Efficient spin transport through native oxides of nickel and permalloy with platinum and gold overlayers

We present measurements of spin pumping detected by the inverse spin Hall effect voltage and ferromagnetic resonance spectroscopy in a series of metallic ferromagnet/normal metal thin film stacks. We compare heterostructures grown in situ to those where either a magnetic or nonmagnetic oxide is introduced between the two metals. The heterostructures, either nickel with a platinum overlayer (Ni/Pt) or the nickel-iron alloy permalloy (Py) with a gold overlayer (Py/Au), were also characterized in detail using grazing-incidence x-ray reflectivity, Auger electron spectroscopy, and both SQUID and alternating-gradient magnetometry. We verify the presence of oxide layers, characterize layer thickness, composition, and roughness, and probe saturation magnetization, coercivity, and anisotropy. The results show that while the presence of a nonmagnetic oxide at the interface suppresses spin transport from the ferromagnet to the nonmagnetic metal, a thin magnetic oxide (here the native oxide formed on both Py and Ni) somewhat enhances the product of the spin-mixing conductance and the spin Hall angle. We also observe clear evidence of an out-of-plane component of magnetic anisotropy in Ni/Pt samples that is enhanced in the presence of the native oxide, resulting in perpendicular exchange bias. Finally, the dc inverse spin Hall voltages generated at ferromagnetic resonance in our Py/Au samples are large, and suggest values for the spin Hall angle in gold of $0.04<{\alpha }_{\mathrm{SH}}<0.22$, in line with the highest values reported for Au. This is interpreted as resulting from Fe impurities. We present indirect evidence that the Au films described here indeed have significant impurity levels.

Figure 1

X-ray reflectivity vs $q_z$ for (a) Ni/Pt, Ni/NiO/Pt, Ni/Ag/Pt and (b) Py/Au, Py/PyOx/Au, and Py/AlOx/Au samples. Black lines represent measured data with red lines the result of refinement to the model density profiles shown in Figs. 2 and 3. The top two plots in each panel are offset upward by a factor of 100 and $10000$.

Figure 2

Refined x-ray scattering-length density (SLD) and depth-profiled Auger electron spectroscopy for (a)–(b) Ni/Pt, (c)–(d) Ni/NiO/Pt, (e)–(f) Ni/Ag/Pt heterostructures. Ni films left exposed to atmosphere for $\sim 24$ hours between FM and NM layer deposition show clear accumulation of O Auger electron signal localized to the interface.

Figure 3

Refined x-ray scattering-length density (SLD) and depth-profiled Auger electron spectroscopy for (a)–(b) Py/Au, (c)–(d) Py/PyOx/Au, (e)–(f) Py/AlOx/Au heterostructures. Py layers left exposed to atmosphere for $\sim 24$ hours between FM and NM layer deposition show clear accumulation of O Auger electron signal localized to the interface.

Figure 4

Inverse spin Hall effect measurements via spin pumping excited by cavity FMR. Top plots of panels (a)–(f) show the FMR response as a function of field with $9.85$ GHz microwave excitation. Bottom plots show the dc spin-pumping voltage measured on the NM layer as a function of field. (a) Ni/Pt, (b) Ni/NiO/Pt, (c) Ni/Ag/Pt, (d) Py/Au, (e) Py/PyOx/Au, (f) Py/AlOx/Au. Note that the presence of the native oxide in both series slightly increases the total $\mathrm{\Delta }{V}_{\mathrm{ISH}}$ (see text), and that the Py/Au series generates large $\mathrm{\Delta }V$. For only the Py/AlOx/Au, the film stack could not be placed exactly at the node of the electric field, and $V_{\mathrm{dc}}$ is dominated by a rectification-driven asymmetric (peak-dip) component. Panel (f) shows the spin-pumping component extracted by fit from the raw signals shown in (g) and (h). In (g) and (h) symbols are measured $V_{\mathrm{dc}}$, solid red lines the fit, and dashed lines show the separate symmetric and antisymmetric components.

Figure 5

FMR data for Ni- and Py-based samples. Example raw FMR response for Ni/Pt and Ni/NiO/Pt samples shown as (a) real and (b) imaginary parts of $S_{21}$ at 20 GHz. The fits shown determine $H_{\mathrm{res}}$ and $\mathrm{\Delta }H$ for each $f$ as shown in (c) and (d) for all samples. Lines in (c) are fits to the out-of-plane Kittel equation [Eq. (3)] and those in (d) are linear fits [Eq. (4)] to extract $\alpha$.

Figure 6

In-plane and out-of-plane magnetization $M$ vs $H$ for Ni heterostructures at $T=300$ K. For Ni the expected $M_s\sim 500\mathrm{emu}/{\mathrm{cm}}^3$.

Figure 7

In-plane and out-of-plane $M$ vs $H$ for Py-based heterostructures at $T=300$ K. For Py the expected $M_s\sim 800\mathrm{emu}/{\mathrm{cm}}^3$.

Figure 8

Summary of evidence of perpendicular exchange bias in the Ni/NiO/Pt sample based on $M$ measured after field cooling from 300 to 10 K under either in-plane or perpendicular external field bias of 3000 Oe. (a) Coercive field for in-plane applied field ($H_{\mathrm{c},\parallel }$) and out-of-plane applied field ($H_{\mathrm{c},\perp }$) as well as perpendicular exchange bias field ($H_{\mathrm{e},\perp }$) vs $T$. In-plane exchange bias was within error bars of $H=0$ at all $T$ and is not shown. (b) $M$ vs $H$ for in-plane applied field shown at various $T$. Larger field scans show a 10 K saturation near 5 kOe. (c) $M$ vs $H$ for out-of-plane applied field shown at various $T$. (d) Schematic view of the interface based on GIXR results (Fig. 2) demonstrating the proposed mechanism for EB and enhanced coercivity. Note that we have not measured the lateral length scale of the roughness, and this schematic therefore only qualitatively represents the morphology of the surfaces. Red arrows indicate points of direct Ni/Pt contact.

Figure 9

Room temperature trends in (a) $\alpha$ (filled square), and $H_c$ [both in-plane (open diamond) and out-of-plane (open circle)], and (b) $M_{\mathrm{eff}}$ (filled square) and $H_k$ (open diamond) for the Ni/Pt sample series. Trends in $\alpha$ are similar to those in $H_c$, and suggest changes in damping are tied to magnetic anisotropy, rather than to spin pumping.

Figure 10

Room temperature trends in (a) $\alpha$ (filled square), and $g_{\uparrow \downarrow ,\mathrm{eff}}$ (open triangle: for intrinsic damping $=0.006$, open pentagon: for intrinsic damping $=0.0077$), and (b) $M_{\mathrm{eff}}$ (filled square) and $H_k$ (open diamond) for the Py/Au sample series. Trends in $\alpha$ here are in line with a spin-pumping origin, and we use two different estimates of intrinsic damping ${\alpha }_0$ to estimate the spin-mixing conductance independently from the electrically detected spin-pumping measurement. These values of $g_{\uparrow \downarrow ,\mathrm{eff}}$ fall in the range expected for transition-metal FM/NM interfaces.

Figure 11

(a) $g_{\uparrow \downarrow ,\mathrm{eff}}{\alpha }_{\mathrm{SH}}$ calculated for each sample using Eq. (11) or (13). All samples show relatively similar values other than Py/AlOx/Au, which is strongly suppressed as expected. (b) $g_{\uparrow \downarrow ,\mathrm{eff}}$ calculated from $g_{\uparrow \downarrow ,\mathrm{eff}}{\alpha }_{\mathrm{SH}}$ using assumed values for ${\alpha }_{\mathrm{SH}}$. For Pt samples the commonly reported ${\alpha }_{\mathrm{SH},\mathrm{Pt}}=0.11$ gives $g_{\uparrow \downarrow ,\mathrm{eff}}$ in the expected range. For Au, we use three values of ${\alpha }_{\mathrm{SH},\mathrm{Au}}$. Values on the order of tenths of a percent give very large $g_{\uparrow \downarrow ,\mathrm{eff}}$, even for the sample with the AlOx layer. ${\alpha }_{\mathrm{SH},\mathrm{Au}}$ of several percent or more yields $g_{\uparrow \downarrow ,\mathrm{eff}}$ in line with expectations for transition-metal FM/NM interfaces.

Figure 12

Electrical resistivity vs $T$ for Au thin films grown in the same chamber from the same source material near the time of the growth of the Au films in samples Py/Au and Py/PyOx/Au (labeled A–C) compared to other samples grown from nominally identical but uncontaminated Au source material. The very large residual resistivity indicates extensive impurities in the A–C films, likely from cross-contamination from the Py source material in the same chamber.