1: %\documentstyle[aps,prbbib,twocolumn,epsf]{revtex}
2:
3:
4: \documentstyle[amssymb,preprint,aps]{revtex}
5: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
6: %TCIDATA{Created=Sat Jun 24 16:24:27 2000}
7: %TCIDATA{LastRevised=Fri Jun 21 17:04:23 2002}
8: %TCIDATA{Language=American English}
9:
10: \begin{document}
11: \draft
12: \title{Theory of Nonequilibrium Coherent Transport through \\
13: an Interacting Mesoscopic Region Weakly Coupled to Electrodes}
14: \author{{\ Yu Zhu and }Tsung-han Lin$^{*}$}
15: \address{{\it State Key Laboratory for Mesoscopic Physics and }\\
16: {\it Department of Physics, Peking University,}{\small \ }{\it Beijing}\\
17: 100871, China}
18: \author{{\ Qing-feng Sun}}
19: \address{{\it Center for the Physics of Materials and}\\
20: {\it Department of Physics, McGill University, Montreal,}\\
21: {\it PQ, Canada H3A 2T8}}
22: \date{}
23: \maketitle
24:
25: \begin{abstract}
26: We develop a theory for the nonequilibrium coherent transport through a
27: mesoscopic region, based on the nonequilibrium Green function technique. The
28: theory requires the weak coupling between the central mesoscopic region and
29: the multiple electrodes connected to it, but allows arbitrary hopping and
30: interaction in the central region. An equation determining the
31: nonequilibrium distribution in the central interacting region is derived and
32: plays an important role in the theory. The theory is applied to two special
33: cases for demonstrations, revealing the novel effects associated with the
34: combination of phase coherence, Coulomb interaction, and nonequilibrium
35: distribution.
36: \end{abstract}
37:
38: %\vskip 0.4in
39:
40: PACS numbers: 73.63.Kv, 85.35.Ds, 73.23.Hk, 73.40.Gk.
41:
42: \baselineskip 20pt %\baselineskip 12pt
43: \newpage
44:
45: In the realm of mesoscopic transport, the basic rules of the traditional
46: electronics breakdown. When the size of device is smaller than the electron
47: mean free path in the material, electrons will behave like a wave rather
48: than particles. Meanwhile, with the reduced size of device, the Coulomb
49: interaction among electrons becomes important. An additional electron has to
50: overcome the repulsion from other electrons in the device before entering
51: it. Moreover, due to lack of inelastic collisions, the thermal distribution
52: in the device is no longer a equilibrium one when electrodes are biased with
53: finite voltages. Therefore, a theory containing the above three
54: ``ingredients''---phase coherence, Coulomb interaction, and
55: nonequilibrium---is of particular interest in the mesoscopic transport.
56:
57: Using nonequilibrium Keldysh formalism, Meir $et\;al.$ derived a formula, in
58: which the current flowing out of an electrode is expressed in term of the
59: Green functions of the central region \cite{current}. The remaining task is
60: to find out these Green functions. Unfortunately, there seems no standard
61: method to derive them, although various approximation schemes were used for
62: special problem in specific system, e.g., large-N expansion \cite{Nexp},
63: truncation for equation of motion \cite{CBO}, introducing an interpolative
64: self-energy \cite{Yeyati}, Ng's ansatz for lesser self-energy \cite{Ng},
65: etc. In fact, too much physics is hidden in the general Hamiltonian, and it
66: is hopeless to invent a theory covering everything. If we restrict ourselves
67: to the weak coupling case, the complex Kondo physics will be ruled out. In
68: these circumstances, the Green functions in the ``atomic limit'' should be a
69: good starting point for the construction of the full Green functions, and we
70: shall show that a general solution to the problem is possible. We note that
71: similar idea has been addressed in the linear response theory by several
72: authors \cite{dia1,dia2}, but hardly investigated in the nonlinear regime
73: \cite{Stafford}.
74:
75: The aim of this paper is to present a scheme for the calculation of the
76: Green functions, and hence establish a theory of nonequilibrium coherent
77: transport through an interacting mesoscopic region. As a price, the coupling
78: between the central mesoscopic region and the electrodes is required to be
79: relatively small. Electron transport through the mesoscopic region is viewed
80: as a summation over various coherent processes via many-body quantum states,
81: weighted by nonequilibrium thermal probabilities. The many-body quantum
82: states in the central region can be found by exact diagonalization, while
83: the nonequilibrium distribution can be determined by an equation derived in
84: the text .
85:
86: The mesoscopic system under consideration is modelled by the Hamiltonian
87: (hereafter $e=\hbar =1$), $H=H_{cent}+\sum_\beta H_\beta +H_T$, where $%
88: H_{cent}=H_{cent}(\{c_i\},\{c_i^{\dagger }\})$ is a general Hamiltonian for
89: the central region with $M$-sites (spin index has been absorbed into the
90: site index $i$), $H_\beta =\sum_\beta (\epsilon _k-V_\beta )a_{\beta
91: k}^{\dagger }a_{\beta k}$ is the Hamiltonian for the $\beta $th electrodes,
92: and $H_T=\sum_{\beta ki}\left[ v_{\beta i}a_{\beta k}^{\dagger
93: }c_i+H.c.\right] $ is the tunnel coupling between them. This Hamiltonian is
94: applicable to a large variety of mesoscopic systems, such as molecular
95: devices, tunneling coupled carbon nanotubes, quantum dot arrays,
96: Aharonov-Bohm rings embedded with quantum dots, etc.
97:
98: Define the Green function of the central region and the self-energy arise
99: from the coupling with $\beta $th electrodes as ${\bf G}%
100: _{ij}^{r,a,<}(t_1,t_2)\equiv \langle \langle c_i(t_1)|c_j^{\dagger
101: }(t_2)\rangle \rangle ^{r,a,<}$ and ${\bf \Sigma }_{\beta
102: ,ij}^{r,a,<}(t_1,t_2)\equiv \sum_kv_{\beta i}^{*}\langle \langle a_{\beta
103: k}(t_1)|a_{\beta k}^{\dagger }(t_2)\rangle \rangle _0^{r,a,<}v_{\beta j}$,
104: where the superscript $r,a,<$ denote for the retarded, advanced, lesser
105: Green function and self-energy, respectively; the subscript $0$ denotes for
106: the Green function in the decoupling limit. Following Meir $et\;al.$, the
107: current flowing out of the $\beta $th electrodes can be expressed in the
108: compact form $I_\beta =\int \frac{d\omega }{2\pi }Tr\left\{ 2%
109: %TCIMACRO{\func{Re}}
110: %BeginExpansion
111: \mathop{\rm Re}%
112: %EndExpansion
113: \left[ {\bf G}(\omega ){\bf \Sigma }_\beta (\omega )\right] ^{<}\right\} $,
114: where $\left[ AB\right] ^{<}\equiv A^rB^{<}+A^{<}B^a$, ${\bf G}%
115: ^{r,a,<}(\omega )$ and ${\bf \Sigma }^{r,a,<}(\omega )$ are the Fourier
116: transformed Green function and self-energy. In the wide bandwidth limit, the
117: self-energy can be evaluated as ${\bf \Sigma }_\beta ^r(\omega )=-\frac{%
118: \text{i}}2{\bf \Gamma }_\beta $ and ${\bf \Sigma }_\beta ^{<}(\omega )=$i$%
119: f_\beta (\omega ){\bf \Gamma }_\beta $, where ${\bf \Gamma }_{\beta
120: ,ij}\equiv 2\pi D_\beta v_{\beta i}^{*}v_{\beta j}$ with $D_\beta $ being
121: the density of states at the Fermi surface of the $\beta $th electrode, $%
122: f_\beta (\omega )\equiv f(\omega -V_\beta )$ with $f(\omega )$ being the
123: Fermi distribution function. We assume that the coupling between the central
124: region and the electrodes is relatively small, i.e., $\Gamma _{\beta ,ij}\ll
125: k_BT$. Under this assumption, the Green function of the central region can
126: {\it approximately} be written in a Dyson equation form ${\bf G=\tilde{g}+%
127: \tilde{g}\Sigma }_0{\bf G}$,{\bf \ }with ${\bf \tilde{g}\equiv }%
128: \lim_{H_T\rightarrow 0}{\bf G}$ being the Green function in the decoupling
129: limit, ${\bf \Sigma }_0\equiv \sum_\beta ({\bf \Sigma }_\beta )$ being the
130: self-energy contributed by the coupling with electrodes. (This approximation
131: is amount to a proper truncation of equation of motion.) Accordingly, the
132: approximate ``Dyson equation'' for ${\bf G}^r$ and the ``Keldysh equation''
133: for ${\bf G}^{<}$ are
134: \begin{eqnarray}
135: {\bf G}^r &=&{\bf \tilde{g}}^r{\bf +\tilde{g}}^r{\bf \Sigma }_0^r{\bf G}^r%
136: {\bf \;,} \\
137: {\bf G}^{<} &=&{\bf G}^r{\bf \Sigma }_0^{<}{\bf G}^a{\bf \;.}
138: \end{eqnarray}
139: Consequently, the current formula can be rewritten in a Landauer type as if
140: in the noninteracting case \cite{current},
141: \begin{equation}
142: I_\beta =\sum_{\alpha \neq \beta }\int \frac{d\omega }{2\pi }T_{\alpha \beta
143: }(\omega )\left[ f_\beta (\omega )-f_\alpha (\omega )\right] \;,
144: \end{equation}
145: with $T_{\alpha \beta }(\omega )\equiv Tr{\bf G}^r{\bf \Gamma }_\alpha {\bf G%
146: }^a{\bf \Gamma }_\beta $.
147:
148: The remaining task is to calculate the retarded Green function in the
149: decoupling limit. It is straightforward to exactly diagonalize $%
150: H_{cent}(\{c_i\},\{c_i^{\dagger }\})$ in the particle occupation bases $%
151: \{(c_M^{\dagger })^{N_M}\cdots (c_2^{\dagger })^{N_2}(c_1^{\dagger
152: })^{N_1}\left| 0\right\rangle $ where $N_i=0$ or $1$, and obtain the 2$^M$
153: eigenstates $\{E_n,\left| n\right\rangle \}$. Once again, under the weak
154: coupling assumption, the density matrix operator of the central region is
155: supposed to have the {\it diagonal} form ${\bf \rho }_{cent}=\sum_nP_n\left|
156: n\right\rangle \left\langle n\right| $, with the constraint $\sum_nP_n=1$.
157: Here, the central region is regarded as ``system'', while the electrodes are
158: ``environment'' in local equilibrium. Given $\{P_n\}$, the decoupled Green
159: functions can be expressed as
160: \begin{eqnarray}
161: \left\langle \left\langle A|B\right\rangle \right\rangle _0^r &=&\sum_{nm}%
162: \frac{P_n+P_m}{\omega -(E_n-E_m)+\text{i}0^{+}}\left\langle
163: m|A|n\right\rangle \left\langle n|B|m\right\rangle \;\;, \\
164: \left\langle \left\langle A|B\right\rangle \right\rangle _0^{<}
165: &=&\sum_{nm}2\pi \text{i\ }P_n\delta \left[ \omega -(E_n-E_m)\right]
166: \left\langle m|A|n\right\rangle \left\langle n|B|m\right\rangle \;\;,
167: \end{eqnarray}
168: where $A$ and $B$ are operators composed of $\{c_i\}$ and $\{c_i^{\dagger
169: }\} $. So the determination of the nonequilibrium distribution $\{P_n\}$
170: lies in the heart of the theory. In the linear response regime, i.e., $%
171: \left| V_\beta -V_{\beta ^{\prime }}\right| \ll k_BT$ and hence $V_\beta
172: \thickapprox V_0$, the central region is in a thermal equilibrium, and the
173: distribution can be written as $P_n=\frac 1Ze^{-(E_n-N_nV_0)/k_BT}$. For the
174: case of nonequilibrium, however, $\{P_n\}$ is determined by the coupling to
175: electrodes with different chemical potentials, and in principle needs a
176: self-consistent calculation. Our strategy is to choose a proper set of
177: observables $\{O\}$ and establish the equations of $\{P_n\}$ by the {\it %
178: stationary} condition
179: \begin{equation}
180: \left\langle \partial _tO\right\rangle =\left\langle \text{i}\left[
181: H,O\right] \right\rangle =0\;\;.
182: \end{equation}
183: We point out that the 2$^M$ conservables $\{O_l\equiv \left| l\right\rangle
184: \left\langle l\right| \}$ ($\left| l\right\rangle $ is the eigenstate of $%
185: H_{cent}$) are ideal candidates for the task. Notice that
186: \begin{equation}
187: \left\langle \text{i}\left[ H,O_l\right] \right\rangle =2%
188: %TCIMACRO{\func{Re} }
189: %BeginExpansion
190: \mathop{\rm Re}%
191: %EndExpansion
192: \int \frac{d\omega }{2\pi }\sum_{\beta k}\sum_{ij}\left[ \langle \langle
193: [c_i,O_l]|c_j^{\dagger }\rangle \rangle v_{\beta j}^{*}\langle \langle
194: a_{\beta k}|a_{\beta k}^{\dagger }\rangle \rangle _0v_{\beta i}\right]
195: ^{<}\;\;,
196: \end{equation}
197: and make the approximation $\langle \langle [c_i,O_l]|c_j^{\dagger }\rangle
198: \rangle \thickapprox \langle \langle [c_i,O_l]|c_j^{\dagger }\rangle \rangle
199: _0$ under the weak coupling approximation, one can derive a set of equations
200: of $\{P_n\}$ as
201: \begin{equation}
202: \sum_\beta \sum_{nm}\left[ P_mf_\beta (E_n-E_m)-P_n\bar{f}_\beta
203: (E_n-E_m)\right] \tilde{\Gamma}_{nm}^\beta Q_{nm}^l=0\;\;\;\;(l=1,2\cdots
204: 2^M)\;,
205: \end{equation}
206: where $n$ and $m$ run over all the eigenstates of $H_{cent}$, $\tilde{\Gamma}%
207: _{nm}^\beta \equiv \sum_{ij}\left\langle m|c_i|n\right\rangle \langle
208: n|c_j^{\dagger }|m\rangle \Gamma _{\beta ,ij}$, $Q_{nm}^l\equiv \delta
209: _{nl}-\delta _{ml}$, and $\bar{f}_\beta (\omega )\equiv 1-f_\beta (\omega )$%
210: . Because $\sum_l\left| l\right\rangle \left\langle l\right| =1$, the 2$^M$
211: conservables can produce 2$^M$-1 independent equations, and the constraint $%
212: \sum_nP_n=1$ should be supplemented for completeness. Eq.(8) is the central
213: result of this work, which determines the nonequilibrium distribution in an
214: interacting system. With $\{P_n\}$ solved from the set of equations, one can
215: calculate both nonequilibrium tunneling current and various quantities of
216: the central region.
217:
218: To sum up, Eq.(1), (3), (4), and (8) consist of the frame for the
219: calculation of the nonequilibrium coherent transport through an interacting
220: mesoscopic region, requiring weak coupling between the central region and
221: the electrodes, but allowing arbitrary interaction and hopping in the
222: central region. Below we shall apply the theory to two special cases for
223: demonstrations.
224:
225: (1) {\it a single quantum dot with multiple levels. }Consider a single
226: quantum dot (QD) connected to electrodes with finite bias voltages, which
227: can be modelled by the Hamiltonian
228: \begin{equation}
229: H_{cent}=\sum_iE_in_i+U\sum_{i<j}n_in_j\;\;,
230: \end{equation}
231: where $n_i\equiv c_i^{\dagger }c_i$ is the particle number operator. For
232: convenience, the particle occupation bases are numbered as $%
233: F=\sum_{i=1}^MN_i^F\cdot 2^{i-1}$ and $\left| F\right\rangle \equiv
234: (c_M^{\dagger })^{N_M^F}\cdots (c_2^{\dagger })^{N_2^F}(c_1^{\dagger
235: })^{N_1^F}\left| 0\right\rangle $. Notice that $H_{cent}$ is already
236: diagonalized in the $\{\left| F\right\rangle \}$ bases, and the eigenenergy
237: of $\left| F\right\rangle $ is $E_{\left| F\right\rangle
238: }=\sum_iE_iN_i^F+U\sum_{i<j}N_i^FN_j^F$. The conservable $\left|
239: l\right\rangle \left\langle l\right| $ can be written explicitly in the form
240: of $\{c_i\}$ and $\{c_i^{\dagger }\}$ as $m_M^l\cdots m_2^lm_1^l$ where $%
241: m_i^l=n_i$ for $N_i^l=1$ and $m_i^l=1-n_i$ for $N_i^l=0$. The equations of $%
242: \{P_F\}$ are simplified to
243: \begin{equation}
244: \sum_{i=1}^M(-1)^{N_i^l}\Gamma _i\left[ h_i(E_{\left| F_1\right\rangle
245: }-E_{\left| F_0\right\rangle })P_{F_0}-\bar{h}_i(E_{\left| F_1\right\rangle
246: }-E_{\left| F_0\right\rangle })P_{F_1}\right] =0\;\;(l=1,2\cdots 2^M)\;,
247: \end{equation}
248: where $\Gamma _i\equiv \sum_\beta \Gamma _{\beta i}$, $h_i(\omega )\equiv
249: \sum_\beta \frac{\Gamma _{\beta i}}{\Gamma _i}f_\beta (\omega )$, $\Gamma
250: _{\beta i}\equiv \Gamma _{\beta ,ii}=2\pi D_\beta |v_{\beta i}|^2$, $%
251: F_1=l-N_i^l\cdot 2^{i-1}+2^{i-1}$, $F_0=l-N_i^l\cdot 2^{i-1}$, and $\bar{h}%
252: _i(\omega )\equiv 1-h_i(\omega )$. The retarded Green function ${\bf \tilde{g%
253: }}^r(\omega )$ is obtained as
254: \begin{equation}
255: \langle \langle c_i|c_j^{\dagger }\rangle \rangle _0^r=\delta _{ij}\sum_F%
256: \frac{P_F}{\omega -\tilde{E}_i^F+\text{i}0^{+}}\;\;,
257: \end{equation}
258: with $\tilde{E}_i^F\equiv E_i+U\sum_{j\neq i}N_j^F$ being the renormalized
259: resonances. Specially, for $M=2$, using $\langle n_1\rangle =P_{01}+P_{11}$,
260: $\langle n_2\rangle =P_{10}+P_{11}$, $\langle n_1n_2\rangle =P_{11}$,
261: Eq.(19) and Eq.(20) are equivalent to
262: \begin{eqnarray}
263: \langle n_i\rangle &=&(1-\langle n_{\bar{i}}\rangle )h_i(E_i)+\langle n_{%
264: \bar{i}}\rangle h_i(E_i+U)\;\;, \\
265: \langle n_1n_2\rangle &=&\frac{\Gamma _1}{\Gamma _1+\Gamma _2}%
266: h_1(E_1+U)\langle n_2\rangle +\frac{\Gamma _2}{\Gamma _1+\Gamma _2}%
267: h_2(E_2+U)\langle n_1\rangle \;\;, \\
268: \langle \langle c_i|c_i^{\dagger }\rangle \rangle _0^r &=&\frac{1-\langle n_{%
269: \bar{i}}\rangle }{\omega -E_i+\text{i}0^{+}}\;+\frac{\langle n_{\bar{i}%
270: }\rangle }{\omega -E_i-U+\text{i}0^{+}}\;,
271: \end{eqnarray}
272: with $\bar{i}=3-i$ for $i=1$ or $2$. Thus, our theory reproduces the correct
273: results for the occupation number $\langle n_i\rangle $ and the retarded
274: Green function $\langle \langle c_i|c_i^{\dagger }\rangle \rangle _0^r$ in
275: the limit of $\Gamma _i\rightarrow 0$ \cite{CBO}, and derive the correlator $%
276: \langle n_1n_2\rangle $ which is otherwise difficult to obtain. Fig.1 shows
277: the equilibrium and nonequilibrium distributions for the quantum dot
278: containing two interacting levels connected with two electrodes. One can see
279: in the plot: (a) The complete Coulomb blockade in equilibrium is partially
280: removed in nonequilibrium, i.e., the blockaded state can be occupied in some
281: ``windows'' of the gate voltage. (b) The correlator $\langle n_1n_2\rangle $
282: is obviously unequal to $\langle n_1\rangle \langle n_2\rangle $ in
283: nonequilibrium, although approximately correct in equilibrium for
284: nondegenerate levels. (c) The total occupation number has fractional steps
285: in nonequilibrium in contrast to integer steps in equilibrium, and the
286: fluctuation of the total number is reminiscent of the shape of tunneling
287: current.
288:
289: Next, we insert the QD to one arm of a Aharonov-Bohm (AB) interferometer
290: \cite{Yacobi}. The other arm (reference arm) is modelled by a quantum point
291: contact, which can be described by adding the term $\sum_k(We^{\text{i}\phi
292: }a_{Lk}^{\dagger }a_{Rk}+H.c.)$ to $H_T$ \cite{Fano}, with $\phi $ being the
293: AB phase induced by magnetic flux. Both the current formula and the equation
294: of $\{P_n\}$ should be modified to include the ``direct'' coupling between
295: electrodes, the details will be presented elsewhere. The background
296: conductance of the reference arm (measured when QD is decoupled from both
297: electrodes) is $G_0=\frac{e^2}h\frac{4x}{(1+x)^2}$ with $x\equiv \pi
298: ^2W^2D_LD_R$. The effective conductance of QD is formally defined as $%
299: G_{dot}=I(V_b,V_g)/V_b-G_0$. Fig.2 shows the curves of $G_{dot}$ in both
300: linear and nonlinear transport regimes. Three features are noteworthy: (a)
301: In linear transport, the conductance is contributed only by the ground
302: state, while in nonlinear transport, the conductance is contributed by both
303: ground and excited states, recognized by the substeps in the conductance
304: plateaus and valleys, which is in agreement with the recent experiment \cite
305: {neqb}. (b) With the increase of the background conductance, $G_{dot}$ vs $%
306: V_g$ curves exhibit Fano pattern, evolving from a peak (plateau) to a pair
307: of peak and dip (plateau and valley), and finally to a dip (valley), which
308: is originated from the constructive and destructive interference between a
309: resonance and the uniform background. (c) The phase analysis shows that the
310: dependence of $G_{dot}$ on $\phi $ across a resonance is quite different
311: between linear and nonlinear transport. In linear transport, the dependence
312: changes from $\cos \phi $ to $\cos (\phi +\pi )$ abruptly across a
313: resonance, and $\cos (2\phi )$ component only appears on the resonance. In
314: nonlinear transport, $\cos (2\phi )$ component is always accompanied with
315: the $\cos \phi $ component, and the crossover from $\cos \phi $ to $\cos
316: (\phi +\pi )$ occurs continuously. The ``nonequilibrium-Fano'' effect
317: discussed here and the ``Kondo-Fano'' effect in \cite{Fano} seem to share
318: some similarities, although the mechanisms are basically different.
319:
320: (2) {\it coupled double quantum dots.} Consider two quantum dots coupled in
321: series with left and right electrodes (N-QD=QD-N), each dot is capacitively
322: coupled to a gate so that the energy level of the dot is tunable \cite{CQD}.
323: The coupled double quantum dots can be modelled by a 4-site Hamiltonian
324: \begin{eqnarray}
325: H_{cent} &=&\sum_{i=1,2}\sum_\sigma E_{i\sigma }n_{i\sigma }+t\sum_\sigma
326: (c_{1\sigma }^{\dagger }c_{2\sigma }+c_{2\sigma }^{\dagger }c_{1\sigma })
327: \nonumber \\
328: &&+\sum_{i=1,2}U_in_{i_{\uparrow }}n_{i_{\downarrow }}+U_{12}(n_{1\uparrow
329: }+n_{1\downarrow })(n_{2\uparrow }+n_{2\downarrow })\;\;.
330: \end{eqnarray}
331: We first diagonalize $H_{cent}$ in the 2$^4$ particle occupation bases. Due
332: to the particle number conservation and spin conservation, the 16
333: dimensional spaces can be divided into several subspaces $%
334: 16=1+(2+2)+(1+1+4)+(2+2)+1$, in which eigenstates are readily solved. In
335: principle, one can find out 2$^4$ conservables written in $\{c_{i\sigma }\}$
336: and $\{c_{i\sigma }^{\dagger }\}$ as done in (1), although it is uneasy and
337: unnecessary. We only need to calculate the effective coupling strength $%
338: \tilde{\Gamma}_{nm}^\beta $ and put them into Eq.(8). With $\{P_n\}$ solved
339: from the 2$^4$ linear equations, ${\bf \tilde{g}}^r$, hence ${\bf G}^r$ and $%
340: {\bf G}^{<}$ are available. Finally, the current formula in the N-QD=QD-N
341: system can be simplified as
342: \begin{equation}
343: I\equiv I_L=-I_R=\sum_\sigma \int \frac{d\omega }{2\pi }\Gamma _L\Gamma
344: _R\left| \langle \langle c_{1\sigma }|c_{2\sigma }^{\dagger }\rangle \rangle
345: ^r\right| ^2\left[ f_L(\omega )-f_R(\omega )\right] \;\;.
346: \end{equation}
347: We present the numerical results of the tunneling current $I$ vs the
348: resonant levels $(E_1,E_2)$ in Fig.3. The bias voltage $V_b$ ($V_L=V_b/2$, $%
349: V_R=-V_b/2$) is fixed as $0.001$, $0.3$, and $0.6$ for the graphs from top
350: to bottom. The graph of $V_b=0.001$ is corresponding to the linear response
351: regime, in which the thermal distribution in the central region is nearly
352: equilibrium. Due to the intradot and interdot Coulomb interactions, the
353: conductance peaks are arranged to a hexagon pattern, consistent with both
354: experiment and the semiclassical Coulomb blockade model. The graph of $%
355: V_b=0.6$ is corresponding to the strong nonequilibrium case, in which
356: thermal distribution of the central region is determined by two reservoirs
357: with different chemical potentials. The effect of finite bias voltage on the
358: tunneling current is two folded: (a) Pull each conductance peak along the
359: direction of $E_1=E_2$, and ultimately emerge them into a ridge. It is easy
360: to understand the pulling effect by thinking of the noninteracting case
361: (i.e., $U_1=U_2=U_{12}=0$), where the conductance through N-QD=QD-N is
362: allowed only when $V_L>E_1\thickapprox E_2>V_R$. (b) Modify the hexagon
363: pattern significantly, and break the symmetry with respect to $E_1=E_2$.
364: This can be attributed to the nonequilibrium occupation of the coupled
365: quantum dots. To proceed, let us cut off the interdot hopping for a moment
366: (i.e., $t=0$). Then the electron filling to the left (right) dot is only
367: related to the chemical potential of the left (right) electrode. The
368: occupation configuration $\{\langle n_1\rangle ,\langle n_2\rangle \}$ vs
369: the energy levels $(E_1,E_2)$ has the same shape of hexagon boundary as in
370: the equilibrium case, except for a displacement of $(\frac{V_b}2,-\frac{V_b}2%
371: )$. Therefore, the symmetry with respect to $E_1=E_2$ is broken when $%
372: V_b\neq 0$. Turning on the interdot hopping will make the problem much more
373: complicated, either $\langle n_1\rangle $ or $\langle n_2\rangle $ are not
374: good quantum number, energy levels of each dot are hybridized into
375: ``molecular orbits'', and the occupation of the ``molecule'' is affected by
376: both of the electrodes. Simple interpretation for this situation is beyond
377: our intelligence.
378:
379: In conclusion, we have presented a theory dealing with nonequilibrium
380: coherent transport through an arbitrary mesoscopic region, possibly
381: containing strong Coulomb interactions. The only restriction of the theory
382: is that the coupling between the central region and electrodes should be
383: sufficient weak so that the central region may be regarded as a single
384: quantum system. The key innovation of the theory is Eq.(8) which determines
385: the nonequilibrium distribution in an interacting system. The general theory
386: is applied to two special cases, a single quantum dot with multiple
387: interacting levels and coupled double quantum dots, and novel behaviors are
388: found in the nonlinear coherent transport.
389:
390: We would like to thank W. Li and Y. F. Yang for the valuable discussions.
391: This project was supported by NSFC\ under Grant No. 10074001 and by the
392: State Key Laboratory for Mesoscopic Physics in Peking University.
393:
394: \smallskip $^{*}$ To whom correspondence should be addressed.
395:
396: %\section* {REFERENCES}
397:
398: \begin{references}
399: \bibitem{current} Y. Meir and N. S. Wingreen, Phys. Rev. Lett. {\bf 68,}
400: 2512 (1992).
401:
402: \bibitem{Nexp} N. E. Bickers, Rev. Mod. Phys. {\bf 59,} 845 (1987).
403:
404: \bibitem{CBO} Y. Meir, N. S. Wingreen, and P. A. Lee, Phys. Rev. Lett. {\bf %
405: 66,} 3048 (1991).
406:
407: \bibitem{Yeyati} A. L. Yeyati, F. Flores, and A. Martin-Rodero, Phys. Rev.
408: Lett. {\bf 83,} 600 (1999).
409:
410: \bibitem{Ng} T. -K. Ng, Phys. Rev. Lett. {\bf 76,} 487 (1996).]
411:
412: \bibitem{dia1} J. M. Kinaret $et\;al.$, Phys. Rev. B {\bf 45, }9489 (1992).
413:
414: \bibitem{dia2} G. Chen, G. Klimeck, and S. Datta Phys. Rev. B {\bf 50,}
415: 8035 (1994).
416:
417: \bibitem{Stafford} We are only aware of the paper by C. A. Stafford, Phys.
418: Rev. Lett. {\bf 77,} 2770 (1996), which is devoted to the identical problem.
419: However, his theory requires the ground state to be nondegenerate, the
420: temperature and the bias voltages are small compared to the excitation
421: energy, while our theory release these restrictions, and is a {\it real}
422: nonequilibrium transport theory.
423:
424: \bibitem{Yacobi} A. Yacoby $et\;al.$, Phys. Rev. Lett. {\bf 74,} 4047
425: (1995).
426:
427: \bibitem{Fano} W. Hofstetter, J. K\"{o}nig, and H. Schoeller, Phys. Rev.
428: Lett. {\bf 87,} 156803 (2001).
429:
430: \bibitem{neqb} M. M. Deshmukh $et\;al.$, Phys. Rev. B {\bf 65, }073301
431: (2002).
432:
433: \bibitem{CQD} F. R. Waugh $et\;al.$, Phys. Rev. Lett. {\bf 75, }705 (1995);
434: C. Livermore $et\;al.$, Science {\bf 274, }1332 (1996).
435: \end{references}
436:
437: \newpage
438:
439: \section*{Figure Captions}
440:
441: \begin{itemize}
442: \item[{\bf Fig. 1}] Equilibrium and nonequilibrium distributions in a QD
443: containing two energy levels indexed by $i=1,2$, connected with two
444: electrodes indexed by $\beta =L,R$. (a), (b) and (c) show the curves of $%
445: \langle n_1\rangle $ (solid) and $\langle n_2\rangle $ (dotted), $\langle
446: n_1n_2\rangle $ (solid) and $\langle n_1\rangle \langle n_2\rangle $
447: (dotted), $\langle n\rangle \equiv \langle n_1\rangle +\langle n_2\rangle $
448: (solid) and $\langle \delta n\rangle \equiv 4\left[ \langle n^2\rangle
449: -\langle n\rangle ^2\right] ^{1/2}$ (dotted) vs the gate voltage $V_g$,
450: respectively. The bias voltage $V_b\equiv V_L-V_R$ is fixed at $V_b=0$ / $%
451: V_b=0.6$ for the left / right panel. Other parameters are: $U=1$, $k_BT=0.02$%
452: , $\Gamma _{1L}=\Gamma _{1R}=\Gamma _{2L}=\Gamma _{2R}=0.001$, $E_1=V_g-0.05$%
453: , $E_2=V_g+0.05$.
454:
455: \item[{\bf Fig. 2}] Linear and nonlinear transport through an AB
456: interferometer embedded with a QD (schematically shown in the inset). (a)
457: and (b) show the curves of $G_{dot}\equiv I(V_b,V_g)/V_b-G_0$ vs the gate
458: voltage $V_g$ for $\phi =0$, with the bias voltage $V_b$ fixed at $0^{+}$
459: and $0.6$, respectively. The solid, dash, and dotted curves correspond to
460: the background conductance of the reference arm $G_0=0$, $0.5\frac{e^2}h$, $%
461: \frac{e^2}h$, respectively. The QD contains three interacting levels, with
462: level spacing $\Delta E=0.1$, interacting constant $U=1$, and $\Gamma
463: _L=\Gamma _R=0.001\ll k_BT=0.02$. (c) and (d) analyze the phase dependence
464: of the $G_{dot}$ at the points marked in (a) and (b). Only the range of $%
465: 0<\phi <\pi $ is shown since $G_{dot}(\phi )=G_{dot}(-\phi )$.
466:
467: \item[{\bf Fig. 3}] Surface graphs of tunneling current $I$ vs the resonant
468: levels $(E_1,E_2)$ in the N-QD=QD-N system, with sideview on the left and
469: topview on the right. The bias voltage is fixed at $V_b=0.001$,$\;0.3$,$\;$%
470: and $0.6$ for the graphs from top to bottom. Other parameters are: $U_1=U_2=1
471: $, $U_{12}=0.3$, $t=0.1$, $k_BT=0.05$, $\Gamma _L=\Gamma _R=0.01$. (In the
472: e-print version, Fig. 3 is separated into Fig. 3a, Fig.3b and Fig. 3c
473: corresponding to $V_b=0.001$,$\;0.3\;$and $0.6$ respectively, to reduce the
474: size of figures.)
475: \end{itemize}
476:
477: \end{document}
478: