david d. awschalom and michael e. flatte- challenges for semiconductor spintronics

8/3/2019 David D. Awschalom and Michael E. Flatte- Challenges for semiconductor spintronics

1/7

nature physics| VOL 3 | MARCH 2007 | www.nature.com/naturephysics 153

REVIEW ARTICLE

DAVID D. AWSCHALOM1 AND MICHAEL E. FLATT21Center for Spintronics and Quantum Computation and Department of Physics,

University of California, Santa Barbara, California 93106, USA2Optical Science and Technology Center, Department of Physics and Astronomy,

and Department of Electrical and Computer Engineering, University of Iowa,

Iowa City, Iowa 52242, USA

e-mail: [email protected]; [email protected]

Semiconductor inormation-processing devices are among themost sophisticated, complex high-perormance structures. As theconstruction cost o a new abrication line approaches $3.5 billion,and 25% o its tools are obsolete and need replacement every threeyears, it is reasonable to question the appeal o alternate methodso inormation processing. Is it even realistic to imagine that anentirely distinct method o perorming logical operations mightbe competitive? Metallic spintronic devices, such as hard disk readheads and magnetic random access memory (MRAM) are one othe most successul technologies o the past decade, with scalingtrends outdoing even CMOS can semiconductor analoguesprovide enough new unctionality to warrant interest? Electro-opticmodulators are also a successul room-temperature technology. Whatis the advantage o replacing these devices with magnetic analogues?

Quantum computing seems to be progressing rapidly with atomsand superconductors why use spins in semiconductors?Metallic spintronic devices1, originating rom the discovery

o giant magnetoresistance (GMR)2,3 in 1988 and the subsequentdevelopment o the spin valve4, can be understood by assuming thatany current o spins is carried by two types o carriers, spin-up andspin-down. Te two-channel picture o spin transport proposedby Mott5 explains the behaviour o magnetoresistive devices610,including GMR and tunnelling magnetoresistance11 (MR), as well asspin injection into metals12. Tis theoretical treatment does not treatspin coherence, in which a persistent component o the spin can bemaintained transverse to an applied magnetic feld or magnetization.Even though respectable spin-coherence times had been measured

via electron spin resonance13 in materials used or spin transport,and although spin coherence was verifed in several o the situations

above (such as spin injection) it was not central to the phenomenaprobed, which we reer to as non-coherent spintronic devices. Asecond generation o metallic spintronic phenomena requires arecognition o the importance o spin coherence, in which spinorientations transverse to the magnetization o magnetic domains,to internal eective magnetic felds, or to the orientations o otherspin populations, are important in the overall dynamics or materialproperties o a system. Examples o coherent spintronic properties

include current-induced precession o the magnetization o magneticmaterials14,15and the spin Hall eect in aluminium16. Tese phenomenawould be used in coherent spintronic devices. By any measure theprogress in metallic spintronics has been exceptionally rapid, withrevolutionary commercial devices available within ten years rom thediscovery o undamental physical eects. More common timelinesrom discovery to the marketplace are 2030 years.

Semiconductor spintronic device physics is progressing alonga similar path to metallic spintronics and has achieved remarkablesuccess in the past decade. A timeline o selected key experimentaldemonstrations since 1994 is shown in Fig. 1. Measurements oexceptionally long room-temperature spin-coherence times in non-magnetic semiconductors (three orders o magnitude longer than innon-magnetic metals)17,18 preceded the convincing demonstration o

high-e ciency semiconductor spin-injection rom a spin-polarizedmaterial1925 and o coherent spin transport in non-magneticsemiconductors26. Initial theories o electronic spin transport,within the two-channel model, emphasized the importance o drior the motion o spin-density packets in semiconductors relativeto metals2731, and have more recently tackled the undamentals odiusive spin transport in magnetic felds32. Te optical accessibilityo spin in semiconductors, which permitted early studies o spindynamics33, has eased the observation o coherent phenomena in spintransport, including spin precession in an internal crystal magneticfeld (without any applied magnetic feld)34, electrically drivenmotion o domain walls35, and the spin Hall eect3639 (beore it wasdiscovered in metals)4044. Rapid progress towards room-temperatureeects has been seen over the past couple o years, suggesting thatroom-temperature devices based on semiconductor spintronics may

Challenges for semiconductor spintronics

High-volume information-processing and communications devices are at present based on

semiconductor devices, whereas information-storage devices rely on multilayers of magnetic metals

and insulators. Switching within information-processing devices is performed by the controlled motionof small pools of charge, whereas in the magnetic storage devices information storage and retrieval is

performed by reorienting magnetic domains (although charge motion is often used for the final stage

of readout). Semiconductor spintronics offers a possible direction towards the development of hybrid

devices that could perform all three of these operations, logic, communications and storage, within the

same materials technology. By taking advantage of spin coherence it also may sidestep some limitations

on information manipulation previously thought to be fundamental. This article focuses on advances

towards these goals in the past decade, during which experimental progress has been extraordinary.


2/7

REVIEW ARTICLE

154 nature physics| VOL 3 | MARCH 2007 | www.nature.com/naturephysics

be soon possible. Here we are not treating the highly successul area o

spins in semiconductors or solid-state quantum computation or areview see re. 45. Our ocus here is on ensembles o spins, especiallyor near-room-temperature operation.

In addition to manipulating spin dynamics within non-magneticmaterials, semiconductor spintronics oers materials possibilities

very unlikely in metal systems. Electrical control o the Curietemperature46 or coercive feld47 o erromagnetic semiconductorshas been demonstrated. Tese phenomena require depleting thecarrier concentration within the materials by a substantial raction, arequirement that would be exceptionally challenging or erromagneticmetals (with carrier concentrations three or our orders o magnitudehigher). Te highly coupled spin-orbit character o the magneticdopants present in these systems provide additional possibilities orcoherent spin manipulation48, using electric felds instead o magnetic

ones to manipulate the spin degree o reedom. Figure 2 shows a metallicspintronic MRAM device rom Freescale, and the demonstration odomain-wall motion due to current in the magnetic semiconductorGaMnAs. Tis eect might eventually permit the development o anMRAM technology based on magnetic semiconductors.

POTENTIAL ADVANTAGES OF SEMICONDUCTOR SPINTRONICS

LOGIC

Charge-based switching devices dierentiate between a 0 and a 1by the location o a small quantity o charge. For example, a feld-eect transistor is on i the channel is charged and thus metallic,and is o i the channel is not charged, and thus insulating. Tisswitching device can be imagined as two wells separated by a

barrier o adjustable height. Tis barrier must be high enough sothat a charge placed in one well will stay there, but must be loweredto move the charge rom one well to another. In separate analysesseveral authors have ound a minimum switching energy to moverom one confguration to another, Ebit = kTln2 ~ 23 meV (re. 49).Tis limit is undamental or charge-based switching devices thatare, once the charge is placed in one o the wells, allowed toreach thermal equilibrium. However, current and projectedsemiconductor logic devices are still ar rom this limit, or theminimum switching energy is derived assuming adiabatic (slow)charge motion, and the desired switching speed o semiconductorlogic devices is ast and continues to increase. Te projected gate-switching energy in 2018 or low-standby-power CMOS and a10 nm gate width is 15 eV, which is three orders o magnitude largerthan the theoretical minimum50.

Inormation encoded in the electron spin orientation, rather

than the position o a pool o charge, is not subject to the abovelimitations on switching energies. A pool o spin-polarized electronswill maintain its spin orientation without the presence o any barrierbetween spin-up and spin-down states or times exceeding 10 ns atroom temperature in common semiconductor materials. Changingthe inormation rom a 1 to a 0 would consist o applying a smallmagnetic feld, which as described below can be a real magnetic feldor an eective feld, to coherently rotate the spin by 180. Tus none othe operations involving electron spin need to raise or lower a barrierto charge motion. I the operations are done coherently the minimumswitching energy derived or charge-based inormation processingdoes not apply. Even when the motion o charge (such as in an electricalgate contact) is used to manipulate spin, the switching energy o a astspin-based device can be much closer to the undamental limit than a

charge-based device51

. Semiconductor spintronic devices thus wouldavoid the above thermodynamic limitation o a minimum switchingenergy by remaining out o equilibrium or long periods o time (othe order o the spin coherence times).

Coherent semiconductor spintronic devices, by virtue o theexceptionally long room-temperature spin-coherence times discoveredin ordinary semiconductor materials17,18, could in principle perormmultiple independent operations beore the carriers reach thermalequilibrium. Intererence o spin packets is one example, whereby twopackets with spin polarizations oriented at 90 to each other generatea new packet with spins oriented at 45 to the two original packets(intererence in a ring structure has been demonstrated52). Routingo spins via the spin Hall eect may be possible, as the sign o the spinHall conductivity depends on the mobility53, and thus could be tuned

either by an applied ordinary voltage or an applied spin-bias54

.Speed is another essential concern or next-generation inormation-processing devices. In charge-based devices the speed is limited by thecapacitance o the device and the drive current. As the semiconductorspintronic device is a coherent one, the speed limitations are given bytypical precession requencies o electron spins, and range rom GHzto Hz. For example, in order to coherently rotate a spin by 180 at aHz rate, an energy splitting o the order o 3 meV must be generatedbetween the up and down spins. Tis energy splitting is an order omagnitude lower than the thermal energy at room temperature. Tisis both a beneft and a danger or semiconductor spintronics. Localthermal equilibrium cannot be relied on to keep the inormation sae,even during a logical operation. Tus the system must be su cientlyisolated rom the environment (stray magnetic felds in particular) toperorm its operation robustly.

1994 1996 1998 2006200420022000

FerromagneticGaMnAs

Electricaldetectionofspin

Singlespinsindiamond

Longspin-c

oherencetimes

Electricalin

jectionofspin

Gatingoffe

rromagnetism

Giantplana

rHalleffect

Singlespinsinnanostructures

Current-ind

ucedspinpolarization

SpinHallef

fect

Opticalmodulationofcoherence

Electricalm

odulationofcoherence

Electron,m

agneticionspin-coherence

Figure 1 Timeline of key experimental discoveries since 1994 in semiconductor spintronics. The pace of progress continues to increase, and with the recent demonstration ofelectrical detection of spin, all essential elements for a semiconductor spintronic technology have been demonstrated under some experimental conditions.


3/7

REVIEW ARTICLE


STORAGE

Non-volatile storage and logic have traditionally been perormedby independent technologies. Te dominant technologicalimplementation o non-volatile storage is through the magneticorientation o media, such as on the platters in hard disks. MRAM,which uses magnetism to store inormation without moving parts,has become available commercially this summer rom Freescale.However MRAM, like other magnetic non-volatile storage, relieson magnetic metals and insulators not semiconductors. I theundamental physical phenomena underlying MRAM can bedemonstrated at room temperature in semiconductors it maybe possible to integrate non-volatile storage directly into logicalprocessors. Spin transer torque, the current-induced reorientation

o the magnetization o a magnet55,56

that is central to second-generation MRAM, has been demonstrated in the magneticsemiconductor GaMnAs, albeit at low temperature (Fig. 2). Evenmore promising is the possibility o integrated magnetic transistordevices, which blend non-volatility or reprogrammability withthe central transistor property o gain. Tese magnetic transistordevices, such as the unipolar spin transistor57 or the magneticbipolar transistor58, have yet to be demonstrated.

Long-lived polarizations are also possible within the nuclearsystem. Substantial eective nuclear magnetic felds (>1 ) can begenerated via the Overhauser eect59,60 and persist or many minuteswith only a small applied magnetic feld (0.15 ), although at lowtemperature. Tese could be used or optoelectronic coupling, or orreprogrammable logic, and have been demonstrated both in ringintererence52 and in all-optical NMR61.

COMMUNICATIONS

Isolation o one optical component o a communications systemrom the other optical components is achieved through one-waytransmission o light. Tis non-reciprocal transmission requires amagnetic component, otherwise the time-reversal invariance o theapparatus will guarantee that the transmission properties or lighttravelling in one direction will be the same as the properties or lighttravelling in the other direction. Tus magneto-optical elements(Faraday rotators) play a key role in the laser systems central tocurrent communications. Although the magnetic orientation osuch an element can be changed with an applied magnetic feld, thematerial properties themselves are inexible. Metals and insulatorscannot be su ciently changed by applied electric felds to provide

markedly dierent magnetic properties. Magnetic semiconductors,however, can have their Curie temperatures signifcantly changed46as well as their magnetic easy axes by an electric feld47. Electricallygatable Faraday rotators would provide a seamless coupling betweenlinear optics and electrical signals, potentially with less optical lossthan typical or electro-optic modulators working via the quantum-confned Stark eect. High-speed optical switches based on spin havealso been suggested and demonstrated at room temperature; here,the very ast spin lietimes in some materials, or the tunable nature othe spin lietime with electric felds, provides advantages over charge-based switches. Spin-based lasing may provide ways to e cientlymodulate high-power semiconductor lasers, as the carrier density inan active region would not need to be changed, the intensity o theemitted coherent light could be modulated by controlling the degreeo spin polarization o the current injected into the active region.

Domain wall

Domain wall

jpulse

I II III

Magnetictunnel

junction

Write line 2ON for sense

OFF forprogram

Iref

I

Isense

Write line 1

I

a b

Figure 2 Electrical reorientation of magnetic domains in metals and semiconductors. a, Schematic diagram of MRAM produced by Freescale. (Figures fromhttp://www.freescale.com/files/memory/doc/white_paper/MRAMWP.pdf). Isense is the sensing current. Iref the reference current.The white arrows represent the free and fixedmagnetization orientations.b, A domain wall in GaMnAs is electrically controlled at low temperatures using current pulses. The domain wall image in the initial position(top panel) and final position (bottom panel) are imaged using MOKE. Three regions (I, II, III) having different GaMnAs thicknesses were made to pattern coercivity and toconfine the domain wall in region II. Jpulse is the current that moves the domain wall from one side of II to the other. This experimental demonstration (from ref. 35)sets thestage for similar device developments in semiconductor spintronics.


4/7

REVIEW ARTICLE


QUANTUM COMPUTATION

Spin-based solid-state approaches or quantum computationprovide the potential or fxing isolated quantum degrees o reedomin space, by embedding quantum dots or ions within a solid

matrix, and then addressing those degrees o reedom with smallelectrical contacts. Progress in understanding the coupling betweenspin degrees o reedom and electrical felds48,6264, principallythrough the spin-orbit interaction but also through nanomagneticabrication65, avoid some o the problems with producing highlylocalized a.c. magnetic felds. Extremely long room-temperaturespin-coherence times, such as 350 s or nitrogen vacancycentresin diamond66, and short gate times67, provide ratios o coherencetimes to gate times that exceed 103. Trough the coupling o thespin degree o reedom o an electron to optical felds there is a clearmethod o coupling photons into and out o a spin-based quantumcomputer68, and those photons can have wavelengths compatiblewith current communications.

MULTIFUNCTIONALITY

As mobile phones and other gadgets become smaller and morepowerul, one could consider whether all the elements o moderninormation manipulation (logic, storage and communication) couldbe combined and perormed on a single chip. Some o the applicationso spintronics described above are in the area o low-power electronics,so such a multiunctional chip might make possible complex butinexpensive submicrometre remote sensors.

CHALLENGES AND ADVANCES IN SEMICONDUCTOR SPINTRONICS

LOGIC

Improving or optimizing a spintronic device requires attention tovery dierent problems than or charge-based devices. Optimizationo charge transport usually means e cient transer o charge current

rom one region to another, or conductivities that are very sensitiveto gate voltages. Tus in some regions large doping densities, and inother regions large electric felds, are useul or shuttling the chargese ciently around a chip. Designers o spin devices have to worry

about the loss o spin polarization or spin coherence, wherein a carrierthat is spin polarized in one direction eectively turns into anothercharge. wo o the accelerants or spin decoherence are large dopingdensities or semiconductors and large electric felds33,69,70. Bothor large doping densities and large electric felds a spin-polarizedcarrier will experience a large, randomly oriented, internal eectivemagnetic feld.

Tese accelerants are now signifcant barriers towards achievingnearly 100% e cient spin injection rom a erromagnetic metal intoa semiconductor and similarly e cient spin detection. Te initialchallenge or spin injection, however, was quite dierent, and wastraced to the substantial dierence in conductivity between mosterromagnets and semiconductors71. Tis dierence made it di cultor the erromagnet to drive a large enough chemical potential

dierence between spin-up and spin-down in the semiconductor.Inserting a spin-dependent interace resistance rom a Schottkybarrier or tunnel barrier resolves this problem7274, and the eecto spin-density dri in the semiconductor also helps29,30. Spininjection rom a erromagnetic metal into a semiconductor hasbeen demonstrated experimentally with this approach using thespin-dependent resistances o Schottky barriers21,23, and oxideinsulators22. Te room-temperature e ciency o spin injection nowexceeds 70% (re. 24), and urther progress is expected as materialsand interace physics improves. Engineering the device dopingand feld profle region will thereore be required in the design ooptimally perorming semiconductor spintronic chips. Shownin Fig. 3 are results or optically detected spin injection into asemiconductor through a Schottky barrier (reported in re. 23) andor electrically detected spin injection into a metal75. A successul

295 K

B (T)

Polarization(%

)

40 K

Fen

QW

p

M

B

20

10

0

4 2 0 2 4

10

202.086

T= 4.2 K2.085

2.084

2.083

2.082

2.081

2.080R()

2.079

2.0780.032

0.031

0.030

0.02960 40 20 0

B (mT)

20 40 60

F1I I

I

V VI

F2

F1

F2

a b

+ +

N N

Figure 3 Spin transport through non-magnetic semiconductors and metals. a, Electroluminescence polarization in an Fe/Al0.1Ga0.9As spin-LED. A schematic of the device isshown in the lower inset. Spin-polarized electrons injected from Fe recombine with holes in the quantum well (QW). The magnetic field (B) rotates the magnetization (M) ofthe Fe film out of the plane (upper inset). Reprinted with permission from ref. 23. Copyright (2005) by the American Physical Society. b, Spin-valve effect (top curve, and leftcartoon) and non-local electrical detection of spin injection (bottom curve, and right cartoon) in a lateral metallic spin-valve device75. R is resistance. In the ordinary spin-valvemeasurement, the current Iflows from ferromagnetic contact F1 through the normal metal N to ferromagnetic contact F2, and the voltage is measured between F1 and F2. Inthe non-local measurement, the current flows from F1 into the normal metal N, and the voltage is measured between the detection contact F2 and a reference contact on thenormal metal. Reprinted with permission from ref. 75. Copyright (2003) by the American Physical Society.


5/7

REVIEW ARTICLE


non-local electrically detected spin-injection experiment sensitiveto precession has also been perormed or a semiconductor76.

Spin injection and detection can be considered the input andread-out stages o a logic device within which the spin is manipulatedby external or internal magnetic felds or by spin-selective scattering.It has been demonstrated that the internal eective magnetic felds insemiconductors with spin-orbit interactions can be used to reorient

spins and also to drive magnetic resonance34

. Tese eective internalmagnetic felds can be manipulated with applied external electricfelds77,78, which implies new gating mechanisms or spin-basedtransistors79. Furthermore the separation o spins can be achievedthrough the recently demonstrated spin Hall eect, frst seen insemiconductors4042,44, later in metals16, and most recently at roomtemperature43. Control o the spin Hall eect via control o the materialmobility may be used to change spin currents in magnitude or evendirection54, using a controllable spin Hall eect to route spins orlogic. Finally, it might be possible to do away with magnetic materialsentirely due to the achievement o spontaneous spin polarization atroom temperature in a non-magnetic semiconductor80,81.

STORAGE

Many o the erromagnetic semiconductor materials have extremelyhigh carrier-doping levels, and controlling the interaces o thesematerials is a great challenge. I novel storage devices based onerromagnetic semiconductors are to be attempted, then achievingerromagnetism in lower-doped semiconductor materials will behighly desirable. Tere are some indications that this might occurnaturally at the edges o erromagnetic materials, as the carriers aredepleted rom the region but magnetism remains82.

Only very recently has there been a report o a pn diode made witha erromagnetic material83,84 previous attempts led to poor diodesbecause the doping level in the intrinsic, or depletion, region was toohigh to support a signifcant voltage. Another recent achievement wasthe demonstration o exchange biasing in magnetic semiconductors85.A central element o metallic MRAM, exchange biasing will be a keyelement o semiconductor spintronics storage technology.

COMMUNICATIONS

Optics and erromagnetism has turned out to be a dirty partnershipso ar in GaMnAs. Te optical lietimes are so short, unlike or non-magnetic semiconductors, that it was some time beore they could bemeasured86. As the desired magneto-optical devices typically requiresubstantial Faraday rotation without signifcant optical losses,magnetic semiconductors have not been successul at dislodging

magnetic insulators rom this niche. Experiments on CdMneand CdMnHge optical isolators, however, suggest competitive values to yttrium iron garnet or the optical rotation relative tooptical loss87,88 in a material that can be monolithically integratedon a semiconductor substrate. A semiconductor waveguide withan integrated erromagnetic metal clad has also shown goodperormance as an optical isolator89.

As the materials become cleaner and more controlled themagneto-optical properties should improve urther. It has beendiscovered that much cleaner GaMnAs could be achieved through

very long low-temperature post-growth annealing.At the same timethe optical properties o very lightly doped GaMnAs seem quite good,even though the material itsel is not doped su ciently to becomeerromagnetic. New discoveries o erromagnetic semiconductors

suggest there should be materials with better optical properties, suchas ZnCre. Whether this material is a carrier-mediated erromagnetor not is not clear yet, although it is dopable and the magnetism hasa large inuence on the optical properties.

QUANTUM COMPUTING

Te achievement o large-scale quantum-inormation processingin any physical system will be a tremendous success. Recentexperimental advances in semiconductor spintronic quantumcomputing include the demonstration o long T1 and T2 times insemiconductor quantum dots (albeit at low temperatures45), thedemonstration o coherent single-spin manipulation in diamond,and numerous examples o gate operations perormed on ensembleso spins9092, but expected to be extended to single-spin manipulationin quantum dots or embedded ions in the near uture.

E

Source Channel Drain Source Magneticsemiconductor

Channel Magneticsemiconductor

Drain

GateGate

Oxide Oxide

E

E

OFF

ON

E

E

OFF Long T1

E

E

E

ONShort T1

E

E

E

Charge-based current gating

(MOSFET)

Spin-based current gating

(Spin-FET)

Figure 4 Schematic structure of a MOSFET (left) and a spin-FET (right). MOSFETs work by raising and lowering a barrier to turn the current on or off. A magneticsemiconductor spin lifetime FET uses static spin-dependent barriers and changes the spin character of the electrons in the channel to turn the current on or off. Reused withpermission from ref. 51. Copyright 2006, American institute of Physics.


6/7

REVIEW ARTICLE


SURPRISES AND (VERY) OPEN QUESTIONS

Do we need magnetic materials for a functional semiconductorspintronic technology?Current-induced spin polarization and the spin Hall eect have bothbeen demonstrated at room temperature43. I the current-inducedspin polarization can be made large enough this eect could replacespin injection rom a erromagnetic metal as the central method

o injecting spins into non-magnetic semiconductors at roomtemperature. Similarly a spin-Hall-eect detector could replace theneed or a erromagnetic detector contact to electrically measure spinpolarization at room temperature. Manipulation o the spins could bedone using internal eective magnetic felds34,93 or perhaps using thespin Hall eect to directly drive resonance94.

Are there interesting semiconductor spintronic devices that do notrequire large T2 times?Although much o the initial interest in semiconductor spintronicswas due to the exceptionally long spin-coherence times seen inordinary semiconductor material, theoretical work indicatingthat T1 times can be tuned by orders o magnitude at roomtemperature70 seems to provide a pathway to a new switching

mechanism or a spin-feld-eect transistor95,96. Comparisons othe scaling properties o this approach to CMOS indicate that spin-based feld-eect transistors could compete with end-o-roadmapCMOS or low-standby-power devices51. Elegant approaches tobalancing precessional transport in lateral devices might permite cient spin manipulation o ensembles, but in the diusiveregime97 rather than the ballistic regime79.

An example o such a spin transistor device is shown in Fig. 4,and compared with a MOSFE. A MOSFE works by controlling theheight o a barrier to electron ow, with the barrier height and widthlargely determined by the desired ono current ratio and leakagecurrent. Te spin-based FE has static spin-dependent barriers(generated by a magnetic insulator or magnetic semiconductor),and works by controlling the spin lietime in the base95. Application

o an electric feld generates an eective magnetic feld98

(a Rashbafeld99), which decoheres the spins. As the spin orientation o thesourcechannel barrier and channeldrain barrier are opposite toeach other, current only ows i the spin lietime in the channelis short. For the spin-FE the inputs and outputs are incoherent,but the transistor mechanism is spintronic. It has been arguedthat the spin lietime FE device shown in Fig. 4 has superiorpower-dissipation properties to end-o-roadmap CMOS51. Terequirements or spin lietimes in the absence o a gate feld andor spin-injection e ciency, however, are very high or this device.Multi-lead devices with coherent inputs and outputs might provideeven better perormance.

What is the connection between spin-coherence times and

optical coherence?Remarkable measurements o the lasing threshold dependenceo a semiconductor microcavity on the T2 time o a spin in aquantum dot inside that cavity suggest that semiconductor spin-coherence could be closely tied to optical coherence in microcavitygeometries100. Tis might provide e cient ways o transmittinginormation rom single photons to single spins, or be used togenerate a hybrid device or quantum-inormation processing withboth optical and spin components.

What might exotic materials provide for semiconductor spintronics?Individual deect centres in diamond have been coherently manipulatedto demonstrate magnetic resonance9092 at room temperature. Rapidadvances in controlling the electrical and optical properties o thissystem may lead to uture diamond-based technologies. Other

wide-bandgap materials such as oxide semiconductors also seempromising or room-temperature spintronic devices.

More physics or more engineering?Perhaps the mechanism or driving the next new technology basedon semiconductor spintronics is already known, and among the wide

variety o interesting device physics eects produced rom the pastdecade o intensive research. It may now be time or circuit designers

and systems engineers to take a closer look at these demonstratedmechanisms to see how a semiconductor spintronics architectureshould be assembled. Even i that is not yet possible, systems andcircuit engineers may greatly help the feld o semiconductorspintronics by identiying those areas that most need progress inorder to achieve a commercially viable technology.

doi:10.1038/nphys551

References1. Ziese, M. & Tornton, M. J. (eds) Spin Electronics (Lecture Notes in Physics series, Vol. 569,

Springer-Verlag, Heidelberg, 2001).

2. Baibich, M. N. et al. Giant magnetoresistance o (001)Fe/(001)Cr magnetic superlattices.Phys. Rev. Lett.61, 24722475 (1988).

3. Binasch, G., Grnberg, P., Saurenbach, F. & Zinn, W. Enhanced magnetoresistance in layeredmagnetic structures with antierromagnetic interlayer exchange.Phys. Rev. B39, 48284830 (1989).

4. Dieny, B. Giant magnetoresistive in so erromagnetic multilayers.Phys. Rev. B43, 12971300 (1991).5. Mott, N. F. Te electric al conductivi ty o transit ion metals.Proc. R. Soc. Lond. A

153, 699717 (1936).6. Camley, R. E. & Barna, J. Teory o giant magnetoresistance eects in magnetic layered structures

with antierromagnetic coupling. Phys. Rev. Lett.63, 664667 (1989).7. Barna, J., Fuss, A., Camley, R. E., Grnberg, P. & Zinn, W. Novel magnetoresis tance eect in

layered magnetic structures: Teory and experiment.Phys. Rev. B42, 81108120 (1990).8. Barthlmy, A. & Fert, A. Teory o the magnetoresistance in magnetic multilayers: Analytic al

expressions rom a semiclassical approach. Phys. Rev. B43, 1312413129 (1991).9. Valet, . & Fert, A. Teory o the perpendicular magnetoresistance in magnetic multilayers.

Phys. Rev. B48, 70997113 (1993).10. Butler, W. H.et al. Conductance and giant magnetoconductance o Co|Cu|Co spin valves:

Experiment and theory. Phys. Rev. B56, 1457414582 (1997).11. Moodera, J. S., Kinder, L. R., Wong, . M. & Meservey, R. Large magnetoresistance at room

temperature in erromagnetic thin flm tunnel junctions.Phys. Rev. Lett.74, 32733276 (1995).12. Johnson, M. & Silsbee, R. H. Spin-injection experiment.Phys. Rev. B37, 53265335 (1988).

13. Schultz, S. & Latham, C. Observation o electron spin resonance in copper.Phys. Rev. Lett.

15, 148151 (1965).

14. soi, M. et al. Excitation o a magnetic multilayer by an electric current.Phys. Rev. Lett.

80, 42814284 (1998).

15. Katine, J. A., Albert, F. J., Buhrman, R. A., Myers, E. B. & Ralph, D. C. Current-drivenmagnetization reversal and spin-wave excitations in Co /Cu /Co pillars.Phys. Rev. Lett.

84, 31493152 (2000).16. Valenzuela, S. O. & inkham, M. Direct electronic measurement o the spin Hall eect.Nature

442, 176179 (2006).17. Kikkawa, J. M., Smorchkova, I. P., Samarth, N. & Awschalom, D. D. Room-temperature spin

memory in two-dimensional electron gases. Science277, 12841287 (1997).18. Kikkawa, J. M. & Awschalom, D. D. Resonant spin amplifcation in n-type GaAs.Phys. Rev. Lett.

80, 43134316 (1998).19. Ohno, Y.et al. Electrical spin injection in a erromagnetic semiconductor heterostructure.Nature

402, 790 792 (1999).20. Fiederling, R. et al. Injection and detection o a spin-polarized current in a light-emitting diode.

Nature402, 787790 (1999).21. Hanbicki, A. ., Jonker, B. ., Itskos, G., Kioseoglou, G. & Petrou, A. E cient electrical spin

injection rom a magnetic metal/tunnel barrier contact into a semiconductor.Appl. Phys. Lett.

80, 12401242 (2002).22. Motsnyi, V. F. et al. Electrical spin injection in a erromagnet/tunnel barrier/semiconductor

heterostructure. Appl. Phys. Lett.81, 265267 (2002).23. Adelmann, C., Lou, X., Strand, J., Palmstrm, C. J. & Crowell, P. A. Spin injection and relaxation in

erromagnet-semiconductor heterostructures. Phys. Rev. B71, 121301(R) (2005).24. Jiang, X. et al. Highly spin-polarized room-temperature tunnel injector or semiconductor

spintronics using MgO(100). Phys. Rev. Lett.94, 056601 (2005).25. Crooker, S. A. et al. Imaging spin transport in lateral erromagnet/semiconductor structures.

Science 309, 21912195 (2005).26. Kikkawa, J. M. & Awschalom, D. D. Lateral drag o spin coherence in gallium arsenide.Nature

397, 139141 (1999).27. Aronov, A. G. & Pikus, G. E. Spin injection into semiconductors.Fiz. Tekh. Poluprovodn.

10, 11771179 (1976); Sov. Phys. Semicond.10, 698700 (1976).28. Flatt, M. E. & Byers, J. M. Spin diusion in semiconductors.Phys. Rev. Lett. 84, 42204223 (2000).

29. Yu, Z. G. & Flatt, M. E. Electric-feld dependent spin diusion and spin injection intosemiconductors. Phys. Rev. B66, 201202 (2002).

30. Yu, Z. G. & Flatt, M. E. Spin diusion and injection in semiconductor structures: Electric feldeects. Phys. Rev. B66, 235302 (2002).

31. Awschalom, D. D., Samarth, N. & Loss, D. (eds).Semiconductor Spintronics and Quantum

Computation (Springer, Heidelberg, 2002).


7/7

REVIEW ARTICLE


32. Qi, Y. & Zhang, S. Spin diusion at fnite electric and magnetic felds.Phys. Rev. B67, 052407 (2003).

33. Meier, F. & Zachachrenya, B. P.Optical Orientation (Modern Problems in Condensed MatterScience series, Vol. 8, North-Holland, Amsterdam, 1984).

34 Kato, Y. K., Myers, R. C., Gossard, A. C. & Awschalom, D. D. Coherent spin manipulation withoutmagnetic felds in strained semiconductors.Nature427, 5053 (2004).

35. Yamanouchi, M., Chiba, D., Matsukura, F. and Ohno, H. Current-induc ed domain-wall switchingin a erromagnetic semiconductor structure. Nature428, 539542 (2004).

36. Dyakonov, M. I. & Perel, V. I. Current-induced spin orientation o electrons in semiconductors.Phys. Lett. A35, 459460 (1971).

37. Hirsch, J. E. Spin Hall eect.Phys. Rev. Lett.83, 18341837 (1999).38. Murakami, S., Nagaosa, N. & Zhang, S.-C. Dissipationless quantum spin current at room

temperature. Science301, 13481351 (2003).39. Sinova, J. et al. Universal intrinsic spin Hall eect. Phys. Rev. Lett.92, 126603 (2004).

40. Kato, Y. K., Myers, R. C., Gossard, A. C. & Awschalom, D. D. Observation o the spin Hall eect insemiconductors. Science 306, 19101913 (2004).

41. Sih, V. et al. Spatial imaging o the spin Hall eect and current-induced polarization intwo-dimensional electron gases. Nature Phys.1, 3135 (2005).

42. Wunderlich, J., Kaestner, B., Sinova, J. & Jungwirth, . Experimental observation o thespin-hall eect in a two-dimensional spin-orbit coupled semiconductor system.Phys. Rev. Lett.

94, 047204 (2005).43. Stern, N. P. et al. Current-induced polarization and the spin Hall eect at room temperature.

Phys. Rev. Lett.97, 126603 (2006).44. Sih, V. et al. Generating spin currents in semiconductors with the spin Hall eect.Phys. Rev. Lett.

97, 096605 (2006).45. Hanson, R., Kouwenhoven, L. P., Petta, J. R., arucha, S. & Vandersypen, L. M. K. Spins in

ew-electron quantum dots. Preprint at http://arxiv.org/cond-mat/0610433 (2006).46. Ohno, H. et al. Electric-feld control o erromagnetism. Nature 408, 944946 (2000).

47. Chiba, D., Yamanouchi, M., Matsukura, F. & Ohno, H. Electrical manipulation o magnetizationreversal in a erromagnetic semiconductor. Science301, 943945 (2003).

48. ang, J.-M., Levy, J. & Flatt, M. E. All-electrical control o single ion spins in a semiconductor.Phys. Rev. Lett.97, 106803 (2006).

49. Landauer, R. Irreversibi lity and heat generation in the computing process.IBM J. Res. Dev.

5, 183191 (1961).

50. International Technology Roadmap for Semiconductors (Semiconductor Industry Association,San Jose, Caliornia, USA, 2003); http://public.itrs.net.

51. Hall, K. C. & Flatt, M. E. Perormance o a spin-based insulated gate feld eect transistor.Appl. Phys. Lett.88, 162503 (2006).

52. Kato, Y., Myers, R. C., Gossard, A. C. & Awschalom, D. D. Electron spin intererometry using asemiconductor r ing structure.Appl. Phys. Lett. 86, 162107 (2005).

53. Engel, H.-A., Halperin, B. I. & Rashba, E. I.Teory o spin Hall conductivity in n-doped GaAs.Phys. Rev. Lett. 95, 166605 (2005).

54. Hankiewicz , E. M., Vignale, G. & Flatt, M. E. Spin-Hall eect in a [110] GaAs quantum well.Phys. Rev. Lett.97, 266601 (2006).

55. Slonczewski, J. Current-driven excitation o magnetic multilayers.J. Magn. Magn. Mater.159, L1L7 (1996).

56. Berger, L. Emission o spin waves by a magnetic multilayer traversed by a current.Phys. Rev. B54, 93539358 (1996).

57. Flatt, M. E. & Vignale, G. Unipolar spin diodes and transistors.Appl. Phys. Lett.

78, 12731275 (2001).

58. Flatt, M. E., Yu, Z. G., Johnston-Halperin, E. & Awschalom, D. D. Teory o semiconductormagnetic bipolar transistors.Appl. Phys. Lett.82, 47404742 (2003).

59. Overhauser, A. W. Polarization o nuclei in metals.Phys. Rev.92, 411415 (1953).60. Paget, D., Lampel, G., Sapoval, B. & Saarov, V. I. Low feld electron-nuclear spin coupling in

gallium arsenide under optical pumping conditions.Phys. Rev. B15, 57805796 (1977).61. Kikkawa, J. M. & Awschalom, D. D. All-optical magnetic resonance in semiconductors.Science

287, 473476 (2000).62. Loss, D. & DiVincenzo, D. P. Quantum computation with quantum dots.Phys. Rev. A

57, 120126 (1998).63. Kato, Y., Myers, R. C., Gossard, A. C., Levy, J. & Awschalom, D. D. Gigahertz electron spin

manipulation using voltage-controlled g-tensor modulation. Science299, 12011204 (2003).64. Petta, J. R. et al. Coherent manipulation o coupled electron spins in semiconductor quantum dots.

Science309, 21802184 (2005).65. okura, Y., van der Wiel, W. G., Obata, . & arucha, S. Coherent single electron spin control in a

slanting Zeeman feld. Phys. Rev. Lett.96, 047202 (2006).66. Gaebel, .et al. Room-temperature coherent coupling o single spins in diamond. Nature Phys.

2, 408413 (2006).67. Jelezko, F., Gaebel, ., Popa, I., Gruber, A. & Wrachtrup, J. Observation o coherent oscillations in a

single electron spin. Phys. Rev. Lett.92, 076401 (2004).68. Leuenberger, M. N., Flatt, M. E. & Awschalom, D. D. eleportation o electronic many-qubit

states via single photons. Phys. Rev. Lett.94, 107401 (2005).69. Dyakonov, M. I. & Perel, V. I. Spin relaxation o conduction electrons in noncentrosymmetric

semiconductors. Sov. Phys. Solid State13, 30233026 (1972).

70. Lau, W. H. & Flatt, M. E. unability o electron spin coherence in III-V quantum wells.J. Appl. Phys.91, 86828684 (2002).

71. Schmidt, G., Ferrand, D., Molenkamp, L. W., Filip, A. . & van Wees, B. J. Fundamental obstacleor electrical spin injection rom a erromagnetic metal into a diusive semiconductor.Phys. Rev. B

62, R4790R4793 (2000).72. Rashba, E. I. Teory o electrical spin injection: unnel contacts as a solution o the conductivity

mismatch problem. Phys. Rev. B 62, R16267R16270 (2000).73. Smith, D. L. & Silver, R. N. Electrical spin injection into semiconductors.Phys. Rev. B

64, 045323 (2001).74. Fert, A. & Jars, H. Conditions or e cient spin injection rom a erromagnetic metal into a

semiconductor. Phys. Rev. B64, 184420 (2001).75. Jedema, F. J., Nijboer, M. S., Filip, A. . & van Wees, B. J. Spin injection and spin accumulat ion in

all-metal mesoscopic spin valves. Phys. Rev. B67, 085319 (2003).76. Lou, X. et al. Electrical detection o spin transport in lateral erromagnetsemiconductor devices.

Nature Phys.3, 197202 (2007).77. Koga, ., Sekine, Y. & Nitta, J. Experimental realizati on o a ballistic spin intererometer based

on the Rashba eect using a nanolithographically defned square loop array.Phys. Rev. B74, 041302 (2006).

78. Bergsten, ., Kobayashi, ., Sekine, Y. & Nitta, J. Experimental demonstrat ion o the time reversalAharonovCasher eect. Phys. Rev. Lett.97, 196803 (2006).

79. Datta, S. & Das, B. Electronic analog o the electro-optic modulator.Appl. Phys. Lett. 56, 665667 (1990).

80. Kato, Y., Myers, R. C., Gossard, A. C. & Awschalom, D. D. Current-induce d spin polarization instrained semiconductors. Phys. Rev. Lett.93, 176601 (2004).

81. Silov, A. Yu.et al. Current-induced spin p olarization at a single heterojunction.Appl. Phys. Lett.85, 59295931 (2004).

82. Rster, C. et al. Very large magnetoresistance in lateral erromagnetic (Ga,Mn)As wires withnanoconstrictions.Phys. Rev. Lett.91, 216602 (2003).

83. Chen, P.et al. All-electrical measurement o spin injection in a magnetic pn junction diode.Preprint at http://arxiv.org/abs/cond-mat/0608453 (2006).

84. Fabian, J., uti, I. & Sarma, S. D. Teory o spin-polarized bipolar transport in magnetic pnjunctions.Phys. Rev. B66, 165301 (2002).

85. Eid, K. F.et al. Exchange biasing o the erromagnetic semiconductor Ga1xMnxAs.Appl. Phys. Lett. 85, 15561558 (2004).

86. Beschoten, B. et al. Magnetic circular dichroism studies o carrier-induced erromagnetism inGa1xMnxAs. Phys. Rev. Lett.83, 30733076 (1999).

87. Zayets, V., Debnath, M. C. & Ando, K. Complete magneto-optical waveguide mode conversion inCd1xMnxe waveguide on GaAs substrate.Appl. Phys. Lett.84, 565567 (2004).

88. Onodera, K., Masumoto, . & Kimura, M. 980 nm compact optical isolators using Cd1xyMnxHgyesingle crystals or high power pumping laser diodes.Electron. Lett.30, 19541955 (1994).

89. Shimizu, H. & Nakano, Y. Fabrication and characterization o an InGaAsP/InP active waveguideoptical isolator with 14.7 dB/mm E mode nonreciprocal attenuation.IEEE J. Lightwave Tech.

24, 3843 (2006).90. Gruber, A., Drbenstedt, A., ietz, C., Fleury, L., Wrachtrup J. & von Borczyskowski, C.

Scanning conocal optical microscopy and magnetic resonance on single deect centers.Science

276, 20122014 (1997).

91. Hanson, R., Gywat, O. & Awschalom, D. D. Room-temperature manipulation and decoherence o asingle spin in diamond. Phys. Rev. B74, 161203(R) (2006).

92. Hanson, R., Mendoza, F. M., Epstein, R. J. & Awschalom, D. D. Polarization and readout o coupledsingle spins in diamond. Phys. Rev. Lett.97, 087601 (2006).

93. Crooker, S. A. & Smith, D. L. Imaging spin ows in semiconductors subject to electric, magnetic ,and strain felds. Phys. Rev. Lett.94, 236601 (2005).

94. Duckheim, M. & Loss, D. Electric-dipole-induced spin resonance in disordered semiconductors.Phys. Rev. Lett.2, 195199 (2006).

95. Hall, K. C., Lau, W. H., Gndodu, K., Flatt, M. E. & Boggess, . F. Nonmagnetic semiconductorspin transistor.Appl. Phys. Lett.83, 29372939 (2003).

96. Cartoi xa, X., ing, D. Z.-Y. & Chang, Y.-C. A resonant spin lietime transistor.Appl. Phys. Lett.

83, 14621464 (2003).

97. Schliemann, J., Egues, J. C. & Loss, D. Nonballistic spin-feld-eect transistor.Phys. Rev. Lett.

90, 146801 (2003).

98. Nitta, J., Akazaki, ., akayanagi, H. & Enoki, . Gate control o spin-orbit interac tion in aninverted In0.53Ga0.47As/In0.52Al0.48As heterostructure. Phys. Rev. Lett.78, 13351338 (1997).

99. Bychkov, Y. A. & Rashba, E. I. Oscillatory eects and the magnetic susceptibility o carriers ininversion layers.J. Phys. C 17, 60396045 (1984).

100. Ghosh, S. et al. Enhancement o spin coherence using Q-actor engineering in semiconductormicrodisc lasers. Nature Mater. 5, 261264 (2006).

AcknowledgementsThe authors would like to acknowledge the support of ONR and DARPA.

Correspondence should be sent to D. D. A. or M. E. F.

Competing InterestsThe authors declare that they have no competing interests.

david d. awschalom and michael e. flatte- challenges for semiconductor spintronics

Documents