Temporal behavior of the inverse spin Hall voltage in a magnetic insulator-nonmagnetic metal structure
M. B. Jungfleisch, A. V. Chumak, V. I. Vasyuchka, A. A. Serga, B. Obry, H. Schultheiss, P. A. Beck, A. D. Karenowska, E. Saitoh, B. Hillebrands
aa r X i v : . [ c ond - m a t . m e s - h a ll ] A p r Temporal behavior of the inverse spin Hall voltage in a magneticinsulator-nonmagnetic metal structure
M. B. Jungfleisch, ∗ A. V. Chumak, V. I. Vasyuchka, A. A. Serga,B. Obry, H. Schultheiss, † P. A. Beck, and B. Hillebrands
Fachbereich Physik and Forschungszentrum OPTIMAS,Technische Universit¨at Kaiserslautern, 67663 Kaiserslautern, Germany
A. D. Karenowska
Department of Physics, Clarendon Laboratory, University of Oxford, OX1 3PU Oxford, United Kingdom
E. Saitoh
Institute for Materials Research, Tohoku University, Sendai 980-8577, Japan (Dated: October 31, 2018)It is demonstrated that upon pulsed microwave excitation, the temporal behavior of a spin-waveinduced inverse spin Hall voltage in a magnetic insulator-nonmagnetic metal structure is distinctlydifferent from the temporal evolution of the directly excited spin-wave mode from which it originates.The difference in temporal behavior is attributed to the excitation of long-lived secondary spin-wavemodes localized at the insulator-metal interface.
Over the last decade, the field of spintronics has risento some prominence. Spintronics is concerned with thedevelopment of devices which exceed the performanceand energy efficiency of conventional charge-based elec-tronics by exploiting the electron’s spin degree of free-dom [1–4]. The information currency in spintronic sys-tems is spin angular momentum. Traditional spintronicarchitectures rely on the electron based spin-transport,however, spin angular momentum can also be transferredby magnons, the quanta of spin waves (collective excita-tions of the spin lattice of a magnetic material). Magnonphysics opens doors to insulator-based spintronic deviceswhich operate with pure spin currents entirely decoupledfrom charge carriers [1, 5]. Spin waves in magnetic in-sulators can propagate over macroscopic distances manyorders of magnitude longer than the spin diffusion lengthstypical of metallic and semiconductor materials [6–8].Spin pumping (which transforms spin waves into spinpolarized electron currents) and the inverse spin Hall ef-fect (iSHE) (which converts spin polarized electron cur-rents into conventional charge currents) are two physicalmechanisms of fundamental importance to the emergingfield of ‘magnon spintronics’. The temporal character-istics of these phenomena will ultimately determine theoperational speeds of magnon spintronic devices [9].In this Letter we report our findings relating to thetemporal behavior of an externally excited spin-wavepulse and a resulting inverse spin Hall voltage in a mag-netic insulator-nonmagnetic metal bi-layer. We demon-strate that there are pronounced differences in the tem-poral evolution of the two signals and show that these dif-ferences may be attributed to the excitation of secondary short-wavelength spin-waves localized at the insulator-metal interface via two-magnon scattering of the exter-nally excited mode.A platinum (Pt) coated yttrium iron garnet (YIG) film
FIG. 1. (Color online) (a) Schematic illustration of the exper-imental setup. (b) Spin pumping scheme and resulting inversespin Hall effect. was used in our experiments. YIG single crystal filmshave the smallest known spin-wave damping [6, 10]. Asa result, magnon currents can be observed in YIG overcentimeter distances [1]. Electron scatter in Platinum isstrongly spin-dependent making it an attractive materialfor iSHE voltage generation [12, 13]. The experimentalsetup is illustrated schematically in Fig. 1(a). It com-prises a 2.1 µ m thick YIG stripe with a 10 nm thick3 × Pt layer deposited on the top. The edge-to- i S H E - i ndu c ed v o l t age / U i S H E U m a x A b s o r bed m i c r o w a v e po w e r / PP FIG. 2. (Color online) The inverse spin Hall voltage (solid redline) and the absorbed microwave power (blue dashed line)are shown as functions of the applied microwave frequency.Data corresponds to a bias magnetic field of H = 1820 Oe.The maximum iSHE voltage U max = 60 µ V is observed at theferromagnetic resonance frequency f = 7 GHz. A decrease inthe spin-pumping efficiency for modes excited at frequenciessmaller than f is evident (see Ref. [11]). edge resistance of the Pt square is 54 Ohms. To ensuregood impedance matching and thus minimal distortionof detected signals, the Pt square was connected to the50 Ohm input of a voltage measuring instrument (seeFig. 1(a)) by a 50 Ohm coaxial cable.Magnetization precession was excited in the YIG layerby a microwave current applied to a copper microstripline of width 600 µ m above the Pt layer. The line wasisolated from the Pt by a thin dielectric coating of cyan-acrylate. Two excitation geometries were investigated.In the first, the microstrip line was placed across theYIG stripe (Fig. 1(a)), in the second (not shown) it wasorientated parallel to it. The temporal behavior of theiSHE voltages observed in the two cases was near iden-tical. Accordingly, we present here only data collectedusing the first geometry.The spin-wave modes excited in the YIG film were de-tected via time-resolved Brillouin light scattering (BLS)spectroscopy [14]. The BLS laser probe beam was fo-cused on the YIG/Pt sample and the intensity of inelas-tically scattered light (which is directly proportional tothe intensity of the scattering spin wave) was analyzedon a 500 ps timescale. For the time-resolved voltage mea-surements we used a wideband (DC to 200 MHz) voltageamplifier FEMTO DHPVA-200 and a 300 MHz band-width Agilent DSO6034A oscilloscope.Our experiments were performed in the following fash-ion: a magnetizing field H was applied across the YIGstripe and magnetization precession was driven by a mag-netic field h ( t ) induced by a microwave current appliedto the microstrip line. Under these conditions, due tothe spin pumping effect of the precessing magnetizationat the YIG/Pt interface, a spin polarized electron cur-rent J s is produced in the Pt layer [1, 13, 15]. As aconsequence of spin-dependent electron scattering in thePt layer [12] this spin polarized current leads in turn toa conventional charge current J c and thus a charge ac- cumulation transverse both to H and J s . Accordingly,an iSHE voltage U iSHE appears across the Pt square (seeFig. 1 (b)). As J c ∝ J s × σ the polarity of the iSHE volt-age can be changed by changing the polarization of thespin current σ via the static magnetization of the YIGstripe.In order to increase the dynamic range of the time-resolved iSHE and BLS measurements we supplied amoderately high microwave power to the microstrip line( P = 100 mW). So as to avoid possible caloric effects,driving microwave pulses of 1 µ s duration were appliedwith a repetition rate of 10 µ s. The rise and fall times ofthese pulses were less than 5 ns.First of all, in order to confirm the origin of the volt-age U iSHE we observed, we verified that its polarity wasindeed dependent on the polarity of the magnetizationdirection. To additionally corroborate our results andto rule out parasitic effects (for example electromagneticinduction in the Pt stripe) we also tested a structurewith a nonmagnetic insulator (gadolinium gallium gar-net, GGG) in the place of the YIG. No voltage was de-tected. Thus, we are able to say with confidence that thevoltages we observed were due to the iSHE.The absorbed microwave power (directly proportionalto the intensity of the excited spin waves) and the iSHEvoltage were measured as the applied microwave fre-quency was varied for a magnetizing field H = 1820 Oe(see Fig. 2). The maximum microwave absorption wasrecorded for the spin-wave mode excited at the frequencyof ferromagnetic resonance (FMR) f = 7 GHz. The ob-served FMR linewidth was around 50 MHz (correspond-ing to 2∆ H FMR = 17 Oe). This value is significantlylarger than the FMR linewidth of 0 . µ s duration withoutsignificant distortion. In order to determine the distor-tion level the convolution of the Fourier transformed in-put pulse and the spectrum of the absorbed microwavepower was calculated. Using the inverse Fourier trans-formation of this convolution, we obtained the calculated‘tailing’ of the spin-wave pulse edges, which was smallerthan 20 ns. As the measured iSHE voltage originatesdirectly from the spin-wave amplitude at the YIG/Pt in-terface (Fig. 2) a similar tail was expected to be observedon the U iSHE pulse. However, the real temporal behaviorboth of the spin-wave amplitude and iSHE voltage provedto be much more complex. Note that time-resolved BLSmeasurements performed at f = 7 GHz established thatexcitation delays between different sample points werealways smaller than 1 ns, confirming the excitation ofquasi-uniform FMR rather than spin-wave modes trav-eling away from the microstrip line. Therefore, no timedelays associated with a nonzero spin-wave propagationtime through the Pt area appear in the iSHE signal.The time resolved measurements were performed at f = 7 GHz where the iSHE voltage was maximum.In Fig. 3 the time profiles of the spin-wave and volt-age pulses are compared. The spin-wave intensity (bluedashed line in Fig. 3) increases rapidly when the mi-crowave pulse is applied at t = 0 ns. After some os-cillations corresponding to a nonlinear transition process(common for relatively high spin-wave intensities [16]),an equilibrium value is reached. When the microwavepulse is switched off at t = 1000 ns, the BLS signal de-creases rapidly. The measured iSHE voltage is shown bya red solid line in Fig. 3. It is clear that the rise and falltimes of the iSHE voltage are appreciably longer thanthose corresponding to the spin-wave intensity.In order to unpick the peculiarities of the temporalevolution of the iSHE signal and to relate these to thedynamics of the precessing magnetization, we show thewaveforms plotted on logarithmic scale in Fig. 4. Thedata of Fig. 4 has several features which warrant clarifi-cation:Firstly, the falling slopes of both the iSHE voltage andof the spin-wave intensity are nonexponential (note thelogarithmic scale).Secondly, the measured iSHE voltage rises and de-creases much more slowly than the spin-wave intensity;for example, during the first 50 ns after the driving mi-crowave pulse has been switched off, the iSHE voltagedecreases only by factor of 3 whereas the spin-wave in-tensity falls by factor of 50.Thirdly, for t > τ iSHE and τ SW are very similar; 460 ns and 420 ns re-spectively.These facts can be understood if one assumes that— B L S s i gna l and i S H E v o l t age ( no r m a li z ed , li nea r sc a l e ) iSHEvoltagespin-waveintensity0.01.01.21.40.80.60.40.2 microwavepulse0 200 400 800 1200600 1000-200 Time (ns)
FIG. 3. (Color online) Comparison of the normalized spin-wave signal measured with Brillouin light scattering spec-troscopy (blue curve) and the iSHE voltage (red curve). Themaximum iSHE voltage U max is 60 µ V. -4 -3 -2 -1 B L S s i gna l and i S H E v o l t age ( no r m a li z ed , l og sc a l e ) Time (ns) t iSHE =460ns t SW =420nsiSHEvoltagespin-waveintensitymicrowavepulse FIG. 4. (Color online) Spin-wave intensity and iSHE voltageas a function of time (logarithmic scale). rather than solely the externally driven spin-wavemode—many modes contribute to the iSHE voltage. Inorder to illustrate how such a model fits with the exper-imental data we consider for simplicity only two groupsof modes. The first group corresponds to the FMRdirectly excited by the microstrip line. This group ischaracterized by a large amplitude (determined by theapplied microwave signal) and a high decay rate dueto its strong coupling to the microstrip antenna (seeFig. 2). A second mode group is excited indirectly viatwo-magnon scattering of the first group by defects andinhomogeneities in the YIG film. This well known mech-anism results in the redistribution of energy from theexternally excited uniform mode into dipolar-exchangespin-wave (DESW) modes with wavelengths determinedby sizes of the scatterers (typically 1 µ m in high qual-ity YIG samples) [6, 10, 17–19]. Due to their extremelyshort wavelengths, the DESW modes are entirely de-coupled from the microstrip line and their relaxation(2∆ H DESW = 0 . − . τ SW of 420 ns measured for t > only by the relaxation of the DESW modes.Therefore, for t > τ iSHE and τ SW are ap-proximately the same. The value of the fall time in thisregion of the signal is a direct evidence for the dipolar-exchange nature of these waves: it corresponds to a reso-nance curve linewidth of 2∆ H DESW = 0 .
14 Oe which fitsvery well with literature data on DESW relaxation.We suggest that the DESW modes, in spite of theirsmall amplitude, make a relatively more significant con-tribution to the iSHE voltage than the directly excitedFMR mode due to their localization close to the YIG/Ptinterface. As a result, after the excited spin-wave inten-sity maximum, the iSHE voltage continues to grow (fortimes t <
500 ns). The same effect is visible after themicrowave pulse has been switched off: the iSHE volt-age continues to be generated by the long-lived DESWgroup.The model we propose accounts well for the main fea-tures of the experimental results even in the case that weconsider contributions from only two spin-wave groupsto the iSHE voltage signal (the slowest and the fastestrelaxing). For quantitative analysis of the transition re-gions (times t <
500 ns and 1000 ns < t < t > t <
500 ns and1000 ns < t < t > indirectly excited short-wavelength dipolar-exchange spinwaves participate in spin pumping at the metal-insulatorinterface as well as directly excited uniform precession.The surface localization of the scattered dipolar-exchangemodes means that despite their low amplitude and inco-herent character, they make a substantial contributionto the iSHE voltage signal. In addition, we can concludethat iSHE voltage signal delays in magnetic insulator-nonmagnetic metal structures are dominated by spin-wave dynamics in the insulator, rather than electron dy-namics in the metal.We thank G. E. W. Bauer for the valuable discussions and the Nano+BioCenter, TU Kaiserslautern, for techni-cal support. A. D. K. is grateful for the support of Mag-dalen College, Oxford. B. O. would like to acknowledgethe DFG for support within the Graduiertenkolleg 792. ∗ jungfl[email protected] † Current adress: Materials Science Division and Centerfor Nanoscale Materials, Argonne National Laboratory,Argonne, Illinois 60439, USA[1] Y. Kajiwara, K. Harii, S. Takahashi, J. Ohe, K. Uchida,M. Mizuguchi, H. Umezawa, H. Kawai, K. Ando, K.Takanashi, S. Maekawa, and E. Saitoh, Nature , 262(2010).[2] S. A. Wolf, D. D. Awschalom, R. A. Buhrman, J.M. Daughton, S. von Moln´ar, M. L. Roukes, A. Y.Chtchelkanova, and D. M. Treger, Science , 5546(2001).[3] I. ˇZutic, J. Fabian, and S. Das Sarma, Rev. Mod. Phys. , 2 (2004).[4] J. F. Gregg, Nature Mater. , 798 (2007).[5] K. Uchida, J. Xiao, H. Adachi, J. Ohe, S. Takahashi, J.Ieda, T. Ota, Y. Kajiwara, H. Umezawa, H. Kawai, G.E. W. Bauer, S. Maekawa, and E. Saitoh, Nature Mater. , 894 (2010)[6] A. A. Serga, A. V. Chumak, and B. Hillebrands, J. Phys.D: Appl. Phys. , 264002 (2010).[7] T. Schneider, A. A. Serga, B. Leven, B. Hillebrands, R.L. Stamps, and M. P. Kostylev, Appl. Phys. Lett. ,022505 (2008).[8] A. Khitun, M. Bao, J. Lee, K. Wang, D. W. Lee, and S.Wang, Materials Research (2007).[9] N.P. Stern, D. W. Steuerman, S. Mack, A.C Gossard,and D.D. Awschalom Nature Physics , 843 (2008).[10] A. G. Gurevich and G. A. Melkov, Magnetization Oscil-lations and Waves (CRC, New York, 1996).[11] C. W. Sandweg, Y. Kajiwara, K. Ando, E. Saitoh, andB. Hillebrands Appl. Phys. Lett. , 252504 (2010).[12] J. E. Hirsch, Phys. Rev. Lett. , 1834 (1999).[13] Y. Tserkovnyak, A. Brataas, and G. E. W. Bauer, Phys.Rev. B , 224403 (2002).[14] S. O. Demokritov, B. Hillebrands, and A. N. Slavin,Phys. Rep. , 441 (2001).[15] S. Mizukami, Y. Ando, and T. Miyazaki, Phys. Rev. B , 104413 (2002).[16] V. S. L’vov, Wave Turbulence under Parametric Ex-citations: Applications to Magnetics (Springer, Berlin,1994).[17] M. Sparks,
Ferromagnetic Relaxation Theory (McGraw-Hill, New York, 1964).[18] G. A. Melkov, V. I. Vasyuchka, Yu. V. Kobljanskyj, andA. N. Slavin, Phys. Rev. B , 224407 (2004).[19] G. A. Melkov, A. D. Dzyapko, A. V. Chumak, and A. N.Slavin, J. Exp. and Theor. Phys.99