1: \documentclass[aps,prl,twocolumn,groupedaddress]{revtex4}
2: %\documentclass[twocolumn,preprintnumbers,amsmath,amssymb]{revtex4}
3: %showpacs
4: \usepackage{graphicx}
5: \usepackage{dcolumn}
6: \usepackage{bm}
7: \begin{document}
8:
9: \title{Spin transference and magnetoresistance amplification in a transistor}
10: \author{H.~Dery}\email{hdery@ucsd.edu}
11: \author{\L.~Cywi{\'n}ski}
12: \author{L.~J.~Sham}
13: \affiliation{Department of Physics, University of California San
14: Diego, La Jolla, California, 92093-0319}
15: %%%%%%%%%%%%%%%%%%
16: \date{\today}
17:
18: \begin{abstract}
19: A current problem in semiconductor spin-based electronics is the
20: difficulty of experimentally expressing the effect of spin-polarized
21: current in electrical circuit measurements. We present a theoretical
22: solution with the principle of transference of the spin diffusion
23: effects in the semiconductor channel of a system with three magnetic
24: terminals. A notable result of technological consequences is the
25: room temperature amplification of the magneto-resistive effect,
26: integrable with electronics circuits, demonstrated by computation of
27: current dependence on magnetization configuration in such a system
28: with currently achievable parameters.
29: \end{abstract}
30: %\pacs{72.25.Dc, 72.25.Mk,85.75.-d}
31: \maketitle
32: %%%%%%%%%%%%%%%%%%
33:
34: Giant magnetoresistance effect has been discovered in
35: heterostructures of ferromagnetic and paramagnetic metal layers
36: \cite{Baibich_PRL88,Binasch_PRB89}. Similar effect has been observed
37: in magnetic tunnel junctions \cite{Moodera_PRL95}. Applications in
38: the spin valve configuration has given rise to important products
39: such as hard disk read-heads and magnetic memories
40: \cite{Prinz_Science98}. Research into spin polarized currents in
41: semiconductors leads to the new field of semiconductor spintronics
42: \cite{Wolf_Science00,Zutic_RMP04} with promise of increased logic
43: functionality of electronic circuits and integration with
44: non-volatile magnetic memory. While room temperature injection from
45: a ferromagnet via a tunnel barrier into a semiconductor has produced
46: reasonable current spin polarization
47: \cite{Hanbicki_APL02,Hanbicki_APL03,Jiang_PRL05,Adelmann_PRB05}, the
48: magneto-resistive effect in a semiconductor spin valve with
49: ferromagnetic metal contacts is predicted to be small
50: \cite{Fert_PRB01,Rashba_EPJB02}.
51: % ABOVE: WITH FM CONTACT HAS BEEN ADDED AT REQUEST OF REF B
52: In order to introduce
53: additional control over the spin-polarized carrier flow, a number of
54: semiconductor-based spin-transistors have been proposed
55: \cite{DattaDas_APL90,Schliemann_PRL03,Ciuti_APL02,Zutic_PRL02,Flatte_APL03},
56: using ideas such as Rashba effect, effective spin reflection,
57: half-metallic ferromagnets, or minority carrier action in junctions
58: with a magnetic semiconductor.
59:
60: In this Letter we present an electrical means of expressing the spin
61: effects, rather than optical means
62: \cite{Crooker_PRL05,Stephens_PRL04}. We study the diffusive spin
63: currents of a nonmagnetic semiconductor (SC) with three
64: ferromagnetic metal (FM) terminals, each capable of injecting or
65: extracting spin-polarized currents. The currents flowing between the
66: contacts depend on the alignment of their magnetizations. The
67: magneto-resistive (MR) effect is defined as a relative change of the
68: current upon flipping of one of the magnetization vectors. In the
69: diffusive regime considered here, the MR effect comes from spin
70: accumulation in the semiconductor due to the spin selectivity of the
71: contacts. The profiles of non-equilibrium spin densities in the
72: semiconductor depend on the magnetic configuration, and the
73: resulting different diffusion currents are the cause of MR. Using
74: more than two terminals provides the capability of transference of
75: spin effects from one controlled diffusion region to another. This
76: transference under voltage and magnetization control will be shown
77: to lead to amplification of the magneto-resistive effect. This
78: control provided by a third conducting and biased terminal is in
79: contrast to the ``nonlocal'' spin valve geometry
80: \cite{Jedema_Nature01,Johnson_Science93}, where the additional
81: contact is a floating voltage probe.
82: \begin{figure}
83: \includegraphics[height=4cm,width=8.6cm]{MCT_scheme_new.eps}
84: \caption{A schematics of the proposed three-terminal device. The
85: channel indicates the current flow region of $n$-doped semiconductor
86: grown on top of an insulating substrate.} \label{fig:scheme}
87: \end{figure}
88:
89: An illustrative application of these ideas is a device we term
90: magnetic contact transistor (MCT), shown in Fig~\ref{fig:scheme}.
91: The principle behind such a system is analogous to the familiar
92: bipolar transistor. Here the two spin diffusion channels whose
93: populations can change by spin-flip take the place of electrons and
94: holes whose populations are changed by recombination. In the latter
95: case, the transistor action results from attaching two p-n diodes
96: back to back into a pnp or npn structure, in which the common base
97: width must be smaller than the recombination diffusion length. A
98: longer base would beget two uncoupled diodes devoid of amplification
99: effects. The two-terminal device in spintronics is a spin-valve, in
100: which the current depends on the relative magnetization directions
101: of two magnetic contacts. By connecting two spin valves with a
102: common source terminal (the middle contact in
103: Fig.~\ref{fig:scheme}), whose width is smaller than spin-diffusion
104: length $L_{sc}$, we create a system capable of amplifying the MR
105: effect of each spin valve. In contrast to the bipolar transistor,
106: the spin current is driven by spin diffusion rather than charge
107: diffusion. In addition to the current control by applied voltages as
108: in the conventional transistor, this spin transistor has control of
109: the spin components by the magnetization configurations of the
110: ferromagnetic electrodes. One strategy in amplifying the MR effect
111: is based on the ability to use the voltage adjustment to yield a
112: zero or near zero current in one magnetic configuration of the
113: ferromagnetic electrodes, and the ability to produce an easily
114: measurable current at the same voltage arrangement by changing the
115: magnetic configuration. The former is simple circuit theory and
116: requires no spin physics but the latter is a result of intricate
117: spin transport physics in the transistor configuration. In our
118: scheme, it requires the coupling of three contacts via spin
119: currents, equivalent to coupling spin transport of two spin valves.
120: The ratio of the ``on'' current to the ``off'' current will be shown
121: below to be robust against electrical noise and to decay with
122: increasing width of the middle contact or, equivalently, the
123: distance between two spin valves beyond the spin diffusion length,
124: indicative of the essential role of spin.
125:
126: The system shown in Fig.~\ref{fig:scheme} is a planar structure. Low
127: mesas beneath the FM terminals are heavily doped, making the
128: Schottky barriers thin ($<$10 nm). As an illustration here, we let
129: the left drain (L) be a ``soft'' magnetic layer whose magnetization
130: can be easily flipped. The middle source (S) and the right drain
131: (R), are magnetized in the same direction. The P and AP
132: configurations denote, respectively, the L magnetization parallel
133: and antiparallel to that of the S and R. The currents $J_{L(R)}$ are
134: flowing in the L (R) part of the channel, and are measured in the
135: L(R) contacts, which are kept at separately controlled voltages
136: $V_{L(R)}$. The required magnetic properties of contacts can be
137: achieved either by pinning the magnetization of middle and right
138: terminals, or by exploiting the contacts' magnetic shape anisotropy,
139: in order to manipulate their coercivities. The magnetization of the
140: left drain can be controlled by external magnetic field (then our
141: system works as a sensor of $B$ field), or by a field created by a
142: pickup current in a wire above the contact. Analogously to CMOS
143: transistors, the MCT has the capacity for digital operation that can
144: be used to trigger a pickup current of another MCT, thus
145: transferring the information from one magnet to another, leading to
146: a new paradigm of computation \cite{Prinz_Science98,Ney_Nature03}.
147:
148: We assume the system to be homogeneous in the $z$ direction (see
149: Fig.~\ref{fig:scheme}) and consider two-dimensional spin diffusion
150: in the semiconductor:
151: \begin{equation}
152: \nabla^2 \mu_{s}(x,y) = \frac{\mu_{s}(x,y) -
153: \mu_{-s}(x,y)}{2L_{sc}^2} \,\, , \label{eq:diffusion}
154: \end{equation}
155: where $\mu_{s}$ is the spin dependent electrochemical potential with
156: $s$=$\pm$ denoting the spin species, and the spin diffusion length
157: $L_{sc}=\sqrt{D\tau_{sp}}$, where $D$ is the diffusion constant and
158: $\tau_{sp}$ is the spin relaxation time. This equation is well known
159: for paramagnetic metals \cite{Hershfield_PRB97}, and holds for
160: non-degenerate semiconductors considered here when the electric
161: field is small \cite{Yu_Flatte_long_PRB02}. This condition is
162: fulfilled here because of the presence of highly resistive barriers
163: at the FM/SC interface. In the following calculations the electric
164: driving force on the current is shown to have negligible influence
165: on spin diffusion. Due to vastly different resistances in metals and
166: semiconductors, we can neglect the spin and spatial dependence of
167: the electrochemical potential in the ferromagnets and replace
168: $\mu_{s}^{FM}$ by a constant given by the bias voltage. Thus, we
169: only need to solve the diffusion equation inside the semiconductor
170: channel. We neglect interfacial spin scattering and use Ohm's law
171: across the contacts, so that the boundary conditions are:
172: \begin{eqnarray}
173: ej^i_s=\frac{\sigma_{sc}}{2}\Big(\widehat{n}\cdot\nabla\mu_{s}\Big)
174: \!=\! \left\{\!\!\!
175: \begin{array}{ll} -G^i_s (eV^i + \mu_s), & \text{contacts} \\ \\ \qquad 0, &
176: \text{otherwise},
177: \end{array} \right. \label{eq:boundaries}
178: \end{eqnarray}
179: where $j^i_s$ denotes the spin $s$ current through the $i^{th}$
180: barrier interface, $\sigma_{sc}$ the semiconductor conductivity,
181: $\widehat{n}$ the outward interface normal and $G^i_{s}$ the
182: spin-dependent barrier conductance. The spin-selective properties of
183: the tunneling barrier ($G_+$$\neq$$G_-$) dominate the spin injection
184: physics \cite{Rashba_PRB00}. This approach is valid for contacts
185: without depletion beyond the thin doped tunneling barrier
186: \cite{Albrecht_PRB03}. Elsewhere we will derive from the 2D
187: diffusion an effective 1D formalism for the lateral spin transport
188: inside a thin layer, with finite-sized metal contacts on top of it.
189: % ON TOP OF IT instead of "on top of the layer" (repetition)
190: The calculations presented below were done with both methods giving
191: essentially the same results. Confining the spatial extent of spin
192: accumulation to the terminals' footprint is optimal for MR
193: amplification. It is achieved by etching away the semiconductor
194: peripheries in Fig.~\ref{fig:scheme}, or by selectively doping the
195: channel with ion implantation.
196:
197: The barrier conductances $G_{s}$ have been calculated for Fe/GaAs
198: system in the simple single-band effective mass model
199: \cite{tunneling_phenomena}. We have assumed triangular Schottky
200: barriers of $\sim$7 nm thickness. At low applied voltages we obtain
201: the conductances of the order of $10^{2}$
202: ($10^{3}$)$\,\Omega^{-1}$cm$^{-2}$ for reverse (forward) bias. The
203: ratio of reverse to forward bias conductance $f$ is taken to be 0.5,
204: a value poorer than the theoretical estimate of 0.2. The ratio of
205: spin-up to spin-down conductance is $G_{+}/G_{-}\simeq 2$. The
206: corresponding spin-injection efficiency coefficient
207: $\alpha=(G_{+}-G_{-})/(G_{+}+G_{-})$ is $30$\% as seen in circular
208: polarization degree of spin light emitting diodes
209: \cite{Hanbicki_APL02,Hanbicki_APL03}.
210: % THE VALUE OF ALPHA HAS BEEN ADDED AT REQUEST OF REF A
211: The spin selectivity of the
212: interface is governed by the ratio of Fermi velocities for up and
213: down spins in Fe. While this simple model lacks details of the
214: atomic structure of the interface \cite{Butler_JAP97}, both the
215: $G_{+}/G_{-}$ ratio and the orders of magnitude of $G_{s}$ are in
216: agreement with the experimental results.
217:
218: It is instructive to look at the behavior of one half of the
219: semiconductor channel decoupled from the other as a two-terminal
220: spin valve. We define $MR$$=$$(J^P-J^{AP})/J^{P}$, with $J^{P}$
221: ($J^{AP}$) denoting the total current through the structure for
222: parallel (antiparallel) alignment of magnetizations of the two
223: terminals. We consider a GaAs layer at room temperature with a
224: conservative value of $L_{sc}$$=$$1~\mu$m. It corresponds to a
225: spin-independent mobility of $5000$ cm$^{2}/$(Vs) and spin
226: relaxation time $\tau_{s}$$=$$80$ ps in the non-degenerate regime at
227: room temperature \cite{Optical_Orientation}. The free carrier
228: concentration is $n$$=$$4\cdot10^{15}$cm$^{-3}$. The lower spin
229: conductance of the forward-biased barrier
230: $G_{b}=1000$$\,\Omega^{-1}$cm$^{-2}$.
231: % DEFINITION AND VALUE OF GB GIVEN ABOVE
232: The dimensions are the following: the contact width $w$ and the
233: separation between the contacts $d$ are both 200 nm, and the
234: thickness of the semiconductor channel is 100 nm. For these
235: parameters we obtain the spin-valve MR$\approx$3\%. Such a small
236: effect is hard to measure and useless for applications. The weak
237: effect can be understood from a simplified 1D transport picture
238: where the analytical solutions are easily available
239: \cite{Fert_PRB01,Rashba_EPJB02}. Although such analysis ignores the
240: lateral geometry issues it gives a qualitative description within an
241: order of magnitude estimate. Even for Schottky barriers as thin as
242: the ones considered here, and for non-degenerate semiconductors, the
243: effective conductance $G_{sc}$$=$$\sigma_{sc}$$/$$L_{sc}$ is much
244: higher than $G_{b}$. The calculation gives, to the lowest order in
245: $G_{b}/G_{sc}$, $MR\sim \alpha^2 \frac{G_{b}}{G_{sc}}
246: \frac{L_{sc}}{d} \frac{f}{f+1}$. In the two-terminal case the
247: difference between P and AP configurations is accommodated by
248: different non-equilibrium spin-density profiles, with a very small
249: change in the total current. Although $J^{P}\simeq J^{AP}$, the
250: electrochemical potential splitting $\Delta \mu =
251: |\mu_{+}^{SC}-\mu_{-}^{SC}|$ near the contact follows the ratio:
252: \begin{equation}
253: \frac{\Delta \mu^{P}}{\Delta \mu^{AP} } \propto
254: \Big(\frac{d}{2L_{sc}}\Big)^{2} \ll 1 \label{eq:delta_mu},
255: \end{equation}
256: whereas the mean value of $\mu_s$ does not change visibly between P
257: and AP. Our three terminal scheme makes effective use of
258: Eq.~(\ref{eq:delta_mu}).
259:
260: \begin{figure}
261: \includegraphics[height=5cm,width=8.6cm]{JR_AP_new.eps}
262: \caption{ \footnotesize{Right drain current density versus source
263: width for the left drain magnetization parallel (P) and
264: antiparallel (AP) to the other two magnets. $V_L=0.1V$ and $V_R$ is
265: adjusted for each $w_{s}$ so that $J_R^{P}=0$. Upper and lower
266: curves of each set show the current densities for $V_R\pm$0.2~mV.
267: The dashed lines are the values for $J^{P}_R$ around zero current.
