1: \documentclass[prl,twocolumn,tightenlines,showpacs]{revtex4}%
2: \usepackage{amsfonts}
3: \usepackage{amsmath}
4: \usepackage{amssymb}
5: \usepackage{graphicx}
6:
7: \textheight=9.55 in
8: \textwidth=7.07 in
9: \topmargin=-0.80 in
10: \oddsidemargin=-0.31 in
11: \evensidemargin=0.in
12:
13: \begin{document}
14: \title{Macroscopic tunnel splittings in superconducting phase qubits}
15: \author{Philip R. Johnson}
16: \email[electronic address: ]{philipj@physics.umd.edu}
17: \author{William T. Parsons}
18: \author{Frederick W. Strauch}
19: \email[electronic address: ]{fstrauch@physics.umd.edu}
20: \author{J.R. Anderson}
21: \author{Alex J. Dragt}
22: \author{C.J. Lobb}
23: \author{F.C. Wellstood}
24: \affiliation{Department of Physics, University of Maryland, College Park, MD 20850}
25:
26: \begin{abstract}
27: Prototype Josephson-junction based qubit coherence times are too short for
28: quantum computing. Recent experiments probing superconducting phase qubits
29: have revealed previously unseen fine splittings in the transition energy
30: spectra. These splittings have been attributed to new microscopic degrees of
31: freedom (microresonators), a previously unknown source of decoherence. We show
32: that macroscopic resonant tunneling in the extremely asymmetric double well
33: potential of the phase qubit can have observational consequences that are
34: strikingly similar to the observed data.
35:
36: \end{abstract}
37: \date{\today}
38: \pacs{74.50.+r, 03.67.Lx, 85.25.Cp, 03.65.Xp}
39: \maketitle
40:
41:
42: Recent experiments by Simmonds \textit{et al.} \cite{Simmonds et al} and
43: Cooper \textit{et al. }\cite{Cooper et al} reveal previously unseen fine
44: splittings in the transition energy spectra of superconducting phase qubits.
45: These splittings are interpreted as resulting from coupling between the
46: circuit's collective dynamical variable (the superconducting phase describing
47: the coherent motion of a macroscopic number of Cooper pairs) and microscopic
48: two-level resonators, hereafter called \emph{microresonators}, within
49: Josephson tunnel junctions. Microresonators may be an important decoherence
50: mechanism \cite{Simmonds et al,Cooper et al,Decoherence mechanisms} for many
51: different superconducting qubit devices \cite{Phase qubit references,Charge
52: qubits,Flux qubits} with broader implications for Josephson junction physics
53: generally. Key questions remain however. Are all of the observed splittings
54: truly a microscopic property of junctions? Could they instead be a macroscopic
55: property of the particular circuit, or a combination of microscopic and
56: macroscopic phenomena?
57:
58: In fact, macroscopic resonant tunneling (MRT) can produce spectral splittings
59: in multiwell systems by lifting degeneracies between the states of different
60: wells. These effects have been probed by Rouse \textit{et al.}, Friedman
61: \textit{et al}., and others \cite{MRT} in superconducting circuits involving
62: asymmetric double wells with a few left well states, and $\lesssim$ 10
63: right-well states. MRT effects have also been demonstrated by Crankshaw
64: \textit{et al.} \cite{Crankshaw et al} in three-junction flux qubits, another
65: system in which spurious splittings have been reported \cite{Plourde and
66: Bertet}. What is not obvious is that MRT effects can be important for systems
67: with extremely asymmetric double well potentials, like the rf SQUID phase
68: qubit \cite{Simmonds et al,Cooper et al}, that have hundreds or thousands of
69: right well states. In this Letter, we analyze the phase qubit in this limit
70: and show that MRT\ produces surprisingly complex observational consequences
71: that are strikingly similar to \emph{some} of the observed data \cite{Simmonds
72: et al,Cooper et al}. MRT is therefore a possible mechanism for fine splittings
73: in a phase qubit and requires further examination.
74:
75: \begin{figure}[t]
76: \begin{center}
77: \includegraphics{Figure1.eps}
78: \end{center}
79: \caption{(a) Circuit diagram for an rf SQUID phase qubit. (b) The device can
80: be tuned via an inductively coupled bias line to give an extremely asymmetric
81: double-well.}%
82: \end{figure}
83:
84: Figure 1(a) shows the circuit schematic for an rf SQUID. The device is a
85: superconducting loop of inductance $L$ interrupted by a single Josephson
86: junction with capacitance $C$ and critical current $I_{c},$ inductively
87: coupled to a flux-bias line. The circuit Hamiltonian is%
88: \begin{equation}
89: H=4E_{C}p^{2}/\hbar^{2}+E_{J}\left( \gamma^{2}/2\beta-\cos\gamma
90: -J\gamma\right) ,\label{Hamiltonian}%
91: \end{equation}
92: where $\gamma$ is the gauge invariant phase difference across the junction,
93: $p=\hbar Q/2e$ is the momentum conjugate to $\gamma$ ($Q$ is the charge on the
94: plates of the capacitor), $\beta=2\pi I_{c}L/\Phi_{0}$ is the modulation
95: parameter ($\Phi_{0}=h/2e$ is the flux quantum), and $J=I/I_{c}$ is the
96: dimensionless current that is induced in the loop by the applied flux bias.
97: The charging energy $E_{C}=e^{2}/2C$ and Josephson energy $E_{J}=I_{c}\Phi
98: _{0}/2\pi$ determine the regime of superconducting qubit behavior; for a phase
99: qubit $E_{J}\gg E_{C}.$
100:
101: \begin{figure*}[tb]
102: \begin{center}
103: \includegraphics{Figure2.eps}
104: \end{center}
105: \caption{(a) Numerically computed spectrum of phase qubit when $I_{c}=8.531$
106: $\mu$A, $C=1.2$ pF, and $L=168$ pH ($\beta=4.355$). Energies are plotted in
107: units of frequency. The inset shows the avoided crossing due to resonant
108: tunnel coupling between the left well state $\left\vert 1\right\rangle _{L}$
109: and a highly excited right well state. (b) The circuit parameters give an
110: asymmetric double well like that shown. (c) Wavefunctions of the $k=641$
111: eigenstate for bias values near the avoided crossing shown in the inset. (d)
112: Solid points are numerically computed sizes and locations of the splittings.
113: Solid lines are splitting sizes derived from WKB theory.}%
114: \end{figure*}
115:
116: The shape of the circuit's potential energy function $U\left( \gamma\right)
117: $ depends on $\beta$ and the bias $J.$ For $\beta\lesssim3\pi/2,$ it is
118: possible to bias the circuit so that the potential has the highly asymmetric
119: double-well shape shown in Fig.~1(b), tuned to give a shallow upper left well
120: with just a few left-localized states, denoted by $\left\vert n\right\rangle
121: _{L}$, and a deep right well with many right-localized states, denoted by
122: $\left\vert m\right\rangle _{R}$. Simmonds \textit{et al.} \cite{Simmonds et
123: al}--motivated by a number of attractive features including reduced
124: quasiparticle generation, tunable anharmoniticity of the left well potential,
125: inductive isolation from and reduced sensitivity to bias noise, and nice
126: read-out properties--have proposed using the rf SQUID with an extremely
127: asymmetric double well potential as a phase qubit \cite{Phase qubit
128: references}.
129:
130: Making a cubic approximation to the left well, we derive the plasma frequency
131: for small oscillations%
132: \begin{equation}
133: \omega_{L}=\omega_{0}\left( 1-\beta^{-2}\right) ^{1/8}\left[ 2\left(
134: J^{\ast}-J\right) \right] ^{1/4},\label{w_L}%
135: \end{equation}
136: where $\omega_{0}=$ $\sqrt{8E_{c}E_{J}/\hbar^{2}},$ and%
137: \begin{equation}
138: J^{\ast}=\left( 1-\beta^{-2}\right) ^{1/2}+\beta^{-1}\arccos\left(
139: -\beta^{-1}\right) >1\label{J_star}%
140: \end{equation}
141: is the critical bias for which the left well vanishes. Note that the effective
142: critical current is $I^{\ast}=I_{c}J^{\ast}>I_{c}.$ The approximate number of
143: left-well states is%
144: \begin{equation}
145: N_{L}=\frac{\Delta U_{L}}{\hbar\omega_{L}}\simeq\frac{2^{3/4}}{3}\sqrt
146: {\frac{E_{J}}{E_{C}}}\left( 1-\beta^{-2}\right) ^{-3/8}\left( J^{\ast
147: }-J\right) ^{5/4},\label{N_L}%
148: \end{equation}
149: where $\Delta U_{L}$ is the barrier height. The level spacing in the right
150: well is approximately $\hbar\omega_{R},$ where $\omega_{R}$ is the right well
151: plasma frequency, and the number of right well states is approximately
152: $N_{R}\simeq\Delta U_{R}/\hbar\omega_{R},$ where $\Delta U_{R}$ is the depth
153: of the right well.
154:
155: Figure 2(a) shows the energy spectrum as $J$ is varied for $0\leq N_{L}\leq6$
156: and $C=1.2$ pF, $L=168$ pH, and $I_{c}=8.531$ $\mu$A, giving $\beta=4.355$,
157: $I^{\ast}=11.659$ $\mu$A, and $\omega_{L}/2\pi\sim10$ GHz. These are the
158: circuit parameters from \cite{Simmonds et al}, assuming that the critical
159: current quoted there is $I^{\ast}$. To obtain the energy spectrum we
160: diagonalize the Hamiltonian in Eq.~(\ref{Hamiltonian}) using a discrete
161: Fourier grid representation \cite{Fourier grid method}, thereby obtaining a
162: numerical solution for the eigenvalues $E_{k}\left( J\right) $ and
163: eigenstates $\left\vert k\left( J\right) \right\rangle $ of the full
164: double-well system versus the bias $J$. A harmonic approximation to the right
165: well yields approximately $500$ states below the left well; the full
166: calculation yields $N_{R}\simeq600-700$ states, depending on the bias
167: \cite{Footnote 2}.
168:
169: In Fig.~2(a) we define the zero of energy to be at the bottom of the left
170: well. We note two different types of energy levels: horizontal ($H$) branches
171: and near vertical ($V$) branches. From our definition of zero energy,
172: eigenvalues corresponding to states mainly localized in the right well [region
173: I of Figs. 2(a) and (b)] fall with increasing $J,$ and are thus nearly
174: vertical. The energy levels in region III correspond to delocalized states
175: fully above the left well. The dashed line in Fig.~2(a) dividing regions II
176: and III indicates the energy at the top of the left-well barrier. In region
177: II, eigenstates whose energies lie along $H$ branches are primarily localized
178: in the left well ($H\sim L$). The number of left-well states at bias $J$ is
179: consistent with $N_{L}$ from Eq.~(\ref{N_L}). Eigenstates whose energies lie
180: along $V$ branches are primarily localized in the right well ($V\sim R$).
181: Their energies fall at essentially the rate of the falling right well. Note
182: that in Fig.~2(a) the \textit{density} of right-well states is comparable to
183: that of the left-well, despite $N_{R}\gg N_{L}.$
184:
185: Every apparent intersection of an $H$ and $V$ energy level in Fig.~2(a) is an
186: avoided crossing (see inset). Degeneracies are lifted by resonant tunneling of
187: left-well states $\left\vert n\right\rangle _{L}$ and right-well states
188: $\left\vert m\right\rangle _{R}.$ Left of an avoided crossing between $k$ and
189: $k+1$ eigenstates we find that $\left\vert k\right\rangle \cong\left\vert
190: n\right\rangle _{L}$ and $\left\vert k+1\right\rangle \cong\left\vert
191: m\right\rangle _{R}$. Right of the crossing the states swap, becoming
192: $\left\vert k\right\rangle \cong\left\vert m\right\rangle _{R}$ and
193: $\left\vert k+1\right\rangle \simeq\left\vert n\right\rangle _{L}.$ At the
194: avoided crossing $\left\vert k\right\rangle \cong\left( \left\vert
195: n\right\rangle _{L}+\left\vert m\right\rangle _{R}\right) /\sqrt{2}$ and
196: $\left\vert k+1\right\rangle \cong\left( \left\vert n\right\rangle
197: _{L}-\left\vert m\right\rangle _{R}\right) /\sqrt{2}.$ Figure 2(c) shows the
198: wavefunctions for the $k=641$ eigenstate before, at, and after the splitting
199: shown in the inset in Fig.~2 (a). The distribution of splitting magnitudes
200: along the first five energy branches are plotted in Fig.~2(d) as solid points.
201: Gaps larger than 1 MHz are within the resolution of recent experiments. Along
202: each left-well energy branch the tunnel splittings are regularly spaced with
203: magnitudes that decrease exponentially with $N_{L}$. We have numerically
204: computed spectra for a variety of circuit parameters, including $I_{c}=2$
205: $\mu$A and $C=0.5$ pF which are comparable to those reported in \cite{Cooper
206: et al}. In each case the spectrum looks qualitatively similar to Fig.~2 (a).
207: We note that the predicted gap sizes are strikingly similar to those reported
208: in \cite{Simmonds et al,Cooper et al} ($\sim$ 1-100 MHz).
209:
210: The complex collection of energy splittings has both direct and indirect
211: effects that should be taken into account when analyzing the experimental
212: data. Consider a double frequency microwave spectroscopic method, like that
213: used in \cite{Simmonds et al}. Microwaves of frequency $\omega_{01}$ are
214: applied to drive the $0\rightarrow1$ transition. Excitation of the $\left\vert
215: 1\right\rangle _{L}$ state is detected with a measurement microwave pulse of
216: frequency $\omega_{13},$ which drives the $1\rightarrow3 $ transition. The
217: $\left\vert 3\right\rangle _{L}$ state's exponentially greater amplitude to be
218: found in the right well compared to the $\left\vert 0\right\rangle _{L}$ and
219: $\left\vert 1\right\rangle _{L}$ states allows an adjacent detection SQUID to
220: easily detect the change in the qubit's flux. This method directly probes
221: splittings along many of the energy branches shown in Fig.~2(d). Cooper
222: \textit{et al.} have introduced a new spectroscopic technique that can probe
223: deeper left wells where $N_{L}>4.$ This method applies a few-nanosecond
224: current pulse changing the bias so that $N_{L}\gtrsim2$ by briefly tilting the
225: potential adiabatically with respect to the left well period $T_{L}\equiv
226: 2\pi/\omega_{L}\sim100$ ps \cite{Cooper et al}. Since the measurement pulse
227: moves left-well states to the right along horizontal ($H$) energy branches
228: [see Fig.~2(a)], read-out should be influenced by the exponentially larger
229: splittings present for smaller $N_{L}$. For example, the measurement pulse may
230: move a deep well state to one of the large splitting degeneracies near
231: $N_{L}\sim2,$ whose presence may produce a significant perturbation on
232: read-out fidelity. Thus the current pulse method is also sensitive to large
233: splittings along multiple energy branches.
234:
235: MRT degeneracies also have very narrow bias value widths. For example, the
236: inset of Fig. 2(a) shows a splitting width of less than $0.1$ nA. The bias
237: widths become only smaller for splittings at larger $N_{L}.$ The horizontal
238: axis of Fig.~2(a) corresponds to more than $300$ nA. Typical experiments
239: sampling only a limited number of bias values likely probe only a subset of
240: the (many) MRT splittings. Changes in experimental conditions (e.g. bias drift
241: and noise, or temperature cycling) may generate surprisingly large shifts in
242: the observed splitting distributions if they result in a different subset of
243: sampled MRT\ degeneracies. These or other features could result in transition
244: spectra with a varying distribution of splitting sizes and bias-value
245: locations which, due to their complexity and variability in time, might appear
246: to have a microscopic origin. Such variations seem more consistent with a
247: model of microscopic critical current fluctuators, suggesting that both MRT
248: and microresonator effects are present \cite{Simmonds and Martinis}. If this
249: is the case, it is important to identify which observed splittings are due to
250: which mechanism.
251:
252: Measured with sufficient resolution, the transition frequency avoided
253: crossings due to MRT should have distinctive characteristics. When driving
254: $0\rightarrow1$ transitions, a splitting in the $\left\vert 0\right\rangle
255: _{L}$ branch should produce crossings like that shown in Fig.~3(a), whereas a
256: splitting in the $\left\vert 1\right\rangle _{L}$ branch should produce
257: crossings like that shown in Fig.~3(b). The observed shapes may be strongly
258: dependent upon the experimental measurement technique. Bias noise could smear
259: out the splittings in the horizontal direction. For splittings in the lower
260: energy branch [Fig.~3(a)] this would leave a distinct frequency gap, givings
261: observed splittings horizontally smeared appearances like those observed in
262: \cite{Simmonds et al,Cooper et al}. In contrast, it is unclear if splittings
263: in the upper branch [Fig.~3(b)] are consistent with observation. Improved
264: experimental resolution that revealed these distinctive avoided-crossing
265: shapes would be compelling evidence for MRT. Han \textit{et al}. have explored
266: other complexities that arise when measuring systems that exhibit MRT
267: \cite{MRT}.
268:
269: \begin{figure}[tb]
270: \begin{center}
271: \includegraphics{Figure3.eps}
272: \end{center}
273: \caption{(a) The distinctive shapes of avoided crossings in the measured
274: transition frequencies for splittings in the lower branch. (b) The avoided
275: crossing transition shape for splittings in the upper branch. (c) The figure
276: shows that $\Delta_{R}\approx\Delta_{L}$ over an extremely large range of
277: double-well circuit parameters. Bold lines show the splitting magnitudes
278: $\Delta$ along the $n_{L}=0,1,$ and $2$ left well energy branches, with
279: $\beta=4.5,$ $N_{L}=3,$ and $I_{c}=10\mu$A.}%
280: \end{figure}
281:
282: We derive an analytic expression for the energy splitting between pair-wise
283: degenerate left and right states in an asymmetric double well in the WKB
284: approximation \cite{WKB methods}. This yields the splitting formula%
285: \begin{equation}
286: \Delta=\sqrt{\frac{2\Delta_{L}\Delta_{R}\left( n_{L}+\frac{1}{2}\right)
287: ^{n_{L}+1/2}\left( m_{R}+\frac{1}{2}\right) ^{m_{R}+1/2}}{\pi\ n_{L}%
288: !m_{R}!e^{n_{L}+m_{R}+1}}}e^{-S},\label{tunnel splitting formula}%
289: \end{equation}
290: where $S=\int_{\gamma_{1}}^{\gamma_{2}}\sqrt{2m\left[ E-V\left(
291: \gamma\right) \right] }d\gamma,$ $m=C\left( \Phi_{0}/2\pi\right) ^{2}$,
292: $\gamma_{1,2}$ are the classical turning points for the barrier given by
293: $V\left( \gamma_{1,2}\right) =E_{n_{L}}$, and $\Delta_{L}\simeq\hbar
294: \omega_{L}$, $\Delta_{R}\simeq\hbar\omega_{R}$ are the left and right well
295: level spacings at energy $E_{n_{L}}$. For deep right wells,
296: Eq.~(\ref{tunnel splitting formula}) becomes independent of $m_{R}.$ In this
297: limit, together with the cubic approximation accurate for shallow left wells,
298: the splittings are approximately%
299: \begin{equation}
300: \Delta\simeq\sqrt{\frac{2^{1/2}\Delta_{L}\Delta_{R}}{n_{L}!\pi^{3/2}}}\left(
301: 432N_{L}\right) ^{\left( n_{L}+1/2\right) /2}e^{-18N_{L}/5}%
302: .\label{cubic approx tunneling splitting formula}%
303: \end{equation}
304: For the right well level spacing we use the WKB estimate $\Delta_{R}=2\pi
305: \hbar/T_{cl}$ \cite{Landau and Lifshitz}, where $T_{cl}$ is the classical
306: period of right-well oscillations with energy $E_{n_{L}}.$ Splittings
307: calculated from Eq.~(\ref{cubic approx tunneling splitting formula}) are shown
308: as solid lines in Fig.~2(d). The agreement with the exact splittings (solid
309: points) is excellent for lower lying states, and surprisingly good for the
310: excited states. Note that the tunnel splitting formula in
311: Eq.~(\ref{cubic approx tunneling splitting formula}) predicts splittings
312: exponentially larger than continuum tunneling rates: $\Delta_{\text{splitting}%
313: }/\Gamma_{\text{tunneling}}\sim\exp\left( 18N_{L}/5\right) ,$ making MRT
314: effects important even when continuum tunneling is negligible.
315:
316: We have compared MRT splittings with
317: Eq.~(\ref{cubic approx tunneling splitting formula}) for a number of numerical
318: examples with $N_{R}\sim100-1000,$ but in principle one can fabricate circuits
319: with many thousands of right-well states. The WKB formula for the splittings
320: and level spacings allows analysis of circuit parameters for very deep right
321: wells where numerical treatment is impractical. Figure 3(c) shows $\hbar
322: \omega_{L}\simeq\Delta_{L}$ (dashed line) and $\Delta_{R}=2\pi\hbar/T_{cl}$
323: (thin-solid line) versus the ratio $E_{C}/E_{J}$ for $I_{c}=10\mu$A, $N_{L}=3
324: $, and $\beta=4.5$ just below the $\beta$ threshold where the potential
325: develops three wells. (For the circuit parameters in Fig.~2 and \cite{Cooper
326: et al}, $E_{C}/E_{J}\sim10^{-4}-10^{-6}$.) The value of $I_{c}$ determines the
327: frequency scale on the left of Fig.~3(c) but leaves the relative positions of
328: the plotted lines essentially unchanged. The top axis shows $N_{R}$ from the
329: harmonic oscillator approximation. Observe that, perhaps unexpectedly,
330: $\Delta_{R}\approx\Delta_{L}$ even for extremely asymmetric double wells. The
331: bold-solid lines show the WKB splitting $\Delta$ when $n_{L}=0,1,$ and $2$.
332: The validity condition for MRT $\Delta\ll\Delta_{R,L}$ is satisfied over a
333: large range of circuit parameters, and for $N_{R}\sim10^{5}$ and greater.
334:
335: Dissipation suppresses resonant tunneling when $\Gamma_{R}\gtrsim\Delta_{R},$
336: where $\Gamma_{R}\simeq N_{R}\hbar/T_{1}$ is the width of excited right well
337: states, and $T_{1}$ is the dissipation time for $\left\vert 1\right\rangle
338: _{R}\rightarrow\left\vert 0\right\rangle _{R}$ \cite{Caldeira and
339: Leggett,Garg}. Using the WKB\ expression for $\Delta_{R},$ we find the
340: condition $N_{R}\lesssim\omega_{L}T_{1}$ for observing MRT. For a phase qubit
341: with $\omega_{L}/2\pi\sim10$ GHz and $T_{1}\sim10-100$ ns, resonant tunneling
342: should be detectable as long as $N_{R}\lesssim600-6000$ states. For the
343: circuit parameters in Fig.~2 $N_{R}\sim600-700$ and for those in \cite{Cooper
344: et al} $N_{R}\sim150-300,$ with a measured $T_{1}\simeq25$ ns. Thus, we do not
345: believe that dissipation will remove the effects of MRT. If the intrinsic
346: dissipation is actually much smaller so that $\Gamma\lesssim\Delta$
347: \cite{Garg}, it should be possible to observe coherent oscillations \cite{Long
348: T1}.
349:
350: In conclusion, we show that significant MRT effects should be present for
351: extremely asymmetric double well phase qubits, and thus MRT should be taken
352: into account in the important effort to fully characterize microresonators or
353: other splittings mechanisms. Our analysis provides tools and can guide
354: experiments to help distinguish between three main possibilities: (1) Both MRT
355: and microresonators are present, (2) MRT effects explain all the observational
356: data, and (3) MRT is entirely absent. We believe that (1) is most likely;
357: however, due to the complexity of effects from MRT, further experiments and
358: detailed modeling are necessary to definitively rule out (2) and (3). Finally,
359: our Letter provides general tools for exploring the quantum mechanics of
360: extremely asymmetric double-well systems.
361:
362: \begin{acknowledgments}
363: We thank R. Simmonds and J. Martinis for useful comments. This work was
364: supported by the NSA, the DCI Postdoctoral Research Program, the NSF QUBIC
365: program, DOE grant DE-FG02-96ER40949, and the University of Maryland's Center
366: for Superconductivity Research.
367: \end{acknowledgments}
368:
369: \begin{thebibliography}{99} %
370: \bibitem {Simmonds et al}R.W. Simmonds \textit{et al}., Phys. Rev. Lett
371: \textbf{93, }077003 (2004).
372:
373: \bibitem {Cooper et al}K.B. Cooper \textit{et al}., Phys. Rev. Lett.
374: \textbf{93}, 180401 (2004).
375:
376: \bibitem {Decoherence mechanisms}D.J. Van Harlingen \textit{et al.}, Phys.
377: Rev. B \textbf{70}, 064517 (2004); F. Meier and D. Loss, cond-mat/0408594 (2004).
378:
379: \bibitem {Phase qubit references}J.M. Martinis, S. Nam, J. Aumentado, and C.
380: Urbina, Phys. Rev. Lett. \textbf{89}, 117901 (2002); Y. Yu \textit{et al}.,
381: Science \textbf{296}, 889 (2002); A. J. Berkley \textit{et al}.,
382: \textit{ibid.} \textbf{300}, 1548 (2003); F.W. Strauch \textit{et al}., Phys.
383: Rev. Lett. \textbf{91} 167005 (2003); H. Xu et al., \textit{ibid. }%
384: \textbf{94}, 027003 (2005).
385:
386: \bibitem {Charge qubits}T. Yamamoto \textit{et al}., Nature (London) 425, 941
387: (2003); Yu. A. Pashkin \textit{et al}., \textit{ibid. }421, 823 (2003); A.
388: Wallraff \textit{et al}., \textit{ibid.} \textbf{431}, 162 (2004); D. Vion
389: \textit{et al}., Science \textbf{296}, 886 (2002).
390:
391: \bibitem {Flux qubits}C. H. van der Wal \textit{et al}., Science \textbf{290},
392: 773 (2000); I. Chiorescu, Y. Nakamura, C.J.P.M. Harmans, and J.E. Mooij,
393: Science \textbf{299}, 1869 (2003); A. Izmalkov \textit{et al}., Phys. Rev.
394: Lett. \textbf{93}, 037003 (2004); Y. Yu \textit{et al}., \textit{ibid.}
395: \textbf{92}, 117904 (2004); L. Tian, S. Lloyd, and T.P. Orlando, Phys. Rev. B
396: \textbf{67}, 220505 (2003).
397:
398: \bibitem {MRT}R. Rouse, S. Han and J.E. Lukens, Phys. Rev. Lett. \textbf{75},
399: 1614 (1995); S. Han, R. Rouse, and J.E. Lukens, \textit{ibid.} \textbf{84},
400: 1300 (2000); J.R. Friedman \textit{et al}., Nature (London) \textbf{406}, 43
401: (2000); D.V. Averin, J.R. Friedman, and J.E. Lukens, Phys. Rev. B \textbf{62},
402: 11802 (2000).
403:
404: \bibitem {Crankshaw et al}D.S. Crankshaw \textit{et al}., Phys. Rev. B
405: \textbf{69}, 144518 (2004).
406:
407: \bibitem {Plourde and Bertet}P. Bertet \textit{et al}., cond-mat/0412485
408: (2004); B.L.T. Plourde \textit{et al.}, cond-mat/0501679 (2005).
409:
410: \bibitem {Fourier grid method}C.C. Marston and G.G. Balint-Kurti, J. Chem.
411: Phys. \textbf{91}, 3571 (1989).
412:
413: \bibitem {Footnote 2}Fig.~1(b) of \cite{Simmonds et al} shows $N_{R}%
414: \approx2500$.
415:
416: \bibitem {Simmonds and Martinis}R. Simmonds and J. Martinis, private communication.
417:
418: \bibitem {WKB methods}We derived Eq. (\ref{tunnel splitting formula}) from
419: Herring's formula assuming a WKB wavefunction under the barrier and matching
420: onto excited harmonic oscillator states in the two wells. This method is
421: described in A. Garg, Am. J. Phys. \textbf{68}, 430 (2000). See also J.M.
422: Schmidt, A.N. Cleland, and J. Clarke, Phys. Rev. B \textbf{43}, 229 (1991).
423:
424: \bibitem {Landau and Lifshitz}L. D. Landau and E. M. Lifshitz, \emph{Quantum
425: mechanics, volume III}, Pergamon Press (London) 1965.
426:
427: \bibitem {Caldeira and Leggett}A.O. Caldeira and A.J. Leggett, Ann. Phys. (New
428: York) \textbf{149}, 374 (1983).
429:
430: \bibitem {Garg}A. Garg, Phys. Rev. B \textbf{51}, 15161 (1995).
431:
432: \bibitem {Long T1}It is interesting to speculate that unexpectedly short
433: dissipation times of $\sim25$ ns in \cite{Cooper et al} result from damping of
434: excited right well states, such that $T_{\text{measured}}\sim$ $T_{1}/N_{R}.$
435: This would imply a $T_{1}$ of a few $\mu$s, consistent with the expected
436: relaxation times (see footnote [11] in \cite{Cooper et al}). The existence of
437: tunnel splittings could also be probed by tuning $T_{1}$ through adjustable
438: coupling to adjacent circuit elements.
439: \end{thebibliography}
440:
441:
442: \end{document}