VOLUME 55) NUMBER 15 PHYSICAL REVIEW LETTERS 7 OCTOBER 1985

Energy-Level Quantization in the Zero-Voltage State

of a Current-Biased Josephson Junction
John M. Martinis, Michel H. Devoret, ' and John Clarke
Department of Physics, University of California, Berkeley, California 94720, and Materials and Molecular
Research Division, Lawrence Berkeley Laboratory, Berkeley, California 94720
(Received 14 June 1985)
We report the first observation of quantized energy levels for a macroscopic variable, namely the
phase difference across a current-biased Josephson junction in its zero-voltage state. The position
of these energy levels is in quantitative agreement with a quantum mechanical calculation based on
parameters of the junction that are measured in the classical regime.

PACS numbers: 03.65.—w, 05.30.—d, 74. 50. + r

Do macroscopic variables obey quantum mechanics? nel junction.

This question, although central to the theory of mea- In this Letter, we describe a novel experiment that
surement, ' has only recently been addressed experi- demonstrates the quantum mechanical nature of the
mentally. An attractive candidate for such experimen- macroscopic variable through energy level quantiza-
tal investigation is the Josephson tunnel junction, a tion. We irradiate the junction with microwaves that,
system in which thermal fluctuations and perturbations at appropriate frequencies, cause the particie to make
due to the environment can be made negligible. In the transitions into excited states. Since the escape rate
case of the current-biased junction, the macroscopic out of the well increases with increasing energy, by
variable is the phase difference, 5, between the super- measuring changes in the escape rate we are able to
conducting order parameters on either side of the bar- demonstrate the existence of quantized energy levels
rier. The junction can be represented as a particle spectroscopically in a way that does not require a pre-
moving in a one-dimensional tilted cosine potential. 2 cise knowledge of all the parameters of the junction.
The zero-voltage state of the junction corresponds to We have designed the experiment so that the admit-
the confinement of the particle to one we11 of this po- tance across the Josephson junction due to its self-
tential. After the particle escapes from this metastable capacitance and current-voltage leads can be represent-
state, it runs freely down the tilted cosine potential, ed to a good approximation by a capacitance C and
and a voltage appears across the junction. For parame- resistance R in parallel [see Fig. 1(b)]. For a junction
ters of experimental interest, the potential well from with a critical current Io and for a bias current I close
which the particle escapes is represented, to a very to Io, the potential-barrier height 4U and plasma fre-
good approximation, by a cubic potential'3 [Fig. 1(a)]. quency co~/27r (the oscillation frequency of the particle
Previously, to demonstrate the quantum mechanical at the bottom of the well) are b, U = — ', J2 Uo (1

nature of 5, experiments 5 have been performed to —I/Io)' 2 and' to~ =co~a[1 —(I/Io)']I 4. Here, ' Uo
investigate the escape rate from the metastable we11 via =ID@a/27r, ro~o= (2mIo/C@II)'t, and C&o= h/2e The.
quantum macroscopic tunneling at low temperatures. dissipation is described through the damping factor
Related experiments have been concerned with a Q =co RC. According to quantum theory, the energy
superconducting ring interrupted by a Josephson tun- levels in the well should be quantized as indicated in
Fig. 1(a); because of the cubic term in the potential,
the spacing between adjacent levels decreases with in-
U' (b) creasing energy in the well. We note that an increase
in the bias current Idecreases the spacings of energy
levels.
9/Nr
hm
i ~V iii I„sir ~I+ I R For our experiments, the Nb-NbO„-PbIn tunnel
P junctions were patterned photolithographically on Si
chips in either a 10&&10-p,m or a 80x10-p, m2 cross-
strip geometry. Our experimental configuration for
experiments in the He temperature range has been
FIG. 1. (a) Cubic potential U vs phase difference 8 show-
ing three energy levels. Transition from the ground state to described previously. 9 In the experiment, the junction
the first excited state induced by a photon of frequency and the last of a chain of low-pass filters for the bias
co/2m is shown. (b) Model of current-biased Josephson circuitry were thermally anchored to the mixing
junction loaded with a resistor and irradiated with an exter- chamber of a dilution refrigerator. The temperature of
nal microwave current source. the mixing chamber was determined with a combina-

1985 The American Physical Society 1S43

VOLUME 55, NUMBER 15 PHYSICAL REVIEW LETTERS 7 OCTOBER 1985

tion of a calibrated germanium thermometer and a crowaves to be resonantly enhanced over that in the
Co orientation thermometer. The accuracy of the absence of microwaves when the microwave frequency
temperature scale and the absence of a significant level coincides with the spacing between two energy levels.
of external noise were checked by use of the junction In practice, we detect this resonance by varying the
itself to measure temperature: By reducing the critical energy-level spacing with the bias current while keep-
current with a magnetic field, we studied the junction ing the microwave frequency fixed. Figures 2 to 4 il-
in the classical limit (kaT/tee~ & I/2n. )' in which the lustrate typical results. In Fig. 2 (a) we show the
zero-voltage state decays via thermal activation. Ex- change in the escape rate due to 2.0-GHz microwave
tensive measurements in this limit down to the lowest irradiation for a junction with parameters chosen so
temperature attainable yielded escape rates that were that the well contained several energy levels
in good agreement with those predicted theoretically" (b, U/h're~ — 6), there was significant population of the
with use of the temperature determined by our ther- lower excited states (k&T/tee~ —
0.3), and the damp-
mometry. ing was low enough (Q —
80) to produce narrow reso-
Microwave power of frequency c0/2vr was injected nances. We observe three peaks in [I (P )
via a separate, filtered coaxial line that was capacitively —I' (0) ]/I (0), indicating that the escape rate is
coupled to the leads of the junction. We used the resonantly enhanced at certain values of the bias
technique of Fulton and Dunkleberger to measure the current. The two peaks at the higher bias currents are
escape rate of the junction out of the zero-voltage state approximately Lorentzian. No further peaks were ob-
as a function of the bias current, with and without mi- served at values of bias current higher than those plot-
crowave power. The microwave power P was adjusted ted in Fig. 2. These discrete resonances are charac-
to give a change in the escape rate I such that teristic of transitions between quantized energy 1evels.
(
[I (P ) —I (0) ]/I (0) 2; values of I (0) were in the
range 10 to 10 s '. We collected typically 10
To investigate the position of these peaks, we have
computed the energy levels by solving the Schrodinger
switching events. After the junction had switched to equation numerically, ' using as input parameters the
the nonzero-voltage state, the bias current was turned measured bias current, the value of Io —30.572
off within 30 iu, s. We ensured that the elapsed time + 0.017 p, A determined from the current dependence
before the next switching event was sufficient for any of I (0) at 28 mK, and the value of C = 47.0 + 3.0 pF
rise in temperature that occurred in the dissipative measured at higher temperatures in the classical re-
state to have become negligible. gime (see Ref. 9 for details of the last two determina-
One expects the escape rate in the presence of mi- tions). The solid lines in Fig. 2(b) show the energy-
level spacings E„„+, vs (n =0, I, 2). The dotted
I I I I I I I l 1

(o)
D ~/2TI'=2 6Hz
O m/2w =5.7GHz

O
O
I

CL l

CL
0.5—
- 22-(b) O.O

(bj
2.0
~ 3.8- (up/2T[
I. 8- I
w 3.6—
50.46 O
34--
FIG. 2. (a) [I'(P) —I (0)]/I (0) vs I for a 80x 10-p, m' 9.37
junction at 28 mK in the presence of 2.0-0Hz microwaves
(kaT/tao=0. 29). Arrows indicate positions of resonances. FIG. 3. (a) [I (P ) —I (0) ]/I' (0) vs I for a 10 x 10-p, mz
Inset represents the corresponding transitions between ener- junction at 18 mK for four microwave frequencies. (b) Cal-
gy levels. (b) Calculated energy-level spacings vs I E„„+t culated energy-level spacing Eo vs I for Io ——9.489 + 0.007
~

for Io —30.572+0.017 p, A and C =47.0+ 3.0 pF. Dotted p, A and C = 6.35+ 0.4 pF. Dotted lines indicate uncertain-
lines indicate uncertainties in the Eo ] curve due to errors ties due to errors in Io and C. Arrows indicate values of bias
in Io and C. Arrows indicate values of bias current at which current at which resonances are predicted. Dashed line indi-
resonances are predicted. cates plasma frequency.
VOLUME 55, NUMBER 15 PHYSICAL REVIEW LETTERS 7 OCTOBER 1985

more, the measured positions of the resonances are

(a)
clearly very different from a classical prediction for the
C) resonant activation of the particle oscillating at the
plasma frequency (dashed line).
(b) To illustrate the evolution from quantum to classical
C3 behavior, in Fig. 4 we show the temperature depen-
I
0.29 dence of the shape of the resonant response to mi-
CL
crowaves for the junction studied in Fig. 3 with a
slightly higher critical current. At the lowest tempera-
O. IO ture (c), at which the junction was firmly in the quan-
tum regime, we observe a single, approximately
I
Lorentzian resonance. At the intermediate tempera-
9.50 ture (b), a shoulder corresponding to the E, 2 transi-
I (p&j
tion begins to appear. At the highest temperature (a),
—
FIG. 4. [I'(P) I'(0) ]/I'(0) vs I for the junction of Fig. the resonance has become asymmetric, the peaks asso-
3 with Ip=9. 57 p, A and C = 6.35 pF at three values of ciated with individual transitions having merged into a
kaT/tee. The microwave frequencies are (a) 4. 5 GHz, (b) continuum; this response is characteristic of that ob-
4. 1 GHz, and (c) 3.7 GHz. served for the classical phenomenon of resonant ac-
tivation. A brief comment on the linewidth of the
resonance is in order. Measurements of Q in the clas-
lines indicate the magnitude of the uncertainty in the sical regime yield Q =30+ 15 for the junction shown
position of the curve corresponding to the Ep tran- &
in Fig. 3. To our knowledge, there is no theory avail-
sition that arises from the estimated errors in Ip and C. able for the effects of damping on the position and
We note that a systematic error in Ip or C will move all broadening of the energy levels of the excited states.
three curves by very nearly the same amount. The in- However, one might reasonably expect the energy lev-
tersection of each of these curves with the horizontal el of an excited state to be broadened by an amount
line corresponding to a microwave frequency of 2.0 Aced so that the relative linewidth of the Ep &
transi-
GHz is the prediction of the bias current at which the tion, Aco/c0, is approximately equal to 1/Q, assuming
resonant peaks should occur. The absolute positions that the lifetime of the excited state against tunneling
of the peaks with respect to the bias current agree with out of the well is sufficiently long. The data presented
the predicted positions to within the experimental un- in Fig. 3(a) yield ru/b, c0 = 50 + 10, in reasonable agree-
certainty. The separations of the peaks along the ment with the value of Q measured in the classical
current axis are in excellent agreement with the limit. We note that the coherence of the ground and
predicted separations. excited states must be maintained over time scales
To study the dependence of the energy difference given by the lifetime of the excited state, approximate-
(Et Ep) on bias current, we made measurements on ly 2zrQ/co~ — 14 ns for the Ep t resonance in Fig.
a second junction, with substantially lower values of Ip 3(a); for the same transition in Fig. 2(a), the coher-
and C, that was more strongly in the quantum regime ence time is about 40 ns.
(kaT/tee — 0. 1, AU/tee~ — 2) at the values of I and T In summary, we have observed microwave-induced
attainable in the experiment. Extensive measure- resonant enhancements of the rate at which a current-
ments of Ip and C in the classical regime yielded biased Josephson tunnel junction escapes from the
lp ——9.489+0.007 p, A and C =6.35 +0.4 pF. At fixed zero-voltage state. The positions of these resonances
bias current over the temperature range from 18 to 25 are in excellent agreement with the predictions of a
mK, the escape rate was temperature independent' in model in which the energy levels of the particle in the
the absence of microwaves, and in very good agree- well are quantized, with no fitting parameters. These
ment with the predictions for macroscopic quantum results provide very strong evidence for the quantum
tunneling. Figure 3 (a) shows resonances in nature of the macroscopic variable 5, and imply that
[I (P) —I (0)]/I (0) observed at 18 mK for four dif- quantum coherence between an excited state and the
ferent microwave frequencies. In Fig. 3(b) we plot the ground state can be maintained for times as long as 40
predicted energy difference between the ground state ns.
and first excited state, together with the estimated un- We are indebted to K. Daly, S. Diamond, D. Esteve,
certainty. The absolute measured positions of the res- R. E. Packard, N. E. Phillips, C. Urbina, and J. Van
onances agree with the predictions to within the exper- Curen for helpful discussion and/or experimental as-
imental uncertainties. The shift in the positions of the sistance. One of us (J.M. M. ) acknowledges the receipt
resonance as the microwave frequency is changed is in of a National Science Foundation graduate fellowship
excellent agreement with the predicted shift. Further- and an IBM predoctoral fellowship, while another

1545
VOLUME 55, NUMBER 15 PHYSICAL REVIEW LETTERS 7 OCTOBER 1985

(M. H. D.) acknowledges partial support from the Com- (1981).

missariat a l'Energy Atomique. This work was sup- 5L. D. Jackel, J. P. Gordon, E. L. Hu, R. E. Howard, L. A.
ported by the Director, Office of Energy Research, Of- Fetter, D. M. Tennant, R. W. Epworth, and J. Kurkijarvi,
fice of Basic Energy Sciences, Materials Sciences Phys. Rev. Lett. 47, 697 (1981).
Division of the U. S. Department of Energy under W. den Boer and R. de Bruyn Ouboter, Physica (Amster-
Contract No. DE-AC03-76SF00098. dam) 9$B, 185 (1980).
Note added. — Since this paper was submitted, two
7R. J. Prance, A. P. Long, T. D. Clark, A. Widom, J. E.
Mutton, J. Sacco, M. W. Potts, G. Megaloudis, and F. Goo-
other papers concerned with experiments on macro- dall, Nature 289, 543 (1981).
scopic quantum tunneling have appeared. ' 8T. A. Fulton and L. N. Dunkelberger, Phys. Rev. B 9,
4760 (1974).
M. H. Devoret, J. M, Martinis, D. Esteve, and J. Clarke,
Phys. Rev. Lett. 53, 1260 (1984).
toI. Affieck, Phys. Rev. Lett. 46, 388 (1981).
M. Buttiker, E. P. Harris, and R. Landauer, Phys. Rev. B
2$, 1268 (1983).
&'~On leave from Service de Physique du Solide et de Res- As a boundary condition on each of the eigenfunctions,
onance Magnetique, Centre d'Etudes Nucleaires de Saclay, P„(h), on the free-running side of the barrier, we set p„= 0
F-91191 Gif-sur- Yvette Cedex, France. at the value of 5 where the energy eigenvalue is equal to the
tA. J. Leggett, Prog. Theor. Phys. (Suppl. ) 69, 80 (1980), potential energy. This approximation leads to a negligible
and in Essays in Theoretical Physics in Honor of Dirk ter Haar error in the eigenvalues.
(Pergamon, Oxford, 1984), p. 95. M. H. Devoret, J. M. Martinis, and J. Clarke, unpub-
2W. C. Stewart, Appl. Phys. Lett. 12, 277 (1968); D. E. lished.
McCumber, J. Appl. Phys. 39, 3133 (1968). t4S. Washburn et al. , Phys. Rev. Lett. 54, 2712 (1985); D.
3A. O. Caldeira and A. J. Leggett, Ann. Phys. (N. Y.) 149, B. Schwartz et al. , in Proceedings of the Third International
374 (1983). Conference on Superconducting Quantum Devices, Berlin,
4R. F. Voss and R. A. Webb, Phys. Rev. Lett. 47, 647 25-28 June 1985 (unpublished).