cond-mat0304205/p4.tex
1: 
2: %\documentclass[preprint]{revtex4}
3: \documentclass[twocolumn,showpacs,preprintnumbers,amsmath,amssymb]{revtex4}
4: \usepackage{graphicx}
5: \begin{document}
6: \title{Gauge Fields, Geometric Phases, and Quantum Adiabatic Pumps
7:       }
8: \author{Huan-Qiang Zhou$^{1*}$, Sam Young Cho$^2$, and Ross H. McKenzie$^{1,2}$}
9: 
10: \affiliation{$^1$Centre for Mathematical Physics,
11:          The University of Queensland, 4072, Australia}
12: \affiliation{$^2$Department of Physics, The University of Queensland,
13:              4072, Australia}
14: \date{\today}
15: 
16: \begin{abstract}
17: %
18:  Quantum adiabatic pumping of charge and spin
19:  between two reservoirs (leads) has  recently
20:  been demonstrated  in nanoscale electronic devices.
21:  Pumping occurs when system parameters
22:  are  varied in a  cyclic manner and  sufficiently
23:  slowly that the quantum system always remains
24:  in its  ground state.  We show that  quantum pumping has a natural
25:  geometric representation in terms of gauge fields
26:  (both  Abelian and non-Abelian)  defined on the
27:  space of system parameters.  We make explicit the
28:  similarities and differences with Berry's
29:  geometric phase. 
30:  Tunneling from a scanning tunneling microscope tip through a magnetic atom
31:  could be used to demonstrate the non-Abelian
32:  character of the gauge field.
33: \end{abstract}
34: \pacs{
35: 03.65.Vf,  %Phases: geometric; dynamic or topological
36: 73.23.-b, %Electronic transport in mesoscopic systems
37: 73.63.-b  %Electronic transport in nanoscale materials and structures 
38: }
39: \maketitle
40: 
41: %%%%%%%%%%%%%%%%%%%%%%% Main Text %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
42: 
43: Normally transport of electrical charge
44: is dissipative (i.e., it produces heat). However, quantum adiabatic
45: pumping \cite{Thouless83} provides a means in nanoscale electronic
46: devices to use novel quantum effects
47: to transport single
48: electrons with minimal dissipation \cite{ag99}.
49: Furthermore, it is also possible to pump
50: electron spin without pumping charge \cite{sc01,a02}.
51: Both charge \cite{s99} and spin \cite{wpmu03} pumping have
52: been recently achieved experimentally, by cyclic variation of
53: the gate voltages that control the shape of an open
54: quantum dot.
55: This motivated extensive theoretical research in this topic,
56: especially on quantum charge pumping
57: \cite{zsa99,saa00,ak00,ae001,mm01}.
58: Quantum spin pumping opens
59: the way for applications in spintronics.
60: It is sometimes suggested, but not explicitly shown, that
61: quantum pumping is related to Berry's phase.
62: As first emphasized
63: by Berry \cite{b84}, discrete quantum systems
64: have the counter-intuitive
65: property that when some of the parameters controlling
66: the system are slowly varied and brought
67: back to their initial value the quantum state of the
68: system is different to the initial state.
69: That is,
70: a quantum state may acquire a geometric phase
71: $\exp (i \gamma _C)$ in addition to the normal dynamic phase
72: $\exp (-(i/ {\hbar}) \int E(t) dt)$.
73: This remarkable discovery may be recast
74: in the language of holonomy theory \cite{s83}.
75: Subsequent work showed
76: that non-Abelian gauge potentials can arise as
77: a result of degeneracies of energy levels of the system
78: \cite{wz84,msw86,z88}.
79: However, it is unlikely that adiabatic pumping, characteristic of
80: quantum {\it open} systems considered in this Letter, 
81: results from Berry's phase
82: for {\it closed discrete} systems.
83: 
84: We present a systematic treatment of
85: quantum adiabatic pumping in open systems in terms of parallel transport
86: and gauge fields
87: (both Abelian and non-Abelian) defined on the system
88: parameter space, 
89: which reveals a unifying concept of geometric phase underlying scattering
90: states. 
91: We make explicit
92: the similarities and differences with 
93: Berry's phase associated with cyclic variations of
94: closed quantum systems (both degenerate and non-degenerate)
95: (see  Table   1).
96:  In the scattering approach developed by Brouwer \cite{b98},
97:  based on an earlier work of B\"uttiker, Thomas,  and Pr\^etre \cite{btp94},
98:  a compact
99: formula was presented for the pumped charge (current) in terms of the parametric derivatives 
100: of the time-dependent scattering matrix subjected to the modulating potential.
101: We show that the pumped charge, given by Brouwer's
102: formula \cite{b98}, is essentially the geometric phase
103: associated with the $U(1)$ subgroup of the gauge group $U(M)$
104: ($M$ is the number of channels in a certain lead),
105: whereas the non-Abelian sector $SU(M)$ describes 
106: the adiabatic pumping associated with the internal degrees
107: of freedom such as spin.
108: Expressions are given for the gauge potentials
109: associated with tunneling from an STM (scanning tunneling microscope)
110: through a magnetic atom.
111: We suggest an experiment 
112: which can be used to illustrate 
113: the non-Abelian character of the gauge field.
114: 
115: {\it The quantum system.}
116: Consider a mesoscopic system with $N$ leads, 
117: and for the $n$-th lead there are $M_n$ channels. 
118: Our aim is to study quantum pumping 
119: by periodically varying a set of the independent external parameters 
120: $X \equiv (X^1, \cdots, X^{\nu}, \cdots, X^p)$ 
121: slowly as a function of time $t$.
122: In the scattering approach, 
123: the $S$ matrix is an ${\cal N} \times {\cal N}$ matrix
124: with ${\cal N}$ the total number of channels, ${\cal N} =\sum _{n=1}^N M_n$.
125: We define vectors 
126: $ {\bf n}_\alpha \equiv (S_{\alpha 1}, \cdots, S_{\alpha  {\cal N}})$
127: in terms of the rows
128: of the scattering matrix $S[X(t)]$ associated with the $n$-th lead. 
129: The unitarity of the scattering matrix implies that these vectors 
130: are orthonormal 
131: ${\bf n}^*_\alpha 
132: \cdot {\bf n}_\beta = \delta _{\alpha\beta}, ~~~\alpha,\beta= 1, \cdots,M_n.$
133: That is, this provides us with a smooth set of (local frame) bases 
134: ${\bf n}_\alpha(t)$. 
135: 
136: {\it The gauge potential.}
137: Assume that the parallel transport law  
138: \begin{equation}
139: \Psi^*_\alpha \cdot d \Psi_\beta =0 \label {pt}
140: \end{equation}
141: holds for (fibre) vectors $\Psi_\alpha$,
142: where $d \Psi_\alpha$ is the variation in $\Psi_\alpha$ 
143: resulting from a variation  $dX$
144: in the external parameters.
145: If $\Psi_\alpha (0) = {\bf n}_\alpha(0)$, 
146: i.e., the initial vector describing 
147: the scattering process in which the incident particle comes 
148: from the $\alpha$-th channel in the $n$-th lead, 
149: then the degeneracy of the channels implies
150: that the transported vector $ \Psi_\alpha(t)$ 
151: {\it must be a linear combination of all} ${\bf n}_\alpha(t)$, 
152: $\Psi_\alpha(t)= \sum_\beta U_{\alpha\beta}(t) {\bf n}_\beta(t)$. 
153: Expressed another way, 
154: the transported vector describes a combined scattering
155: process in which particles come from all channels in the $n$-th lead.
156: Obviously, $U(t)$ is unitary, i.e., $U(t) \in U(M_n)$. 
157: Physically,
158: this means certain information about where the incident particles come
159: from is lost during parallel transport, and is
160: encoded in the unitary matrix $U(t)$. 
161: Inserting into the
162: parallel transport law in Eq. (\ref{pt}), we have 
163: \begin{equation}
164: (U^{-1}d U)_{\alpha\beta}= -{\bf n}^*_\beta \cdot d {\bf n}_\alpha.
165: \label{vpt}
166: \end{equation}
167: Since ${\bf n}_\alpha$ varies 
168: as the parameters $X^{\nu}$ vary with time, 
169: we can thus define the gauge potential 
170: $
171:  A_{\alpha\beta\nu} \equiv 
172:  {\bf n}^*_\beta \cdot {\partial}_\nu {\bf n}_\alpha,
173:  \label{eq2}
174: $
175: where $\partial_\nu \equiv \partial/\partial{X^\nu}$
176: so that 
177: $$ (U^{-1}d U)_{\alpha\beta}= -\sum_\nu A_{\alpha\beta\nu}  d X^\nu.$$
178: This can be integrated in terms of exponential integrals. 
179: For the period $\tau$ of an adiabatic cycle, we have
180: \begin{equation}
181: (U(\tau))_{\alpha\beta}= ({\rm P} \exp (-\oint 
182:   \sum_{\nu} A_\nu d X^\nu))_{\alpha\beta}, 
183: \label{u}
184: \end{equation}
185: where ${\rm P}$ denotes path ordering.
186: Defining $A \equiv \sum_{\nu} A_\nu dX^\nu$, 
187: one can see it is Lie algebra $u(M_n)$ valued and thus anti-Hermitian.
188: $A$ plays the role of a gauge potential, 
189: as in the case of Berry's phase \cite{wz84}
190: for closed (discrete) quantum systems. 
191: 
192: {\it Gauge transformation.}
193: The gauge group $U(M_n)$ 
194: originates from the unitary freedom in choosing local bases 
195: ${\bf n}_\alpha (\alpha= 1, \cdots, M_n)$,
196: $$
197: {\bf n'}_\alpha (t) = \sum_\beta \omega _{\alpha\beta}(t){\bf n}_\beta (t).
198: $$
199: This amounts to different choices of the scattering matrix: 
200: $S'(t) =\Omega (t) S(t)$ with $\Omega (t)$ a diagonal block matrix, 
201: the $n$-th block of which is 
202: an $M_n \times M_n$ unitary matrix $\omega (t)$. 
203: Physically speaking, left multiplication of the scattering matrix $S(t)$
204: by $\Omega(t)$ just redistributes the scattering particles among
205: different incoming channels 
206: associated with a certain lead, which does not affect
207: correlations at the scatterer and so the physics remains the
208: same. 
209: The gauge potential $A(t)$ transforms as
210: \begin{equation}
211: A'(t) = d \omega \omega^{-1} + \omega A \omega^{-1}. \label{gt}
212: \end{equation}
213: The gauge field strength  defined by
214: $F \equiv dA - A \wedge A$   
215: transforms covariantly $F' = \omega F \omega^{-1}$. 
216: Therefore, a $U(M_n) \equiv U(1) \times SU(M_n)$ gauge field is
217: defined on a $p$-dimensional
218: parameter space which drives the quantum pumping.
219: The trace of $U(\tau)$ given by Eq. (\ref{u}) is gauge-invariant.
220: 
221: {\it Justification of parallel transport.}
222: Now we need to justify the assumption of the parallel transport 
223: law in Eq. (\ref{pt}). 
224: Physically, by  ``adiabatic'' we mean that 
225: the dwell time $\tau_p$ during which particles scatter off the scatterer
226: is much shorter than the time period
227: $\tau= 2\pi/\omega_a$ during which 
228: the system completes the adiabatic cycle.
229: Here $\omega_a$ is a slow frequency characterizing the adiabaticity and
230: $\tau_p$ is related to
231: (but not determined alone by) Wigner time delay matrix
232: $\tau_w(s,E) \equiv -i S^\dagger(s,E) \partial S(s,E)/\partial E $ with $E$
233: being the energy of scattering particles 
234: and $s$ being the so called {\em epoch} defined as $s=\omega_a t$ \cite{ae003}.
235: Then
236: the response to
237: the variation of the particle
238: distribution in a certain channel {\it is only limited to} 
239: channels associated with the same lead. 
240: That means we ignore any responses which involve channels
241: in different leads. 
242: Such a response can be treated as dissipation,
243: a correction to the adiabatic limit. Then, the parallel transport law 
244: in Eq. (\ref{pt})
245: follows from the parallel transport law for the wave function.
246: The latter is a solution of the time-dependent Schr\"odinger equation
247: $i \hbar \partial_t |\Phi(t)\rangle = H(t) |\Phi(t) \rangle$. 
248: As is well known, the Schr\"odinger equation induces a parallel transport law
249: ${\rm Im}\langle\phi(t)|\partial_t \phi(t)\rangle=0$ \cite{aa87}, 
250: with $|\phi(t)\rangle \equiv \exp (i\int h(t)dt)
251: |\Phi(t)\rangle$, where 
252: $h(t) = \langle\Phi|H|\Phi\rangle/\langle \Phi|\Phi\rangle$. 
253: We can write the wave function $|\phi(t)\rangle $ as a linear combination of
254: {\it all} scattering states associated with a certain lead in the adiabatic
255: case. 
256: Formally, $|\phi(t)\rangle = \sum _\alpha c_\alpha |\psi_\alpha(t)\rangle$, 
257: with $|\psi_\alpha(t) \rangle $ denoting the scattering states
258: in which the scattered particles come from channels associated with the
259: $n$-th lead,
260: and $c_\alpha$ being arbitrary constants.
261: Then we have $\langle \psi_\alpha(t) | \partial_t \psi_\beta (t) \rangle=0$.
262: The adiabatic assumption implies that
263: $|\psi_\alpha(t)\rangle$ may be expanded in terms of instantaneous 
264: asymptotic scattering states, 
265: $|\psi_\alpha(t)\rangle 
266:  = \sum_\beta U_{\alpha\beta}(t) |\psi^S_\beta(t)\rangle$,
267: with
268: $|\psi^S_\alpha(t)\rangle 
269: = |\alpha\rangle_{in}+\sum^{\cal N}_{\beta=1} S_{\alpha\beta}(t) 
270:   |\beta\rangle_{out}$,
271: $\alpha=1, \cdots, M_n$.
272: Here, $|\alpha \rangle _{in}$ and $|\beta \rangle_{out}$ denote, 
273: respectively, the incoming and outgoing scattering states,
274: which are normalized such that they carry a unit flux. 
275: Substituting into
276: the parallel transport law for $|\psi_\alpha\rangle$, 
277: one gets Eq. (\ref{vpt}) which is equivalent to the parallel transport law
278: for row vectors of the scattering matrix.
279: 
280: {\it Quantum adiabatic pumping.}
281: In order to establish the connection between the geometric phase
282: above and the quantum pumping charge,
283: we need to consider the time-reversed scattering states
284: $|\hat\psi^S_\alpha(t)\rangle 
285: = |\hat{\alpha}\rangle_{in}
286:  +\sum^{\cal N}_{\beta=1} S_{\beta\alpha}(t) |\hat{\beta}\rangle_{out}$
287: with $\hat{}$ denoting the counterparts under time reversal operation
288: \cite{note},
289: which constitute a solution of the Schr\"odinger equation for
290: the time-reversed Hamiltonian $\hat{H}$ at any given (frozen) time
291: {\it at the epoch scale} \cite{ae003}.
292: This gives rise to another gauge potential 
293: $\hat{A}_{\alpha\beta\nu}
294:  \equiv \hat{\bf n}^*_\beta \cdot \partial_\nu \hat{\bf n}_\alpha$
295: with $\hat{\bf n}_\alpha \equiv (S_{1\alpha}, \cdots, S_{{\cal N} \alpha})$,
296: i.e., the column vectors of the scattering matrix $S(t)$.
297: In this case, the gauge group arises from
298: redistribution of scattering particles among different outgoing channels.
299: %
300: If $\hat {\bf n}'_\alpha(t) = \sum_\beta \hat \omega_{\alpha\beta}
301:    \hat {\bf n}_\beta(t)$, then the gauge transformation takes
302: $\hat A'(t) = d \hat \omega \hat \omega^{-1} 
303: + \hat \omega \hat A \hat \omega^{-1}$.
304: The gauge fields $A$ and $\hat A$ are connected via time reversal operation.
305: If we identify the emissivity into 
306: the $\alpha$-th channel in the $n$-th lead as 
307: ${\rm Im} [{\hat A}_{\alpha\alpha}/2\pi]$ \cite{btp94}, 
308: then we immediately reproduce
309: Brouwer's formula \cite{b98} describing charge pumping, 
310: which turns out to be associated
311: with the Abelian subgroup $U(1)$,
312: \begin{equation}
313: Q= \frac {e}{2\pi} {\rm Im} \oint {\rm Tr} {\hat A}, \label {pc}
314: \end{equation}
315: with $Q$ being the charge transferred 
316: into the $n$-th lead during one cycle. 
317: That is, the charge transferred 
318: during adiabatic pumping 
319: is essentially the geometric phase 
320: associated with the charge sector $U(1)$. 
321: This also explains why 
322: Planck's constant $\hbar$ does not  occur in the
323: adiabatic quantum pumped charge (current), a peculiar feature different
324: from the Landauer-B\"uttiker conductance. 
325: However, as is well known, the geometric phase
326: is determined only up to a multiple of $2\pi$. 
327: This concerns global geometric properties,
328: i.e., the winding number of the overall phase of 
329: the gauge transformation in Eq. (\ref{gt}),
330: $ N\equiv 1/(2\pi i) \oint {\rm Tr} (d\hat \omega \hat \omega^{-1})$.
331: The requirement that all physical 
332: observables be invariant under the gauge transformation leads 
333: us to the conclusion that $O(Q)=O(Q-eN)$, 
334: with $O$ denoting any observable. 
335: This result has been noticed by Makhlin and Mirlin \cite{mm01},
336: without proper justification, 
337: for the counting statistics in quantum charge pumps 
338: (see also, Ref. \cite{ak00}). 
339: 
340: To see the effects caused by  non-Abelian gauge potentials, 
341: we need to consider gauge invariant quantities.
342: Besides ${\rm Tr} \hat{U}(\tau)$, we see that both the determinant 
343: and eigenvalues of $\hat{U}(\tau)$ are gauge invariant. 
344: Actually, there are $M_n$ independent gauge invariant quantities
345: such as the eigenvalues $\exp[i\gamma_\alpha]$.
346: On the other hand, there are $M_n$ independent simultaneous
347: observables such as
348: the pumping currents $I_\alpha$ flowing into the $\alpha$-th channel, 
349: which must be gauge invariant.
350: Therefore, one may expect that
351: the pumping currents $I_\alpha$ are some functions of $\gamma_1, \cdots, \gamma_{M_n}$.
352: Because our argument only relies on gauge invariance and 
353: does not depend on any details of the system, such functions must be
354: model-independent. 
355: Guided by the results for the so-called
356: ``Abelianized" non-Abelian gauge potentials, i.e.,
357: the gauge potentials which
358: turn out to be diagonal in a certain gauge, % \cite{hqz},
359: we have 
360: \begin{equation}
361: I_\alpha =-\frac{1}{2\pi \tau} \gamma_\alpha.
362: \end{equation}
363: This connects the physical observables with the eigenvalues of the geometric
364: matrix phase.
365: Especially, the charge pumping current $I_c$ corresponding to
366: the Abelian sector $U(1)$ is
367: $I_c \equiv \sum_{\alpha} I_\alpha =-1/(2\pi \tau) \sum_\alpha \gamma_\alpha$.
368: One may verify that this is consistent with Eq. (\ref{pc}) since
369: $\dot{Q}(t) = e I_c$ and $d \det \hat{U}(t) =-{\rm Tr}
370:  \hat{A}(t) \det \hat{U}(t)$. % \cite{hqz}.
371: Alternatively, $Q= e \; {\rm Im} \ln \det U(\tau)$.
372: Similarly, we may define generalized ``spin" pumping currents associated with
373: the Cartan subalgebra of the non-Abelian sector $SU(M_n)$.
374: The simplest non-Abelian case $U(2)$ is relevant to 
375: the charge and spin pumping.
376: 
377: %
378: 
379: {\it Tunneling through a single magnetic spin.}
380: Consider the Hamiltonian 
381: which describes a system 
382: consisting of two leads coupled to a single site, the spin of which
383: has an exchange interaction $J$ with a magnetic spin
384: \cite{zb02},
385: \begin{eqnarray}
386: H &= &\sum_{k \in {L,R},\sigma} \epsilon_{k\sigma} 
387:         c^\dagger _{k \sigma} c_{k \sigma}
388: + J \sum _{\sigma,\sigma'} d^\dagger_\sigma \Omega _{\sigma \sigma'} d_{\sigma'}
389:  \nonumber \\
390:  && \hspace*{1cm}
391: +\sum_{k \in {L,R},\sigma;\sigma'} 
392:  (V_{k\sigma,\sigma'} c^\dagger _{k\sigma} d_{\sigma'} +{\rm H.c.}).
393: \label {ham}
394: \end{eqnarray}
395: Here $c^\dagger_{k\sigma}$ and $c_{k\sigma}$ are, respectively, 
396: the creation and destruction 
397: operators of an electron with momentum $k$ 
398: and spin $\sigma$ in either the left ($L$) or
399: the right ($R$) lead, and $d^\dagger_\sigma$ 
400: and $d_\sigma$ are the counterparts of the single 
401: electron with spin $\sigma$ at the spin site.  
402: The quantity $\epsilon_{k\sigma}$ are the single 
403: particle energies of conduction electrons in the two leads, 
404: which we will assume
405: $\epsilon_{k\sigma} = v_F (|k|-k _F)$ 
406: with the convention that $v_F=1$, and the momentum is measured from
407: the Fermi surface for electrons in leads. 
408: The electrons on the spin site are connected to those in the two leads 
409: with the tunneling matrix 
410: elements $V_{k\sigma,\sigma'}$. 
411: For simplicity, we assume symmetric tunneling barriers between
412: the local spin and the leads, and only keep the spin-conserved coupling; viz.
413: $V_{L++}=V_{L--}=V_{R++}=V_{R--}=V$ and $V_{L+-}=V_{L-+}=V_{R+-}=V_{R-+}=0$.
414: The entries of the 
415: coupling matrix $\Omega$
416: take the form
417: $\Omega _{++} =-\Omega _{--}=\cos \theta$
418: and
419: $\Omega _{+-} = \Omega_{-+}^* =\sin \theta \exp (-i \phi)$.
420: The model is exactly soluble as far as the scattering
421: matrix is concerned. 
422: 
423: Once the scattering matrix is determined,
424: our general formalism leads us to 
425: the non-Abelian gauge potential,
426: \begin{equation}
427: \hat A= \hat A_\theta d \theta + \hat A_\phi d \phi,
428: \label{nb}
429: \end{equation}
430: where 
431: $\hat A_\theta \equiv \hat A^1_\theta \sigma^1 + \hat A^2_\theta \sigma^2 
432:   +\hat A_\theta^3 \sigma^3$
433: with 
434: $\hat A^1_\theta = i(\sin (\delta_1 -\delta _2) \cos \theta \cos \phi + 
435: (1-\cos (\delta_1-\delta_2)) \sin \phi)/4$,
436: $\hat A_\theta^2 = i(\sin (\delta_1 -\delta _2) \cos \theta \sin \phi - 
437: (1-\cos (\delta_1-\delta_2)) \cos \phi)/4$, and
438: $\hat A_\theta^3 =-i(\sin (\delta_1 -\delta _2) \sin \theta)/4$,
439: and
440: $\hat A_\phi \equiv \hat A^1_\phi \sigma^1 + \hat A^2_\phi \sigma^2 
441:  +\hat A_\phi^3 \sigma^3$ with
442: $
443: \hat A^1_\phi = -i(\sin (\delta_1 -\delta _2) \sin \theta \sin \phi -
444: (1-\cos (\delta_1-\delta_2)) \sin \theta \cos \theta \cos \phi)/4$,
445: $\hat A^2_\phi= i(\sin (\delta_1 -\delta _2) \sin \theta \cos \phi + 
446: (1-\cos (\delta_1-\delta_2)) \sin \theta \cos \theta \sin \phi)/4$,
447: and
448: $\hat A^3_\phi=-i(1-\cos (\delta_1 -\delta _2)) \sin ^2 \theta /4$.
449: Here $\delta_i \ (i=1,2)$ are 
450: the phase shifts defined by $\delta_1= -2 \tan ^{-1} (\Gamma/(k-J))$
451: and $\delta_2= -2 \tan ^{-1} (\Gamma/(k+J))$ with the tunneling rate
452: $\Gamma \equiv V^2$.
453: 
454: 
455: The gauge field strength $\hat F$ then takes the form
456: $
457: \hat F=-i(1-\cos (\delta_1-\delta_2)){\vec n}\cdot {\vec \sigma} d \Omega/4$.
458: Here 
459: $d \Omega = \sin \theta d \theta \wedge d \phi$ is the invariant area element 
460: and
461: ${\vec n}=(\sin\theta \cos\phi, \sin\theta \sin\phi, \cos\theta)$
462: is the direction of the magnetic spin.
463: Obviously, this is just a simple rotation of the standard form 
464: $\hat F=-i (\alpha^2-1) \sigma^3  d\Omega /2$.
465: Up to a gauge transformation,
466: this is the same non-Abelian gauge potential,
467: found by Moody {\it et al.} \cite{msw86} for a diatomic molecule. 
468: This is
469: consistent with a theorem, proved in \cite{fm80},
470: stating that
471: the rotationally invariant connection on the sphere is essentially unique. 
472: To establish the relation between $\alpha$ and $\cos(\delta_1-\delta_2)$,
473: we need to calculate the gauge invariant quantity 
474: ${\rm Tr} \hat F \wedge * \hat F$,
475: with $*\hat F$ being the dual of $\hat F$.
476: Then we have $\alpha^2=(3-\cos(\delta_1-\delta_2))/2$.
477: When $J=0$, i.e., in the absence of the direct exchange interaction
478: between electrons and the local spin, the gauge field is a pure gauge
479: because $\hat F=0$. Since the Pauli matrices are traceless, we have 
480: ${\rm Tr} \hat A =0$, meaning that charge pumping is absent in the model
481: under consideration. That implies $\gamma_+ = -\gamma_-$.
482: Therefore,
483: the spin pumping current defined by 
484: $I_s=I_+-I_-$ becomes $I_s= -\gamma_+/(\pi \tau)$.  
485: 
486: %
487: %To observe the effects of the non-Abelian gauge field, it is 
488: %necessary to choose a cycle which varies both $\theta$ and $\phi$. 
489: %The analysis of Zee \cite{z88} for the gauge field in Eq. (\ref{nb})
490: %showed that
491:  One can compute a phase factor $\hat{U}_{SR}$ which is obtained 
492:  from the time-reversed counterpart of Eq. (\ref{u})
493:  for a ``spherical rectangle (SR)''.
494:  From $1/2{\rm Tr} \hat{U}_{SR} = \cos \pi \tau I_s$,
495:  the spin pumping current $I_s$ may be extracted and 
496: %
497:  is shown in Fig. \ref{fig1}
498:  as a function of system parameters for two different paths,
499:  $C_1$ and $C_2$. 
500:  The spin pumping current takes its maximum
501:  value around the resonant scattering lines $k= \pm J$. 
502: 
503:  {\it Possible experiments.}
504:  A scanning tunneling microscope (STM) has been used to
505:  detect a quantum mirage around a single magnetic
506:  cobalt atom placed on a non-magnetic metallic 
507:  copper surface \cite{Manoharan00}.
508:  Electron spin resonance (ESR)-STM experiments \cite{Farle98}  
509:  have advanced to the point that they have spatial resolution
510:  at the level of a few spins \cite{Ralph01}.
511:  The STM setup as shown in Fig. \ref{fig1} {\sf A}
512:  should make it possible
513:  to observe gauge invariant spin pumping via a single magnetic atom
514:  on the surface of the substrate.
515:  To measure the spin pumping current $I_s$, 
516:  one could replace one of the leads by a ferromagnetic one.
517:  The spin pumping current can then be measured via the charge pumping currents.
518: 
519: 
520: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
521:  {\it Acknowledgments.}
522:   We thank Urban Lundin, Gerard Milburn, and Christoph Renner
523:   for valuable discussions.  
524:   Thanks also go to Markus B\"uttiker, Claudio Chamon,
525:   and Frank Wilczek for enlightening comments and encouragement.  
526:   This work was supported by the Australian Research Council.
527: 
528: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
529: %\begin{thebibliography}{99}
530:  \begin{references}
531: 
532:  \bibitem[*] [E-mail: hqz@maths.uq.edu.au
533: 
534:  \bibitem{Thouless83}
535:  D. J. Thouless, 
536:  Phys. Rev. B {\bf 27}, 6083 (1983). 
537: 
538:  \bibitem{ag99}
539:   B. L. Altshuler and L. I. Glazman, 
540:   Science {\bf 283}, 1864 (1999).
541: 
542:  \bibitem{sc01}
543:  P. Sharma and C. Chamon, 
544:  Phys. Rev. Lett. {\bf 87}, 096401 (2001);
545: %  
546: % \bibitem{msc02}
547:   E. R. Mucciolo, C. Chamon, and C. M. Marcus,
548:   Phys. Rev. Lett. {\bf 89}, 146802 (2002).
549: 
550:  \bibitem{a02}
551:  T. Aono, Phys. Rev. B {\bf 67}, 155303 (2003); 
552: % cond-mat/0205395;
553:  M. Governale, F. Taddei, and R. Fazio, 
554:  cond-mat/0211211.
555: 
556:  \bibitem{s99}
557:  M. Switkes, C. M. Marcus, K. Campman, and A. C. Gossard,
558:  Science {\bf 283}, 1905 (1999);
559: %\bibitem{Brouwer01}
560:  For an alternative interpretation, see
561:  P. W. Brouwer, 
562:  Phys. Rev. B {\bf 63}, 121303 (2001).
563: 
564: 
565:  \bibitem{wpmu03}
566:  S. K. Watson, R. M. Potok, C. M. Marcus, and V. Umansky, 
567:   cond-mat/0302492.
568: 
569:  \bibitem{zsa99}
570:  F. Zhou, B. Spivak, and B. Altshuler, 
571:  Phys. Rev. Lett. {\bf 82}, 608 (1999).
572: 
573:  \bibitem{saa00}
574:  T. A. Shutenko, I. L. Aleiner, and B. L. Altshuler, 
575:  Phys. Rev. B {\bf 61}, 10366 (2000).
576: 
577:  \bibitem{ak00}
578:  A. Andreev and A. Kamenev, 
579:  Phys. Rev. Lett. {\bf 85}, 1294 (2000).
580: 
581:  \bibitem{ae001}
582:  J. E. Avron {\it et al.}, 
583:  Phys. Rev. B {\bf 62}, R10618 (2000);
584: %
585: % \bibitem{ae002}
586:  J. E. Avron {\it et al.}, %A. Elgart, G. M. Graf, and L. Sadun, 
587:  Phys. Rev. Lett. {\bf 87}, 236601 (2001).
588: 
589:  \bibitem{mm01}
590:  There is a difference of factor 2 between our result and equation (5) in
591:  Y. Makhlin and A. D. Mirlin, 
592:  Phys. Rev. Lett. {\bf 87}, 276803 (2001).
593: 
594: 
595:  \bibitem{b84}
596:  M. V. Berry,
597:  Proc. Roy. Soc. London. A {\bf 392}, 45 (1984).
598: 
599:  \bibitem{s83}
600:  B. Simon, 
601:  Phys. Rev. Lett. {\bf 51}, 2167 (1983).
602: 
603: 
604:  \bibitem{wz84}
605:  F. Wilczek and A. Zee,  
606:  Phys. Rev. Lett. {\bf 52}, 2111 (1984).
607: 
608: 
609:  \bibitem{msw86}
610:  J. Moody, A. Shapere, and F. Wilczek,  
611:  Phys. Rev. Lett. {\bf 56}, 893 (1986).
612: 
613:  \bibitem{z88}
614:  A. Zee,  
615:  Phys. Rev. A {\bf 38}, 1 (1988).
616: 
617:  \bibitem{b98}
618:  P. W. Brouwer, 
619:  Phys. Rev. B {\bf 58}, R10135 (1998).
620: 
621:  \bibitem{btp94}
622:  M. B\"uttiker, H. Thomas, and A. Pr\^etre, 
623:  Z. Phys. B {\bf 94}, 133 (1994);
624: % 
625:  see also,
626:  M. Moskalets and M. B\"uttiker,
627:  Phys. Rev. B 66, 035306 (2002).
628: 
629:  \bibitem{ae003}
630:  J. E. Avron {\it et al.}, %A. Elgart, G. M. Graf, and L. Sadun, 
631:  J. Math. Phys. {\bf 43}, 3415 (2002).
632: 
633: \bibitem{aa87}
634:  Y. Aharonov and J. Anandan,  
635:  Phys. Rev. Lett. {\bf 58}, 1593 (1987);
636: %
637: % \bibitem{sb88}
638:  J. Samuel and R. Bhandari,
639:  Phys. Rev. Lett. {\bf 60}, 2339 (1998).
640: 
641:  \bibitem{note}
642:  We emphasize that the time reversal operation here is
643:  carried out for the Hamiltonian at the frozen time, 
644:  which may be or may not be broken, depending on
645:  the Hamiltonian concerned.
646:  However, there is another time reversal operation
647:  at large time scale, which is always broken during pumping.
648: 
649: %\bibitem{hqz}
650: % H.-Q. Zhou, S. Y. Cho, and R. H. McKenzie, {\it unpublished}.
651: 
652:  \bibitem{zb02}
653:  J. X. Zhu and A. V. Balatsky, 
654:  Phys. Rev. Lett. {\bf 89}, 286802 (2002).
655:  
656: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
657: 
658:  \bibitem{fm80}
659:  M. Forgacs and N. Manton, 
660:  Commun. Math. Phys. {\bf 72}, 15 (1980);
661: %
662: % \bibitem{Witten77}
663:  E. Witten,  
664:  Phys. Rev. Lett. {\bf 38}, 121 (1977);
665: %
666: % \bibitem{Gu81}
667:  C. H. Gu, 
668:  Phys. Rep. {\bf 80}, 251 (1981).
669: 
670: 
671:  \bibitem{Manoharan00} 
672:  H. C. Manoharan, C. P. Lutz, and D. M. Eigler,
673:  Nature {\bf 403}, 512 (2000).
674: 
675:  \bibitem{Farle98}
676:  M. Farle, 
677:  Rep. Prog. Phys. {\bf 61}, 755 (1998);
678: %
679: % \bibitem{Bode98}
680:  M. Bode, M. Getzlaff, and R. Wiesendanger,
681:  Phys. Rev. Lett. {\bf 81}, 4256 (1998);
682: %
683: % \bibitem{Heinze00}
684:  S. Heinze {\it et al.},
685:  Science {\bf 288}, 1805 (2000).
686: 
687:  \bibitem{Ralph01}
688:  D. Ralph,
689:  Science {\bf 291}, 999 (2001);
690: %
691: % \bibitem{Weber01}
692:  W. Weber, S. Riesen, and H. C. Siegmann,
693:  Science {\bf 291}, 1015 (2001).
694: 
695:  \end{references}
696: %\end{thebibliography}
697: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
698: 
699: \begin{widetext}
700: \end{widetext}
701: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
702: \begin{table}[hb]
703: %\begin{ruledtabular}
704: \begin{center}
705: \begin{tabular}{c|c}
706:  \hline \hline
707:  Berry's Phase & Scattering (Pumping) Geometric Phase \\ \hline\hline
708:  Closed systems & Open systems \\
709:  Wave functions $|\psi_\alpha\rangle$ 
710:   & Row (column) vectors ${\bf n}_\alpha$ ($\hat {\bf n}_\alpha$)
711:      of the $S$ matrix \\
712:  Energy levels $E_n$ & Leads $n$ \\
713:  $M$ degeneracies &  $M$ channels \\
714:  Discrete spectrum (bound states) & Continuous spectrum (scattering states) \\
715:  Eigenstates in the $n$-th level & Channels in the $n$-th lead\\ 
716:  Parallel transport due to adiabatic theorem 
717:  &
718:  \begin{minipage}{8.5cm}
719:  Parallel transport due to 
720:  adiabatic charge and spin pumps
721: %Parallel transport due to formula {\it a la} B\"uttiker et
722: %al \cite{btp94} and Brouwer \cite{b98}
723:  \end{minipage} \\
724:  Gauge potential 
725:   $A_{\alpha\beta\nu}=\langle \psi^*_\beta| \partial_\nu 
726:                |\psi_\alpha\rangle$ 
727:   & Gauge potentials 
728:    $A_{\alpha\beta\nu}={\bf n}^*_\beta \cdot 
729:       \partial_\nu{\bf n}_\alpha$ and 
730:    $\hat A_{\alpha\beta\nu}=\hat {\bf n}^*_\beta \cdot 
731:       \partial_\nu\hat {\bf n}_\alpha$ \\
732:  $U(M)$ bundle & $U(M)$ bundle \\
733:  \begin{minipage}{5.5cm}
734:  Gauge group $U(M)$ arising from different choices of bases
735:  \end{minipage}
736:   & 
737:  \begin{minipage}{8.5cm}
738:  Gauge group $U(M)$ arising from redistribution of the scattering particles
739:  among different channels 
740:  \end{minipage}
741:  \\
742:  External parameters $X=(X^1, \cdots, X^p)$ 
743:   & External parameters $X=(X^1, \cdots, X^p)$  \\
744:   \hline \hline 
745: \end{tabular}
746: \end{center}
747: \caption{Comparison of Berry's phase and the quantum scattering (pumping)
748: geometric phase.}
749: %\end{ruledtabular}
750: \label{table}
751: \end{table}
752: %%%%%%%%%%%%%%%%%%%%%TABLE%%%%%%%%%%%%%%%%%%%%%%%%%%
753: 
754: %%%%%%%%%%%%%%%%%%%%%% FIGURES %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
755: 
756: %%%%%%%%%%%%%%%%%%%%%%%%% FIG1 %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
757: \begin{figure}
758:  \vspace*{10.0cm}
759:  \special{psfile=Fig1.ps vscale=55 hscale=55 hoffset=-35 voffset=-80}
760: %\includegraphics[height=10cm, width=8.3cm]{fig1.jpg}
761: %\includegraphics[height=12cm, width=9cm]{fig1.jpg}
762: \caption{The dependence of the spin pumping current 
763:         $I_s$ (times $\tau$, the period of cyclic variation)  
764:           on system parameters
765:           for two leads connected to a single magnetic spin whose direction
766:           is slowly varied around the path shown on the right.
767:           {\sf A.} Left: Schematic of the magnetic spin coupled to left (L)
768:              and right (R) leads.
769:          The magnetic spin ${\bf S}$ precesses around the direction
770:          of the magnetic field ${\bf B}$.
771:              Right: An equivalent 
772:              scanning tunneling microscope experimental setup.
773:          The pumping cycles on the parameter $(\theta,\phi)$-sphere 
774:          are, respectively,
775:          taken to be 
776:          $C_1 (\theta_1,\phi_1,\theta_2,\phi_2 ) = (\pi/8,0,\pi/2,\pi)$
777:          for {\sf B}
778:          and
779:          $C_2 (\theta_1,\phi_1,\theta_2,\phi_2 ) = (\pi/8,0,7\pi/8,\pi)$
780:          for {\sf C}.
781:          }
782: \label{fig1}
783: \end{figure}
784: %%%%%%%%%%%%%%%%%%%%%%Fig. 1%%%%%%%%%%%%%%%%%%%%%%%%%%
785: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
786: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
787: \end{document}
788: