Observation of the intrinsic inverse spin Hall effect from ferromagnet

Transcription

1 Observation of the intrinsic inverse spin Hall effect from ferromagnet Ayaka Tsukahara #, Yuta Kitamura #, Eiji Shikoh, Yuichiro Ando, Teruya Shinjo and Masashi Shiraishi * # Graduate School of Engineering Science, Osaka University, Toyonaka , Japan *corresponding author: Masashi Shiraishi (shiraishi@ee.es.osaka-u.ac.jp) # These authors equally contributed to this study. Integration of spin injection, modulation, and detection into a single elementary device is the ultimate goal of spintronics technology 1. In conventional semiconductor spintronic devices 2, 3, these processes operate separately in different parts of the device; namely spin injection into semiconductor from ferromagnetic reservoir, modulation of spin current in semiconductor channel, and spin detection by using non-magnetic metal electrodes. Although, advancements in spin injection techniques such as spin pumping 4-7, and spin detection via inverse spin Hall effect (ISHE) by using metals with large spin-orbit interaction (SOI) 8-10 have realized a more efficient and sensitive semiconductor spintronic devices, the basic device structure remains the same. Here, we report a discovery that a single Permalloy (Ni 80 Fe 20, Py) film can act as spin reservoir (for spin injection), channel 1

2 for pure spin current (generated by itself), and spin detector (by virtue of its SOI) simultaneously to itself. We used spin pumping to generate pure spin current in the Py, and detected it by measuring electromotive force (EMF). Our measurements and calculations reveal that the EMF is induced by the intrinsic ISHE in Py. Our study provides insight into the role of ISHE in the mechanism of this phenomenon, which can be utilized for possible development of simpler spintronics devices with high functionality. Figure 1 shows the schematic structure of the Permalloy (Py)/SiO 2 device. Py (Ni 80 Fe 20 ) layer is deposited on thermally oxidized Si substrate (SiO 2 /Si), where SiO 2 thickness, t = 500 nm. The length, l, and width, w, of the Py layer is 1.5 mm and 4.0 mm, respectively. The thickness, d, of Py is varied from 10 to 50 nm. Two electrodes (separated by 2.5 mm gap) are attached on the Py layer for electromotive force (EMF) measurement. Ferromagnetic resonance (FMR) measurements are performed using electron spin resonance (ESR) system. Magnetic field, H, is applied to the Py layer at an angle, θ H, as shown in Fig. 1. See Methods section for details of device fabrication and measurement set-up. Figure 2a shows FMR spectra, di(h)/dh as a function of H-H FMR of the device with 10-nm-thick Py layer, under excitation power, P MW, of 200 mw, for θ H = 0 o (top), 90 (middle), 2

3 180 (bottom), where I is microwave absorption intensity in arbitrary unit. Although. FMR signals were observed for all θ H, EMF is observed only at 0 and 180, and not at 90 o, as shown in Fig. 2b. Also, the polarity reversal of EMF is observed when θ H, is changed from 0 to 180, as indicated by the inversion of the EMF line shape in Fig. 2b (top and bottom). These findings suggest that EMF is induced by spin-related dynamics in Py, because the experimental results are similar to ISHE-induced EMF in conventional spin injection devices Since EMF is measured in Py layer only, this suggests that spin current is induced inside Py, itself, and ISHE is responsible for the EMF. To explore this phenomenon, we apply the conventional fitting function used for analyzing the induced EMF in spin injection devices 9 to obtain the contribution of ISHE and anomalous Hall effect (AHE) to the EMF in our device, because AHE is also induced in Py. Figure 2c shows the experimentally obtained EMF at θ H = 0 (open circle) as a function of H-H FMR. Here, surprisingly, the fitting curve (solid red line) agrees perfectly with the experimental result, suggesting that spin current is induced in Py layer. To determine the contribution of ISHE and AHE to the EMF, we used the deconvoluted fitting function with independent contributions from ISHE and AHE (see Supplementary Information, SI). We found that the EMF is mainly induced by ISHE as indicated by the V ISHE fit (solid blue line), where V ISHE = 19.7 µv, and the contribution of AHE to EMF is negligibly small, as indicated by V AHE fit (solid green line). Moreover, the EMF as a function of H-H FMR for various P MW in Fig. 2d 3

4 shows that EMF increases linearly with P MW, (see solid circles in black in inset), which corroborates that the observed EMF is induced by ISHE, similar to that of other spin injection devices To understand how EMF is induced by ISHE, it is important to consider first how spin current is generated and flows in the Py layer. Upon the application of microwave into the system, the precession of magnetization in the whole Py layer generates pure spin current to conserve spin angular momentum. Since, EMF (V) results fromv ~ J s!!, where J s and σ are the spin current and spin angular momentum, respectively, which indicated by the polarity reversal of EMF in Fig. 2b, the spin current should flow towards the Py/SiO 2 interface. For this to occur, spin density gradient must be present in Py layer, in which the spin density is larger than in the vicinity of Py/SiO 2 interface. This is realized because spin relaxation processes at the Py/SiO 2 interface dominate the system, which is larger than that in Py layer. Thus, the local spin damping at the Py/SiO 2 interface induces spin density gradient, which allows diffusive flow of pure spin current in the Py layer. We note that no spin current flows if the spin density in the Py layer is uniform. Therefore, when spin current flows into the Py layer it couples with the SOI of Py and generates ISHE, which results in EMF. 4

5 If the above scenario is correct, a choice of substrates can verify this mechanism, because one can suppress the size of the spin density gradient in Py layer by using a substrate with small spin damping. To verify this, we chose Yttrium-Iron-Garnet (YIG, 1 µm thick) on Gadolinium-Gallium-Garnet (GGG) as a substrate, because (1) YIG has an electrical resistivity comparable to SiO 2, (2) the magnetization damping of YIG is small (α ~ , where α is the Gilbert damping constant), and (3) the resonant magnetic field of YIG under FMR is different from that of Py. As shown in Fig. 3, the EMF for Py/YIG sample is strongly suppressed under similar conditions imposed for Py/SiO 2 devices, and the EMF is estimated to be 2.7 µv by using the conventional fitting function described in SI, which is just one seventh of that observed in Py/SiO 2 device (19.7 µv). This result reveals that the amount of the spin current generated in the Py is governed primarily by the spin damping at the Py/SiO 2 interface. It is known that microwave application to spin pumping devices induces thermal agitation and temperature increase, which can contribute to spurious effects in the EMF induction in Py. One possible spurious effect is the anomalous Nernst effect 16, 17, in which temperature gradient, vertical to film plane, generates charge current that flows through the Py, yielding a lateral EMF, perpendicular to the magnetization of Py layer. In fact, the angular dependence of the EMF, induced by the anomalous Nernst effect, exhibits the same behaviour 5

6 as that observed in this study. However, we can conclude that the anomalous Nernst effect is negligibly small in our device, because the EMF in Py/YIG device is much smaller than that of Py/SiO 2 device, although the thermal conductivities of SiO 2 and YIG are in the same order (thermal conductivity, k, for SiO 2 and YIG are 1.2 and 6.0 Wm -1 K -1, respectively, and that of air is Wm -1 K ). We note that the EMF from Py/YIG device should be larger, if the anomalous Nernst effect governs this phenomenon, because the temperature gradient in Py layer is larger, due to larger thermal conductivity of the YIG. The findings discussed above allow us to rule out the possibility that the anomalous Nernst effect contributes to ISHE-induced EMF. We also note that the Al cap layer covering the Py/YIG device has negligible contribution to spurious effects 21, as shown in SI. We estimated the spin Hall angle, θ SHE, in our Py/SiO 2 device by extending the phenomenological model for the conventional spin pumping in Py/NM (Pt and Pd) devices, as reported elsewhere 22, 23. Figure 4a shows the schematic model of spin current diffusive flow in Py/SiO 2 device, where a parallelepiped-shaped Py layer is placed in contact with SiO 2 substrate. Under FMR condition, pure spin current density, j s 0 (inverted cone-shaped protrusion in green in Fig. 4a), is generated from an infinitesimal layer, dy, at the position ζ from the Py/SiO 2 interface, and flows toward the interface, as observed in the experiment. Then, j s 0 is converted to charge 6

7 current via ISHE, which is typically within spin diffusion length, λ F. In the calculation, we estimated j s 0 by using j 0 s = g r 2 2 γ h [4πM γ + 2 8πα [(4πM s s (4πM 2 2 ) γ s 2 2 ) γ 2 + 4ω ] 2 + 4ω ], (1) where h,!, and α are the microwave magnetic field, Dirac constant, and Gilbert damping constant, respectively, and ω (=2πf, where f is the microwave frequency) is the angular frequency of the magnetization precession, which is the established expression for j s 0 in conventional spin pumping devices 22. g r!" is the real part of the mixing conductance, given by g r!" = 2 3M s!d gµ B " (W F #W int ), (2) where g, µ B, W F, and W int are the g factor, Bohr magneton, FMR spectral width for the Py layer with the SiO 2, and FMR spectral width for Py without flow of a spin current, respectively. In the case of Py/YIG device, although spin damping is strongly suppressed in Py/YIG device, we introduce the value of the FMR spectral width of Py/YIG for the estimation of θ SHE, since the suppression is not perfect here. Since, j s 0 is uniformly generated in the Py layer, the total charge current can be calculated by integrating the contribution from each dy, in the thickness of Py layer, in y direction. Here, spin current decays along y-direction towards Py/SiO 2 interface as, j' ( y) s sinh[( ς y)/ λf ] j sinh[( ς / y)] =, (3) where j s (y) is the spin current density at the position y, and λ F is the spin diffusion length of Py. 0 S 7

8 The j s (y) is converted to charge current density, j c (y) via ISHE as, j' c (y) =! SHE ( 2e! ) j' s (y). (4) Therefore, the average charge current density, due to the generation of pure spin current at ζ, is expressed as, j' c 1 ζ 2e λf ζ 0 ( ) j' c ( y) dy θshe( ) tanh( ) js ζ =. (5) 0 ζ 2λ F Thus, the total charged current induced in the Py layer is calculated by integrating j' c, j c 1 d = < j' c > dζ = θ d 0 SHE 2e λf ( ) d d 0 js 0 tanh( ζ 2λF ) dζ, (6) ζ where the decay and the conversion of the pure spin current are assumed to be uniform in the Py layer. Thus, the EMF (V ISHE ) can be calculated by using the equation, V ISHE = d l j c w R = θ σ F 2e λ d F 0 tanh( ζ 2λF ) SHE( ) js dζ 0 d ζ, (7) where R and σ F are the resistance and the conductivity of the Py, respectively. The calculations indicate that EMF depends on the thickness, d, of Py layer, which is an important finding, in order to estimate the spin Hall angle. Figure 4b shows the EMF of as a function of d in Py/SiO 2 device. The closed circles in red are the experimentally obtained EMF. The solid lines are theoretical lines for the EMF predicted for various θ SHE i.e (green), (blue) and (black), where λ F is set to 3 nm in Py at RT 10, and j s 0 = 1.06! 10-9 J/m 2 is calculated from Eq. (1). Apparently, the theoretical line for θ SHE =0.018 reproduces the experimental results very well. 8

9 Furthermore, to verify the obtained θ SHE in Fig. 4b, we estimated θ SHE by measuring the AHE of Py by using simple Hall measurement scheme (Fig. 5a). Since, the FMR measurements revealed that the magnetization of Py layer is perpendicular to the film plane at 1,000 mt, the linear relationship between the Hall voltage and H above 1,000 mt can be ascribed to the Hall effect. Figure 5b shows the Hall voltage as a function of H from -2,000 mt to 2,000 mt. The experimentally obtained Hall voltage is the open circles in black. The red line is the linear fit employed above 1,000 mt, but extended down to 0 mt to obtain (via ordinate intercept) the Hall voltage, V y due to AHE in Py, which is found to be 19.2 µv. The anomalous Hall resistivity, ρ SHE, is expressed as! AHE = V y d / I x = " AHE!! 2 24, 25, where! is the resistivity under the zero magnetic field,!!"# is the anomalous Hall conductivity, and I x is the applied dc current. The relationship between the conductivities of the spin Hall effect and the AHE is described as! SHE = ( 1 P )! AHE, where P is the spin polarization of Py 24. Since, θ SHE is expressed as! SHE = " SHE " F, the spin Hall angle of the Py is estimated to be , because P of Py is typically 0.02± The estimated θ SHE agrees well with the value (θ SHE = 0.018) obtained from the fitting results in Fig. 4b, which indicates that our simple model in Fig. 4a captures the essence of the mechanism of ISHE-induced EMF in Py/SiO 2 device. Our experiment and calculation provide information on how to control and utilize this effect for 9

10 possible development of simpler spintronics devices with high functionality. Methods Two different types of substrates were used for device fabrication in this study; thermally oxidized Si (500 nm in thick) / Si substrate, and Yttrium Iron Garnet (YIG, 1 µm in thick) on Gadolinium-Gallium-Garnet (GGG) substrate. After surface cleaning of the substrates with acetone and isopropanol, Ni 80 Fe 20 (Py) were deposited by using electron beam evaporation at room temperature (RT). The dimension of the Py layer was 1.5 mm 4.0 mm, and the thickness, d, was varied from 10 to 50 nm. Two leading wires for measuring electromotive force were attached to the edge of the Py film with Ag paste. The sample was placed near the center of the TE 102 cavity, where the magnetic-field component of the microwave mode was maximum and the electric-field component was minimum in the electron spin resonance (ESR) system (Bruker EMX10/12). Microwave mode with frequency, f, of GHz, and static external magnetic field were applied to the samples. The electromotive force was measured by using a nanovoltmeter (KEITHLEY 2182A) and all measurements were performed at room temperature. For the anomalous Hall measurements, a Hall-bar-shaped Py layer ( d = 50 nm ) with the dimension of µm 2 was fabricated on thermally oxidized Si substrate (SiO 2 /Si) by using electron beam lithography, lift-off process, and electron beam evaporation. A DC electric 10

11 current (I x = 1 ma) and a perpendicular magnetic field, H, were applied in a Physical Properties Measurement System (Quantum Design). The measurements were also performed at room temperature. 11

12 Acknowledgements This research was partly supported by a Grant-in-Aid for Scientific Research from the MEXT, Japan and by the Global COE program of Core Research and Engineering of Advanced Materials Interdisciplinary Education Center for Materials Research. Additional information The authors declare no competing financial interests. 12

13 References 1. Wolf, S. A. et al. Spintronics: A spin-based electronics vision for the future. SCIENCE 294, (2001). 2. Appelbaum, I. et al. Electronic measurement and control of spin transport in silicon. Nature 447, (2007). 3. Suzuki, T., Room-Temperature Electron Spin Transport in a Highly Doped Si Channel. Appl. Phys. Express 4, (2011). 4. Mizukami, S., Ando, Y. & Miyazaki, T. Effect of spin diffusion on Gilbert damping for a very thin permalloy layer in Cu/permalloy/Cu/Pt films. Phys. Rev. B 66, (2002). 5. Tserkovnyak, Y., Brataas, A. & Bauer, G. E. W. Enhanced Gilbert Damping in Thin Ferromagnetic Films. Phys Rev. Lett. 88, (2002). 6. Brataas, A., Tserkovnyak, Y., Bauer, G. E. W. & Halperin, B. I. Spin battery operated by ferromagnetic resonance. Phys Rev. B 66, (R) (2002). 7. Tserkovnyak, Y., Brataas, A., Bauer, G. E. W. & Halperin, B. I., Nonlocal magnetization dynamics in ferromagnetic heterostructures. Rev. Mod. Phys. 77, (2005). 8. Valenzuela, S. O. & Tinkham, M. Direct electronic measurement of the spin Hall effect. Nature 442, (2006). 9. Saitoh, E., Ueda, M., Miyajima, H. & Tatara, G. Conversion of spin current into charge 13

14 current at room temperature: Inverse spin-hall effect. Appl. Phys. Lett. 88, (2006). 10. Kimura, T., Otani, Y., Sato, T., Takahashi, S., Maekawa, S. Room temperature reversible spin Hall effect. Phys Rev. Lett. 98, (2007). 11. Ando, K., et al. Angular dependence of inverse spin Hall effect induced by spin pumping investigated in a Ni 81 Fe 19 /Pt thin film. Phys. Rev. B 78, (2008). 12. Ando, K., et al. Electrically tunable spin injector free from the impedance mismatch problem. Nature Mater. 10, (2011). 13. Shikoh, E., et al. Spin-pumping-induced spin transport in p-type Si at room temperature. submitted. 14. Koike, M., et al. Dynamical Spin Injection into p-type Germanium at Room Temperature. Appl. Phys. Express accepted for publication. 15. Kajiwara, Y., et al. Transmission of electrical signals by spin-wave interconversion in a magnetic insulator. Nature 464, (2010). 16. Huang, S. Y., et al. Intrinsic Spin-Dependent Thermal Transport. Phys Rev. Lett. 107, (2011). 17. Weiler, M., et al. Local Charge and Spin Currents in Magnetothermal Landscapes, Phys Rev. Lett. 108, (2012). 18. Kato, R. & Hatta, I. Thermal conductivity measurement of thermally-oxidized SiO 2 films on a silicon wafer using a thermo-reflectance technique. International Journal of Thermophysics 14

15 26, 1 (2005). 19. Padture, N. P. & Klemens, P. G. Low Thermal Conductivity in Garnets. J. Am. Ceram. Soc. 80, (1997). 20. Lemmon, E. W. & Jacobsen, R. T. Viscosity and Thermal Conductivity Equations for Nitrogen, Oxygen, Argon, and Air. International Journal of Thermophysics 25, 1(2004). 21. Parker, W. J., et al. Flash method of determining thermal diffusivity, heat capacity, and thermal conductivity. J. Appl. Phys. 32, 1679 (1961). 22. Ando, K. & Saitoh, E. Inverse spin-hall effect in palladium at room temperature. J. Appl.Phys. 108, (2010). 23. Ando, K., et al. Inverse spin-hall effect induced by spin pimping in metallic system. J.Appl. Phys. 109, (2011). 24. Naito, T., Hirashima, D. S., & Kontani, H. Tight-binding study of anomalous Hall effect in ferromagnetic 3d transition metals. Phys. Rev. B 81, (2010). 25. Shhindler, A. I. & Salkovitz, E. I. Ferromagnetic Hall Coefficient of Nickel Alloys. Physical Review 99, (1955). 15

16 Figure legends Figure 1 Sample structure. A schematic illustration of the Ni 80 Fe 20 (Py)/SiO 2 devices. The dimension of the Py layer is 1.5 mm (l) 4.0 mm and the thickness, d, is varied from 10 to 50 nm. Two electrodes are attached on the Py layer by using Ag paste. Electrodes separation width, w, is 2.5 mm. In the measurements, the sample is placed near the center of a TE 102 microwave cavity of an ESR system with a frequency of f = GHz for obtaining FMR. The static external magnetic field, H, is applied at an angle of θ H to the Py film plane. x, y and z are the direction of the coordinates. Figure 2 Electromotive forces from the Py/SiO 2 sample. a, FMR spectra, di(h)/dh, of the device with 10-nm-thick Py layer for θ H = 0, 90, 180, as a function of H-H FMR, where I is microwave absorption intensity in arbitrary unit. The microwave power is mw, and the ferromagnetic resonance field, H FMR and peak-to-peak width, W F, of the FMR signal is estimated to be mt and 7.8 mt, respectively. b, The H dependences the electromotive force (EMFs), V, for θ H =0 o, 90 o, and 180 o. Whereas the FMR spectra can be observed in every measurement, the EMFs were observed only when θ H =0 o and 180 o, and the signal polarity was reversed, which corresponds to the symmetry of the ISHE. c, H dependence 16

17 of the electromotive force, V, when θ H =0 o. The open circles are the experimental data, and the red solid line is fitting line obtained by using a conventional fitting function that considers the contribution from the ISHE and the AHE (see SI). The blue and green lines are the fitting lines for the ISHE signals from the Py and the AHE signals, respectively. The signal from the ISHE of the Py was quite dominant, and its intensity was calculated to be 19.7 µv. d, The H dependence of the V under different microwave excitation powers when θ H =0 o. The inset shows microwave power dependence of the V ISHE and the V AHE, where the V ISHE and V AHE are the magnitudes of the electromotive forces due to the ISHE and the AHE, respectively. Figure 3 An electromotive force from Al/Py/YIG. a, The H dependence of the FMR signal, di(h)/dh, when θ H =0 o for Al/Py/YIG sample (solid line in orange). The FMR signal for Py/SiO 2 /Si sample is also displayed as black solid line. A thin Al layer (4 nm in thick) was deposited onto the Py (10 nm in thick) in order to prevent oxidation of the Py. The procedures of the measurements of EMFs and the geometry of the sample are the same as those of the Py(10 nm)/sio 2 /Si sample. H FMR and W F is estimated to be mt and 3.4 mt, respevtively. b, The H dependence of the electromotive force, V, measured for the Py(10nm)/YIG, when θ H =0 o. The open circles are the experimental data and the red solid line is a fitting line obtained by using a conventional fitting function that considers 17

18 the contribution from the ISHE and the AHE (see also Supplementary Information). The blue and green lines are the fitting lines for the ISHE signals from the Py and the AHE signals, respectively. The notable is that the EMF by the ISHE from this sample is quite small and the I SHE was estimated to be 2.7 µv. Figure 4 A schematic of a simple model for estimating the spin Hall angle of the Py. a, Pure spin current, j s 0, is generated from an infinitesimal layer, dy, at the position of ζ from the interface of the Py/SiO 2 under the FMR, and flows to the interface as observed in the experiment. The j s 0 is converted to a charge current due to the ISHE typically within a length scale of spin diffusion length, λ F. Since the j s 0 is uniformly generated in the Py, the total charge current can be calculated by integrating the contribution from the layer, dy, in the thickness of the Py in the y direction. b, The V ISHE as a function of the Py thickness for the Py/SiO 2 sample. The solid lines are theoretical lines of the EMF predicted for a variety of the θ SHE (0.002, and 0.030), where the spin diffusion length in the Py at RT was set to be 3 nm 10 and the spin current density calculated from Eq. (1) was 1.06! 10-9 J/m 2. Red closed circles show the experimental data and the error bars show the standard deviation. Figure 5 Hall voltage measurements for the Py/SiO 2. 18

19 a, A schematic illustration of the Py/SiO 2 sample for Hall effect measurement. The dimension of the Py layer was 60 µm 700 µm and the thickness, d, was set to be 50 nm. A dc current, I, of 1 ma and a perpendicular magnetic field, H, are applied. b, The H dependence of the Hall voltage for the Py/SiO 2 sample. The linear relationship between the Hall voltages and the H above 1 T is reasonably explained by the Hall effect and the red line is a linear fit of the data above 1 T. The anomalous Hall voltage of the Py layer is obtained from the ordinate intercept of the fitting line. 19

20 Figures - V + Ni 80 Fe 20 Thermally oxidized Si substrate Fig.1 A. Tsukahara et al.

21 a b c d Fig.2 A. Tsukahara et al

22 a b Fig.3 A. Tsukahara et al

23 a dy b θ SHE = V(µV) θ SHE = θ SHE = the thickness d of (nm) 0 the Ni Fe d (nm) Fig. 4 A. Tsukahara et al

24 a V Ni 80 Fe 20 Thermally oxidized Si substrate b V (µv) ma fitting curve magnetic field H (mt) Fig 5 A. Tsukahara et al

25 Supplementary Information Observation of the intrinsic inverse spin Hall effect in Ni 80 Fe 20 Ayaka Tsukahara #, Yuta Kitamura #, Eiji Shikoh, Yuichiro Ando, Teruya Shinjo and Masashi Shiraishi * # Graduate School of Engineering Science, Osaka University, , Toyonaka, Japan A. Deconvolution of the electromotive force signals Since, contribution from anomalous Hall effect (AHE) is included in the EMF, deconvolutiton of the signals is important. In general, EMF-H curves have been analyzed by using following fitting function [1] ;! 2 V = V ISHE (H " H FMR ) 2 +! +V "2!(H " H FMR ) 2 AHE (H " H FMR ) 2 +! 2, (S1) Here, Γ denotes the damping constant and H FMR is the ferromagnetic resonance field (115.6 mt for θ H = 0 in this study). The first term, which has a symmetrical Lorentzian shape, corresponds to the contributions from the ISHE, whereas the second term, which has an asymmetrical shape,

26 corresponds to that from the AHE. For 10nm-thick Py/SiO 2 /Si sample, whose EMF is shown in Fig. 1b in the main text, V ISHE is estimated to be 19.7 µv, which is larger than that from ferromagnetic metal / nonmagnetic metal samples such as Py/Pd in previous studies. The ratio V ISHE /V AHE is estimated to be 7.3, indicating that the EMF is due to ISHE of Py is dominant. B. Electromotive force from Py/YIG/GGG samples In Fig. 3 of the paper, a 4 nm-thick Al layer has been deposited on Py/YIG/GGG sample in order to prevent oxidation of the Py layer and obtain an intrinsic FMR spectra of the Py layer. We also revealed that the EMF, due to ISHE in the Py layer, is not sensitive to existence of Al layer. Figure S1 shows FMR spectrum and EMF for Py/YIG/GGG sample without Al layer. The separation of the two contacts is ~1.3 mm. The signal shape of the EMF is almost the same with that of the Py/YIG/GGG sample with Al layer. In fact, the magnitude ratio V ISHE /V AHE is estimated to be 0.19, which is close to that of the Py/YIG/GGG sample with Al layer (0.17). In addition, although thermal conductivity of Al (k = 200 Wm -1 K -1 [2] ) is much better than that of air, the estimated spin current for Al (4 nm)/py/ SiO 2 /Si sample was almost the same as that observed in the Py/SiO 2.

27 Fig. S1 The H dependence of a the FMR signal, di(h)/dh, and b the electromotive force, V, when θ H =0 o for Py/YIG sample. The open circles in Fig. S1b are the experimental data and the red solid line is a fitting line obtained by using a conventional fitting function that considers the contribution from the ISHE and the AHE. The blue and green lines are the fitting lines for the ISHE signals from the Py and the AHE signals, respectively. References [1] Saitoh, E., Ueda, M., Miyajima, H. & Tatara, G. Conversion of spin current into charge current at room temperature: Inverse spin-hall effect. Appl. Phys. Lett. 88, (2006). [2] Parker, W. J., et al. Flash method of determining thermal diffusivity, heat capacity, and

28 thermal conductivity. J. Appl. Phys. 32, 1679 (1961).