R ES E A RC H | R E PO R TS

QUANTUM MECHANICS experimental fluctuations of the double-well bias,

and we experimentally find that the contrast and
damping improve with increasing J (18).
Two-particle quantum interference in We now consider the theoretical expectation
for an equivalent dynamical experiment starting

tunnel-coupled optical tweezers with two particles, one per well. For perfect cool-
ing and spin preparation of the isolated atoms,
the particles will be indistinguishable—that is, all
A. M. Kaufman,1,2 B. J. Lester,1,2 C. M. Reynolds,1,2 M. L. Wall,1,2 M. Foss-Feig,3 degrees of freedom besides their position (left or
K. R. A. Hazzard,1,2 A. M. Rey,1,2 C. A. Regal1,2* right) will have been made the same. We know
there is a particle in the left well and there is a par-
The quantum statistics of atoms is typically observed in the behavior of an ensemble via ticle in the right well, but we cannot associate any
macroscopic observables. However, quantum statistics modifies the behavior of even two additional label to the particles. The bosonic atoms
particles. Here, we demonstrate near-complete control over all the internal and external will then, necessarily, occupy the spatially sym-
degrees of freedom of two laser-cooled 87Rb atoms trapped in two optical tweezers. This metric state jS〉 ≡ p1ffiffi2 ðjL〉1 jR〉2 þ jR〉1 jL〉2 Þ, where
controllability allows us to observe signatures of indistinguishability via two-particle the ket subscript is a particle label. For poor
interference. Our work establishes laser-cooled atoms in optical tweezers as a promising cooling or spin preparation, the atoms can be
route to bottom-up engineering of scalable, low-entropy quantum systems. distinguished by a degree of freedom other than

Q
their position; hence, the atoms can antisymme-
uantum interference between possible de- low the quantum dynamics between initial and trize in the additional degree of freedom and, in
tection paths for two indistinguishable final states that are both known with single-site turn, have a projection onto the antisymmetric
spatial state jA〉 ≡ p1ffiffi2 ðjL〉1 jR〉2 − jR〉1 jL〉2 Þ. The bos-

Downloaded from http://science.sciencemag.org/ on August 9, 2018

particles yields information about quan- resolution.
tum statistics and correlations (1, 2). An In Fig. 2, A to C, we study the single-atom onic state can then be written as a mixture of the
example is the Hong-Ou-Mandel (HOM) tunneling dynamics by only considering experi- states jyT 〉 ¼ p1ffiffi2 ðjS〉jcþ 〉 T jA〉jc− 〉Þ, where jcT 〉 ¼
effect, which reveals bosonic quantum statistics ments that, after stochastic loading (14), yield a p1ffiffi ðjc〉1 jc〉2 T jc〉1 jc〉2 Þ and fc; cg describe the other
2
through a coalescence effect that causes two in- single atom in the left or right well in the first degree of freedom such as motional state fn; n′g
distinguishable photons incident on different image (fig. S1). After imaging and cooling, the or spin f↑; ↓g. Two atoms in either of the jyT 〉
ports of a beam splitter to emerge on the same, atom is in the 3D motional ground state and the states are distinguishable because the additional
yet random, output port (3). The HOM effect has jF ¼ 2; mF ¼ 2〉 ≡ j↑〉 spin state, where F and mF degree of freedom fc; cg is uniquely correlated
been observed with photons (3–5) and in an are the total angular momentum quantum num- with the atoms’ positions: For the jyþ 〉 (jy− 〉) state,
analogous experiment with electrons (6). ber and its projection along a quantization axis, the atom on the left is in state jc〉 (jc〉) and the atom
Here, we observe two-particle interference respectively. The tweezer depths and spacing on the right is in state jc〉 (jc〉). The ability to mea-
akin to the HOM effect with independently pre- (Fig. 2A) are then decreased rapidly (slowly) with sure indistinguishability arises from the different
pared bosonic atoms in tunnel-coupled optical respect to the tunnel coupling (trap frequency) dynamics exhibited in the symmetric and antisym-
tweezers (Fig. 1A). We control the single-atom to prepare an initial state ideally localized in metric cases. The symmetric spatial state dynam-
quantum state by laser cooling each atom to its the left or right well (fig. S2). The tunnel coupling ically evolves as jS〉 → jS〉cosð2JtÞ þ piffiffi2 ðjL〉1 jL〉2 þ
motional ground state and then directly observe is given by J ¼ −〈RjHsp jL〉ℏ, where jL〉 (jR〉) is the jR〉1 jR〉2 Þsinð2JtÞ. The antisymmetric state jA〉
the effects of their quantum indistinguishability. lowest-energy, localized state in the left (right) undergoes destructive interference that prevents
Whereas the role of quantum statistics in mac- well, and Hsp is the single-particle Hamiltonian the two atoms from being in the same well, and
roscopic ensembles of fermionic and bosonic (18). We control the double-well energy bias by hence displays no tunneling dynamics.
atoms can be observed with Hanbury Brown and varying the relative intensity of each tweezer. In Fig. 2D, we show the ideal dynamics for the
Twiss interference experiments (7–13), our sys- On the tunneling resonance, an atom prepared distinguishable (purple), symmetric (black), and
tem allows the study of nonclassical few-atom in the left well undergoes the coherent dynamics antisymmetric (brown) cases. We consider the
states with single-atom control. jL〉 → cosðJtÞjL〉 þ isinðJtÞjR〉 (19–22). After an observable P11 ðtÞ, which is the likelihood to mea-
Our results depend on the mobility of two evolution time t in the presence of tunneling, the sure the atoms in separate wells as a function
wavelength-scale optical tweezers and single- depth of the traps is rapidly increased to freeze of t and is analogous to looking at coincidence
site imaging, which are realized using the ap- the atom distribution, the traps are pulled apart, counts on a pair of photon detectors. The sym-
paratus illustrated in Fig. 1B (14–16). For laser and the single-atom location is imaged. metric jS〉 state yields unity contrast oscillations
cooling 87 Rb atoms to the three-dimensional Figure 2B demonstrates resonant coherent tun- of P11 ðtÞ, whereas jA〉 yields a time-independent
(3D) ground state (Fig. 1C) (16, 17) and imaging neling as measured by recording the likelihood P11 ðtÞ ¼ 1. Hence, the distinguishable states jyT 〉
in position-resolved potentials, our tweezers are of observing the atom in the left well (PL ) as a result in a P11 ðtÞ that is the average of these dy-
positioned far apart compared to the focused function of time for an atom starting in the left namics and, as such, does not attain a value below
spot radius of 710 nm. For tunneling, the tweezers well (blue) or the right well (red). The tunneling 0:5. The tunneling between the wells yields an
are brought close together such that there is a can be tuned by varying either the tweezer effective atom beam splitter where t varies the
small, tunable overlap of the single-particle wave spacing or the overall potential depth (fig. S3C). reflection and transmission coefficients. Of par-
functions. Our full experimental sequence con- In Fig. 2B, a fit to the data reveals J=2p ¼ 262ð4Þ ticular importance are the times tb ≡ 2p=8J and
sists of the following steps: We image the initial Hz; Fig. 2C shows data in which J is increased to odd-integer multiples thereof (green dashed lines).
atom positions, laser cool with Raman sideband 348ð4Þ Hz. One contribution to the subunity os- At t ¼ tb , the tunneling splits each atom equally
cooling, perform tunneling experiments, and cillation contrast is atom loss due to background between the wells and thereby realizes a balanced
then image the atoms again. Hence, we can fol- collisions; in the duration of our experiments, the atom beam splitter according to the transformations
loss probability (Ploss ) ranges from 0.03 to 0.05 jL〉 → p1ffiffi2 ðjL〉 þ ijR〉Þ and jR〉 → p1ffiffi2 ðjR〉 þ ijL〉.
1
JILA, National Institute of Standards and Technology and and is known precisely for each experiment (fig. Accordingly, for the jS〉 state, P11 ðtb Þ vanishes in
University of Colorado, Boulder, CO 80309, USA. S1, table S1, and Fig. 2, B and C, gray regions). We analogy to the indistinguishable photons incident
2
Department of Physics, University of Colorado, Boulder, CO also observe damping of the tunneling oscillations on separate ports of a balanced beam splitter in
80309, USA. 3Joint Quantum Institute and the National
Institute of Standards and Technology, Gaithersburg, MD
and finite initial contrast that is not accounted the original HOM experiment (3). However, dis-
20899, USA. for by particle loss (t ≈ 10 ms for J=2p ¼ 262 Hz) tinguishable atoms, like distinguishable photons,
*Corresponding author. E-mail: [email protected] (fig. S3B). These effects are most likely due to will yield P11 ¼ 0:5 when equally beam split.

306 18 JULY 2014 • VOL 345 ISSUE 6194 sciencemag.org SCIENCE

 RE S EAR CH | R E P O R T S

In Fig. 2E, we experimentally investigate the In analyzing the results of the two-particle values differ by 6s (18): APdist ¼ 0:282ð12Þ and
population dynamics observed with two particles. dynamics, our goal is to compare P11 ðtÞ from our AP11 ¼ 0:46ð2Þ.
We plot P11 ðtÞ for cases in which the stochastic two-particle measurement to that of a theoretical A full treatment of the observed P11 ðtÞ must
loading results in two atoms, one in each well (black expectation for uncorrelated, distinguishable also consider potential effects of interactions be-
squares); these points are taken in the same ex- atoms, which we refer to as Pdist ðtÞ. Pdist at any tween the atoms. In many experiments with atoms
perimental sequence as the single-particle data time can be calculated directly from correspond- in optical lattices, the on-site interaction energy
in Fig. 2B. In our atom detection protocol, we ing single-particle data via Pdist ¼ PL1 PR2 þ PR1 PL2 U is the dominant scale (19, 24); however, we in-
1ð2Þ
image scattered light from the two well-separated (purple circles in Fig. 2E) (18). Here, PL cor- tentionally operate in a regime where U is smaller
traps onto a charge-coupled device (CCD) array. responds to measuring an atom in the left well than J. For the data shown in Fig. 2E, U ¼
During the 25- to 50-ms imaging time, the atoms when an atom starts in the left (right) well, i.e., 0:44ð4ÞJ (18). In Fig. 2F, we demonstrate two-
1ð2Þ
are cooled by polarization gradient cooling, during the blue (red) data of Fig. 2B, and PR is the particle oscillations for experimental conditions
which light-assisted atomic collisions result in a corresponding information for measuring an atom of an even smaller relative interaction U ¼
signal corresponding to either zero or one atom in the right well. A calculation of Pdist directly from 0:22ð2ÞJ, with measurements AP11 ¼ 0:48ð2Þ and
(14, 22, 23). P11 is determined by the distinct sig- the single-particle points inherently contains APdist ¼ 0:306ð18Þ. The similarity of these results
nature in which the image indicates one atom in both loss and finite single-particle contrast. to those in Fig. 2E suggests that interactions are
each well. If the experiment yields two atoms in For example, Pdist ðtb Þ reaches a minimum value not a relevant scale in either experiment. A care-
one well, P20 or P02, this is manifest by final images consistent with ð1 − Ploss Þ2 =2 ≈ 0:5 − Ploss , and ful theoretical analysis also demonstrates that
that yield zero atoms, or in some cases one atom in the amplitude of Pdist ðtÞ is consistent with the existing interactions between distinguishable atoms
a single well (fig. S1) (18). To accurately interpret expectation of one-half the product of the single- could not mimic the observed signal of two-particle
P11 , we take into account signal depletion caused particle contrasts (18). We can compare the am- interference (18) (figs. S5 and S6).
by the single-particle loss described earlier (Ploss ). plitude of oscillation for the distinguishable We now study the behavior at the balanced
tunneling point tb , where, in analogy to HOM

Downloaded from http://science.sciencemag.org/ on August 9, 2018

This effect reduces the maximum value that can expectation (purple circles) to our two-particle
be achieved by the measured P11 to ð1 − Ploss Þ2 . measurement (black squares). We find that these experiments, the two-particle interference is most
readily observed. Using multiple experimental
knobs, we can vary the indistinguishability of the
atoms and observe a variation in P11 ðtb Þ ¼ Pb .
We start by studying the dependence of Pb on the
relative spin state of the two atoms using two
distinct methods. In the first method, after cool-
ing the atoms, we apply a variable-length micro-
wave pulse that couples the j↑〉 and j↓〉 spin states
in only the right well (fig. S4). This is accom-
plished by shifting the transition in the left well
out of resonance using a circularly polarized,
tightly focused laser spot (18). Upon p rotation of
the right spin, the atoms become distinguishable,
and we expect the two-particle interference dip
of P11 below Pdist to disappear. The observed de-
pendence on the microwave pulse area is shown
in Fig. 3A; for comparison, we show that Pdist
(purple circles), calculated from the single-particle
measurements, remains constant and near
ð1 − Ploss Þ2 =2. We study multiple spin rota-
tions to show that the interference is recovered
after a 2p rotation. We find that the frequency of
oscillation is 32:6ð6ÞkHz, which is in agreement
with the measured microwave Rabi frequency of
32:05ð18ÞkHz. The displayed fit to the data deter-
mines Pbmin ¼ 0:314ð14Þ, and the amplitude of the
variation in Pb is 0:15ð2Þ. Taking into account the
spin rotation fidelity, we expect an amplitude of
0:84ðð1 − Ploss Þ2=2 − Pbmin Þ ¼ 0:130ð13Þ (18), which
is consistent with the measured value. The agree-
ment further substantiates the irrelevance of
interactions, which are nearly constant as the
two-particle spin state is varied (25), and hence
not responsible for the observed variation in Pb .
Fig. 1. Quantum interference of two particles. (A) The initial two-particle state is a ground state spin- In the second spin study, we couple the j↑〉 and
up atom in each optical tweezer, denoted by jS〉 ≡ p1ffiffi2 ðjL〉1jR〉2 þ jR〉1jL〉2Þ, where the ket subscript is a j↓〉 spin states of atoms in both wells using a pair
particle label, and the single-particle states jL〉 (jR〉) correspond to an atom localized in the left (right) well. of Raman beams. This global rotation avoids any
For a tunneling time that realizes a balanced atom beam splitter, destructive interference yields the state systematic effects that might be introduced by
piffiffi ðjL〉 jL〉 þ jR〉 jR〉 Þ. (B) Experimental setup. The apparatus for realizing tunneling between optical single-site addressing. During the time (25 ms)
2 1 2 1 2
tweezers uses high numerical aperture optics combined with control of the tweezers’ positions and depths between the Raman pulse and the tunneling, the
with acousto-optic modulators (AOMs). The same objective that creates the focused tweezer potentials atoms lose their spin coherence, and hence the
also collects 780 nm fluorescence from the optically trapped atoms. (C) The sideband cooling is spin state of each atom is in an incoherent mix-
accomplished with lasers driving coherent (green) and spontaneous (blue) Raman transitions that couple ture. Ideally, odd-integer multiples of a p=2-pulse
to the atomic motion and spin states jF ¼ 1; mF ¼ 1〉 ≡ j↓〉 and jF ¼ 2; mF ¼ 2〉 ≡ j↑〉. yield an equal mixture of all possible two-atom

SCIENCE sciencemag.org 18 JULY 2014 • VOL 345 ISSUE 6194 307

R ES E A RC H | R E PO R TS

Fig. 2. Single- and two-particle tunneling. (A) While the tweezers are dynamics initiated at t ¼ 0 and in the symmetric state jS〉, the distinguish-
23ð1Þ MHz deep, the atom is imaged, cooled and optically pumped to j↑〉. For able states jyT〉, and the antisymmetric state jA〉. The dashed green lines
tunneling experiments, the tweezers are swept together such that the two correspond to tb. (E) Measured two-particle dynamics during the same experimen-

Downloaded from http://science.sciencemag.org/ on August 9, 2018

Gaussian functions (double-well minima) are centered ≈800 nm (≈600 nm) tal sequence as in (B). Likelihood to measure exactly one atom in each well (P11) for
apart, and the total trap intensity is decreased to a single-well depth of either the initial condition of an atom prepared in each well (black squares). Distinguishable
96 kHz or 60 kHz. (B) Resonant tunneling oscillations at 2J for an 808-nm expectation Pdist calculated from the single-particle data in (B) (purple circles).The
Gaussian function spacing and a 96-kHz depth. Blue circles (red triangles) gray shaded region above the dashed black line indicates the expected reduction
are the expectation value PL1ð2Þ for finding an atom in the left well given an from atom loss. (F) Same as (E) except with a larger value of J=U using the double-
initial single atom in the left (right) well. The gray shaded region indicates well parameters of (C). In all plots, the shaded regions are the 95% confidence
atom loss Ploss. (C) Same as B except with an 805-nm Gaussian function interval for a sinusoidal fit.The error bars are the standard error in the measurement;
spacing and a depth of 60 kHz. (D) Ideal P11ðtÞ for two-particle tunneling each black (red or blue) data point is the mean of ≈140 (100) atom measurements.

spin states, and hence the two-particle interfer-

ence dip should be reduced by 1=2. As a function
of Raman pulse area (Fig. 3B), the frequency of
oscillation is 65:5T1:2 kHz, which, as expected, is
twice the measured Raman Rabi frequency of
32:3ð3Þ kHz. The displayed fit to the data deter-
mines Pbmin ¼ 0:296ð10Þ, and the amplitude of the
variation in Pb is 0:085ð15Þ. Taking into account
the relative spin rotation fidelity of 0:90ð3Þ, we
expect an amplitude of ð0:902 =2Þ½ð1 − Ploss Þ2 =2 −
Pbmin
¼ 0:069ð6Þ, which is consistent with the
measured amplitude.
Last, we study the dependence of Pb on the Fig. 3. Controlling two-particle distinguishability.
motional state of the atoms in Fig. 3C. During the In all plots, the black squares are P11ðtbÞ = Pb, the
last stage of cooling, we vary the frequency dCool purple circles are the expectation for distinguish-
of one Raman beam that controls the cooling able particles calculated directly from the single-
along the weak axis (z) of both tweezer wells. atom tunneling [PdistðtbÞ], and the dashed black
For a separable potential, motional excitation along line marks ð1 − PlossÞ2=2. (A) Before tunneling, we
this axis would leave the single-particle tunnel- apply a microwave drive that couples j↑〉 and j↓〉
ing unaffected. For our nonseparable tweezer for one of the atoms in a two-particle experiment.
potential, we expect and observe some variation in In the trap where J=2π ¼ 348 Hz, the tunneling
the tunneling (18), but near tb the single-particle time is fixed at t ¼ 0:99 ms (second realization of
tunneling still results in a relatively constant Pdist tb). (B) Before tunneling, we apply a global co-
(purple circles), which is consistent with the distin- herent drive of varied pulse area to couple j↑〉 and
guishable expectation ð1 − Ploss Þ2 =2 ¼ 0:4660ð14Þ. j↓〉 and then allow for decoherence. In the trap where J=2π ¼ 262 Hz, the tunneling time is fixed at
For the two-particle measurements, at the pri- t ¼ 0:45 ms. In (A) and (B), the solid line and shaded band are sinusoidal fits and the associated 95% confidence
mary sideband cooling resonance (dCool ¼ 0), we interval. (C) We vary the detuning (dCool) of the cooling beams of motion along the z axis. In the trap where
observe a dip to Pbmin ¼ 0:28ð2Þ, which is a value J=2π ¼ 262 Hz, the tunneling time is fixed at t ¼ 0:45 ms. The two shaded regions correspond to frequency
below the distinguishable expectation. ranges of efficient (1st sideband) and less efficient (2nd sideband) cooling. For all plots, each black data point is the
In all of our measurements, the value of Pbmin is average of ≈360 measurements, and each set of measurements corresponding to a purple point is the average of
observed to be finite, and it is useful to consider the ≈240 measurements. All error bars are the standard error in the measurement.
origin of imperfections that could lead to the ob-
served value. If residual atom temperature along be 85% (16, 18)] would correspond to Pbmin ¼ 0:12. We have demonstrated an experimental system
the weak z axis were the only contributing factor, Hence, it is likely that the finite Pbmin is further with which we achieve quantum control over the
the central value of our single-atom ground-state enlarged by technical fluctuations similar to those motion, position, and spin of single neutral atoms.
fraction [measured by sideband spectroscopy to that lead to our finite single-particle contrast. Through a two-particle interference experiment,

308 18 JULY 2014 • VOL 345 ISSUE 6194 sciencemag.org SCIENCE

 RE S EAR CH | R E P O R T S

we have shown that it is possible to create a low- OPTICAL METROLOGY

entropy bosonic state by individually placing
atoms in their motional ground state with laser
cooling (26–28). Our results lay a foundation for
linear quantum computing with atoms (29), inter-
Electro-optical frequency division
ferometric highly sensitive force detection (30),
control of neutral atoms in nanoscale optical de-
and stable microwave synthesis
vices (31, 32), and quantum simulation with laser-
Jiang Li,1* Xu Yi,1* Hansuek Lee,1 Scott A. Diddams,2 Kerry J. Vahala1†
cooled atoms in scalable optical tweezer arrays.
Optical frequency division by using frequency combs has revolutionized time keeping
RE FE RENCES AND N OT ES and the generation of stable microwave signals. We demonstrate optical frequency division
1. R. J. Glauber, Phys. Rev. 130, 2529–2539 (1963). and microwave generation by using a tunable electrical oscillator to create dual combs
2. R. Hanbury Brown, R. Q. Twiss, Nature 177, 27 (1956).
3. C. K. Hong, Z. Y. Ou, L. Mandel, Phys. Rev. Lett. 59,
through phase modulation of two optical signals that have a stable difference frequency.
2044–2046 (1987). Phase-locked control of the electrical oscillator by means of optical frequency division
4. J. Beugnon et al., Nature 440, 779–782 (2006). produces stable microwaves. Our approach transposes the oscillator and frequency reference
5. C. Lang et al., Nat. Phys. 9, 345–348 (2013). of a conventional microwave frequency synthesizer. In this way, the oscillator experiences
6. E. Bocquillon et al., Science 339, 1054–1057 (2013).
7. M. Yasuda, F. Shimizu, Phys. Rev. Lett. 77, 3090–3093
large phase noise reduction relative to the frequency reference. The electro-optical approach
(1996). additionally relaxes the need for highly linear photodetection of the comb mode spacing.
8. S. Fölling et al., Nature 434, 481–484 (2005). As well as simplicity, the technique is also tunable and scalable to higher division ratios.

T
9. M. Schellekens et al., Science 310, 648–651 (2005).
10. A. Öttl, S. Ritter, M. Köhl, T. Esslinger, Phys. Rev. Lett. 95, he photomixing of two highly coherent ference for the microwave source (Fig. 1A). These
090404 (2005).

Downloaded from http://science.sciencemag.org/ on August 9, 2018

11. T. Rom et al., Nature 444, 733–736 (2006).
laser signals is a well-known approach to laser lines are produced by Brillouin oscillation
12. T. Jeltes et al., Nature 445, 402–405 (2007). generate a stable radio frequency (RF) or in a single high-quality-factor (Q) microcavity. How-
13. S. S. Hodgman, R. G. Dall, A. G. Manning, K. G. H. Baldwin, microwave signal (1, 2). Recently, however, ever, the lines could also result from any stable
A. G. Truscott, Science 331, 1046–1049 (2011). a different approach to all-optical signal optical references, including various types of
14. N. Schlosser, G. Reymond, I. Protsenko, P. Grangier, Nature
411, 1024–1027 (2001).
generation has been demonstrated that may dual-mode lasers (10–15), two lasers locked to
15. F. Serwane et al., Science 332, 336–338 (2011). revolutionize applications that require high- distinct optical modes of a reference cavity (16),
16. A. M. Kaufman, B. J. Lester, C. A. Regal, Phys. Rev. X 2, 041014 stability microwaves. Rather than photomixing or lasers stabilized to atomic transitions (17). The
(2012). stabilized laser signals to directly produce a mi- laser lines enter the frequency divider portion
17. C. Monroe et al., Phys. Rev. Lett. 75, 4011–4014 (1995).
18. Materials and methods are available as supplementary
crowave signal, the approach uses an octave- of the signal generator, where they are phase-
materials on Science Online. spanning, self-referenced frequency comb to modulated by a pair of modulators at a frequency
19. S. Fölling et al., Nature 448, 1029–1032 (2007). divide a stable optical reference frequency down set by a voltage-controlled electrical oscillator
20. M. Anderlini et al., Nature 448, 452–456 (2007). to microwave or RF rates (3). Frequency dividers (VCO). The sideband spectrum created by the
21. J. Sebby-Strabley et al., Phys. Rev. Lett. 98, 200405
(2007).
are widely used in electronics to generate new fre- phase modulators is further broadened through
22. C. Weitenberg et al., Nature 471, 319–324 (2011). quencies from a single base oscillator or to coher- pulse-forming and self-phase modulation in an
23. W. S. Bakr, J. I. Gillen, A. Peng, S. Fölling, M. Greiner, Nature ently link different frequency bands. As an ancillary optical fiber (8, 18, 19). The comb of lines extend-
462, 74–77 (2009). benefit, all frequency dividers reduce the phase ing from each laser line results in a pair of side-
24. R. Ma et al., Phys. Rev. Lett. 107, 095301 (2011).
25. M. Egorov et al., Phys. Rev. A 87, 053614 (2013).
noise spectral density of the output signal relative bands near the midpoint of the frequency span.
26. V. Vuletić, C. Chin, A. J. Kerman, S. Chu, Phys. Rev. Lett. 81, to the input by the square of the division ratio. These are optically filtered and detected. The de-
5768–5771 (1998). The new optical frequency dividers perform divi- tected beat-note signal contains the phase noise
27. D. S. Weiss et al., Phys. Rev. A 70, 040302 (2004). sion by a factor of ∼50,000 (the ratio of optical to of the VCO, but magnified by the optical division
28. A. Steffen et al., Proc. Natl. Acad. Sci. U.S.A. 109, 9770–9774
(2012).
microwave frequencies), so that phase-noise reduc- factor. It therefore provides a suitable error sig-
29. P. Kok, K. Nemoto, T. C. Ralph, J. P. Dowling, G. J. Milburn, tion is greater than 109. Moreover, reference-cavity nal for phase-lock loop control of the VCO.
Rev. Mod. Phys. 79, 135–174 (2007). stabilized lasers exhibit a superior, fractional fre- We contrast this microwave source that is
30. V. Giovannetti, S. Lloyd, L. Maccone, Science 306, quency stability in comparison with electrical based on electro-optical frequency division (EOFD)
1330–1336 (2004).
oscillators (4, 5). Optical dividers applied to such with a conventional microwave source that is
31. E. Vetsch et al., Phys. Rev. Lett. 104, 203603 (2010).
32. J. D. Thompson et al., Science 340, 1202–1205 (2013). signals thereby generate microwave signals with based on electrical frequency division of a VCO
an exceedingly low phase noise level (3). (20). In the conventional approach (Fig. 1B),
ACKN OW LEDG MEN TS We present a way to generate high-performance the VCO provides the highest frequency in the
We thank T. P. Purdy for useful discussions. This work was supported microwave signals through optical frequency di- system. It is stabilized through electrical frequen-
by the David and Lucile Packard Foundation and the National Science
Foundation (NSF) under grant no. 1125844. C.A.R. acknowledges
vision (OFD) by using a cascade of direct phase cy division and phase comparison with a lower-
support from the Clare Boothe Luce Foundation, A.M.K. and C.M.R. modulation and self-phase modulation to create frequency reference oscillator, such as a quartz
from the National Defense Science and Engineering Graduate an optical comb (6–9). Because the spectral line oscillator. A consequence is that the stabilized
Fellowships, and B.J.L. from the NSF Graduate Research Fellowship spacing is set by the electrical oscillator used to VCO has a phase noise level that is always higher
Program. K.R.A.H., M.L.W., and A.M.R. acknowledge funding from
NSF-Physics at the Information Frontier, Army Research Office
drive the phase modulators (as opposed to an than the reference oscillator phase noise by the
and Air Force Office of Scientific Research. K.R.A.H. and M.F.-F. optical resonator), the method of microwave syn- square of their frequency ratio (the frequency
acknowledge support from a National Research Council thesis has similarities to conventional micro- division ratio). In contrast, our optical version
postdoctoral fellowship. wave synthesizers while also leveraging the power reverses the positions of the reference and the
of OFD so as to reduce phase noise. VCO in the frequency domain. Specifically, the
SUPPLEMENTARY MATERIALS In our approach, two laser lines having good reference is provided by the frequency difference
www.sciencemag.org/content/345/6194/306/suppl/DC1 relative frequency stability provide an optical re- of the laser lines, and this frequency difference
Materials and Methods is made much greater than the frequency of the
Figs. S1 to S6 1
Table S1
T. J. Watson Laboratory of Applied Physics, California VCO (in the present implementation, this is a
Institute of Technology, Pasadena, CA 91125, USA. 2Time nondetectable rate set at ~150 times the VCO
References (33–36)
and Frequency Division, National Institute of Standards and
frequency). Moreover, this reference frequency
22 December 2013; accepted 6 June 2014 Technology, Boulder, CO 80305, USA.
Published online 26 June 2014; *These authors contributed equally to this work. †Corresponding is divided down to the VCO frequency, as op-
10.1126/science.1250057 author. E-mail: [email protected] posed to dividing the VCO frequency down to

SCIENCE sciencemag.org 18 JULY 2014 • VOL 345 ISSUE 6194 309

Two-particle quantum interference in tunnel-coupled optical tweezers
A. M. Kaufman, B. J. Lester, C. M. Reynolds, M. L. Wall, M. Foss-Feig, K. R. A. Hazzard, A. M. Rey and C. A. Regal

Science 345 (6194), 306-309.

DOI: 10.1126/science.1250057originally published online June 26, 2014

Bosons of a feather flit together

Bosons are a type of particle that likes to congregate. This property has a major effect on the behavior of identical
bosons. Kaufman et al. demonstrated quantum interference of two bosonic Rb atoms placed in two neighboring quantum
wells (see the Perspective by Thompson and Lukin). They prepared the atoms in exactly the same state so that there
would be no way to tell them apart except for which well each atom was in. They then monitored the probability of the two

Downloaded from http://science.sciencemag.org/ on August 9, 2018

atoms still being in separate wells. At certain times, the probability had a characteristic dip signifying that the bosons
preferred to be in the same well.
Science, this issue p. 306; see also p. 272

ARTICLE TOOLS http://science.sciencemag.org/content/345/6194/306

SUPPLEMENTARY http://science.sciencemag.org/content/suppl/2014/06/25/science.1250057.DC1
MATERIALS

RELATED http://science.sciencemag.org/content/sci/345/6194/272.full
CONTENT

REFERENCES This article cites 35 articles, 7 of which you can access for free
http://science.sciencemag.org/content/345/6194/306#BIBL

PERMISSIONS http://www.sciencemag.org/help/reprints-and-permissions

Use of this article is subject to the Terms of Service

Science (print ISSN 0036-8075; online ISSN 1095-9203) is published by the American Association for the Advancement of
Science, 1200 New York Avenue NW, Washington, DC 20005. 2017 © The Authors, some rights reserved; exclusive
licensee American Association for the Advancement of Science. No claim to original U.S. Government Works. The title
Science is a registered trademark of AAAS.