Académique Documents
Professionnel Documents
Culture Documents
he spin Hall effect (SHE) was predicted 40 years ago1,2. Theorists Dyakonov and Perel proposed that an unpolarized electrical current should lead to a transverse spin current in systems with the relativistic spinorbit coupling. In their picture, spinorbit coupling enters SHE via the Mott scattering of electrons on unpolarized impurities, which results in spatial separation of electrons with opposite spins. The effect has a Hall symmetry, because the polarization axis of the spins is perpendicular to the plane of the transverse spin current and the driving longitudinal electrical current. Concepts for the experimental detection of the phenomenon were introduced by Hirsch3 and Zhang 4 almost 30years after the original theoretical
a
2 150 ns (a.u.) 1 0 1
b
2 1
Reectivity (a.u.) 2 3 4 5
100
50 Position (m)
50
100
Figure 1 | Observation of SHE by the magneto-optical Kerr microscope. a,b, Two-dimensional images of spin polarization density ns (a) and reectivity (b), for the unstrained GaAs sample measured at temperature 30K and applied driving electric eld 10mVm1. Arbitrary units are used for ns as the measured Kerr rotation signal is not calibrated to the corrresponding spin density for this measurement. Figure reproduced with permission from ref.21, 2004 AAAS.
work. Hirsch proposed a device in which a spin current is generated by SHE in one part and injected into another part where it is detected by the inverse spin Hall effect (iSHE). In iSHE, the spin current generates a transverse current of charge and when accumulated at the edges of the sample the charge can be detected electrically 3. Relying on both SHE and iSHE simultaneously for observing the phenomenon turned out to be experimentally challenging, and the method was realized only recently 5. The proposal3 that SHE has an inverse counterpart has, nevertheless, played a key role in establishing the basic physics of the phenomenon and in utilizing the effect as a tool for both the electrical injection and electrical detection of spin currents in non-magnetic materials. Zhang 4 suggested that the edge spin accumulation produced by SHE could be detected electrically using an attached ferromagnetic probe6. The method is based on measuring the dependence of the electrochemical potential at the detection ferromagnetic electrode on the relative orientation of the magnetization of the electrode, and the accumulated spins in the non-magnetic system underneath it. It took several years to demonstrate the viability of this method7. Nevertheless, in a broader context, the idea of connecting SHE with the more mature field, which utilizes ferromagnets for injection and detection of spins in non-magnetic systems, has fuelled numerous important studies of spin Hall devices. The experimental discovery of SHE was prompted by theoretical work that approached the SHE physics from a different angle. Inspired by studies of the intrinsic nature of the closely related anomalous Hall effect in ferromagnets8,9, Murakami et al.10 and Sinovaetal.11 predicted that a spin-dependent transverse deflection of electrons in non-magnetic systems can originate directly from the relativistic band structure of the conductor without involving the Mott scattering. The intrinsic SHE proposal triggered an intense theoretical debate, which is summarized in several review articles1220. Unlike Hirsch3 and Zhang 4 who considered the extrinsic, scattering induced SHE1 and electrical detection schemes designed for metals, the intrinsic SHE proposals focused on semiconductors and suggested that the optical activity of these materials be utilized for detecting SHE. In particular, the circularly polarized electroluminescence was suggested in ref.10 and the magneto-optical Kerr effect in refs10,11. These methods were used in the first measurements of the SHE phenomenon. Kato et al.21 employed a magneto-optical Kerr microscope to scan the spin polarization across the channel (Fig.1), whereas Wunderlich etal.22 used coplanar pn diodes to detect circularly polarized electroluminescence at opposite edges of the spin Hall channel (Fig.2; ref.23). Remarkably, Wunderlichetal. ascribed the observed signal to the intrinsic SHE whereas Kato etal. to the extrinsic SHE.
Institute of Physics ASCR, v.v.i., Cukrovarnick 10, 162 53 Praha 6, Czech Republic, 2School of Physics and Astronomy, University of Nottingham, Nottingham NG7 2RD, UK, 3Hitachi Cambridge Laboratory, Cambridge CB3 0HE, UK. *e-mail: jungw@fzu.cz
1
382
n channel ILED
2DHG n p n
Ip
ILED
ILED
e
Polarization (%) 2 1 0 1 2 1.500 1.505 1.510 Energy (eV) 1.515 1.520
b Ip
Down
Figure 2 | Observation of SHE by the circularly polarized electroluminescence of coplanar pn diodes. a, Schematic of the lateral pn junction with the channel current Ip and the diode current ILED for detecting spin accumulation. b, Light emission from the pn junction recorded by a charged-coupled device camera. The black dashed line shows the position of the pn junction. c, Electron microscope image of the microdevice with symmetrically placed pndiodes at both edges of the 2D hole gas channel. d,e, Emitted light polarization of recombined light in the pn junction for the channel and diode current ow indicated by arrows in (c). Figure reproduced with permission from ref.23, 2005 APS.
NATURE MATERIALS | VOL 11 | MAY 2012 | www.nature.com/naturematerials
383
x (m) ^ x ^ y
3 ^ [100] z
0 y (m)
Figure 3 | Observation of iSHE using the two-colour optical pump-and-probe technique. a, Illustration of orthogonally polarized and 2 pulses producing a pure spin current (double headed arrow) along the beam polarization direction ( x). The charge current due to iSHE (curved arrows) along leads to electron accumulation near one edge of the illuminated region. b, Measured charge accumulation due to iSHE. N is the excess charge. Figure reproduced with permission from ref.31, 2006 APS.
384
NATURE MATERIALS | VOL 11 | MAY 2012 | www.nature.com/naturematerials
IPH
1 0 1 x =1 m
0.8 0.6 IPH (A) 0.4 0.2 0.0 0.5 0.0 1.0
10 0.0
3 0.5 VG (V)
Figure 4 | iSHE-based transistor. a, Schematic of the spin-injection Hall effect measurement set-up with optically injected spin-polarized electrical current propagating through the Hall bar and corresponding experimental Hall effect signals at crosses H1 and H2. The Hall resistances, RH=VH/IPH, where VH is the Hall voltage and IPH is the photocurrent for the two opposite helicities of the incident light are plotted as a function of the focused light spot position, that is, of the position of the injection point. VB is the pn junction bias voltage, + and are the helicity of the polarized light. IPH is independent of the helicity of the incident light and varies only weakly with the light spot position. b, Same as (a) for iSHE in which the electrical current is closed before the rst detecting Hall cross H1. c, Schematic of the spin Hall transistor and experimental Hall signals as a function of the gate voltage (VG) at a Hall cross placed behind the gate electrode for two light spot positions with a relative shift of 1m. The red and dashed black lines correspond to measurements with the relative shift of the laser spot of 1m. Figure reproduced with permission from ref.38, 2010 AAAS.
thereby populating the 2D electron gas. The 1.5- and 10-m-wide channels in the experimental devices22,23 were formed in the unetched part of the epilayer with the 2D hole gas. The etched-off regions of the wafer outside the channel make the channel edges act as coplanar pn diodes with an effective recombination width of ~100nm. The measured circular polarization of the electroluminescence peak (Fig.2d,e) reverses sign when the sign of the SHE-generating electrical current along the 2D hole channel is reversed. Also consistent with the SHE phenomenology, opposite polarizations were detected at opposite edges of the 2D hole gas channel. The amplitude of the measured SHE signals (at 4K) reached ~1% in these experiments. The 2D hole gas studied in refs 22,23 was in the strong spinorbit coupling regime, so/4, which favours the intrinsic SHE mechanism. A quantitative microscopic description of the measured edgespin-accumulation signal was developed and further experimentally tested23. Theoretical analysis pointed out that the length scale of the edge spin accumulation is defined in the strong spinorbit coupling regime by the spinorbit precession length Lso = Fso, where so= /so is the precession time of the spin in the internal spin orbit field and F is the Fermi velocity. With increasing strength of the spinorbit coupling, the edge-spin-accumulation region narrows down and, simultaneously, the amplitude of the spin polarization increases. For the experimental parameters of the 2D hole gas, Lso ~10nm and the calculated amplitude of the edge spin polarization was 8%, in good agreement with the 1% polarization of the measured electroluminescence signal, which was averaged over the ~100 nm sensitivity range of the coplanar light-emitting diode. Measurements in a device with a 10m wide 2D hole gas channel confirmed the theoretical expectation that the SHE edge spin accumulation is independent of the channel width (for widths larger than Lso)23. Subsequent magneto-optical measurements of SHE in the n-GaAs 3D epilayers have demonstrated experimentally that the SHE-induced spin accumulation is due to a transverse spin current, which can drive spin polarization tens of micrometres into a region in which there is minimal electric field25. The work proved experimentally that SHE can be used as a source of spin current generated in a non-magnetic system. A systematic doping dependence of the SHE angle was studied in n-GaAs 3D epilayers with electron densities nin
NATURE MATERIALS | VOL 11 | MAY 2012 | www.nature.com/naturematerials
the range 1.810163.31017cm3 and the results were found to be consistent with theoretical predictions for the extrinsic SHE30. The measured SHE angles of ~51045103 increased with increasing doping with a tendency to saturate at the high doping end of the studied set of samples at a value corresponding to ~1% edge spin polarization. It was concluded from this systematic analysis that the spin accumulation is reduced by an enhanced spin relaxation due to the DyakonovPerel spin-relaxation mechanism, whereas the spin current induced by SHE is enhanced with increasing n (ref.30). The spin Hall effect was also observed in other semiconductor systems including n-ZnSe 3D epilayers26, and InGaN/GaN superlattices27. The optically generated and detected iSHE was observed in an intrinsic GaAs (at 80K) using a two-colour optical coherence control technique with orthogonally polarized 1,430nm and 715nm fs pulses31. The method, shown in Fig.3a, is based on quantum interference between absorption pathways for one- and two-photon absorption connecting the same initial and final states, and allows for a pure spin-source current to be generated in an intrinsic semiconductor with no drift currents induced by an applied electric field, but rather with ballistic carriers decelerated by scattering. For generating a pure spin-current, a coherent pulse centred at frequency with phase is normally incident along the z direction and linearly polarized along the x direction that can be arbitrary with respect to the crystal axes as the effect is not strongly sensitive to the crystal orientation. A co-propagating 2 pulse with phase 2 is linearly polarized along the orthogonal direction. Excited spin-up electrons are polarized along z and move preferentially in one direction along x , whereas spin-down electrons move in the opposite direction. Together they generate a spin current jsz cos(), where =22. This pure spin-current is the source of a transverse charge current arising due to iSHE. The resulting charge accumulation was measured optically by differential transmission, that is, the transmission with and without pump beams, of a linearly polarized probe pulse. Consistent with the iSHE phenomenology, the excess charge (N) on one side and the deficit on the other side of the sample, shown in Fig.3b, was observed along the direction perpendicular to the driving spin current. The ballistic nature of transport in these experiments in intrinsic GaAs has been fully exploited in recent femtosecond time-resolved
385
V+
I+
I+
V+
c
0.1
d
LFM = 2 m 1 2 0.4 VII (m) RSHE Sin 0.2 B(T) 0.0 0.2 0.4
RSHE (m)
0.0
0 B M
0 VII (m) 0 1
2 0
0.1
2 4 2 0 B(T) 2 4 2 B(T) 3
Figure 5 | iSHE detection in a metal. a,b, Devices with electrical spin injection used for spin detection by iSHE (a) and by the non-local spin-valve effect (b). Arrows indicate the direction of current. The light-coloured electrodes in the micrographs are CoFe ferromagnets of width 400nm (FM1) and 250nm (FM2). The dark-coloured Hall cross is made of Al. c, Spin Hall resistance (open symbols) and magnetization angle (black line) measured as a function of the applied out-of-plane magnetic eld. Inset: Magnetization direction of the FM electrode relative to the substrate. d, Non-local spin-valve resistance V/I measured as a function of the out-of-plane magnetic eld (up to 3.25T in main gure and up to 0.5T in inset). Arrows show relative orientations of the magnetization of the injection and detection electrodes. Open symbols show the antiparallel conguration and lled symbols show the parallel conguration. LFM is the distance between the two ferromagnetic electrodes. Figure reproduced from ref.34, 2006 NPG.
measurements32. These measurements allowed the momentum scattering time 0.45ps to be inferred, and with a much shorter time delay of the probe pulses, the transverse charge current could be observed in real time. The measurements showed that the charge current was generated well before the first scattering event, providing a direct demonstration of the intrinsic iSHE. A traditional way of generating spin-polarized photo-carriers in semiconductors is by absorption of circularly polarized light 63. Because of the optical selection rules, the out-of-plane spin polarization of photo-carriers is determined in this technique by the helicity and degree of the circular polarization of vertically incident light. Wunderlich etal.37,38 used the same type of a lateral pn diode they had used previously 22,23 to utilize optical spin injection by a circularly polarized laser beam for observing the iSHE and for employing the effect in experimental opto-spintronic and spin-transistor devices. In their original SHE measurements22,23, the pn junctions were fabricated along the edges of the 2D hole channel and under forward bias could detect the spin state of recombining electrons and holes through polarized electroluminescence. In their later work37,38, on the other hand, the spin Hall channel was fabricated in the etched part of the epilayer with the 2D electron gas, the channel was oriented perpendicular to the pn junction, and the diode was under zero or reverse bias, operating as a photocell as shown in Fig.4. The optical activity of the lateral diode confined to a submicro metre depletion region, combined with a focused (~1m) laser beam, allowed for a well-localized injection of spin-polarized photo-electrons into
386
the planar 2D electron gas channel. The Hall signals were detected electrically on multiple Hall-crosses patterned along the channel. Two regimes of operation of the device were distinguished: One was referred to as the spin-injection Hall effect regime, in which the reverse-bias charge current is drained behind the Hall crosses at the opposite end of the channel from the pn junction injection point (Fig.4a). The other regime corresponds to the iSHE measurement, because in this case the charge current is drained before the Hall crosses, allowing only the pure spin-current to diffuse further in the channel (Fig.4b). In both cases the measured transverse electrical signals were consistent with the spin Hall phenomenology 37,38. The sign of the voltage was opposite for opposite helicities of the incident light, that is, opposite spin polarizations of injected photo-electrons. Moreover, the amplitude of the electrical signals was found to depend linearly on the degree of circular polarization of the light, rendering the device an electrical polarimeter 37. The electrical signals were observable over a wide temperature range with spin Hall angles of 103102. The measured 2D electron gas was in the weak spinorbit coupling regime, so/~101, and the measured data were consistent with the extrinsic spin Hall mechanism37. A distinct feature of the iSHE experiments in the 2D electron gas is the observed spin precession due to internal Rashba and Dresselhaus spinorbit fields37,38. As the spin diffusion length scales approximately 38 as Lso2/w, it was possible to observe a few spin precessions in channels of a width w=1m for Lso~1m of the studied 2D electron gas. The corresponding oscillations of the spin Hall
NATURE MATERIALS | VOL 11 | MAY 2012 | www.nature.com/naturematerials
Valenzuela and Tinkham34 performed experiments (at 4K) in which a spin current injected from a ferromagnetic electrode into a nonmagnetic metal strip was detected by iSHE and by the non-local spin-valve effect using a ferromagnetic probe electrode. The electrical injection of the spin current results from a non-equilibrium spatially dependent splitting of spin-up and spin-down electrochemical potentials on either side of the electrically biased ferromagnet/ normal metal contact 6. In the device shown in Fig.5, the ferromagnetic CoFe electrode FM1 is used to inject spin-polarized electrons through an Al2O3 tunnel barrier into an Al channel. The tunnel barrier is important because it enhances the polarization of the injected electrons. In Fig.5b, the electrical current is directed from the injection electrode FM1 to the left, away from the ferromagnetic CoFe electrode FM2, causing a pure spin-current to diffuse towards FM2. The non-local spin-valve voltage is then measured between the ferromagnetic probe FM2 and the right side of the Al strip. The CoFe electrodes are magnetized in the plane at zero magnetic field owing to the strong thin-film shape anisotropy. The large positive nonlocal spin-valve voltage at zero magnetic field shown in Fig.5d, corresponds to parallel magnetizations of the injection and detection CoFe electrodes, that is, parallel spin polarization in the Al channel with the magnetization in the FM2 probe above it. The large negative spin-valve voltage at zero field corresponds to the antiparallel configuration. At weak out-of-plane magnetic fields the polarization of the ferromagnetic electrodes remains in-plane and the measured spin-valve signal shows the in-plane component of the injected spins in the Al channel, which precesses owing to the Hanle effect. At large
a
0.1
0.05 Vc /l (m) RSHE H 0.05 M 0.1 200 77 K 100 0 0 H (mT) 100 200 0.2 200 le Vc Vs /l (m) H
Figure 6 | iSHE and SHE detection in a metal. a, The change in Hall resistance RSHE= Vc / I due to iSHE at 77K. The black and grey curves show measurements for the two opposite sweeps of the magnetic eld. b, Spin-accumulation signal Vs / I generated by SHE at 77K. Insets: measurement set-up. NiFe, Cu and Pt are in grey, pink and yellow, respectively. 0H is the applied magnetic eld, M is the magnetization of NiFe and Ie is the applied electrical current. Figure reproduced with permission from ref.7 2007 APS.
NATURE MATERIALS | VOL 11 | MAY 2012 | www.nature.com/naturematerials
387
Brne et al.5 performed a non-local electrical measurement in nanoscale H-shaped structures built on high-mobility HgTe/ (Hg,Cd)Te quantum wells with a top-gate electrode. The experiment was in the spirit of the original proposal of Hirsch3 for simultaneously employing SHE and iSHE to detect the phenomenon electrically. The geometry of the microdevice64 and the measured iSHE signals are shown in Fig.8a. The idea behind these transport measurements is as follows. When an electric current flows in one of the legs of the H-bar structure, a transverse spin current due to SHE is induced in the connecting part. Subsequently, this spin current produces, owing to iSHE, a non-local voltage difference in the opposite leg of the H-bar structure, which can be measured by a voltmeter. Sweeping the gate voltage in the sample allowed the strength of the Rashba spinorbit splitting to be varied by both the electrical field across the quantum well and the Fermi level position in the quantum well. In the sample it was possible to electrically tune the carrier density from strongly n-type, through insulating, down to a
a
Microwave
b
FMR = 90
c
= 90 H dV(H)/dH (a.u.)
dI(H)/dH (a.u.)
Ni81Fe19 /Pt
10 V/T
0 Ni81Fe19
Figure 7 | Observation of iSHE in a metal device with spin injection from a ferromagnet by the ferromagnetic resonance spin pumping. a, Schematic of the NiFe/Pt sample system used in the study and of the spin-pumping effect and iSHE. Jc is the charge current, Js is the spin current and is the spin orientation of injected electrons. b, Magnetic-eld dependence of the ferromagnetic resonance (FMR) signal for the NiFe/Pt bilayer lm and a Ni lm. I is the microwave absorption intensity. c, Magnetic-eld dependence of dV(H)/dH for the NiFe/Pt sample. V is the electrical-potential difference between the electrodes on the Pt layer. Figure reproduced with permission from ref.33, 2006 AIP.
388
NATURE MATERIALS | VOL 11 | MAY 2012 | www.nature.com/naturematerials
R21,36 ()
300 V 200
c
0 VNL (V)
d 1.0
0.5 VH (V) 0.0 5 10 15
100
0.5 1.0 50
1.5
1.0
0.5
0 Vg* (V)
0.5
1.0
1.5
2.0
20 50
25
0 Bx (mT)
25
50
25
0 Bx (mT)
25
50
Figure 8 | Electrical spin Hall devices in semiconductors. a, Spin injection by SHE and detection by iSHE in a gated H-shaped device. Inset: schemtic showing the measurement set-up for current injection (I) and voltage probes (V). The black curve shows the non-local iSHE resistance signal. The blue curve indicates the residual voltage owing to current spreading. bd, Semiconductor iSHE device with electrical modulation of the spin signal. Schematic showing the experimental set-up (b). Experimental non-local spin-valve (VNL) and iSHE (VH) signals in the in-plane eld Bx measured at constant spininjection bias current IB=300A and at three different drift currents ID depicted in b (c,d). Figures reproduced from: a, ref.5 2010NPG; bd, ref.61.
was detected by iSHE and the spin polarization was simultaneously measured using an additional Fe electrode. The spins in the channel were manipulated by the Hanle spin precession induced by an applied magnetic field and by a drift induced by an applied electrical bias65. As shown in Fig.8bd, the output iSHE and non-local spinvalve signals are suppressed or enhanced depending on the electrical bias, that is, the device acts as an electrically controlled spin modulator. Qualitatively, the modulation of the spin signal can be explained by a shift of the injected spin polarization profile from the injection electrode in the direction towards the detection electrodes in the case of the positive drift current or away from the detectors in the case of the negative drift. Electrical tuning of the spin signal in a semiconductor has also been recently demonstrated by Ando and colleagues62. In this experiment, spins were injected from NiFe into GaAs through a Schottky contact using ferromagnetic resonance spin pumping. Tuning of the spin-pumping efficiency was achieved by applying a bias voltage across the NiFe/GaAs Schottky barrier and interpreted as a consequence of a suppressed or enhanced spin coupling across the interface. The ferromagnet/semiconductor spin-pumping experiments in ref. 62 were also performed on samples with an ohmic contact between NiFe and GaAs. The measurements demonstrated that the resistance mismatch problem in ohmic metal/semiconductor spininjection devices can be circumvented by using the ferromagnetic resonance spin-pumping technique.
Outlook
Experiments in spin Hall devices performed so far have established the basic physics of SHE and iSHE, showed that the intrinsic and extrinsic mechanisms can contribute to the phenomenon, and that SHE and iSHE are universal to metal and semiconductor systems (3D and 2D) with spinorbit coupling. As the strength of the spinorbit coupling increases, the spin current that arises from SHE increases. At the same time, however, the spin lifetime decreases. Finding the optimal compromise between these two counteracting effects of the spinorbit coupling remains an important open
NATURE MATERIALS | VOL 11 | MAY 2012 | www.nature.com/naturematerials
problem in the field. Research directions recently initiated5,37,38 are among the experiments that may contribute to a better understanding of this problem. They showed that spin Hall devices can operate in the ballistic regime, that is, when the typical dimensions of the device are smaller than the scattering mean free path5. Also, for the diffusive spin Hall devices it was pointed out that there are regimes in which spin coherence can be strongly enhanced38. One regime is when the width of the channel is comparable to or smaller than the precession length in the internal spinorbit field66. In the other regime, the width of the channel is not relevant, and enhanced spin coherence occurs as a result of the single-particle transport analogue of the spin helix state67,66 realized at 2D Rashba and Dresselhaus spinorbit fields of equal or similar strengths. The utility of SHE and iSHE as an electrical spin injector and detector in non-magnetic systems can allow a variety of new spintronic functionalities to be explored. One of the recently proposed areas is opto-spintronics. It was shown, for example, that iSHE acts as a direct electrical detection tool for the polarization of light 37,56, and because of the universality of the spin Hall phenomenon, the approach can be implemented in semiconductors active over a wide range of wavelengths. Other proposed applications include spinphotovoltaic cells, switches, invertors and interconnects. Several types of devices combining electrical spin detection by iSHE with electrical spin manipulation have been recently demonstrated, which makes spin transistors another potentially fruitful area of spin Hall research. Some devices are reported to act as a spin analogue of a field-effect transistor 5,38,62. The gate electrode in these devices can control the spinorbit coupling 5,38 or the interfacial spin interactions62 by changing the electric field strength across the film or the Fermi level position. The device reported in ref. 61 is different as it uses an additional drift current to control the output spin signal. Finally we recall the research direction suggested by recent reports of the SHE-controlled magnetization dynamics4951. These are the pure spin-current analogues of the spin-transfer-torque phenomena that have for decades played a key role in basic spintronics research and in bringing spintronics to commercial applications68.
389
References
Acknowledgements
We acknowledge support from EU grants ERC Advanced Grant 268066-0MSPIN and FP7-215368 SemiSpinNet, and from Czech Republic grant AV0Z10100521 Praemium Academiae.
Additional information
The authors declare no competing financial interests. Reprints and permissions information is available online at http://www.nature.com/reprints. Correspondence should be addressed to T.J.
NATURE MATERIALS | VOL 11 | MAY 2012 | www.nature.com/naturematerials