268: The solid and dotted lines show the currents for $J^{AP}_R$ for two
269: different spin-flip times. The drain width, separation between the
270: contacts and the channel thickness are
271: \protect{$w$=$d$=$2h$=$0.2~\mu$m}. }}\label{fig:transistor_effect}
272: \end{figure}
273:
274: For a certain ratio $V_R/V_L$, the right contact current in the P
275: configuration can be quenched ($J_{R}^{P}=0$), while the AP current
276: remains finite. Thus, the readout of the magnetic configuration is
277: digitized. From simple circuit theory, when the barrier resistances
278: dominate, the voltage ratio $V_R/V_L$ which quenches $J_{R}^{P}$ is
279: estimated to be $1/(rf+1)$, where $r$ is the ratio of source to
280: drain widths. We examine the tolerance to error or noise by varying
281: this ratio by $\pm 0.2\%$, consistent with the Johnson noise for the
282: barrier resistance with contact area of 1~$\mu$m$^2$ at sub-GHz
283: frequencies. Fig.~\ref{fig:transistor_effect} shows the noise
284: margins for the right terminal current as function of the width of
285: the source contact $w_{s}$. For large source width, equivalent to
286: uncoupled spin valves, the MR amplification effect is lost. For
287: ultra-narrow contacts the effect is also small due to the increased
288: resistance of the contact \cite{Fert_PRB01}. In-between there is an
289: optimal value of $w_{s}$ (well below $L_{sc}$) where the MR effect
290: measured in the right contact easily exceeds hundreds of percent.
291: The resulting current densities calculated at room temperature are
292: of the order of 1 A/cm$^{2}$, which could be directly measured in
293: sub-micron contacts or further amplified, leading to a robust
294: read-out of the left drain magnetization direction.
295: % TYPICAL LEFT CURRENT ADDED HERE (TO ADDRESS QUESTION 6 OF REF A):
296: During the read-out the left contact current density is about 100
297: A/cm$^{2}$ for the voltages used in
298: Fig.~\ref{fig:transistor_effect}.
299:
300: % ONE LONG PARAGRAPH SPLIT INTO TWO HERE:
301: The amplification of the MR effect depends on the robustness of a
302: finite current $J_{R}^{AP}$ when the magnetic configuration is
303: changed to AP. An explanation based on our calculation is as
304: follows. The source of the different currents is the difference in
305: the spin splitting of the electrochemical potential, $\Delta$$\mu$,
306: at the R contact in the P and AP configurations at the optimal
307: voltage ratio. In P, $\Delta$$\mu^{P}$ is small. In AP,
308: $\Delta$$\mu^{AP}$ is large, for the reason in
309: Eq.~(\ref{eq:delta_mu}), which still qualitatively holds in the MCT.
310: % ADDED ``which still qualitatively holds in the MCT''
311: This disrupts the balance of the spin
312: currents on two sides of the source, resulting in a sizable current,
313: $J_R^{{AP}} \sim (G_{+}-G_{-})\Delta \mu^{{AP}}$. In the
314: semiconductor channel, the electrochemical potential of one spin
315: population rises a good fraction above $-eV_R$ and the other one
316: drops an equal amount below. Thus, the current is fully
317: spin-polarized. Fig.~\ref{fig:2d_resolved} shows the two dimensional
318: spatial distribution of the $x$-component current densities in the
319: semiconductor channel in the $100$~nm section left of the right
320: contact. The middle contact width is at the optimal value shown in
321: Fig.~\ref{fig:transistor_effect}.
322: % SENTENCE BELOW ADDED TO AVOID READERS CONFUSION WHEN CAREFULLY COMPARING THE NUMBERS FROM FIGS 2 AND 3
323: Note that the difference in total current densities between
324: Figures~\ref{fig:transistor_effect} and \ref{fig:2d_resolved} comes
325: from the ratio of channel thickness $h$ and R contact width $w$. In
326: Fig.~\ref{fig:2d_resolved}, the zero and 4 A/cm$^2$ respective
327: values in the middle of the color scale for the left and the right
328: upper panels show the zero current in P and a robust finite current
329: in AP. The amplification of the spin current from P to AP is visible
330: in the lower panels. By adding and subtracting the current densities
331: in the upper and lower panels of each column, one can see that the
332: spin currents in the P case flow in the opposite directions at the
333: opposite edges of the channel, resulting in the zero net charge
334: current.
335:
336: \begin{figure}
337: \includegraphics[height=6cm,width=8.6cm]{JR_2D.eps}
338: \caption{ \footnotesize{ (color). Spatially resolved $x$ component
339: of the current density at the right half of the right channel. Note
340: the different scales for each figure. The upper panels show the zero
341: and finite net charge currents for the parallel and antiparallel
342: configurations, respectively ($0$ versus $\sim$$3.5$Acm$^{-2}$). The
343: lower panels show the amplification of $\Delta J$$=$$j_+$$-$$j_-$
344: due to the difference in electrochemical potential splitting. The
345: parameters are as in Fig.~\ref{fig:transistor_effect} with optimal
346: source width and $\tau_{sp}$$=$$80$ ps. }}\label{fig:2d_resolved}
347: \end{figure}
348:
349: In practical designs, the ``soft'' L contact might have different
350: properties from the remaining two. The exact left-right symmetry is
351: not required for the digital effect; any asymmetry between the
352: terminals can be counteracted by adjusting the voltages. Moreover,
353: the zero current state can be stabilized using the feedback loops
354: adjusting $V_{L}$ and $V_{R}$. Although the current from the L
355: contact is larger than the read-out R current, the L output needs
356: not be wasted. It can, for instance, be used to amplify the output
357: from the R contact using external circuits.
358:
359:
360: In all the modeling presented we have used experimentally available
361: properties of FM/SC structures. The formalism and ideas used here
362: can be applied as well to an all-metallic system, with a
363: paramagnetic metal channel. However, vastly different parameters
364: makes the control of the MR amplification effect more difficult. The
365: requirements of having a sizable AP electrochemical potential
366: splitting, of being able to control the voltages with desired
367: accuracy, and of keeping the current densities at reasonable levels,
368: are hard to reconcile. Thus, the semiconductor-based system is more
369: naturally suited for demonstration of voltage-controlled spin
370: transference.
371:
372:
373:
374: In summary, we have constructed a room temperature theory of the
375: transference of the spin polarization in the current between a pair
376: of electrodes to another pair provided they are connected by spin
377: diffusion. The three-terminal structure effectively exploits the
378: spin accumulation in the semiconductor channel. One result, the
379: amplification of the magneto-resistive effect by voltage control,
380: may be of practical importance as the electrical read-out of
381: magnetic memory integrable to an electronics circuit. Such a device
382: may be used in a scheme of ``magnetic computation'', working as a
383: building block of reprogrammable logic gate \cite{Ney_Nature03}.
384: Such a synergy of information processing and non-volatile storage
385: represents a possible fulfillment of the promises of spintronics.
386: Further properties which may result from the spin transference need
387: to be explored in the future.
388:
389: \begin{acknowledgments}
390: This work is supported by NSF under Grant No. DMR-0325599.
391: \end{acknowledgments}
392:
393: %\bibliography{refs}
394: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%
395: \begin{thebibliography}{99}
396: \bibitem{Baibich_PRL88} M. N. Baibich \textit{et al.}, Phys. Rev. Lett. \textbf{61}, 2472 (1988).
397: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
398: \bibitem{Binasch_PRB89} G. Binasch, P. Gr{\"u}nberg, F. Saurenbach, and W. Zinn, Phys. Rev.
399: B \textbf{39}, R4828 (1989).
400: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
401: \bibitem{Moodera_PRL95} J. S. Moodera, L. R. Kinder, T. M. Wong, and R. Meservey, Phys. Rev.
402: Lett. \textbf{74}, 3273 (1995).
403: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
404: \bibitem{Prinz_Science98} G. A. Prinz, Science \textbf{282}, 1660 (1998).
405: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
406: \bibitem{Wolf_Science00} S. A. Wolf \textit{et al.}, Science \textbf{294}, 1488 (2001).
407: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
408: \bibitem{Zutic_RMP04} I. {\u Z}uti{\'c}, J. Fabian, and S. D. Sarma, Rev. Mod.
409: Phys. \textbf{76}, 323 (2004).
410: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
411: \bibitem{Hanbicki_APL02} A. T. Hanbicki \textit{et al.}, Appl. Phys.
412: Lett. \textbf{80}, 1240 (2002).
413: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
414: \bibitem{Hanbicki_APL03} A. T. Hanbicki \textit{et al.}, Appl. Phys.
415: Lett. \textbf{82}, 4092 (2003).
416: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
417: \bibitem{Jiang_PRL05} X. Jiang \textit{et al.}, Phys. Rev. Lett. \textbf{94}, 056601 (2005).
418: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
419: \bibitem{Adelmann_PRB05} C. Adelmann, X. Lou, J. Strand, C. J. Palmstrom, and P. A. Crowell, Phys. Rev. B \textbf{71}, 121301(R) (2005).
420: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
421: \bibitem{Fert_PRB01} A. Fert and H. Jaffr{\`es}, Phys. Rev. B \textbf{64}, 184420 (2001).
422: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
423: \bibitem{Rashba_EPJB02} E. I. Rashba, Eur. Phys. J. B \textbf{29}, 513 (2002).
424: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
425: \bibitem{DattaDas_APL90} S. Datta and B. Das, Appl. Phys. Lett \textbf{56}, 665 (1990).
426: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
427: \bibitem{Ciuti_APL02} C. Ciuti, J. P. McGuire, and L. J. Sham, Appl. Phys. Lett \textbf{81}, 4781 (2002).
428: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
429: \bibitem{Zutic_PRL02} I. {\u Z}uti{\'c}, J. Fabian and S. DasSarma, Phys. Rev. Lett. \textbf{88}, 066603 (2002).
430: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
431: \bibitem{Flatte_APL03} M. E. Flatt{\'e}, Z. G. Yu, E. Johnston-Halperin
432: and D. D. Awschalom, Appl. Phys. Lett \textbf{82}, 4740 (2003).
433: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
434: \bibitem{Schliemann_PRL03} J. Schliemann, J. C. Egues and D.
435: Loss, Phys. Rev. Lett. \textbf{90}, 146801 (2003).
436: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
437: \bibitem{Crooker_PRL05} S. A. Crooker and D. L. Smith, Phys. Rev. Lett. \textbf{94}, 236601 (2005).
438: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
439: \bibitem{Stephens_PRL04} J. Stephens, J. Berezovsky, J. P. McGuire, L. J. Sham, A. C. Gossard, D. D. Awschalom, Phys. Rev. Lett. \textbf{93}, 097602 (2004).
440: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
441: \bibitem{Jedema_Nature01} F. J. Jedema, A. T. Filip, and B. J. van Wees, Nature \textbf{410}, 345
442: (2001).
443: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
444: \bibitem{Johnson_Science93} M. Johnson, Science \textbf{260}, 320
445: (1993).
446: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
447: \bibitem{Ney_Nature03} A. Ney, C. Pampuch, R. Koch, and K. H. Ploog, Nature \textbf{425}, 485
448: (2003).
449: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
450: \bibitem{Hershfield_PRB97} S. Hershfield and H. L. Zhao, Phys. Rev. B \textbf{56}, 3296
451: (1997).
452: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
453: \bibitem{Yu_Flatte_long_PRB02} Z. G. Yu and M. E. Flatt{\'e}, Phys. Rev. B \textbf{66}, 235302
454: (2002).
455: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
456: \bibitem{Rashba_PRB00} E. I. Rashba, Phys. Rev. B \textbf{62}, R16267 (2000).
457: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
458: \bibitem{Albrecht_PRB03} J. D. Albrecht and D. L. Smith, Phys. Rev. B \textbf{68}, 035340
459: (2003).
460: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
461: \bibitem{tunneling_phenomena} \textit{Tunneling Phenomena}, edited by E. Burnstein and S. Lundqvist (Plenum Press, New York,
462: 1969).
463: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
464: \bibitem{Butler_JAP97} W. H. Butler \textit{et al.}, J. Appl. Phys. \textbf{81}, 5518
465: (1997).
466: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
467: \bibitem{Optical_Orientation} \textit{Optical Orientation}, edited by F. Meier and B. P. Zakharchenya (Nort-Holland, New York,
468: 1984).
469: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
470: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
471: \end{thebibliography}
472: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
473:
474: \end{document}
475: