1: %% LyX 1.2 created this file.  For more info, see http://www.lyx.org/.

2: %% Do not edit unless you really know what you are doing.

3: \documentclass[twocolumn,english,aps,prb,showpacs]{revtex4}

4: \usepackage[T1]{fontenc}

5: \usepackage[latin1]{inputenc}

6: \usepackage{graphicx}

7: \usepackage{amssymb}

8: 

9: \makeatletter

10: 

11: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%% LyX specific LaTeX commands.

12: \providecommand{\LyX}{L\kern-.1667em\lower.25em\hbox{Y}\kern-.125emX\@}

13: %% Bold symbol macro for standard LaTeX users

14: \newcommand{\boldsymbol}[1]{\mbox{\boldmath $#1$}}

15: 

16: 

17: \usepackage{babel}

18: \makeatother

19: \begin{document}

20: 

21: \newcommand{\kt}[1]{\left|#1\right\rangle }

22: 

23: 

24: 

25: \newcommand{\br}[1]{\left\langle #1\right|}

26: 

27: 

28: 

29: \newcommand{\ph}{\hat{\phi }}

30: 

31: 

32: 

33: \newcommand{\pd}{\hat{\psi }^{\dagger }}

34: 

35: \newcommand{\ps}{\hat{\psi }}

36: 

37: 

38: 

39: \title{Dephasing in sequential tunneling through a double-dot interferometer

40: {\small }\\

41: }

42: 

43: 

44: \author{Florian Marquardt and C. Bruder}

45: 

46: 

47: \affiliation{Departement Physik und Astronomie, Universität Basel, Klingelbergstr.

48: 82, CH-4056 Basel}

49: 

50: 

51: \date{March 19th, 2003}

52: 

53: 

54: \email{Florian.Marquardt@unibas.ch}

55: 

56: \begin{abstract}

57: We analyze dephasing in a model system where electrons tunnel sequentially

58: through a symmetric interference setup consisting of two single-level

59: quantum dots. Depending on the phase difference between the two tunneling

60: paths, this may result in perfect destructive interference. However,

61: if the dots are coupled to a bath, it may act as a which-way detector,

62: leading to partial suppression of the phase-coherence and the reappearance

63: of a finite tunneling current. In our approach, the tunneling is treated

64: in leading order whereas coupling to the bath is kept to all orders

65: (using $P(E)$ theory). We discuss the influence of different bath

66: spectra on the visibility of the interference pattern, including the

67: distinction between {}``mere renormalization effects'' and {}``true

68: dephasing''.

69: \end{abstract}

70: 

71: \pacs{73.23.Hk, 71.38.-k, 03.65.Yz}

72: 

73: \maketitle

74: 

75: \section{Introduction}

76: 

77: The destruction of quantum-mechanical phase coherence due to coupling

78: of a system to an irreversible bath is a subject important not only

79: because of its connection to fundamental issues (the quantum measurement

80: process and the quantum-classical transition) but also because of

81: its role in the suppression of phenomena resulting from quantum interference

82: effects, such as those studied in mesoscopic physics (including Aharonov-Bohm

83: interference, weak localization and universal conductance fluctuations).

84: Recently, the field of mesoscopic physics in particular has seen a

85: revival of interest in these questions, due to surprising experimental

86: findings\cite{mohantywebb} concerning a possible saturation of the

87: weak-localization dephasing rate at low temperatures, that have not

88: yet been explained convincingly. Apart from investigations dealing

89: directly with the problem of weak localization in a disordered system

90: of interacting electrons, several toy models have been analyzed\cite{cedraschi,buettrevb,GZ_PB,OurABring,NagaevBuettikerHO,Guinea,GSZ_modelsLowTDeph,ImryInexistenceZPDP}

91: to answer the question whether decoherence at zero temperature is

92: possible at all, contrary to the expectations based on perturbation

93: theory. One of the difficulties faced by models involving discrete

94: levels consists in the fact that destruction of phase coherence for

95: a superposition of excited states of finite excitation energy is perfectly

96: possible even at zero temperature, due to spontaneous emission of

97: energy into the bath. It is only in the zero-frequency limit of the

98: linear response in a system with a continuous spectrum (relevant for

99: weak-localization and other equilibrium transport experiments) that

100: perturbation theory suggests in general a vanishing dephasing rate,

101: because then the perturbation does not supply energy to the system,

102: such that at $T=0$ the system is not able to leave a trace in the

103: bath, which is considered to be the prerequisite for decoherence.

104: 

105: Some questions of interest concerning dephasing, especially in connection

106: with mesoscopic systems and low temperatures, are the following ones:

107: How reliable is the simple classical picture of a phase being randomized

108: by fluctuating external noise\cite{sai}? In particular, what is the

109: meaning of the zero-point fluctuations of the bath in this picture,

110: as opposed to the thermal fluctuations dominating at frequencies lower

111: than the temperature? When do the former lead to {}``mere renormalization

112: effects'' and how is it possible to distinguish these from {}``true''

113: dephasing? Under which circumstances is the suppression of off-diagonal

114: terms in the reduced system density matrix itself already a good indicator

115: of dephasing? How reliable are simple arguments based on Golden Rule

116: and energy conservation, related to the connection between dephasing

117: and the trace left in the bath by the particle ({}``which-way''

118: detection)? When does perturbation theory fail qualitatively, what

119: is the influence of non-Markoffian behaviour? How does the dephasing

120: rate depend on the energy supplied by an external perturbation (frequencies

121: excited in linear response, bias voltage applied in a transport measurement)?

122: What is the influence of the Pauli principle in a system of degenerate

123: fermions? How strong are the qualitative differences in behaviour

124: resulting from different bath spectra?

125: 

126: In this work, we will present a model that is able to give insights

127: into most of these questions. 

128: 

129: %

130: \begin{figure}

131: \begin{center}\includegraphics[  height=5cm]{geometry2new.eps}\end{center}

132: 

133: 

134: \caption{\label{DDdoubleslitsetup}The double-dot {}``double-slit'' setup,

135: with a fixed phase difference $\varphi $ between the two paths and

136: under the influence of a fluctuating environment.}

137: \end{figure}

138: 

139: 

140: Our model represents a kind of mesoscopic double-slit setup. It consists

141: of two single-level quantum dots which are tunnel-coupled to two leads,

142: with a possible phase difference between the two interfering paths

143: (see Fig. \ref{DDdoubleslitsetup}). Due to destructive interference

144: (at $\varphi =\pi $), the tunneling current may be suppressed completely,

145: provided the two dot-levels are degenerate and the setup is symmetric

146: in the two interfering paths. Coupling the dots to a bath may partly

147: destroy the phase coherence and re-enable the electrons to go through

148: the setup. For a symmetric setup, with equal coupling strength between

149: the bath and each of the two dots, mere renormalization effects will

150: not be able to lift the destructive interference in this way. Thus,

151: a finite tunneling current may be taken as a genuine sign of dephasing.

152: This criterion for dephasing has been employed before in a model of

153: dephasing due to spin-flip transitions in first-order tunneling transport

154: through one or two dots\cite{koeniggefen}, as well as for cotunneling

155: through an Aharonov-Bohm ring coupled to a fluctuating magnetic flux\cite{OurABring}. 

156: 

157: The influence of phonons on sequential tunneling through two quantum

158: dots \emph{in series} has been studied experimentally in Ref.~\onlinecite{KouwenhouwenDD}.

159: There, inelastic transitions induced by piezoelectric coupling to

160: acoustic phonons in GaAs have been essential for obtaining a finite

161: current through the two off-resonant dot levels. This kind of setup

162: has been analyzed theoretically in Refs.~\onlinecite{StoofNazarov,brune,ziegler,BrandesKramer,aguado,Debald,KeilSchoellerRTRG}.

163: On the other hand, we will be analyzing tunneling through two dots

164: placed \emph{in parallel}. Early theoretical investigations of this

165: problem (without a fluctuating environment) include Refs.~\onlinecite{akera,shahbazyan}.

166: Recently, a parallel-dot tunneling setup has been realized experimentally

167: in Ref.~\onlinecite{KotthausDD}, with an emphasis on spectroscopy of the

168: {}``molecular states'' of the double-dot system (with inter-dot

169: tunneling present). In our model of an interference setup, we choose

170: to describe a situation without tunneling between the dots (but with

171: Coulomb-repulsion). In addition, we want to concentrate on interference

172: effects in the orbital motion and therefore consider the case of spin-polarized

173: transport. This model - in the absence of a fluctuating environment

174: - has been investigated previously in Ref.~\onlinecite{BoeseDD}. Other

175: recent theoretical works concerning tunneling through dots in a parallel

176: geometry have mostly investigated spin and Kondo physics \cite{LossShenjaDD,mahnsoo,BoeseHofstetterSchoeller},

177: but also dephasing by spin-flip transitions\cite{koeniggefen}. Some

178: works have treated the influence of phonons in tunneling interference

179: structures\cite{HauleBonca,key-2}, but no systematic discussion of

180: dephasing and the visibility of the interference pattern has been

181: given. Some while ago, dephasing by \emph{nonequilibrium} current

182: noise has been investigated experimentally\cite{key-1} and theoretically\cite{aleinerwhichpath}

183: in a setup with a single quantum-dot placed into one arm of an Aharonov-Bohm

184: interferometer.

185: 

186: Our analysis of dephasing in sequential tunneling through a double-dot

187: will take into account the system-bath coupling exactly, while we

188: treat the tunnel-coupling only in leading order. The presence of the

189: Fermi sea in the leads introduces some aspects related to the Pauli

190: principle and to the behaviour of systems with a continuous spectrum

191: that cannot be analyzed in simpler models of dephasing in discrete

192: systems coupled to a bath.

193: 

194: The work is organized as follows: After setting up the model (Sec.

195: \ref{sec:The-model}), we will present a qualitative discussion of

196: its main features (\ref{sec:Qualitative-discussion}). In particular,

197: we will discuss the relation between entanglement, dephasing and renormalization

198: effects. Subsequently, we derive a general formula for the tunneling

199: decay rate of an electron that has been placed on the two dots in

200: a symmetric superposition of states (Sec. \ref{sec:Decay-rate-and}).

201: This is done by building on the concepts of the $P(E)$ theory of

202: tunneling in a dissipative environment \cite{ingoldhabil,SchoenPE}.

203: Following this, we will evaluate the dependence of the tunneling rate

204: on the bias voltage and the bath spectra (Sec. \ref{sec:Evaluation-for-different}).

205: We will interpret the results in terms of {}``renormalization effects''

206: and {}``true dephasing'' (Sec. \ref{sec:Discussion-of-the}). Building

207: on these sections, we will finally derive a master equation for the

208: case of weak tunnel coupling (Sec. \ref{DDseqTunnelingSection}),

209: which allows us to calculate the sequential tunneling current as a

210: function of bias voltage, temperature, and phase difference (Sec.

211: \ref{sec:Evaluation-of-the}).

212: 

213: The most important results derived in this work are the following:

214: Equation (\ref{DDGammabasic}) is the general expression for the phase-dependent

215: tunneling decay rate in presence of the fluctuating environment. It

216: forms the basic input for the master equation (Eqs. (\ref{DDrhoppeq})-(\ref{DDrhopmEq})),

217: that describes sequential tunneling through the double-dot, where

218: the resulting current can be obtained from Eq. (\ref{DDcurrentexpression}).

219: The visibility of the interference pattern, which is defined by the

220: phase-dependence of the current, is given in Eq. (\ref{DDvisibilityI}).

221: It is connected with the visibility obtained from the phase-dependence

222: of the tunneling rate itself (Eqs. (\ref{eq:visDef}), (\ref{eq:visGamm})). 

223: 

224: 

225: \section{The model}

226: 

227: \label{sec:The-model}We consider a Hamiltonian describing two degenerate

228: single-level quantum dots, with respective single-particle states

229: $\left|+\right\rangle $ and $\left|-\right\rangle $ (spin is excluded

230: for simplicity, since we are interested in dephasing of the electronic

231: motion). Each of them is tunnel-coupled to two electrodes (with the

232: same strength for both dots), but involving a possible phase difference

233: between the tunnel amplitudes (see Fig. \ref{DDdoubleslitsetup}).

234: In addition, the potential difference between the two dots is given

235: by a fluctuating field $\hat{F}$, whose dynamics is derived from

236: a linear bath. It represents the fluctuations due to phonons or Nyquist

237: noise. The system-bath coupling strength is taken to be the same for

238: both dots, while the sign is opposite, such that the bath can distinguish

239: between an electron being on dot $\left|+\right\rangle $ or $\left|-\right\rangle $:

240: 

241: \begin{eqnarray}

242: \hat{H} & = & \epsilon (\hat{n}_{+}+\hat{n}_{-})+\hat{F}(\hat{n}_{+}-\hat{n}_{-})+U\hat{n}_{+}\hat{n}_{-}+\nonumber \\

243:  &  & \hat{H}_{L}+\hat{H}_{R}+\hat{H}_{B}+\hat{V}\label{DDhamiltonian}

244: \end{eqnarray}

245: 

246: 

247: Here $\hat{n}_{\pm }$ are the particle numbers on the two dots (equal

248: to $0$ or $1$). The bath Hamiltonian $\hat{H}_{B}$ describes a

249: set of uncoupled harmonic oscillators. It governs the dynamics of

250: the fluctuating potential $\hat{F}$, which is assumed to be linear

251: in the oscillator coordinates. The coupling between electron and bath

252: is of the form of the {}``independent boson model''\cite{mahan}.

253: For the case of exactly one electron on the double-dot, and in the

254: absence of tunneling, it corresponds to a spin-boson model with {}``diagonal

255: coupling''. In this model, no transition between different levels

256: is brought about by the bath, such that pure dephasing results. $U$

257: denotes the Coulomb repulsion energy, which we will take to be so

258: large that double-occupancy is forbidden. Note that the degeneracy

259: of the two dot-levels is important in the following: It is necessary

260: to ensure complete destructive interference at $\varphi =\pi $ (compare

261: also the discussion in Sec. \ref{DDseqTunnelingSection}). 

262: 

263: The terms $\hat{H}_{L}$ and $\hat{H}_{R}$ contain the energies of

264: the electrons in the left and right reservoirs: 

265: 

266: \begin{equation}

267: \hat{H}_{L(R)}=\sum _{k}\epsilon _{k}\hat{a}_{L(R)k}^{\dagger }\hat{a}_{L(R)k}\, .\end{equation}

268: The tunneling between the dots and the leads is described by $\hat{V}=\hat{V}_{L}+\hat{V}_{R}$,

269: with

270: 

271: \begin{equation}

272: \hat{V}_{R}=\sum _{k}t_{k}^{R}\hat{a}_{Rk}^{\dagger }(\hat{d}_{+}+e^{i\varphi }\hat{d}_{-})+h.c.\label{DDtunnelHam}\end{equation}

273: 

274: 

275: for the right junction, and 

276: 

277: \begin{equation}

278: \hat{V}_{L}=\sum _{k}t_{k}^{L}\hat{a}_{Lk}^{\dagger }(\hat{d}_{+}+\hat{d}_{-})+h.c.\label{eq:DDtunnelL}\end{equation}

279: 

280: 

281: for the left junction.

282: 

283: Here $\hat{d}_{\pm }$ are the annihilation operators for the two

284: dots ($\hat{n}_{\pm }=\hat{d}_{\pm }^{\dagger }\hat{d}_{\pm }$) and

285: the phase-factor of $e^{i\varphi }$ controls the interference between

286: tunneling events along either the upper or lower path. The tunneling

287: phase difference might be thought of as arising due to the Aharonov-Bohm

288: phase from a magnetic flux penetrating the region between the quantum

289: dots. 

290: 

291: Note that the tunneling matrix elements $t_{k}^{R(L)}$ are assumed

292: not to depend on the dot state $\kt{+}$ or $\kt{-}$ in our model.

293: This means that the dots are close enough such that they couple to

294: the same point on the lead electrodes, to within less than a Fermi

295: wavelength. Obviously there could be no appreciable interference effect

296: if the dots were separated by some larger distance (in which case

297: the $k$-dependence of matrix elements would be different for the

298: two states). The same idealized assumption underlies several similar

299: models (see, e.g., Refs.~\onlinecite{koeniggefen,BoeseDD,BoeseHofstetterSchoeller}).

300: The effect of an arbitrary dot separation has been discussed in some

301: detail in Ref.~\onlinecite{shahbazyan}. 

302: 

303: The present model, without the bath, has been analyzed previously

304: in Ref.~\onlinecite{BoeseDD} (see also Sec. IV.C of Ref.~\onlinecite{koeniggefen}).

305: There, an orbital type of Kondo effect was found in equilibrium, for

306: $\varphi =\pi $, when the level energy was below the chemical potential.

307: This arises because at $\varphi =\pi $ there are two states of the

308: double-dot that couple only to the left and the right lead, respectively

309: (denoted by $\kt{e}$ and $\kt{o}$ in the following). These degenerate

310: states form the pseudospin responsible for the Kondo effect. However,

311: that mechanism will be irrelevant for our analysis, as we consider

312: the transport situation where the (renormalized) level energy lies

313: between the chemical potentials of the left and the right lead. Therefore,

314: the degeneracy is effectively lifted by the bias voltage (which will

315: be assumed to be much larger than the tunneling rate), and only the

316: state coupling to the left lead would be occupied at $\varphi =\pi $. 

317: 

318: 

319: \section{Qualitative discussion}

320: 

321: \label{sec:Qualitative-discussion}In this and the following three

322: sections, we first analyze the escape of a single electron into the

323: right lead, where the electron is assumed to start out in a symmetric

324: superposition of the two dot levels, which has been formed by an electron

325: tunneling onto the dots from the left lead. In the situation without

326: any bath, this is the state $\left|e\right\rangle \equiv (\left|+\right\rangle +\left|-\right\rangle )/\sqrt{2}$. 

327: 

328: Without dephasing, the tunneling decay out of state $\kt{e}$ is made

329: impossible in the case of perfect destructive interference at $\varphi =\pi $,

330: while maximal constructive interference is present for $\varphi =0$.

331: It should be noted that the attribution of the phase factor to one

332: of the tunnel couplings represents a certain choice of gauge, which

333: affects the wave functions in the following discussion but none of

334: the physically observable quantities that are derived as a result

335: of the master equation in Section \ref{DDseqTunnelingSection}.

336: 

337: For simplicity, we will assume a zero-temperature situation throughout

338: the following qualitative discussion, with a bias $eV>0$ applied

339: between the two dots and the lead in such a way that the electron

340: is allowed to tunnel into the lead (see Fig. \ref{DDevpic}). In addition,

341: since we will describe the tunneling decay to the right, we will only

342: consider the coupling $\hat{V}_{R}$ to the right lead in this section

343: and drop the index $R$ for now.

344: 

345: %

346: \begin{figure}

347: \begin{center}\includegraphics[  height=5cm]{energydiagram.eps}\end{center}

348: 

349: 

350: \caption{\label{DDevpic}The ground state $\kt{\chi _{+}}$ ($\kt{\chi _{-}}$)

351: which the bath assumes in the presence of an electron on dot $\kt{+}$

352: ($\kt{-}$), shown schematically for a single bath oscillator (see

353: main text). After the electron has tunneled into the lead, $\kt{\chi _{-}}$

354: becomes a superposition of excited states (dashed), while the state

355: $\kt{\chi _{0}}$ represents the ground state of the bath in the new

356: potential.}

357: \end{figure}

358: 

359: 

360: Without the bath and for perfect constructive interference ($\varphi =0$),

361: the tunneling decay rate $\Gamma $ will take on its maximum value

362: of $2\Gamma _{0}$, with

363: 

364: \begin{equation}

365: \Gamma _{0}\equiv 2\pi D\left\langle \left|t_{k}\right|^{2}\right\rangle \, ,\label{DDbaretunnelrate}\end{equation}

366: 

367: 

368: where $D$ is the lead density of states at the Fermi energy, $\left\langle \left|t_{k}\right|^{2}\right\rangle $

369: is the angular average of $\left|t_{k}\right|^{2}$ at this energy.

370: The bias voltage $V$ does not enter in this case, as long as it is

371: positive (permitting decay). For $\varphi =\pi $, $\Gamma $ vanishes

372: due to perfect destructive interference. In general, we have:

373: 

374: \begin{equation}

375: \Gamma =\Gamma _{0}(1+\cos \varphi )\, .\label{eq:nobath}\end{equation}

376: 

377: 

378: If the bath is included in the description, the following happens:

379: 

380: First of all, the energy of a single extra electron on any of the

381: two dots will be renormalized from its initial value of $\epsilon $,

382: since the bath relaxes to a ground state of lower energy in presence

383: of the electron. We will assume that the value of $\epsilon $ has

384: been chosen exactly to compensate for this energy change, which is

385: given by $-\int _{0}^{\infty }d\omega \, \left\langle \hat{F}\hat{F}\right\rangle _{\omega }/\omega $

386: (see App. \ref{indepBosonApp}). Then, the energy of an electron on

387: the dot (and the bath in its new ground state) is the same as that

388: of the electron being in the lead, at the Fermi energy of $\epsilon _{F}\equiv 0$

389: (for $V=0$). 

390: 

391: Tunneling of an electron from the dots to the lead will not change

392: the bath state, but it will displace the origin of the harmonic oscillators

393: comprising the bath, since the coupling to $\hat{F}$ is switched

394: off ($\hat{n}_{+}-\hat{n}_{-}$ changes to zero). Therefore, the original

395: ground state of the bath (in presence of the electron) will become

396: a superposition of excited states in the new bath potential (in absence

397: of the electron; see Fig. \ref{DDevpic}). On the other hand, since

398: energy conservation has to be fulfilled with respect to the total

399: energy of the electrons and the bath before and after the tunneling

400: event, only those excited bath states can be reached whose energies

401: are not greater than $eV$, the energy supplied to the electron by

402: the bias voltage. This leads to the Coulomb-blockade type \emph{suppression}

403: of the tunneling rate at low bias voltages, for $\varphi =0$. Physically,

404: this effect is just the same as that described by Franck-Condon overlap

405: integrals evaluated between vibronic states for electronic transitions

406: in molecules. Qualitatively, this effect is independent of the interference

407: setup, since it already occurs for tunneling through a single dot

408: coupled to a bath.

409: 

410: In contrast, for the case of destructive interference ($\varphi =\pi $),

411: the bath may actually \emph{enhance} the tunneling rate from its initial

412: value of $0$, since it partly destroys the phase coherence that is

413: a presupposition for perfect interference. An electron coming from

414: the left lead will form the following entangled state with the bath,

415: instead of the symmetric superposition $\left|e\right\rangle =(\left|+\right\rangle +\left|-\right\rangle )/\sqrt{2}$:

416: 

417: \begin{equation}

418: (\left|+\right\rangle \left|\chi _{+}\right\rangle +\left|-\right\rangle \left|\chi _{-}\right\rangle )/\sqrt{2}\, .\label{DDentangled}\end{equation}

419: 

420: 

421: Here the states $\left|\chi _{\pm }\right\rangle $ denote the respective

422: ground states of the bath for a bath Hamiltonian given by $\hat{H}_{B}\pm \hat{F}$,

423: which are related to each other by a parity transformation (This also

424: means we assume by definition there to be no phase factor between

425: these states; e.g. both may be assumed to have real-valued positive

426: wave functions). Actually, the entangled state considered here will

427: be formed only if the electron is given barely enough energy to enter

428: the double-dot at all (i.e. chemical potential of the left lead infinitesimally

429: larger than the renormalized level position). Otherwise, excited bath

430: states may be created even at this step. These complications will

431: be taken care of in the complete discussion of the sequential tunneling

432: current (Section \ref{DDseqTunnelingSection}). There, it will turn

433: out that the tunneling decay rate derived in the following, based

434: on our physically motivated ansatz (\ref{DDentangled}), is exactly

435: the rate that enters the full master equation. Thus, we proceed with

436: the ansatz (\ref{DDentangled}) for the initial entangled state, in

437: order to calculate the rate for such an electron to tunnel into the

438: right lead.

439: 

440: The bath measures (to some extent) which dot the electron resides

441: on, such that the reduced system density matrix (for the electron

442: on the two dots) becomes mixed and its off-diagonal elements get suppressed

443: by the overlap factor $\left\langle \chi _{+}|\chi _{-}\right\rangle $.

444: Put differently, the phase factor between the two dot states in the

445: wave function of the electron (initially equal to $+1$) becomes uncertain.

446: Therefore, there is a finite probability of 

447: 

448: \begin{equation}

449: P_{o}=(1-\left\langle \chi _{+}|\chi _{-}\right\rangle )/2\label{eq:Podd}\end{equation}

450: 

451: 

452: to find the electron in the antisymmetric (odd) state $\left|o\right\rangle \equiv (\left|+\right\rangle -\left|-\right\rangle )/\sqrt{2}$.

453: At $\varphi =\pi $, where tunneling decay of the symmetric superposition

454: $\kt{e}$ is blocked due to destructive interference, the state $\kt{o}$

455: is allowed to decay into the lead, at the maximal rate of $2\Gamma _{0}$.

456: In this way, the interference-induced blockade of electron tunneling

457: is lifted by dephasing. 

458: 

459: However, this simple picture is true only for large bias voltages,

460: when energy conservation permits any final state of the bath after

461: the tunneling event. If the maximum energy supplied to the electron

462: is limited, the suppression discussed above (for the case of $\varphi =0$)

463: will apply again. In particular, if the bias voltage is turned to

464: zero, energy conservation only allows the state $\left|\chi _{0}\right\rangle $

465: to be reached, which is the ground state of the bath in the absence

466: of any electrons on the dots. Then, the tunneling rate is exactly

467: zero again, despite the fact that the reduced density matrix of the

468: electron may be mixed to a strong extent. The reason is the following:

469: When the overlap of the entangled state (\ref{DDentangled}) with

470: the state $\left|\chi _{0}\right\rangle $ is taken, the two overlap

471: factors $\left\langle \chi _{0}|\chi _{+}\right\rangle $ and $\left\langle \chi _{0}|\chi _{-}\right\rangle $

472: turn out to be the same, \emph{if} the coupling of the bath to the

473: two dots is symmetric (i.e. of equal strength, only of opposite sign),

474: which we have assumed in writing down the Hamiltonian, Eq. (\ref{DDhamiltonian}).

475: Therefore, the electronic state resulting from the projection of (\ref{DDentangled})

476: onto $\kt{\chi _{0}}$ is equal to the symmetric combination, whose

477: decay is forbidden. Thus, the combination of energy conservation and

478: Pauli blocking prevents a finite tunneling rate at zero bias voltage,

479: in spite of the mixed state of the electron coupled to the bath. In

480: this limit the entanglement between electron and bath only leads to

481: renormalization effects (such as the change in tunneling rate), but

482: not to genuine dephasing. If the coupling were asymmetric, then destructive

483: interference could be lost even without dephasing (merely due to renormalization),

484: just as it would be the case for initially asymmetric bare tunnel

485: couplings. That is why the asymmetric case is uninteresting for our

486: purposes of distinguishing renormalization effects from real dephasing. 

487: 

488: However, whether we are indeed able to claim that dephasing actually

489: vanishes in the limit of low bias voltages will depend on the behaviour

490: of the tunneling rate as a function of $V$ and on the comparison

491: of the cases $\varphi =0$ and $\varphi =\pi $. Here, the bath spectrum,

492: and, above all, its low-frequency properties, enter. In order to be

493: able to discuss $\Gamma (V,\varphi )$ quantitatively, we will make

494: use of the concepts of the $P(E)$ theory of tunneling in a dissipative

495: environment.

496: 

497: 

498: \section{Decay rate and connection to $P(E)$ theory}

499: 

500: \label{sec:Decay-rate-and}The tunneling rate $\Gamma $ will be calculated

501: using the standard Fermi Golden Rule, i.e. lowest order perturbation

502: theory in the bare tunneling rate $\Gamma _{0}$, but taking into

503: account exactly the bath coupling. In deriving the formula for $\Gamma $,

504: it turns out to be useful to assume that the bath oscillators do \emph{not}

505: get shifted in the tunneling event (unlike the qualitative considerations

506: from above), but it is rather the bath states which get displaced

507: (in the opposite direction). Obviously, this amounts to the same,

508: as long as we are interested only in overlap integrals of different

509: bath states after the event. To that end, we introduce the displacement

510: operator $\exp (i\hat{\phi })$, which transforms the bath ground

511: state of $\hat{H}_{B}$ into that of $\hat{H}_{B}+\hat{F}$. Here

512: $\hat{\phi }$ is a suitable hermitian operator that is linear in

513: the bosonic variables of the bath. In fact, this amounts to performing

514: the canonical transformation of the independent boson model\cite{mahan},

515: see Appendix \ref{indepBosonApp}. In terms of the two dot states

516: $+$ and $-$, we have $\hat{F}_{+}=\hat{F}$ and $\hat{F}_{-}=-\hat{F}$,

517: as well as $\ph _{+}=\ph $ and $\ph _{-}=-\ph $. The transformation

518: eliminates the system-bath coupling from the Hamiltonian, but gives

519: rise to modified dot operators $\hat{d}_{\pm }'=e^{\pm i\ph }\hat{d}_{\pm }$

520: in the transformed tunnel Hamiltonian $\hat{V}_{R}'$ (see Eq. (\ref{IBdtransform})). 

521: 

522: We will assume the tunnel-coupling to be sufficiently weak, such that

523: we can use lowest-order perturbation theory to calculate the tunneling

524: decay rate:

525: 

526: \begin{equation}

527: \Gamma =2\pi \sum _{f}\left|\left\langle f|\hat{V}'_{R}|i\right\rangle \right|^{2}\delta (E_{f}-E_{i})\, ,\end{equation}

528: 

529: 

530: where the initial state $\left|i\right\rangle $ is given by the configuration

531: involving the electron residing in the symmetric superposition on

532: the dots, the unperturbed Fermi sea in the lead and the bath in its

533: ground state $\left|i_{B}\right\rangle $. The bath ground state has

534: become independent of the position of the electron, due to the above-mentioned

535: transformation. At finite temperatures, an additional thermal average

536: over the initial bath state and the initial state of the electrons

537: in the lead has to be performed. The energies and eigenstates refer

538: to the Hamiltonian without tunnel coupling. Applying the new tunneling

539: Hamiltonian $\hat{V}'_{R}$ to the initial state, we obtain the following

540: expression:

541: 

542: \begin{eqnarray}

543: \Gamma =\pi \sum _{k,f_{B}}\left|t_{k}\right|^{2}(1-f(\epsilon _{k}+eV))\times  &  & \nonumber \\

544: \left|\left\langle f_{B}|e^{+i\hat{\phi }}+e^{i\varphi }e^{-i\hat{\phi }}|i_{B}\right\rangle \right|^{2}\delta (E_{f}^{B}-E_{i}^{B}+\epsilon _{k})\, , &  & \label{DDgammaanfang}

545: \end{eqnarray}

546: 

547: 

548: Here $f(\cdot )$ is the Fermi function (for chemical potential equal

549: to zero), and $E_{f,i}^{B}$ are the energies of the initial and final

550: bath states. The energy supplied to the bath is equal to the energy

551: lost by the electron (given by $-\epsilon _{k}$, since the renormalized

552: dot energy is zero). Following the usual derivation of the $P(E)$

553: theory \cite{ingoldhabil,SchoenPE}, we express the energy-conserving

554: $\delta $ function as an integral over time and also replace the

555: sum over lead states $k$ by an integral over the energy $E=-\epsilon _{k}$

556: supplied to the bath, finally yielding:

557: 

558: \begin{eqnarray}

559: \Gamma =\Gamma _{0}\int _{-\infty }^{+\infty }dE(1-f(-E+eV))\int _{-\infty }^{+\infty }\frac{dt}{2\pi }e^{iEt}\times  &  & \nonumber \\

560: \frac{1}{2}\left\langle (e^{-i\hat{\phi }(t)}+e^{-i\varphi }e^{i\hat{\phi }(t)})(e^{i\hat{\phi }}+e^{i\varphi }e^{-i\hat{\phi }})\right\rangle  &  & \label{DDGammavorher}

561: \end{eqnarray}

562: 

563: 

564: For the case of arbitrary temperature, the brackets denote a thermal

565: average over the initial bath state $\left|i_{B}\right\rangle $.

566: We introduce the definitions:

567: 

568: \begin{equation}

569: P_{(-)}(E)=\frac{1}{2\pi }\int _{-\infty }^{+\infty }dt\, e^{iEt}\, e^{\pm \left\langle \hat{\phi }(t)\hat{\phi }\right\rangle -\left\langle \hat{\phi }^{2}\right\rangle }\, .\label{DDPEDef}\end{equation}

570: 

571: 

572: This permits us to write down our final result for the tunneling decay

573: rate in terms of $P_{(-)}(E)$:

574: 

575: \begin{equation}

576: \Gamma =\Gamma _{0}\int _{-\infty }^{+\infty }dE\, (1-f(-E+eV))\, (P(E)+\cos (\varphi )P_{-}(E))\label{DDGammabasic}\end{equation}

577: 

578: 

579: The formula given here constitutes the basic expression for the decay

580: rate as a function of bias voltage and interference phase $\varphi $.

581: It represents the appropriate modification of Eq. (\ref{eq:nobath})

582: in presence of a bath. 

583: 

584: Note that for the slightly more general case of arbitrarily correlated

585: fluctuating potentials $\hat{F}_{+}$ and $\hat{F}_{-}$ attached

586: to the dots (i.e. an interaction of the form $\hat{F}_{+}\hat{n}_{+}+\hat{F}_{-}\hat{n}_{-}$),

587: the function $P_{-}(E)$ would contain the cross-correlator of the

588: associated phases $\hat{\phi }_{+}$ and $\hat{\phi }_{-}$, while

589: $P(E)$ would depend on the autocorrelator of $\hat{\phi }_{+}$ or

590: $\hat{\phi }_{-}$ (assumed to be the same, for the setup to remain

591: symmetric). In contrast to the model treated here, such an interaction

592: would also involve fluctuations of the sum of energies of the dot-levels.

593: However, they would only add to the renormalization effects mentioned

594: previously and do not contribute to dephasing by themselves, since

595: such fluctuations cannot distinguish between the two interfering paths.

596: 

597: By using the definitions

598: 

599: \begin{equation}

600: \gamma _{(-)}\equiv \Gamma _{0}\int dE\, (1-f(-E+eV))\, P_{(-)}(E)\, ,\label{DDgammaDefinition}\end{equation}

601: 

602: 

603: we can write

604: 

605: \begin{equation}

606: \Gamma =\gamma +\cos (\varphi )\gamma _{-}\, .\end{equation}

607: 

608: 

609: The strength of the dependence of $\Gamma $ on the phase $\varphi $

610: may be taken as a signature of phase coherence in our model. We define

611: the {}``visibility'' of the interference pattern in the usual way,

612: by

613: 

614: \begin{equation}

615: \upsilon \equiv (\Gamma _{max}-\Gamma _{min})/(\Gamma _{max}+\Gamma _{min}),\label{eq:visDef}\end{equation}

616: 

617: 

618: which is equal to the ratio

619: 

620: \begin{equation}

621: \upsilon =\frac{\gamma _{-}}{\gamma }\, .\label{eq:visGamm}\end{equation}

622: 

623: 

624: The visibility $\upsilon $ will be $1$ whenever the destructive

625: interference is perfect, and it is zero if there is no dependence

626: of $\Gamma $ on $\varphi $.

627: 

628: The effects of the bath on the decay rate are encoded in the functions

629: $P(E)$ and $P_{-}(E)$, whose general properties we will discuss

630: now. In the next section, we will evaluate them for different types

631: of bath spectra. 

632: 

633: As usual, the function $P(E)$ describes the probability (density)

634: that an electron will emit the energy $E$ into the bath while tunneling

635: into the lead. It is real, nonnegative and normalized to unity \cite{SchoenPE,ingoldhabil}. 

636: 

637: At large times $\left|t\right|\rightarrow \infty $, the correlation

638: function $\left\langle \hat{\phi }(t)\hat{\phi }\right\rangle $ in

639: the exponent of the integral (\ref{DDPEDef}) will decay towards zero,

640: for a continuous bath spectrum. This means that the integrand of $P(E)$

641: approaches the value of $z\equiv \exp (-\left\langle \hat{\phi }^{2}\right\rangle )$,

642: starting from $1$ at $t=0$. Therefore, $P(E)$ contains a {}``quasiparticle

643: $\delta $ peak'' of strength $z$ at $E=0$, if $z$ does not vanish.

644: It corresponds to the probability $z$ of having no energy transfer

645: at all from the electron to the bath (similar to the recoil-free emission

646: of a $\gamma $ ray by a nucleus inside a crystal, i.e. the Mössbauer

647: effect).

648: 

649: The function $P_{-}(E)$ in front of the $\cos (\varphi )$ term in

650: Eq. (\ref{DDGammabasic}) is different: The integrand of $P_{-}(E)$

651: will increase at large times, towards the value of $z$, starting

652: from $z^{2}$ at $t=0$. The function $P_{-}(E)$ is real-valued (because

653: of $\left\langle \hat{\phi }(t)\hat{\phi }\right\rangle =\left\langle \hat{\phi }\hat{\phi }(t)\right\rangle ^{*}$),

654: but it can become negative. Therefore, it cannot be interpreted as

655: a probability density, in contrast to $P(E)$. Its normalization is

656: given by:

657: 

658: \begin{equation}

659: \int dE\, P_{-}(E)=z^{2}\, .\end{equation}

660: 

661: 

662: If $z$ is nonzero, $P_{-}(E)$ also has a $\delta $ peak at $E=0$,

663: of weight $z$, just as $P(E)$. As a consequence, in the case of

664: destructive interference ($\varphi =\pi $), the tunneling rate $\Gamma $

665: at $V\rightarrow 0,\, T=0$ still vanishes even in the presence of

666: the bath, since the $\delta $ peaks contained in $P(E)$ and $P_{-}(E)$

667: cancel exactly in the integral (\ref{DDGammabasic}). The physical

668: reason for this coherence has been discussed at the end of the previous

669: section. 

670: 

671: In the case of constructive interference ($\varphi =0$), at $T=0$

672: and for $V\rightarrow 0$, the integration over $E$ will only capture

673: the $\delta $ peaks contained in $P_{(-)}(E)$, yielding $\Gamma =2z\Gamma _{0}$.

674: Thus, the tunneling rate is suppressed by the constant factor $z$

675: from its noninteracting value. However, this may be interpreted as

676: a mere renormalization of the effective tunnel coupling, since the

677: visibility $\upsilon $ of the interference pattern is still equal

678: to unity. In order to connect this result to the qualitative discussion

679: from above, we note that the overlap of the two different bath ground

680: states that are adapted to the absence or presence of an electron

681: on dot $\pm $, is given by: 

682: 

683: \begin{equation}

684: \left\langle \chi _{0}|\chi _{\pm }\right\rangle =\left\langle \chi _{0}\left|e^{\pm i\hat{\phi }}\right|\chi _{0}\right\rangle =\exp (-\left\langle \hat{\phi }^{2}\right\rangle /2)=z^{1/2}\, ,\end{equation}

685: 

686: 

687: Therefore, the magnitude squared of this overlap, that determines

688: the probability of tunneling without exciting any bath mode, is equal

689: to $z$.

690: 

691: On the other hand, for sufficiently large bias voltages (much larger

692: than the cutoff frequency of the bath spectrum), the normalization

693: conditions for $P_{(-)}(E)$ yield

694: 

695: \begin{equation}

696: \Gamma =\Gamma _{0}(1+z^{2}\cos (\varphi ))\, .\label{DDgammahighv}\end{equation}

697: 

698: 

699: The visibility is given by $\upsilon =z^{2}$. In this limiting case,

700: where the restrictions due to energy conservation and the Pauli principle

701: are no longer important, the tunneling rate $\Gamma $ at the point

702: $\varphi =\pi $ of destructive interference does not vanish. It takes

703: the value $\Gamma _{0}(1-z^{2})$, which is small if the effects of

704: the bath are weak ($z$ near to $1$) and is equal to one half the

705: ideal maximum value $2\Gamma _{0}$ for a bath that is sufficiently

706: strong to destroy phase coherence completely ($z=0$), leading to

707: an incoherent mixture of symmetric and antisymmetric states on the

708: two dots. In the latter case, the visibility vanishes (even for arbitrary

709: voltages), since then $P_{-}(E)$ is equal to zero, which makes $\Gamma $

710: independent of $\varphi $. This will be true for the Ohmic bath,

711: to be discussed in the next section. 

712: 

713: As explained above, the reduced density matrix of the electron on

714: the dots coupled to the bath predicts a finite probability of $P_{o}=(1-\left\langle \chi _{+}|\chi _{-}\right\rangle )/2$

715: to find the electron in the antisymmetric state if one starts out

716: from the symmetric superposition before coupling it to the bath. The

717: overlap factor of the bath states involved in this probability can

718: be expressed as 

719: 

720: \begin{equation}

721: \left\langle \chi _{+}|\chi _{-}\right\rangle =\left\langle \chi _{0}|(e^{-i\hat{\phi }})^{2}|\chi _{0}\right\rangle =z^{2}\, .\label{DDchipm}\end{equation}

722: 

723: 

724: Comparing with the result $\Gamma (\varphi =\pi )=\Gamma _{0}(1-z^{2})$

725: given above, it may be observed that the decay rate at sufficiently

726: large bias voltages is indeed determined directly by the probability

727: to find the electron in the state whose decay is not forbidden by

728: destructive interference (as has been argued already at the end of

729: the previous section, near Eq. (\ref{eq:Podd})). It is only in this

730: limiting case, where an arbitrary amount of energy is available for

731: excitation of the bath, that the suppression of interference effects

732: in the transport situation is correctly deduced from the electron's

733: reduced density matrix in the presence of the bath. Formally, this

734: holds because the sum over final bath states $f_{B}$ in Eq. (\ref{DDgammaanfang})

735: is not restricted any more and corresponds to the insertion of a complete

736: set of basis states. Thus, one obtains, directly from Eq. (\ref{DDgammaanfang}):

737: 

738: \begin{equation}

739: \Gamma =\frac{\Gamma _{0}}{2}\left\langle \chi _{+}+e^{-i\varphi }\chi _{-}|\chi _{+}+e^{i\varphi }\chi _{-}\right\rangle \, ,\label{eq:highvexplain}\end{equation}

740: 

741: 

742: which reduces to Eq. (\ref{DDgammahighv}) when the overlaps are evaluated,

743: using Eq. (\ref{DDchipm}). Physically, the case of high bias voltage

744: corresponds to a kind of infinitely fast von Neumann projection measurement

745: that determines the state of the electron, revealing the fluctuations

746: due to the bath. In contrast, at low bias voltages (low energy supply),

747: a kind of {}``weak'' measurement is carried out that takes a longer

748: amount of time, such that only the low-frequency fluctuations of the

749: bath are important for dephasing.

750: 

751: 

752: \section{Evaluation for different bath spectra}

753: 

754: \label{sec:Evaluation-for-different}We will restrict the discussion

755: to $T=0$ at first. 

756: 

757: The simplest example for the bath is a single harmonic oscillator

758: of frequency $\omega $. This offers an approximate description of

759: the interaction with optical phonon modes ({}``Einstein model'').

760: In this case, $P(E)$ and $P_{-}(E)$ can be obtained easily by expanding

761: the exponential in a Taylor series and using $\left\langle \hat{\phi }(t)\hat{\phi }\right\rangle =\left\langle \hat{\phi }^{2}\right\rangle \exp (-i\omega t)$,

762: before the integration over time is performed. For $P(E)$, the resulting

763: series of $\delta $ peaks at harmonics of $\omega $ corresponds

764: to all possible processes where the electron emits any number $n$

765: of phonons into the bath while tunneling into the lead. The expression

766: for $P_{-}(E)$ is the same, apart from alternating signs in front

767: of the $\delta $ functions: 

768: 

769: \begin{equation}

770: P_{(-)}(E)=z\sum _{n=0}^{\infty }\frac{\left\langle \pm \hat{\phi }^{2}\right\rangle ^{n}}{n!}\delta (E-n\omega )\, .\end{equation}

771: 

772: 

773: Thus, every process involving the transfer of an even number of quanta

774: to the bath will not ruin the destructive interference at $\varphi =\pi $,

775: since the corresponding contributions from $P(E)$ and $P_{-}(E)$

776: cancel in Eq. (\ref{DDGammabasic}). This is because the coupling

777: between electron and bath is of the type $(\hat{n}_{+}-\hat{n}_{-})\hat{F}$,

778: which gives a different sign of the interaction amplitude for a phonon

779: emission process, depending on the dot. Therefore, the amplitude of

780: emission of an \emph{even} number of phonons will \emph{not} depend

781: on the dot, it is insensitive to the state of the electron, and the

782: amplitudes of the electron tunneling from $\left|+\right\rangle $

783: and $\left|-\right\rangle $ will still interfere destructively.

784: 

785: In contrast, emission processes involving an odd number of quanta

786: introduce a negative sign for an electron starting in state $\left|-\right\rangle $,

787: {}``detecting'' the path (or rather, the initial state) of the electron

788: and interfering \emph{constructively} with the processes from $\left|+\right\rangle $.

789: This lifts the destructive interference and makes $\Gamma \neq 0$

790: at $\varphi =\pi $. However, below the frequency $\omega $ of the

791: oscillator, destructive interference at $\varphi =\pi $ is still

792: perfect since no quantum can be emitted, while the magnitude of $\Gamma $

793: at $\varphi =0$ is renormalized by the factor $z$, as has been discussed

794: above in general for the limiting case $V\rightarrow 0$. The same

795: holds true for any bath with a finite excitation gap, at $T=0$. This

796: is shown in Figs. \ref{DDGamma0} and \ref{DDGammaPi}, to be discussed

797: in the next section.

798: 

799: We now pass on to arbitrary bath spectra. At first, we will cover

800: the case $z\neq 0$ ({}``weak baths''), when we can apply perturbation

801: theory to discuss the behaviour of $P_{(-)}(E)$ at low energy transfers

802: $E$ (and, consequently, that of $\Gamma $ at low voltages). A Taylor-expansion

803: of the exponent in Eq. (\ref{DDPEDef}) yields:

804: 

805: \begin{eqnarray}

806: P_{(-)}(E)=\frac{z}{2\pi }\sum _{n=0}^{\infty }\frac{1}{n!}\int _{-\infty }^{+\infty }dt\, e^{iEt}\left[\pm \left\langle \hat{\phi }(t)\hat{\phi }\right\rangle \right]^{n} &  & \nonumber \\

807: =z\sum _{n=0}^{\infty }\frac{(\pm 1)^{n}}{n!}(\left\langle \hat{\phi }\hat{\phi }\right\rangle _{\omega }*\ldots *\left\langle \hat{\phi }\hat{\phi }\right\rangle _{\omega })(E) &  & \label{DDPfolding}

808: \end{eqnarray}

809: 

810: 

811: The repeated convolution product contains $n$ times the correlator

812: $\left\langle \hat{\phi }\hat{\phi }\right\rangle _{\omega }$, for

813: $n=0$ it is to equal $\delta (E)$, and the negative sign holds for

814: $P_{-}(E)$. 

815: 

816: For the following discussion, we prescribe the spectrum of the fluctuating

817: potential $\hat{F}$ to be a power-law in frequency $\omega $ (at

818: $T=0$), with exponent $s$:

819: 

820: \begin{equation}

821: \left\langle \hat{F}\hat{F}\right\rangle _{\omega }^{T=0}=2\alpha \omega _{c}\left(\frac{\omega }{\omega _{c}}\right)^{s}\theta (\omega _{c}-\omega )\theta (\omega ),\label{DDUUlaw}\end{equation}

822: 

823: 

824: The dimensionless parameter $\alpha $ characterizes the bath strength.

825: In order to be able to rely on perturbation theory, we have to ensure

826: $z>0$. Since $\left\langle \hat{\phi }\hat{\phi }\right\rangle _{\omega }=\left\langle \hat{F}\hat{F}\right\rangle _{\omega }/\omega ^{2}$,

827: the variance of the fluctuating phase, $\left\langle \hat{\phi }^{2}\right\rangle $,

828: will be finite only for $s>1$ (at $T=0$, otherwise $s>2$). In that

829: case, we have $z=\exp (-2\alpha /(s-1))$. This means the perturbative

830: analysis presented above is restricted to a super-Ohmic bath, $s>1$.

831: The case of the Ohmic bath will be discussed separately further below.

832: 

833: After keeping only terms up to second order in the expansion of $P_{(-)}(E)$

834: given in Eq. (\ref{DDPfolding}), we get

835: 

836: \begin{equation}

837: P(E)+P_{-}(E)=z(2\delta (E)+(\left\langle \hat{\phi }\hat{\phi }\right\rangle _{\omega }*\left\langle \hat{\phi }\hat{\phi }\right\rangle _{\omega })(E)+\ldots )\, ,\end{equation}

838: 

839: 

840: for the symmetric combination, that will determine the prefactor of

841: $1+\cos (\varphi )$ in the expression for $\Gamma $, Eq. (\ref{DDGammabasic}),

842: and

843: 

844: \begin{equation}

845: P(E)-P_{-}(E)=2z\left\langle \hat{\phi }\hat{\phi }\right\rangle _{E}+\ldots \end{equation}

846: 

847: 

848: for the antisymmetric combination (determining the prefactor of $1-\cos (\varphi )$).

849: Inserting these into (\ref{DDGammabasic}), using the power law for

850: $\left\langle \hat{\phi }\hat{\phi }\right\rangle _{\omega }=\left\langle \hat{F}\hat{F}\right\rangle _{\omega }/\omega ^{2}$

851: given by (\ref{DDUUlaw}), and performing the energy integrals, we

852: find, for sufficiently low voltages ($2\alpha (eV/\omega _{c})^{s-1}\ll s-1$): 

853: 

854: \begin{eqnarray}

855: \Gamma \approx \frac{\Gamma _{0}}{2}z\{(1+\cos (\varphi ))(1+\frac{\alpha ^{2}C_{s}}{(s-1)}\left(\frac{eV}{\omega _{c}}\right)^{2(s-1)})+ &  & \nonumber \\

856: (1-\cos (\varphi ))\frac{2\alpha }{s-1}\left(\frac{eV}{\omega _{c}}\right)^{s-1}\}\, . &  & \label{DDweakgamma}

857: \end{eqnarray}

858: 

859: 

860: The numerical prefactor $C_{s}$ is defined as $\int _{0}^{1}(y(1-y))^{s-2}dy$. 

861: 

862: From Eq. (\ref{DDweakgamma}), we see that the destructive interference

863: at $\varphi =\pi $ is perfect at $V=0$, but gets lifted when increasing

864: the bias voltage, with a power $V^{s-1}$. In contrast, the decay

865: rate $\Gamma $ at $\varphi =0$ starts out from the constant value

866: of $2z\Gamma _{0}$ and grows as $V^{2(s-1)}$. Therefore, the visibility

867: $\upsilon $ starts out at $1$ for $V=0$ but decreases as:

868: 

869: \begin{equation}

870: \upsilon \approx 1-\frac{4\alpha }{s-1}\left(\frac{eV}{\omega _{c}}\right)^{s-1}\, .\label{viasapprox}\end{equation}

871: 

872: 

873: For $s\downarrow 1$, the range in bias voltage $V$ where these approximate

874: expressions hold shrinks to zero (at constant $\alpha $ and $\omega _{c}$).

875: At $s=1$, i.e. for the Ohmic bath, the probability $z$ of not emitting

876: energy into the bath vanishes completely. As discussed above, this

877: means that there is no $\varphi $-dependence at all in $\Gamma $,

878: and, consequently, the visibility is zero at all bias voltages. Furthermore,

879: the tunneling rate vanishes for $eV\rightarrow 0$, even at $\varphi =0$.

880: This is the well-known Coulomb-blockade type of behaviour for tunneling

881: in presence of Ohmic dissipation \cite{devoret}. At higher bias voltages,

882: the blockade is removed and $\Gamma $ grows towards $\Gamma _{0}$.

883: The growth at low voltages is determined by the power-law behaviour

884: of $P(E)$, which rises as $c\omega _{c}^{-2\alpha }E^{2\alpha -1}$,

885: where the exponent is determined by the bath-strength rather than

886: the exponent $s=1$ of the bath spectrum. The dimensionless prefactor

887: $c$ must be found from the normalization condition for $P(E)$ and

888: depends only on $\alpha $ (and the type of cutoff in the bath spectrum).

889: Therefore, in the case of the Ohmic bath we have, at low $V$ and

890: $T=0$: 

891: 

892: \begin{equation}

893: \Gamma (V)=\Gamma _{0}\frac{c}{2\alpha }\left(\frac{eV}{\omega _{c}}\right)^{2\alpha }\, .\end{equation}

894: 

895: 

896: Finally, we briefly discuss the case of finite temperatures, $T>0$.

897: 

898: In that case, the variance of $\hat{\phi }$ is given by

899: 

900: \begin{equation}

901: \left\langle \hat{\phi }^{2}\right\rangle =\int _{0}^{\infty }d\omega \, \left\langle \hat{\phi }\hat{\phi }\right\rangle _{\omega }^{(T=0)}\coth \left(\frac{\omega }{2T}\right)\, ,\end{equation}

902: 

903: 

904: which yields

905: 

906: \begin{equation}

907: \left\langle \hat{\phi }^{2}\right\rangle \approx \left\langle \hat{\phi }^{2}\right\rangle ^{(T=0)}+4\alpha \left(\frac{T}{\omega _{c}}\right)^{s-1}\int _{0}^{\infty }\frac{y^{s-2}}{e^{y}-1}dy\, .\end{equation}

908: 

909: 

910: The approximation of extending the integral to infinity holds for

911: temperatures much smaller than the bath cutoff $\omega _{c}$. This

912: formula gives the temperature-dependence of the renormalization factor

913: $z=\exp \left(-\left\langle \hat{\phi }^{2}\right\rangle \right)$.

914: The second integral diverges for $s\leq 2$, because $z=0$ for these

915: cases, in contrast to $T=0$ where $z=0$ only for $s\leq 1$. Again,

916: this results in complete absence of the interference effect in the

917: tunneling rate $\Gamma (V,\varphi )$ (because $P_{-}(E)$ vanishes).

918: It may seem surprising that an infinitesimally small temperature can

919: yield such a drastic qualitative change (for $1<s\leq 2$), compared

920: to the zero-temperature case, since the difference should be observable

921: only at very large times $t\gg 1/T$. However, it must be remembered

922: that our analysis is carried out for the limit $\Gamma _{0}\rightarrow 0$,

923: where the average decay time of the given state is inifinitely large.

924: In other words, the limits $T\rightarrow 0$ and $\Gamma _{0}\rightarrow 0$

925: do not commute for such relatively strong baths. At finite $\Gamma _{0}$,

926: the transition from one to the other case should turn out to be smooth,

927: but this goes beyond the present analysis. 

928: 

929: Apart from the change in $z$ with temperature, there are two other

930: important differences to the case $T=0$: First of all, even at $V\rightarrow 0$

931: the electron may emit energy into the bath, due to the thermal smearing

932: of the Fermi surface in the lead (lifting of Pauli blocking). Secondly,

933: it may now also absorb some energy during the tunneling process. Both

934: facts will, in general, lead to a finite tunneling decay rate at $\varphi =\pi ,\, V\rightarrow 0$

935: for any bath, where, at $T=0$, the rate had vanished in any case. 

936: 

937: We can approximate the visibility $\upsilon $ at $V\rightarrow 0$

938: and finite $T$ by using the expansion (\ref{DDPfolding}). Inserting

939: the resulting expressions for $\gamma _{(-)}$ (\ref{DDgammaDefinition})

940: into $\upsilon =\gamma _{-}/\gamma $, we obtain

941: 

942: \begin{equation}

943: \upsilon (T,V\rightarrow 0)\approx 1-4\int d\epsilon \, \left\langle \ph \ph \right\rangle _{\epsilon }f(\epsilon )\, .\end{equation}

944: 

945: 

946: We evaluate the integral for a power-law bath spectrum in the limit

947: $T\ll \omega _{c}$:

948: 

949: \begin{eqnarray}

950: \int d\epsilon \, \left\langle \ph \ph \right\rangle _{\epsilon }f(\epsilon ) & = & \nonumber \\

951: =\int _{0}^{\infty }d\epsilon \, \frac{\left\langle \ph \ph \right\rangle _{\epsilon }^{T=0}}{\sinh (\beta \epsilon )} &  & \nonumber \\

952: \approx 2\alpha \omega _{c}^{1-s}\int _{0}^{\infty }d\epsilon \, \frac{\epsilon ^{s-2}}{\sinh (\beta \epsilon )}\, . &  & 

953: \end{eqnarray}

954: 

955: 

956: This yields:

957: 

958: \begin{equation}

959: 1-\upsilon (T,V\rightarrow 0)\approx 32\, \alpha \, \left(\frac{T}{\omega _{c}}\right)^{s-1}(\frac{1}{2}-2^{-s})\Gamma (s-1)\zeta (s-1)\, ,\label{DDvisfiniteT}\end{equation}

960: 

961: 

962: where $\Gamma $ is the Euler gamma function, and $\zeta $ the Riemann

963: zeta function. Therefore, the decrease of the visibility with increasing

964: temperature $T$ (and $V\rightarrow 0$) is governed by the same power-law

965: as that for increasing bias voltage $V$ at $T=0$, see Eq. (\ref{viasapprox}).

966: 

967: 

968: \section{Discussion of the results}

969: 

970: \label{sec:Discussion-of-the}The following discussion relates to

971: the results obtained for $T=0$, that are plotted in the figures.

972: 

973: %

974: \begin{figure}

975: \begin{center}\includegraphics[  height=7cm]{arrayNew.eps}\end{center}

976: 

977: 

978: \caption{\label{DDbathspectra}The bath spectrum $\left\langle \hat{F}\hat{F}\right\rangle _{E}$

979: (bottom) and the resulting functions $P(E)$ (top) and $P_{-}(E)$

980: (middle), plotted vs. energy $E$, for different baths. Energies are

981: measured in units of the {}``bath cutoff'' $\omega _{c}$. Energy

982: axis is the same in all panels (starting at $E=0$, horizontal tick

983: distance: $1$); vertical tick distance in all panels is $0.5$. \emph{a}:

984: $s=1.5,\, \alpha =0.25$; \emph{b}: {}``acoustic phonons'', $s=3,\, \alpha =1$;

985: \emph{c}: {}``optical phonons'', Bath with gap; \emph{d}: $s=1,\, \alpha =0.25$;

986: \emph{e}: $s=1,\, \alpha =0.75$ (d,e are {}``Ohmic'' baths of different

987: strength, $z=0$)}

988: \end{figure}

989: 

990: 

991: %

992: \begin{figure}

993: \begin{center}\includegraphics[  height=7cm]{Gamm0NEW.eps}\end{center}

994: 

995: 

996: \caption{\label{DDGamma0}Decay rate $\Gamma $ as a function of bias voltage

997: $V$ for the case of \emph{constructive} interference ($\varphi =0$),

998: at $T=0$. Curves correspond to different bath spectra shown in Fig.

999: \ref{DDbathspectra}. Dashed lines correspond to approximation Eq.

1000: (\ref{DDweakgamma}). The initial Coulomb-blockade type suppression

1001: to a value of $\Gamma /2\Gamma _{0}=z$ ($z=0$ for the Ohmic bath

1002: d,e) is lifted with increasing bias voltage, saturating at $\Gamma /2\Gamma _{0}=(1+z^{2})/2$.

1003: Inset depicts energy diagram with definition of bias voltage for this

1004: situation.}

1005: \end{figure}

1006: 

1007: 

1008: %

1009: \begin{figure}

1010: \begin{center}\includegraphics[  height=7cm]{GammaPINEW.eps}\end{center}

1011: 

1012: 

1013: \caption{\label{DDGammaPi}Decay rate $\Gamma (V)$ for the case of \emph{destructive}

1014: interference $(\varphi =\pi )$, at $T=0$. Dashed lines refer to

1015: Eq. (\ref{DDweakgamma}). Due to dephasing, the decay rate becomes

1016: finite at finite voltages, saturating at $\Gamma /2\Gamma _{0}=(1-z^{2})/2$.

1017: For the Ohmic bath (d,e) the dependence is exactly equal to that for

1018: $\varphi =0$ (Fig. \ref{DDGamma0}).}

1019: \end{figure}

1020: %

1021: \begin{figure}

1022: \begin{center}\includegraphics[  height=7cm]{Vis.eps}\end{center}

1023: 

1024: 

1025: \caption{\label{DDVis}Visibility $\upsilon =(\Gamma _{\varphi =0}-\Gamma _{\varphi =\pi })/(\Gamma _{\varphi =0}+\Gamma _{\varphi =\pi })$

1026: as a function of bias voltage $V$ for different bath spectra (see

1027: Fig. \ref{DDbathspectra}). For the Ohmic bath (cases d,e) $\upsilon \equiv 0$.

1028: Dashed lines correspond to Eq. (\ref{viasapprox}). Inset illustrates

1029: change in interference pattern $\Gamma (\varphi )$ upon switching

1030: on the interaction with the bath.}

1031: \end{figure}

1032: 

1033: 

1034: In Fig. \ref{DDbathspectra}, several different types of bath spectra

1035: $\left\langle \hat{F}\hat{F}\right\rangle _{E}$ are shown. Cases

1036: (a),(b),(d) and (e) are power-laws of the form given in Eq. (\ref{DDUUlaw}),

1037: for a cutoff frequency of $\omega _{c}=1$. The last two (d,e) are

1038: of Ohmic type ($s=1,\, z=0$), which corresponds physically to gate

1039: voltage fluctuations due to Nyquist noise. Case (c) represents a bath

1040: with an excitation gap (for example optical phonons), with a spectrum

1041: given by an inverted parabola. In the limit of infinitely small spectral

1042: bandwidth, it would correspond to the single harmonic oscillator (Einstein

1043: mode) discussed above. Case (b), with a bath spectrum rising as $\omega ^{3}$,

1044: corresponds to the experimentally relevant case of piezoelectric coupling

1045: to acoustic phonons, which was determined to be the major inelastic

1046: mechanism in the experiments of Ref.~\onlinecite{KouwenhouwenDD} on double-dots

1047: in GaAs (see Ref.~\onlinecite{BrandesKramer} for a theoretical analysis

1048: deriving this spectrum for wavelengths larger than the dot distance).

1049: The spectra for the first three cases (a,b,c) have been chosen to

1050: give the same renormalization factor, $z=1/e$. The same figure shows

1051: the resulting functions $P(E)$ and $P_{-}(E)$. These have been obtained

1052: using the integral equation described in Refs.~\onlinecite{minnhagen,ingoldhabil}.

1053: We recall that the low-energy behaviour of $P(E)$ is given by $\left\langle \hat{\phi }\hat{\phi }\right\rangle _{E}=\left\langle \hat{F}\hat{F}\right\rangle _{E}/E^{2}$

1054: for the cases with $z\neq 0$, where perturbation theory may be applied.

1055: In case (c), the alternating signs of the different contributions

1056: to $P_{-}(E)$ may be observed, whose physical meaning has been explained

1057: above for the limiting case of the harmonic oscillator.

1058: 

1059: We now briefly mention some numerical estimates for the bath strengths

1060: as they may occur in experimental situations.

1061: 

1062: In GaAs, the lack of inversion symmetry leads to piezoelectric fields

1063: proportional to the lattice deformation, whose effect on electrons

1064: at low frequencies is much stronger than that of the usual deformation

1065: potential (where it is only the \emph{potential} that is proportional

1066: to the deformation). For the piezoelectric coupling\cite{gant} to

1067: acoustic phonons in GaAs, one finds (compare Ref.~\onlinecite{BrandesKramer})

1068: $\left\langle \hat{\phi }\hat{\phi }\right\rangle _{\omega }^{T=0}=W\, \omega /(c_{s}/d)^{2}$

1069: for $\omega \ll c_{s}/d$, where $c_{s}\approx 5\cdot 10^{3}\, m/s$

1070: is an estimate for the average velocity of longitudinal sound waves

1071: in GaAs, and $d$ denotes the distance between the quantum dots. We

1072: obtain $W=const\cdot (eh_{14}/4\pi )^{2}/(\hbar \rho c_{s}^{3})$,

1073: where $eh_{14}=1.4\, eV/nm$ is the single piezo-electric modulus

1074: in the cubic $T_{d}$ structure of $GaAs$ and $\rho =5.3\cdot 10^{3}\, kg/m^{3}$

1075: the mass density. The numerical constant is of order $1$ and accounts

1076: for the details of the sound wave dispersion relation as well as the

1077: orientation of the crystal axes with respect to the vector separating

1078: the quantum dots. Inserting these values, $W$ is found to be on the

1079: order of $0.01$. In order to obtain the renormalization factor $z$,

1080: the spectrum $\left\langle \hat{\phi }\hat{\phi }\right\rangle _{\omega }$

1081: must be integrated over all frequencies (see above), i.e. up to the

1082: cutoff frequency $\omega _{c}$. The effective cutoff frequency $\omega _{c}\propto c_{s}/d_{0}$

1083: is determined by the extent $d_{0}$ of the dot wave functions (for

1084: $d_{0}=100nm$ one obtains $\omega _{c}\sim 50\, GHz$). Given the

1085: present values, and assuming $d_{0}\approx d$, this leads to estimates

1086: for $\int \left\langle \hat{\phi }\hat{\phi }\right\rangle _{\omega }d\omega $

1087: on the order of $0.01$, yielding $z=\exp (-\left\langle \hat{\phi }^{2}\right\rangle )$

1088: near $1$. Note that the distance $d$ between the dots cancels in

1089: the estimate for $z$, as long as the cutoff frequency is assumed

1090: to be given by $\omega _{c}\propto c_{s}/d$. However, as $\omega _{c}$

1091: might be considerably larger than $c_{s}/d$ (if $d_{0}\ll d$), one

1092: could also obtain a $z$ that deviates more strongly from unity.

1093: 

1094: For the Ohmic bath, we may imagine the quantum dots placed inside

1095: a capacitor $C$ connected to a circuit of resistance $R$, such that

1096: the potential difference $2\hat{F}$ between the dots would be given

1097: by the fluctuating voltage drop across the capacitor. This leads to

1098: a bath spectrum $\left\langle \hat{F}\hat{F}\right\rangle _{\omega }^{T=0}=\pi (R/R_{Q})\hbar ^{2}\omega /(1+(RC\omega )^{2})$,

1099: with $R_{Q}=h/e^{2}$ the quantum of resistance. Therefore, the dimensionless

1100: coupling constant $\alpha $ introduced above would be equal to $\alpha =(\pi /2)R/R_{Q}$,

1101: which can have values both larger and smaller than $1$. 

1102: 

1103: Finally, for optical phonons, we use the Fröhlich interaction Hamiltonian

1104: (Ref. \onlinecite{mahan}) with a dimensionless Fröhlich coupling

1105: constant of $\alpha =0.07$ (GaAs) to obtain the rough estimate $\left\langle \hat{F}\hat{F}\right\rangle _{\omega }^{T=0}=\delta (\omega -\omega _{LO})\cdot (1meV)^{2}\cdot (100nm/d_{0})$,

1106: with $\omega _{LO}\approx 5\cdot 10^{13}Hz$. This yields a $z$ deviating

1107: from unity by about $10^{-3}$.

1108: 

1109: However, in the plots we have chosen $z=1/e$ for illustrative purposes.

1110: 

1111: The resulting behaviour of $\Gamma (\varphi ,V)$ at $T=0$, calculated

1112: from Eq. (\ref{DDGammabasic}), is shown in Figs. \ref{DDGamma0}

1113: and \ref{DDGammaPi}. In the case of constructive interference ($\varphi =0$,

1114: Fig. \ref{DDGamma0}), the decay rate for the {}``weak baths'' (a,b,c)

1115: starts out from $\Gamma /2\Gamma _{0}=z$ at $V=0$ and goes to $\Gamma /2\Gamma _{0}=(1+z^{2})/2$

1116: at $eV/\omega _{c}\gg 1$. The initial deviation from the constant

1117: value of $z$ at low voltages is given by the power-law $V^{2(s-1)}$

1118: contained in Eq. (\ref{DDweakgamma}). In contrast, the decay rate

1119: for the Ohmic bath (d,e) starts at $\Gamma =0$, rising with a power-law

1120: and saturating at a value of $\Gamma /2\Gamma _{0}=1/2$, corresponding

1121: to an equal admixture of odd and even states in the reduced density

1122: matrix of the electron coupled to the bath. For destructive interference

1123: ($\varphi =\pi $, Fig. \ref{DDGammaPi}), the behaviour of (a) and

1124: (b) at low voltages is given by $V^{s-1}$ (see Eq. (\ref{DDweakgamma})),

1125: while the decay rate of the Ohmic bath (d,e) remains the same as that

1126: for $\varphi =0$. In the special case (c) of the gapped bath, we

1127: observe perfect destructive interference up to the excitation threshold

1128: of the bath at $eV=\omega _{c}$, where $\Gamma (\varphi =\pi ,V)$

1129: increases in a stepwise manner for the first time, with the next increase

1130: at $eV=3\omega _{c}$. Note that, on the other hand, $\Gamma (\varphi =0,\, V)$

1131: increases at even multiples of the excitation gap. The difference

1132: comes about because it is only the emission of an odd number of phonons

1133: into the bath that reveals the location of the electron, as discussed

1134: above. This feature would be absent if the two dots were coupled to

1135: two independent baths, whereas the other qualitative properties would

1136: remain the same.

1137: 

1138: From the decay rates at $\varphi =0$ and $\varphi =\pi $, we may

1139: calculate the visibility $\upsilon $ of the {}``interference pattern''

1140: that is defined by the dependence of $\Gamma $ on $\varphi $. The

1141: result is shown in Fig. \ref{DDVis}. As we have noted before, the

1142: visibility is always zero for the Ohmic bath. On the other hand, for

1143: the {}``weak baths'', it is perfect (equal to $1$) at $V\rightarrow 0$,

1144: due to the perfect destructive interference, regardless of the suppression

1145: factor $z$ appearing in $\Gamma (\varphi =0)$. In general, the visibility

1146: decreases towards higher bias voltages before saturating at the limiting

1147: value of $z^{2}$. However, in contrast to intuitive expectation,

1148: the decrease may be nonmonotonous, i.e. the visibility of the interference

1149: effect may actually be enhanced by increasing the supply of energy

1150: available to the electron, although the decay rate $\Gamma $ always

1151: increases monotonously at any $V$. This is particularly striking

1152: in case (c), where the visibility drops down to zero in a certain

1153: range before rising again. The decrease down to the exact value of

1154: $0$ is related to the special choice of $\left\langle \hat{\phi }^{2}\right\rangle =1$

1155: ($z=1/e$), which gives equal strengths of the peak at $E=0$ and

1156: the first peak around $E=\omega _{c}$, which then are able to cancel

1157: in the integral $\gamma _{-}$ over $P_{-}(E)$ that is proportional

1158: to the visibility. However, the physical reason for a dip in visibility

1159: is rather generic: In that energy range, the decay rate $\Gamma $

1160: for $\varphi =\pi $ has already increased due to dephasing, while

1161: the blockade-type suppression of the value of $\Gamma $ for $\varphi =0$

1162: has not yet been lifted. This is a consequence of the even-odd effect

1163: discussed above. 

1164: 

1165: 

1166: \section{Sequential tunneling through the double-dot}

1167: 

1168: \label{DDseqTunnelingSection}Up to now, we have discussed in detail

1169: the influence of the bath on the tunneling decay rate of an electron

1170: which has been placed onto the two dots in the symmetric superposition.

1171: In order to complete the picture, we have to calculate the sequential

1172: tunneling current through such a double-dot interference setup. This

1173: will be done by deriving and solving a master equation for the reduced

1174: density matrix of the double-dot system, taking into account the system-bath

1175: coupling exactly, while the tunnel-coupling is treated in leading

1176: order. We are interested specifically in the nonlinear response, i.e.

1177: in how an increasing bias voltage helps to destroy the phase coherence.

1178: The tunneling rates calculated previously will serve as input to the

1179: master equation.

1180: 

1181: However, in order to facilitate the understanding of the results,

1182: we first turn to a qualitative description of the situation without

1183: the bath.

1184: 

1185: At $\varphi =\pi $, tunneling is completely blocked, since the left

1186: reservoir only couples to the even state $\kt{e}$, while the right

1187: reservoir couples to the antisymmetric (odd) superposition, $\kt{o}$.

1188: At $\varphi =0$, both reservoirs couple to $\kt{e}$, whereas $\kt{o}$

1189: is completely decoupled from the leads (compare the discussion in

1190: Ref.~\onlinecite{BoeseHofstetterSchoeller}). This means that a current

1191: may flow if $\kt{o}$ is empty. However, if $\kt{o}$ is filled, the

1192: current vanishes, because double-occupancy is forbidden in our model.

1193: Since there is no way to change the occupation of $\kt{o}$, the stationary

1194: density-matrix of the double-dot at $\varphi =0$ will be any convex

1195: combination of these two possibilities (at $T=0$, in the absence

1196: of other relaxation paths). At any value of $\varphi $ in between

1197: these extremes, there is always the state $\kt{\Psi }=(\kt{+}-e^{-i\varphi }\kt{-})/\sqrt{2}$,

1198: whose decay into the right lead is blocked by destructive interference.

1199: As there is a nonvanishing overlap between $\kt{\Psi }$ and the state

1200: $\kt{e}$ which is reached by tunneling from the left lead, one will

1201: observe an accumulation of population in $\kt{\Psi }$, until the

1202: current is blocked again. This argument holds at $T=0$, while at

1203: finite temperatures the electron can decay towards the left lead and

1204: make a new attempt. Therefore, in this simple picture, the stationary

1205: current at $T=0$ would be zero at any $\varphi $ except for $\varphi =0$,

1206: where it is undefined. 

1207: 

1208: However, one has to take into account that the coupling to the reservoirs

1209: does not only lead to decay but also to an effective tunnel coupling

1210: between $\kt{+}$ and $\kt{-}$. Although this cannot change the blockade

1211: of the current at $\varphi =\pi $ (leading only to an energy shift

1212: of $\kt{e}$ vs. $\kt{o}$), it does lift the blockade at other values

1213: of $\varphi $. This is because the blocked state $\kt{\Psi }$ is

1214: no longer stationary, such that an electron will not remain there

1215: forever. The degeneracy at $\varphi =0$ still remains. Therefore,

1216: in the ideal case without coupling to a bath, we expect the current

1217: to vanish at $\varphi =\pi $ and to rise towards a maximal amplitude

1218: near $\varphi =0$. According to the previous argument, at $T=0$

1219: this maximal amplitude will be determined by the effective tunnel-coupling

1220: between the dot states. 

1221: 

1222: Introducing the bath will then lead to renormalization effects and

1223: spoil the perfect destructive interference at higher values of the

1224: bias voltage (or temperature), qualitatively in the same way as it

1225: has been explained above. We will show that the actual visibility

1226: $\upsilon _{I}$ of the current interference pattern $I(\varphi )$

1227: is given by a monotonous function of the visibility $\upsilon $ introduced

1228: above for the tunneling rate (at symmetric bias). 

1229: 

1230: We start with the Hamiltonian that is obtained after applying the

1231: unitary transformation of the independent boson model (\ref{IBHprime})

1232: onto our Hamiltonian (\ref{DDhamiltonian}):

1233: 

1234: \begin{equation}

1235: \hat{H}'=\epsilon '(\hat{n}_{+}+\hat{n}_{-})+U'\hat{n}_{+}\hat{n}_{-}+\hat{H}_{B}+\hat{H}_{L}+\hat{H}_{R}+\hat{V}'\, \end{equation}

1236: 

1237: 

1238: Here $\epsilon '$ is the (renormalized) energy of the two states,

1239: which we will take to be $\epsilon '=0$ from now on. $U'$ is the

1240: interaction constant that involves both the Coulomb repulsion as well

1241: as the effective attractive interaction induced by the bath. We assume

1242: $U,U'\gg T,eV$, such that double-occupancy is forbidden. 

1243: 

1244: The term which we will treat as a perturbation is given by $\hat{V}'$,

1245: describing the tunneling to the left and the right leads in the presence

1246: of the bath. It is the transformed version of $\hat{V}$ (compare

1247: Eqs. (\ref{DDtunnelHam}) and (\ref{eq:DDtunnelL}) and Appendix \ref{indepBosonApp}),

1248: where the additional fluctuating phase factors $\exp (\pm i\ph )$

1249: have been introduced:

1250: 

1251: \begin{equation}

1252: \hat{V}'=\sum _{j=l,r}\sum _{\alpha =+,-}\hat{j}_{\alpha }\hat{d}_{\alpha }+h.c.\, ,\label{DDseqtunnVprime}\end{equation}

1253: 

1254: 

1255: where

1256: 

1257: \begin{eqnarray}

1258: \hat{l}_{\pm } & = & e^{\pm i\ph }\hat{l}\\

1259: \hat{l} & = & \sum _{k}t_{k}^{L}\hat{a}_{Lk}^{\dagger }\\

1260: \hat{r}_{+} & = & e^{+i\ph }\hat{r}\\

1261: \hat{r}_{-} & = & e^{-i\ph }e^{i\varphi }\hat{r}\\

1262: \hat{r} & = & \sum _{k}t_{k}^{R}\hat{a}_{Rk}^{\dagger }\, .\label{DDseqrDef}

1263: \end{eqnarray}

1264: 

1265: 

1266: As usual, the current through the device does not only depend on the

1267: rates for electrons to tunnel into and out of the dots, but also on

1268: the stationary state which the system assumes in the nonequilibrium

1269: situation, i.e. under an applied bias voltage.

1270: 

1271: We will now derive a master equation for the reduced density matrix

1272: $\hat{\rho }$ of the double-dot system, which contains the populations

1273: $\rho _{++},\rho _{--},\rho _{00}$ ({}``$0$'' denoting {}``no

1274: electron'') and the coherences $\rho _{+-}$ and $\rho _{-+}$ (with

1275: $\rho _{00}=1-\rho _{++}-\rho _{--}$, $\rho _{\alpha 0}=\rho _{0\alpha }=0$

1276: for $\alpha \neq 0$, and $\rho _{-+}=\rho _{+-}^{*}$). We cannot

1277: simply use the standard kind of master equation, since we have to

1278: deal with two degenerate levels $\kt{+}$ and $\kt{-}$, and it is

1279: important that a tunneling event may create a coherent superposition

1280: of $\kt{+}$ and $\kt{-}$ (for example the even state $\kt{e}$).

1281: Such a master equation - for degenerate levels - has also been employed

1282: in Ref.~\onlinecite{koeniggefen} (without coupling to the bath, and evaluated

1283: in the linear-response regime). The equation is different from that

1284: employed in the {}``orthodox'' theory of sequential tunneling, where

1285: no coherent superpositions are involved. Note that for a finite tunnel-coupling

1286: the levels could be treated as degenerate as long as their energetic

1287: distance is much smaller than the level-broadening due to tunneling.

1288: However, as we consider the limit $\Gamma _{0}\rightarrow 0$, we

1289: need to have exactly equal energies. Otherwise, the energy of the

1290: hole that is created in the left electrode would betray the dot state

1291: which the electron has entered, thus preventing any coherent superposition

1292: to form.

1293: 

1294: Given the initial reduced density matrix $\hat{\rho }(0)$, and assuming

1295: the state of the environment (bath and reservoirs) to be independent

1296: of the electronic state on the dot at $t=0$, we obtain the time-evolution

1297: $\hat{\rho }(t)$ by tracing over the environmental degrees of freedom

1298: ({}``$E$''):

1299: 

1300: \begin{eqnarray}

1301: \hat{\rho }(t) & = & tr_{E}[\hat{T}e^{-i\int _{0}^{t}ds\, \hat{V}'(s)}\hat{\rho }(0)\otimes \hat{\rho }_{E}\tilde{\hat{T}}e^{i\int _{0}^{t}ds\, \hat{V}'(s)}]\nonumber \\

1302:  & = & \hat{\rho }(0)-\nonumber \\

1303:  &  & \int _{0}^{t}dt_{1}\int _{0}^{t_{1}}dt_{2}\, tr_{E}[\hat{V}'(t_{1})\hat{V}'(t_{2})\hat{\rho }(0)\otimes \hat{\rho }_{E}+h.c.]\nonumber \\

1304:  &  & +\int _{0}^{t}dt_{1}\int _{0}^{t}dt_{2}\, tr_{E}[\hat{V}'(t_{1})\hat{\rho }(0)\otimes \hat{\rho }_{E}\hat{V}'(t_{2})]\, \, +\ldots 

1305: \end{eqnarray}

1306: 

1307: 

1308: Physically, by using factorized initial conditions, we neglect correlations

1309: between subsequent tunneling events which could be due to excitations

1310: in the electrodes or in the bath: Since the tunneling rate is very

1311: small, these excitations will have traveled away from the double-dot

1312: until the next event takes place. The entanglement between electron

1313: and bath (discussed in the previous sections) would preclude factorized

1314: initial conditions, if it were not treated indirectly in this approach

1315: (via the unitary transformation). Note that we do not have to make

1316: any secular approximation at this point, unlike the usual derivation

1317: of a master equation \cite{Blum}. It turns out that all contributions

1318: only depend on the time-difference $t_{1}-t_{2}$ anyway, because

1319: the dot levels are degenerate. Therefore, in the long-time limit $t\rightarrow \infty $,

1320: the integration over $(t_{1}+t_{2})/2$ results in a factor $t$,

1321: and the endpoints of the integrals over $t_{1}-t_{2}$ may be extended

1322: to $\infty $. This yields the desired master equation that will determine

1323: the stationary $\hat{\rho }$, as well as the current, in the limit

1324: of weak tunnel coupling. 

1325: 

1326: In the expectation values of products $\hat{V}'\hat{V}'$ only those

1327: contributions remain which combine $\hat{d}_{\alpha }\hat{j}_{\alpha }$

1328: (tunneling out of the dots) with $\hat{j}_{\beta }^{\dagger }\hat{d}_{\beta }^{\dagger }$

1329: (tunneling onto the dots):

1330: 

1331: \begin{eqnarray}

1332: \frac{d\hat{\rho }}{dt}=-\sum _{\alpha ,\beta ,j}\int _{0}^{\infty }ds\, \left\{ \hat{d}_{\alpha }(s)\hat{d}_{\beta }^{\dagger }\hat{\rho }\left\langle \hat{j}_{\alpha }(s)\hat{j}_{\beta }^{\dagger }\right\rangle +h.c.\right. &  & \nonumber \\

1333: \left.+\hat{d}_{\alpha }^{\dagger }(s)\hat{d}_{\beta }\hat{\rho }\left\langle \hat{j}_{\alpha }^{\dagger }(s)\hat{j}_{\beta }\right\rangle +h.c.\right\}  &  & \nonumber \\

1334: +\sum _{\alpha ,\beta ,j}\int _{-\infty }^{+\infty }ds\, \left\{ \hat{d}_{\alpha }(s)\hat{\rho }\hat{d}_{\beta }^{\dagger }\left\langle \hat{j}_{\beta }^{\dagger }\hat{j}_{\alpha }(s)\right\rangle +\right. &  & \nonumber \\

1335: \left.\hat{d}_{\alpha }^{\dagger }(s)\hat{\rho }\hat{d}_{\beta }\left\langle \hat{j}_{\beta }\hat{j}_{\alpha }^{\dagger }(s)\right\rangle \right\} \, . &  & 

1336: \end{eqnarray}

1337:  (Note that there is no minus sign from fermion operator re-ordering

1338: in this factorization of dot and reservoir part, as the reservoir

1339: fermion operators are dragged past an even number of dot operators;

1340: compare e.g. \cite{schoellerRTRG}; alternatively, it is also possible

1341: to define them as commuting operators, since there is no interaction

1342: between them). We get for the individual matrix elements (for brevity,

1343: the summation over $j=l,r$ is implied):

1344: 

1345: \begin{eqnarray}

1346: \dot{\rho }_{++} & = & -\rho _{++}\int _{-\infty }^{+\infty }ds\, \left\langle \hat{j}_{+}^{\dagger }(s)\hat{j}_{+}\right\rangle \nonumber \\

1347:  &  & +\rho _{00}\int _{-\infty }^{+\infty }ds\, \left\langle \hat{j}_{+}\hat{j}_{+}^{\dagger }(s)\right\rangle \nonumber \\

1348:  &  & -\rho _{-+}\int _{0}^{\infty }ds\, \left\langle \hat{j}_{+}^{\dagger }(s)\hat{j}_{-}\right\rangle -h.c.\, ,\label{DDrhoppeqGeneral}

1349: \end{eqnarray}

1350: 

1351: 

1352: \begin{eqnarray}

1353: \dot{\rho }_{+-} & = & -\rho _{+-}\int _{0}^{\infty }ds\, \left\langle \hat{j}_{+}^{\dagger }(s)\hat{j}_{+}\right\rangle \nonumber \\

1354:  &  & -\rho _{+-}\int _{0}^{\infty }ds\, \left\langle \hat{j}_{-}^{\dagger }\hat{j}_{-}(s)\right\rangle \nonumber \\

1355:  &  & +\rho _{00}\int _{-\infty }^{+\infty }ds\, \left\langle \hat{j}_{-}\hat{j}_{+}^{\dagger }(s)\right\rangle \nonumber \\

1356:  &  & -\rho _{++}\int _{0}^{\infty }ds\, \left\langle \hat{j}_{+}^{\dagger }\hat{j}_{-}(s)\right\rangle \nonumber \\

1357:  &  & -\rho _{--}\int _{0}^{\infty }ds\, \left\langle \hat{j}_{+}^{\dagger }(s)\hat{j}_{-}\right\rangle \, .\label{DDrhopmEqGeneral}

1358: \end{eqnarray}

1359: 

1360: 

1361: The equation for $\rho _{--}$ follows from that for $\rho _{++}$

1362: by interchanging indices $+$ and $-$. 

1363: 

1364: Now we have to evaluate environment correlators, such as the prefactor

1365: of $\rho _{++}$ in the second equation (e.g. for $j=r$):

1366: 

1367: \begin{equation}

1368: \left\langle \hat{r}_{+}^{\dagger }\hat{r}_{-}(s)\right\rangle =e^{i\varphi }\left\langle e^{-i\ph }e^{-i\ph (s)}\right\rangle \left\langle \hat{r}^{\dagger }\hat{r}(s)\right\rangle \, .\label{eq:rrphiphi}\end{equation}

1369: 

1370: 

1371: By introducing the bare tunneling rates $\Gamma _{R(L)0}=2\pi \, D_{R(L)}\left\langle \left|t_{k}^{R(L)}\right|^{2}\right\rangle $

1372: (compare Eq. (\ref{DDbaretunnelrate})), we get, using Eq. (\ref{DDseqrDef})

1373: (remember $\hat{r}$ \emph{creates} a reservoir electron):

1374: 

1375: \begin{equation}

1376: \left\langle \hat{r}^{\dagger }\hat{r}(s)\right\rangle =\frac{\Gamma _{R0}}{2\pi }\int d\epsilon \, (1-f_{R}(\epsilon ))\, e^{+i\epsilon s}\, .\end{equation}

1377: 

1378: 

1379: Here we have neglected any energy-dependence of the tunnel-coupling

1380: and electrode DOS, assuming the relevant voltages and temperatures

1381: to be sufficiently small (but see below). The bath correlator in (\ref{eq:rrphiphi})

1382: evaluates to $\exp (-\left\langle \ph \ph (s)\right\rangle -\left\langle \ph ^{2}\right\rangle )$,

1383: which can be expressed by using the definition (\ref{DDPEDef}) for

1384: $P_{-}(\omega )$. There, we have to set $s\mapsto -s$ because of

1385: the reversed order in the $\ph $-correlator:

1386: 

1387: \begin{equation}

1388: e^{-\left\langle \ph \ph (s)\right\rangle -\left\langle \ph ^{2}\right\rangle }=\int d\omega \, P_{-}(\omega )e^{i\omega s}\, .\end{equation}

1389: 

1390: 

1391: Therefore, we obtain:

1392: 

1393: \begin{eqnarray}

1394: \int _{0}^{\infty }ds\, \left\langle \hat{r}_{+}^{\dagger }\hat{r}_{-}(s)\right\rangle = &  & \nonumber \\

1395: e^{i\varphi }\frac{\Gamma _{R0}}{2}\int d\epsilon \, (1-f_{R}(\epsilon ))\, \tilde{P}_{-}^{*}(-\epsilon )\, , &  & \label{DDexamplerr}

1396: \end{eqnarray}

1397: 

1398: 

1399: with

1400: 

1401: \begin{eqnarray}

1402: \tilde{P}_{-}(\epsilon )=\frac{1}{\pi }\int d\omega \, P_{-}(\omega )\int _{0}^{\infty }ds\, e^{i(\epsilon -\omega )s}= &  & \nonumber \\

1403: P_{-}(\epsilon )+\frac{i}{\pi }\int d\omega \, \frac{P_{-}(\omega )}{\epsilon -\omega }\, . &  & 

1404: \end{eqnarray}

1405: 

1406: 

1407: The integral in the second line is understood as a principal-value

1408: integral. In order to abbreviate expressions like this, we introduce

1409: the following definitions for the effective in- and out-tunneling

1410: rates as well as the effective tunnel couplings generated by the electrodes:

1411: 

1412: \begin{eqnarray}

1413: \gamma _{L(-)} & \equiv  & \Gamma _{L0}\int d\epsilon \, (1-f_{L}(\epsilon ))\, P_{(-)}(-\epsilon )\label{DDgammaseconddef}\\

1414: \gamma _{L(-)}^{in} & \equiv  & \Gamma _{L0}\int d\epsilon \, f_{L}(\epsilon )\, P_{(-)}(\epsilon )\\

1415: \Delta _{L} & \equiv  & -\frac{\Gamma _{L0}}{\pi }\int _{-\infty }^{\Lambda }d\epsilon \, (1-f_{L}(\epsilon ))\, \int d\omega \, \frac{P_{-}(\omega )}{\epsilon +\omega }\\

1416: \tilde{\gamma }_{L-} & \equiv  & \gamma _{L-}[P\mapsto \tilde{P}]=\gamma _{L-}+i\Delta _{L}\, .\label{DDgammatildedef}

1417: \end{eqnarray}

1418: 

1419: 

1420: Analogous definitions hold for $L\mapsto R$. Eq. (\ref{DDgammaseconddef})

1421: is equivalent to the definition (\ref{DDgammaDefinition}) used for

1422: $\gamma _{(-)}$ in previous sections. Note that the effective tunnel

1423: coupling $\Delta _{L(R)}$ depends on $P_{-}$, because it arises

1424: from transitions between the states $\kt{+}$ and $\kt{-}$, via an

1425: intermediate lead state. In the expression for $\Delta _{L(R)}$,

1426: the energy-dependence of the density of states and the tunnel coupling

1427: to the reservoir electrode should be kept in order to have a convergent

1428: integral. We will take this into account by introducing an effective

1429: upper energy cutoff $\Lambda $ in the integral. Using these definitions,

1430: (\ref{DDexamplerr}) is equal to $\exp (i\varphi )\tilde{\gamma }_{R-}^{*}/2$.

1431: 

1432: One might wonder why the effective tunnel couplings $\Delta _{L(R)}$

1433: do depend on the occupation of electron states in the reservoirs.

1434: After all, in the non-interacting case, it is possible to calculate

1435: such a change of the effective single-particle Hamiltonian prior to

1436: filling in the electron states. Alternatively, in a calculation that

1437: already takes into account occupation factors, there would be two

1438: contributions which add up to an integral that does not depend on

1439: the Fermi function. However, we consider the interacting case $U=\infty $,

1440: such that (even without the bath) one of these contributions is missing

1441: (since it would involve intermediate states with double occupancy).

1442: The resulting logarithm is analogous to that which appears in the

1443: Kondo problem. This effective tunnel coupling has also been discussed

1444: in Ref.~\onlinecite{BoeseDD}, for the case without a bath. There, the upper

1445: cutoff $\Lambda $ was provided by the Coulomb coupling $U$, since

1446: for higher energies double-occupancy is no longer forbidden and the

1447: non-interacting case takes over (where two contributions arise that

1448: cancel each other). If we take the limit $U\rightarrow \infty $,

1449: then $\Lambda $ will be set by a cutoff in the tunnel matrix elements

1450: (or the electron reservoir's density of states).

1451: 

1452: The general master equation for the reduced density matrix of the

1453: double-dot, derived in the limit of weak tunnel coupling but arbitrary

1454: electron-bath coupling, follows by inserting the definitions (\ref{DDgammaseconddef})-(\ref{DDgammatildedef})

1455: into Eqs. (\ref{DDrhoppeqGeneral}) and (\ref{DDrhopmEqGeneral}):

1456: 

1457: \begin{eqnarray}

1458: \dot{\rho }_{++} & = & -\rho _{++}(\gamma _{L}+\gamma _{R})\nonumber \\

1459:  &  & +\rho _{00}(\gamma _{L}^{in}+\gamma _{R}^{in})\nonumber \\

1460:  &  & -\frac{\rho _{-+}}{2}(e^{i\varphi }\tilde{\gamma }_{R-}+\tilde{\gamma }_{L-})-h.c.\, ,\label{DDrhoppeq}

1461: \end{eqnarray}

1462: 

1463: 

1464: \begin{eqnarray}

1465: \dot{\rho }_{--} & = & -\rho _{--}(\gamma _{L}+\gamma _{R})\nonumber \\

1466:  &  & +\rho _{00}(\gamma _{L}^{in}+\gamma _{R}^{in})\nonumber \\

1467:  &  & -\frac{\rho _{+-}}{2}(e^{-i\varphi }\tilde{\gamma }_{R-}+\tilde{\gamma }_{L-})-h.c.\, ,\label{DDrhommeq}

1468: \end{eqnarray}

1469: 

1470: 

1471: \begin{eqnarray}

1472: \dot{\rho }_{+-} & = & -\rho _{+-}(\gamma _{L}+\gamma _{R})\nonumber \\

1473:  &  & +\rho _{00}(e^{i\varphi }\gamma _{R-}^{in}+\gamma _{L-}^{in})\nonumber \\

1474:  &  & -\frac{\rho _{++}}{2}(e^{i\varphi }\tilde{\gamma }_{R-}^{*}+\tilde{\gamma }_{L-}^{*})\nonumber \\

1475:  &  & -\frac{\rho _{--}}{2}(e^{i\varphi }\tilde{\gamma }_{R-}+\tilde{\gamma }_{L-})\, .\label{DDrhopmEq}

1476: \end{eqnarray}

1477: 

1478: 

1479: The ingredients of the master equation obtained here may be interpreted

1480: as follows: 

1481: 

1482: One part of the right hand side corresponds to the unitary time-evolution

1483: generated by the effective tunneling Hamiltonian

1484: 

1485: \begin{equation}

1486: \hat{H}_{eff}^{T}=\frac{1}{2}(e^{i\varphi }\Delta _{R}+\Delta _{L})\kt{+}\left\langle -\right|+h.c.\, .\end{equation}

1487: 

1488: 

1489: Furthermore, the in-tunneling contributions in the equations for $\rho _{++}$

1490: and $\rho _{--}$ depend on $P(E)$, while that for $\rho _{+-}$

1491: is determined by $P_{-}(E)$, since it describes the creation of a

1492: coherent superposition of $\kt{+}$ and $\kt{-}$ (which is hindered

1493: by the bath). This term would be absent in the usual master equation.

1494: In particular, if $\gamma _{L-}^{in}\rightarrow \gamma _{L}^{in}$,

1495: which will be the case at $T=0$ for vanishing bias between the dots

1496: and the left electrode, an electron tunneling from the left lead will

1497: end up in the coherent superposition where $\rho _{+-}=\rho _{++}=\rho _{--}$.

1498: Taking into account that we are working in a transformed basis, this

1499: describes just the entangled state (\ref{DDentangled}), confirming

1500: the starting point of our earlier discussion. Note that the out-tunneling

1501: contribution for $\rho _{++}$ also depends on $\rho _{+-}$, for

1502: example. This reflects the fact that a superposition between the two

1503: states may be blocked from decaying into the lead, while each state

1504: separately can decay.

1505: 

1506: The stationary density matrix is obtained by demanding $d\hat{\rho }/dt=0$

1507: (and using the relations $\rho _{00}=1-\rho _{++}-\rho _{--}$ and

1508: $\rho _{-+}=\rho _{+-}^{*}$). This will give us the density matrix

1509: in zeroth order $\Gamma _{0}^{0}$ in the bare tunnel coupling, which

1510: we need to calculate the current in leading order $\Gamma _{0}^{1}$.

1511: 

1512: We can obtain the current from the contribution of the left electrode

1513: to the change $\dot{\rho }_{++}+\dot{\rho }_{--}$ in the double-dot

1514: occupation (i.e. keeping only terms that stem from the left electrode

1515: in the master equation). This is equal to the right-going current

1516: in the stationary limit:

1517: 

1518: \begin{eqnarray}

1519: \frac{I}{e}=(\dot{\rho }_{++}+\dot{\rho }_{--})_{L}= &  & \nonumber \\

1520: 2\rho _{00}\gamma _{L}^{in}-\gamma _{L}(\rho _{++}+\rho _{--})-2\gamma _{L-}Re[\rho _{+-}]\, . &  & \label{DDcurrentexpression}

1521: \end{eqnarray}

1522: 

1523: 

1524: An alternative way of deriving the current would be to start from

1525: the general Meir-Wingreen formula\cite{meirwingreen} which expresses

1526: the current in terms of the exact Green's functions of the double

1527: dot, to be calculated in presence of the tunnel-coupling and the bath.

1528: This has been the approach of Ref.~\onlinecite{koeniggefen} for the case

1529: without the bath, and we have checked (\ref{DDcurrentexpression})

1530: to give the same result in that case.

1531: 

1532: 

1533: \section{Evaluation of the sequential tunneling current and the visibility}

1534: 

1535: \label{sec:Evaluation-of-the}In order to evaluate the current as

1536: a function of temperature $T$, bias voltage $V$ and phase difference

1537: $\varphi $, we will now specialize to the case of symmetric bias

1538: and left-right symmetric tunnel couplings ($\Gamma _{R0}=\Gamma _{L0}=\Gamma _{0}$).

1539: All essential features (in particular the perfect destructive interference

1540: in absence of the bath) are independent of this assumption. We will

1541: find that the current is symmetric under $\varphi \mapsto -\varphi $

1542: even for the nonlinear response considered here, due to the symmetry

1543: of the model (compare Ref.~\onlinecite{koeniggefen} for a systematic analysis

1544: of phase-locking in a variety of interference geometries).

1545: 

1546: %

1547: \begin{figure}

1548: \begin{center}\includegraphics[  height=7cm]{Ishape.eps}\end{center}

1549: 

1550: 

1551: \caption{\label{DDIshapeFigure}The current $I$ for different values of the

1552: visibility $\upsilon =\gamma _{-}/\gamma =0.8,\, 0.9,\, 0.99,\, 0.999,\, 0.9999$

1553: (from top to bottom). The limits $\varphi \rightarrow 0$ and $\upsilon \rightarrow 1$

1554: do not commute. Other parameters held fixed: $\lambda =e^{-\beta \mu }=0.2$

1555: and $\delta _{L}=\delta _{R}=-1$. }

1556: \end{figure}

1557: 

1558: 

1559: We find from Eqs. (\ref{DDgammaseconddef})-(\ref{DDgammatildedef}),

1560: using $f(\epsilon )=1-f(-\epsilon )$:

1561: 

1562: \begin{equation}

1563: \gamma _{R(-)}=\gamma _{L(-)}^{in}=\gamma _{(-)}\equiv \Gamma _{0}\int d\epsilon \, f(\epsilon -\mu )\, P_{(-)}(\epsilon )\, ,\end{equation}

1564: 

1565: 

1566: where $\mu =eV/2$ is the chemical potential of the left reservoir.

1567: This is definition (\ref{DDgammaDefinition}), with $eV$ replaced

1568: by $\mu =eV/2$ (since we deal with the symmetric bias case). Furthermore,

1569: we use the condition of detailed balance, $P_{(-)}(-E)=\exp (-\beta E)\, P_{(-)}(E)$

1570: (see, for example, Ref.~\onlinecite{ingoldhabil}), which leads to

1571: 

1572: \begin{equation}

1573: \gamma _{L(-)}=\gamma _{R(-)}^{in}=e^{-\beta \mu }\gamma _{(-)}\, .\end{equation}

1574: 

1575: 

1576: The effective tunnel couplings are still different (because of the

1577: different Fermi distributions):

1578: 

1579: \begin{equation}

1580: \Delta _{L(R)}=-\frac{\Gamma _{0}}{\pi }\int _{-\infty }^{\Lambda }d\epsilon \, f(-(\epsilon \mp \mu ))\int d\omega \, \frac{P_{-}(\omega )}{\epsilon +\omega }\, .\end{equation}

1581: 

1582: 

1583: The lower sign belongs to the right electrode. 

1584: 

1585: For the special case of $T=0$, electrons always enter from the left

1586: and go to the right, such that we have $\gamma _{L}=\gamma _{L-}=\gamma _{R}^{in}=\gamma _{R-}^{in}=0$

1587: and $\gamma _{R(-)}=\gamma _{L(-)}^{in}=\gamma _{(-)}$, with

1588: 

1589: \begin{equation}

1590: \gamma _{(-)}=\Gamma _{0}\int _{0}^{\mu }d\epsilon \, P_{(-)}(\epsilon )\, .\end{equation}

1591: 

1592: 

1593: The effective tunnel couplings are, at $T=0$:

1594: 

1595: \begin{equation}

1596: \Delta _{L(R)}=-\frac{\Gamma _{0}}{\pi }\int d\omega \, P_{-}(\omega )\, \ln \left[\frac{\Lambda +\omega }{|\mu \pm \omega |}\right]\, .\end{equation}

1597: 

1598: 

1599: Note that, without any bath present, $\Delta _{L(R)}$ will have a

1600: logarithmic singularity at $\mu \rightarrow 0$, for $T=0$. The upper

1601: cutoff $\Lambda $ will be given by the minimum of the Coulomb repulsion

1602: energy $U$ and the bandwidth of the reservoir's electronic energy

1603: band (or by some cutoff in the tunnel matrix elements). For the purposes

1604: of our discussion, we assume $\Lambda \gg \mu ,\omega $.

1605: 

1606: In the the limit of high bias voltages ($\omega \ll \Lambda ,\mu $),

1607: we obtain effective tunnel couplings whose magnitude goes as $z^{2}$

1608: and decreases logarithmically with increasing $\mu $:

1609: 

1610: \begin{equation}

1611: \Delta _{L}\approx \Delta _{R}\approx -\frac{\Gamma _{0}}{\pi }\ln \left[\frac{\Lambda }{\mu }\right]\int d\omega \, P_{-}(\omega )\, =-z^{2}\frac{\Gamma _{0}}{\pi }\ln \left[\frac{\Lambda }{\mu }\right]\, .\end{equation}

1612: 

1613: 

1614: By solving the master equation for the stationary density matrix and

1615: inserting the result into Eq. (\ref{DDcurrentexpression}), we obtain

1616: the expression for the current through the double dot in terms of

1617: all of the quantities mentioned previously. In general (at arbitrary

1618: $T$), it is found that the current may be written as the product

1619: of $\gamma $ with a dimensionless function of the phase difference

1620: $\varphi $ and the ratios $\upsilon =\gamma _{-}/\gamma $, $\delta _{L(R)}=\Delta _{L(R)}/\gamma $

1621: and $\beta \mu $:

1622: 

1623: \begin{equation}

1624: I=e\gamma \, I_{0}[\varphi ,\beta \mu ,\upsilon ,\delta _{L},\delta _{R}]\, .\end{equation}

1625: 

1626: 

1627: The complete expression for $I_{0}$ is very cumbersome, although

1628: it may be found analytically by straightforward solution of the master

1629: equation (it is listed for $T=0$ in Appendix \ref{DDseqappendix}).

1630: Therefore, let us first discuss the situation without coupling to

1631: a bath. In that case, we obtain 

1632: 

1633: \begin{equation}

1634: \delta _{L}=\delta _{R}\equiv \delta =-\frac{\Gamma _{0}}{\pi }\int _{-\infty }^{\Lambda }\frac{d\epsilon }{\epsilon }\, f(\mu -\epsilon )\, \end{equation}

1635:  and $\gamma =\gamma _{-}=\Gamma _{0}f(-\mu )$. The current turns

1636: out to be (with $\lambda \equiv e^{-\beta \mu }$):

1637: 

1638: \begin{equation}

1639: \frac{I}{e\gamma }=\frac{4(1-\lambda )(\delta ^{2}+\lambda )\cos ^{2}(\frac{\varphi }{2})}{3\delta ^{2}+2(1+\lambda +\lambda ^{2})+3\delta ^{2}\cos (\varphi )}\, .\label{DDidealcurrent}\end{equation}

1640: 

1641: 

1642: Several points should be noticed about this expression: Firstly, the

1643: destructive interference at $\varphi =\pi $ remains perfect regardless

1644: of temperature, because there are no current-carrying states at all.

1645: At zero temperature ($\lambda =0$), the maximal amplitude of the

1646: current is $I_{max}/e\gamma =2\delta ^{2}/(3\delta ^{2}+1)$, which

1647: vanishes when the effective tunnel coupling $\delta $ goes to zero.

1648: This has been explained above as a consequence of the possible transition

1649: into a current-blocking state, which can only be undone by the effective

1650: tunnel coupling. At finite temperatures ($\lambda >0$), the maximal

1651: current is nonzero even for $\delta \rightarrow 0$, where it approaches

1652: the value of $I_{max}/e\gamma =2\lambda (1-\lambda )/(1+\lambda +\lambda ^{2})$.

1653: This has a maximum at around $T\sim \mu $. It vanishes for larger

1654: temperatures as $\mu /T$, which is to be expected for tunneling through

1655: a localized level (decreasing derivative of the Fermi function). In

1656: addition, the shape of $I(\varphi )$ depends on $\delta $ and $\lambda $,

1657: with a sharper minimum at $\varphi =\pi $ in the case of larger $|\delta |$.

1658: In the limit of $\delta \rightarrow 0$, the current becomes a pure

1659: cosine. At finite temperatures (as well as for $\upsilon \neq 1$)

1660: the behaviour is similar, except for the finite amplitude of the current

1661: at $\delta \rightarrow 0$.

1662: 

1663: Now we turn to the situation including the bath. The general expression

1664: for the current is very lengthy, and we will omit it here. However,

1665: it turns out that the maximal and minimal current are functions merely

1666: of $\upsilon $ and $\lambda =e^{-\beta \mu }$, while they are independent

1667: of $\delta _{L,R}$.

1668: 

1669: The amplitude of the minimal current (at $\varphi =\pi $) is given

1670: by

1671: 

1672: \begin{equation}

1673: \frac{I(\varphi =\pi )}{e\gamma }=\frac{2(1+\lambda )(1-\lambda ^{2})(1-\upsilon ^{2})}{3(1+\lambda )^{2}+(1-\lambda )^{2}\upsilon ^{2}}\, ,\end{equation}

1674: while the maximal current (at $\varphi =0$) is

1675: 

1676: \begin{equation}

1677: \frac{I(\varphi =0)}{e\gamma }=\frac{2}{3}(1-\lambda )\, .\label{DDmaxcurrentgeneral}\end{equation}

1678: It should be noted that the expression (\ref{DDidealcurrent}) for

1679: the current in the ideal case seems to contradict this simple formula.

1680: However, that is because the limits $\varphi \rightarrow 0$ and $\upsilon \rightarrow 1$

1681: do not commute. This is shown in Fig. \ref{DDIshapeFigure}. It means

1682: that for $T=0$ and $\delta _{L,R}\rightarrow 0$ the maximal current

1683: calculated according to (\ref{DDmaxcurrentgeneral}), which is independent

1684: of $\delta _{L,R}$, and the {}``typical'' amplitude of the current

1685: ($\propto \delta _{L}^{2}$) may deviate strongly. The peculiar behaviour

1686: near $\varphi =0$ seems to be connected to the physical degeneracy

1687: of the case $\varphi =0,\, \upsilon =1$ which has been discussed

1688: above.

1689: 

1690: From these formulas, we obtain the visibility, defined in terms of

1691: the current:

1692: 

1693: \begin{equation}

1694: \upsilon _{I}\equiv \frac{I(\varphi =0)-I(\varphi =\pi )}{I(\varphi =0)+I(\varphi =\pi )}\, .\end{equation}

1695: 

1696: 

1697: It can be expressed entirely by the visibility $\upsilon $ defined

1698: previously in terms of the tunneling rates (Eqs. (\ref{eq:visDef}),

1699: (\ref{eq:visGamm})), as well as the temperature-dependent factor

1700: $\lambda =e^{-\beta \mu }$ ($\mu =eV/2$):

1701: 

1702: \begin{equation}

1703: \upsilon _{I}=\frac{2(1+\lambda +\lambda ^{2})\upsilon ^{2}}{3(1+\lambda )^{2}-(1+4\lambda +\lambda ^{2})\upsilon ^{2}}\, .\label{DDvisibilityI}\end{equation}

1704: 

1705: 

1706: This is a monotonous mapping of $\upsilon $ to the interval $[0,1]$,

1707: with only a weak dependence on $\lambda $. The other parameters $\delta _{L},\delta _{R}$

1708: only modify the amplitude and shape of the current pattern $I(\varphi )$.

1709: Therefore, all the statements about the visibility made in the previous

1710: discussion of the tunneling decay out of the symmetric superposition

1711: continue to hold up to this monotonous transformation (and with $eV$

1712: replaced by $\mu =eV/2$). In particular, at $T=0$, we have

1713: 

1714: \begin{equation}

1715: \upsilon _{I}=\frac{2\upsilon ^{2}}{3-\upsilon ^{2}}\, .\end{equation}

1716: 

1717: 

1718: %

1719: \begin{figure}

1720: \begin{center}\includegraphics[  height=7cm]{visAll.eps}\end{center}

1721: 

1722: 

1723: \caption{\label{DDvisIvsMu}The visibility $\upsilon _{I}$ of the pattern

1724: $I(\varphi )$, for piezoelectric coupling to acoustic phonons (b)

1725: (solid line) and for the optical phonon bath (c) (dashed line), plotted

1726: vs. $\mu =eV/2$, at different temperatures $T/\omega _{c}=0.01,\, 0.05,\, 0.1,\, 0.2,\, 0.4,\, 0.5$

1727: (top to bottom). Inset depicts energy diagram for tunneling in this

1728: situation.}

1729: \end{figure}

1730: The dependence of the visibility $\upsilon _{I}$ on the bias voltage

1731: $eV=2\mu $, the temperature $T$ and the bath spectrum is displayed

1732: in Fig. \ref{DDvisIvsMu}, for bath spectra of type (b) and (c). The

1733: decrease of $\upsilon _{I}$ at $\mu =0$ with increasing temperature

1734: $T$ in case (b) is well approximated by Eq. (\ref{DDvisfiniteT})

1735: for $\upsilon (T,V\rightarrow 0)$ (employing the relation $\upsilon _{I}=\upsilon ^{2}/(2-\upsilon ^{2})$

1736: for $\mu =0$). (The functions $P_{(-)}(E)$ for finite temperatures

1737: have been calculated numerically using the fast Fourier transform,

1738: from the defining equation (\ref{DDPEDef})).

1739: 

1740: Note that for bath spectra with $z=0$ (i.e. exponent $s\leq 1$ at

1741: $T=0$ and $s\leq 2$ at $T>0$) the visibility vanishes entirely

1742: (at any $V$), as has been explained in the previous sections. We

1743: have already pointed out that this picture is expected to change if

1744: one treats the tunnel-coupling to higher order. However, we have to

1745: leave this analysis for the future. One possible approach to a nonperturbative

1746: (but still approximate) treatment of both the tunnel-coupling and

1747: the system-bath coupling at the same time seems to be the numerical

1748: {}``real-time renormalization group'' scheme \cite{schoellerRTRG}.

1749: 

1750: 

1751: \section{Conclusions}

1752: 

1753: We have analyzed dephasing in tunneling through two parallel single-level

1754: quantum dots with a fluctuating energy difference between the dots.

1755: The disappearance of perfect destructive interference in a symmetric

1756: setup has been taken as a criterion for {}``genuine'' dephasing,

1757: as opposed to mere renormalization. The coupling to the bath has been

1758: taken into account exactly, via the {}``independent boson model''

1759: and the concepts of the {}``$P(E)$ theory'' of tunneling in a dissipative

1760: environment, while the tunnel coupling has been treated in leading

1761: order. 

1762: 

1763: We have discussed in detail the behaviour of the density matrix of

1764: a single electron that has been placed in a superposition of the two

1765: dot levels. The bath measures (to some extent) the position of the

1766: electron, such that the electron's density matrix becomes mixed. However,

1767: this allows direct conclusions about the {}``incoherent current''

1768: only in the limit of high bias voltages, corresponding to a fast {}``projection''

1769: measurement of the electron's state. For lower voltages, only the

1770: low-frequency part of the bath spectrum contributes to the lifting

1771: of destructive interference. Thus, for any {}``weak bath'', whose

1772: spectrum falls off fast towards low frequencies, the visibility of

1773: the interference effect becomes perfect in the limit of low bias voltages

1774: $V$ and temperatures $T$, when the energy supplied to the electron

1775: is vanishingly small. This is the case for a fluctuation spectrum

1776: $\propto \omega ^{s}$ with $s>1$ ($s>2$) for $T=0$ ($T>0$). The

1777: visibility may show a nonmonotonous behaviour as a function of bias

1778: voltage. For {}``stronger'' spectra (smaller exponent $s$), including

1779: the Ohmic bath ($s=1$), there is the well-known zero-bias anomaly

1780: (suppression of the tunneling current at low voltages), which affects

1781: equally both the cases of constructive and destructive interference.

1782: Therefore, the visibility vanishes exactly at any bias voltage in

1783: our approach, where the tunnel coupling has been treated only in leading

1784: order. Although there is always a suppression of the magnitude of

1785: the tunnel current for the case of constructive interference, this

1786: may be interpreted as a mere renormalization of the effective tunnel-coupling,

1787: since the perfect destructive interference is not affected and since

1788: it occurs even for a bath with an excitation gap. The full dependence

1789: of the sequential tunneling current $I(\varphi )$ on voltage, temperature,

1790: bath spectrum and phase difference $\varphi $ between the interfering

1791: paths has been derived by setting up a master equation for the state

1792: of the double-dot (which is special due to the degeneracy of dot levels). 

1793: 

1794: The major questions that have remained open in our analysis are related

1795: to the behaviour at stronger tunnel coupling. In particular, the perfect

1796: destructive interference may also be overcome by correlated tunneling

1797: of several particles (with an intermediate {}``virtual'' excitation

1798: of the bath), and this process will therefore contribute to dephasing,

1799: although it is expected to be suppressed strongly at low voltages

1800: and temperatures. Likewise, the visibility for the Ohmic bath (or

1801: other strong baths), which turns out to be zero in the present approximation,

1802: may be changed at low bias voltages and temperatures comparable to

1803: the tunneling rate. This will require other methods to analyze the

1804: competition between strong tunnel coupling and system-bath coupling.

1805: 

1806: \begin{acknowledgments}

1807: We thank J. König, H. Schoeller, D. Loss and H. Grabert for useful

1808: discussions. This work has been supported through the Swiss NSF and

1809: the Swiss NCCR for Nanoscience.

1810: \end{acknowledgments}

1811: \appendix

1812: 

1813: 

1814: \section{Independent boson model}

1815: 

1816: \label{indepBosonApp}For reference purposes, we describe here the

1817: canonical transformation employed in the independent boson model.

1818: See Ref.\onlinecite{mahan} for more details (concerning the case

1819: of at most a single particle). Consider a set of electronic levels

1820: $j$ that couple to bath operators $\hat{F}_{j}$ which are assumed

1821: to be linear in the coordinates (and momenta) of a bath of harmonic

1822: oscillators, $\hat{H}_{B}$:

1823: 

1824: \begin{equation}

1825: \hat{H}=\sum _{j}(\varepsilon _{j}+\hat{F}_{j})\hat{n}_{j}+\hat{H}_{B}\, .\label{indepbosonham}\end{equation}

1826: 

1827: 

1828: Here $\varepsilon _{j}$ is the unperturbed level energy and $\hat{n}_{j}=\hat{d}_{j}^{\dagger }\hat{d}_{j}$

1829: is the number of particles on level $j$. The fluctuating fields are

1830: characterized completely by their power spectra at $T=0$, 

1831: 

1832: \begin{equation}

1833: \left\langle \hat{F}_{l}\hat{F}_{j}\right\rangle _{\omega }^{T=0}\equiv \frac{1}{2\pi }\int _{-\infty }^{+\infty }dt\, e^{i\omega t}\, \left\langle \hat{F}_{l}(t)\hat{F}_{j}\right\rangle ^{T=0}\, .\end{equation}

1834: 

1835: 

1836: Here we will restrict ourselves to the case where the different variables

1837: commute, $[\hat{F}_{l},\hat{F}_{j}]=0$. As a consequence, the spectrum

1838: $\left\langle \hat{F}_{l}\hat{F}_{j}\right\rangle _{\omega }^{T=0}$

1839: is real-valued, but there may still be correlations. 

1840: 

1841: The most straightforward solution proceeds via a unitary transformation\cite{mahan}

1842: (essentially a gauge transformation). One introduces the fluctuating

1843: phases $\ph _{j}$, whose time-derivatives are given by the $\hat{F}_{j}$:

1844: 

1845: \begin{equation}

1846: \dot{\ph }_{j}\equiv i[\hat{H}_{B},\ph _{j}]=-\hat{F}_{j}\, .\label{IBphidef}\end{equation}

1847: 

1848: 

1849: The exponent generating the unitary transformation is defined as:

1850: 

1851: \begin{equation}

1852: \hat{\chi }=\sum _{j}\ph _{j}\hat{n}_{j}\, .\end{equation}

1853: 

1854: 

1855: Applying the transformation to the Hamiltonian in Eq. (\ref{indepbosonham})

1856: yields:

1857: 

1858: \begin{equation}

1859: \hat{H}'=e^{-i\hat{\chi }}\hat{H}e^{+i\hat{\chi }}=\sum _{j}\varepsilon _{j}\hat{n}_{j}-\sum _{lj}J_{lj}\hat{n}_{l}\hat{n}_{j}+\hat{H}_{B}\, .\label{IBHprime}\end{equation}

1860: 

1861: 

1862: The coupling between system and bath has been eliminated, resulting

1863: in an effective interaction between particles on the different levels,

1864: with:

1865: 

1866: \begin{equation}

1867: J_{lj}=\int _{0}^{\infty }d\omega \, \frac{\left\langle \hat{F}_{l}\hat{F}_{j}\right\rangle _{\omega }^{T=0}}{\omega }\, .\label{IBcoupling}\end{equation}

1868: 

1869: 

1870: The $J_{lj}$ are real-valued and independent of temperature. For

1871: $l=j$ they describe energy shifts of single-particle levels. The

1872: canonical transformation also changes the particle annihilation and

1873: creation operators,

1874: 

1875: \begin{equation}

1876: \hat{d}_{j}'=e^{-i\hat{\chi }}\hat{d}_{j}e^{+i\hat{\chi }}=e^{i\ph _{j}}\hat{d}_{j}\, ,\label{IBdtransform}\end{equation}

1877: 

1878: 

1879: and $\hat{d}_{j}'^{\dagger }=\hat{d}_{j}^{\dagger }e^{-i\ph _{j}}$.

1880: This will affect all Green's functions and, therefore, also the time-evolution

1881: of the single-particle density matrix. In addition, it becomes important

1882: if a tunneling part is added to the Hamiltonian, where the operators

1883: $\hat{d}_{j}^{(\dagger )}$ appear, such that they have to be transformed

1884: according to (\ref{IBdtransform}). However, since the phases $\hat{\phi }_{j}$

1885: and the particle operators $\hat{d}_{j}^{(\dagger )}$ commute (even

1886: at different times, when evolved according to $\hat{H}'$), the evaluation

1887: of Green's functions always splits into a part referring to the particles

1888: and a separate average over the bath operators. This is the major

1889: simplification brought about by the {}``diagonal coupling'' between

1890: system and bath.

1891: 

1892: 

1893: \section{Current expression for sequential tunneling through the double-dot}

1894: 

1895: \label{DDseqappendix}At $T=0$, for the symmetric situation, the

1896: current $I$ is given by $I=e\gamma \, I_{0}[\upsilon ,\delta _{L},\delta _{R}]$,

1897: with:

1898: 

1899: \begin{eqnarray}

1900: I_{0}[\upsilon ,\delta _{L},\delta _{R}] & = & 2\cdot [-\delta _{L}^{2}+(\upsilon ^{2}-1)(1+\delta _{R}^{2})+\nonumber \\

1901:  &  & 2\delta _{L}\delta _{R}(\upsilon ^{2}-1)\cos \varphi +\delta _{L}^{2}\upsilon ^{2}\cos ^{2}\varphi ]\cdot \nonumber \\

1902:  &  & [-3\delta _{L}^{2}+2\delta _{L}\delta _{R}\upsilon ^{2}+(1+\delta _{R}^{2})(\upsilon ^{2}-3)+\nonumber \\

1903:  &  & 2(\upsilon ^{2}(1+\delta _{L}^{2}+\delta _{R}^{2})+\delta _{L}\delta _{R}(\upsilon ^{2}-3))\cos \varphi +\nonumber \\

1904:  &  & \delta _{L}(\delta _{L}+2\delta _{R})\upsilon ^{2}\cos ^{2}\varphi ]^{-1}

1905: \end{eqnarray}

1906: 

1907: 

1908: \begin{thebibliography}{10}

1909: \bibitem{mohantywebb}P. Mohanty, E.~M.~Q. Jariwala, and R.~A. Webb, Phys. Rev. Lett.

1910: \textbf{78}, 3366 (1997).

1911: \bibitem{cedraschi}P. Cedraschi, V. V. Ponomarenko, and M. Büttiker, Phys. Rev. Lett.

1912: \textbf{84}, 346 (2000).

1913: \bibitem{buettrevb}P. Cedraschi and M. Büttiker, Phys. Rev. B \textbf{63}, 165312 (2001);

1914: Annals of Physics, \textbf{289}, 1 (2001).

1915: \bibitem{GZ_PB}A. D. Zaikin and D. S. Golubev, Physica B \textbf{280}, 453 (2000).

1916: \bibitem{OurABring}F. Marquardt and C. Bruder, Phys. Rev. B \textbf{65}, 125315 (2002).

1917: \bibitem{NagaevBuettikerHO}K. E. Nagaev and M. Büttiker, Europhys. Lett. \textbf{58}, 475 (2002).

1918: \bibitem{Guinea}F. Guinea, Phys. Rev. B \textbf{65}, 205317 (2002).

1919: \bibitem{GSZ_modelsLowTDeph}D.~S.~Golubev, G.~Schön, and A.~D.~Zaikin, cond-mat/0208548 (2002).

1920: \bibitem{ImryInexistenceZPDP}Y. Imry, cond-mat/0202044 (2002).

1921: \bibitem{sai}A. Stern, Y. Aharonov, and Y. Imry, Phys. Rev. A \textbf{41}, 3436

1922: (1990).

1923: \bibitem{koeniggefen}J. König and Y. Gefen, Phys. Rev. Lett. \textbf{86}, 3855 (2001);

1924: Phys. Rev. B \textbf{65}, 045316 (2002). 

1925: \bibitem{KouwenhouwenDD}T. Fujisawa et al., Science \textbf{282}, 932 (1998).

1926: \bibitem{StoofNazarov}T. H. Stoof and Y. V. Nazarov, Phys. Rev. B \textbf{53}, 1050 (1996).

1927: \bibitem{brune}Ph. Brune, C. Bruder, and H. Schoeller, Phys. Rev. B \textbf{56},

1928: 4730 (1997); Physica E \textbf{1}, 216 (1997).

1929: \bibitem{ziegler}R. Ziegler, C. Bruder, and H. Schoeller, Phys. Rev. B \textbf{62},

1930: 1961 (2000).

1931: \bibitem{BrandesKramer}T. Brandes and B. Kramer, Phys. Rev. Lett. \textbf{83}, 3021 (1999).

1932: \bibitem{aguado}R. Aguado and L. P. Kouwenhoven, Phys. Rev. Lett. \textbf{84}, 1986

1933: (2000).

1934: \bibitem{Debald}S. Debald, T. Brandes, and B. Kramer, Phys. Rev. B \textbf{66}, 041301

1935: (2002).

1936: \bibitem{KeilSchoellerRTRG}M. Keil and H. Schoeller, cond-mat/0205308 (2002).

1937: \bibitem{akera}H. Akera, Phys. Rev. B \textbf{47}, 6835 (1993).

1938: \bibitem{shahbazyan}T. V. Shahbazyan and M. E. Raikh, Phys. Rev. B \textbf{49}, 17123

1939: (1994).

1940: \bibitem{KotthausDD}A. W. Holleitner et al., Science \textbf{297}, 70 (2002).

1941: \bibitem{BoeseDD}D. Boese, W. Hofstetter, and H. Schoeller, Phys. Rev. B \textbf{64},

1942: 125309 (2001).

1943: \bibitem{LossShenjaDD}D. Loss and E. V. Sukhorukov, Phys. Rev. Lett. \textbf{84}, 1035 (2000).

1944: \bibitem{mahnsoo}M.-S. Choi, C. Bruder, and D. Loss, Phys. Rev. B \textbf{62}, 13569

1945: (2000).

1946: \bibitem{BoeseHofstetterSchoeller}D. Boese, W. Hofstetter, and H. Schoeller, Phys. Rev. B 66, 125315

1947: (2002).

1948: \bibitem{HauleBonca}K. Haule and J. Bonca, Phys. Rev. B \textbf{59}, 13087 (1999).

1949: \bibitem{key-2}L. G. Mourokh, N. J. M. Horing, and A. Yu. Smirnov, Phys. Rev. B \textbf{66},

1950: 085332 (2002).

1951: \bibitem{key-1}E. Buks et al., Nature \textbf{391}, 871 (1998).

1952: \bibitem{aleinerwhichpath}I. L. Aleiner, N. S. Wingreen, and Y. Meir, Phys. Rev. Lett. \textbf{79},

1953: 3740 (1997).

1954: \bibitem{ingoldhabil}G.-L. Ingold and Y. V. Nazarov, in: \emph{Single Charge Tunneling},

1955: ed. by H. Grabert and M. Devoret, NATO ASI Series B, vol. 294, Plenum,

1956: New York (1992).

1957: \bibitem{SchoenPE}G. Schön, in: T. Dittrich et al., \emph{Quantum transport and dissipation},

1958: Wiley-VCH, Weinheim (1998).

1959: \bibitem{mahan}G. D. Mahan: \emph{Many-Particle Physics}, Plenum, New York (1981),

1960: pp. 269-310.

1961: \bibitem{devoret}M. H. Devoret et al., Phys. Rev. Lett. \textbf{64}, 1824 (1990).

1962: \bibitem{minnhagen}P. Minnhagen, Phys. Lett. A \textbf{56}, 327 (1976).

1963: \bibitem{gant}V.~F. Gantmakher and Y.~B. Levinson, \emph{Carrier scattering in

1964: metals and semiconductors}, North-Holland, Amsterdam (1987), § 3.5. 

1965: \bibitem{Blum}K. Blum, \emph{Density Matrix Theory and Applications}, Plenum, New

1966: York (1996).

1967: \bibitem{schoellerRTRG}H. Schoeller, Lect. Notes Phys. \textbf{544}, 137 (2000). (cond-mat/9909400)\bibitem{meirwingreen}Y. Meir and N. Wingreen, Phys. Rev. Lett. \textbf{68}, 2512 (1992).

1968: \end{thebibliography}

1969: 

1970: \end{document}

1971: