanisms have been proposed to explain the decrease in K₁—including the possible impact of intergrain exchange decoupling [6], surface anisotropy arising from the grain–oxide interface [7], and stacking faults within the magnetic grains [8–10]—while it remains ambiguous which one plays a more important role or whether they have contributed equally in the thin films.

Spin–orbit interaction is believed to play a dominant role in the magnetic anisotropy [11, 12]. Particularly in magnetic grains with dimensions of a few nanometers, the orbital contribution to the total magnetic moment is no longer negligible, which could influence the spin–orbit interaction and in turn has an impact on the magnetic anisotropy [13–15]. It is therefore important to study the orbital moment and the spin–orbit interaction in such granular CoPt : oxide thin films, in order to provide a clear picture of the underlying mechanism governing the above-mentioned decrease in the magnetocrystalline anisotropy. X-ray magnetic circular dichroism (XMCD) is an ideal tool for this purpose due to its unique capability of separating spin- and orbital moments. Combined with sum rules analysis [16], XMCD has been shown to be a powerful tool to study subtle changes in spin and orbital moments and magnetic anisotropy of Co particles [17–19].

In this work, we use angle-dependent XMCD to quantify the spin and orbital moments of Co, which represent the main source of magnetization in CoPt : oxide thin films [20]. By studying the dependence of the microscopic magnetic moments on the oxide volume fraction, we have observed an increase in orbital moment normal to the grain–oxide interface with increasing oxide volume fraction, caused by the symmetry breaking at the grain–oxide interface, which plays a dominant role in the K₁ drop.

A series of samples, glass/Ta(5 nm)/Ru(13 nm)/Co₈₀Pt₂₀+WO₃(total of 13 nm)/C(7 nm), have been deposited at room temperature using dc and RF magnetron sputtering in a Unaxis M12 sputter tool with base pressure below 1 × 10⁻⁶ Pa. The magnetic layer was formed by co-sputtering Co, Pt, and oxide targets. The oxide volume fraction (OVF), ranging from 0.08 to 0.19, was controlled by the oxide co-sputtering power. The microstructural and magnetic characterization has been described in our previous publications [2, 3].

XMCD measurements, including Co L-edge XMCD spectra as well as XMCD hysteresis loops, were carried out in total-electron-yield mode on bending magnet beamline 6.3.1 at the Advanced Light Source, Berkeley, using circularly polarized x-rays with a degree of polarization of 55%. Two different experimental geometries were used, as illustrated in the left panel of figure 1, i.e. at normal incidence (γ = 0°) and at grazing incidence (γ = 60°) of the x-ray beam, which allows for extracting the anisotropic behaviour of the magnetic moments. In each case the magnetic field was always along the incident x-ray beam, and all measurements were taken at room temperature.

Zoom In Zoom Out Reset image size

Figures 2(a) and (b) display the angle-dependent XMCD hysteresis loops of the samples with OVF = 0.08 and 0.15, respectively, which were obtained by recording the peak height of the Co L₃ signal at ∼778 eV divided by the Co L₂ signal at ∼793 eV as a function of the applied magnetic field. The loops show a pronounced hysteresis, and the difference between the curves at γ = 0° and 60° shows that the easy axis of K₁ is normal to the film plane, as depicted in the right panel of figure 1.

Zoom In Zoom Out Reset image size

From the above hysteresis measurements, we have obtained the coercivity of the samples, i.e. H_C,0° (γ = 0°) and H_C,60°(γ = 60°), as shown in figure 2(c). The calculated values of H_C,60° from H_C,0°, based on the domain-wall-motion (DWM, H_C,60° = H_C,0°/cos60°) and the Stoner–Wohlfarth (SW, H_C,60° = H_C,0°cos60°sin60°) models are also plotted in the figure for comparison [21, 22]. We found that the experimentally obtained H_C,60° value agrees well with the calculation by the SW model. This indicates that the reversal mechanism of these samples tends to comply with the SW model wherein coherent rotation dominates [22], which suggests that the grains are magnetically decoupled and their magnetization reversal is independent of each other [23]. More importantly, we noticed that such a relationship exists for all OVF values, which rules out the effect of intergrain coupling/decoupling as the main cause for the change of K₁ in the studied thin films.

From the angle dependence of the XMCD hysteresis loops, we calculated the Co magnetic anisotropy energy (MAE) by using

$$\begin{equation*} {\rm MAE}=\frac{\int_0^{M_{\rm S}} {H\rmd M_{\theta_{1}} -\int_0^{M_{\rm S}} {H\rmd M_{\theta_{2}}}}}{\sin^{2}(\theta_{1} -\theta_{2})}\mbox{ [24]}, \end{equation*}$$

where H is the applied magnetic field, θ₁ = 0°, θ₂ = 60° and M_S is the total magnetic moment involving the orbital and spin magnetic moments of Co, which can be obtained from the XMCD spectra and will be shown in the following paragraphs. As shown in figure 2(d), the value of the Co MAE decreases from ∼0.18 meV/atom to ∼0.08 meV/atom as OVF increases from 0.08 to 0.19. Although the Co MAE is estimated from the model with domain nucleation involved, which may not be remarkably accurate, its decreasing trend with increasing OVF is consistent with that of K_1,2 thus suggesting that the main cause of the K₁ reduction is the drastically reduced Co MAE in the studied thin films [3].

A detailed analysis of the Co orbital and spin moments as a function of OVF is required, in order to extract the microscopic origin of the drop in K₁. Accordingly, angle-dependent x-ray absorption spectra (XA) and XMCD spectra were taken at the Co L_2,3 edge by reversing the magnetic field of 2 T. As shown in figure 3, the spectra have been normalized to the step height of the absorption spectra at 815 eV, i.e. well above the L₂ edge, where dichroic effects are negligible. From the XA spectra, we see that there is no signature of oxide formation, which would give otherwise multiplet features [25].

Zoom In Zoom Out Reset image size

The Co 3d effective spin moment, m_S, and orbital moment, m_L, have been evaluated from the XMCD spectra using the sum rules analysis [16]. The m_S and m_L values obtained at different x-ray incidence angles, are plotted in figures 4(a) and (b), respectively, as a function of OVF. Note that the values at γ = 60°, m_S,60° and m_L,60°, have been corrected for incomplete magnetic saturation, i.e. by multiplication by a factor of M_2T,0°/M_2T,60°, where M_2T,0° (≈ saturation magnetization) and M_2T,60° represent the 0° and 60° magnetization obtained at a 2 T field by vibrating sample magnetometry.

Zoom In Zoom Out Reset image size

As seen in figure 4(a), m_S,0° ≈ m_S,60°, indicating that m_S is almost isotropic. This also means that the magnetic dipole term in the sum rules, which causes an anisotropy in the spin distribution, can be safely neglected here [26, 27]. In sharp contrast, m_L is highly anisotropic, as shown in figure 4(b): m_L,0° > m_L,60°, which is consistent with the hysteresis loops in figures 2(a) and (b), where the easy axis of K₁ is perpendicular to the film plane, i.e. along the direction of γ = 0°.

Considering the different trends of m_L,0° and m_L,60° in figure 4(b), it can be concluded that m_L,0° is nearly independent of the OVF, suggesting that the effect of stacking faults could be negligible here, since it has been reported that a considerable reduction in the orbital moment along the easy axis of K₁ is expected due to the collapse of the stacking ordering [10]; while m_L,60° shows a considerable increase of ∼0.06 μ_B/atom with increasing OVF (∼44% increase from OVF = 0.08 to OVF = 0.19), which is an essential result of this work, as we will discuss in the following paragraphs.

Based on the relation m_L,60° = m_L,0°cos²60° + m_L,90°sin²60°, where m_L,0° and m_L,90° refer to the orbital moments parallel and perpendicular to the sample surface normal, respectively [28], we extracted the value of m_L,90°, as also shown in figure 4(b). In our case, the direction of m_L,90° is normal to the grain–oxide interface, as illustrated in figure 4(c), with the orbital anisotropy, Δm_L = m_L,0° − m_L,90°, shown as well.

Figure 4(b) shows that m_L,90° increases rapidly with increasing OVF, reaching a value of ∼0.20 μ_B/atom at OVF = 0.19, where the orbital moment is nearly isotropic. This high value corresponds to a large orbital-to-spin moment ratio of ∼0.15, comparable to values reported for various CoPt thin films [20, 29]. Since m_L,0° is virtually constant, the strongly increasing trend of m_L,90° leads to a distinctly decreasing behaviour of Δm_L with increasing oxide content. As shown in figure 4(c), Δm_L decreases from ∼0.12 μ_B/atom (OVF = 0.08) to ∼0.02 μ_B/atom (OVF = 0.19), which can be correlated to the reductions in the Co MAE (from ∼0.18 to ∼0.08 meV/atom) and K₁ (from ∼0.8 to ∼0.6 × 10⁷ erg cm³) when the OVF increases. Thus, the above analysis provides a microscopic origin, namely that the orbital moments become increasingly isotropic, with increasing OVF, for the reduction of K₁ in the films we have characterized.

The underlying mechanism can be understood as follows: as the OVF increases, the magnetic grains become smaller, such that the grain–oxide interface, i.e. the interface between the grains and WO₃, starts to play an increasingly important role. At the grain–oxide interface, the effective coordination number of Co atoms is reduced. Compared to bulk Co, which features a high crystal symmetry, the lower symmetry at the interface produces a strong uniaxial crystal field. This may be expected to lift the orbital degeneracy, resulting in different occupations of the in-plane and out-of-plane 3d orbitals [30]. In turn, an enhancement in m_L,90° would result [13], which has been predicted theoretically and found experimentally in low-dimensional systems like nanoclusters [13, 31, 32], surfaces and interfaces [33–35]. Consequently, the higher the OVF is, the stronger the symmetry breaking at the interface, and the larger the m_L,90°. Hence, our observed m_L,60° enhancement, or equivalently the increase of m_L,90°, with increasing OVF, is mainly ascribed to the reduced symmetry at the grain–oxide interfaces, as was also demonstrated in Co nanoparticles on Pt(1 1 1) surface showing a giant magnetic anisotropy [36]. This kind of anisotropy has its easy axis orthogonal to the interface, caused by an enhanced spin–orbit interaction in that direction, which partly counteracts the overall perpendicular magnetic anisotropy K₁ here [7, 8].

Summarizing, the present work clarifies the microscopic origin of the reduced magnetocrystalline anisotropy in Co₈₀Pt₂₀ : oxide thin films when the oxide volume fraction increases. For the studied thin films, the angle-dependent XMCD analysis showed that the orbital moment becomes increasingly isotropic with the increasing oxide content, caused by a symmetry reduction at the grain–oxide interface, which is the main reason for the drop in the magnetocrystalline anisotropy. Keeping in mind that the effect of symmetry breaking is more pronounced in smaller grains with higher surface-to-volume ratio, our results reveal an intimate correlation between the grain size and the magnetocrystalline anisotropy, which is instructive for obtaining an optimum oxide volume fraction, beneficial for the stable [37] and high-density recording simultaneously.

The authors would like to thank Dr Elke Arenholz and Dr Catherine Jenkins for their kind support at BL 6.3.1. The Advanced Light Source is supported by the Director, Office of Science, Office of Basic Energy Sciences, of the US Department of Energy under Contract No DE-AC02-05CH11231.