Optik 26

Spatial correlation diagnostics for atoms in
optical lattices
J. Grondalski∗,† , P. M. Alsing† , and

I. H. Deutsch∗
Department of Physics and Astronomy∗ ,
Albuquerque High Performance Computing Center† ,
University of New Mexico, Albuquerque, NM
[email protected]
http://panda30.phys.unm.edu/Deutsch/Homepage.html
Abstract: We explore the use of first and second order same-time

atomic spatial correlation functions as a diagnostic for probing the small
scale spatial structure of atomic samples trapped in optical lattices. As-
suming an ensemble of equivalent atoms, properties of the local wave
function at a given lattice site can be measured using same-position
first-order correlations. Statistics of atomic distributions over the lat-
tice can be measured via two-point correlations, generally requiring the
averaging of multiple realizations of statistically similar but distinct
realizations in order to obtain sufficient signal to noise. Whereas two-
point first order correlations are fragile due to phase fluctuations from
shot-to-shot in the ensemble, second order correlations are robust. We
perform numerical simulations to demonstrate these diagnostic tools.
c 1999 Optical Society of America
OCIS codes: (020.0020) Atomic and molecular physics; (030.0030) Coherence and
statistical optics; (110.0110) Imaging systems; (140.3320) Laser cooling
References and links

1. P. S. Jessen and I. H. Deutsch, “Optical Lattices,” Adv. At. Mol. Opt. Phys. 37, 95-138 (1996).
2. Ivan H. Deutsch and Poul S. Jessen, “Quantum-state control in optical lattices,” Phys. Rev. A
57, 1972-1986 (1998).
3. Qian Niu, Xian-Gen Zhao, G. A. Georgakis, and M. G. Raizen, “Atomic Landau-Zener Tunneling
and Wannier Stark Ladders in Opical Potenitals,” Phys. Rev. Lett. 76, 4504-4507 (1996).
4. S. K. Dutta, B. K. Teo, and G. Raithel, “Tunneling Dynamics and Guage Potentials in Optical
Lattices,” Phys. Rev. Lett. 83, 1093-1936 (1999).
5. Gavin K. Brennen, Cartlon M. Caves, Poul S. Jessen, and Ivan H. Deutsch, “Quantum Logic
Gates in Optical Lattices,” Phys. Rev. Lett. 82, 1060-1063 (1999).
6. D. Jaksch, H.-J. Briegel, J. I. Cirac, C. W. Gardiner, and P. Zoller, “Entanglement of Atoms via
Cold Controlled Collisions,” Phys. Rev. Lett. 82, 1975-1978 (1999).
7. Anders Sorensen and Klaus Molmer, “Spin-Spin Interactions and Spin Squeezing in an Opical
Lattice,” Phys. Rev. Lett. 83, 2274-2277 (1999).
8. S. Lukman Winoto, Marshall T. DePue, Nathan E. Bramall, and David S. Weiss, “Laser cooling
at high density in deep far-detuned optical lattices,” Phys. Rev. A 59, R19-R22 (1999).
9. B. P. Anderson and M. A. Kasevich, “Macroscopic Quantum Interference from Atomic Tunnel
Arrays,” Science 282, 1686-1689 (1998).
10. Dai-Il Choi and Qian Niu, “Bose-Einstein Condensates in an Optical Lattice,” Phys. Rev. Lett.
82, 2022-2025 (1999).
11. Kirstine Berg-Sørenson and Klaus Mølmer, “Bose-Einstein condensates in spatially periodic po-
tentials,” Phys. Rev. A 58, 1480-1484 (1998).
12. E. V. Goldstein, P. Pax, and P. Meystre, “Dipole-dipole in three-dimensional optical lattices,”
Phys. Rev. A 53, 2604-2615 (1996).
13. D. Jaksch, C. Bruder, J. I. Cirac, C. W. Gardiner, and P. Zoller, “Cold Bosonic Atoms in Optical
Lattices,” Phys. Rev. Lett. 81, 3108-3111 (1998).
14. Klaus Drese and Martin Holthaus, ”Exploring a Metal-Insulator Transition with Ultrcold Atoms
in Standing Light Waves,” Phys. Rev. Lett. 2932, 2932-2935 (1997).
#12687 - $15.00 US Received September 08, 1999; Revised November 08, 1999
(C) 1999 OSA 22 November 1999 / Vol. 5, No. 11 / OPTICS EXPRESS 249
15. M. H. Anderson, J. R. Ensher, M. R. Matthews, C. E. Wieman, and E. A. Cornell, “Observation
of Bose-Einstein Condensation in a Dilute Atomic Vapor,” Science 269, 198-201 (1995).
16. J. E. Thomas and L. J. Wang, “Quantum theory of correlated-atomic-position measurements by
resonance imaging,” Phys. Rev. A 49, 558-569 (1994).
17. K. B. Davis, M. O. Mewes, M. R. Andrews, N. J. van Druten, D. M. Kurn, and W. Ketterle, “Bose-
Einstein Condensation in a Gas of Sodium Atoms,” Phys. Rev. Lett. 75, 3969-3973 (1995).
18. M. R. Andrews, M.-O. Mewes, N. J. van Druten, D. S. Durfee, D. M. Kurn, W. Ketterle, “Direct,
Nondestructive Observation of a Bose Condensate,” Science 273, 84-87 (1996)
19. M. R. Andrews, C. G. Townsend, H.-J. Miesner, D. S. Durfee, D. M. Kurn, W. Ketterle, “Obser-
vation of Interference Between Two Bose Condensates,” Science 275, 637-641 (1997).
20. G. Birkl, M. Gatzke, I. H. Deutsch, S. L. Rolston, and W. D. Phillips, “Bragg Scattering from
Atoms in Optical Lattices,” Phys. Rev. Lett. 75, 2823-2826 (1995).
21. Hideyuki Kunugita, Tetsuya Ido, and Fujio Shimizu, “Ionizing Collisional Rate of Metastable
Rare-Gas Atoms in an Optical Lattice,” Phys. Rev. Lett. 79, 621-624 (1997).
22. C. Orzel, M. Walhout, U. Sterr, P. S. Julienne, and S. L. Rolston, “Spin polarization and quantum-
statistical effects in ultracold ionizing collisions,” Phys. Rev. A 59, 1926-1935 (1999).
23. P. D. Lett, R. N. Watts, C. I. Westbrook, W. D. Phillips, P. L. Gould, and H. J. Metcalf, “Ob-
servation of atoms laser cooled below the Doppler limit,” Phys. Rev. Lett. 61, 169-172 (1988).
24. Benjamin Chu, Laser Light Scattering, Second Edition (Academic Press, San Diego, 1991).
25. Masami Yasuda and Fujio Shimizu, “Observation of Two-Atom Correlation of an Ultracold Neon
Atomic Beam,” Phys. Rev. Lett. 77, 3090-3093 (1996).
26. M. Henny, S. Oberholzer, C. Strunk, T. Heinzel, K. Ensslin, M. Holland, C. Schönenberger, “The
Fermionic Hanbury Brown and Twiss Experiment,” Science 284, 296-298 (1999).
27. William D. Oliver, Jungsang Kim, Robert C. Liu, Yoshihisa Yamamoto, “Hanbury Brown and
Twiss-Type Experiment with Electrons,” Science 284, 299-301 (1999).
28. Claude Cohen-Tannoudji, Bernard Diu, Franck Laloë, Quantum Mechanics Vol. 1 (John Wiley &
Sons, New York, 1977).
29. S. E. Hamann, D. L. Haycock, G. Klose, P. H. Pax, I. H. Deutsch, and P. S. Jessen, “Resolved-
Sideband Raman Cooling to the Ground State of an Optical Lattice,” Phys. Rev. Lett. 80,
4149-4152 (1998).
30. Joseph W. Goodman, Statistical Optics (John Wiley & Sons, New York, 1985).
31. Tomographic TOF techniques exist which do not lose this phase information.
U. Janicke and M. Wilkens, “Tomography of atom beams,” J. Mod. Opt. 42, 2183-2199 (1995).
Ch. Kurtsiefer, T. Pfau, and J. Mlynek, “Measurement of the Wigner function of an ensemble of
helium atoms,” Nature 386, 150-153 (1997).
32. B. Saubaméa, T. W. Hijmans, S. Kulin, E. Rasel, E. Peik, M. Leduc, and C. Cohen-Tannoudji,
“Direct Measurement of the Spatial Correlation Function of Ultracold Atoms,” Phys. Rev. Lett.
79, 3146-3149 (1997).
33. E. V. Goldstein, O. Zobay, and P. Meystre, “Coherence of atom matter-wave fields,” Phys. Rev.
A 58, 2373-2384 (1998).
34. Rodney Loudon, The Quantum Theory of Light, Second Edition (Oxford University Press, New
York, 1983).
1 Introduction
Optical lattices, periodic arrays of microscopic potentials induced by the ac Stark ef-
fect of interfering laser beams, can be used to trap ultra-cold atoms [1]. This system
has found application in coherent control of atomic wave packets [2] including atomic
tunneling [3,4], and proposals for quantum computing [5,6] and quantum simulation
[7]. Furthermore, the optical lattice provides a clean environment for studies of many-
body effects in a periodic potential as experimenters achieve ever higher atomic densities
through special cooling techniques [8] and by loading Bose-Einstein condensates (BECs)
into lattices [9]. When densities are high, atom-atom interactions may lead to the for-
mation of small scale structure [10-12] and/or nonclassical atomic-occupation statistics
within each well as is seen in a conductor-insulator phase transition [13, 14].
Spatial distributions of cold atomic gases are usually probed directly by absorption
spectroscopy [15], near-resonance fluorescence spectroscopy [16], or off-resonance spec-
troscopy [17,18]. The resolution of these imaging techniques is fundamentally limited by
the wavelength of the external probe laser. In addition, the photons used to probe the
atomic distribution impart on average an energy of (h̄k)2 /(2M ) (h̄k is the photon mo-
mentum and M is the mass of the atom). For cold atomic samples, these “recoil kicks”
generally heat the sample very quickly, although in the case of off-resonance spectroscopy
this heating has been suppressed by a factor of order 100, allowing for multiple images
before the sample is destroyed [18]. These imaging techniques typically integrate the
signal along one dimension of the atomic cloud thus measuring column densities. Some
experiments have achieved three-dimensional resolution through tomographic methods
[19]. Spatial information can also be inferred by measuring Bragg reflection from an
atomic sample [20] or atomic collision rates [21,22].
Given the range of new applications proposed for optical lattices, we consider di-
agnostic tools appropriate for experimental investigations which do not have the lim-
itations of many the techniques described above. In this article we consider Time of
Flight (TOF) imaging [23] whereby atoms initially trapped in an optical lattice are
released suddenly and are allowed to freely expand for a time t, after which they are
counted with point-like detectors, such as a microchannel plate array (MCP) in the case
of meta-stable noble gas atoms [21,22]. Recoil heating is eliminated because no external
fields are used and additional restrictions on the resolution imposed by a probe laser
no longer apply. We consider same-time spatial correlations of order one and two in the
detection plane and find Fourier relations between these functions and the initial atomic
distribution. These relations are completely separable in three dimensions allowing for
the possibility of 3D resolution.
The remainder of this article is organized as follows. In Sec. 2.1 we review the re-
lationship between TOF measurements and the wave function of a trapped atom. We
show that for an atomic sample of equivalent atoms, properties of the local wave function
can be deduced. Such a diagnostic is useful for studying coherent wave packet motion
in, e.g., a double well potential. In the next two sections we examine ways to measure
the distribution of atoms throughout the lattice (i. e. the small scale structure) with
two point correlation functions. Such a diagnostic would be useful for, e.g., quantum
computing where interactions are induced between neighboring atoms. In Sec. 2.2 we
consider a Young’s double slit experiment and establish the relationship between the
complex coherence factor and the spatial distribution of the atomic sample. We show
that the mean position of the atom is mapped onto the phase of the complex coherence
factor, a quantity susceptible to shot-to-shot fluctuations in the sample. In Sec. 2.3 we
consider the possibility of deducing spatial information from the next higher correlation
function which is shown to be a much more robust diagnostic. Second order correlations
of scattered laser light have been used successfully to measure the size and shape of
macromolecules in solution as well as in colloidal suspensions [24]. In contrast to the
photon coincidences measured in these experiments, we propose the use of atom coin-
cidence counts, similar to recent experiments that employed time-delayed coincidence
counting to analyze degenerate Bose [25] and Fermi [26,27] gases. There exists a Fourier
relationship between the second order correlation function at the detector plane and the
probability for a pair of atoms to have a certain separation, independent of the pair’s
absolute location in the lattice. In Sec. 3 we present the results of numerical simulations
of ensemble averaged TOF experiments. Averaging washes out information about the
mean positions of atoms contained in the first order correlation function but important
information is retained in second order. Finally, in Sec. 4 we summarize our results.
2 Spatial Correlation Functions
2.1 Local Wave Function via Atomic Density Measurements
The normalized first order correlation function at a single position, g (1) (x, x), is a mea-
sure of the atomic density n(x). Such correlations can be measured using the well known
time of flight (TOF) technique in which a trapped atomic sample is “suddenly” released
from an optical lattice in a time short compared to the characteristic period of oscillation
of a trapped atom. The atoms then expand ballistically in free space until they reach the
detection plane. The arrival times of the atoms are measured and the initial momentum
distribution of the atoms is inferred [23]. A quantum mechanical propagator formalism
[28], similar to the diffraction theory of physical optics, can be applied to obtain the
wave function, Ψ(x, t), at the detection plane from the initial wave function, Ψ(x0 , t0 ),
with unprimed variables (x, t) denoting space-time coordinates of the detection plane
and the primed variables (x0 , t0 ) denoting the optical lattice plane,
Z
Ψ(x, t) = dx0 K(x, t; x0 t0 ) Ψ(x0 , t0 ). (1)
For free space, the propagator K(x, t; x0 t0 ) is given by

√
0 0 −i x − x0 2
K(x, t; x t ) = exp [iπ( ) ], (2)
L L
where L is an effective length given by, L2 = h(t − t0 )/M , which is a measure of the
time evolved between the initial and final state. Consider, for example, a Gaussian with
initial width σ 0 , located at x0i . An atom allowed to freely expand for a sufficiently long
time after the trapping lasers are turned off can be considered to be in the far field and
the wave function at a later time, Ψ(x, t), can be shown to be related to the initial wave
function through a Fourier transform,
1 1 −iπ( x0 −x0i )2 −( x0 −x0i )2

Ψ(x, t) ' ( )4 e L e 2σ . (3)
2πσ 2
Here σ = L2 /(4πσ 0 ) is the width of the Gaussian in the detection plane.
Now consider an atom initially localized at lattice site i in the state, Φ(x0 −x0i , t0 = 0).
The wave function in the far field is,
√
−i 2π x2
Ψ(x, t) = exp [i 2 ( + x x0i )] F [Φ(x0 )]u= x2 . (4)
L L 2 L
Here u is the reciprocal coordinate. The mean location of the initial wave packet, x0i , is
mapped onto the phase of the final wave function as a consequence of the well known shift
theorem from Fourier analysis. Thus, in the far field the detected signal is proportional
to the absolute value squared of the momentum space wave function and is insensitive
to the initial position of the atomic wave packet. For a large collection of incoherent
atoms, each in an arbitrary mixed state, the measured TOF signal is given by a statistical
average,
XR
1
n(x) = pj 2 |F [Φj (x0 )]u= x2 |2 , (5)
j=1
L L
where pj is the classical probability for the j th local wave function to occur. If the
atomic sample consists of R localized atoms in quantum mechanical pure states which
differ only by translation to a given lattice site, as recently demonstrated by Hamman
et al. [29], then one can partially reconstruct the initial wave function of the individual
atoms from the Wiener-Khintchine theorem subject to the usual limitations imposed
by the loss of phase information caused by taking the modulus [30,31]. This somewhat
surprising result can be made intuitive if one considers an optical analogy. Given light
illuminating a small random cluster of identical pinholes, as long as the photodetector
is in the far field and the distance between the pinholes is not too large, the intensity
will simply be the diffraction pattern of a single pinhole resulting from the incoherent
sum of the individual, completely overlapping, diffraction patterns.
Non-trivial dynamical information about coherent atomic wave packet motion can
be deduced from the results above. Suppose one arranges a 1D lattice of N double
wells using a configuration of counterpropagating lasers with wavelength λ whose linear
polarization have a relative angle θ [2]. The ith double well is located at x0i and the well
separation is set by ∆ξ 0 = (λ/(2π)) tan−1 (tan(θ)/2). With the appropriate cooling and
preparation of the initial state, the wave packet dynamics is essentially restricted to a
two dimensional Hilbert space spanned by two macroscopically separated Gaussians Φ0
of width, σ 0 , centered about, ±∆ξ 0 /2. A general state is given by the wave packet,
Ψ(x0 ) = c1 Φ0 (x0 − ∆ξ 0 /2) + eiφ c2 Φ0 (x0 + ∆ξ 0 /2) (6)
Here, c1 and c2 are the (real) probability amplitudes and φ is the relative phase between
the Gaussians. If this wave packet is allowed to freely expand, the atomic density at
the detector plane will have a Gaussian envelope with fringes whose spacing is given by,
df = L2 /∆ξ 0 , (See Fig. 1),
r
2 1 x 2 x∆ξ 0
|Ψ(x)| = exp (−2( ) ) (1 + 2c 1 c 2 cos(2π + φ)), (7)
2πσ 2 2σ L2
where σ is defined above. Observation of the fringes requires that the size of the sample,
S, be much smaller than the fringe separation, S << df (i. e. we are sufficiently in the far
field). In the absence of decoherence the complete initial wave function can be inferred.
The relative phase can be deduced from the shift of the center of the diffraction pattern
with respect to the fringe envelope and the probability amplitudes can be obtained from
the visibility of the fringes as a function of time. In the presence of decoherence the visi-
bility of these fringes will decay with time and will not exhibit recurrences characteristic
of the macroscopic superposition state. This could be a useful diagnostic to measure the
decohering effects of the lattice environment [2]. Similar interference patterns have been
measured in TOF experiments to extract the temperature of an atomic sample, cooled
via velocity-selective-coherent-population-trapping (VSCPT) [32]. The effectiveness of
the above diagnostic depends on one’s ability to prepare identical pure states. Intensity
inhomogeneities of the trapping laser or magnetic field gradients will cause wave func-
tions at different lattice sites to vary slightly, thus broadening the signal, although this
effect could be reduced through the use of apertures.
2.2 Spatial Distribution via First Order Visibility Measurements

We consider next the normalized first order correlation function, at different positions,
g (1) (x1 , x2 ), defined as the complex degree of coherence [30]. The modulo of this function,
V (∆x) = |g (1) (∆x/2, −∆x/2)| (8)
corresponds to the visibility of fringes formed by a Young’s double slit experiment

with slit spacing ∆x = x2 − x1 . In Michelson stellar interferometry one measures the
visibility of the interference pattern as a function of ∆x to deduce the spatial intensity
∆ξ ′
x′
L2
∆ξ ′
x
D
Fig. 1. Schematic of a TOF experiment. The atomic wave functions initially consists
of a superposition of two Gaussians separated
√ by ∆ξ 0 with relative phase φ = 0 and
probability amplitudes c1 = c2 = 1/ 2. The resulting completely incoherently
overlapping fringe pattern, has a fringe spacing of L2 /∆ξ 0 .
distribution of the source [30]. We consider the atom-optic version here as a diagnostic
of the distribution of atoms throughout the lattice (i.e. the small scale structure).
Suppose that a 1D lattice has N sites with lattice constant w0 . For simplicity we
take the wells to be harmonic at each site and the atomic state to be thermal so that
the local wave function is Gaussian. An atom in the vibrational ground state, initially
located at x0j = jw0 (j = 0, · · · , N − 1), will expand into an approximate plane wave in
the far field (See Fig. 2),
w0
Ψ(x, t) ∝ eikj x where kj = j2π . (9)
L2
The atomic field impinging on the double slit detector plane can thus be modeled as a
set of N discrete plane waves with a mode spacing of ∆k = 2πw0 /L2 . We define creation
and annihilation operators, â†j and âj , for the j th mode. The position space annihilation
operator b̂i at detector position xi is given by,
1 X
b̂i = √ âj eikj xi . (10)
N j
We will take our atoms to be Bosons (true for most laser cooled species) with creation
and annihilation operators satisfying the usual canonical commutation relations, though
the analysis for first order correlation functions is identical for the case of Fermions.
We further simplify our analysis by making some assumptions about the preparation
of the atomic sample. In a typical laser cooling experiment, near resonant scattering
and collisions prohibit two atoms from occupying the same lattice site. We therefore
only consider the case where a lattice site is either empty or contains one atom. In
x′
xj
w′
kj
kj
z
∆x
x1 x2 x
Fig. 2. Schematic of an experiment that measures same-time two-point first order

spatial correlations of an atomic field. By measuring the visibility of the fringes as
a function of slit spacing one can deduce the atomic distribution.
full analogy with quantum optics we assume that the apparatus detects an atom by
removing it from the field (i.e. an MCP), so that atom detection can be described
by normally ordered creation and annihilation operators. We will not consider here
other definitions of coherence based on other detection schemes (e.g. fluorescence and
nonresonant imaging [33]).
The detection time can be approximated as instantaneous if it is much shorter than
the coherence time, given by the length of the wave packet in the z-direction divided
by the group velocity. Furthermore, if we assume that an equal flux impinges on each
slit, then the complex degree of coherence is [34],
G(1) (x1 , x2 )
g (1) (x1 , x2 ) = p , (11)
G(1) (x1 , x1 ) G(1) (x2 , x2 )
G(1) (x1 , x2 ) = hb̂†1 b̂2 i

1 X i (kj x2 −k` x1 ) †
= e hâ` âj i.
N
j,`
Given an atomic field we can evaluate this expression and associate it with detection
through a double slit.
For concreteness consider a single atom located at x0j . The state of this system is
described by the one-atom Fock state, |0, · · · , 1kj , · · · , 0i, so that the complex coherence
factor is given by,
g (1) (x1 , x2 ) = ei kj (x2 −x1 ) . (12)
From this expression we see that the mean position of the atom in the lattice maps
onto the phase of the correlation function or the shift of the zero-delay of the fringes in
a double-slit experiment. It is simple to generalize this for ensemble averages of many
atomic field states,
N
X −1 N
X −1
(1) (1) (1)
g` = Pj ei (kj `∆x) = Pj ei (j∆k `∆x) , (13)
j=0 j=0
(1)
where g` = g (1) (− 2` ∆x, 2` ∆x) (` = 0, · · · , N − 1), is the complex degree of coherence
(1)
for a slit spacing of `∆x and Pj is the probability for lattice site j to contain an atom.
The fundamental slit spacing ∆x is set by the size of the atomic sample, (N − 1)w0 ; a
slit spacing of (N − 1)∆x is necessary to resolve two atoms separated by a w0 . Thus,
we see that the atomic probability distribution is related to the complex degree of
coherence through a discrete Fourier transform. This is the atom-optic analog of the
Van Cittert-Zernike theorem [30]. The inverse relationship is,
N −1
(1) 1 X (1) −i (j∆k `∆x)
Pj = g` e . (14)
2πN
`=0
In principle, Eqs. (13)-(14) can completely determine the spatial distribution of an

atomic sample from an ensemble of double slit experiments. However, unlike the exam-
ple considered in Sec. 2.1 where a Young-type interference pattern was built into the
initial wave function of the double well and seen in same-position first order correlations,
in this case coherence is sampled at two spatially separated points masked by the double
slit. This is a low flux measurement requiring ensemble averaging of multiple realiza-
tions of similarly prepared systems to acquire sufficient signal-to-noise. Furthermore,
even with high atomic flux, one must repeat the measurement at a variety of different
slit spacings to deduce the visibility dependence. Thus, Eqs. (13)-(14) are a useful di-
agnostic only if the atomic distribution can be exactly reproduced for each run of the
experiment. In general, however, the atomic distribution will vary from shot-to-shot,
causing uncontrollable phase shifts in the fringe pattern which will wash out the spatial
information, even for atomic samples that are statistically similar. By contrast, the sec-
ond order correlation function is more robust because it is insensitive to this phase. An
analogous situation exists in optics. The Michelson stellar interferometer which mea-
sures first-order field correlations is very sensitive to atmospheric fluctuations whereas
the Hanbury-Brown Twiss interferometer which measures intensity correlations is more
stable [30]. For this reason we consider higher order spatial correlations.
2.3 Spatial Distribution via Coincidence Count Measurements

We now consider the atom-optic analog of Hanbury-Brown Twiss stellar interferometry
which uses photon coincidence counting as a function of ∆x to deduce the spatial in-
tensity distribution of the source [30]. The normalized same-time second order spatial
correlation function at different positions, g (2) (x1 , x2 ), corresponds to atom coincidence
counts between two point-like detectors located at x1 and x2 in the detection plane [34]
(See Fig. 3),
G(2) (x1 , x2 ; x2 , x1 )
g (2) (x1 , x2 ) = , (15)
G (x1 , x1 ) G(1) (x2 , x2 )
(1)
G(2) (x1 , x2 ; x2 , x1 ) = hb̂†1 b̂†2 b̂2 b̂1 i

1 X i ((kj −kj0 )x1 +(k` −k`0 )x2 ) † †
= e hâj 0 â`0 â` âj i.
N2 0 0
j,j ,`,`
x′
x j x ′j + ∆x
w′
w′
kj k j + ∆k
∆x
x1 x2 x
D1 D2
C
Fig. 3. Schematic for an experiment that measures same-time second order spatial
correlations of an atomic field. The interference arises from the two possible ways
that the atoms can be jointly detected, denoted by the solid and dotted paths.
Again for concreteness consider two Bosons separated by one lattice site, x0j and
x0j +w0 , with state vector, |0, · · · , 1kj , 1kj +∆k , · · · , 0i, where ∆k = 2πw0 /L2 . Inserting this
in Eq. (15) gives, 2g (2) (x1 , x2 )−1 = cos(∆k(x2 −x1 )). The difference in position between
the two atoms maps onto the spatial period of the cosinusoidially varying coincidence
counts. For Fermions, the fringes receive a π phase shift due to their anticommutation
relations. The interference exists even with no quantum entanglement between the two
atoms. The interference term arises from the two possible indistinguishable paths that
lead to joint detection (See Fig. 3). This relation can be generalized for field states of
(2)
R atoms distributed throughout N lattice sites. We define g` = g (2) (− 2` ∆x, 2` ∆x) for
two detectors separated by `∆x (` = 0, · · · , N − 1), given the state a†1 a†2 , · · · , a†R |0i we
find
N
X −1
R2 (2) (2)
g` − 1 = Pj cos(j∆k `∆x). (16)
R(R − 1) j=0
(2)
Pj is the probability for two atoms to be separated by j lattice constants jw0 (j =
0, · · · , N − 1), or equivalently, for two atomic plane waves to impinge the detector plane
(2)
with a mode spacing of j∆k. We see that g0 = 1 for R = 2, reflecting the perfect
(2)
second-order coherence for the state of exactly two atoms and g0 → 2 as R → ∞, which
is the usual bunching factor associated with a highly chaotic macroscopic distribution
(2)
[34]. Note, Pj depends only on the relative mode spacing and is independent of the
absolute location of a pair of atoms. Such information is mapped onto the phase of the
first order correlation function as discussed in Sec. 2.2 but is irrelevant to the second
order correlation function.
With the assumption of large R and analytically extending the sum to negative
values of j (j = −N, · · · , N − 1), since g (2) is symmetric under reflections about zero,
we obtain the following discrete Fourier relation,
N −1 (2)
(2)
X Pj
g` −1= ei j∆k `∆x (17)
2
j=−N
and its inverse

(2) N −1
Pj 1 X (2)
= (g` − 1) e−i j∆k `∆x . (18)
2 2π(2N )
`=−N
(2)
Similar relations can be obtained for Fermions by making the transformation (g` −1) →
(2)
−(g` − 1). For the particular case that the atomic positions satisfy Gaussian statistics,
it is well known that g (2) (x1 , x2 ) is related to the first order autocorrelation of the initial
atomic distribution and can be obtained through the Wiener-Khintchine theorem [30].
We will not restrict our attention to this case in the results presented below.
To determine the practical resolution of this detection scheme, note that a pair of
atoms spaced by w0 in the optical lattice plane will result in a coincidence count period
of, Λ = L2 /w0 . The smallest atomic separation in the lattice plane that can be resolved
is determined by the width of the Gaussian envelope in the detection plane which
modulates the interference fringes. For the situation of cesium atomic wave packets
freely expanding for ∼ 1 sec, w0 ' 0.5µm, and σ 0 ' 30 nm, then Λ ' σ ' 1 cm (where
σ 0 and σ were defined in Sec. 2.1). Also, for any case of practical significance, the small
scale information will be contained in the “wings” of the Gaussian in the detection
plane where the phase varies quadratically. The above Fourier relations will still be
valid, however, if the detectors are located on the wave front of constant phase, since
the Van-Cittert Zernike theorem is valid under a Fresnel approximation [30]. This can be
accomplished with a curved paraboloid detection surface, symmetric coincidence counts,
or a flat detection surface with appropriate electronic time delays introduced. Finally,
these relations are completely separable and can be extended to three dimensions. For
the z-direction (the direction that the atoms are falling) same-time spatial coincidences
are replaced by same-position, temporal coincidences, with the transformation given by
the appropriate dynamical equations. If the atoms are falling under the influence of
gravity, then x = gt2 /2, and a detector spacing of ∆x corresponds to a detection delay
time of gt∆t.
3 Results
In principle, the density operator, ρ̂, offers a complete description of the atomic distribu-
tion. For our assumptions about the nature of the sample with no coherences between
different lattice site, we need only consider diagonal matrix elements,
P (n) (x01 , x02 , · · · , x0n ) = h1k1 , 1k2 , · · · , 1kn |ρ̂|1k1 , 1k2 , · · · , 1kn i, (19)
corresponding to the nth order joint probability detection.
In this work we will restrict our attention solely to the first and second order proba-
(1)
bilities. We have seen that g (1) (x1 , x2 ) allows for a direct measure of Pj , the probability
0 (2)
for an atom to be located at xj , and g (x1 , x2 ) provides for a direct measure of the
(2)
Pj , the joint probability for two atoms to be separated by j lattice spacings, w0 ,
independent of the absolute location of the pair,
N
X
(2)
Pj = P (2) (x0` + jw0 , x0` ). (20)
`=1
(a) (d)
1 1
Natom
Natom
0.5 0.5
0 0
0 50 100 150 200 250 0 50 100 150 200 250
(b) (e)
1 1
Natom
Natom
0.5 0.5
0 0
0 50 100 150 200 250 0 50 100 150 200 250
(c) (f)
1 1
Natom
Natom
0.5 0.5
0 0
0 50 100 150 200 250 0 50 100 150 200 250
j j
Fig. 4. Atomic distribution bunched around a chosen “seed point”. First column (a)-
(c) shows samples with a fixed seed at site N = 128. The second column, (d)-(f),
shows the atomic distribution for a randomly varying seed point.
(1) (2)
The spatial information contained in Pj and Pj is different. To see this, factor the
joint probability under the sum using Bayes’ rule,
N
X
(2) (1)
Pj = P` P (x0` + jw0 |x0` ), (21)
`=1
where P (x0` + jw0 |x0` ) is the conditional probability for an atom to be located at x0` + jw0
given that an atom is at x0` . Only in the special case that the lattice sites are statistically
(1) (2)
independent so that P (x0` + jw0 |x0` ) = P`+j , do we obtain Pj from an autocorrelation
(1)
of Pj ,
N
X
(2) (1) (1)
Pj = P` P`+j . (22)
`=1
This is an example of the Wiener-Khintchine theorem [30].

We have performed computer simulations of TOF experiments that measure the
first and second order spatial correlations discussed above. The lattice had 256 lattice
sites and a fill factor of ' 10%. The simulation consisted of ensemble averaging the
complex coherence factor and atom coincidence counts over 500 runs. The averaged data
(1) (2)
is inverted with the appropriate Fourier relation given above to obtain Pj and Pj
respectively. In the first simulation, the atoms were distributed in the lattice according
to a conditional “bunched” distribution in which a seed point x0` was picked and then
the atomic distribution was conditioned on it in such a way as to cluster around it,
r
0 0 1 −2( x0 −x0` )2
P (x |x` ) = e 2τ . (23)
2πτ
In order to elucidate the effects of shot-to-shot phase fluctuations, the seed point was
picked in two ways. First we fixed the seed at lattice site N = 128 for each run of the
gedenken experiment and in the second we let the initial seed point vary randomly (See
Fig. 4).
(1) (2)
The results for Pj and Pj given the fixed seed point are shown in Fig. 5a-b and
we find that both correlation functions contain useful spatial information. When the
seed point is varied randomly for each run of the experiment, so that the atoms tended
(a) (c)
0.04 0.04
0.03 0.03
(1)
P(1)
Pj
j
0.02 0.02
0.01 0.01
0 0
0 50 100 150 200 250 0 50 100 150 200 250
(b) (d)
0.04 0.04
0.03 0.03
P(2)
P(2)
j
j
0.02 0.02
0.01 0.01
0 0
0 50 100 150 200 250 0 50 100 150 200 250
j j
(1)
Fig. 5. Probability for an atom to be located a site j, Pj , and for two atoms to be
(2)
separated by j sites, Pj ,
for a fixed seed point, (a) and (b), and for a randomly
varying seed point, (c) and (d).
(a) (c)
0.04 0.04
0.03 0.03
P(2)
P(2)
j
j
0.02 0.02
0.01 0.01
0 0
0 20 40 60 80 100 0 20 40 60 80 100
(b) (d)
0.06 0.04
0.05
0.03
0.04
P(2)
P(2)
j
j
0.03 0.02
0.02
0.01
0.01
0 0
0 20 40 60 80 100 0 20 40 60 80 100
j j
Fig. 6. Probability for atoms to be separated by j lattice sites, obtained by Fourier

transform of simulated coincidence count measurements. Results are shown for
atomic distributions which are (a) a random , (b)“bunched”, (c) “anti-bunched”,
and (d) macroscopic variation on a “super-lattice”.
to cluster in a different parts of the lattice, we see that the spatial information contained
(1) (2)
in Pj completely washes out, while the spatial information in Pj is unaffected (Fig.
5c-d). The first order correlations g (1) (x1 , x2 ) depend on the absolute locations of atoms
and on average it sees a randomly filled lattice. In contrast, g (2) (x1 , x2 ) measures only
the relative locations of atoms independent of the absolute location of the cluster.
To illustrate the capability of atomic coincidence counting, we carried out simulations
for various distributions: random, “bunched”, “anti-bunched”, “macroscopic-periodic”.
The results are compared in Fig. 6. Note that our coincidence period was taken to be
just large enough to resolve the lattice spacing. Larger detection areas would result in a
“picket fence” distribution which would explicitly display the periodicity of the lattice.
An important simplification arises when the atomic sample is characterized by a
conditional probability with the functional form
P (x0 |x0` ) = f (x0 − x0` ). (24)
Upon substituting this into Eq. (21) one finds,
(2)
X (1)
Pj = P` f (x0` + jw0 − x0` ) = f (jw0 ). (25)
`
In this case of stationary statistics, the joint probability, as measured by g (2) (x1 , x2 ), is
a direct measure of the relative conditional probabilities of the lattice, independent of
global properties of the lattice such as intensity or magnetic field inhomogeneities.
4 Summary
Given very cold atoms trapped in an optical lattices, their wave nature becomes man-
ifest. We have explored the possibility of exploiting this feature to image the atomic
distribution using both first and second, same-time, atomic spatial correlation func-
tions in TOF diagnostics. Although we have analyzed these correlation functions in one
dimension, the derived relations still hold in three dimensions because the free space
Hamiltonian is completely separable. One finds that information about a single atomic
wave function can be inferred from the atomic density in the detection plane for a lattice
filled with atoms whose quantum mechanical states are identical up to translation. The
ability to measure this quantity depends on having a lattice potential that is homoge-
neous, but for a typical sample, the signal obtained from a single run of this type of
TOF experiment is large. Information about the spatial distribution of atoms through-
out the lattice can be obtained from both first and second order correlation functions
at different spatial points through Fourier relations that connect the measured signal
and the initial atomic distribution. In contrast to atomic density measurements in the
detection plane, visibility and atom coincidence counts in the detection plane require
the necessity of averaging many TOF experiments in order to obtain sufficient signal
and this generally leads to fluctuations which wash out the interference fringes associ-
ated with g (1) (x1 , x2 ). In contrast, g (2) (x1 , x2 ) is more robust and can be a more useful
diagnostic. We have performed computer simulations of these detection schemes which
illustrate the salient features.
Acknowledgements
The authors gratefully acknowledge Sudakar Prasad, Steven Rolston, Simone Kulin,
and Gavin Brennen for many useful discussions. This research was supported by NSF
Grant No. PHY-9732456 and the Albuquerque High Performance Computing Center.

Optik 26

Uploaded by

Copyright:

Available Formats

Optik 26

Uploaded by

Document Information

Original Description:

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Optik 26

Uploaded by

Copyright:

Available Formats

Spatial correlation diagnostics for atoms in

J. Grondalski∗,† , P. M. Alsing† , and

Abstract: We explore the use of first and second order same-time

References and links

For free space, the propagator K(x, t; x0 t0 ) is given by

1 1 −iπ( x0 −x0i )2 −( x0 −x0i )2

Ψ(x0 ) = c1 Φ0 (x0 − ∆ξ 0 /2) + eiφ c2 Φ0 (x0 + ∆ξ 0 /2) (6)

2.2 Spatial Distribution via First Order Visibility Measurements

V (∆x) = |g (1) (∆x/2, −∆x/2)| (8)

corresponds to the visibility of fringes formed by a Young’s double slit experiment

Fig. 2. Schematic of an experiment that measures same-time two-point first order

G(1) (x1 , x2 ) = hb̂†1 b̂2 i

In principle, Eqs. (13)-(14) can completely determine the spatial distribution of an

2.3 Spatial Distribution via Coincidence Count Measurements

G(2) (x1 , x2 ; x2 , x1 ) = hb̂†1 b̂†2 b̂2 b̂1 i

and its inverse

This is an example of the Wiener-Khintchine theorem [30].

Fig. 6. Probability for atoms to be separated by j lattice sites, obtained by Fourier

P (x0 |x0` ) = f (x0 − x0` ). (24)

You might also like