PHYSICAL REVIEW B 74, 245319 2006

Tunable spin Hall effect by Stern-Gerlach diffraction

Jun-Qiang Lu and X.-G. Zhang
Center for Nanophase Materials Sciences, and Computer Science and Mathematics Division, Oak Ridge National Laboratory, Oak Ridge, Tennessee 37831, USA

Sokrates T. Pantelides
Department of Physics and Astronomy, Vanderbilt University, Nashville, Tennessee 37235, USA and Oak Ridge National Laboratory, Oak Ridge, Tennessee 37831, USA Received 6 November 2006; published 18 December 2006 We propose an effect based on simultaneous real- and reciprocal-space magnetic inhomogeneities, combining features of the Stern-Gerlach and both the intrinsic and extrinsic spin Hall effects. The known difculties of directly observing the spin Hall effect are circumvented as spin currents generated from the effect are well-dened, dissipative, and detectable. Simulations of a specic system are used to illustrate wide tunability, allowing formation of a periodic spin lattice distinct from the charge lattice, selective polarization ipping, or spin current detection. Similar effects can be produced by photons and neutrons. DOI: 10.1103/PhysRevB.74.245319 PACS number s : 72.25.Dc, 71.70.Ej, 85.75. d

I. INTRODUCTION

The pursuit of spintronics ultimately depends on our ability to steer spin currents and detect or ip their polarization. The 1922 Stern-Gerlach SG experiment1 rst demonstrated that electron currents can be steered by the Zeeman force, B , caused by a spatially inhomogeneous magnetic F= is the electron magnetic moment. The eld B r , where effect was recently observed for electrons in a semiconductor nanostructure, where the spatially inhomogeneous magnetic eld was generated by a ferromagnetic layer.2 In a similar vein, the SG effect has been demonstrated for a beam of light passing through a medium where the photons morph into polaritons and experience a spatially inhomogeneous effective magnetic eld.3 In the last few years, there has been intense interest in the manipulation of spins without the need for ferromagnetic materials or the application of external magnetic elds, relying solely on spin-orbit SO coupling and external electric elds.410 An SG lter was recently proposed5 in which an electron current gets split by a Zeeman force from an effective magnetic eld produced by SO interactions through the Rashba Hamiltonian11 and made inhomogeneous by the application of suitable external electric elds. The spin Hall effect SHE is another altogether different way to manipulate spins without relying on the Zeeman force. It exploits the momentum dependence of the SO effective magnetic eld.12 In the extrinsic SHE6 one relies on impurities with strong SO coupling to produce skew scattering of electron currents. In the intrinsic SHE,710 one enhances SO coupling by an external electric eld applied to a two-dimensional electron system 2DES . Thus unlike the SG effect that relies on a magnetic eld that is inhomogeneous in real space, the SHE is caused by an effective magnetic eld that is inhomogeneous in reciprocal space. The experimental observation of the SHE has proved to be quite challenging.1315 Kato et al.13 reported an unambiguous signature of SHE, namely the accumulation of opposite spins in directions transverse to the electron current, in
1098-0121/2006/74 24 /245319 5

a semiconductor two-dimensional electron system 2DES , but the origins of the effect are not completely clear. The authors attributed the observations to the extrinsic SHE, though alternative opinions have since been expressed.10 Observations of the intrinsic SHE remain an open issue.16 A key difculty is the fact that in a 2DES with strong SO coupling, spin is not a good quantum number so that spin currents are not well dened. In this paper, we propose to combine the principles that underlie the SG experiment and the SHE to arrive at an interesting phenomenon that exploits a designed real-space inhomogeneity with the intrinsic reciprocal-space dependence of the effective SO magnetic eld in a 2DES. The effect avoids both the randomness of impurity distribution that is inherent in the extrinsic SHE and the intrinsic-SHE difculties of dening spin currents in a SO-coupled region: the SO-coupled scatterers are conned in a nanoscale diffraction grating in a 2DES, offering wide tunability of end results. The diffraction grating can be implemented either by an array of SO-coupled quantum dots and a uniform external electric eld perpendicular to the 2DES or by a uniform layer of SO-coupled material and a patterned gate to produce an electric-eld-induced grating. Unlike the intrinsic SHE, the spin currents generated by such a setup are well dened outside the SO region. The transverse charge currents and spin currents are real, dissipative, and detectable. The grating acts as the SG lter of Ref. 5, but we can now control the number of emerging beams. In addition, the diffraction design takes advantage of the SHE-like spin-dependent velocities arising from the momentum-dependent effective magnetic eld. The proposed tunable SHE TSHE can be used to produce a variety of interesting effects that allow the generation, manipulation, and/or detection of spin currents and spin-polarized electrical currents, including polarization ipping. The underlying physics is similar to what one would get by a periodic array of SO-coupled impurities, i.e., in some sense the effect is akin to the extrinsic SHE, but their spatial arrangement in a grating brings into action the SG force. In addition, by doing away with both SO impurities and quantum dots QDs , the effect can be generated entirely
2006 The American Physical Society

245319-1

LU, ZHANG, AND PANTELIDES

PHYSICAL REVIEW B 74, 245319 2006

by a patterned external electric eld, which imposes the SG force on the intrinsic SHE. We demonstrate the TSHE with numerical simulations for a model system. We also show that the effect is quite general and can be produced by a general class of Hamiltonians that have odd powers both of spin and momentum. It can in principle be realized in diffraction phenomena of photons, neutrons, and even ion beams.
II. TUNABLE SPIN HALL EFFECT IN 2DES

We start with the theory for a semiconductor 2DES. In the presence of SO coupling, the electrons or holes can be described by the Rashba Hamiltonian11 H = p2/2m + p e = p2/2m + p e , 1

where m is the effective mass, p = i , is the Pauli spin matrix vector, is a coupling constant, and e is a unit vector in the direction of the external electric eld, which controls the value of . The effective magnetic eld is Beff = / B p e ,12 where B is the Bohr magneton. In Ref. 5, the authors explored the effect of an inhomogeneous external electric eld, which makes a function of r, giving rise to a Zeeman force that splits electron currents SG lter . The effective magnetic eld, however, is intrinsically also a function of p, which produces a spin-dependent transverse velocity given by v=
pH

FIG. 1. The schematics of the basic concept of the tunable spin Hall effect in a two-dimensional electron system: a spin-polarized electron beam is asymmetrically diffracted by an array of quantum dots with Rashba spin-orbit coupling. The shaded region is the quantum dot array, which has been modeled as a simple oscillatory function in the simulations.

p + m

pB

eff

p + m

e.

To see the effects of both a real- and reciprocal-space dependence of Beff r , p , consider a spin-polarized current in the x direction, being injected into a 2DES in an x-z plane, at a particular energy, typically the Fermi energy EF, with crystal momentum p = kF and velocity = kF / m, as shown in Fig. 1. In the presence of an array of SO-coupled QDs shaded region in the z direction with width a and spacing b, there will be diffracted beams at angles given by sin n = 2 n / kFb, with transverse momenta pzn = 2n / b. The distribution of diffraction angles are symmetric between left +z and right z as they are resulting from the periodic z . The diffracted beams, however, acquire additional spindependent transverse velocities:
vzn =

H/ pz = pzn /m

The two terms arise from the real-space and reciprocal-space inhomogeneities of the effective magnetic eld, respectively. The second term, arising from SO coupling, breaks the spin symmetry in the diffracted beams. Equation 3 displays the main features of the TSHE and reveals the options for control: the periodicity of the SO gratingeither an array of SO-coupled dots, or simply a patterned external electric eld applied to a SO-coupled regioncontrols the number of diffracted beams and the strength of the electric eld controls the magnitude of the spin-dependent transverse velocities. One might even envision a superposed gradient on the external electric eld to produce additional variety of outcomes. In addition, by patterning a periodic gate with a very small period, one can apply electric elds in a way that the period

of the effective grating can be a multiple of the physical period of the patterned gate. By a suitable choice of the grating parameters, several distinct objectives can be pursued. At one extreme, one can rely on the interference of diffracted beams to produce a two-dimensional periodic lattice of maxima and minima. The charge lattice, dened by the maxima or minima of the total electron density and the spin lattice, dened by the maxima or minima of the spin polarization, depend on the polarization of the incoming beam and are different from each other. They offer interesting opportunities for applications. At the other extreme, one can choose parameters so that there is only one left-diffracted beam and one right-diffracted beam, in addition to the principal beam that goes through unhindered. The asymmetry produces a transverse charge or spin current. The transverse currents can have spinpolarization opposite to the injected one, accomplishing controllable spin ipping. More importantly, by measuring the net transverse charge current from the diffracted beam, one can deduce the polarization of the principal beam, opening the possibility of a spin-polarization detector. We have applied the above theory to a potentially realizable device shown schematically in Fig. 1. A periodic array of SO-coupled QDs produce a periodic x , z whose overall amplitude is controlled by an external electric eld along the y direction. The complete effective-mass Hamiltonian of the 2DES has the form

H=

p2 + 2m

x,z

p ey L,

i 2

x,z ey 4

+ Vc x,z + Vs x,z

245319-2

TUNABLE SPIN HALL EFFECT BY STERN-GERLACH

PHYSICAL REVIEW B 74, 245319 2006

FIG. 2. Color online Spatial distributions of the wave densities x , z in the region near the quantum dot array, for an injected beam spin polarized in the a x or b z directions.

where the second term contains the Rashba SO coupling and the third term is needed for current conservation. Vc is the spin-independent part of the crystal potential, which includes the connement potential of QDs and the disordered impurity scattering potential. The term Vs is the SO coupling potential from Vc, where L is the orbital angular momentum of electron.6 We have modeled the effect of the diffraction grating on by an analytical form x , z that is oscillatory in the z direction with period b: x,z = k0/m 1 + cos kbz 0 if a if x x 0; 0, 5 where k0 sets the SO coupling strength, and kb = 2 / b. In pursuing numerical solutions, we checked the effect of the potentials Vc and Vs. We found that the connement potential plays a secondary role, causing a superposed diffraction pattern. Disordered impurity scattering, on the other hand, if strong enough, can destroy the diffraction patterns altogether, producing only skew scattering as in Ref. 6. In order to avoid the connement potential and the disordered scattering, the diffraction grating can be produced by a patterned external electric eld, using a SO-coupled layer that, to the extent that is possible, contains no impurities. In order to focus on the TSHE, we do not include any impurity scattering and set the potentials Vc and Vs to zero.
III. APPLICATION TO INGAAS/INALAS 2DES

a or x

jected beam is spin polarized in the x direction, the diffraction pattern, shown in Fig. 2 a , is clearly spatially asymmetric. In contrast, if the injected beam is spin polarized in the z direction, as shown in Fig. 2 b , the diffraction pattern is symmetric because of the absence of asymmetric transverse term in the velocity, Eq. 3 . The spin-dependent transverse charge current j = jL jR is plotted in Fig. 3 a . For 0 a 20 nm, if the injected beam is polarized in the x direction, then j is positive, which means that the left-diffracted beam has a larger ux than the right-diffracted beam. j is always 0 if the injected beam is polarized in the z direction. The ux of the principal beam, j0, and its spin polarization, P0, are shown in Fig. 3 b . Here, P0 = j j / j0, and are relative to the spin direction of 0 0 the injected beam. j0 is always greater than 0.95. The polar-

In order to vividly view the TSHE in the 2DES, we rst present the results for InGaAs/ InAlAs 2DES,5,17,18 with a choice of parameters that produces only two diffracted beams in addition to the principal beam that goes through unhindered. The numerical solutions are obtained by expanding the electron wave functions in a plane-wave basis and solving the diffraction problem using the LippmannSchwinger equation. The spatial wave density distribution x , z = x , z 2 in the region near the QD array is shown in Fig. 2. We take the effective mass of the electrons m = 0.05me, the Fermi wave vector of the injected electrons kF = 0.353 nm1.17,18 The QDs, with the SO coupling strength k0 = 0.05kF,17 width a =15 nm , and period b =20 nm , are drawn schematically in the middle of Fig. 2 a . If the in-

FIG. 3. Color online a The transverse charge current j, b the principle beam j0 and its polarization P0 for an injected beam spin polarized in the x, y, or z directions, as a function of the width of the quantum dot array a. Inset: the angle of the spin precession after passing through the array for an injected beam spin polarized in the x or y direction.

245319-3

LU, ZHANG, AND PANTELIDES

PHYSICAL REVIEW B 74, 245319 2006

FIG. 4. Color online Spatial distributions of the x component of spin polarization, Px x , z , in the region near the quantum dot array, for an injected beam spin polarized in the a x or b z directions.

ization P0 is exactly 1.0 if the injected beam is spin-polarized in the z direction. If the injected beam is spin polarized in the x direction, P0 decreases with a, but it is still above 0.85 at a = 15 nm, the width corresponding to the maximum j in Fig. 3 a . Exploiting the transverse charge current in combination with the preservation of the spin polarization of the principal beam, this effect can be the basis for a spinpolarization detector design. The nonzero j in Fig. 3 a for the injected beam polarized in the y direction is not anticipated by Eq. 3 . However, both this effect and the changes of P0 in Fig. 3 b can be explained by spin precession due to the effective magnetic eld, when electrons travel through the QD array. For the principal beam, the electron momentum is always in the x direction, yielding an effective magnetic eld along the z eff direction, Bz = px / . There is no spin precession if the injected beam is spin polarized in the z direction. Thus P0 remains exactly 1.0 in this case, as shown in Fig. 3 b . When the injected beam is spin polarized in either the x or the y direction, the spin precesses around the z axis, thus away from its original direction. This precession explains the nonzero j even if the injected beam is spin polarized in the y direction. The spin precession also explains the decrease in P0 shown in Fig. 3 b for the injected beam spin polarized in either the x or the y direction. The angle of spin precession, can be directly calculated from P0 as, = cos1 P0. The angle scales linearly with the distance a, as plotted in the inset of Fig. 3 b . On the other hand, the angular rate of the spin eff precession under the effective magnetic eld is = 2Bz / . Thus the angle of spin precession when the electron emerges from the QD array can be calculated. The result is = t=
eff 2Bz t/

Px x , z = x , z x , z / x , z , and denote the +x and x directions. Clearly, in both Figs. 4 a and 4 b , the spin polarization after the diffraction is different from that of the injected beam, which conrms the presence of spin precession. If the injected beam is spin polarized in the z direction, although the principal beam spin does not precess, the diffracted beam spin does, as shown in Fig. 4 b . The spin precession causes the spin lattices in Fig. 4 to be totally different from the charge lattices in Fig. 2. A more interesting feature in Fig. 4 b is the antisymmetric spin pattern. If the injected beam is spin polarized in the z direction, although the diffracted uxes are symmetric between left and right Fig. 2 b , their spin polarizations are not Fig. 4 b . This effect causes a transverse spin current, somewhat similar to the SHE, but with an important difference that the spin current is outside of the SO coupling region, which makes it dissipative and detectable. Because the asymmetric diffraction depends on the spinpolarization of the injected beam, the TSHE effect can be easily used to design spintronics applications. Generally, applications based on the TSHE effect can benet from two features: rst, the applications can be controlled electrically as the SO coupling strength is proportional to the electric eld; second, different functions can be realized through adjusting the structure parameters of the QD array, as indicated earlier.

IV. POTENTIAL APPLICATION IN NEUTRON AND PHOTON SCATTERING

= 2k0 pxt/m = 2k0a.

For our system, this expression give = 0.0353a, which agrees perfectly with the slope in the inset of Fig. 3 b . The agreement conrms the presence of the spin precession and the effective magnetic eld due to the SO coupling inside the QD array. Spin precession also leads to the different spin orientations of the different beams after diffraction, resulting in the separation of the spin lattice and charge lattice. Figure 4 shows the spatial distribution of the x component of the spin polarization, Px x , z , of an injected beam spin polarized in the x direction Fig. 4 a or z direction Fig. 4 b . Here,

The TSHE is more general than the Rashba Hamiltonian of Eq. 1 , which applies to semiconductor 2DES. It can be produced by corresponding Hamiltonians that describe photon or neutron beams, which have odd powers both of spin and momentum. For example, the SO coupling Hamiltonian due to the motion of neutrons through an external electric eld E,19 HSO = s p E , 7

is linear in both neutron spin s and momentum p. Here, = e / m2c2 is a constant, and = 1.91 is the gyromagnetic ratio of neutron. For photons, a SO-coupling Hamiltonian,

245319-4

TUNABLE SPIN HALL EFFECT BY STERN-GERLACH

PHYSICAL REVIEW B 74, 245319 2006

2

HSO =

p2 p2 /p2, x y

=2

0 px p y /p

, 8

ACKNOWLEDGMENTS

where 0 is a constant, was recently used to study the optical spin Hall effect.20 The Hamiltonian is linear in spin and bilinear in momentum. Thus the TSHE can potentially be realized in such systems as well.

This research was conducted at the CNMS sponsored at ORNL by the Division of Scientic User Facilities and by BES/DMS MS and ASCR/MICS, U.S. DOE. The work was further supported by DOE Grant No. FDEFG0203ER46096, and by the McMinn Endowment at Vanderbilt University.

Gerlach and O. Stern, Z. Phys. 9, 349 1922 . Wrobel, T. Dietl, A. Lusakowski, G. Grabecki, K. Fronc, R. Hey, K. H. Ploog, and H. Shtrikman, Phys. Rev. Lett. 93, 246601 2004 . 3 L. Karpa and M. Weitz, Nat. Phys. 2, 332 2006 . 4 S. Datta and B. Das, Appl. Phys. Lett. 56, 665 1990 . 5 J. I. Ohe, M. Yamamoto, T. Ohtsuki, and J. Nitta, Phys. Rev. B 72, 041308 R 2005 . 6 J. E. Hirsch, Phys. Rev. Lett. 83, 1834 1999 . 7 E. I. Rashba, Phys. Rev. B 68, 241315 R 2003 . 8 E. I. Rashba, Phys. Rev. B 70, 201309 R 2004 . 9 J. Sinova, D. Culcer, Q. Niu, N. A. Sinitsyn, T. Jungwirth, and A. H. MacDonald, Phys. Rev. Lett. 92, 126603 2004 . 10 B. A. Bernevig and S. C. Zhang, Phys. Rev. Lett. 95, 016801 2005 . 11 E. I. Rashba, Fiz. Tverd. Tela Leningrad 2, 1224 1960 Sov. Phys. Solid State 2, 1109 1960 ; Y. A. Bychkov and E. I.
2 J.

1 W.

Rashba, J. Phys. C 17, 6039 1984 . Q. Shen, Phys. Rev. Lett. 95, 187203 2005 . 13 Y. K. Kato, R. C. Myers, A. C. Gossard, and D. D. Awschalom, Science 306, 1910 2004 . 14 J. Wunderlich, B. Kaestner, J. Sinova, and T. Jungwirth, Phys. Rev. Lett. 94, 047204 2005 . 15 S. O. Valenzuela and M. Tinkham, Nature London 442, 176 2006 . 16 G. E. W. Bauer, Science 306, 1898 2004 . 17 V. M. Ramaglia, D. Bercioux, V. Cataudella, G. D. Filippis, and C. A. Perroni, J. Phys.: Condens. Matter 16, 9143 2004 . 18 J. Nitta, T. Akazaki, H. Takayanagi, and T. Enoki, Phys. Rev. Lett. 78, 1335 1997 . 19 M. Blume, Phys. Rev. 133, A1366 1964 . 20 A. Kavokin, G. Malpuech, and M. Glazov, Phys. Rev. Lett. 95, 136601 2005 .
12 S.

245319-5