PHYSICAL REVIEW LETTERS week ending

PRL 100, 103601 (2008) 14 MARCH 2008

Nondestructive Probing of Rabi Oscillations on the Cesium Clock Transition

near the Standard Quantum Limit
P. J. Windpassinger,1 D. Oblak,1 P. G. Petrov,1,* M. Kubasik,2,1 M. Saffman,3 C. L. Garrido Alzar,1,† J. Appel,1
J. H. Müller,1 N. Kjærgaard,1,‡ and E. S. Polzik1
1
QUANTOP, Niels Bohr Institute, University of Copenhagen, Denmark
2
ICFO, Institut de Ciències Fotòniques, Barcelona, Spain
3
Department of Physics, University of Wisconsin, Madison, Wisconsin 53706 USA
(Received 9 November 2007; published 13 March 2008)
We report on the nondestructive observation of Rabi oscillations on the Cs clock transition. The internal
atomic state evolution of a dipole-trapped ensemble of cold atoms is inferred from the phase shift of a
probe laser beam as measured using a Mach-Zehnder interferometer. We describe a single color as well as
a two-color probing scheme. Using the latter, measurements of the collective pseudospin projection of
atoms in a superposition of the clock states are performed and the observed spin fluctuations are shown to
be close to the standard quantum limit.

DOI: 10.1103/PhysRevLett.100.103601 PACS numbers: 42.50.Lc, 03.65.Ta, 06.30.Ft, 42.50.Nn

The problem of a two-level atom interacting resonantly In this Letter, we demonstrate a measurement of Rabi
with a coherent radiation field is of general importance to oscillations and the pseudospin of the cesium clock tran-
the field of atomic and optical physics, and quantum optics sition which is fast, nondestructive, and allows us to ap-
[1]. Two-level quantum systems and the ability to manipu- proach the standard quantum limit of sensitivity. The state-
late them coherently are essential to today’s atomic clocks dependent index of refraction of a cold atomic ensemble is
[2] and tomorrow’s quantum computers [3]. The coherence measured using a Mach-Zehnder (MZ) interferometer
lifetimes in these systems can be measured via the decay of (shown schematically in Fig. 1), which compares the phase
population oscillations when subjected to an external reso- accumulated by a laser beam in a probe arm to that of a
nant drive field. Such Rabi oscillations are routinely ob- spatially separated reference arm. The atoms are located in
served in quantum dots [4], Josephson junction qubits [5], the probe arm and modify the optical path length according
nitrogen-vacancy centers in diamond [6], as well as for to their internal quantum state. As compared to the polar-
trapped ions [7] and atoms [8]. In the case of trapped ions ization scheme [12] the MZ interferometer provides a
and atoms the internal quantum state j"i or j#i is typically refractive index change per atom which is larger since
detected via spontaneously scattered photons when prob- the reference beam is not phase shifted by the atoms. We
ing the atoms with laser light resonant with a transition employ a pulsed probing scheme to take advantage of ac
from one of the states to an auxiliary level. This method photo detectors characterized by very low electronic noise
instantaneously destroys the coherence between states j"i
and j#i interrupting the Rabi oscillations at the time of
probing. To sample the Rabi oscillations at given instances
of time requires at least as many repreparations of the
quantum system [9]. Rather than performing a completely
destructive measurement on an atomic sample, it is pos-
sible to gain information about the quantum state in a more
gentle fashion by using off-resonant probe light and con-
sidering the ensemble of atoms as a refractive medium.
Such nondestructive probing could, e.g., prove advanta-
geous to optical lattice clocks which are presently limited
by the time it takes to reprepare atomic samples [10,11].
Recently, a dispersive measurement of Rabi oscillations on
the cesium clock transition was demonstrated [12]. In that FIG. 1 (color online). Schematic drawing of the experimental
setup. A probe laser beam enters the Mach-Zehnder interfer-
work Rabi flopping between the Cs clock states was ob-
ometer via a port of the beam splitter BS1. One part of the beam
served by continuously probing the state-dependent bire- propagates through the Cs atoms as they interact with the
fringence of a cold atomic sample. The authors state that radiation field of the microwave source MW. The other part of
they were not able to reach the sensitivity sufficient for the beam propagates on a reference path with no atoms. The two
observation of the atomic quantum spin noise which sets parts are recombined at the beam splitter BS2 and the phase shift
the ultimate limit on sensitivity of such a measurement from the atoms is detected as the differential signal of detectors
[13]. D1 and D2 at the two output ports of BS2.

0031-9007=08=100(10)=103601(4) 103601-1 © 2008 The American Physical Society

 PHYSICAL REVIEW LETTERS week ending
PRL 100, 103601 (2008) 14 MARCH 2008

compared to typical dc detectors. This allows us to reduce the two arms have equal length. After loading the dipole
the number of probe photons and thus to minimize the trap we apply a small vertical bias field
1 G to define a
destructiveness of the atomic measurement. In particular, quantization axis, and optically pump the atoms to the
our detection of light is shot noise limited, which in turn F  4; mF  0  j"i clock state using vertically polar-
allows for the best detection of atomic spin noise. Finally, ized laser light driving 6S1=2 F  4;mF
0 ! 6P3=2 F0 
rather than being released from a magneto-optical trap 4  transitions. An additional repumping laser beam
(MOT), our atomic sample is confined by an optical dipole resonant with the 6S1=2 F  3 ! 6P3=2 F0  4  tran-
trap resulting in a denser sample, available for an extended sitions ensures that the atoms do not accumulate in the F 
period of time. As shown in [14], the optical depth of the 3 ground level. We end up with up to 80% of the atoms in
atomic ensemble is the key parameter for optimum gain of j"i with an efficiency likely to be limited by mF coherences
information about the collective quantum state versus its created by our narrow-bandwidth (
250 kHz) pumping
destruction. lasers [16].
Our experimental setup has been described in detail in We address the clock states with 9.2 GHz microwave
[15]. In brief, Cs atoms are loaded from a MOT into an radiation generated by a precision synthesizer. The orien-
optical dipole trap, where the atomic sample has a diameter tation of the linearly polarized magnetic field vector of the
of
60
m and contains up to 5  105 atoms, achieving microwave field is parallel to the guiding bias field of the
an optical depth around 25 as suggested by [14]. The atoms so that it only drives magnetic dipole  transitions.
trapped sample is located in one arm of our MZ interfer- We have full control over the power, frequency, relative
ometer as shown in Fig. 1. The geometric path length phase, and duration of the microwave field, so that we can
difference between the two arms of the interferometer is produce any target state on the generalized Bloch sphere of
locked by a piezoactuated mirror using a weak auxiliary far the effective two-level system [1]. As discussed in [15,17]
off-resonance laser beam counterpropagating the atomic the phase shift of a light beam with polarization state q and
probe beam. To minimize the influence of the laser phase wavelength  propagating a distance l through an atomic
noise of the locking and the probe beams, the interferome- gas with population NF;mF in each hyperfine substate
ter is aligned to the so-called white light position, where F; mF  confined to a volume V is

X
  
0 F0 1 F 2 J J0 1 2 =2
F;F0

  0 NF;mF 2F  12F  1 0 ; (1)
mF q mF F0 F I
2  =22
F;mF ;F0 ;m0F {z} |{z} F;F0
|
Wigner 3j symbol Wigner 6j symbol

where  is the linewidth of the transition,
F;F0 is the tuations of the probe light from the empty interferometer
detuning of the incident light with respect to the LJ F ! are limited by the shot noise [15].
L0J0 F0  transition, and 0  3l2 2J0  1=4V; the Figure 2 shows examples of microwave induced Rabi
Wigner 3j and 6j symbols are defined in [18]. The atom flopping between the clock states as observed with our
probing light is derived from an external grating stabilized interferometer. The probe light beam interacting with the
diode laser frequency locked to the atomic line. The fre- atoms has a power of 140 nW distributed into 0:5 and
quency of the probe laser is
4;5  160 MHz blue detuned 1:0
s pulses for the two traces shown, respectively (cor-
from the Cs D2 line F  4 ! F0  5 transition. Since the responding to 3  105 and 6  105 photons/pulse). A
phase-shift decreases with 1=
, this probe light is only marked difference in oscillation frequency and decay be-
sensitive to the atomic population in the F  4 state and tween the two traces can be observed as a result of the
the expression (1) reduces to (assuming mF  0 and light rather modest change in the probe pulse duration by a
polarized along the quantization axis q  0): factor of 2. The decay of oscillation visibility is caused
by several processes. Even in the absence of the probe light
5 =2
F;F0 a decay will be present due to dephasing between atoms

  0 N4;0 2 ; (2) caused by inhomogeneous differential light shifts from the
36
F;F0  =22
dipole trapping laser beam and small spatial variations in
the microwave driving field. The addition of probe laser
which is the j "i clock state contribution to the phase shift. light leads to a small amount of spontaneous photon scat-
Atomic probing is performed with light pulses with a tering which can be subdivided into inelastic Raman and
typical duration of a microsecond produced with a standard elastic Rayleigh events [19]. The former leads to complete
acousto-optical modulator. We record the pulses in a ho- decoherence while the effect of the latter depends on the
modynelike setup with a low noise differential photo de- internal state of the atoms in that atomic ensembles at the
tector. The detector output is digitized with a high poles of the Bloch sphere (i.e., all atoms in either the j"i or
bandwidth oscilloscope and the pulse areas extracted using the j#i state) are not affected since elastic scattering takes
numerical integration. It is crucial for the minimal destruc- an atom back to its original internal state. Even with the
tiveness of the measurement that the observed phase fluc- probing being close to nondestructive in the sense of little
103601-2
 PHYSICAL REVIEW LETTERS week ending
PRL 100, 103601 (2008) 14 MARCH 2008

what corresponds to a spin echo sequence, the dephasing

can be minimized.
As mentioned, our optical pumping scheme is not able to
transfer all of the atoms to the (F  4, mF  0) clock state.
In Fig. 2 this manifests itself as a fringe visibility limited to
54%. To purify the atomic polarization further we apply a
microwave -pulse transferring these atoms to the F 
3; mF  0  j#i clock state and blow away the remaining
atoms in the (F  4, mF
0) states using resonant light
on the 6S1=2 F  4 ! 6P3=2 F0  5 cycling transition.
In Fig. 3(a) we show Rabi oscillations as recorded for such
a purified ensemble initially in the (F  3, mF  0) clock
state. The atoms are probed with pulses containing &105
FIG. 2 (color online). Rabi oscillations (average of 10 experi- photons every 2:3
s corresponding to almost 50 times per
mental cycles) on the clock transition for an ensemble of Cs Rabi cycle. Since this is far beyond what can be resolved in
atoms [optically pumped towards the (F  4, mF  0) state] as the figure we include a zoom-in on the first millisecond of
inferred from the phase shift of probe laser light. The oscillations evolution in Figs. 3(b) and 3(c). Figure 3(b) shows a single
are recorded using probe pulses with a repetition period of 6
s
and duration of 0:5 and 1:0
s, respectively. In the latter case a
experimental realization with clear Rabi oscillations albeit
pronounced damping of oscillations is observed, however, with with some noise. Of course, by averaging over more ex-
coherent dynamics at longer times (see inset). In the former case perimental realizations [Fig. 3(c)] an improved signal-to-
the fringe contrast for the normalized phase shift is limited to noise ratio is achieved. Clearly, the data presented in
about 54% as result of
30% remaining atoms in (F  4, mF
Fig. 3(b) demonstrate the capabilities of our method in
0) states. that the coherent evolution of an ensemble can be followed
‘‘nondestructively’’ in ‘‘real time.’’
spontaneous photon scattering, the dispersive interaction As is well known, a two-level quantum system is for-
(responsible for the observable phase shift of probe light) mally equivalent to a spin 1=2 particle so that our phase-
will act as to wash out Rabi oscillations as a result of a shift measurement of the atomic ensemble can be inter-
frequency shift over the sample. An ac Stark shift is preted as a collective pseudospin projection observation.
imposed onto the atoms every time the probe light is Experimentally, we can prepare an ensemble of atoms in an
applied [20] changing the phase between clock states equal coherent superposition of j"i and j#i by applying a
superpositions as j#i  j"i 哬 j#i  ei j"i. In our experi- microwave =2 pulse to our atomic sample after the
ment the average value of  per photon is on the order of purification stage. It is possible to achieve a zero phase
4  107 rad which also characterizes the magnitude of shift for such a coherent spin state (CSS) in the balanced
the phase spread over the atomic ensemble as a result of the interferometer by employing a two-color probing scheme.
Gaussian probe beam profile. This inhomogeneous dephas- To determine the population in j#i, we use an additional
ing of atoms is responsible for the dramatic increase in probe laser at
3;2  135 MHz, red detuned from the
effective Rabi frequency and envelope decay when dou- F  3 ! F  2 transition and coupled to the interferome-
bling the probe photon number in Fig. 2 [21]. By only ter via the same optical fiber as the F  4 ! F0  5
applying the probe pulses every half oscillation period, in probing laser. The detunings are arranged such that with

FIG. 3 (color online). Purified sample of atoms (a) Rabi cycling on the clock transition recorded using 0:2
s probe pulses with a
repetition period of 2:3
s. Oscillations are still visible after a time of 8 ms (a running average filter over 8 pulses has been included at
this point of time to enhance the oscillations) corresponding to having applied
3500 probe pulses. The trace is an average of 50
realizations of the experiment. (b) A single experiment recording of Rabi oscillations during the first millisecond. (c) A zoom-in on the
first millisecond of the trace in (a), i.e., an average of 50 traces as shown in (b). At time 0 ms the phase shift is zero as a result of all the
atoms being transferred to the (F  3, mF  0) clock state and the probe only being sensitive to atoms in the F  4 states.
103601-3
 PHYSICAL REVIEW LETTERS week ending
PRL 100, 103601 (2008) 14 MARCH 2008

ally. When used in a feedback scheme the information

gained in the first measurement can be used to construct
unconditionally squeezed target states. Such engineered
nonclassical collective states open up the possibility to
surpass the projection noise limit for atomic clocks oper-
ated using a Ramsey sequence as discussed in [17] for our
experimental configuration.
In conclusion, we have measured the state-dependent
refractive index of a trapped ensemble of Cs atoms and
shown the ability to follow coherent processes such as Rabi
oscillations in ‘‘real time.’’ Using a shot noise limited two-
color probing scheme we have measured the noise of the
pseudospin projection of atoms in a superposition between
the clock states. The recorded atomic pseudospin noise is
FIG. 4 (color online). The fluctuations in the interferometer near the standard quantum limit —the projection noise,
phase for ensembles in an equal superposition of the clock states indicating our ability to perform a QND measurement on
as a function of the interferometer phase shift with all atoms at a
cold dipole-trapped atoms.
pole of the Bloch sphere (this phase shift is proportional to the
atom number). The quantum projection noise estimated from our This work was funded by the Danish National Research
experimental parameters is shown as a line. The inset shows the Foundation, as well as EU Grants QAP, COVAQUIAL, and
distribution of the measured pseudospin projection, its width EMALI. N. K. acknowledges the support of the Danish
increasing with atom number. National Research Council and helpful advise from
Stefan Kuhr.

equal powers of the two probes, we obtain a zero mean

phase shift for equal populations in both levels, irrespective
*Present address: Department of Physics, Ben Gurion
of the number of atoms in the ensemble. Using both probe
University, Be’er-Sheva 84105, Israel.
beams (colors) at the same time thus gives information on †
Present address: LCF, Institut d’Optique, 91127 Palaiseau,
the population number difference of the two hyperfine France.
states. It is known [14,17] that when the probe power is ‡
[email protected]
increased the pseudospin measurements become limited by [1] L. Allen and J. H. Eberly, Optical Resonance and Two-
the projection noise of atoms (the quantum fluctuations of Level Atoms (Dover, New York, 1987).
the CSS [22]). In order to reach this regime we employ a [2] J. Vanier and C. Audoin, The Quantum Physics of Atomic
single pulse with 3:6  107 photons, roughly correspond- Frequency Standards (Adam Hilger, Bristol, 1989).
ing to the total photon number used during the entire [3] P. Zoller et al., Eur. Phys. J. D 36, 203 (2005).
observation cycle of Fig. 3(b). Figure 4 shows the observed [4] T. H. Stievater et al., Phys. Rev. Lett. 87, 133603 (2001).
atomic pseudospin fluctuations as a function of the atom [5] J. M. Martinis et al., Phys. Rev. Lett. 89, 117901 (2002).
number along with an estimate of the atomic quantum [6] F. Jelezko et al., Phys. Rev. Lett. 92, 076401 (2004).
projection noise. The data presented results from the analy- [7] D. Leibfried et al., Rev. Mod. Phys. 75, 281 (2003).
[8] M. R. Matthews et al., Phys. Rev. Lett. 81, 243 (1998).
sis of
34 000 repetitions of the experimental cycle. As
[9] S. Kuhr et al., Phys. Rev. A 72, 023406 (2005).
can be seen, for low atom numbers our measurements of [10] M. Takamoto et al., Nature (London) 435, 321 (2005);
the collective pseudospin projection show the expected A. D. Ludlow et al., Phys. Rev. Lett. 96, 033003 (2006);
quantum noise level and characteristic linear scaling of R. Le Targat et al., ibid. 97, 130801 (2006).
atomic noise (variance) with atom number. For higher [11] D. Meiser, J. Ye, and M. J. Holland, arXiv:0707.3834.
atom numbers, however, nonquantum noise sources con- [12] S. Chaudhury et al., Phys. Rev. Lett. 96, 043001 (2006).
tribute by a non-negligible part. We attribute this to the [13] D. J. Wineland et al., Phys. Rev. A 46, R6797 (1992).
relative amplitude and phase fluctuations of the two inde- [14] K. Hammerer et al., Phys. Rev. A 70, 044304 (2004).
pendent probes. Reduction of this classical (quadratic) [15] P. G. Petrov et al., Phys. Rev. A 75, 033803 (2007).
noise component will take us to the regime where the [16] P. Tremblay and C. Jacques, Phys. Rev. A 41, 4989 (1990).
atomic projection noise becomes comparable to the shot [17] D. Oblak et al., Phys. Rev. A 71, 043807 (2005).
[18] I. I. Sobelman, Atomic Spectra and Radiative Transitions
noise of the probe light, which would be of considerable
(Springer, Berlin, 1979).
interest since our dispersive light-atom interaction has a [19] R. Ozeri et al., Phys. Rev. Lett. 95, 030403 (2005).
quantum nondemolition (QND) character [23]. A measure- [20] P. D. Featonby et al., J. Phys. B 31, 375 (1998).
ment of the collective pseudospin projection could hence [21] P. Windpassinger et al., arXiv:0801.3242.
be used to predict the outcome of a subsequent measure- [22] G. Santarelli et al., Phys. Rev. Lett. 82, 4619 (1999).
ment beyond the standard quantum limit. Such predictive [23] A. Kuzmich, N. P. Bigelow, and L. Mandel, Europhys.
power squeezes the variance of the projection condition- Lett. 42, 481 (1998).

103601-4