Introduction

Spin-orbit coupling (SOC) is a relativistic effect and relates the spin moment of an electron to its orbital momentum via a momentum-dependent effective magnetic field. In the presence of SOC, a charge current without any spin polarization can be converted into a pure spin current (the flow of spin angular momentum) and vice versa, known as the direct spin Hall effect and the inverse spin Hall effect (ISHE)1,2,3,4,5,6,7,8,9,10,11,12,13,14,15,16,17,18. The presence of these phenomena in non-magnetic semiconductors had already been recognized in 1970s (refs 19, 20, 21, 22) but recent theoretical1,2 and experimental investigations3,4 have given a boost to this field. Spin current-based electronics with low-energy consumption has since been discussed, where spin-current injection and detection, using charge/spin conversion by direct spin Hall effect and ISHE, played a key role. A variety of materials have subsequently been explored to realize efficient charge/spin conversion5,6,7,8,9,10,11,12,13,14,15,16,17,18. The playground for the exploration was first limited to semiconductors but then extended to metals, where we can take advantage of the less pronounced effect of interfacial barriers. Heavy transition metals such as Pt6,7,8,9,10,14, Au9,12,13, and Pd8,9,11 and also Cu-based alloys17,18 were found to exhibit a particularly large spin Hall angle αSH (the maximum yield of the charge/spin conversion), 0.01–0.1 at room temperature, owing to their pronounced SOC effects. These effects give these metals an advantage for application as a spin injector. In addition to the large αSH, their low electrical resistivity (ρC), typically ρC=10−7–10−5 Ω cm, allows large charge currents to be passed and hence to inject a large spin current without serious Joule heating.

Spin currents can be detected by using the ISHE. In contrast to the case for spin injectors, the low ρC of the heavy metals with a large αSH is a disadvantages when it comes to spin detection. In the ISHE, the electric voltage ΔVISHE generated by a spin current IS is in proportion to ρC (refs 23, 24):

Spin Hall resistivity given by determines the efficiency of spin-current detection. This means that a sensitive detection of spin current could be achieved in materials with both a large αSH and a high ρC. In typical heavy metals, their low ρC imposes strong constraints in tuning the value of ρSH as it remains as low as <1 μΩ cm. One of the obvious approaches to improve the performance of spin detectors could be to increase the ρC by alloying. Alloying can in fact increase not only the ρC but also the αSH through an extrinsic spin Hall effect25,26,27. At room temperature where the inelastic scattering of charge carriers is dominant, however, it should be hard to anticipate a drastic increase of ρC by alloying, for example, more than an order of magnitude increase as compared with the pristine material. Therefore, the improvement of ρSH would not be achieved only by alloying.

In this work, we propose 5d transition metal oxides (TMOs) as alternative metals to realize both large αSH and large ρC, which should enhance ρSH. The uniqueness of 5d TMOs is characterized by the extremely strong SOC~0.5–1 eV originating from the predominant 5d character of the conduction band. The SOC in 5d TMOs is in fact as strong as to reconstruct the electronic structures drastically, as first discussed in the case of Jeff=1/2 Mott state Sr2IrO4 (ref. 28). The analogous predominant SOC effect has been commonly observed in a broad range of Ir oxides including Ba2IrO4 (ref. 29), CaIrO3 (ref. 30) and Ir2O4 (ref. 31). In those Ir oxides, in addition to the strong SOC, the localized character of d orbitals gives rise to a moderately high charge resistivity ρC even in the metallic state. Typical ρC values of conductive 5d Ir TMOs such as IrO2 and SrIrO3 at room temperature are of the order of 10−4–10−3 Ω cm, at least one order of magnitude higher than those of normal s metals. As the variety of materials continues to grow rapidly, 5d TMOs offer a unique opportunity to explore the giant ρSH for efficient spin detection. We selected a simple binary oxide, rutile IrO2 (refs 32, 33, 34), as the basis for our exploration. IrO2 with Ir4+ shares the same 5d5 configuration with many other Ir oxides, where the strong SOC dominates the electronic states and the resultant electronic properties34. This binary oxide has been long used as an electrode in various device applications, ranging from non-volatile ferroelectric memories to electrochemical devices. The excellent electrode properties are due to the formation of a clean, well-defined interface with other materials owing to its high chemical and thermal stability, and superior barrier properties for oxygen diffusion. In such a clean interface, it may be possible to inject a spin current through a diffusion process; IrO2 is therefore an ideal platform for testing the ISHE of 5d TMOs. Here, we describe successful spin-current injection into IrO2 using a lateral spin-valve (LSV) device geometry and demonstrate the ISHE with a remarkably high ρSH as a metal not only in the polycrystalline state but also in the practically important amorphous state.

Results

Device structure

We fabricated cross-junction-type devices consisting of an IrO2 wire and a permalloy(Ni80Fe20, Py)/Ag/Py LSV35, shown in Fig. 1a,b (see also Methods). Spin currents are passed in the Ag strip by a non-local charge current injection (IC) from the ferromagnetic Py electrode and are perpendicularly injected in part into the IrO2 wire through the diffusion (Fig. 1a). This effect, called spin absorption, enables us to conduct a quantitative analysis of the ISHE as have been demonstrated for various materials7,8,17,18. We examined the potential of polycrystalline and amorphous IrO2 wires as the spin detector element. Their resistivities were 2.0 × 10−4 Ω cm and 5.7 × 10−4 Ω cm at 300 K, respectively, which are 1–2 orders of magnitude higher than those of metals that have been studied to date as a spin Hall material.

Figure 1: Inverse spin Hall effect measurements using a spin absorption effect in a lateral spin-valve geometry.
figure 1

(a) The structure of device used for ISHE measurements is shown schematically. A LSV with two Py electrodes (Py1: spin source, Py2: spin detector) bridged by an Ag spin transport layer was formed on an IrO2 wire. Spin-polarized charge current, IC, was injected along the arrow to accumulate pure spin currents in the Ag. The diffusion of spin current into the IrO2 wire, namely, the spin absorption, gives rise to the ISHE in IrO2. (b) Scanning electron microscopy image of a typical device. Scale bar, 500 nm. The dotted lines indicate the side edges of the IrO2 wire for clarity.

Full size image

To check that the Ag layer is not oxidized by directly contacting with IrO2, we first characterized the Ag/IrO2 interface by means of interface resistance measurements36. The current–voltage characteristics of the Ag/IrO2 interface showed an ohmic behaviour, typical of metal/metal interface. The slope of the data yielded a resistance–area product (RA) of 2.3 fΩ m2 at 300 K, which is as low as that of the metallic, transparent Ag/Py interface (RA=1.9 fΩ m2 at 300 K). An estimate of the thickness of the possibly oxidized Ag layer, using the resistivities of Ag2O and AgO in the literature37, never exceeded 0.1 nm. These observations imply that the Ag/IrO2 interface in the present devices is very sharp as in metal/metal interfaces.

Spin-current injection into IrO2

The spin absorption by IrO2 across the interface was then confirmed by non-local spin-valve (NLSV) experiments7,8,17,18,35. The schematic diagram of the measurement is shown in Fig. 2a. The pure spin currents accumulated in the Ag strip by injecting spin-polarized charge currents IC from Py1 to Ag is detected as a voltage VS (=V+−V−) between Py2 (V+) and Ag (V−) (upper panel). The spin accumulation signal ΔRS is given by the difference in VS/IC (RS) between parallel (high RS) and antiparallel (low RS) magnetizations of the two Py wires. If an IrO2 middle wire is in contact with the Ag, the accumulated spin diffuses in part into it (lower panel), resulting in the reduction in ΔRS (ref. 7). Figure 2b and c show field dependences of RS at 300 and 10 K, respectively. The decrease of ΔRS by the insertion of the IrO2 middle wire indicates the spin absorption effect. This is more clearly demonstrated in the temperature (T) dependence of ΔRS with (ΔRSwith) and without (ΔRSwithout) the middle wire, as displayed in Fig. 2d. At all the temperatures, ΔRSwith is systematically smaller than ΔRSwithout, while both are enhanced at low temperatures owing to the increase of the spin-diffusion length of Ag (λAg)35. By analysing the data based on the three-dimensional spin-diffusion model18 (see Supplementary Methods and Supplementary Fig. S1), we estimated a spin-diffusion length in polycrystalline IrO2, =3.8 nm at 300 K and 8.4 nm at 10 K, which are comparable to those of Pt (3–10 nm at 300 K (refs 6, 7, 8, 9, 10, 14)) and CuIr (5–20 nm at 10 K (ref. 17)) where relatively large spin Hall angles were observed. This indicates that spin scattering of IrO2 is indeed very strong as expected from the 5d conduction band34.

Figure 2: Non-local spin-valve measurements.
figure 2

(a) The set-up of measurement is schematically depicted. The magnetic field H is applied along the easy axis of the Py electrodes. (b,c) NLSV signals measured at 300 K and 10 K for Py/Ag/Py LSVs without and with the polycrystalline IrO2 middle wire. (d) Temperature dependences of the NLSV signals.

Full size image

ISHE and ρSH of IrO2

By injecting spin currents through the spin absorption effect, we successfully observed ISHE in IrO2. In Fig. 3a,b, we plot the room-temperature ISHE signal RISHE (≡ΔVISHE/IC) for polycrystalline and amorphous samples as a function of the magnetic field (The temperature dependence of the ISHE is shown in Supplementary Fig. S2 and is explained in Supplementary Note 1). Here the magnetic field was applied along the hard axis of the Py spin source (see Fig. 1a), and ΔVISHE was induced along the long-axis direction of the IrO2 wire. Reflecting the magnetization process in hard axis of Py, RISHE linearly increased up to ~1,500 Oe and then saturated. The antisymmetric response to the applied magnetic field excludes any spurious effects such as anisotropic magnetoresistance of Py; the ISHE of IrO2 is responsible for the resistance change (2ΔRISHE) between positive and negative magnetic fields. The in-plane angular dependence of magnetic field effect on RISHE was confirmed fully consistent with the ISHE origin. Figure 3c shows the variation in RISHE under a magnetic field direction rotated by φ with respect to the ordinary set-up (Fig. 3d, inset). The sign change of RISHE occurs between φ=0 and π by magnetization switching of Py and concomitant polarization reversal of the spin current. As shown in Fig. 3d, the resistance change 2ΔRISHE vanishes at φ=π/2 and exhibits a cos φ relation. This characteristic angular dependence is the hallmark of ISHE7,12, evidencing spin-dependent electron scattering in IrO2.

Figure 3: Inverse spin Hall effect of IrO2 at 300 K.
figure 3

(a) The ISHE signal RISHE (≡ΔVISHE/IC) for a polycrystalline IrO2 wire with a width w of 170 nm was measured as a function of the applied magnetic field H. (b) The results for amorphous IrO2 samples. Two devices with different w values (100 and 170 nm) were measured. The 100 nm-width device showed larger RISHE because of the increase of spin-current density. RISHE of the two amorphous devices are smaller than that of the polycrystalline sample shown in a. (c,d) Angular dependence of RISHE and the resistance change 2ΔRISHE by the ISHE. The in-plane azimuthal angle φ with respect to the long axis of Ag is rotated to change the polarization direction of the spin current (inset in d). The solid line in d is a cos φ curve fitted to the results.

Full size image

Because of the large difference in ρC between Ag and IrO2, the charge current induced in the IrO2 wire by the ISHE is partially shunted by the adjacent Ag layer. To take into account such geometric effects and to precisely determine ρSH, we adopted the three-dimensional spin-diffusion model (see also Supplementary Methods). Using the ISHE signal 2ΔRISHE and data obtained from the NLSV measurements, αSH and ρSH are, respectively, calculated to be 0.040 and 8.0 μΩ cm for polycrystalline IrO2, and 0.065 and 37.5 μΩ cm for amorphous IrO2 at 300 K.

Figure 4 summarizes spin Hall resistivity ρSH and electrical resistivity ρC for a variety of metals with a large spin Hall angle αSH reported so far, together with that for IrO2. It mimics the point of this work in that ρSH of IrO2 is distinctly large compared with typical heavy metals and their alloys of which ρSH is ~0.16–1.2 μΩ cm (refs 6, 7, 8, 9, 10, 11, 12, 13, 14, 17, 18), and is now comparable to the large ρSH that was recently discovered in ultrathin films of β-phase (A15 crystal structure) W (ref. 15) and Ta (ref. 16).

Figure 4: Spin Hall resistivity of metals.
figure 4

Experimentally measured values of spin Hall resistivity ρSH for various metals are plotted as a function of electrical resistivity ρC. Data include Pt (refs 9, 10) and Pd (ref. 11) evaluated at room temperature by spin pumping, Pt (ref. 14) and atomic layer thin films of β-W (ref. 15) and β-Ta (ref. 16) at room temperature by spin-transfer-torque ferromagnetic resonance, Au at 295 K using a planar Hall device12, Ta, Nb, Pd, Mo, Pt (ref. 8), 12% Ir-doped Cu (ref. 17), 0.5% Bi-doped Cu (ref. 18) at 10 K, and polycrystalline and amorphous IrO2 at 300 K (this work) by spin absorption.

Full size image

Discussion

To generate a large electric voltage using the ISHE of IrO2 from existing metal-based spin-current devices, it is important to reduce the shunting effect at the Ag/IrO2 interface due to their large difference in ρC. One of the promising approaches is to use Al (ref. 38) and Mg (ref. 39) as the non-magnetic spin transport layer. Their ρC values are one order of magnitude higher than that of Ag we used, of the order of 10 μΩ cm at room temperature, while a comparably long spin-diffusion length is obtained. Another interesting route is a spin transport layer based on a conductive oxide composed of light elements with small SOC. For example, in ZnO (ref. 40), doping with Al yields ρC of the order of 100 μΩ cm (ref. 41), which can match with that of IrO2. The high chemical stability of IrO2 could be advantageous in fabricating such oxide-based junctions.

The discovery of the ISHE in amorphous IrO2, which can be grown at room temperature, is a significant step towards spintronics devices based on such oxide materials. The comparably large αSH in the amorphous form may point to that spin scattering in this compound occurs at the length scale of as short as a unit cell, potentially enabling the spin detection at the atomic layer thickness. Although the physics of spin scattering in amorphous states has not been understood yet, the availability of the amorphous form should make it easier to study SOC-induced transport phenomena in 5d TMOs not used before in spintronics.

In conclusion, 5d TMOs are the most promising materials for spin detection in spintronics. We demonstrated that the combination of strong SOC and a moderately high ρC is a key to improve the performance of spin detection. These achievements in the simple binary oxide manifest the potential of 5d TMOs as a new class of spintronic materials. Strong SOC in this class of materials produces a rich variety of intriguing physical properties, including correlated topological insulator42, Weyl semi-metal43 and Kitaev spin liquid44. Our discovery of the large spin Hall resistivity may make 5d TMOs even more fascinating.

Methods

Device fabrication

The Py/Ag/Py LSV with an IrO2 middle wire was fabricated on a SiO2/Si substrate. Au/Ti leads for electrical measurements were prepared on the substrate by a photolithography and an e-beam deposition. IrO2 film was grown by a reactive sputtering with a 99.9% pure Ir target. A wire structure of IrO2 was formed using a posi-type resist patterned by an e-beam lithography. The typical wire width and thickness were 170 and 15 nm, respectively. During the deposition, the substrate was not heated, and the pressure was fixed at 0.7 Pa of Ar/O2 (90:10) mixture gas. X-ray diffraction analysis indicated that the as-deposited IrO2 film was amorphous. The amorphous film turned into polycrystalline after a post-annealing at 400 °C in air. The LSV with a Py–Py separation of 700 nm (centre-to-centre) was then formed onto the IrO2 wire by the shadow evaporation method35,36. To prevent the surface oxidation of Ag, a 2-nm-thick MgO protection layer was e-beam deposited on the surface without breaking vacuum. The width and thickness of wire were 170 and 100 nm for Ag, and 170 and 20 nm for Py, respectively. Their ρC and spin-diffusion length were 3.0 μΩ cm and 330 nm for Ag and 47 μΩ cm and 5 nm for Py, respectively, at 300 K. One of the two Py electrodes was made longer than the other (not shown in Fig. 1b) to make the contrast in the fields for magnetization reversal.

Electrical measurements

All electrical measurements were performed by a standard lock-in technique in a He flow cryostat. An alternating current IC of 200 μA with a frequency of 79 Hz was used for NLSV measurements. In ISHE measurements, IC of 400 μA was applied. We confirmed the linearity between the output voltage ΔVISHE and IC in the range of 100–400 μA.

Additional information

How to cite this article: Fujiwara, K. et al. 5d iridium oxide as a material for spin-current detection. Nat. Commun. 4:2893 doi: 10.1038/ncomms3893 (2013).