Spin pumping blocked by single-layer graphene

Spin pumping, the generation of spin currents from precessing magnetization, has been studied in ferromagnetic/paramagnetic bilayers. In the bilayer, the angular momentum of precessing magnetization in the ferromagnetic layer can be transferred to conduction electron spins in the paramagnetic layer under ferromagnetic resonance (FMR), which emits a spin current into the adjacent paramagnetic material.^1–3) This technique has been used to generate nonequilibrium spin currents in a wide range of materials from ferromagnetic metals, semiconductors, and insulators.^4–7)

Recent studies have demonstrated experimentally that spin pumping is driven by short-range spin–exchange coupling at ferromagnetic/paramagnetic interfaces.⁸⁾ The evidence has been obtained by measuring spin pumping by inserting an insulating thin barrier at the interface; the strength of spin pumping decreases exponentially with increasing barrier thickness in Pt/barrier/Y₃Fe₅O₁₂ heterostructures.⁸⁾ Suppression of spin pumping has also been observed in metallic junctions by inserting a thin MgO barrier or a nano-oxide layer at the interface.^9,10)

In this letter, we demonstrate that spin pumping in a metallic junction is blocked even by single-layer graphene (SLG). Spin transport in single or a few layers of graphene has been the subject of intense interest in spintronics, owing to the low intrinsic spin–orbit interaction, as well as the low hyperfine interaction of the electron spins with the carbon nuclei.^11–13) Although graphene is highly conductive in the plane, it exhibits low conductivity perpendicular to the plane, making it a suitable material for tunnel barriers.¹⁴⁾ In this work, we found that spin pumping in a Ni₈₁Fe₁₉/Pt junction is strongly suppressed by inserting SLG at the interface; the magnetization damping of a Ni₈₁Fe₁₉/SLG/Pt trilayer is as small as that of a Ni₈₁Fe₁₉ film, whereas the damping of a Ni₈₁Fe₁₉/Pt bilayer is twice as large as that of the Ni₈₁Fe₁₉ film. This finding reveals that the atomic monolayer acts as an efficient barrier that blocks spin-current emission from the ferromagnetic metal, illustrating the crucial role of short-range spin–exchange coupling at ferromagnetic/paramagnetic interfaces in spin pumping.

A sample Ni₈₁Fe₁₉/SLG/Pt trilayer with a 2 × 3 mm² rectangular shape was fabricated by the following process. SLG was grown on a Cu foil with low-pressure chemical vapor deposition (CVD) at 1000 °C with a mixture of methane, argon, and hydrogen gases with partial pressures of 20, 40, and 20 Pa, respectively. A 50-nm-thick Pt layer was deposited on a 2 × 3 mm² sapphire (0001) substrate by radio frequency (rf) magnetron sputtering in an Ar atmosphere, and then the entire Pt surface was covered with SLG by transferring the SLG from a Cu foil via the conventional poly(methyl methacrylate)-assisted method.^15,16) To remove residual impurities, the SLG/Pt bilayer was subsequently annealed at 300 °C under high vacuum. Finally, a 10-nm-thick Ni₈₁Fe₁₉ layer was deposited at ambient temperature by electron beam deposition after annealing at a higher temperature of 400 °C. Ni₈₁Fe₁₉ and Ni₈₁Fe₁₉/Pt bilayer samples with similar shapes were fabricated by sputtering (Pt) and electron-beam evaporation (Ni₈₁Fe₁₉). The thicknesses of the Pt and Ni₈₁Fe₁₉ films were 10 and 10 nm, respectively. For the FMR measurement, the Ni₈₁Fe₁₉/SLG/Pt trilayer was placed on a coplanar wave guide, where a microwave was applied to the input of the signal line. The signal line was 500 µm wide, and the gaps between the signal line and the ground lines were designed to match the characteristic impedance of 50 Ω. An in-plane external magnetic field H was applied parallel to the signal line. We measured the external magnetic field dependence of the microwave absorption I for different microwave frequencies f. All the measurements were performed at room temperature.

Figure 1 shows the Raman spectra of the SLG/Pt bilayer after 300 °C annealing and the SLG/SiO₂ sample that was fabricated by transferring the CVD-grown SLG onto a 300-nm-thick SiO₂/Si substrate. The spectra were collected from randomly selected areas of 100 × 100 µm² on SLG by scanning a laser beam (488 nm wavelength) with a spot area of about 1 µm². The Raman signals were detected from the entire scan area in a similar manner, indicating that the Pt surface is fully covered with graphene. As in the SLG/SiO₂ sample, the strong enhancement of graphene Raman signals on a 300-nm-thick SiO₂ layer is useful for characterization of graphene.¹⁷⁾ The four peaks assigned to the D, G, D + D'', and 2D bands of graphene are seen in the spectra. The sharp peaks at 1554 and 2327 cm⁻¹ are due to O₂ and N₂ in air, respectively. The large 2D band/G band intensity ratio I(2D)/I(G) of about 2 in SLG/SiO₂, the symmetric shape of the 2D band, and the small D band intensity in SLG/Pt and SLG/SiO₂ show high-quality SLG even after the transfer process.¹⁸⁾ The peak positions of the G and 2D bands (1589 and 2712 cm⁻¹) in SLG/Pt are shifted to the higher-frequency side by 1 and 5 cm⁻¹, respectively, compared to the positions (1588 and 2707 cm⁻¹) in SLG/SiO₂. The blue shifts of the G and 2D bands and also the significant decrease in I(2D)/I(G) in SLG/Pt (see the figure) can be attributed to the p-type doping of SLG, which is relevant to the larger work function of Pt (5.3–5.9 eV) compared to that of SLG (4.5 eV).¹⁹⁾ Furthermore, the visible broadenings of the G and 2D bands in SLG/Pt suggest the existence of strong interfacial interactions between SLG and Pt.²⁰⁾

Zoom In Zoom Out Reset image size

In Fig. 2(a), we show the external magnetic field (H) dependence of the microwave absorption I under 10 mW microwave excitation with a frequency of f = 6.0 GHz. We observed FMR absorption around μ₀H_FMR = 53.1 mT. The resonance field μ₀H_FMR varies systematically with the microwave excitation frequency f, as shown in Fig. 2(b). By fitting the f dependence of μ₀H_FMR using the Kittel formula, we obtain the saturation magnetization μ₀M_s = 825 mT for the Ni₈₁Fe₁₉/SLG/Pt trilayer. The FMR was also measured for a Ni₈₁Fe₁₉/Pt film, in which SLG is missing, and a Ni₈₁Fe₁₉ film. The FMR measurements revealed that the saturation magnetization of the Ni₈₁Fe₁₉ film is not changed by attaching the SLG/Pt or Pt layers.

Zoom In Zoom Out Reset image size

In Fig. 3, we show the microwave frequency (f) dependence of the FMR linewidth μ₀ΔH for the Ni₈₁Fe₁₉/Pt and Ni₈₁Fe₁₉ films. Figure 3 shows that μ₀ΔH is reproduced well by the sum of the linewidth due to Gilbert damping α and the linewidth ΔH₁ due to inhomogeneity:

$$\begin{equation} \Delta H = \frac{2\pi\alpha}{\gamma\mu_{0}} f + \Delta H_{1}, \end{equation} \tag{ 1 }$$

where γ is the gyromagnetic ratio. Equation (1) shows that the damping constant α can be determined from the slope of the linear fit to the experimental data. Thus, Fig. 3 shows that the magnetization damping α for the Ni₈₁Fe₁₉ film is enhanced by attaching the Pt film. This enhancement can be attributed to spin pumping at the Ni₈₁Fe₁₉/Pt interface, because spin pumping emits a spin current due to the precessing magnetization, resulting in enhanced angular momentum dissipation from the ferromagnetic layer. The spin current emitted from the ferromagnetic layer is expressed as^1,2)

$$\begin{equation} \mathbf{j}_{\text{s}} = \frac{\hbar}{4\pi}g_{\text{eff}}^{\uparrow \downarrow}\mathbf{m}(t) \times \frac{d\mathbf{m}(t)}{dt}, \end{equation} \tag{ 2 }$$

where m = M/M_s. Further, $g_{\text{eff}}^{ \uparrow \downarrow }$ is the effective spin-mixing conductance, which is proportional to the damping enhancement $\Delta \alpha = \alpha _{\text{F/N}} - \alpha _{\text{F}}$ due to spin pumping as²¹⁾

$$\begin{equation} \Delta\alpha = \frac{g\mu_{\text{B}}}{4\pi M_{\text{s}}d}g_{\text{eff}}^{\uparrow\downarrow}, \end{equation} \tag{ 3 }$$

where d is the thickness of the Ni₈₁Fe₁₉ layer, g is the g factor, and μ_B is the Bohr magneton. Further, α_F/N and α_F are the damping for the Ni₈₁Fe₁₉/Pt and Ni₈₁Fe₁₉ films, respectively. From the measured f dependence of μ₀ΔH shown in Fig. 3, the damping enhancement Δα due to spin pumping at the Ni₈₁Fe₁₉/Pt interface is Δα = 1.49 × 10⁻². From this value and the measured saturation magnetization, we obtained the effective spin-mixing conductance for the Ni₈₁Fe₁₉/Pt junction, $g_{\text{eff}}^{ \uparrow \downarrow } = 3.39 \times 10^{19}$ m⁻², which is consistent with previous reports.^9,22)

Zoom In Zoom Out Reset image size

The enhancement of the magnetization damping disappears when SLG is inserted. The f dependence of μ₀ΔH for the Ni₈₁Fe₁₉/SLG/Pt trilayer is also shown in Fig. 3. The magnetization damping for the Ni₈₁Fe₁₉/SLG/Pt trilayer determined from the experimental result is α = 8.69 × 10⁻³. This value is much smaller than α = 1.49 × 10⁻² for the Ni₈₁Fe₁₉/Pt film and is comparable to α = 7.41 × 10⁻³ for the Ni₈₁Fe₁₉ film, demonstrating that the damping enhancement due to spin pumping at the Ni₈₁Fe₁₉/Pt interface is strongly suppressed by inserting SLG.

In a linear response regime, the spin conductance $g_{\text{s}} = g_{\text{eff}}^{ \uparrow \downarrow }/4\pi$ dictating the spin pumping efficiency is given by^7,23)

$$\begin{equation} g_{\text{s}} = \frac{2J_{\text{sd}}^{2}S_{0}^{2}N_{\text{int}}}{\hbar^{2}N_{\text{SS}}}\sum_{\mathbf{k}}\,\frac{1}{\Omega_{\text{rf}}}\text{Im}[\chi_{\mathbf{k}}^{\text{R}}(\Omega_{\text{rf}})], \end{equation} \tag{ 4 }$$

where J_sd is the s–d exchange interaction constant at the ferromagnet (F)/spin sink (SS) interface, S₀ is the localized spin magnitude of the F layer, N_int is the number of F spins at the F/SS interface, N_SS is the number of lattice sites in the SS layer, and Ω_rf = 2πf denotes the frequency of the applied rf field. The transverse spin susceptibility is given by $\chi _{\mathbf{k}}^{\text{R}}(\omega ) = \chi _{0}/(1 + \lambda ^{2}k^{2} - i\tau _{\text{sf}}\omega )$, where χ₀ denotes the uniform static susceptibility with the spin diffusion length λ and spin relaxation time τ_sf. For Ω_rfτ_sf ≪ 1, which is satisfied in the present study, $\text{Im} [\chi _{\mathbf{k}}^{\text{R}}(\Omega _{\text{rf}})] \simeq \chi _{0}\tau _{\text{sf}}\Omega _{\text{rf}}(1 + \lambda ^{2}k^{2})^{ - 2}$, and the k-summation in Eq. (4) is constant. Although it is reasonably anticipated that insertion of SLG does not greatly affect the "bulk" properties, i.e., N_SS and τ_sf, of the SS (Pt or SLG/Pt complex), the magnitude of the spin–exchange coupling J_sd is more sensitive to the electronic properties of the F/SS interface. At a ferromagnetic insulator/SS interface,⁷⁾ J_sd has been estimated as two orders of magnitude smaller than those of typical ferromagnetic metal/SS interfaces. Insertion of an oxide or Schottky barrier into the F/SS junction further reduces J_sd, causing the exponential decay of the spin pumping efficiency observed in a recent experiment.^8–10) SLG has been found to work as an insulating interlayer to prevent impedance mismatch between a ferromagnetic metal and a semiconductor;^24–26) thus, the suppression of spin pumping observed in the present study is attributed mainly to the reduction in the spin–exchange coupling at the F/SS interface. In addition, because the spin resistance of SLG is much larger than those of Ni₈₁Fe₁₉ and Pt, the insertion of SLG prevents the effective transfer of spin currents at the interface, which can also contribute to the suppression of spin pumping in the present system.

During the preparation of this paper, a similar result in a Ni₈₁Fe₁₉/SLG/Cu trilayer was reported.²⁷⁾ Because Pt acts as a good spin sink owing to the strong spin–orbit interaction the result presented in this paper provides further clear evidence of the spin pumping blocking effect.

In summary, we investigated the FMR of Ni₈₁Fe₁₉, Ni₈₁Fe₁₉/Pt bilayer, and Ni₈₁Fe₁₉/SLG/Pt trilayer films at different microwave frequencies. The FMR measurements demonstrate that the enhanced magnetization damping due to spin pumping in the Ni₈₁Fe₁₉/Pt bilayer is dramatically suppressed by inserting SLG. This finding illustrates the crucial role of short-range spin–exchange coupling at a ferromagnetic/paramagnetic interface in the dynamical generation of spin currents. It also provides a route for controlling magnetization damping, which is affected by spin pumping, in magnetic tunnel junctions using SLG, offering a way to optimize the magnetic response of spintronic devices.

Acknowledgments