Professional Documents
Culture Documents
Wojciech Wasilewski,1 Kasper Jensen,1 Hanna Krauter,1 Jelmer J. Renema,1 and Eugene S. Polzik1
1
Niels Bohr Institute, Danish Quantum Optics Center QUANTOP,
Copenhagen University, Blegdamsvej 17, 2100 Copenhagen , Denmark
Ultimate sensitivity of metrology and sensing is defined by the fundamental principles of quantum
physics. The quantum state of a system determines the magnitude of its quantum fluctuations. The
optimal balance between the information about the system obtained during the measurement and
the destructive back action of the measurement on the measured object is again defined by the laws
of quantum mechanics. Here we demonstrate an atomic magnetometer operating at the fundamental
limit of sensitivity set by the quantum projection noise (PN) of atoms and beyond it, when assisted
by atomic entanglement. Using techniques developed for quantum information processing [1, 2, 3] we
apply an optimal sequence of state preparation, evolution, and the back-action evading measurement
to achieve the PN limited sensitivity. We furthermore experimentally demonstrate that Einstein-
arXiv:0907.2453v2 [quant-ph] 12 Oct 2009
A number of measurements in physics can be viewed frame after the rf pulse is ΓBRF Jx T2 [1 − exp(−τ /T2 )]/2.
as a measurement of a small angle of rotation of a vec- Here T2 is the spin decoherence time during the rf pulse
tor. For NA atoms in a state with the magnetic quantum and Γ = Ω/B = 2.2 · 1010 rad/(sec·Tesla) for caesium.
number mF = F along a quantization axis x the vector Equating the mean value to the PN uncertainty we get
is a collective magnetic moment (spin) of the ensemble for the minimal detectable field under the PN limited
with the length Jx = F NA . A magnetic field along the y measurement
axis causes a rotation of J~ in the x−z plane. Polarization h p i−1
of light propagating along z will be rotated proportional Bmin = Γ F NA /2 · T2 1 − e−τ /T2 . (1)
to Jz due to the Faraday effect. Ultra-sensitive atomic
magnetometry is based on the measurement of this polar- Note that for the best sensitivity to the BRF with a
ization rotation [4]. From a quantum mechanical point given δ the atomic bandwidth should be narrow: δ =
of view, this measurement is limited by quantum fluc- τ −1 ≥ T2 −1 (see Supplementary Information for details).
tuations (shot noise) of light, the quantum backaction A long T2 also helps to take advantage of the entangle-
noise of light onto atoms and the projection noise (PN) ment of atoms. Entangled states are fragile and have a
of atoms. PN originates from the Heisenberg uncertainty shorter lifetime T2E < T2 . As we demonstrate here, un-
relation δJz ·δJy ≥ Jx /2, and corresponds to the minimal der the condition δ −1 < T2E < T2 , that is for broadband
2 RF fields, entanglement improves the sensitivity. Similar
transverse spin variances δJz,y = Jx /2 = F NA /2 for un-
correlated atoms in a coherent spin state [5]. Quantum conclusions have been reached theoretically for atomic
entanglement as the means of enhancing the performance clocks in [8, 10] and for dc magnetometry in [9].
of metrology and sensing has been widely discussed theo- In our experiment a long magnetometer coherence time
retically [5, 6, 7, 8, 9, 10, 11, 12]. In [13, 14] entanglement T2 > 30msec is achieved by the time resolved quantum
of a few ions have been used in spectroscopy. Recently spin state preparation (optical pumping), evolution, and
proof-of-principle measurements with larger atomic en- measurement, as in quantum information processing pro-
sembles, which go beyond the PN limit have been imple- tocols [2, 3]. In this way T2 is not reduced by the optical
mented in interferometry with 103 atoms [15] and Ram- pumping and/or measurement-induced decoherence.
sey spectroscopy [16, 17] with up to 105 atoms. PN limited sensitivity requires, besides elimination of
the technical noise, the reduction of the back action noise
In this Letter we demonstrate quantum limited de- of the measurement and of the shot noise of the probe
tection of a radio-frequency (rf) magnetic field by an light.
atomic vapour magnetometer operating with thousands The back action noise comes from the Stark shift im-
of billions of atoms. In the magnetometer J~ precesses at posed by quantum fluctuations of light polarization on
the Larmor frequency Ω/2π = 322kHz around a dc field Jylab in the laboratory frame when Jzlab is measured. Since
B = 0.92G applied along the x axis and an rf field with
Jzlab and Jylab experience Larmor precession, both of them
the frequency Ω is applied in the z − y plane (Fig. 1a).
accumulate the back action noise [1, 18]. As proposed in
In the experiment (Fig. 2a) the magnetometer detects [1] the effect of this noise can be canceled if two atomic
an rf pulse with a constant amplitude BRF and duration ensembles with orientations Jx and −Jx are used. Fig. 2a
τ (Fourier limited full width half maximum bandwidth presents the sketch of the magnetometer layout which
δ = 0.88τ −1 ≈ τ −1 ). The mean value of the projection is used for the back action evading measurement of the
of the atomic spin on the y − z plane in the rotating BRF .
2
sion, Cs-Cs collisions and probe light-induced collisions. The maximal spin projection is reached in the time
Increasing the size of the cells may help to reduce some
of those effects since a larger optical depth will then be 1
tmax = ln (δT2 ) / δ − ≈ min (T2 , 1/δ) (5)
achieved for a given density of atoms. T2
which is on the order of the smallest of the atomic life-
time T2 and the inverse rf bandwidth 1/δ. At this
Methods time the largest
p signal to noise ratio is reached S/N =
|J⊥ (tmax )| / Jx /2.
Projection noise calibration. The probe light changes Using the time tmax we obtain an estimate for the PN
the atomic state, as described by the counterpart of limited sensitivity which has the following scaling with
Eq. (2): the two parameters, the magnetometer bandwidth 1/T2
r and the rf bandwidth δ:
p 2F NA in √
Jˆy−
0
= (Jˆy1
0
−Jˆy2
0
) = (Jˆy1 −Jˆy2 ) 1 − ξ 2 κ2 +κ Ŝ3c , δ for δ 1/T2 using the time t ≈ 1/δ.
Φ sens ∝ p
(3) 1/T2 for δ 1/T2 using the time t ≈ T2 .
where 0 denotes the atomic state after the interaction. (6)
in
Ŝ3c is the 322kHz cosine component of the Stokes oper- For
√ large rf bandwidths the sensitivity is proportional
ator given by the difference of the number of the right to δ and independent of T2 . For smallp rf bandwidths
hand and left hand circular polarized photons. Using an the sensitivity is proportional to 1/T2 and inde-
electro-optical modulator we can create a certain aver- pendent of δ. The magnetometer is therefore most
age value in sensitive to rf fields with small bandwidths δ ≤ 1/T2 .
p of Ŝ3c and then calibrate it in the units of shot2
noise Φ/2 by homodyne detection. To determine κ Similar conclusions have been reached theoretically for
for a certain light power andp atom number, we first send atomic clocks in [8, 10] and for dc magnetometry in [9].
a 1msec pulse with hŜ3c in
i = Φ/2, which is mapped on From the above considerations we can determine when
entanglement will improve the sensitivity. In general,
the atoms. Then we flip the sign of Jˆy2 and Jˆz2 by mo-
an entangled state will have shorter lifetime than a
mentarily increasing the dc magnetic field and send a sec-
state which is not entangled: T2E < T2 , where T2E is
ond 1msec light pulse. Now hJˆy− 0 out
i is mapped on hŜ2s i,
the lifetime of the entangled state. Since a reduction
following equation (2), and measurement on the second
out of the atomic lifetime degrade the sensitivity for small
pulse yields hŜ2s i = κ2 . Using κ2 we find Var S2c,s
out
rf bandwidths, entanglement only helps in the opposite
from Eq. (2) and by comparing it to the measured noise case of large rf bandwidths δ 1/T2E .
level we calibrate the detection efficiency η = 0.80(5).
Finally we calibrate the atomic displacement Experimental data showing the improvement due to
(hJˆy i, hJˆz i) caused
√ by a particular BRF field in entanglement is presented Fig. 4c in the main text. In
units of PN, F NA . After applying the BRF we the experiment we applied a rf pulse with duration τ and
send a short probe pulse with a known value of constant amplitude for which the FWHM bandwidth is
κ. Then the mean value of the homodyne signal is 0.88/T ≈ τ −1 . In Fig. 4c we plot the signal to noise ratio
√ p
hS̃ˆ2out i = κ η Φ/2F NA (hJy1 + Jy2 i + ihJz1 − Jz2 i). (S/N ) times the rf bandwidth δ = 1/τ as a function of
An RF pulse with the same strength and duration can the rf bandwidth, squares represent data without the use
then be used to measure κ and evaluate the noise of of entanglement and circles represent data with the use
measurements at different probe powers. of entanglement (created by a measurement prior to the
rf pulse). For rf bandwidths δ = 0.2 − 2 kHz the signal to
noise ratio is clearly improved by the usage of entangle-
Supplementary information ment. The improvement increases with the bandwidth
due to the smaller lifetime of the entangled state. Since
A. Sensitivity and bandwidth
the sensitivity is inversely proportional to the signal to
noise ratio, the improvement in the signal to noise ratio
is directly an improvement of the sensitivity.
To examine the connection between magnetometer sen- To guide the eye, fits to the data are plotted in Fig. 4c.
sitivity and the bandwidth of the rf field we analyze the For the case of no entanglement, the expected depen-
case of an exponentially decaying rf field with amplitude dence on δ and T2 is (S/N ) · δ ∝ δ · T2 1 − e−1/δT2 ,
BRF e−δt . From the norm square of the Fourier trans- since a rf pulse with constant amplitude creates the trans-
form, we find the FWHM of the rf field to be δ/π (with verse spin component |J⊥ (τ )| ∝ T2 1 − e−τ /T2 . A fit to
units Hz). After applying the rf field for a time t, the this function is shown as a dashed line, where in the fit
projection of the atomic spin on the y − z plane is T2 = 32 ms was fixed and the proportionality constant
was the only free parameter. For the case of entangle-
ΓBrf Jx e−δt − e−t/T2 ment the data is fitted to a slightly different model, where
|J⊥ (t)| = · 1 . (4)
T2 − δ
2 it is assumed that the entanglement (corresponding to
5
the noise) decays at a rate T2E different than the decay C. Experimental optimization of the signal to noise
rate T2 of the mean value. ratio
[1] Julsgaard, B., Kozhekin, A. & Polzik, E. S. Experimen- [5] Wineland, D. J., Bollinger, J. J., Itano, W. M., Moore,
tal long-lived entanglement of two macroscopic objects. F. L. & Heinzen, D. J. Spin squeezing and reduced quan-
Nature 413, 400 (2001). tum noise in spectroscopy. Phys. Rev. A 46, R6797–
[2] Julsgaard, B., Sherson, J., Cirac, J. I., Fiurasek, J. & R6800 (1992).
Polzik, E. S. Experimental demonstration of quantum [6] Petersen, V., Madsen, L. B. & Mølmer, K. Magnetometry
memory for light. Nature 432, 482–486 (2004). with entangled atomic samples. Phys. Rev. A 71, 012312
[3] Sherson, J. F. et al. Quantum teleportation between light (2005).
and matter. Nature 443, 557–560 (2006). [7] Stockton, J. K., Geremia, J. M., Doherty, A. C. &
[4] Budker, D. & Romalis, M. Optical magnetometry. Nat Mabuchi, H. Robust quantum parameter estimation: Co-
Phys 3, 227–234 (2007). herent magnetometry with feedback. Phys. Rev. A 69,
6
032109 (2004).
[8] Huelga, S. F. et al. Improvement of frequency standards
with quantum entanglement. Phys. Rev. Lett. 79, 3865–
3868 (1997).
[9] Auzinsh, M. et al. Can a quantum nondemolition mea-
surement improve the sensitivity of an atomic magne-
tometer? Phys. Rev. Lett. 93, 173002 (2004).
[10] André, A., Sørensen, A. S. & Lukin, M. D. Stability of
atomic clocks based on entangled atoms. Phys. Rev. Lett.
92, 230801 (2004).
[11] Kominis, I. K. Sub-shot-noise magnetometry with a
correlated spin-relaxation dominated alkali-metal vapor.
Phys. Rev. Lett. 100, 073002 (2008).
[12] Cappellaro, P., Lukin, M. D. Quantum correlation in
disordered spin systems:Applications to magnetic sens-
ing. Phys. Rev. A 80, 032311 (2009).
[13] Liebfried, D., et al. Toward Heisenberg-limited spec-
troscopy with multiparticle entangled states. Science
304, 1476–1478 (2004).
[14] Roos, C. F., Chwalla, M., Kim, K., Blatt, R. ‘Designer
atoms’ for quantum metrology. Nature 433, 443:316–319
(2006).
[15] Esteve, J., Gross, C., Weller, A., Giovanazzi, S. &
Oberthaler, M. K. Squeezing and entanglement in a bose-
einstein condensate. Nature 455, 1216–1219 (2008).
[16] Schleier-Smith, M. H., Leroux, I. D. & Vuletic, V.
Reduced-quantum-uncertainty states of an ensemble of
two-level atoms. arXiv:0810.2582.
[17] Appel, J. et al. Mesoscopic atomic entanglement for pre-
cision measurements beyond the standard quantum limit.
Proceedings of the National Academy of Sciences. pub-
lished online, doi:10.1073/pnas.0901550106.
[18] Savukov, I. M., Seltzer, S. J., Romalis, M. V. & Sauer,
K. L. Tunable atomic magnetometer for detection of
radio-frequency magnetic fields. Phys. Rev. Lett. 95,
063004 (2005).
[19] Kupriyanov, D. V., Mishina, O. S., Sokolov, I. M., Juls-
gaard, B. & Polzik, E. S. Multimode entanglement of
light and atomic ensembles via off-resonant coherent for-
ward scattering. Phys. Rev. A 71, 032348 (2005).
[20] Wasilewski, W. et al. Generation of two-mode squeezed
and entangled light in a single temporal and spatial
mode. Optics Express 17, 14444 (2009).
[21] Lee, S.-K., Sauer, K. L., Seltzer, S. J., Alem, O. & Ro-
malis, M. V. Subfemtotesla radio-frequency atomic mag-
netometer for detection of nuclear quadrupole resonance.
Applied Physics Letters 89, 214106 (2006).
[22] Seton, H.-C., Hutchison, J. M., Bussel, M. Liquid helium
cryostat for SQUID-based MRI receivers. Cryogenics 45,
348 (2005).
[23] Sherson, J., Julsgaard, B. & Polzik, E. S. Advances in
Atomic, Molecular, and Optical Physics, chap. Determin-
istic atom-light quantum interface (2006).
[24] Duan, L.-M., Lukin, M. D., Cirac, J. I. & Zoller, P. Long-
distance quantum communication with atomic ensembles
and linear optics. Nature 414, 413 (2001).