Single-Spin Addressing in an Atomic Mott Insulator

Christof Weitenberg
1
, Manuel Endres
1
, Jacob F. Sherson
1
, Marc Cheneau
1
,
Peter Schau
1
, Takeshi Fukuhara
1
, Immanuel Bloch
1,2
, and Stefan Kuhr
1
1
Max-Planck-Institut f ur Quantenoptik, Hans-Kopfermann-Str. 1, 85748 Garching, Germany
2
Ludwig-Maximilians-Universit at, Schellingstr. 4/II, 80799 M unchen, Germany
(Dated: 9 January 2011)
Ultracold atoms in optical lattices are a versatile tool to investigate fundamental properties of
quantum many body systems. In particular, the high degree of control of experimental parame-
ters has allowed the study of many interesting phenomena such as quantum phase transitions and
quantum spin dynamics. Here we demonstrate how such control can be extended down to the most
fundamental level of a single spin at a specic site of an optical lattice. Using a tightly focussed
laser beam together with a microwave eld, we were able to ip the spin of individual atoms in
a Mott insulator with sub-diraction-limited resolution, well below the lattice spacing. The Mott
insulator provided us with a large two-dimensional array of perfectly arranged atoms, in which we
created arbitrary spin patterns by sequentially addressing selected lattice sites after freezing out the
atom distribution. We directly monitored the tunnelling quantum dynamics of single atoms in the
lattice prepared along a single line and observed that our addressing scheme leaves the atoms in
the motional ground state. Our results open the path to a wide range of novel applications from
quantum dynamics of spin impurities, entropy transport, implementation of novel cooling schemes,
and engineering of quantum many-body phases to quantum information processing.
The ability to observe and control the position of sin-
gle atoms on a surface of a solid via scanning tunnelling
and atomic force microscopy has revolutionised the eld
of condensed matter physics [1, 2]. In few-atom systems,
coherent control of single particles in e.g. an ion chain
has proven crucial for the implementation of high-delity
quantum gates and the readout of individual qubits in
quantum information processing [3]. Bringing such lev-
els of control to the regime of large scale many-body sys-
tems has been a longstanding goal in quantum physics.
In the context of ultracold atoms in optical lattices, a
major challenge has been to combine degenerate atomic
samples with single-site addressing resolution and single-
atom sensitivity. This full control is essential for many
applications in condensed matter physics, such as the
study of spin impurities [4] and quantum spin dynamics
[5, 6] within quantum magnetism, entropy transport, the
implementation of novel cooling schemes [7, 8] or digital
quantum simulations based on Rydberg atoms [9]. For
scalable quantum information processing, a Mott insula-
tor with unity lling provides a natural quantum register
with several hundreds of qubits. In order to exploit the
full potential of such a large scale system for quantum
computation, coherent manipulation of individual spins
is indispensable, both within a circuit-based [10] or a
one-way quantum computer architecture [11, 12].
The quest to address atoms on single sites of an opti-
cal lattice has a long history [7, 1322]. In one dimen-
sion, single-site addressing was accomplished optically

present address: Department of Physics and Astronomy, Univer-

sity of Aarhus, DK-8000 Aarhus C, Denmark.

Electronic address: stefan.kuhr@mpq.mpg.de

in a long-wavelength lattice [23], in which however tun-
nelling was completely suppressed, and using magnetic
resonance techniques in a sparsely lled short-wavelength
lattice [24, 25]. In two dimensions, an electron beam was
used to depopulate sites of a Bose-Einstein condensate
loaded into an array of potential tubes, each containing
up to 80 atoms [26]. In this case, coherent spin manip-
ulation was not possible and the readout was done by
averaging over more than 100 single images. None of the
experiments so far has shown single-atom spin control
in strongly correlated systems together with high delity
single-atom detection.
Here we report on the achievement of this goal, by
demonstrating single-site-resolved addressing and control
of the spin states of individual atoms in a Mott insula-
tor in an optical lattice. The Mott insulator provided
us with an almost perfect initial two-dimensional array
of atoms in the same spin state. Apart from few ther-
mal defects, each lattice site contained a single atom in
its motional ground state [27, 28]. Using a tightly fo-
cused laser beam, we introduced a controlled dierential
energy shift between two atomic spin states at a given lat-
tice site. Microwave radiation resonant with this shifted
transition then allowed us to selectively address the spin
of a single atom [7, 17] with high delity. We thereby ob-
tained sub-diraction-limited spatial resolution well be-
low the lattice spacing. By moving the addressing laser
beam to dierent lattice sites and by inducing spin-ips
in the Mott insulator, we were able to deterministically
create arbitrary two-dimensional spin patterns of indi-
vidual atoms, thereby realising a scalable single-atom
source [2931]. Furthermore, we investigated how much
our single-spin manipulation aects the motional state of
the atoms by directly monitoring the tunnelling dynam-
ics of single atoms after addressing them. Averaging over
a
r
X
i
v
:
1
1
0
1
.
2
0
7
6
v
1


[
c
o
n
d
-
m
a
t
.
q
u
a
n
t
-
g
a
s
]


1
1

J
a
n

2
0
1
1
2
FIG. 1: Addressing scheme. a, Atoms in a Mott insula-
tor with unity lling arranged on a square lattice with pe-
riod a
lat
= 532 nm were addressed using an o-resonant laser
beam. The beam was focussed onto individual lattice sites by
a high-aperture microscope objective (not shown) and could
be moved in the xy plane with an accuracy of better than
0.1 a
lat
. b, Energy diagram of atoms in the lattice for the
two hyperne states |0 and |1. The

-polarized addressing
beam locally induces a light shift LS of state |1, bringing
it into resonance with a microwave eld. A Landau-Zener
sweep (central frequency MW, sweep width MW) transfers
the addressed atoms from |0 to |1.
several snapshots after dierent tunnelling times, we fully
reconstructed the characteristic spatial probability distri-
bution of the single-atom wave function and its coherent
evolution over more than 20 lattice sites. We were able
to discriminate the dynamics of the atoms in the zeroth
and in the rst band and found that most of the atoms
remained in the motional ground state after addressing.
Experimental setup
In our experiments, we prepared a two-dimensional
sample of ultracold
87
Rb atoms in an optical lattice, con-
ned in a single antinode of a vertical standing wave along
the z direction. Two pairs of counterpropagating laser
beams (wavelength = 1064 nm) along the horizontal x
and y directions provided a square lattice with period of
a
lat
= /2 = 532 nm (for details see Ref. [28]). Starting
from a Bose-Einstein condensate we raised the poten-
tial in the x and y lattice axes within 75 ms to values of
V
x,y
= 23(2) E
r
(the number in parenthesis denotes the
1 uncertainty of the last digit), where E
r
= h
2
/(2m
2
)
is the recoil energy, and m denotes the atomic mass of
87
Rb. In this way, the interaction energy of the particles
with respect to their kinetic energy was increased such
that the system undergoes a quantum phase transition
to a Mott insulating state [3234]. Due to the external
harmonic connement, the Mott insulator exhibits a shell
structure with xed integer atom numbers increasing in
a step-like manner from the outer regions of the system
to the inner core [35, 36]. The initial state for all ex-
periments presented in this paper was a single shell with
only one atom per lattice site, which in our system was
realised for atom numbers smaller than 400.
We detected the atoms using uorescence imaging via
a high-resolution microscope objective with numerical
aperture of NA = 0.68. An optical molasses induced uo-
rescence light and simultaneously laser-cooled the atoms.
Light-assisted collisions lead to rapid losses of atom pairs,
such that we only detected the atom number modulo two
on each lattice site [27, 28]. With about 5,000 collected
photons per atom, we identied individual atoms in the
lattice with an excellent signal-to-noise ratio. Even in
the regions of high atomic density, we determined the
presence or absence of an atom for each lattice site with
> 99.5% delity using a special reconstruction algorithm
[28].
Addressing single lattice sites
In order to address the atoms in the lattice, we used
an o-resonant laser beam focused by the high-resolution
imaging system onto individual lattice sites (Fig. 1). The
laser beam causes a dierential light shift of the two
relevant hyperne levels |0 |F = 1, m
F
= 1 and
|1 |F = 2, m
F
= 2 and tunes the addressed atom
into resonance with an external microwave eld at 6.8
GHz. The

-polarized addressing beam had a wave-

length of 787.55 nm, between the D
1
and D
2
lines, in
order to obtain a large dierential light shift between
the two hyperne levels. For perfect

-polarization,
this magic wavelength generates a light shift only for
state |1, and leaves state |0 unaected. The beam
had a diameter of 600 nm full-width at half-maximum
(FWHM) and could be moved in the object plane over
the entire eld of view by changing its angle of incidence
into the microscope objective with a two-axis piezo mir-
ror. We were able to position the beam with an accu-
racy better than 0.1 a
lat
using an independent calibration
3
e
d
f
a
2 m
b
c
x
y
FIG. 2: Single-site addressing. a, Experimentally obtained uorescence image of a Mott insulator with unity lling in which
the spin of selected atoms was ipped from |0 to |1 using our single-site addressing scheme. Atoms in state |1 were removed
by a resonant laser pulse before detection. The lower part shows the reconstructed atom number distribution on the lattice.
Each circle indicates a single atom, the points mark the lattice sites. b-f, Same as a, but a global microwave sweep exchanged
the population in |0 and |1, such that only the addressed atoms were observed. The line in b shows 14 atoms on neighbouring
sites, the images c-f contain 29, 35, 18 and 23 atoms, respectively. The single isolated atoms in b,e,f were placed intentionally
to allow for the correct determination of the lattice phase for the feedback on the addressing beam position.
measurement of its position together with a feedback that
tracks the slowly varying lattice phases (see Appendix).
If the addressing laser beam is perfectly centered onto
one lattice site (see Fig. 1b), the dierential light shift
is
LS
/(2) 70 kHz, whereas a neighbouring atom
only experiences 10% of the peak intensity. The result-
ing dierence in light shifts can be well resolved spec-
trally by our microwave pulses. In order to ip the spin,
we performed Landau-Zener sweeps (see Appendix) of

MW
/(2) = 60 kHz width and 20 ms duration yielding a
near at-top frequency spectrum with a maximum pop-
ulation transfer eciency of 95%.
As a rst experiment, we sequentially ipped the spin
of the atoms at selected lattice sites in our Mott insulator
with unity lling and spin state |0 (Fig. 2a). The lattice
depths were rst changed to V
x
= 56 E
r
, V
y
= 90 E
r
and
V
z
= 70 E
r
in order to completely suppress tunnelling
even when the addressing beam locally perturbs the lat-
tice potential. For each lattice site, we then switched on
the addressing laser beam with an s-shaped ramp within
2.5 ms, which is adiabatic with respect to the on-site
oscillation frequency of 30 kHz. Subsequently, a mi-
crowave pulse with the parameters described above pro-
duced spin-ips from |0 to |1. The addressing laser was
switched o again within 2.5 ms, before its position was
changed in 5 ms to address the next lattice site. For the
image of Fig. 2a, this procedure was repeated 16 times in
order to ip the spins along a line. Finally, a 5 ms push-
out laser pulse, resonant with the F = 2 to F

= 3 tran-
sition, removed all addressed atoms in state |1. In order
to reveal only the spin-ipped atoms, the spin states of all
atoms were ipped by a global microwave sweep before
the push-out laser was applied (Fig. 2c,d). However, due
to the nite transfer eciency of the global sweep, some
atoms remaining in state |0 were clearly visible in addi-
tion to the addressed ones. To avoid this problem when
detecting the addressed atoms, we initially transferred
the whole sample to state |1 by a microwave sweep and
then shone in a repumping laser that completely depop-
ulated state |0. Then, we used our addressing scheme
to transfer selected atoms back to |0 and subsequently
pushed out the atoms in |1, yielding typical images as
shown in Fig. 2b,e,f.
Spin-ip delity
We quantied the success rate of our addressing scheme
by again producing a series of spin-ips along the y lattice
axis in a Mott insulator with unity lling (see Fig. 3). The
experimental procedure was the same as described above
for the realisation of Fig. 2a, in which the addressed sites
were detected as empty sites. From the reconstructed
atom number distribution (Fig. 3a), we determined the
4
a
c
Pointing offset x (lattice sites)
0.2 0.4 0 -0.2 -0.4
0.8
0.6
0.4
0.2
0
1
FWHM = 330(10) nm
H
o
l
e

p
r
o
b
a
b
i
l
i
t
y

p
0
Fidelity = 0.95(2)
y
x
x
b
FIG. 3: Addressing delity. The spin-ip probability was
measured by sequentially addressing a series of 16 neighbour-
ing sites along the y lattice axis (red circles in a) in a Mott
insulator with unity lling. The red data points in c show
the resulting hole probability p0(x) as a function of the
pointing oset x, as dened in b. Each point was obtained
by averaging over 4 7 pictures (total 50 100 addressed
lattice sites), taking only those sites into account which lie
well within a Mott shell with unity lling. The displayed
error bars show the 1 statistical uncertainty, given by the
Clopper-Pearson condence limits. The data was tted by
a at-top model function (see Appendix) and yields a full-
width at half-maximum a = 330(20) nm, an edge sharpness
of s = 50(10) nm, and a peak delity of 95(2)%. The oset
was xed at the 6(2)% probability of thermally activated holes
as deduced from the neighbouring and next neighbouring sites
(blue shaded regions in a,b and blue points in c).
probability p
0
(x) of nding an empty site as a function
of the pointing oset x between the addressing beam
and the center of the lattice site (see Fig. 3b). We also
investigated the eect of the addressing on neighbouring
atoms, which ideally should remain unaected. For this
purpose, we monitored the probability of nding a hole
at the sites next to the addressed ones (dark blue regions
in Fig. 3a,b and points in Fig. 3c). In order to distinguish
accidentally ipped neighbouring atoms from holes that
originate from thermal excitations of the initial Mott in-
sulator [28], we also monitored the probability of nding
a hole at the second next neighbours (light blue regions
and points in Fig. 3). As both yielded the same hole
probability of 6(2)%, we attribute all holes to thermal
excitations and conclude that the probability of address-
ing a neighbouring atom is indiscernibly small. We tted
the hole probability p
0
(x) of the addressed site with a
at-top model function (see Appendix), keeping the o-
set xed at the thermal contribution of 6%. From the t,
we derived a spin-ip delity of 95(2)%, a full-width at
half-maximum of
a
= 330(10) nm and an edge sharpness
of
s
= 50(10) nm (Fig. 3c). These values correspond to
60% and 10% of the addressing beam diameter, demon-
strating that our method reaches sub-diraction-limited
resolution, well below the lattice spacing.
The observed maximum spin-ip delity is currently
limited by the population transfer eciency of our mi-
crowave sweep. The edge sharpness
s
originates from
the beam pointing error of 0.1 a
lat
and from variations
in the magnetic bias eld. The latter causes frequency
uctuations of 5 kHz, which translate into an eective
pointing error of 0.05 a
lat
at the maximum slope of the
addressing beam prole. The resolution
a
could in prin-
ciple be further reduced by a narrower microwave sweep,
at the cost of a larger sensitivity to the magnetic eld
uctuations. A larger addressing beam power would re-
duce this sensitivity, but we observed that this deformed
the lattice potential due to the imperfect

-polarization,
allowing neighbouring atoms to tunnel to the addressed
sites.
Coherent tunneling dynamics
The preparation of an arbitrary atom distribution
opens up new possibilities for exploring coherent quan-
tum dynamics at the single-atom level. As an example,
we studied the tunnelling dynamics in a one-dimensional
lattice (see Fig. 4) which allowed us to determine how
much our addressing scheme aects the vibrational state
of the atoms. We started by preparing a single line of
up to 18 atoms along the y direction before we lowered
the lattice along the x direction to V
x
= 5.0(5)E
r
within
200 s. At the same time, the other lattices were lowered
to V
y
= 30 E
r
and V
z
= 23 E
r
, which reduced the exter-
nal connement along the x direction, but still suppressed
tunnelling in the y and z directions. After a varying hold
time t, allowing the atoms to tunnel along x, the atomic
distribution was frozen by a rapid 100 s ramp of all lat-
tice axes to 70 E
r
. By averaging the resulting atomic
distribution along the y direction and repeating the ex-
periment several times, we obtained the probability dis-
tribution of nding an atom at the dierent lattice sites
(Fig. 4, bottom row).
This probability distribution samples the single-atom
wave function after a coherent tunnelling evolution. We
observed how the wave function expands in the lattice
and how the interference of dierent paths leads to dis-
tinct maxima and minima in the distribution, leaving
for example almost no atoms at the initial position after
a single tunnelling time (Fig. 4c). This behaviour diers
5
b
10 5 0 5 10
0
0.5
1
x Position (lattice sites)
10 5 0 5 10
0
0.1
0.2
0.3
0.4
10 5 0 5 10
0
0.1
0.2
0.3
0.4
10 5 0 5 10
0
0.1
0.2
0.3
0.4
x Position (lattice sites)
P
r
o
b
a
b
i
l
i
t
y 1.5 ms 3.0 ms
c d a
x
0.0 ms 0.75 ms
x Position (lattice sites) x Position (lattice sites)
y
FIG. 4: Ground state tunnelling dynamics. a, Atoms were prepared in a single line along the y direction before the lattice
along the x axis was lowered, allowing the atoms to tunnel in this direction (b-d). The top row shows snapshots of the atomic
distribution after dierent hold times. White circles indicate the lattice sites at which the atoms were prepared (not all sites
initially contained an atom). The bottom row shows the respective position distribution obtained from an average over 10 20
of such pictures, the error bars give the 1 statistical uncertainty. A single t to all distributions recorded at dierent hold
times (red curve) yields a tunnelling coupling of J
(0)
/ = 940(20) Hz, a trap frequency of trap/(2) = 103(4) Hz and a trap
oset of x
os
= 6.3(6) a
lat
.
markedly from the evolution in free space, where a Gaus-
sian wave packet disperses without changing its shape,
always preserving a maximum probability in the center.
For longer hold times, an asymmetry in the spatial dis-
tribution becomes apparent (Fig. 4d), which originates
from an oset between the bottom of the external har-
monic connement and the initial position of the atoms.
We describe the observed tunnelling dynamics by a sim-
ple Hamiltonian including the tunnel coupling J
(0)
be-
tween two neighbouring sites and an external harmonic
connement, parameterized by the trap frequency
trap
,
and the position oset x
os
(see Appendix). A single t
to all probability distributions recorded at dierent hold
times yields J
(0)
/ = 940(20) Hz,
trap
/(2) = 103(4) Hz
and x
os
= 6.3(6) a
lat
. This is in agreement with the
trap frequency
trap
/(2) = 107(2) Hz obtained from an
independent measurement via excitation of the dipole
mode without the x lattice, whose contribution to the
external connement is negligible compared to the other
two axes. From J
(0)
, we calculated a lattice depth of
V
x
= 4.6(1) E
r
, which agrees with an independent cal-
ibration via parametric heating. The expansion of the
wave packet can also be understood by writing the ini-
tial localized wave function as a superposition of all
Bloch waves of quasi-momentum q, with /a
lat
q
/a
lat
. To each quasi-momentum q, one can assign a
velocity v
q
=
1

E
q
, determined by the dispersion rela-
tion E(q) = 2J
(0)
cos(qa
lat
) of the lowest band. The
edges of the wave packet propagate with the largest oc-
curring velocity v
max
= 2J
(0)
a
lat
/ = 1.88(4) a
lat
/ms, in
agreement with our data.
Our measurements constitute the rst observation of
the ground state tunnelling dynamics of massive parti-
cles on a lattice with single-site resolution. A similar
continuous-time quantum walk has been demonstrated
with photons in an array of evanescently coupled pho-
tonic waveguides [37]. For massive particles, a discrete
quantum walk of single atoms has been observed us-
ing a sequence of spin manipulations and spin-dependent
transports in an optical lattice [38] and also with trapped
ions [39]. Without single-particle and single-site resolu-
tion, a continuous-time quantum walk in the ground state
has been observed for ultracold fermionic atoms by mea-
suring their ballistic expansion in a weak lattice [40].
In a second tunnelling experiment, we observed the
faster dynamics of atoms in the rst excited band (see
Fig. 5). For this, we deliberately excited the atoms by
introducing a pointing oset x of the addressing beam,
which caused a shift of the potential wells during the
switch-on. We repeated the same tunnelling experiment
as above with a hold time of t = 1 ms for dierent point-
ing osets x. For a small pointing oset (x = 0.1 a
lat
in Fig. 5b) we observed a narrow distribution, compared
to a much broader one for a large oset (x = 0.4 a
lat
in Fig. 5a). We attribute this to dierent fractions f of
atoms in the rst band which is characterized by the
higher tunnelling rate J
(1)
. We tted the distribution of
Fig. 5a to a two-band model (see Appendix) and found
J
(1)
/ = 6.22(6) kHz. This is in excellent agreement with
the expected value of J
(1)
/ = 6.14(6) kHz from a band
6
Pointing offset x (lattice sites)
x Position (lattice sites)
F
r
a
c
t
i
o
n

f

i
n

f
i
r
s
t

b
a
n
d
0 0.2 0.4 -0.2 -0.4
0.2
0.6
0.4
b
a
P
r
o
b
a
b
i
l
i
t
y
P
r
o
b
a
b
i
l
i
t
y
15 5 0 -15 -10 -5 10
15 5 0 -15 -10 -5 10
0.04
0.1
0.3
0.2
0
0.08
0

x = -0.4 a
lat
x = 0.1 a
lat

FIG. 5: Tunnelling dynamics of the rst excited band.

Some atoms were excited to the rst band by a pointing oset
x of the addressing beam. a and b show the atomic position
distribution after 1 ms tunnelling time for x = 0.4 a
lat
and
x = 0.1 a
lat
, respectively. We tted the data with a model
that includes atoms in the zeroth band (red line) and a frac-
tion f in the rst band (orange line) (see Appendix). The
right insets of a, b show corresponding original images. The
left inset of b shows f versus x with a broad minimum of
f = 13(2)%, indicating that most of the atoms are left in the
ground state.
structure calculation in which we used J
(0)
as an input
parameter to calculate the lattice depth. Our measure-
ment of the fraction of excited atoms f as a function of
the pointing oset x (inset in Fig. 5b) shows that the
atoms are strongly heated for large pointing osets. By
contrast, only a small fraction of the atoms is excited to
the rst band for small pointing osets |x| 0.1 a
lat
,
yielding a ground state population of 1 f = 87(2)%.
Discussion
In summary, we have demonstrated full two-
dimensional single-site and single-atom spin control in
an optical lattice with sub-diraction-limited spatial res-
olution. Starting from a Mott insulator with unity lling,
we achieved a spin-ip delity of 95(2)% with negligible
inuence on the neighbouring lattice sites. Our scheme
leaves most of the atoms in the vibrational ground state.
The control of single spins in a strongly correlated many-
body system on a lattice opens many new possibilities for
studying quantum dynamics and quantum phases. Our
technique will allow us to create out-of-equilibrium states
or local perturbations in order to observe the ensuing
dynamics of the many-body system, such as spin-charge
separation [5, 6] or spin impurity dynamics beyond Lut-
tinger liquid theory [4]. Our studies of the tunnelling
dynamics at the single-atom level can be extended to cor-
related particle tunnelling [4143], also in higher dimen-
sions, or to observe transport across local impurities [44]
or potential barriers. The tunneling can also be used to
circumvent the pairwise losses during the imaging [27, 28]
by letting the atoms of a one-dimensional system spread
along the perpendicular direction in order to obtain a
suciently low density. Further prospects are the imple-
mentation of novel cooling schemes relying on the local
removal of regions with high entropy [8, 45]. The single-
spin control in our large systems with several hundreds
of atoms also opens new perspectives for scalable quan-
tum computing. Combining single-qubit manipulation
with local readout and a global entanglement operation
in a spin-dependent lattice [46, 47] would be the basis of
a one-way quantum computer [11, 12]. For the circuit
model of a quantum computer, two-qubit operations can
be realised by Rydberg gates between selected atom pairs
in the lattice [48, 49].
Acknowledgements
We thank Wolfgang Ketterle for stimulating discus-
sions and valuable ideas. We acknowledge the help of
Rosa Glockner and Ralf Labouvie during the construc-
tion of the experiment. We acknowledge funding by
MPG, DFG, Stiftung Rheinland-Pfalz f ur Innovation,
Carl-Zeiss Stiftung, EU (NAMEQUAM, AQUTE, Marie
Curie Fellowships to J.F.S. and M.C.), and JSPS (Post-
doctoral Fellowship for Research Abroad to T.F.).
[1] G. Binnig and H. Rohrer, Rev. Mod. Phys. 59, 615
(1987).
[2] F. J. Giessibl, Rev. Mod. Phys. 75, 949 (2003).
[3] R. Blatt and D. Wineland, Nature 453, 1008 (2008).
[4] M. B. Zvonarev, V. V. Cheianov, and T. Giamarchi,
Phys. Rev. Lett. 99, 240404 (2007).
[5] A. Recati, P. O. Fedichev, W. Zwerger, and P. Zoller,
Phys. Rev. Lett. 90, 020401 (2003).
[6] A. Kleine, C. Kollath, I. P. McCulloch, T. Giamarchi,
and U. Schollwock, Phys. Rev. A 77, 013607 (2008).
[7] D. S. Weiss, J. Vala, A. V. Thapliyal, S. Myrgren,
U. Vazirani, and K. B. Whaley, Phys. Rev. A 70,
040302(R) (2004).
[8] J.-S. Bernier, C. Kollath, A. Georges, L. De Leo, F. Ger-
bier, C. Salomon, and M. Kohl, Phys. Rev. A 79,
061601(R) (2009).
7
[9] H. Weimer, M. M uller, I. Lesanovsky, P. Zoller, and H. P.
B uchler, Nature Phys. 6, 382 (2010).
[10] M. A. Nielsen and I. L. Chuang, Quantum Computation
and Quantum Information (Cambridge University Press,
Cambridge, 2000).
[11] R. Raussendorf and H. J. Briegel, Phys. Rev. Lett. 86,
5188 (2001).
[12] H. J. Briegel, D. E. Browne, W. D ur, R. Raussendorf,
and M. Van Den Nest, Nature Phys. 5, 19 (2009).
[13] R. Dumke, M. Volk, T. M uther, F. Buchkremer, G. Birkl,
and W. Ertmer, Phys. Rev. Lett. 89, 097903 (2002).
[14] S. Bergamini, B. Darquie, M. Jones, L. Jacubowiez,
A. Browaeys, and P. Grangier, J. Opt. Soc. Am. B 21,
1889 (2004).
[15] M. Saman, Opt. Lett. 29, 1016 (2004).
[16] T. Calarco, U. Dorner, P. S. Julienne, C. J. Williams,
and P. Zoller, Phys. Rev. A 70, 012306 (2004).
[17] C. Zhang, S. L. Rolston, and S. Das Sarma, Phys. Rev.
A 74, 042316 (2006).
[18] J. Joo, Y. L. Lim, A. Beige, and P. L. Knight, Phys. Rev.
A 74, 042344 (2006).
[19] J. Cho, Phys. Rev. Lett. 99, 020502 (2007).
[20] A. V. Gorshkov, L. Jiang, M. Greiner, P. Zoller, and
M. D. Lukin, Phys. Rev. Lett. 100, 093005 (2008).
[21] N. Lundblad, J. M. Obrecht, I. B. Spielman, and J. V.
Porto, Nature Phys. 5, 575 (2009).
[22] K. Shibata, S. Kato, A. Yamaguchi, S. Uetake, and
Y. Takahashi, Appl. Phys. B 97, 753 (2009).
[23] R. Scheunemann, F. S. Cataliotti, T. W. H ansch, and
M. Weitz, Phys. Rev. A 62, 051801(R) (2000).
[24] D. Schrader, I. Dotsenko, M. Khudaverdyan, Y. Mirosh-
nychenko, A. Rauschenbeutel, and D. Meschede, Phys.
Rev. Lett. 93, 150501 (2004).
[25] M. Karski, L. F orster, J.-M. Choi, A. Steen,
N. Belmechri, W. Alt, D. Meschede, and A. Widera, New
J. Phys. 12, 065027 (2010).
[26] P. W urtz, T. Langen, T. Gericke, A. Koglbauer, and
H. Ott, Phys. Rev. Lett. 103, 080404 (2009).
[27] W. S. Bakr, A. Peng, M. E. Tai, R. Ma, J. Simon, J. I.
Gillen, S. F olling, L. Pollet, and M. Greiner, Science 329,
547 (2010).
[28] J. F. Sherson, C. Weitenberg, M. Endres, M. Cheneau,
I. Bloch, and S. Kuhr, Nature 467, 68 (2010).
[29] N. Schlosser, G. Reymond, I. Protsenko, and P. Grangier,
Nature 411, 1024 (2001).
[30] S. Kuhr, W. Alt, D. Schrader, M. M uller, V. Gomer, and
D. Meschede, Science 293, 278 (2001).
[31] T. Gr unzweig, A. Hilliard, M. McGovern, and M. F. An-
dersen, Nature Phys. 6, 951 (2010).
[32] M. P. A. Fisher, P. B. Weichman, G. Grinstein, and D. S.
Fisher, Phys. Rev. B 40, 546 (1989).
[33] D. Jaksch, C. Bruder, J. I. Cirac, C. Gardiner, and
P. Zoller, Phys. Rev. Lett. 81, 3108 (1998).
[34] M. Greiner, O. Mandel, T. Esslinger, T. W. H ansch, and
I. Bloch, Nature 415, 39 (2002).
[35] S. F olling, A. Widera, T. M uller, F. Gerbier, and
I. Bloch, Phys. Rev. Lett. 97, 060403 (2006).
[36] G. K. Campbell, J. Mun, M. Boyd, P. Medley, A. E.
Leanhardt, L. Marcassa, D. Pritchard, and W. Ketterle,
Science 313, 649 (2006).
[37] H. B. Perets, Y. Lahini, F. Pozzi, M. Sorel, R. Moran-
dotti, and Y. Silberberg, Phys. Rev. Lett. 100, 170506
(2008).
[38] M. Karski, L. F orster, J.-M. Choi, A. Steen, W. Alt,
D. Meschede, and A. Widera, Science 325, 174 (2009).
[39] F. Z ahringer, G. Kirchmair, R. Gerritsma, E. Solano,
R. Blatt, and C. F. Roos, Phys. Rev. Lett. 104, 100503
(2010).
[40] U. Schneider, L. Hackerm uller, J. P. Ronzheimer, S. Will,
S. Braun, T. Best, I. Bloch, E. Demler, S. Mandt,
D. Rasch, et al., arXiv:1005.3545v1 (2010).
[41] K. Winkler, G. Thalhammer, F. Lang, R. Grimm,
J. Hecker Denschlag, A. J. Daley, A. Kantian, H. P.
B uchler, and P. Zoller, Nature 441, 853 (2006).
[42] S. F olling, S. Trotzky, P. Cheinet, M. Feld, R. Saers,
A. Widera, T. M uller, and I. Bloch, Nature 448, 1029
(2007).
[43] A. Peruzzo, M. Lobino, J. C. F. Matthews, N. Matsuda,
A. Politi, K. Poulios, X.-Q. Zhou, Y. Lahini, N. Ismail,
K. W orho, et al., Science 329, 1500 (2010).
[44] A. Micheli, A. J. Daley, D. Jaksch, and P. Zoller, Phys.
Rev. Lett. 93, 140408 (2004).
[45] J. F. Sherson and K. Mlmer, arXiv:1012.1457v1 (2010).
[46] D. Jaksch, H.-J. Briegel, J. I. Cirac, C. W. Gardiner, and
P. Zoller, Phys. Rev. Lett. 82, 1975 (1999).
[47] O. Mandel, M. Greiner, A. Widera, T. Rom, T. H ansch,
and I. Bloch, Nature 425, 937 (2003).
[48] T. Wilk, A. Gaetan, C. Evellin, J. Wolters, Y. Miroshny-
chenko, P. Grangier, and A. Browaeys, Phys. Rev. Lett.
104, 010502 (2010).
[49] L. Isenhower, E. Urban, X. L. Zhang, A. T. Gill,
T. Henage, T. A. Johnson, T. G. Walker, and
M. Saman, Phys. Rev. Lett. 104, 010503 (2010).
[50] M. Garwood and L. DelaBarre, J. Magn. Reson. 153, 155
(2001).
Appendix
Calibration of the addressing beam position
To move the addressing laser beam in the object plane,
we changed the angle of the beam entering from the re-
verse direction into the microscope objective using a two-
axis piezo mirror. The device has an angular resolution
of 5 rad, corresponding to a theoretical resolution in the
object plane of 0.02 a
lat
10 nm. In order to position
the addressing laser beam onto the atoms with high pre-
cision, we measured calibration functions that translate
the two control voltages of the piezo mirror into image
coordinates. This calibration was performed by replacing
the far detuned addressing laser beam by a near resonant
molasses beam that follows the identical beam path. Us-
ing in addition the x and y molasses beams, we took a
uorescence image of a large thermal atom cloud in the
vertical lattice alone and observed a strongly enhanced
signal at the position of the focused beam. We deter-
mined the position of this uorescence maximum with
an uncertainty of 0.2 pixels in our images, corresponding
to 0.05 a
lat
= 25 nm in the object plane. The long term
drifts of the addressing beam position are on the order of
0.1 a
lat
per hour, which we took into account by regular
recalibration of the beam position.
8
Lattice phase feedback
In order to compensate slow phase drifts of the opti-
cal lattice, we applied a feedback on the position of the
addressing beam. We determined the two lattice phases
along x and y after each realisation of the experiment
by tting the position of isolated atoms. Averaging over
the positions of typically 1 5 isolated atoms per image
allowed us to determine the lattice phase to better than
0.05 a
lat
. For the determination of the phase, we used the
lattice constant and the lattice angles determined from a
uorescence image with many isolated atoms [28]. Since
our phase drifts were slower than 0.04 a
lat
between two
successive realisations of the experiment (25 s cycle time),
we used the lattice phase from the last image to correct
the addressing beam position. This feedback was done by
adding appropriate osets to the piezo control voltages.
Microwave sweeps
Our microwave sweeps are HS1-pulses [50] with time-
dependent Rabi frequency (t) and detuning (t) given
by
(t) =
0
sech
_

_
2t
T
p
1
__
(1a)
(t) =

MW
2
tanh
_

_
2t
T
p
1
__
, (1b)
where
0
/(2) = 3 kHz is the maximum Rabi-frequency,
= 5 is a truncation factor, T
p
= 20 ms is the pulse
length, and
MW
/(2) = 60 kHz is the sweep width. The
detuning (t) is measured relative to the center of the
sweep at
MW
=
0

MW
(see Fig. 1b). Here,
0
is the
bare resonance between the two hyperne states, includ-
ing the shift of -570 kHz due to the magnetic bias eld
along the z direction and
MW
/(2) = 75 kHz denotes
the oset of the sweep center.
Spin-ip delity
In order to determine the spin-ip delity, we tted the
hole probability p
0
as a function of the pointing oset x
(see Fig. 3b) to a at-top model function given by
p
0
(x) =
A
2
_
erf
_
x+a/2
s
_
+ erf
_

xa/2
s
__
+B. (2)
Here, erf(x) = 2/

_
x
0
e

2
d is the error function,

a
denotes the full-width at half-maximum of the at-
top prole and
s
the edge sharpness. We chose this
model function since our HS1-pulses (see above) produce
a at-top population transfer prole, the edges of which
are dominated by randomly uctuating quantities (beam
pointing and magnetic elds) following Gaussian statis-
tics. The addressing delity is dened as F = A/(1 B)
taking into account that the maximum hole probability
p
max
0
= A + B also includes holes from thermal defects.
These yield a hole with probability B at unsuccessfully
addressed sites which occur with probability 1 F, such
that p
max
0
= F + (1 F)B.
Single-particle tunneling dynamics
We describe the coherent tunneling dynamics on k =
2n + 1 lattice sites by the Hamiltonian

H
(0)
= J
(0)
n

i =n
_
a
+
i
a
i+1
+ a
+
i
a
i1
_
+V
ext
n

i =n
(i x
os
)
2
a
+
i
a
i
, (3)
where J
(0)
is the tunnel coupling in the lowest band,
a
+
i
( a
i
) is the creation (annihilation) operator for a
particle at site i. The strength of the external har-
monic potential with trapping frequency
trap
is given
by V
ext
=
1
2
m
2
trap
a
2
lat
, and x
os
describes a position o-
set with respect to the bottom of the harmonic potential.
The single particle wave function and its coherent time
evolution are given by

(0)
(t) =
n

i =n
c
(0)
i
(t) a
+
i

0
_
= exp
_
i

H
(0)
t/
_

(0)
(0), (4)
with the initial condition
(0)
(0) = a
+
0

0
_
and the vac-
uum state

0
_
. The resulting probability of nding the
particle at lattice site i after time t is P
(0)
i
(t) = |c
(0)
i
(t)|
2
.
For analyzing the data of Fig. 4, we calculated the time
evolution for k = 17 lattice sites.
Tunneling in the rst excited band
When including tunnelling in the rst band, we assume
an incoherent sum P
tot
(t) of the distributions P
(0)
(t) of
the zeroth and P
(1)
(t) of the rst band as
P
tot
(t) = (1 f)P
(0)
(t) +fP
(1)
(t). (5)
The Hamiltonian

H
(1)
in the rst band and the coher-
ent dynamics are identical to the ones of the zeroth band
(Eqs. 3 and 4), except for a dierent tunnel coupling J
(1)
.
When tting this model to our data, we kept
trap
, x
os
and J
(0)
xed at the values obtained from the data dis-
played in Fig. 4. We extracted J
(1)
from the data of
Fig. 5a and used this value to t the results for other
pointing osets. For the data in Fig. 5, the parameters
of our microwave sweep were such that also neighbour-
ing atoms were addressed. We took this into account by
summing over two distinct probability distributions with
a second starting position in the direction of the pointing
oset.