Vol 464 | 22 April 2010 | doi:10.

1038/nature08919

LETTERS
Nonlinear atom interferometer surpasses classical
precision limit
C. Gross1, T. Zibold1, E. Nicklas1, J. Estève1{ & M. K. Oberthaler1

Interference is fundamental to wave dynamics and quantum In linear interferometry, the phase precision for an N-atom coherent
mechanics. The quantum wave properties of particles are spin state can be explained by classical statistics. The situation is equi-
exploited in metrology using atom interferometers, allowing for valent to N individual measurements on a single particle. At the most
high-precision inertia measurements1,2. Furthermore, the state-of- sensitive point of the interferometer, each particle has an equal chance
the-art time standard is based on an interferometric technique pffiffiffiffi in state jaæ or state jbæ; the uncertainty for N particles
of being measured
known as Ramsey spectroscopy. However, the precision of an is thus Dn 5 N /2. The maximal slope, hn/hQ 5 N/2, is determined
interferometer is limited by classical statistics owing to the finite by the visibility, , which is close to one for a macroscopically popu-
number of atoms used to deduce the quantity of interest3. Here we lated coherent spin state. The resultingpffiffiffiffi minimal phase error for a
show experimentally that the classical precision limit can be sur- classical measurement, DQ 5 1/ N , is known as the standard
passed using nonlinear atom interferometry with a Bose–Einstein quantum limit. In the case of correlated particles, this classical limit
condensate. Controlled interactions between the atoms lead to can be exceeded; doing so is the subject of the emerging field of
non-classical entangled states within the interferometer; this quantum metrology. The idea is to achieve improved scaling of the
represents an alternative approach to the use of non-classical input interferometric phase sensitivity with the number of quanta, N, which
states4–8. Extending quantum interferometry9 to the regime of are typically atoms or photons, using entangled quantum states within
large atom number, we find that phase sensitivity is enhanced by the interferometer9,18. Here the fundamental quantum limit is known
15 per cent relative to that in an ideal classical measurement. Our as Heisenberg limit, where the phase error is DQ 5 1/N.
nonlinear atomic beam splitter follows the ‘one-axis-twisting’ In this Letter, we report a direct experimental demonstration of
scheme10 and implements interaction control using a narrow interferometric phase precision beyond the standard quantum limit
Feshbach resonance. We perform noise tomography of the in a novel nonlinear Ramsey interferometer. Two hyperfine states, jaæ
quantum state within the interferometer and detect coherent spin and jbæ, of a 87Rb Bose–Einstein condensate act as the two modes of
squeezing with a squeezing factor of 28.2 dB (refs 11–15). The the interferometer. The atoms in the two states are trapped in the
results provide information on the many-particle quantum state, wells of a deep one-dimensional optical lattice (deep enough that
and imply the entanglement of 170 atoms16. there is no tunnelling coupling). Absorption imaging at high spatial
The concept of interferometry relies on the splitting of a quantum resolution allows the direct measurement of the populations in jaæ
state into two modes, a period of free evolution and, finally, the recom- and jbæ in each well. The two hyperfine states are detected after each
bination of the modes for readout. The key observable, the accumu- other with a delay of 780 ms. Controlled interparticle interactions
lated relative phase, is inferred from the measured population using a Feshbach resonance allow for the realization of a nonlinear
difference between the two output states17. A prominent example of beam splitter in which the state at its output ports is a coherent spin
linear interferometry is Ramsey spectroscopy, a technique used to squeezed state (Fig. 1b). Coherent spin squeezed states are a special
define the current time standard. The concept underlying Ramsey kind of many-particle entangled state for which the minimal inter-
pffiffiffiffi
spectroscopy is shown in Fig. 1a. A quantum state of N atoms is ferometric phase error is DQ 5 jS/ N , where j2S 5 N(DJz)2/
transformed unitarily by a linear beam splitter, resulting in a coherent (ÆJxæ2 1 ÆJyæ2) (refs 11, 12) is the coherent spin squeezing factor. In
spin state jYæ / (^ a{ z^b{ )N j0æ, where ^a{ and ^b{ are the creation opera- the symmetric two-mode case, the total spin is J 5 N/2 for a system
tors for the modes a and b. This beam splitter is typically realized using consisting of N quanta. For large N and small population difference,
a p/2 microwave pulse coupling two internal atomic states. The mis- n = N, the orthogonal spin components are related to the population
match between the frequency of the microwave field and the frequency difference, n, and relative phase, Q, between the two modes as follows:
of the atomic transition to be measured leads to an accumulated phase, Jz 5 n, Jx < N/2cos Q and Jy < N/2sin Q. Here we choose a coordinate
Q, after the evolution time, t, and to a quantum state jYæ / system such that ÆJyæ 5 0 when measuring the mean spin length
(^a{ zeiQ ^b{ )N j0æ. To convert the relative phase, Q, into an observable ÆJxæ2 1 ÆJyæ2. The spin representation allows the visualization of
population difference, n 5 (Na 2 Nb)/2, where Na and Nb are the quantum states on a generalized Bloch sphere as shown in Fig. 1c.
respective populations of the two modes a and b, the two states are For the realization of the nonlinear beam splitter, we follow the
coupled again before readout. Varying the relative phase by 2p causes one-axis-twisting scheme proposed in ref. 10. The detailed interfero-
the observed population difference to change sinusoidally, producing metric sequence is shown in Fig. 2a. A fast p/2 pulse produces a
a so-called Ramsey fringe. The interferometer’s phase error, DQ 5 Dn/ coherent spin state with ÆJzæ 5 0, and subsequent evolution under
(hn/hQ), is determined from the root mean squared error, Dn, of the the influence of interactions causes a shearing effect transforming
population difference and the slope of the interference signal, hn/hQ. the circular uncertainty region into an elliptical one as detailed in
Best experimental performance is achieved at the zero crossing (n 5 0) Fig. 3a. The resulting quantum state is a coherent spin squeezed state
of the Ramsey fringe, where the slope is maximal. where the squeezed direction forms at an angle a0 relative to the Jz
1
Kirchhoff-Institut für Physik, Universität Heidelberg, Im Neuenheimer Feld 227, 69120 Heidelberg, Germany. {Present address: Laboratoire Kastler Brossel, CNRS, UPMC, Ecole
Normale Supérieure, 24 rue Lhomond, 75231 Paris, France.

1165
©2010 Macmillan Publishers Limited. All rights reserved
LETTERS NATURE | Vol 464 | 22 April 2010

a b c

| b〉

| a〉

χJz2
Nonlinear
beam splitter
| b〉

ϕ ϕ | a〉

ϕ ϕ

| b〉

〈Jz 〉 〈Jz 〉
| a〉

Figure 1 | Comparison of linear and nonlinear interferometry. a, Schematic squeezed state improves the precision of this interferometer beyond the
of a classical linear interferometer. The input state (top) is brought into a classical limit. c, We prepare six independent Bose–Einstein condensates of
87
superposition of two modes by the first beam splitter, which is indicated in Rb in a one-dimensional optical lattice. Two hyperfine states form a two-
the Bloch sphere representation. This coherent spin state evolves freely and mode system, that is, the interferometer, in each well. The Bloch sphere
acquires a phase shift, Q, proportional to the quantity to be measured until representation of the quantum state in one well is shown for all atoms in
the two modes are mixed by a second beam splitter, where the acquired state | aæ (blue), in state | bæ (red) and in a coherent superposition of the two
relative phase translates into an observable mean population imbalance, states (purple). An imaging system with high spatial resolution (,1 mm)
ÆJz(Q)æ, at the two output ports. b, Here the input state is split by a nonlinear allows the detection of the condensate in each well individually using high-
beam splitter, leading to an entangled quantum state, that is, a coherent spin intensity absorption imaging30. State-selective time-delayed imaging causes
squeezed state with reduced fluctuations in the relative phase, Q. This phase the atomic clouds to have different shapes.

axis. The final step in realizing the nonlinear beam splitter is the optical lattice potential13, we are able to perform six independent experi-
rotation of the uncertainty ellipse around its centre by a 5 p/2 1 a0 ments in parallel (Fig. 1c). This results in increased statistics for a given
to prepare a phase squeezed state. The quality of the output state measurement time. The single traps are almost spherical and have
critically depends on the choice of the axis about which this rotation dipole frequencies of vtrap 5 2p 3 425 Hz. An adiabatic passage is used
is made. Therefore, we add a spin echo pulse in the evolution period to sweep the atoms to the state jaæ 5 j1, 1æ. Using two-photon com-
to correct for technical dephasing. After the beam splitting process, bined microwave and radio-frequency pulses, we couple the jaæ and
we allow for t 5 2 ms of phase accumulation time and then recouple jbæ 5 j2, 21æ states with a Rabi frequency of V. The single-photon
the states jaæ and jbæ using a p/2 pulse with a controlled phase Q before detuning to the j2, 0æ state is d 5 2200 kHz. The chosen states have a
readout of the population imbalance. narrow Feshbach resonance at a magnetic field of B 5 9.10 G, which
The Ramsey fringe resulting from this nonlinear interferometric allows us to control the interspecies s-wave scattering length, aab, and
sequence is shown in Fig. 2b (red), and from it we deduce a visibility therefore to access the miscible regime20–23.
of 5 0.92 6 0.02. For a direct comparison with the linear interfero- For our trap geometries and interaction regime, the single-spatial-
meter, we replace the nonlinear beam splitter with a standard linear mode approximation has been shown to be well justified24. The
one — realized using a single p/2 pulse — and keep the total number dynamics is described by the Josephson Hamiltonian, H/B
of atoms at the output constant. We obtain a fringe with a visibility of 5 xJz2 1 VJc 1 Dv0Jz, where Jc 5 Jxcos c 1 Jysin c is the component
5 0.98 6 0.02 (Fig. 2b, blue). The solid lines above and below the of angular momentum in the direction in which the coupling pulses
two fringes in Fig. 2b show the total spin length, 6N/2, for reference. are applied, x / aaa 1 abb 2 2aab is the effective nonlinearity arising
To determine the phase error, we analysed the region around the zero from intra- and interspecies interaction25, Dv0 is the detuning from
crossing in more detail, as shown in Fig. 2c. We performed 120 the resonance frequency, v0, and B is Planck’s constant divided by
experimental runs per phase setting and measured DJz for a linear 2p. The ratio of the background scattering lengths, aaa:aab:abb
interferometer (Fig. 2c, blue) and a nonlinear interferometer (Fig. 2c, 5 100:97.7:95, results in a very small effective nonlinear interaction,
red). It is important to note that none of the data shown in Fig. 2c has but using the Feshbach resonance decreases aab such that
been corrected for noise. The grey shaded areas indicate the 2-s.d. x 5 2p 3 0.063 Hz at B 5 9.13 G. The Rabi frequency, V, can be
bounds for an ideal classical measurement, that is, one in which there switched rapidly between 0 and 2p 3 600 Hz, allowing for fast dia-
is no excess noise due to the measurement process. The solid lines are batic coupling of the states. We keep the magnetic field constant
linear fits to the upper and lower ends of the 2-s.d. error bars of the throughout the whole interferometric sequence, but during the
measured data. The difference between the slopes of these lines and coupling pulses the interaction xJz2 can be neglected because the
the ideal measurement is caused by the slightly decreased visibility. In system is in the Rabi regime, that is, xN/V = 1. Differential energy
the case of nonlinear interferometry, reduced noise in the population shifts, Dv0, between the jaæ and jbæ states are mainly due to un-
difference is apparent. The performance of this interferometer in controlled magnetic field fluctuations. We use active magnetic field
phase estimation is 15% superior to that in the ideal classical case, stabilization to suppress the low-frequency components of the mag-
as shown by the dashed lines indicating the respective phase errors in netic noise to sub-milligauss levels. This is of utmost importance
Fig. 2c. This is remarkable because the phase error that we measure because the performance of the nonlinear beam splitter crucially
for the linear interferometer is 24% higher than that expected in an relies on the final rotation around an axis stable through the centre
ideal apparatus. This excess noise is due to readout noise in the of the noise ellipse.
detection scheme13,19 and experimental imperfections in the coupling In addition to the interferometry experiment, we characterize the
pulses (Supplementary Information). quantum state produced by the nonlinear beam splitter by noise
The experimental sequence starts with a Bose–Einstein condensate of tomography26. A rotation by a around the centre of the quantum
2,300 87Rb atoms in the jF 5 1, mF 5 21æ hyperfine state in an optical state maps spin directions orthogonal to the mean spin orientation to
dipole trap. By splitting the trap into six using a one-dimensional the Jz readout axis, such that their fluctuations can be measured.
1166
©2010 Macmillan Publishers Limited. All rights reserved
NATURE | Vol 464 | 22 April 2010 LETTERS

a Jz
α0 ϕ
Jy
Jx

Jz
0
〈Jz〉

–N π/2–y π–y (π/2 + α0)x π/2–x+ϕ

9 ms 330 μs 9 ms 2 μs t
Nonlinear beam splitter

b c
100
200

100
Atom number difference, Jz

0 0

〈Jz〉
–100

∆ϕnl
–200
∆ϕl
–100

–180 –90 0 90 180 –20 0 20

Figure 2 | Direct experimental demonstration of precision beyond the 0.98 6 0.02, for the linear interferometer, to 0.92 6 0.02, for the nonlinear
standard quantum limit. a, The nonlinear Ramsey interferometric sequence interferometer. The solid lines show the total atom number, 6N/2,
in the Bloch sphere representation. The states before and after the coupling measured for each phase setting, as a reference. c, To determine the
pulses are represented by grey and red shading, respectively. Rotation angles performance of the nonlinear interferometer, we repeat the population-
of the different pulses and their rotation axes (subscripts) are as indicated. difference measurement for a given phase 120 times (red), concentrating on
The curve shows the temporal evolution of Jz, its width indicating the the region shown in the dotted box in b. The solid lines are fits through the
corresponding variance. The nonlinear beam splitter realized with the first lower and upper ends of the 2-s.d. error bars. The grey shaded areas show the
three pulses produces a phase squeezed state. The last pulse, with phase Q, corresponding uncertainty regions for an ideal linear interferometer with
mixes the | aæ and | bæ modes before readout of the population imbalance. b, A full visibility. For the nonlinear interferometer, we find a phase error, DQnl,
Ramsey fringe observed when scanning the phase, Q, over a full period. The 15% smaller than that, DQl, for the ideal linear scheme. This is remarkable
blue data correspond to linear interferometry (the first three pulses are because for classical linear interferometry (blue), technical noise causes
replaced with a single p/2 pulse) and the red data correspond to nonlinear phase errors 24% larger than those expected for an ideal measurement. For
interferometry. Technical imperfections cause the decrease in visibility from clarity, the two measurements are vertically displaced.
Figure 3a shows a close-up of the results for small angles. We find The measured Heisenberg uncertainty product 4(DJz)2(DJy)2/ÆJxæ2,
minimal fluctuations for a0 5 16u; the corresponding number of conjugate variances, is 1.65 6 0.35, which is only slightly larger
squeezing factor is j2N 5 2(DJz)2/J 5 26.9z0:8{0:9 dB, where we average than the one expected for a minimal-uncertainty state. A fit to the
results for a 5 16u and a 5 17u to increase the statistics to 634 mea- data assuming evolution under the one-axis-twisting Hamiltonian10
surements. Here photon shot noise due to the imaging process is with the nonlinearity x as a free parameter shows very good agree-
removed13,19 (Fig. 3a, blue). The red data in Fig. 3a show the result ment with theory (Fig. 3b) for x 5 2p 3 0.063 Hz. The discrepancy
of also subtracting technical noise due to coupling-pulse imperfec- between fit and data in the number squeezed region is mainly due to a
tions and magnetic field fluctuations (Supplementary Information). loss of 15% of the atoms during the nonlinear evolution24,28. To put
Using this correction, we deduce a number squeezing factor of the best number squeezing factor, j2N 5 28.2 dB, in context with the
j2N 5 28.2z0:9
{1:2 dB, which is close to the atom-loss-limited theoretical interferometer experiment, it is important to note that there the
optimum for our internal system, ,210 dB (ref. 24). performance of the nonlinear beam splitter was j2N 5 24.3 dB or
Maximum fluctuations of j2N,max 5 110.3z0:3 {0:4 dB are observed for j2N 5 22.1 dB, where the first value was inferred from the popu-
measurement along the axis orthogonal to the squeezed direction and lation-difference fluctuations at the interferometer output with sub-
limit the coherence, and visibility, of the nonlinear beam splitter traction of known technical noise and the second value was inferred
output to 5 Æcos Qæ 5 exp(2j2N,max /2N) 5 0.986 6 0.001, assum- without subtraction of known technical noise. The degradation of the
ing a Gaussian distribution of the phase fluctuations and the validity performance can be explained by the drift of the magnetic fields
of the two-mode approximation27 (Supplementary Information). (approximately 5 mG per day) during the measurement period
1167
©2010 Macmillan Publishers Limited. All rights reserved
LETTERS NATURE | Vol 464 | 22 April 2010

a b
Jz 12
Jx
Jy
3

χJz2

Number squeezing, ξN2 (dB)

Number squeezing, ξN2 (dB)

Ω Jx 0
–6
0.5
α

–9 20
–6

ξN2
80
–12
ΔJz 0 10 20 30 40 170
0
Rotation angle, α (º) 0.95 〈cos ϕ〉 1
–12
0 90 180 270 360
Rotation angle, α (º)

Figure 3 | Characterization of the quantum state within the nonlinear beam splitter. The dotted box indicates the region detailed in a. The largest
interferometer. a, One-axis-twisting dynamics of a coherent spin state fluctuations we measure have a number squeezing factor of
evolving under a pure xJz2 Hamiltonian. By analogy with the non-interacting j2N,max 5 110.3z0:3
{0:4 dB, resulting in an uncertainty product of the conjugate
Hamiltonian H/B 5 Dv0Jz, we expect eigenstates of Jz to rotate around the Jz variances 1.65 6 0.35 times larger than expected for a minimal-uncertainty
axis at a rate proportional to hH/hJz. In the interacting case, this gives a state. The black line is a fit to the data allowing for one free parameter in the
rotation rate proportional to Jz. Because a coherent spin state with two-mode approximation. Because the theory does not include the 15%
| ÆJzæ | , N/2 can be described as a superposition of several of these atom loss, it overestimates the optimal suppression of number fluctuations.
eigenstates, the twisting effect shown in the left-hand column appears. After Nevertheless, we find good agreement confirming the expected
a fixed evolution time of 18 ms, a coupling pulse rotates the quantum state interdependence between number squeezing and purely interaction-driven
around its centre and the fluctuations DJz are detected. The graph shows the phase dispersion. Knowledge of the minimal and maximal spin fluctuations
observed number squeezing factor, j2N , at the output of this nonlinear beam in two orthogonal directions allows for a statement on the many body
splitter as a function of rotation angle, a. The blue data have been corrected entanglement present in the system12. The inset shows theoretical limits for
for photon shot noise and the red data additionally take the technical noise j2N for different minimal non-separable block sizes of the N-particle density
into account (Supplementary Information). For clarity, 1-s.d. error bars are matrix (grey lines)16. These sizes equal the numbers of entangled atoms. The
given for the red data only. We observe a best number squeezing factor of red data point is the result of the noise tomography and indicates
j2N 5 28.2z0:9
{1:2 dB. b, Noise tomography of the output state of the nonlinear entanglement of more than 80 atoms with a 3-s.d. confidence level.

(24 h here but only 3 h in the measurement of the best number independently realized internal-state spin squeezing on an atom chip
squeezing). through controlled interactions using state-dependent potentials31.
With knowledge of the fluctuation in two conjugate spin direc-
Received 18 November 2009; accepted 4 February 2010.
tions, it is possible to make statements about entanglement in the Published online 31 March 2010.
spin system12,16,29. For distinguishable particles, entanglement is
defined as the non-separability of the overall density matrix. The 1. Gustavson, T. L., Bouyer, P. & Kasevich, M. A. Precision rotation measurements
with an atom interferometer gyroscope. Phys. Rev. Lett. 78, 2046–2049 (1997).
problem of indistinguishability of the atoms in a Bose–Einstein con- 2. Fixler, J., Foster, G., McGuirk, J. & Kasevich, M. Atom interferometer
densate can in principle be overcome by a local operation whereby measurement of the Newtonian constant of gravity. Science 315, 74–77 (2007).
the particles are localized to distinguishable positions in space with- 3. Santarelli, G. et al. Quantum projection noise in an atomic fountain: a high stability
out affecting the spin properties of the collective system16. Because cesium frequency standard. Phys. Rev. Lett. 82, 4619–4622 (1999).
4. Hald, J. Sørensen, J. L., Schori, C. & Polzik, E. S. Spin squeezed atoms: a
local measurements cannot generate entanglement, it must have been
macroscopic entangled ensemble created by light. Phys. Rev. Lett. 83, 1319–1322
present beforehand. The values of squeezed and antisqueezed fluc- (1999).
tuations in two orthogonal spin directions imply a lower bound for 5. Kuzmich, A., Mandel, L. & Bigelow, N. P. Generation of spin squeezing via
the block size of the largest non-separable part of the density matrix16. continuous quantum nondemolition measurement. Phys. Rev. Lett. 85, 1594–1597
The measured number squeezing, j2N , and coherence, Æcos Qæ, imply (2000).
6. Meyer, V. et al. Experimental demonstration of entanglement-enhanced rotation
the entanglement of 170 atoms (red data point in inset of Fig. 3b), angle estimation using trapped ions. Phys. Rev. Lett. 86, 5870–5873 (2001).
and we can exclude entanglement of fewer than 80 atoms with a 3-s.d. 7. Leibfried, D. et al. Toward Heisenberg-limited spectroscopy with multiparticle
confidence level. entangled states. Science 304, 1476–1478 (2004).
We have directly demonstrated the feasibility of nonlinear atom 8. Roos, C. F., Chwalla, M., Kim, K., Riebe, M. & Blatt, R. ‘Designer atoms’ for
quantum metrology. Nature 443, 316–319 (2006).
interferometry beyond the standard quantum limit using a macro- 9. Giovannetti, V., Lloyd, S. & Maccone, L. Quantum-enhanced measurements:
scopic ensemble of atoms. Precise characterization of the output state beating the standard quantum limit. Science 306, 1330–1336 (2004).
of the nonlinear beam splitter shows that coherent spin squeezing of 10. Kitagawa, M. & Ueda, M. Squeezed spin states. Phys. Rev. A 47, 5138–5143 (1993).
j2S 5 j2N /Æcos Qæ2 5 28.2 dB is obtainable. In principle (that is, if 11. Wineland, D., Bollinger, J., Itano, W. & Heinzen, D. Squeezed atomic states and
there is no excess technical noise), this allows for a 61% increase in projection noise in spectroscopy. Phys. Rev. A 50, 67–88 (1994).
12. Sørensen, A., Duan, L., Cirac, J. & Zoller, P. Many-particle entanglement with
phase sensitivity over classical linear interferometry. This is a signifi- Bose–Einstein condensates. Nature 409, 63–66 (2001).
cant step towards useful spin squeezing in atom interferometry, as to 13. Esteve, J., Gross, C., Weller, A., Giovanazzi, S. & Oberthaler, M. K. Squeezing and
our knowledge previously reported atomic coherent spin squeezing entanglement in a Bose–Einstein condensate. Nature 455, 1216–1219 (2008).
has been limited to 25.6 dB (refs 13–15, 26). The extension of many- 14. Leroux, I. D., Schleier-Smith, M. H. & Vuletić, V. Implementation of cavity
squeezing of a collective atomic spin. Phys. Rev. Lett. 104, 073602 (2010).
particle atom interferometry to the nonlinear regime is thus an 15. Appel, J. et al. Mesoscopic atomic entanglement for precision measurements
advance towards applied quantum atom optics using coherent inter- beyond the standard quantum limit. Proc. Natl Acad. Sci. USA 106, 10960–10965
actions between atoms. We note that the group of P. Treutlein has (2009).
1168
©2010 Macmillan Publishers Limited. All rights reserved
NATURE | Vol 464 | 22 April 2010 LETTERS

16. Sørensen, A. & Mølmer, K. Entanglement and extreme spin squeezing. Phys. Rev. 29. Tóth, G., Knapp, C., Gühne, O. & Briegel, H. J. Optimal spin squeezing
Lett. 86, 4431–4434 (2001). inequalities detect bound entanglement in spin models. Phys. Rev. Lett. 99,
17. Cronin, A. D., Schmiedmayer, J. & Pritchard, D. E. Optics and interferometry with 250405 (2007).
atoms and molecules. Rev. Mod. Phys. 81, 1051–1129 (2009). 30. Reinaudi, G., Lahaye, T., Wang, Z. & Guéry-Odelin, D. Strong saturation
18. Pezzé, L. & Smerzi, A. Entanglement, nonlinear dynamics, and the Heisenberg absorption imaging of dense clouds of ultracold atoms. Opt. Lett. 32, 3143–3145
limit. Phys. Rev. Lett. 102, 100401 (2009). (2007).
19. Esteve, J. et al. Observations of density fluctuations in an elongated Bose gas: ideal 31. Riedel, M. F. et al. Atom-chip-based generation of entanglement for quantum
gas and quasicondensate regimes. Phys. Rev. Lett. 96, 130403 (2006). metrology. Nature doi:10.1038/nature08988 (this issue); preprint at Æhttp://
20. Widera, A. et al. Entanglement interferometry for precision measurement of arxiv.org/abs/1003.1651æ (2010).
atomic scattering properties. Phys. Rev. Lett. 92, 160406 (2004).
21. Erhard, M., Schmaljohann, H., Kronjäger, J., Bongs, K. & Sengstock, K. Supplementary Information is linked to the online version of the paper at
Measurement of a mixed-spin-channel Feshbach resonance in 87Rb. Phys. Rev. A www.nature.com/nature.
69, 032705 (2004).
Acknowledgements We thank J.-P. Ronzheimer for technical assistance
22. Kaufman, A. M. et al. Radio-frequency dressing of multiple Feshbach resonances.
throughout the realization of this project and acknowledge discussions with Y. Li
Phys. Rev. A 80, 050701 (2009).
23. Sinatra, A. & Castin, Y. Binary mixtures of Bose-Einstein condensates: phase
and A. Sinatra. We gratefully acknowledge support from the Deutsche
dynamics and spatial dynamics. Eur. Phys. J. D 8, 319–332 (2000).
Forschungsgemeinschaft, the German-Israeli Foundation, the Heidelberg Center of
24. Li, Y., Treutlein, P., Reichel, J. & Sinatra, A. Spin squeezing in a bimodal condensate:
Quantum Dynamics, the ExtreMe Matter Institute and the European Commission
spatial dynamics and particle losses. Eur. Phys. J. B 68, 365–381 (2009).
Future and Emerging Technologies Open Scheme project MIDAS (Macroscopic
25. Steel, M. & Collett, M. Quantum state of two trapped Bose-Einstein condensates Interference Devices for Atomic and Solid-State Systems). C.G. acknowledges
with a Josephson coupling. Phys. Rev. A 57, 2920–2930 (1998). support from the Landesgraduiertenförderung Baden-Württemberg.
26. Fernholz, T. et al. Spin squeezing of atomic ensembles via nuclear-electronic spin Author Contributions All authors contributed extensively to the work presented in
entanglement. Phys. Rev. Lett. 101, 073601 (2008). this paper.
27. Ferrini, G., Spehner, D., Minguzzi, A. & Hekking, F. W. J. Anomalous decoherence
rate of macroscopic superpositions in Bose Josephson junctions. Preprint at Author Information Reprints and permissions information is available at
Æhttp://arxiv.org/abs/0911.0655æ (2009). www.nature.com/reprints. The authors declare no competing financial interests.
28. Li, Y., Castin, Y. & Sinatra, A. Optimum spin squeezing in Bose-Einstein Correspondence and requests for materials should be addressed to M.K.O.
condensates with particle losses. Phys. Rev. Lett. 100, 210401 (2008). (quantum.metrology@matterwave.de).

1169
©2010 Macmillan Publishers Limited. All rights reserved