1: \documentclass[aps,amsmath,prb,twocolumn,a4paper,showpacs,floatfix]{revtex4}
2: \usepackage{graphicx}
3: \newcommand{\comment}[1]{}
4: \newcommand{\beq}{\begin{equation}}
5: \newcommand{\eneq}{\end{equation}}
6: \newcommand{\bea}{\begin{eqnarray}}
7: \newcommand{\enea}{\end{eqnarray}}
8: \newcommand{\bc}{\begin{center}}
9: \newcommand{\ec}{\end{center}}
10: \newcommand{\vt}{ \varphi}
11: \newcommand{\ct}{ \cos \Theta}\newcommand{\st}{ \sin \Theta}
12: \newcommand{\cz}{ \cos \chi}
13: \newcommand{\sz}{ \sin \chi}
14: \newcommand{\cphi}{ \cos \Phi}
15: \newcommand{\sphi}{ \sin \Phi}
16: \newcommand{\nn}{\nonumber}
17: \newcommand{\wc}{\omega_c}
18: \newcommand{\met}{\frac{1}{2}}
19: \newcommand{\ddt}[1]{\frac{#1}{dt}}
20: \newcommand{\df}{\dot {\varphi}}
21: \newcommand{\dP}{\dot {\Phi}}
22: \newcommand{\dTh}{\dot {\Theta}}
23: \newcommand{\spin}[2]{|{#1}_1,{#2}_2 \rangle }
24: \newcommand{\mr}{\mathrm}
25: \newcommand{\cmt}{ \cos \frac{ \vartheta}{2}}
26: \newcommand{\smt}{ \sin \frac{ \vartheta}{2}}
27: \newcommand{\cmqt}{ \cos ^2\frac{ \vartheta}{2}}
28: \newcommand{\smqt}{ \sin ^2\frac{ \vartheta}{2}}
29: \newcommand{\bu}{\bar{u}}
30: \newcommand{\bv}{\bar{v}}
31: \newcommand{\UU}{\breve{{\cal{U}}}}
32: \newcommand{\GG}{\breve{G}}
33: \newcommand{\ovr}{\overline{R}}
34: \newcommand{\wo}{\omega_o}
35: \newcommand{\wu}{\omega_1}
36: \newcommand{\drond}[1]{\frac{\partial}{\partial #1} }
37: \newcommand{\EE}{\mathbf{E}}
38: \newcommand{\pp}{\mathbf{p}}
39: \newcommand{\sig}{\mathbf{\sigma}}
40: \newcommand{\rr}{\mathbf{r}}
41: \newcommand{\CRE}[1]{C^{\dagger}_{#1}}
42: \newcommand{\DES}[1]{C_{#1}}
43: \newcommand{\LAG}[3]{L_{#1}^{#2}\left(#3 \right)}
44: \newcommand{\COE}{\mathcal{C}}
45: \newcommand{\h}{\hbar}
46: \newcommand{\bd}{\begin{displaymath}}
47: \newcommand{\ed}{\end{displaymath}}
48: \newcommand{\de}{\delta}
49: \newcommand{\f}{\varphi}
50: \newcommand{\ww}{\omega}
51: \newcommand{\DD}{\Delta}
52: \newcommand{\la}{\vec \lambda}
53: \newcommand{\ug}{\underline{\hat{g}}}
54: \newcommand{\up}{\uparrow}
55: \newcommand{\da}{\downarrow}
56: \newcommand{\hd}{\hat d}
57: \newcommand{\hsp}[2]{\langle {#1}_1,{#2}_2 | }
58: \newcommand{\T}{\mathcal{T}}
59: \newcommand{\ang}{ i\omega _o t }
60: \newcommand{\bwt}{\begin{widetext}}
61: \newcommand{\ewt}{\end{widetext}}
62: \begin{document}
63:
64: \title{Quantum Rings with Rashba spin orbit coupling: a path integral
65: approach}
66:
67: \author{P. Lucignano$^{1,2}$}\author{D. Giuliano$^{1,3}$}
68: \author{A. Tagliacozzo$^{1,2}$}
69:
70: \affiliation{$^1$ Dipartimento di Scienze Fisiche Universit\`a degli
71: studi di Napoli "Federico II", Napoli, Italy}
72: \affiliation{$^2$ Coherentia-INFM,
73: Monte S.Angelo - via Cintia, I-80126 Napoli, Italy}
74: \affiliation{$^3$ Dipartimento di Fisica, Universit\`a della Calabria and
75: I.N.F.N., Gruppo collegato di Cosenza, Arcavacata di Rende
76: I-87036, Cosenza, Italy}
77:
78:
79:
80: \pacs{03.65.Vf,% Phases: geometric, dynamic, topological
81: 72.10.-d,% Electronic transport and scattering mechanisms
82: 73.23.-b,% Electronic transport in mesoscopic system
83: 71.70.Ej % SO Coupling, Zeeman, Stark...
84: }
85: \date{\today}
86: \begin{abstract}
87: We employ a path integral real time approach to compute the DC
88: conductance and spin polarization for electrons transported across a
89: ballistic Quantum Ring with Rashba spin-orbit interaction. We use a
90: piecewise semiclassical approximation for the particle orbital motion
91: and solve the spin dynamics exactly, by accounting for both Zeeman
92: coupling and spin-orbit interaction at the same time. Within our
93: approach, we are able to study how the interplay between Berry phase,
94: Ahronov Casher phase, Zeeman interaction and weak localization
95: corrections influences the quantum interference in the conductance
96: within a wide range of externally applied fields. Our results are
97: helpful in interpreting recent measurements on interferometric rings.
98: \end{abstract}
99:
100: \maketitle
101:
102: \section{Introduction}
103:
104: In classical physics, a charged particle moving in an external
105: magnetic field $B$ feels a (Lorentz) force only in the regions in
106: which $B$ is different from zero. Instead, in 1959 Ahronov and Bohm
107: (AB)\cite{aharonovbohm} showed that the wave packet representing the
108: state of a quantum particle can be influenced by an external vector
109: potential $\vec{A}$, even if the corresponding magnetic field is zero,
110: provided that the particle is moving in a space with a nontrivial
111: topology (holes). To be more specific, the wavefunction of the charged
112: particle may acquire a nonzero phase, when undergoing a closed path in
113: a space threaded by an external magnetic flux.
114:
115: Nowadays, mesoscopic quantum rings (QR's) allow to have direct access to
116: the phase of the electron wavefunction, since their size is smaller
117: than the distance over which the phase randomizes, as a consequence of
118: scattering against impurities and interactions. The total rate of
119: particles coming from the two arms of the ring and interfering at the
120: exit contact can be directly probed by measuring the QR conductance.
121: Indeed, interference effects have been observed in metal QRs many years
122: ago\cite{webb}. In addition, since electrons are spinful particles,
123: the spin part of the wavefunction is influenced by the magnetic field
124: via the Zeeman term in the Hamiltonian. Moreover, a more subtle
125: effect arises when there is a magnetic field non-orthogonal to the
126: plane of the orbiting particle because, as a consequence of the
127: orbital motion, its spin dynamics is instantaneously governed by a
128: time-dependent Hamiltonian \cite{anandans}. This time dependence ends
129: up in an extra phase acquired by the particle wavefunction which is
130: named after Berry
131: \cite{berry,anandan}, who put in foreground its topological properties
132: when the orbits are closed.
133:
134: Recently, semiconductor technology allows to grow samples (for
135: instance made by InAs or InGaAs) with sizeable spin orbit interaction
136: (SOI). Rashba\cite{rashba} first pointed out that the SOI strength can be
137: controlled by means of voltage gates. This feature has been recently
138: found experimentally \cite{meijer,miller}. Because of the SOI, an electric
139: field $E$ orthogonal to the orbit provides a momentum dependent effective
140: magnetic field felt by the electron spin in orbital motion. Such a
141: field can be equally well described by a vector potential that adds an
142: extra phase to the wavefunction\cite{aronov}. This phase was spelled
143: out by Ahronov and Casher \cite{ac}, as a ``dual'' AB effect, with
144: charge and spin interchanged (together with $B$ and $E$).
145:
146: During the last years, the effects of SOI on the AB oscillations have
147: been observed in semiconductor based QR by several groups
148: \cite{yau,nitta,morpurgo,kato}. It has been recognized that, in the
149: presence of both $B$ and $E$ orthogonal to the orbit plane, the
150: effective, momentum dependent total $B$ field is tilted w.r. to the vertical
151: direction. The resulting Berry phase influences the interference
152: pattern. Such a result is of the utmost interest because the Rashba-SOI
153: (RSOI) turns out to be a tool to tune the Berry phase and, ultimately,
154: the conductance, as well as the spin polarization of the outgoing
155: electrons. In Ref. \cite{molenkamp}, it is clearly shown that the AB
156: interference fringes can be modified by tuning an external
157: electrostatic potential.
158:
159: In a recent publication\cite{noiletter} we have included all these
160: phenomena affecting the interference of an electron ballistically
161: transported in a ring, also accounting for some dephasing at the
162: contacts. We have also shown that the AB peak in the Fourier transform
163: of the magnetoconductance displays satellite peaks due to the RSOI. It
164: has been suggested that satellite peaks observed in recent
165: experiments\cite{shayegan} have the same origin. In the present work
166: we review the theory by presenting the full calculation. In addition,
167: we include the semiclassical paths leading to weak localization
168: corrections\cite{nonloso}, and we discuss the rotation of the spin
169: polarization, during the transport across the ring.
170:
171: During the last years, several theoretical techniques have been
172: employed to study quantum transport in a QR. In Ref.s \cite{loss}, an
173: imaginary time path integral approach\cite{feynman} is developed to
174: study the conductance of a strictly one dimensional (1D) QR and its
175: conductance fluctuations in the diffusive limit. In
176: Ref. \cite{tserkov} a real time path integral approach is applied in
177: the limit of negligible Zeeman splitting. Several papers have
178: discussed the conductance properties and the spin selective transport
179: of QR's in the strictly 1D ballistic limit, by means of a spin
180: dependent scattering matrix approach
181: \cite{molnar,frustaglia,shen,dario,citro}. In the absence
182: of the magnetic flux, the conductance shows quasi-periodic
183: oscillations in the SOI stength, which can be modified by switching
184: the magnetic field on. Numerical calculations \cite{souma,lozano,wu}
185: have shown that in the 2D case there are only quantitative
186: modifications of the 1D results that do not qualitatively affect the
187: physics.
188:
189: In this paper we extend the real time path integral approach
190: previously developed in Ref.\cite{noiletter} to study the conductance
191: and the spin transport properties of a ballistic quantum ring in the
192: presence of both RSOI and of an external magnetic flux orthogonal to
193: the ring plane. We use a ``piecewise'' saddle point approximation for
194: the orbital motion, keeping the full quantum dynamics of the
195: spin. This approach allows us to take into account, in a
196: nonperturbative way, both the RSOI and AB phase and to include also
197: the Zeeman spin splitting.
198:
199: Our numerical approach evaluates all paths contributing to the quantum
200: propagator. The scattering at the leads can be forward or backward,
201: according to the probability amplitudes given by the S-matrix. Weak
202: localization corrections can be easily extracted from our result. We
203: also allow for some diffusiveness at the contacts by adding a random
204: phase factor in the motion.
205:
206: The DC conductance is derived from the Landauer formula
207: \cite{landauer} $ {\cal{G}} ={e^2}/{\h} \sum_{\sigma\sigma'}
208: \left|A(\sigma; \sigma'|E)\right|^2$, where $A(\sigma ;\sigma'|E)$ is
209: the probability amplitude for an electron entering the ring with
210: energy $E$ and spin polarization $\sigma '$ to exit with spin
211: polarization $\sigma $. We also report the change in the spin
212: polarization the electron transported across the ring.
213:
214: The structure of the paper is as follows:
215: \begin{itemize}
216: \item
217: In Section II we introduce the Feynman propagator for a spinful
218: electron injected at the Fermi energy in the ring.
219: \item
220: In Section III we discuss the topology of the allowed paths and the
221: scattering of the electron at the leads.
222: \item
223: In Section IV we represent our path integral in the coherent spin
224: basis\cite{haldane,plet} and derive the saddle point equations of motion,
225: whose classical counterpart is described in detail in Appendix A. This
226: allows us to justify the choice of a piecewise semiclassical
227: approximation for the orbital motion of the electron in the ring.
228: \item
229: In Section V we present the details of the calculation by rewriting
230: the path integral as a collection of single arm propagators. These are
231: the building blocks to be calculated in the next section.
232: \item
233: In Section VI we analyze how the orbital motion affects the full
234: quantum dynamics of the electron spin for each arm of the ring and
235: chirality. The spin propagator is derived in Appendix B, in the basis
236: corresponding to the rotating reference frame in the spin space.
237: \item
238: In Section VII we discuss the dependence of the conductance on the external
239: fields and on the overall transmission across the ring.
240: \item
241: In Section VIII we focus on the spin polarization of the outcoming
242: electron.
243: \item
244: Section IX includes a short summary and our conclusions.
245: \end{itemize}
246: %
247: \section{The transmission amplitude}
248:
249: Our model Hamiltonian will be the two-dimensional Hamiltonian for a
250: particle with spin-1/2 $\vec{S}$, in an orthogonal magnetic field,
251: with spin-orbit coupling to an orthogonal electric field (Rashba
252: coupling). It is given by
253: %
254: \bea
255: H[ \vec{p} , \vec{r} , \vec{S} ] =
256: \frac{1}{2m}\left(\vec p+\frac{e}{c}\vec A_0\right)^2
257: - \omega_{c}\: S_{z}+{H}_{so}
258: \label{hamilt} \\
259: H_{so}= \frac{2\alpha}{\hbar ^2}\left( {\hat z} {\times}
260: \left({\vec p+\frac{e}{c} \vec A_0}\right) \right){\cdot}
261: {\vec{S}} \:\:\:\: , \nonumber
262: \enea
263: %
264: where $\alpha$ is the spin orbit coupling constant, in units $ eV
265: \:$\AA, $\vec{S} = \hbar \vec{\sigma } /2 $
266: (${\sigma_x,\sigma_y,\sigma_z}$ are the Pauli matrices),
267: $\vec{A}_0(\vec{r}) = \frac{B}{2} ( - y , x , 0 ) $ is the vector
268: potential generating the uniform field $B$, normal to the ring
269: surface, taken in the symmetric gauge, $\omega_c = ge B / 2mc$ is the
270: cyclotron frequency. In real nanostructures based on InAs or InGaAs
271: the $g$ factor can strongly deviate from the value of two. However
272: the result we present here, are fully general as they depend on the
273: ratio $\alpha/\hbar \omega_c R$ which can be tuned by acting on
274: $\alpha$. Since we will assume only a single channel to be avaliable
275: for electron propagation across the ring, we will picture the single
276: channel ring as a $1-d$ circle of radius $R$, connected to two
277: leads. Accordingly, the position of the particle within the ring is
278: parametrized by the angle $\vt$ and the vector potential has just the
279: azimutal component $A_\vt = \phi / 2\pi R $, where $\phi$ is the
280: magnetic flux threading the ring.
281:
282: In order to study the conduction properties of the ring, one needs the
283: propagation amplitude for an electron entering the ring with spin
284: polarization $\mu_0$ to exit with spin polarization $\mu_f$, at energy
285: $E_0$. This is given by
286:
287: \beq
288: A(\mu _f ; \mu _0 | E_0 ) =
289: \int_{0}^{\infty} \: \frac{d t_f}{\tau _0} \: e^{i \frac{E_{0}
290: t_f}{\hbar}} \:
291: \langle {\vec r}_{f} , \mu_{f} , t_{f}|{\vec r}_{0} , \mu_{0 } , t_{0} \rangle
292: \:\:\:\: ,
293: \label{ampclas1}
294: \eneq
295: %
296:
297: where $\langle {\vec r}_{f} , \mu_{f} , t_{f}|{\vec r}_{0} ,
298: \mu_{0 } , t_{0} \rangle$ is the amplitude for a particle entering the
299: ring at the point $\vec{r}_0$ and at the time $t_0$ with spin
300: polarization $\mu_0$ to exit at the point $\vec{r}_f$ at the
301: time $t_f$ with spin polarization $\mu_f$. In our tensor product notation,
302: we define $|{\vec r} , \mu \rangle=|{\vec r}\rangle\:\otimes\:|\mu\rangle\ $.
303: $\tau _0 =mR^2 /(2\hbar) $ is the
304: time scale for the quantum motion.
305:
306: In order to compute $\langle {\vec r}_{f} , \mu_{f} , t_{f}|{\vec
307: r}_{0} , \mu_{0 } , t_{0} \rangle$ , we resort to a path integral
308: representation for the orbital part of the amplitude. Since we
309: parametrize the orbital motion of the particle in terms of the angle
310: $\varphi$, we provide the pertinent Lagrangian, ${\mathcal{L}}_{orb
311: }$, as a function of $\varphi , \dot{\varphi}$. It is given by
312: %
313: \beq
314: {\mathcal{L}}_{orb}[\varphi(t), \dot \varphi(t),\vec \sigma]=
315: \frac{m}{2}R^2\dot\varphi^2(t)-\frac{\phi}{\phi_{0}} \hbar
316: \dot\varphi(t) + \frac{\alpha^2\:m}{2 \hbar^2}+ \frac{\hbar^2}{8 m
317: R^2} \:\:\:\: .
318: \label{lag}
319: \eneq
320: %
321: The last two contributions to Eq.(\ref{lag}) are a constant, coming from
322: the spin-orbit term, and the Arthurs\cite{morette} term, which is
323: required when a path integration is performed in cylindrical
324: coordinates. Since both contributions are constant, they can be lumped
325: into the incoming energy $E_0$ and therefore they will be omitted
326: henceforth.
327: %
328: By taking into account the spin degree of freedom, as well, we represent
329: the propagation amplitude as
330: %
331: \bwt
332: %
333: \beq
334: \langle {\vec r}_{f}
335: , \mu_{f} , t_{f}|{\vec r}_{0} , \mu_{0 } , t_0 \rangle = \langle
336: {\vec r}_{f} , \mu_{f} | e^{ - i \int_{t_0}^{t_f } \: d t \: H } | {\vec
337: r}_{0} , \mu_{0 } , t_0 \rangle =\int_{\varphi ( t_0 ) =
338: \varphi_0}^{\varphi ( t_f ) = \varphi_f} \!\!\!\!{\cal D} \varphi\: e
339: ^{- i \int_{0}^{t_f} d t \; \left[ \tau_0 \dot{\varphi} ^2 - q
340: \dot{\varphi}\right ]} \: \langle \mu _f | \hat{U}_{spin} ( t_f, t_0 ) |
341: \mu _0 \rangle \:\:\:\: ,
342: \label{parte}
343: \eneq
344: \ewt
345: %
346: where $q=\phi / \phi_o $, $\phi_o $ being the
347: flux quantum $hc/e$.
348: %
349: \beq
350: \hat{U}_{spin} ( t_f, t_0 ) = {\bf T} \exp \left [ -\frac{i}{\hbar} \:
351: \int _{t_0}^{t_f} dt' \: \hat{H} _{spin} (t') \right ] \:\:\: .
352: \eneq
353: %
354: is the full spin propagator and the spin Hamiltonian
355: $ \hat{H} _{spin} (t)$ is given by
356: %
357: \beq
358: \frac{1}{\hbar } \hat{H}_{spin} ( t ) =
359: \left[ \begin{array}{cc} \frac{ \omega _c}{ 2} &
360: \gamma \dot\varphi e^{ - i \varphi ( t ) } \\
361: \gamma \dot\varphi e^{ i \varphi (t) } &
362: - \frac{ \omega _c}{ 2} \end{array} \right]
363: \:\:\:\: ,
364: \label{three}
365: \eneq
366:
367: %
368: with $\gamma = 2\alpha \tau _0 /(\hbar R ) $\cite{nota}.
369: %
370: In Sections IV and V we show that the amplitude of Eq.(\ref{parte}) can
371: be approximated by choosing a piecewise semiclassical orbital motion
372: for the particle in each arm of the ring, while keeping the full
373: quantum dynamics of the spin.
374: %
375: In particular, we will see that, within the physically relevant range
376: of parameters, the orbital motion can be approximated as a uniform
377: rotation (with constant angular velocity), which makes the spin
378: dynamics to be the one of a spin-1/2 in an effective, rotating,
379: external magnetic field. Yet, in order to explain how we deal with
380: quantum backscattering at the contacts between ring and arms and
381: corresponding dephasing effects, we will introduce our formalism in
382: the next section, by discussing a simplified version of our problem: a
383: spinless electron propagating across a mesoscopic ring.
384: %
385: \section{Feyman's paths for a Spinless particle transmitted across a ring}
386: %
387: In this section we introduce our formalism by computing the
388: transmission amplitude for a spinless electron of mass $m$ and charge $-e$,
389: traveling across the ring in an
390: orthogonal magnetic field.
391: %
392: For a realistic device, at each lead one has to take into account three
393: possible scattering processes, consistently with the conservation of the
394: total current. This is described in terms of a unitary $S-$matrix
395: that, in the symmetric case in which the two arms are symmetric, is
396: given by
397: %
398: \beq
399: \mathcal S=\left(
400: \begin{array} {c c c}
401: -\met (1+\bar r) & \met (1-\bar r) & \sqrt{\met(1-\bar r^2)}\\
402: \met (1-\bar r) &-\met (1+\bar r) & \sqrt{\met(1-\bar r^2)}\\
403: \sqrt{\met(1-\bar r^2)} & \sqrt{\met(1-\bar r^2)} & \bar r
404: \end{array}\right)
405: \label{matr}
406: \:\:\:\: .
407: \eneq
408: %
409: The numerical labeling of the S-matrix elements referring to the three
410: terminals of each contact fork, are explained in Fig.(\ref{noWLpaths},
411: 1a). Assuming, for simplicity, that the scattering matrix is the same
412: for both leads, Eq.(\ref{matr}) will hold both at the left-hand lead,
413: and at the right-hand lead of the ring.
414:
415: In particular, $\mathcal S_{3,3}=\bar r$ is the reflection amplitude
416: for a wave coming from the left lead, $\mathcal S_{1(2),1(2)}=-\met
417: (1+\bar r)$ the reflection amplitude for a wave incoming from the
418: upper/lower arm, $\mathcal S_{1(2),2(1)}= \met (1-\bar r)$ is the
419: transmission amplitude from the upper (lower) to the lower (upper) arm
420: and $\mathcal S_{1(2),3}=\mathcal S_{3,1(2)}=\sqrt{\met(1-\bar r^2)}$
421: is the transmission amplitude from the upper/lower arm to outside of
422: the ring.
423:
424:
425: In Fig.(\ref{noWLpaths}) we show the simplest paths of the electrons
426: in the ring including forward scattering at the contacts, only.
427: The contribution to the total amplitude coming from such paths, in
428: which the electron enters the ring at an angle $\varphi_0$ at time
429: $t_0$ and exits at $\varphi_f$ at time $t_f$, is given by
430: %
431: %\bwt
432: \bea
433: {\cal{A}}(\vt_{f}, t_f; \vt _0, t_0 ) = \sum _{n=-\infty}^{+\infty}
434: \int _{\vt (t_0)=
435: \vt _i}^{\vt (t_f)= \vt _f +2\pi n }\!\!\!\!\!\!\! {\cal{D}}\vt (\tau ) \:
436: e^{-i{\cal{S}}[\vt[t)]/\hbar }\nonumber\\
437: =\sum _{n=-\infty}^{+\infty}
438: e^{-iq \:(\vt_f-\vt_i +2\pi n ) }
439: \int _{\vt (t_0 )=
440: \vt _i}^{\vt (t_f)= \vt _f +2\pi n }\!\!\!\!\!\!\!
441: \!\!\!\!\!\!\!
442: \!\!\!\!\!\!\!
443: \!\!\!\!\!\!\!
444: {\cal{D}}\vt (t ) \:
445: e^{-i\frac{mR^2}{2\hbar } \int _{t_0}^{t_f} dt \: \dot \vt ^2 (t )}
446: \:,
447: \label{spinless}
448: \enea
449: %\ewt
450: %%
451: where we have summed over paths in which the electron winds $n+1/2$ times in
452: the ring, before exiting it.
453: Positive (negative) $n$ values imply clockwise (counterclockwise)
454: propagation along the ring.
455:
456: This propagator can be evaluated exactly
457: \cite{moran}. However we report here just the saddle point
458: evaluation, for comparison with the spinful case. Minimizing the
459: action gives the classical equation of motion (together with the pertinent
460: boundary conditions for a path winding $n+1/2$ times):
461:
462: \beq
463: \ddot \vt (t )
464: = 0 \:\:\:
465: ; \vt (t_i ) = \vt _i \: , \vt (t_f ) = \vt_f + 2\pi n
466: \:\:\:\: .
467: \eneq
468: %
469: Let us assume that the particle is injected in the ring at $\vt_0 =0 $
470: and comes out at $\vt_f = \pi $ in a time $T = t_f$.
471: Eq.(\ref{spinless}) gives:
472: %
473: \bea {\cal{A}}(\pi ,0, t_f ) =
474: e^{i\pi q } \: \sqrt{\frac{\tau_0}{\pi i t _f}} \: {\sum
475: _{n=-\infty}^{+\infty }}' e^{-i \pi ^2 (2|n|-1) ^2 \tau_0 / t_f -i
476: 2\pi n q } \nonumber\\
477: \:\:\:\: ,
478: \label{less}
479: \enea
480: %
481: where the prime in the sum takes into account the fact that one does not
482: sum over $n=0$, and the square root at the prefactor accounts for the
483: gaussian fluctuations. Of course, this propagator is periodic in $q$ of
484: period $q=1$ up to a minus sign.
485:
486: More involuted paths arise if one takes into account backscattering
487: processes in which the electron can get backscattered within the same
488: ring's arm from which it is coming. For instance, the paths $(2f)$ and
489: $(2h)$, as well as $(2g)$ and $(2i)$ in Fig.(\ref{WLpaths}), include
490: looping in opposite directions around the ring's hole. Interference
491: between clockwise and counterclockwise windings leads to weak
492: localization corrections. We denote these corresponding paths
493: (including also $(2c)$ and $(2d)$ ) as ``reversed paths''$(RP)$. In our
494: approach, all order paths are numerically generated up to the
495: convergency and the $S-$matrix of Eq.(\ref{matr} ) is implemented in
496: the numerical algorithm.
497: %
498: \begin{figure}[!htp]
499: \centering \includegraphics[width=\columnwidth]{paths2b.eps}
500: \caption{(color online) First and second order paths included in the
501: calculation of the transmission amplitude across the ring,
502: from left to right, including
503: forward scattering only. Numbers $1,2,3(1',2',3')$ in Fig. 1a refer to
504: the labeling of the terminals in Eq.(\ref{matr}.)}
505: \label{noWLpaths}
506: \end{figure}
507: %
508: %
509: \begin{figure}[!htp]
510: \centering
511: \includegraphics[width=\columnwidth]{paths3.eps}
512: \caption{Second order paths of the transmission amplitude from left to
513: right including backscattering at the leads. Paths $(2f)$ and
514: $(2h)$, as well as $(2g)$ and $(2i)$ contribute to the weak
515: localization corrections.}
516: \label{WLpaths}
517: \end{figure}
518: In conclusion, we have established that the paths contributing to the
519: transmission amplitude can be built up by adding four types of
520: elementary paths: $u_\rightarrow$ forward orbiting in the upper arm of
521: the ring ($\vt \in ( 0,\pi )$), $ u_\leftarrow$ backward orbiting in
522: the upper arm of the ring ($\vt \in ( \pi,0 )$), $d_\rightarrow$
523: forward orbiting in the lower arm of the ring ($\vt \in ( 2 \pi,\pi )$), $
524: d_\leftarrow$ backward orbiting in the lower arm of the ring ($\vt \in
525: ( \pi, 2 \pi )$). In Section V, we will generalize such an approach, while
526: in the next Section we discuss the Feynman path-integral
527: representation of the propagation amplitude for a spinful electron
528: propagating along one of these elementary paths.
529: %
530: \section{Quantum amplitude and semiclassical orbital motion
531: for a spinful electron in the ring}
532: %
533: In this section, we construct the Feynman representation in the basis
534: of the coeherent spin states, for the propagation amplitude of an
535: electron moving along one of the arms of the ring with a given
536: chirality. To discuss the saddle-point approximation we need the
537: equations of motion coming from the condition that the action is
538: stationary. The coherent spin state basis provides a straightforward
539: route to perform a semiclassical approximation involving both orbital,
540: and spin degrees of freedom, at the same time. In particular, we will
541: show under which conditions the classical orbital motion $\dot{\vt}=
542: cnst $, can be retained, as in the spinless case. Accordingly, the
543: spin dynamics will be that of a ``quantum magnetic moment'' in a
544: time-dependent external magnetic field.
545:
546: Let $\Omega$ denote the orientation of the spin $\vec S$ and $|\Omega
547: \rangle$ be the coherent state such that:
548: %
549: \beq
550: \langle \Omega | \hat{\vec{S}}| \Omega \rangle \equiv \vec{S}[\Omega ] =
551: \hbar \: S \: \left[ \begin{array}{c} \sin \Theta \cos \Phi \\
552: \sin \Theta \sin \Phi \\ \cos \Theta \end{array} \right]
553: \:\:\:\: ,
554: \label{ringo2.2.2}
555: \eneq
556: \noindent
557: %
558: for the three components of the spin vector, respectively.
559: %
560: The full propagator in the coherent spin state representation is given
561: by:
562: %
563: \bwt
564: %
565: \bea
566: \langle \varphi_f , \Omega _f , t_f \: | \varphi_0 , \Omega _0 , 0
567: \rangle =\int_{\varphi ( 0 ) = \varphi_0}^{\varphi ( t_f ) = \varphi_f}
568: \!\!\!\!{\cal D} \varphi\: e ^{- i \int_{0}^{t_f} d t \;
569: \left[ \tau_0 \dot{\varphi} ^2 - q \dot{\varphi}\right ]}
570: \int_{\Theta ( 0 ) = \Theta_0}^{\Theta ( t_f )
571: = \Theta_f} \!\!\!\!
572: {\cal D} \Theta \int_{\Phi ( 0 ) = \Phi_0}^{\Phi ( t_f ) = \Phi_f}
573: \!\!\!\!{\cal D} \Phi\:
574: \: e^{-\frac{i}{\hbar}
575: {\cal{S}}_{spin}\left [ \Theta , \dot{\Theta} ; \Phi ,
576: \dot{\Phi} | \vt , \dot{\vt} \right ] }
577: \:\:\:\: ,
578: \label{propo}
579: \enea
580: %
581: where:
582: %
583: \bea
584: {\cal{S}}_{spin} \left [ \Theta , \dot{\Theta} ; \Phi ,
585: \dot{\Phi} | \vt , \dot{\vt} \right ] /\hbar = \int_{0}^{t_f} d t \;
586: \left \{ \frac{ (1-\cos\Theta )}{2} \dot \Phi + {\cal {L}}_{spin}
587: \left [ \Theta , \dot{\Theta} ; \Phi ,
588: \dot{\Phi}| \varphi , \dot{\varphi} \right ] \right \}
589: \:\:\:\: .
590: \label{prospin}
591: \enea
592:
593: %
594: with:
595: %
596: \bea
597: {\cal L}_{spin} \left [ \Theta , \dot{\Theta} ; \Phi ,
598: \dot{\Phi} | \varphi , \dot{\varphi} \right ] =
599: \frac{ \omega_c}{2} \cos \Theta - \gamma
600: \dot{\varphi} \sin \Theta \cos ( \varphi - \Phi )
601: \:\:\:\: .
602: \label{venti}
603: \enea
604:
605: %
606: \ewt
607: %
608: The Lagrangian of Eq.(\ref{venti}) corresponds to the coherent
609: spin-state representation of the classical Lagrangian derived from
610: Eq.(\ref{hamilt})
611: %
612: \bea
613: {\cal L} [ \varphi , \dot{\varphi} , \vec{S} ] =
614: \nonumber \\
615: \frac{m}{2} \dot\vec{r} ^2 -
616: \frac{e}{c}\: \dot\vec{r} \cdot \vec{A} + 2m \alpha \left [
617: \hat z\cdot \left ( \dot\vec{r} \times \vec{S}\right ) \right ] +
618: \frac{ m \alpha^2}{\hbar^2} + \omega_c S_z \: .
619: \label{sette}
620: \enea
621: %
622: In Eq.(\ref{sette}) we have introduced the constraint that the orbital electron
623: motion takes place along a one-dimensional circle by parametrizing the
624: trajectories with the angle $\varphi$ as $ x = R \cos \varphi ; \; y =
625: R \sin \varphi $.
626: %
627: The additional {\sl Berry phase } term \cite{aurbach} in
628: Eq.(\ref{prospin}) arises from the fact that different spin coherent
629: states are not orthogonal to each other since, to leading order in
630: $\epsilon$, the scalar product between spin-coherent states at times
631: $t_j , t_j +\epsilon $, $ | \Omega ( t_j ) \rangle$ and $ | \Omega (
632: t_{ j } + \epsilon ) \rangle$ is given by
633: %
634: \beq
635: \langle \Omega ( t_j + \epsilon ) | \Omega ( t_j ) \rangle \approx
636: \exp \left[ \frac{i}{2} [ 1 - \cos \Theta ( t_j ) ] \dot{\Phi} ( t_j )
637: \epsilon \right]
638: \:\:\:\: .
639: \label{ringostar}
640: \eneq
641:
642: %
643: Let us look for the trajectories in orbital and spin space which make
644: the action stationary. This requires solving the Eulero-Lagrange
645: equations for the Lagrangian $ {\cal L} [ \varphi , \dot{\varphi} ;
646: \Theta , \dot{\Theta} ; \Phi , \dot{\Phi} ]$ appearing in Eq.(\ref{propo}),
647: which are given by
648: %
649: \begin{eqnarray}
650: \frac{\dot{\Theta}}{2} \sin \Theta + \gamma
651: \dot{\varphi} \: \sin \chi \: \sin \Theta &=& 0
652: \label{em1}\:, \\
653: \sin \Theta [- \dot{\Phi} + \omega_c ] +
654: 2 \gamma
655: \dot{\varphi} \: \cos \Theta\: \cos \chi &=& 0
656: \label{em2}\:, \\
657: \ddot{ \varphi} - \frac{ \gamma}{2\tau _0} [ \dot{\Theta} \: \cos \Theta
658: \: \cos \chi + \dot{\Phi} \: \sin \Theta\: \sin \chi ]
659: &=& 0 \label{em3}
660: \: ,
661: \end{eqnarray}
662:
663: %
664: with $\chi = \varphi -\Phi $.
665:
666: In order to extract the relevant physics from the above equations, we
667: multiply Eq.(\ref{em3}) by $\dot{\varphi}$, and, by use of
668: Eqs.(\ref{em1},\ref{em2}), we rewrite Eq.(\ref{em3}) as:
669: %
670: \beq \dot{\varphi} \ddot{
671: \varphi} =- \frac{ \omega _c}{4\tau _0} \: \dot{\Theta} \: \sin\Theta
672: \:\:\:\: .
673: \label{orbo}
674: \eneq
675: %
676: A straightforward time integration gives:
677: %
678: \beq
679: \tau_0 \frac{ ( \dot{\varphi} )^2 }{2} - \frac{\omega_c}{4} \cos \Theta = cnst
680: \:\:\:\: ,
681: \label{orbo2}
682: \eneq
683:
684: which states that the total energy is conserved. According to
685: Eq.(\ref{orbo2}), the particle energy only includes the orbital kinetic
686: term and the Zeeman term. This is obvious, since the force associated
687: to the spin orbit coupling, being gyroscopic, does no work.
688: %
689: A change in the precession angle $\Theta$ implies acceleration in the
690: orbital motion. According to the spin Hamiltonian of Eq.(\ref{three})
691: the RSOI is responsible for flipping of the spin, while the Zeeman
692: coupling tends to stabilize the spin direction. Hence, there are two
693: physically relevant limits, in which the orbital motion fully
694: decouples from the spin dynamics, according to the inequality
695: $\omega_c/2 <(>) \dot \varphi$, as we are going to show next.
696: Both limits involve an orbital motion with constant velocity
697: $\dot \varphi $ and a spin orientation given by the angles
698: $\Theta = {\rm constant} \:\:\: , \:\:
699: \varphi - \Phi = 0 \: (mod.\: \pi ) $.
700:
701: {\sl a) Vanishing Zeeman coupling}: This case has already been considered
702: and it has been shown that it can be exactly solved
703: analytically\cite{tserkov}. In this case Eq.(\ref{orbo})
704: allows for the semiclassical solution $\dot\varphi =cnst $ and quantum
705: fluctuations of the spin do not interfere with the orbital motion.
706: The spin is tilted by the constant angle $\tan \Theta = 2\gamma $ and
707: precesses with a constant frequency $\dot \Phi = \dot \varphi $.
708:
709: {\sl b) Large Zeeman coupling}: flipping of the spin and quantum
710: fluctuations of the spin are strongly disfavoured, again corresponding
711: to the classical
712: saddle point configuration $\Theta = {\rm constant} \:\:\: , \:\:
713: \varphi - \Phi = 0, \pi $ with
714: %
715: \beq
716: \pi/t_f=\dot{\varphi} = \pm \dot{\Phi} =
717: \omega_c \:\frac{1}{1-2\gamma \: {\rm ctan} \Theta}\:\:\:.
718: \label{orbo3}
719: \eneq
720: %
721: The spin precesses around the $z$-axis, with the same angular velocity
722: as the orbital motion. When $\gamma \to 0 $, $\Theta $ can be
723: vanishingly small, with $\dot\Phi =\omega _c $.
724:
725: It may appear that the simple spin precession provided by
726: Eq.(\ref{orbo}) with $\Theta = cnst $ is a solution of
727: Eq.s(\ref{em1},\ref{em2},\ref{em3}) for any ratio $\omega _c
728: /\dot\varphi $. Were this the case, there would be no need to invoke
729: the limitations of case $b)$ as stated above. However, a careful
730: analysis of the stability of this saddle point solution shows that the
731: simple spin precession is a minimum of the action only when the Zeeman
732: coupling is strong. In particular, if $\omega _c /\dot\varphi >>1 $,
733: an additional condition $\omega _c \tau _o/\gamma ^2 >1 $ has to be
734: satisfied. Otherwise the frequency of the fluctuations around the
735: saddle point solution does not keep real and the analysis of small
736: oscillations in the parameter space around the saddle point solution
737: breaks down.
738:
739: This shows that a classical orbital motion with constant velocity is
740: compatible with both limits of large and small ratios of the Zeeman
741: coupling to the RSOI strength. However the two limiting cases do not
742: belong to the same saddle point. In particular there will be a
743: crossover region connecting the two limits in which quantum
744: fluctuations of the spin may induce changes in the orbital velocity of
745: the particle.
746:
747: In the rest of the paper we will choose $\dot \varphi=const$ piecewise
748: and parametrize the quantum dynamics of the spin with the value of the
749: velocity obtained by the stationary phase condition discussed in the
750: next section. As discussed above, our approximation cannot reproduce
751: faithfully the intermediate region of parameters ranging the two
752: limits of large and small Zeeman coupling $\omega_c / \gamma
753: \dot\varphi <(>) 1$. We checked numerically the reliability of our
754: approximation, by numerically integrating
755: Eqs.(\ref{em1},\ref{em2},\ref{em3}) and have have found that it is
756: satisfactory in most of the parameter range.
757:
758: By putting $\dot\varphi =cnst $ into Eqs.(\ref{em1},\ref{em2}), they
759: become, as we show in detail in Appendix A, the classical equations of
760: motion for a magnetic moment in a time dependent magnetic field
761: ${\cal{B}}
762: \equiv ({\cal{B}}_+,{\cal{B}}_-,{\cal{B}}_z ) = \left ( \gamma
763: \dot{\vt} \: e^{i\vt}, \gamma \dot{\vt} \: e^{-i\vt},-\omega _c/2
764: \right ) $. This can be easily understood from the fact that
765: minimizing the action in Eq.(\ref{propo}) directly w.r. to the spin
766: components provides:
767: %
768: \beq
769: \vec{S}\times \frac{d\vec{S}}{dt} = - \frac{\delta H }{\delta
770: \vec{S}} \:\:\:\: ,
771: \eneq
772: %
773: which has to be solved together with the constraint of constant
774: modulus: $\vec{S}\cdot d\vec{S} /dt = 0$ ( see Eq.(\ref{appe1}).
775: %
776:
777: Among the other possible saddle-points, a solution of the motion
778: equations can be found with the particle trapped within the ring arm
779: (turning points of $\varphi(t)$ are at $\varphi = 0,\pi $) if $\gamma
780: $ is large enough. This solution doesn't seem to be practical as it
781: requires fine tuning of the external parameters with transfer of
782: energy from the spin motion to the orbital motion and viceversa. We
783: have not investigated it in detail, but we expect that it could
784: provide resonant tunneling across the ring.
785: %
786: \section{Saddle point approximation and looping in the ring}
787: %
788: In this Section we show how we implement the piecewise saddle point
789: solution for the orbital motion, $\dot{\vt} = cnst$, to study the
790: coherent propagation of the electron inside the ring.
791: %
792: In the following, we will denote by ``loop'' and ``looping
793: trajectory`` both trajectories that wind around the ring (closed), and
794: paths in which the particle moves forth and back in one of the ring
795: arms (open) (see Fig.(\ref{WLpaths})). Of course, the amplitudes
796: differs very much in these two types of looping . Indeed, the net spin
797: rotation is different between the two paths and, also, the
798: Ahronov-Bohm ($AB$) phase is absent in the latter ones .
799: %
800:
801: %
802: \begin{figure}[!htp]
803: \centering \includegraphics[width=\columnwidth]{tree.eps}
804: \caption{(color online) Cayley tree describing the way in which higher
805: order paths are built in the numerical code. Full lines correspond to
806: propagation in the upper arm ( $u$ ) or lower arm ( $d$) in the
807: forward ($\rightarrow$) or backward ( $\leftarrow$) direction. The
808: nodes represent the leads of the ring. The exit nodes are marked by
809: a broken line at each order $n$. The heavy line correspond to the
810: path reported in Eq.(\ref{ugly}). }
811: \label{tree}
812: \end{figure}
813: %
814: In general, at order $n$, we will have $2^{(2|n|-1)}$ trajectories of
815: a particle entering the ring at $\vt = 0 $ and exiting at $\vt = \pi $
816: and each of them will include $2|n|-1 (n = \pm 1,\pm 2,...) $
817: elementary paths (or ``stretches'') of the type
818: $u_\rightarrow,u_\leftarrow,d_\rightarrow,d_\leftarrow$, as classified
819: at the end of Section II. Looping at order $|n|$ involves $2 |n|$
820: scattering processes at the contacts. In the case of spinful
821: electrons the $S$-matrix is doubled ($6\times 6$) w.r. to the one
822: given in Eq.\ref{matr}. Of course, the $S$-matrix at the contacts is
823: sample dependent and any special choice is arbitrary. In the
824: following we neglect possible asymmetries in the up-down channel as
825: well as accidental spin flipping in traversing the contacts. Hence, we
826: make the simplifying assumption that the $S$-matrix is block diagonal
827: of the form given in Eq.\ref{matr} for both spins.
828:
829: Each time the trajectory impinges at a contact without leaving the
830: ring, the $S-$ matrix of Eq.(\ref{matr}) alledges for either forward,
831: or backward scattering. Let us denote with
832: $u_{\rightarrow(\leftarrow)}(t_i,t_j)$ the forward (backward)
833: propagation amplitude in the upper arm from time $t_j$ to time $t_{i}$
834: ($d_{\rightarrow(\leftarrow)}(t_i,t_j)$ for the lower arm). The
835: $u(d)$ amplitudes include the Ahronov-Bohm phase and the spin
836: evolution, but not the dynamical phase, which is factored out (see
837: below). As shown in section II, to first order ($|n|=1$) there are
838: only two possible paths (see Fig.(\ref{noWLpaths} 1a,1b)), the
839: propagation amplitude is, then, the sum of the the corresponding two
840: amplitudes:
841: %
842: \bwt
843: %
844: \beq
845: A_1(\mu _f ; \mu _0 | E_0 ) = \int_0^\infty \: \frac{d t_f}{ \tau_0}
846: e^{ i \frac{E_0 t_f}{\hbar}}
847: \left\langle\mu_f,t_f\right|\biggl [ ( \mathcal S_{3'1} u_\rightarrow(t_f,t_0) \mathcal S_{13} + \mathcal
848: S_{3'2} d_\rightarrow(t_f,t_0) \mathcal S_{23} ) \biggr ]
849: \left|\mu_0,t_0\right\rangle \:
850: e^{-i\frac{mR^2}{2\hbar } \int _{t_0}^{t_f} dt \: \dot \vt ^2 (t )}
851: \label{prio}
852: \eneq
853: %
854: The amplitudes have to be summed all together, order by order. The
855: key observation appearing in the symbolical writing of Eq.(\ref{ugly})
856: is that the dynamical phase, at a given order, does not depend on the
857: chirality of the motion and can be factored out. On the contrary, the
858: Ahronov-Bohm phase depends on the chirality, and the spin evolution
859: depends on both the chirality and on the modulus of the propagation
860: velocity.
861: %
862: Beyond the first order, there is a net increase in the number and type
863: of the paths to be summed together. In Fig.(\ref{tree}), we
864: pictorially sketch all the possible paths by mean of a Cayley tree.
865: Each node represents a lead of the ring and, according to the
866: scattering matrix of Eq.(\ref{matr}), the electron can be
867: backreflected into the ring's arm it is coming from with an amplitude
868: $S_{ii}$, or, with an amplitude $S_{ij}$, it can be either transmitted
869: to the other arm, or outside of the ring. In the tree, the
870: transmission out of the ring is possible at all the nodes crossing the
871: dashed lines. Each dashed line is labeled by the order $n$ of the
872: interference in the ring. As an example, we explicitely write down one
873: of the possible second order paths (the bold red line marked by $\cal
874: {P}$ in Fig.(\ref{tree}), which corresponds to
875: Fig.(\ref{WLpaths},2h)):
876: %
877: \bea
878: A_2^{\cal{P}} (\mu _f ; \mu _0 | E_0 )=\!
879: \int_0^\infty \!\! \frac{d t_f}{ \tau_0} \! e^{ i \frac{E_0 t_f}{\hbar}}
880: \int _{t_0}^{t_f} \!\!\!dt_2 \!
881: \int _{t_0}^{t_2}\!\! dt_1 \!
882: \left\langle\mu_f,t_f\right|
883: \mathcal S_{3'2} d_\rightarrow(t_f,t_2) \mathcal S_{22} d_\leftarrow(t_2,t_1) \mathcal
884: S_{21} u_\rightarrow (t_1,t_0)\mathcal S _{13}
885: \left|\mu_0,t_0\right\rangle
886: e^{-i\frac{mR^2}{2\hbar } \int _{t_0}^{t_f} dt \dot \vt ^2 (t )} \:.
887: \label{ugly}
888: \enea
889: %
890:
891:
892: At the saddle point with uniform velocity, $\dot{\vt} = 2 \pi \:
893: (2|n|-1) / t_f $, we classify the collection of sequences belonging to
894: a certain $n$ as $\{{\cal{C}}_n\} $, and denote the superposition of
895: amplitudes (e.g., the ones in the big parenthesis of Eq.(\ref{prio})
896: to first order ) as $ {\cal{F}}
897: \left [\mu _f, \mu _0; q, \dot{\vt} | {\cal{C}}_n \right ] $. In this
898: way, we can rewrite the full amplitude of Eq.(\ref{ampclas1}) as:
899: %\bwt
900: \beq
901: \mathcal A(\mu _f ; \mu _0 | E_0 ) =
902: \int_{0}^{\infty} \: d t_f \: e^{-i E_{0}
903: t_f/\hbar}
904: \sqrt{\frac{\tau _0}{\pi i t_f } } \:
905: \: {\sum _{n=-\infty}^{+\infty }}' \: \sum _{\{{\cal{C}}_n\}}
906: \: {\cal{F}} \left [ \mu _f, \mu _0; q, \dot{\vt} = 2 \pi \: (2 |n|-1) / t_f
907: | {\cal{C}}_n \right ] \:
908: e^{ -i \pi ^2 (2|n| -1) ^2 \tau _0 / t_f }
909: \: .
910: \label{coll}
911: \eneq
912: \ewt
913: %
914: The series in Eq.(\ref{coll}) is uniformly convergent. Thus, we may
915: swap the integral with the sum, and integrate term by term.
916: The integral contributing to order $n$ is given by
917: %
918: \bea I_n = \int_{0}^{\infty} \: d t_f \:
919: \sqrt{\frac{\tau _0}{\pi i t_f }} \: e^{-i E_{0} t_f/\hbar -i \pi ^2
920: (2|n| -1) ^2 \tau _0 /t_f } \: \nonumber\\ =\: \sqrt{\frac{1}{\pi i }}
921: \: \int_{0}^{\infty} \: d x \: e^{-i\epsilon \: x^2 -i \pi ^2 (2|n|
922: -1) ^2 /x^2 }
923: \label{stat}
924: \enea
925: %
926: with $\epsilon = E_{0} \tau _0/\hbar $. We compute it approximately
927: within stationary phase contribution. Since the phase of the exponent
928: of the integrand is stationary at $\bar{t}_n = \epsilon ^{(-1/2)} \:
929: \pi \:(2|n|-1) \tau _0$, by inserting this value in the phase and
930: integrating out the gaussian fluctuations, we readily get:
931: %
932: \bea
933: I_n \approx e^{-i \sqrt{\epsilon} 2 \pi (2|n|-1) }
934: \sqrt{\frac{1}{\pi i }} \:
935: \int_{-\infty}^{\infty} \: d (\delta x ) \:
936: e^{-i \epsilon \:( \delta x )^2} \nonumber\\
937: \sim
938: \frac{1}{i \sqrt{2\epsilon }}
939: \: e^{-i \sqrt{\epsilon} 2 \pi (2|n|-1)^2 } \:\:\: .
940: \enea
941: %
942: ($ -i \epsilon ^{-1/2} $ is the usual factor appearing in the
943: one-dimensional free particle Green's function in real space and
944: energy). This approximation, when plugged into Eq.(\ref{coll}),
945: provides the final result:
946: %
947: \bwt
948: \beq
949: A(\mu _f ; \mu _0 | E_0 ) =
950: \frac{1}{i \sqrt{ 2\epsilon } } \:
951: {\sum _{n=-\infty}^{+\infty }}' \: \biggl ( \sum _{\{{\cal{C}}_n\}}
952: \: {\cal{F}} \biggl [ \mu _f, \bar{t}_n , \mu _0; q, \dot{\vt} =
953: 2 \pi \: ( 2|n|-1 ) / \bar{t}_n
954: \biggl | \biggr . {\cal{C}}_n \biggr ] \: \biggr ) \:
955: \: e^{-i \sqrt{\epsilon} 2 \pi (2|n|-1) }
956: \: .
957: \label{colf}
958: \eneq
959: \ewt
960: %
961: The enumeration of the trajectory configurations belonging to the
962: collection $ {\cal{C}}_n $, to order $n$, is numerically performed
963: order by order.
964:
965: In the next Section, we discuss the elementary
966: propagators for each of the four stretches,
967: $u_{\rightarrow},u_{\leftarrow},d_{\rightarrow},d_{\leftarrow} $, as
968: defined in Section III. This allows us to construct the functional $
969: {\cal{F}} $ for each incoming and outgoing spin polarization.
970:
971:
972: \section{Quantum spin dynamics of the electron propagating in the ring}
973: In this Section, we provide the explicit formula for the spin
974: contribution to the total propagation amplitude, given by
975: \beq
976: \hat{U}_{\rm spin} ( t , t' ) = {\bf T} \exp \left[ - \frac{i}{ \hbar}
977: \int_{ t' }^{ t } \: d \tau \: H_{\rm spin} ( \tau ) \right]
978: \:\:\:\: .
979: \label{a.2.1}
980: \eneq
981:
982: As discussed in detail in appendix A, within the saddle point
983: approximation, $ H_{\rm spin} ( t )$ corresponds to the Hamiltonian of
984: a spin-1/2 in a time-dependent external magnetic field. It can be
985: written as (from now on, we will denote by $\omega_o$ the frequency of
986: the orbital motion, that is, the stationary phase value of
987: $\dot{\varphi}$)
988:
989: \beq
990: \hat{H}_{\rm spin} ( t ) = \left[ \begin{array}{cc} r \cos \vartheta &
991: r \sin \vartheta e^{ - i \omega_o t } \\
992: r \sin \vartheta e^{ i \omega_o t } &
993: - r \cos \vartheta \end{array} \right]
994: \label{three_1}
995: \:\:\:\: ,
996: \eneq
997:
998: with
999: \begin{eqnarray}
1000: r \cos \vartheta = \frac{ \omega _c}{ 2}\: , \:\:\:\:\:
1001: r \sin \vartheta = \gamma \omega_o \: , \:\:\:\:\:
1002: \varphi (t) = \omega_o t \: , \label{def}\\
1003: r = \met \sqrt{ \omega _c^2+4 \gamma ^2 \omega_o ^2} \: ,
1004: \:\:\:\:\: \tan \vartheta = \frac{2\gamma \omega_o }{ \omega _c}
1005: \:\: .
1006: \nonumber
1007: \end{eqnarray}
1008:
1009: Including only the AB phase implies $\vartheta = 0 $, while including
1010: only RSOI implies $\vartheta \to \pi /2 $.
1011:
1012: It is useful to solve for the spin dynamics in the representation of
1013: the instantaneous eigenstates of $\hat{H}_{\rm spin} (t)$. At fixed
1014: $t$, its eigenvalues are given by $\pm \epsilon = \pm r$, while the
1015: corresponding eigenvectors take the form:
1016: \begin{eqnarray}
1017: | + , t \rangle = \left ( \begin{array}{c}
1018: \cmt \\
1019: \smt \: e^{\ang } \end{array} \right ) \: , \:\:\:\:\:\:
1020: | - , t \rangle = \left ( \begin{array}{c}
1021: -\smt \: e^{ - \ang} \\
1022: \cmt \end{array} \right ) \: .
1023: \label{four}
1024: \end{eqnarray}
1025:
1026: Thus, the matrix diagonalizing $\hat H_{\rm spin} ( t ) $ at time $t$
1027: is
1028: \begin{eqnarray}
1029: \hat B ( t ) \equiv \left[ \begin{array}{cc}
1030: \cmt & - \smt \: e^{ - \ang} \\
1031: \smt \: e^{ \ang} & \cmt
1032: \end{array} \right]
1033: \:\:\:\: .
1034: \label{five}
1035: \end{eqnarray}
1036: The matrix $\hat B ( t )$ encodes the adiabatic contribution to the
1037: evolution of $ | \Psi ( t ) \rangle$. Therefore, in order to write
1038: down the Schr\"odinger equation with Hamiltonian $\hat{H}$,
1039: \[
1040: \biggl\{ i \frac{ \partial}{ \partial t} - \hat{H} ( t ) \biggr\}
1041: | \Psi ( t ) \rangle = 0
1042: \]
1043:
1044: in the adiabatic basis, one has to strip off from the state $ | \Psi (
1045: t ) \rangle$ its adiabatic evolution, operating with $\hat{B}^\dagger
1046: (t)$, so to get:
1047: \beq
1048: \biggl[ i \frac{ \partial}{ \partial t}
1049: - \hat{B}^\dagger ( t ) \hat{H} ( t ) \hat{B} ( t ) +
1050: \hat{B} ( t )^\dagger i \frac{ \partial}{ \partial t} \hat{B} ( t )
1051: \biggr] \hat{B}^\dagger ( t ) | \Psi ( t ) \rangle = 0 \:\: .
1052: \label{addi2}
1053: \eneq
1054:
1055: Eq.(\ref{addi2}) may be rewritten in a 2$\times$2 matrix formalism.
1056: Let $ \left( \begin{array}{c} u_+ \\ u_- \end{array} \right )$ be the
1057: components of $ | \Psi ( t ) \rangle$ in the adiabatic basis. The
1058: corresponding system of differential equations reads
1059:
1060: \beq
1061: i \frac{d}{dt}\left ( \begin{array}{c}
1062: u_+ \\
1063: u_- \end{array} \right ) \:=
1064: \hat{H} _A(t) \:
1065: \left ( \begin{array}{c}
1066: u_+ \\
1067: u_-
1068: \end{array} \right )\label{addi6}
1069: \eneq
1070: where we have defined
1071:
1072: \bea
1073: \hat{H}_A=
1074: \left( \begin{array}{cc}
1075: r+ \wo \smqt & \met \wo \sin\vartheta e^{-i \wo t} \\
1076: \met \wo \sin\vartheta e^{i \wo t} & - r - \wo \smqt \end{array}
1077: \right)\:.
1078: \label{berryham}
1079: \enea
1080: The extra term appearing on the diagonal w.r.to the hamiltonian of
1081: Eq.(\ref{three_1}) is just the Berry phase:
1082: \beq
1083: \langle + , t | i \frac{d}{ d t } | + , t \rangle = -
1084: \langle - , t | i \frac{d}{d t } | - , t \rangle = \omega _o \: \smqt \:\: .
1085: \eneq
1086: %
1087: Eq.(\ref{addi6}) is solved in Appendix B and the full propagator
1088: in the representation of the instantaneous eigenvectors reads:
1089: \bwt
1090:
1091: \beq
1092: U(t,t')\:=\:
1093: \left ( \begin{array}{cc}
1094: (\cos(\epsilon (t-t'))- i \eta \sin(\epsilon (t-t')))e^{i/2 \varphi
1095: (t-t')} &
1096: -i \beta \sin (\epsilon (t-t'))e^{i/2 \varphi(t+t')} \\
1097: -i \beta \sin (\epsilon (t-t'))e^{-i /2 \varphi (t+t')} &
1098: (\cos(\epsilon (t-t'))+ i \eta \sin(\epsilon (t-t'))) e^{-i/2 \varphi
1099: (t-t')}
1100: \end{array} \right )\;.
1101: \label{propagatorespin}
1102: \eneq
1103: %
1104: \ewt
1105: where $ \epsilon=\pm\sqrt{ (r + \frac{\wo}{2}\cos
1106: \vartheta )^2+s^2}$ and $s = \frac{\omega _o}{2} \sin \vartheta $. Also:
1107: %
1108: \[\beta=\frac{\wo}{2\epsilon } \sin\vartheta \:\:\: ,
1109: \:\:\:\eta =\frac{ r + \frac{\wo}{2}\cos
1110: \vartheta }{ \epsilon}\: . \]
1111: %
1112: This is the propagator in the adiabatic basis. Therefore, in order to
1113: switch to the fixed spin basis, one should write
1114: $U_{spin}(t,t')=B(t)U(t,t')B^\dagger(t') $, where $B(t) $ is given by
1115: Eq.(\ref{five}).
1116: The four elementary stretches imply the following substitutions in
1117: the propagator of Eq.(\ref{propagatorespin}):
1118:
1119: $u_\rightarrow$) {\sl forward orbiting in the upper arm of
1120: the ring :} $\:\:\:\:\:\:\varphi (t) = \omega _o t $ and $\vartheta \to \vartheta $.
1121:
1122: $ u_\leftarrow$) {\sl backward orbiting in
1123: the upper arm of the ring :} $\:\:\:\:\:\:\varphi (t) = \pi - \omega _o t $ and
1124: $\vartheta \to -\vartheta $.
1125:
1126: $d_\rightarrow$) {\sl forward orbiting in the lower arm of the ring :}
1127: $\:\:\:\:\:\:\varphi (t) = 2\pi - \omega _o t $ and $\vartheta \to -\vartheta $.
1128:
1129: $ d_\leftarrow$) {\sl backward orbiting in the lower arm of the ring :}
1130: $\:\:\:\:\:\:\varphi (t) = \pi + \omega _o t $ and $\vartheta \to \vartheta $.
1131:
1132: \section{the conductance}
1133: %
1134: In this section, we derive the DC conductance ${\cal{G}}$ across the ring, at
1135: the Fermi energy. Within Landauer's approach, ${\cal{G}}$ is given by
1136: %
1137: \beq
1138: {\cal{G}}=\frac{e^2}{\h}\sum_{\sigma,\sigma'}
1139: \left| \mathcal A (\sigma;\sigma'|E_0)\right|^2
1140: \:\:\:\: .
1141: \label{ec.1}
1142: \eneq
1143: %
1144: We will here consider the dependence on the external magnetic field
1145: ($\phi /\phi _0$) and on the spin-orbit strength $k_{SO} R
1146: $\cite{nota} both in absence and in presence of dephasing at the
1147: contacts. The various amplitudes in Eq.(\ref{ec.1}) have been
1148: numerically computed, as discussed in Sec.s(III-VI).
1149: %
1150: \begin{figure}[!htp]
1151: \centering \includegraphics[width=\columnwidth]{GvsB.eps}
1152: \caption{(color online) $a) $ Magnetoconductance of an ideal ring as a
1153: function of the magnetic field $\phi/\phi_0$. The Fourier transform
1154: (FT) of the magnetoconductance ({\sl in the inset}) shows only the
1155: AB peak at freq. $\phi_0^{-1} $. $b) $ Magnetoconductance of a
1156: realistic ring in which the S matrix of Eq.(\ref{matr}) regulates the
1157: scattering of the electron at the leads. As a consequence of the
1158: backscattering, the FT of the magnetoconductance ({\sl inset})
1159: shows higher order frequencies.}
1160: \label{GvsB}
1161: \end{figure}
1162: %
1163: In Fig.(\ref{GvsB}) we show the magnetoconductance of the ring in the
1164: absence of RSOI ($k_{SO}R =0$).
1165: %
1166: In panel a) of Fig.(\ref{GvsB}), only the path of
1167: Fig. (\ref{noWLpaths}, $1a)$ has been considered,
1168: i.e., full transmission across the ring is
1169: assumed, as it would be the case for ideal coupling to the leads. The
1170: corresponding Fourier spectrum is showed in the corresponding inset.
1171: The well known
1172: Ahronov-Bohm sinusoidal pattern implies that just the fundamental
1173: frequency $\phi_0^{-1}$ appears.
1174:
1175: To make the model more realistic, we allow for higher order looping of
1176: the electron within the ring. In Ref.\cite{noiletter}, only the paths
1177: of the kind of Fig.(\ref{noWLpaths},2a),\ref{noWLpaths},2b)) were
1178: included. Here, we consider also the paths of the kind of
1179: Fig.(\ref{WLpaths}) in which the electron can be backscattered at the
1180: leads. We use here $\bar r=0$ in the scattering matrix between the
1181: arms and the leads, which means that no backreflection in the incoming
1182: lead is present. The magnetoconductance of the ring, in this regime,
1183: is showed in Fig. (\ref{GvsB}b). Because of the inclusion of time
1184: reversed paths (TRP) within the ring, we see that higher
1185: order frequencies appear in the Fourier spectrum. In particular, the
1186: inset shows a peak at $2/\phi_0$ which is the signature of weak
1187: localization\cite{nonloso}.
1188: %
1189: \begin{figure}[!htp]
1190: \centering \includegraphics[width=\columnwidth]{barge.eps}
1191: \caption{(color online)Conductance as a function of $\phi/\phi_0$ {\sl
1192: (left panels)} and $k_{SO}R$ {\sl (right panels)} for ideal {\sl
1193: (black curves)} and realistic {\sl (red curves)} contacts. An
1194: increasing amount of dephasing at the contacts is also included: {\sl
1195: from top to bottom:} $\zeta=\pi/3,\pi,2\pi$}.
1196: \label{GvsBSO}
1197: \end{figure}
1198: %
1199: We also include some dephasing due to diffusiveness in the contacts by
1200: adding a random phase $z\in (-\zeta,\zeta ) $ for each scattering at
1201: the leads. In Fig.(\ref{GvsBSO}), we report the conductance {\cal{G}}
1202: {\it vs.} $ \phi / \phi_0$ , with $k_{\text {SO}}R = 0$ {\sl (left
1203: panels)} or {\it vs.} $k_{SO} R$ with $\phi/\phi_0=0$ {\sl (right
1204: panels)}. These are averaged over $N=1000$ realizations of disorder,
1205: and plotted increasing the window of phase randomness ($\zeta = \pi/3
1206: ,\pi, 2\pi$ from top to bottom). The black curves refer to the case of
1207: Fig(\ref{GvsB} $a)$) (ideal contacts) while the red curves refer to
1208: the case of realistic contacts (Fig(\ref{GvsB} $b)$), with
1209: $\bar{r}=0$.
1210:
1211: By comparing the top left panel of Fig.(\ref{GvsBSO}) with
1212: Fig.(\ref{GvsB}), we see that the ring is rather insensitive to small
1213: dephasing at the contacts. By increasing the amount of dephasing
1214: {\sl (middle and bottom left panels in Fig.(\ref{GvsBSO}))} we find that the
1215: sensitiveness is larger in the case of realistic contacts.
1216: This is due to the fact that for realistic coupling, the
1217: electrons in the ring can experience higher order paths, since it scatters
1218: with the leads many times.
1219:
1220: In the right panel of Fig.(\ref{GvsBSO}), we plot the DC conductance
1221: {\it vs.} $k_{\text {SO}}R$ at $\phi/\phi_0 =0 $ for both ideal
1222: contacts and realistic contacts (and $\bar{r}=0$) {\sl (black
1223: and red lines in each box)}, with an increasing phase randomization
1224: {\sl (boxes
1225: from top to bottom with $\zeta = \pi/3 ,\pi, 2\pi$)}, averaged over
1226: $N=1000$ realizations. In the case of ideal contacts and little
1227: dephasing {\sl (top right panel black curve)}, we see again the
1228: quasiperiodic oscillation of the conductance reproducing the
1229: localization conditions at the expected values of $k_{\text {SO}}R$
1230: \cite{frustaglia,molnar,souma,dario,noiletter}. When including higher
1231: order processes, interference effects give rise to a slightly
1232: different pattern. In the case of realistic contacts, we note that
1233: the device is seriously affected by dephasing, mainly because
1234: including the TRPs contributing to the transmission amplitude
1235: increases the number of scattering processes
1236: at the leads. Indeed, when the
1237: dephasing is quite large, it gives rise to
1238: random oscillations that are not
1239: averaged out, so that they wash out the conductance oscillations.
1240: The effect takes
1241: place for
1242: $\zeta\sim\pi$ when TRPs are included, in contrast to $\zeta \sim 2\pi$
1243: when the TRPs are absent. As regular magnetoconductance oscillations
1244: are experimenally
1245: observed \cite{nitta,yau,morpurgo,kato,molenkamp} with little
1246: precentage of contrast between maxima and minima, we conclude that, in
1247: real samples, dephasing is ubiquitous.
1248:
1249: \section{Spin Transmission}
1250:
1251: In this Section we calculate the rotation of the spin of the electron
1252: transmitted through the ring. We first consider an incoming electron
1253: beam with in-plane spin polarization (let's say, polarized along the
1254: $x$ direction ). The spin rotation is measured by calculating
1255: the average value of the outgoing spin:
1256: %
1257: \beq
1258: \langle S_z\rangle= \frac{\left\langle\Psi_{out}\right|S_z\left|\Psi_{out}\right\rangle}{\left\langle
1259: \Psi_{out}\right.\left|\Psi_{out}\right\rangle }\;.
1260: \label{szout}
1261: \eneq
1262: %
1263: Since in the previous Section we have shown that higher order looping
1264: just adds subleading higher order harmonics to the conductance, here we
1265: focus on the case of ideal contacts, that is, we include in the
1266: calculation only paths as the ones of Fig.(\ref{noWLpaths},1a). The
1267: in plane polarization can be considered as a superposition of equal
1268: weighted z-polarized spin components. In the absence of RSOI
1269: ($ k_{SO} R =0 $),
1270: opposite spin polarizations do not interfere with each other. As a
1271: consequence, the total expected $\langle S_z \rangle $ component keeps
1272: zero at the exit. Fig.(\ref{spintramiscela}) shows the
1273: magnetoconductance for increasing $ k_{SO} R$, and the
1274: corresponding expected spin component polarized along the $z-$axis
1275: at the exit of the ring. The
1276: Zeeman term is on the diagonal of the spin Hamiltonian of Eq.(\ref{three_1})
1277: and operates to keep the $z-$components of the spin polarization fixed,
1278: while the RSOI is offdiagonal and tends to favor inplane spin polarization.
1279: This implies that when the magnetic field increases
1280: ($\phi /\phi _0 >> k_{SO} R$) the spin polarization gets frozen to the
1281: incoming polarization.
1282: This is the reason why,in Fig.(\ref{spintramiscela}), at high fields,
1283: the trasmitted spin polarization is inplane. Incidentally we observe
1284: that this result should not be expected in real systems in which spin
1285: relaxation can occur due to electron-phonon interaction or other mechanisms
1286: as hyperfine interaction with nuclear spins.
1287: Spin relaxation would induce flipping of that spin component that is
1288: energetically unfavourable and the final
1289: transmission of the spin will result to be partly out of the $x-y$ plane.
1290: On the contrary,
1291: when $\phi /\phi _0 \sim
1292: k_{SO} R $ the competition of the Zeeman and the RSOI induces
1293: a coherent rotation of the spin while the electron travels
1294: along the ring.
1295: Fig.(\ref{spintramiscela}) shows that, when $\phi /\phi _0 \sim
1296: k_{SO} R$, the spin is moved significantly out of the $x-y $ plane,
1297: consistently affecting the AB oscillations.
1298: %
1299: \begin{figure}[!htp]
1300: \centering
1301: \includegraphics[width=\columnwidth]{spintramiscela.eps}
1302: \caption{Magnetoconductance and expectation value of the outgoing $\hat
1303: z $ spin component for an incoming spin in the $x$ direction at
1304: different values of the RSOI strength ($k_{SO}R = 0,1,2,4 $ indicated
1305: in the pictures).}
1306: \label{spintramiscela}
1307: \end{figure}
1308: %
1309: To better understand what happens when $\phi /\phi _0 \sim k_{SO} R$,
1310: we isolate the spin ``up'' polarization for the incoming particle
1311: in what follows ($\left\langle\Psi_{in}\right|S_z\left|\Psi_{in}\right\rangle/\left\langle
1312: \Psi_{in}\right.\left|\Psi_{in}\right\rangle=1$) and
1313: we separately plot in Fig.(\ref{spintraup}) the two contributions to
1314: the conductance $G_{up-up} $ ({\sl full line}) and $ G_{down-up} $
1315: ({\sl dotted line}), for opposite polarizations of the outgoing
1316: electron. $G_{up-up} $ is the contribution to the conductance
1317: arising from the particle flux that mantains the same polarization at the exit as the incoming one, while $ G_{down-up} $ refers to a particle flux
1318: having the opposite polarizations at the exit with respect to the one
1319: at the entrance. Of course, when $ k_{SO} R = 0 $, the electron spin is in
1320: the ``up''direction for any $\phi/\phi_0$. When both RSOI and magnetic
1321: field are present, with $\phi /\phi _0 >> k_{SO} R$, the spin
1322: polarization is steadly in the $z-$ direction, except for sharp
1323: reversals at flux values $\phi _0 m/2 $ ($m$ integer). However,
1324: $G_{down-up} $ is vanishingly small at these places, together with
1325: $G_{up-up} $. Therefore the conductance vanishes at these points
1326: anyhow, and the transmitted spin polarization is fully up, except for these
1327: points. On the contrary, in the parameter intervals characterized by
1328: $\phi /\phi _0 \sim k_{SO} R$, both $ G_{up-up} $ and $ G_{down-up} $
1329: are non vanishing (see Fig.\ref{spintraup}), so that the spin is rotated
1330: at the exit, with
1331: nonvanishing transmission amplitude.
1332: %
1333: \begin{figure}[!htp]
1334: \centering \includegraphics[width=\columnwidth]{spintraup.eps}
1335: \caption{Separate contributions to the magnetoconductance for
1336: different outcoming spin polarizations, $ G_{up-up} $ ({\sl full line})
1337: and $ G_{down-up} $ ({\sl dotted line} ) compared to the expected
1338: value of the outgoing $\hat z $ spin component. The incoming spin is
1339: polarized ``up'' . Different values of the RSOI strength are reported
1340: ($k_{SO}R = 0,1,2,4 $ indicated in the picture ).}
1341: \label{spintraup}
1342: \end{figure}
1343: %
1344: %
1345:
1346: We now examine in more detail the expected dependence of the outgoing
1347: spin polarization on the RSOI, at zero magnetic field. For an
1348: incoming electron polarized with spin ``up'', Fig.(\ref{spinwire}(a))
1349: shows that a large enough RSOI produces a rotation of the spin which
1350: points down at the exit for any $k_{SO}R$ value. Meanwhile, the total
1351: conductance oscillates with $k_{SO}R$. This finding also appears in
1352: Ref.\cite{souma} and is quite remarkable, because it is the
1353: consequence of the interference between the two arms of the ring. In
1354: order to point out the role of quantum interference, we discuss here,
1355: for comparison, what happens by transporting a spinful electron along
1356: a single arm of the ring (the upper one). When just one arm is
1357: considered, (see Fig(\ref{spinwire}b)) spin polarization oscillates,
1358: as a function of $k_{SO} R $ \cite{datta}, while the conductance is
1359: always unitary, because of the conservation of the particle flux. In
1360: the limit of large $k_{SO} R$ the propagation amplitude for the
1361: travelling electron acquires a simple analytical form. From
1362: Eq.(\ref{propagatorespin}) we see that, in the representation of the
1363: instantaneous spin eigenvectors, the spin propagator at the exit time
1364: $t_f$ (with $\omega _o t_f = \pi )$, in this limit is diagonal:
1365: %
1366: \beq
1367: U^{u\rightarrow}(t_f,0)=
1368: \left(
1369: \begin{array}{c c}
1370: i e^{-i\pi\gamma} & 0\\
1371: 0 & -i e^{i\pi\gamma}
1372: \end{array}
1373: \right) \:\: ,
1374: \eneq
1375: %
1376: so that the spin appears not to be rotated at the exit in the rotating
1377: reference frame. However, in order to move from the representation of
1378: the instantaneous spin eigenvectors to the reference basis, one has to
1379: perform the transformation with the unitary $B$ matrix of
1380: Eq.(\ref{five}) , with $\vartheta = \pi /2$. The spin part of the
1381: propagator for an electron travelling into the upper arm (upper path
1382: of Fig.(\ref{noWLpaths},1a)) is then:
1383: %
1384: \bea
1385: U^{u\rightarrow}_{spin}(t_f,0)=
1386: B(t_f)U^{u\rightarrow}(t_f,0)B^\dagger(0)
1387: \nonumber\\
1388: =\left(
1389: \begin{array}{c c}
1390: i \cos(\pi \gamma) & \sin(\pi \gamma)\\
1391: - \sin(\pi \gamma) & -i \cos(\pi \gamma)
1392: \end{array}
1393: \right) \:\: .
1394: \enea
1395: %
1396: If we inject $up$ electrons in the upper arm only, the expectation
1397: value of the the outcoming $S_z$ defined in Eq.(\ref{szout}) is:
1398: $\langle S_z\rangle=\cos(2\pi\gamma)=\cos(\pi k_{SO} R)$ (note that
1399: for large enough $\gamma$ this result well approximates the red-full
1400: line in Fig.(\ref{spinwire}b)). The conductance is unitary, $G=2e^2/h$
1401: (the black-dashed line in Fig.(\ref{spinwire}b), as no interference
1402: takes place.
1403: %
1404:
1405: We now go back to the transmission along both arms simultaneously and
1406: examine the resulting interference. According to the rules given
1407: after Eq.(\ref{propagatorespin}), in the same limiting case as above,
1408: the propagator accounting for transmission of incoming $up$ spins
1409: across the ring is:
1410: %
1411: \beq
1412: U^{u\rightarrow+d\rightarrow}_{spin}(t_f,0)=
1413: \left(
1414: \begin{array}{c c}
1415: 0 & 2 \sin(\pi \gamma)\\
1416: - 2\sin(\pi \gamma) & 0
1417: \end{array}
1418: \right) \:\: ,
1419: \eneq
1420: %
1421: so that the spin at the exit is reversed. In fact, in the expectation
1422: value of Eq.(\ref{szout}), the oscillations in the numerator
1423: compensate those in the denominator, eventually giving $\langle
1424: S_z\rangle=-1$ (for large enough $\gamma$ this result well approximate
1425: the red-full line in Fig.(\ref{spinwire}a)). The conductance however
1426: oscillates according to $G/(2e^2/h)= 2\sin^2(\pi k_{SO}R/2)$, as
1427: plotted in the black-dashed line in (Fig.\ref{spinwire}a)).
1428:
1429: It is quite remarkable that this result is only found at zero magnetic
1430: field. Indeed, no matter how small $B$ is, the time reversal symmetry
1431: is broken and the spin oscillates with $k_{SO}R $ (see
1432: Fig.(\ref{brokensymmetry},a). However, for very small magnetic field
1433: these oscillations are confined close to special values of $ k_{SO} R
1434: = 2l $($l$ integer) and display a Lorentzian shape around these
1435: points. The role of the magnetic field is to make the oscillations
1436: broader.
1437:
1438: To summarize, there are two limiting conditions in the outgoing spin
1439: polarization, for incoming ``up'' spin polarization: $a)$ {\sl the
1440: zero RSOI } which leaves the incoming spin polarization unchanged,
1441: provided no relaxation takes place; $b)$ {\sl the zero magnetic flux
1442: case } in which the RSOI produces a flip of the spin at the exit. It
1443: is interesting that when the flux $\phi $ is an integer number of flux
1444: quanta $ \phi _0$, the crossover between case $a)$ and case $b) $,
1445: with $k_{SO} R$ increasing from the value zero to values $ k_{SO} R >>
1446: \phi /\phi _0 $ is rather sharp. This is shown in
1447: Fig. (\ref{brokensymmetry},b) where the expectation value of the
1448: outcoming spin is plotted vs. $k_{SO} R$ for different integer values
1449: of $\phi/\phi_0$. For values of $\phi/\phi_0 >> k_{SO} R $ the
1450: outgoing spin polarization is the same as that at the entrance, ( $up$
1451: in the picture). On the contrary, by increasing $k_{SO} R$ at non
1452: zero $B$ field, we see again a pattern similar to the one of
1453: Fig. (\ref{brokensymmetry},a), but shifted to higher values of $k_{SO}
1454: R$.
1455:
1456: \begin{figure}[!htp]
1457: \centering
1458: \includegraphics[width=\columnwidth]{spinwire.eps}
1459: \caption{ (color online) Conductance ({\sl broken line}) and spin
1460: polarization ({\sl full line}) of the outgoing electron for $B=0$, as
1461: a function of the RSOI. The spin of the incoming electrons is
1462: polarized ``$up$'' : $a)$ The ring case with ideal contacts. $b)$ a
1463: single wire of the same length and curvature as one of the ring arms.}
1464: \label{spinwire}
1465: \end{figure}
1466: %
1467: \begin{figure}[!htp]
1468: \centering
1469: \includegraphics[width=\columnwidth]{SvsAvBs.eps}
1470: \caption{(color online) Expectation value of the outgoing spin for
1471: incoming spin ``$up$'' polarized electrons as a function of the RSOI
1472: for different values of the magnetic flux : $a)$ $\phi/\phi_0$ zero or
1473: very small; $b)$ increasing integer values of $\phi/\phi_0$. }
1474: \label{brokensymmetry}
1475: \end{figure}
1476:
1477: \section{Conclusions}
1478:
1479:
1480: To conclude, we have employed a path integral real time approach to
1481: compute the DC conductance of a ballistic one dimensional mesoscopic
1482: ring in both external electrical and magnetic fields orthogonal to the
1483: ring plane.The spinful electron experiences a Rashba spin orbit
1484: interaction and the Zeeman term. We employ a piecewise saddle point
1485: approximation for the orbital motion, but we implement the full
1486: scattering matrix at the leads and sum over all the possible higher
1487: order paths up to convergency of the result. Our approach goes beyond
1488: other recent semiclassical calculations. Our theory is
1489: nonperturbative and separates the adiabatic spin dynamics from the non
1490: adiabatic one by using the rotating frame for the spin travelling
1491: around the ring. In practice, we diagonalize the time dependent spin
1492: Hamiltonian in the representation of the spin eigenvectors of the
1493: instantaneous Hamiltonian. This allows us to explore a wide range of
1494: Hamiltonian parameters, ranging from the limit of strong magnetic
1495: field and weak Rashba SOI to the opposite case. In both extreme
1496: regimes our piecewise saddle point approximation is very efficient as
1497: quantum fluctuations with flipping of the spin has little influence on
1498: the orbital motion. This is also seen from the number of paths
1499: required to gain full convergency. As explained in Sec. $V$ the
1500: separation of adiabatic from non adiabatic spin dynamics shows that in
1501: the intermediate regimes our approximation is less justified, but
1502: nevertheless, the results it produces seem to be in rather good
1503: agreement with recent numerical calculations
1504: \cite{frustaglia, molnar, shen,
1505: dario,souma} and experiments\cite{nitta, yau,
1506: morpurgo,kato,molenkamp}. When we include also time reversed paths,
1507: the Fourier transform of the magnetoconductance shows the typical
1508: $\phi_0/2$ peak due to weak localization\cite{nonloso}.This would be the
1509: only surviving contribution if an ensemble average or different rings
1510: were measured\cite{koga}.
1511:
1512: We have also allowed for nonideal couplings between ring and leads as
1513: we account for dephasing effects due to diffusiveness at the
1514: contacts. The results satisfactorily compare with experiments where
1515: the contrast between maxima and minima in the interference fringes is
1516: always few tens of percentage of the background DC signal.
1517:
1518: \begin{acknowledgments}
1519: We acknowledge financial support by the Italian Ministry of Education
1520: (PRIN) and by the CNR within ESF Eurocores Programme FoNE (contract
1521: N. ERAS-CT-2003-980409).
1522: \end{acknowledgments}
1523:
1524:
1525: \appendix
1526: \section{ Motion of a classical spin in a rotating magnetic field}
1527:
1528: In this appendix we derive the classical equations of motion for the
1529: spin from the Lagrangian in Eq.(\ref{venti}), by assuming for the
1530: orbital coordinate the saddle point solution $\dot{\varphi}$ =
1531: constant. Once the orbital motion is dealt with in this way, the
1532: Lagrangian for the spin degrees of freedom is given by (besides a
1533: constant contribution)
1534: %
1535: \beq
1536: \tilde{\cal L} [ \Theta , \Phi , \dot{\Phi } ] / \hbar = \left( \frac{1 -
1537: \cos \Theta }{2} \right) \dot{\Phi} + \vec{{\cal B}} \cdot \vec{S}
1538: \:\:\:\: ,
1539: \label{al1}
1540: \eneq
1541:
1542: where the effective time dependent magnetic field is
1543: ${\cal{B}} \equiv ({\cal{B}}_+,{\cal{B}}_-,{\cal{B}}_z ) =
1544: \left ( \gamma \dot{\vt} \: e^{i\vt}, \gamma \dot{\vt} \:
1545: e^{-i\vt},-\omega _c/2 \right ) $.
1546: %
1547: To derive the equations of motion from a variational principle, we
1548: write the Berry phase term in the total spin action as
1549: %
1550: \beq
1551: \tilde{S}_B =
1552: \int_{ \Phi ( 0 ) }^{ \Phi ( t_f )} \: d t \: \left( \frac{1 -
1553: \cos \Theta }{2} \right) d \Phi = \int_\Sigma \sin \Theta \: d \Theta \wedge
1554: d \Phi
1555: \:\:\:\: ,
1556: \label{al3}
1557: \eneq
1558:
1559: %
1560: where $\Sigma$ is the spherical triangle with vertices given by the
1561: north pole on the sphere and by the points with coordinates $( \Theta
1562: ( 0 ) , \Phi ( 0 ) ) $ , $ ( \Theta ( t_f ) , \Phi ( t_f ) )$. Let $
1563: ( t , u ) $ be a parametrization of the spherical triangle, such that
1564: $ \vec{S} ( t , 1 ) = \vec{S} ( t )$, and $\vec{S} ( t , 0 ) = (0 , 0
1565: , 1)$. Thus, one may rewrite the action $\tilde{S}_B$ in
1566: Eq.(\ref{al3}) as
1567: %
1568: \beq
1569: \tilde{S}_{B} = \int_0^T \: d t \: \int_0^1 \: d u \: {\bf S} \cdot \left[
1570: \frac{ \partial {\bf S}}{ \partial t} \times \frac{ \partial {\bf S}}{
1571: \partial u} \right]
1572: \:\:\:\: .
1573: \label{e.9}
1574: \eneq
1575:
1576: %
1577: To derive the equations of motion, we consider a variation
1578: $ \vec{S} ( t , u ) \rightarrow \vec{S} ( t , u ) + \delta \vec{S}
1579: ( t , u )$ such that $\vec{S} ( T , u ) $
1580: and $\vec{S} ( 0 , u )$ are "locked", that is, $ \delta \vec{S} ( 0 , u )
1581: = \delta \vec{S} ( T , u ) = 0$.
1582: %
1583: Since $ [ \vec{S} ( t , u ) ]^2 = 1$ $\forall t , u $, one gets $\vec{
1584: S} \cdot \frac{ \partial \vec{ S}}{ \partial t} = \vec{ S} \cdot
1585: \frac{ \partial \vec{ S}}{ \partial u} = 0 $, As a consequence,
1586: $\frac{ \partial \vec{ S}}{ \partial t} \times \frac{ \partial \vec{
1587: S}}{ \partial u} $ is parallel to $\vec{S}$. As $\delta \vec{S} \cdot
1588: \vec{S} = 0$, this implies that
1589: %
1590: \beq
1591: \int_0^T \: d t \: \int_0^1 \: d u \: \delta \vec{S} \cdot \left[
1592: \frac{ \partial \vec{ S}}{ \partial t} \times \frac{ \partial \vec{ S}}{
1593: \partial u} \right] = 0
1594: \:\:\:\: .
1595: \label{e.9.b}
1596: \eneq
1597:
1598: %
1599: Thus, by integrating by parts we see that the only nonzero variation of
1600: $\tilde{S}_B$ is given by the boundary term
1601: %
1602: \beq
1603: \delta \tilde{S}_B = \int_0^{t_f } \: d t \: \delta \vec{S} ( t )
1604: \cdot \left[ \vec{S} ( t ) \times \frac{ \partial \vec{S} ( t ) }{
1605: \partial t} \right]
1606: \:\:\:\: ,
1607: \label{e.12b}
1608: \eneq
1609:
1610: %
1611: where we have used the fact that $\vec{S} ( t , 1 ) = \vec{S} ( t )$.
1612: %
1613: By equating to zero the total variation of the action, one obtains
1614: %
1615: \beq
1616: \vec{S}\times \frac{d\vec{S}}{dt} =\vec{\mathcal B}
1617: \:\:\:\: ,
1618: \label{appe1}
1619: \eneq
1620: that is, the classical equations of motion we used in section IV.
1621: %
1622: To show that Eqs.(\ref{appe1}), when the spin components are expressed
1623: in polar coordinates, are equivalent to Eqs.(\ref{em1},\ref{em2}),
1624: let us set $\omega = \gamma \dot{\vt} $ and $ \Omega = -\omega _c/2$.
1625: Also, we define
1626: %
1627: \bea
1628: S_{+}=S_{x}+iS_{y}\hspace*{2cm}{\mathcal B}_{+}={\mathcal B}_{x} +
1629: i{\mathcal B}_{y}\nonumber\\ S_{-}=S_{x}-iS_{y}\hspace*{2cm}{\mathcal
1630: B}_{-}={\mathcal B}_{x}-i{\mathcal B}_{y}\vspace*{1cm}\nonumber
1631: \:\:\:\: .
1632: \enea
1633: In terms of the new variables, the equations of motion are given by
1634: %
1635: \bea
1636: \frac{d S_{z}}{dt}=\frac{i}{2}({\mathcal B}_{+}S_{-}-{\mathcal B}_{-}S_{+})
1637: \nonumber\\
1638: \frac{d S_{+}}{dt}= i({\mathcal B}_{z} S_{+}-{\mathcal B}_{+}S_{z})\nonumber\\
1639: \frac{d S_{-}}{dt}=-i({\mathcal B}_{z} S_{-}-{\mathcal B}_{-}S_{z}) \label{e2}
1640: \:\:\:\: .
1641: \label{appe2}
1642: \enea
1643: %
1644: or:
1645: %
1646: \bea
1647: \frac{d m(t)}{dt} =i\:\left ( \Omega -\dot\varphi \right )\:p(t)-2\:i
1648: \omega \:S_{z} \nonumber \\
1649: \frac{d p(t)}{dt} =i\:\left (\Omega -\dot{\varphi} \right )\:m(t)
1650: \nonumber\\
1651: \frac{dS_{z}}{dt} =-\frac{i\omega }{2}\:{m(t)}\:
1652: \label{e4}
1653: \enea
1654: %
1655: where
1656: %
1657: \bea
1658: p(t)=S_{+}e^{(-i\varphi)}+S_{-}e^{(i\varphi)}\nonumber \\
1659: m(t)=S_{+}e^{(-i\varphi)}-S_{-}e^{(i\varphi)}\nonumber
1660: \:\:\:\: ,
1661: \enea
1662: %
1663: and $ 1= 4|S|^2=4 {S_{z}}^2+ (p^2-m^2 ) $.
1664: %
1665: By introducing $b= \Omega -\dot \vt $, we obtain:
1666: %
1667: \bea
1668: \frac{d(m(t) + p(t))}{dt} &=& i b \left( m(t) + p(t) \right) -2 i\omega
1669: \: S_z (t) \label{m3.1}\\
1670: \frac{d {S}_z(t)}{dt} &=& - i \omega \: m(t) \label{m3.2}
1671: \enea
1672: %
1673: Resorting to the polar coordinates $(\Theta ,\Phi )$ for the spin $\vec{S}$,
1674: we get:
1675: %
1676: \bea
1677: \left[\dTh \ct +i \left( \dP- \Omega \right)\st \right] e^{i \chi}
1678: + i \ct \alpha \df=0 \label{una} \\
1679: \left[\dTh - \omega \: \sin{\chi}\right] \st = 0
1680: \:\:\:\: .
1681: \label{due}
1682: \enea
1683: %
1684: Eq.(\ref{due}) is the same as Eq.(\ref{em1}). The real part
1685: of Eq.(\ref{una}) is proportional to the imaginary part: both give
1686: Eq.(\ref{em2}) when equated to zero, which completes the proof.
1687: %
1688: \section{The spin propagator}
1689: %
1690: In order to find the propagator of the Berry Hamiltonian $ \hat{H}_A $
1691: of Eq.(\ref{berryham}),
1692: we solve
1693: the system of differential Eq.(\ref{addi6}), in the
1694: representation of the instantaneous eigenvectors:
1695: \bwt
1696: \beq
1697: i \frac{d}{dt}\left ( \begin{array}{c}
1698: u_+ \\
1699: u_- \end{array} \right ) \:=
1700: \left ( \begin{array}{cc}
1701: r+ \wo \smqt & \wo \smt \cmt e^{-i \wo t} \\
1702: \wo \smt \cmt e^{i \wo t}& -r- \wo \smqt
1703: \end{array} \right )
1704: \left ( \begin{array}{c}
1705: u_+ \\
1706: u_-
1707: \end{array} \right )\label{addi6_1}
1708: \:\:\:\: .
1709: \eneq
1710: %
1711: \ewt
1712: %
1713: To solve Eq.(\ref{addi6_1}), first of all, we switch to a
1714: time-independent coefficient matrix by defining:
1715: %
1716: \beq
1717: \left ( \begin{array}{c}
1718: y_+ \\
1719: y_- \end{array} \right ) \:=
1720: \left ( \begin{array}{cc}
1721: e^{ +i \frac{ \wo }{2} t } & 0 \\
1722: 0 & e^{- i \frac{\wo}{2}t }
1723: \end{array} \right )
1724: \left ( \begin{array}{c}
1725: u_+ \\
1726: u_-
1727: \end{array} \right )\:.
1728: \label{T}
1729: \eneq
1730: %
1731: By setting
1732: \[Y=
1733: \left ( \begin{array}{c}
1734: y_+ \\
1735: y_- \end{array} \right ) \:\:\: ; \:\:
1736: W=\left ( \begin{array}{c}
1737: u_+ \\
1738: u_-
1739: \end{array} \right )\:,\]
1740: we define the matrix $T$ through
1741: \beq Y\:=\:T\:W,\;\;\;W\:=\:T^{-1}\:Y\:.\eneq
1742:
1743: Eqs.(\ref{addi6_1}) now read:
1744: \bea
1745: i \frac{ d y_+}{ d t } ( t ) = (r-\frac{\wo}{2}cos(\vartheta)) y_
1746: + ( t ) + \frac{ \wo}{ 2 }
1747: \sin \vartheta y_- ( t ) \:,\nonumber \\
1748: i \frac{ d y_-}{ d t } ( t ) =
1749: + \frac{ \wo}{ 2 } \sin \vartheta y_+ ( t )
1750: +(-r+\frac{\wo}{2}cos(\vartheta)) y_- ( t )\:.
1751: \enea
1752: Now we define $ r' = r - \frac{ \wo}{2} \cos \vartheta $ and $s=
1753: \frac{ \wo}{2} \sin \vartheta $, so that in matrix form we have:
1754: \begin{eqnarray}
1755: i \frac{d}{dt}
1756: \left ( \begin{array}{c}
1757: y_+ \\
1758: y_- \end{array} \right ) \:=
1759: \left ( \begin{array}{cc}
1760: r' & s \\
1761: s & -r'
1762: \end{array} \right )
1763: \left ( \begin{array}{c}
1764: y_+ \\
1765: y_- \end{array} \right ) \:,
1766: \label{dicia}
1767: \end{eqnarray}
1768: in a compact form we can rewrite the last equation as:
1769: \beq
1770: i \frac{d}{dt}Y\:=\:C\:Y\:,
1771: \label{matr1}
1772: \eneq
1773: which defines the matrix $C$.\\ We now decouple the
1774: previous system of equation by diagonalizing the matrix $C$.
1775: Its eigenvalues are $\lambda=\pm \epsilon=\pm\sqrt{r'^2+s^2}$ and the
1776: matrix that diagonalizes $C$ is
1777: \beq
1778: P=\left ( \begin{array}{cc}
1779: 1 & \frac{r'-\epsilon}{s} \\
1780: \frac{\epsilon-r'}{s} & 1
1781: \end{array} \right ) \:.
1782: \eneq
1783: Its inverse is
1784: \beq
1785: P^{-1}=\left ( \begin{array}{cc}
1786: \frac{s^2}{2\epsilon(\epsilon -r')} & \frac{(\epsilon-r')s}{2
1787: \epsilon(\epsilon -r')} \\
1788: -\frac{(\epsilon-r')s}{2\epsilon(\epsilon -r')} & \frac{s^2}{2\epsilon(
1789: \epsilon -r')}
1790: \end{array} \right ) \:.
1791: \eneq
1792:
1793: Eq.(\ref{matr1}) now reads:
1794: \beq
1795: i \frac{d}{dt} P^{-1}\: Y\:=\:P^{-1}\:C\:P\:P^{-1}\:Y\:,
1796: \eneq
1797: which, by defining $V\:=\:P^{-1}\:Y$, becomes
1798: \beq
1799: i \frac{d}{dt} V\:=\left ( \begin{array}{cc}
1800: \epsilon & 0 \\
1801: 0 & -\epsilon
1802: \end{array} \right ) \: V\:.
1803: \eneq
1804: Its formal solution is:
1805: \beq
1806: V(t)\:=\left ( \begin{array}{cc}
1807: e^{-i\epsilon (t-t')} & 0 \\
1808: 0 & e^{i\epsilon (t-t')}
1809: \end{array} \right ) \: V(t')\:,
1810: \eneq
1811: or, in matrix form
1812: \beq
1813: V(t)\:=\:S(t-t')\:V(t')\:.
1814: \eneq
1815: Now we apply inverse transformations, in order to obtain the full
1816: Schr\"odinger propagator, that is the matrix transformation between
1817: ($W(t')$ and $W(t)$).
1818: \[
1819: W(t)\:=\:T^{-1}(t)\:P\:S\:P^{-1}\:T(t')\:W(t')
1820: \;\;,\]
1821: where $P$ is time independent.
1822: The full evolution operator in the adiabatic basis is:
1823: \beq
1824: U(t,t')\:=\:T^{-1}(t)\:P\:S\:P^{-1}T(t')\:;
1825: \eneq
1826: By performing all the matrix products, we obtain the result
1827: of Eq.(\ref{propagatorespin}), as
1828: given in the text.
1829:
1830:
1831: \begin{thebibliography}{99}
1832:
1833: \bibitem{aharonovbohm} Y. Aharonov, D. Bohm, Phys. Rev. {\bf 115} 485
1834: (1959).
1835:
1836: \bibitem{webb} S. Washburn and R. A. Webb, Rep. Prog. Phys. {\bf 55}, 1311
1837: (1992).
1838:
1839: \bibitem{anandans} J. Anandan, Science {\bf 297}, 1656 (2002).
1840:
1841: \bibitem{berry} M. V. Berry, Proc. R. Soc. London, A {\bf 392}, 45 (1984).
1842:
1843: \bibitem{anandan}Y. Aharonov, J. Anandan, Phys.
1844: Rev. Lett. {\bf 58}, 1593 (1987).
1845:
1846: \bibitem{rashba}E.I. Rashba, Fiz. Tverd. Tela {\bf 2}, 1224 (1960) [Sov.Phys. -
1847: Solid State {\bf 2}, 1109 (1960),Y.A. Bychkov, E.I. Rashba, J.Phys.{\bf C17},
1848: 6039 (1984).
1849:
1850: \bibitem{meijer} F. E. Meijer, A. F. Morpurgo, T. M. Klapwijk, T. Koga
1851: and J. Nitta, Phys. Rev. B {\bf 70} 201307 (2004).
1852:
1853: \bibitem{miller} J. B. Miller, D. M. Zumb\"uhl, C. M. Marcus,
1854: Y. B. Lyanda-Geller, D. Goldhaber-Gordon, K. Campman
1855: and A. C. Gossard, Phys. Rev. Lett. {\bf 90} 076807
1856: (2003).
1857:
1858: \bibitem{aronov} A. G. Aronov and Y. L. Lyanda-Geller, Phys. Rev.
1859: Lett.{\bf 70}, 343 (1993).
1860:
1861: \bibitem{ac} Y.Ahronov and A. Casher Phys. Rev. Lett. {\bf 53}, 319
1862: (1984).
1863:
1864: \bibitem{nitta}J. Nitta, T.Koga, F. E. Meijer, Physica E {\bf 18}, 143
1865: (2003); F. E. Meijer, J. Nitta, T. Koga,
1866: A.F. Morpurgo, T. M. Klapwijk, Physica E {\bf 22},
1867: 402, (2004); M.J. Yang, C.H. Yang, K.A. Cheng and
1868: Y.B. Lyanda-Geller, cond-mat/0208260.
1869:
1870: \bibitem{yau} J. B. Yau, E. P. de Poortere, and M. Shayegan,
1871: Phys. Rev. Lett. {\bf 88}, 146801 (2002).
1872:
1873: \bibitem{morpurgo} A. F. Morpurgo, J. P. Heida, T. M. Klapwijk,
1874: B. J. van Wees and G. Borghs, Phys. Rev. Lett. {\bf
1875: 80}, 1050 (1998).
1876:
1877: \bibitem{kato} Y.K. Kato, R.C. Meyers, A.C. Gossard, D.D. Awshalom
1878: cond/mat 2005
1879:
1880: \bibitem{molenkamp}M. K\"onig, A. Tschetschetkin, E.M. Henkiewicz,
1881: J. Sinova, V. Hock, V. Daumer, M. Sch\"afer,
1882: C.R. Becker, H. Buhmann and L.W. Molenkamp,
1883: Phys. Rev. Lett. {\bf 96} 0760804 (2006).
1884:
1885: \bibitem{noiletter} R.Capozza, D.Giuliano, P. Lucignano, and A. Tagliacozzo,
1886: Phys. Rev. Lett. {\bf 95}, 226803 (2005).
1887:
1888: \bibitem{shayegan} B. Habib, E. Tutuc and M. Shayegan cond-mat/0612638.
1889:
1890: \bibitem{nonloso} A review is given by C. W. J. Beenakker and H. van
1891: Houten, Solid State Phys., {\bf 44}, 1 (1991).
1892:
1893: \bibitem{loss} D. Loss, P. Goldbart and A. V. Balatsky, Phys. Rev. Lett.
1894: {\bf 65}, 1655 (1990); H. A. Engel and D. Loss, Phys. Rev. B {\bf 62},
1895: 10238 (2000).
1896:
1897: \bibitem{feynman}R. P. Feynmann, \emph{Quantum Mechanics and Path Integral},
1898: Mc Graw-Hill, New York 1965.
1899:
1900: \bibitem{tserkov} Y. Tserkovnyak and A. Brataas cond-mat/0611086
1901:
1902: \bibitem{frustaglia} D. Frustaglia, K. Richter, Phys. Rev. B {\bf 69} 235310
1903: (2004).
1904:
1905: \bibitem{molnar} B. Moln\'ar, F. M. Peeters and P. Vasilopoulos,
1906: Phys. Rev. B { \bf 69}, 155335 (2004); X.F.Wang and P. Vasilopoulos
1907: Phys. Rev. B {\bf 72}, 165336 (2005).
1908:
1909: \bibitem{shen} S.Q. Shen, Z.J. Li and Z. Ma, Appl. Phys. Lett. B
1910: {\bf 84}, 996 (2004).
1911: \bibitem{dario} D.Becioux, D.Frustaglia and M.Governale, Phys. Rev. B
1912: {\bf 72}, 113310 (2005).
1913:
1914: \bibitem{citro} R. Citro, F. Romeo and M. Marinaro Phys. Rev. B {\bf 74},
1915: 115329 (2006).
1916:
1917: \bibitem{souma} S.Souma and B.K.Nikoli\'c, Phys.Rev.B{\bf 70},195346(2004).
1918:
1919: \bibitem{lozano} G.S. Lozano, M.J. Sanchez, Phys. Rev. B {\bf 72},
1920: 205315 (2005).
1921:
1922: \bibitem{wu} B.H. Wu and J.C. Chao, Phys. Rev. B {\bf 74}, 115313 (2006)
1923:
1924: \bibitem{landauer} R. Landauer, IBM J. Res. Dev.{\bf 1} 223 (1957);
1925: M. Buttiker, IBM J. Res. Dev.{\bf 32} 317 (1988).
1926:
1927: \bibitem{haldane}F. D. M. Haldane Phys. Rev. Lett. {\bf 50}, 1153
1928: (1983).
1929:
1930: \bibitem{plet}M. Pletyukhov, Ch. Amann, M. Metha and M. Brack, Phys. Rev. Lett.
1931: {\bf 89 }, 116601 (2002); O.Zaitsev, D. Frustaglia and
1932: K. Richter, Phys. Rev. Lett. {\bf94}, 026809 (2005).
1933:
1934: \bibitem{morette} C. Morette-de Witt, A. Maheshwari and B. Nelson,
1935: Phys. Rep. {\bf 50}, 55 (1979).
1936:
1937: \bibitem{nota} In the current notation $k_{SO}R=2m\alpha
1938: R/\hbar=2\gamma$
1939:
1940: \bibitem{moran} G. Morandi and E. Menossi, {Eur. J. Phys.} {\bf 5}
1941: 49 (1984).
1942:
1943: \bibitem{aurbach} A. Aurbach, F. Berruto and L. Capriotti, Field
1944: theory for low-dimensional systems Ed.s G. Morandi,
1945: P.Sodano, A.Tagliacozzo and V.Tognetti
1946: (Springer New York 2000).
1947:
1948: \bibitem{datta} S. Datta and B. Das, Appl. Phys. Lett. {\bf 56}, 665
1949: (1990).
1950: \bibitem{koga} T.Koga, Y.Sekine and J.Nitta, Phys.Rev.B{\bf 74}, 041302
1951: (2006).
1952:
1953: \end{thebibliography}
1954:
1955: \end{document}
1956: