1: %&LaTeX

2: \documentstyle[preprint,prb,aps,epsfig]{revtex} 

3: \begin{document}

4: \tighten

5: \title{An Electromechanical Which--Path Interferometer}

6: \author{A. D. Armour\cite{auth}} 

7: \address{The Blackett Laboratory, Imperial College of Science,

8: Technology and Medicine, London SW7~2BZ, United Kingdom}

9: \author{M. P. Blencowe\cite{auth2}}

10: \address{Department of Physics and Astronomy, Dartmouth College,

11: Hanover, New Hampshire 03755}

12: \date{\today}

13: \maketitle

14: \begin{abstract}

15: We investigate the possibility of an electromechanical which--path 

16: interferometer, in 

17: which electrons travelling through an 

18: Aharonov--Bohm  

19: ring incorporating a quantum dot in one of the arms are dephased by an interaction 

20: with the fundamental flexural mode of a 

21: radio frequency cantilever. The cantilever is positioned so that its tip

22:  lies just above the dot and a bias is applied so that an electric field

23:   exists between the dot and the tip. This electric field is modified when an 

24:   additional electron hops onto the dot, coupling the 

25:    flexural mode of the cantilever and the microscopic electronic 

26:   degrees of freedom. We analyze the transmission properties of this system 

27:   and the dependence of interference fringe visibility on the cantilever--dot 

28:   coupling and on the mechanical properties of the cantilever. 

29:   The fringes are progressively 

30:   destroyed as the interaction with the cantilever is turned up, in 

31:   part due to dephasing arising  from the entanglement of the electron and 

32:   cantilever states and also due to the thermal smearing that results from 

33:   fluctuations in the state of the cantilever. When the dwell time of the 

34:   electron on the dot is comparable to or longer than the cantilever period,

35:   we find coherent 

36:   features in the transmission amplitude. These features are washed out 

37:   when the cantilever is decohered by its coupling to the environment. 

38:   \end{abstract}

39: \pacs{PACS numbers: 85.85.+j, 85.35.Ds, 73.63.Kv, 73.23.Hk, 03.65.Yz }

40: 

41: \section{Introduction}

42: 

43: Which--path devices such as the canonical

44:  two--slit interference experiment, where a  measurement is made of the

45: path a particle takes,  

46:  have a long 

47: history going back as far as the early debates about 

48: complementarity.\cite{hist,Fey} 

49: Recent interest in their investigation has been stimulated

50: partly by advances in experimental

51: techniques, which have lead to the realization of several different

52: varieties of which--path systems in the laboratory,\cite{Buks,op} and

53: partly by accompanying developments in the theory of quantum

54: measurement.\cite{SS,Scully,Walls} However, it is  the

55: realization that a which--path experiment provides a very convenient

56: model system for developing and testing fundamental ideas about

57: decoherence in mesoscopic systems,\cite{Stern,hacr} that has  increased the

58: level of interest within the solid state physics community in particular.

59: 

60: 

61: Which--path experiments in solid state systems were recently pioneered

62: by Buks {\emph{et al.}}\cite{Buks} The solid state analog

63:  of the two--slit interference experiment is the measurement of the oscillations

64:   in the current passing  through an Aharonov--Bohm (AB) ring as a function 

65:   of the applied magnetic field. The path taken by an electron may be probed by 

66:   placing a measuring device close to one of the arms. Buks {\emph{et al.}}

67:    incorporated a quantum dot in one of the arms in order to slow the electrons 

68:    down, with a neighboring quantum point contact (QPC) serving as a 

69:    which--path detector.

70: An electron travelling around the arm of the ring containing the dot 

71:  dwells on the dot for a finite amount of time before 

72: moving on. The presence of the dot alone does not destroy 

73: the coherence of the electron transport through the ring, so long as the dwell time 

74: of the electrons is sufficiently short,\cite{Yac} but it does provide time for the 

75: electron to interact with an external measuring device. A QPC adjacent to the dot 

76: functions as a measuring device since it can be biased so that its 

77: conductance is very sensitive to changes in the occupancy of the dot: the 

78: passage of an electron via the path including the dot leaves behind which--path 

79: information in the QPC device, although actual knowledge of which path 

80: an electron took, so--called `true which--path information', can only be obtained 

81: via further measurement.\cite{Hack} In accord with theoretical

82: predictions,\cite{Le,Al,Gurvitz} the experiment demonstrated that

83:  the interference fringes are degraded when the interaction 

84: between the electrons and the measuring device is sufficiently 

85: sensitive

86: for information to be obtained that would help, even in principle,

87:  to determine which of the two possible paths an electron

88: took.    

89: 

90: The experiment of Buks {\emph{et al.}} has close parallels with 

91: the well--known thought experiment (see, e.g., Sec. 1-6 of Ref.\ 

92: \onlinecite{Fey}), in which 

93: a light source

94: is used to detect through which slit an electron passes in a  two--slit

95: interference experiment. In both cases dephasing is effected by

96: scatterers, each of which interact once, and once only, with the interfering

97: particle and whose interactions with each other may safely be ignored.

98: For the electron--light scattering scheme,\cite{Fey} the scatterers are photons

99: which probe the electron's state directly, whereas in the

100: solid--state experiment the scatterers are electrons in the QPC

101: which interact with the electron on the dot via electrostatic coupling. 

102: However, because the electron has a finite dwell time on the

103: dot, it has time to interact with  more than one scatterer in the QPC, 

104: so that in this case dephasing can be achieved via a series of 

105: very weak interactions rather than a single strong event, as is the

106: case in the photon thought experiment.

107: 

108: 

109: The present work is concerned with a variation on the system 

110: considered by Buks 

111: {\emph{et al.}}, 

112: in which a radio frequency mechanical cantilever is used, rather than a

113: QPC, to determine which path an electron takes. A coupling  between electrons 

114: residing on the dot (which is again on one of the AB interferometer arms) and 

115:  the

116: cantilever, whose tip is suspended over the dot, can be set up by developing a 

117: uniform electric field between the tip of the cantilever and the base of the dot. 

118: Electrons on the dot couple approximately linearly to the cantilever position, thus leading 

119: to a coupling between the flexural phonon modes of the cantilever and the

120: occupancy of the dot. Furthermore, because the coupling strength decays rapidly 

121: with increasing frequency  

122: and  also the flexural mode spectrum of a cantilever 

123: is quite sparse, it turns out that for micron scale  cantilevers  only the fundamental

124: flexural mode is relevant. Therefore, at low 

125: temperatures the cantilever can be

126: treated as a single quantum mechanical oscillator.  

127: 

128: 

129: As a consequence of the cantilever having effectively only 

130: one degree of freedom,  the electromechanical which--path 

131: interferometer exhibits qualitatively different behavior from the QPC 

132: which--path device. In particular, we 

133: find that the dephasing behavior of 

134: the electron due to the cantilever depends on the 

135: relative magnitudes of the electron dwell time on the dot and the cantilever 

136: period. When the  dwell time is short compared to the cantilever 

137: period, the dephasing occurs in a way analogous to Einstein's recoiling slit

138: thought--experiment.\cite{hist,SS} In contrast, when the dwell time of

139: the electron on the dot is comparable to or longer than the period of the 

140: cantilever,

141: a description of the dephasing in terms of the entanglement of the

142: cantilever and electron states becomes more appropriate than a

143: semiclassical picture of momentum transfer.

144: The effectively harmonic nature of  the cantilever motion

145: means that the degree of entanglement between the cantilever and 

146: electronic states

147: must be periodic, so long as the cantilever interacts only with 

148: the electron on the dot. If the dwell time

149:  of the electrons on the dot could be tuned to the cantilever period, 

150:  then we would 

151:    be able to erase all which--path information held in the state of the 

152:    cantilever. 

153:    This would give us direct 

154:     evidence that the 

155:    coupled cantilever and dot were behaving as a single coherent quantum system. 

156:  However, in practice electron dwell times on a quantum dot have a distribution

157:   of values and so we can only obtain indirect evidence for the quantum 

158:  coherence of the  cantilever--dot system, such as the coherent exchange of 

159:   energy quanta between the electron and the cantilever which give rise to 

160:  side resonances in the elastic transmission amplitude of the device. 

161:  

162:  The environment of the cantilever influences its behavior  in two important ways: 

163:  over short timescales it destroys the cantilever's phase coherence, whilst

164:  over longer timescales it damps the cantilever's motion.

165:  The electromechanical which--path system we propose acts as a probe of 

166:  the cantilever's decoherence due to it's interaction

167:   with the environment. We show that the cantilever's decoherence 

168:   inhibits the coherent

169:   exchange of energy between the cantilever and the dot and hence manifests 

170:   itself by washing out the side resonances in the elastic transmission 

171:   amplitude.  

172:      

173: 

174: 

175: The outline of this paper is as follows. In Sec. II we give our basic 

176: model for the electromechanical

177: which--path device and show that the interference properties of the AB

178: ring, modified to include the dot and cantilever, can be obtained by calculating

179: the elastic transmission amplitude of the arm containing the dot. In 

180: Sec. III 

181: we describe a simple tight--binding model for an electron on a  dot which is linearly

182: coupled to the cantilever (treated as a quantum oscillator) and to

183:  propagating states in the leads. We show that recent work on

184: inelastic resonant tunneling can be adapted to obtain the elastic

185: transmission amplitude. 

186: 

187: We present results for our model in Sec. IV, where we examine

188: dephasing as a function of the dot--cantilever coupling in both

189: the regime where the electron dwell time is short compared to the

190: cantilever period and  where they are comparable. 

191: 

192: In Sec. V we examine the influence of the cantilever's

193: environment on the  dephasing produced by the cantilever when the

194: dwell time is comparable to the period of the cantilever. We obtain a modified

195: expression for the transmission through the arm with the dot when the 

196: cantilever is coupled to an environment which consists of a bath of oscillators.  

197: 

198: 

199: In Sec. VI we outline the principal practical constraints on an

200: electromechanical which--path device and discuss its feasibility 

201: using currently available technology. We draw our conclusions in Sec. 

202: VII and discuss ways in which the present work could be extended.

203: Appendix A contains a detailed, classical analysis of the cantilever--dot 

204: coupling and the flexural mode spectrum of the cantilever which underpin 

205: the simplified model employed in the main text. Finally, appendix B contains 

206: details of a general calculation of the effect of the cantilever environment on

207: the transmission amplitude of the dot. 

208: 

209: \section{Theoretical Model}

210: 

211: The ultimate goal of our analysis of the electromechanical which--path device

212: is to obtain an expression for how

213: the magnetoresistance of the AB ring varies as a function of the coupling

214: between the cantilever and an electron on the dot. In particular, we

215: want to determine how  the amplitude and phase of the

216: oscillation in the current through the device as the magnetic field is

217: varied (i.e., the interference fringes) depend on the voltage between the

218: cantilever and the dot. 

219: In order to simplify the analysis, we will consider only the case

220: where the magnetoresistance is measured as an average over a 

221: time  much longer than any other timescale in the

222: problem.    

223:    

224: Our model builds  on the theoretical analysis of the

225: QPC which--path experiment,  carried out by a number of

226: groups.\cite{Le,Al,Gurvitz}  However,

227: there are also close parallels between the system we

228: consider and a quantum optical system analyzed recently,\cite{BJK,Bose} in which 

229: radiation is used to drive an oscillator--mounted mirror into non--classical states. 

230: 

231: Since we are interested in the behavior of the system averaged over

232: time, it is sufficient to work within the Landauer framework, so that

233: our task of obtaining the average magnetoresistance is equivalent to

234: calculating the transmission characteristics of the device.\cite{Datta}

235: The simplest way of thinking about the transmission probability through  

236:  the AB ring is in terms of a two--slit experiment: the total 

237: transmission probability, $\mathcal{T}$,  is given by the coherent sum of 

238: amplitudes for 

239: transmission through the left and right arms, $t_l$ and $t_r$,

240: \begin{equation}

241: {\mathcal{T}}=|t_l+t_r|^2.\label{sm}

242: \end{equation}

243: This approach was employed by Yacoby {\emph{et al.}}\cite{Yac} in their 

244: analysis of the first experiments on AB rings with a quantum dot 

245: included in one arm. However, in that case this simple 

246: formalism proved

247: inadequate as it neglects the contribution of electrons which have

248: undergone multiple reflections. The contribution to the transmission

249: of multiply reflected electrons leads to the enforcement of the

250: Onsager--type relation for the conductance from source (S) to 

251: drain (D): $G_{DS}(B)=G_{DS}(-B)$, which must apply for a

252: two--terminal device.\cite{But1,But2,Imry,Yac2}

253: If the AB ring is modified to absorb electrons which are scattered 

254: backwards in the device, then the Onsager--type condition is no longer satisfied 

255: and the simple--minded formula (\ref{sm}) is valid. Buks 

256: {\emph{et al.}} incorporated  just such a modification in 

257: their AB ring and we shall limit our analysis to this case.    

258:  We write 

259: $\tilde{t}_{\mathrm QD}$ for the transmission through the arm containing the dot and 

260: $t_0$ for the other amplitude. The presence of a magnetic flux, $\Phi$, induces 

261: an additional  relative phase shift between the two paths and so the 

262: transmission probability in the absence of the cantilever can be written as

263: \begin{equation}

264: {\mathcal{T}}=\left|t_0+\tilde{t}_{\mathrm QD}{\mathrm{e}}^{i 

265: 2\pi\Phi/\Phi_0}\right|^2

266: ={\mathcal{T}}^{(0)}+2{\mathrm{Re}}\left[ t^*_0\tilde{t}_{\mathrm QD}{\mathrm{e}}^{i 

267: 2\pi\Phi/\Phi_0}\right], \label{T}

268: \end{equation}

269: where $\Phi_0$ is the flux quantum and ${\mathcal{T}}^{(0)}$ is the 

270: flux independent term of the transmission probability. In practice, 

271: the transmission through the arm with the dot is much

272: less than that through the arm without the dot,  

273: $|\tilde{t}_{\mathrm QD}|\ll|t_0|$, and so ${\mathcal{T}}_0\approx |t_0|^2$.

274: 

275: When there is a non--zero interaction between the cantilever and the

276: electron on the dot, we must  explicitly include the fact that the

277: transmission depends on the initial  state of the

278: cantilever. Since the cantilever contains a macroscopic number 

279: of atoms we assume that

280: its initial state can always be described as a thermal mixture. Such a

281: procedure will necessarily lead us to neglect short time correlations

282: between the cantilever states, but this will be unimportant if the

283: magnetoresistance is averaged over a time which is long compared with

284: the characteristic timescale for the thermalization of the

285: cantilever's motion.\cite{steady} With the initial state of the 

286: cantilever assumed to be thermal we have,

287: \begin{equation}

288: {\mathcal{T}}=\sum_i\sum_f \rho_i\left[ \langle i|t_0 

289: +\hat{t}_{\mathrm QD}{\mathrm{e}}^{i 2\pi\Phi/\Phi_0} | f\rangle \left(\langle

290:   i|t_0 +\hat{t}_{\mathrm QD}{\mathrm{e}}^{i

291: 2\pi\Phi/\Phi_0} | f\rangle\right)^*\right], \label{T2} 

292: \end{equation}

293: where $\hat{t}_{\mathrm QD}$ is an operator on cantilever states only, $\rho_i$ is

294: the usual thermal weight ($\rho_i={\mathrm{e}}^{-\beta

295:   \epsilon_i}/\sum_j{\mathrm{e}}^{-\beta \epsilon_j}$, with

296: $\beta=1/k_{\mathrm{B}}T$ and $\epsilon_i$ the energy of the state),

297:  and we have made no  assumption about the final state

298:   of the cantilever mode. Because only the dot arm interacts with the cantilever

299: the interference term is diagonal in the cantilever modes,

300: \begin{equation}

301: 2{\mathrm{Re}}\left[ \sum_i \rho_i t^*_0\langle i|\hat{t}_{\mathrm QD}|i 

302: \rangle {\mathrm{e}}^{i 2\pi\Phi/\Phi_0}\right]=

303: 2{\mathrm{Re}}\left[t^*_0\langle\tilde{t}_{\mathrm QD}\rangle 

304:  {\mathrm{e}}^{i 2\pi\Phi/\Phi_0}\right],

305: \end{equation}  

306: where $\langle \tilde{t}_{\mathrm QD}\rangle$ is averaged over the cantilever thermal

307: state. Thus, only elastic scattering processes contribute to the 

308: interference term. 

309: 

310: At finite temperatures we also have to average the transmission amplitude for 

311: transport through the arm containing the dot over the Fermi 

312: distribution:\cite{Datta}

313: \begin{equation}

314: \langle\tilde{t}_{\mathrm QD}\rangle

315: =\int_{0}^{\infty} {d}\epsilon\left( -\frac{\partial f}{\partial 

316: \epsilon}\right) \langle t_{\mathrm QD}(\epsilon)\rangle. \label{therm}

317: \end{equation}

318: 

319: 

320: 

321: The knowledge of the transmission amplitude allows us to obtain the

322: amplitude of the periodic oscillations in the current, or in the

323: language of a generic two--slit experiment, the visibility of the

324: interference fringes. The visibility, $v$, in a two--slit experiment is 

325:  defined in terms of the maximum and minimum signal (in this case 

326: current), measured at the peaks and valleys of the fringes,

327:  respectively:\cite{vis}

328: \begin{equation}

329: v=\frac{{\mathrm{max.}}-{\mathrm{min.}}}{{\mathrm{max.}}+{\mathrm{min.}}}.

330: \end{equation}

331: 

332: In the case of the electromechanical which--path device considered here, the 

333: current is proportional to the transmission probability, $\mathcal{T}$, 

334: given in Eq. (\ref{T2}) above. We can write the transmission amplitudes through 

335: the two arms in modulus--argument form, 

336: $\langle\tilde{t}_{\mathrm QD}\rangle=|\langle\tilde{t}_{\mathrm QD}

337: \rangle|{\mathrm{e}}^{i\alpha}$ and 

338: ${t}_0^*=|t_{0}|{\mathrm{e}}^{-i\beta}$, so that the interference 

339: part of the transmission probability takes the form

340: \begin{equation}

341: 2{\mathrm{Re}}\left[t_0^*

342: \langle\tilde{t}_{\mathrm QD}\rangle{\mathrm{e}}^{i2\pi\Phi/\Phi_0}\right]

343: =2|{t}_0||\langle\tilde{t}_{\mathrm QD}\rangle|\cos\left( 2\pi\Phi/\Phi_0+ \alpha-\beta 

344: \right).

345: \end{equation}

346: Hence in this case, the visibility of the fringes is

347: \begin{equation}

348: v=\frac{2|{t}_0||\langle\tilde{t}_{\mathrm QD}\rangle|}{{\mathcal{T}}^{(0)}}

349: \simeq \frac{2|\langle\tilde{t}_{\mathrm QD}\rangle|}{|{t}_0|}. 

350: \end{equation}

351: The phase of the

352: transmission amplitude  determines the phase of the fringe

353: pattern as a function of the magnetic field. Any change in the phase

354: of the transmission amplitude should therefore be detectable as a change in the

355: phase of the whole interference pattern. 

356: 

357: In the absence of the dot, and for the ideal case of a ring in which both arms 

358: are identical, the transmission amplitudes for both arms are 

359: the same so that ${\mathcal{T}}_0=2 |{t}_0|^2$ and the visibility is unity. 

360: This demonstrates the utility of $v$ as a measure:  it does not depend 

361: on the value of the total transmission probability through the device,

362: but just on its interferometric properties.

363: 

364: In practice, a fringe visibility close to unity cannot be

365: achieved. Apart from the obvious difficulty in constructing a ring in which both 

366: arms are identical, there are two further reasons why the `intrinsic' 

367: visibility is less than unity.\cite{Yac} Firstly, there is more than one conduction 

368: channel in each arm and so the transport is never really a `one electron' problem. 

369: Secondly, thermal smearing can play an important r\^{o}le, since the thermal 

370: smearing length is typically comparable to the size of the ring at temperatures

371:  of around 100mK.\cite{Yac} However, for our device 

372:  the reduction in fringe visibility arising from sources 

373: other than the cantilever is irrelevant: all we 

374: require is that there should be a measurable change in the visibility of the 

375: fringes as the electric field coupling the dot and cantilever is turned on.

376: 

377: 

378: \section{Transmission Amplitude}

379: 

380: In order to obtain the transmission amplitude for the arm of the AB

381:  ring containing the dot, we employ a standard tight--binding model for the dot 

382: and the leads to which it is coupled. In

383: the Coulomb blockade regime, the quantum dot is modelled by a single localised 

384: state, at energy $\epsilon_0$, which is coupled to sets of non--interacting,

385:  propagating, states in

386:  the leads. The cantilever is treated as a single quantum oscillator mode of frequency

387:   $\omega_0$, although the generalisation to include a spectrum of modes is 

388:   straightforward, as we discuss below.   

389: The total Hamiltonian of the interferometer arm can be written as the

390: sum of two parts 

391: \begin{equation}

392: {\mathcal{H}}={\mathcal{H}}_0+{\mathcal{H}}_1 \label{Ham1},

393: \end{equation}

394: with non--interacting part

395: \begin{equation}

396: {\mathcal{H}}_0=\epsilon_0\hat{c}^{\dagger}\hat{c}+\sum_k\left( 

397: \epsilon_{kL}\hat{c}^{\dagger}_{kL}\hat{c}_{kL}+\epsilon_{kR}\hat{c}^{\dagger}_

398: {kR}\hat{c}_{kR}\right)+\hbar \omega_0 \hat{a}^{\dagger}\hat{a},

399: \end{equation}

400: and interacting part

401: \begin{equation}

402: {\mathcal{H}}_1=-\lambda \hat{c}^{\dagger}\hat{c}(\hat{a}+\hat{a}^{\dagger})

403: +\sum_k V_{kL}\left( 

404: \hat{c}^{\dagger}_{kL}\hat{c}+\hat{c}^{\dagger}\hat{c}_{kL}\right)

405: +\sum_k V_{kR}\left( 

406: \hat{c}^{\dagger}_{kR}\hat{c}+\hat{c}^{\dagger}\hat{c}_{kR}\right) \label{coup},

407: \end{equation}

408: where $\hat{c}$ and $\hat{a}$ operate on the states of the electron on the 

409: dot and the cantilever, respectively. The states in the left--hand(right--hand) 

410: leads have energies $\epsilon_{kL}$($\epsilon_{kR}$) and are operated on by 

411: $\hat{c}_{kL}$($\hat{c}_{kR}$), where the index $k$ runs over propagating 

412: states in the leads. The matrix elements for hopping onto(off) the dot 

413: are 

414: given by $V_{kL}$($V_{kR}$). The interaction between the electron on

415: the dot and the cantilever is modelled as a linear coupling between the

416: displacement of the flexural mode and the occupation of the dot, since

417: when the electron is not on the dot there is no displacement of the

418: cantilever.\cite{justify} We

419: analyze the interaction between the cantilever and the dot in some detail in 

420: appendix A, where the form of the interaction is derived in terms 

421: of the normal modes of the cantilever. We find that the coupling constant, 

422: $\lambda$, is given by the relation $\lambda=\xi 

423: eE\sqrt{\hbar/2m\omega_0}$ [see Eq.\ (\ref{coupling})], 

424: where $\xi$ is a geometrical factor of order one, $E$ is the electric field 

425: experienced by the additional electron on the dot, $m$ is the cantilever mass 

426: and $e (>0)$ the electronic charge.

427: 

428: 

429: The energy-dependent amplitude, $t_{\mathrm QD}(\epsilon)$, is calculated using

430: the usual  

431: $S$--matrix formalism\cite{W89} employed in transport theory. 

432: The amplitude for transmission from a state of energy $\epsilon$ to one of 

433: energy $\epsilon'$ is given by the element of the 

434: $S$--matrix\cite{W89} linking propagating states in the left and right--hand leads

435: \begin{equation}

436: \langle \epsilon', R| S|\epsilon, L \rangle =t_{\mathrm QD}(\epsilon,\epsilon'),

437: \end{equation} 

438: where $\langle  \epsilon', R|$($|  \epsilon, L\rangle$) is the state in the 

439: right(left)--hand lead with energy $\epsilon'$($\epsilon$). The total 

440: transmission amplitude at energy $\epsilon$ is then obtained by 

441: integrating over the final state energies

442: \begin{equation}

443: t_{\mathrm QD}(\epsilon)=\int_0^{\infty}t_{\mathrm QD}(\epsilon,\epsilon')d 

444: \epsilon'.\label{fermi}

445: \end{equation} 

446: 

447: The $S$--matrix element for a single electron tunneling from left to right 

448: through 

449: a single, localised, state which is coupled to a cantilever mode can be 

450: calculated using either the methods described by Wingreen {\emph{et al.}} 

451: \cite{W89} or those of Glazman and Shekhter.\cite{GS}

452:  For initial and final states of the cantilever

453: given by $\alpha_i$ 

454: and $\alpha_f$ respectively, the relevant matrix element is

455: \begin{eqnarray}

456: \langle \epsilon', \alpha_f, R|S|\epsilon, \alpha_i, L \rangle=&-&i\int\int 

457: \frac{dt_1dt_2}{\hbar^2}{\mathrm{e}}^{-\eta(|t_1|+|t_2|)}\cr

458:  &\times& \langle \epsilon', \alpha_f, R|

459: {\mathrm{e}}^{i{\mathcal{H}}_0t_2/\hbar}{\mathcal{H}}_1 

460: \hat{G}_R(t_2-t_1){\mathcal{H}}_1{\mathrm{e}}^{-i{\mathcal{H}}_0t_1/\hbar}

461: |\epsilon, \alpha_i, L \rangle,

462: \end{eqnarray}

463: where $\hat{G}_R(t)=-i\Theta(t){\mathrm{e}}^{-i{\mathcal{H}}t/\hbar}$

464: and $\eta$ is the usual small positive real number inserted to ensure 

465: convergence. 

466: 

467: We want to calculate the visibility of the fringes and so we need only calculate 

468: the coherent part of the transmission probability, and hence the elastic 

469: transmission amplitude. To evaluate the elastic transmission amplitude we  consider 

470: processes in which the state of the cantilever remains unchanged and so 

471: we write $\alpha_i=\alpha_f=\alpha$ and calculate an average over an ensemble 

472: of 

473: states of the cantilever (see the discussion in Sec. II above). Thus

474: \begin{equation}

475: \langle S \rangle=-V_R(\epsilon)V_L(\epsilon')\int\int 

476: \frac{{d}t_1dt_2}{\hbar^2}{\mathrm{e}}^{-\eta(|t_1|+|t_2|)}

477: \times {\mathrm{e}}^{i(\epsilon' t_2-\epsilon t_1)/\hbar}\Theta(t_2-t_1)\langle 

478: 0,\alpha| \hat{c}(t_2)\hat{c}^{\dagger}(t_1)|0, \alpha \rangle 

479: \end{equation}

480: with 

481: $|V_{L(R)}(\epsilon)|^2=\sum_k|V_{kL(R)}|^2\delta(\epsilon-\epsilon_{kL(kR)})$ 

482: and where  $|0,\alpha\rangle$ is the state with no electrons on the 

483: localised level and  the cantilever in state $\alpha$.

484: 

485: We assume that the coupling to the leads is independent of energy over

486: the range of interest and that it is symmetric so that  we can write

487: \begin{equation}

488: \Gamma=\Gamma_{L(R)}=2\pi|V_{L(R)}|^2.

489: \end{equation}

490: Since we are considering only the elastic part of the transmission

491: amplitude the Green function will be invariant with respect to time.

492:  Hence, we change variables to 

493: $\tau=t_2-t_1$ and $t_0=t_1$, so that the Green function takes the form

494: \begin{equation}

495: G^R(\tau,t_0)=-i\Theta(\tau)\langle 

496: \hat{c}(\tau+t_0)\hat{c}^{\dagger}(t_0)\rangle

497: \end{equation}

498: and the  invariance with respect to translations in time is 

499: equivalent to  the statement that the value of the Green function is 

500: independent of the choice of $t_0$. Averaging over $t_0$, and taking 

501: the limit $\eta \rightarrow 0^+$, 

502: the overall expression for the $S$--matrix element is then

503: \begin{equation}

504: \langle \epsilon', R|S |\epsilon , L\rangle=

505: -i\frac{\Gamma}{2 

506: \pi}\int_0^{\infty} \frac{d\tau}{\hbar} 

507: {\mathrm{e}}^{i\epsilon\tau/\hbar}G^R(\tau) 

508: \times 2\pi\delta(\epsilon-\epsilon').

509: \end{equation}

510: 

511: Thus, the final expression for the transmission amplitude of the dot is 

512: \begin{equation}

513: \langle t_{\mathrm QD}(\epsilon)\rangle=-i\Gamma

514: \int_0^{\infty}\frac{d\tau}{\hbar}

515: {\mathrm{e}}^{i\epsilon\tau/\hbar}G^R(\tau), \label{tqd}

516: \end{equation}

517: where the transmission amplitude is understood to be time--averaged in

518: the sense described in Sec. II, and the 

519: relevant Green function is 

520: \begin{equation}

521: G^R(\tau)=-i\Theta(\tau)\langle\hat{c}(\tau)\hat{c}^{\dagger}(0)\rangle, \label{tqd22} 

522: \end{equation}

523: where the expectation value is over a thermal distribution state of the 

524: cantilever  with no electrons present. This result for the 

525: transmission amplitude  is  very similar to that

526: studied by Aleiner {\emph{et al.}},\cite{Al} in their analysis of the

527: QPC which--path device.

528: 

529: The Green function can be evaluated using either operator algebra

530: techniques,\cite{GS} or many body perturbation theory,\cite{W89} and 

531: one obtains

532: \begin{equation}

533: G^R(\tau)=G^R_0(\tau){\mathrm{e}}^{-\phi(\tau)}, \label{tqd2}

534: \end{equation}

535: where 

536: \begin{equation}

537: G^R_0(\tau)=-i\Theta(\tau){\mathrm{e}}^{(-i\epsilon_0-\Gamma/2)\tau/\hbar},

538: \end{equation}

539: and it has been assumed that the renormalisation of the dot energy,

540: $\epsilon_0$, due to coupling to the leads can be ignored.

541: The factor due to coupling to the cantilever, $\phi(\tau)$, can be obtained by

542: calculating the contribution for the cantilever beginning and ending in a

543: particular coherent state, $|\nu\rangle$, and then carrying out a thermal

544: average over all coherent states with appropriate weightings,

545: \begin{equation}

546: {\mathrm{e}}^{-\phi(\tau)}=

547: {\mathrm{e}}^{i(\lambda/\hbar\omega_0)^2[\omega_0\tau-\sin(\omega_0 \tau)]}

548: {\mathrm{e}}^{-(\lambda/\hbar\omega_0)^2[1-\cos(\omega_0\tau)]}

549: \int\frac{{d}^2\nu}{\pi\overline{n}}{\mathrm{e}}^

550: {(\lambda/\hbar\omega_0)(\nu^*\mu-\nu\mu^*)}{\mathrm{e}}

551: ^{-|\nu|^2/\overline{n}},\label{ephi}

552: \end{equation}

553: where $\mu={\mathrm{e}}^{i\omega_0 t}-1$. 

554: Evaluating the thermal average, we find

555: \begin{equation}

556: \phi(\tau)

557: =\left(\frac{\lambda}{\hbar \omega_0}\right)^2\left\{i[\sin(\omega_0 

558: \tau)-\omega_0 \tau]+[1-\cos(\omega_0 \tau)](1+2\overline{n})\right\}, 

559: \label{ex2}

560: \end{equation}

561: with $\overline{n}$ the thermal occupation of the cantilever mode. 

562: 

563: It is

564: clear that the thermal average leads to a much more rapidly decaying

565: term, due to the extra factor of $2\overline{n}$. If the

566: cantilever remained in a coherent state throughout then it would be far less

567: effective, compared to the thermal state, at reducing the visibility of the 

568: fringes. This is because each coherent state affects the transmission amplitude 

569: in two different ways: the magnitude of the transmission is reduced by an amount 

570: which is independent of which state the cantilever is in and also

571:  a phase shift is induced whose size depends 

572: sensitively on the cantilever state.   

573: When we carry out the thermal average we are in effect averaging over a range

574: of different phase shifts. Such a procedure effectively destroys the

575: interference fringes whenever the thermal state includes contributions

576: in which the phase differs by $\sim 2\pi$, irrespective of the magnitude of the 

577: transmission amplitude for each of the coherent states which constitute the 

578: thermal mixture. The thermal averaging timescale is given by the damping time 

579: of the cantilever which will typically be much 

580: larger than the dwell time of the electron on the dot. In this case, the loss of 

581: fringe visibility arising from thermal smearing is  not due 

582: to which--path detection of individual electrons: a measurement of the  current 

583: averaged over times shorter than the damping time but longer than the 

584: dwell time would resolve  AB fringes with phase fluctuating in time.  In 

585: contrast, the state-independent reduction in the transmission 

586: amplitude would give rise to a reduction in the fringe visibility even for 

587: the time-resolved measurement, signifying which--path detection.   

588: However, since we are

589:  working in a steady state regime in which we consider measurements made over 

590: very long times we will not be able to make an explicit distinction between 

591: which--path detection and thermal smearing in this work.  

592: 

593: 

594: We can repeat this calculation for the more general case of coupling

595: to a whole series of non-interacting cantilever modes. The result has the

596: same basic structure as before, but the cantilever factor, $\phi(\tau)$, 

597: is modified and now takes the form of a sum of the contributions from each

598:  mode,\cite{W89,GS}  

599: \begin{equation}

600: \phi(\tau)=\sum_i \left(\frac{\lambda_i}{\hbar \omega_i}\right)^2

601: \left[i(\sin(\omega_i 

602: \tau)-\omega_i \tau)+(1-\cos(\omega_i 

603: \tau))(1+2\overline{n}_i)\right],\label{phinoenv}

604: \end{equation}

605: where $\lambda_i$, $\omega_i$, and $\overline{n}_i$ are the coupling

606: constant, frequency, and thermal occupation number of the $i$th cantilever

607: mode, respectively. However, we  find that for the purposes of

608: demonstrating the electromechanical which--path device we need only

609: consider a single mode, as is discussed in appendix A.

610: 

611: \section{Results for Isolated Cantilever} 

612: 

613: The important question which we need to answer is the following:

614: When does the cantilever destroy the AB fringes? There are two

615: very different regimes which we can explore by varying the

616: relative sizes of characteristic dwell time of the electron on the dot, 

617: $\tau_d=\hbar/\Gamma$, and the fundamental frequency of the cantilever: 

618: \begin{enumerate}

619: \item $\omega_0\tau_d\ll1$: in this case the  

620: periodic behavior of  the cantilever will not be relevant and the entanglement

621:  built up between the cantilever and electron states during the dwell time will

622:  be irreversible, leading to dephasing. 

623:   

624: \item $\omega_0\tau_d\sim 1$: in this regime the periodicity of the

625:   cantilever means that the 

626:  entanglement of the cantilever and dot states caused by their interaction 

627:  may be partially undone, or erased,

628:  although  the distribution of electron dwell times will make the effect 

629:  impossible to observe directly. 

630: \end{enumerate}

631: 

632: The first regime is easier to analyze theoretically, as the

633: calculations can be done analytically and invite direct comparisons

634: with the sliding--slit  thought experiment. The case where the dwell

635: time of the electron on the dot is comparable to the period of the

636: cantilever has to be analyzed numerically. It is also complicated by

637: the presence of the cantilever's environment which makes the entanglement of the

638: cantilever and dot states irreversible by breaking the overall periodicity in the 

639: expression (\ref{tqd}) for the transmission amplitude. In this

640: section, we calculate the magnitude and phase of the transmission

641: amplitude in both regimes, neglecting the effect of the

642: environment which we address in detail in Sec. V. This approach enables

643: a clear picture to be built up of exactly how the environment affects the 

644: behavior of the cantilever--dot system.    

645: 

646: 

647: \subsection{Regime Where $\omega_0\tau_d\ll1$}  

648: 

649: Since $\omega_0\tau_d\ll1$, we can simplify Eq.\

650: (\ref{tqd}) for

651: the transmission amplitude through the arm with the dot 

652: by expanding the harmonic functions to quadratic order in $\omega_0\tau_d$.

653: We obtain

654: \begin{equation}

655:  \langle t_{\mathrm QD}(\epsilon)\rangle \simeq-\Gamma 

656: \int_0^{\infty}\frac{{d}\tau}{\hbar}

657: {\mathrm{e}}^{[i(\epsilon-\epsilon_0)-\Gamma/2]\tau/\hbar

658: -(eE\Delta x_{\mathrm th}\tau)^2/2\hbar^2}, \label{exp}

659: \end{equation}

660: where $\Delta x_{\mathrm th}=\sqrt{(2\overline{n}+1)(\hbar/2m\omega_0)}$ is the

661: thermal position uncertainty of the cantilever and we have taken the 

662: geometrical factor in the coupling constant (\ref{coupling}) to be unity for simplicity.

663: The integral on the righthand side  can now be evaluated to obtain

664: \begin{equation}

665:  \langle t_{\mathrm QD}(\epsilon)\rangle \simeq -\sqrt{\frac{\pi}{2}}

666: \left(\frac{\Gamma}{eE\Delta x_{\mathrm th}}\right)

667: \exp\left(\frac{[\Gamma/2-i(\epsilon-\epsilon_0)]^2}

668: {2(eE\Delta {x_{\mathrm th}})^2}\right){\mathrm{Erfc}}

669: \left(\frac{\Gamma/2-i(\epsilon-\epsilon_0)}{\sqrt{2} eE\Delta x_{\mathrm th} }\right).

670: \end{equation}

671: 

672: If we ignore, for the moment, the thermal width of the electron energy

673: distribution in the leads, then we can deduce from this expression a 

674: rough criterion for dephasing 

675: in the region close to the electronic  resonance:

676: \begin{equation}

677: eE\Delta x_{\mathrm th}> \Gamma,

678: \label{dephasecond1}

679: \end{equation}

680: or written in another way,

681: \begin{equation}

682: eE\tau_{d}>\frac{\hbar}{\Delta x_{\mathrm th}}.

683: \label{dephasecond2}

684: \end{equation}

685: Practical considerations (see Sec. VI) restrict the dwell time to be 

686: a few ns or less and place an upper bound on the electric field: $E \sim 10^5$V/m. 

687: Thus, the destruction of the 

688: interference fringes requires

689:  $\Delta x_{\mathrm th}>10^{-2}$\AA, a value which is  certainly  achievable.

690: 

691: As previously discussed, 

692: the loss of fringe visibility is not only associated with which--path 

693: detection, but with

694: thermal smearing as well.  

695: We can obtain the condition for which--path detection 

696: by setting the cantilever temperature to zero in Eq. 

697: (\ref{dephasecond2}):

698: \begin{equation}

699: eE \tau_d >\frac{\hbar}{\Delta x_{\mathrm zp}}=2\Delta p_{\mathrm zp},

700: \label{whichpathcond}

701: \end{equation}

702: where $\Delta x_{\mathrm zp}=\sqrt{\hbar/2m\omega_{0}}$ denotes the 

703: zero-point position uncertainty. Because the reduction in the 

704: transmission amplitude due to each 

705: different coherent state is the same [c.f., Eq. (\ref{ephi})], this condition holds 

706: independently of which coherent state the cantilever is in.

707: Thus, we find that for which--path detection to occur the

708: classical impulse delivered to the cantilever during the dwell time must exceed

709: twice the zero--point momentum uncertainty of the cantilever.

710: 

711: 

712: 

713: Our result is  equivalent to that obtained by Bohr in his

714: famous discussion of Einstein's

715: sliding--slit {\emph{gedanken}} experiment.\cite{hist} In that case Bohr argued

716: that in order to

717: detect the passage of an electron through a given slit, the momentum 

718: uncertainty in the

719: slit must be less than the impulse transferred by the passing

720: electron, thus necessitating a corresponding 

721: latitude  in position of the slits (via the uncertainty principle)

722: which in turn  washes out the fringes by causing large phase shifts 

723: (i.e., $\sim 2\pi$) for

724: successive electrons. However, these phase shifts are not associated with

725: any kind of thermal fluctuation, instead they arise from the position uncertainty

726: of the quantum state of the slits.  

727: 

728: 

729:   

730: 

731: At finite temperatures we must take into account

732: not only the thermal state of the cantilever, but also the fact that

733:  the electron energies are  spread over a range $\sim 4k_{\mathrm{B}}T$

734: around the Fermi energy. We must therefore average

735: the transmission amplitude over energy, weighted by the derivative of the

736: Fermi distribution function [see Eq. (\ref{therm})]:

737: \begin{equation}

738: \langle \tilde{t}_{\mathrm QD} \rangle =  \int_{0}^{\infty}{d}\epsilon

739: \frac{\langle t_{\mathrm QD}(\epsilon) \rangle}{4k_{\mathrm{B}}T}

740: {\mathrm{sech}}^2\left( \frac{\epsilon-\epsilon_0}

741: {2k_{\mathrm{B}}T}\right),

742: \end{equation}

743: where we have assumed that the bias across the device is small enough for a 

744: linear response approach to be valid and that the average Fermi energy in the 

745: leads is tuned to the resonance $\epsilon_0$.

746: The effect of this procedure is to reduce both the coherent

747: transmission amplitude itself and the influence the cantilever has on

748: it. The explanation for this unexpected feature can be found by comparing the

749:  transmission amplitude, $\langle t_{\mathrm QD}(\epsilon)\rangle$, at and away 

750:  from the 

751: resonance. Whilst the transmission amplitude at, or close to,  resonance 

752: decays rapidly with increasing 

753: coupling to the cantilever, the situation is reversed when the electron energy

754:  is far from resonance, where the transmission amplitude is {\emph{enhanced}} by

755: the interaction with the cantilever. The reduction in  dephasing efficiency 

756: due to the thermal averaging over the electron 

757: energy distribution makes an interesting contrast to the effect of the 

758: thermal average over cantilever states, which increases the dephasing rate 

759: substantially.

760:       

761: 

762: Fig.~1 illustrates the effect of increasing the dimensionless coupling constant 

763: $\kappa=\lambda/\hbar\omega_0$ on the magnitude of the resonant transmission 

764: amplitude $|\langle t_{\mathrm QD}(\epsilon_0)\rangle|$, with

765: and without averaging over the thermal width of the electron energy

766: distribution in the leads. For this example, we have taken

767: $\omega_0=140$~MHz and $T=$20~mK (giving a thermal occupation 

768: $\overline{n}=18$), $m=8\times10^{-20}$~kg, and assumed 

769:  $\kappa$ to have a maximum value of about 3, consistent

770:  with the analysis of the cantilever--dot coupling in appendix A. 

771:  The effect of the thermal broadening of electron energies in reducing 

772:  the efficiency of the cantilever to cause dephasing is clear.

773:  

774:  It is important to note  that our  

775: treatment is entirely restricted to a one--electron picture of transport. 

776: Such an approach is valid so long as the electron gas in the leads and the dot

777: itself is non-degenerate. When the electron temperature drops towards 

778: zero the interactions between electrons can no longer be ignored and so our

779: model and the results which follow from it become inapplicable. At 

780: $T=0$ and

781: in the weak bias limit, 

782: inelastic transport through the dot would become impossible: with the absence of 

783: unoccupied states below the Fermi level and  the cantilever in 

784: the ground state an electron is unable to exchange energy quanta with 

785: the cantilever.

786:  

787: 

788: \subsection{Regime Where $\omega_0\tau_d\sim 1$}           

789: 

790: In the regime where the dwell time of the electron on the dot

791: approaches the period of the cantilever, the system takes on

792: a rather different character, as the dephasing interaction between the

793: cantilever and the electron on the dot is no longer a brief scattering

794: event, but a sustained interaction. In this regime we would 

795: expect to see evidence of the periodic behavior of the cantilever,

796: such as side resonances in the elastic transmission amplitude, but

797: in practice these would have to compete with thermal effects which tend to wash out

798: fine structure in the transmission characteristics.

799: 

800: In order to calculate the transmission amplitude, 

801: it is convenient to recast our earlier expression (\ref{tqd}) 

802: in the form

803: \begin{eqnarray}

804: \langle t_{\mathrm QD}(\epsilon) \rangle&=& -\frac{\Gamma}{\hbar\omega_0}

805: \frac{{\mathrm{e}}^{-(\lambda/\hbar\omega_0)^2(2\overline{n}+1)}}

806: {1-{\mathrm{e}}^{-(2\pi/\omega_0)[i(\epsilon_0-(\lambda^2/\hbar\omega_0)

807: -\epsilon)/\hbar)+\Gamma/2\hbar]}}

808: \nonumber \\ 

809: &&\times \int_0^{2\pi}{d}\tau'

810: \exp\Biggl\{\frac{-\tau'}{\omega_0}\left[\frac{i}{\hbar}

811: \left(\epsilon_0-\frac{\lambda^2}{\hbar\omega_0}-\epsilon \right)

812: +\frac{\Gamma}{2\hbar}  \right]\nonumber \\       

813: &&-i\left(\frac{\lambda}{\hbar\omega_0}\right)^2 \sin (\tau')

814: +(2\overline{n}+1)\left(\frac{\lambda}{\hbar\omega_0}\right)^2 

815: \cos (\tau')\Biggr\},\label{calc}

816: \end{eqnarray}

817: where $\tau'=\omega_0 \tau$. Of course, to model a practical experiment, we also 

818: need to average over the energy, $\epsilon$, to take into

819: account the effect of electron temperature. However, we

820:  look first at the dependence of the transmission amplitude on the incident energy. 

821:  Although this  would require `monochromatic' 

822: electrons,  important insight is gained into the behavior of the system

823: which in practice may be obscured by thermal effects.

824: 

825: In Fig.~2, the magnitude of the resonant transmission amplitude is plotted

826: against the coupling $\kappa$ for $\Gamma/\hbar\omega_0$=2, 1, and

827:  0.5, and with $\overline{n}=18$. The behavior is similar in all three cases for 

828:  small $\kappa$, with a rapid decay in the magnitude of the transmission amplitude,

829:   due to dephasing caused by a combination of  which--path detection and

830:  thermal fluctuations in the state of the cantilever.

831:   Notice, however, that

832:  for $\Gamma/\hbar\omega_0$=1,  oscillations eventually develop in the 

833: transmission, and that these become even more pronounced 

834: for $\Gamma/\hbar\omega_0$=0.5. 

835: 

836: 

837: 

838: 

839: 

840: In order to clarify the origin of the oscillations in

841:  $|\langle t_{\mathrm QD}(\epsilon)\rangle|$,

842:  Fig.~3 plots the evolution of the amplitude and phase of the

843: transmission amplitude as functions of both the coupling constant,

844: $\kappa$, and the detuning energy $\epsilon-\epsilon_0$, for 

845: $\Gamma/\hbar \omega_0=0.5$ and $\overline{n}=18$.  

846: It is now clear that there are side resonances at

847: $\epsilon-\epsilon_{0}+\lambda^2/\hbar\omega_0=

848: \pm\hbar\omega_0,\pm2\hbar\omega_0,...$, and that

849: these resonances drift in energy as the value of $\kappa$

850: is increased.\cite{W89,JW} 

851: It is this drift  which is responsible for the oscillations

852:  in $|\langle t_{\mathrm QD}(\epsilon)\rangle|$ as a function of the coupling, $\kappa$, 

853: for given fixed $\epsilon$. 

854: 

855: Note that if the cantilever temperature is set equal to 

856: zero, i.e., it is in 

857: its ground state, 

858: $\overline{n}=0$, then side resonances are still observed, but only on one 

859: side of the main peak, at $\epsilon-\epsilon_{0}+\lambda^2/\hbar\omega_0

860: =+\hbar\omega_0,

861: +2\hbar\omega_0,...$.This is because in its ground state the cantilever  

862: can only absorb energy quanta.

863: 

864: 

865:  

866: Under the conditions $k_{\mathrm B}T\gg\hbar\omega_{0}$, 

867: $\kappa\sim 1$, and near resonance, Eq. 

868: (\ref{calc}) has the following asymptotic approximation:

869: \begin{equation}

870:     \langle t_{\mathrm QD}(\epsilon) \rangle 

871:     \sim 

872:     -\frac{\Gamma}{2\lambda}\sqrt{\frac{\pi\hbar\omega_{0}}{k_{\mathrm 

873:     B}T}}

874:     \left\{\frac

875: {1+{\mathrm{e}}^{-(2\pi/\hbar\omega_0)[i(\epsilon_0-\lambda^2/\hbar\omega_0

876: -\epsilon)+\Gamma/2\hbar]}}{1-{\mathrm{e}}^{-(2\pi/\hbar\omega_0)

877: [i(\epsilon_0-\lambda^2/\hbar\omega_0

878: -\epsilon)+\Gamma/2\hbar]}}\right\}.\label{asympt}

879: \end{equation} 

880: The various above-discussed features in the transmission 

881: amplitude dependence on coupling and energy can be clearly seen in 

882: this simplifying approximation.   

883: 

884: 

885: Side resonances are a familiar feature in tunneling problems, but unlike those

886: found here they are usually 

887: associated with inelastic processes. However, as was  first

888: shown by Jauho and Wingreen,\cite{JW} who considered the classical--oscillator

889: version of the problem we address, such resonances can also occur in the

890: elastic transmission amplitude.  Side resonances in the elastic transmission 

891: indicate

892: coherent or virtual exchange of energy quanta between the electron on the dot 

893: and the cantilever, with no net energy interchange over the dwell 

894: time of the electron.  

895: In a fully quantum mechanical system, these processes must rely on the coupled 

896: cantilever--electron system maintaining its phase coherency and so we may expect 

897: that the influence on the cantilever of its environment should be detectable, 

898: at least in principle,  via its effect on these resonances. 

899: 

900: The side resonances have a separation in energy of $\hbar\omega_0$, but  

901: for any practically realizable system the thermal energy scale will be 

902: much larger, as we discuss in detail in Sec. VI. However, in  

903: the regime where $\hbar\omega_0\ll k_{\mathrm{B}} T$,

904: averaging over the thermal distribution of electron energies will wash 

905: out the side resonances.

906:  This means that it would not be

907: possible to find any trace of the coherent exchange of energy quanta between

908: the cantilever and the electron on the dot in the transmission 

909: characteristics. Only if the

910: electron energy width could be lowered or the frequency raised to the

911: point where $\hbar\omega_0\sim k_{\mathrm{B}}T$ would we then expect

912: such features to be visible. Note, in the case of photon assisted tunneling the photon

913: energy can be made comparable with $k_{\mathrm{B}} T$  

914: without difficulty.\cite{JW}

915: 

916: 

917: \section{Influence of Environment on Cantilever}

918: Thus far, we have treated the cantilever--dot system as isolated,

919: thereby ignoring the effect of the environment on their degrees of

920: freedom. This approximation is valid when the dwell time of the electron on 

921: the dot is  short compared to the decoherence time of the 

922: cantilever, 

923: as will most certainly be the case for $\omega_0\tau_d\ll1$. However, there is no 

924: reason  to assume

925: that this will also be the case when $\omega_0\tau_d\sim1$.  Indeed, we

926: would like to know how the environment of the cantilever affects the coherent 

927: oscillations which occur in the transmission amplitude as a function of the 

928: coupling strength  (for monochromatic electrons). Intuitively, we expect that the 

929: decoherence arising from the cantilever's environment  should wash out the side 

930: resonances in the transmission amplitude, but a detailed calculation is required 

931: to determine  the cantilever quality ($Q$) factor range

932:  for this to 

933: occur.  

934: 

935: 

936: In general,

937: both the interaction between the dot electron and its local environment (other

938: electrons, phonons, photons, etc. in the dot region) and the interaction

939: between the cantilever and its environment (the other collective vibrational modes,

940: as well as internal electronic degrees of freedom, external photons, 

941: gas molecules, etc.) will

942: contribute to the decoherence. However, since we can always measure the

943: properties of the fringes with the dot--cantilever interaction turned

944: off for an arbitrary dwell time ($\tau_d=\hbar/\Gamma$), we need only consider

945: the additional effect of the cantilever's environment, as the electron's environment

946: can be included via a renormalisation of the zero electric field

947: transmission amplitude.

948: 

949: 

950: 

951: \subsection{Estimate of Cantilever Decoherence Time}

952: 

953: We can obtain a rough estimate of the time--scale over which superpositions of 

954: cantilever states, resulting from the coupling to the dot electron, 

955: are likely to be decohered by the environment by using

956: a simple, heuristic approach which models the environment as an infinite set of 

957: harmonic oscillators.\cite{UZ,Bose} This 

958: approach leads to the prediction that for

959: a system with  a classical damping rate, $\gamma_c$, a linear superposition of

960:  two different coherent

961: states whose centers are a distance $\Delta x$ apart, where $\Delta x$

962: is greater than the thermal

963: de Broglie wavelength, $\lambda_{\mathrm th}=\hbar/\sqrt{2m k_{\mathrm{B}}T}$,

964: will decohere at a rate $\gamma_d$, given by

965: \begin{equation}

966: \gamma_d\sim\frac{2m\gamma_c k_{\mathrm{B}}T(\Delta x)^2}{\hbar^2}.\label{rot}

967: \end{equation}

968: 

969: However, we cannot apply this heuristic rule directly to the coupled 

970: cantilever--dot system as the displacement of the cantilever is a continuously 

971: varying function. An estimate for the decoherence rate of the cantilever can 

972: only be obtained using this method if the further approximation of 

973: averaging over the cantilever displacement is made.\cite{Bose} 

974: If the cantilever starts in any given 

975: coherent state when the electron arrives on the dot, then the

976: state of the cantilever at time $t$ later will be a different coherent state

977: centered a distance $\Delta x$ apart from the point on which the initial 

978: state was centred, where 

979: \begin{equation}

980: \Delta x(t)=\sqrt{\frac{2\hbar}{m\omega_0}}\ 

981: \kappa[1-\cos(\omega_0 t)].

982: \end{equation}

983: Thus we can obtain an estimate for the decoherence rate $\gamma_{d}$ of the 

984: cantilever due to its

985: environment for the case where $\tau_d\omega_0=1$ by averaging this displacement 

986: over the cantilever period:\cite{Bose}

987: \begin{equation}

988: \gamma_d \sim\frac{1}{(2\pi/\omega_0)}\frac{4\kappa^2\gamma_c k_{\mathrm{B}}T}

989: {\hbar\omega_0}\int_0^{2\pi/\omega_0}{d}t[1-\cos(\omega_0 

990: t)]^2

991: =6\kappa^2\gamma_c\frac{k_{\mathrm{B}}T}{\hbar 

992: \omega_0}.\label{decoherence}

993: \end{equation} 

994: For a nanotube cantilever of frequency $\sim140$MHz (see appendix A), 

995: the $Q$--factor\cite{vib}

996: can be of order $500$ and so the classical damping rate, $\gamma_c$, will be 

997: of order $3\times 10^5$~$\mathrm{s}^{-1}$. Thus, at a temperature of 

998: 20~mK, our

999: heuristic expression for the decoherence rate of the nanotube cantilever  gives 

1000: $1/\gamma_d\sim 3\times 10^{-9}$~s for $\kappa\simeq 3$. This result 

1001: signals that  the decoherence of the cantilever due to its environment has an effect

1002: whenever the dwell time approaches a magnitude of order $1$ns.

1003: 

1004: Whilst the heuristic model we have outlined is very useful for

1005: estimating whether or not the cantilever's environment is relevant for

1006: the calculation of the transmission amplitude, it is not expected to

1007: be very accurate, as it is an approximation even of the rule--of--thumb given 

1008: by   Eq. (\ref{rot}).

1009: In order to improve on this estimate, 

1010: we must  enlarge our simple model by adding to the system 

1011: Hamiltonian (\ref{Ham1}) the cantilever's 

1012: environment, modelled as an infinite bath of harmonic oscillators with 

1013: linear coupling to the cantilever.  

1014: Modelling the environment in this way is of course itself a fairly

1015: serious approximation. However, this approximation is ubiquitous in

1016: one form or another throughout the theory of open quantum 

1017: systems.\cite{CL,UZ,O,JZ} By extending our calculation to include this

1018:  model of the cantilever's environment, we can obtain predictions

1019: for how the transmission properties of our which--path device depend

1020: on the decoherence rate of the cantilever. Thus, we can use our

1021: theory to predict how changing the $Q$--factor of the cantilever

1022: affects the interference fringes.         

1023: 

1024: \subsection{Transport Properties with Environmental Coupling}

1025: 

1026: The standard model of the environment which we use is  an infinite bath of harmonic 

1027: oscillators which interact linearly with the cantilever, but do not interact 

1028: with each other. We have to assume an infinite bath of oscillators in order to 

1029: have a reservoir which remains in thermal equilibrium despite contact with the 

1030: cantilever. The infinite oscillator--bath model of the environment is also 

1031: equivalent to a quantum Langevin formalism in which the cantilever operators 

1032: have an equation of motion containing a damping term and a thermal noise 

1033: operator arising from the reservoir. 

1034: 

1035: We begin by considering just the interaction between the cantilever and its

1036: environment, before going on to include the effect of an electron on the dot and

1037: hence obtaining the transmission amplitude including the environment.

1038: The cantilever-environment Hamiltonian can be written as\cite{MW,Lou}

1039: \begin{equation}

1040: {\mathcal{H}}^{c}=\hbar \omega_0\hat{a}^{\dagger}\hat{a}

1041: +\sum_{\omega} \hbar \omega 

1042: \hat{A}^{\dagger}(\omega)\hat{A}(\omega) 

1043: +\sum_{\omega}[g(\omega)\hat{a}^{\dagger}\hat{A}(\omega) + 

1044: g^*(\omega)\hat{A}^{\dagger}(\omega)\hat{a}], \label{envham}

1045: \end{equation}

1046: where the cantilever states are again operated on by 

1047: $\hat{a}$, the bath oscillator of frequency $\omega$ is operated on by

1048:   $\hat{A}(\omega)$ and the coupling constants $g(\omega)$ depend on

1049:   the bath oscillator frequency.

1050: 

1051:  Using the Hamiltonian ${\mathcal{H}}^c$, the  equation 

1052: of motion for $\hat{a}$ in the interaction picture is 

1053: readily obtained,\cite{MW}

1054: \begin{equation}

1055: \dot{\hat{a}}(t)=(-i \omega-\gamma)\hat{a}(t)-\hat{F}(t),

1056: \end{equation}

1057: where 

1058: \begin{equation}

1059: \hat{F}(t)=i\sum_\omega g(\omega)\hat{A}(\omega,0){\mathrm{e}}^{-i \omega t}

1060: \end{equation}

1061: and the damping coefficient, $\gamma$, is given by

1062: \begin{equation}

1063: \gamma=\pi \eta_{b}(\omega_0)|g(\omega_0)|^2,

1064: \end{equation}

1065: where $\eta_{b}(\omega)\delta \omega$ is the number of bath oscillators in the 

1066: spectral range $\delta\omega$. The coefficient $\gamma$ provides the bridge 

1067: between the model and experiment, as it is simply related to the rate at which the 

1068: system loses energy after being excited (i.e., the classical damping rate),\cite{Lou}

1069: $\gamma_c=2\gamma$.

1070: 

1071: 

1072: Integration of the equation of motion for $\hat{a}(t)$, and a similar one for 

1073: $\hat{a}^{\dagger}(t)$, leads to explicit expressions for 

1074: $\hat{a}(t)$ and $\hat{a}^{\dagger}(t)$:

1075: \begin{equation}

1076: \hat{a}(t)=\hat{a}(0){\mathrm{e}}^{(-i\omega_0-\gamma)t}-{\mathrm{e}}

1077: ^{(-i\omega_0- 

1078: \gamma)t}\int_0^t dt' \hat{F}(t'){\mathrm{e}}^{(i\omega_0+\gamma)t'}, 

1079: \label{plaina}

1080: \end{equation} 

1081: \begin{equation}

1082: \hat{a}^{\dagger}(t)=\hat{a}^{\dagger}(0){\mathrm{e}}^{(i\omega_0-\gamma)t}-

1083: {\mathrm{e}}^{(i 

1084: \omega_0-\gamma)t}\int_0^t dt' 

1085: \hat{F}^{\dagger}(t'){\mathrm{e}}^{(-i\omega_0+\gamma)t'}. \label{adag}

1086: \end{equation}  

1087: 

1088: The time $t=0$ holds a special place in the theory since it is the time  at 

1089: which the interactions between the cantilever and the bath are turned on and 

1090: so the cantilever and the bath are apparently quite independent at this instant.

1091:  However, this  need not be the case and certainly would not be appropriate 

1092:  for the system  we are considering. In

1093: order to specify the model completely we need to give the expectation values 

1094: at $t=0$ for both the cantilever and the bath. If we set the initial 

1095: expectation values of the cantilever to be those of a thermal state at the

1096: same temperature as the bath, then we can describe the case in which the

1097: cantilever has been interacting with the bath for a time much longer than the 

1098: damping time, $1/\gamma_c$, {\emph{before}} $t=0$ and so is in equilibrium with

1099: the bath. Our only assumption is that

1100: quantum correlations between the cantilever and the bath can be neglected

1101: at $t=0$, the usual assumption made in calculating decoherence rates. 

1102: 

1103: 

1104:  

1105:  The expectation values are over  

1106: products of  cantilever and environment states.  We eventually want to calculate the 

1107: transmission amplitude for an electron interacting with

1108: a cantilever which is initially in a thermal state

1109: and so we choose to define the $t=0$ expectation values as:

1110: \begin{equation}

1111: \langle...\rangle=\frac{{\mathrm{Tr}}[...{\mathrm{e}}

1112: ^{-{\mathcal{H}}^c_0/k_{\mathrm{B}}T}]}

1113: {{\mathrm{Tr}}[{\mathrm{e}}^{-{{\mathcal{H}}^c_0}/k_{\mathrm{B}}T}]},

1114: \end{equation}

1115: where the Hamiltonian without interactions is 

1116: \begin{equation}

1117: {\mathcal{H}}^c_0=\hbar \omega_0\hat{a}^{\dagger}\hat{a}

1118: +\sum_{\omega} \hbar \omega 

1119: \hat{A}^{\dagger}(\omega)\hat{A}(\omega),

1120: \end{equation}

1121: and $T$ defines the fixed temperature of the environment.

1122: Using this definition we find

1123: \begin{eqnarray}

1124: \langle \hat{A}(\omega,0)\rangle&=&0 \nonumber \\

1125: \langle \hat{A}(\omega,0)\hat{A}(\omega',0)\rangle&=&0 \\

1126: \langle \hat{A}^{\dagger}(\omega,0)\hat{A}(\omega',0)\rangle&=&\delta_{\omega 

1127: \omega'}N(\omega),\nonumber 

1128: \end{eqnarray}   

1129: with

1130: \begin{equation}

1131: N(\omega)=\frac{1}{{{\mathrm{e}}}^{\hbar \omega/k_{\mathrm{B}}T}-1}.

1132: \end{equation}

1133: 

1134: For the cantilever itself the values of the zero-time correlation functions 

1135: represent the initial conditions of the problem. In 

1136: thermal equilibrium with the environment at temperature $T$,   

1137: \begin{eqnarray}

1138: \langle \hat{a}(0)\rangle&=&0\nonumber \\

1139: \langle \hat{a}(0)\hat{a}(0)\rangle&=&0 \\

1140: \langle \hat{a}^{\dagger}(0)\hat{a}(0)\rangle&=&\overline{n},\nonumber

1141: \end{eqnarray}   

1142: with

1143: \begin{equation}

1144: \overline{n}\equiv N(\omega_0)=\frac{1}{{\mathrm{e}}^{\hbar 

1145: \omega_0/k_{\mathrm{B}}T}-1}.

1146: \end{equation}

1147: 

1148: The purpose of extending the analysis to include the interactions between the

1149: cantilever and its environment is to see how they modify the 

1150: interaction between the cantilever and the electron {\emph{whilst it is on the dot}}. 

1151: Therefore we can choose our

1152: origin of  time, and hence the definition of the initial cantilever state, to 

1153: be the time when the electron jumps onto the dot. There is no need to explicitly

1154: include the interaction with the bath before the electron hops onto the dot as it is

1155: already implicitly included by assuming the cantilever is in a thermal state 

1156: at temperature $T$. However, when the electron is on the dot it drives 

1157: the cantilever away from equilibrium and we now need to include the 

1158: interaction with the environment explicitly to model the behavior 

1159: of the cantilever--dot system as accurately as possible. 

1160: 

1161: 

1162: 

1163: We obtain the transmission amplitude for the dot, including the 

1164: effect of the cantilever's environment by applying the $S$--matrix we used in

1165:  Sec. III to a generalization of the Hamiltonian (\ref{Ham1})

1166:  which includes the coupling of the cantilever to 

1167:  the bath of oscillators,

1168: \begin{equation}

1169: {\mathcal{H}}=\epsilon_0\hat{c}^{\dagger}\hat{c}+{\mathcal{H}}^c

1170: +{\mathcal{H}}_1+\sum_k\left( 

1171: \epsilon_{kL}\hat{c}^{\dagger}_{kL}\hat{c}_{kL}+\epsilon_{kR}\hat{c}^{\dagger}_

1172: {kR}\hat{c}_{kR}\right), 

1173: \end{equation}

1174: where ${\mathcal{H}}_1$, given by eqn (\ref{coup}), describes the 

1175: electron--cantilever coupling and the cantilever--environment part, ${\mathcal{H}}^c$, 

1176: is given by Eq.\ (\ref{envham}). 

1177:  

1178: 

1179: Using the methods of Glazman and Shekhter,\cite{GS} we can again separate out 

1180: the electronic 

1181: part of the transmission amplitude from the average over cantilever states, 

1182: and so we find

1183: \begin{eqnarray}

1184: \langle t_{\mathrm QD}(\epsilon) \rangle &=& 

1185: -\Gamma\int_0^{\infty}\frac{dt}{\hbar}{\mathrm{e}}^{[i(\epsilon-\epsilon_0)

1186: -\Gamma/2]t/\hbar}

1187: \langle{\mathrm{T}}_t 

1188: {\mathrm{e}}^{-i \int_0^{t}W_I(t')dt'/\hbar} \rangle.

1189: \end{eqnarray}

1190: The term in angled brackets on the right--hand side is known as the influence

1191:  functional, 

1192: ${\mathrm{T}}_t$ is the time ordering operator, and $W_I(t)$  is the

1193:  electron--cantilever coupling defined in the interaction picture 

1194: \begin{equation}

1195: W_I(t)=-{\mathrm{e}}^{i{\mathcal{H}}^c t/\hbar}

1196: \lambda(\hat{a}^{\dagger} +\hat{a})

1197: {\mathrm{e}}^{-i{\mathcal{H}}^c t/\hbar}

1198: =-\lambda[ \hat{a}^{\dagger}(t) +\hat{a}(t)],

1199: \end{equation}

1200: where $\hat{a}(t)$ and $\hat{a}^{\dagger}(t)$ are given by Eqs. (\ref{plaina}) and 

1201: (\ref{adag}) above.

1202: 

1203: 

1204: Here, we shall

1205: carry out a much simpler, approximate calculation of the influence 

1206: functional which involves performing a second order expansion in 

1207: $\lambda$ and then re-exponentiating. In appendix B, it is shown that 

1208: this approximate calculation in fact coincides with the exact result 

1209: obtained from a full calculation. 

1210: 

1211: The expansion to second order gives 

1212: \begin{equation}

1213: \langle{\mathrm{T}}_t {\mathrm{e}}^{-i \int_0^{t}W_I(t')dt'/\hbar} \rangle=

1214: 1-i\int_0^t \frac{dt'}{\hbar}\langle W(t')\rangle 

1215: +(-i)^2\int_0^t \frac{dt'}{\hbar}\int_0^{t'}\frac{dt''}{\hbar}\langle 

1216: W(t')W(t'')\rangle. \label{if}

1217: \end{equation}  

1218: The next step is to evaluate the correlation 

1219: functions which arise in the first and second order terms in the 

1220: expansion of the influence functional. At first order

1221: \begin{equation}

1222: \langle W(t') \rangle=-\lambda\langle  [\hat{a}(t')+\hat{a}^{\dagger}(t')]\rangle,

1223: \end{equation}

1224:  and at second order

1225: \begin{equation}

1226: \langle W(t')W(t'')\rangle = \lambda^2\{ \langle 

1227: \hat{a}(t') \hat{a}(t'')\rangle+ \langle \hat{a}(t') 

1228: \hat{a}^{\dagger}(t'')\rangle   

1229: +\langle \hat{a}^{\dagger}(t') \hat{a}(t'')\rangle + \langle 

1230: \hat{a}^{\dagger}(t') \hat{a}^{\dagger}(t'')\rangle \},

1231: \end{equation} 

1232: where $t'\geq t''$.

1233: 

1234: We can calculate all the correlation functions 

1235:  using the initial conditions defined above and the standard results of 

1236: quantum Langevin theory.\cite{MW,Lou}

1237: We  start by observing that three of the five correlation functions are 

1238: zero:

1239: \begin{equation}

1240: \langle \hat{a}(t')\hat{a}(t'')\rangle=\langle 

1241: \hat{a}^{\dagger}(t')\hat{a}^{\dagger}(t'')\rangle=0 \label{a}

1242: \end{equation}

1243: and

1244: \begin{equation}

1245: \langle [\hat{a}(t')+\hat{a}^{\dagger}(t')]\rangle=0.

1246: \end{equation}

1247: These results follow from the definitions above, since correlators of the type 

1248: $\langle \hat{a}(0)\hat{F}(t) \rangle$  decouple into products of one-time 

1249: correlation functions and the definitions imply that $\langle \hat{F}(t) 

1250: \rangle=\langle \hat{F}(t')\hat{F}(t'') \rangle=0$. 

1251: For the other two correlation functions, we have

1252: \begin{eqnarray}

1253: \langle \hat{a}(t')\hat{a}^{\dagger}(t'')\rangle&=&  (\overline{n}+1)

1254: {\mathrm{e}}^{-i\omega_0 (t'-t'')} {\mathrm{e}}^{-\gamma (t'-t'')} 

1255: \label{b} \\

1256: \langle \hat{a}^{\dagger}(t')\hat{a}(t'')\rangle&=&  \overline{n}

1257: {\mathrm{e}}^{i\omega_0 (t'-t'')} {\mathrm{e}}^{-\gamma (t'-t'')} \label{c}.

1258: \end{eqnarray}

1259: Thus, we again find

1260: \begin{equation}

1261: \langle t_{\mathrm QD}(\epsilon) \rangle=-\Gamma\int_0^{\infty}\frac{dt}{\hbar}

1262: {\mathrm{e}}^{[i(\epsilon-\epsilon_0)

1263: -\Gamma/2]t/\hbar}{\mathrm{e}}^{-\phi(t)}, 

1264: \label{exact}

1265: \end{equation} 

1266: but now with 

1267: \begin{equation}

1268: \phi(t)=\frac{\lambda^2}{\hbar^2} \int_0^t 

1269: {dt'}\int^{t'}_{0}{dt''}{{\mathrm{e}}}^{-\gamma (t'-t'')}\left[ (\overline{n}+1)

1270: {\mathrm{e}}^{-i\omega_0 (t'-t'')}+\overline{n}

1271: {\mathrm{e}}^{i\omega_0 (t'-t'')}\right], 

1272: \label{phi}

1273: \end{equation}

1274: where we have re-exponentiated the expansion in $\lambda$.

1275: 

1276: 

1277: 

1278: Carrying out a change of variables to $\tau=t'-t''$, the integrals are readily 

1279: evaluated to give

1280: \begin{equation}

1281: \phi(t)=\left(\frac{\lambda}{\hbar}\right)^2\left[ 

1282: \frac{(\overline{n}+1)}{(\gamma+i\omega_0)}\left(t+\frac{{\mathrm{e}}^{-(\gamma+i

1283: \omega_0)t}-1}{\gamma+i\omega_0}\right)+\frac{\overline{n}}{(\gamma-i\omega_0)}

1284: \left(t+\frac{{\mathrm{e}}^{-(\gamma-i\omega_0)t}-1}{\gamma-i\omega_0}\right)

1285: \right]. \label{phix}

1286: \end{equation}

1287: Setting $\gamma=0$, one may verify that (\ref{phix}) indeed reduces 

1288: to expression (\ref{ex2}) in the absence of the environment.

1289: 

1290: Under the conditions $Q\gg 1$, $k_{\mathrm B}T\gg\hbar\omega_{0}$, 

1291: $\kappa\sim 1$, and near resonance, the transmission amplitude 

1292: (\ref{exact}) has the following asymptotic approximation:

1293: \begin{equation}

1294:     \langle t_{\mathrm QD}(\epsilon) \rangle 

1295:     \sim 

1296:     -\frac{\Gamma}{2\lambda}\sqrt{\frac{\pi\hbar\omega_{0}}{k_{\mathrm 

1297:     B}T}}

1298:     \left\{\frac

1299: {1+{\mathrm{e}}^{-(2\pi/\hbar\omega_0)[i(\epsilon_0-\lambda^2/\hbar\omega_0

1300: -\epsilon)+\Gamma/2\hbar+(\kappa^{2}\gamma_{c}k_{\mathrm 

1301: B}T/\hbar\omega_{0})]}}{1-{\mathrm{e}}^{-(2\pi/\hbar\omega_0)

1302: [i(\epsilon_0-\lambda^2/\hbar\omega_0

1303: -\epsilon)+\Gamma/2\hbar+(\kappa^{2}\gamma_{c}k_{\mathrm 

1304: B}T/\hbar\omega_{0})]}}\right\}.\label{asymptenv}

1305: \end{equation} 

1306: Note that this approximation differs from the earlier-derived one for 

1307: the transmission amplitude 

1308: in the absence of the environment, Eq. (\ref{asympt}), merely by the 

1309: replacement of $\Gamma/2\hbar$ with $\Gamma/2\hbar+(\kappa^{2}\gamma_{c}k_{\mathrm 

1310: B}T/\hbar\omega_{0})$. Thus, a decoherence rate $\gamma_{d}$ can be identified 

1311: as the term $\kappa^{2}\gamma_{c}k_{\mathrm B}T/\hbar\omega_{0}$, which 

1312: agrees with the earlier-derived estimate 

1313: (\ref{decoherence}) up to an overall numerical factor. 

1314: Approximation (\ref{asymptenv}) clearly shows the washing out of

1315: the coherent, oscillatory behavior by the 

1316: environment when the decoherence time $1/\gamma_{d}$ is shorter than 

1317: the cantilever period. When the decoherence time exceeds the dwell 

1318: time $\hbar/\Gamma$, then the former has a 

1319: negligible effect on the transmission properties.      

1320: 

1321: 

1322: 

1323: 

1324: \subsection{Results for Cantilever Coupled to Environment}

1325: The expression for the transmission amplitude including the coupling to

1326: the cantilever's environment [Eq.\ (\ref{exact})] can be integrated

1327: numerically, thereby allowing us to explore the effect on the

1328: transmission characteristics of varying the cantilever's

1329: $Q$--factor under more general 

1330: conditions than those for which approximation (\ref{asymptenv}) is 

1331: justified. Fig.~4 shows the magnitude 

1332: and phase of the

1333: transmission  amplitude through the dot for $\Gamma/\hbar \omega_0=0.5$,

1334: $\overline{n}=18$ and $Q=\omega_0/\gamma_{c}=50$. The diagram 

1335: should be compared with

1336: Fig.~3 showing the behavior of the same system without

1337: environmental coupling. Fig.~5 shows the behavior at resonance of

1338: $|\langle t_{\mathrm QD}(\epsilon)\rangle|$ for 

1339: $Q=50$ and 500, as well as the $\gamma_{c}/\omega_0 =0$  (no environment)

1340: case for comparison.

1341: 

1342: It is clear from the figures that the coupling to the environment tends to

1343: destroy the side resonances in the transmission amplitude, as well as

1344: the associated features in the phase. This is because the environment acts to 

1345: degrade the coherent 

1346: superposition of cantilever states into which the interaction with the electron tries

1347: to drive the cantilever. Furthermore, the figures show that these 

1348:  environmental effects 

1349: become increasingly important as the cantilever--electron coupling,

1350: $\kappa$, is increased. This is because the larger $\kappa$ is, the greater 

1351: is the 

1352:  difference between the 

1353: states in the superposition into which the cantilever is driven, and consequently 

1354: the faster is the rate at which the superposition decoheres.        

1355: 

1356:  

1357: 

1358: \section{Practical Considerations}

1359: 

1360: 

1361: The model parameter ranges actually allowed in an experiment  are limited by 

1362: practical constraints. Up until now we 

1363: have only referred to these very loosely and have concentrated instead on the

1364: range of behavior which can occur in the which--path system under fundamental 

1365: constraints alone.

1366: 

1367: Probably

1368: the most important practical constraint affects the upper range of the electric field

1369: which can be developed between the cantilever and the dot. 

1370: The maximum allowable field is typically $\sim 10^5$~V/m before 

1371: breakdown of the two--dimensional electron gas occurs due to  

1372: deconfinement,\cite{Buksp} and so we

1373: must take this as the largest possible value in considering the

1374: practicality of the system. 

1375: 

1376: The temperature of

1377:  the system is limited by the difficulty of cooling conduction electrons  

1378:  to ultralow temperatures. It becomes extremely difficult to reduce the 

1379:  electron temperature  below about 20~mK,

1380:  because the electrons become practically decoupled from acoustic 

1381:  phonons. 

1382:  Therefore we take 20~mK to be 

1383:  the temperature minimum. 

1384: 

1385: The frequency range of the cantilever is crucial 

1386: for observing quantum coherent behavior such as  side resonances, as 

1387: well as their destruction 

1388: due to 

1389: decoherence. The lower limit 

1390: on the frequency range is set by the requirement that the cantilever 

1391: period should be comparable to the dwell time $\tau_d=\hbar/\Gamma$ of the electron on the quantum dot,  

1392: while the 

1393: upper limit is set

1394:  by the requirement that the interaction between the cantilever and

1395: the electron on the dot be sufficiently strong, given the limits on

1396: the electric field,  to lead to decoherence,

1397: again on a time--scale  comparable to the dwell time.

1398: The interaction between

1399: the electron on the dot and the cantilever is discussed in more

1400: detail in appendix A, along with the geometrical factors which arise in

1401: the calculation of the effect of the electric field on the flexural cantilever

1402: modes.

1403: The dwell time  is

1404: limited by the rate at which processes other than the interaction 

1405: with the cantilever cause decoherence, as well as how small a current can be measured 

1406: through the dot. Of course, the decoherence rate

1407: of all the background processes is very difficult to estimate

1408: theoretically; in experiment it can be done by

1409: observing the visibility of the fringes as a function of dwell time

1410: with the cantilever interaction switched off. This then provides a

1411: baseline with which to compare all later measurements where we are

1412: interested in the effect of the cantilever. Previous experiments 

1413: carried out at $T=0.1$~K by

1414: Yacoby {\emph{et al.}},\cite{Yac} which were designed to measure the phase of

1415: the transmission amplitude through a dot in the absence of any

1416: external probe, show that dwell times as long 10~ns and currents as low as 

1417: $\sim10^{-11}$~A lead to fringes which are still detectable.

1418: 

1419: It turns out that the best compromise between the two 

1420: competing frequency limits is achieved by using say a carbon nanotube 

1421: cantilever with a frequency of 100--200~MHz, giving a maximum 

1422: coupling $\kappa=\lambda/\hbar\omega_{0}\sim 3$ (appendix A). The   

1423: problem is not so much in finding a cantilever with a high enough frequency, but rather

1424: in obtaining a large enough coupling from the electric field to cause a 

1425: detectable amount of dephasing.  

1426: 

1427: 

1428: 

1429: In the light of these practical constraints it is clear that only some 

1430:  of the theoretical results obtained in our

1431: analysis would be observable in an experiment using current

1432:  technology. The overall destruction of the

1433:  AB interference fringes as the coupling between the cantilever and the

1434:  electron on the dot is increased should be detectable, both in the

1435:  regime where $\Gamma/\hbar\omega_0\gg 1$ and $\Gamma/\hbar\omega_0\sim1$.

1436: However, the restrictions on the temperature and cantilever

1437:  frequency imply that an experiment would have to be performed in the regime

1438:  where $k_{\mathrm{B}}T\gg \hbar\omega_0$. This means that the thermal

1439:  width of electron energies will wash out the side resonances 

1440:  in the transmission which were found to occur when 

1441: $\Gamma/\hbar\omega_0\sim1$. Under these circumstances, the effect of

1442:  the cantilever's environment on the transmission characteristics

1443:  would be obscured. However, we emphasise again that these

1444:  limitations are not fundamental: if the width of the electron

1445:  energies could be reduced or if the restriction on the maximum

1446:  cantilever frequency could be relaxed (by finding a way to increase

1447:  the coupling between the electron on the dot and the cantilever, for

1448:  example) then the coherent, quantum electromechanical features which our

1449:  analysis predicts should be observable. 

1450:  Whilst it is not clear how

1451:  the coupling constant between the cantilever and the dot could be

1452:  increased, it is possible to reduce the thermal width of

1453:  electron energies by using a double quantum dot system 

1454:  rather than a single one as we have considered here.\cite{Buks2}  

1455:    

1456: It is also  possible to conduct experiments in which  an ac bias is applied 

1457: to the AB ring. Such an experiment would provide an alternative  way of 

1458: investigating

1459: the interaction between the cantilever and its environment: if the ac frequency 

1460: is higher than the rate $\gamma_{c}$ at which the cantilever state changes due to 

1461: thermal fluctuations then the thermally-induced fluctuations in the phase of the 

1462: interference fringes should  be detectable, and distinguishable from 

1463: the destruction of the interference fringes due to which-path 

1464: detection. However, a detailed analysis of such

1465: an experiment requires extending the theory developed here to include

1466: time dependence, and so goes well beyond the present analysis.

1467: 

1468: 

1469: \section{Conclusions and Discussion}

1470: 

1471: We have carried out a theoretical analysis of a possible solid state

1472: which--path interferometer, in which electronic and mechanical degrees of freedom 

1473: become coupled. The visibility and phase of the interference

1474: fringes in the system depend strongly on the coupling between the

1475: electron on the dot and the cantilever. The reduction in visibility 

1476: with increasing coupling is 

1477: due in part to

1478: the cantilever measuring the path taken by the electron and also due 

1479: to thermal

1480: fluctuations in the state of the cantilever.

1481: 

1482: When the dwell time of the electron on the dot is short compared to

1483: the period of the cantilever, the system behaves in a way which is

1484: analogous to Einstein's celebrated recoiling slit experiment. In

1485: contrast, when the dwell time is comparable to the cantilever's period,

1486:  the cantilever and the electron on the

1487: dot   show signs of behaving  as a single coherent quantum system,

1488:  so long as the electrons incident on the dot have a

1489: sufficiently narrow energy width.  The coherency of the

1490: cantilever--dot system is inferred from the appearance of side

1491: resonances in the transmission characteristics. Including  the cantilever's

1492:  environment in the analysis, we find that the side

1493:  resonances are washed out   for small enough  oscillator 

1494:  $Q$--factor, while the average decrease in the fringe visibility with

1495:  increasing coupling to the cantilever is not affected.

1496: 

1497: The basic feature of the reduction of fringe visibility as a function

1498: of the coupling between the electron on the dot and the cantilever

1499: should be observable in an experiment using currently available technology. 

1500: However, the 

1501: more delicate features   such as the

1502: side resonances in the transmission amplitude for long dwell times 

1503: and the effect of the 

1504:  cantilever's environment on these resonances will usually be

1505:  obscured by the  energy thermal width of electrons incident on the dot.

1506: 

1507: There are two important ways in which our analysis could be

1508: extended. Most straightforwardly, we could examine what effect using a

1509: double dot, rather than a single one, would have on the behavior of the

1510: system. In particular, it would be interesting to see to what extent the

1511: electron energy width could be reduced, whilst maintaing

1512: an overall, measurable current through the device. 

1513: The second way our

1514: analysis could be extended would be to go beyond our steady state

1515: treatment to obtain the time dependence of the transmission

1516: amplitude. A time dependent analysis would allow us to make predictions

1517: about the way in

1518: which the thermal fluctuations in the state of the cantilever cause

1519: fluctuations in the phase of the interference fringes. Such

1520: fluctuations may prove to be observable in ac experiments and may also

1521: provide us with another way of inferring information about how the

1522: cantilever interacts with its environment.

1523:   

1524: \acknowledgements

1525: M.P.B. would especially like to thank Eyal Buks for many helpful 

1526: conversations and invaluable suggestions. Very helpful conversations 

1527: with Sougato Bose are also gratefully acknowledged.

1528: A.D.A. thanks Angus MacKinnon for a series of very useful discussions.  Funding was provided by the 

1529: EPSRC under Grant No. GR/M42909/01.

1530: 

1531: \appendix

1532: \section{Cantilever}

1533: 

1534: The cantilever in the which--path experiment we propose must fulfil a

1535: number of quite stringent properties: it must have a fundamental 

1536: frequency in the range 100--200~MHz and it must also  

1537: be a conductor. These requirements can be satisfied conveniently

1538:  by using a carbon nanotube as a cantilever, rather than a device which has

1539: been fabricated via some kind of etching process from a much larger substrate. 

1540: 

1541: Carbon nanotubes have a number of remarkable physical properties which

1542: are beginning to exploited for practical purposes. Recent experiments

1543: have seen them employed as hyper--sensitive tips in AFM experiments,

1544:  with the nanotube attached to the end of a conventional AFM tip to

1545:  extend the effective range of resolution of the device.\cite{AFM} A similar

1546:  apparatus could be used to bring a nanotube into position to act as

1547:  the cantilever in our which--path experiment, either in the

1548:  geometrical configuration explicitly considered above or some

1549:  variation of it  which would lead to slightly different geometrical

1550:  factors, but not change the underlying linear form of the

1551:  cantilever--dot interaction.   

1552:  

1553: In this appendix we carry out an analysis of the coupling between the

1554:  cantilever

1555: and the electron on the dot, and the normal modes of a nanotube cantilever.  

1556: We verify the

1557:   linear dependence of the cantilever energy on the electric field assumed 

1558:  in the text and derive the form of the coupling constant, $\lambda$.

1559:   By examining the mode spectrum of a nanotube cantilever, we 

1560:  confirm the possibility of obtaining a cantilever with a fundamental flexural

1561:  mode of order $10^8$~Hz, and justify our assumption that for any given cantilever it 

1562:  is sufficient to consider just the lowest mode.  

1563: 

1564:  

1565: \subsection{Cantilever--Dot Coupling}

1566: 

1567: We begin by determining the relation between the

1568: maximum electric field at the surface of the dot

1569: and the voltage applied to the conducting cantilever, before analysing the 

1570: details of the effect of the electron on the cantilever energy.   

1571: 

1572: We will consider a cantilever of length $L\sim1\ \mu$m, positioned so

1573: that its tip lies over the centre of the dot and at a height $d\sim

1574: 0.1\ \mu$m.\cite{justify} When an extra electron is added to the dot it

1575: will cause a (classical) vertical displacement of the tip $z\ll d$. We treat the

1576: cantilever as a conducting needle, since it's length will be far greater than 

1577: either of the other two dimensions. An applied voltage induces a line

1578: charge density, $\sigma$, on the cantilever, but because the cantilever is 

1579: necessarily finite in length, the charge density is not entirely uniform and so

1580: for an exact treatment we should write the line charge density as $\sigma(x)$ 

1581: where $x$

1582: runs along the length of the cantilever from the tip ($x=0$). However,

1583: for a cantilever with a large enough aspect ratio, the charge density can be 

1584: approximated as constant with only a small error [the

1585: error is of order $\delta$ with $\delta^{-1}=2\ln(L/r)$, where $r$ is

1586: the radius, for a rod shaped cantilever\cite{jack}]. For the cantilever we are

1587: considering the radius may be as small as $1.5$nm, as we discuss

1588: below, and so the error in assuming a uniform charge density will be

1589: less than 10\%, which is acceptable as our aim is to obtain an order

1590: of magnitude estimate for the  interaction strength.    

1591: 

1592: The charge induced on the cantilever leads to an electric field at the

1593: surface of the dot, 

1594: \begin{equation}

1595: {\mathbf{E}}=\int_0^{L}\frac{\sigma(-x\hat{\mathbf{i}}+d\hat{\mathbf{j}})

1596: {d}x}{4\pi\epsilon_0(x^2+d^2)^{3/2}},

1597: \end{equation} 

1598: taking $\sigma=cV$ where $c$ is an unimportant  constant and $V$

1599: is the voltage applied to the cantilever. The unit vectors are defined

1600: so that $\hat{\mathbf{i}}$ runs along the cantilever away from its tip and

1601: $\hat{\mathbf{j}}$ points down from the tip towards the dot.

1602: Since $L\gg d$, we find

1603: \begin{equation}

1604: |{\mathbf{E}}|\simeq \frac{2^{1/2}cV}{4\pi\epsilon_0 d}.

1605: \end{equation}

1606: 

1607: We also need to know how the potential energy of the cantilever varies

1608: for small vertical displacements, $z$, about the equilibrium position,

1609: $d_0$, where $d=d_0-z$. For a section of the cantilever of

1610: length $\Delta x$ centered at $x$, the potential energy due to the

1611: displacement is

1612: \begin{equation}

1613: \Phi(x)=\frac{qcV\Delta x}{4\pi\epsilon_0\sqrt{(d_0-z)^2+x^2}},

1614: \end{equation}

1615: where $q=-e$ is the excess charge on the dot. Since $d_0\gg z$, we obtain 

1616: the $z$--dependent part of the potential energy as

1617: \begin{equation}

1618: \Phi_z(x)=z\frac{qcVd_0\Delta x}{4\pi\epsilon_0[d_0^2+x^2]^{3/2}}.

1619: \end{equation} 

1620: This is of course linear in $z$ as we anticipated in our model. However,

1621: in order to obtain the effective coupling between the electric field

1622: and each of the cantilever modes we need to rewrite the $z$--dependent

1623: part of the potential energy in terms of the normal modes of the

1624: cantilever and so we turn now to the mode spectrum of the cantilever.

1625: 

1626: 

1627:  

1628: \subsection{Cantilever Modes}

1629: 

1630: 

1631: A single--walled nanotube cantilever  can be obtained with lengths $\sim

1632: 1\ \mu$m, diameters $\sim 3$~nm, and a Young's modulus $\sim 1$~TPa.\cite{nano} 

1633: Recent experiments,\cite{vib,modes2} have shown that the

1634: flexural modes of such nanotubes have frequencies which lie in the MHz--GHz

1635: regime. We can treat a nanotube as a rigid hollow rod and so obtain

1636: the frequencies of the flexural modes,\cite{modes}  

1637: \begin{equation}

1638: \omega_i=\frac{\beta^2_i}{2L^2}\sqrt{ \frac{Y (a^2+b^2)}{\rho}},

1639: \end{equation}   

1640: where $a$ and $b$ are the outer and inner  diameters, $L$ the length,

1641: $Y$ the Young's modulus and $\rho$ the density of the tube. The factors

1642: $\beta_i$ arise from geometrical considerations and are the solutions

1643: of the equation $\cos(\beta_i)\cosh(\beta_i)=-1$. 

1644: 

1645: The energy of the cantilever can be written in terms of its classical normal

1646: modes as 

1647: \begin{equation}

1648: H=\sum_i\left(\frac{1}{2}m\omega^2_i q^2_i 

1649: +\frac{p^2_i}{2m}\right)I_i,

1650: \end{equation}

1651: where 

1652: \begin{equation}

1653: I_i=\frac{1}{L}\int_0^L\Gamma_i^2{d}x

1654: \end{equation}

1655: and

1656:  \begin{eqnarray}

1657: \Gamma_i(L-x)=&&[\cos(\beta_i)+\cosh(\beta_i)][\sin(\beta_i

1658: x/L)-\sinh(\beta_i x/L)]\nonumber\\

1659: &&-[\sin(\beta_i)+\sinh(\beta_i)][\cos(\beta_i x/L)-\cosh(\beta_i x/L)]

1660: \end{eqnarray}

1661: are the vibrational mode

1662: eigenfunctions.\cite{modes} If we modify the definitions of the

1663: canonical variables slightly to define $q_i'=q_iI_i^{1/2}$ and 

1664: $p_i'=p_iI_i^{1/2}$, then the Hamiltonian takes the conventional form 

1665: \begin{equation}

1666: H=\sum_i\left(\frac{1}{2}m\omega^2_i (q'_i)^2 +\frac{(p'_i)^2}{2m}\right).

1667: \end{equation}

1668: We can add in the additional potential energy due to small amplitude

1669: deflections in the electric field by expanding the displacement $z$ in

1670: terms of the normal modes,

1671: \begin{equation}

1672: V(z)=-\sum_i\frac{ e cV d_0}{4\pi\epsilon_0 I_i^{1/2}}

1673: \left(\int_0^L  \frac{\Gamma_i(x){d}x}{[(x-L)^2+d_0^2]^{3/2}}

1674: \right)q'_i, 

1675: \end{equation}

1676: where the origin of $x$ has been shifted.

1677: 

1678: We can  quantise the full Hamiltonian for the cantilever in the

1679: usual way, and so we are able to associate a position operator

1680: of the form  

1681: \begin{equation}

1682: \hat{q}'_i=\left(\frac{\hbar}{2m\omega_i}\right)^{1/2}(\hat{a}^{\dagger}_i

1683: +\hat{a}_i)

1684: \end{equation}

1685: with each mode.

1686: Thus, it follows that the interaction between the electron on the dot and

1687: the cantilever can be written in the form proposed [Eq.\ (\ref{coup})]

1688: with a  coupling constant between the cantilever and the

1689: dot which depends on the mode we are considering:

1690: \begin{equation}

1691: \lambda_i= e 

1692: E\xi_i\sqrt{\frac{\hbar}{2m\omega_i}},\label{coupling}

1693: \end{equation}

1694: where $\xi_i$ is a dimensionless constant of order unity defined

1695: by the relation

1696: \begin{equation}

1697: \xi_i=\frac{d_0^2}{(2I_i)^{1/2}}

1698: \int_0^L  \frac{\Gamma_i(x){d}x}{[(x-L)^2+d_0^2]^{3/2}}.

1699: \end{equation}

1700: 

1701: We can see from Eqs. (\ref{tqd}) and (\ref{ex2}) that the

1702:  effect  of a particular cantilever mode on the electron

1703: interference in the which--path device depends on

1704: $(\lambda_i/\hbar\omega_i)^2$, rather than on $\lambda_i$ alone. Thus, the

1705: ratio of the coupling between the fundamental and the $i$th excited mode goes

1706: as $(\omega_0/\omega_i)^{3}=(\beta_0/\beta_i)^{6}$. Because the electric field

1707: is strongly limited, we will always be working in the regime where the

1708: coupling constant is just sufficient to cause detectable effects. 

1709: Therefore, since $(\beta_0/\beta_1)^6\sim0.005$, our assumption that only

1710: the fundamental mode is relevant is  justified.   

1711: 

1712: As a concrete example, we consider a nanotube cantilever of length 

1713: $1.4$~$\mu$m and 

1714: outer radius $3.3$~nm.\cite{nano}

1715: In this case, the fundamental frequency is $140$~MHz and the mass is of order 

1716: $8\times 10^{-20}$~kg. If the tip--dot distance is set at $\sim 

1717: 0.1$~$\mu$m, then 

1718: $\xi_0\sim 1.3$ and for a maximum electric field of $10^5$~V/m, the corresponding  

1719: maximum coupling constant $\kappa=\lambda/\hbar\omega_0\sim 3$. 

1720: 

1721: 

1722: 

1723: \section{Non--Perturbative Calculation of Environmental Coupling}

1724: 

1725: In Sec. V, we calculated the transmission amplitude through the dot, 

1726: including the effect of the cantilever's environment, by expanding to 

1727: second order in the interaction between the cantilever and 

1728: the electron on the dot, $\lambda$, and then re-exponentiating.

1729: In  this appendix, we justify this result with a full calculation.

1730: The fact that a second order expansion in $\lambda$ leads to the exact 

1731: result is at first somewhat surprising. 

1732: However, the important step is the re-exponentiation of the truncated series 

1733: expansion (\ref{if}): we implicitly equate the 

1734: influence functional with not just the first few terms in an expansion, but 

1735: with an infinite series of terms which are themselves 

1736: composed of products of the second order terms we evaluated. This procedure is 

1737: generally acceptable as an approximation in the limit of small $\lambda$, but

1738:  in this particular  case the series generated in fact coincides with the exact form 

1739:  obtained from a linked cluster expansion.\cite{Mahan}

1740:  

1741:  

1742: We begin by adopting the notation

1743: \begin{eqnarray*}

1744: \hat{O}(t)&=&-{\mathrm{e}}^{(-i\omega_0- \gamma)t}\int_0^t dt' 

1745: \hat{F}(t'){\mathrm{e}}^{(i\omega_0+\gamma)t'} \\

1746: \hat{O}^{\dagger}(t)&=&-

1747: {\mathrm{e}}^{(i 

1748: \omega_0-\gamma)t}\int_0^t dt' 

1749: \hat{F}^{\dagger}(t'){\mathrm{e}}^{(-i\omega_0+\gamma)t'},

1750: \end{eqnarray*}

1751: so that 

1752: \begin{eqnarray*}

1753: \hat{a}(t)&=&\hat{a}(0){\mathrm{e}}^{(-i\omega_0-\gamma)t}+\hat{O}(t) \\

1754: \hat{a}^{\dagger}(t)&=&\hat{a}^{\dagger}(0){\mathrm{e}}^{(+i\omega_0-\gamma)t}+

1755: \hat{O}^{\dagger}(t).

1756: \end{eqnarray*}

1757: The operators $\hat{O}(t)$ and $\hat{O}^{\dagger}(t)$ operate only on the 

1758: variables of the oscillator bath and they both commute with $\hat{a}(0)$ and 

1759: $\hat{a}^{\dagger}(0)$ which operate on the states of the cantilever alone. 

1760: 

1761: The object which we wish to evaluate is the influence functional

1762: \begin{equation}

1763: \left\langle{\mathrm{T}}_t {\mathrm{e}}^{-i \int_0^{t}W_I(t')dt'/\hbar} 

1764: \right\rangle

1765: =\left\langle {\mathrm{T}}_t 

1766: {\mathrm{e}}^{i\lambda 

1767: \int_0^{t}\left[\hat{a}(0){\mathrm{e}}^{(-i\omega_0-\gamma)t'}+\hat{a}

1768: ^{\dagger}(0){\mathrm{e}}^{(+i\omega_0-\gamma)t'}\right]dt'/\hbar}\right\rangle

1769: \left\langle {\mathrm{T}}_t 

1770: {\mathrm{e}}^{i\lambda

1771:  \int_0^{t}\left[\hat{O}(t')+\hat{O}^{\dagger}(t')\right]dt'/

1772:  \hbar}\right\rangle, \label{c2}

1773: \end{equation}

1774: where we have exploited the commutation properties of the operators to 

1775: factor the expression into two terms, which we can now write as $C_1\times C_2$.

1776: 

1777: 

1778: The method we use to evaluate $C_1$ and $C_2$ is based on 

1779: the application of Wick's theorem, as described in the books by

1780:  Mahan\cite{Mahan}  

1781: and Louisell.\cite{Lou} We can write the first term as

1782: \begin{equation}

1783: C_1=\left\langle {\mathrm{T}}_t 

1784: {\mathrm{e}}^{i\lambda 

1785: \left[\int_0^{t}\hat{a}(0){\mathrm{e}}^{(-i\omega_0-\gamma)t'}+\hat{a}

1786: ^{\dagger}(0){\mathrm{e}}^{(+i\omega_0-\gamma)t'}\right]dt'/\hbar}\right\rangle=

1787: \sum_{n=0}^{\infty}i^n U_n(t),

1788: \end{equation}

1789: where 

1790: \begin{eqnarray*}

1791: U_n(t)&=&\frac{(\lambda/\hbar)^n}{n!}\int_0^t dt_1 \ldots\int_0^t

1792:  dt_n 

1793: \left\langle {\mathrm{T}}_t \left[

1794: \hat{a}(0){\mathrm{e}}^{(-i\omega_0-\gamma)t_1}+\hat{a}

1795: ^{\dagger}(0){\mathrm{e}}^{(+i\omega_0-\gamma)t_1}\right]\ldots\right. \\

1796: &&\times\left.\left[\hat{a}(0){\mathrm{e}}^{(-i\omega_0-\gamma)t_n}+\hat{a}

1797: ^{\dagger}(0){\mathrm{e}}^{(+i\omega_0-\gamma)t_n}\right]\right\rangle. 

1798: \end{eqnarray*}

1799: An obvious simplification arises from the fact that the averages over odd 

1800: numbers of operators will always vanish and so we can replace $n$ by the even 

1801: index $2m$. 

1802: 

1803: We now apply Wick's theorem which allows us to write the average of products 

1804: of pairs of operators as products of the averages of pairs of operators. Thus 

1805: for $U_{2m}$, we have

1806: \begin{equation}

1807: U_{2m}=\frac{(\lambda/\hbar)^{2m}i^{2m}}{(2m)!}\int_0^t dt_1 \ldots

1808: \int_0^tdt_{2m} \sum_C\left\{ D_0(t_1-t_i)...D_0(t_j-t_{2m}) \right\}, 

1809: \end{equation}

1810: where the summation is over all possible pairing combinations of the $2m$ time 

1811: labels and 

1812: \begin{equation}

1813: iD_0(t_1-t_2)=\left\langle {\mathrm{T}}_t 

1814: \left[\hat{a}(0){\mathrm{e}}^{(-i\omega_0-\gamma)t_1}+\hat{a}

1815: ^{\dagger}(0){\mathrm{e}}^{(+i\omega_0-\gamma)t_1}\right] 

1816: \left[\hat{a}(0){\mathrm{e}}^{(-i\omega_0-\gamma)t_2}+\hat{a}

1817: ^{\dagger}(0){\mathrm{e}}^{(+i\omega_0-\gamma)t_2}\right]\right\rangle.

1818: \end{equation}

1819: The summation over all possible combinations for each term allows us to 

1820: re-exponentiate so that $

1821: C_1={\mathrm{e}}^{-\phi_0(t)}$, 

1822: where

1823: \begin{equation}

1824: \phi_0(t)=\frac{i}{2}\left(\frac{\lambda}{\hbar}\right)^2\int_0^tdt_1

1825: \int_0^tdt_2 D_0(t_1-t_2).

1826: \end{equation}

1827: It is important to notice that the exact form of the function of time and 

1828: frequency multiplying the operators $\hat{a}(0)$ and $\hat{a}^{\dagger}(0)$ is 

1829: unimportant. These functions give the operators their individual time labels, 

1830: but because they are just algebraic functions they do not affect the validity of 

1831: Wick's theorem. 

1832: 

1833: 

1834: Now we must consider the second term,

1835: \begin{equation}

1836: C_2=\left\langle{\mathrm{e}}^{i\lambda\int_0^t\left[\hat{O}(t)+\hat{O}^{\dagger}

1837: (t)\right]{d}t'/\hbar}\right\rangle. 

1838: \end{equation}

1839: We can make progress by separating out the underlying operators of the 

1840: oscillators in the bath:

1841: \begin{eqnarray}

1842: \hat{O}(t)&=&-{\mathrm{e}}^{(-i\omega_0-\gamma)t}\int_0^tdt'\sum_{\omega}

1843: \hat{A}(\omega,0)g(\omega){\mathrm{e}}^{-i\omega t'}

1844: {\mathrm{e}}^{(i\omega_0 +\gamma)t} \\

1845: &=&\sum_{\omega}\hat{A}(\omega,0)g(\omega)\left[-{\mathrm{e}}^{(-i\omega_0

1846: -\gamma)t} \int_0^tdt'{\mathrm{e}}^{-i\omega t'}

1847: {\mathrm{e}}^{(i\omega_0 +\gamma)t'}\right] \\

1848: &=&\sum_{\omega}\hat{A}(\omega,0)g(\omega)f(\omega,\omega_0,\gamma,t)

1849: \end{eqnarray} 

1850: and similarly,

1851: \begin{equation}

1852: \hat{O}^{\dagger}(t)=\sum_{\omega}\hat{A}^{\dagger}

1853: (\omega,0)g^*(\omega)f^*(\omega,\omega_0,\gamma,t).

1854: \end{equation}

1855: We do not need to calculate $f$ explicitly since it is just an algebraic

1856: function and so always commutes. Because the oscillators in the heat bath 

1857: are all non--interacting, almost all the $\hat{A}(\omega,0)$ and 

1858: $\hat{A}^{\dagger}(\omega,0)$ operators commute. The only exceptions are 

1859: annihilation and creation operators of the same frequency. This means that we 

1860: can write the expectation value as a product over all the frequencies,

1861: \begin{equation}

1862: C_2=\prod_{\omega}\left\langle 

1863: {\mathrm{e}}^{i\lambda\int_0^t\left[\hat{A}(\omega,0)g(\omega)f+\hat{A}^{\dagger}

1864: (\omega,0)g^*(\omega)f^*\right]dt'/\hbar}\right\rangle. 

1865: \end{equation}

1866: 

1867: The advantage of decoupling $C_2$ into a product of terms is that each of these 

1868: terms can be handled in the same way as $C_1$. Because $\hat{A}(\omega,0)$ and 

1869: $\hat{A}^{\dagger}(\omega,0)$ are boson operators, Wick's theorem can again be 

1870: applied so that eventually we obtain

1871: \begin{equation}

1872: C_2=\prod_{\omega}{\mathrm{e}}^{-\phi_{\omega}(t)},

1873: \end{equation}

1874: where

1875: \begin{equation}

1876: \phi_{\omega}(t)=\frac{i\lambda^2}{2\hbar^2}\int_0^tdt_1

1877: \int_0^tdt_2 D_{\omega}(t_1-t_2),

1878: \end{equation}

1879: with 

1880: \begin{eqnarray}

1881: iD_{\omega}(t_1-t_2)&=&\left\langle {\mathrm{T}}_t 

1882: \left[\hat{A}(\omega,0)g(\omega)f(\omega,\omega_0,\gamma,t_1)+\hat{A}

1883: ^{\dagger}(\omega,0)g^*(\omega)f^*(\omega,\omega_0,\gamma,t_1)\right]\right. \\ \nonumber

1884: &&\times\left. 

1885: \left[\hat{A}(\omega,0)g(\omega)f(\omega,\omega_0,\gamma,t_2)+\hat{A}

1886: ^{\dagger}(\omega,0)g^*(\omega)f^*(\omega,\omega_0,\gamma,t_2)\right]\right\rangle.

1887: \end{eqnarray}

1888: 

1889: The overall expression for the influence functional can now be written in a 

1890: simplified form

1891: \begin{equation}

1892: \left\langle{\mathrm{T}}_t {\mathrm{e}}^{-i 

1893: \int_0^{t}W_I(t')dt'/\hbar}\right\rangle=

1894: {\mathrm{e}}^{-\left[\phi_0(t)+\sum_{\omega}\phi_{\omega}(t)\right]}

1895: ={\mathrm{e}}^{-\phi(t)},

1896: \end{equation}

1897: where

1898: \begin{equation}

1899: \phi(t)=\frac{i\lambda^2}{\hbar^2}\int_0^tdt_1

1900: \int_0^{t_1}dt_2 \left\{D_0(t_1-t_2)+\sum_{\omega}D_{\omega}

1901: (t_1-t_2)\right\}.

1902: \end{equation}

1903: However, because the operators in the bath are independent, only the averages 

1904: including  

1905: pairs of operators from the same oscillator are non--zero. Thus, we can simplify 

1906: the sum over $D_{\omega}(t)$ functions: 

1907: \begin{eqnarray}

1908: \sum_{\omega}D_{\omega}(t_1-t_2)&=&\sum_{\omega}\left\langle 

1909: \left[\hat{A}(\omega,0)g(\omega)f(\omega,\omega_0,\gamma,t_1)+\hat{A}^{\dagger}

1910: (\omega,0)g^*(\omega)f^*(\omega,\omega_0,\gamma,t_1)\right]\right.\\ \nonumber

1911: &&\left.\times\left[\hat{A}(\omega,0)g(\omega)

1912: f(\omega,\omega_0,\gamma,t_2)+\hat{A}^{\dagger}

1913: (\omega,0)g^*(\omega)f^*(\omega,\omega_0,\gamma,t_2)\right]

1914: \right\rangle\\

1915: &=&\left\langle  

1916: \left[\hat{O}(t_1)+\hat{O}^{\dagger}(t_1)\right]

1917: \left[\hat{O}(t_2)+\hat{O}^{\dagger}(t_2)\right] 

1918: \right\rangle.

1919: \end{eqnarray}

1920: 

1921: 

1922: Since the $\hat{O}(t)$ and $\hat{O}^{\dagger}(t)$ operators commute with 

1923: $\hat{a}(0)$ and $\hat{a}^{\dagger}(0)$ we can complete the process of 

1924: recombination to obtain

1925: \begin{equation}

1926: i\left\{D_0(t_1-t_2)+\sum_{\omega}D_{\omega}(t_1-t_2)\right\}

1927: =\left\langle\left[\hat{a}(t_1)+\hat{a}^{\dagger}(t_1)\right]

1928: \left[\hat{a}(t_2)+\hat{a}^{\dagger}(t_2)\right]\right\rangle. \label{dv}

1929: \end{equation}

1930: 

1931: We can now use our previous results in Eqs. (\ref{a}), (\ref{b}), and (\ref{c}) 

1932: to evaluate the averages in Eq. (\ref{dv}):

1933: \begin{equation}

1934: \left\langle\left[\hat{a}(t_1)+\hat{a}^{\dagger}(t_1)\right]

1935: \left[\hat{a}(t_2)+\hat{a}^{\dagger}(t_2)\right]\right\rangle

1936: ={\mathrm{e}}^{-\gamma (t_1-t_2)}\left[ (\overline{n}+1)

1937: {\mathrm{e}}^{-i\omega_0 (t_1-t_2)}+\overline{n}

1938: {\mathrm{e}}^{i\omega_0 (t_1-t_2)}\right].

1939: \end{equation}

1940: Thus, our final expression for the influence functional  is

1941: \begin{equation}

1942: \left\langle{\mathrm{T}}_t {\mathrm{e}}^{-i \int_0^{t}W_I(t')dt'/\hbar} 

1943: \right\rangle

1944: ={\mathrm{e}}^{-\phi(t)},

1945: \end{equation}

1946: with

1947: \begin{equation}

1948: \phi(t)=\frac{\lambda^2}{\hbar^2}\int_0^tdt_1

1949: \int_0^{t_1}dt_2 {{\mathrm{e}}}^{-\gamma (t_1-t_2)}\left[ 

1950: (\overline{n}+1)

1951: {\mathrm{e}}^{-i\omega_0 (t_1-t_2)}+\overline{n}

1952: {\mathrm{e}}^{i\omega_0 (t_1-t_2)}\right]. \label{phic}

1953: \end{equation}

1954: Comparing Eq. (\ref{phic}) with Eq. (\ref{phi}), it is clear that the expression

1955:  we obtained for the influence functional in Sec. V is exact.

1956: 

1957: 

1958: 

1959: 

1960: \begin{references}

1961: \bibitem[*]{auth}Electronic address: a.armour@ic.ac.uk

1962: \bibitem[\dag]{auth2}Electronic address: miles.p.blencowe@dartmouth.edu

1963: 

1964: \bibitem{hist}N. Bohr, in {\emph{Albert Einstein: Philosopher

1965:       Scientist}}, edited by P.A. Schilpp, (Library of Living

1966:   Philosophers, Evanston, Ill., 1949), vol. 7, p. 200.

1967: \bibitem{Fey} R. Feynman, R. Leighton and M. Sands, {\emph{The Feynman

1968:       Lectures on Physics}} (Addison Wesley, Mass.,  1965), vol. III.

1969: \bibitem{Buks} E. Buks, R.  Schuster, M. Heiblum, D. Mahalu, and

1970:       V. Umansky, {{Nature}} {\bf{391}}, 871 (1998). 

1971: \bibitem{op} S. D\"{u}rr, T. Nonn, and G. Rempe,   {{Nature}} 

1972: {\bf{395}}, 33 (1998). 

1973: \bibitem{SS}W. K. Wooters and W. H. Zurek, {{Phys. Rev. D}} {\bf{19}}, 473 

1974: (1979).

1975: \bibitem{Scully} M. O. Scully, B.--G. Englert, and H. Walther,  {{Nature}} 

1976: {\bf{351}}, 111 (1991). 

1977: \bibitem{Walls} P. Storey, S. Tan, M. Collett, and D. Walls,  {{Nature}} 

1978: {\bf{367}}, 626 (1994). 

1979: \bibitem{Stern} A. Stern, Y. Aharanov, and Y. Imry, {{Phys. Rev. A}} 

1980: {\bf{41}}, 3436 (1990).

1981: \bibitem{hacr} For a review of previous work on phase coherent transport

1982: and controlled dephasing see G. Hackenbroich, /cond-mat/0006361

1983: \bibitem{Yac}A. Yacoby, M. Heiblum, D. Mahalu, and H. Shtrikman 

1984: {{Phys. Rev. Lett.}} {\bf{74}}, 4047 (1995). 

1985: \bibitem{Hack} G. Hackenbroich, B. Rosenow, and H. A. Weidenm\"{u}ller,

1986:   {{Europhys. Lett.}} {\bf{44}}, 693 (1998).

1987: \bibitem{Le} Y. Levinson, {{Europhys. Lett.}} {\bf{39}}, 299 (1997).

1988: \bibitem{Al} I. L. Aleiner, N. S. Wingreen, and Y. Meir,

1989:  {{Phys. Rev. Lett.}} {\bf{79}}, 3740 (1997).

1990: \bibitem{Gurvitz} S. A. Gurvitz, {{Phys. Rev. B}} {\bf{56}}, 15215 (1997).

1991: \bibitem{BJK} S. Mancini, V. I. Man'ko, and P. Tombesi,  {{Phys. Rev. A}} 

1992: {\bf{55}}, 3042 (1997);

1993: S. Bose, K. Jacobs, and P. L. Knight , {{Phys. Rev. A}} 

1994: {\bf{56}}, 4175 (1997).

1995: \bibitem{Bose}S. Bose, K. Jacobs, and P. L. Knight,

1996: {{Phys. Rev. A}}  {\bf{59}}, 3204 (1999).

1997: \bibitem{Datta} S. Datta, {\emph{Electronic Transport in Mesoscopic

1998:       Structures}}, (Cambridge University Press, Cambridge, U.K., 1995).

1999: \bibitem{But1} M. B\"{u}ttiker, IBM J. Res. Develop. {\bf{32}}, 317 (1988).

2000: \bibitem{But2} A. Levy Yeyati and   M. B\"{u}ttiker,

2001:     {{Phys. Rev. B}} {\bf{52}}, R14360 (1995).

2002: \bibitem{Imry} Y. Imry, {\emph{Introduction to Mesoscopic Physics}}, 

2003:   (Oxford University Press, Oxford, U.K., 1997). 

2004: \bibitem{Yac2}A. Yacoby, R. Schuster, and M. Heiblum,  

2005: {{Phys. Rev. B}} {\bf{53}}, 9583 (1996).

2006: \bibitem{steady}Our assumption is that the which--path device is 

2007: operated in a steady--state mode, whereby measurements are only made 

2008: after any transitory behavior has decayed so that the probability 

2009: distribution of (initial) cantilever states is stationary. 

2010: This steady--state distribution 

2011: is further assumed to be the usual thermal distribution. These assumptions will 

2012: only

2013: be strictly correct in the limit as the current goes to zero, but can be expected 

2014: to provide  predictions which are at least qualitatively correct for the average

2015:  behavior over long times  even with 

2016: finite currents.

2017: \bibitem{vis}P. D. D. Schwindt, P. G. Kwiat, and B.--G. Englert, 

2018:  {{Phys. Rev. A}} {\bf{60}}, 4285 (1999).

2019: \bibitem{justify} We assume that all positions are measured from the

2020:   equilibrium height of the cantilever above the dot, including the

2021:   effect of the residual electrons on the dot which do not participate 

2022:  in the transport. 

2023: \bibitem{W89}N. S. Wingreen, K. W. Jacobsen, and J. W. Wilkins, 

2024: {{Phys. Rev. B}} {\bf{40}}, 11834 (1989).

2025: \bibitem{GS}L. I. Glazman and R. I. Shekhter, Zh. Eksp. Teor. Fiz. 

2026: {\bf 94}, 292 (1987) 

2027: [{{Sov. Phys.---JETP}} {\bf{67}}, 163 (1988)].

2028: \bibitem{JW}A.--P. Jauho and N. S. Wingreen, 

2029:   {{Phys. Rev. B}} {\bf{58}}, 9619 (1998).

2030: \bibitem{UZ}W. G. Unruh and W. H. Zurek, 

2031: {{Phys. Rev. D}} {\bf{40}}, 1071 (1989).

2032: \bibitem{vib}P. Poncharal, Z. L. Wang, D. Ugarte, and W. A. de Heer, 

2033:  {{Science}} {\bf{283}}, 1513 (1999).

2034: \bibitem{CL} A. O. Caldeira and A. J. Leggett, 

2035: {{Ann. Phys. (N.Y.)}} {\bf{149}}, 374 (1983).

2036: \bibitem{O}R. Omn\`{e}s, {\emph{The Interpretation of Quantum Mechanics}},

2037: (Princeton University Press, Princeton, NJ, 1994).  

2038: \bibitem{JZ}E. Joos and H. D. Zeh, {{Z. Phys. B}} {\bf{59}} 223 (1985).

2039: \bibitem{MW}L. Mandel and E. Wolf, {\emph{Optical Coherence and Quantum Optics}},

2040: (Cambridge University Press, Cambridge, U.K., 1995). 

2041: \bibitem{Lou}W. Louisell, {\emph{Quantum Statistical Properties of Radiation}},

2042: (J. Wiley, New York, 1973).

2043: \bibitem{Buksp}E. Buks (private communication).

2044: \bibitem{Buks2}D. Sprinzak, E. Buks, M. Heiblum, and H. Shtrikman, 

2045:  {{Phys. Rev. Lett.}} {\bf{84}}, 5820 (2000). 

2046: \bibitem{AFM} R. M. D. Stevens, N. A. Frederick, B. L. Smith, D. E. Morse,

2047:  G. D. Stucky, and P. K. Hansma,  {{Nanotechnology}}

2048:   {\bf{11}}, 1  (2000).

2049: \bibitem{jack}J. D. Jackson, {{Am. J.  Phys.}}

2050: {\bf{68}}, 789 (2000). 

2051: \bibitem{nano} M.M.J. Treacy, T. W. Ebbesen, and J. M. Gibson,

2052:   {{Nature}} {\bf{381}}, 678 (1996).

2053: \bibitem{modes2}B. Reulet, A. Y. Kasumov, M. Kociak, R. Deblock,

2054: I. I. Khodos, Y. B. Gorbatov, V. T. Volkov, C. Journet, and H. Bouchiat,

2055:  {{Phys. Rev. Lett.}} {\bf{85}}, 2829 (2000). 

2056: \bibitem{modes}H.--J. Butt and M. Jaschke,  {{Nanotechnology}}

2057:   {\bf{6}}, 1  (1995).

2058: \bibitem{Mahan}G. Mahan, {\emph{Many--Particle Physics}}, 

2059: (2nd edition, Plenum Press, New York, 1990).

2060: \end{references}

2061: 

2062: 

2063: 

2064: \begin{figure}

2065: \caption{Magnitude of the transmission amplitude, $|\langle t_{\rm QD} \rangle|$

2066: at resonance, for a

2067:   cantilever with $\Gamma/\hbar\omega_0=10$ as a function of the dimensionless

2068:   coupling constant

2069: $ \kappa=\lambda/\hbar\omega_0$. The dashed curve is obtained without

2070: an average over incident electron energies, while the full curve includes

2071: the averaging. The amplitudes are normalized to one at $\kappa=0$.  }

2072: \label{fig:one}

2073: \end{figure}

2074: 

2075: 

2076: \begin{figure}

2077: \caption{Magnitude of the resonant transmission amplitude, $|\langle 

2078: t_{\mathrm QD}\rangle|$,

2079:  as a function of coupling constant $\kappa$, for  

2080: $\Gamma/\hbar\omega_0$=2 (dotted curve), 1 (dashed curve) and 0.5 (full curve).}

2081: \label{fig:two}

2082: \end{figure}

2083: 

2084: 

2085: \begin{figure}

2086: \caption{(a) Magnitude of the transmission amplitude, $|\langle 

2087: t_{\mathrm QD}\rangle |$, and

2088:   (b) the phase, for a

2089:   cantilever with $\Gamma/\hbar\omega_0=0.5$  as a function of coupling constant

2090: $ \kappa$ and the energy detuning 

2091: $\epsilon-\epsilon_{0}$. The average over the electron energy

2092:   distribution has not been performed.}

2093: \label{fig:three}

2094: \end{figure}

2095: 

2096: 

2097: 

2098: \begin{figure}

2099: \caption{(a) Magnitude of the transmission amplitude, $|\langle t_{\mathrm QD}\rangle|$, and

2100:   (b) the phase, for a

2101:   cantilever with $\Gamma/\hbar\omega_0=0.5$  as a function of coupling 

2102:   constant, including

2103:   the effect of the cantilever's environment with 

2104:   $Q=50$, as a function of $\kappa$ and  

2105:   $\epsilon-\epsilon_{0}$.}

2106: \label{fig:four}

2107: \end{figure}

2108: 

2109: \begin{figure}

2110: \caption{Magnitude of the resonant transmission amplitude, $|\langle 

2111: t_{\mathrm QD}\rangle|$, 

2112:  including the effect of the cantilever's environment as a function of 

2113:  $\kappa$, for  

2114: $\Gamma/\hbar\omega_0=0.5$ and $Q=50$ (dotted curve)

2115:  and  500

2116: (dashed curve). The latter curve has been shifted upwards by 0.5 and 

2117: the former shifted upwards by 1 for clarity.

2118: The case without environmental coupling where  

2119: $\gamma_{c}/\omega_0$=0 (full curve) is included for comparison.}

2120: \label{fig:five}

2121: \end{figure}

2122: 

2123: \vfill

2124: \eject

2125: \mbox{\epsfig{file=which-pathplots1.nb.ps, width=6in}}

2126: \mbox{\epsfig{file=which-pathplots2.nb.ps, width=6in}}

2127: \mbox{\epsfig{file=wp3a.ps, width=6in}}

2128: \mbox{\epsfig{file=wp3b.ps, width=6in}}

2129: \mbox{\epsfig{file=wp4a.ps, width=6in}}

2130: \mbox{\epsfig{file=wp4b.ps, width=6in}}

2131: \mbox{\epsfig{file=which-pathplots5.nb.ps, width=6in}}

2132: 

2133: 

2134: 

2135: 

2136: \end{document}

2137: