1: %documentstyle[rmp,aps,floats,psfig,preprint]{revtex} 

2: \documentclass[rmp,aps,floatfix,twocolumn,unsortedaddress]{revtex4}

3: %\documentclass[rmp,aps,floatfix,preprint,unsortedaddress]{revtex4}

4: \usepackage{epsfig}

5: \begin{document}

6: 

7: 

8: %\twocolumn[

9: %\hsize\textwidth\columnwidth\hsize\csname@twocolumnfalse\endcsname

10: 

11: \title{Spintronics: Fundamentals and applications}

12: 

13: \author{Igor \v{Z}uti\'{c}\footnote{Electronic address: igorz@physics.umd.edu.

14: Present address: Center for Computational Materials Science,

15: Naval Research Laboratory, Washington, D.C. 20375, USA}}

16: \affiliation{Condensed Matter Theory Center,

17: Department of Physics, University of Maryland at College

18: Park, College Park, Maryland 20742-4111, USA}

19: 

20: \author{Jaroslav Fabian\footnote{Electronic address: 

21: jaroslav.fabian@uni.graz.at}}

22: \affiliation{Institute for Theoretical Physics, Karl-Franzens University,

23: Universit\"atsplatz 5, 8010 Graz, Austria }

24: 

25: \author{S. Das Sarma}

26: \affiliation{Condensed Matter Theory Center,

27: Department of Physics, University of Maryland at College

28: Park, College Park, Maryland 20742-4111, USA}

29: 

30: 

31: %\maketitle

32: \begin{abstract}

33: {Spintronics, or spin electronics, involves the study of active control

34: and manipulation of spin degrees of freedom in solid-state systems.

35: This article reviews

36: the current status of this subject, including both recent advances

37: and well-established results.

38: The primary focus is on the basic physical principles underlying the

39: generation of carrier spin polarization, spin dynamics, and spin-polarized

40: transport in semiconductors and metals.

41: Spin transport differs from charge transport in that spin is

42: a nonconserved quantity in solids due to spin-orbit and hyperfine coupling.

43: The authors discuss in detail spin decoherence mechanisms in metals and 

44: semiconductors. Various theories of spin injection

45: and spin-polarized transport are applied to hybrid structures relevant to

46: spin-based devices and fundamental studies of materials properties.

47: Experimental work is reviewed with the emphasis on projected applications, 

48: in which

49: external electric and magnetic fields and illumination by light

50: will be used to control spin and charge dynamics to create new functionalities 

51: not feasible or ineffective with conventional electronics.}

52: \end{abstract}

53: \maketitle

54: 

55: \setcounter{tocdepth}{4}

56: \tableofcontents

57: 

58: 

59: \section{\label{sec:I} Introduction}

60: %===========================================

61: 

62: 

63: \subsection{\label{sec:IA} Overview}

64: %-------------------------------------------

65: 

66: Spintronics is a multidisciplinary field

67: whose central theme is the

68: active manipulation of spin degrees of freedom in 

69: solid-state systems.\footnote{While there are proposals for spintronic

70: devices based on deoxyribonucleic acid (DNA) molecules

71: \cite{Zwolak2002:APL}, the whole device, which includes 

72: electrodes, voltage/current source, etc., is still a solid-state system.}

73: In this article the term spin stands for either the spin of a single

74: electron $\bf s$, which can be detected by its magnetic moment 

75: $-g \mu_B {\bf s}$

76: ($\mu_B$ is the Bohr magneton and $g$ is the electron $g$ factor, in a

77: solid generally different from the free electron value of 

78: $g_0=2.0023$),

79: or the average spin of an ensemble of electrons, manifested by magnetization.

80: The control of spin is then a control of either the population and

81: the phase of the spin of an ensemble of particles, or a coherent

82: spin manipulation of a single or a few-spin system.

83: The goal of spintronics is to understand the interaction

84: between the particle spin and its solid-state environments

85: and to make useful devices using the acquired knowledge. 

86: Fundamental studies of spintronics include investigations of spin 

87: transport in electronic materials, as well as 

88: understanding spin dynamics and spin relaxation. 

89: Typical questions that are posed are 

90: (a) what is an effective way to polarize

91: a spin system?

92: (b) how long is the system able to remember its spin

93: orientation? (c) how 

94: can spin be detected? 

95: 

96: Generation of spin polarization usually means creating a nonequilibrium

97: spin population. This can be achieved in several ways. While

98: traditionally spin has been oriented using optical techniques in which 

99: circularly polarized photons transfer their angular momenta

100: to electrons, for device applications electrical spin injection

101: is more desirable. In electrical spin injection a 

102: magnetic electrode is connected

103: to the sample. When the current drives spin-polarized electrons

104: from the electrode to the sample, nonequilibrium spin accumulates

105: there. The rate of spin accumulation depends

106: on spin relaxation, 

107: the

108: process of bringing the accumulated spin population 

109: back to equilibrium. There are several relevant mechanisms

110: of spin relaxation, most involving spin-orbit coupling to  provide

111: the spin-dependent potential, in combination with momentum

112: scattering providing a randomizing force. Typical time scales

113: for spin relaxation in electronic systems are 

114: measured in 

115: nanoseconds, 

116: while the range is from pico to microseconds.

117: Spin detection, also 

118: part of a generic spintronic scheme, typically

119: relies on sensing the changes in the signals caused by the

120: presence of nonequilibrium spin in the system. The common

121: goal in many spintronic devices is to maximize the 

122: spin detection sensitivity  to the point it detects 

123: not the spin itself, but changes in the spin states.  

124: 

125: Let us illustrate the generic spintronic scheme on a prototypical

126: device, the Datta-Das spin field effect

127: transistor (SFET) \cite{Datta1990:APL}, depicted in Fig.~\ref{fig:DD}. 

128: The scheme shows the structure of the usual FET, 

129: with a drain, a source, a narrow channel, and a gate for controlling the 

130: current.

131: The gate either allows 

132: the current to flow (ON) or does not (OFF). The

133: spin transistor is similar in that the result is also a

134: control of the charge current through the narrow channel. 

135: The difference, however, is in the physical realization

136: of the current control. In the Datta-Das SFET the source

137: and the drain are ferromagnets

138: acting as the injector and detector of the electron spin. 

139: The drain injects electrons with spins parallel to the

140: transport direction. The electrons are transported ballistically

141: through the channel. When they arrive at the drain, their spin is 

142: detected. In a simplified picture, the electron can enter 

143: he drain (ON) if its spin 

144: points in the same direction as the spin of the drain. Otherwise

145: it is scattered away (OFF). The role of the gate is to generate an

146: effective magnetic field 

147: (in the direction of $\bf\Omega$ in Fig.~\ref{fig:DD}), 

148: arising from the 

149: spin-orbit coupling in the substrate material,

150: from the confinement geometry of the transport channel, and the

151: electrostatic potential of the gate.

152: This 

153: effective magnetic field causes

154: the electron spins to precess. By modifying the voltage, one can cause

155: the precession to lead to either parallel or antiparallel 

156: (or anything between) electron spin at the drain, effectively

157: controlling the current. 

158: 

159: \begin{figure}

160: \centerline{\psfig{file=zutic_fig01.eps,width=1\linewidth,angle=0}}

161: \caption{Scheme of the Datta-Das spin field-effect transistor (SFET). 

162: The source (spin injector) and the drain (spin detector) are ferromagnetic 

163: metals or semiconductors,

164: with parallel magnetic moments. 

165: The injected spin-polarized electrons with wave vector ${\bf k}$ 

166: move ballistically

167: along a quasi-one-dimensional channel formed by, for example, 

168: an InGaAs/InAlAs heterojunction in a plane normal to ${\bf n}$. Electron spins

169: precess about the precession vector ${\bf \Omega}$, which arises from  

170: spin-orbit coupling and which is defined by the structure and 

171: the materials properties of the channel. The magnitude

172: of ${\bf \Omega}$ is tunable by the gate voltage $V_G$ 

173: at the top of the channel.  The current is large if the electron 

174: spin at the drain points in the initial direction (top row), for

175: example, if the precession period is much larger than the 

176: time of flight, and small if the direction is reversed (bottom).

177: }

178: \label{fig:DD}

179: \end{figure}

180: 

181: Even though the name {\it spintronics} is rather novel,\footnote{The term

182: was coined by S. A. Wolf in 1996, as a name for a 

183: DARPA initiative for novel magnetic materials and devices.} 

184: contemporary research in spintronics relies closely on a long

185: tradition of results obtained in diverse areas of physics (for

186: example, magnetism, semiconductor physics, superconductivity, optics,

187: and mesoscopic physics) and establishes new connections between its

188: different subfields \cite{Rashba2002:JS,Zutic2002:JS}.

189: We review here both well-established results and the physical

190: principles relevant to the present and future applications. 

191: Our strategy is to give a comprehensive view of what

192: has been accomplished, focusing in detail on a few

193: selected topics that we believe are representative for

194: the broader subject within which they appear. For example, 

195: while discussing the generation of spin polarization, we survey

196: many experimental and theoretical studies of both optical orientation

197: and electrical spin injection and present a detailed and

198: self-contained formalism of electrical spin injection. Similarly,

199: when we discuss spin relaxation, we give a catalog 

200: of important work, while studying spin relaxation in 

201: the cases of Al and GaAs as representative of the whole field.

202: Finally, in the section on spin devices we give detailed

203: physical principles of several selected devices, such as, 

204: for example, the above-mentioned Datta-Das SFET.

205: 

206: There have been many other reviews written on spintronics, 

207: most focusing on a particular aspect of the field. We divide

208: them here, for an easier orientation, into two groups,

209: those that cover the emerging applications\footnote{Reviews on

210: emerging application include those of 

211: \cite{Wolf2000:IEEE,

212: DasSarma2000:SM,DasSarma2000:IEEE,DasSarma2000:DRC,%

213: DasSarma2001:SSC,DasSarma2001:AS,Zutic2002:JS,Rashba2002:JS,Wolf2001:S,%

214: Oestreich2002:SST,Zutic2002:PROC}.} 

215: and those covering already well-established schemes 

216: and materials\footnote{Established schemes and materials are

217: reviewed by 

218: \cite{Prinz1995:PT,Prinz1998:S,Gregg1997:JMMM,Bass1999:JMMM,%

219: Ansermet1998:JPCM,Gijs1997:AP,Tedrow1994:PR,Daughton1999:JPDAP,Stiles2003:P}.}

220: The latter group, often described as {\it magnetoelectronics} 

221: typically covers paramagnetic and ferromagnetic metals and insulators,

222: which utilize magnetoresistive effects, 

223: realized, for example, as

224: magnetic read heads in computer hard drives, nonvolatile magnetic

225: random access memory (MRAM), and circuit isolators

226: \cite{Wang2002:JAP}. These more established aspects of spintronics

227: have been also addressed in several books\footnote{See, for example, the

228: books of 

229: \cite{Hartmann:2000,Parkin:2002,Shinjo:2002,Maekawa:2002,Hirota:2002,%

230: Levy:2002,Chtchelkanova:2003,Ziese:2001}}

231: and will be discussed in another review,\footnote{In preparation

232: by S. S. P. Parkin for Review Modern Physics.} complementary to

233: ours.

234: 

235: Spintronics also benefits from a large class of emerging materials, such as

236: ferromagnetic semiconductors \cite{Ohno1998:S,Pearton2003:JAP},

237: organic semiconductors \cite{Dediu2002:SSC}, organic ferromagnets 

238: \cite{Pejakovic2002:PRL,Epstein2003:MRS},

239: high temperature superconductors \cite{Goldman1999:JMMM},

240: and carbon nanotubes \cite{Tsukagoshi1999:N,Zhao2002:APL},

241: which can bring novel functionalities to the traditional devices.

242: There is a continuing need for fundamental studies before the

243: potential of spintronic applications is fully realized.

244: 

245: After an overview, Sec.~\ref{sec:I} covers some basic historical and 

246: background material, part of which has already  been extensively covered 

247: in the context of magnetoelectronics

248: and will not be discussed further in this review. Techniques 

249: for generating spin polarization, focusing on optical spin orientation and

250: electrical spin injection, are described in Sec.~\ref{sec:II}. 

251: The underlying mechanisms responsible for the

252: loss of spin orientation and coherence, which impose fundamental limits on 

253: the length  and time scales in spintronic devices, are addressed in 

254: Sec.~\ref{sec:III}.

255: Spintronic applications and devices, with the emphasis on 

256: those based on semiconductors, are discussed in Sec.~\ref{sec:IV}.

257: The review concludes with a look at future prospects

258: in Sec.~\ref{sec:V}

259: and with the table (Tab. II) listing the most common abbreviations 

260: used in the text. 

261: 

262: 

263: \subsection{\label{sec:IB} History and background}

264: %-----------------------------------------------------------

265: 

266: \subsubsection{\label{sec:IB1} 

267: Spin-polarized transport and magnetoresistive effects}

268: 

269: In a pioneering work, \textcite{Mott1936:PRCa,Mott1936:PRCb} provided

270: a basis for our understanding of spin-polarized transport.  Mott

271: sought an explanation for an unusual behavior of resistance in

272: ferromagnetic metals. He realized that at sufficiently low

273: temperatures, where magnon scattering becomes vanishingly small,

274: electrons of majority and minority spin, with magnetic moment parallel

275: and antiparallel to the magnetization of a ferromagnet, respectively,

276: do not mix in the scattering processes. The conductivity can then be

277: expressed as the sum of two independent and unequal parts for two

278: different spin projections--the current in ferromagnets is

279: spin polarized.  This is also known as the two-current model and has

280: been extended by \textcite{Campbell1967:PM,Fert1968:PRL}. It 

281: continues, in its modifications, to provide an explanation for various

282: magnetoresistive phenomena \cite{Valet1993:PRB}.

283: 

284: Tunneling measurements played a key role in early experimental work on

285: spin-polarized transport. Studying N/F/N junctions, where N was a

286: nonmagnetic\footnote{Unless explicitly specified, we shall use the

287: terms ``nonmagnetic'' and ``paramagnetic'' interchangeably, i.e.,

288: assume that they both refer to a material with no long-range

289: ferromagnetic order and with Zeeman-split carrier spin subbands in

290: an applied magnetic field.} metal and F was an Eu-based ferromagnetic

291: semiconductor \cite{Kasuya1968:RMP,Nagaev:1983}, revealed that I-V

292: curves could be modified by an applied magnetic field

293: \cite{Esaki1967:PRL} and show potential for developing a solid-state

294: spin-filter. When unpolarized current is passed across a 

295: ferromagnetic semiconductor, the current becomes spin-polarized

296: \cite{Moodera1988:PRL,Hao1990:PRB}.

297: 

298: \begin{figure}

299: \centerline{\psfig{file=zutic_fig02.eps,width=\linewidth,angle=0}}

300: \caption{Schematic illustration of electron tunneling in

301: ferromagnet/insulator/ferromagnet (F/I/F) tunnel junctions: 

302: (a) Parallel and (b) antiparallel orientation of magnetizations 

303: with the corresponding spin-resolved density of the d states

304: in ferromagnetic metals that have exchange spin splitting $\Delta_{ex}$.

305: Arrows in the two ferromagnetic regions are determined by the

306: majority-spin subband. Dashed lines depict spin-conserved tunneling.

307: }

308: \label{intro:1}

309: \end{figure}

310: 

311: A series of experiments

312: \cite{Tedrow1971:PRLa,Tedrow1973:PRB,Tedrow1994:PR} in

313: ferromagnet/insulator/superconductor (F/I/S) junctions has unambiguously

314: proved that the tunneling current remains spin-polarized even outside

315: of the ferromagnetic region.\footnote{It has been shown

316: that electrons photoemitted from ferromagnetic gadolinium remain 

317: spin polarized \cite{Busch1969:PRL}.}

318: The Zeeman-split quasiparticle density of

319: states in a superconductor \cite{Tedrow1970:PRL,Fulde1973:AP}

320: was used as a detector of spin polarization of conduction electrons in

321: various magnetic materials.  

322: \textcite{Julliere1975:PL} measured  

323: tunneling conductance of F/I/F junctions, where I was 

324: an amorphous Ge. 

325: By adopting the

326: \textcite{Tedrow1971:PRLa,Tedrow1973:PRB} analysis 

327: of the tunneling

328: conductance from F/I/S to the F/I/F junctions,  

329: \textcite{Julliere1975:PL} formulated a model for a change of

330: conductance between the parallel ($\uparrow \uparrow$) and

331: antiparallel ($\uparrow \downarrow$) magnetization in the two

332: ferromagnetic regions F1 and F2, as depicted in Fig.~\ref{intro:1}.

333: The corresponding tunneling magnetoresistance\footnote{Starting with

334: \textcite{Julliere1975:PL} an equivalent expression $(G_{\uparrow

335: \uparrow}-G_{\uparrow \downarrow})/G_{\uparrow \uparrow}$ has also also

336: used by different authors and often referred to as junction

337: magnetoresistance (JMR) \cite{Moodera1999:JMMM}.}  

338: (TMR) in an 

339: F/I/F 

340: magnetic tunnel   

341: junction (MTJ) is defined as 

342: \begin{equation}

343: TMR= \frac{\Delta R}{R_{\uparrow \uparrow}}

344: =\frac{R_{\uparrow \downarrow}-R_{\uparrow \uparrow}}{R_{\uparrow \uparrow}}  

345: =\frac{G_{\uparrow \uparrow}-G_{\uparrow \downarrow}}{G_{\uparrow \downarrow}}, 

346: \label{eq:tmr}

347: \end{equation}

348: where conductance G and resistance R=1/G are labeled by the relative

349: orientations of the magnetizations in F1 and F2 (it is possible

350: to change the relative orientations, between $\uparrow \uparrow$ and

351: $\uparrow \downarrow$, even at small applied magnetic fields $\sim$ 10

352: G). 

353: TMR is a particular manifestation of a magnetoresistance (MR) that 

354: yields a change of electrical resistance in the presence of an  

355: external magnetic field.\footnote{The concept of TMR was proposed

356: independently by R. C. Barker in 1975 [see \textcite{Meservey1983:JMMM}] 

357: and by \textcite{Slonczewski1976:IBM}, 

358: who envisioned its use for 

359: magnetic bubble memory \cite{Parkin:2002}.}

360: Historically, the anisotropic

361: MR in bulk ferromagnets such as Fe and Ni was discovered

362: first, dating back the to experiments of Lord Kelvin

363: \cite{Thomson1857:PRSL}.  Due to spin-orbit interaction, electrical

364: resistivity changes with the relative direction of the charge current

365: (for example, parallel or perpendicular) with respect to the direction

366: of magnetization.

367: 

368: Within Julli{\`{e}}re's model, which assumes constant tunneling matrix elements

369: and that electrons tunnel without spin flip, Eq.~(\ref{eq:tmr}) yields

370: \begin{equation}

371: TMR=\frac{2P_1 P_2}{1-P_1 P_2},

372: \label{eq:julliere}

373: \end{equation}

374: where the polarization $P_i=({\cal N}_{M i} - {\cal N}_{m i})/

375: ({\cal N}_{M i} + {\cal N}_{m i})$

376: is expressed in terms of

377: the spin-resolved density of states ${\cal N}_{M i}$ and 

378: ${\cal N}_{m i}$, for majority and minority spin in F$_i$, respectively. 

379: Conductance in Eq.~(\ref{eq:tmr}) can then 

380: be expressed as

381: \cite{Maekawa1982:IEEE} $G_{\uparrow \uparrow} \sim {\cal N}_{M 1}

382: {\cal N}_{M 2}+ {\cal N}_{m 1} {\cal N}_{m 2}$ and $G_{\uparrow

383: \downarrow} \sim {\cal N}_{M 1} {\cal N}_{m 2}+ {\cal N}_{m 1}

384: {\cal N}_{M 2}$ to give Eq.~(\ref{eq:julliere}).\footnote{In

385: ~\ref{sec:IV} we address some limitations of the Julli{\`{e}}re's

386: model and its potential ambiguities to identify precisely which spin

387: polarization is actually measured.}  While the early results of

388: \textcite{Julliere1975:PL} were not confirmed, TMR at 4.2 K was

389: observed using NiO as a tunnel barrier by \textcite{Maekawa1982:IEEE}.

390: 

391: The prediction of Julli{\`{e}}re's model illustrates the spin-valve effect:

392: the resistance of a device can be changed by manipulating the relative

393: orientation of the magnetizations {\bf M$_1$} and {\bf M$_2$}, in

394: F1 and F2, respectively.  

395: Such  orientation 

396: can be preserved even in the absence of a power

397: supply and the spin-valve effect,\footnote{The term was coined by

398: \textcite{Dieny1991:PRB} in the context of GMR,

399: by invoking an analogy with the physics of the TMR.}  later

400: discovered in multilayer structures displaying the giant

401: magnetoresistance (GMR) effect,\footnote{The term ``giant'' reflected the

402: magnitude of the effect 

403: (more than $\sim 10$ \%), as compared to the better known anisotropic

404: magnetoresistance ($\sim 1$ \%).} \cite{Baibich1988:PRL,Binasch1989:PRB} can be

405: used for nonvolatile memory applications

406: \cite{Hartmann:2000,Parkin:2002,Hirota:2002}. GMR structures are often

407: classified according whether the current flows parallel (CIP) or

408: perpendicular (CPP) to the interfaces between the different layers,

409: as depicted in Fig.~\ref{gmr:1}.  Most of the GMR applications use the

410: CIP geometry, while the CPP version, first realized by

411: \cite{Pratt1991:PRL}, is easier to analyze theoretically

412: \cite{Gijs1997:AP,Levy:2002} and relates to the physics of the TMR

413: effect \cite{Mathon1997:PRB}. The size of magnetoresistance in the GMR

414: structures can be expressed analogously to  Eq.~(\ref{eq:tmr}), where

415: parallel and antiparallel orientations of the magnetizations in the

416: two ferromagnetic regions are often denoted by ``P'' and ``AP,''

417: respectively (instead of $\uparrow \uparrow$ and $\uparrow

418: \downarrow$). 

419: Realization of a large room temperature GMR 

420: \cite{Parkin1991:PRL,Parkin1991:APL} enabled a quick transition from basic 

421: physics to commercial applications in magnetic recording \cite{Parkin2003:PIEEE}.

422: 

423: One of the keys to the success of the

424: MR-based-applications is their ability to 

425: control\footnote{For example, with small magnetic field \cite{Parkin:2002} 

426: or at high

427: switching speeds \cite{Schumacher2003:PRLa,Schumacher2003:PRLb}.} the

428: relative orientation of {\bf M$_1$} and {\bf M$_2$}. An interesting

429: realization of such control was proposed independently by 

430: \textcite{Berger1996:PRB} and \textcite{Slonczewski1996:JMMM}. 

431: While in GMR or TMR structures the relative orientation of magnetizations

432: will affect the flow of spin-polarized current, they predicted

433: a reverse effect. The flow of spin-polarized current can transfer angular

434: momentum from carriers to ferromagnet and alter the orientation 

435: of the corresponding magnetization, even in the absence of an applied

436: magnetic field.

437: This phenomenon, known as spin-transfer

438: torque, has since been extensively studied both theoretically and 

439: experimentally

440: \cite{Bazily1998:PRB,Tsoi1998:PRLa,Myers1999:S,Sun2000:PRB,%Tsoi1998:PRLb

441: Stiles2002:PRB,Wanital2000:PRB}

442: and current-induced magnetization reversal was demonstrated at room 

443: temperature \cite{Katine2000:PRL}. 

444: It was also shown that the magnetic field generated by passing the

445: current through a CPP GMR device could produce room temperature

446: magnetization reversal \cite{Bussman1999:APL}.

447: \begin{figure}

448: \centerline{\psfig{file=zutic_fig03.eps,width=0.33\linewidth,angle=-90}}

449: \caption{Schematic illustration of (a) the current in plane (CIP)

450: (b) the current perpendicular to the plane (CPP) giant magnetoresistance

451: geometry.} 

452: \label{gmr:1}

453: \end{figure}

454: In the context of ferromagnetic semiconductors additional control of

455: magnetization was demonstrated optically (by shining light)

456: \cite{Koshihara1997:PRL,Oiwa2002:PRL,Boukari2002:PRL} and electrically

457: (by applying gate voltage)

458: \cite{Ohno2000:N,Park2002:S,Boukari2002:PRL} to perform switching

459: between the ferromagnetic and paramagnetic states.

460: 

461: 

462: Julli{\`{e}}re's model also justifies the  

463: continued quest for highly spin-polarized materials -- they

464: would provide large magnetoresistive effects, desirable for device 

465: applications. In an extreme case, spins would be completely polarized even in

466: the absence of magnetic field. Numerical support for the existence of such

467: materials--the so called half-metallic ferromagnets\footnote{Near the Fermi

468: level they behave as metals only for one spin, the density of states 

469: vanishes completely for the other spin.} was provided 

470: by \textcite{deGroot1983:PRL},

471: and these materials were reviewed by \textcite{Pickett2001:PT}.

472: In addition to ferromagnets, such as CrO$_2$ \cite{Soulen1998:S,Parker2002:PRL}

473: and manganite perovskites \cite{Park1998:N}, there is evidence for high

474: spin polarization in III-V ferromagnetic semiconductors like 

475: (Ga,Mn)As \cite{Braden2003:P,Panguluri2003:P}.

476: The challenge remains

477: to preserve such spin polarization above room temperature and in junctions

478: with other materials, since the surface (interface) and bulk magnetic

479: properties can be significantly different 

480: \cite{Falicov1990:JMR,Fisher1967:PRL,Mills1971:PRB}. 

481: 

482: While many existing spintronic applications

483: \cite{Hartmann:2000,Hirota:2002} are based on the GMR effects, the

484: discovery of large room-temperature TMR

485: \cite{Moodera1995:PRL,Miyazaki1995:JMMM} has renewed interest in the

486: study of magnetic tunnel junctions 

487: which are now the basis for the

488: several MRAM prototypes\footnote{Realization of the early MRAM

489: proposals used the effect of anisotropic magnetoresistance

490: \cite{Pohm1987:IEEETM,Pohm1988:IEEETM}.} 

491: \cite{Parkin1999:JAP,Tehrani2000:IEEE}. 

492: Future generations of magnetic read heads are expected to  

493: use MTJ's instead of CIP GMR.

494: To improve the switching

495: performance of related devices it is important to reduce the junction

496: resistance, which determines the RC time constant of the MTJ cell.

497: Consequently, semiconductors, which would provide a lower tunneling

498: barrier than the usually employed oxides, are being investigated both as the

499: non-ferromagnetic region in MTJ's and as the basis for an

500: all-semiconductor junction that would demonstrate large TMR at low

501: temperatures \cite{Tanaka2002:SST,Tanaka2001:PRL}. Another

502: desirable property of semiconductors has been demonstrated by

503: the extraordinary large room-temperature MR in hybrid

504: structures with metals reaching 750 000\% at a magnetic field of 4 T 

505: \cite{Solin2000:S} which could lead to improved magnetic

506: read heads \cite{Solin2002:APL,Moussa2003:JAP}. MR effects of

507: similar magnitude have also been found in hybrid metal/semiconductor

508: granular films \cite{Akinaga2002:SST}. Another 

509: approach to obtaining large room-temperature magnetoresistance 

510: ($>100$\% at $B\sim 100$ G)

511: is to fabricate ferromagnetic regions separated by 

512: a nanosize contact. For simplicity, such a structure could be thought of 

513: as the limiting case of the CPP GMR scheme in Fig.~\ref{gmr:1}(b).

514: This behavior, also known as ballistic magnetoresistance, has already 

515: been studied in a large number of materials and geometries 

516: \cite{Garcia1999:PRL,Tatara1999:PRL,Imamura2000:PRL,Chung2002:PRL,%

517: Bruno1999:PRL,Versluijs2002:PRL}.

518: 

519: 

520: \subsubsection{\label{sec:IB2} 

521: Spin injection and optical orientation}

522: 

523: Many  materials in their ferromagnetic state can have a substantial degree of 

524: {\it equilibrium} carrier spin polarization. However, as illustrated

525: in Fig.~\ref{fig:DD}, this alone is 

526: usually  not sufficient for spintronic applications, which typically 

527: require current 

528: flow and/or manipulation of the {\it nonequilibrium} spin 

529: (polarization).\footnote{Important exceptions are tunneling devices operating 

530: at low bias

531: and near {\it equilibrium} spin. Equilibrium polarization 

532: and the current flow can be potentially realized, for example,

533: in spin-triplet superconductors and thin-film ferromagnets 

534: \cite{Konig2001:PRL},

535: accompanied by dissipationless spin currents. 

536: Using an analogy with the quantum Hall effect, it has been suggested that the 

537: spin-orbit interaction could lead to dissipationless spin currents in

538: hole-doped semiconductors \cite{Murakami2003:P}. 

539: \textcite{Rashba2003:PRB} has pointed out that similar dissipationless spin

540: currents in thermodynamic equilibrium, due to spin-orbit interaction,

541: are not transport currents which could be employed for transporting spins

542: and spin injection.

543: It is also 

544: instructive to compare several earlier proposals that use spin-orbit coupling 

545: to generate spin currents, discussed in Sec.~\ref{sec:IIA}.

546: } 

547: The importance of generating {\it nonequilibrium} spin is not limited to 

548: device applications; it can also be used as a sensitive spectroscopic

549: tool to study a wide variety of fundamental properties ranging from spin-orbit 

550: and hyperfine interactions \cite{Meier:1984} to the pairing symmetry of high 

551: temperature superconductors 

552: \cite{Tsuei2000:RMP,Vasko1997:PRL,Wei1999:JAP,Ngai2003:P} 

553: and the

554: creation of  spin-polarized beams to measure parity violation in high energy 

555: physics \cite{Pierce:1984}.

556: 

557: Nonequilibrium spin is the result of some source

558: of pumping arising from transport, optical, or resonance methods. 

559: Once the pumping is turned off the spin will return to its equilibrium value.

560: While for most applications it is desirable to 

561: have long spin relaxation times, it has been demonstrated that short spin 

562: relaxation times are useful in the implementation of fast switching 

563: \cite{Nishikawa1995:APL}. 

564: 

565: Electrical spin injection, an example of a transport method for generating 

566: nonequilibrium spin, has already been realized

567: experimentally by

568: \textcite{Clark1963:PRL}, 

569: who drove a direct current

570: through a sample of InSb in the presence of

571: constant applied magnetic filed.

572: The principle was based on the Feher effect,\footnote{The importance and 

573: possible applications

574: of the Feher effect \cite{Feher1959:PRL} to polarize electrons was discussed

575: by \cite{DasSarma2000:IEEE,Suhl2002:P}.}

576: in which the hyperfine coupling between the electron and nuclear spins, 

577: together with

578: different temperatures representing electron velocity  and  

579: electron spin populations, is responsible for the dynamical

580: nuclear polarization \cite{Slichter:1989}.\footnote{Such an effect can be 

581: thought of as a generalization of the 

582: Overhauser effect \cite{Overhauser1953b:PR} in which the use of 

583: a resonant microwave excitation causes 

584: the spin relaxation of the nonequilibrium  

585: electron 

586: population through hyperfine coupling to lead to the spin polarization of 

587: nuclei.

588: \textcite{Feher1959:PRL} suggested several other methods, instead 

589: of microwave excitation, that could 

590: produce a  

591: nonequilibrium electron population and yield a dynamical polarization of 

592: nuclei [see also

593: \textcite{Weger1963:PR}].} 

594: Motivated by the work of \textcite{Clark1963:PRL},

595: \textcite{Tedrow1971:PRLa,Tedrow1973:PRB}, and the principle of

596: optical orientation \cite{Meier:1984},

597: \textcite{Aronov1976:JETPL,Aronov1976:SPJETP} 

598: and \textcite{Aronov1976:SPS} established several key concepts

599: in electrical spin injection from ferromagnets into metals,

600: semiconductors\footnote{In an earlier

601: work spin injection of minority carriers was proposed in a 

602: ferromagnet/insulator/$p$-type semiconductor structure. Measuring polarization

603: of electroluminescence was suggested as a technique for detecting injection

604: of polarized carriers in a semiconductor \cite{Scifres1973:SSC}.}

605: and superconductors. 

606: When a charge current flowed across 

607: the F/N junction (Fig.~\ref{intro:2}) 

608: \textcite{Aronov1976:JETPL} predicted 

609: that spin-polarized carriers in a ferromagnet would

610: contribute to the net current of magnetization entering the nonmagnetic

611: region and would lead to nonequilibrium magnetization $\delta M$, 

612: depicted in Fig.~\ref{intro:2}(b), 

613: with the spatial extent

614: given by the spin diffusion length

615: \cite{Aronov1976:SPS,Aronov1976:JETPL}.\footnote{Supporting the findings of 

616: \textcite{Clark1963:PRL}, Aronov calculated

617: that the electrical spin injection would polarize nuclei and lead to a 

618: measurable effect in the electron spin resonance (ESR). Several decades

619: later related experiments on spin injection are also examining other

620: implications of dynamical nuclear polarization 

621: \cite{Johnson2000:APL,Strand2003:PRL}.} 

622: Such $\delta M$, which is also

623: equivalent to a {\it nonequilibrium} spin accumulation, was first measured 

624: in metals by \textcite{Johnson1985:PRL,Johnson1988:PRBb}.

625: In the steady state $\delta M$ is realized as the balance between spins 

626: added by 

627: the magnetization current and spins removed by spin 

628: relaxation.\footnote{The spin diffusion length 

629: is an important quantity

630: for CPP GMR. The thickness of the N region in Fig.~\ref{gmr:1}

631: should not exceed the spin diffusion length, otherwise the 

632: information on the orientation of the magnetization in F1 will

633: not be transferred to the F2 region.}

634: \begin{figure}

635: \centerline{\psfig{file=zutic_fig04.eps,width=\linewidth,angle=0}}

636: \caption{Pedagogical illustration of the concept of electrical

637: spin injection from a ferromagnet (F) into a normal metal (N).

638: Electrons flow from F to N:

639: (a) schematic device geometry; (b) magnetization M as a function of

640: position. Nonequilibrium magnetization $\delta M$ (spin accumulation)

641: is injected into a normal metal;

642: (c) contribution of 

643: different spin-resolved densities of states to 

644: charge and spin transport across the F/N interface.

645: Unequal filled levels in the density of states depict spin-resolved

646: electrochemical potentials different from the equilibrium value

647: $\mu_0$.}

648: \label{intro:2}

649: \end{figure}

650: 

651: Generation of nonequilibrium spin polarization and spin accumulation

652: is also possible by optical methods known as optical orientation or

653: optical pumping. In optical orientation, the angular momentum of absorbed 

654: circularly polarized 

655: light is transferred to the medium. Electron orbital momenta are directly 

656: oriented by light and through spin-orbit interaction electron spins 

657: become polarized. In ~\ref{sec:IIB} we focus on the optical orientation in 

658: semiconductors, a well-established technique \cite{Meier:1984}.

659: In a pioneering work \textcite{Lampel1968:PRL} demonstrated 

660: that spins in silicon can be optically oriented (polarized). 

661: This technique is derived from optical pumping proposed by 

662: \textcite{Kastler1950:JDP} in which optical irradiation changes the relative

663: populations within the  Zeeman and the hyperfine levels of the ground states 

664: of atoms.

665: While there are similarities with previous studies of free atoms 

666: \cite{Cohen-Tannoudji:1966,Happer1972:RMP},

667: optical orientation in semiconductors has important differences related

668: to the strong coupling between the electron and nuclear

669: spin and macroscopic number of particles 

670: \cite{Paget1977:PRB,Hermann1985:APF,Meier:1984}.  

671: Polarized nuclei can exert large magnetic 

672: fields ($\sim 5$ T) 

673:  on electrons. 

674: In bulk III-V semiconductors, such as GaAs, optical orientation

675: can lead to 50\% polarization of electron density which could be further 

676: enhanced 

677: in quantum structures of reduced dimensionality or by applying a strain.

678: A simple reversal in the polarization of the illuminating light (from

679: positive to negative helicity) also reverses the sign of the electron density

680: polarization. Combining these properties of optical orientation with the 

681: semiconductors tailored to have a negative electron affinity allows 

682: photoemission of spin-polarized electrons to be used as a powerful detection 

683: technique in high-energy physics and for investigating surface 

684: magnetism \cite{Pierce:1984}.

685: 

686: 

687: \section{\label{sec:II} Generation of spin polarization}

688: %====================================================================

689: 

690: \subsection{\label{sec:IIA} Introduction}

691: %--------------------------------------------------------------------

692: 

693: Transport, optical, and resonance methods (as well as their combination) 

694:  have all been used to create nonequilibrium spin. 

695: After introducing the concept of spin polarization in solid-state systems

696: we give a pedagogical picture of electrical spin injection and

697: detection of polarized carriers. While electrical spin injection

698: and optical orientation will be discussed in more detail later in

699: this section, we also survey here several other techniques for polarizing

700: carriers. 

701: 

702: Spin polarization not only of electrons, but also of holes, nuclei, 

703: and excitations can be defined as 

704: \begin{equation}

705: P_X=X_s/X,

706: \label{eq:polar}

707: \end{equation}                                                                     

708: the ratio of the difference 

709: $X_s=X_\lambda-X_{-\lambda}$ and the sum $X=X_\lambda+X_{-\lambda}$, 

710: of the spin-resolved   

711: $\lambda$ components for a particular quantity $X$. 

712: To avoid ambiguity as to what precisely is meant by spin polarization 

713: both the choice of the spin-resolved components and the relevant

714: physical quantity $X$ need to be specified.

715: Conventionally, $\lambda$ is taken to be $\uparrow$ or $+$

716: (numerical value +1) for spin up, $\downarrow$ or $-$ (numerical

717: value -1) for spin down, with respect to the chosen axis of 

718: quantization.\footnote{For example, 

719: along the spin angular momentum, applied magnetic field, magnetization, or 

720: direction of light propagation.}

721: In ferromagnetic metals it is customary to refer to $\uparrow$ ($\downarrow$) 

722: as carriers with magnetic moment parallel (antiparallel) to the magnetization 

723: or, 

724: equivalently, as carriers with majority (minority) spin \cite{Tedrow1973:PRB}. 

725: In semiconductors the terms majority and minority usually refer to relative 

726: populations of the carriers while $\uparrow$ or $+$ and $\downarrow$ or $-$

727: correspond to the quantum numbers $m_j$ with respect to the $z$-axis taken 

728: along the direction of the light propagation or along the applied magnetic 

729: field \cite{Meier:1984,Jonker2003:P}.

730: It is important to emphasize that both the magnitude and the sign of

731: the spin polarization in Eq.~(\ref{eq:polar}) depends of the choice of 

732: $X$, relevant to the detection technique employed, say optical vs. transport 

733: and bulk vs. surface measurements \cite{Mazin1999:PRL,Jonker2003:P}.

734: Even 

735: in the same homogeneous material the measured $P_X$ can vary for

736: different $X$, and it is crucial to identify which physical 

737: quantity---charge current, carrier density, conductivity, 

738: or the density of states---is being measured experimentally.

739: 

740: \begin{figure}

741: \centerline{\psfig{file=zutic_fig05.eps,width=\linewidth,angle=0}} 

742: \caption{Spin injection, spin accumulation, and spin detection:

743: (a) two idealized completely polarized ferromagnets F1 and F2

744: (the spin down density of states ${\cal N}_\downarrow$ is zero

745: at the energy of electrochemical potential $E=\mu_0$) 

746: with parallel

747: magnetizations are separated by the nonmagnetic region N;

748: (b) density-of-states diagrams for spin injection from F1 into N,

749: accompanied by the spin accumulation--generation of the nonequilibrium

750: magnetization $\delta M$. At F2

751: in the limit of low impedance ($Z$=0) 

752: spin is detected

753: by measuring the spin-polarized

754: current across the N/F2 interface. In the limit of high impedance 

755: ($Z=\infty$) spin is detected

756: by measuring the voltage $V_s\sim \delta M$ developed across the N/F2 

757: interface;

758: (c) spin accumulation in a device in which  

759: a superconductor (with the superconducting gap $\Delta$) is

760: occupying the region between F1 and F2.}

761: \label{inj:1}

762: \end{figure}                                                                                       

763: The spin polarization of electrical current or carrier density, 

764: generated in a nonmagnetic region,

765: is typically used to describe the efficiency of electrical spin injection.

766: \textcite{Silsbee1980:BMR}  suggested that the nonequilibrium 

767: density polarization in the N region, or equivalently the

768: nonequilibrium magnetization, acts as the source of spin electromotive force 

769: (EMF) and produces a measurable ``spin-coupled'' 

770: voltage $V_s \propto \delta M$. Using this concept, also referred to

771: as  {\it spin-charge coupling},

772: \textcite{Silsbee1980:BMR} proposed a detection technique  

773: consisting of two ferromagnets F1 and F2 (see Fig.~\ref{inj:1})

774: separated by a nonmagnetic region.\footnote{A similar geometry was also 

775: proposed independently by \textcite{deGroot1983:PA}, where F1 and F2 were two

776: half-metallic ferromagnets with the goal of implementing  spin-based

777: devices to amplify and/or switch current.} 

778: F1 serves as the spin injector (spin aligner)

779: and F2 as the spin detector. This could be called the polarizer-analyzer 

780: method, 

781: the optical counterpart of the transmission of light through 

782: two optical linear polarizers.

783: From  Fig.~\ref{inj:1} it follows that

784: the reversal of the magnetization

785: direction in one of the ferromagnets would lead either to 

786: $V_s \rightarrow -V_s$,

787: in an open circuit (in the limit of large impedance Z), or to the reversal of

788: charge current $j \rightarrow -j$, in a short circuit (at small Z), 

789: a consequence

790: of Silsbee-Johnson spin-charge coupling 

791: \cite{Silsbee1980:BMR,Johnson1987:PRB,Johnson1988:PRBa}.

792: Correspondingly, as discussed in the following sections,

793: the spin injection could be detected through the spin accumulation signal 

794: either as a voltage or the resistance change when the magnetizations 

795: in F1 and F2 are changed from parallel to antiparallel alignment. 

796: 

797: Since the experiments demonstrating

798: the spin accumulation of conduction electrons

799: in metals \cite{Johnson1985:PRL}, spin injection has been realized

800: in a wide range of materials. While in Sec.~\ref{sec:IIC} we focus on related

801: theoretical work motivated by potential applications, experiments on

802: spin injection have also stimulated proposals for examining the fundamental

803: properties of electronic systems.\footnote{For example, studies          

804: probing the  spin-charge separation

805: in the non-Fermi liquids have been proposed by

806: \cite{Kivelson1990:PRB,Zhao1995:PRB,Si1997:PRL,Si1998:PRL,Balents2000:PRL,Balents2001:PRB}. 

807: Spin and charge are carried by

808: separate excitations and can lead to spatially separated spin and 

809: charge currents \cite{Kivelson1990:PRB}.}

810: 

811: The generation of nonequilibrium spin polarization has a long

812: tradition in magnetic resonance methods

813: \cite{Abragam:1961,Slichter:1989}. However, transport methods

814: to generate carrier spin polarization are not limited to electrical

815: spin injection. For example, they also include scattering of unpolarized

816: electrons in the presence of spin-orbit coupling 

817: \cite{Mott:1965,Kessler:1976} and in materials that lack the

818: inversion symmetry \cite{Levitov1985:SPJETP}, adiabatic 

819: \cite{Mucciolo2002:PRL,Sharma2003:PRB,Watson2003:PRL} and nonadiabatic  

820: quantum spin pumping \cite{Zheng2002:P} [for an instructive description

821: of parametric pumping see \cite{Brouwer1998:PRB}],

822: and the proximity effects \cite{Ciuti2002:PRL}.  

823: 

824: It would be interesting to know what the limits are on

825: the magnitude of various spin polarizations. Could we have a completely

826: polarized current [$P_j\rightarrow \infty$, see Eq.~(\ref{eq:polar})],

827: with only a spin current ($j_\uparrow-j_\downarrow$) and no charge current

828: ($j_\uparrow+j_\downarrow=0$)?

829: While it is tempting to

830: recall the Stern-Gerlach experiment and try to set up magnetic

831: drift through  inhomogeneous magnets \cite{Kessler:1976}, 

832: this would most likely work

833: only as a transient effect \cite{Fabian2002b:PRB}. It was proposed

834: already by \textcite{Dyakonov1971:PL,Dyakonov1971:JETPLa} that a

835: transverse spin current (and transverse spin polarization in a

836: closed sample) would form as a result of spin-orbit coupling-induced

837: skew scattering in the presence of a longitudinal electric field.

838: This interesting effect, also called the {\it spin Hall effect} 

839: \cite{Hirsch1999:PRL,Zhang2000:PRL},

840: has yet to be demonstrated. An alternative scheme for producing pure

841: spin currents was proposed by \textcite{Bhat2000:PRL},

842: motivated by the experimental demonstration of

843: phase coherent control of charge currents 

844: \cite{Atanasov1996:PRL,Hache1997:PRL}

845: and carrier population \cite{Fraser1999:PRL}. 

846: A quantum-mechanical

847: interference between one- and two-photon absorptions of orthogonal

848: linear polarizations creates an opposite ballistic flow of spin up and

849: spin down electrons in a semiconductor. Only a spin

850: current can flow, without a charge current, as

851: demonstrated by \textcite{Stevens2003:PRL} and \textcite{Hubner2003:PRL}, 

852: who were able to achieve coherent 

853: control of the spin current direction and magnitude by the

854: polarization and the relative phase of two exciting laser light fields.

855: 

856: Charge current also can be driven by circularly polarized

857: light \cite{Ivchenko:1997}. 

858: Using the principles of optical orientation (see Sec.~\ref{sec:IB2}

859: and 

860: further discussion in Sec.~\ref{sec:IIB}) in semiconductors of reduced 

861: dimensionality or lower symmetry, 

862: both  the direction and  the magnitude 

863: of a generated charge current 

864: can be controlled by

865: circular polarization of the light. 

866: This is called the circular photo-voltaic effect \cite{Ganichev2003:JPCM},

867: which can be viewed as a transfer of the angular momentum of photons to 

868: directed motion of electrons. This could also be called a spin

869: corkscrew effect, since a nice mechanical analog is a corkscrew

870: whose rotation generates linear directed motion. A related effect,

871: in which the photocurrent is driven, is called the spin-galvanic effect

872: \cite{Ganichev2003:JPCM}. The current here is causes by the difference

873: in spin-flip scattering rates for electrons with different spin states

874: in some systems with broken inversion symmetry.

875: A comprehensive survey of the related effects 

876: from the circular photo-galvanic effect \cite{Asnin1979:SSC}

877: to recent demonstrations in semiconductor quantum wells 

878: \cite{Ganichev2001:PRL,Ganichev2002:PRL,Ganichev2002:N,Ganichev2003:PRB}]

879: is given by \textcite{Ganichev2003:JPCM}.

880: 

881: There is a wide range of recent theoretical proposals for devices

882: that would give rise to a spin electromotive force 

883: \cite{Long2002:APL,Zutic2001:APL,Zutic2001:PRB,Governale2003:PRB,%

884: Brataas2002:PRB,Malshukov2003:PRB,Ting2003:APL}, 

885: often

886: referred to as spin(-polarized) pumps, cells, or batteries. However,

887: even when it is feasible to generate pure spin current, this does 

888: not directly imply that 

889: it would be dissipationless. 

890: In the context of superconductors, it has been 

891: shown that Joule heating can arise from pure spin current 

892: flowing through a Josephson junction \cite{Takahashi2001:JAP}.

893: 

894: \subsection{\label{sec:IIB} Optical spin orientation}

895: %-----------------------------------------------------------------------------

896: 

897: 

898: \begin{figure}

899: \centerline{\psfig{file=zutic_fig06.eps,width=\linewidth,angle=0}}

900: \caption{(a) Interband transitions in GaAs: 

901: (a) schematic band structure of GaAs near the center 

902: of the Brillouin zone ($\Gamma$ point), where E$_g$ 

903: is the band gap and $\Delta_{so}$ is

904: the spin-orbit splitting; 

905: CB, conduction band; HH, valence heavy hole; LH, light hole; SO,

906: spin-orbit-split-off 

907: subbands; $\Gamma_{6,7,8}$ are the corresponding symmetries at the $k=0$ 

908: point, more precisely, the irreducible 

909: representations of the tetrahedron group $T_d$ \cite{Ivchenko:1997};

910: (b) selection rules for interband transitions between 

911: $m_j$ sublevels 

912: for circularly polarized light $\sigma^+$ 

913: and $\sigma^-$ (positive and negative helicity). The circled numbers

914: denote the relative transition intensities that apply for both

915: excitations (depicted by the arrows) and radiative recombinations.} 

916: \label{oo:1}

917: \end{figure}

918: 

919: In a semiconductor the photoexcited spin-polarized electrons and holes 

920: exist for the time $\tau$ before they recombine. If a fraction of the 

921: carriers' initial 

922: orientation survives longer than the recombination time, that is, 

923: if $\tau < \tau_s$, \footnote{In Si this condition is not

924: fulfilled. Instead of measuring the luminescence polarization, 

925: \textcite{Lampel1968:PRL} has used NMR to detect optical spin orientation.}

926: where $\tau_s$ is the spin relaxation time (see Sec.~\ref{sec:III}), 

927: the luminescence  (recombination radiation) will be partially polarized.  

928: By measuring the circular polarization of the luminescence it is 

929: possible to study the spin dynamics of the nonequilibrium carriers 

930: in semiconductors \cite{Oestreich2002:SST} and to extract such 

931: useful quantities as the spin orientation, the recombination time, or  

932: the spin relaxation time of the carriers

933: \cite{Parsons1969:PRL,Garbuzov1971:ZhETF,Ekimov1970:ZhETF,Meier:1984}.

934: 

935: We illustrate the basic principles of optical orientation

936: by the example of GaAs which is representative of a large

937: class of III-V and II-VI zincblende semiconductors. The

938: band structure is depicted in Fig.~\ref{oo:1}(a).

939: The band gap is $E_g=1.52$ eV at $T=0$  $K$, while the spin 

940: split-off band is separated from the light and 

941: heavy hole bands by $\Delta_{so}=0.34$ eV. 

942: We denote the Bloch states according to the total angular

943: momentum $J$ and its projection onto the positive

944: $z$ axis $m_j$: $|J,m_j \rangle$. Expressing the wave

945: functions with the symmetry of $s$, $p_x$, $p_y$, and $p_z$

946: orbitals as $|S\rangle $, $|X\rangle $, $|Y\rangle$, and $|Z\rangle$, 

947: respectively, the band wave functions can be written as listed 

948: in Table~\ref{tab:1} 

949: \cite{Pierce1976:PRB} [with minor typos removed, 

950: see also \cite{Kittel:1963}]. 

951: 

952: To obtain the excitation (or recombination) probabilities  

953: consider photons arriving in the $z$ direction. 

954: Let $\sigma^{\pm}$ represent the helicity of the exciting light. 

955: When we represent the dipole operator corresponding to the 

956: $\sigma^{\pm}$ optical transitions as \footnote{For an outgoing light 

957: in the $-z$ direction the helicities are reversed.}

958: $\propto (X\pm iY) \propto Y^{\pm 1}_1$, where $Y^m_l$  is the spherical

959: harmonic, it follows from Table~\ref{tab:1} that  

960: \begin{equation}

961: \label{eq:three}

962: \frac{|\langle 1/2,-1/2 | Y^1_1 |3/2, -3/2 \rangle|^2}

963: {|\langle 1/2,1/2 | Y^1_1 |3/2, -1/2 \rangle|^2}=3

964: \end{equation}

965: for the relative intensity of the $\sigma^+$ transition

966: between the heavy ($|m_j=3/2|$) and the light ($|m_j=1/2|$) hole subbands 

967: and the conduction band.

968: Other transitions are analogous. The relative transition rates

969: are indicated  in Fig.~\ref{oo:1}(b). The same selection rules apply

970: to the optical orientation of shallow impurities 

971: \cite{Parsons1969:PRL,Ekimov1970:ZhETF}. 

972: 

973: 

974: \begin{table}

975: \begin{tabular}{lll}

976: \hline

977: \hline

978: & &  \\

979: Symmetry   &  $|J,m_j \rangle$ & Wave function  \\

980: & &  \\

981: \hline 

982: $\Gamma_6$ & $|1/2,1/2 \rangle$ &  $|S \! \uparrow  \rangle$  \\

983:  & $|1/2,-1/2 \rangle$ &  $|S \! \downarrow  \rangle$  \\

984: \hline

985: $\Gamma_7$ & $|1/2,1/2 \rangle$ &  

986: $|-(1/3)^{1/2}[\, (X+iY) \! \downarrow  - Z \! \uparrow] \, \rangle$  \\

987:  & $|1/2,-1/2 \rangle$ &  

988: $|(1/3)^{1/2}[\, (X-iY) \! \uparrow  + Z \! \downarrow]  \, \rangle$  \\

989: \hline

990: $\Gamma_8$ & $|3/2,3/2 \rangle$ &  

991: $|(1/2)^{1/2}(X+iY) \! \uparrow \rangle$  \\

992:  & $|3/2,1/2 \rangle$ &  

993: $|(1/6)^{1/2}[\, (X+iY) \! \downarrow  + 2 Z \! \uparrow] \, \rangle$  \\

994:  & $|3/2,-1/2 \rangle$ &  

995: $|-(1/6)^{1/2}[\, (X-iY) \! \uparrow  - 2 Z \! \downarrow] \, \rangle$  \\

996:  & $|3/2,-3/2 \rangle$ &  

997: $|(1/2)^{1/2}(X-iY) \! \downarrow \rangle$  \\

998: \hline

999: \hline

1000: \end{tabular}

1001: \caption{Angular and spin part of the wave function at $\Gamma$.}

1002: \label{tab:1}

1003: \end{table}

1004:   

1005: The spin polarization of the excited electrons\footnote{Although holes

1006: are initially polarized too, they lose spin orientation very

1007: fast, on the time scale of the momentum relaxation time 

1008: (see Sec.~\ref{sec:IIID1})

1009: However, it was suggested that manipulating hole spin by short electric

1010: field pulses, between momentum scattering events, could be useful for

1011: ultrafast spintronics applications \cite{Dargys2002:PRB}.

1012: }.

1013: depends on the photon energy $\hbar\omega$. For $\hbar\omega$ between $E_g$ 

1014: and $E_g+\Delta_{so}$, only the light and heavy hole subbands contribute. 

1015: Denoting by $n_{+}$ and $n_{-}$ the density of electrons polarized 

1016: parallel ($m_j=1/2$) and

1017: antiparallel ($m_j=-1/2$) to the direction of light propagation, we

1018: define the 

1019: spin polarization as (see Sec.~\ref{sec:IIA})

1020: \begin{equation} \label{eq:pn}

1021: P_n=(n_+-n_-)/(n_++n_-).

1022: \end{equation}

1023: For our example of the zincblende structure, 

1024: \begin{equation}

1025: P_n=(1-3)/(3+1)=-1/2

1026: \end{equation}

1027: is the spin polarization at the moment of photoexcitation. The spin is 

1028: oriented against the direction of light propagation, since there are more 

1029: transitions from the heavy hole than from the light hole subbands. 

1030: The circular polarization of the luminescence is defined as

1031: \begin{equation}

1032: \label{eq:circ}

1033: P_{\rm circ}=(I^+ -I^-)/(I^+ + I^-),

1034: \end{equation}

1035: where $I^\pm$ is the radiation intensity for the helicity $\sigma^{\pm}$.

1036: The polarization of the $\sigma^+$ photoluminescence is then

1037: \begin{equation}

1038: P_{\rm circ}=\frac{(n_+ + 3n_-)-(3n_+ + n_-)}{(n_+ + 3n_-)+(3n_+ + n_-)}

1039:              =-\frac{P_n}{2}=\frac{1}{4}.

1040: \label{eq:1/4}       

1041: \end{equation}

1042: 

1043: If the excitation involves transitions from the spin split-off band, 

1044: that is, if  $\hbar \omega \gg E_g + \Delta_{so}$, the electrons

1045: will not be spin polarized ($P_n=P_{\rm circ}=0$), underlining the 

1046: vital role of spin-orbit coupling for spin orientation.  On the

1047: other hand, Fig.~\ref{oo:1} 

1048: suggests that a removal of the heavy/light

1049: hole degeneracy can substantially increase $P_n$ \cite{Dyakonov:1984}, 

1050: up to the limit of complete spin polarization. 

1051: An increase in $P_n$ and $P_{\rm circ}$

1052: in GaAs strained due to a lattice mismatch with a substrate, or due

1053: to confinement in quantum well heterostructures,

1054: has indeed been demonstrated \cite{Oskotski1997:PLDS,Vasilev1993:SM},

1055: detecting $P_n$ greater than 0.9. 

1056: 

1057: While photoexcitation with  circularly polarized light creates 

1058: spin-polarized electrons, the nonequilibrium spin decays due to both 

1059: carrier recombination and spin relaxation. The steady-state degree

1060: of spin polarization depends on the balance between the spin 

1061: excitation and decay. Sometimes a distinction is made 

1062: \cite{Pierce1976:PRB,Meier:1984}

1063: between the terms optical {\it spin orientation} and optical {\it spin 

1064: pumping}. The former term is used in relation to the minority carriers

1065: (such as electrons in p-doped samples) and represents the 

1066: orientation of the excited carriers. The latter term is reserved for the

1067: majority carriers (electrons in n-doped samples), representing

1068: spin polarization of the ``ground'' state. Both spin orientation

1069: and spin pumping were demonstrated in the early investigations

1070: on p-GaSb \cite{Parsons1969:PRL} and  p- and n-Ga$_{0.7}$Al$_{0.3}$As 

1071: \cite{Ekimov1970:ZhETF,Ekimov1971:JETPL,Zakharchenya1971:JETPL}. Unless

1072: specified otherwise, we shall use the term optical orientation to 

1073: describe both

1074: spin orientation and spin pumping.

1075: 

1076: 

1077: To derive the steady-state expressions for the spin polarization due to optical

1078: orientation, consider the simple model of carrier recombination and spin 

1079: relaxation

1080: (see Sec.~\ref{sec:IVA4}) in a homogeneously doped semiconductor.

1081: The balance between direct electron-hole recombination and optical 

1082: pair creation can be written as 

1083: \begin{equation} 

1084: \label{eq:np}

1085: r(np-n_0p_0)=G,

1086: \end{equation}

1087: where $r$ measures the recombination rate, the electron and hole densities are 

1088: $n$ and

1089: $p$, with index zero denoting the equilibrium values, 

1090: and $G$ is the electron-hole

1091: photoexcitation rate. Similarly, the balance between spin relaxation and 

1092: spin generation is expressed by

1093: \begin{equation} 

1094: \label{eq:sp}

1095: rsp+s/\tau_s=P_{n}(t=0) G,

1096: \end{equation}

1097: where $s=n_+-n_-$ 

1098: is the electron spin density and $P_{n}(t=0)$ is the spin polarization

1099: at the moment of photoexcitation, given by Eq.~(\ref{eq:pn}). 

1100: Holes are assumed to lose their

1101: spin orientation very fast, so they are treated as unpolarized. The first

1102: term in Eq.~(\ref{eq:sp}) describes the disappearance of the spin density 

1103: due to carrier recombination, while the second term describes the

1104: intrinsic spin relaxation. From Eqs.~(\ref{eq:np}) and (\ref{eq:sp}) we 

1105: obtain the

1106: steady-state electron polarization as \cite{Zutic2001:PRB}

1107: \begin{equation} \label{eq:p}

1108: P_n=P_n(t=0)\frac{1-n_0p_0/np}{1+1/\tau_srp}.

1109: \end{equation}

1110: 

1111: 

1112: In a p-doped sample $p\approx p_0$, $n\gg n_0$, and Eq.~(\ref{eq:p}) gives 

1113: \begin{equation}

1114: \label{eq:orient}

1115: P_n=P_n(t=0)/(1+\tau/\tau_s),

1116: \end{equation}

1117: where $\tau=1/rp_0$ is the electron lifetime.\footnote{After the illumination

1118: is switched off, the electron spin density, or equivalently the nonequilibrium

1119: magnetization, will decrease exponentially with the inverse time constant

1120: $1/T_s=1/\tau+1/\tau_s$ \cite{Parsons1969:PRL}.}

1121: The steady-state polarization is independent of the illumination intensity,

1122: being reduced from the initial spin polarization $P_n(t=0)$.

1123: \footnote{The effect of a finite length for the light absorption on $P_n$ 

1124: is discussed 

1125: by \textcite{Pierce:1984}. The absorption length $\alpha^{-1}$ is

1126: typically a micron for GaAs.   It varies with frequency

1127: roughly as $\alpha(\hbar \omega) \propto (\hbar \omega - E_g)^{1/2}$ 

1128: \cite{Pankove:1971}.}

1129: The polarization of the photoluminescence is 

1130: $P_{\rm circ}=P_n(t=0) P_n$ \cite{Parsons1969:PRL}.

1131: Early measurements of  $P_n=0.42 \pm 0.08$ in  GaSb \cite{Parsons1969:PRL} 

1132: and 

1133: $P_n=0.46 \pm 0.06$  in  Ga$_{0.7}$Al$_{0.3}$As \cite{Ekimov1970:ZhETF}

1134: showed an effective spin orientation close to the maximum 

1135: value of $P_n(t=0)=1/2$ for a bulk unstrained  zincblende structure,

1136: indicating that $\tau/\tau_s \ll 1$. 

1137: 

1138: For spin pumping in an n-doped sample, where $n\approx n_0$ and $p\gg p_0$, 

1139: Eqs.~(\ref{eq:np}) and (\ref{eq:p}) give \cite{Dyakonov1971:JETPL}

1140: \begin{equation}

1141: \label{eq:pump}

1142: P_n=P_n(t=0)/(1+n_0/G\tau_s).

1143: \end{equation}

1144: In contrast to the previous case, the carrier (now hole) lifetime 

1145: $\tau=1/rn_0$

1146: has  no effect on $P_n$. However, $P_n$ depends on the

1147: photoexcitation intensity $G$, as expected for a pumping process. 

1148: The effective

1149: carrier lifetime is $\tau_J=n_0/G$, 

1150: where $J$ represents the intensity of the illuminating light. 

1151: If it is comparable to or shorter than $\tau_s$, 

1152: spin pumping is very effective. Spin pumping works because the photoexcited 

1153: spin-polarized electrons do not need to recombine with holes. There

1154: are plenty of unpolarized electrons in the conduction band available

1155: for recombination. The spin is thus pumped in to the electron system.

1156: 

1157: 

1158: When magnetic field {\bf B} is applied 

1159: perpendicular to the axis of spin orientation (transverse magnetic field),

1160: it will induce spin precession with the 

1161: Larmor frequency $\Omega_L=\mu_B g B/\hbar$, where $\mu_B$ is the Bohr

1162: magneton and $g$ is the electron $g$ factor.\footnote{In our convention

1163: the $g$ factor of free electrons is positive, 

1164: $g_0=2.0023$ \cite{Kittel:1996}.}

1165: The spin precession, together with the random character of carrier generation

1166: or diffusion, leads to the spin dephasing (see Sec.~\ref{sec:IIIA1}).

1167: Consider spins excited by circularly polarized light (or by any means

1168: of spin injection) at a steady rate. In a steady rate a balance between

1169: nonequilibrium spin generated and spin relaxation is maintained, resulting

1170: in a net magnetization. If a transverse magnetic field is applied,

1171: the decrease of the steady-state magnetization can have two sources:

1172: (a) spins which were excited at random time and (b) random diffusion

1173: of spins towards a detection region. Consequently, spins precess along

1174: the applied field acquiring random phases relative to those which were

1175: excited or have arrived at different times.

1176: As a result, the projection

1177: of the electron spin along the exciting beam will decrease with the 

1178: increase of transverse magnetic field, leading to depolarization of the  

1179: luminescence. 

1180: This is also known as the Hanle effect \cite{Hanle1924:ZP}, in analogy

1181: to the depolarization of the resonance fluorescence of gases. The Hanle

1182: effect was first measured in semiconductors by \textcite{Parsons1969:PRL}. 

1183: The steady-state spin polarization of the precessing electron spin 

1184: can be calculated by solving the Bloch-Torrey equations 

1185: \cite{Bloch1946:PR,Torrey1956:PR}, 

1186: Eqs.~(\ref{eq:relax:bloch1})--(\ref{eq:relax:bloch3}) describing the 

1187: spin dynamics

1188: of diffusing carriers.

1189: 

1190: In p-doped semiconductors the Hanle curve 

1191: shows a Lorentzian decrease of the polarization \cite{Parsons1969:PRL},

1192: $P_n(B)=P_n(B=0)/(1+\Omega_L T_s)^2$,

1193: where $P_n(B=0)$ is the polarization at $B=0$ from Eq.~(\ref{eq:orient})

1194: and $T_s^{-1}$ is the effective spin lifetime given by 

1195: $1/T_s=1/\tau+1/\tau_s$;

1196: see footnote 26.

1197: Measurements of the Hanle curve in GaAlAs were used 

1198: by  \textcite{Garbuzov1971:ZhETF}    

1199: to separately determine both $\tau$ and $\tau_s$ at various temperatures.

1200: The theory of the Hanle effect in n-doped semiconductors was developed by

1201: \textcite{D'yakonov1976:FTP} who showed the  

1202: non-Lorentzian decay of the luminescence for the

1203: regimes  both of low ($\tau_J / \tau_s \gg  1$) 

1204: and high ($\tau_J / \tau_s \ll 1$) intensity of the exciting light. 

1205: At high fields $P_n(B)\propto 1/B^{1/2}$, consistent

1206: with the experiments of \textcite{Vekua1976:FTP} in Ga$_{0.8}$Al$_{0.2}$As, 

1207: showing a Hanle curve different from the usual $P_n(B)\propto 1/B^2$ Lorentzian 

1208: behavior \cite{Dyakonov:1984}. Recent findings on the Hanle effect in

1209: nonuniformly doped GaAs and reanalysis of some earlier studies are

1210: given by \textcite{Dzihoev2003:FTT}.

1211: 

1212: \subsection{\label{sec:IIC} Theories of spin injection}

1213: %-------------------------------------------------------------------------------

1214: 

1215: Reviews on spin injection have covered materials ranging from semiconductors 

1216: to high temperature superconductors and have 

1217: addressed the implications for device 

1218: operation as well as for fundamental studies in solid 

1219: state systems.\footnote{See, for example, 

1220: \cite{Johnson2002:SST,Schmidt2002:SST,Jedema2002:JS,Wei2002:JS,%

1221: Goldman1999:JMMM,Goldman2001:JS,Tang:2002,Johnson2001:JS,Osofsky2000:JS,%

1222: Maekawa2001:MSE}}.  

1223: In addition to degenerate conductors, examined in these works, we also give

1224: results for nondegenerate semiconductors in which the violation of local 

1225: charge neutrality, electric fields, and carrier band bending require

1226: solving the Poisson equation. The notation introduced here emphasizes the

1227: importance of different (and inequivalent) spin polarizations arising in 

1228: spin injection.

1229: 

1230: \subsubsection{\label{sec:IIC1} F/N junction}

1231: %---------------------------------------------------------------------------

1232: 

1233: 

1234: A theory of spin injection across a ferromagnet/normal metal (F/N) interface

1235: was first offered by \textcite{Aronov1976:JETPL}. Early work also

1236: included spin injection into a semiconductor (Sm) 

1237: \cite{Aronov1976:SPS,Masterov1979:SPS}

1238: and a superconductor (S) \cite{Aronov1976:SPJETP}. Spin injection in 

1239: F/N junctions

1240: was subsequently studied in detail by 

1241: \textcite{Johnson1987:PRB,Johnson1988:PRBa},\footnote{Johnson and Silsbee base 

1242: their approach on irreversible thermodynamics and consider also  the effects of 

1243: a temperature gradient on spin-polarized transport, omitted in this section.} 

1244: \textcite{vanSon1987:PRL},

1245: \textcite{Valet1993:PRB}, \textcite{Hershfield1997:PRB}, and others.

1246: Here we follow the approach of \textcite{Rashba2000:PRB,Rashba2002:EPJ} and 

1247: consider

1248: a steady-state\footnote{Even some dc spin injection experiments are actually

1249: performed at low (audio-frequency) bias. Generalization to ac spin injection, 

1250: with a harmonic time dependence, was studied by \textcite{Rashba2002:APL}.}

1251: flow of electrons along the $x$ direction in a three-dimensional (3D) 

1252: geometry consisting of a metallic ferromagnet (region $x<0$) and a 

1253: paramagnetic metal or

1254: a degenerate semiconductor (region $x>0$).

1255: 

1256: \begin{figure}

1257: \centerline{\psfig{file=zutic_fig07.eps,width=0.9\linewidth,angle=0}}

1258: \caption{Spatial variation of the electrochemical potentials near a

1259: spin-selective resistive interface at an F/N junction. At the interface $x=0$

1260: both the spin-resolved electrochemical potentials ($\mu_\lambda$, 

1261: $\lambda=\uparrow,\downarrow$, denoted with solid lines) and the average 

1262: electrochemical potential ($\mu_{F}$, $\mu_{N}$, dashed lines) are 

1263: discontinuous. The spin diffusion length $L_{sF}$ and $L_{sN}$

1264: characterizes the decay of  $\mu_s=\mu_\uparrow - \mu_\downarrow$ 

1265: (or equivalently the decay of spin accumulation and  

1266: the nonequilibrium magnetization) away from the interface and into the bulk 

1267: F and N regions, respectively.} 

1268: \label{inj:2}

1269: \end{figure}

1270: The two regions, F and N, form a contact at $x=0$,

1271: as depicted in Fig.~\ref{inj:2}.

1272: The relative magnitudes of three characteristic resistances  

1273: per unit area\footnote{For this simple

1274: geometry various resistances have a common factor of the cross-sectional area, 

1275: which

1276: can be factored out. This is no longer possible for a more complicated geometry 

1277: \cite{Takahashi2003:PRB}.} determine the degree of current polarization 

1278: injected 

1279: into a nonmagnetic material. These are the contact resistance $r_c$ 

1280: and the two characteristic 

1281: resistances $r_N$ $r_F$, each given by the ratio of the spin diffusion 

1282: length and the 

1283: effective bulk conductivity in the corresponding region.

1284: Two limiting cases  correspond to the transparent limit, 

1285: where $r_c\rightarrow 0$, and the low-transmission limit, where 

1286: $r_c \gg r_N,r_F$.

1287: 

1288: Spin-resolved quantities are labeled by 

1289: $\lambda=1$ or $\uparrow$ for spin up, $\lambda=-1$ or $\downarrow$ 

1290: for spin down along the chosen quantization axis. 

1291: For a free electron, spin angular momentum

1292: and magnetic moment are in opposite directions, and what

1293: precisely is denoted by ``spin up'' varies in the 

1294: literature \cite{Jonker2003:P}.

1295: Conventionally, in metallic systems 

1296: \cite{Tedrow1973:PRB,Gijs1997:AP},

1297: spin up refers to carriers with majority spin. This means that the

1298: spin (angular momentum) of such carriers is

1299: antiparallel to the magnetization.

1300: Spin-resolved charge current (density)

1301: in a diffusive regime can be expressed as 

1302: \begin{equation}

1303: \label{eq:jq}

1304: j_\lambda=\sigma_\lambda \nabla \mu_\lambda,

1305: \end{equation}

1306: where $\sigma_\lambda$ is conductivity and the electrochemical potential is

1307: \begin{equation}

1308: \label{eq:elchem}

1309: \mu_\lambda=(q D_\lambda/\sigma_\lambda) \delta n_\lambda-\phi,

1310: \end{equation}

1311: with $q$ proton charge, $D_\lambda$ diffusion coefficient, 

1312: $\delta n_\lambda=n_\lambda-n_{\lambda 0}$ the change of electron density 

1313: from the equilibrium value for spin $\lambda$, 

1314: and $\phi$ the electric potential.~\footnote{More generally, 

1315: for a noncollinear magnetization, 

1316: $j_\lambda$ becomes a second-rank tensor 

1317: \cite{Johnson1988:PRBa,Margulis1994:PB,Stiles2002:PRB}.}

1318: 

1319: In the steady state the continuity equation is

1320: \begin{equation}

1321: \label{eq:jcont}

1322: \nabla j_\lambda=

1323: \lambda q \left[\frac{\delta n_\lambda}{\tau_{\lambda-\lambda}}

1324:                -\frac{\delta n_{-\lambda}}{\tau_{-\lambda \lambda}} \right],

1325: \end{equation}

1326: and $\tau_{\lambda\lambda'}$ is the average time for flipping a $\lambda$-spin

1327: to $\lambda'$-spin.

1328: For a degenerate conductor\footnote{In the nondegenerate case of

1329: Boltzmann statistics, the Einstein relation implies 

1330: that the ratio of the diffusion coefficient and the mobility is $k_BT/q$.} 

1331: the Einstein relation 

1332: is

1333: \begin{equation}

1334: \sigma_\lambda=q^2 {\cal N}_\lambda D_\lambda, 

1335: \label{eq:einstein}

1336: \end{equation}

1337: where $\sigma=\sigma_\uparrow+\sigma_\downarrow$ and 

1338: ${\cal N}={\cal N}_\uparrow+{\cal N}_\downarrow$

1339: is the density of states. Using a detailed balance 

1340: ${\cal N}_\uparrow/ \tau_{\uparrow \downarrow}=

1341: {\cal N}_\downarrow/ \tau_{\downarrow \uparrow}$  

1342: \cite{Hershfield1997:PRB,Kravchenko2002:JETP} 

1343: together with Eqs.~(\ref{eq:elchem}) and (\ref{eq:einstein}),

1344: the continuity equation can be expressed as

1345: \begin{equation}

1346: \label{eq:jq'}

1347: \nabla j_\lambda=\lambda q^2 \frac{{\cal N}_\uparrow {\cal N}_\downarrow}

1348: {{\cal N}_\uparrow+{\cal N}_\downarrow}

1349: \frac{\mu_\lambda-\mu_{-\lambda}}{\tau_s},

1350: \end{equation}

1351: where $\tau_s=\tau_{\uparrow \downarrow} \tau_{\downarrow \uparrow}/ 

1352: (\tau_{\uparrow \downarrow}+\tau_{\downarrow \uparrow})$ 

1353: is the spin relaxation time. Equation (\ref{eq:jq'})  

1354: implies the conservation of charge 

1355: current $j=j_\uparrow+j_\downarrow=const.$, while the spin counterpart,

1356: the difference of the spin-polarized currents 

1357: $j_s=j_\uparrow-j_\downarrow$ is position dependent. 

1358: Other ``spin quantities,'' $X_s$,

1359: unless explicitly defined, are analogously expressed 

1360: with the corresponding (spin) polarization given by $P_X=X_s/X$.

1361: For example, the current polarization\footnote{This is 

1362: also referred to as a

1363: spin injection coefficient \cite{Rashba2000:PRB,Rashba2002:EPJ}.} 

1364: $P_j=j_s/j$, generally different

1365: from the density polarization

1366: $P_n=(n_\uparrow-n_\downarrow)/n$, is related to the conductivity polarization 

1367: $P_\sigma$ as 

1368: \begin{equation}

1369: \label{eq:Pj}

1370: P_j=2(\sigma_\uparrow \sigma_\downarrow/\sigma) \nabla \mu_s/j+P_\sigma

1371: \end{equation}

1372: where $\mu_s=\mu_\uparrow-\mu_\downarrow$. In terms of the

1373: average electrochemical potential $\mu=(\mu_\uparrow+\mu_\downarrow)/2$, 

1374: $P_\sigma$ further satisfies

1375: \begin{equation}

1376: \label{eq:mu}

1377: \nabla \mu=-P_\sigma \nabla \mu_s/2+j/\sigma.

1378: \end{equation}

1379: From Eqs.~(\ref{eq:elchem}) and (\ref{eq:jq'}) it follows that $\mu_s$

1380: satisfies the diffusion equation

1381: \cite{vanSon1987:PRL,Valet1993:PRB,Hershfield1997:PRB,Schmidt2000:PRB} 

1382: \begin{equation}

1383: \label{eq:mus}

1384: \nabla ^2 \mu_s=\mu_s/L_s^2,

1385: \end{equation}

1386: where the spin diffusion length is $L_s=(\overline{D}\tau_s)^{1/2}$ with

1387: the spin averaged diffusion coefficient 

1388: $\overline{D}=(\sigma_\downarrow D_\uparrow 

1389: +\sigma_\uparrow D_\downarrow)/\sigma

1390: ={\cal N}({\cal N}_\downarrow/D_\uparrow+{\cal N}_\uparrow/D_\downarrow)^{-1}$. 

1391: Using Eq.~(\ref{eq:elchem}) and the local charge quasineutrality 

1392: $\delta n_\uparrow+\delta n_\downarrow=0$

1393: shows that $\mu_s$ is proportional to the nonequilibrium 

1394: spin density $\delta s=\delta n_\uparrow - \delta n_\downarrow$ 

1395: ($s=s_0+\delta s=n_\uparrow-n_\downarrow$)

1396: \begin{equation}

1397: \label{eq:spin}

1398: \mu_s=\frac{1}{2 q} \frac{{\cal N}_\uparrow+{\cal N}_\downarrow} 

1399: {{\cal N}_\uparrow {\cal N}_\downarrow}

1400: \delta s.

1401: \end{equation}                                                                 

1402: Correspondingly,  $\mu_s$ is often referred to as the (nonequilibrium) 

1403: {\it spin 

1404: accumulation}\footnote{Spin accumulation is also relevant to a number of

1405: physical phenomena outside the scope of this article, for example, to the

1406: tunneling rates in the quantum Hall regime 

1407: \cite{MacDonald1999:PRL,Chan1999:PRL}.} 

1408: and is used to explain the GMR effect in CPP structures 

1409: \cite{Johnson1991:PRL,Valet1993:PRB,Hartmann:2000,Hirota:2002,Gijs1997:AP}. 

1410: 

1411: The preceding equations are simplified for the N region 

1412: by noting that 

1413: $\sigma_\lambda=\sigma/2$, $\sigma_s=0$, and 

1414: $D_\lambda=\overline{D}$.

1415: Quantities pertaining to a particular region are denoted by the index 

1416: F or N.

1417: 

1418: Equation ~(\ref{eq:mus}) has also been used to study the diffusive 

1419: spin-polarized transport and spin accumulation in  ferromagnet/superconductor 

1420: structures \cite{Jedema1999:PRB}. 

1421: Some care is needed to establish the appropriate

1422: boundary conditions at the F/N interface.

1423: In the absence of spin-flip scattering\footnote{The effects

1424: of non-conserving interfacial scattering on spin injection were considered 

1425: in \cite{Valet1993:PRB,Fert1996:PRB,Rashba2002:EPJ}.}

1426: at the F/N interface (which can arise, for example, 

1427: due to spin-orbit coupling or magnetic impurities) 

1428: the spin current is continuous 

1429: and thus $P_{jF}(0^-)=P_{jN}(0^+)\equiv P_j$ (omitting

1430: $x=0^\pm$ for brevity, and superscripts $\pm$ in other quantities). 

1431: These boundary conditions were used by 

1432: \textcite{Aronov1976:SPS,Aronov1976:JETPL} 

1433: without relating $P_j$ to the effect of the F/N contact or 

1434: material parameters in the F region.

1435: 

1436: Unless the F/N contact is highly transparent, $\mu_\lambda$ is discontinuous

1437: across the interface 

1438: \cite{Johnson1988:PRL,Valet1993:PRB,Hershfield1997:PRB,Rashba2000:PRB} 

1439: and the boundary condition is

1440: \begin{equation}

1441: \label{eq:jbc}

1442: \ j_\lambda(0)=\Sigma_\lambda [\mu_{\lambda N}(0)-\mu_{\lambda F}(0)],

1443: \end{equation}

1444: where 

1445: \begin{equation}

1446: \label{eq:sigma}

1447: \Sigma=\Sigma_\uparrow+\Sigma_\downarrow

1448: \end{equation}

1449: is the contact conductivity. For a free-electron model 

1450: $\Sigma_\uparrow \neq \Sigma_\downarrow$ can be simply 

1451: inferred from the effect of the exchange energy, which would 

1452: yield spin-dependent Fermi wave vectors and transmission

1453: coefficients. A microscopic determination of the 

1454: corresponding contact resistance [see Eq.~(\ref{eq:rc})]

1455: is complicated by the influence of disorder,

1456: surface roughness, and different scattering mechanisms and

1457: is usually obtained from  model calculations 

1458: \cite{Schep1997:PRB,Stiles2000:PRB}. Continued work on

1459: the first-principles calculation of F/N interfaces

1460: \cite{Stiles1996:JAP,Erwin2002:PRB} is needed for 

1461: a more detailed understanding of spin injection.

1462: From Eqs.~(\ref{eq:jbc}) and (\ref{eq:sigma})  

1463: it follows that

1464: \begin{eqnarray}

1465: \label{eq:musbc}

1466: \mu_{sN}(0)-\mu_{sF}(0)=2r_c(P_j-P_\Sigma)j, \\

1467: \label{eq:mubc}

1468: \mu_N(0)-\mu_F(0)=r_c ( 1-P_\Sigma P_j ) j,

1469: \end{eqnarray}

1470: where the effective contact resistance is 

1471: \begin{equation}

1472: \label{eq:rc}

1473: r_c=\Sigma/4\Sigma_\uparrow \Sigma_\downarrow.

1474: \end{equation}

1475: The decay of $\mu_s$, away from the interface, is characterized by the

1476: corresponding spin diffusion length

1477: \begin{equation}

1478: \label{eq:decay}

1479: \mu_{sF}=\mu_{sF}(0) e^{x/L_{sF}}, \quad

1480: \mu_{sN}=\mu_{sN}(0) e^{-x/L_{sN}}.  

1481: \end{equation}

1482: A nonzero value for $\mu_{sN}(0)$ implies the existence of nonequilibrium

1483: magnetization $\delta M$ in the N region (for noninteracting electrons

1484: $q \mu_s=\mu_B \delta M/\chi$, where $\chi$ is 

1485: the magnetic susceptibility).

1486: Such  a $\delta M$, as a result of electrical spin injection, was proposed

1487: by \textcite{Aronov1976:SPS} and first measured in metals by 

1488: \textcite{Johnson1985:PRL}.

1489: 

1490: By applying Eq.~(\ref{eq:Pj}), separately, to the F and N regions, 

1491: one can obtain the amplitude of spin accumulation in terms 

1492: of the current and density of states spin polarization 

1493: and the effective resistances $r_F$ and $r_N$, 

1494: \begin{equation}

1495: \label{eq:mus0}

1496: \mu_{sF}(0)=2r_F \left [P_j-P_{\sigma F} \right ]j, \quad

1497: \mu_{sN}(0)=-2r_N P_j j, 

1498: \end{equation}

1499: where

1500: \begin{equation}

1501: \label{eq:rs}

1502: r_N=L_{sN}/\sigma_N, \quad

1503: r_F=L_{sF} \sigma_F/(4\sigma_{\uparrow F}\sigma_{\downarrow F}).

1504: \end{equation}

1505: From Eqs.~(\ref{eq:mus0}) and 

1506: (\ref{eq:musbc}) the current polarization can be obtained as 

1507: \begin{equation}

1508: \label{eq:gamma}

1509: P_j=\left [r_c P_\Sigma+ r_F P_{\sigma F} \right] /r_{FN},

1510: \end{equation}

1511: where $r_{FN}=r_F+r_c+r_N$ is the effective equilibrium

1512: resistance of the F/N junction. 

1513: It is important to emphasize that a measured highly polarized current,

1514: representing an efficient spin injection, does not itself imply

1515: a large spin accumulation or a large density polarization, typically

1516: measured by optical techniques. In contrast to the derivation of $P_j$

1517: from Eq.~(\ref{eq:gamma}), determining $P_n$

1518: requires using 

1519: Poisson's

1520: equation or a condition of the local charge 

1521: quasineutrality.\footnote{Carrier density

1522: will also be influenced by the effect of screening, which changes with

1523: the dimensionality of the spin injection geometry \cite{Korenblum2002:PRL}.} 

1524: 

1525: It is useful to note\footnote{\textcite{Rashba2002:PC}.}

1526: that Eq.~(\ref{eq:gamma}), written as Eq.~(18) in \cite{Rashba2000:PRB}

1527: can be mapped to Eq.~(A11) from \cite{Johnson1987:PRB}, where it

1528: was first derived.\footnote{The substitutions are $P_j \rightarrow \eta^*$, 

1529: $P_\sigma \rightarrow p$,

1530: $P_\Sigma \rightarrow \eta$, $r_c \rightarrow [G(\xi-\eta^2)]^{-1}$,

1531: $r_N \rightarrow \delta_n/\sigma_n\zeta_n$,

1532: $r_F \rightarrow \delta_f/\sigma_f(\zeta_f-p_f^2)$, 

1533: $L_{sN,F} \rightarrow  \delta_{n,F}$,

1534: and $n,f$ label N and F region, respectively. 

1535: $\eta$, $\zeta_n$, and $\zeta_f$ are of the order of unity.

1536: To ensure that resistances and the spin diffusion lengths in 

1537: \textcite{Johnson1987:PRB} are positive, one must additionally have 

1538: $(\xi-\eta^2)>0$ and

1539: $(\zeta_i-p_i^2)>0$, $i=n,f$ (for normal and ferromagnetic regions, 

1540: respectively). In particular, assuming $\xi=\zeta_n=\zeta_f=1$ 

1541: a detailed correspondence between  Eq.~(\ref{eq:gamma}) and  Eq.~(A11) in 

1542: \cite{Johnson1987:PRB} is recovered. For example,  

1543: $r_c \rightarrow [G(\xi-\eta^2)]^{-1}$

1544: yields Eq.~(\ref{eq:rc}), where $\Sigma \rightarrow G$.}

1545: An equivalent form for $P_j$ in Eq.~(\ref{eq:gamma}) was obtained by 

1546: \textcite{Hershfield1997:PRB} and for $r_c=0$ results from 

1547: \textcite{vanSon1987:PRL} are recovered.

1548: 

1549: In contrast to normal metals \cite{Johnson1985:PRL,Johnson1988:PRBb} 

1550: and superconductors, for which  injection has been reported in 

1551: both  conventional 

1552: \cite{Johnson1994:APL}, and high temperature superconductors 

1553: \cite{Hass1994:PC,Vasko1997:PRL,Dong1997:APL,Yeh1999:PRB},

1554: creating a substantial current polarization by direct electrical spin 

1555: injection 

1556: from a metallic ferromagnet

1557: into a semiconductor proved to be more difficult

1558: \cite{Monzon1999:JMMM,Hammar1999:PRL,Filip2000:PRB,Zhu2001:PRL}.

1559: 

1560: By examining  

1561: Eq.~(\ref{eq:gamma}) we can both infer some possible limitations and

1562: deduce several experimental strategies for effective spin injection

1563: i.e. to increase $P_j$ into semiconductors.

1564: For a perfect Ohmic contact $r_c=0$, the typical resistance mismatch 

1565: $r_F \ll r_N$

1566: (where F is a metallic ferromagnet)

1567: implies inefficient spin injection with $P_j \approx r_F/r_N \ll 1$, 

1568: referred to

1569: as the {\it conductivity mismatch} problem by \textcite{Schmidt2000:PRB}. 

1570: Even in  the absence of the resistive contacts, effective spin injection into 

1571: a semiconductor can be achieved if the resistance mismatch is reduced 

1572: by using for spin injectors either a magnetic semiconductor or a highly 

1573: spin-polarized ferromagnet.\footnote{From Eq.~(\ref{eq:rs}) a half-metallic 

1574: ferromagnet implies a large $r_F$.}  

1575: 

1576: While there was early experimental evidence \cite{Alvarado1992:PRL}

1577: that employing resistive (tunneling) contacts could lead to 

1578: an efficient spin injection\footnote{The influence of the resistive contacts on 

1579: spin injection can also be inferred by 

1580: explicitly considering resistive contacts

1581: \cite{Johnson1987:PRB,Hershfield1997:PRB}.}

1582: a systematic understanding was provided by \textcite{Rashba2000:PRB}

1583: and supported with the subsequent experimental and theoretical studies 

1584: \cite{Smith2001:PRB,Fert2001:PRB,Rashba2002:EPJ,Takahashi2003:PRB,%

1585: Johnson2003:PRB,Johnson:2003}.

1586: As can be seen from Eq.~(\ref{eq:gamma}) the spin-selective resistive 

1587: contact $r_c \gg r_F,r_N$ (such as a tunnel or Schottky 

1588: contact) would contribute to effective spin injection with 

1589: $P_j\approx P_\Sigma$ 

1590: being dominated by 

1591: the effect

1592: $r_c$ and not the ratio $r_F/r_N$.\footnote{A similar result was stated

1593: previously by \textcite{Johnson1988:PRBa}.}

1594: This limit is also instructive to illustrate the principle of 

1595: spin filtering \cite{Esaki1967:PRL,Moodera1988:PRL,Hao1990:PRB,Filip2002:APL}.

1596: In a spin-discriminating transport process

1597: the resulting degree of spin polarization is changed. Consequently

1598: the effect of spin filtering, similar to spin injection, leads  

1599: to the generation of (nonequilibrium) spin 

1600: polarization. \footnote{While most of the schemes resemble a CPP geometry 

1601: [Fig.~\ref{gmr:1}(b)], there are also proposals for generating highly polarized 

1602: currents in a CIP-like geometry [Fig.~\ref{gmr:1}(a)] 

1603: \cite{Gurzhi2001:FNT,Gurzhi2003:P}.}

1604: For example, 

1605: at low temperature EuS and EuSe, discussed in 

1606: Sec.~\ref{sec:IVC}, can act as spin-selective barriers.

1607: In the extreme case,

1608: initially spin-unpolarized carriers (say, injected from a nonmagnetic material)

1609: via spin-filtering could attain a complete polarization.

1610: For a strong spin-filtering contact $P_\Sigma > P_{\sigma F}$, the sign of the

1611: spin accumulation (nonequilibrium magnetization) is reversed in the F and

1612: N regions, near the interface [recall Eq.~(\ref{eq:musbc})], in contrast

1613: to the behavior sketched in Fig.~\ref{inj:2}, where $\mu_{s F,N} > 0$.

1614: 

1615: The spin injection process alters the potential drop across the F/N interface

1616: because differences of spin-dependent electrochemical potentials on

1617: either side of the interface generate an effective resistance $\delta R$.

1618: By integrating Eq.~(\ref{eq:mu}) for N and F regions, separately,

1619: it follows  that $Rj=\mu_N(0)-\mu_F(0)+P_{\sigma F}\mu_{sF}(0)/2$,

1620: where $R$ is the junction resistance.

1621: Using Eqs.~(\ref{eq:mubc}), (\ref{eq:rs}), and (\ref{eq:gamma}) 

1622: allows us to express 

1623: $R=R_0+ \delta R$, where $R_0=1/\Sigma$ ($R_0=r_c$ if 

1624: $\Sigma_\uparrow=\Sigma_\downarrow$) 

1625: is the equilibrium resistance, in the absence of spin injection, and 

1626: \begin{eqnarray} 

1627: \delta R= [r_N ( r_F P^2_{\sigma F} +r_c P^2_{\Sigma} ) 

1628:         +r_F r_c ( P_{\sigma F} - P_{\Sigma})^2]  / r_{FN}, 

1629: \label{eq:delR} 

1630: \end{eqnarray}

1631: where $\delta R>0$ is the nonequilibrium resistance.

1632: Petukhov has shown \cite{Jonker2003:MRS} that Eqs.~(\ref{eq:gamma})

1633: and (\ref{eq:delR}) could be obtained by considering an equivalent

1634: circuit scheme with two resistors 

1635: $\tilde{R}_\uparrow$,

1636: $\tilde{R}_\downarrow$ connected in parallel, where 

1637: $\tilde{R}_\lambda=L_{sF}/\sigma_{\lambda F}+1/\Sigma_\lambda+2L_{sN}/\sigma_N$

1638: and $\tilde{R}_\uparrow + \tilde{R}_\downarrow=4r_{FN}$.

1639: For such a resistor scheme, by noting that 

1640: $j_\uparrow \tilde{R}_\uparrow= j_\downarrow \tilde{R}_\downarrow$,

1641: Eq.~(\ref{eq:gamma}) is obtained as 

1642: $P_j=-P_{\tilde{R}}\equiv-(\tilde{R}_\uparrow-\tilde{R}_\downarrow)

1643: /(\tilde{R}_\uparrow+\tilde{R}_\downarrow)$. $\delta R$ in

1644: Eq.~({\ref{eq:delR}) is then obtained as the difference between the

1645: total resistance of the nonequilibrium spin-accumulation region of the

1646: length $L_{sF}+L_{sN}$ [given by the equivalent resistance 

1647: $\tilde{R}_\uparrow \tilde{R}_\downarrow/ 

1648: (\tilde{R}_\uparrow + \tilde{R}_\downarrow)$] and the equilibrium 

1649: resistance for the same region, $L_{sF}/\sigma_F + L_{sN}/\sigma_N$.

1650: %\textcite{Rashba2002:EPJ} has pointed out that the nontrivial 

1651: %contribution of individual conductivities in Eq.~(\ref{eq:delR})

1652: %is likely to rule out a simple equivalent scheme that would represent

1653: %the effect of spin injection.  In a resistor scheme \cite{Gijs1997:AP}  

1654: %based on the two-current model (see \ref{sec:IB1}), 

1655: %extensively used to describe GMR,

1656: %individual resistances connected in series or in parallel yield 

1657: %positive contributions, unlike the 

1658: %term $-2r_F r_c P_{\sigma F} P_{\Sigma}$, which contributes to $\delta R$ 

1659: %in Eq.~(\ref{eq:delR}).

1660: 

1661: The concept of the excess resistance $\delta R$ can also be explained as 

1662: a consequence of the Silsbee-Johnson spin-charge coupling 

1663: \textcite{Silsbee1980:BMR,Johnson1985:PRL,Johnson1987:PRB}

1664: and illustrated by considering the simplified schemes in  

1665: Figs.~\ref{inj:1} and \ref{inj:2}.

1666: Accumulated spin 

1667: near the F/N interface, 

1668: together with a finite spin relaxation  and a finite spin diffusion,

1669: impedes the 

1670: flow of spins and acts as a ``spin bottleneck'' 

1671: \cite{Johnson1991:PRL}. A rise of $\mu_{sN}$ must be accompanied by the rise of 

1672: $\mu_{sF}$ [their precise alignment at the interface is given in 

1673: Eq.~(\ref{eq:musbc})] 

1674: or there will be a backflow of the nonequilibrium spin back into the F region. 

1675: Because both spin and charge are carried by electrons 

1676: in spin-charge coupling 

1677: the backflow of spin 

1678: driven by diffusion creates an additional resistance for the charge flow 

1679: across the 

1680: F/N interface. Based on an analogy with the charge transport across a 

1681: clean N/superconductor (S) interface (see Sec.~\ref{sec:IVA3}) 

1682: \textcite{vanSon1987:PRL} explained $\delta R$ by invoking

1683: the consequences of current conversion from spin-polarized, at 

1684: far to the left of the F/N interface, to completely unpolarized, at far right 

1685: in the N region.

1686: 

1687: The increase in the total resistance with spin injection can be most

1688: dramatic if the N region is taken to be a superconductor (S);

1689: see Fig.~\ref{inj:1}(c).  

1690: Spin injection depletes the superconducting condensate and can

1691: result in the switching to a normal state of much higher resistance 

1692: \cite{Vasko1997:PRL,Dong1997:APL,Yeh1999:PRB,Wei1999:JAP,Takahashi1999:PRL}.

1693: A critical review of possible spurious effects 

1694: in reported  experiments 

1695: \textcite{Gim2001:JAP} 

1696: has also stimulated the development of a novel detection technique

1697: which uses scanning tunneling spectroscopy combined with 

1698: pulsed quasiparticle  spin injection to minimize 

1699: Joule heating \cite{Ngai2003:P} (see Sec.~\ref{sec:IVA1}).

1700: In the S region the quasiparticle energy is 

1701: $E_k=(\xi_k^2+\Delta^2)^{1/2}$, where $\xi_k$ is the single particle

1702: excitation energy 

1703: corresponding to the wave vector ${\bf k}$

1704: and $\Delta$ is the superconducting gap

1705: [see Fig.~\ref{inj:1}(c)]. Such a dispersion relation results in a

1706: smaller diffusion coefficient and a longer spin flip time than

1707: in the N region, 

1708: while their product, the spin diffusion length, remains the same

1709: \cite{Yamashita2002:PRB}. Consequently,

1710: Eq.~(\ref{eq:mus}) also applies to  the diffusive 

1711: spin-polarized transport and spin accumulation in  ferromagnet/superconductor 

1712: structures \cite{Jedema1999:PRB,Yamashita2002:PRB}. 

1713: Opening of the superconducting gap implies that a superconductor is a

1714: low carrier system for spin, which is carried by quasiparticles 

1715: \cite{Takahashi2003:PRB}.

1716: 

1717: In the preceding analysis, appropriate for bulk, homogeneous,

1718: three-dimensional N and F regions and degenerate (semi)conductors,

1719: Poisson's equation was not invoked and the local charge neutrality

1720: $\delta n_\uparrow+\delta n_\downarrow$ was used only to derive

1721: Eq.~(\ref{eq:spin}).\footnote{For spin injection in

1722: nondegenerate semiconductors (with the carriers obeying the 

1723: Boltzmann statistics)

1724: there can be large effects due to 

1725: built-in fields and deviation

1726: from local charge neutrality, as discussed in Sec.~\ref{sec:IIC3}.}

1727: Focusing on bulk samples in which both the size of the F and N regions and the

1728: corresponding spin diffusion lengths are much larger than the Debye 

1729: screening length, one can find that 

1730: the quasineutrality condition, combined with the Eqs.~(\ref{eq:elchem})

1731: and (\ref{eq:einstein}), yields

1732: \begin{equation}

1733: \label{eq:phi}

1734: \phi=-\mu-P_{\cal N} \mu_s/2,

1735: \end{equation}

1736: where the density of states spin polarization of $P_{\cal N}$ vanishes 

1737: in the N region.

1738: At the contact $x=0$ there is a potential drop, even when $r_c=0$,

1739: which can be evaluated from

1740: Eqs.~(\ref{eq:mubc}) and (\ref{eq:phi}) as

1741: \begin{equation}

1742: \label{eq:contact}

1743: \phi_N(0)-\phi_F(0)=-r_c [1-P_\Sigma P_j] j + P_{{\cal N}F}(0)\mu_{sF}(0)/2.

1744: \end{equation}

1745: The creation of nonequilibrium spin in the N region results in the 

1746: spin EMF in the F/N structure 

1747: which 

1748: can be used to detect electrical spin injection,

1749: as depicted in Fig.~\ref{inj:1}. Within a simplified semi-infinite geometry

1750: for the F and N regions, we consider an effect of spin 

1751: pumping in the N region, realized either by electrical spin injection from

1752: another F region [as shown in Fig.~\ref{inj:1}(b)] or by optical pumping 

1753: (see Sec.~\ref{sec:IIB}).  The resulting potential drop can calculated by

1754: modifying $\mu_{sN}$ in Eq.~(\ref{eq:decay}), 

1755: \begin{equation}

1756: \mu_{sN}=\mu_{sN}(\infty)

1757: +[\mu_{sN}(0)-\mu_{sN}(\infty)]e^{-x/L_{sN}},  

1758: \end{equation}

1759: where $\mu_{sN}(\infty)$ 

1760: represents the effect of homogeneous spin pumping

1761: in the N region. To calculate the open circuit voltage ($j=0$)  the continuity

1762: of spin current at $x=0$ should be combined with the fact that $P_j j=j_s$. 

1763: From Eq.~(\ref{eq:Pj}) it follows that

1764: \begin{equation}

1765: \label{eq:js0}

1766: j_s(0)=2\frac{\sigma_\uparrow \sigma_\downarrow}{\sigma_F} 

1767: \frac{\mu_{sF}(0)}{L_{sF}}=

1768: -\frac{1}{2}\sigma_N \frac{\mu_{sN}(0)-\mu_{sN}(\infty)}{L_{sN}}, 

1769: \end{equation}                                                            

1770: while the discontinuity of $\mu_s$ in Eq.~(\ref{eq:musbc}) 

1771: yields\footnote{A missprint

1772: in $\mu_{sF}(0)$ from \textcite{Rashba2002:EPJ} has been corrected.}

1773: \begin{eqnarray} \nonumber

1774: \mu_{sF}(0)&=&(r_F/r_{FN})\mu_{sN}(\infty), \quad

1775: j_s(0)=\mu_{sN}(\infty)/2 r_{FN}, \\ 

1776: \label{eq:musNF}

1777: \mu_{sN}(0)&=&[(r_c+r_F)/r_{FN}]\mu_{sN}(\infty).

1778: \end{eqnarray} \nonumber

1779: By substituting this solution into Eq.~(\ref{eq:contact}),

1780: we can evaluate

1781: the contact 

1782: potential drop

1783: can be evaluated as

1784: \begin{equation}

1785: \label{eq:contact2}

1786: \phi_N(0)-\phi_F(0)= \left [ r_F P_{{\cal N}F}+r_c P_\Sigma \right ]

1787: \mu_{sN}(\infty)/2 r_{FN}. 

1788: \end{equation}

1789: The total potential drop (recall $j=0$) at the F/N junction\footnote{A similar

1790: potential drop was also calculated across

1791: a ferromagnetic domain wall \cite{Dzero2003:PRB}.} is \cite{Rashba2002:EPJ}

1792: \begin{equation}

1793: \label{eq:drop}

1794: \Delta \phi_{FN}=\phi_N(\infty)-\phi_F(-\infty)=P_j\mu_{sN}(\infty)/2. 

1795: \end{equation}

1796: where $P_j$ is given in Eq.~(\ref{eq:gamma}).

1797: In the context of the spin-detection scheme from Fig.~\ref{inj:1} and high

1798: impedance measurements at the N/F2 junction, the spin-coupled voltage $V_s$ 

1799: \cite{Silsbee1980:BMR,Johnson1985:PRL} was also found to be proportional

1800: to current polarization and the spin accumulation 

1801: ($\mu_s \propto \delta s \propto \delta M$) \cite{Johnson1988:JAP}.

1802: 

1803: 

1804: \subsubsection{\label{sec:IIC2} F/N/F junction}

1805: %---------------------------------------------------------------------------

1806: \label{F/N/F}

1807: 

1808: The above 

1809: analysis of the F/N bilayer can be readily extended to the geometry

1810: in which the infinite F regions are separated by an N region 

1811: of thickness $d$. 

1812: The quantities pertaining to the two ferromagnets

1813: are defined as in the case of an F/N junction and labeled

1814: by the superscripts L and R (left and right regions, respectively).

1815: It follows from Eq.~(\ref{eq:Pj}), 

1816: by assuming the continuity of the spin current at L,R, 

1817: that the difference of the spin-resolved

1818: electrochemical potential, responsible for the spin accumulation, is

1819: \begin{eqnarray} 

1820: \label{eq:musL}

1821: \mu_{sF}^L&=&2 r_F^L (P_j^L-P_{\sigma F}^L) j e^{x/L_{sF}^L}, \quad x<0, \\ 

1822: \nonumber

1823: \mu_{sN}&=&2 r_N \left [ P_j^R \cosh(x/L_{sN}) 

1824: -P_j^L\cosh[(d-x)/L_{sN}] \right  ]  \\   

1825: &\times& j/\sinh(d/L_{sN}), \quad 0<x<d, \\ 

1826: \label{eq:musN}

1827: \mu_{sF}^R&=&-2 r_F^R (P_j^R-P_{\sigma F}^R) j e^{(d-x/L_{sF}^R)}, \quad x>d, 

1828: \label{eq:musR}

1829: \end{eqnarray}

1830: where the current spin polarization $P_j^{L,R}$ at the two contacts

1831: in the F/N/F geometry can be expressed \cite{Rashba2002:EPJ} in terms of 

1832: the $P_j$ calculated for F/N junction with the infinite F and N regions

1833: in Eq.~(\ref{eq:Pj}) and the appropriate effective resistances. 

1834: By $P_{j\infty}^{L,R}$ we denote the $P_j$ calculated in Eq.~(\ref{eq:gamma})

1835: for at left

1836: and right contact (with the appropriate parameters for the F/N/F junction)

1837: as if it were surrounded by the infinite F and N regions.

1838: Analogously to the F/N junction, the consequence of the spin injection

1839: is the increase of the resistance $R=R_0+\delta R$, as compared to the 

1840: equilibrium value $R_0=(\Sigma^L)^{-1}+(\Sigma^R)^{-1}$. 

1841: The nonequilibrium resistance $\delta R$ is also always

1842: positive for spin-conserving contacts \cite{Rashba2002:EPJ,Rashba2000:PRB},

1843: in agreement with experiments on all-semiconductor trilayer structures

1844: \cite{Schmidt2001:PRL}; see Sec.~\ref{sec:IID3}.

1845: 

1846: Many applications 

1847: based on magnetic multilayers rely on the spin-valve effect

1848: in which

1849: the resistance changes due to the relative orientations of the magnetization

1850: in the two F regions. The geometry considered here is relevant for CPP GMR 

1851: \cite{Parkin:2002,Bass1999:JMMM,Gijs1997:AP} and all-metallic

1852: spin injection of \textcite{Johnson1985:PRL}.

1853: In particular, the resistance change between 

1854: antiparallel and parallel magnetization orientations in the two 

1855: ferromagnets can

1856: be expressed using current polarization of an infinite F/N junction 

1857: $P_{j\infty}^{L,R}$

1858: \cite{Rashba2002:EPJ}:

1859: \begin{equation}

1860: \label{eq:DelR}

1861: \Delta R= R_{\uparrow \downarrow} - R_{\uparrow \uparrow} =

1862: 4 P_{j\infty}^L P_{j\infty}^R \frac{r_{FN}^L r_{FN}^R r_N}{{\cal D} 

1863: \sinh(d/L_{sN})}, 

1864: \end{equation}

1865: where $r_{F}^{L,R}$, $r_c^{L,R}$, and $r_N$ are defined as in 

1866: the case of an F/N junction

1867: and

1868: \begin{eqnarray}

1869: \label{eq:det}

1870: {\cal D}&=&(r_F^L+r_c^L)(r_c^R+r_F^R)  \\ \nonumber

1871: &+& r_N^2 +r_N(r_F^L+r_c^L+r_c^R+r_F^R) \coth(d/L_{s N}).  

1872: \end{eqnarray}

1873: Up to a factor of 2, Eq.~(\ref{eq:DelR}) has also been obtained by 

1874: \textcite{Hershfield1997:PRB} using Onsager relations. 

1875: In the limit of a thin N region, 

1876: $d/L_{sN} \rightarrow 0$, 

1877: $\Delta R$ remains finite. In the opposite limit, for $d \gg L_{sN}$, 

1878: \begin{equation}

1879: \Delta R \sim  P_{j\infty}^L P_{j\infty}^R \exp(-d/L_{sN}).

1880: \label{rd}

1881: \end{equation}

1882: For a symmetric F/N/F junction, 

1883: where $r_{c,F}^{L}=r_{c,F}^{R}$, it follows that

1884: \begin{equation}

1885: \label{eq:DelRS}

1886: \Delta R=\frac{4 r_N (r_c P_\Sigma + r_F P_{\sigma F})}

1887: { {\cal D} \sinh(d/L_{sN})}. 

1888: \end{equation}

1889: 

1890: Considering the spin injection from F into a ballistic 

1891: N region in the presence of diffusive interfacial scattering,

1892: where the phase coherence is lost and the Boltzmann equation can be 

1893: applied,

1894: it is 

1895: instructive to reconsider the effect of contact 

1896: resistance \cite{Kravchenko2003:PRB}. We introduce the Sharvin resistance 

1897: $R_{\rm Sharvin}$ 

1898: \cite{Sharvin1965:ZETF}, arising in ballistic transport between the 

1899: two infinite

1900: regions connected by a contact (an orifice or a narrow and short constriction)

1901: of  radius much smaller than the mean free path, $a \ll l$. 

1902: In a 3D geometry the resistance is

1903: \begin{equation}

1904: \label{eq:sharvin}

1905: R_{\rm Sharvin}=\frac{4 \rho l}{3 \pi a^2}=\left [ \frac{e^2}{h} 

1906: \frac{k^2 A}{2 \pi} \right ]^{-1}, 

1907: \end{equation}

1908: where 

1909: $h/e^2\approx 25.81$ k$\Omega$ is the quantum of resistance per spin, 

1910: $A$ is the contact

1911: area, and $k$ is the Fermi wave vector.

1912: The opposite limit,  of  diffusive transport through the contact with

1913: $a \gg l$, corresponds to the the Maxwell or Drude resistance

1914: $R_{\rm Maxwel}=\rho/2 a$. The studies of intermediate cases provide an

1915: interpolation scheme between the $R_{\rm Maxwell}$ and $R_{\rm Sharvin}$ 

1916: for  various ratios of $a/l$

1917: \cite{Wexler1966:PPSL,Jansen1980:JPCSSP,deJong1994:PRB,Nikolic1999:PRB}.

1918: Following \textcite{Kravchenko2003:PRB} the effective contact resistance 

1919: $r_c=r_{c\uparrow}+r_{c\downarrow}$ 

1920: (recall that it is defined per unit area)

1921: is obtained as

1922: \begin{equation}

1923: \label{eq:contact22}

1924: r_{c\lambda}=(4 R_{\rm Sharvin}/A) (1-t^L_\lambda-t^R_\lambda)/t^L_\lambda,

1925: \end{equation}

1926: where $t^{L,R}_\alpha$ represent the transmission coefficients 

1927: for the electrons 

1928: reaching the contact from the left and from the right and satisfy

1929: $t^L+t^R \le 1$.

1930: For $r_c$ which would exceed the resistance of the N and F bulk regions 

1931: the spin injection efficiency can attain 

1932: $P_j\sim (r_{c \uparrow}-r_{c \downarrow})/r_c$

1933: \cite{Kravchenko2003:PRB},

1934: showing, similarly to the diffusive regime, the importance of the 

1935: resistive contacts

1936: to efficient spin injection. Connection with the results in the 

1937: diffusive regime can be obtained \cite{Kravchenko2003:PRB}

1938: by identifying 

1939: $r_{c\lambda}=1/4 \Sigma_\lambda$, where the contact conductivity 

1940: $\Sigma_\lambda$ was introduced in  Eq.~\ref{eq:sigma}. 

1941: 

1942: While most of the experimental results on spin injection are feasible in the 

1943: diffusive regime, there are many theoretical studies treating the

1944: ballistic case and phase-coherent transport both in F/N and

1945: F/N/F junctions

1946: \cite{Mireles2001:PRB,Matsuyama2002:PRB,Hu2001:PRB,Hu2001:PRL}.

1947: Simple models in which the N region is a degenerate semiconductor often adopt an

1948: approach developed first for charge transport in  junctions involving 

1949: superconductors, discussed in Sec.~\ref{sec:IVA3}. Considering spin-orbit

1950: coupling and the potential scattering at the F/N interface modeled by the 

1951: $\delta$-function, \textcite{Hu2001:PRL} have examined ballistic

1952: spin injection in the F/N junction. They show that even a 

1953: spin-independent barrier 

1954: can be used to enhance the spin injection and lead to an increase in the

1955: conductance polarization. First-principles calculations were also used for 

1956: ballistic spin injection from a ferromagnetic metal into a 

1957: semiconductor \cite{Wunnicke2002:PRB,Mavropoulos2002:PRB,Zwierzycki2003:PRB}.

1958: In the limit of coherent (specular) scattering\footnote{The 

1959: wave-vector 

1960: component 

1961: along the interface is conserved during scattering.}

1962: and high interfacial quality it was shown that 

1963: different band structure in the F and the N regions would contribute to a 

1964: significant 

1965: contact resistance and  an efficient spin injection \cite{Zwierzycki2003:PRB}.

1966: 

1967: 

1968: \subsubsection{\label{sec:IIC3} Spin injection through the space-charge region}

1969: %----------------------------------------------------------------------------

1970: 

1971: 

1972: Interfaces making up a  semiconductor often

1973: develop a space-charge region---a region of local macroscopic charges.

1974: Typical examples are the Schottky contact and the depletion layer in 

1975: {\it p-n} junctions. 

1976: While phenomenological models, such as the one introduced 

1977: in Sec.~\ref{sec:IIC1}, capture a remarkable wealth of spin injection physics, 

1978: they carry little information about spin-dependent processes

1979: right at the interfaces. Microscopic studies of spin-polarized transport and

1980: spin-resolved tunneling through space-charge regions are still 

1981: limited in scope. The difficulty lies in the need

1982: to consider self-consistently simultaneous charge accumulation and

1983: electric-field generation (through Poisson's equation), both affecting 

1984: transport. 

1985: Non-self-consistent analyses of a Schottky barrier spin injection 

1986: were performed in \cite{Prins1995:JPCM,Albrecht2002:PRB,Albrecht2003:PRB}, 

1987: while

1988: \textcite{Osipov2003:P} proposed an efficient spin 

1989: injection method using a $\delta$-doped Schottky contact.

1990: 

1991: Let us now consider spin injection through the depletion

1992: layer in magnetic {\it p-n} junctions 

1993: \cite{Zutic2002:PRL,Fabian2002:PRB,Zutic2003:APL}.

1994: The physics is based on drift and diffusion\footnote{Tunneling or field 

1995: emission becomes important, for example, 

1996: in thin Schottky barriers

1997: or in {\it p-n} junctions and heterostructures at large reverse biases 

1998: \cite{Kohda2001:JJAP,Johnston-Halperin2002:PRB,vanDorpe2003:P}.} 

1999: limited

2000: by carrier recombination and spin relaxation, as described in more detail

2001: in Sec.~\ref{sec:IVA4}. The transport equations are solved

2002: self-consistently  with Poisson's equation, 

2003: taking full account of electric field due to accumulated charges.

2004: Additional examples of magnetic {\it p-n} junctions are discussed

2005: in Sec.~\ref{sec:IVD}. 

2006: 

2007: The system is depicted in Fig.~\ref{deplete}. The {\it p-n} junction 

2008: has a magnetic 

2009: $n$-region\footnote{Equilibrium magnetization can be 

2010: a consequence of doping with magnetic impurities, yielding large carrier 

2011: $g$ factors, and applying magnetic

2012: field, or of using a ferromagnetic semiconductor 

2013: \cite{Ohno1998:S,Pearton2003:JAP}.}

2014: with a net equilibrium electron spin $P_{n0}^R$, where $R$ stands

2015: for the right (here $n$) region. Holes are assumed to be unpolarized. 

2016: An important issue to be resolved is whether

2017: there will be spin accumulation in the 

2018: $p$-region 

2019: if a forward bias is applied to the junction.

2020: In other words, will spin be injected across the depletion layer? Naively 

2021: the answer is yes, since spin is carried by electrons, but the result shown 

2022: in Fig.~\ref{deplete} suggests a more complicated answer. 

2023: At small biases there is

2024: no spin injection. This is the normal limit of diode operation,

2025: in which

2026: where the

2027: injected carrier density through the depletion region is still smaller 

2028: than the equilibrium carrier density. Only with

2029: bias increasing to the high-injection limit (typically above 1 V)  

2030: is spin injected.

2031:  

2032: \begin{figure}

2033: \centerline{\psfig{file=zutic_fig08.eps,width=1\linewidth,angle=0}}

2034: \caption{Spin injection through the space-charge region of a magnetic 

2035: {\it p-n}

2036: junction. The geometry 

2037: is depicted in the inset, which shows  

2038: a junction

2039: with a spin-split conduction band in the 

2040: $n$-region

2041: with spin-polarized electrons (solid circles) and 

2042: unpolarized holes (empty circles).

2043: Under applied forward bias $V$ the charge current flows to the right. 

2044: The curves, labeled by $V$, show the electron density polarization profiles 

2045: $P_n(x)$ for the depicted geometry and GaAs materials parameters. 

2046: The equilibrium density polarization in the $n$-region 

2047: is about 0.5.

2048: At low bias (0.8 V) 

2049: there is no

2050: spin injection. Spin injection, manifested by the

2051: increase of $P_n$ in the 

2052: $p$-region, 

2053: appears only at large biases

2054: (1.2 and 1.5 V), 

2055: where it is driven by electric drift \cite{Zutic2002:PRL}.

2056: Spin polarization of the current is discussed by  

2057: \textcite{Zutic2001:APL,Fabian2002:PRB}.

2058: Adapted from \onlinecite{Zutic2002:PRL}. 

2059: }

2060: \label{deplete}

2061: \end{figure}

2062: 

2063: The explanation for the absence of spin injection at small biases 

2064: and for nondegenerate doping levels (Boltzmann statistics are applicable)

2065: is as follows.

2066: On the $n$ side, there are more spin up than spin down electrons,

2067: $n_\uparrow > n_\downarrow$.

2068: If $2q\zeta$ is the spin splitting of the

2069: conduction band, $n_\uparrow(\zeta)/n_\uparrow(\zeta=0)=\exp(q\zeta/k_BT)$.

2070: Under a forward bias, electrons 

2071: flow to the 

2072: $p$-region. 

2073: The flow is limited by thermal activation over

2074: the barrier (given by the built-in electrostatic potential  

2075: minus bias),

2076: which is, for the spin up electrons, greater by $q\zeta$. For Boltzmann

2077: statistics, the rate of transmission of spin up electrons

2078: over the barrier is 

2079: $\sim \exp(-q\zeta/k_BT)$.

2080: Since current is proportional to both the carrier density and the transmission

2081: rate, the two exponential factors cancel out. Similarly for spin down.

2082: As a result, the spin-resolved current is unaffected by $2q\zeta$ 

2083: and there is no spin current flowing through the depletion

2084: layer. There is no spin accumulation.

2085: Spin injection appears only at large biases, where it is driven

2086: by electric drift leading to nonequilibrium 

2087: spin population already in the $n$-region \cite{Fabian2002:PRB,Zutic2002:PRL}. 

2088: In addition to spin injection, spin extraction has also been

2089: predicted in magnetic {\it p-n} junctions with a magnetic

2090: $p$-region 

2091: \cite{Zutic2002:PRL}.  Under a large bias, spin is extracted 

2092: (depleted) from the nonmagnetic $n$-region. 

2093: 

2094: Electric field in the bulk regions next to the space charge is important only 

2095: at large biases. It affects not only spin density, but spin diffusion

2096: as well. That spin injection efficiency can increase in the presence

2097: of large electric fields due to an increase in the spin diffusion length 

2098: (spin drag) was first shown by \textcite{Aronov1976:SPS},

2099: and was later revisited by other authors.\footnote{See, for example,

2100: \cite{Zutic2001:PRB,Fabian2002:PRB,Margulis1994:PB,Martin2003:PRB,%

2101: Yu2002:PRBa,Flensberg2001:PRB,Vignale2003:SST,Bratkovsky2003:P}.} 

2102: To be important, 

2103: the electric field

2104: needs to be very large,\footnote{The critical magnitude is

2105: obtained by dividing a typical energy, such as the thermal or Fermi 

2106: energy, by $q$ and by the spin diffusion length. At room temperature the thermal 

2107: energy is 25 meV, while the spin diffusion length can be several

2108: microns.} more than 100 V/cm 

2109: at room temperature. While such large fields are usually present inside the

2110: space-charge regions, they exist in the adjacent bulk regions only at

2111: the high injection limit and  

2112: affect transport and spin injection.

2113: In addition to electric drift, magnetic drift,

2114: in magnetically inhomogeneous semiconductors, can also enhance spin injection

2115: \cite{Fabian2002:PRB}. 

2116: 

2117: The following formula was obtained for spin injection at small

2118: biases \cite{Fabian2002:PRB}:

2119: \begin{equation}

2120: P_{n}^L=\frac{P_{n0}^L[1-(P_{n0}^R)^2]+

2121: \delta P_{n}^R(1-P_{n0}^LP_{n0}^R)}

2122: {1-(P_{n0}^R)^2+\delta P_{n}^R(P_{n0}^L-P_{n0}^R)},

2123: \label{eq:aL}

2124: \end{equation}

2125: where $L$ (left) and $R$ (right) label the edges of the

2126: space-charge (depletion) region of a {\it p-n} junction. Correspondingly, 

2127: $\delta P_n^R$ represents  the nonequilibrium

2128: electron  polarization, evaluated at $R$, arising from a 

2129: spin source. The case

2130: discussed in Fig.~\ref{deplete} is for $P_{n0}^L=\delta P_n^R=0$. Then

2131: $P_n^L=0$, in accord with the result of no spin injection. For a homogeneous

2132: equilibrium magnetization ($P_{n0}^L=P_{n0}^R$), 

2133: $\delta P_{n}^L=\delta P_{n}^R$;

2134: the nonequilibrium spin polarization is the same across the depletion layer.

2135: Equation (\ref{eq:aL}) demonstrates that only {\it nonequilibrium} spin, 

2136: already

2137: present in the bulk region, can be transferred through the depletion

2138: layer at small biases \cite{Zutic2001:PRB,Fabian2002:PRB}.  

2139: Spin injection of nonequilibrium spin is also very effective if 

2140: it proceeds from the 

2141: $p$-region 

2142: \cite{Zutic2001:PRB},

2143: which is the case for a spin-polarized solar cell \cite{Zutic2001:APL}. 

2144: The resulting spin 

2145: accumulation in the $n$-region 

2146: extends the spin diffusion range,

2147: leading to spin amplification---increase of the spin population away from 

2148: the spin source. 

2149: These results were also confirmed in the junctions

2150: with two differently doped $n$-regions 

2151: \cite{Pershin2003:PRL,Pershin2003:Pb}.

2152: Note, however, that  

2153: the term ``spin polarization density'' used in

2154: \textcite{Pershin2003:PRL,Pershin2003:Pb} 

2155: is actually the spin density $s=n_\uparrow-n_\downarrow$,

2156: not the spin polarization $P_n$. 

2157: 

2158: Theoretical understanding of spin injection has focused largely on 

2159: spin density while neglecting spin phase, which is 

2160: important for some 

2161: proposed spintronic applications. The problem of

2162: spin evolution in various transport modes (diffusion, tunneling,

2163: thermionic emission) remains to be investigated. Particularly relevant

2164: is the question of whether spin phase is conserved during spin injection.

2165: \textcite{Malajovich2001:N} showed, by studying spin evolution 

2166: in transport through a 

2167: n-GaAs/n-ZnSe

2168: heterostructure, 

2169: that the phase can indeed be preserved.

2170: 

2171: 

2172: \subsection{\label{sec:IID} Experiments on spin injection}

2173: %===========================================================================

2174: 

2175: \subsubsection{\label{sec:IID1} Johnson-Silsbee spin injection}

2176: %---------------------------------------------------------------------------

2177: 

2178: 

2179: The first spin polarization of electrons by electrical spin injection 

2180: \cite{Johnson1985:PRL} was demonstrated in a ``bulk wire'' of 

2181: aluminum on which 

2182: an array of thin film permalloy (Py) pads (with 70 \% nickel and 30 \% iron)  

2183: was deposited spaced in multiples of 50 $\mu$m, center to center 

2184: \cite{Johnson1988:PRBb}

2185: to serve as spin injectors and detectors. In one detection scheme

2186: a single ferromagnetic pad was used as a spin injector while the distance

2187: to the spin detector was altered by selecting different Py pads to

2188: detect $V_s$ and through the spatial decay of this spin-coupled voltage 

2189: infer $L_{sN}$.\footnote{The spin relaxation time in a ferromagnet is 

2190: often assumed to be

2191: very short.  Correspondingly, in the analysis of the experimental data, 

2192: both the spin 

2193: diffusion length and $\delta M$ are taken to vanish in the F region  

2194: \cite{Silsbee1980:BMR,Johnson1985:PRL,Johnson1988:PRBa,Johnson1988:PRBb}.}

2195: This procedure is illustrated in 

2196: Fig.~\ref{injexp:1}, where 

2197: the separation between the spin injector and detector $L_x$ is variable. 

2198: 

2199: \begin{figure}

2200: \centerline{\psfig{file=zutic_fig09.eps,width=1.\linewidth,angle=0}}

2201: \caption{Schematic top view of nonlocal, quasi-one-dimensional 

2202: geometry used by \textcite{Johnson1985:PRL}: F1 and F2, 

2203: the two metallic ferromagnets having magnetizations in the $x-z$ plane;  

2204: dotted lines, equipotentials characterizing electrical current flow;

2205: grey shading, diffusing population of nonequilibrium spin-polarized

2206: electrons injected at $x=0$, with darker shades corresponding to higher density 

2207: of polarized electrons. From \onlinecite{Johnson2002:SST}.} 

2208: \label{injexp:1}

2209: \end{figure}

2210: \textcite{Johnson1985:PRL} 

2211: point out that in the depicted geometry  

2212: there

2213: is no flow of the charge current for $x >0$ and that 

2214: in the absence of 

2215: nonequilibrium spins a voltage measurement between $x=L_x$ and $x=b$ gives zero.

2216: Injected spin-polarized electrons will diffuse symmetrically (at low current

2217: density the effect of electric fields can be neglected), and the 

2218: measurement of voltage will give a spin-coupled signal $V_s$ related to the 

2219: relative orientation of magnetizations in F1 and F2.\footnote{This method for

2220: detecting the effects of spin injection is also referred to as a

2221: potentiometric method.}

2222: The results, corresponding to the polarizer-analyzer detection and the

2223: geometry of Fig.~\ref{injexp:1}, 

2224: are given in Fig.~\ref{injexp:2}. An in-plane field 

2225: (${\bf B}$ $\|$ ${\bf \hat{z}}$), 

2226: of a magnitude several times larger

2227: than a typical field for magnetization reversal, $B_0 \approx 100$ G,

2228: is applied to define the direction of magnetization in the injector 

2229: and detector.  

2230: As the field sweep is performed, from negative to positive values, 

2231: at $B_{01}$ there is 

2232: a reversal of magnetization in one of the ferromagnetic films accompanied 

2233: by a sign change in the spin-coupled signal. As $B_z$ is further

2234: increased, at approximately $B_{02}$, there is 

2235: another reversal of magnetization,

2236: resulting in parallel orientation of F1 and F2 and a $V_s$ 

2237: of magnitude similar to that for the previous parallel orientation 

2238: when $B_z < B_{01}$. 

2239: 

2240: 

2241: \begin{figure}

2242: \centerline{\psfig{file=zutic_fig10.eps,width=1.\linewidth,angle=0}}

2243: \caption{Spin injection data from bulk Al wire sample.

2244: Negative  magnetic field is applied parallel to the magnetization (-$z$ axis) 

2245: in the two ferromagnetic regions. As the field is increased, at $B_{0,1}$ 

2246: magnetization in one of the ferromagnetic regions is reversed, and at $B_{0,2}$

2247: the magnetization in the other region is also reversed 

2248: (both are along +$z$ axis).

2249: Inset: amplitude of the observed Hanle signal as a function of orientation 

2250: angle $\phi$ of magnetic field. From \onlinecite{Johnson1985:PRL}.} 

2251: \label{injexp:2}

2252: \end{figure}

2253: 

2254: A more effective detection of the spin injection 

2255: is realized through measurements of the Hanle effect, 

2256: also discussed Secs.~\ref{sec:IIB} and \ref{sec:IIIA2}, and described by 

2257: the Bloch-Torrey equations \cite{Bloch1946:PR,Torrey1956:PR}

2258: [see Eqs.(~\ref{eq:relax:bloch1})--(\ref{eq:relax:bloch3})].

2259: The inset of Fig.~\ref{injexp:2} summarizes results from a

2260: series of Hanle experiments on a single sample. For the

2261: Hanle effect ${\bf B}$ must have a component perpendicular to the orientation 

2262: axes

2263: of the injected spins. Only projection of ${\bf B}$ perpendicular to 

2264: the spin axis

2265: applies a torque and dephases spins. The magnitude of ${\bf B}$, applied at

2266: an angle $\phi$ to the $z$-axis in the $y-z$ plane, is small enough that the

2267: magnetizations in ferromagnetic thin films remain in the $x-z$ plane (see 

2268: Fig.~\ref{injexp:1}). If, at ${\bf B}=0$, injected nonequilibrium magnetization

2269: is $\delta M(0) \hat{z}$ 

2270: then at finite field $\delta {\bf M}$ precesses about

2271: ${\bf B}$ with the cone of angle $2\phi$. After averaging over several cycles,

2272: only $\delta M(0) \cos \phi$, 

2273: the component $\|$ ${\bf B}$, will survive.

2274: The voltage detector\footnote{Recall from the discussion leading to 

2275: Eq.~(\ref{eq:drop}) that the spin-coupled signal $\propto \delta M$.}

2276: senses the remaining part of the magnetization projected on the axis

2277: of the detector 

2278: $\delta M(0) \cos \phi \times \cos\phi$ \cite{Johnson1988:PRBa}.

2279: The predicted angular dependence for the  amplitude of the Hanle signal 

2280: (proportional to the  depolarization of $\delta M$ in a finite field) 

2281: $[\delta M(0) -\delta M(0) \cos^2\phi]$ 

2282: is plotted in the inset together

2283: with the measured data.\footnote{The range of the angle $\phi$, 

2284: in the inset, is corrected from the one originally given in Fig.~3 of 

2285: \textcite{Johnson1985:PRL}.} Results confirm the first application of

2286: the Hanle effect to dc spin injection. 

2287: 

2288: The Hanle effect was also studied theoretically by solving

2289: the Bloch-Torrey equations for an arbitrary orientation, characterized by the

2290: angle $\alpha$, between the magnetization

2291: in F1 and F2 \cite{Johnson1988:PRBa}. From the Hanle curve [$V_s(B_\bot)$]

2292: measured at $T=4.3$ ($36.6$) K,  

2293: the parameters $L_s=450$ ($180$) $\mu$m and 

2294: $P_\Sigma=0.06$ ($0.08$) were extracted.\footnote{The fitting parameters are 

2295: $\tau_s$, $P_\Sigma$, and $\alpha$

2296: \cite{Johnson1988:PRBb}, and since the diffusion coefficient is obtained from 

2297: Einstein's relation $L_s$ is known.}

2298: This spin injection technique using a few pV resolution of

2299: a  superconducting quantum interference device (SQUID) 

2300: and with an 

2301: estimated $P_\Sigma\approx 0.07$ provided an accuracy able to detect 

2302: $P_n \approx 5\times 10^{-12}$, causing 

2303: speculating on that

2304: a single-spin sensitivity might be possible in  smaller samples 

2305: \cite{Johnson1985:PRL,Johnson1988:PRBb}.

2306: While in a good conductor, such as Al, the observed 

2307: resistance  change 

2308: $\Delta R$ was small ($\sim n\Omega$), 

2309: the relative change at low temperatures and for 

2310: $L_x \ll L_s$ was $\Delta R/R \approx 5 \%$, where $\Delta R$  

2311: is defined as in Eq.~(\ref{eq:tmr}), determined by the relative orientation of 

2312: the magnetization in F1 and F2, and $R$ is the

2313: Ohmic resistance \cite{Johnson2002:SST}.

2314: Analysis from Sec.~\ref{sec:IIC2} shows that the measurement 

2315: of $\Delta R$ could 

2316: be used to determine the product of injected current polarizations 

2317: in the two F/N junctions.

2318: 

2319: The studies of spin injection were extended to the thin-film geometry, 

2320: also known as the ``bipolar spin switch'' or ``Johnson spin transistor''

2321: \cite{Johnson1993:S,Johnson1993:PRL} 

2322: similar to the one depicted in Fig.~\ref{inj:1}(a). 

2323: The measured spin-coupled signals

2324: \footnote{$d \sim 100$ nm 

2325: was much smaller then the separation between F1 and F2 in bulk Al wires 

2326: \cite{Johnson1985:PRL}, and the amplitude of the Hanle effect

2327: was about $10^4$ larger \cite{Johnson2002:SST}.}

2328: in Au films were larger than the values obtained in bulk Al 

2329: wires \cite{Johnson1985:PRL,Johnson1988:PRBb}.

2330: A similar trend, 

2331: $V_s\sim 1/d$, potentially important for applications,

2332: was already anticipated by \textcite{Silsbee1980:BMR}.

2333: The saturation of this increase can be 

2334: inferred from Eqs.~(\ref{eq:DelR}) and (\ref{eq:det}) for 

2335: $d \ll L_{sN}$ 

2336: and has been discussed by \textcite{Hershfield1997:PRB} and

2337: \textcite{Fert1996:PRB}. 

2338: 

2339: When polarizer-analyzer detection was used, one of the fitting parameters

2340: from the measured data $P_\Sigma$ sometimes exceeded 1---which 

2341: corresponds to complete

2342: interfacial polarization. The origin of this discrepancy remains to 

2343: be fully resolved

2344: \cite{Hershfield1997:PRB,Fert1996:PRB,Johnson1993:PRL,Johnson2002:SST,Geux2000:APPA}.

2345: Results obtained from the Hanle effect, on similar samples, gave the 

2346: expected  $P_\Sigma <1$ values \cite{Johnson2002:SST}.\footnote{Theoretical 

2347: estimates for $V_s$  from which $P_\Sigma > 1$ was inferred

2348: are modified when one considers the Coulomb interaction and 

2349: proximity effects---near the N/F interface the spin splitting 

2350: of the carrier bands in the N region will be finite even at equilibrium.

2351: Model calculations \cite{Chui1995:PRL,Chui1995:PRB}, 

2352: which treat the F/N/F junction as a whole, show that the magnetic 

2353: susceptibility 

2354: $\chi$ in N can be much smaller than the free electron value and can increase 

2355: the predicted 

2356: $V_s \propto 1/\chi$. These corrections to the free-electron picture 

2357: of an F/N/F junction

2358: are smaller for larger $d$, as in the bulk-wire geometry of 

2359: \textcite{Johnson1985:PRL},

2360: where theoretical estimates of $V_s$ did not lead to $P_\Sigma > 1$.}  

2361: 

2362: A modification of the bipolar spin switch structure was used to demonstrate the

2363: spin injection into a niobium film \cite{Johnson1994:APL},

2364: realizing the theoretical assertion of \textcite{Aronov1976:SPJETP}

2365: that nonequilibrium spin could be injected into a superconductor.

2366: Two insulating Al$_2$O$_3$ films were inserted between F1 and F2 

2367: (both made of Py) 

2368: and a Nb film [see Fig.~\ref{inj:1}(b)]. The measurements were performed 

2369: near the superconducting transition temperature $T_c$

2370: with the data qualitatively similar, above and below $T_c$, to the spin-coupled

2371: voltage, as obtained in the magnetic-field sweep from Fig.~\ref{injexp:2}.

2372: The results were interpreted as support for enhanced  depletion of the 

2373: superconducting condensate (and correspondingly the reduction of 

2374: the critical current $I_c$) by spin-polarized  quasiparticles, as compared 

2375: to the usual spin-unpolarized quasiparticle injection. Related measurements 

2376: were recently performed in a CPP geometry \cite{Gu2002:PRB}, and the

2377: penetration depth of the 

2378: quasiparticle in the Nb films was measured to be $\sim 16$ nm, 

2379: as compared to $2$ nm in 

2380: \cite {Johnson1994:APL}. The corresponding temperature dependence of CPP GMR is

2381: well explained by the theory of \textcite{Yamashita2003:PRB} and the 

2382: modification

2383: of Andreev reflection (see Sec.~\ref{sec:IVA3}) by spin polarization.

2384: 

2385: The spin injection technique of Johnson and Silsbee was also applied to 

2386: semiconductors.

2387: Initial experiments on using a metallic ferromagnet to inject spin into a 

2388: two-dimensional

2389: electron gas (2DEG) showed only a very low ($\sim 1 \%$) efficiency

2390: \cite{Hammar1999:PRL}  

2391: for which various

2392: explanations were offered \cite{Monzon2000:PRL,vanWees2000:PRL,Hammar2000:PRL}. 

2393: However, stimulated by the proposal of \textcite{Rashba2000:PRB}

2394: to employ spin-selective diffusive contacts (Sec.~\ref{sec:IIC1}), 

2395: the subsequent 

2396: measurements have showed substantially more efficient spin injection 

2397: into a 2DEG

2398: after an insulating layer was inserted \cite{Hammar2001:APL,Hammar2002:PRL}. 

2399: The geometry

2400: employed is depicted in Fig.~\ref{injexp:1}. 

2401: In interpreting 

2402: the results, the spin-orbit coupling and the energy-independent density

2403: of states 

2404: at the Fermi level were taken into account \cite{Silsbee2001:PRB}. 

2405: This topic is reviewed by \textcite{Tang:2002}. 

2406: 

2407: \begin{figure}

2408: \centerline{\psfig{file=zutic_fig11.eps,width=0.5\linewidth,angle=-90}}

2409: \caption{Schematic representation of (a) local and (b) nonlocal geometry

2410: used to measure the effects of spin injection and spin accumulation.}

2411: \label{exp:3}

2412: \end{figure} 

2413: 

2414: \subsubsection{\label{sec:IID2} Spin injection into metals}

2415: %---------------------------------------------------------------------------

2416: 

2417: An important part of the operation of  CPP GMR structures is the presence 

2418: of nonequilibrium spin polarization in nonmagnetic metallic regions.

2419: Studies of spin-injection parameters

2420: in such systems have been reviewed by \textcite{Bass1999:JMMM} and

2421: \textcite{Gijs1997:AP}.   

2422: However, until recently, except for the work of Johnson and Silsbee,

2423: there were few  other experimental studies directly concerned with 

2424: spin injection into metals. A series of 

2425: experiments, \cite{Jedema2001:N,Jedema2002:JS,Jedema2002:Na,Jedema2002:APL} 

2426: at both low (4.2 K) and room temperature,

2427: were performed using  

2428: the van der Pauw geometry depicted in Fig.~\ref{exp:3}.

2429: In various structures \cite{Jedema2002:T}

2430: the two ferromagnetic regions (made of Py, Co, or Ni) 

2431: were chosen to be of different sizes to provide different  

2432: coercive fields, allowing an independent reversal of magnetization 

2433: in F1 and F2.

2434: The cross-shaped nonmagnetic region was made of Al or Cu \cite{Jedema2002:T}. 

2435: Nonlocal measurements, similar to the approach shown in Figs.~\ref{inj:1} and

2436: \ref{injexp:1} [discussed in \cite{Johnson1988:PRBb,Johnson1993:PRL}], were

2437: shown to simplify the extraction of spurious effects (for example,

2438: anisotropic magnetoresistance  and the Hall signal)

2439: from effects intrinsic to spin injection, as compared

2440: to the local or conventional  spin-valve geometry. 

2441: 

2442: In the first type of experiment the cross-shaped region was deposited directly

2443: over the F region 

2444: (Fig.~\ref{exp:3}), 

2445: and the spin-coupled resistance $\Delta R$, defined

2446: analogously to Eq.~(\ref{eq:tmr}),

2447: was measured as a function of an in-plane magnetic filed. A theoretical 

2448: analysis

2449: \cite{Jedema2001:N, Jedema2002:JS} was performed assuming no interfacial

2450: resistance ($r_c=0)$ and the continuity of the electrochemical 

2451: potentials at the F/N interface (see Sec.~\ref{sec:IIC1}).

2452: For a spin injection from Py into Cu, the maximum current polarization 

2453: was obtained to be $P_j \approx 0.02$ at 4.2 K. The results for $\Delta R$

2454: \cite{Jedema2001:N} scaled to the size of the samples

2455: used by \textcite{Johnson1993:S,Johnson1993:PRL} were interpreted to be 

2456: 3-4 orders of magnitude smaller. As discussed in Secs.~\ref{sec:IIC1} and 

2457: \ref{sec:IIC2}, the presence of interfacial

2458: spin-selective resistance can substantially change the spin injection 

2459: efficiency 

2460: and influence the resistance mismatch between the F and N regions

2461: [see Eq.~(\ref{eq:Pj})]. 

2462: Estimates of how these considerations would  affect the results of 

2463: \textcite{Jedema2001:N} 

2464: were given by \textcite{Jedema2002:Nb}

2465: as well as by others 

2466: \cite{Takahashi2003:PRB,Johnson2003:PRB}, 

2467: who analyzed 

2468: the importance 

2469: of multidimensional geometry. 

2470: In addition to

2471: comparing characteristic

2472: values of contact resistance

2473: obtained on different samples,\footnote{For example, the measured

2474: resistance of clean F/N contacts in CPP GMR \cite{Bussmann1998:IEEETM} was 

2475: used to infer that there is also a large contact resistance in 

2476: all-metal spin injection experiments \cite{Johnson2002:N}.}

2477: for a conclusive

2478: understanding it will be crucial to have 

2479: {\it in situ} measurements.

2480: 

2481: In analyzing data for the van der Pauw cross, a two-dimensional geometry 

2482: has an important effect---while the electric current is following  

2483: the paths depicted in Fig.~\ref{exp:3}, the spin current, through 

2484: the diffusion of nonequilibrium spin,

2485: would have similar flow in all four arms \cite{Johnson2002:SST}. This

2486: is different from the usual (quasi)one-dimensional analysis in which 

2487: spin and charge currents flow along the same paths. 

2488: For a full

2489: understanding of the van der Pauw cross geometry,  two-dimensional 

2490: modeling might be necessary \cite{Johnson2003:PRB,Takahashi2003:PRB}.

2491: 

2492: In the second type of experiment, tunneling contacts were fabricated 

2493: by inserting Al$_2$O$_3$ as an insulator into the regions where F1 and F2 

2494: overlapped with the cross. By applying a transverse field B$_z$ 

2495: (see Fig.~\ref{exp:3}) the precession of the injected nonequilibrium spin 

2496: was controlled and the amplitude of the Hanle effect was measured

2497: \cite{Jedema2002:Na,Jedema2002:APL},  as  outlined in Sec.~{\ref{sec:IID1}.

2498: From Co/Al$_2$O$_3$/Al/Al$_2$O$_3$/Co structures

2499: $L_s \approx 0.5$ $\mu$m was extracted at room temperature.

2500: The analysis of the Hanle signal was performed  by averaging contributions

2501: of different lifetimes \cite{Dyakonov:1984b}. This proved to be

2502: equivalent to the 

2503: \textcite{Johnson1988:PRBb}

2504: solution to the Bloch-Torrey equations.

2505: 

2506: 

2507: \subsubsection{\label{sec:IID3}  All-semiconductor spin injection}

2508: \label{magneticSM}

2509: 

2510: 

2511: \begin{figure}

2512: \centerline{\psfig{file=zutic_fig12.eps,width=0.8\linewidth,angle=0}}

2513: \caption{Schematic device geometry and band diagram of a spin LED:

2514: (a) Recombination of

2515:  spin-polarized electrons injected from

2516: the (II,Mn)VI spin aligner and  unpolarized holes injected from the 

2517: p-doped GaAs, in the intrinsic GaAs quantum well, 

2518: producing circularly polarized  light; (b) conduction and valence bands of 

2519: a spin aligner in an external magnetic field; 

2520: (c) sketch of the corresponding

2521: band edges and band offsets in the device geometry.

2522: In the quantum well, spin down electrons and unpolarized holes

2523: are depicted by solid and empty circles, respectively. 

2524: Adapted from \onlinecite{Fiederling1999:N}.}

2525: \label{injexp:3}

2526: \end{figure} 

2527: 

2528: If a magnetic semiconductor could be used as a robust spin injector 

2529: (spin aligner) 

2530: into a  nonmagnetic semiconductor it would facilitate the integration of 

2531: spintronics 

2532: and semiconductor-based electronics. Comparable resistivities of magnetic 

2533: and nonmagnetic semiconductors could provide efficient spin injection 

2534: [see Eq.~(\ref{eq:gamma}), with $r_F\approx r_N$] even without using  

2535: resistive contacts. Ultimately, for a wide range of applications and 

2536: for compatibility with

2537: complementary metal-oxide semiconductors (CMOS) 

2538: \cite{Wong1999:PIEEE}, it would be 

2539: desirable to be able to inject spin into silicon at room temperature.

2540: 

2541: Early studies \cite{Viglin1991:FTT,Osipov1990:PZETF,Viglin1997:PLDS,Osipov1998:PL}, 

2542: which have since largely been ignored, used a Cr- and Eu-based 

2543: chalcogenide ferromagnetic semiconductor (FSm) \cite{Nagaev:1983} as the spin 

2544: injector.\footnote{These materials, while more difficult to fabricate 

2545: than the subsequent 

2546: class of III-V ferromagnetic semiconductors,  have desirable properties of 

2547: providing injection of spin-polarized electrons 

2548: (with spin lifetimes typically

2549: much longer than for holes) and large spin splitting 

2550: [$\sim 0.5$ eV at 4.2 K for  n-doped HgCr$_2$Se$_4$ \cite{Nagaev:1983}]

2551: with nearly complete spin polarization 

2552: and a Curie temperature $T_C$ of up to 130 K 

2553: (HgCr$_2$Se$_4$) \cite{Osipov1998:PL}.}

2554: The experiments were motivated

2555: by the theoretical work of \textcite{Aronov1976:SPS,Aronov1976:SPJETP} 

2556: predicting that the ESR signal, proportional to the steady-state

2557: magnetization, would be changed by spin injection. The measurements of 

2558: \textcite{Viglin1991:FTT,Osipov1990:PZETF} prompted a 

2559: related prediction \cite{Margulis1994:PB} that spin injection could be 

2560: detected through changes in electric dipole spin resonance (EDSR).

2561: EDSR is the spin-flip resonance absorption for conduction electrons

2562: at Zeeman frequency, which is excited by the electric-field 

2563: vector of

2564: an incident electromagnetic wave. The theory of EDSR, developed 

2565: by \textcite{Rashba1961:SPSS} is extensively reviewed by

2566: \textcite{Rashba:1991}.    

2567: 

2568: Ferromagnetic semiconductor 

2569: spin injectors 

2570: formed {\it p-n} and {\it n-n} heterostructures with a 

2571: nonmagnetic semiconductor

2572: InSb. The choice of InSb is  very suitable due to its large 

2573: negative ($\sim -50$) $g$ factor \cite{McCombe1971:PRB}, 

2574: for detecting the effects of spin injection through ESR. The observed absorption

2575: and emission of microwave power \cite{Osipov1998:PL} was tuned by an 

2576: applied magnetic field (from 35 GHz at $\approx$ 400 G up to 1.4 THz at 20 kG)

2577: and only seen when electrons flowed  from FSm into an Sm region. 

2578: The injection to the lower Zeeman level increased the ESR absorption,

2579: while injection to the higher Zeeman level, leading to  population 

2580: inversion,

2581: generated  microwave emission. 

2582: 

2583: 

2584: The most recent experiments using semiconductor spin injectors can be grouped 

2585: into two different classes, In one approach (II,Mn)VI paramagnetic 

2586: semiconductors 

2587: were employed as the spin aligners. 

2588: These included

2589: CdMnTe \cite{Oestreich1999:APL}, BeMnZnSe \cite{Fiederling1999:N}, and

2590: ZnMnSe \cite{Jonker2000:PRB}.  

2591: In the second approach 

2592: ferromagnetic semiconductors like 

2593: (Ga,Mn)As \cite{Ohno1999:N,Chun2002:PRB,Mattana2002:PRL} were used.

2594: Both approaches were also employed to inject spins into 

2595: CdSe/ZnSe \cite{Seufert2004:PRB} and InAs

2596: \cite{Chye2002:PRB} quantum dots, respectively.

2597: 

2598: In (II,Mn)VI materials, at low Mn-concentration and at low temperatures,

2599: there is a  giant Zeeman splitting $\Delta E = g^* \mu_B H$ 

2600: \cite{Furdyna1988:JAP,Gaj:1988}

2601: of the conduction  band,

2602: in which $g^*$ is the effective electron  $g$ factor.

2603: Such splitting arises due to 

2604: {\it sp-d} exchange between the spins of

2605: conduction electrons and the S=5/2 spins of the localized Mn$^{2+}$ ions.

2606: The $g^*$ factor for $H \ne 0$

2607: can exceed\footnote{At low temperatures ($\sim1$ K)

2608: Cd$_{0.95}$Mn$_{0.05}$Se has $|g^*|>500$

2609: \cite{Dietl:1994}, while in n-doped (In,Mn)As $|g^*|>100$

2610: at 30 K \cite{Zudov2002:PRB}. Such large $g$ factors, in the

2611: presence of a highly inhomogeneous magnetic filed could lead to the

2612: charge carrier localization \cite{Berciu2003:PRL}.} 100

2613: and is given by \cite{Furdyna1988:JAP,Brandt1984:AP}

2614: \begin{equation}

2615: g^*=g+\alpha M/(g_{Mn} \mu_B^2 H),

2616: \label{eq:geff}

2617: \end{equation}

2618: where $g$ is the $H=0$ II-VI ``band'' value $g$, generally different from the

2619: free-electron value, magnetization

2620:  M$\propto$ $\langle S_z \rangle\propto B_s[(g_{Mn} \mu_B S H)/(k_B T)]$, B$_s$ 

2621: is the Brillouin function 

2622: \cite{Ashcroft:1976}, and $\alpha$ 

2623: is the exchange integral for $s$-like $\Gamma_6$ electrons

2624: (see Table I in Sec.~\ref{sec:IIB}), given by 

2625: \cite{Furdyna1988:JAP} 

2626: \begin{equation}

2627: \alpha\equiv\langle S|J_{sp-d}|S \rangle/V_0,

2628: \label{eq:alpha}

2629: \end{equation}

2630: where $J_{sp-d}$ is the electron-ion exchange coupling,

2631: and V$_0$ is the volume of an elementary cell.

2632: From Eqs.~(\ref{eq:geff}) and (\ref{eq:alpha}) it follows that 

2633: $g^*=g^*(H)$

2634: can even change its sign. 

2635: Similar analysis applies also to $g$ factors of holes, with the 

2636: Zeeman splitting of a valence band being typically several times larger

2637: than that of a conduction band \cite{Brandt1984:AP}.

2638: 

2639: (II,Mn)VI materials can be incorporated in high quality heterostructures 

2640: with different optically active III-V nonmagnetic semiconductors which,

2641: by providing 

2642: circularly polarized luminescence, can also serve

2643: as spin detectors. In this case 

2644: carriers

2645: are excited by electrical means and we speak of 

2646: electroluminescence  rather then photoluminescence.

2647: The selection rules for the recombination light are

2648: the same as discussed in Sec.~\ref{sec:IIB}.

2649: 

2650: Figure \ref{injexp:3} depicts a scheme for

2651: realization of all-semiconductor electrical spin injection and  

2652: optical detection \cite{Fiederling1999:N,Jonker2000:PRB}. 

2653: Displayed is a spin light-emitting diode (LED) \cite{Jonker1999:PA}

2654: in a Faraday geometry where both the applied B-field and the 

2655: direction of propagation of the

2656: emitted light 

2657: lie along the growth direction.

2658: Similar to a an ordinary

2659: LED \cite{Sze:1981}, electrons and holes recombine

2660: (in a quantum well or a {\it p-n} junction) and produce electroluminescence. 

2661: However,

2662: in a spin LED,  as a consequence of radiative recombination of spin-polarized 

2663: carriers, the emitted light is circularly polarized. 

2664: In experiments of  

2665: \textcite{Fiederling1999:N,Jonker2000:PRB}, at B$\approx 1$ T, 

2666: $T\approx 4$ K, 

2667: and forward bias, electrons entering from the $n$-contact were almost 

2668: completely polarized 

2669: in the spin down state as they left the spin aligner and are injected 

2670: across the (II,Mn)VI/AlGaAs interface. The electrons further traveled

2671: (by drift and diffusion) to an intrinsic GaAs quantum well (QW) 

2672: where they recombined with the unpolarized holes, which were injected from the 

2673: p-doped GaAs.\footnote{The spatial separation and  spin relaxation

2674: between the spin injection and the point of spin detection (in QW)

2675: make a fully quantitative analysis of the injected polarization more difficult. 

2676: It would be valuable to perform realistic calculations of a spin-polarized

2677: transport and spin injection which would treat the whole spin LED as a single

2678: entity.} 

2679: 

2680: The efficiency of  electrical spin injection across 

2681: the (II,Mn)VI/AlGaAs interface was studied \cite{Fiederling1999:N}

2682: using $P_{\rm circ}$ (defined in Sec.~\ref{sec:IIB}) of electroluminescence, 

2683: as a function of 

2684: B and  the thickness of the magnetic spin aligner 

2685: (0 nm, 3 nm, and 300 nm, respectively).

2686: $P_{\rm circ}$ increased with the

2687: thickness of the magnetic layer, suggesting the finite spin relaxation time

2688: needed for initially unpolarized electrons to relax into the lower 

2689: (spin down) Zeeman level. 

2690: The results of \textcite{Jonker2000:PRB} were similar to those of 

2691: \textcite{Fiederling1999:N} 

2692: for the thickest magnetic region. The behavior of $P_{\rm circ}$(B), up to the 

2693: saturation value (B$\approx 3$ T), 

2694: could be well explained by the magnetization

2695: described with the Brillouin function \cite{Furdyna1988:JAP,Gaj:1988}, 

2696: expected

2697: for the (II,Mn)VI semiconductors. In Fig.~\ref{injexp:3} the injected spin down 

2698: electrons are majority electrons with their magnetic

2699: moments parallel to the applied magnetic field.

2700: The principles of optical orientation discussed 

2701: in Sec.~\ref{sec:IIB} and the 

2702: selection 

2703: rules for GaAs sketched in Fig.~\ref{oo:1} are used to infer $P_n$ in a QW.

2704: 

2705: For a QW of approximately the same width (150 nm) the conversion of 

2706: $P_{\rm circ}$ to $P_n$  used by

2707: \textcite{Fiederling1999:N}  differed by a factor of 2 

2708: from that used by \textcite{Jonker2000:PRB}.

2709: \textcite{Fiederling1999:N}

2710: assumed that confinement effects were negligible, leading to the selection

2711: rules for a bulk GaAs 

2712: (recall $P_{\rm circ}=-P_n/2$, from Sec.~\ref{sec:IIB}).

2713: The maximum $P_{\rm circ}\approx 43 \%$ was interpreted 

2714: as implying nearly $90$\% 

2715: polarized injected electrons. \textcite{Jonker2000:PRB} 

2716: inferred $|P_n| \approx 50$\%, 

2717: from  $P_{\rm circ}=-P_n$ \cite{Weisbuch:1991}, as a consequence of QW 

2718: confinement and 

2719: lifting of the degeneracy 

2720: between light and heavy hole states 

2721: in the valence band ($\approx 5-6$ meV), see Fig.~\ref{oo:1}.

2722: Both results clearly demonstrated a 

2723: robust low-temperature

2724: spin injection using the spin LED's. 

2725: Subsequent studies 

2726: \cite{Park2000:APL,Jonker2001:APL,Stroud2002:PRL}

2727: have supported the lifting of degeneracy 

2728: between the light and heavy hole bands. The corresponding data are 

2729: shown in Fig.~\ref{injexp:4}.

2730: Similar values

2731: of $P_{\rm circ}$ were also measured in a resonant tunneling diode based

2732: on ZnMnSe \cite{Waag2001:JS,Gruber2001:APL}. Spin injection using the

2733: spin LED's,

2734: described above, is not limited to structures grown by molecular-beam epitaxy

2735: (MBE). It is also feasible using air-exposed interfaces \cite{Park2000:APL}

2736: similar to the actual fabrication conditions employed in conventional

2737: electronics. 

2738: 

2739: \begin{figure} 

2740: \centerline{\psfig{file=zutic_fig13.eps,width=1.0\linewidth}}

2741: \caption{(a)  Electroluminescence in a spin light-emitting diode (LED):

2742: (a) Electroluminescence (EL) spectra from a surface-emitting spin LED 

2743: with a Zn$_{0.94}$Mn$_{0.06}$Se contact for selected values of applied 

2744: magnetic field, 

2745: analyzed for  $\sigma^{\pm}$ (positive and negative helicity);

2746: the magnetic field is applied along the 

2747: surface normal (Faraday geometry) and the spectra are dominated by the heavy 

2748: hole exciton;

2749: (b)  magnetic field dependence of the EL circular polarization. 

2750: Adapted from \onlinecite{Jonker2001:APL}.}         

2751: \label{injexp:4}

2752: \end{figure} 

2753: 

2754: The robustness of measured $P_{\rm circ}$ was studied by intentionally

2755: changing the density of linear defects, from stacking faults at

2756: the ZnMnSe/AlGaAs  

2757: interface \cite{Stroud2002:PRL}. An approximate linear 

2758: decrease of $P_{\rm circ}$ with the density of stacking faults

2759: was shown to be

2760: consistent with the influence of spin-orbit interaction as

2761: modeled by Elliot-Yafet scattering (see Sec.~\ref{sec:IIIB1}) at the interface.

2762: The nonspherically symmetric defect potential (entering the spin-orbit 

2763: interaction) causes  a highly anisotropic loss of 

2764: spin polarization.  At small angles to the axis of growth

2765: [see Fig.~(\ref{injexp:3})], the probability of the spin flip

2766: of an injected electron is very high,  leading to a small spin polarization.

2767: These findings illustrate the importance of

2768: interface quality and the effect of defects on the spin

2769: injection efficiency, an issue not limited to semiconductor heterostructures.

2770: Related information is currently being sought by spatial imaging of the spin 

2771: polarization 

2772: in spin LED's \cite{Thruber2002:APL,Thruber2003:JMR} 

2773: using magnetic resonance

2774: force microscopy \cite{Sidles1995:RMP}.

2775: 

2776: (III,Mn)V ferromagnetic semiconductors are also used to inject spin  

2777: in a spin-LED structures as depicted in 

2778: Fig.~\ref{injexp:3}. Spin injection can be achieved with no external

2779: field, and the reports of high $T_C$ in some compounds suggest  

2780: that all-semiconductor spin LED's could operate at room temperature.

2781: The drawback, however, 

2782: is that the most 

2783: (III,Mn)V's have spin-polarized holes 

2784: (rather than electrons) as the main carriers 

2785: which, due to spin-orbit coupling, lose their polarization

2786: very quickly after being injected into a nonmagnetic semiconductor. 

2787: Consequently, results of the spin injection show only a small degree 

2788: of hole polarization. 

2789: 

2790: In the experiment of \textcite{Ohno1999:N}, 

2791: an intrinsic GaAs spacer of thickness $d$ was 

2792: introduced

2793: between the spin aligner (Ga,Mn)As 

2794: and the (In,Ga)As quantum well. 

2795: The electroluminescence  

2796: in a QW was measured perpendicular to

2797: the growth direction [the easy magnetization axis of (Ga,Mn)As and the

2798: applied magnetic field were both perpendicular to the growth

2799: direction]. The corresponding relation between the $P_{\rm circ}$ 

2800: and

2801: hole density  polarization 

2802: $P_p$ is not

2803: straightforward; 

2804: the analysis was  performed only on the 

2805: electroluminescence 

2806: [for possible difficulties see \textcite{Fiederling2003:APL}].

2807: A small measured signal ($P_{\rm circ}\sim 1 \%$ at $5$ K), 

2808: consistent with the 

2809: expectation for holes as the injected spin-polarized carriers,

2810: was also obtained in an additional experiment \cite{Young2002:APL}.

2811: $P_{\rm circ}$ was approximately independent of the GaAs thickness 

2812: ($d=20-420$ nm), a 

2813: behavior which remains to be understood considering that the 

2814: hole spins 

2815: should relax fast \cite{Hilton2002:PRL} 

2816: as they are

2817: transfered across the nonmagnetic semiconductor.\footnote{A possible 

2818: exception is QW, in which the effects of quantum confinement and 

2819: quenching spin-orbit coupling lead to longer $\tau_s$.}

2820: In contrast, for a repeated experiment \cite{Young2002:APL}

2821: using a Faraday

2822: geometry (as in Fig.~\ref{injexp:3}), with both measured electroluminescence  

2823: and $B$ along the growth direction, the same change of thickness

2824: $P_{\rm circ}$ was reduced from $7 \%$ to $0.5\%$. A highly

2825: efficient spin injection of $P_n \approx 80\%$ in GaAs has been realized

2826: using (Ga,Mn)As

2827: as a spin injector in a Zener diode structure \cite{vanDorpe2003:P}. 

2828: The detection

2829: employed the technique of an oblique Hanle effect, discussed in 

2830: the next section.

2831: 

2832: All-electrical spin injection studies of trilayer structures

2833: (II,Mn)VI/II-VI/(II,Mn)VI have displayed up to 25\% MR at $B\approx5$ T and

2834: $T=4$ K \cite{Schmidt2001:PRL}. A strong suppression of this MR signal

2835: at applied bias of $\sim10$ mV was attributed to the nonlinear regime

2836: of spin injection, in  which the effects of band bending 

2837: and charge accumulation

2838: at the (II,Mn)VI/II-VI interface were important \cite{Schmidt2002:P}.

2839: It would be instructive to analyze these measurements by adopting 

2840: the approach discussed in the context of magnetic {\it p-n} junctions

2841: (Secs.~\ref{sec:IIC3} and \ref{sec:IVD}), which self-consistently

2842: incorporates the effects of band bending and deviation from 

2843: local charge neutrality.

2844: 

2845: 

2846: \subsubsection{\label{sec:IID4} Metallic ferromagnet/semiconductor junctions}

2847: %---------------------------------------------------------------------------

2848: 

2849: A large family of  metallic ferromagnets, some of them highly spin polarized,

2850: offer the 

2851: possibility of spin injection at room temperature, even in the 

2852: absence of

2853: applied magnetic field. Spin injection into  (110) GaAs at room temperature 

2854: has been already 

2855: demonstrated 

2856: using vacuum tunneling from a polycrystalline Ni STM tip and 

2857: optical detection via circularly polarized luminescence 

2858: \cite{Alvarado1992:PRL,Alvarado1995:PRL}. 

2859: It was shown that the minority spin electrons (spin $\downarrow$ in 

2860: the context of

2861: metals; see Sec.~\ref{sec:IIA}) 

2862: in 

2863: Ni produced the dominant contribution to the tunneling current, and the    

2864: resulting polarization was inferred to be $P_n=(-31 \pm 5.6)\%$ 

2865: \cite{Alvarado1992:PRL}. 

2866: Even thought the spin injection in  future spintronic 

2867: devices will likely be implemented by some means other than 

2868: vacuum tunneling, this result 

2869: supports the importance of the tunneling contact for efficient 

2870: spin injection, as discussed in Sec.~\ref{sec:IIC1}. 

2871: Similar studies of spatially 

2872: resolved spin injection,

2873: sensitive to the topography of the GaAs surface, have employed 

2874: a single-crystal 

2875: Ni (100) 

2876: tip \cite{LaBella2001:S}.  At 100 K nearly fully 

2877: spin-polarized injection of electrons

2878: was reported.

2879: However,

2880: further analyses 

2881: of the measurements of 

2882: $P_{\rm circ}$ have substantially reduced these estimates 

2883: to $\approx 25$\% \cite{Egelhoff2002:S,LaBella2002:S}.

2884: 

2885: Direct spin injection from a ferromagnet into 

2886: a 2DEG,\footnote{For a 

2887: comprehensive review of 2DEG,  

2888: see \textcite{Ando1982:RMP}.} 

2889: motivated by the proposal of \textcite{Datta1990:APL} proposal,

2890: initially  

2891: showed only small effects \cite{Lee1999:JAP,Hammar1999:PRL,Gardelis1999:PRB}, 

2892: with $\Delta R/R \sim 1\%$,

2893: or effects within the noise \cite{Filip2000:PRB}. Such inefficiency could be 

2894: attributed to the resistance mismatch in the F and N regions, discussed 

2895: in Secs.~\ref{sec:IIC1} and \ref{sec:IIC2}. 

2896: The possibility of spurious effects arising from the Hall and 

2897: anisotropic magnetoresistance signals in

2898: similar structures was suggested earlier  \cite{Monzon1997:APL}

2899: as well as after the initial 

2900: experiments \cite{Monzon2000:PRL,vanWees2000:PRL,Tang:2002}.

2901: Control measurements have been performed to address these issues

2902: \cite{Hammar2000:PRL,Hammar2000:PRB}. This debate about the 

2903: presence/absence of spin injection effects via Ohmic contacts stimulated

2904: further studies, 

2905: but the experimental focus has since shifted to  other 

2906: approaches.

2907: 

2908: Spin injection via Schottky contacts at room temperature was demonstrated

2909: in a Fe/GaAs junction by \textcite{Zhu2001:PRL}, 

2910: who reported 

2911: detection of 

2912: $P_{\rm circ} \approx 2\%$ using spin LED structures and optical detection as 

2913: described in \ref{sec:IID3}.

2914: These studies were extended \cite{Ramsteiner2002:PRB}

2915: by using MBE to grow 

2916: MnAs,  a ferromagnetic metal,  

2917: on top of the GaAs to provide high-quality interfaces \cite{Tanaka2002:SST}. 

2918: There was no preferential behavior for spin injection using different 

2919: azimuthal orientations of the epitaxial MnAs layer, which could have been

2920: expected from the symmetry between 

2921: conduction-band wave functions in MnAs and GaAs.

2922: The tunneling properties of a Schottky barrier

2923: were discussed by 

2924: \textcite{Meservey1982:JAP,Gibson1985:JAP,Prins1995:JPCM,Kreuzer2002:APL}.

2925: The measured I-V curves display a complicated behavior 

2926: \cite{Hirohata2001:PRB,Isakovic2001:PRB} which can be significantly affected

2927: by the interface (midgap) states at the Schottky barrier \cite{Jonker1997:PRL}.

2928: A theoretical explanation of this behavior is still lacking.  

2929:  

2930: \begin{figure} 

2931: \centerline{\psfig{file=zutic_fig14.eps,width=1.0\linewidth}}

2932: \caption{Electroluminescence (EL) and polarization due to spin injection

2933: from Fe Schottky contact: (a) EL spectra 

2934: from a surface-emitting 

2935: spin LED with an Fe/AlGaAs

2936: Schottky tunnel contact  for selected values of applied magnetic field, 

2937: analyzed for $\sigma^{\pm}$ 

2938: circular polarization. The large difference in intensity

2939: between these components indicates successful spin injection from the Fe 

2940: into the GaAs

2941: QW, and reveals an electron spin polarization in the QW of 32\%.  

2942: The magnetic field is

2943: applied along the surface normal (Faraday geometry).  

2944: The spectra are dominated by the heavy

2945: hole (HH) exciton.  Typical operating parameters are 1 mA and 2 V.

2946: (b)  Magnetic-field dependence of the EL circular polarization of the 

2947: HH exciton.  The

2948: polarization tracks the hard axis magnetization of the Fe contact, 

2949: and saturates at an

2950: applied magnetic-field value 4$\pi$M = 2.2 T, at which the Fe 

2951: magnetization is entirely along

2952: the surface normal. From \onlinecite{Hanbicki2003:P}.}  

2953: \label{injexp:5}

2954: \end{figure} 

2955: 

2956: As discussed in Sec.~\ref{sec:IIC1}, tunnel contacts formed between a metallic 

2957: ferromagnet and a semiconductor can provide effective spin injections.

2958: Optical detection in spin LED structures, as discussed 

2959: in Sec.~\ref{sec:IID3},

2960: was used to show carrier polarization of $P_n \approx 30 \%$ using 

2961: Fe as a spin injector \cite{Hanbicki2003:P}.

2962: Experimental results are given in Fig.~\ref{injexp:5}.

2963: Some care has to be taken in defining the efficiency of the spin injection,

2964: normalized to the polarization of a ferromagnet, as used in related previous

2965: experiments~\cite{Hanbicki2002:APLa,Jansen2002:APL,Hanbicki2002:APLb}. 

2966: Furthermore,

2967: there are often different conventions for defining the sign of $P_n$ 

2968: \cite{Hanbicki2003:P},  used in the context of semiconductors and ferromagnetic 

2969: metals, as discussed in Sec.~\ref{sec:IIA}. 

2970: \textcite{Jiang2004:P} have demonstrated that MgO can be a suitable choice of an

2971: insulator for highly efficient spin injection into GaAs. Spin LED with 

2972: GaAs/AlGaAs quantum well was used to detect $P_n \approx 50 \%$

2973: at 100 K injected from CoFe/MgO (100) tunnel injector. While quantum

2974: well emission efficiency limits detection at higher temperatures ($>100$ K),

2975: the same tunnel injector should also be suitable for efficient spin

2976: injection even at room temperature. 

2977: 

2978: 

2979: An oblique Hanle effect \cite{D'yakonov1975:SPJETP} (see also 

2980: Secs.~\ref{sec:IIB}

2981: and \ref{sec:IID1})

2982: was used \cite{Motsnyi2002:APL,Motsnyi2003:PRB,VanDorpe2002:P}

2983: to detect spin injection, giving up to $P_n\approx 16 \%$, 

2984: at room temperature.

2985: The geometry used is similar to that sketched in Fig.~\ref{injexp:3}(a),

2986: with an insulating layer (AlO$_x$) separating the ferromagnetic spin injector

2987: and the (Al,Ga)As/GaAs spin LED.

2988: The magnetization easy axis lies in the plane of 

2989: the ferromagnet. 

2990: An oblique

2991: magnetic field is applied to give a net out-of-plane component of injected

2992: spin which could contribute to the emission of circularly polarized light.

2993: This approach allows one to apply a magnetic field several times smaller 

2994: than would be 

2995: needed to pull the magnetization out of plane

2996: [for Fe it is $\approx 2$ T \cite{Hanbicki2002:APLa}] and to measure polarized

2997: luminescence in a Faraday geometry. Furthermore, using standard measurements

2998: of the Hanle curve, one can extract separately the spin lifetime

2999: and carrier recombination time.

3000: 

3001: Hot-electron spin injection above the Schottky barrier is another method for 

3002: using a high polarization of metallic ferromagnets  

3003: to create a nonequilibrium 

3004: spin in a semiconductor even at room temperatures. 

3005: Typically such injection is performed in three-terminal, transistor-like

3006: devices, as discussed in Sec.~\ref{sec:IVE3}.

3007: 

3008: Direct electrical spin injection has also been demonstrated in organic 

3009: semiconductors \cite{Dediu2002:SSC}. MR measurements were performed in 

3010: an F/N/F junction, where F is 

3011: La$_{0.7}$Sr$_{0.3}$MnO$_3$ 

3012: (LSMO), a colossal magnetoresistive manganite, and N is 

3013: is sexithienyl (T$_6$), a  

3014: $\pi$-conjugated rigid-rod oligomer organic semiconductor. The decrease of MR

3015: with increasing thickness of the N region was used to infer 

3016: $L_{sN} \sim 100$ nm at room temperature. The resulting spin 

3017: diffusion length is a combination of low mobility, 

3018: $\sim$ 10$^{-4}$cm$^2$V$^{-1}$s$^{-1}$

3019: (about $\sim$ 10$^7$ times smaller than for the bulk GaAs),

3020: and long spin relaxation times, $\sim$ $\mu$s,\footnote{This is

3021: also a typical value

3022: for other organic semiconductors \cite{Krinichnyi2000:SM}, 

3023: a consequence of weak spin-orbit coupling

3024: \cite{Davis2003:JAP}.} as compared to the usual III-V inorganic semiconductors.

3025: Motivated by these findings \textcite{Xiong2004:N} have replaced one of

3026: the LSMO electrodes by Co. Different coercive fields in the two ferromagnetic

3027: electrodes allowed them to measure a spin-valve effect with

3028: $\Delta R / R \sim 40 \%$ at 11 K.

3029: Related theoretical studies of the ferromagnetic metal/conjugated polymer

3030: interfaces were reported by \textcite{Xie2003:PRB}. 

3031: 

3032: \section{\label{sec:III} Spin relaxation and spin dephasing}

3033: %===========================================================================

3034: 

3035: \subsection{\label{sec:IIIA} Introduction}

3036: %------------------------------------------------------------------------

3037: 

3038: Spin relaxation and spin dephasing are processes that lead to spin 

3039: equilibration

3040: and are thus of great importance for spintronics. The fact that 

3041: nonequilibrium

3042: electronic spin in metals and semiconductors lives relatively long (typically a

3043: nanosecond), allowing for spin-encoded information to travel macroscopic 

3044: distances, is 

3045: what makes spintronics a viable option for technology. After introducing the 

3046: concepts of spin relaxation and spin 

3047: dephasing times $T_1$ and $T_2$, 

3048: respectively, 

3049: which

3050: are commonly called $\tau_s$ throughout this paper, we

3051: discuss four major physical mechanisms responsible for spin equilibration in 

3052: {\it nonmagnetic} electronic

3053: systems: Elliott-Yafet, D'yakonov-Perel', Bir-Aronov-Pikus, and 

3054: hyperfine-interaction 

3055: processes. We then survey recent works on electronic spin relaxation in 

3056: nonmagnetic metals and semiconductors, using the important examples of

3057: Al 

3058: and GaAs for illustration.

3059: 

3060: \subsubsection{\label{sec:IIIA1} $T_1$ and $T_2$}

3061: 

3062: Spin relaxation and spin dephasing of a spin ensemble are traditionally defined 

3063: within the

3064: framework of the Bloch-Torrey equations~\cite{Bloch1946:PR,Torrey1956:PR} 

3065: for magnetization dynamics. For mobile electrons, spin 

3066: relaxation time $T_1$ (often called longitudinal or spin-lattice time) and

3067: spin dephasing time $T_2$ (also transverse or decoherence time) are

3068: defined via the equations for the spin precession, decay, and diffusion of 

3069: electronic 

3070: magnetization $\bf M$ in an applied magnetic field ${\bf B}(t)=B_0\hat{\bf 

3071: z}+{\bf B}_1(t)$, 

3072: with a static longitudinal component $B_0$ (conventionally in the $\hat{z}$ 

3073: direction) and, frequently, a transverse oscillating part ${\bf B}_1$ 

3074: perpendicular to $\hat{\bf z}$

3075:  ~\cite{Torrey1956:PR,Kaplan1959:PR}:

3076: \begin{eqnarray} 

3077: \label{eq:relax:bloch1}

3078: \frac{\partial M_{x}}{\partial t}&=&\gamma ({\bf M} \times {\bf B})_x-

3079: \frac{M_x}{T_2}+D\nabla^2 M_x, \\

3080: \label{eq:relax:bloch2} 

3081: \frac{\partial M_{y}}{\partial t}&=&\gamma ({\bf M} \times {\bf B})_y-

3082: \frac{M_y}{T_2}+D\nabla^2 M_y, \\

3083: \label{eq:relax:bloch3}

3084: \frac{\partial M_{z}}{\partial t}&=&\gamma 

3085: ({\bf M} \times {\bf B})_z-\frac{M_z-M^0_z}{T_1}+D\nabla^2 M_z.

3086: \end{eqnarray}

3087: Here $\gamma=\mu_B g/\hbar$ is the electron gyromagnetic ratio 

3088: ($\mu_B$ is the Bohr 

3089: magneton 

3090: and $g$ 

3091: is the electronic $g$ factor), $D$ is the diffusion coefficient (for simplicity

3092: we assume an isotropic or a cubic solid with scalar $D$), and 

3093: $M_z^0=\chi B_0$ is the thermal equilibrium magnetization with $\chi$ 

3094: denoting the system's static susceptibility. 

3095: The Bloch equations are phenomenological, describing quantitatively 

3096: very well the dynamics of mobile electron spins (more properly, magnetization)

3097: in experiments such as conduction electron spin resonance and optical 

3098: orientation. 

3099: Although relaxation and decoherence processes in a 

3100: many-spin system are generally

3101: too complex to be fully described by only two parameters, 

3102: $T_1$ and

3103: $T_2$ are nevertheless an extremely robust and convenient measure 

3104: for quantifying 

3105: such processes in many cases of interest. 

3106: To obtain microscopic expressions for spin relaxation and dephasing times,

3107: one starts with a microscopic description of the spin system (typically

3108: using the density-matrix approach), derives the magnetization dynamics, 

3109: and compares 

3110: it 

3111: with the Bloch equations to extract $T_1$ and $T_2$.

3112: 

3113: Time $T_1$ is the time it takes for the longitudinal magnetization to reach 

3114: equilibrium. 

3115: Equivalently, it is the time of thermal equilibration of the spin population 

3116: with 

3117: the lattice.  

3118: In $T_1$ processes an energy has to 

3119: be taken from the spin system, usually by phonons, to the lattice. 

3120: Time $T_2$ is classically the time it takes for an 

3121: ensemble of transverse electron spins, initially precessing in phase  

3122: about the 

3123: longitudinal

3124: field, to lose their phase due to spatial and temporal fluctuations of the 

3125: precessing frequencies. 

3126: For an ensemble of mobile electrons the measured $T_1$ and $T_2$ come about by

3127: averaging spin over the thermal distribution of electron momenta. Electrons

3128: in different momentum states have not only different spin-flip characteristics,

3129: but also slightly different $g$ factors and thus different precession 

3130: frequency. This is analogous to precession frequencies fluctuations

3131: of localized spins due to inhomogeneities in the static field $B_0$. 

3132: However, since momentum scattering (analogous to intersite hopping or exchange 

3133: interaction 

3134: of localized spins) typically proceeds much faster than spin-flip scattering,

3135: the $g$-factor-induced broadening is inhibited by motional 

3136: narrowing\footnote{Motional (dynamical) 

3137: narrowing is an inhibition of  phase change

3138: by random fluctuations \cite{Slichter:1989}. 

3139: Consider a spin rotating with frequency

3140: $\omega_0$. The spin phase changes by 

3141: $\Delta \phi = \omega_0 t$ 

3142: over time $t$. If the spin is subject to a random force that 

3143: makes spin precession equally likely clockwise and anticlockwise,

3144: the average spin phase does not change, but the root-mean-square phase change

3145: increases with time as 

3146: $(\langle \Delta^2 \phi \rangle)^{1/2} \approx 

3147: (\omega_0 \tau_c) (t/\tau_c)^{1/2}$,

3148: where $\tau_c$ is the correlation time of the random force, or the average

3149: time of spin precession in one direction. 

3150: This is valid for rapid fluctuations, $\omega_0 \tau_c \ll 1$.

3151: The phase relaxation time  

3152: $t_\phi$ is defined as the time over which the phase 

3153: fluctuations reach unity: $1/t_\phi=\omega_0^2 \tau_c$.}

3154: and

3155: need not be generally considered as contributing to $T_2$ 

3156: [see, however,~\cite{Dupree1967:PSS}]. Indeed, motional narrowing of the 

3157: $g$-{factor

3158: fluctuations, $\delta g$, 

3159: gives a contribution to $1/T_2$ of the order of 

3160: $\Delta \omega^2\tau_p$, where the $B_0$-dependent precession frequency

3161: spread is 

3162: $\Delta \omega = (\delta g/g) \gamma B_0$ 

3163: and $\tau_p$ is the momentum scattering time. For $B_0$ fields of the order 

3164: 1 T, 

3165: scattering times of 1 ps, and $\delta g$ as large as 0.01, 

3166: the ``inhomogeneous broadening'' is a microsecond, which is much 

3167: more than 

3168: the observed values for $T_2$. Spatial inhomogeneities of $B_0$, 

3169: like those coming 

3170: from hyperfine fields, are inhibited by motional narrowing, too, due to the

3171: itinerant nature 

3172: of electrons. For localized electrons (e.g., for donor states in 

3173: semiconductors),

3174: spatial inhomogeneities play an important role 

3175: and are often observed to affect 

3176: $T_2$. 

3177: To describe such reversible phase losses, which can potentially be 

3178: eliminated by spin-echo experiments, sometimes the symbol $T_2^*$ 

3179: ~\cite{Hu2001:lanl}

3180: is used to describe spin dephasing of ensemble spins, while the symbol $T_2$ is

3181: reserved for irreversible loss of the ensemble spin phase. 

3182: In general, $T_2^* \le T_2$, 

3183: although for conduction electrons to a very good approximation $T_2^*=T_2$. 

3184: 

3185: In isotropic and cubic solids $T_1=T_2$ if $\gamma B_0 \ll 1/\tau_c$, 

3186: where $\tau_c$ is the so-called correlation or interaction time: $1/\tau_c$ is

3187: the rate of change of the effective dephasing magnetic field 

3188: (see footnote 70).  

3189: Phase losses occur during time intervals of $\tau_c$.

3190: As shown below, 

3191: in electronic systems $\tau_c$ is given either by $\tau_p$ or by the

3192: time of the interaction of electrons with phonons and holes. Those

3193: times are typically smaller than a picosecond, so $T_1=T_2$ is fulfilled for 

3194: magnetic

3195: fields up to 

3196: several tesla. The equality between the relaxation and 

3197: dephasing

3198: times was noticed first in the context of 

3199: NMR~\cite{Bloch1946:PR,Wangsness1953:PR}

3200: and later extended to electronic spin systems 

3201: ~\cite{Pines1955:PR,Andreev1958:JETP}.

3202: A qualitative reason for $T_1=T_2$ is that if the phase acquires a random 

3203: contribution 

3204: during a short time interval $\tau_c$, the energy uncertainty of the spin 

3205: levels 

3206: determining the longitudinal spin is greater than the Zeeman splitting $\hbar 

3207: \gamma B_0$ 

3208: of the levels. The splitting then does not play a role in dephasing, and 

3209: the dephasing field will act equally on the longitudinal and

3210: transverse spin. In classical terms, spin that is 

3211: oriented along the direction of the 

3212: magnetic

3213: field can precess a full period about the perpendicular fluctuating field, 

3214: feeling

3215: the same dephasing fields as transverse components. As the external field 

3216: increases,

3217: the precession angle of the longitudinal component is reduced, inhibiting 

3218: dephasing.

3219: 

3220: The equality of the two times is very convenient for comparing

3221: experiment and theory, since measurements usually yield $T_2$, 

3222: while theoretically it is often more convenient to calculate $T_1$. In many

3223: cases a single symbol $\tau_s$ is used for spin relaxation and dephasing

3224: (and called indiscriminately either of these terms), 

3225: if it does not matter what 

3226: experimental situation is involved, or if one is working at small magnetic 

3227: fields.\footnote{Sometimes one finds as spin relaxation

3228: time $2\tau_s$. 

3229: While this is correct for just one spin state, 

3230: conventionally by spin relaxation we mean magnetization relaxation, 

3231: in which each spin flip adds

3232: to the equilibration of both spin up and spin down states, doubling the

3233: rate for magnetization relaxation.} Throughout this paper we adopt this

3234: notation to avoid unnecessary confusion.

3235: 

3236: If the system is anisotropic, the equality $T_1=T_2$ no longer holds, even in

3237: the case of full motional narrowing of the spin-spin

3238: interactions and $g$-factor 

3239: broadening. Using simple qualitative analysis

3240: \textcite{Yafet:1963} showed that, while there is no general relation

3241: between the two times, the inequality $T_2 \le 2 T_1$ holds, and that

3242: $T_2$ changes with the direction by at most a factor of 2. In the motionally

3243: narrowed case this difference between $T_1$ and $T_2$ can be considered

3244: as arising from the tensorial nature of the spin relaxation rate. Specific

3245: examples of this will be discussed in studying spin relaxation in 

3246: two-dimensional semiconductor heterostructures (Sec.~\ref{sec:IIIB2}).

3247: 

3248: Finally, we discuss the connection between $\tau_s$ and the single-spin 

3249: decoherence 

3250: time\footnote{To distinguish ensemble and individual spin dephasing, we use

3251: the term decoherence in relation to single, or a few, spins.},

3252: $\tau_{sc}$, which is the single-spin 

3253: correlation time. Time $\tau_{sc}$ becomes important for applications 

3254: in spin-based 

3255: quantum computing \cite{Loss1998:PRA,Hu2000:PRA}, 

3256: where spin coherence needs to last for at least $10^4-10^6$ gate

3257: operations for the computation to be fault tolerant \cite{Preskill1998:PRSL}. 

3258: The relative magnitudes of $\tau_s$ and $\tau_{sc}$ depend on many factors. 

3259: Often

3260: $\tau_{sc} < \tau_s$, as is the case for a direct exchange interaction causing 

3261: single-spin decoherence, while contributing to ensemble dephasing only as a

3262: dynamical averaging factor (the exchange interaction preserves total spin).  

3263: An

3264: analogy with momentum scattering may be helpful. Electron-electron collisions 

3265: lead to

3266: individual momentum equilibration while conserving the total momentum 

3267: and hence 

3268: do not contribute 

3269: to charge current

3270: (unless umklapp processes are taken into account).

3271: A momentum scattering time obtained 

3272: from conductivity (analogous to $\tau_s$) 

3273: would thus

3274: be very different from a single-state momentum time (analogous to 

3275: $\tau_{sc}$).

3276: It is not clear at present how much $\tau_s$ and $\tau_{sc}$ differ for 

3277: different 

3278: materials under different conditions, although it is often, with little 

3279: justification, 

3280: assumed that they are similar. As \textcite{Dzhioev2002:PRB} 

3281: recently suggested,

3282: $\tau_{sc}$ can be smaller than $\tau_s$ by several orders of magnitude 

3283: in n-GaAs at low doping densities where electrons are localized in 

3284: donor states, see also Sec.~\ref{sec:IIID3a}. 

3285: We note that it is $\tau_s$ that is relevant for spintronics (spin transport) 

3286: applications, while $\tau_{sc}$ is relevant for solid-state quantum computing. 

3287: Much remains to be learned

3288: about $\tau_{sc}$.

3289: 

3290: 

3291: \subsubsection{\label{sec:IIIA2} Experimental probes}

3292: %---------------------------------------------------------------------------

3293: 

3294: Experiments detecting spin relaxation and decoherence of conduction electrons 

3295: can be grouped into two

3296: broad categories: (a) those measuring spectral characteristics of magnetization 

3297: depolarization

3298: and (b) those measuring time or space correlations of 

3299: magnetization.

3300: 

3301: Experiments of type (a) are exemplified by conduction-electron spin 

3302: resonance (CESR) 

3303: and optical orientation 

3304: in combination with the Hanle effect. CESR was the first technique used to 

3305: detect the spin

3306: of conduction electrons in metals ~\cite{Griswold1952:PR,Feher1955:PR}

3307: and donor states in semiconductors like Si \cite{Feher1959b:PR}. 

3308: In GaAs, spin resonance techniques are aided by other measurements, e.g., 

3309: optical detection \cite{Weisbuch1977:PRB}, photoconductivity 

3310: \cite{Stein1983:PRL},

3311: or magnetoresistance \cite{Dobers1988:PRB}.

3312: The technique measures signatures of resonant absorption of microwaves by a 

3313: Zeeman-split

3314: spin system of conduction electrons. Typically changes in surface 

3315: impedance and 

3316: in 

3317: the transmission coefficient of the sample are observed. By comparing the 

3318: absorption

3319: resonance curve with theory~\cite{Dyson1955:PR,Kaplan1959:PR} 

3320: one can obtain

3321: both  the carrier $g$ factor and $T_2$. 

3322: 

3323:  

3324: Optical spin orientation~\cite{Meier:1984} is a method

3325: of exciting spin-polarized carriers (electrons and holes) in direct-gap 

3326: semiconductors like GaAs by absorption of circularly polarized light 

3327: (see Sec.~\ref{sec:IIB}). 

3328: The injected spin polarization can be detected by observing circularly 

3329: polarized luminescence resulting from recombination of the spin-polarized 

3330: electrons and 

3331: holes.  

3332: Since in a steady state the excited spin polarization depends not only on 

3333: $\tau_s$

3334: but also on the carrier recombination time, the Hanle effect,

3335: polarization of luminescence by a transverse magnetic field 

3336: (see Sec.~\ref{sec:IID1}),

3337: is employed to deduce 

3338: $\tau_s$ unambiguously.

3339: 

3340: The Hanle effect has also been a great tool for investigating $T_2$ in metals,

3341: in connection with electrical spin injection. The advantage of optical 

3342: orientation

3343: over CESR is that, if carrier lifetime is known, zero-field 

3344: (or zero-$g$-factor 

3345: material) 

3346: data can be measured. In addition, smaller $\tau_s$ values can be detected. 

3347: 

3348: Type (b) experiments measure magnetization in a time or space domain. 

3349: The most important examples are the Johnson-Silsbee spin injection  experiment,

3350: the time-resolved (pump-probe) photoluminescence (in which the probe

3351: is used on the same principle as optical orientation),

3352: and time-resolved

3353: Faraday and (magneto-optic) Kerr rotation. The last two methods can

3354: follow coherent (in the ensemble sense) dynamics of electron spin 

3355: precession. 

3356: 

3357: Spin injection experiments \cite{Johnson1985:PRL} detect the amount of 

3358: nonequilibrium magnetization by observing charge response 

3359: (see Sec.~\ref{sec:IID1}). 

3360: In the Johnson-Silsbee scheme, electrons

3361: are first injected by electrical spin injection from a ferromagnetic

3362: electrode into a metal or semiconductor. As the spin diffuses throughout the 

3363: sample, another ferromagnetic electrode detects the amount of spin as a 

3364: position-dependent quantity. The detection is by means of 

3365: spin-charge coupling,

3366: whereby an EMF appears across the paramagnet/ferromagnet

3367: interface in proportion to the nonequilibrium magnetization 

3368: \cite{Silsbee1980:BMR}. 

3369: By fitting the

3370: spatial dependence of magnetization to the exponential decay formula, 

3371: one can extract the

3372: spin diffusion length $L_s$ and thus the spin relaxation 

3373: time $T_1=L_s^2/D$. 

3374: The Hanle effect, too, can be used in combination with

3375: Johnson-Silsbee spin injection, yielding directly $T_1=T_2$, which

3376: agrees with $T_1$ determined from the measurement of $L_s$.

3377: A precursor to the Hanle effect in spin injection was 

3378: transmission-electron

3379: spin resonance (TESR), in which

3380: nonequilibrium electron spin excited in the skin 

3381: layer of a metallic sample is transported to the other side, emitting 

3382: microwave

3383: radiation. In very clean samples and at low temperatures, electrons

3384: ballistically propagating 

3385: through

3386: the sample experience Larmor precession

3387: resulting in Larmor waves, seen as an oscillation of the transmitted

3388: radiation amplitude with changing static magnetic field~\cite{Janossy1980:PRB}.

3389: 

3390: Time-resolved photoluminescence

3391: detects, after creation of spin-polarized carriers, circular polarization of 

3392: the recombination light. This technique was used, for example, 

3393: to detect a 500 ps spin coherence time $T_2$ of free excitons in a 

3394: GaAs quantum well \cite{Heberle1994:PRL}.

3395: The Faraday and (magneto-optic) Kerr effects are the rotation of the 

3396: polarization

3397: plane of a linearly polarized light, transmitted through (Faraday) 

3398: or reflected by

3399: (Kerr)  a magnetized sample. The Kerr effect is more useful for 

3400: thicker and nontransparent samples or for thin films fabricated on thick 

3401: substrates.  The angle of rotation is proportional

3402: to the amount of magnetization in the direction of the incident light. 

3403: Pump (a circularly polarized laser pulse) and probe experiments employing 

3404: magneto-optic effects can now follow, with 100 fs resolution, the evolution

3405: of magnetization as it dephases in a transverse magnetic field. 

3406: Using lasers for spin excitation has the

3407: great advantage of not only detecting, 

3408: but also 

3409: manipulating spin dephasing, as shown using Faraday rotation on bulk 

3410: GaAs and GaAs/ZnSe 

3411: heterostructures \cite{Kikkawa2001:PE,Awschalom1999:PT}. 

3412: The Kerr effect was used, for example, to investigate the spin dynamics of bulk 

3413: CdTe \cite{Kimel2000:PRB}, 

3414: and Faraday rotation was used to study spin coherence in nanocrystals of 

3415: CdSe \cite{Gupta2002:PRB} and coherent control of spin dynamics of

3416: excitons in GaAs quantum wells \cite{Heberle1996:IEEE}.

3417: 

3418: 

3419: \subsection{\label{sec:IIIB} Mechanisms of spin relaxation}

3420: %==========================================================================

3421: 

3422: Four  mechanisms for spin relaxation of conduction electrons have been found

3423: relevant for metals and semiconductors: the Elliott-Yafet (EY),

3424: D'yakonov-Perel' (DP), Bir-Aronov-Pikus (BIP), and hyperfine-interaction 

3425: (HFI) 

3426: mechanisms.\footnote{We do not 

3427: consider magnetic scattering, that is, 

3428: scattering due to an exchange interaction between conduction electrons

3429: and magnetic impurities.} In the EY

3430: mechanism electron spins relax because the electron wave functions normally

3431: associated with a given spin have an admixture of the

3432: opposite spin states, due to spin-orbit coupling induced by ions. The DP

3433: mechanism explains spin dephasing in solids without a center of symmetry.

3434: Spin dephasing occurs because electrons feel an effective

3435: magnetic field, resulting from the lack of inversion symmetry and from the

3436: spin-orbit interaction, which 

3437: changes in random directions every time the electron

3438: scatters to a different momentum states. The BIP mechanism is 

3439: important for p-doped semiconductors, in which the electron-hole exchange 

3440: interaction

3441: gives rise to fluctuating local magnetic fields flipping electron spins. 

3442: Finally,

3443: in semiconductor heterostructures (quantum wells and quantum dots) based on

3444: semiconductors with a nuclear magnetic moment, 

3445: it is the hyperfine interaction 

3446: of the electron spins and nuclear moments which dominates spin dephasing of 

3447: localized or confined electron spins. An informal review of spin 

3448: relaxation of conduction 

3449: electrons can be found 

3450: in \textcite{Fabian1999:JVST}.

3451: 

3452: \subsubsection{\label{sec:IIIB1} Elliott-Yafet mechanism}

3453: %-------------------------------------------------------------------------

3454: 

3455: \textcite{Elliott1954:PR} was the first to realize that conduction-electron 

3456: spins can relax via

3457: ordinary momentum scattering (such as by phonons or impurities) 

3458: if the lattice ions induce spin-orbit

3459: coupling in the electron wave function. In the presence 

3460: of 

3461: the spin-orbit interaction 

3462: \begin{equation}\label{eq:relax:vso}

3463: V_{so}=\frac{\hbar}{4 m^2 c^2} \nabla V_{sc} \times \hat{\bf p} \cdot 

3464: {\bf \hat{\bf \sigma}},

3465: \end{equation}

3466: where $m$ is the free-electron mass, $V_{sc}$ is the scalar (spin-independent) 

3467: periodic lattice 

3468: potential, $\hat{\bf p}\equiv -i\hbar \nabla$

3469: is the linear momentum operator, and ${\bf \hat{\bf \sigma}}$ 

3470: are the Pauli matrices, 

3471: single-electron (Bloch) wave functions in a solid are no longer the 

3472: eigenstates of 

3473: $\hat{\sigma}_z$, but rather a mixture of the Pauli spin up 

3474: $|\uparrow \rangle$ 

3475: and spin down $|\downarrow \rangle$

3476: states. If the solid possesses a center of symmetry, the case of 

3477: elemental metals which Elliott considered, the Bloch states 

3478: of ``spin up'' and ``spin down'' electrons with the lattice momentum 

3479: $\bf k$ and band index $n$  

3480: can be written as \cite{Elliott1954:PR}

3481: \begin{eqnarray}

3482: \Psi_{{\bf k}n\uparrow}({\bf r})&=&\left [a_{{\bf k}n} ({\bf r}) 

3483: |\uparrow\rangle

3484: +b_{{\bf k}n} ({\bf r}) |\downarrow\rangle  \right ]e^{i{\bf k}\cdot {\bf r}}, 

3485: \\

3486: \Psi_{{\bf k}n\downarrow}({\bf r})&=&\left [a^*_{-{\bf k}n} ({\bf r}) 

3487: |\downarrow\rangle  

3488: -b^*_{-{\bf k}n} ({\bf r}) |\uparrow\rangle  \right ] e^{i{\bf k}\cdot {\bf r}}, 

3489: \end{eqnarray}

3490: where we write the explicit dependence of the complex 

3491: lattice-periodic 

3492: coefficients $a$ and $b$ 

3493: on the radius vector $\bf r$.

3494: The two Bloch states are degenerate: they are connected by spatial 

3495: inversion and time reversal~\cite{Elliott1954:PR}. That it makes sense to call

3496: $\Psi_{{\bf k}n\uparrow}$ and  $\Psi_{{\bf k}n\downarrow}$, respectively,

3497: spin up 

3498: and spin down 

3499: states follows from the fact that in most cases the typical values of $|a|$ are 

3500: close to unity,

3501: while $|b| \ll 1$. 

3502: 

3503: Indeed, consider a band structure in the absence of spin-orbit coupling.

3504: Turning $V_{so}$ on couples electron states of opposite 

3505: Pauli spins which are 

3506: of the same ${\bf k}$ (because $V_{so}$ has the period of the lattice), 

3507: but different $n$. Because the spin-orbit interaction is normally much 

3508: smaller than the

3509: distance between the bands, perturbation theory gives 

3510: \begin{equation} \label{eq:relax:lambda}

3511: |b|\approx \lambda_{so}/\Delta E \ll 1,

3512: \end{equation}

3513: where $\Delta E$ is the energy distance between the band state in question and 

3514: the

3515: state (of the same momentum) 

3516: in the nearest band, and $\lambda_{so}$ is the amplitude of the matrix element 

3517: of $V_{so}$ 

3518: between the two states. 

3519: The spin-orbit coupling alone does not lead to spin relaxation. However, in 

3520: combination 

3521: with

3522:  momentum scattering, the spin up [Eq.~(56)] and spin down  [Eq.~(57)] %iz lab

3523: states

3524: can couple and lead to spin relaxation.

3525: 

3526: Momentum scattering is typically caused by impurities (at low $T$) and phonons 

3527: (at high

3528: $T$). There is another important spin-flip scattering mechanism that involves 

3529: phonons. 

3530: A periodic, lattice ion-induced spin-orbit interaction is modified by phonons 

3531: and can 

3532: directly couple the (Pauli) spin up and spin down states. Such processes 

3533: were first 

3534: considered 

3535: for a jellium model by \textcite{Overhauser1953:PR} [see also 

3536: \cite{Grimaldi1997:PRB}], 

3537: and for band structure systems by \textcite{Yafet:1963}. They must be combined

3538: with the Elliott processes discussed above to form a consistent picture of 

3539: phonon-induced spin relaxation, especially at low 

3540: temperatures~\cite{Yafet:1963}, 

3541: where the

3542: two processes have similar magnitudes. At larger $T$,

3543: phonon-modified $V_{so}$ 

3544: is not 

3545: important for polyvalent metals, whose spin relaxation is dominated by spin hot 

3546: spots---states

3547: with anomalously large $|b|$---as shown by the explicit calculation 

3548: of \textcite{Fabian1999:PRL}. Spin hot spots are discussed in more detail in

3549: Sec.~\ref{sec:IIIC}.  

3550: 

3551: We now give a recipe, useful for elemental metals and semiconductors, 

3552: for calculating phonon-induced $1/\tau_s$ 

3553: from the known band and phonon structure. 

3554: The corresponding theory was systematically developed by \textcite{Yafet:1963}.

3555: The spin relaxation rate due to phonon scattering according 

3556: to the EY mechanism

3557: can be expressed through the spin-flip Eliashberg function 

3558: $\alpha_S^2F(\Omega)$

3559: as ~\cite{Fabian1999:PRL}

3560: \begin{equation} \label{eq:relax:ey}

3561: 1/\tau_s=8\pi T\int_{0}^{\infty} d\Omega \alpha_s^2F(\Omega)\partial 

3562: N(\Omega)/\partial T, 

3563: \end{equation}

3564: where $N(\Omega)=[\exp(\hbar \Omega/k_B T)-1]^{-1}$ is the phonon distribution 

3565: function.

3566: The spin-flip Eliashberg function gives the contribution of the phonons with 

3567: frequency

3568: $\Omega$ to the spin-flip electron-phonon interaction, 

3569: \begin{equation} \label{eq:relax:ef}

3570: \alpha_s^2F(\Omega)=\frac{g_S}{2M\Omega}\sum_{\nu}\langle\langle 

3571: g_{{\bf k}n\uparrow, {\bf k}'n'\downarrow}^{\nu}\delta (\omega_{{\bf q}\nu}-

3572: \Omega)

3573: \rangle_{{\bf k}n}\rangle _{{\bf k}'n'}.

3574: \end{equation}

3575: Here $g_S$ is the number of electron states per spin and atom at the Fermi 

3576: level, 

3577: $M$ is the ion mass, $\omega_{{\bf q}\nu}$ is the frequency of phonons

3578: with momentum ${\bf q}={\bf k}-{\bf k}'$ and branch index $\nu$,

3579: and the spin-flip matrix element is

3580: \begin{equation} 

3581: g_{{\bf k}n\uparrow, {\bf k}'n'\downarrow}^{\nu}\equiv |{\bf u}_{{\bf 

3582: q}\nu}\cdot

3583: \left (\Psi_{{\bf k}n\uparrow}, \nabla V \Psi_{{\bf k}'n'\downarrow} 

3584: \right )  |^2,

3585: \end{equation}

3586: where ${\bf u}_{{\bf q}\nu}$ is the polarization of the phonon with momentum 

3587: $\bf q$ and in branch $\nu$. 

3588: The brackets $\langle ... \rangle_{{\bf k}n}$ in Eq.~(\ref{eq:relax:ef}) 

3589: denote Fermi-surface averaging. The Bloch wave functions for this 

3590: calculation are chosen 

3591: to satisfy $(\Psi_{{\bf k}n\uparrow}, \hat {\sigma}_z

3592: \Psi_{{\bf k}n\uparrow})=-(\Psi_{ {\bf k}n\downarrow}, \hat {\sigma}_z 

3593: \Psi_{{\bf k}n\downarrow})$.

3594: The periodic lattice-ion interaction $V$ contains both scalar and 

3595: spin-orbit parts: $V=V_{sc}+V_{so}$.

3596: 

3597: There are two important relations giving  an order-of-magnitude estimate

3598: of $\tau_s$, as well as 

3599: its temperature dependence: the Elliott and the

3600: Yafet relations.

3601: The Elliott relation relates $\tau_s$ to the shift

3602: $\Delta g$ of the electronic $g$ factor from the free-electron value of

3603: $g_0=2.0023$~\cite{Elliott1954:PR}:

3604: \begin{eqnarray} \label{eq:relax:elliott}

3605: 1/\tau_s \approx (\Delta g)^2/\tau_p,

3606: \end{eqnarray}

3607: where $\tau_p$ is the momentum relaxation time.

3608: This relation follows from the fact that for a momentum scattering

3609: interaction $V_i$ the spin-flip scattering probability in the

3610: Born approximation is proportional to

3611: $|(\Psi_{{\bf k}n \uparrow},

3612: V_i \Psi_{{\bf k}'n' \downarrow})|^2 \approx |b|^2\times|(\Psi_{{\bf k}n

3613: \uparrow},

3614: V_i \Psi_{{\bf k}'n' \uparrow})|^2$. Realizing that the spin-conserving

3615: scattering

3616: probability gives the momentum relaxation rate, after a 

3617: Fermi-surface averaging we

3618: get the  estimate

3619: \begin{equation} \label{eq:relax:aux}

3620: 1/\tau_s \approx \langle b^2 \rangle  /\tau_p.

3621: \end{equation}

3622: 

3623: On the other hand, $\Delta g$ is determined by the expectation value of 

3624: $\hat{l}_z$, 

3625: the z-component of the orbital momentum 

3626: in a Bloch state. Without the spin-orbit 

3627: interaction this expectation value is 

3628: zero. 

3629: Considering the spin-orbit interaction to be a small parameter,

3630: we find by 

3631: perturbation theory $\Delta g \approx |b|$, 

3632: which combines with Eq.~(\ref{eq:relax:aux}) to give the Elliott relation 

3633: Eq.~(\ref{eq:relax:elliott}). An empirical test of the Elliott relation 

3634: for simple

3635: metals when spin relaxation is due to thermal phonons 

3636: ~\cite{Beuneu1978:PRB} gives the revised estimate 

3637: \begin{equation} \label{eq:revisedE}

3638: 1/\tau_s \approx 10\times (\Delta g)^2 /\tau_p. 

3639: \end{equation}

3640: 

3641: The Elliott relation is only a very rough estimate of $\tau_s$. The

3642: experimentally relevant ratio  $\tau_p/\tau_s$ depends on the scattering

3643: mechanism. The ratio is different for scattering off impurities,

3644: boundaries, or phonons, although one would expect it to be

3645: within an order of magnitude. For example, scattering by heavy impurities 

3646: induces an

3647: additional spin relaxation channel where spin flip is due to the 

3648: spin-orbit interaction induced by the impurities. 

3649: Equation (\ref{eq:relax:aux}) then

3650: does not hold.  

3651: Scattering by phonons is too complex to be simply equated 

3652: with the ratio

3653: $\tau_p/\tau_s$ 

3654: for impurity or boundary scattering. However, the ratio is comparable

3655: for scattering by light impurities and by boundaries. The ratio

3656: $\tau_p/\tau_s$ for impurity and phonon scattering in Al and Cu is

3657: compared by \textcite{Jedema2003:PRB}.

3658: 

3659: The Yafet relation is only qualitative, connecting the temperature 

3660: dependence of 

3661: $1/T_1$ with 

3662: that of the resistivity $\rho$:

3663: \begin{equation} \label{eq:relax:yafet}

3664: 1/T_1(T) \sim  \langle b^2 \rangle \rho (T),

3665: \end{equation}

3666: as follows directly from Eq.~(\ref{eq:relax:aux}) after realizing that 

3667: $1/\tau_p(T)\sim \rho (T) $. By careful symmetry considerations 

3668: \textcite{Yafet:1963} 

3669: proved that 

3670: $1/T_1 \sim T^5$ at low temperatures, similarly to the Bloch-Gr\"{u}neisen 

3671: law for resistivity,

3672: justifying Eq.~(\ref{eq:relax:yafet}) over a large temperature range.   

3673: Yafet's $T^5$ law stems from the nontrivial fact that for spin-flip 

3674: electron-phonon scattering 

3675: $g_{{\bf k}n\uparrow,{\bf k}'n\downarrow}^\nu \sim

3676: ({\bf k}-{\bf k}')^4$ as ${\bf k} \rightarrow {\bf k}'$ ~\cite{Yafet:1963}, 

3677: while only a quadratic dependence holds for the spin-conserving scattering. 

3678: This corresponds to the long-wavelength

3679: behavior $\alpha_S^2 F(\Omega)\sim \Omega^4$ of the spin-flip Eliashberg 

3680: function. The Yafet relation was tested experimentally by 

3681: \textcite{Monod1979:PRB}. This work led to a deeper understanding of spin 

3682: relaxation processes

3683: in polyvalent metals \cite{Silsbee1983:PRB,Fabian1998:PRL}.

3684: 

3685: Realistic calculations of the EY $\tau_s$ in semiconductors  

3686: can be performed analytically using approximations of the 

3687: band and phonon structures, as

3688: most important states are usually  around high symmetry points. Here we give 

3689: a formula for the spin relaxation of conduction electrons with energy 

3690: $E_{\bf k}$

3691: in the frequently studied case of III-V semiconductors 

3692: ~\cite{Chazalviel1975:PRB,Pikus:1984}:

3693: \begin{equation} \label{eq:relax:chazalviel}

3694: \frac{1}{\tau_s(E_{\bf k})}= A\left 

3695: (\frac{\Delta_{so}}{E_g+\Delta_{so}}\right )^2\left 

3696: (\frac{E_{\bf k}}{E_g}\right )^2\frac{1}

3697: {\tau_p(E_{\bf k})}, 

3698: \end{equation}

3699: where $\tau_p(E_{\bf k})$ 

3700: is the momentum scattering time at energy $E_{\bf k}$, 

3701: $E_g$ is the energy gap,

3702: and $\Delta_{so}$ is the spin-orbit splitting of the valence band 

3703: (see Fig.~\ref{oo:1}). 

3704: The Numerical factor

3705: $A$, which is of order 1, depends on the dominant scattering 

3706: mechanism (charge or neutral impurity, phonon, electron-hole).

3707: Analytic formulas

3708: for the EY mechanism due to electron-electron scattering 

3709: are given by

3710: \textcite{Boguslawski1980:SSC}.

3711: 

3712: Equation~(\ref{eq:relax:chazalviel}) shows that the EY mechanism is important 

3713: for small-gap semiconductors with large spin-orbit splitting (the prototypical 

3714: example is InSb). For degenerate electron densities

3715: the spin relaxation time is given by Eq.~(\ref{eq:relax:chazalviel}) 

3716: with $E_{\bf k}=E_F$,

3717: while for nondegenerate densities the thermal averaging leads to a substitution 

3718: of thermal energy $k_B T$ for $E_{\bf k}$ and thermal-averaged momentum 

3719: relaxation

3720: time for $\tau_p$. To estimate $\tau_s$ from 

3721: Eq.~(\ref{eq:relax:chazalviel}) one needs to know $\tau_p$. 

3722: It often suffices to know the doping or temperature dependence of $\tau_p$ 

3723: to decide on the relevance of the EY mechanism \cite{Pikus:1984}.  

3724: 

3725: The temperature dependence of $1/\tau_s$ for metals and degenerate  

3726: semiconductors follows the dependence of $1/\tau_p$. 

3727: In metals this means a constant at low $T$ 

3728: and a linear increase at large $T$. 

3729: For nondegenerate semiconductors $1/\tau_s(T)\sim T^2/\tau_p(T)$. In the 

3730: important case

3731: of scattering by charged impurities ($\tau_p\sim T^{3/2}$) $1/\tau_s\sim 

3732: T^{1/2}$. 

3733: The magnetic field dependence of the EY spin relaxation  has not been

3734: systematically investigated. At low temperatures, where cyclotron 

3735: effects become important, 

3736: one needs to average over cyclotron trajectories on the Fermi 

3737: surface need to obtain $1/\tau_s$. 

3738: We expect that such averaging leads, in general, to an 

3739: increase in $1/T_1$, especially in systems where spin hot 

3740: spots are important (see Sec.~\ref{sec:IIIC}).

3741: 

3742: 

3743: \subsubsection{\label{sec:IIIB2} D'yakonov-Perel' mechanism}

3744: %---------------------------------------------------------------------------

3745: 

3746: 

3747: An efficient mechanism of spin relaxation due to spin-orbit coupling

3748: in systems lacking inversion symmetry 

3749: was found by \textcite{Dyakonov1972:SPSS}. Without

3750: inversion symmetry the momentum states of the spin up and spin down electrons 

3751: are not degenerate: $E_{{\bf k}\uparrow}\ne E_{{\bf k}\downarrow}$.  

3752: Kramer's theorem still dictates that $E_{{\bf k}\uparrow}= E_{-{\bf 

3753: k}\downarrow}$. Most prominent examples of materials without inversion

3754: symmetry 

3755: come from groups

3756: III-V (such as GaAs) and II-VI (ZnSe) semiconductors, where 

3757: inversion symmetry is broken by the presence of two distinct atoms in the

3758: Bravais lattice. Elemental semiconductors like Si possess inversion symmetry

3759: in the bulk, so the DP mechanism does not apply to them. In heterostructures

3760: the symmetry is broken by the presence of asymmetric confining potentials.  

3761: 

3762: Spin splittings 

3763: induced by  

3764: inversion asymmetry 

3765: can be described by introducing an

3766: intrinsic ${\bf k}$-dependent magnetic field ${\bf B}_{i}({\bf k})$

3767: around which electron spins 

3768: precess with Larmor frequency ${\bf \Omega}({\bf k})=(e/m) {\bf B}_i({\bf k})$.

3769: The intrinsic field derives from the spin-orbit coupling in the band-structure.

3770: The corresponding Hamiltonian term describing the precession of

3771: electrons in the conduction band is

3772: \begin{equation} \label{eq:HDP}

3773: H({\bf{k}})=\frac{1}{2}\hbar {\bf \hat{\sigma}} \cdot{\bf \Omega}(\bf{k}),

3774: \end{equation}

3775: where ${\bf \hat{\sigma}}$ are the Pauli matrices. Momentum-dependent 

3776: spin precession 

3777: described

3778: by $H$, together with momentum scattering 

3779: characterized by momentum 

3780: relaxation time $\tau_p$,\footnote{In the qualitative reasonings below we use 

3781: $\tau_p$

3782: instead of the effective correlation time $\tilde{\tau}$ for ${\bf \Omega}$ 

3783: during momentum 

3784: scattering; $\tilde\tau$ is defined later, in Eq.~(\ref{eq:relax:taul}).}

3785: leads to spin dephasing. While the microscopic expression for 

3786: $\bf \Omega({\bf k})$

3787: needs to be obtained from the band structure, treating the effects of 

3788: inversion asymmetry by introducing intrinsic precession helps to give a 

3789: qualitative understanding

3790: of spin dephasing. It is important to 

3791: note, however, that the analogy with real

3792: Larmor precession is not complete. An applied magnetic field induces 

3793: a macroscopic spin polarization and magnetization, while $H$ of 

3794: Eq.~(\ref{eq:HDP})

3795: produces an equal number of spin up and spin down states.  

3796: 

3797: Two limiting 

3798: cases can be considered: (i) $\tau_p\Omega_{av}\agt 1$ and (ii) 

3799: $\tau_p\Omega_{av}\alt 1$, 

3800: where $\Omega_{av}$ is an average magnitude of the intrinsic 

3801: Larmor frequency $\Omega({\bf k})$ over the actual momentum

3802: distribution. Case (i) corresponds to the situation in which

3803: individual electron 

3804: spins precess a full cycle before

3805: being scattered to another momentum state. The total spin in this regime 

3806: initially dephases 

3807: reversibly due to the anisotropy in ${\bf \Omega}({\bf k})$. The spin dephasing 

3808: rate,\footnote{The reversible decay need not be exponential.}

3809: which depends on the

3810: distribution of values of ${\bf \Omega}({\bf k})$, is in general

3811: proportional to the width $\Delta \Omega$ of the distribution: 

3812: $1/\tau_s \approx \Delta \Omega$. The spin is irreversibly lost after time 

3813: $\tau_p$, when randomizing scattering takes place.

3814: 

3815: Case (ii) is what is usually meant by the D'yakonov-Perel' 

3816: mechanism. This regime can be viewed from the point of view of individual 

3817: electrons as a 

3818: spin precession about fluctuating magnetic fields, whose magnitude and 

3819: direction change randomly with the average time step of $\tau_p$. 

3820: The electron spin rotates about 

3821: the intrinsic field at an angle $\delta \phi = \Omega_{av}\tau_p$, 

3822: before experiencing 

3823: another field

3824: and starting to rotate with a different speed and in a different direction. 

3825: As a result, the spin phase follows a random walk: after time $t$, which 

3826: amounts to

3827: $t/\tau_p$ steps of the random walk, the phase

3828: progresses by $\phi(t) \approx \delta\phi\sqrt{t/\tau_p}$. Defining $\tau_s$ as 

3829: the time

3830: at which $\phi(t)=1$, the usual motional narrowing result is obtained:

3831: $1/\tau_s=\Omega_{av}^2 \tau_p$ (see footnote 70).

3832: The faster the momentum relaxation, 

3833: the slower 

3834: the

3835: spin dephasing. The difference between cases (i) and (ii) is that in 

3836: case (ii)

3837: the electron spins form an ensemble that directly samples the 

3838: distribution of 

3839: $\Omega({\bf k})$, while

3840: in case (ii) it is the distribution of the {\it sums} 

3841: of the intrinsic 

3842: Larmor frequencies (the 

3843: total 

3844: phase of a spin after many steps consists of a sum of randomly selected 

3845: frequencies

3846: multiplied by $\tau_p$), which, according to the central limit theorem, has a 

3847: significantly reduced variance. Both limits (i) and (ii) 

3848: and the transition between

3849: them have been experimentally demonstrated in n-GaAs/AlGaAs quantum wells by 

3850: observing temporal spin oscillations over a large range of temperatures 

3851: (and thus $\tau_p$)

3852:  \cite{Brand2002:PRL}.

3853: 

3854: A more rigorous expression for $\tau_s$ in regime (ii) 

3855: has been obtained by solving the kinetic rate equation for a spin-dependent 

3856: density matrix 

3857: \cite{Dyakonov1971:SPJETP,Dyakonov1972:SPSS}.

3858: If the band structure is isotropic and scattering is both elastic and 

3859: isotropic, 

3860: evolution 

3861: of the z-component of spin ${\bf s}$

3862: is  

3863: \cite{Pikus:1984}

3864: \begin{equation}\label{eq:relax:sz}

3865: \dot{s}_z =

3866: -\tilde{\tau_l}\left [s_z \overline{\left ( \Omega^2 -\Omega_z^2 \right )}

3867: -s_x \overline{\Omega_x\Omega_z}-s_y \overline{\Omega_y\Omega_z}\right ],

3868: \end{equation}

3869: where the bar denotes averaging over directions of $\bf k$. Analogous 

3870: expressions

3871: for $\dot{s}_x$ and $\dot{s}_y$ can be written by index permutation. The 

3872: effective

3873: momentum scattering time is introduced as

3874: \begin{equation}\label{eq:relax:taul}

3875: 1/\tilde{\tau_l}=\int_{-1}^1 W(\theta)\left [1-P_l\left (\cos\theta\right 

3876: )\right ] d\cos\theta,

3877: \end{equation}

3878: where $W(\theta)$ is the rate of momentum scattering through angle $\theta$ at 

3879: energy

3880: $E_{\bf k}$, and $P_l$ is the Legendre polynomial whose order $l$ 

3881: is the power 

3882: of $\bf k$ in

3883: $\Omega({\bf k})$. [It is assumed that $\Omega({\bf k})\sim k^l$ in 

3884: Eq.~(\ref{eq:relax:sz}).] 

3885:  In two dimensions $P_l(\cos\theta)$ is replaced by $\cos(l\theta)$

3886: in Eq.~(\ref{eq:relax:taul}) for the $l$th polar harmonic of 

3887: $\bf \Omega({\bf k})$

3888: ~\cite{Pikus1995:PRB}.

3889: Since it is useful to express the results in terms of the known momentum 

3890: relaxation

3891: times\footnote{In fact,  normal (not umklapp) electron-electron collisions 

3892: should also be included in the effective spin randomization time 

3893: $\tilde{\tau}$, though they 

3894: do not

3895: contribute to the momentum relaxation time which appears in the measured 

3896: mobility~\cite{Glazov2002:JETPL,Glazov2003:P}.}

3897:  $\tau_p=\tilde{\tau}_1$, one defines 

3898: \footnote{ \textcite{Pikus:1984} 

3899: initially define $\gamma_l$ as here, but 

3900: later evaluate it, inconsistently, as the inverse $\gamma_l \rightarrow 

3901: \gamma_l^{-1}$.} 

3902: $\gamma_l=\tau_p/\tilde{\tau_l}$ to measure the effectiveness of momentum 

3903: scattering

3904: in randomizing Larmor frequencies; $\tilde{\tau}_l$ accounts for the relative 

3905: angle between 

3906: ${\bf \Omega}$ before and after scattering. Generally $\gamma_l > 1$ for $l>1$, 

3907: that is, momentum scattering is more effective in randomizing spins than 

3908: in randomizing momentum. 

3909: 

3910: 

3911: Comparing with the Bloch-Torrey equations 

3912: (\ref{eq:relax:bloch1})--(\ref{eq:relax:bloch3}), for $\bf{B}=0$ and no spin 

3913: diffusion, we see that spin decay is described by the tensor 

3914: $1/\tau_{s,ij}$ 

3915: (here $i$ and $j$ are the Cartesian coordinates) whose diagonal 

3916: $1/\tau_{s,ii}$

3917: and off-diagonal $1/\tau_{s,ij}$, for $i\ne j$, terms

3918: are

3919: \begin{equation} \label{eq:relax:tauii}

3920: 1/\tau_{s,ii}=\gamma_l^{-1}\tau_p(\overline{\Omega^2}-

3921: \overline{\Omega_i^2}),\,\,\,

3922: 1/\tau_{s,ij}=-\gamma_l^{-1}\tau_p\overline{\Omega_i\Omega_j}.

3923: \end{equation}

3924: In general, spin dephasing depends on the spin direction and on the 

3925: dephasing rates of 

3926: the perpendicular

3927: spin components. Equations (\ref{eq:relax:tauii}) are valid for small magnetic 

3928: fields, 

3929: satisfying $\Omega_0 \tau_p \ll 1$, where $\Omega_0$ is the Larmor frequency of 

3930: the

3931: external field. 

3932: 

3933: The most important difference between the EY and the DP mechanism is their 

3934: opposite 

3935: dependence on $\tau_p$. While increased scattering intensity makes the EY 

3936: mechanism 

3937: more effective, it decreases the effectiveness of the

3938: the DP processes. In a sense the two 

3939: mechanisms are 

3940: similar to collision broadening and motional narrowing in 

3941: NMR \cite{Slichter:1989}. Indeed, in the EY 

3942: process

3943: the precession frequency is conserved between collisions and the loss of

3944: phase occurs only in 

3945: the short time {\it during} collision. The more 

3946: collisions there are,

3947: the greater is the loss of phase memory, in analogy with collision 

3948: broadening of 

3949: spectral

3950: lines. On the other hand, in DP spin dephasing, spin phases are randomized 

3951: {\it between} collisions, since electrons precess with different frequencies 

3952: depending

3953: on their momenta. Spin-independent collisions with impurities or phonons do not 

3954: lead

3955: to phase randomization during the collision itself, but help to establish the 

3956: random-walk-like evolution of the phase, leading to motional narrowing.

3957: While these two mechanisms coexist in systems lacking inversion symmetry,

3958: their relative strength depends on many factors. Perhaps the most robust trend

3959: is that the DP mechanism becomes more important with increasing band

3960: gap and increasing temperature. 

3961: 

3962: In the rest of the section we apply Eq.~(\ref{eq:relax:tauii})

3963: to the study of spin dephasing in bulk and two-dimensional III-V semiconductor 

3964: heterostructures.

3965: 

3966: \paragraph{\label{sec:IIIB2a} Bulk III-V semiconductors.}

3967: %--------------------------------------------------------------------------- 

3968: 

3969: In bulk III-V semiconductors the intrinsic Larmor frequency vector of

3970: Eq.~(\ref{eq:HDP}) 

3971: due

3972: to the

3973: lack of inversion symmetry is \cite{Dyakonov1971:SPJETP}

3974: \begin{equation} \label{eq:relax:dress}

3975: {\bf \Omega}({\bf{k}})=\alpha\hbar^2(2m_c^3E_g)^{-1/2}\bf{\kappa},

3976: \end{equation}

3977: where 

3978: \begin{eqnarray}

3979: {\bf\kappa} = \left [k_x(k_y^2-k_z^2),k_y(k_z^2-k_x^2),

3980: k_z(k_x^2-k_y^2)\right ].

3981: \end{eqnarray}

3982: Here $k_i$ are the lattice wave-vector components along the crystal principal 

3983: axes.

3984: The material-specific parameters are the band gap $E_g$ and the 

3985: conduction electron mass

3986: $m_c$; $\alpha$ is a dimensionless parameter specifying

3987: the strength of the spin-orbit interaction.

3988: The spin splitting described by Eq.~(\ref{eq:relax:dress})

3989: is proportional to the cube of the lattice momentum, as was first found by 

3990: \textcite{Dresselhaus1955:PR}. For GaAs $\alpha\approx 

3991: 0.07$~\cite{Marushchak1984:SPSS}. Spin splitting of conduction and

3992: heavy and light holes in GaAs quantum wells, due to bulk inversion

3993: asymmetry was calculated by \textcite{Rashba1988:PLA}.

3994: 

3995: Substituting Eq.~(\ref{eq:relax:dress}) for ${\bf\Omega}$ in 

3996: Eq.~(\ref{eq:relax:tauii}), 

3997: and using $\overline{\kappa_i\kappa_j}=(4/105)k^6\delta_{ij}$, we obtain the 

3998: expected

3999: result that the off-diagonal elements of $1/\tau_{s,ij}$ vanish for cubic 

4000: systems

4001: and the diagonal elements are all equal to~\cite{Pikus:1984} 

4002: \begin{equation}\label{eq:relax:DP}

4003: 1/\tau_s(E_{\bf k})=

4004: \frac{32}{105}\gamma_3^{-1}\tau_p(E_{\bf k}) \alpha^2\frac{E_{\bf 

4005: k}^3}{\hbar^2 E_g}.

4006: \end{equation}

4007: The above expression describes DP spin dephasing of degenerate ($E_{\bf k}=E_F$) or 

4008: hot\footnote{This is strictly true only if the spin relaxation of 

4009: the hot electrons 

4010: is faster

4011: than energy relaxation by optical phonon emission, which is rarely the case. 

4012: One has to consider either the spin relaxation at different energy levels 

4013: during

4014: the cascade process of optical phonon emission or, if the optical 

4015: phonon emission is

4016: particularly fast, spin relaxation only during the final stages of energy 

4017: relaxation

4018: by acoustic phonon emission (see \textcite{Pikus:1984}.} electrons

4019: in bulk III-V semiconductors. For impurity scattering $\gamma_3\approx 

4020: 6$, for

4021: acoustic phonons $\gamma_3\approx 1$, while for optical polar phonons 

4022: $\gamma_3\approx 41/6$.

4023: The temperature dependence of $1/\tau_s$ comes from the temperature dependence 

4024: of 

4025: $\tau_p$. For the important case of charged impurity scattering ($\tau_p\sim 

4026: T^{3/2}$),

4027: $1/\tau_s\sim T^{3/2}$. 

4028: 

4029: Compared to the EY expression, Eq.~(\ref{eq:relax:chazalviel}),

4030: the DP spin dephasing increases much faster with 

4031: increasing

4032: electron energy and is expected to be dominant at large donor doping levels and

4033: at high temperatures. The EY mechanism can be dominant in small-band-gap and 

4034: large spin-orbit-splitting materials. 

4035: The two mechanisms can also be easily distinguished by 

4036: their

4037: opposite dependence on momentum relaxation. Contrary to the EY mechanism, 

4038: greater impurity density will decrease the importance of the DP processes. 

4039: The  most frequently used ways

4040: to distinguish between various methods of spin relaxation are comparing the

4041: electron density (through the variation of the Fermi energy) 

4042: and the temperature dependences of $1/\tau_s$ with the

4043: theoretical estimates. Since the prefactors may vary with different scattering

4044: mechanisms, it is best to deduce $\tau_p(E_{\bf k})$  and $\tau_p(T)$ 

4045: from mobility measurements and use Eqs.~(\ref{eq:relax:chazalviel}) and 

4046: (\ref{eq:relax:DP})

4047: or the equations given below for $1/\tau_s(T)$.

4048: 

4049: Another interesting distinction between the two mechanisms is revealed 

4050: by the 

4051: dependence

4052: of their spin diffusion length $L_s=\sqrt{D\tau_s}$ on momentum scattering. 

4053: Since $D\sim \tau_p$, for EY $L_s\sim \tau_p$, while for DP $L_s$ does 

4054: not depend on 

4055: the momentum

4056: scattering time and for a degenerate electron system should be  a constant 

4057: independent of

4058: $T$, of the order of $v_F/\Omega_{av}$.  

4059: We do not know of an experimental verification of this distinction.

4060: 

4061: If the electrons obey nondegenerate statistics, which is the usual case of 

4062: p-doped 

4063: materials, thermal averaging over the Boltzmann distribution 

4064: gives~\cite{Pikus:1984}. 

4065: \begin{equation}\label{eq:relax:DPt}

4066: 1/\tau_s=Q \tau_m \alpha^2 \frac{(k_B T)^3}{\hbar^2 E_g},

4067: \end{equation} 

4068: where $\tau_m=\langle \tau_p(E_{\bf k}) E_{\bf k}\rangle/\langle E_{\bf k} 

4069: \rangle$.

4070: The coefficient $Q$, which is of order 1, is 

4071: \begin{equation}

4072: Q=\frac{16}{35}\gamma_3^{-1} \left ( \nu+\frac{7}{2} \right )

4073: \left (\nu+\frac{5}{2} \right ),

4074: \end{equation}

4075: where the power law $\tau_p\sim E^{\nu}_{\bf k}$ is assumed for momentum 

4076: relaxation time.

4077: For scattering by ionized impurities $Q\approx 1.5$, while scattering by polar 

4078: optical

4079: or piezoelectric phonons gives $Q\approx 0.8$, and scattering by acoustic 

4080: phonons 

4081: (deformation potential) $Q \approx 2.7$~\cite{Pikus:1984}. The temperature 

4082: behavior of

4083: DP spin dephasing in nondegenerate samples is 

4084: $1/\tau_s\sim T^3\tau_m(T)$. 

4085: For 

4086: scattering

4087: by charged impurities $1/\tau_s\sim T^{9/2}$.

4088: 

4089: Application of longitudinal (to the initial spin direction) 

4090: magnetic field suppresses the DP mechanism~\cite{Pikus:1984} for two 

4091: reasons: (i) The $B$-field 

4092: suppresses

4093: precession along the transverse intrinsic 

4094: fluctuating fields when $\Omega_L\tau_p > 1$, where $\Omega_L$ is 

4095: the Larmor precession due to $B$. (ii)  $\Omega_{\bf k}$ 

4096: is orbitally averaged,

4097: which 

4098: has a similar effect to averaging by random scattering,  when 

4099: $\Omega_c\tau_p>1$, where $\Omega_c$ is the cyclotron frequency. 

4100: Since for conduction electrons  $m_c \ll m_e$, it follows that

4101: $\Omega_c \gg \Omega_L$, the orbital motion induced by $B$ is the cause for 

4102: suppression

4103: of spin relaxation in semiconductors.    

4104: %

4105: %Theoretically spin relaxation and spin dephasing in bulk n-type GaAs was 

4106: %investigated

4107: %in \cite{Wu2000:PSS,Wu2001:JS} by solving numerically kinetic equations. Only 

4108: %the DP effect was studied. Acoustic and longitudinal phonon and impurity 

4109: %scattering 

4110: %were included. 

4111: %The main findings are: (i) Spin dephasing increases with increasing impurity 

4112: %density up to a critical density after which the dephasing time decreases. 

4113: %(ii) As a function of temperature, spin dephasing time increases with 

4114: %increasing 

4115: %temperature at small impurity content, but decreases with increasing 

4116: %temperature

4117: %at large impurity density (the latter point is in contrast to previous 

4118: %calculations).

4119: %It is also found that spin dephasing time increases with decreasing electron

4120: %density. (iii) Spin dephasing time decreases with increasing electron density,

4121: %and (iv) spin dephasing time increases with increasing magnetic field (only 

4122: %spin,

4123: %no orbital effects included).

4124: 

4125: 

4126: \paragraph{\label{sec:IIIB2b} Two-dimensional III-V semiconductor systems.}

4127: %----------------------------------------------------------------------

4128: 

4129: In two-dimensional III-V semiconductor systems (quantum wells and 

4130: heterostructures) there are 

4131: two distinct Hamiltonian terms that contribute

4132: to DP spin dephasing: the bulk inversion asymmetry term $H_{\rm BIA}$  

4133: and the structure inversion asymmetry term, $H_{\rm SIA}$, which appears 

4134: only in 

4135: asymmetric systems. Both $H_{\rm BIA}$ and

4136: $H_{\rm SIA}$ lead to spin splitting of the conduction band linear in 

4137: ${\bf k}$. The two 

4138: terms, however, predict a different dependence of $\tau_s$ on the 

4139: quantum-well 

4140: orientation relative to the principal axes. Figure \ref{fig:omega} shows 

4141: the vector field patterns of the intrinsic magnetic fields for both

4142: bulk and spin inversion asymmetry.

4143: 

4144: The bulk inversion asymmetry term comes from the bulk Dresselhaus 

4145: spin splitting, Eq.~(\ref{eq:relax:dress}). 

4146: Treating wave vectors ${\bf k}$ in Eq.~(\ref{eq:relax:dress}) as operators

4147: $\hat{\bf k}=-i\nabla$, and evaluating $\bf\Omega$ as the expectation value in 

4148: the 

4149: confined states, leads to momentum quantization along the confinement  

4150: unit vector

4151: ${\bf n}$ of the quantum well (QW). In the following, ${\bf k}$ denotes the 

4152: wave vector for a Bloch state propagating in the plane, and 

4153: ${k_n^2}\equiv\langle (\hat{\bf k}\cdot {\bf n})^2\rangle$ 

4154: denotes the expectation value 

4155: of the square 

4156: of the component of the wave-number 

4157: operator 

4158: normal to the plane 

4159: in the lowest-subband state. For a rectangular QW of width $a$, 

4160: $k_n^2=(\pi/a)^2$.

4161: For a triangular well of confining potential $eEz$, 

4162: $k_n^2 \approx 0.7794 (2m_e E/\hbar^2)^{2/3}$ [see, for example,

4163: \textcite{deSousa2003:PRBc}].

4164: Quantum averaging of $\kappa$ can be done using the formula

4165: \begin{equation}

4166: \langle \hat{k}_i \hat{k}_j \hat{k}_l \rangle 

4167: = k_n^2 \left (k_i n_j n_l + k_j n_l n_i + k_l n_i n_j  \right ) 

4168: + k_i k_j k_l. 

4169: \end{equation}

4170: This readily gives~\cite{Dyakonov1986:SPS}

4171: \begin{equation}\label{eq:relax:kappa2d}

4172: \kappa_x= k_n^2  \left [2 n_x\left (n_y k_y -n_z k_z \right ) + 

4173: k_x \left ( n_y^2 -n_z^2 \right ) \right ],

4174: \end{equation}

4175: and similarly for $\kappa_y$ and $\kappa_z$ by index permutation. 

4176: Terms cubic in $k$ were omitted from the above equation, assuming that for 

4177: narrow QW's 

4178: $k^2\ll k_n^2$. The explicit knowledge of $\kappa$ is useful in qualitative 

4179: analysis

4180: of spin dephasing for particular orientations of QW's.

4181: 

4182: 

4183: The spin dephasing tensor $1/\tau_{s,ij}$, defined in 

4184: Eq.~(\ref{eq:relax:tauii}), 

4185: is readily evaluated using 

4186: Eqs.~(\ref{eq:relax:dress}) and 

4187: (\ref{eq:relax:kappa2d})\footnote{Averaging over the directions

4188: of $\bf k$ in a plane perpendicular to $\bf n$ can be performed by using 

4189: $\overline{k_ik_j}=

4190: (k^2/2)(\delta_{ij}-n_in_j)$.}

4191: \begin{eqnarray}\label{eq:relax:nu}

4192: 1/\tau_{s,ij}=\left (\delta_{ij}{\rm Tr} \hat{\nu} 

4193: - \nu_{ij}  \right )/\tau_{s}^0(E_{\bf k}),  

4194: \end{eqnarray}

4195: where 

4196: \begin{equation}\label{eq:relax:DK}

4197: \frac{1}{\tau_{s}^0(E_{\bf k})}=\frac{\alpha^2 \hbar^2 

4198: \left (k_n^2\right) ^2}{2 m_c^2 E_g} 

4199: E_{\bf k}\tau_p(E_{\bf k}).

4200: \end{equation}

4201: The tensor $\hat{\nu}$ depends on the orientation of $\bf n$ with

4202: respect to the principal crystal 

4203: axes~\cite{Dyakonov1986:SPS}\footnote{A trivial

4204: typo in $\nu_{xx}$ is corrected.} 

4205: \begin{eqnarray}

4206: \nu_{xx}&=&4n_x^2(n_y^2+n_z^2)-(n_y^2-n_z^2)^2(9n_x^2-1), \\

4207: \nu_{xy}&=&n_xn_y\left[9(n_x^2-n_z^2)(n_y^2-n_z^2)-2(1-n_z^2)\right],

4208: \end{eqnarray}

4209: and analogously for other components. 

4210: 

4211: \begin{figure}

4212: \centerline{\psfig{file=zutic_fig15.eps,width=1\linewidth,angle=0}}

4213: \caption{Vector fields ${\bf\Omega}({\bf k})\sim \kappa({\bf k})$ 

4214: on the Fermi surface (here a circle) 

4215: for the structure 

4216: inversion asymmetry

4217: (SIA)

4218: and bulk 

4219: inversion asymmetry (BIA). 

4220: Since 

4221: ${\bf\Omega}({\bf k})$

4222: is also the spin quantization axis, the vector pattern is also the 

4223: pattern of the spin on 

4224: the Fermi surface. As 

4225: the opposite spins have different energies, the Fermi circle becomes 

4226: two concentric circles with opposite spins. This is shown here only for 

4227: the SIA case,

4228: but the analogy extends to all examples. The field for BIA [110] lies 

4229: perpendicular to

4230: the plane, with the magnitude varying along the Fermi surface. All other 

4231: cases have 

4232: constant fields lying in the plane.  

4233: }

4234: \label{fig:omega}

4235: \end{figure}

4236: 

4237: We follow \textcite{Dyakonov1986:SPS} in discussing the three important 

4238: cases of 

4239: [001], [111], and [110] quantum wells. For [001], 

4240: \begin{equation}

4241: {\bf\Omega}({\bf k})\sim\kappa=k_n^2 (-k_x,k_y,0).

4242: \end{equation}

4243: While the magnitude of ${\bf\Omega}({\bf k})$ is  

4244: constant over the Fermi surface,

4245: the directions follow a ``breathing'' pattern as shown in Fig.~\ref{fig:omega}. 

4246: The spin relaxation times follow from 

4247: Eq.~(\ref{eq:relax:DK}): 

4248: $1/\tau_{s,xx}=1/\tau_{s,yy}=1/2\tau_{zz}=1/\tau_{s}^0$.

4249: Defining $1/\tau_{s\parallel}$ and $1/\tau_{s,\perp}$ as spin dephasing 

4250: rates of 

4251: spins parallel and perpendicular to the plane, one obtains

4252: \begin{equation}

4253: 1/\tau_{s,\parallel}=1/2\tau_{s,\perp}=1/\tau_{s}^0.

4254: \end{equation}

4255: As expected for the case of the in-plane field, the lifetime of a spin 

4256: parallel to 

4257: the plane is twice that of the a spin perpendicular to the plane.

4258: 

4259: For [111] QW's,

4260: \begin{equation}

4261: {\bf\Omega}({\bf k})\sim\kappa = 2/\sqrt{3} k_n^2 ({\bf k}\times {\bf n}).

4262: \end{equation}

4263: The intrinsic magnetic field lies in the plane, having a constant magnitude 

4264: (refer to Fig.~\ref{fig:omega}). 

4265: Spin relaxation rates are now 

4266: $1/\tau_{s,ii}=16/9\tau_{s}^0$ and $1/\tau_{s,i\ne j}=4/9\tau_{s}^0$. 

4267: By diagonalizing $1/\tau_{ij}$  we obtain

4268: \begin{equation}

4269: 1/\tau_{s\parallel}=1/2\tau_{s,\perp}=4/3\tau_{s}^0.

4270: \end{equation}

4271: As for the [001] case, a perpendicular spin dephases twice as fast as a 

4272: parallel one, since 

4273: ${\bf\Omega}({\bf k})$ lies in the plane.

4274: 

4275: The most interesting case is the [110] orientation for which 

4276: $1/\tau_{s,xx}=1/\tau_{s,yy}=1/2\tau_{s,zz}=-1/\tau_{s,xy}=1/8\tau_{s}^0$. 

4277: Other

4278: off-diagonal components vanish. Diagonalizing the tensor  gives

4279: \begin{equation}

4280: 1/\tau_{s,\parallel}=1/4\tau_{s}^0,\,\,\,1/\tau_{s,\perp}=0.

4281: \end{equation}

4282: The perpendicular spin does not dephase. This is due to the fact that $\kappa$, 

4283: unlike the previous cases, is always normal to the plane 

4284: (see Fig.~\ref{fig:omega}), 

4285: and thus cannot

4286: affect the precession of the perpendicular spin. Indeed, 

4287: \begin{equation}

4288: {\bf\Omega}({\bf k})\sim\kappa = k_n^2 \left (k_x/2) (-1,-1, 0 \right ),

4289: \end{equation}

4290: where it is used that ${\bf k}\cdot {\bf n}=0$. Spin dephasing in [110] QW's can 

4291: still be 

4292: due to the cubic terms in $k$ left out of Eq.~(\ref{eq:relax:kappa2d}) or to 

4293: other spin relaxation mechanisms. Note that the magnitude of 

4294: ${\bf\Omega}({\bf k})$

4295: changes along the Fermi surface. 

4296: Electrons moving along [001] experience

4297: little spin dephasing.

4298: 

4299: The structure inversion asymmetry term arises from the Bychkov-Rashba spin 

4300: splitting ~\cite{Bychkov1984:JETPL,Bychkov1984:JPC,Rashba1960:SPSS} 

4301: occurring in asymmetric QW's 

4302: or in deformed bulk systems. The corresponding Hamiltonian is that

4303: of Eq.~(\ref{eq:HDP}), with the precession vector

4304: \begin{equation} \label{eq:relax:BR}

4305: {\bf\Omega}({\bf k})=2\alpha_{BR} ({\bf k}\times {\bf n}). 

4306: \end{equation}

4307: Here $\alpha_{BR}$ is a parameter depending on spin-orbit coupling and the 

4308: asymmetry

4309: of the confining electrostatic potentials arising from the growth 

4310: process of the

4311: heterostructure. The splitting can also arise in nominally symmetric 

4312: heterostructures

4313: with fluctuations in doping density \cite{Sherman2003:APL}.  

4314: The Bychkov-Rashba field always lies

4315: in the plane, having a constant magnitude. As for the bulk inversion 

4316: asymmetry case, the

4317: structure inversion asymmetry leads to a splitting of the Fermi surface, 

4318: according

4319: to the direction of the spin pattern---parallel or antiparallel to 

4320: $\bf{\Omega}({\bf k})$,

4321: as shown in Fig.~\ref{fig:omega}. Perhaps the most appealing fact about 

4322: structure

4323: inversion asymmetry is that $\alpha_{BR}$ can be tuned electrostatically, 

4324: potentially

4325: providing an effective spin precession control without the need for magnetic 

4326: fields 

4327: \cite{Rashba2003:PRL,Levitov2003:PRB}.

4328: This has led to one of the pioneering spintronic proposals 

4329: by \textcite{Datta1990:APL} (see Sec.~\ref{sec:IVE1}). 

4330: Note that for the [111] orientation the bulk

4331: and structure inversion asymmetry terms have the same form. 

4332: 

4333: 

4334: Using the same procedure as for bulk inversion asymmetry, 

4335: we describe the 

4336: spin relaxation rate by Eq.~(\ref{eq:relax:nu}) as 

4337: \begin{equation}

4338: 1/\tau_{s}^0=4\alpha_{BR}^2\frac{m_c}{\hbar^2} E_{\bf k} \tau_p

4339: \end{equation}

4340: and

4341: \begin{equation}\label{eq:relax:nusia}

4342: \nu_{ij}=1-n_in_j.

4343: \end{equation}

4344: Since the intrinsic precession vector $\sim {\bf \Omega}({\bf k})$ for the 

4345: structure inversion

4346: asymmetry always lies in the plane, 

4347: a perpendicular spin should dephase 

4348: twice as fast as a spin in the plane. Indeed,

4349: by diagonalizing $1/\tau_{s,ij}$ one finds that 

4350: \begin{equation}

4351: 1/\tau_{s,\parallel}=1/2\tau_{s,\perp}=1/\tau_s^0

4352: \end{equation}

4353: holds for all QW orientations ${\bf n}$. 

4354: This interesting fact qualitatively

4355: distinguishes structure from bulk inversion asymmetry and can be used in 

4356: assessing the 

4357: relative importance of

4358: the Dresselhaus and Rashba spin splittings in III-V heterostructure systems. 

4359: If bulk and structure inversion asymmetry are of similar importance, 

4360: the interference terms from the cross 

4361: product $\overline{\Omega_{\rm BIA} \Omega_{\rm SIA}}$ can lead to spin 

4362: dephasing 

4363: anisotropies within the plane, as was shown for [001] 

4364: QW's  

4365: in~ \cite{Averkiev1999:PRB,Kainz2003:PRB}. This plane anisotropy can be 

4366: easily seen by adding

4367: the corresponding vector fields in Fig.~\ref{fig:omega}. Another interesting

4368: feature of bulk and structure inversion asymmetry fields is that injection

4369: of electrons along a quasi-one-dimensional channel can lead to large relaxation

4370: times for spins oriented along $\bf{\Omega}({\bf k})$, where ${\bf k}$ is

4371: the wave vector for the states in the channel \cite{Hammar2002:PRL}.

4372: 

4373: Model spin dephasing calculations based on structure inversion asymmetry 

4374: were carried out 

4375: by \textcite{Pareek2002:PRB}. 

4376: Calculations of $\tau_s$ based on the DP mechanism, 

4377: with structural asymmetry due to doping fluctuations in the heterostructure 

4378: interface were performed by \textcite{Sherman2003:APL}.

4379: 

4380: Research on spin inversion asymmetry is largely motivated by Datta-Das

4381: spin field-effect transistor proposal (see Secs.~\ref{sec:IA} and

4382: \ref{sec:IVE1}) in which $\alpha_{BR}$ is tailored by a gate.

4383: This tailoring, however, has been controversial and the microscopic

4384: origin of the Bychkov-Rashba Hamiltonian, and thus the interpretation

4385: of experimental results on splitting in semiconductor heterostructures, 

4386: has been debated. The Bychkov-Rashba Hamiltonian is often interpreted 

4387: as arising from the electric field of the confining potential, assisted by

4388: external bias, which acts on a moving electron in a transverse direction.

4389: The relativistic transformation then gives rise to a magnetic field

4390: (spin-orbit coupling) acting on the electron spin. The parameter 

4391: $\alpha_{BR}$ is then assumed to be directly proportional to the 

4392: confining electrical field. This is in general wrong, since the average

4393: electric force acting on a confined particle of uniform effective mass

4394: is zero.

4395: 

4396: The asymmetry that gives rise to structure inversion asymmetry is the

4397: asymmetry in the band structure (including spin-orbit coupling)

4398: parameters of a heterostructure, such as the effective mass, or the

4399: asymmetry in the penetration of the electron wave function into the

4400: barriers \cite{Silva1997:PRB}. The difficulty in understanding the 

4401: influence of the external gates is caused by the lack of the

4402: understanding of the influence of the gate field on the asymmetry

4403: of the well. For a clear qualitative explanation of the involved

4404: physics see \textcite{Pfeffer1999:PRB} and \textcite{Zawadzki2001:PRB}.

4405: Band-structure {\bf $k \cdot p$} calculations of $\alpha_{BR}$ for

4406: quantum wells in GaAs/AlGaAs heterostructures can be found in 

4407: \textcite{Pfeffer1995:PRB}, \textcite{Pfeffer1997:PRB},

4408: \textcite{Wissinger1998:PRB}, and \textcite{Kainz2003:PRB};

4409: a calculation for InGaAs/InP quantum wells is reported in 

4410: \textcite{Engels1997:PRB} and \textcite{Schapers1998:JAP},

4411: in InSb/InAlSb asymmetric quantum well it can be found in 

4412: \textcite{Pfeffer2003:PRB}, and in p-InAs metal-oxide-semiconductor

4413: field-effect transistor channel in \textcite{Lamari2003:PRB}.

4414: Adding to the controversy, \textcite{Majewski:2003} have recently

4415: calculated the structure inversion asymmetry by local density functional

4416: methods and concluded that the induced spin splitting arises from

4417: microscopic electric fields in asymmetric atomic arrangements at the 

4418: interfaces, so that a large Bychkov-Rashba term can be present in 

4419: otherwise symmetric quantum wells with no common atom.

4420: 

4421: Interpretation of experimental data on structure inversion asymmetry is

4422: difficult, especially at determining the zero magnetic field spin splitting

4423: (usually seen in Shubnikov- de Haas oscillations), which is masked by

4424: Zeeman splitting at finite fields. In addition, the splittings are small,

4425: typically less than 1 meV. The Bychkov-Rashba parameter was measured in

4426: GaSb/InAs/GaSb quantum wells 

4427: ($\alpha_{BR} \approx 0.9 \times 10^{-11}$ eV m for 75 $\AA$  

4428: thick well)

4429: by \textcite{Luo1990:PRB}; in InAlAs/InGaAs/InAlAs quantum wells (20 nm),

4430: where also the gate voltage is obtained: $\alpha_{BR}$ ranged from

4431: $10^{-11}$ eV m at the depleting voltage of -1 V, to $5 \times 10^{-12}$

4432: eV m at +1.5 V. Weak antilocalization studies of InAlAs/InGaAs/InAlAs

4433: quantum wells have recently been used to study electron density dependence

4434: of $\alpha_{BR}$ by \textcite{Koga2002:PRL}. Gate dependence of $\alpha_{BR}$

4435: was also measured in modulation-doped InP/InGaAs/InP quantum wells

4436: \cite{Engels1997:PRB,Schapers1998:JAP}. The observed values are several

4437: $10^{-12}$ eV m. On the other hand, there are experimental reports that

4438: either fail to observe the expected spin splitting due to 

4439: Bychkov-Rashba field, or interpret the splitting differently [see, for example,

4440: \textcite{Brosig1999:PRB} and \textcite{Rowe2001:PRB}].

4441: Furthermore, measurements of \textcite{Heida1998:PRB} show a constant

4442: $\alpha_{BR} \approx 0.6 \times 10^{-11}$ eV m, independent of gate

4443: voltage, in asymmetric AlSb/InAS/AlSb quantum wells, demonstrating that

4444: control of $\alpha_{BR}$ may be difficult. In order to unify the

4445: different views on what exactly the Bychkov-Rashba spin splitting means

4446: and how the spin splitting can be tuned with gate voltage, more

4447: experimental efforts need to be devoted to this interesting topic.

4448: 

4449: \subsubsection{\label{sec:IIIB3} Bir-Aronov-Pikus Mechanism} 

4450: %---------------------------------------------------------------------------

4451: 

4452: Spin relaxation of conduction electrons in p-doped semiconductors can also 

4453: proceed through scattering, accompanied by spin exchange, by holes, 

4454: as was first 

4455: shown by \textcite{Bir1976:SPJETP}. 

4456: 

4457: The exchange interaction between electrons and holes is governed by the 

4458: Hamiltonian

4459: \begin{equation}

4460: H= A{\bf S}\cdot {\bf J} \delta({\bf r}),

4461: \end{equation}

4462: where $A$ is proportional to the exchange integral between the conduction and 

4463: valence states,

4464: ${\bf J}$ is the angular momentum operator for holes, ${\bf S}$ is the electron 

4465: spin operator,

4466: and ${\bf r}$ is the relative position of electrons and holes.

4467: 

4468: The spin-flip scattering probability depends on the state of the holes 

4469: (degenerate or

4470: nondegenerate, bound on acceptors or free, fast or slow). We present below the 

4471: most 

4472: frequently used formulas when assessing the relevance of the BAP mechanism. 

4473: The formulas are 

4474: valid for 

4475: the usual cases of heavy holes $m_v \gg m_c$. For electron spin relaxation 

4476: due to exchange with 

4477: nondegenerate

4478: holes, 

4479: \begin{equation}\label{eq:relax:BAP1}

4480: \frac{1}{\tau_s}= \frac{2}{\tau_0}N_a a_B^3 \frac{v_{\bf k}}{v_B} \left 

4481: [\frac{p}{N_a}

4482: |\psi(0)|^4 + \frac{5}{3}\frac{N_a-p}{N_a} \right],

4483: \end{equation}

4484: where $a_B$ is the exciton Bohr radius $a_B=\hbar^2\epsilon/e^2m_c$,

4485: $p$ is the density of free holes, $\tau_0$ is an exchange splitting parameter:

4486: $\hbar/\tau_0=(3\pi/64)\Delta_{\rm ex}^2/E_B$ (with $E_B$ denoting the Bohr 

4487: exciton

4488: energy, $E_B=\hbar^2/2m_ca_B^2$ and $\Delta_{\rm ex}$ the exchange splitting 

4489: of the

4490: excitonic ground state), and $v_B=\hbar/m_ca_B$; $|\psi(0)|^2$ is Sommerfeld's 

4491: factor, which

4492: enhances the free hole contribution. 

4493: For an unscreened Coulomb potential

4494: \begin{equation}

4495: |\psi(0)|^2 = \frac{2\pi}{\kappa} \left 

4496: [1-\exp\left(-\frac{2\pi}{\kappa}\right) 

4497: \right ]^{-1},

4498: \end{equation}

4499: where $\kappa=E_{\bf k}/E_B$. For a completely screened 

4500: potential $|\psi(0)|^2=1$.

4501: 

4502: If holes are degenerate and the electrons' velocity $v_{\bf k}$ 

4503: is greater than the

4504: Fermi velocity of  the holes', then  

4505: \begin{equation}\label{eq:relax:BAP2}

4506: \frac{1}{\tau_s}=\frac{3}{\tau_0}p a_B^3 \frac{v_{\bf k}}{v_B}\frac{k_B 

4507: T}{E_{Fh}},

4508: \end{equation}

4509: where $E_{Fh}$ is the hole Fermi energy. For degenerate holes $|\psi(0)|^2$ is 

4510: of order 1.

4511: If electrons are thermalized, $v_{\bf k}$ needs to be replaced by the thermal 

4512: velocity $v_e=(3k_B T/m_c)^{1/2}$.

4513: 

4514: The temperature dependence of $\tau_s$ is dominated by the temperature 

4515: dependence of $|\psi(0)|^2$

4516: as well as by $p$. The dependence on the 

4517: acceptor density

4518: is essentially $1/\tau_s \sim N_a$ for nondegenerate/bound holes from 

4519: Eq.~(\ref{eq:relax:BAP1}) and $1/\tau_s \sim N_a^{1/3}$ for degenerate holes 

4520: from 

4521: Eq.~(\ref{eq:relax:BAP2}). In between, $1/\tau_s$ is only weakly dependent 

4522: on $N_a$.

4523: For GaAs $a_B\approx 114$ \AA, $E_B \approx 4.9$ meV, 

4524: $v_B\approx 1.7 \times 10^7$ cm$\cdot$s$^{-1}$, 

4525: $\tau_0\approx 1\times 10^{-8}$ s, and 

4526: $\Delta_{\rm ex}\approx 4.7 \times 10^{-5}$ eV~\cite{Aronov1983:SPJETP}.

4527: 

4528: The BAP mechanism coexists with the the EY and DP mechanisms 

4529: in p-doped materials lacking

4530: inversion symmetry. The three mechanisms can be distinguished by their

4531: unique density and temperature dependences. A general trend is that the

4532: BAP dominates in heavily doped samples at small temperatures. At large

4533: temperatures even for large acceptor densities, the DP mechanism can become

4534: more important, due to its increased importance at large electron energies. 

4535: Specific examples of the domain of importance for the three

4536: mechanisms are discussed in Sec.~\ref{sec:IIID1}.  

4537: Model calculations of BAP processes for electrons in p-doped bulk and quantum 

4538: wells were

4539: performed by \textcite{Maialle1996:PRB,Maialle1997:PRB}. 

4540: 

4541: Another potentially relevant mechanism for spin relaxation of donor-bound 

4542: electrons in p-doped semiconductors is the

4543: exchange interaction with holes bound to acceptors~\cite{Dyakonov1974:SPJETP}. 

4544: The exchange interaction provides an effective magnetic field for electron 

4545: spins 

4546: to precess, leading to inhomogeneous dephasing. Both electron hopping and hole 

4547: spin-flip motionally narrow the precession.

4548: 

4549: \subsubsection{\label{sec:IIIB4} Hyperfine-interaction mechanism}

4550: %--------------------------------------------------------------------------

4551: 

4552: The hyperfine interaction, which is the magnetic interaction between 

4553: the magnetic moments of electrons and nuclei, provides an important 

4554: mechanism~\cite{Dyakonov1974:SPJETP} for ensemble spin dephasing

4555: and single-spin decoherence of localized electrons, such as those confined 

4556: in quantum dots (QD) or bound on donors. The interaction is too weak to cause 

4557: effective 

4558: spin relaxation

4559: of free electrons in metals or in bulk semiconductors~\cite{Overhauser1953:PR},

4560: as it is strongly dynamically narrowed by the itinerant nature of 

4561: electrons (see Sec.~\ref{sec:IIIA1}).

4562: In addition to spin dephasing, the hyperfine interaction

4563: is relevant for spintronics as a means to couple, in a controlled way, 

4564: electron 

4565: and nuclear spins \cite{Dyakonov:1984}.

4566: 

4567: Localized electrons are typically spread over many lattice sites 

4568: ($10^4$--$10^6$),

4569: experiencing the combined magnetic moments of many nuclei. In GaAs all the 

4570: lattice nuclei carry the

4571: magnetic moment of 3/2 spin, while in Si the most abundant isotope, $^{28}$Si,

4572: carries no spin and the hyperfine interaction is due to $^{29}$Si 

4573: (natural abundance 4.67\%) 

4574: or the frequent donor $^{31}$P, both of nuclear spin $1/2$. As a result, an 

4575: electron bound

4576: on a shallow donor in Si experiences only around 100 magnetic nuclei, and the 

4577: effects of the hyperfine interaction are considerably smaller than in GaAs.

4578: 

4579: The effective Hamiltonian for the hyperfine interaction is the 

4580: Fermi contact potential 

4581: energy~\cite{Slichter:1989} 

4582: \begin{equation}

4583: H=\frac{8\pi}{3} \frac{\mu_0}{4\pi}g_0 \mu_B \sum_i \hbar \gamma_{n,i} 

4584: {\bf S}\cdot{\bf I}_i

4585: \delta ({\bf r}- {\bf R}_i),

4586: \end{equation}

4587: where $\mu_0$ is the vacuum permeability, $g_0=2.0023$ is the free-electron 

4588: $g$ factor, $\mu_B$

4589: is the Bohr magneton, $i$ is the label for nuclei at positions 

4590: ${\bf R}_i$, ${\bf S}$,  and ${\bf I}_i$ are, respectively,  electron and 

4591: nuclear spin operators expressed in the units of $\hbar$, and $\gamma_{n,i}$ 

4592: is the nuclear 

4593: gyromagnetic ratio.

4594: We stress that it is the electron $g$ factor $g_0$ and not the effective

4595: $g$ that appears in the hyperfine interaction, Eq.~(96), %label iz

4596: as shown by \textcite{Yafet1961:JPCS} [see also \textcite{Paget1977:PRB}].

4597: It follows that the spin of an electron in an orbital state $\psi({\bf r})$ 

4598: experiences 

4599: magnetic field 

4600: \begin{equation}

4601: {\bf B}_n=\frac{2\mu_0}{3} \frac{g_0}{g} \sum_i \hbar \gamma_{n,i} {\bf I}_i 

4602: |\psi({\bf R}_i)|^2,

4603: \end{equation}

4604: where $g$ is the effective $g$ factor of the electron. The electron Zeeman 

4605: splitting 

4606: due to the average $B_n$  

4607: corresponds to a field of  $\sim 1$ 

4608: T or thermal energy of 1 K, for a complete nuclear 

4609: polarization~\cite{Paget1977:PRB}.

4610: 

4611: There are three important regimes in which the hyperfine interaction leads to 

4612: spin dephasing of 

4613: localized

4614: electrons:

4615: 

4616: (i) In the limit of small orbital and spin correlation between 

4617: separated 

4618: electron states and nuclear spin states, spatial variations in ${\bf B}_n$ 

4619: lead to 

4620: inhomogeneous dephasing of 

4621: the spin ensemble, with the rate proportional to the r.m.s. of $B_n$, 

4622: given by the 

4623: corresponding

4624: thermal or nonequilibrium distribution of the nuclear spins. 

4625: Such inhomogeneous dephasing is seen by electron spin resonance (ESR) 

4626: experiments 

4627: on donor states both in Si \cite{Feher1959b:PR} and in GaAs \cite{Seck1997:PRB}. 

4628: This

4629: effect can be removed by spin echo experiments (in Si donor states performed, 

4630: for example, by

4631:  \textcite{Gordon1958:PR}). The spread in the Larmor precession period due to 

4632: the variance in 

4633: $B_n$ in GaAs is estimated to be around 1 ns 

4634: \cite{Merkulov2002:PRB,Dzhioev2002:PRL}).

4635: 

4636: (ii) Temporal fluctuations in $B_n$, 

4637: which can occur due to nuclear dipole-dipole 

4638: interactions, lead to irreversible spin dephasing and decoherence of electron 

4639: spins.

4640: Such processes are sometimes referred to as spectral diffusion, since the 

4641: electron Zeeman levels split by $B_n$ undergo random shifts \cite{Sousa2002:P}.

4642: The typical time scale for the fluctuations in GaAs is 

4643: given by the nuclear Larmor precession period in the field of neighboring 

4644: nuclei and

4645: is of order 100 $\mu$s~\cite{Merkulov2002:PRB}. Nuclear moments 

4646: also precess (and orient) in the magnetic fields of polarized electrons, an 

4647: effect important in 

4648: optical orientation \cite{Meier:1984}, where the feedback from this precession 

4649: can 

4650: be directly observed through the modulated precession of electron spins. The 

4651: time

4652: scale for the Larmor precession of nuclear spins in hyperfine fields is 1 $\mu$s in 

4653: GaAs 

4654: \cite{Merkulov2002:PRB}, so this effect does not lead to motional narrowing of 

4655: $B_n$;

4656: electron spins precess many times before the nuclear spin flips.

4657: 

4658: (iii) In the presence of strong orbital correlations (electron hopping or 

4659: recombination with

4660: acceptor hole states) or spin (direct exchange interaction) between

4661: neighboring electron states, spin precession due to $B_n$ is motionally 

4662: narrowed.

4663: While the direct spin exchange interaction does not cause ensemble spin 

4664: relaxation 

4665: (the total 

4666: spin is preserved in spin flip-flops), it leads to individual spin decoherence, 

4667: which 

4668: can be much faster than what is inferred from $T_2$. This 

4669: effect 

4670: is much more pronounced in GaAs than in Si, since the donor states spread to 

4671: greater

4672: distances, and thus even in the low-doping limits ($\approx 10^{14}$ cm$^{-3}$ 

4673: donors)

4674: the exchange interaction can be rather large, masking the effects of temporal 

4675: fluctuations

4676: of $B_n$ (see Sec.~\ref{sec:IIID3}). 

4677: Many useful parameters for evaluating effective magnetic fields and precession 

4678: frequencies

4679: due the HFI mechanism in GaAs are given by \textcite{Paget1977:PRB}.

4680: 

4681: Ensemble spin dephasing due to the HFI mechanism in an external magnetic 

4682: field has been studied 

4683: by \textcite{Dyakonov1974:SPJETP}, who found suppression of $1/\tau_s$ 

4684: if the external field 

4685: is greater than $B_n$ for regime (i), or a smaller Larmor precession 

4686: period than the correlation time 

4687: for random changes 

4688: in $B_n$, in regime (iii).

4689: due to the external field. 

4690: 

4691: Calculations of $\tau_s$ using the HFI mechanism  were performed for shallow 

4692: donor states in Si 

4693: at low 

4694: temperatures and magnetic fields~\cite{Saikin2002:NL}, for electron spins in 

4695: QD's~\cite{Merkulov2002:PRB,Semenov2002:P}, and even for the case of 

4696: conduction 

4697: electrons in semiconductors (revisited by 

4698: \textcite{Pershin2003:P}). Spin relaxation processes due 

4699: to phonon-assisted HFI  were investigated in GaAs QD's, but were found to be 

4700: ineffective \cite{Erlingsson2001:PRB}. Unfortunately, there are still 

4701: too few experimental 

4702: data to make conclusions about the merits of specific models of the 

4703: HFI mechanism. 

4704: 

4705: Spin decoherence times for single-electron spins were recently computed 

4706: for  

4707: case 

4708: (ii) by \textcite{Khaetskii2002:PRL,Khaetskii2002:P},

4709: who studied spin coherence time $\tau_{sc}$ of a single electron spin 

4710: in the regime in which the electron Larmor period due to $B_n$ is much

4711: shorter than the correlation time of the nuclear magnetic field fluctuations.

4712: Realistic estimates 

4713: of HFI spin dephasing in GaAs QD's were given 

4714: by \textcite{Sousa2003:PRB,Sousa2002:P}, who offer

4715: reasons why mechanism (ii)

4716: should dominate spin decoherence in GaAs QD's of radius smaller than 100 nm.

4717: For instance,  in a 50 nm wide QD the estimated $\tau_{sc}$

4718: is $\approx 50$ $\mu$s, 

4719: large enough 

4720: for quantum computing applications (see Sec.~\ref{sec:IVF}).

4721: This claim is supported by the recent measurement of the spin dephasing

4722: time of about 60 ms of an isolated spin in a phosphorus donor in

4723: isotopically pure $^{28}$Si, by spin echo measurements 

4724: \cite{Tyryshkin2003:PRB}. 

4725: 

4726: 

4727: \subsection{\label{sec:IIIC} Spin relaxation in metals}

4728: %--------------------------------------------------------------------------

4729: 

4730: The spin relaxation time of conduction electrons 

4731: in metals has been measured by both CESR and spin injection techniques. 

4732: Typical values of $\tau_s$ were found to be 0.1 to 1 ns, but the range

4733: of observed values is large, from pico- to microseconds. To our knowledge the 

4734: longest $\tau_s$ reported for a metal--a microsecond--was found in high-purity

4735: sodium below 10 K~\cite{Kolbe1971:PRB}. 

4736: 

4737: The majority of simple metals are believed to follow the EY mechanism of spin 

4738: relaxation,

4739: with the possible exception of Li~\cite{Damay1976:PRB}. This is

4740: supported by several facts: 

4741: 

4742: (i) The EY processes give the right order of 

4743: magnitude

4744: for $\tau_s$ ~\cite{Elliott1954:PR,Yafet:1963},

4745: while other known possible spin relaxation mechanisms lead to 

4746: much greater $\tau_s$ than what is observed~\cite{Overhauser1953:PR}.

4747: 

4748: (ii) The temperature dependence of $\tau_s$ is consistent with

4749: the EY mechanism: $1/\tau_s$ is constant at low temperatures, indicating 

4750: impurity

4751: spin-flip scattering, while at high temperatures $1/\tau_s$ grows linearly with 

4752: increasing

4753: $T$, consistent with phonon-induced spin relaxation. 

4754: 

4755: (iii) The Elliott relationship, 

4756: Eq.~(\ref{eq:relax:elliott}), has been tested for many important metals and 

4757: found 

4758: to be valid over many orders of magnitude of $\Delta g$~\cite{Beuneu1978:PRB} 

4759: (this

4760: reference contains a useful collection of data for $\Delta g$). For the 

4761: majority of metals tested 

4762: (alkali and noble), a best fit gives the quantitative formula, 

4763: Eq.~(\ref{eq:revisedE}) 

4764: \cite{Beuneu1978:PRB}. 

4765: (iv) The Yafet relation, Eq.~(\ref{eq:relax:yafet}), is satisfied for 

4766: most metals 

4767: with the known 

4768: temperature dependence of $\tau_s$~\cite{Monod1979:PRB,Fabian1999:JVST}.

4769: The initially suggested deviation from the Yafet relation for several 

4770: polyvalent 

4771: metals 

4772: (Al, Pd, Be, and Mg) was later resolved by spin-hot-spot theory 

4773: \cite{Silsbee1983:PRB,Fabian1998:PRL,Fabian1999:JVST}, to be described

4774: below.

4775: This work showed that the magnitudes of the spin-mixing probabilities 

4776: $b^2$, taken from 

4777: atomic physics to test Eq.~(\ref{eq:relax:yafet}), should not be used in the

4778: solid-state environment. Various band-structure anomalies (spin hot spots), 

4779: such as crossing of the Brillouin-zone boundaries, 

4780: accidental degeneracy points, 

4781: or symmetry points

4782: on the Fermi surface, can increase the atomic-physics-derived $b^2$ by several 

4783: orders of magnitude, strongly enhancing spin relaxation in polyvalent metals 

4784: as compared to simple 

4785: estimates~\cite{Fabian1998:PRL,Fabian1999:JAP}.

4786: 

4787: (v) A realistic, first-principles calculation for Al \cite{Fabian1999:PRL} 

4788: (see Sec.~\ref{sec:IIIC}) using

4789: Eq.~(\ref{eq:relax:ey}) shows  

4790: excellent

4791: agreement with experiment. 

4792:  

4793: CESR measurements of $\tau_s$ in various metals are numerous.\footnote{A list

4794: of selected metal includes

4795: Li \cite{Feher1955:PR,Orchard-Webb1970:PSS,Damay1976:PRB};

4796: Na \cite{Feher1955:PR,Vescial1964:PR,Kolbe1971:PRB},

4797: K \cite{Walsh1966:PR};

4798: Rb \cite{Schultz1966:PRL,Walsh1966:PRL};

4799: Cs  \cite{Schultz1966:PRL,Walsh1966:PRL};

4800: Be \cite{Cousins1965:PL,Orchard-Webb1970:PSS}, Mg \cite{Bowring1971:PSS};

4801: Cu \cite{Schultz1965:PRL,Dunifer1976:PRB,Monod1982:JP};

4802: Au \cite{Monod1977:JLTP};

4803: Zn \cite{Stesmans1981:PRB}; 

4804: Al \cite{Lubzens1976:PRL}; 

4805: graphite \cite{Wagoner1960:PR,Matsubara1991:PRB};

4806: Rb$_{3}$C$_{60}$ \cite{Janossy1993:PRL}; MgB$_2$ \cite{simon2001:PRL}.

4807: Various data on CESR $\tau_s$ are collected in 

4808: \cite{Beuneu1978:PRB,Monod1979:PRB}.} 

4809: 

4810: The spin injection technique (see Sec.~\ref{sec:II}) was also used to measure 

4811: $\tau_s$ for various metals, 

4812: including Al 

4813: \cite{Johnson1985:PRL,Jedema2002:Na,Jedema2002:APL,Jedema2003:PRB}, 

4814: Au \cite{Elezzabi1996:PRL},

4815: Cu \cite{Jedema2001:N,Jedema2003:PRB}, and Nb \cite{Johnson1994:APL}. 

4816: In addition to CESR and spin injection, information about spin-orbit

4817: scattering 

4818: times 

4819: $\tau_{so}$ (see below)

4820: in various (but mostly noble) 

4821: metals at low temperatures has been also 

4822: obtained 

4823: from weak localization magnetoresistance measurements 

4824: on thin films~\cite{Bergmann1982:ZP} and 

4825: tunneling spectroscopy of metallic nanoparticles~\cite{Petta2001:PRL}.

4826: Surface-scattering spin relaxation times in normal metals and 

4827: superconductors 

4828: are

4829: collected by \textcite{Meservey1978:PRL}. Interesting results were

4830: obtained by injecting spin into superconductors. Using YBa$_2$Cu$_3$O$_{7-

4831: \delta}$, for

4832: example, data were interpreted \cite{Fu2002:PRB} to infer that the 

4833: in-plane spin relaxation time is unusually long, 

4834: about 100 $\mu$s at low temperatures to 1 $\mu$s close to the superconducting

4835: transition temperature. For quasiparticles moving along the c-axis, 

4836: $\tau_s$ is more likely to be the usual spin-orbit-induced spin 

4837: relaxation time, having the values of 10-100 ps.

4838: The microscopic origin of quasiparticle spin relaxation in cuprate 

4839: superconductors is not yet known. 

4840: 

4841: There is one more important time scale, the spin-orbit scattering

4842: time $\tau_{so}$, that is often invoked in mesoscopic transport

4843: as a characteristic of spin relaxation processes. We discuss it briefly

4844: in connection to $\tau_s$.

4845: The spin-orbit scattering time is the scattering time of

4846: Bloch electrons by the spin-orbit potential induced by impurities.

4847: The spin-orbit

4848: part of the Fourier transform of the impurity potential can be written 

4849: as $i c({\bf k}-{\bf k}') ({\bf k} \times {\bf k'})\cdot {\bf \sigma}$,

4850: where $c({\bf q})$ is proportional to the Fourier transform of the 

4851: impurity potential.

4852: The spin-orbit scattering time then is \cite{Werthamer:1969}

4853: \begin{equation}

4854: 1/\tau_{so}=\frac{2\pi}{\hbar} N_i \langle |c({\bf k}-{\bf k}')|^2 

4855: |{\bf k}\times {\bf k}' | \rangle^2

4856: {\cal N}(E_F),

4857: \end{equation}

4858: where $N_i$ is the impurity concentration, ${\cal N}(E_F)$   

4859: is the density of states per spin at

4860: the Fermi level, and the angle brackets denote Fermi-surface averaging. 

4861: As a parameter $\tau_{so}$ also includes the spin-orbit coupling of the host 

4862: lattice, 

4863: in the

4864: sense of the EY mechanism. Note, however, that $\tau_s \ne \tau_{so}$, even 

4865: at low

4866: temperatures where the impurity scattering dominates spin relaxation, 

4867: since the spin-orbit

4868: scattering includes both spin-flip and spin-conserving processes, which, 

4869: for

4870: isotropic scattering rates are in the ratio 2:1. In addition, 

4871: the spin relaxation rate is twice the spin-flip scattering rate, since each 

4872: spin flip

4873: equilibrates both spins equally.  For isotropic systems 

4874: $1/\tau_s\approx 4/(3\tau_{so})$.

4875: For a discussion of the effects of the DP processes on weak localization, 

4876: see \textcite{Knap1996:PRB}.}

4877: 

4878: 

4879: 

4880: 

4881: We illustrate spin relaxation in metals on the case of Al, whose $\tau_s$ was

4882: measured by CESR and spin injection, and numerically calculated

4883: from first principles. The case is instructive since it illustrates 

4884: both the general principles of the EY mechanism as well as the predicting 

4885: power 

4886: of, and the need for realistic band-structure calculations of $\tau_s$. 

4887: 

4888: Spin relaxation in Al was originally observed in CESR experiments, in which 

4889: $\tau_s$

4890: was measured at low temperatures, from 1 to 100 K \cite{Lubzens1976:PRL}.

4891: The spin relaxation rate $1/\tau_s$ was found to be independent of temperature 

4892: below 10-20 K; at higher temperatures $1/\tau_s$ increases linearly with 

4893: increasing $T$. 

4894: The same behavior was later observed in the original spin injection 

4895: experiment~\cite{Johnson1985:PRL,Johnson1988:PRBb}.

4896: Recently, $\tau_s$  was measured by spin injection at room temperature

4897: ~\cite{Jedema2002:Na,Jedema2002:APL,Jedema2003:PRB}. Unlike the CESR 

4898: and the original spin injection experiments, which were performed on 

4899: bulk samples, the room-temperature measurement

4900: used  thin Al films, observing strong spin relaxation due to surface

4901: scattering.

4902: 

4903: 

4904: 

4905: 

4906: Spin relaxation in Al, as well as in  other polyvalent metals, at first 

4907: appeared \cite{Monod1979:PRB}, in that a simple application

4908: of the Yafet relation, Eq.~(\ref{eq:relax:yafet}), yielded estimates for 

4909: $1/\tau_s$ 

4910: two orders of magnitude smaller than the observed data. Consider Na as a 

4911: reference.

4912: The atomic $\lambda_{so}/\Delta E$ [cf. Eq.~(\ref{eq:relax:lambda})] for 

4913: Na and Al are 

4914: within about 10\% of each other

4915: ~\cite{Beuneu1978:PRB,Monod1979:PRB}, yet the corresponding $\tau_s$ for Al is 

4916: about two 

4917: orders of magnitude smaller than that for 

4918: Na~\cite{Feher1955:PR,Vescial1964:PR}. 

4919: This anomaly extends to the $g$ factors as well.

4920: For Na, $\Delta g_{Na} \approx -8 \times 10^{-4}$ and 

4921: for

4922: Al it is six times greater, $\Delta g_{Al}\approx -5\times 10^{-3}$, while

4923: one would expect them to differ also by about 10\%. Note, however, that

4924: the Elliott relation, Eq.~(\ref{eq:relax:elliott}), is unaffected by this

4925: discrepancy, as it predicts that $\tau_s$(Na})/$\tau_s$(Al)$\approx 40$.  

4926: It was later suggested~\cite{Silsbee1983:PRB} that this is due to 

4927: accidental degeneracies in the two-band Fermi surface of Al. 

4928: 

4929: A full theoretical description, supported by first-principles

4930: calculations, of spin relaxation in Al and other polyvalent metals 

4931: led to the

4932: spin-hot-spots 

4933: theory~\cite{Fabian1998:PRL,Fabian1999:PRL,Fabian1999:JAP,Fabian1999:JVST}. 

4934: Spin hot spots are states on the Fermi surface that have anomalously 

4935: large spin mixing probabilities $|b|^2\approx 

4936: (\lambda_{so}/\Delta E)^2$, 

4937: arising from small energy gaps $\Delta E$. Quite generally, such

4938: states occur near Brillouin-zone boundaries and accidental degeneracy points,

4939: or at high-symmetry points. The condition for a spin hot spot is both a small 

4940: band gap

4941: $\Delta E$ and nonvanishing  

4942: $\lambda_{so}$.\footnote{At some symmetry points 

4943: $|b|$ may be very small. This occurs in the noble metals which 

4944: have Fermi states at 

4945: the Brillouin-zone boundaries, where $\Delta E$ is large,

4946: but the corresponding $\lambda_{so}$ is very small 

4947: due to symmetry.} 

4948: 

4949: If an electron hops in or out of a spin hot spot, the chance of spin flip 

4950: dramatically increases. Although the total area of spin hot spots on the 

4951: Fermi 

4952: surface

4953: is small, their contribution to $1/\tau_s$ is dominant, due to the large value

4954: of their $|b|^2$ in the Fermi surface average $\langle |b|^2 \rangle$, as

4955: was shown by analytical arguments~\cite{Fabian1998:PRL,Fabian1999:JAP}.

4956: A realistic numerical calculation~\cite{Fabian1999:PRL} for Al, 

4957: also showed that both

4958: the accidental degeneracies considered by~\cite{Silsbee1983:PRB} and states 

4959: close to 

4960: the Brillouin-zone

4961: boundaries dominate spin relaxation. 

4962: 

4963: A realistic calculation of $\tau_s$ in Al, based on pseudopotentials 

4964: and a realistic phonon description, has been performed by 

4965: \textcite{Fabian1999:PRL} and 

4966: compared to the experimental data available for $T<100$ 

4967: K~\cite{Johnson1985:PRL,Johnson1988:PRBb}. 

4968: Figure \ref{fig:T1_Al} shows both the experiment and the theory. 

4969: In the experimental data only the phonon contribution to $1/\tau_s$ is 

4970: retained

4971: \cite{Johnson1985:PRL}; the constant background impurity scattering is removed.

4972: The figure shows a rapid decrease of $\tau_s$ with increasing $T$ at low $T$, 

4973: where

4974: the agreement between experiment and theory is very good. 

4975: Above 200 K (the Debye temperature $T_D\approx 400$ K)

4976: the calculation predicts a linear dependence $\tau_s [{\rm ns}] 

4977: \approx 24\times

4978: T^{-1}[\rm K^{-1}]$.

4979: In the phonon-dominated linear regime the EY mechanism predicts that 

4980: the ratio $a^{ph}=\tau_p/\tau_s$ does not depend on $T$

4981: (see Sec.~\ref{sec:IIIB1}). 

4982: The calculated value is 

4983: $a^{ph}_{th}=1.2\times 10^{-4}$~\cite{Fabian1999:PRL}, 

4984: showing that $10^4$ phonon scatterings are needed to randomize electron spin.

4985: 

4986: 

4987: An important step towards extending spin injection capabilities was undertaken 

4988: recently

4989: by achieving spin injection into Cu and Al at room temperature

4990: ~\cite{Jedema2001:N,Jedema2002:Na,Jedema2002:APL,Jedema2003:PRB}; the 

4991: measured data 

4992: are unique

4993: in providing reliable values for spin diffusion lengths and spin relaxation 

4994: times in 

4995: these two important metals at room temperature. 

4996: The measured values for Al are 

4997: somewhat 

4998: sensitive to the experimental procedure/data analysis: $\tau_s= 85$ ps

4999: ~\cite{Jedema2002:Na} and $\tau_s=124$ ps ~\cite{Jedema2003:PRB}, as compared 

5000: to $\tau_s=90$ ps predicted by the theory at $T=293$ K. The room temperature 

5001: experimental

5002: data are included in Fig.~\ref{fig:T1_Al} for comparison. They nicely confirm 

5003: the 

5004: theoretical prediction. Less sensitive to data analysis is the ratio $a^{ph}$, 

5005: for

5006: which the experiments give $1.1\times 10^{-4}$~\cite{Jedema2002:Na} and 

5007: $1.3\times 10^{-4}$ \cite{Jedema2003:PRB}, comparing favorably with the

5008: theoretical  $a^{ph}_{th}=1.2\times 10^{-4}$.    

5009:  

5010: \begin{figure}

5011: \centerline{\psfig{file=zutic_fig16.eps,width=1\linewidth,angle=0}}

5012: \caption{Measured and calculated $\tau_s$ in Al. The low-$T$ measurements are 

5013: CESR \cite{Lubzens1976:PRL} and 

5014: spin injection \cite{Johnson1985:PRL}. 

5015: Only the phonon contribution is shown, as adapted from

5016: \textcite{Johnson1985:PRL}. 

5017: The solid line is the first-principles calculation, not a fit to the 

5018: data, \cite{Fabian1998:PRL}.

5019: The data at $T=293$ K are results from room-temperature spin injection 

5020: experiments of \textcite{Jedema2002:Na,Jedema2003:PRB}. 

5021: Adapted from \onlinecite{Fabian1999:PRL}.}

5022: \label{fig:T1_Al}

5023: \end{figure}

5024: 

5025: 

5026: Spin relaxation in Al depends rather strongly on magnetic fields at low $T$.

5027: CESR measurements ~\cite{Lubzens1976:PRL,Dunifer1976:PRB} 

5028: show that at temperatures below 100 K, $1/\tau_s$ increases linearly

5029: with increasing $B$. A specific sample~\cite{Lubzens1976:PRL}

5030: showed a decrease of $\tau_s$ from about 

5031: 20 ns to 1 ns, upon increase in $B$ from 0.05  to 1.4 T. 

5032: It was proposed that the observed behavior was due to cyclotron motion 

5033: through spin hot spots ~\cite{Silsbee1983:PRB}. The reasoning is as follows.

5034: Assume that there is considerable spread (anisotropy) 

5035: $\delta g \approx \Delta g$ 

5036: of the $g$ factors over the Fermi surface. Such a situation is common in

5037: polyvalent metals, whose spin hot spots have anomalously 

5038: large spin-orbit coupling.

5039: In a magnetic field the electron spins precess correspondingly with rates

5040: varying by $\delta \Omega_L \approx (\delta g/g) \Omega_L$, where 

5041: $\Omega_L$ is the Larmor frequency. 

5042: Motional narrowing leads to $1/\tau_s \approx (\delta\Omega_L)^2 \tau_c$,

5043: where $\tau_c$ is the correlation time for the random changes in $g$. 

5044: At small magnetic fields $\tau_c=\tau_p$ and $1/\tau_s \sim B^2 \tau_p$. 

5045: Such 

5046: a quadratic dependence of $1/\tau_s$ on $B$ is a  typical motional 

5047: narrowing case

5048: and has been observed at low temperatures in Cu \cite{Lubzens1976:PRL}.  

5049: As the field increases $\tau_c$ becomes the time of flight through 

5050: spin hot spots, in which case $\tau_c \sim 1/B$. As a result $1/\tau_s$ 

5051: acquires

5052: a component linear in $B$, in accord with experiment. 

5053: 

5054: 

5055: In an effort to directly detect phonon-induced spin flips in Al, an interesting 

5056: experiment was devised ~\cite{Lang1996:PRL,Grimaldi1996:PRL} 

5057: using the Zeeman splitting of the energy gap in Al superconducting tunnel 

5058: junctions. 

5059: Although the experiment failed, due to overwhelming spin-flip 

5060: boundary scattering,

5061: it showed the direction for future research in studying spin-flip 

5062: electron-phonon

5063: interactions.

5064: 

5065: \subsection{ \label{sec:IIID} Spin relaxation in semiconductors}

5066: %--------------------------------------------------------------------------

5067: 

5068: Although sorting out different spin-relaxation mechanisms of conduction 

5069: electrons in semiconductors

5070: is a difficult task, it has generally been 

5071: observed that the EY mechanism is relevant in small-gap and large-spin-orbit 

5072: coupling semiconductors, while the DP processes are responsible for spin 

5073: dephasing in middle-gap 

5074: materials and at high temperatures. In heavily p-doped samples the BAP  

5075: mechanism 

5076: dominates at lower temperatures, while DP at higher. In low-doped systems the

5077: DP dominates over the whole temperature range 

5078: where electron states are extended.  

5079: Spin relaxation of bound electrons proceeds 

5080: through the hyperfine interaction. Finally, spin relaxation of holes is due to 

5081: the EY processes.

5082: In bulk III-V or II-VI materials, for holes $\tau_s\approx \tau_p$, 

5083: since the valence spin and orbital states are completely

5084: mixed. However, in two-dimensional systems, where the heavy and light hole 

5085: states are split,

5086: hole spin relaxation is much less effective.

5087: 

5088: \subsubsection{\label{sec:IIID1} Bulk semiconductors}

5089: %--------------------------------------------------------------------------

5090: 

5091: There is a wealth of useful data on $\tau_s$ in 

5092: semiconductors.\footnote{References for selected semiconductors include

5093: the following:

5094: {p-GaAs}: 

5095: \cite{Fishman1977:PRB,Marushchak1984:SPSS,Seymour1981:PRB,Aronov1983:SPJETP,%

5096: Zerrouati1988:PRB,Sanada2002:APL};

5097: {n-GaAs}: See \ref{sec:IIID3};

5098: {p-Al$_{x}$Ga$_{1-x}$As}: \cite{Garbuzov1971:ZhETF,Clark1975:PRB};

5099: {p-GaSb}: \cite{Aronov1983:SPJETP,Safarov1980:JPSJ,Sakharov1981:SPSS};

5100: {n-GaSb}: \cite{Kauschke1987:PRB};

5101: {n-InSb}: \cite{Chazalviel1975:PRB};

5102: {InAs}: \cite{Boggess2000:APL};

5103: {p-InP}: \cite{Gorelenok1986:SPS};

5104: {n-InP}: \cite{Kauschke1987:PRB};

5105: {n-GaN}: \cite{Fanciulli1993:PRB,Beschoten2001:PRB}.}

5106: 

5107: 

5108: A comprehensive theoretical investigation of spin dephasing in bulk 

5109: semiconductors

5110: (both p- and n-types), applied to GaAs, GaSb, 

5111: InAs, and InSb, has been carried out by \textcite{Song2002:PRB} by using the 

5112: EY, DP, and BAP mechanisms.

5113: The calculation uses analytical formulas like 

5114: Eq.~(\ref{eq:relax:chazalviel}), 

5115: while

5116: explicitly evaluating $\tau_p$ for different momentum scattering processes 

5117: at different control parameters (temperature and density), but only 

5118: for nondegenerate 

5119: electron systems (Boltzmann statistics) and zero magnetic field. 

5120: The main results are as follows: in n-type III-V 

5121: semiconductors DP dominates at $T \agt 5$ K. At lower $T$ 

5122: the EY mechanism becomes relevant. This crossover temperature appears to be 

5123: quite 

5124: insensitive to the electron density, being between 1 and 5 K in most 

5125: investigated III-V 

5126: semiconductors for donor densities greater than $10^{14}$ cm$^{-3}$ 

5127: \cite{Song2002:PRB}. 

5128: For p-type materials the dominant mechanisms are DP and BAP, with the 

5129: crossover 

5130: temperature 

5131: sensitive to the acceptor density. For example, in p-GaAs at room temperature, 

5132: the DP mechanism dominates below $10^{18}$ cm$^{-3}$, while at 

5133: large densities 

5134: the BAP 

5135: mechanism dominates. In small-gap InSb the DP mechanism appears to be 

5136: dominant for all 

5137: acceptor densities at temperatures above 50 K. The strong disagreement with 

5138: experiment found at low T (at 5 K, to be specific) points to our still limited  

5139: theoretical understanding of spin relaxation in semiconductors. 

5140: The discrepancy likely arises 

5141: from neglect of

5142: HFI effects. 

5143: 

5144: Spin relaxation times as long as 300 ns were recently obtained in bulk, 

5145: $\approx 100$ nm wide, GaAs at low 4.2 K, placed in the proximity of QW's

5146: ~\cite{Dzhioev2001:JETPL,Dzhioev2002:PRL}. The samples were low doped 

5147: ($\approx 10^{14}$ cm$^{-3}$ uncompensated donor density), so that optical 

5148: orientation detected $\tau_s$ 

5149: of

5150: electrons bound on donors. At such low donor concentrations the HFI 

5151: mechanism is 

5152: responsible

5153: for spin relaxation. The unusually large $\tau_s$ is attributed to the 

5154: presence 

5155: of additional

5156: conduction electrons in the structure, coming from the barriers separating 

5157: the 

5158: sample and

5159: the nearest QW. The hyperfine interaction is then motionally narrowed by the 

5160: exchange 

5161: interaction

5162: between the donor-bound and conduction electrons. Upon depletion of the 

5163: conduction electrons

5164: from the sample by resonant excitations in the QW,  

5165: $\tau_s$ decreased to 5 ns~\cite{Dzhioev2002:PRL}, implying that the 

5166: effects of the

5167: static hyperfine fields on bound-electrons spin precessions are not 

5168: reduced by 

5169: motional narrowing. 

5170: 

5171: 

5172: Spin relaxation of holes in bulk III-V materials is very fast due to a

5173: complete mixing of orbital and spin degrees of freedom in the valence band. 

5174: The 

5175: EY mechanism predicts that hole $\tau_s$ is similar to hole $\tau_p$;

5176: this is a common assumption when considering hole contribution 

5177: to spin-polarized

5178: transport. Hole spin lifetime in undoped GaAs has been measured by optical 

5179: orientation and time-resolved spectroscopy \cite{Hilton2002:PRL}. 

5180: The observed value

5181: at room temperature is $\tau_s \approx 110$ fs, consistent with the 

5182: theoretical 

5183: assumption. 

5184: 

5185: Spin relaxation of conduction electrons in strained III-V crystals was studied 

5186: experimentally and theoretically by 

5187: \textcite{Dyakonov1986:SPJETP}. Spin relaxation under strain is enhanced 

5188: and becomes

5189: anisotropic due to the 

5190: strain-induced spin splitting of the conduction band, which is linear in $k$, 

5191: similarly to the bulk inversion asymmetry in two-dimensional systems 

5192: (see Sec.~\ref{sec:IIIB2}). 

5193: It was found that $1/\tau_s\sim\sigma^2$, 

5194: where $\sigma$ is the applied stress, and that $\tau_s$ is only weakly 

5195: temperature dependent. Spin relaxation of photoholes in strain crystals has 

5196: been 

5197: studied in 

5198: \cite{Dyakonov1974:SPS}, with the conclusion that the hole spin along 

5199: the strain axis can 

5200: relax (by the EY processes) on time scales much longer than in 

5201: unstrained samples, 

5202: due to 

5203: the lifting of heavy and light hole degeneracy. 

5204: 

5205: 

5206: Compared to III-V or II-VI, much less effort has been devoted to 

5207: investigation of $\tau_s$ in bulk Si. The reason is that CESR is thus far

5208: the only technique capable of effective detection of spin relaxation 

5209: in Si. Optical orientation is rather weak~\cite{Lampel1968:PRL} due to the 

5210: indirect band-gap structure, while robust 

5211: spin injection in Si is yet to be demonstrated. 

5212: Spin relaxation in Si is slow due to the presence of inversion symmetry 

5213: (the DP mechanism is not applicable) and lack of a nuclear moment for 

5214: the main Si isotope. Earlier experimental studies~\cite{Feher1959b:PR} were

5215: concerned with the hyperfine-interaction-dominated spin dephasing in donor 

5216: states.

5217: 

5218: 

5219: A comprehensive experimental study of low-doped Si (P donors were present 

5220: at the 

5221: levels

5222: $7.5\times 10^{14} \le N_d \le 8\times 10^{16}$ cm$^{-3}$), 

5223: at temperatures $20 < T < 300$ K,

5224: was performed by \textcite{Lepine1970:PRB}. Three distinct temperature regimes 

5225: were 

5226: observed:

5227: 

5228: (a) ($20 < T < 75$ K) Here $\tau_s$ decreases with increasing $T$.

5229: The HFI mechanism dominates: electrons are bound to the ground donor states, 

5230: while

5231: thermal excitations to higher states and the exchange interaction with 

5232: conduction

5233: electrons motionally narrows the hyperfine interaction. 

5234: 

5235: (b) ($75 < T < 150$ K) In this temperature range 

5236: $\tau_s$ continues to decrease with increasing $T$, the effect caused by the 

5237: spin-orbit interaction in the first excited donor state being motionally 

5238: narrowed by thermal motion. 

5239: 

5240: (c) ($T> 150$ K) Here

5241: $1/\tau_s$ increases with $T$, in accord with the EY mechanism. 

5242: The observed room-temperature CESR linewidth is

5243: about 8 G, corresponding to the electron spin

5244: lifetime of 7 ns.

5245: 

5246:  

5247: \subsubsection{\label{sec:IIID2} Low-dimensional semiconductor structures}

5248: %-------------------------------------------------------------------------

5249: 

5250: 

5251: The importance of low-dimensional semiconductor systems (quantum wells, wires, 

5252: and dots)

5253: lies in their great flexibility in manipulating charge and, now, also spin 

5254: properties

5255: of the electronic states. Studies of spin relaxation in those systems are driven

5256: not only by the need for 

5257: fundamental understanding of spin relaxation and decoherence, but also 

5258: by the goal of finding ways to reduce or otherwise control spin 

5259: relaxation and coherence in general. For a survey of spin relaxation 

5260: properties 

5261: of semiconductor 

5262: quantum wells, see \textcite{Sham1993:JPCM}.

5263: 

5264: 

5265: Spin relaxation in semiconductor heterostructures is caused by random magnetic

5266: fields originating either from the base material or from the heterostructure 

5267: itself. All four mechanisms of spin relaxation can be important, depending on 

5268: the material, doping, and geometry. The difference from the bulk is 

5269: the localization of the wave function into two, one, or zero dimensions

5270: and the appearance of structure-induced random magnetic fields. 

5271: Of all the mechanisms, the DP and HFI are believed to be most relevant. 

5272: 

5273: The most studied systems are GaAs/AlGaAs QW's. 

5274: The observed $\tau_s$ varies from nanoseconds

5275: to picoseconds, depending on the range 

5276: of control parameters such as temperature, QW width or confinement 

5277: energy $E_1$, carrier 

5278: concentration, mobility, magnetic field, or bias.\footnote{Here 

5279: is a list of selected references with useful data on $\tau_s$ in 

5280: GaAs/AlGaAs QW's:

5281: confinement energy dependence has been studied by 

5282: \textcite{Tackeuchi1996:APL,Britton1998:APL,Ohno1999:PRL,Malinowski2000:PRB,%

5283: Ohno2000:PE,Endo2000:JJAP};

5284: temperature dependence is treated by 

5285: \textcite{Wagner1993:PRB,Ohno1999:PRL,Malinowski2000:PRB,Ohno2000:PE,%

5286: Adachi2001:PE};

5287: carrier concentration dependence is studied by \textcite{Sandhu2001:PRL};

5288: dependence on mobility is examined by \textcite{Ohno1999:PRL}; 

5289: and dependence on magnetic field 

5290: in studied by \textcite{Zhitomirskii1993:JETPL}.}

5291: 

5292: Spin relaxation has also been investigated in In/GaAs 

5293: \cite{Paillard2001:PRL,Cortez2002:PE}, in an InAs/GaSb superlattice 

5294: \cite{Olesberg2001:PRB}, in InGaAs \cite{Guettler1998:PRB}, in 

5295: GaAsSb multiple QW's by \cite{Hall1999:APL}.

5296: II-VI QW's (specifically ZnCdSe) were

5297: studied by \textcite{Kikkawa1997:S}, finding $\tau_s\approx 1$ ns, 

5298: weakly dependent

5299: on both mobility and temperature (in the range $5 < T < 270$ K). Electron and 

5300: hole

5301: spin dephasing have also been investigated in dilute magnetic 

5302: semiconductor QW's doped

5303: with Mn ions \cite{Crooker1997:PRB,Camilleri2001:PRB}. 

5304: 

5305: 

5306: Reduction of spin relaxation by inhibiting the BAP electron-hole exchange 

5307: interaction through spatially

5308: separating the two carriers has been demonstrated in 

5309: $\delta$-doped p-GaAs:Be/AlGaAs 

5310: \cite{Wagner1993:PRB}. The observed $\tau_s$ was $\approx 20$ ns 

5311: at $T < 10$ K, 

5312: which is indeed unusually large.

5313: The exchange interaction was also studied at room temperature, observing

5314: an increase of $\tau_s$ with bias voltage which increases spatial separation 

5315: between electrons

5316: and holes, reducing the BAP effects \cite{Gotoh2000:JAP}. In the fractional 

5317: quantum Hall 

5318: effect regime it was demonstrated \cite{Kuzma1998:S} that nonequilibrium spin 

5319: polarization in 

5320: GaAs QW's can survive for tens of $\mu$s. Spin lifetime was 

5321: also found to be enhanced 

5322: in GaAs QW's

5323: strained by surface acoustic waves \cite{Sogawa2001:PRL}. A theoretical study 

5324: \cite{Kiselev2000:PRB} proposed that spin dephasing in 2DEG can be 

5325: significantly  suppressed by constraining the system to finite stripes, several 

5326: mean free paths wide.  

5327: 

5328: 

5329: Theoretical studies focusing 

5330: on spin dephasing in III-V and II-VI systems include those of 

5331: \cite{Wu2000:PRB,Wu2001:JS,Wu2002:SSC,Lau2001:PRB,Puller2002:P,Lau2002:JAP,%

5332: Bronold2002:PRB,Krishnamurthy2003:APL}.

5333: Spin relaxation due to the DP mechanism with bulk inversion asymmetry term 

5334: in the important case of GaAs/AlGaAs 

5335: rectangular QW's was investigated by Monte-Carlo simulations 

5336: \cite{Bournel2000:APL} at room 

5337: temperature, including interface roughness scattering. Nice agreement with 

5338: experiment was

5339: found for $\tau_s(E_1)$, where $E_1$ is the confinement energy. 

5340: Interface roughness becomes important at large

5341: values of $E_1$, where scattering increases $\tau_s$ (see also 

5342: \textcite{Sherman2003:APL}). 

5343: 

5344: Spin relaxation and spin coherence of spin-polarized photoexcited 

5345: electrons and holes in symmetric p- and n-doped 

5346: and undoped GaAs/AlGaAs quantum 

5347: wells was investigated using rate equations 

5348: \cite{Uenoyama1990:PRL,Uenoyama1990:PRB}. 

5349: It was shown that in these heterostructures hole spin relaxation proceeds 

5350: slower 

5351: than electron-hole recombination. Hole relaxation is found to occur mostly 

5352: due to 

5353: acoustic phonon emission. The ratio of the spin-conserving to spin-flip hole 

5354: relaxation

5355: times was found to be 0.46, consistent with the fact that 

5356: luminescence is polarized even in n-doped quantum wells at times greater 

5357: than the

5358: momentum relaxation time. Similar observations hold for  strained 

5359: bulk GaAs, where hole spin relaxation is also reduced. Spin relaxation of 

5360: holes in quantum wells 

5361: was calculated \cite{Ferreira1991:PRB,Bastard1992:SS} using the interaction 

5362: with ionized

5363: impurities and s-d exchange in semimagnetic semiconductors. It was shown that 

5364: size quantization 

5365: significantly reduces 

5366: spin relaxation of holes, due to the 

5367: lifting of

5368: heavy and light hole degeneracy. The observed spin lifetimes for holes at 

5369: low 

5370: temperatures

5371: reached up to 1 ns, while at $T > 50$ K in the same samples $\tau_s$ got

5372: smaller than

5373: 5 ps \cite{Baylac1995:SSS}. 

5374: 

5375: Spin dynamics and spin relaxation of excitons in GaAs 

5376: \cite{Munoz1995:PRB,Vina2001:PB} and ZnSe \cite{Kalt2000:PSS} were

5377: investigated 

5378: experimentally and theoretically \cite{Maialle1993:PRB,Sham1988:PE}.

5379: Coherent spin dynamics in magnetic semiconductors was considered by 

5380: \textcite{Linder1998:PE}.

5381: 

5382: Spin relaxation in Si heterostructures has been investigated by electron 

5383: spin resonance in modulation doped Si/SiGe QW's.  

5384: Very high mobility (about $10^5$ cm$^2\cdot$V$^{-1}\cdot$s$^{-1}$) samples with 

5385: $n\approx 3\times

5386: 10^{11}$ cm$^{-2}$ free electrons forming a 2DEG show, 

5387: at $T=4.2$ K, 

5388: $T_1$ up to 30 $\mu$s \cite{Sandersfeld2000:TSF,Graeff1999:PRB} 

5389: and $T_2$ of the order of 100 ns~\cite{Graeff1999:PRB},

5390: depending on the orientation of $B$ with respect to the QW growth direction. 

5391: Spin relaxation

5392: was attributed to Bychkov-Rashba spin splitting in these asymmetric wells, 

5393: estimating

5394: the corresponding $\hbar\alpha_{BR}$ 

5395: in Eq.~(\ref{eq:relax:BR}) to be around $1\times 

5396: 10^{-14}$ eV$\cdot$m~\cite{Wilamowski2002:PE,Wilamowski2002:PRB}. 

5397: Si/Ge heterostructures may have enhanced 

5398: rates of spin relaxation due to 

5399: the leakage of the electron wave function to Ge, 

5400: which is heavier than Si and has 

5401: greater spin-orbit interaction. 

5402: Recent studies \cite{Wilamowski2004:PRB} have confirmed the dominant role

5403: of the DP spin relaxation mechanism, leading to the microsecond spin 

5404: relaxation times. The spin dephasing is argued to be strongly suppressed

5405: by cyclotron motion in high-mobility samples (see Sec.~\ref{sec:IIIB2a}

5406: for a brief discussion of the influence of magnetic field on $\tau_s$).

5407: Spin-orbit coupling in symmetric Si/SiGe quantum wells has been studied 

5408: theoretically by \textcite{Sherman2003:PRB}.

5409: 

5410: In quantum dots the relevant spin relaxation mechanism is still being 

5411: debated,

5412: as the mechanisms 

5413: (EY and DP) effective for conduction electrons are ineffective for states

5414: localized in QD's~\cite{Khaetskii2000:PRB,Khaetskii2001:PRB,Khaetskii2001:PE}.

5415: It is believed, however, that similar to electrons bound on donors, the dominant

5416: mechanism is a HFI process

5417: \cite{Merkulov2002:PRB,Semenov2002:P,Sousa2003:PRB,Sousa2002:P,%

5418: Khaetskii2002:PRL}.

5419: Unfortunately, experiments on CdSe QD's (of diameter 22-80 \AA) show strong 

5420: inhomogeneous dephasing

5421: ($\tau_s\approx 3$ ns at $B=0$, while $\tau_s \approx 100$ ps at 4 T) 

5422: \cite{Gupta1999:PRB}, masking the intrinsic spin dephasing processes.

5423: Recently a lower bound, limited by the signal-to-noise ratio, on $T_1$

5424: of 50 $\mu$s has been measured at 20 mK in a one-electron quantum dot

5425: defined in 2DEG GaAs/AlGaAs heterostructure by 

5426: \textcite{Hanson2003:lanl}. The magnetic field of 7.5 T was oriented 

5427: parallel to the plane heterostructure. While the actual value of

5428: $T_1$ may be orders of magnitude larger, the observed bound suffices

5429: for performing elementary quantum gates (see Sec.~\ref{sec:IVF}).

5430: 

5431: \subsubsection{\label{sec:IIID3} Example: spin relaxation in GaAs}

5432: %---------------------------------------------------------------------

5433: 

5434: We review recent experimental results on spin relaxation in bulk n-GaAs% 

5435: \footnote{p-GaAs is extensively discussed by \textcite{Meier:1984}}

5436: and GaAs-based low-dimensional systems.

5437: 

5438: 

5439: 

5440: \paragraph{\label{sec:IIID3a} Bulk n-GaAs.}

5441: %---------------------------------------------------------------------------

5442: 

5443: The importance of GaAs for spintronics and quantum computing applications 

5444: has been 

5445: recently

5446: underlined by the discovery of rather long spin relaxation times (of the order 

5447: of 100 ns)

5448: in n-doped samples, as well as by the development of experimental 

5449: techniques to 

5450: manipulate spin precession in this semiconductor in a coherent 

5451: manner~\cite{Awschalom2001:PE,Oestreich2002:SST}.

5452: 

5453: 

5454: Both optical orientation and time-resolved Faraday rotation spectroscopy have 

5455: been used to measure $\tau_s$ in bulk n-GaAs. 

5456: In the earliest observations of optical spin orientation of

5457: electrons, in 

5458: n-Ga$_{0.7}$Al$_{0.3}$As with $N_d\approx 1\times 10^{16}$ cm$^{-3}$ at 4.2 K,

5459: it was found that $\tau_s\approx 1.2 $ ns~\cite{Ekimov1971:JETPL}.

5460: A much larger spin lifetime was found by optical orientation on n-GaAs 

5461: ~\cite{Dzhioev1997:PSS}, 

5462: where for $N_d\approx 1\times 10^{15}$ cm$^{-3}$ the observed

5463: $\tau_s\approx 42$ ns. 

5464: Faraday rotation studies~\cite{Kikkawa1998:PRL,Awschalom2001:PE} 

5465: found even longer  spin lifetimes. At the doping density 

5466: $N_d=1\times 10^{16}$ cm$^{-3}$ of Si donors and  $T=5$ K, the observed 

5467: $\tau_s\approx 130$ ns at zero magnetic field. 

5468: At greater and smaller doping densities, spin relaxation time is significantly 

5469: reduced: for both a nominally undoped sample and for $N_d=1\times 10^{18}$ 

5470: cm$^{-3}$, $\tau_s\approx 0.2$ ns. A comprehensive theoretical investigation

5471: of $\tau_s$ in bulk n-GaAs is reported by \textcite{Wu2000:PSS}, and

5472: \textcite{Wu2001:JS}, who solved numerically kinetic equations in the

5473: presence of magnetic filed. Only the DP mechanism was considered, 

5474: acting with longitudinal phonon and impurity scattering.

5475: 

5476: A recent comprehensive study of $\tau_s$ based on optical orientation 

5477: revealed a nice, albeit 

5478: complex, picture of spin relaxation in bulk n-GaAs over a large range of doping 

5479: levels~\cite{Dzhioev2002:PRB}. Figure \ref{fig:Dzhioev} summarizes these 

5480: findings. 

5481: The spin relaxation time rises with increasing

5482: $N_d$ at small doping levels, reaching its first maximum (180 ns) at around 

5483: $3\times 

5484: 10^{15}$ cm$^{-3}$; $\tau_s$ then decreases until $N_d=N_{dc}=2\times 10^{16}$ 

5485: m$^{-3}$,

5486: where a sudden increase brings $\tau_s$ to another maximum, reaching $\approx 

5487: 150$ ns. 

5488: At still higher doping levels $\tau_s$ decreases strongly with increasing 

5489: doping.

5490: 

5491: \begin{figure}

5492: \centerline{\psfig{file=zutic_fig17.eps,width=1\linewidth,angle=0}}

5493: \caption{Spin relaxation time in n-GaAs as a function of donor density $N_d$ 

5494: (labeled as $n_D$ here) at

5495: low temperatures:  empty symbols, the optical orientation data of 

5496: \textcite{Dzhioev2002a:PRB}; solid circles are the time-resolved 

5497: Faraday 

5498: rotation data

5499: of  \textcite{Kikkawa1998:PRL,Awschalom2001:PE};

5500: open triangles, single-spin decoherence times 

5501: $\tau_{sc}\approx \tau_c$ due to the exchange interaction 

5502: between electron spins on 

5503: neighboring donors;  solid lines, parameter-free theoretical estimates 

5504: with

5505: labels indicating the dominant spin relaxation mechanisms;

5506: dotted line a 

5507: fit to the experimental data on the exchange correlation time 

5508: (triangles) $\tau_c$, 

5509: using a simple model of the exchange coupling between donor states; 

5510: dashed vertical 

5511: line, the metal-insulator transition at 

5512: $N_{dc}=2\times 10^{16}$ cm$^{-3}$. 

5513: From \onlinecite{Dzhioev2002a:PRB}. 

5514: }

5515: \label{fig:Dzhioev}

5516: \end{figure}

5517: 

5518: The above picture is valid at $T \le 5$ K, where isolated shallow donors 

5519: are not normally ionized, and the sample is a Mott insulator at small dopings. 

5520: Conductivity

5521: is due to hopping between donor states. Beyond the critical density 

5522: $N_{dc}\approx 

5523: 2\times 10^{16}$ cm$^{-3}$ (the dashed vertical line in Fig.~\ref{fig:Dzhioev})

5524: the donor states start to overlap and form an impurity 

5525: conduction band---electronic

5526: states delocalize and the sample becomes metallic. Figure~\ref{fig:Dzhioev} 

5527: shows that

5528: it is the rather narrow window around the metal-to-insulator transition where 

5529: the largest $\tau_s$ are found.  

5530: 

5531: \begin{figure}

5532: \centerline{\psfig{file=zutic_fig18.eps,width=1\linewidth,angle=0}}

5533: \caption{Measured magnetic-field dependence of the spin dephasing time 

5534: (here denoted as $T_2^*$ to indicate the likely presence of inhomogeneous 

5535: broadening; 

5536: see Sec.~\ref{sec:IIIA1}) 

5537: for bulk n-GaAs

5538: at 5 K. Doping levels, varying from insulating 

5539: ($N_d < N_{dc}=

5540: 2\times 10^{16}$ cm$^{-3}$) to metallic ($N_d > N_{dc}$), are

5541: indicated. Adapted from \onlinecite{Kikkawa1998:PRL}.} 

5542: \label{fig:KA1}

5543: \end{figure}

5544: 

5545: 

5546: At $N_d > N_{dc}$ the DP mechanism dominates.  Equation

5547: (\ref{eq:relax:DP}) for degenerate electrons explains the observed data

5548: rather well.  Indeed, considering that $E_F\sim N_d^{2/3}$ and

5549: assuming the Brooks-Herring formula for the impurity scattering

5550: $(1/\tau_p \sim N_d/E_F^{3/2}$), one obtains $\tau_s\sim 1/N_d^2$,

5551: which is observed in Fig.~\ref{fig:Dzhioev}. The EY mechanism, 

5552: Eq.~(\ref{eq:relax:chazalviel}), would give $\tau_s\sim N_d^{-4/3}$.  The

5553: data on the insulating side are consistent with the HFI mechanism: the

5554: precession due to local random magnetic fields from the nuclear moments is

5555: motionally narrowed by the exchange interaction, which increases with

5556: increasing $N_d$ (that is, with increasing overlap between donor

5557: states). The theoretical estimates~\cite{Dzhioev2002:PRB} agree well

5558: with the data.  The behavior of $\tau_s$ in the intermediate regime, 

5559: $3\times 10^{15}$ cm$^{-3} < N_d < N_{dc}$, where $\tau_s$ decreases

5560: with increasing $N_d$, was proposed by~\cite{Kavokin2001:PRB} to be

5561: due to motional narrowing of the antisymmetric exchange interaction

5562: \footnote{The anisotropic exchange interaction of the

5563: Dzyaloshinskii-Moriya form, $\bf{S}_1\times {\bf S}_2$ 

5564: \cite{Dzyaloshinskii1958:PCS,Moriya1960:PR}

5565: appears as a result of spin-orbit coupling in semiconductors 

5566: lacking inversion symmetry.} 

5567: between bound electrons, with the correlation time $\tau_c$

5568: provided by the usual ${\bf S}_1\cdot{\bf S}_2$, direct exchange. This new

5569: mechanism of spin relaxation, which should be generally present for

5570: bound electrons in systems lacking inversion symmetry (such as III-V

5571: and II-VI), although still being

5572: investigated~\cite{Gorkov2003:PRB,Kavokin2002:P}, appears to give

5573: a satisfactory explanation for the experimental data.

5574: 

5575: In addition to the doping dependence of $\tau_s$, both the temperature and

5576: the magnetic field dependences of spin relaxation in bulk n-GaAs have been

5577: studied by \textcite{Kikkawa1998:PRL}.  Figure \ref{fig:KA1} shows

5578: $\tau_s(B)$ for samples with varying doping levels at $T=5$ K. The spin

5579: relaxation time increases with $B$ in the metallic regime, the

5580: behavior qualitatively consistent with the predictions of the DP

5581: mechanism.  In contrast, $\tau_s$ in the insulating samples decreases

5582: with increasing $B$.  Bound electrons are more susceptible to $g$-factor

5583: anisotropies (due to the distribution of electron energies over donor

5584: states) and local magnetic field variations (due to the hyperfine

5585: interactions).  These anisotropies are amplified by increasing $B$

5586: and motionally narrowed by the exchange interaction. It is thus

5587: likely that $\tau_s\sim B^{-2}\tau_c(B)$, where the exchange

5588: correlation time $\tau_c$ depends on $B$ through magnetic orbital

5589: effects on the bound electron wave functions (magnetic confinement

5590: reduces the extent of the bound orbital, thus reducing the exchange

5591: integrals between neighboring donor states). However, no satisfactory

5592: quantitative explanation for $\tau_s(B)$ in insulating samples

5593: exists.

5594: 

5595: \begin{figure}

5596: \centerline{\psfig{file=zutic_fig19.eps,width=1\linewidth,angle=0}}

5597: \caption{Measured temperature  dependence of the spin dephasing time 

5598: for bulk n-GaAs doped with $N_d=1\times 10^{16}$ cm$^{-3}$ Si donors, at 

5599: $B=0$ and $B=4$ T. 

5600: The sample is insulating at low

5601: $T$ and nondegenerate at high $T$ ($T\agt 50$ K, assuming $\approx 4$ meV for

5602: the donor binding energy), where donors are ionized.

5603: Adapted from \onlinecite{Kikkawa1998:PRL}. 

5604: }

5605: \label{fig:KA2}

5606: \end{figure}

5607: 

5608: 

5609: Figure \ref{fig:KA2} plots $\tau_s(T)$ for an insulating sample with

5610: $N_d=1\times 10^{16}$ cm$^{-3}$ at $B=0$ and $B=4$ T.   

5611: For the

5612: zero-field data the initial decrease of $\tau_s$ with $B$ is very

5613: rapid, dropping from 130 ns at 5 K to less than 1 ns at 150 K. However, 

5614: the sample held at $B=4$ T 

5615: shows at first a rapid

5616: increase with increasing $T$, and then a decrease at $T\approx 50$ K.

5617: The decrease of $\tau_s$ with increasing $T$ above 50 K has been found

5618: to be consistent with the DP mechanism~\cite{Kikkawa1998:PRL}, taking

5619: $\tau_s\sim T^{-3}$ in Eq.~(\ref{eq:relax:DPt}), while extracting

5620: the temperature dependence of $\tau_p$ from the measurement of mobility. 

5621: The DP

5622: mechanism for conduction electrons was also observed in p-GaAs in the

5623: regime of nondegenerate hole densities $N_a\approx 10^{17}$ cm$^{-3}$

5624: at temperatures above  100 K~\cite{Aronov1983:SPJETP},

5625: after the contribution from the BAP mechanism was subtracted using a

5626: theoretical prediction. From the observed mobility

5627: it was found that $\tau_p(T)\sim T^{-0.8}$, so that 

5628: according to the

5629: DP mechanism $\tau_s\sim

5630: T^{-2.2}$, which is indeed consistent with the experimental data.  The

5631: origin of $\tau_s(T)$ below 50 K in Fig.~\ref{fig:KA2} is less

5632: obvious.  At low $T$, electrons are localized, so in order to explain

5633: the experimental data the theory should include ionization of donors.

5634: The increase with increasing $T$ of $\tau_s$ at 4 T  

5635: invokes a

5636: picture of motional narrowing in which the correlation time decreases

5637: with increasing $T$ much faster than the dispersion of local Larmor

5638: frequencies. We do not know of a satisfactory quantitative explanation

5639: for these experimental results.\footnote{There is a 

5640: discrepancy in the data presented in Figs. \ref{fig:KA1}

5641: and \ref{fig:KA2}. Take the $N_d=1\times 10^{16}$ cm$^{-3}$ sample.

5642: While Fig.~\ref{fig:KA1} reports $\tau_s\approx 3$ ns at 5 K and 4 T, 

5643: $\tau_s$ is only about 1 ns in Fig.~\ref{fig:KA2}.

5644: The reason for this difference \cite{Kikkawa2003:PC}

5645: turns out to be electronically-induced nuclear polarization 

5646: \cite{Kikkawa2000:S}.

5647: At low temperatures and large magnetic fields, nuclear polarization

5648: develops via the Overhauser effect inhomogeneously throughout the

5649: electron spin excitation region. The inhomogeneous magnetic

5650: field due to polarized nuclei causes inhomogeneous broadening

5651: of the electronic $\tau_s$. The measured spin dephasing time is indeed

5652: $T_2^*$, rather than the intrinsic $T_2$. Furthermore, since nuclear

5653: polarization typically takes minutes to develop, the measured

5654: $T_2^*$ depends on the measurement ``history.'' This is the reason

5655: why two different measurements, reported in Figs. \ref{fig:KA1}

5656: and \ref{fig:KA2}, show different $T_2^*$ under otherwise equivalent

5657: conditions. The nuclear polarization effect is also part of the

5658: reason why the $T_2(T)$ at 4 T 

5659: sharply deviates from that

5660: at zero field at small $T$. The technique should give consistent

5661: results at small fields and large temperatures, as well as 

5662: in heavily doped samples where the nuclear fields are motionally

5663: narrowed by the itinerant nature of electrons.

5664: }

5665: Similar behavior of

5666: $\tau_s(T)$ 

5667: in insulating samples was found in GaN~\cite{Beschoten2001:PRB}.

5668: 

5669: 

5670: The temperature dependence of $\tau_s$ for samples with 

5671: $N_d \gg N_{dc}$ has been reported \cite{Kikkawa1998:PRL} to be very weak, 

5672: indicating, for

5673: these degenerate electron densities, that $\tau_p(T)$ is only  weakly

5674: dependent on $T$. What can be expected for $\tau_s$ at room temperature?  The

5675: answer will certainly depend on $N_d$. Recent experiments on time-

5676: resolved Kerr rotation~\cite{Kimel2001:PRB} suggest that 5 ps$ < \tau_s <

5677: 10 $ ps for undoped GaAs and 15 ps $< \tau_s < 35$ ps for a heavily doped

5678: n-GaAs with $N_d=2\times 10^{18}$ cm$^{-3}$.

5679: 

5680: For spintronic applications to make use of the large $\tau_s$

5681: observed in bulk n-GaAs one is limited to both very small temperatures

5682: and small doping levels. Although this may restrict the design of 

5683: room-temperature spintronic devices, such a regime seems acceptable for

5684: spin-based quantum computing (see Sec.~\ref{sec:IVF}), 

5685: where one is interested in the spin lifetime

5686: of single (or a few) electrons, bound to impurities or confined to

5687: quantum dots.  How close is $\tau_s$ to the 

5688: individual spin lifetime $\tau_{sc}$?

5689: There is no clear answer yet.  Ensemble spin dephasing seen for

5690: insulating GaAs samples appears to be due to motional narrowing of the

5691: hyperfine interaction. The randomizing processes are spin flips due to

5692: the direct exchange, leading to the correlation time $\tau_c$, which can

5693: be taken as a measure for the lifetime $\tau_{sc}$ of the individual spins. 

5694: Extracting

5695: this lifetime from the experiment is not easy, but the obvious trend

5696: is the smaller the $\tau_c$, the larger the $\tau_s$.  For a specific

5697: model of spin relaxation in bound electron states $\tau_c$ was

5698: extracted experimentally by \textcite{Dzhioev2002:PRB} by detecting

5699: the changes in the spin polarization due to longitudinal magnetic fields.  The

5700: result is shown in Fig.~\ref{fig:Dzhioev}.  The two times, $\tau_c$

5701: and $\tau_s$ differ by orders of magnitude. For the doping levels where

5702: $\tau_s$ is {\it greater} than 100 ns, $\tau_c$ is {\it smaller} than   

5703: 0.1 ns.

5704: Unfortunately, the useful time

5705: for spin quantum computing would be extracted in the limit of very

5706: small dopings, where the data are still sparse. For an informal recent review

5707: of $\tau_s$ in n-GaAs, see \textcite{Kavokin2002:PSS}.

5708: 

5709: Closely related to spin relaxation is spin diffusion. 

5710: \textcite{Hagele1998:APL} observed the transport of a spin 

5711: population---longitudinal spin drift---in i-GaAs over a length

5712: scale greater than 4 $\mu$m in electric

5713: fields up to 6 kV/cm and at low temperatures. 

5714: This was followed by a remarkable result of \textcite{Kikkawa1999:N},

5715: the observation of the drift of precessing electron 

5716: spins---transverse spin drift---in GaAs with $N_d=1\times 10^{16}$ cm$^{-3}$, 

5717: over 100 $\mu$m in moderate electric fields (tens of V/cm) at $T=1.6$ K,

5718: setting the length scale for the spin dephasing.

5719: By directly analyzing the spreading and drifting of the electron

5720: spin packets in time, 

5721: Kikkawa and Awschalom obtained

5722: the spin diffusion 

5723: (responsible for spreading) and electronic

5724: diffusion

5725: (drift by electric field) 

5726: coefficients.

5727: It was found that

5728: the former is about 20 times as large as the latter. These results are

5729: difficult to interpret, since the sample is just below the metal-to-insulator

5730: transition, where charge is transported via hopping, but they suggest that 

5731: spin diffusion is strongly

5732: enhanced through the exchange interaction. Investigations of this type

5733: in even smaller doping limits may prove important for understanding

5734: single-spin coherence.

5735: 

5736: \paragraph{GaAs-based quantum wells.}

5737: %-----------------------------------------------------------------

5738: 

5739: We discuss selected experimental results on spin relaxation in 

5740: GaAs/Al$_x$Ga$_{1-x}$As QW's,

5741: presenting  the temperature and confinement energy dependence of $\tau_s$.

5742: 

5743: Figure \ref{fig:QW1} plots the temperature dependence of $1/\tau_s$ in

5744: the interval of $90 < T < 300$ K for QW's of width $L$ ranging from 6

5745: to 20 nm \cite{Malinowski2000:PRB}.  The wells, with $x=0.35$ and

5746: orientation along [001], were grown on a single wafer to minimize

5747: sample-to-sample variations when comparing different wells. The

5748: reported interface roughness was less than the exciton Bohr radius of

5749: 13 nm.  In these structures the excitonic effects dominate at $T < 50$ K

5750: (with the reported $\tau_s\approx 50$ ps), while the exciton ionization is

5751: complete roughly at $T > 90$ K, so the data presented are for free

5752: electrons. Spin relaxation was studied  using pump-probe optical

5753: orientation spectroscopy with a 2 ps time resolution and the typical

5754: excitation intensity/pulse of $10^{10}$

5755: cm$^{-2}$.

5756: 

5757: \begin{figure}

5758: \centerline{\psfig{file=zutic_fig20.eps,width=1\linewidth,angle=0}}

5759: \caption{Measured temperature dependence of the conduction-electron-spin 

5760: relaxation rate 

5761: $1/\tau_s$

5762: in GaAs/AlGaAs QW's of varying widths: the dashed curve,  data for

5763: a low-doped ($N_a=4\times 10^{16}$ cm$^{-3}$) 

5764: bulk p-GaAs \cite{Meier:1984}; solid line, the 

5765: $\tau_s \sim T^2$ dependence. From \onlinecite{Malinowski2000:PRB}.} 

5766: \label{fig:QW1}

5767: \end{figure}

5768: 

5769: As Fig.~\ref{fig:QW1} shows, $\tau_s$ depends rather weakly on $T$ for

5770: the narrow wells with $L < 10$ nm.

5771: For the well with $L=15$ nm, after

5772: being approximately constant (or somewhat decreasing) as $T$ increases

5773: to about 200 K, $\tau_s$ 

5774: increases with increasing $T$ at greater temperatures.

5775: The increase is consistent with the $1/\tau_s \sim T^2$ behavior.  The

5776: thickest well increases with the same power law, $1/\tau_s\sim T^2$, over 

5777: the whole temperature range.  In order to make a reliable comparison

5778: with theoretical predictions (the expected mechanism is that of DP in

5779: two-dimensional systems), one needs to know the behavior of $\tau_p(T)$.

5780: The DP mechanism predicts, for the nondegenerate electron densities

5781: employed in the experiment, that $1/\tau_s \sim T^3 \tau_p$ [see 

5782: Eq.~(\ref{eq:relax:DPt})] in the bulk and wide QD's, the condition being that

5783: thermal energy is greater than the subband separation), and $1/\tau_s

5784: \sim T E_1^2 \tau_p$ from Eq.~(\ref{eq:relax:DK}) for the

5785: bulk inversion asymmetry

5786: after making thermal averaging ($E_{\bf k} \rightarrow k_B T$),

5787: when one realizes

5788: that confinement energy $E_1$ is $\sim \langle k_n^2 \rangle$.

5789: When one assumes that momentum relaxation in these elevated temperatures

5790: is due to scattering by phonons, $\tau_p$ should be similar in bulk

5791: and low-dimensional structures. From the observed high-temperature

5792: bulk $\tau_s(T)$ (at low temperatures $\tau_s$ is affected by BAP

5793: processes) one can estimate $\tau_p \sim 1/T$, which is consistent

5794: with the constant $\tau_s$ for the narrow wells, 

5795: and with the quadratic dependence for the wide wells.

5796: At low $T$, in addition

5797: to the BAP mechanism, $\tau_s$ will deviate from that in the bulk due

5798: to impurity scattering. The EY and BAP mechanisms were found not

5799: to be relevant to the observed data \cite{Malinowski2000:PRB}.

5800: 

5801: \begin{figure}

5802: \centerline{\psfig{file=zutic_fig21.eps,width=1\linewidth,angle=0}}

5803: \caption{Measured room-temperature dependence of $1/\tau_s$ on the

5804: confinement

5805: energy $E_1$ for GaAs/AlGaAs QW's. The solid line is a quadratic fit, 

5806: showing 

5807: behavior

5808: consistent with the DP mechanism. From \onlinecite{Malinowski2000:PRB}.} 

5809: \label{fig:QW2}

5810: \end{figure}

5811: 

5812: Figure \ref{fig:QW2} shows the dependence of $1/\tau_s$ on the

5813: experimentally determined confinement energy $E_1$ for a variety of

5814: QW's on the same wafer \cite{Malinowski2000:PRB}.  The data are 

5815: at room

5816: temperature. The spin relaxation time 

5817: varies from somewhat less than

5818: 100 ps for wide QW's, approximating the bulk data (cf.

5819: \textcite{Kimel2001:PRB} where 15 ps $< \tau_s < 35$ ps 

5820: was found for a heavily

5821: doped n-GaAs), to about 10 ps in most confined structures. The

5822: downturn for the highest-$E_1$ well (of width 3 nm) is most likely due

5823: to the increased importance of interface roughness at such small

5824: widths \cite{Malinowski2000:PRB}.  Confinement strongly enhances spin

5825: relaxation. This is consistent with the DP mechanism for

5826: two-dimensional systems, in which the spin precession about the intrinsic

5827: magnetic fields 

5828: (here induced by bulk inversion asymmetry) 

5829: increases as 

5830: $E_1^2$ with increasing

5831: confinement. The observed data in Fig.~\ref{fig:QW2} are consistent

5832: with the theoretical prediction.

5833: 

5834: Similarly to bulk GaAs, spin relaxation in GaAs QW's was found to be reduced at 

5835: carrier

5836: concentrations close to the metal-to-insulator transition

5837: ($n\approx 5\times 10^{10}$ cm$^{-2}$) \cite{Sandhu2001:PRL}.

5838: 

5839: 

5840: \section{\label{sec:IV} Spintronic devices and applications}

5841: %=========================================================================

5842: 

5843: In this section we focus primarily on the physical principles 

5844: and materials issues

5845: for various device schemes, which, while not yet commercially viable, are likely

5846: to influence future spintronic research and possible applications.

5847: 

5848: 

5849: \subsection{\label{sec:IVA} Spin-polarized transport}

5850: %--------------------------------------------------------------------------

5851: 

5852: \subsubsection{\label{sec:IVA1} F/I/S tunneling}

5853: %--------------------------------------------------------------------------

5854: 

5855: 

5856: Experiments reviewed by \textcite{Tedrow1994:PR} in

5857: ferromagnet/insulator/superconductor (F/I/S) junctions have

5858: established a sensitive technique for measuring the spin polarization $P$

5859: of magnetic thin films and, at the same time, 

5860: has demonstrated that the

5861: current will remain spin-polarized after tunneling through an

5862: insulator. 

5863: These experiments also stimulated more recent

5864: imaging techniques based on the spin-polarized STM  

5865: (see \textcite{Johnson1990:JAP,Wiesendanger1990:PRL}; and a review,

5866: \textcite{Wiesendanger:1998}) 

5867: with the ultimate goal of imaging spin

5868: configurations down to the atomic level.

5869: 

5870: The degree of spin polarization 

5871: is important for many 

5872: applications such as determining the magnitude of tunneling magnetoresistance

5873: (TMR) in magnetic tunnel junctions (MTJ) [recall Eq.~(\ref{eq:julliere})].

5874: Different probes for spin polarization generally

5875: can measure significantly different values even in experiments

5876: performed on the same homogeneous sample. 

5877: In an actual MTJ, measured polarization is {\it not} an intrinsic

5878: property of the F region and could depend on interfacial

5879: properties and the choice of insulating barrier.  Challenges in 

5880: quantifying $P$, discussed here in the context of F/I/S tunneling, even

5881: when F is a simple ferromagnetic metal, should serve as a caution for

5882: studies of novel, more exotic, spintronic materials.

5883: 

5884: F/I/S tunneling conductance is shown in 

5885: Fig.~\ref{tunnel}, where for simplicity we assume that the spin-orbit

5886: and spin-flip scattering (see Sec.~\ref{sec:IIIC})  

5887: can be neglected, a good approximation for

5888: Al$_2$O$_3$/Al \cite{Tedrow1971:PRLb,Tedrow1994:PR}, a common choice

5889: for I/S regions.  For each spin the normalized BCS density of states

5890: is ${\tilde {\cal N}}_S(E)=Re(|E|/2\sqrt{E^2-\Delta^2})$, where E is the

5891: quasiparticle excitation energy and $\Delta$ the superconducting

5892: gap.\footnote{Here we focus on a conventional $s$-wave superconductor

5893: with no angular dependence in $\Delta$.} The BCS density of states

5894: is split in a

5895: magnetic field H, applied parallel to the interface, due to a shift in

5896: quasiparticle energy $E \rightarrow E \pm \mu_B H$, for $\uparrow$

5897: ($\downarrow$) spin parallel (antiparallel) to the field, where

5898: $\mu_B$ is the Bohr magneton.  The tunneling conductance is normalized

5899: with respect to its normal state value--for an F/I/N junction,

5900: $G(V)\equiv (dI/dV)_S/(dI/dV)_N=G_\uparrow(V)+G_\downarrow(V)$, where

5901: $V$ is the applied bias. This conductance can be expressed 

5902: by generalizing analysis of

5903: \textcite{Giaever1961:PR} as

5904: \begin{eqnarray}

5905: G(V)&=&\int^{\infty}_{-\infty}\frac{1+P}{2} 

5906: \frac{{\tilde {\cal N}}_{S}(E+\mu H) \beta dE}

5907: {4 \cosh^2[\beta(E+qV)/2]}  \\ \nonumber  

5908: &+& \int^{\infty}_{-\infty} \frac{1-P}{2} 

5909: \frac{{\tilde{\cal N}}_{S}(E-\mu H) \beta dE}

5910: {4 \cosh^2[\beta(E+qV)/2]}. 

5911: \label{eq:tg}

5912: \end{eqnarray}

5913: 

5914: Here $\beta=1/k_BT$, $k_B$ is the Boltzmann constant, $T$ is the temperature, 

5915: and $q$ is the proton charge. The factors $(1\pm P)/2$ represent 

5916: the difference in

5917: tunneling probability 

5918: between for 

5919: $\uparrow$ and $\downarrow$ electrons.

5920: While a rigorous determination of $P$,

5921: in terms of materials parameters,

5922: would require a full calculation

5923: of spin-dependent tunneling, including the appropriate boundary conditions

5924: and a detailed understanding of the interface properties, it is

5925: customary to make some simplifications. Usually $P$ can be identified as 

5926: \cite{Worledge2000:PRB,Maekawa:2002}

5927: \begin{equation}

5928: P \rightarrow P_G=(G_{N\uparrow} -G_{N\downarrow})/

5929: (G_{N\uparrow}+G_{N\downarrow}),

5930: \label{eq:PG}

5931: \end{equation}

5932: the spin polarization of the normal-state conductance (proportional

5933: to the weighted average of the density of states in F and S and 

5934: the square of the  

5935: tunneling matrix element), 

5936: where $\uparrow$ is the electron spin with the magnetic moment parallel 

5937: to the applied field (majority electrons in F). 

5938: With the further simplification of spin-independent and constant 

5939: tunneling matrix element \cite{Tedrow1971:PRLa,Tedrow1994:PR},

5940: Eq.~(\ref{eq:PG}) can be expressed as

5941: \begin{equation}

5942: P \rightarrow P_{\cal N}=({\cal N}_{F\uparrow} -{\cal N}_{F\downarrow})/

5943: ({\cal N}_{F\uparrow} + {\cal N}_{F\downarrow}),

5944: \label{eq:PF}

5945: \end{equation}

5946: the spin polarization of the tunneling density of states in the F region 

5947: at the Fermi level.

5948: 

5949: \begin{figure} 

5950: \centerline{\psfig{file=zutic_fig22.eps,width=\linewidth,angle=0}}

5951: \caption{Ferromagnet/insulator/superconductor tunneling in an applied magnetic

5952: filed: (a) Zeeman splitting of the BCS density of states as a function of 

5953: applied bias; 

5954: (b) normalized spin-resolved conductance (dashed lines) and the total 

5955: conductance (solid line)

5956: at finite temperature.} 

5957: \label{tunnel}

5958: \end{figure}

5959: Spin polarization $P$ of the F electrode can be deduced \cite{Tedrow1994:PR} 

5960: from the asymmetry

5961: of the conductance amplitudes at the four peaks in Fig.~\ref{tunnel} (b) 

5962: [for $P$=0, G(V)=G(-V)].

5963: In CrO$_2$/I/S tunnel junctions, nearly complete

5964: spin polarization $P_G>0.9$ was measured \cite{Parker2002:PRL}. 

5965: Only two of the four peaks sketched

5966: in Fig.~\ref{tunnel}, have been observed, indicating no features due to the

5967: minority spin up to H=2.5 T. \textcite{Parkin2004:P} have shown that

5968: by replacing an aluminum oxide (a typical choice for an insulating 

5969: region) with magnesium oxide, one can significantly increase the 

5970: spin polarization in F/I/S junctions. Correspondingly, extraordinarily large

5971: values of TMR ($>200\%$ at room temperature) can be achieved even

5972: with conventional ferromagnetic CoFe) electrodes.

5973: 

5974: The assumption of spin-conserving tunneling can be generalized 

5975: \cite{Tedrow1994:PR,Worledge2000:PRB,Monsma2000:APLa,Monsma2000:APLb}  

5976: to extract $P$ in the presence of spin-orbit and spin-flip scattering. 

5977: Theoretical 

5978: analyses \cite{Fulde1973:AP,Maki1964:PTP,Bruno1973:PRB} using 

5979: many-body techniques

5980: show that the  spin-orbit scattering would smear the Zeeman-split 

5981: density of states, eventually merging the

5982: four peaks into two, while the magnetic impurities \cite{Abrikosov1960:ZETP}

5983: act as pair breakers and reduce the value of $\Delta$. 

5984: Neglecting the spin-orbit

5985: scattering was shown to lead to the extraction of higher $P$ 

5986: values \cite{Tedrow1994:PR,Monsma2000:APLa}.

5987: 

5988: With a few exceptions \cite{Worledge2000:PRL},

5989: F/I/S conductance measurements \cite{Tedrow1994:PR}

5990: have revealed positive $P$--the dominant contribution of 

5991: majority spin electrons for different ferromagnetic films 

5992: (for example, in Fe, Ni, Co and Gd). However, electronic

5993: structure calculations typically give  

5994: that ${\cal N}_{F\uparrow}<{\cal N}_{F\downarrow}$ and $P_{\cal N}<0$

5995: [for Ni and Co  ${\cal N}_{F\uparrow}/{\cal N}_{F\downarrow} \sim 1/10$ 

5996: \cite{Butler2001:PRB}].

5997: Early theoretical work addressed this apparent

5998: difference,\footnote{For a list of references see 

5999: \textcite{Tedrow1973:PRB,Tedrow1994:PR}.}

6000: and efforts to understand precisely what is being experimentally measured 

6001: have continued.

6002: 

6003: \textcite{Stearns1977:JMMM} suggested that only itinerant, 

6004: freelike electrons

6005: will contribute to tunneling, while nearly localized electrons, 

6006: with a large 

6007: effective mass,

6008: contribute to the total density of states but not to G(V) 

6009: [see also \textcite{Hertz1973:PRB}

6010: and, for spin-unpolarized tunneling, \textcite{Gadzuk1969:PR}].

6011: From the assumed parabolic dispersion of the spin subbands with fixed spin 

6012: splitting, Stearns related the measured polarization 

6013: to the magnetic 

6014: moment, giving positive 

6015: $P \rightarrow P_k=(k_{F\uparrow}-k_{F\downarrow})/(k_{F\uparrow}

6016: +k_{F\downarrow})$,

6017: the spin polarization of the projections of Fermi wave vectors 

6018: perpendicular

6019: to the interface.

6020: Similar arguments, for inequivalent 

6021: density-of-states contributions  

6022: to G(V), were generalized

6023: to more complex electronic structure.

6024: \textcite{Mazin1999:PRL} showed 

6025: the importance of the tunneling matrix elements which have

6026: different Fermi velocities for different bands 

6027: [see also \cite{Yusof1998:PRB},

6028: in the context of tunneling in 

6029: a high temperature superconductor (HTSC)]. 

6030: Consequently, $P_G$ could even have an opposite sign from $P_{\cal N}$--which, 

6031: for example, would be measured by spin-resolved photoemission. 

6032: 

6033: Good agreement between  

6034: tunneling data and electronic structure calculations was illustrated by 

6035: the example of Ni$_x$Fe$_{1-x}$ \cite{Nadgorny2000:PRB}, showing, however, 

6036: that $P$ is not directly related to the  magnetic moment 

6037: \cite{Meservey1976:PRL}. 

6038: The difference between bulk and the surface densities of states 

6039: of the ferromagnet 

6040: (probed in tunneling measurements) \cite{Oleinik2000:PRB}, 

6041: the choice of tunneling barrier \cite{DeTeresa1999:PRL}, and details of the

6042: interfacial properties, which can change over time \cite{Monsma2000:APLb}, have

6043: all been shown to affect the measured $P$ directly. 

6044: 

6045: Tedrow-Meservey technique is also considered as a

6046: probe to detect spin injection in Si, where optical methods, due to

6047: the indirect gap, would be ineffective. F/I/S tunneling was also studied

6048: using amorphous Si (a-Si) and Ge (a-Ge) as a barrier. While 

6049: with a-Si  some spin polarization was detected \cite{Meservey1982:JAP} 

6050: no  spin-polarized tunneling was observed using a-Ge  barrier 

6051: \cite{Gibson1985:JAP},

6052: in contrast to the first reports of TMR \cite{Julliere1975:PL}.

6053: 

6054: Spin-dependent tunneling was also studied using a 

6055: HTSC

6056: electrode as a detector

6057: of spin polarization \cite{Vasko1998:APL,Chen2001:PRB}. 

6058: While this can significantly extend the temperature

6059: range in the tunneling experiments, a lack of  understanding of 

6060: HTSC's makes such structures more a test ground for fundamental physics

6061: than a quantitative tool for 

6062: quantitatively 

6063: determining

6064: $P$.

6065: There are also several important differences between studies using

6066: HTSC's and 

6067: conventional low-temperature superconductors. The superconducting

6068: pairing symmetry no longer yields an isotropic energy gap, and even 

6069: for the BCS-like picture the density of states should be accordingly modified. 

6070: A sign change of the pair potential can result in $G(V=0) >0$

6071: for $T\rightarrow 0$ even for a strong tunneling barrier

6072: and give rise to a zero-bias conductance peak 

6073: \cite{Hu1994:PRL,Kashiwaya1995:PRL,Wei1998:PRL}.

6074: This is explained

6075: by the two-particle process of Andreev reflection (discussed further in

6076: Sec.~\ref{sec:IVA3}),

6077: which, in addition to the usual quasiparticle tunneling, 

6078: contributes to the I-V characteristics of a F/I/S junction 

6079: \cite{Zutic1999:PRBb,Zhu1999:PRB,Zutic2000:PRB,Kashiwaya1999:PRB,Hu1999:PRBb}

6080: [a simpler N/I/S case is reviewed by \textcite{Hu1998:PRB} and

6081: \textcite{Kashiwaya2000:RPP}]. 

6082: The suppression of a zero-bias conductance peak, measured by an STM,

6083: was recently used to detect spin injection into a HTSC \cite{Ngai2003:P}.

6084: 

6085: \subsubsection{\label{sec:IVA2} F/I/F tunneling}

6086: %------------------------------------------------------------------------

6087: 

6088: 

6089: In the preface to a now classic 

6090: reference on spin-unpolarized tunneling

6091: in solids, \textcite{Duke:1969} concludes that (with only a few exceptions)

6092: the study of tunneling is an art and not a science. 

6093: Perhaps this is also an apt 

6094: description

6095: for the 

6096: present state of experiment on spin-polarized tunneling between 

6097: two ferromagnetic

6098: regions. Even for MTJ's with standard ferromagnetic metals, the bias and the

6099: temperature dependence of the TMR, as identification of the relevant spin 

6100: polarization

6101: remain to be fully understood. In a brief review of current findings we

6102: intend to identify questions that could arise as new materials for

6103: MTJ's are being considered.

6104: 

6105: A resurgence in interest in the study of MTJ's, following a hiatus after the

6106: early work by \textcite{Julliere1975:PL,Maekawa1982:IEEE}, was 

6107: spurred by the 

6108: observation of large room-temperature TMR 

6109: \cite{Moodera1995:PRL,Miyazaki1995:JMMM}.

6110: This discovery has 

6111: opened the possibility of using MTJ's for fundamental studies

6112: of surface magnetism and room-temperature spin polarization in various 

6113: ferromagnetic electrodes 

6114: as well as to suggesting applications such as highly sensitive 

6115: magnetic-field sensors, magnetic read heads, 

6116: and nonvolatile magnetic memory applications. 

6117: 

6118: It is instructive to notice the similarity between the schematic geometry 

6119: and the direction of current flow in  an MTJ and that in CPP GMR 

6120: (recall Figs.~\ref{intro:1} and 

6121: \ref{gmr:1}),

6122: which only differ in the middle layer being an insulator and a metal, 

6123: respectively.

6124: By considering the limit of ballistic transport in 

6125: CPP GMR\footnote{Related

6126: applications are usually in a diffusive regime.}

6127: it is possible to give a unified picture of both TMR and CPP GMR by

6128: varying  the strength of the hopping integrals \cite{Mathon1997:PRB} in a

6129: tight-binding representation.

6130: 

6131: \textcite{Julliere1975:PL} modified Eq.~(\ref{eq:tg}) in the limit

6132: $V \rightarrow 0$, $T \rightarrow 0$,

6133: and applied it to study F/I/F tunneling. 

6134: The two F regions are treated as uncoupled with the spin-conserving 

6135: tunneling across the barrier. This effectively leads to the two-current

6136: model proposed by \textcite{Mott1936:PRCa} and also applied to 

6137: the CPP GMR geometries \cite{Valet1993:PRB,Gijs1997:AP}.

6138: The values for $P$ extracted

6139: from F/I/S measurements are in a good agreement with the observed 

6140: TMR  values (typically positive, as expected from P$_{1,2}>0$). 

6141: However, Julli{\`{e}}re's formula\footnote{For its limitations and extensions

6142: see comprehensive reviews by \textcite{Moodera1999:JMMM,Moodera1999:ARMS}.}

6143: does not provide an explicit TMR dependence on bias and temperature.

6144: 

6145: Julli{\`{e}}re's result can be obtained as a

6146: limiting case from a more general Kubo/Landauer approach \cite{Mathon1999:PRB}

6147: with the assumption that the component of the wave vector parallel to

6148: the interface ${\bf k}_\|$ is not conserved (incoherent tunneling).

6149: Such a loss of coherence is good approximation for simply capturing the

6150: effects of disorder for amorphous Al$_2$O$_3$, a common choice 

6151: for the I region

6152: with metallic ferromagnets. 

6153: Despite its simplicity, the Julli{\`{e}}re's model for the TMR 

6154: has continued to be used for interpreting

6155: the spin polarization in various MTJ's. 

6156: Recent examples include 

6157: F regions made of manganite perovskites displaying colossal 

6158: magnetoresistance (CMR) \cite{Bowen2002:P} (suggesting $P_{\cal N}$$>$0.95);

6159: magnetite (Fe$_3$O$_4$) \cite{Hu2002:PRL} (with $P$$<$0 and TMR$<$0); 

6160: III-V ferromagnetic semiconductors \cite{Chun2002:PRB}; 

6161: a nonmagnetic semiconductor used as a tunneling barrier 

6162: \cite{Kreuzer2002:APL};

6163: Co/carbon nanotube/Co MTJ \cite{Tsukagoshi1999:N};

6164: and resonant tunneling in F/I/N/F junctions \cite{Yuasa2002:S}.

6165: 

6166: For novel materials, in which the electronic structure calculations and an

6167: understanding of the interfacial properties are not available, 

6168: Julli{\`{e}}re's 

6169: formula 

6170: still provides useful insights.

6171: A quantitative understanding of MTJ's challenges similar to those 

6172: discussed for F/I/S tunneling, including determining precisely which spin 

6173: polarization is relevant and the related issue of reconciling the 

6174: (typically positive) sign 

6175: of  the observed TMR with the electronic structure 

6176: \cite{Tsymbal1997:JPCM,Oleinik2000:PRB,Bratkovsky1997:PRB,MacLaren1997:PRB,%

6177: Mathon1997:PRB,LeClair2002:PRL}.  

6178: 

6179: In an approach complementary to Julli{\`{e}}re's, 

6180: \textcite{Slonczewski1989:PRB} 

6181: considered F/I/F as a single quantum-mechanical system in a free-electron 

6182: picture. 

6183: When matching the two-component wave functions at interfaces, 

6184: coherent tunneling was  

6185: assumed, with conserved ${\bf k}_\|$, relevant to epitaxially grown MTJ's 

6186: \cite{Mathon2001:PRB} and the I region was modeled by a square 

6187: barrier.\footnote{A formally analogous problem was considered by 

6188: \textcite{Griffin1971:PRB} in an N/I/S system where the two-component

6189: wave functions represented electron-like and hole-like quasiparticles 

6190: rather then the two spin projections; see Sec.~\ref{sec:IVA3}.}

6191: The resulting TMR can be expressed as in

6192: Eq.~(\ref{eq:julliere}) but with the redefined polarization 

6193: \begin{equation}

6194: P \rightarrow P_k 

6195: (\kappa^2-k_{F\uparrow}k_{F\downarrow})/(\kappa^2

6196: +k_{F\uparrow}k_{F\downarrow}),

6197: \label{eq:slon}

6198: \end{equation}

6199: where $P_k$, as defined by \textcite{Stearns1977:JMMM}, is also

6200: $P_{\cal N}$ (in a free-electron picture) and $i \kappa$ is the usual imaginary

6201: wave vector through a square barrier. Through the dependence of

6202: $\kappa$ on $V$ the resulting polarization in Slonczewski's model can change

6203: sign. A study of a similar geometry using a Boltzmann-like approach shows

6204: \cite{Chui1997:PRB}

6205: that the spin splitting of electrochemical potentials persists in the F region

6206: all the way to the F/I interface, implying 

6207: $\kappa_\uparrow \neq \kappa_\downarrow$

6208: and an additional voltage dependence of the TMR. Variation of  

6209: the density of states [inferred from the

6210: spin-resolved photoemission data \cite{Park1998:PRL,Park1998:N}]  within the

6211: range of applied bias in MTJ's of Co/SrTiO$_3$/La$_{0.7}$Sr$_{0.3}$MnO$_3$ 

6212: (Co/STO/LSMO) 

6213: \cite{DeTeresa1999:PRL}, 

6214: together with Julli{\`{e}}re's model, was used to explain

6215: the large negative TMR (-50\% at 5 K), 

6216: which would even change sign for positive

6217: bias (raising the Co Fermi level above the corresponding one of LSMO). 

6218: The bias dependence of the TMR was also attributed to the density of states

6219: by extending the

6220: model of a trapezoidal tunneling barrier \cite{Brinkman1970:JAP} to the 

6221: spin-polarized

6222: case \cite{Xiang2002:PRB}. 

6223: 

6224: The decay of TMR with temperature can be attributed to several causes. 

6225: Early theoretical work on N/I/N tunneling 

6226: \cite{Anderson1966:PRL,Appelbaum1966:PRL}

6227: [for a detailed discussion and a review of

6228: related experimental results see \textcite{Duke:1969}]

6229: showed that the presence of magnetic impurities  in the tunneling

6230: barrier produces temperature dependent conductance--referred to as 

6231: zero-bias anomalies. 

6232: These findings, which considered both spin-dependent and 

6233: spin-flip scattering,

6234: were applied to  fit the decay of the TMR with temperature

6235: \cite{Miyazaki:2002,Jansen2000:PRB,Inoue1999:JMMM}.

6236: Hot electrons  localized at F/I interfaces were predicted, 

6237: to create magnons, or collective spin excitations,

6238: near the F/I interfaces, and suppress the TMR \cite{Zhang1997:PRL}.

6239: Magnons were observed \cite{Tsui1971:PRL} in an antiferromagnetic (AFM)

6240: NiO barrier in single crystal Ni/NiO/Pb tunnel junctions and were suggested 

6241: \cite{Moodera1995:PRL} as the cause of

6242: decreasing TMR with $T$ by spin-flip scattering.

6243: Using an $s$-$d$ exchange  (between itinerant $s$ and nearly localized $d$ 

6244: electrons)

6245: Hamiltonian, it was shown \cite{Zhang1997:PRL} that, at  $V\rightarrow 0$,

6246: $G(T)-G(0) \propto T \ln T$, for both 

6247: $\uparrow \uparrow$ and $\uparrow \downarrow$

6248: orientations.

6249: A different temperature dependence of TMR was

6250: suggested by \textcite{Moodera1998:PRL}.

6251: It was related to the decrease of the surface magnetization 

6252: \cite{Pierce1982:PRB,Pierce:1984}

6253: $M(T)/M(0) \propto T^{3/2}$. Such a temperature dependence 

6254: [known as the Bloch's law and reviewed by \textcite{Krey2003:P}],

6255: attributed to magnons, was also obtained for TMR 

6256: \cite{MacDonald1998:PRL}.

6257: An additional decrease of TMR with $T$ was expected due to the spin-independent 

6258: part of G(T)  \cite{Shang1998:PRB}, seen also in N/I/N junctions. 

6259: 

6260: Systematic studies of MTJ's containing a semiconductor (Sm) region

6261: (used as a tunneling barrier and/or  an F electrode) have begun only 

6262: recently.\footnote{The early F/Ge/F results \cite{Julliere1975:PL} were not 

6263: reproduced and other metallic structures involving Si, Ge, GaAs, and GaN 

6264: as a barrier have shown 

6265: either no \cite{Gibson1985:JAP,Loraine2000:JAP,Boeve2001:JMMM} 

6266: or only a small \cite{Meservey1982:JAP,Jia1996:IEEE,Kreuzer2002:APL} 

6267: spin-dependent signal.}

6268: To improve the performance of MTJ's it is desirable to reduce the junction

6269: resistance. A smaller $RC$ constant would allow faster switching 

6270: times in MRAM (for a detailed discussion see \textcite{DeBoeck2002:SST}). 

6271: Correspondingly, using a semiconducting barrier could prove 

6272: an alternative strategy for difficult fabrication of ultrathin 

6273: ($<$1 nm ) oxide barriers 

6274: \cite{Rippard2002:PRL}. Some F/Sm/F MTJ's have been

6275: grown epitaxially, and the amplitude of TMR can be studied as function

6276: of the crystallographic orientation of a F/Sm interface.

6277: For an epitaxially grown Fe/ZnSe/Fe MTJ electronic structure calculations

6278: have predicted  \cite{MacLaren1999:PRB} large TMR (up to $\sim$1000 \%), 

6279: increasing

6280: with ZnSe thickness. However, the observed TMR in Fe/ZnSe/Fe$_{0.85}$Co$_{0.15}$

6281: was limited  below 50 K, reaching 15 \% at 10 K for junctions of higher

6282: resistance and lower defect density\footnote{Interface defects could diminish 

6283: measured TMR. We recall (see Sec.~\ref{sec:IID3}) that at a ZnMnSe/AlGaAs interface 

6284: they limit the spin injection efficiency \cite{Stroud2002:PRL} and from 

6285: Eq.~(\ref{eq:delR}) infer a reduced spin-valve effect.}

6286: \cite{Gustavsson2001:PRB}. 

6287: Results on ZnS, 

6288: another II-VI semiconductor, demonstrated a TMR of $\sim$5 \% at room 

6289: temperature \cite{Guth2001:APL}.

6290: 

6291: \begin{figure} 

6292: \centerline{\psfig{file=zutic_fig23.eps,width=0.9\linewidth}}

6293: \caption{All-semiconductor magnetic tunnel junction:

6294: (a) magnetization of Ga$_{1-x}$Mn$_x$As (x=4.0\%,50 nm)/AlAs

6295: (3 nm)/ Ga$_{1-x}$Mn$_x$As (x=3.3\%,50 nm) trilayer measured by a SQUID at 8 K.

6296: The sample size is 3 $\times$ 3 mm$^2$. 

6297: Magnetization shown is normalized with respect to the saturation value M$_s$.

6298: (b) 

6299: TMR curves of a Ga$_{1-x}$Mn$_x$As (x=4.0\%,50 nm)/AlAs

6300: (1.6 nm)/ Ga$_{1-x}$Mn$_x$As (x=3.3\%,50 nm) tunnel junction of 200 $\mu$m

6301: in diameter. Bold solid curve, 

6302: sweep of the magnetic

6303: field from positive to negative;

6304: dashed curve, sweep from

6305: negative to positive; thin solid curve,

6306: a minor loop. From \onlinecite{Tanaka2001:PRL}.}

6307: \label{tmr0}

6308: \end{figure}

6309: 

6310: There is also a possibility of using 

6311: all-semiconductor F/Sm/F single-crystalline MTJ's

6312: where F is a ferromagnetic semiconductor. These would simplify integration

6313: with the existing conventional semiconductor-based electronics and allow

6314: flexibility of various doping profiles and fabrication of quantum structures,

6315: as compared to the conventional all-metal MTJ's. Large TMR ($>$70 \% at 8 K),

6316: shown in Fig.~\ref{tmr0},

6317: has been measured in an epitaxially grown (Ga,Mn)As/AlAs/(Ga,Mn)As junction 

6318: \cite{Tanaka2001:PRL}. The results are consistent with the $k_\|$ being 

6319: conserved in the tunneling process \cite{Mathon1997:PRB}, with the decrease of 

6320: TMR with $T$ expected from the spin-wave excitations 

6321: \cite{Shang1998:PRB,MacDonald1998:PRL},

6322: discussed above. TMR is nonmonotonic with thickness in AlAs 

6323: (with the peak at

6324: $\sim$ 1.5 nm). For a given AlAs thickness, double MTJ's were also shown

6325: to give similar TMR values and were used to determine electrically

6326: the spin injection in GaAs QW \cite{Mattana2002:PRL}.

6327: However, 

6328: a room-temperature effect 

6329: remains to be demonstrated 

6330: as 

6331: the available

6332: well-characterized ferromagnetic semiconductors do not have 

6333: as high a ferromagnetic

6334: transition temperature. 

6335: 

6336: A lower barrier in F/Sm/F MTJ's can have important implications in 

6337: determining the actual values of TMR. The standard four-probe technique for

6338: measuring I and V 

6339: has been known to give spurious values when the resistance 

6340: of the F electrodes is non-negligible to the junction resistance.

6341: The tunneling current in that regime has been shown to be highly 

6342: nonuniform\footnote{Nonuniform tunneling current has been studied

6343: in nonmagnetic junctions \cite{Pederson1967:APL}, 

6344: CPP multilayers 

6345: \cite{Lenczowski1994:JAP},    

6346: and conventional MTJ's \cite{Moodera1996:APL,Rzchowski2000:PRB}.} 

6347: and the measured apparent resistance $R_m=V/I$  (different from the

6348: actual junction resistance R$_J$)  can even attain negative values 

6349: \cite{Pederson1967:APL,Moodera1996:APL}. The important implications

6350: for MTJ are the possibility of large overestimates in the TMR amplitude

6351: \cite{Moodera1999:JMMM} and a desirable hysteresis effect--at H=0

6352: the two values of resistance can be used for various nonvolatile 

6353: applications \cite{Moodera1996:APL}. 

6354: 

6355: A detailed understanding of MTJ's will also require knowing the influence

6356: of the interface and surface roughness \cite{Itoh1999:JPSP}. Even in 

6357: the spin-unpolarized case it is known that the full quantum-mechanical 

6358: approach 

6359: \cite{Tesanovic1986:PRL} can lead to qualitatively different 

6360: results from the 

6361: usual quasiclassical picture and from averaging out the spatial information on the 

6362: length scale of the inverse Fermi wave vector.

6363: 

6364: A comprehensive review of tunneling phenomena and magnetoresistance in 

6365: granular materials, ferromagnetic single-electron transistors, 

6366: and double tunnel junctions

6367: is given by \textcite{Maekawa:2002}. 

6368: A theoretical study of F/I/F junctions,

6369: in which the I region is a quantum dot,

6370: shows the importance of Coulomb interactions, which could lead to spin 

6371: precession

6372: even in the absence of an applied magnetic field \cite{Konig2003:PRL}.

6373: 

6374: \subsubsection{\label{sec:IVA3} Andreev reflection}

6375: %------------------------------------------------------------------------

6376: 

6377: 

6378: Andreev reflection \cite{Andreev1964:SPJETP} 

6379: is a scattering process, at an interface with a superconductor, 

6380: responsible for a conversion between a dissipative quasiparticle

6381: current and a dissipationless supercurrent [see also early work

6382: by \textcite{deGennes1963:PL}]. 

6383: For a spin-singlet superconductor

6384: an incident electron (hole) of spin

6385: $\lambda$ is reflected as a

6386: hole (electron) belonging to the opposite spin subband $\overline{\lambda}$, 

6387: back

6388: to the nonsuperconducting region, while a Cooper pair is transferred to the 

6389: superconductor. 

6390: This is a phase-coherent scattering process in which the reflected particle

6391: carries the information about both the phase of 

6392: the incident particle

6393: and the macroscopic phase of the superconductor.\footnote{For instructive

6394: reviews see \textcite{Lambert1998:JPCM,Pannetier2000:JLTP}.} Andreev

6395: reflection  thus is responsible for a proximity effect where the 

6396: phase correlations are introduced to a nonsuperconducting 

6397: material \cite{Demler1997:PRB,Halterman2002:PRB,Bergeret2001:PRL,%

6398: Izyumov2002:PU,Fominov2003:T}.

6399: The probability for Andreev reflection at low bias voltage 

6400: ($qV \lesssim \Delta$), which is 

6401: related to the square of the normal-state transmission, 

6402: could

6403: be ignored for  low-transparency junctions with conventional

6404: superconductors, as discussed in Sec.~\ref{sec:IVA1}. In contrast, 

6405: for high-transparency junctions (see the discussion 

6406: of Sharvin conductance in Sec.~\ref{sec:IIC2}),

6407: single-particle tunneling vanishes [recall Eq.~(\ref{eq:tg})]

6408: at low bias and $T=0$ and Andreev

6409: reflection is the dominant process. A convenient description is

6410: provided by the 

6411: Bogoliubov-de Gennes equations \cite{deGennes:1989},

6412: \begin{eqnarray}

6413: \left[\begin{array}{cc} H_\lambda & \Delta \\

6414: \Delta^* & -H^*_{\overline{\lambda}}\end{array} \right]

6415: \left[ \begin{array}{c} u_\lambda \\

6416: v_{\overline{\lambda}} \end{array} \right]=E

6417: \left[ \begin{array}{c} u_\lambda \\

6418: v_{\overline{\lambda}} \end{array} \right],

6419: \label{eq:BdG}

6420: \end{eqnarray} 

6421: and by matching the wave functions at the boundaries (interfaces) between

6422: different regions.

6423: Here $H_\lambda$ is the single-particle Hamiltonian 

6424: for spin $\lambda=\uparrow, \downarrow$ and  $\overline{\lambda}$

6425: denotes a spin opposite to $\lambda$ \cite{deJong1995:PRL,Zutic2000:PRB}.

6426: $\Delta$ is the pair potential \cite{deGennes:1989},  $E$ the excitation energy 

6427: and $u_\lambda$, $v_{\overline{\lambda}}$ are the electronlike quasiparticle 

6428: and holelike quasiparticle amplitudes, 

6429: respectively.\footnote{Equation (\ref{eq:BdG}) 

6430: can be simply modified to include the spin flip and

6431: spin-dependent interfacial scattering \cite{Zutic1999:PRBa}.}   

6432: \textcite{Griffin1971:PRB} have solved the Bogoliubov-de Gennes 

6433: equations with 

6434: square or a $\delta$-function barriers  of  varying strength 

6435: at an N/S interface. They obtained a  

6436: result that interpolates between the clean and the tunneling limits.

6437: \textcite{Blonder1982:PRB} used a similar approach,

6438: known as the Blonder-Tinkham-Klapwijk method, in which 

6439: the two limits correspond to $Z\rightarrow0$ and $Z\rightarrow \infty$,

6440: respectively, and $Z$ is the strength of the  $\delta$-function barrier.

6441: The transparency of this

6442: approach\footnote{A  good agreement \cite{Yan2000:PRB}

6443: was obtained with the more rigorous nonequilibrium 

6444: Keldysh technique \cite{Keldysh1965:SPJETP,Rammer1986:RMP}, for an illustration 

6445: of how such a technique can be used to study spin-polarized transport in a 

6446: wide range of heterojunctions see \textcite{Melin2002:EPJB,Zeng2003:P}.}

6447: makes it suitable for the study of  

6448: ballistic spin-polarized transport and spin injection even in the

6449: absence of a  superconducting region

6450: \cite{Hu2001:PRL,Matsuyama2002:PRB,Hu2001:PRB,Schapers2001b:PRB}.

6451: 

6452: It is instructive to note a similarity between the

6453: two-component transport in N/S junctions  (for

6454: electronlike and holelike quasiparticles) and F/N junctions 

6455: (for spin $\uparrow$, $\downarrow$), which both lead to current conversion, 

6456: accompanied by the additional boundary resistance

6457: \cite{Blonder1982:PRB,vanSon1987:PRL}.  

6458: In the N/S junction Andreev reflection is responsible for the conversion 

6459: between the 

6460: normal and the 

6461: supercurrent, characterized by the superconducting coherence length,

6462: while in the F/N case a conversion between spin-polarized and unpolarized

6463: current is characterized by the spin diffusion length.

6464: 

6465: For spin-polarized carriers, with  different populations in two spin

6466: subbands, only a fraction of the incident electrons from a majority subband

6467: will have a  minority subband partner in order to be  Andreev

6468: reflected. This can be simply quantified at zero bias and $Z=0$,

6469: in terms of the total number of scattering channels (for each $k_\|$) 

6470: $N_\lambda=k_{F\lambda}^2 A/4 \pi$ at the Fermi level. Here A is the 

6471: point-contact 

6472: area, and $k_{F\lambda}$ 

6473: is the spin-resolved Fermi wave vector. 

6474: A spherical Fermi surface in the F and S regions, with no (spin-averaged)

6475: Fermi velocity mismatch, is assumed. 

6476: When S is in the normal state, the 

6477: zero-temperature Sharvin conductance is

6478: \begin{equation}

6479: G_{FN}=\frac{e^2}{h}(N_\uparrow+N_\downarrow),

6480: \label{eq:sharvin2} 

6481: \end{equation}

6482: equivalent to $R_{\rm Sharvin}^{-1}$, from Eq.~(\ref{eq:sharvin}).

6483: In the superconducting state all of the $N_\downarrow$ 

6484: and only $(N_\downarrow/N_\uparrow)N_\uparrow$ scattering channels 

6485: contribute to Andreev reflection across the F/S interface and transfer 

6486: charge $2e$, yielding \cite{deJong1995:PRL}

6487: \begin{equation}

6488: G_{FS}=\frac{e^2}{h} \left(2N_\downarrow + \frac{2N_\downarrow}{N_\uparrow} 

6489: N_\uparrow \right)=4\frac{e^2}{h}N_\downarrow.

6490: \label{eq:dejong}

6491: \end{equation}

6492: The suppression of the normalized zero-bias conductance 

6493: at $V=0$ and $Z=0$ \cite{deJong1995:PRL},

6494: \begin{equation}

6495: G_{FS}/G_{FN}=2(1-P_G)

6496: \label{eq:pc}

6497: \end{equation}

6498: with the increase in the spin polarization $P_G \rightarrow 

6499: (N_\uparrow-N_\downarrow)/(N_\uparrow+N_\downarrow)$,

6500: was used as a sensitive transport technique to detect spin polarization 

6501: in a point contact  \cite{Soulen1998:S}. 

6502: Data are 

6503: given in Fig.~\ref{point}. A

6504: similar  study, using a  thin-film nanocontact geometry 

6505: \cite{Uphaday1998:PRL}, emphasized the importance of

6506: fitting the conductance data over a wide range of applied bias, not only

6507: at $V=0$, in order to extract the spin polarization of the F region more 

6508: precisely.

6509: 

6510: \begin{figure} 

6511: \centerline{\psfig{file=zutic_fig24.eps,width=\linewidth,angle=0}}

6512: \caption{The differential conductance for several spin-polarized materials,

6513: showing the suppression of Andreev reflection with increasing $P_G$.

6514: The vertical lines denote the bulk superconducting gap for Nb: 

6515: $\Delta$($T=0$)=1.5 meV. 

6516: Note that NiMnSb, one of the Heusler alloys originally proposed 

6517: as half-metallic ferromagnets \cite{deGroot1983:PRL}, shows only

6518: partial spin polarization. From \onlinecite{Soulen1998:S}.} 

6519: \label{point}

6520: \end{figure}

6521: The advantage of such techniques is the detection of polarization

6522: in a much wider range of materials than  those which

6523: can be grown for detection in F/I/S or F/I/F tunnel junctions. 

6524: A large number of

6525: experimental results using spin-polarized Andreev reflection

6526: has since been reported 

6527: \cite{Nadgorny2001:PRB,Ji2001:PRL,Parker2002:PRL,Bourgeois2001:PRB,%

6528: Panguluri2003:Pb},  

6529: including the first direct measurements 

6530: \cite{Braden2003:P,Panguluri2003:P,Panguluri2004:P} of

6531: the spin polarization in (Ga,Mn)As and (In,Mn)Sb.\footnote{Similar 

6532: measurements were

6533: also suggested by \textcite{Zutic1999:PRBa} to yield information 

6534: about the FSm/S interface. A more complete analysis should also quantify

6535: the effects of spin-orbit coupling.}

6536: However,

6537: for a quantitative interpretation of the measured polarization, important

6538: additional factors (similar to the limitations discussed for the application

6539: of Julli{\`{e}}re's formula in Sec.~\ref{sec:IVA2}) need to be incorporated 

6540: in the picture 

6541: provided by Eq.~(\ref{eq:pc}).

6542: For example, the Fermi surface may not be spherical [see the discussion of 

6543: \textcite{Mazin1999:PRL}, specifying what type of spin polarization is

6544: experimentally measured and also that of \textcite{Xia2002:PRL}]. 

6545: The roughness or the size of the  F/S interface 

6546: may lead to a diffusive component of the transport 

6547: \cite{Jedema1999:PRB,Falko1999:PZETF,Mazin2001:JAP}.

6548: As a caution concerning the  possible difficulties in analyzing experimental data,

6549: we mention some subtleties that arise even for the simple model

6550: of a spherical Fermi surface used to describe both F and S regions.

6551: Unlike charge transport in N/S junctions \cite{Blonder1983:PRB} 

6552: in a Griffin-Demers-Blonder-Tinkham-Klapwijk approach, Fermi velocity mismatch

6553: between the F and the S regions, 

6554: does not simply increase the value of effective $Z$.

6555: Specifically, at $Z=V=0$ and normal incidence

6556: it is possible to have perfect transparency even when all the

6557: Fermi velocities differ, satisfying 

6558: $(v_{F\uparrow} v_{F\downarrow})^{1/2}=v_S$,

6559: where $v_S$ is the Fermi velocity in a superconductor 

6560: \cite{Zutic1999:PRBa,Zutic1999:PRBb,Zutic2000:PRB}. In other words, 

6561: unlike in Eq.~(\ref{eq:pc}), the spin polarization 

6562: (nonvanishing exchange energy) 

6563: can {\it increase} the subband conductance, 

6564: for fixed Fermi velocity mismatch. 

6565: Conversely, at a fixed exchange energy, an increase in Fermi velocity mismatch  

6566: could increase the subgap conductance.\footnote{Similar results were 

6567: also obtained when 

6568: F and S region were separated with a quantum dot 

6569: \cite{Zhu2002:PRB,Feng2003:PRB,Zeng2003:P} 

6570: and even in a 1D tight-binding model with  no spin polarization 

6571: \cite{Affleck2000:PRB}.}

6572: In a typical interpretation of a measured conductance, complications can

6573: then arise in trying to disentangle the influence of parameters $Z$, $P_G$,

6574: and Fermi velocity mismatch from the nature of the point contacts 

6575: \cite{Kikuchi2002:PRB} 

6576: and the role of inelastic scattering \cite{Auth2003:PRB}.

6577: Detection of $P$ in HTSC's is even possible with a large barrier or a

6578: vacuum between the F and S regions, 

6579: as proposed by \textcite{Wang2002b:P} 

6580: using resonant Andreev reflection and 

6581: a $d$-wave superconductor.\footnote{Interference effects between the

6582: quasi-electron and quasi-hole scattering trajectories that feel

6583: pair potentials of different sign lead to a large 

6584: conductance near zero bias, even at large interfacial barrier

6585: (referred to as a zero-bias conductance peak in Sec.~\ref{sec:IVA1}).

6586: } 

6587: 

6588: Large magnetoresistive effects are predicted for crossed Andreev reflection

6589: \cite{Deutscher2000:APL}, when the two F regions, separated within the 

6590: distance of

6591: the superconducting coherence length,\footnote{Recent theoretical

6592: findings suggest that the separation should not exceed the 

6593: Fermi wavelength \cite{Yamashita2003:P}.} are on the same side of the

6594: S region. Such structures have also been theoretically studied to

6595: understand the implications of nonlocal correlations 

6596: \cite{Apinyan2002:EPJB,Melin2002:EPJB}. 

6597: 

6598: \subsubsection{\label{sec:IVA4} Spin-polarized drift and diffusion}

6599: %------------------------------------------------------------------------------

6600: 

6601: Traditional semiconductor devices such as field-effect transistors, 

6602: bipolar diodes and transistors, or semiconductor solar cells rely 

6603: in great part on carriers (electrons and holes) whose motion

6604: can be described as drift and diffusion, limited by carrier recombination. In 

6605: inhomogeneous devices where charge buildup is rule, the 

6606: recombination-limited 

6607: drift-diffusion is supplied by Maxwell's equations, to be solved in a 

6608: self-consistent manner. 

6609: Many proposed spintronic devices as well as experimental 

6610: settings for spin injection (see Sec.~\ref{sec:II}) can be described by 

6611: both carrier and spin 

6612: drift and diffusion, limited by carrier recombination

6613: and spin relaxation \cite{Fabian2002:PRB,Zutic2002:PRL}. 

6614: In addition, if spin precession is important for device operation, spin 

6615: dynamics need to be explicitly incorporated into the transport equations 

6616: \cite{Qi2003:PRB}. Drift of the spin-polarized carriers can be due not only 

6617: to the electric field, but also 

6618: to magnetic fields. We illustrate spin-polarized drift and diffusion 

6619: on the transport model of spin-polarized bipolar transport, 

6620: where bipolar refers to the presence of electrons and holes, 

6621: not spin up and down. 

6622: A spin-polarized unipolar transport can be obtained as a limiting case by

6623: setting the electron-hole recombination rate to zero and considering only

6624: one type of carrier (either electrons or holes).

6625: 

6626: Consider electrons and holes whose density is commonly denoted here as 

6627: $c$ (for carriers), moving in the electrostatic potential $\phi$ which 

6628: comprises both the external bias $V$ and the internal built-in fields 

6629: due to 

6630: charge inhomogeneities. Let the equilibrium spin splitting of the 

6631: carrier band be 

6632: $2q\zeta_{c}$. The spin $\lambda$ resolved carrier charge-current density is

6633: \cite{Zutic2002:PRL}

6634: \begin{equation}

6635: \label{eq:jcl}

6636: {\bf j}_{c\lambda}=

6637: -q\mu_{c\lambda}c_{\lambda}\nabla \phi {\pm}qD_{c\lambda}\nabla c_\lambda

6638: -q\lambda\mu_{c\lambda}c_{\lambda}\nabla \zeta_c,

6639: \end{equation}

6640: where $\mu$ and $D$ stand for mobility and 

6641: diffusion

6642: coefficients, 

6643: the upper sign is for electrons  and the lower sign is for holes. The first term on the

6644: right hand side describes drift caused by the total electric field, 

6645: the second term represents diffusion, while the last term stands for 

6646: magnetic drift---carrier drift in inhomogeneously split 

6647: bands.\footnote{Equation~(\ref{eq:jcl}) can be viewed as the generalization

6648: of the Silsbee-Johnson spin-charge coupling 

6649: \cite{Johnson1987:PRB,Heide2001:PRL,Wegrowe2000:PRB} to bipolar transport 

6650: and to systems with spatially inhomogeneous charge density.}

6651: More transparent are the equations for the total charge, 

6652: $j=j_{\uparrow}+

6653: j_{\downarrow}$, and spin, $j_s=j_{\uparrow}-j_{\downarrow}$, 

6654: current densities:

6655: \begin{eqnarray} \label{eq:jc}

6656: &{\bf j}_c&=-\sigma_c \nabla\phi - 

6657: \sigma_{sc}\nabla \zeta_c \pm qD_c \nabla c \pm qD_{sc}\nabla s_c,   \\

6658: \label{eq:js}

6659: &{\bf j}_{sc}&=-\sigma_{sc} \nabla\phi - 

6660: \sigma_{c}\nabla \zeta_c \pm qD_{sc} \nabla c \pm qD_{c}\nabla s_c,

6661: \end{eqnarray}

6662: where the carrier density $c=c_\uparrow+c_\downarrow$ 

6663: and spin $s_c=c_\uparrow-c_\downarrow$, 

6664: and we introduced the carrier charge and spin conductivities 

6665: $\sigma_c=q(\mu_c c+ \mu_{sc} s_c)$

6666: and $\sigma_{sc}=q(\mu_{sc}c+\mu_cs_c)$, 

6667: where $\mu_c=(\mu_{c\uparrow}+\mu_{c\downarrow})/2$

6668: and $\mu_{cs}=(\mu_{c\uparrow}-\mu_{c\downarrow})/2$ 

6669: are charge and spin mobilities, and similarly

6670: for the diffusion coefficients. Equation (\ref{eq:jc}) 

6671: describes the spin-charge coupling 

6672: in bipolar transport in inhomogeneous magnetic semiconductors. 

6673: Spatial variations in 

6674: spin density can cause charge currents. 

6675: Similarly, it follows from Eq.~(\ref{eq:js})

6676: that spatial variations in carrier densities can lead to spin currents.  

6677: 

6678: Steady-state carrier recombination and spin relaxation processes 

6679: are described 

6680: by the continuity equations for the spin-resolved carrier densities:

6681: \begin{equation} 

6682: \label{eq:cont}

6683: \nabla\cdot \frac{{\bf j}_{c\lambda}}{q}= 

6684: \pm w_{c\lambda} (c_{\lambda}\bar{c}-

6685: c_{\lambda 0}\bar{c}_0)\pm 

6686: \frac{c_{\lambda}-c_{-\lambda}-\lambda \tilde{s}_c}{2\tau_{sc}}.

6687: \end{equation}

6688: Here $w$ is the spin-dependent recombination rate, the bar denotes

6689: a complementary carrier ($\bar{n}=p$, for example), $\tau_{sc}$ 

6690: is the spin relaxation

6691: time of the carrier $c$ (not to be confused with the single spin

6692: decoherence time discussed in Sec.~\ref{sec:IIIA1}),

6693: and $\tilde{s}_c=P_{c0}c$ is the nonequilibrium spin

6694: density, which appears after realizing that spin relaxation equilibrates 

6695: spin while

6696: preserving carrier density. Finally, the set of equations is 

6697: completed with

6698: Poisson's equation, 

6699: \begin{equation}\label{eq:poisson}

6700: \varepsilon\Delta \phi=-\rho,

6701: \end{equation} 

6702: connecting the electric field and charge $\rho=q(p-n+N_d-N_a)$,

6703: where $N_d$ and $N_a$ are the donor and acceptor densities, respectively, and

6704: $\varepsilon$ is

6705: the dielectric constant. 

6706: 

6707: In many important cases Eqs.~(\ref{eq:jcl}), (\ref{eq:cont}), 

6708: and (\ref{eq:poisson}) 

6709: need to be solved self-consistently, which usually requires

6710: numerical techniques \cite{Zutic2002:PRL}. 

6711: In some cases it is possible to extract the relevant physics in limiting 

6712: cases analytically,

6713: usually neglecting electric field or magnetic drift. 

6714: In unipolar spin-polarized transport one does not need to consider carrier 

6715: recombination. It also often suffices to study pure spin diffusion, 

6716: if the built-in electric fields are small. Unipolar spin-polarized transport 

6717: in inhomogeneous systems in the presence of electric fields was analyzed by 

6718: \textcite{Fabian2002:PRB,Yu2002:PRBa,Yu2002:PRBb,Martin2003:PRB,Pershin2003:PRL}.

6719: Spin-polarized drift and diffusion in 

6720: model GaAs quantum wires was studied by \textcite{Sogawa2000:PRB}, while 

6721: ramifications of magnetic drift for unipolar transport were 

6722: studied by \textcite{Martin2003:PRB,Fabian2002:PRB}. Bipolar transport 

6723: in the presence of electrical drift and/or 

6724: diffusion has been studied by 

6725: \textcite{Zutic2002:PRL,Fabian2002:PRB,Beck2002:PE,Flatte2000:PRL}.

6726: Transient dynamics of spin drift and diffusion was considered by 

6727: \textcite{Fabian2002b:PRB}.

6728: Recently an interesting study \cite{Saikin2003:JAP} was reported on a 

6729: Monte-Carlo simulation of quantum-mechanical 

6730: spin dynamics limited by spin relaxation, in which quasiclassical 

6731: orbital transport 

6732: was carried out for the in-plane

6733: transport in III-V heterostructures where spin precession is due to  bulk

6734: and structure inversion asymmetry (see Sec.~\ref{sec:IIIB2}).

6735: 

6736: 

6737: \subsection{\label{sec:IVB} Materials considerations}

6738: 

6739: Nominally highly spin-polarized materials, as discussed in the previous

6740: sections, could provide both effective spin injection into nonmagnetic

6741: materials and large MR effects, important for nonvolatile applications.

6742: Examples include  half-metallic oxides such as CrO$_2$, Fe$_3$O$_4$, 

6743: CMR materials, and double perovskites \cite{Kobayashi1998:N} 

6744: [for reviews of half-metallic materials see 

6745: \cite{Pickett2001:PT,Fang2002:JAP}].

6746: Ferromagnetic semiconductors \cite{Nagaev:1983}, known since  CrBr$_3$ 

6747: \cite{Tsubokawa1960:JPSJ}, have been demonstrated to be highly spin polarized.

6748: However, more recent interest in ferromagnetic semiconductors was spurred by the 

6749: fabrication of (III,Mn)V compounds.\footnote{Ferromagnetic order 

6750: with Mn-doping was

6751: obtained previously, for example, in (Sn,Mn)Te \cite{Escorne:1974}, 

6752: (Ge,Mn)Te \cite{Cochrane1974:PRB} and (Pb,Sn,Mn)Te \cite{Story1986:PRL}.}

6753: After the initial discovery of (In,Mn)As 

6754: \cite{Munekata1989:PRL,Ohno1992:PRL,Munekata1991:JCG}, 

6755: most of the research 

6756: has focused on (Ga,Mn)As \cite{Ohno1996:APL,VanEsch1997:PRB,Hayashi1997:JCG}.  

6757: In contrast to (In,Mn)As and

6758: (Ga,Mn)As with high carrier density ($\sim$ 10$^{20}$ cm$^{-3}$),

6759: a much lower carrier density in (Zn,Cr)Te \cite{Saito2002:JAP}, 

6760: a II-VI ferromagnetic 

6761: semiconductor with Curie temperature $T_C$ near room temperature 

6762: \cite{Saito2003:PRL},

6763: suggests that transport properties can be effectively controlled by

6764: carrier doping. Most of the currently studied ferromagnetic semiconductors

6765: are p-doped with holes as spin-polarized carriers, which typically

6766: leads to lower mobilities and shorter spin relaxation times than in

6767: n-doped materials. It is possible to use selective doping to substantially

6768: increase $T_C$, as compared to the uniformly doped bulk ferromagnetic

6769: semiconductors \cite{Nazmul2003:PRB}.

6770: 

6771: Early work on (Ga,Mn)As \cite{DeBoeck1996:APL} showed the low solubility 

6772: of Mn and the formation of  

6773: magnetic nanoclusters 

6774: characteristic of many 

6775: subsequent compounds and different magnetic impurities. 

6776: The presence of such 

6777: nanoclusters often complicates accurate

6778: determination of  $T_C$ as well as of whether 

6779: the compound is actually in a single phase. Consequently, the reported

6780: room-temperature ferromagnetism  in an increasing number of compounds

6781: reviewed by \textcite{Pearton2003:JAP} is not universally accepted.

6782: Conclusive evidence for intrinsic ferromagnetism in

6783: semiconductors is highly nontrivial. For example, early work reporting 

6784: ferromagnetism even at nearly 900 K in  La-doped CaBa$_6$ 

6785: \cite{Young1999:N,Ott2000:PB,Tromp2001:PRL}, was later revisited

6786: suggesting extrinsic effect \cite{Bennett2003:P}. 

6787: It remains to be understood what the limitations are for using 

6788: extrinsic ferromagnets and, for example,  whether they can be 

6789: effective spin injectors. 

6790: 

6791: \begin{figure}

6792: \centerline{\psfig{file=zutic_fig25.eps,width=1.0\linewidth,angle=0}}

6793: \caption{Photoinduced ferromagnetism in a (In,Mn)As/GaSb heterostructure:

6794: (a) light-irradiated 

6795: sample displaying photoinduced

6796: ferromagnetism--direction of light irradiation is

6797: shown by an arrow; (b) band-edge profile of (In,Mn)As/GaSb heterostructure.

6798: $E_c$, conduction band, $E_v$, valence band; $E_F$,  

6799: Fermi level, respectively. From \onlinecite{Koshihara1997:PRL}.} 

6800: \label{light:1}

6801: \end{figure}

6802: 

6803: A high $T_C$ 

6804: and almost complete spin polarization in 

6805: bulk samples are alone not sufficient for successful applications. Spintronic 

6806: devices typically rely on inhomogeneous doping, structures of reduced 

6807: dimensionality, 

6808: and/or structures containing different materials. 

6809: Interfacial properties, as discussed in the

6810: previous sections, can significantly influence the magnitude of 

6811: magnetoresistive effects\footnote{In magnetic multilayers GMR is typically

6812: dominated by interfacial scattering \cite{Parkin1993:PRL}, while in MTJ's 

6813: it is the surface rather than the bulk electronic structure which 

6814: influences the relevant spin polarization.}

6815: and the efficiency of spin injection.

6816: Doping properties and possibility of fabricating a wide range of structures

6817: allow spintronic applications beyond MR effects, for example,

6818: spin transistors, spin lasers, and spin-based quantum computers 

6819: (Sec.~\ref{sec:IVF}).

6820: Materials properties of hybrid F/Sm heterostructures, relevant to

6821: device applications, were reviewed by \textcite{Samarth2003:SSC}.

6822: 

6823: Experiments in which the ferromagnetism is induced 

6824: optically 

6825: \cite{Koshihara1997:PRL,Oiwa2002:PRL,Wang2003:P} and electrically 

6826: \cite{Ohno2000:N,Park2002:S} provide a method for distinguishing the 

6827: carrier-induced

6828: ferromagnetism, based on the exchange interaction between the 

6829: carrier and the magnetic impurity spins,

6830: %(based on the exchange interaction $J {\bf S} \cdot {\bf s}_i$ between

6831: %the carrier spins ${\bf S}$ and the localized magnetic spins ${\bf s}_i$)

6832: from ferromagnetism that originates from magnetic nanoclusters. 

6833: Such experiments also suggest a possible nonvolatile multifunctional

6834: devices with tunable, optical, electrical, and magnetic properties. 

6835: Comprehensive surveys of magneto-optical materials and applications, not

6836: limited to semiconductors, are given by \textcite{Zvezdin:2003,Sugamo:2000}.

6837: 

6838: Photoinduced ferromagnetism was demonstrated by \textcite{Koshihara1997:PRL} 

6839: in p-(In,Mn)As/GaSb 

6840: heterostructure, shown in Figs.~\ref{light:1}, and \ref{light:2}. 

6841: Unpolarized light penetrates through a thin (In,Mn)As layer and is absorbed 

6842: in the GaSb layer. A large band bending across the heterostructures separates,

6843: by a built-in field, electrons and holes. The excess holes generated in a 

6844: GaSb layer are effectively transferred to the p-doped (In,Mn)As layer where 

6845: they enhance the ferromagnetic spin exchange among Mn ions, resulting

6846: in a paramagnetic-ferromagnetic transition.

6847: 

6848: 

6849: \begin{figure}

6850: \centerline{\psfig{file=zutic_fig26.eps,width=0.85\linewidth,angle=0}}

6851: \caption{Magnetization curves for (In,Mn)As/GaSb at 5 K observed before 

6852: (open circles)

6853: and after (solid circles) light irradiation. Solid line show a theoretical 

6854: curve. (b) Hall resistivity at 5 K before (dashed lines) 

6855: and after (solid lines) 

6856: light irradiation. From \onlinecite{Koshihara1997:PRL}.} 

6857: \label{light:2}

6858: \end{figure}

6859: The increase in magnetization, measured by a SQUID, 

6860: is shown in Fig.~\ref{light:2}(a) and in

6861: Fig.~\ref{light:2}(b) the corresponding Hall resistivity 

6862: \begin{equation}

6863: \rho_{\rm Hall}=R_0 B+ R_S M,

6864: \label{eq:hall}

6865: \end{equation}

6866: is shown, where the $R_0$ is the ordinary and $R_S$ 

6867: the anomalous Hall coefficient, 

6868: respectively.

6869: Typical for (III,Mn)V compounds, $\rho_{\rm Hall}$ is dominated by the 

6870: anomalous

6871: contribution, $\rho_{\rm Hall} \propto M$. 

6872: 

6873: 

6874: \begin{figure}

6875: \centerline{\psfig{file=zutic_fig27.eps,width=1.0\linewidth}}

6876: \caption{Electric-field control of ferromagnetism.

6877: R$_{\rm Hall}$ vs field curves under three different gate biases.

6878: Application of V$_G$=0, +125, and -125 V results in a qualitatively different 

6879: field

6880: dependence of R$_{\rm Hall}$ measured at 22.5 K (sample B):

6881: almost horizontal dash-dotted line, paramagnetic response 

6882: are

6883: partially depleted from the channel (V$_G$=+125 V);

6884: dashed lines, clear hysteresis at low fields

6885: ($<$0.7 mT) as holes are accumulated in the channel (V$_G$=-125 V);

6886: solid line, R$_{\rm Hall}$ curve measured at V$_G$=0 V before 

6887: application of $\pm$ 125 V, dotted line, 

6888: R$_{\rm Hall}$ after 

6889: application of $\pm$ 125 V.

6890: Inset, the same curves shown at higher magnetic fields.

6891: From \onlinecite{Ohno2000:N}.} 

6892: \label{ohno}

6893: \end{figure}

6894: 

6895: 

6896: A different type of photoinduced magnetization was measured in ferromagnetic 

6897: (Ga,Mn)As.\footnote{Previous studies in paramagnetic (II,Mn)VI materials 

6898: have shown that

6899: nonequilibrium spin-polarized carriers can change the orientation of

6900: magnetic spins in (Hg,Mn)Te \cite{Krenn1985:PRL,Krenn1989:PRB} and in 

6901: (Cd,Mn)Te

6902: \cite{Awschalom1987:PRL}.}

6903: In a Faraday geometry (recall \ref{sec:IID3}), by changing the polarization

6904: of a circularly polarized light, 

6905: one can modulate

6906: the Hall resistance and thus the induced 

6907: magnetization

6908: by up to 15\% of the saturation value \cite{Oiwa2002:PRL}.

6909: Additional experiments on photoinduced magnetization rotation 

6910: \cite{Munekata2003:JS,Oiwa2003:JS} suggest that the 

6911: main contribution of carrier spin to such rotation is realized by generating

6912: an effective magnetic field through the $p$-$d$ exchange interaction, 

6913: rather than by spin-transfer torque as discussed in 

6914: Secs.~\ref{sec:IB1} and \ref{sec:V}

6915: \cite{Moriya2003:P}.

6916: In GaAs-Fe composite films an observation of room temperature photoenhanced 

6917: magnetization was used to demonstrate that a magnetic force can be changed

6918: by light illumination \cite{Shishi2003:APL}.

6919: 

6920: Electrically induced ferromagnetism was realized  by applying gate

6921: voltage V$_G$ to change the hole concentration in $d$=5nm thick (In,Mn)As

6922: used as a magnetic channel in a metal-insulator semiconductor FET structure.

6923: Below a metal gate and an insulator the (In,Mn)As channel was grown on top

6924: of a InAs/(Al,Ga)Sb/AlSb and GaAs substrate. 

6925: In Fig.~\ref{ohno},

6926: the corresponding data for $R_{\rm Hall}=\rho_{\rm Hall}/d \propto M$ 

6927: [recall Eq.~(\ref{eq:hall}), show that the ferromagnetism can be switched on

6928: and off, as an electric analog of the manipulation of M 

6929: from Fig.~\ref{light:2}.

6930: Subsequent work by \textcite{Park2002:S} showed that in MnGe ferromagnetism 

6931: can

6932: be manipulated at higher temperature and at significantly lower gate voltage

6933: (at $\sim$ 50 K and $\sim$ 1 V). The combination of light and electric-field

6934: control of ferromagnetism was used in modulation-doped p-type

6935: (Cd,Mn)Te QW \cite{Boukari2002:PRL}. It was demonstrated that 

6936: illumination by light in $p-i-n$ diodes would  enhance the spontaneous

6937: magnetization, while illumination in $p-i-p$ structures would 

6938: destroy ferromagnetism.

6939: 

6940: 

6941: \begin{figure}

6942: \centerline{\psfig{file=zutic_fig28.eps,width=1.0\linewidth}}

6943: \caption{Voltage-controlled spin precession: (a) time-resolved Kerr rotation

6944: measurements of electron spin precession in a quantum well at different

6945: gate voltages $V_G$ with Al concentration of 7\% at 5 K and B=6 T;

6946: displacement of the electron wave function

6947: towards the back gate into regions with more Al concentration

6948: as a positive

6949: voltage $V_G$ is applied between back and front gate;

6950: leading to an increase of $g$. 

6951: At $V_G$=2 V, no precession is

6952: observed, corresponding to $g$=0. From \onlinecite{Salis2001:N}.} 

6953: \label{mater:1}

6954: \end{figure}

6955: 

6956: 

6957: In semiconductors $g$ factors, which determine the 

6958: spin splitting of carrier bands (and consequently influence the spin dynamics 

6959: and spin resonance), can be very different from the free-electron value. 

6960: With strong spin-orbit coupling in narrow-band III-V's they 

6961: are $\approx$-50 

6962: for InSb and $\approx$-15 for InAs, while, 

6963: as discussed in Sec.~\ref{sec:IID3} the doping 

6964: with magnetic impurities can give even $|g^*|\sim 500$. Manipulation

6965: of the $g$ factor in a GaAs/AlGaAs quantum well (QW)  

6966: in Fig.~\ref{mater:1}, relies on the results for a bulk Al$_x$Ga$_{1-x}$As, 

6967: the variation of Al concentration changes the $g$ factor 

6968: \cite{Chadi1976:PRB,Weisbuch1977:PRB} 

6969: $g$=-0.44 for x=0 and $g$=0.40 for x=0.3. Related experiments on 

6970: modulation-doped 

6971: GaAs/Al$_{0.3}$Ga$_{0.7}$As have shown that by applying V$_G$ 

6972: on can shift the electron

6973: wave function in the QW and produce $\sim$1\% change

6974: in the $g$ factor \cite{Jiang2001:PRB}. Subsequently, 

6975: in an optimized Al$_x$Ga$_{1-x}$As quantum well, where

6976: $x$ varied gradually across the structure, much larger changes were

6977: measured--when $V_G$ is changed, the electron wave function  

6978: efficiently samples 

6979: different regions with different $g$ factors \cite{Salis2001:N}. 

6980: Figure~\ref{mater:1}(a) gives

6981: the time-resolved Kerr rotation data (the technique is discussed 

6982: in Sec.~\ref{sec:III}) 

6983: data which can be described as 

6984: $\propto \exp(-\Delta t/T_2^*)\cos(\Omega \Delta t)$,

6985: where $\Delta t$ is the delay time between the circularly polarized pump

6986: and linearly polarized probe pulses, T$_2^*$ is the

6987: transverse electron spin lifetime with inhomogeneous broadening, and 

6988: the angular precession frequency $\Omega=\mu_B g  B/\hbar$ can be used to 

6989: determine the $g$ factor. 

6990: It is also possible to manipulate $g$ factors dynamically using 

6991: time-dependent $V_G$ \cite{Kato2003:S}. The anisotropy of $g$ factor

6992: ($g$ tensor) allows voltage control of both the magnitude and the

6993: direction of the spin precession vector ${\bf \Omega}$.

6994: 

6995: \subsection{\label{sec:IVC} Spin filters}

6996: %-----------------------------------------------------------------------

6997: 

6998: Solid state spin filtering (recall the similarity with

6999: spin injection from Sec~\ref{sec:IIC1}) was first realized in N/F/N

7000: tunneling. 

7001: It was shown by \textcite{Esaki1967:PRL} that the magnetic tunneling through

7002: (ferro)magnetic semiconductor Eu chalcogenides 

7003: \cite{Nagaev:1983,Kasuya1968:RMP,vonMolnar1967:JAP}, 

7004: such as EuSe\footnote{At zero magnetic field

7005: EuSe is an antiferromagnet, 

7006: and at moderate fields it becomes a ferromagnet with T$_C$$\approx$ 5K.} 

7007: and EuS,\footnote{At zero magnetic field, exchange

7008: splitting of a conduction band in bulk EuS is $\approx$ 0.36 eV

7009:  \cite{Hao1990:PRB}.}

7010: could be modified by an applied magnetic field. The change in $I-V$ curves 

7011: in the

7012: N/F/N structure, where N is a normal metal and F is a ferromagnet, was 

7013: explained by the influence of the magnetic field on the height of the barrier 

7014: formed at the N/F interface (for EuSe, the barrier height was lowered by

7015: 25\% at 2 T). 

7016: The large spin splitting of the Eu chalcogenides was subsequently 

7017: employed in the absence of

7018: applied field with EuS \cite{Moodera1988:PRL} and nearly 100\%

7019: spin polarization was reached  

7020: at $B$=1.2 T with EuSe \cite{Moodera1993:PRL}. 

7021: These spin-filtering properties of the 

7022: Eu chalcogenides, used together with one-electron quantum dots, were proposed 

7023: as the basis for a method to convert 

7024: single spin into single charge 

7025: measurements\footnote{This method could already be 

7026: realized using single-electron transistors or 

7027: quantum point contacts.} and provide an important ingredient in realizing a 

7028: quantum computer \cite{DiVincenzo1999:JAP}, see Sec.~\ref{sec:IVF}.

7029: 

7030: Zeeman splitting in semiconductor heterostructures and superlattices 

7031: (enhanced by large $g$ factors) \cite{Egues1998:PRL,Guo2001:PRB},

7032: in quantum dots \cite{Recher2000:PRL,Deshmukh2002:PRL,Borda2003:PRL}, 

7033: and nanocrystals  

7034: \cite{Efros2001:PRL} provide effective spin filtering and spin-polarized

7035: currents. Predicted quantum size effects and resonance tunneling 

7036: \cite{Duke:1969}   %iz p.79 Duke

7037: also have their spin-dependent counterparts. The structures studied

7038: are typically double-barrier resonant tunneling diodes 

7039: (for an early

7040: spin-unpolarized study see \textcite{Tsu1973:APL}), with either 

7041: Zeeman splitting or using ferromagnetic materials, in which

7042: spin filtering can be tuned by an applied bias.\footnote{See, for example, 

7043: \cite{Brehmer1995:APL,Mendez1986:PRB,Giazotto2003:APL,Ohno1998:S,% 

7044: Petukhov2002:PRL,Ting2002:APL,Slobodskyy2003:PRL,petukhov1998:ASS,%

7045: Aleiner1991:JETPL, Vurgaftman2003:PRB}.} 

7046: 

7047: Several other realizations of spin filtering have been investigated,

7048: relying on  spin-orbit coupling.\footnote{These include the work of 

7049: \cite{Kiselev2001:APL,Governale2002:PRB,Voskoboynikov1998:PRB,%

7050: Koga2002b:PRL,Silva1999:PRB,Voskoboynikov1999:PRB,Perel2003:PRB}.} 

7051: or hot-electron transport across ferromagnetic regions,\footnote{See

7052: \cite{vanDijken2002:PRB,Monsma1995:PRL,Rippard2000:PRL,Upadhyay1999:APL,%

7053: Oberli1998:PRL,Cacho2002:PRL,Filipe1998:PRL}.} discussed in more detail 

7054: in Sec.~\ref{sec:IVE3}.

7055: A choice of particular atomically ordered F/Sm interfaces was suggested

7056: to give a strong spin-filtering effect \cite{Kircenzow2001:PRB},

7057: limited by the spin-orbit coupling and interfacial spin-flip scattering.

7058: 

7059: \begin{figure} 

7060: \centerline{\psfig{file=zutic_fig29.eps,width=1.0\linewidth}}

7061: \caption{Mesoscopic spin filter: 

7062: (a) micrograph and circuit showing the

7063: polarizer-analyzer configuration used in the experiment

7064: of \textcite{Folk2003:S}. The emitter (E)

7065: can be formed into either a quantum dot or a quantum point contact (QPC). 

7066: The collector

7067: (C), is a single point contact. Electrons are focused from E to C through

7068: the base region (B), using a small perpendicular magnetic field. Gates

7069: marked with ``x'' are left undepleted 

7070: when E is operated as a QPC;

7071: (b) base-collector voltage (V$_C$) showing two focusing peaks;

7072: (c) focusing peak height at B$_\|$=6 T with spin-selective collector

7073: QPC conductance ($g_C$=$0.5e^2/h$), comparing E as QPC at $2e^2/h$ (dashed

7074: curve) and E as a quantum dot with both leads at $2e^2/h$ (solid curve). 

7075: Adapted from \onlinecite{Folk2003:S}.} 

7076: \label{filter:1}

7077: \end{figure}

7078: 

7079: Mesoscopic spin filters have also been suggested 

7080: \cite{Frustaglia2001:PRL,Joshi2001:PRB,Ionicioiu2003:PRB,Avdonin2003:P}

7081: and we discuss here a particular realization.

7082: In an applied magnetic field 

7083: two quantum point contacts (QPC), an emitter (E)

7084: and a collector (C), fabricated on top of a high-mobility 

7085: 2DEG in GaAs/AlGaAs,

7086: can act as spin polarizer and analyzer 

7087: \cite{Potok2002:PRL}.\footnote{QPC was also used 

7088: to locally create and probe nonequilibrium nuclear spin

7089: in GaAs/AlGaAs heterostructures 

7090: in the quantum Hall regime \cite{Wald1994:PRL}.} 

7091: In a ballistic regime and at $T=70$ mK 

7092: (mean free path $\approx$ 45 $\mu$m $>>$ sample size $\approx$ 1.5 $\mu$m)

7093: magnetic focusing\footnote{Suggested by \textcite{Sharvin1965:ZETF}

7094: as a technique to study Fermi surfaces; see also \textcite{vanHouten1989:PRB}.} 

7095: with $B_\bot$ results in base-collector voltage peaks, when the separation

7096: of the two QPCs is an even multiple of the cyclotron radius $m^*v_F/eB_\bot$,

7097: where $m^*$ is the effective mass, and v$_F$ the Fermi velocity.

7098: These results are illustrated in Figs.~\ref{filter:1}(a) and (b) 

7099: on a slightly modified structure \cite{Folk2003:S}

7100: where, by applying a gate voltage, one can cause the emitter to form either 

7101: a quantum dot or QPC.  

7102: Effective spin filtering, due to the large in-plane field $B_\|$, can 

7103: be tuned by the gate voltage which changes the conductance of QPC.

7104: The resulting effect of spin filtering modifies the collector voltage

7105: $V_C$  \cite{Potok2002:PRL}, 

7106: \begin{equation}

7107: V_C=\alpha(h/2e^2)I_E(1+P_{I_E}P_{T_C}),

7108: \label{eq:filter}

7109: \end{equation}

7110: where $0< \alpha <1$ parameterizes the imperfections in focusing,

7111: $P_{I_E}$ and $P_{T_C}$ are the spin polarization [recall Eq.~(\ref{eq:polar})] 

7112: of the emitter current $I_E$, and the collector transmission coefficient $T_C$

7113: is related to the collector conductance by $g_C=(2e^2/h)T_C$. 

7114: In Eq.~(\ref{eq:filter}) we note a recurring form for a spin-valve

7115: effect. The measured signal involves the product of two different spin 

7116: polarizations, for example, similar to TMR in Eq.~(\ref{eq:julliere}) 

7117: or to spin-charge coupling due to nonequilibrium spin 

7118: [recall Eqs.~(\ref{eq:DelR}) and (\ref{eq:md})].

7119: Another mesoscopic spin filter with few-electron quantum dot 

7120: (GaAs/AlGaAs-based) was used to demonstrate a nearly complete spin

7121: polarization which could be reversed by adjusting gate voltages 

7122: \cite{Hanson2003:P}.

7123: 

7124: \subsection{\label{sec:IVD} Spin diodes}

7125: %--------------------------------------------------------------------------

7126: 

7127: Spin diodes are inhomogeneous two-terminal devices whose electronic or optical 

7128: properties depend on the spin polarization of the carriers. Such devices

7129: were envisaged long before the emergence of 

7130: spintronics. \textcite{Solomon1976:SSC},

7131: for example, proposed and demonstrated a silicon {\it p-n} junction whose 

7132: current was modified by changing the spin polarization of the recombination

7133: centers. In a magnetic field both the mobile carriers and the recombination

7134: centers have an equilibrium spin polarization due to the Zeeman splitting.

7135: The current in a {\it p-n} junction depends on the recombination rate,

7136: which, in turn, depends on the relative orientation of the spin 

7137: of the carriers and the centers \cite{Lepine1972:PRB}. The trick to 

7138: modifying the current is to decrease (even saturate) the spin polarization

7139: of either the electrons or the centers by electron spin

7140: resonance. Indeed, \textcite{Solomon1976:SSC} found a variation of

7141: $\approx 0.01\%$ of the saturation current at small biases where recombination

7142: in the space-charge region dominates. Similar experiments could be used to 

7143: detect nonequilibrium spin due to (potential) spin injection in Si,

7144: where optical methods are ineffective, but also in other semiconductors

7145: where electrical detection would be desirable.\footnote{Spin diodes can 

7146: also probe fundamental properties of electronic

7147: systems. The diode demonstrated by \textcite{Kane1992:PRB} is 

7148: based on a junction

7149: between two coplanar AlGaAs/GaAs 2DEG's, one with $\nu<1$ and 

7150: the other with $\nu>1$, 

7151: where $\nu$ is the Landau-level filling; that is, the two regions have

7152: opposite spins at the Fermi level. The current crossing such a junction,

7153: which has a diode property due to the existence of a built-in field in

7154: the contact, is accompanied by a spin flip. Interestingly, the

7155: current is also time dependent, due to the current-induced dynamic nuclear

7156: polarization.} 

7157: 

7158: Several spin diodes have recently been proposed or demonstrated 

7159: with the goal of either maximizing the sensitivity of the $I-V$ characteristics

7160: to spin and magnetic field, or to facilitating spin injection and its

7161: detection through semiconductor interfaces comprising a magnetic semiconductor

7162: as the injector. 

7163: Magnetic tunneling diodes have been used for spin injection 

7164: from a ferromagnetic to a nonmagnetic semiconductor, in p-GaMnAs/n-GaAs

7165: {\it p-n} 

7166: junctions~\cite{Kohda2001:JJAP,Johnston-Halperin2002:PRB,vanDorpe2003:P}.

7167: As discussed in Sec.~\ref{sec:IID3},

7168: {\it p-n} 

7169: heterostructures have combined Cr- or Eu-based ferromagnetic

7170: semiconductors and InSb  \cite{Osipov1998:PL,Viglin1997:PLDS}.

7171: Spin light-emitting diodes (recall Figs.~\ref{injexp:3}

7172: and \ref{injexp:4}) 

7173: were employed for injecting and detecting spins 

7174: in semiconductors,

7175: while resonant tunneling diodes have been demonstrated as effective spin 

7176: injectors (Sec.~\ref{sec:IID3}) and spin filters (Sec.~\ref{sec:IVC}).

7177: A magnetic unipolar diode has been proposed by

7178: \textcite{Flatte2001:APL}

7179: to simulate the working of ordinary diodes, but with 

7180: homogeneous monopolar

7181: doping (either donors or acceptors, not both). The role of 

7182: inhomogeneous doping in the {\it p-n} junction is played by the inhomogeneous 

7183: spin splitting

7184: of the carrier band, with the spin up and spin down carriers 

7185: playing roles similar

7186: to those of the electrons and holes in bipolar diodes. 

7187: Si-based {\it p-i-n} diode sandwiched between two ferromagnetic

7188: metals was suggested to allow controlling the device performance

7189: by an externally applied magnetic field \cite{Dugaev2003:PE}.

7190: Finally, 

7191: \textcite{Zutic2002:PRL} 

7192: have proposed the magnetic bipolar diode described below.

7193: 

7194: 

7195: The magnetic bipolar diode\footnote{Not to be confused with the usual 

7196: magnetic diodes

7197: which are ordinary diodes in a magnetic field. The $I-V$ characteristics of

7198: such diodes, depend on the magnetic field through small orbital effects 

7199: on diffusion coefficient, 

7200: not through the spin effects described here.} (MBD) is a {\it p-n} 

7201: junction diode with one or both regions magnetic 

7202: \cite{Zutic2002:PRL,Fabian2002:PRB}.

7203: The MBD is the prototypical device of bipolar

7204: spintronics, a subfield of spintronics in which both electrons and holes 

7205: take part 

7206: in carrier transport, while either electrons or holes (or both) are spin 

7207: polarized (see Sec.~\ref{sec:IVA4}). 

7208: Examples of nonmagnetic bipolar spintronic devices are the 

7209: spin-polarized {\it p-n} 

7210: junction \cite{Zutic2001:PRB} and the spin solar cell \cite{Zutic2001:APL}. 

7211: These devices offer 

7212: opportunities for effective spin injection, 

7213: spin amplification (see Sec.~\ref{sec:IIC3}), 

7214: or spin capacity---the effect of 

7215: changing, by voltage, nonequilibrium spin density \cite{Zutic2001:PRB}. 

7216: The advantages of 

7217: magnetic bipolar spintronic devices 

7218: \cite{Zutic2002:PRL,Fabian2002:PRB,Zutic2003:APL,Fabian2002:P}

7219: lie in the combination of equilibrium 

7220: magnetism and nonequilibrium spin and effective methods to manipulate

7221: a minority carrier population.

7222: The most useful effects of the 

7223: spin-charge coupling in MBD's are the 

7224: spin-voltaic

7225: and the giant-magnetoresistive

7226: effects, which are enhanced over those of metallic systems by the 

7227: exponential dependence of the current on bias voltage.

7228: 

7229: \begin{figure}

7230: \centerline{\psfig{file=zutic_fig30.eps,width=1\linewidth,angle=0}}

7231: \caption{Scheme of a magnetic bipolar diode. 

7232: The $p$-region (left) is magnetic, indicated by the spin

7233: splitting $2q\zeta$ of the

7234: conduction band. The $n$-region (right) is nonmagnetic, but spin 

7235: polarized by a spin source: Filled circles, spin-polarized electrons;

7236: empty circles, unpolarized holes.

7237: If the nonequilibrium spin in the 

7238: $n$-region is oriented parallel (top figure) 

7239: to the equilibrium spin in the 

7240: $p$-region, large forward current flows. 

7241: If the relative orientation is antiparallel (bottom), the current drops

7242: significantly. Adapted from \onlinecite{Zutic2002:PRL}. 

7243: }

7244: \label{fig:md}

7245: \end{figure}

7246: 

7247: A scheme of an MBD is shown in Fig.~\ref{fig:md} 

7248: (also see Fig.~\ref{deplete}). 

7249: The $p$-region is magnetic, by 

7250: which we mean that it has a spin-split conduction band with the spin

7251: splitting (Zeeman or exchange) 

7252: $2q\zeta$ $\sim k_BT$. 

7253: Zeeman splitting can be significantly enhanced by 

7254: the large $g^*$ factors of magnetically doped (Sec.~\ref{sec:IIC3})

7255: or narrow-band-gap semiconductors (Sec.~\ref{sec:IVB}).

7256: Using an MBD with a ferromagnetic semiconductor slightly above its $T_C$ 

7257: is also expected to give large $g^*$ factors.

7258: The $n$-region 

7259: is nonmagnetic, but 

7260: electrons can be 

7261: spin-polarized 

7262: by a spin source (circularly polarized light or 

7263: magnetic electrode). The interplay between the equilibrium spin of polarization 

7264: $P_{n0}=\tanh(q\zeta/k_B T)$ 

7265: in the $p$-region, and the nonequilibrium 

7266: spin source of polarization $\delta P_n$ 

7267: in the $n$-region, at the edge of the depletion layer, 

7268: determines the $I-V$  characteristics of the diodes. 

7269: It is straightforward to generalize these considerations to include 

7270: the spin-polarized holes \cite{Fabian2002:PRB}.

7271: 

7272: The dependence of the electric current $j$ on $q\zeta$ and $\delta P_n$ was 

7273: obtained by both numerical and analytical methods. 

7274: Numerical calculations \cite{Zutic2002:PRL} 

7275: were performed by self-consistently solving for the drift-diffusion, 

7276: continuity, as well as carrier recombination and 

7277: spin-relaxation equations, 

7278: discussed in  

7279: Sec.~\ref{sec:IVA4}. While the numerical calculations are indispensable in the

7280: high-injection limit,\footnote{The small bias or low-injection limit 

7281: is the regime of applied bias in which the density of the carriers injected

7282: through the depletion layer (the minority carriers) is much smaller than 

7283: the equilibrium density of the majority carriers. 

7284: Here and in Sec.~\ref{sec:IVE2} the terms majority and

7285: minority refer to the relative carrier (electron or hole) population and

7286: not to spin. The large bias or high-injection limit is the regime where 

7287: the injected carrier density

7288: becomes comparable to the equilibrium density. This occurs

7289: at forward biases comparable to the built-in potential, typically  1 V.}

7290: valuable insight and  

7291: analytical formulas can be obtained in the

7292: low-injection limit, where the Shockley theory \cite{Shockley:1950} for 

7293: ordinary {\it p-n}

7294: junctions was generalized by \textcite{Fabian2002:PRB} for the magnetic

7295: case.  

7296: 

7297: To illustrate the $I-V$ characteristics of MBD's, consider the low-injection 

7298: limit in the configuration of Fig.~\ref{fig:md}. 

7299: The electron contribution to the total electric current is

7300: \cite{Fabian2002:PRB}

7301: \begin{equation}\label{eq:md}

7302: j_n \sim n_{0}(\zeta) 

7303: \left [e^{qV/k_B T} \left (1+\delta P_n P_{n0}\right ) -1 \right ],

7304: \end{equation}

7305: where $V$ is the applied bias (positive for forward bias) 

7306: and 

7307: $n_{0}(\zeta)=(n_i^2/N_a)\cosh(q\zeta/k_B T)$ 

7308: is the equilibrium number of electrons 

7309: in the $p$-region, 

7310: dependent on the splitting, the intrinsic carrier 

7311: density $n_i$, and the acceptor doping $N_a$. Equation (\ref{eq:md}) 

7312: generalizes the 

7313: Silsbee-Johnson spin-charge coupling \cite{Silsbee1980:BMR,Johnson1985:PRL}, 

7314: originally proposed for ferromagnet/paramagnet metal interfaces, to 

7315: the case of magnetic {\it p-n} junctions.  The advantage

7316: of the spin-charge coupling in {\it p-n} junctions, as opposed to metals or 

7317: degenerate systems, is the nonlinear voltage dependence 

7318: of the nonequilibrium carrier and spin densities \cite{Fabian2002:PRB}, 

7319: allowing for 

7320: the exponential enhancement of the effect with increasing $V$. 

7321: Equation (\ref{eq:md}) can be understood qualitatively

7322: from Fig.~\ref{fig:md} \cite{Fabian2002:PRB}. 

7323: In equilibrium, $\delta P_n=0$ 

7324: and $V=0$, no current flows through the

7325: depletion layer, as the electron currents from both sides of the junction 

7326: balance out. The balance is disturbed either by applying bias or 

7327: by selectively populating different spin states,

7328: making the flow of one spin species greater than that of the other. 

7329: In the latter case, 

7330: the effective barriers for crossing

7331: of electrons from the $n$ to the $p$ side is different for spin

7332: up and down electrons (see Fig.~\ref{fig:md}).

7333: Current can flow even at $V=0$ when $\delta P_n\ne 0$. 

7334: This is an example of the spin-voltaic effect (a spin analog

7335: of the photo-voltaic effect), in which  nonequilibrium

7336: spin causes an EMF \cite{Zutic2002:PRL,Zutic2003:P}.

7337: In addition, the direction of the zero-bias current 

7338: is controlled by the relative sign of 

7339: $P_{n0}$ and $\delta P_n$. 

7340: 

7341: \begin{figure}

7342: \centerline{\psfig{file=zutic_fig31.eps,width=1\linewidth,angle=0}}

7343: \caption{Giant magnetoresistance (GMR) effect in magnetic diodes. 

7344: Current/spin-splitting characteristics ($I$-$\zeta$) 

7345: are calculated self-consistently 

7346: at $V$=0.8 V

7347: for the diode from  Fig.~\ref{fig:md}. 

7348: Spin splitting $2q\zeta$ on the p-side is normalized to $k_B T$. 

7349: The solid curve corresponds to a switched-off spin source. The current is

7350: symmetric in $\zeta$. With spin source on (the extreme case of 100\% spin 

7351: polarization injected into the 

7352: $n$-region 

7353: is shown), the current is a strongly asymmetric 

7354: function of $\zeta$, displaying large GMR, shown by the dashed curve. 

7355: Materials parameters of GaAs were applied.

7356: Adapted from \onlinecite{Zutic2002:PRL}.}

7357: \label{fig:gmrd}

7358: \end{figure}

7359: 

7360: MBD's can  display an interesting  GMR-like effect, which 

7361: follows from Eq.~(\ref{eq:md}) \cite{Zutic2002:PRL}. The 

7362: current depends strongly on the relative orientation of the 

7363: nonequilibrium spin and the equilibrium magnetization. 

7364: Figure \ref{fig:gmrd} plots $j$,

7365: which also includes the contribution from holes, as a function of 

7366: $2q\zeta/k_B T$ for 

7367: both the unpolarized, $\delta P_n=0$, and fully polarized, $\delta P_n=1$, 

7368: $n$-region. 

7369: In the first case $j$ is

7370: a symmetric function of $\zeta$, increasing exponentially with increasing 

7371: $\zeta$ due to the increase in the 

7372: equilibrium minority carrier density $n_0(\zeta)$. 

7373: In unipolar systems, where transport is due to the majority carriers, 

7374: such a modulation of the current is not likely, as the majority carrier

7375: density is fixed by the density of dopants. 

7376: 

7377: If $\delta P_n \ne 0$, the current will depend on the sign of 

7378: $P_{n0}\cdot \delta 

7379: P_n$.  For parallel nonequilibrium (in the $n$-region)

7380: and equilibrium spins (in the $p$-region), most electrons cross 

7381: the depletion layer through the 

7382: lower barrier (see Fig.~\ref{fig:md}), increasing the current. 

7383: In the opposite case of antiparallel relative orientation, 

7384: electrons experience a larger barrier and the current is inhibited. 

7385: This is demonstrated in Fig.~\ref{fig:gmrd}

7386: by the strong asymmetry in $j$. The corresponding GMR ratio, the 

7387: difference between  $j$ for parallel and antiparallel orientations,

7388: can also be calculated analytically from Eq.~(\ref{eq:md}) as

7389: $2|\delta P_n P_{n0}|/(1-|\delta P_n P_{n0}|)$ \cite{Fabian2002:PRB}. 

7390: If, for example, $|P_{n0}|=|\delta 

7391: P_n|=0.5$, the relative change is 66\%. 

7392: The GMR effect should be useful for measuring

7393: the spin relaxation rate of bulk semiconductors \cite{Zutic2003:APL},

7394: as well as for detecting nonequilibrium spin in the nonmagnetic

7395: region of the {\it p-n} junction.\footnote{This could be a

7396: way to detect spin injection into Si, where optical detection

7397: is ineffective.}

7398: 

7399: Although practical MBD's are still to be fabricated and the predicted effects

7400: tested, magnetic {\it p-n} junctions have already been demonstrated. 

7401: Indeed, \textcite{Wen1968:IEEETM}\footnote{We thank M. Field for bringing this

7402: reference to our attention.} were perhaps

7403: the first to show that a ferromagnetic {\it p-n} junction, based on the

7404: ferromagnetic semiconductor CdCr$_2$Se$_4$ doped with Ag acceptors and In 

7405: donors, could act as a diode. Heavily doped p-GaMnAs/n-GaAs junctions

7406: were fabricated 

7407: \cite{Ohno2000:ASS,Kohda2001:JJAP,Johnston-Halperin2002:PRB,Arata2001:PE,%

7408: vanDorpe2003:P} 

7409: to demonstrate tunneling interband spin injection. 

7410: Incorporation of (Ga,Mn)As layer in the intrinsic region of {\it p-i-n}

7411: GaAs diode was shown to lead to an efficient photodiode, in which the

7412: Mn ions function as recombination centers \cite{Teran2003:APL}.

7413: It would be interesting to see such devices combined with a spin

7414: injector in the bulk regions.

7415: Recently, 

7416: \textcite{Tsui2003:APL}

7417: have shown that the current in p-CoMnGe/n-Ge magnetic heterojunction diodes 

7418: can indeed be controlled by magnetic field.

7419: To have functioning MBD's at room temperature, and to

7420: observe the above predicted phenomena, several important challenges

7421: have to be met:

7422: 

7423: (i) Zeeman or exchange splitting needs to sufficiently large to 

7424: provide equilibrium spin polarization,

7425: $\gtrsim 1-10\%$. This may be difficult at room temperature, unless

7426: the effective $g$ factor is $\sim 100$ at $B \sim 1$ T 

7427: (Sec.~\ref{sec:IID3}). The use of ferromagnetic semiconductors

7428: is limited by their $T_C$ (Sec.~\ref{sec:IVB}).

7429: 

7430: (ii) For a strong spin-charge coupling [recall the discussion of 

7431: Eq.~(\ref{eq:md})] a nondegenerate carrier density

7432: is desirable, which, while 

7433: likely in (Zn,Cr)Te, is not easily realized in many other ferromagnetic

7434: semiconductors that are typically heavily doped (Sec.~\ref{sec:IVB}).

7435: 

7436: (iii) An effective integration of magnetic and nonmagnetic structures 

7437: into single devices \cite{Samarth2003:SSC} is needed.

7438: 

7439: (iv) The samples need to be smaller than the spin diffusion lengths,

7440: requiring high carrier mobility and long spin relaxation (easier to 

7441: realize for spin-polarized electrons).

7442: 

7443: (v) The effects of actual device structures, such as two- and/or 

7444: three-dimensional spin flow, interface contacts, spin-dependent band offsets 

7445: and band bendings, strong spin relaxation in the depletion layers, etc., 

7446: will need to be understood.

7447: 

7448: By combining two magnetic {\it p-n} junctions in series one can

7449: obtain a magnetic bipolar transistor (Sec.~\ref{sec:IVE2}), a 

7450: three terminal device which offers spin-dependent amplification.

7451: 

7452: \subsection{\label{sec:IVE} Spin transistors}

7453: %---------------------------------------------------------------------------

7454: 

7455: We review several proposals for spin transistors 

7456: that have at least one semiconductor region and 

7457: that aim at integrating spin 

7458: and charge transport within traditional device schemes of the field-effect and 

7459: junction transistors. 

7460: Three important cases are discussed in detail: the Datta-Das  

7461: spin field-effect transistor, the magnetic bipolar transistor, and

7462: the hot-electron spin transistor. 

7463: 

7464: Various spin transistors that contain metallic (and insulating)

7465: regions have been proposed 

7466: \cite{Johnson1993:S,Bauer2003:APL,Zvezdin2003:TP,You2000:JAP}. 

7467: There is also a large category of

7468: the spin single-electron transistors, first realized by 

7469: \textcite{Ono1996:JPSJ}, and later investigated in 

7470: \cite{Barnas1998:PRL,Korotkov1999:PRB,Ciorga2002:APL,Martinek2002:PRB}. 

7471: Spin single-electron

7472: transistors can be viewed as an extension of magnetic tunneling 

7473: (see Sec.~\ref{sec:IVA2})

7474: to double tunnel junctions, where the Coulomb blockade becomes important 

7475: \cite{Takahashi1998:PRL}. For a review of spin single-electron transistors see 

7476: \cite{Maekawa:2002}. 

7477: 

7478: \subsubsection{\label{sec:IVE1} Spin field-effect transistors}

7479: %--------------------------------------------------------------------------------------------

7480: 

7481: \textcite{Datta1990:APL} proposed what became the prototypical spintronic 

7482: device scheme, 

7483: the Datta-Das spin field-effect transistor (SFET) (see Fig.~\ref{fig:DD}). 

7484: The device is based on spin injection and spin detection by a ferromagnetic 

7485: source and 

7486: drain, and on spin precession about the built-in structure inversion asymmetry 

7487: (Bychkov-Rashba) field ${\bf\Omega}$, Eq.~(\ref{eq:relax:BR}), 

7488: in the asymmetric, quasi-one-dimensional channel 

7489: of an ordinary field-effect transistor. The attractive feature of the

7490: Datta-Das SFET is that spin-dependent device operation is controlled

7491: not by external magnetic fields, but by gate bias, which controls the 

7492: spin precession rate. 

7493: 

7494: 

7495: The structure of the Datta-Das SFET is shown in Fig.~\ref{fig:DD}. 

7496: Consider a 2DEG  confined along the plane of the unit vector $\bf n$. 

7497: The precession axis of $\bf\Omega$

7498: lies always in the channel plane (see Sec.~\ref{sec:IIIB2}), so 

7499: the results 

7500: (unlike those for bulk inversion asymmetry) 

7501: are insensitive 

7502: to the relative orientation of ${\bf n}$ and the principal crystal axes. 

7503: Equation (\ref{eq:relax:BR}) determines the evolution of the 

7504: expectation value for a spin perpendicular to the plane, 

7505: $s_{n}={\bf s}\cdot {\bf n}$, and a spin parallel to the in-plane 

7506: $\bf k$, $s_{\parallel}={\bf s}\cdot{\bf k}/k$:  

7507: \begin{equation}

7508: ds_n/dt=2\alpha_{BR} k s_{\parallel}, \,\,\,\,

7509: ds_\parallel/dt=-2\alpha_{BR} k s_{n},

7510: \end{equation}

7511: where $\alpha_{BR}$ is the structure inversion asymmetry coefficient 

7512: appearing in 

7513: Eq.~(\ref{eq:relax:BR}).

7514: The average spin component along 

7515: $\bf\Omega$, $s_{\perp}={\bf s}\cdot ({\bf k} \times {\bf n})/k$, 

7516: is constant. 

7517: As a result, $s_\parallel=s_{0\parallel}\cos(\omega t)$, where 

7518: $\omega=2\alpha_{BR} k$ 

7519: and the injected spin at the source is labeled with zero. 

7520: If $\varphi$ is the angle between $\bf k$ and the source-drain axis,

7521: the electron will reach the drain at time $t'=L m_c/(\hbar k\cos\varphi)$, 

7522: with the spin 

7523: $s_\parallel$

7524: precessing at

7525: the angle $\phi=2\alpha_{BR} m_c L/\hbar$. The average spin at the drain 

7526: in the direction

7527: of magnetization is $s_{\parallel}(t')\cos\varphi +

7528: s_{0\perp}{\bf m}\cdot ({\bf k}\times {\bf n})$, so the current is 

7529: modulated by 

7530: $1 - \cos^2\varphi\sin^2(\phi/2)$, the probability of finding the spin in 

7531: the direction of magnetization $\bf m$.   

7532: 

7533: Note that $\phi$ does not depend on the momentum (or energy) of the 

7534: carriers. As 

7535: the spread 

7536: $\varphi$ in the momenta increases, the modulation effect decreases.  

7537: The largest effect is seen for $\varphi=0$, where the current modulation 

7538: factor is $\cos^2(\phi/2)$. It was therefore proposed~\cite{Datta1990:APL} 

7539: that $\varphi$

7540: be limited by further confining the electron motion along $\varphi=0$ using 

7541: a one-dimensional channel as a waveguide. 

7542: Spin modulation of the current becomes ineffective if transport is 

7543: diffusive. 

7544: Taking typical values~\cite{Nitta1997:PRL,Koga2002:PRL}

7545: for 

7546: $\hbar\alpha_{BR}\approx 1\times 10^{-11}$ eV$\cdot$m, and 

7547: $m_c=0.1 m_e$, current modulation should be observable at source-drain 

7548: separations of $L\agt 1$ $\mu$m, setting the scale for ballistic transport.

7549: The device will work best with 

7550: narrow-gap materials ~\cite{Lommer1988:PRL}

7551: like InAs, in which the structure inversion asymmetry dominates the spin 

7552: precession \cite{Luo1988:PRB,Luo1990:PRB,Das1989:PRB}. 

7553: Another option is using Si heterostructures, in which bulk inversion 

7554: asymmetry is

7555: absent. However, the small magnitude of the spin-orbit interaction makes 

7556: $\alpha_{BR}$ in Si probably rather weak.

7557: 

7558: The Datta-Das SFET is yet to be realized. There are at least four important

7559: difficulties in observing the proposed effects. 

7560: 

7561: (i) The effective spin 

7562: injection

7563: of spin-polarized carriers from the ferromagnetic source into a 2DEG  

7564: is nontrivial (see Sec.~\ref{sec:IID4}).

7565: 

7566: (ii) Ballistic spin-polarized transport should be realized through the channel

7567: with $\it uniform$ $\alpha_{BR}$ by eliminating undesirable electric fields 

7568: due to interface inhomogeneities. 

7569: 

7570: (iii) The parameter $\alpha_{BR}$ should 

7571: be effectively controllable by the gate. 

7572: 

7573: (iv) 

7574: The structure inversion asymmetry should dominate over the bulk inversion

7575: asymmetry, and the spin precession rate must be 

7576: large enough ($\hbar\alpha_{BR} \agt 10^{-11}$ eV$\cdot$m) to allow

7577: at least a half precession during the ballistic transport. 

7578: 

7579: These four factors present a great challenge to fabricating a 

7580: Datta-Das SFET at room temperature, limiting the design to 

7581: special materials and very clean interfaces.

7582: However, the modulation of $\alpha_{BR}$ 

7583: by biasing voltage (iii) 

7584: has been already convincingly demonstrated in 

7585: In$_{0.53}$Ga$_{0.47}$As/In$_{0.52}$Al$_{0.48}$As 

7586: QW's~\cite{Nitta1997:PRL,Hu1999:PRB,Grundler2000:PRL} 

7587: [for GaAs/AlGaAs 2DEG see

7588: also \cite{Miller2003:PRL}].  Initial experimental 

7589: investigations of magnetoresistance in the Datta-Das SFET systems 

7590: were performed by \textcite{Gardelis1999:PRB}. Recently spin precession

7591: in the the Datta-Das SFET, including the bulk inversion asymmetry,

7592: was investigated by \cite{Winkler2003:lanl} using $k\cdot p$

7593: model calculations. It is not surprising that the conductance through the

7594: transistor, in the present orientation-dependent bulk inversion asymmetry,

7595: depends rather strongly on the crystallographic orientation of the

7596: two-dimensional channel \cite{Lusakowski2003:PRB}. For more discussion

7597: of the Dresselhaus bulk inversion asymmetry and the Bychkov-Rashba structure

7598: asymmetry see Sec.~\ref{sec:IIIB2b}.

7599: 

7600: The Datta-Das SFET has generated great interest in mesoscopic spin-polarized 

7601: transport

7602: in the presence of structure inversion asymmetry. 

7603: Model calculations using the tight-binding formulation of $H_{\rm SIA}$

7604: (recall Sec.~II.B.2)

7605: were reported by \textcite{Pareek2002:PRB}.

7606: Further theoretical investigations on the theme of the Datta-Das spin 

7607: transistor

7608: can be found in \textcite{Matsuyama2002:PRB,Nikolic2001:P} 

7609: and in an extensive review by \textcite{Bournel2000:APF}. 

7610: Distinct SFET's have also been suggested, even in the absence of 

7611: ferromagnetic

7612: regions which are replaced by a rotating external magnetic field of 

7613: uniform strength

7614: \cite{Wang2002:P}. \textcite{Ciuti2002:APL} proposed 

7615: a ferromagnetic-oxide-semiconductor 

7616: transistor, with a nonmagnetic source and drain, 

7617: but with two ferromagnetic gates in series above 

7618: the base channel. The relative orientation of the gates' magnetization 

7619: leads to 

7620: magnetoresistance effects. An SFET that can operate in the diffusive regime, 

7621: in the presence of both bulk and structure inversion asymmetry, 

7622: has been considered by \textcite{Schliemann2002:P}.

7623: 

7624: 

7625: \subsubsection{\label{sec:IVE2} Magnetic bipolar transistor}

7626: %--------------------------------------------------------------------------

7627: 

7628: The magnetic bipolar transistor (MBT) is a bipolar transistor

7629: with spin-split carrier bands and, in general, an injected

7630: spin 

7631: \cite{Fabian2002:P,Fabian2003:Pa,Fabian2003:Pb}. 

7632: A related device structure

7633: was already proposed by \textcite{Gregg1997:JMMM} in a push for

7634: silicon-based spintronics. In this proposal (also called

7635: SPICE for spin-polarized injection current emitter)

7636: the semiconductors have no equilibrium spin, while

7637: the 

7638: spin source 

7639: is provided by a ferromagnetic spin

7640: injector attached to the emitter, and another ferromagnetic

7641: metal, a spin detector,

7642: is attached to the base/collector junction to modulate

7643: the current flow. In both configurations the aim is to

7644: control current amplification 

7645: by spin and magnetic field.

7646: 

7647: A scheme of a particular  MBT is shown in Fig.~\ref{fig:mbt}. 

7648: Such a three-terminal device can be thought of as consisting of 

7649: two magnetic {\it p-n} junctions connected in series. Materials 

7650: considerations discussed in Sec.~\ref{sec:IVD} also apply  

7651: to an MBT in order to provide a sufficient equilibrium polarization

7652: in a magnetic base $P_{B0}$. 

7653: While nonmagnetic, the emitter has a nonequilibrium polarization $\delta P_E$ 

7654: from a spin source, similar to the magnetic diode case in Fig.~\ref{fig:md}.

7655: Only the spin polarization of electrons is assumed. Applying the generalized

7656: Shockley theory to include spin 

7657: effects \cite{Fabian2002:PRB}, a theory of MBT was developed 

7658: by \textcite{Fabian2002:P,Fabian2003:Pb}. 

7659: Later, simplified schemes of MBT [not including the effect of nonequilibrium

7660: spin ($\delta P_E=0$)] were also considered by \textcite{Flatte2003:APL}

7661: and \textcite{Lebedeva2003:JAP}.

7662: 

7663: \begin{figure}

7664: \centerline{\psfig{file=zutic_fig32.eps,width=1\linewidth,angle=0}}

7665: \caption{

7666: Scheme of an {\it n-p-n} 

7667: magnetic bipolar transistor with magnetic base (B),

7668: nonmagnetic emitter (E), and collector (C). 

7669: Conduction and valence bands are separated by the energy gap $E_g$.

7670: The conduction band has a spin splitting $2q\zeta$, leading

7671: to equilibrium spin polarization $P_{B0}=\tanh(q\zeta/k_B T)$. 

7672: Carriers and depletion regions are represented as in Fig.~\ref{fig:md}.

7673: In the so called forward active regime, where the transistor

7674: can amplify currents, the E-B junction is forward biased (here

7675: with voltage $V_{BE}>0$ lowering the built-in potential $V_{bi}$), while the

7676: B-E junction is reverse biased ($V_{BC}<0$). The directions

7677: of the current flows are indicated. Electrons flow from E to B, 

7678: where they either recombine with holes (dashed 

7679: lines) or continue to be swept by the electric field in the B-E 

7680: depletion layer towards C. Holes form mostly the base current, $I_B$,

7681: flowing to the emitter. The current amplification

7682: $\beta=I_C/I_B$ 

7683: can be controlled by $P_{B0}$ 

7684: as well as by the nonequilibrium spin in E. 

7685: Adapted from \onlinecite{Fabian2003:Pa}.

7686: } 

7687: \label{fig:mbt}

7688: \end{figure}

7689: The current amplification (gain) $\beta=I_C/I_B$ (see Fig.~\ref{fig:mbt})

7690: is typically $\sim 100$ in practical transistors.

7691: This ratio depends on many factors, such as the doping 

7692: densities, carrier lifetimes,  

7693: diffusion coefficients,

7694: and structure geometry. 

7695: In an MBT $\beta$ also depends on the spin splitting 

7696: $2q \zeta$ (see Fig.~\ref{fig:mbt})

7697: and the nonequilibrium polarization $\delta P_E$. 

7698: This additional dependence of $\beta$

7699: in an MBT is called magnetoamplification \cite{Fabian2003:Pb}.

7700: An important prediction is that the nonequilibrium spin can be 

7701: injected at low bias 

7702: all the way from the emitter, through the base, to the collector 

7703: \cite{Fabian2002:P,Fabian2003:Pb}

7704: in order to make possible an

7705: effective control of $\beta$ by $\delta P_E$. 

7706: 

7707: The calculated dependence of the gain on the spin splitting for 

7708: $\delta P_E=0.9$ is 

7709: shown in Fig.~\ref{fig:gain}, for GaAs and Si materials parameters. 

7710: The gain is

7711: very sensitive to the equilibrium magnetization in Si, while the 

7712: rapid carrier 

7713: recombination in GaAs prevents more effective control of the transport 

7714: across the base.

7715: In Si it is the spin injection at the emitter-collector depletion 

7716: layer which 

7717: controls the current. As the spin-charge coupling is most effective 

7718: across 

7719: the depletion layer (see Sec.~\ref{sec:IVD}), this 

7720: coupling is essential for the current in Si. 

7721: In the limit of slow carrier recombination \cite{Fabian2002:P},

7722: \begin{equation} \label{eq:beta}

7723: \beta \sim \cosh(q\zeta/k_B T) (1 + \delta P_E P_{B0}).  

7724: \end{equation}

7725: Both magnetic field (through $\zeta$) and nonequilibrium spin affect the 

7726: gain, an implication of the spin-voltaic effect

7727: \cite{Zutic2002:PRL,Zutic2003:P}.

7728: The sensitivity of the current to spin can be used to measure  

7729: the injected spin polarization. If no 

7730: spin source 

7731: is present ($\delta P_E=0$),

7732: there is no spin-charge coupling in the space-charge regions, unless

7733: at least two regions are magnetic. The only remaining effects 

7734: on the $I-V$ characteristics come from the sensitivity of the carrier 

7735: densities 

7736: in the equilibrium  

7737: magnetic regions to $\zeta$ [see Eq.~(\ref{eq:beta})

7738: for the case of $\delta P_E=0$]. 

7739: 

7740: \begin{figure}

7741: \centerline{\psfig{file=zutic_fig33.eps,width=1\linewidth,angle=0}}

7742: \caption{Calculated gain dependence of an MBT as a function of

7743: base spin splitting $2q\zeta$, given in units of thermal energy 

7744: $k_B T$.  The nonequilibrium

7745: spin polarization in the emitter is $\delta P_E=0.9$. 

7746: Si (solid) and GaAs (dashed) 

7747: materials parameters were applied.

7748: Adapted from \onlinecite{Fabian2002:P}. 

7749: } 

7750: \label{fig:gain}

7751: \end{figure}

7752: 

7753: The MBT is, in effect, a magnetic heterostructure transistor, since its 

7754: functionality depends

7755: on tunability of the band structure of the emitter, base, or collector. The 

7756: advantage 

7757: of MBT, however, is that the band structure is not built-in, but can be tuned 

7758: {\it during} 

7759: the device operation by magnetic field or spin injection signals. 

7760: The challenges to demonstrate the predicted phenomena in MBT are 

7761: similar to those of magnetic bipolar diodes, see Sec.~\ref{sec:IVD}.

7762: 

7763: \subsubsection{\label{sec:IVE3} Hot-electron spin transistors}

7764: %-----------------------------------------------------------------------

7765: 

7766: \label{hotspin}

7767: Spin transistors that

7768: rely on transport of hot (nonthermalized) carriers

7769: have the potential to serve of several different purposes. 

7770: On the one hand, they could be used as a diagnostic tool to characterize

7771: spin- and energy-dependent interfacial properties, scattering processes, 

7772: and electronic structure, relevant to spintronic 

7773: devices.\footnote{These efforts 

7774: are motivated in part by the success of (spin-insensitive) 

7775: ballistic-electron-emission microscopy in providing high spatial and

7776: energy resolution of 

7777: properties of metal/semiconductor interfaces 

7778: \cite{Kaiser1988:PRL,Bonnell:2001,Smith2000:PRB}. A subsequent

7779: variation--a ballistic-electron-magnetic microscopy, which also uses an STM 

7780: tip to inject hot carriers, is capable of resolving magnetic features at 

7781: a $\sim10$ nm length scale \cite{Rippard1999:APL,Rippard2000:PRL}.}

7782: On the other hand, hot-electron transistors are also of interest for

7783: their ability to sense magnetic fields, their possible

7784: memory applications, 

7785: and a their potential as a source

7786: of ballistic hot-electron spin injection.

7787: Below we discuss two representative examples,

7788: a spin-valve transistor and a magnetic tunneling transistor.

7789: 

7790: The spin-valve or Monsma transistor provided an early demonstration of a

7791: hot-electron 

7792: spin transistor and realization of a hybrid spintronic

7793: device that integrates metallic ferromagnets and semiconductors 

7794: \cite{Monsma1995:PRL,Monsma1998:S}.

7795: A three terminal structure 

7796: \footnote{Similar to other hot-electron 

7797: spin devices, the term

7798: transistor characterizes their three-terminal structure rather

7799: than the usual functionality of a conventional semiconductor transistor.

7800: In particular, a semiconductor bipolar transistor,

7801: which also has an emitter/base/collector structure, typically has

7802: a sizable current gain--a small change in the base current 

7803: leads to a large change in the collector current (see Sec.~\ref{sec:IVE2}). 

7804: However, only a small

7805: current gain $\sim2$ (due to large current in a metal base)

7806: was predicted in magnetic tunnel-junction-based devices \cite{Hehn2002:PRB}.}

7807: consisted of a metallic base (B) made

7808: of a ferromagnetic multilayers in a CPP geometry [as depicted in 

7809: Fig.~\ref{gmr:1}(a)] surrounded by a silicon emitter (E) and 

7810: collector (C) with two Schottky contacts, formed at E/B and B/C 

7811: interfaces.\footnote{Another realization of a spin-valve transistor

7812: combines a GaAs emitter with a Si collector \cite{Dessein2000:JAP}.} 

7813: Forward bias $V_{EB}$ controls the emitter current $I_E$

7814: of spin-unpolarized electrons, which are injected

7815: into a base region as hot carriers. 

7816: The scattering processes in the base,

7817: together with the reverse bias $V_{BC}$,

7818: influence how many of the injected electrons can overcome the B/C Schottky

7819: barrier and contribute to the collector current $I_C$. 

7820: Similar to the physics of GMR structures~\cite{Levy:2002,Gijs1997:AP}

7821: scattering in the base region strongly depends on the relative orientation 

7822: of the magnetizations in the ferromagnetic layers.

7823: %iz check majority direction

7824: Electrons with spin which has magnetic moment opposite (antiparallel)

7825: to the magnetization of a ferromagnetic layer typically are scattered

7826: more than electrons with parallel magnetic moments,

7827: resulting

7828: in a spin-filtering effect which can be described in terms of spin-dependent

7829: mean free path % $l_\lambda$, $\lambda=\uparrow,\downarrow$ 

7830: \cite{Rendell1980:PRL,Pappas1991:PRL,Hong2000:PRB}. 

7831: Generally, both elastic and 

7832: inelastic

7833: scattering processes determine the effective spin-dependent mean free path,

7834: sometimes also referred to as the attenuation length.\footnote{For electrons 

7835: with sufficiently high excess energy, a scattering process

7836: (influencing the mean free path) does not necessarily remove the electron

7837: from the collector current. The attenuation length, which can be determined

7838: by measuring the base layer thickness dependence of the collector current

7839: [see 

7840: \cite{Rippard1999:APL,Rippard2000:PRL,Vlutters2001:PRB,vanDijken2002:PRB}]

7841: can therefore differ from the effective mean free path.}

7842: The magnetorestive response is usually expressed using magneto current (MC),

7843: defined as the change in collector current, normalized to the 

7844: minimum value

7845: \begin{equation}

7846: MC=(I_{C\uparrow \uparrow} -I_{C\uparrow \downarrow})/

7847: I_{C\uparrow \downarrow},

7848: \label{eq:mc}

7849: \end{equation}

7850: analogous to the expression for GMR or TMR structures

7851: [recall Eq.~(\ref{eq:tmr})],

7852: where $\uparrow \uparrow$ (parallel) and $\uparrow \downarrow$ 

7853: (antiparallel) denote the relative orientation of the magnetizations.

7854: The large values of MC\footnote{These values substantially 

7855: exceed the CPP GMR value 

7856: for the same magnetic multilayer used in the base.} ($>$200\%) 

7857: and the sensitivity of $\sim$130\% per G measured  at room temperature 

7858: \cite{AnilKumar2000:JMMM} demonstrate a capability for magnetic-field sensors. 

7859: Several important challenges, raised by the operation of

7860: the spin-valve transistor, need to be addressed to better realize 

7861: the potential of hot-electron 

7862: transistors. These challenges 

7863: include increasing the small collector current and determining 

7864: whether the spin injection of hot carriers into semiconductors is 

7865: feasible. Furthermore, it would be desirable to fabricate

7866: structures in which semiconductor regions played an active

7867: role, not limited to energy selection (via Schottky barriers)

7868: of the carriers injected into the base and collector regions.

7869: 

7870: \begin{figure} 

7871: \centerline{\psfig{file=zutic_fig34.eps,width=0.85\linewidth,angle=0}}

7872: \caption{Schematic energy diagram of a magnetic tunneling transistor.

7873: Region 1 is the emitter, region 2 

7874: the Al$_2$O$_3$ tunnel barrier of height $\phi$, and region 3 the base.

7875: Together they form a magnetic tunnel junction.

7876: Region 4 is a semiconductor collector that has a Schottky barrier

7877: at the interface with the base. 

7878: From \onlinecite{vanDijken2003:PRL}. 

7879: }

7880: \label{mtt}

7881: \end{figure}

7882: 

7883: An alternative class of hot-electron transistors, often referred to as 

7884: magnetic 

7885: tunneling transistors, has a tunneling junction instead of a Schottky 

7886: barrier 

7887: emitter

7888: \cite{Mizushima1997:IEEETM,Yamauchi1998:PRB,Sato2001:APL,vanDijken2003:PRL,%

7889: vanDijken2002:PRB,Jiang2003:PRL,vanDijken2002:APL}. The

7890: addition of a tunnel junction, combined with a variable $V_{EB}$, allows

7891: exploration of   

7892: hot-electron transport over an energy range

7893: of several eV. At large $V_{EB}$ bias, the ratio $I_C/I_E$, important 

7894: for the device performance, can be substantially increased 

7895: over that of

7896: the 

7897: spin-valve transistor 

7898: \cite{vanDijken2003:APLa,vanDijken2003:APLb,Sato2001:APL}.

7899: 

7900: A particular realization is depicted in Fig.~\ref{mtt}. Different

7901: coercive fields in the regions 1 and 3 ensure independent switching of

7902: of the corresponding magnetizations in the applied

7903: magnetic field. The magnetocurrent MC, defined in Eq.~(\ref{eq:mc}),

7904: shows a nonmonotonic behavior with $V_{EB}$ \cite{vanDijken2003:PRL}

7905: influenced by the conduction-band structure of a collector. 

7906: In GaAs, in addition to the direct conduction band minimum

7907: at the $\Gamma$ point [recall Fig.~\ref{oo:1} (a)], there

7908: are indirect minima at $L$ points at higher energy \cite{Blakemore1982:JAP}.

7909: After an initial decrease of MC with electron energy, 

7910: at $V_{EB} \approx 0.3$ V 

7911: larger than the base/collector Schottky barrier

7912: there is an onset of hot-electron transport 

7913: into $L$ valleys 

7914: accompanied by an increase in MC \cite{vanDijken2003:PRL}. 

7915: 

7916: A large magnetocurrent alone, measured in various hot-electron 

7917: spin transistors 

7918: \cite{Monsma1995:PRL,Monsma1998:S,Sato2001:APL,vanDijken2003:APLb},

7919: is not sufficient to demonstrate spin injection in a semiconductor 

7920: collector. 

7921: For conclusive evidence spin detection in a collector region is needed. 

7922: This was first achieved \cite{Jiang2003:PRL} 

7923: using optical detection with a spin LED 

7924: structure\footnote{Analogous to the spin LED from Fig.~\ref{injexp:3},

7925: in which GaAs collector served as an $n$-type spin aligner and InGaAs/GaAs

7926: was used for a quantum well.}

7927: added to the collector in Fig.~\ref{mtt}.

7928: Measurements at $T=1.4$ K and $B=2.5$ T, after a background subtraction, 

7929: showed majority spin injection  with $P_{\rm circ} \approx 10$ \%.

7930: 

7931: In another realization of a magnetic tunnel transistor, more similar

7932: to the original spin-valve transistor, the emitter was nonmagnetic (Cu) 

7933: while the

7934: base was a magnetic multilayer (F1/N/F2) \cite{vanDijken2003:APLb}.

7935: The resulting strong spin-filtering effect can be inferred  from the 

7936: transmitted hot carriers with a

7937: spin-dependent exponential decay  within the F$_i$, $i=1,2$ layer.

7938: Unpolarized electrons, injected from the emitter, 

7939: after passing an F1 layer of thickness $t$ acquire an effective

7940: transmitted polarization 

7941: \begin{equation}

7942: P_{N1}=\frac{N_\uparrow-N_\downarrow}

7943: {N_\uparrow+N_\downarrow}=

7944: \frac{e^{-t/l_\uparrow}-e^{-t/l_\downarrow}}

7945: {e^{-t/l_\uparrow}+e^{-t/l_\downarrow}},

7946: \end{equation}

7947: where 

7948: $N_\uparrow$ and $N_\downarrow$ 

7949: represent the number of transmitted electrons

7950: with majority or minority spin and $l_\uparrow$ and $l_\downarrow$

7951: are the corresponding attenuation length (the polarization $P_{N2}$ has

7952: an analogous form). 

7953: The resulting magnetocurrent can be

7954: expressed as \cite{vanDijken2003:APLb}

7955: \begin{equation}

7956: MC=2 P_{N1}P_{N2}/(1-P_{N1} P_{N2}),

7957: \label{eq:mc2}

7958: \end{equation}

7959: which is analogous to Eq.~(\ref{eq:julliere}) 

7960: for TMR using Julli{\`{e}}re's model,

7961: but with the redefined definition of spin polarization. 

7962: At $V_{EB}=0.8$ V and at $T=77$ K, the measured

7963: MC exceeds $3400$ \%, while  with Eq.~(\ref{eq:mc2}) the

7964: polarization of the transmitted electrons can be estimated to exceed 90 \%,

7965: even with a ferromagnet only a $\sim 3$ nm thick \cite{vanDijken2003:APLb}.

7966: A theoretical analysis of spin injection and spin filtering in

7967: magnetic tunneling transistors was given by \textcite{Rashba2003:P}

7968: who extended the approach for ballistic spin injection 

7969: \cite{Kravchenko2003:PRB}

7970: (Sec.~\ref{sec:IIC2}) to include the effects of hot-electron transport

7971: and inelastic scattering.

7972: 

7973: Future studies of hot-electron 

7974: spin transistors are expected to

7975: result in increased spin injection even at room temperatures and

7976: to utilize other semiconductor collectors. It would be particularly 

7977: desirable to demonstrate hot-electron 

7978: spin injection in Si

7979: and facilitate an integration with the CMOS technology.

7980: 

7981: 

7982: \subsection{\label{sec:IVF} Spin qubits in semiconductor nanostructures}

7983: %-------------------------------------------------------------------------

7984: 

7985: A potentially revolutionary idea in spintronics is the possibility

7986: of using the two-level nature of 

7987: electron spin to create a solid-state quantum computer 

7988: \cite{DiVincenzo1995:S,DasSarma2001:SSC,Nielsen:2000}.  

7989: The basic unit in a quantum computer is the quantum bit (or qubit), 

7990: the quantum

7991: analog of the binary bit in a classical digital computer.

7992: A qubit is essentially

7993: a controllable quantum two-level system 

7994: \cite{DasSarma2001:AS,Nielsen:2000}.  

7995: While the dimensionality (2$^n$) of the Hilbert space of $n$ electron spins

7996: is the same as the number of  configurations of a corresponding classical

7997: system, a quantum system can be

7998: in a superposition of all the basis states, effectively performing 

7999: (via a unitary evolution) many classical computations in parallel.  

8000: Several spin-based quantum computer

8001: schemes have been proposed and extensively studied\footnote{See, for example,  

8002: \cite{Loss1998:PRA,Privman1998:PL,Kane1998:N,Hu2000:PRA,Koiller2002:PRL,%

8003: Koiller2003:PRL,Friesen2003:PRB,Piermarocchi2002:PRL,DiVincenzo2000:FP,%

8004: Kane2000:FP,Hu2001:PRA,Burkard1999:PRB,Troiani2003:PRL,%

8005: Skinner2003:PRL,Levy2002:PRL,Meier2003:PRL,Vrijen2000:PRA}.}  

8006: A common theme in these proposals is the idea of

8007: manipulating the dynamics of a single (or a few) electron spin(s) in

8008: semiconductor nanostructures (e.g., quantum dots), with the reasonable

8009: hope that the predicted behavior will extend to many-spin systems,

8010: requisite for practical quantum computation. 

8011: 

8012: \begin{figure}

8013: \centerline{\psfig{file=zutic_fig35.eps,width=1\linewidth,angle=0}}

8014: \caption{The \textcite{Loss1998:PRA} proposal for spin-based solid-state 

8015: quantum computing. Electrons are localized in

8016: electrostatically defined quantum dots, with coupling between 

8017: electron spins---via the exchange interaction---allowed by tunneling 

8018: between the dots. The tunneling is controlled by gate

8019: voltage. The figure shows two electrons localized in the regions defined

8020: by the gates (shaded). Single-qubit

8021: operations are realized by single-spin precessions (circles), 

8022: performed by applying 

8023: local magnetic fields (here perpendicular to the page) to each dot. 

8024: Two-qubit operations are done 

8025: through the exchange interaction indicated by the dashed curves. 

8026: The scheme works according to the time-dependent Hamiltonian 

8027: $H(t)=\sum_{i,j}' J_{i,j}(t) {\bf S}_i\cdot {\bf S}_j 

8028: +\mu_B g \sum {\bf B}_i(t) {\bf S}_i$, 

8029: where the first summation, over all neighboring spin pairs, 

8030: describes the local

8031: exchange interaction ($J$ is the exchange coupling), 

8032: while the second describes

8033: the single spin operations by local magnetic fields. Variations to 

8034: this scheme are described by \textcite{Burkard2000:FP}.  

8035: }

8036: \label{fig:QC}

8037: \end{figure}

8038: 

8039: The control of spin dynamics and entanglement [many-spin quantum 

8040: correlations \cite{Nielsen:2000}]

8041: at the single-spin level in a semiconductor quantum dot structure is a 

8042: formidable task,

8043: which has not been achieved even at mK temperatures, 

8044: although impressive experimental advances have recently been made

8045: \cite{Fujisawa2002:N,Elzerman2003:PRB,Hanson2003:lanl}.

8046: The current architectures for spin-based quantum computing 

8047: employ GaAs quantum dots \cite{Loss1998:PRA} or

8048: Si (or Si-Ge) systems \cite{Kane1998:N}, with different 

8049: variations. 

8050: The basic idea (see Fig.~\ref{fig:QC}) is to manipulate

8051: the spin states of a single electron using external magnetic fields (or

8052: microwaves) for single-qubit operations and to utilize the quantum exchange

8053: coupling between two neighboring electrons to carry out two-qubit

8054: operations.  

8055: 

8056: State-of-the-art 

8057: techniques, to measure a single spin in a solid,

8058: such as magnetic resonance force microscopy 

8059: \cite{Sidles1995:RMP,Mamin2003:P,Barbic2002:JAP} 

8060: or the spin-selective single-electron-transistor spectroscopy

8061: \cite{Ono2002:S}, are 

8062: still not sensitive enough 

8063: for quantum computing operations.  

8064: However, recently a single shot readout of the spin of an individual electron

8065: has been demonstrated using an

8066: electrical pump-probe measurement \cite{Kouwenhoven2004:PC}.

8067: A single electron with an unknown spin was trapped in a quantum dot

8068: for a few miliseconds. At the end of the trapping time the spin was

8069: measured by quickly shifting the Zeeman resolved spin states towards

8070: the Fermi energy. A spin to charge conversion allowed for an

8071: electrical readout of the spin.

8072: 

8073: The real motivation for using the spin

8074: as a qubit is its long coherence time, microseconds or longer

8075: in operational experimental conditions \cite{Sousa2003:PRB}, to be

8076: contrasted with typical picosecond coherence times for charge or orbital

8077: states in solids. 

8078: Interest in spin-based quantum computing

8079: will only increase as we understand

8080: more about spin coherence and relaxation from other spintronic studies. 

8081: The broad

8082: subject of spin-based quantum computing, 

8083: which is related to the areas of quantum measurement and quantum 

8084: decoherence \cite{Zurek2003:RMP}

8085: is beyond the scope of 

8086: this review.

8087: 

8088: 

8089: 

8090: \section{\label{sec:V} Outlook}

8091: 

8092: 

8093: We have reviewed selected topics on spintronics, emphasizing both the

8094: fundamental aspects of spin dynamics, transport,

8095: and relaxation, and the potential applications.

8096: While the current push for spintronics

8097: is driven by the prospect of technological applications, the fundamental

8098: spin physics, which has a longstanding tradition in the solid-state

8099: community, is by itself exciting and worth pursuing. Furthermore,

8100: even though many proposed spintronic device schemes may turn out

8101: to be impractical in the end, their importance lies in stimulating

8102: interesting experimental and theoretical research.

8103: 

8104: There are many challenges and open questions to be tackled by future

8105: research, in particular a robust

8106: spin injection into silicon.\footnote{Small signals attributed to spin 

8107: injection have already been reported \cite{Jia1996:IEEE}.}

8108: While GaAs is of great technological

8109: importance, the control of spin in silicon would raise hopes for

8110: seamless integration of spintronics with the current

8111: information technology. In addition, the small magnitude of the

8112: spin-orbit interaction and the absence of inversion symmetry

8113: lead to relatively long room-temperature spin lifetimes (of about

8114: 10 ns; see Sec.~\ref{sec:IIID1}), relaxing some constraints on the

8115: operational length and time scales. Important materials advances

8116: have been realized in improving the compatibility of Si/III-V

8117: structures \cite{Sieg1998:APL}

8118: suggesting a possibility that the existing control

8119: of spin in GaAs or in III-V ferromagnetic semiconductors might be

8120: extended to Si.

8121: 

8122: Future progress in spin-polarized transport will be largely driven

8123: by the materials advances. In the context of semiconductors, considering

8124: all-semiconductor structures rather than the hybrid structures with

8125: metallic ferromagnets will depend on the improvements in ferromagnetic

8126: semiconductors, for example, whether they 

8127: can achieve higher mobility, 

8128: higher Curie temperature,\footnote{There still remain many challenges in

8129: accurately predicting Curie temperature. First principles results 

8130: suggest that dominant models of ferromagnetism in semiconductors

8131: cannot be used to explain a variation of Curie temperature across

8132: different materials \cite{Erwin2004:NM}. For reviews of ferromagnetic

8133: semiconductor theories outside the scope of this article, see,

8134: for example, \textcite{Nagaev:1983}; \textcite{Bhat2002:JS};

8135: \textcite{Dietl2002:SST}; \textcite{Sanvito2002:JS}, 

8136: \textcite{DasSarma2003:PRB}; \textcite{Koning:2003}; 

8137: \textcite{Timm2003:JPCM}.}

8138: and a simple fabrication of high quality

8139: interfaces with nonmagnetic materials. What is missing, even in the

8140: currently available materials, is a systematic understanding of the

8141: effects of magnetic interfaces and materials inhomogeneities on 

8142: spin-polarized transport. A comprehensive transport calculation

8143: in the actual devices

8144: with realistic electronic structure of

8145: the studied materials would provide valuable insights into 

8146: both the  spin polarization being measured and

8147: how it is reduced from the moment it was generated.

8148: 

8149: Spin relaxation and spin dephasing of conduction electrons

8150: is a rather field, with the basic principles well understood.

8151: What is needed are accurate band-structure-derived calculations

8152: of spin relaxation times, in both metals and semiconductors.

8153: The same can be said of

8154: $g$ factor, calculation of which from first

8155: principles is a nontrivial task that has not been accomplished even

8156: for the elemental metals. An important and still debated issue

8157: is spin relaxation and decoherence of localized or

8158: confined electrons, where the hyperfine-interaction mechanism

8159: dominates. Furthermore, single-spin relaxation and

8160: decoherence, and their relation to the ensemble spin dephasing,

8161: need to be 

8162: pursued further 

8163: in the context of quantum computing. 

8164: A first step towards understanding

8165: single-spin relaxation  is the recent experiment of

8166: \textcite{Hanson2003:lanl} in a one-electron quantum dot.

8167: 

8168: While dynamic nuclear polarization induced by electron spin can often

8169: be a nuisance for detecting intrinsic spin dynamics 

8170: (see Sec.~\ref{sec:IIID3}),

8171: the interaction between electron and nuclear spins 

8172: \cite{Paget:1984,Fleisher:1984,Vagner2003:P,Smet2002:N} 

8173: is of fundamental

8174: importance for spintronics. An NMR of the nuclear spin polarized

8175: by spin-polarized photoexcited electrons has already been used to detect

8176: the nonequilibrium electron spin in Si \cite{Lampel1968:PRL}. On the

8177: other hand, an NMR signal can be detected optically through measuring

8178: changes in the circular polarization of photoluminescence

8179: due to resonant variations of the nuclear field 

8180: \cite{Dyakonov:1984}, as shown first in

8181: p-doped Ga$_{0.7}$Al$_{0.3}$As \cite{Ekimov1972:JETPL}. 

8182: The early work of \textcite{Lampel1968:PRL}, 

8183: and \textcite{Ekimov1972:JETPL} established the basic

8184: principles for a series of experiments that demonstrated various 

8185: realizations of an all-optical NMR. The role of the resonant 

8186: radio waves is played by periodically optically excited electron spins

8187: \cite{Salis2001:PRL,Kalevich1986:FTT,Kalevich1981:FTT,Kalevich1980:FTT,%

8188: Kikkawa2000:S,Eickhoff2002:PRB,Fleisher:1984b}.

8189: Electron-nuclear spintronics is likely to become relevant for

8190: quantum computation and for few-spin manipulations, which can benefit

8191: from long nuclear spin coherence times (even lasting minutes).

8192: 

8193: The range of potential spintronic applications 

8194: goes beyond the use of

8195: large magnetoresistive

8196: effects.

8197: \textcite{Rudolph2003:APL}, for example, have demonstrated the operation of

8198: a spin laser. The laser is a vertical-cavity surface-emitting laser 

8199: (VCSEL), optically pumped in the gain medium, here two InGaAs quantum

8200: wells, with 50\% spin-polarized electrons. The electrons recombine with 

8201: heavy holes, which are effectively unpolarized, emitting circularly

8202: polarized light (see Sec.~\ref{sec:IIB}). The threshold electrical current, 

8203: extracted from the pump power for the lasing operation, was found to be 

8204: 0.5 A$\cdot$cm$^{-2}$,

8205: which is 23\% below the threshold current of the spin-unpolarized VCSEL.

8206: Furthermore, for a fixed pump power, the emission power of the laser

8207: changed by 400\% upon changing the degree of circular polarization of

8208: the pump laser. The reason for the decrease in threshold is the selective

8209: coupling of spin-polarized electrons to photons with one helicity.

8210: While the experiment was conducted at 6 K, a room-temperature

8211: operation and an electrically pumped laser should be viable as 

8212: well.\footnote{The requirement is that the spin relaxation time be longer 

8213: than

8214: the carrier recombination time, and that the spin injection spot and the

8215: gain medium be within the spin transport length.}

8216: 

8217: \begin{figure}

8218: \centerline{\psfig{file=zutic_fig36.eps,width=1.0\linewidth}}

8219: \caption{(a) Nanopilar device: schematic diagram of

8220: a nanopilar device

8221: operating at room temperature. 

8222: The direction of magnetization

8223: is fixed (pinned) in the thick Co film and free in the thin Co film;

8224: (b) differential resistance $dV/dI$ of a nanopilar device as

8225: a function of applied field; (c) $dV/dI$ of the device as a

8226: function of applied current. 

8227: The arrows in panels (b) and (c) represent the direction of

8228: magnetic field and current sweeps, respectively. 

8229: For positive current, 

8230: electrons flow from the thin to the thick Co film.

8231: Adapted from \onlinecite{Albert2002:PRL}.} 

8232: \label{ralph}

8233: \end{figure}

8234: 

8235: The demonstration that the flow of spin-polarized carriers, rather

8236: than applied magnetic field, can also be used to manipulate

8237: magnetization of ferromagnetic materials brings the exciting

8238: prospect of a novel class of spintronic devices. In addition

8239: to reversal of magnetization, which is a key element in realizing

8240: various

8241: magnetoresitive applications, the driving of a spin-polarized current

8242: can lead to coherent 

8243: microwave oscillations in nanomagnets \cite{Kiselev2003:P}. 

8244: Spin-transfer

8245: torque (Sec.~\ref{sec:IB1}) 

8246: has already been realized in several 

8247: experimental geometries. These include nanowires 

8248: \cite{Wegrowe1999:EL,Kelly2003:PRB}, 

8249: point contacts 

8250: \cite{Tsoi1998:PRLa,Yi2003:PRL,Tsoi2000:N,Tsoi2002:PRL}, 

8251: nanoconstrictions \cite{Myers1999:S,Rippard2003:P}, 

8252: and nanopilars \cite{Katine2000:PRL,Albert2002:PRL,Urazhdin2003:PRL} 

8253: (see Fig.~\ref{ralph}), all involving metallic ferromagnets.

8254: The common feature of all these geometries is a need for very 

8255: large current densities ($\sim10^7$ Acm$^{-2}$).  

8256: Ongoing

8257: experiments \cite{Munekata2003:PC,Chiba2004:P,Yamanouchi2004:N}

8258: to demonstrate 

8259: spin-transfer torque 

8260: (together with other cooperative phenomena) 

8261: in ferromagnetic semiconductors, which have

8262: much smaller magnetization than their metallic counterparts, are expected

8263: to also require much smaller switching currents. Based on the findings

8264: in electric-field controlled ferromagnetism (see Fig.~\ref{ohno}),

8265: it has been demonstrated that the reversal of magnetization in (In,Mn)As can

8266: be manipulated by modifying the carrier density, using a gate voltage

8267: in a FET structure \cite{Chiba2003:S}.

8268: 

8269: \acknowledgments{We thank R. H. Buhrman, K. Bussmann, R. de Sousa, 

8270: M. I. D'yakonov,  

8271: S. C. Erwin, M. E. Fisher, A. M. Goldman, K. Halterman, X. Hu, S. V. Iordanskii,

8272: M. Johnson, B. T. Jonker, K. Kavokin, J. M. Kikkawa,  

8273: S. Maekawa, C. M. Marcus, I. I. Mazin, B. D. McCombe, J. S. Moodera, 

8274: H. Munekata, 

8275: B. E. Nadgorny, H. Ohno, 

8276: S. S. P. Parkin,  

8277: D. C. Ralph, E. I. Rashba, W. H. Rippard, V. I. Safarov, G. Schmidt, 

8278: E. Sherman, 

8279: R. H. Silsbee, D. D. Smith, 

8280: M. D. Stiles, O. T. Valls, S. van Dijken, 

8281: H. M. van Driel, T. Venkatesan, S. von Moln\'{a}r, and J. Y. T. Wei 

8282: for useful discussions.  

8283: We thank D. D. Awschalom, N. C. Emley, R. T. Harley, M. Johnson, B. T. Jonker, 

8284: K. Kavokin, J. M. Kikkawa, C. M. Marcus, H. Munekata, Y. Ohno, A. Oiwa, 

8285: S. S. P. Parkin, R. J. Soulen, Jr.,  and M. Tanaka for providing us with 

8286: illustrative

8287: figures from their published works, and A. Kaminski and D. J. Priour

8288: for help with preparation of the manuscript. This work was supported by 

8289: DARPA, the US ONR, and the NSF-ECS.} 

8290: 

8291: \begin{table}

8292: \begin{tabular}{ll}

8293: \hline

8294: \hline

8295: 2DEG  & two dimensional electron gas \\

8296: BAP   & Bir-Aronov-Pikus \\

8297: BCS   & Bardeen Cooper Schrieffer \\

8298: BIA   & bulk inversion asymmetry \\

8299: BTK   & Blonder-Tinkham-Klapwijk \\

8300: CESR  & conduction electron spin resonance \\

8301: CIP   & current in plane \\

8302: CMOS  & complementary metal oxide semiconductor \\

8303: CMR   & colossal magnetoresistance \\

8304: CPP   & current perpendicular to plane \\

8305: DNA   & deoxyribonucleic acid \\

8306: DP    & D'yakonov-Perel' \\

8307: EDSR  & electron dipole spin resonance \\

8308: EMF   & electromotive force \\

8309: ESR   & electron spin resonance \\

8310: EY    & Elliot-Yafet \\

8311: F     & ferromagnet \\

8312: FSm   & ferromagnetic semiconductor \\

8313: GMR   & giant magnetoresistance \\

8314: FET   & field effect transistor  \\

8315: HFI   & hyperfine interaction \\

8316: I     & insulator \\

8317: HTSC  & high temperature superconductor \\

8318: LED   & light emitting diode \\

8319: MBD   & magnetic bipolar diode \\

8320: MBE   & molecular beam epitaxy \\

8321: MBT   & magnetic bipolar transistor \\

8322: MC    & magnetocurrent \\

8323: MR    & magnetoresistance \\

8324: MRAM  & magnetic random access memory \\

8325: MTJ   & magnetic tunnel junction \\

8326: N     & normal (paramagnetic) metal \\

8327: NMR   & nuclear magnetic resonance \\

8328: QD    & quantum dot \\

8329: QPC   & quantum point contact \\

8330: QW    & quantum well \\

8331: S     & superconductor \\

8332: SET   & single electron transistor  \\

8333: SFET  & spin field effect transistor  \\

8334: SIA   & structure inversion asymmetry \\

8335: Sm    & semiconductor \\

8336: SQUID & superconducting interference quantum device \\

8337: STM   & scanning tunneling microscope \\

8338: TESR  & transmission electron spin resonance \\

8339: \hline

8340: \hline

8341: \end{tabular}

8342: \caption{List of acronyms used in the text.}

8343: \label{tab:2}

8344: \end{table}

8345: 

8346: \bibliographystyle{apsrmp}

8347: \bibliography{references}

8348: 

8349: \end{document}

8350: