The generation and detection of spin currents play a critical role in the field of spintronics. To date, the spin pumping effect and the spin Seebeck effect (SSE) are two widely used approaches to generate spin currents in nonmagnetic / ferromagnetic heterostructures.^[1–10] The spin current can be detected by the inverse spin Hall effect (ISHE) which converts a spin current into an electric voltage through spin–orbit coupling (SOC) in nonmagnetic materials.^[11–21] Usually, some 5d metals are used as spin detector materials due to their enhanced intrinsic SOC and large spin Hall angle θ_SH (typical 0.01–0.1), such as Pt and Au.^[1,2,7–11] Meanwhile, some oxides with larger resistivity are proposed to serve as spin detectors in order to achieve larger ISHE signal, according to the V_ISHE ∝ ρ_c θ_SHI_s relationship.^[22–25] Fujiwara et al. reported that IrO₂ can be used for spin detection and the spin Hall resistivity ρ_SH = ρ_c θ_SH of IrO₂ is even larger than that of typical heavy metals and their alloys.^[22] Qiu et al. observed a strong ISHE signal in indium tin oxide (ITO)/yttrium iron garnet (YIG) bilayers by spin pumping, and found that the mixing conductance at the ITO/Py interface is very close to that of a metal/metal interface.^[23,24] Further, a smaller V_ISHE in IrO₂/YIG was achieved in the longitudinal SSE (LSSE) configuration, and attributed to the weak interfacial spin mixing conductance.^[25] So far, there are only a limited number of reports on using oxides for spin detection, and more experimental investigations are essential to realize high conversion of spin current into electric voltage by different approaches.

ITO is the most widely used transparent oxide that shows a much higher resistivity than metals, which is beneficial to attain a large spin Hall resistivity. In this study, we investigate the generation of spin current in ITO/YIG bilayer by the method of spin pumping and LSSE. The ISHE of ITO/YIG driven by spin pumping under different magnetization configurations was observed, demonstrating that ITO can be used as a spin detect material. However, we did not observe the ISHE signal in the presence of a longitudinal temperature gradient in ITO/YIG, which is probably due to almost the same thermal conductivities of ITO and YIG and particular interface structure.

First, a 100-nm-thick ITO film was deposited on single crystalline (111) YIG films by magnetron sputtering at room temperature. The YIG film of 3.23 μm thickness was grown by the method of liquid phase epitaxy on Gd₃Ga₅O₁₂ (GGG) substrates. Then, the stripe (length: 3 mm, width: 0.3 mm) and Hall bar (length: 5 mm, width: 3 mm, arm width: 0.3 mm) structures of the ITO layer were prepared by standard photolithography for spin pumping and SSE measurements, respectively, as shown in Fig. 1(a). In addition, a Pt layer of 10 nm thickness with the same stripe and Hall bar structures was prepared on YIG as a reference sample. Silver colloid was employed to electrically connect the ITO surface and Pt wire as an electrode, and then was baked at 423 K for about 30 min to achieve Ohmic contacts. For the spin pumping experiment, a microwave excitation with the frequency of 3.5 GHz was generated and applied to the ITO/YIG samples by a microwave line with diameter around 30 μm. A magnetic field H was provided by a two-dimensional (2D) vector field. The ferromagnetic resonance (FMR) spectra of all the samples were taken by a N5230A PNA-L network analyzer from Agilent Technologies. The DC voltage induced by ISHE in ITO was measured as a function of H. For the LSSE experiment, a longitudinal temperature gradient ∇T_z was formed out of the plane in the z direction by sandwiching the sample between a large Cu block and a heater.^[26] The temperature difference ΔT_z is defined as the corresponding temperature difference between the ITO surface and the bottom of the sample substrate. Before measurements, the transport setup was mounted in a cryostat of low vacuum to effectively reduce the interference signals outside. All of the experiments were performed at room temperature.

Fig. 1.

Figure Option ViewDownloadNew Window

Fig. 1. (a) Schematic diagram of the ITO(100 nm)/YIG sample and electrical measurement geometry. The stripe and the Hall bar were used for spin pumping and LSSE measurements, respectively. (b) ISHE voltage as a function of H along y axis for the as-prepared ITO/YIG, annealed ITO/YIG, and Pt/YIG. (c) FMR linewidths of YIG, Pt/YIG, as-prepared ITO/YIG, and annealed ITO/YIG structure as a function of the microwave frequency.

Initially, a 20 mW microwave excitation of 3.5 GHz was applied on an as-prepared ITO(100 nm)/YIG sample (Fig. 1(a)). The ISHE voltage V_ISHE was measured as a function of magnetic field H with H along the y axis in the film plane. In Fig. 1(b), it is noted that two anti-symmetric voltage peaks with opposite signs appear around ±700 Oe. The ISHE voltage exited by spin pumping is ascribed to the generation of spin currents in ITO, which is generated by spin pumping in YIG and injected into the adjacent ITO layer due to ISHE.^[7–9] The single peak feature is different from the multiple peak characteristics reported in ITO/La:YIG, which was attributed to the spin wave resonant effect.^[23] For comparison, the magnetic field dependence of V_ISHE of a Pt(10 nm)/YIG sample is shown in Fig. 1(b). Note that the sign of the voltage peak of Pt/YIG is reversed relative to that of ITO/YIG, which is attributed to the opposite sign of θ_SH.

Figure 1(c) shows the FMR linewidths of the YIG, Pt/YIG, and as-prepared ITO/YIG structure as a function of the microwave frequency. It is shown that the linewidth increases linearly with the frequency, from which the damping parameter α can be estimated to be 1.44 × 10⁻¹ and 1.45 × 10⁻¹ for the as-prepared ITO/YIG and Pt/YIG structure, respectively.^[27,28] In general, the spin Hall angle θ_SH and interfacial spin mixing conductance g_↑↓ can be deduced from

Consequently, θ_SH and g_↑↓ of Pt/YIG are calculated to be 8.46 × 10⁻¹ and 8.06 × 10¹⁸ m⁻², which are comparable to the values in the previous reports,^[11,31] indicating the validity and reliability of the spin pumping measurement. For the as-prepared ITO/YIG, θ_SH and g_↑↓ are calculated to be −2.88 × 10⁻⁴ and 7.66 × 10¹⁸ m⁻², respectively. In the previous reports, θ_SH and g_↑↓ of ITO/Py and IrO₂/YIG were estimated to be 0.0065 ± 0.001 and (1.1 ± 0.2) × 10¹⁹ m⁻², 4 × 10⁻¹ and 1.2 × 10¹⁶ m⁻², respectively.^[22,24,25] By comparison, the absolute value of θ_SH of ITO/YIG is much smaller than that of Pt/YIG (8.46 × 10⁻¹) and IrO₂/YIG (4 × 10⁻¹). Meanwhile, g_↑↓ of ITO/YIG is comparable to that of Pt/YIG, and even two orders of magnitude larger than that of IrO₂/YIG. The typical value of resistivity of the as-prepared ITO in our experiment is about 4.46 × 10⁻⁴ Ω⋅m. Therefore, the large ISHE signal of the as-prepared ITO/YIG arises from the large value of g_↑↓ of ITO/YIG and the large resistivity of ITO, according to Eq. (1). Furthermore, this conclusion prompts that the ISHE signal may be enhanced by annealing the ITO/YIG sample so as to further vary g_↑↓ and resistivity of ITO. To validate this assumption, we annealed the ITO/YIG sample at 340 °C in oxygen atmosphere for 10 min, then measured the transport properties. In Fig. 1(b), it is seen that the ISHE voltage of the annealed ITO/YIG structure is reduced compared with that of the as-prepared ITO/YIG. The θ_SH and g_↑↓ of the annealed ITO/YIG sample are estimated to be −2.92 × 10⁻³ and 3.47 × 10¹⁹ m⁻², respectively, which are obviously larger than those of the as-prepared sample (−2.88 × 10⁻⁴ and 7.66 × 10¹⁸ m⁻²). After the annealing process, the ITO film resistivity decreases from 4.46 × 10⁻⁴ Ω⋅m for the as-prepared ITO to about 8.62 × 10⁻⁶ Ω⋅m. The different trends of variation of θ_SH, g_↑↓, and resistivity of ITO jointly lead to the reduced ISHE signal of the annealed ITO/YIG. Therefore, it is expected to achieve higher spin-to-charge conversion by optimizing the annealing temperature in the future.

To obtain insight of the characteristics of ISHE, we measured the angular dependence of V_ISHE of the as-prepared ITO/YIG. Figure 2(a) shows the results of V_ISHE as a function of H at different ϕ with H applied in the YIG film plane. As ϕ changes from 270° to 180°, V_ISHE at the resonant field (around 700 Oe) is gradually reduced from the maximum to zero, and reverses its sign as ϕ changes from 180° to 90°. The values of H_res and V_ISHE in Fig. 2(a) have been extracted to be plotted as a function of ϕ, as shown in Figs. 2(b) and 2(c), respectively. It is obviously seen that H_res exhibits a cos2ϕ dependence, as a result of a weak uniaxial in-plane magnetic anisotropy field of YIG (about 72.53 Oe).^[32] For V_ISHE, the obvious sinϕ relation is a characteristic of ISHE since V ∝ J_s × M.^[8,23] Figure 2(d) shows the microwave power dependence of V_ISHE at ϕ = 270°. It can be seen that V_ISHE increases linearly with P, which was expected for the ISHE induced by spin pumping.^[8,33] Thus, these results confirm that the observed ISHE voltage arises from FMR spin pumping due to the magnetic field applied in the YIG film plane. For the annealed ITO/YIG samples, the similar results were qualitatively obtained.

Fig. 2.

	Figure Option ViewDownloadNew Window
	Fig. 2. (a) Angular dependent VISHE–H characteristics of ITO/YIG with in-plane magnetic field. (b) Hres and (c) VISHE as a function of ϕ. (d) VISHE as a function of microwave excitation power P.

To further investigate the dependence of the spin polarization vector on ISHE, the V_ISHE signal of the as-prepared ITO/YIG was measured with the magnetic field applied out-of-plane. Figure 3(a) shows the field dependence of V_ISHE at various θ. It is clearly seen that V_ISHE increases from the minimum to maximum as θ changes from 0° to 90°, and H_res varies in the opposite trend. The values of H_res and V_ISHE in Fig. 3(a) are plotted as a function of θ in Figs. 3(b) and 3(c), respectively. The similar angular symmetries for H_res and V_ISHE exhibit the out-of-plane ISHE behaviors. By comparison of Figs. 2(b) and 3(b), it is noted that the H_res for the out-of-plane geometry is much larger than that of in-plane due to the extra force to counteract the demagnetization field of YIG. Due to the enhanced demagnetizing field with H applied out-of-plane, the direction of the magnetization of YIG does not always follow that of H,^[8] leading to the deviation of the V_ISHE–θ curve from sinθ as shown in Fig. 3(b). As a result, the angular dependences of H_res and V_ISHE measured in both in-plane and out-of-plane geometries indicate that the spin current can be generated in YIG by the spin pumping excitation, and ITO can be used as a spin detecting material.

Fig. 3.

	Figure Option ViewDownloadNew Window
	Fig. 3. (a) Angular dependent VISHE–H characteristics of ITO/YIG with out-of-plane magnetic field. (b) Hres and (c) VISHE as a function of θ.

Beside the spin pumping effect, it is well known that spin currents can also be generated by a temperature gradient in NM/FM heterostructures, which is manifested as the LSSE.^[34–36] To explore the efficient method of generating spin currents, we studied the ISHE by applying a longitudinal temperature gradient to the as-prepared ITO/YIG structure.^[28] Figure 4 shows the thermal voltage as a function of the magnetic field. It is obvious that the ISHE signal is absent in the presence of the longitudinal temperature differences of 50 K and 100 K, revealing that the spin current is hardly excited through heat gradient in the as-prepared ITO/YIG. This is also true for the annealed ITO/YIG. In contrast, a giant ISHE signal can be achieved in the Pt/YIG structure when a temperature difference ΔT_z of 50 K is applied, as illustrated in Fig. 4.^[2,37]

Fig. 4.

	Figure Option ViewDownloadNew Window
	Fig. 4. The H dependence of thermal voltage of ITO/YIG at (a) ΔTz = 50 K and (b) 100 K. Insert in (a) shows the V–H curves of Pt/YIG at ΔTz = 50 K.

We suggest that almost the same thermal conductivity of the ITO and substrate (ITO: ∼ 5.95 W⋅m⁻¹⋅K⁻¹, YIG: 7.4 W⋅m⁻¹⋅K⁻¹, GGG: 9 W⋅m⁻¹ ⋅K⁻¹) plays a critical role in the LSSE measurement.^[38,39] As a temperature gradient is applied across the ITO/YIG/GGG heterostructure, the huge thickness difference (ITO: 100 nm, YIG/GGG: 0.5 mm) leads to that most heat is dropped across the YIG/GGG substrate. Hence, the temperature gradient across the ITO/YIG interface is negligible, producing no spin currents and ISHE signal in ITO.^[5,40–43] On the contrary, the remarkable difference of thermal conductivity between Pt (77.8 W⋅m⁻¹⋅K⁻¹) and YIG/GGG (6–9 W⋅m⁻¹⋅K⁻¹) causes a large temperature gradient across the Pt/YIG interface, giving rise to the generation of spin currents and the corresponding ISHE signal in Pt.^[36,37] A weak V_ISHE signal has been observed in IrO₂/YIG, and attributed to the smaller g_↑↓ (1.2 × 10¹⁶ m⁻²).^[25] Usually, the smaller g_↑↓ originates from the rough interface of the layered heterostructures. However, the rough interface also gives rise to a large interfacial thermal resistance,^[5,40–43] forming a sizable temperature gradient at the IrO₂/YIG interface and facilitating to introduce spin currents. Therefore, it is necessary to find a compromise between the interface mixing conductance g_↑↓ and interfacial thermal resistance via adjusting the interfacial structure so as to achieve large LSSE. It is expected to realize high spin to charge conversion for all oxide systems through interface structural modification in the future. Recently, Flebus and Cornelissen et al. suggested that a spin superfluidity can be generated in the bulk of YIG and manifest as SSE,^[44–47] implying that there may be other factor hindering the visibility of the LSSE signal in our experiment. One possibility is that the spin mixing conductance may have a large negative temperature coefficient,^[48,49] so that at higher measurement temperature, the magnon current is blocked by the interface, then the spin mixing conductance will be eliminated. Thereafter, we measured the thermal voltage of ITO/YIG as a function of H in the temperature range of 150 K to 300 K, but no visible ISHE signal was found. The LSSE measurement in the lower temperature regime will be performed in the future, so as to reveal the relative contribution of interface and YIG bulk toward ISHE.

In summary, we investigated the ISHE in ITO/YIG by the methods of spin pumping and spin Seebeck effect. It is shown that the ISHE voltage can be generated in ITO by spin pumping under different magnetization configurations. Moreover, the enhancement of spin Hall angle and interfacial spin mixing conductance can be realized by appropriate thermal treatment. However, the ISHE voltage is hardly seen in the presence of a longitudinal temperature gradient, which can be ascribed to the almost equal thermal conductivity of ITO and YIG, as well as the specific interface structure, or the large negative temperature dependent spin mixing conductance. These results provide vital information to realize effective spin-current-driven devices by engineering the interfacial structures.