1: \documentclass[pra]{revtex4}

2: \usepackage{graphicx}

3: 

4: \begin{document}

5: 

6: 

7: \title{Running-phase state in a Josephson washboard potential}

8: 

9: \author{G.~S.~Paraoanu}

10: \affiliation{NanoScience Center and Department of Physics, University of Jyv\"askyl\"a,

11: P.O.~Box 35 (YFL), FIN-40014 University of Jyv\"askyl\"a, FINLAND}

12: 

13: 

14: \begin{abstract}

15: 

16:  

17: We investigate the dynamics of the phase variable of an ideal underdamped Josephson junction in switching 

18: current 

19: experiments. These experiments have provided the first evidence for macroscopic quantum tunneling in large 

20: Josephson junctions and are currently used for state read-out of superconducting qubits.

21: We calculate the shape of the resulting macroscopic wavepacket and find that

22: the propagation of the wavepacket long enough after a switching event leads to an average 

23: voltage increasing linearly with time.

24: 

25: 

26: 

27: \end{abstract}

28: 

29: \maketitle

30: 

31: %\section{Introduction}

32: 

33: 

34: 

35: The dynamics of a large underdamped Josephson junction characterized by a 

36: capacitance $C$ and Josephson energy $E_{J}$ can be described by the motion of a particle in a 

37: washboard potential $U(\gamma )= E_{J}(1 - \cos\gamma ) + I\bar\Phi_0 \gamma $. 

38: The particle has $C$ as the mass, the flux $\bar{\Phi}_{0}\gamma$ as the coordinate

39: and the charge on the capacitor $Q$ as the canonically conjugate momentum. Here $\gamma$ is the

40: phase difference of the superconducting order parameter across the junction and $\bar \Phi_{0} = 

41: \Phi_{0} /2\pi = \hbar/2e$ is the flux quanta divided by $2\pi$. Much attention has been given since the 

42: discovery of the Josephson effect to the switching dynamics of the junction in the thermal activation 

43: regime and in the macroscopic tunneling (MQT) regime. Surprisingly, while the description of the state

44: of the junction before a switching event and calculations of the corresponding probability has been a topical issue

45: for many decades, what happens with the quantum state of the junction after tunneling did not receive that 

46: much attention. It is argued \cite{tinkham,switch} that the junction ends up in a running-phase state, 

47: with the voltage $Q/C$ increasing until it becomes sufficiently large so that the transport could be 

48: done through quasiparticle excitations. 

49: However, a quantum mechanical description of this state is missing. Much of what we understand about 

50: the running-wave state, for instance

51: the physics of the retrapping current, comes from assuming a quasi-classical dynamics.

52: 

53: In this paper we give an explicit formula for the macroscopic wavefunction of an ideally underdamped

54:  junction after a MQT switching event. 

55: If the switching probability is exponential, 

56: which is the case for all the theoretical models and also confirmed experimentally,

57: one expects \cite{suppl} the following expression for the dynamics of the wavefunction

58: \begin{equation}

59: |\Psi (t)\rangle = e^{-\Gamma t /2}e^{-i\omega_0 t}|\Psi_{0}\rangle + |\Psi_{out}(t)\rangle.

60: \label{psii}

61: \end{equation} 

62: In this equation, $\Psi_{0}$ is the initial state, coresponding to a bound state inside one of the 

63: metastable wells, while $\Psi_{out}(t)$ is the wavefunction of the particle corresponding to states in the 

64: continuum, outside the well (Fig. \ref{tfig}). This expression gives indeed an exponentially decreasing 

65: probability for

66: the particle to be inside the well, with lifetime $\Gamma ^{-1}$.

67: In the following, we are interested in the structure of $\Psi_{out}(t)$. 

68: 

69: To solve this problem, the standard approach is to start with a wavepacket localized initially in one of the 

70: metastable wells, and then expand it and evolve it in the eigenfunctions of the full Hamiltonian.

71: This procedure works for simple potentials \cite{deltapotential}, but even in these cases the solutions are

72: complicated. Fortunately, unlike problems in scattering theory, in condensed

73: matter the frequent situation is that we do not need an exact solution of the Scr\"odinger problem for tunneling,

74: but rather we are interested in the most generic features of it. In most cases in solid state physics, 

75: tunneling is simply

76: treated as a process that annihilates a particle on some mode of a solid and creates one on another mode.

77: We will approach our problem in the same spirit \cite{cond}.

78: A good approximation in MQT is that no other state within the well is involved

79: with the exception of the state with energy $\omega_0$ in which the system is prepared, $|\Psi_0 \rangle$; 

80: therefore one can write a reduced Hamiltonian of the form

81: \begin{equation}

82: H = \hbar\omega_0 |\Psi_0 \rangle\langle \Psi_0 | + \int \hbar\epsilon |\psi_\epsilon \rangle\langle 

83: \psi_\epsilon | + \int d\epsilon \left[ k(\omega_0, \epsilon ) |\Psi_0 \rangle\langle 

84: \psi_\epsilon |

85: + k(\epsilon , \omega_0 ) |\psi_\epsilon \rangle\langle \Psi_0 |\right], \label{hammi}

86: \end{equation}

87: where by $\{\psi_\epsilon  \}$ we denote the continuum of eigenvectors outside the barrier. 

88: We then write the wavefunction in the form

89: \begin{equation}

90: |\Psi (t) \rangle = a (t) e^{-i\omega_0 t} |\Psi_0\rangle + \int d\epsilon b (\epsilon, t) e^{-i\epsilon t}

91: |\psi_\epsilon \rangle ,

92: \label{lapl}

93: \end{equation}

94: with $a(0) = 1$, $b (0) = 0$. Inserting this expression in the Schr\"odinger equation we get

95: an integro-differential equation for $a (t)$. The Laplace transform of this equation reads

96: \begin{equation}

97: {\cal L} [a] (s) = \frac{1}{s + {\cal L}[\chi ](s)},\label{laplace}

98: \end{equation}

99: where

100: \begin{equation}

101: \chi (t) = \frac{1}{\hbar ^2}\int d\epsilon |k (\epsilon ,\omega_0 )|^2e^{i (\omega_0 - \epsilon )t}.

102: \end{equation}

103: In general, the tunneling matrix element $k(\omega_0, \epsilon)$ 

104: depend on the energies $\epsilon$ and they are determined by

105: the overlap of the left and right wavefunctions under the barrier

106: \cite{cond}.

107: We notice that since typically the lifetime of the metastable states is much larger than the

108: oscillation period in the well (in other words the last term in the Hamiltonian is a perturbation),

109:  the states $\{|\epsilon \rangle \}$ which contribute effectively to tunneling

110: are located in a relatively small energy interval compared to the plasma oscillation frequency, 

111: therefore the shape of these states under the barrier

112: is approximately identical. We can then take

113: the tunneling matrix element as being a complex constant; but since we

114: will be interested exclusively in the outgoing component, the

115: relative phase between the wavefunction inside the well and that outside

116: will not play any role. We have confirmed this assumption also by 

117: expanding the initial wavefunction in terms of the WKB solution

118: of the washboard potential calculated in \cite{josephson}.

119: Therefore

120: we take $k = \hbar\sqrt{\Gamma /2\pi}$ real; 

121: we obtain ${\cal L}[\chi] (s) = \Gamma /2$= constant, which turns out to be

122:  the decay probability of the system. Indeed, the inverse Laplace transform of  Eq. (\ref{laplace}) gives 

123: precisely the classical exponential decay law

124: \begin{equation}

125: a (t) = e^{-\Gamma t/2}.

126: \end{equation}

127: The outgoing wavepacket becomes 

128: \begin{equation}

129: |\Psi_{out} (t) \rangle = -i\sqrt{\frac{\Gamma }{2\pi}}\int d\epsilon \frac{e^{(-\Gamma /2 + i \omega_0 )t}- e^{-i\epsilon t}}{\Gamma /2 + i (\omega_{0} -\epsilon )}

130: |\psi _\epsilon^{\pm} \rangle .

131: \label{q}

132: \end{equation}

133: To conclude this derivation, we find that with the identification $k = \hbar\sqrt{\Gamma /2\pi}$ the Hamiltonian

134: Eq. (\ref{hammi}) becomes a model Hamiltonian for decay in the continuum which can be solved exactly, with 

135: solution given by Eqs. (\ref{psii}) and (\ref{q}). A similar type of model Hamiltonian has been obtained in 

136: \cite{twopotential,kurizkikofman}.

137: One can show, using the properties of the 

138: Lorentz distribution, that

139: these wavefunctions are correctly normalized, as explained above. 

140: 

141: Let us now single out one component of the wave $\Psi_{out} (t)$, namely

142: \begin{equation}

143: \Psi^{\rightarrow}_{out} = i\sqrt{\frac{\Gamma }{2\pi}}\int d\epsilon \frac{e^{-i\epsilon t}}

144: {\Gamma /2 + i (\omega_{0} -\epsilon )}

145: |\psi _\epsilon^{+} \rangle .

146: \end{equation}

147: We first notice that the normalization of the total function $|\Psi_{out}\rangle$ is such that 

148: $\langle \Psi_{out}|\Psi_{out}\rangle = 1 - \exp (-\Gamma t)$, which reflects correctly the fact that

149: the probability of finding the particle outside comes from an exponential decay law, 

150: while that of $|\Psi^{\rightarrow}_{out}\rangle$ 

151: is such that $\langle \Psi^{\rightarrow}_{out}|\Psi^{\rightarrow}_{out}\rangle = 1$. In the following

152: we will see that $|\Psi^{\rightarrow}_{out} \rangle$ plays indeed a special role. 

153: \begin{figure}[htb]

154: \includegraphics[width=84truemm]{tfigure.eps}

155: \caption{Tunneling out of one of the metastable wells of the washboard potential. }

156: \label{tfig}

157: \end{figure}         

158: To move on, we notice that part of the expression for the outgoing phase contains a term 

159: which decays exponentially on a time scale $\Gamma^{-1}$. These terms are associated with the 

160: fast components of the localized wavefunction which would escape first. 

161: Although a calculation that includes these terms is no doubt interesting, especially for the problem of

162: non-exponential decay rates \cite{deltapotential}, 

163: in what follows 

164: we will regard them as transient oscillatory effects whose presence will be difficult to assess experimentally 

165: anyway, and we will neglect their contribution.  

166: In the WKB approximation, far enough from the classical turning point, the eigenvalues 

167: $\{ |\psi _\epsilon \rangle \}$ have the form (up to a normalization factor and constant phase factors 

168: due to matching to the region left of the classical turning point) of incoming and outgoing scattering states

169: \begin{equation}

170: \psi^{\pm}_\epsilon (\gamma ) \approx \sqrt{\frac{e}{C\bar\Phi _0 V_{\epsilon}(\gamma )}}\exp\left[\pm i\frac{C}{e}

171: \int_{\gamma_0}^{\gamma}V_{\epsilon }(\varphi)d\varphi \right] ,\label{wkb}

172: \end{equation}

173: where 

174: \begin{equation}

175: V_{\epsilon}(\gamma ) = \frac{\hbar\omega_p}{\sqrt{2}e}\sqrt{\frac{2e\epsilon  + I \gamma }

176: {I_{c0}} + 1 - \cos\gamma }.

177: \end{equation}

178: The physical meaning of this voltage is that it corresponds to the (classical) energy accumulated 

179: on the capacitor when the phase difference across the junction is $\gamma$ and the initial energy

180: of the system is $\epsilon$; indeed, $CV^{2}_{\epsilon}(\gamma )/2 = \hbar\epsilon - U(\gamma )$. 

181: We now use the fact that for values of $\gamma$ outside the well and far enough from the classical turning

182: point the inequality $|\hbar\epsilon - \hbar\omega_{0}|\ll \hbar\omega_0 - U(\gamma )$ holds. 

183: We can therefore take $V_\epsilon (\gamma ) = V_{\omega_0}(\gamma )\stackrel{\rm not}{=}

184: V_{0}(\gamma )$ in the denominator of Eq. (\ref{wkb})

185: and approximate the exponent as

186: \begin{equation}

187: V_{\epsilon} (\gamma ) \approx V_{0}(\gamma )\left[1 + \frac{\epsilon - \omega_{0}}{2( 

188: \omega_{0}- \hbar^{-1}U( \gamma ))}\right].

189: \end{equation}

190: With these approximations, using Eqs. (\ref{q}) and (\ref{wkb}) we can performe the integral over 

191: the angular frequencies $\epsilon$; as a result, the contribution of the in-going scattering states in zero, 

192: while the out-going scattering states build up a wavepacket of the form 

193: \begin{equation} 

194: \Psi^{\rightarrow}_{out}(\gamma ,t) = \frac{{\cal N}}{\sqrt{V_{0}(\gamma )}}

195: \exp\left[\frac{i}{\hbar}\int_{\gamma_{0}}^{\gamma } 

196: C \tilde{V}(\varphi )\bar\Phi_0 d\varphi -

197: \left(i\omega_{0} + 

198: \frac{\Gamma}{2}\right)\left(t-\int_{\gamma_0}

199: ^{\gamma}

200: \frac{\bar \Phi_{0}d\varphi }{V_{0}(\varphi)}\right)\right]

201: \Theta\left[t - \int_{\gamma_0}^{\gamma}

202: \frac{\bar \Phi_{0}d\varphi }{V_{0}(\varphi)}\right]. 

203: \label{out}

204: \end{equation}

205: Here ${\cal N}$ is a normalization factor which can be obtained through $\int_{-\infty}^{\infty} 

206: d(\bar\Phi_{0}\gamma )|\Psi^{\rightarrow}_{out}(\gamma ,t)|^2 = 1$

207: with the mention that we make a negligible error by extending the integral to $-\infty$, 

208: {\it i.e.} before the well region (where the actual values are exponentially small). The voltage

209: $\tilde{V}(\gamma )$ is defined as

210: $\tilde{V}(\gamma )= V_{0}(\gamma ) - \hbar\omega_0 /C V_{0}(\gamma )\approx V_{0}(\gamma )$.

211: It is interesting to see also what happens with the rest of the components of $|\Psi_{out}\rangle$.

212: Although they do contribute to the normalization as discussed before, they are decaying both in time and 

213: away from 

214: the barrier as $\exp [-\frac{\Gamma}{2}\int_{\gamma_0}^{\gamma}V_{0}^{-1}(\varphi)

215: \bar \Phi_{0}d\varphi ]$

216: which, as we will see below, would give far from the barier a factor of $\exp [-

217: \Gamma /2(t + \sqrt{2\gamma}/\omega_{p})]$. It is clear that these terms can be neglected starting

218: roughly from a time $\Gamma^{-1}$.   

219: The wavefunction Eq. (\ref{out}) contains all the information about the dynamical evolution of the state of the

220: circuit containing the Josephson junction, and it is the main result of this paper. 

221: 

222: In a typical experiment, 

223: the voltage across the junction is 

224: monitored by a voltmeter at room temperature. A fundamental issue is to find a microscopic mechanism for the junction-voltmeter interaction and 

225: a suitable theory of quantum measurement that would

226: model the collapse of the wavefunction; this is however beyond the scope of this paper. 

227: Still, Eqs. (\ref{out}) and (\ref{prob}) give a quite clear qualitative  

228: picture of what happens: the 

229: particle rolls down the washboard potential with a quasi-classical speed given by energy conservation

230: $CV^{2}_{0}(\gamma )/2 = \hbar\omega_0 - U(\gamma )$. Quantum mechanics enters in the picture through the tunneling rate; we expect the results

231: of the measurements 

232: to have a spread dermined by $\Gamma$.

233: One can assume that the measurement projects the outgoing state onto eigenvalues of the voltage

234: operator; therefore the  

235: probability of recording the value $V$ at the moment $t$ will be given by the standard

236: quantum mechanics recipe

237: \begin{equation}

238: P(V, t) = 

239: \frac{1}{2\pi\hbar}\vert \int_{-\infty}^{\infty} d(\bar\Phi_0\gamma )\Psi_{out}^{*}(\gamma ,t)\exp{(iVC\gamma /2e)} \vert ^2.

240: \label{prob}

241: \end{equation}

242:  As an example, had the outside the well  potential $U$ been

243: zero, we would have gotten for the charge $CV$, by performing the integration in Eq. (\ref{prob}), a 

244: standard Cauchy-Breit-Wigner distribution centered around $CV_0$ and full width at half maximum 

245: $\Gamma\hbar /V_0$

246: \begin{equation}

247: P(CV)= \frac{1}{\pi}\frac{\Gamma\hbar}{2 V_0}\left[ (CV-CV_0)^2 + 

248: \left( \frac{\Gamma \hbar}{2V_0}\right)^2 \right]^{-1}.

249: \end{equation} 

250: 

251: %A better way to measure the voltage would be however through a less invasive method, such as

252: %coupling the junction capacitively to an RF-SET read-out.

253: 

254: %Therefore the same phenomena happens in Eq. [\ref{out}]: there will be a spreading of the values

255: %measured by the voltmeter around the instantaneous classical values $V_{0}(\gamma (t) )$, where

256: %$\gamma (t)$ is given by solving the equation $\int_{\gamma_0}^{\gamma (t)}

257: %V^{-1}_{0}(\varphi ) \bar\Phi_0 d\varphi = t$.

258: 

259: Considering again the case of a junction with a washboard potential $U(\gamma )$, we notice that

260: a good approximation is $U (\gamma ) \approx E_{J}\gamma $. This comes from the fact that switching is 

261: typically observed at values of the bias current close to the critical current of the junction, as well

262: as from the observation that for times larger than $\Gamma ^{-1}$ the wavepacket is concentrated at

263: large values of $\gamma \gg 1$, in which case the $\cos\gamma$ term in the potential is negligible. 

264: In other words, the particle gets soon so fast that the "speed bumps" created by the Josephson 

265: effect are not slowing it down significantly. This can be checked {\it a posteriori}. A first observation

266: is that the relevant quantity for the dynamics of the center of the wavepacket is the argument of the

267: $\Theta $ function; the condition that this argument vanishes sets the maximum value

268: of $|\Psi_{out}|^2$ and gives a phase $\omega_{p}t/2\gg 1$ for $t$ larger than $\Gamma^{-1}$. 

269: A legitimate concern is whether the wavefunction does not spread faster than it moves downwards. This 

270: is not the 

271: case, as we will see below: the spread of the wavefunction increases linearly with time, while the 

272: average coordinate (phase) is advancing as $t^2$. With these observations, the normalization constant

273: can be calculated, and the outgoing wavepacket becomes

274: \begin{equation}

275: \psi^{\rightarrow}_{out} (\gamma ,t) = \sqrt{\frac{\Gamma}{\sqrt{2\gamma}\bar\Phi_{0}\omega_{p}}}\exp\left[i\frac{\hbar\omega_{p}

276: \gamma^{3/2}}{6\sqrt{2}E_{c}} - \left(i\frac{\omega_0}{\omega_p} + \frac{\Gamma}{2\omega_p}\right)\left(

277: t\omega_{p}-\sqrt{2\gamma}\right)\right]\Theta\left(t\omega_p - \sqrt{2\gamma}\right).\label{super}

278: \end{equation}

279: A plot of the wavefunction is given in Fig. \ref{wf}. 

280: 

281: \begin{figure}[htb]

282: \includegraphics[width=84truemm]{wavefunction.eps}

283: \caption{Contourplot $(\gamma, t)$ of the modulus of the running-phase wavefunction Eq. (\ref{super}) for $\omega_p = 60 \Gamma = 30$ GHz.}

284: \label{wf}

285: \end{figure} 

286: 

287: The average phase (flux) corresponding to this wavepacket can be obtained

288: \begin{equation}

289: \langle \gamma \rangle (t)= \omega_{p}^2(t^{2}/2 - \Gamma ^{-1}t + \Gamma^{-2}),\label{av}

290: \end{equation}

291: and we notice that the dominant term is quadratic in t. 

292: The spread of the flux variable is given by (we keep only the dominant term here)

293: \begin{equation}

294: \sqrt{\langle \gamma^2\rangle (t) - \langle \gamma \rangle^2(t)} = \omega_{p}^{2}\Gamma^{-1}t.

295: \end{equation}

296: To get the average voltage we can use Ehrenfest theorem; we obtain

297: \begin{equation}

298: \langle V(t) \rangle = \bar\Phi_{0}\omega_{p}^{2}(t -\Gamma^{-1}).  \label{avv}

299: \end{equation}  

300: The dominant term for the voltage is linear in time and satisfies the classical energy

301: conservation $C\langle V\rangle ^2(t) /2 = \hbar \omega_{0} - U(\langle \gamma \rangle (t))\approx I_{0}\langle

302: \gamma \rangle (t)$.  

303: 

304: Let us now analyze what happens in typical switching current experiments, as they are done now in the

305: context of superconducting qubits \cite{qubit}: the bias current of the junction is increased

306: fast to a value that allows tunneling, it is kept there for a time $0 < \tau < \Gamma^{-1}$, then it is lowered

307: to a value that suppresses tunneling. This value has to be large enough so that the experimentalist can get a 

308: reliable reading of voltage on the quasiparticle branch if the junction has switched; in practice, it can still satisfy

309: $I\approx I_0$. Although the change 

310: of the bias current has a major effect with respect to tunneling through the barrier, where the 

311: the tunneling rate decreases exponentially with the height of the barrier, from the point of view of the

312: structure of the running-phase state it amounts only to a modification of the parameter $\Gamma$.  

313: Finally,  

314: the current is put to zero and, after waiting long enough for retrapping to occur, the whole cycle can be 

315: repeated. In our model, the essential physics is that after the time $\tau$, the tunneling matrix element $t$ is zero,

316: therefore the system evolves only under the action of $H_0$. The wavefunction is "cut" into two separate pieces,

317: one which is (almost) the bound state $\Psi_0$ inside the well, the other being the wavepacket in the continuum

318: which evolves as

319: \begin{equation}

320: |\Psi_{out} (t) \rangle = -i\sqrt{\frac{\Gamma }{2\pi}}\int d\epsilon \frac{e^{(-\Gamma /2 + i \omega_0 )\tau }

321: e^{-i\omega_{j} (t-\tau )} 

322: - e^{-i\epsilon t}}{\Gamma /2 + i (\omega_{0} -\epsilon )}

323: |\psi _\epsilon^{\pm}\rangle ,

324: \end{equation}

325: with normalization $\langle \Psi_{out} (t)|\Psi_{out} (t) \rangle = 1- e^{-\Gamma \tau}$. Now, for $t-\tau >

326: \Gamma^{-1}$ we can see

327: that the outgoing function consists of two consecutive (separated by the time $\tau$) and dephased (with $\omega_0 \tau $)

328: outgoing wavepackets with the structure of $|\Psi^{\rightarrow}_{out} (t)\rangle$

329: which propagate at the same speed across the phase coordinate $\gamma$. The second wavepacket, which has a probability amplitude

330: smaller by a factor of $e^{-\Gamma /2\tau}$, results from the

331: waves localized near the barrier during the time $\tau$ when tunneling was in progress. 

332: After integration over energy, we get

333: \begin{equation}

334: |\Psi_{out} (t) \rangle = e^{-(\Gamma /2 + i\omega_0)\tau }|\Psi^{\rightarrow}_{out} (t -\tau )\rangle

335: - |\Psi^{\rightarrow}_{out} (t)\rangle ,\label{xx}

336: \end{equation}

337: where $|\Psi^{\rightarrow}_{out} (t)\rangle$ is given by Eqs. (\ref{out}) and (\ref{super}). To check that 

338: the normalization 

339: $\langle \Psi_{out} (t)|\Psi_{out} (t) \rangle = 1- e^{-\Gamma \tau}$ remains valid, we notice that 

340: in the region of overlap of the two wavepackets, which coincides with the domain where 

341: $|\Psi^{\rightarrow}_{out}(t-\tau )\rangle$

342: is finite $t-\tau > \int_{\gamma_0}^{\gamma}\bar \Phi_{0} V_{0}^{-1}(\varphi)d\varphi $, there exists a very 

343: simple relation between them: 

344: $|\Psi^{\rightarrow}_{out} (t) \rangle = \exp[-i\omega_0\tau -\Gamma \tau /2]|\Psi^{\rightarrow}_{out} (t-\tau )\rangle$.

345: Using this property and the previous expressions Eqs. (\ref{av}) and (\ref{avv}), we can calculate the average

346: phase and voltage on the state Eq. (\ref{xx})

347: \begin{equation}

348: \langle \gamma \rangle (t) = \frac{1}{2}\omega_{p}^{2}[1- e^{-\Gamma \tau}]t^{2} - 

349: e^{-\Gamma \tau}\omega_{p}^{2}\tau t - \frac{1}{2}e^{-\Gamma \tau}\omega_{p}^{2}, 

350: \end{equation}

351: and 

352: \begin{equation}

353: \langle V \rangle (t) = \bar\Phi_{0}\omega_{p}^{2}t\left(1- e^{-\Gamma \tau }\right)

354: -\bar\Phi_{0}\omega_{p}^{2}e^{-\Gamma \tau}\tau .

355: \end{equation}

356: In Fig. \ref{prediction} we present a plot of the average voltage as a function of the time $\tau$.

357: We see that for values of $\tau$ of the same order or larger than the lifetime $\Gamma^{-1}$

358: the average voltage at t flattens, reflecting the fact that the junction has switched, as in the case

359: of Eq. (\ref{avv}).  

360: \begin{figure}[htb]

361: \includegraphics[width=84truemm]{prediction.eps}

362: \caption{The average voltage as a function of the time $\tau$ for $\omega_p = 30$ GHz, $\Gamma = 1$ ns 

363:  and $t = 10$ ns.}

364: \label{prediction}

365: \end{figure}         

366: 

367: For designing an experiment to test these predictions, several remarks should be made. 

368: In the case of real junctions, the Josephson energy and the plasma frequency can be reduced 

369: by using a SQUID configuration and by adding capacitors in parallel with the junctions.

370: This makes the time evolution of the switching state slower and therefore easier to detect.

371: %Finally, we notice as before that we do not need to prepare the system at t=0 in the state corresponding

372: %to $I=0$; the tunneling can be suppresses to a good approximation on the time scale of the experiment

373: %(at most 40 ns as we will see below) also with a suitably large value of the critical current.

374: An important limitation on time comes from the fact that as soon as the voltage reaches the quasiparticle

375: branch (at twice the value of the gap) our analysis is not valid. The other limitation is technological: 

376: even with a good dilution refrigerator,

377: thermalizing the junction is very difficult at low temperatures. With a good high-power refrigerator

378: with base temperature of about 5 mK, we assume an optimistic value of 10 mK for the effective temperature

379: of the electrons. This temperature corresponds to a crossover angular frequency of 8.66 GHz between the MQT

380: and the thermal

381: activation transition. 

382: A plasma frequency of $\omega_p = 30$ GHz (zero bias current) will thus keep us safely in the MQT regime when 

383: the current is raised up to about half a percent close to the critical current, according to the formula that 

384: gives the plasma oscillation frequency at a finite bias current \cite{tinkham,switch}. For Nb, with gap of 1.4 meV, this 

385: corresponds to a time of approximately 10 ns, as given by Eq. (\ref{avv}). A voltage increase on this timescale 

386: can be detected with standard experimental techniques

387: Suppose now that we choose to work at currents about 5\% less than the critical current.  

388: We still have to satisfy the condition $t>\Gamma^{-1}$; an inspection of the formula that gives 

389: the tunneling rate for underdamped junctions (see {\it e.g.} \cite{switch}) shows that switching rates of about 500 MHz and more 

390: (with the restriction $\Gamma \ll \omega_p$) can be achieved for $E_J /\omega_p$ of the order of 30, values which can be obtained easily with 

391: large junctions.   

392:  

393: 

394: 

395: %In our calculation we have worked with an idealized model of underdamped junction, therefore

396: %neglecting completely the effect of dissipation. Although 

397: %overdamped real junctions can be fabricated, the quality factor is typically limited to values of 

398: %the order of 100. In the standard 

399:  %RCSJ model

400: %\begin{equation}

401: %C\frac{d^2}{dt^2}\left({\bar \Phi_0} \phi \right) = I - I_{0}\sin( \phi ) -\frac{1}{R}\frac{d}{dt}

402: %\left({\bar \Phi_0} \phi\right) + I_n (t) \label{rcsj}

403: %\end{equation}

404: %where we have expressed everything in term of the "flux" coordinate $ \bar \Phi_0\phi $.

405: %The value of the resistance $R$ can be determined experimentally; typical values are in the range of 100-200$\Omega$.

406: %Classically, we notice that dissipation becomes important when the friction force becomes of the order of magnitude

407: %of the force applied; the limitation is that the maximum current through the resistor when the junction reaches the 

408: %quasiparticle threshold should be less than the critical current of the junction. For Nb, this gives a minimum 

409: %critical current of the order of 10-20$\mu$A, which can be achieved in large junctions. 

410: %A more delicate discussion refers to the role of finite temperature. 

411: %The temperature manifests itself

412: %through fluctuations in the biasing current with white-noise correlations

413: %$\langle I_{n}(t)I_{n}(0)) \rangle  = 2K_{B}TR^{-1}\delta (t)$; 

414: %the average of this noise is zero, $\langle I_{n}\rangle = 0 $.

415: %We notice that although the condition for MQT is fulfilled, this does not guarantee that the evolution will

416: %preserve the purity of the state. Consider the quantized version of Eq. [\ref{rcsj}], the so-called quantum Lagrange equation. We introduce

417: %the analog of de Broglie thermal wavelength, which in our case has dimensions of flux:

418: %\begin{equation}

419: %\Phi_{T} = \frac{h}{\sqrt{2\pi CK_{B}T}} = \frac{2}{\sqrt{\pi}}\Phi_{0}\sqrt{\frac{E_{c}}{K_{B}T}}

420: %\end{equation} 

421: %For an initial superposition of two states, the decoherence time is typically set by \cite{zurek}

422: %\begin{equation}

423: %\tau _{D} = RC \left( \frac{2\pi \Phi_{T}}{\Phi_{0}\delta \phi}\right)^2 = \frac{8\pi e^2 R}{K_{B}T(\delta\phi)^2}

424: %\end{equation}

425: %Assuming that this result sets correctly the orders of magnitude also for our problem,

426: %we see that the out-going state and the state inside the barrier will decohere fast. 

427: 

428: %To estimate the quantity $\delta\phi$ we consider a junction biased so that the MQT is experimentally observable

429: %(less than 1\% of the critical current). Then $\delta\phi$ is the phaseshift of an eigenfunction when the energy 

430: %changes with $K_B T$. This gives $8\pi /(\delta \phi)^2 \approx 10^4$, resulting in a characteristic

431: %decoherence time $\tau_D \approx 10ns$. This is large enough for the pure-state quantum behavior of the particle to be 

432: %observed. It is also consistent with the values obtained from Ramsey interference experiments with superconducting 

433: %quantum bits.

434: 

435: %For most of the existing experiments with switching junctions, $\Gamma ^-1$ is in fact larger than the time it takes

436: %for the switching voltage to reach the quasiparticle branch. Even if our calculations are not valid in this limit 

437: %because of transient effects we see that there will be anyway a growing outgoing component 

438: %of the wavepacket moving towards the quasiparticle branch. 

439: %We see that in the case of real-life experiment the environment decides whether the junction has switched or not 

440: %faster than the time scale $\Gamma ^-1$ at which switching itself takes place. In classical terms, this 

441: %sounds almost paradoxical: the information that the junction has switched or not becomes available faster than 

442: %the switching event itself. It is as if the apparatus provides the result of the measurement of a process

443: %before the process happens. 

444: 

445: %The time that the particle spends in the running-phase state, before reaching the

446: %quasiparticle gap, is quite short. This is due to the fact 

447: %that for underdamped junctions the speed of the particle down the washboard potential increases fast. This makes

448: %ny measurement of the runing-phase state a challenging experimental problem but even so, oscilloscopes 

449: %and amplifiers with bandwidths of tens of GHz exist and could be used for the observation of the pulses.

450: %However, a much better experimental strategy is to reduce the plasma frequency by using instead of

451: %a single junction a dc-SQUID system in a magnetic field. This will suppress $\omega_p$ to almost 

452: %any suitable value (in practice one-two orders of magnitude). Finally, to keep the quality factor above 1,

453: %a good solution is to add a capacitor in parallel with the SQUID. With this arrangement, the exponential 

454: %running-phase time can be set in the range of tens of nanoseconds, thus enabling the experimental 

455: %observation.

456:  

457: 

458: 

459: 

460: %\section{Conclusion}

461: 

462: 

463:  

464: %\section{Acknowledgements}

465:   

466: G.~S.~P. was supported by an EU  Marie Curie Fellowship (HPMF-CT-2002-01893); this work is also

467: part of the SQUBIT-2 project (IST-1999-10673), the 

468: Academy of Finland TULE No.7205476, and the 

469: Center of Excellence in Condensed Matter and Nuclear Physics at 

470: the University of Jyv\"askyl\"a.  

471: 

472: 

473: 

474: \begin{thebibliography}{99}

475: 

476: \bibitem{tinkham} M.~Tinkham, {\it Introduction to Superconductivity}, 2nd ed. (McGraw-Hill

477: Inc., New York, 1996).

478: 

479: \bibitem{switch}J.~Clarke, A.~N.~Cleland, M.~H.~Devoret, D.~Esteve, and J.~M.~Martinis,

480: Science {\bf 239}, 992 (1988).

481: 

482: \bibitem{suppl} A.~J.~Leggett, Suppl. Progr. Theor. Phys. {\bf 69}, 80 (1980).

483: 

484: \bibitem{deltapotential} R.~G.~Winter, Phys. Rev. {\bf 123}, 1503 (1961); 

485: C.~B.~Chiu, E.~C.~G.~Sudarshan, and 

486: B.~Misra, Phys. Rev. D {\bf 16}, 520 (1977); S.~De~Leo and P.~P.~Rotelli, quant-ph/0401145.

487: 

488: \bibitem{cond} J.~Bardeen, Phys.~Rev.~Lett. {\bf 6}, 67 (1961);

489: M.~Galperin, D.~Segal, and A.~Nitzan, J.~Chem.~Phys. {\bf 111}, 1569 (1999).

490: 

491: \bibitem{josephson}K.~S.~Chow, D.~A.~Browne, and V.~Ambegaokar, Phys. Rev. B {\bf 37}, 1624 (1988);

492: S.~Takagi, {\it Macroscopic Quantum Tunneling} (Cambridge University Press, Cambridge, 2002).

493: 

494: \bibitem{twopotential} S.~A.~Gurvitz and G.~Kalbermann, Phys.~Rev.~Lett. {\bf 59}, 262 (1987);

495: S.~A.~Gurwitz, Phys.~Rev.~A {\bf 38}, 1747 (1988).

496: 

497: \bibitem{kurizkikofman} A.~Barone, G.~Kurizki, and A.~G.~Kofman, Phys.~Rev.~Lett. {\bf 92}, 200403

498: (2004).

499: 

500: \bibitem{qubit} 

501: %Y.~Nakamura, Y.~A.~Pashkin, and J.~S.~Tsai, 

502:  %Nature, {\bf 398}, 786 (1999); 

503:  %J.~E.~Mooij, T.~P.~Orlando, L.~Levitov, L.~Tian, 

504:  %C.~H.~van~der~Val, and S.~Lloyd, Science {\bf 285}, 1036 (1999); 

505:  %J.~R.~Friedman, V.~Patel, W.~Chen, S.~K.~Tolpygo, and J.~E.~Lukens, Nature {\bf 406},

506:  %43 (2000); 

507:  D.~Vion, A.~Aassime, A.~Cottet, P.~Joyez, H.~Pothier, C.~Urbina, 

508:  D.~Esteve, and M.~H.~Devoret, Science {\bf 296}, 886 (2002); I.~Chiorescu, Y.~Nakamura, 

509:  C.~J.~P.~M.~Harmans, and J.~E.~Mooij, Science {\bf 299}, 1869 (2003); J.~Claudon, F.~Balestro, F.~W.~J.~Hekking, and 

510:  O.~Buisson, Phys. Rev. Lett. {\bf 93}, 187003 (2004); J.~Sjostrand, J.~Walter,

511:  D.~Haviland, H.~Hansson, A.~Karlhede, cond-mat/04066510.

512:  %; Yu.~A.~Paskin, T.~Yamamoto,

513:  %O.~Astafiev, Y.~Nakamura, D.~V.~Averin, and J.~S.~Tsai, Nature {\bf 421}, 823 (2003).

514: 

515: 

516: 

517: %\bibitem{zurek} W.~H.~Zurek, in {\it Frontiers of Nonequilibrium Quantum Statistical Physics},

518: %NATO ASI vol. B 135, pp. 145-149, G.~T.~Moore and M.~O.~Scully, eds., (Plenum 1986); quant-ph/0302044.

519: 

520: 

521: \end{thebibliography}

522: 

523: \end{document}

524: