IOP PUBLISHING JOURNAL OF PHYSICS B: ATOMIC, MOLECULAR AND OPTICAL PHYSICS
J. Phys. B: At. Mol. Opt. Phys. 43 (2010) 051003 (5pp) doi:10.1088/0953-4075/43/5/051003
FAST TRACK COMMUNICATION
Atom chip for BEC interferometry
RJSewell
1,3
,JDingjan
1
, F Baumg
¨
artner
1
, I Llorente-Garc
´
ıa
1
,
S Eriksson
1,4
,EAHinds
1
, G Lewis
2
, P Srinivasan
2
,ZMoktadir
2
,
C O Gollasch
2
and M Kraft
2
1
Centre for Cold Matter, Blackett Laboratory, Imperial College, Prince Consort Road,
London SW7 2BW, UK
2
School of Electronics and Computer Science, University of Southampton, Highfield, Southampton,
SO17 1BJ, UK
E-mail: ed.hinds@imperial.ac.uk
Received 7 January 2010, in final form 20 January 2010
Published 11 February 2010
Online at stacks.iop.org/JPhysB/43/051003
Abstract
We have fabricated and tested an atom chip that operates as a matter wave interferometer. In
this communication we describe the fabrication of the chip by ion-beam milling of gold
evaporated onto a silicon substrate. We present data on the quality of the wires, on the current
density that can be reached in the wires and on the smoothness of the magnetic traps that are
formed. We demonstrate the operation of the interferometer, showing that we can coherently
split and recombine a Bose–Einstein condensate with good phase stability.
Atom chips are microfabricated devices that control electric,
magnetic and optical fields in order to trap and manipulate
cold atom clouds and Bose–Einstein condensates (BECs)
[1–4]. Such devices have significant potential for applications
in sensing, metrology and quantum information processing.
Although BEC was first created on an atom chip 10 years ago
[5], the fabrication of functional devices has posed significant
technical challenges that have only recently been overcome.
In particular, BEC interferometry on an atom chip has been
demonstrated using static magnetic fields in combination with
radio-frequency [6], optical [7] or microwave fields [8].
In this communication we report on the fabrication and
initial testing of a working BEC interferometer. We have
fabricated test batches of atom chips using a variety of
techniques for depositing gold on silicon and for etching
the required wire structures. After evaluating the quality of
these, we settled on a fabrication method using electron beam
evaporation of a gold layer followed by ion-beam milling to
define the wires. We have used one of the atom chips fabricated
in this way to make a BEC interferometer.
The atom chip that we have fabricated is shown in figure 1.
Four parallel Z-shaped wires produce the necessary dc and rf
3
Present address: ICFO–Institut de Ciencies Fotoniques, Mediterranian
Technology Park, 08860 Castelldefels (Barcelona) Spain.
4
Present address: Department of Physics, Swansea University, Singleton
Park, Swansea SA2 8PP, UK.
fields for trapping and manipulating BECs near the surface of
the chip. The wires in the outer pair are 100 μm wide and have
a separation of 300 μm (centre-to-centre). The inner wires are
50 μm wide with 85 μm separation. The central section of
the wires along the z-axis, above which the BEC is produced,
is 7 mm long. Two more wires are patterned onto the chip
parallel to the ends of the Z-wires along the x-axis. These are
used to provide additional axial trap depth and to adjust the
field strength at the trap minimum.
In order to load this chip, cold
87
Rb atoms from a low-
velocity intense source (LVIS) are first captured 4 mm from
the surface in a magneto-optical trap (MOT) [9]. The gold
surface serves as a mirror that reflects some of the laser cooling
light, allowing the MOT to be formed close to the chip. The
atoms are then passed to the magnetic trap, where the cloud is
further cooled by forced evaporation using an rf field to eject
the most energetic atoms. A BEC is formed at approximately
500 nK and this provides the coherent matter wave for our
interferometer. The procedure is similar to that described in
our previous publications [10].
Although atom chips are made using standard
microfabrication techniques, the experimental requirements
impose a number of unusual design constraints. In order to
create magnetic traps with sufficient depth, the wires must be
able to carry several amperes for a period of 10–20 s. At the
same time they should be as small as possible to manipulate
0953-4075/10/051003+05$30.00 1 © 2010 IOP Publishing Ltd Printed in the UK
J. Phys. B: At. Mol. Opt. Phys. 43 (2010) 051003 Fast Track Communication
5mm
100µm
x
z
Figure 1. The atom chip used in our interference experiments. Four
parallel Z-wires occupy the central region of the chip and there are
two additional end wires. The surrounding gold pads form a mirror
surface used in pre-cooling the atoms in a MOT. Inset: an optical
microscope image of the centre of the chip showing the four parallel
trapping wires (in gold). The silicon substrate can be seen in the
gaps between the wires (in grey). The roughness of the silicon
substrate is due to over-etching during the ion-beam milling.
the atoms on a small length scale. This means that they must
be able to carry very large current densities. The wires must
also be very smooth on length scales of up to several hundred
microns in order to minimize transverse currents that lead
to roughness of the magnetic trap and fragmentation of cold
clouds (see below). Finally, in order to facilitate pre-cooling
of the atoms in a MOT, the entire chip surface must have a
mirror finish.
Gold is our material of choice since it has a low resistivity
and high reflectivity at the relevant laser wavelength. We
use a silicon substrate in order to take advantage of mature
fabrication techniques that will allow for future integration
of optical elements such as fibres, cavities, waveguides and
pyramid MOTs into a single device [11–14]. Silicon has
good thermal conductivity, but must be covered in a thin oxide
layer to ensure that the wires are electrically isolated from the
substrate.
The wires can be fabricated using standard lithographic
techniques or by adapting thin film hybrid technology [4, 15].
High-quality wires have been made by applying a lift-off
technique to an evaporated gold layer [16–18]. Evaporated
gold has a good surface finish, and the lift-off technique
gives the wires good edge definition when they are thin.
However, the method is not well developed for films as
thick as ours. Other groups have patterned a thin gold
film photolithographically, then grown thick wires by
electrochemical deposition [19–21]. Thicker wires can be
fabricated using this technique, but the homogeneity of the
gold and definition of the edges are not as good. In our
previous work we have studied techniques for fabricating
gold wires using sputtered and electrochemically deposited
gold layers that are subsequently patterned by wet-etching or
ion-beam milling [22]. We have also investigated growing
wires by electrochemical deposition into a mould formed by
a lithographically patterned photoresist [23]. In the light of
all these studies, we have fabricated our BEC interferometer
SiO
2
Au
Cr
Si
Photoresist
wet oxidation
evaporation of Cr seed laye
r
evaporation of Au
spinning of photoresist
reflow of resist
UV photolithography
ion beam milling
plasma ash
Figure 2. Fabrication process flow for electron beam evaporation of
a thick gold film followed by ion-beam milling.
chip using a thick evaporated gold layer patterned by ion-beam
milling.
We use a 4 inch p-doped [1 0 0] silicon wafer with a
resistivity of 17–30 cm to produce 16 atom chips. The wafer
is cleaned using the standard Radio Corporation of America
procedures and fuming nitric acid. It then undergoes wet
oxidation to produce a 100 nm thick oxide layer, the first of
several steps illustrated in figure 2. A 40 nm adhesion layer of
chromium is then evaporated over the whole surface, followed
by 3 μm of gold deposited in five steps of 600 nm to avoid
overheating the evaporator. After cleaning in fuming nitric
acid, a 2.2 μm thick layer of HPR504 photoresist is spun onto
the gold at 500 rpm for 10 s, followed by 30 s at 1500 rpm.
This is given a soft bake at 90
◦
C for 120 s, then it is patterned
by UV lithography using a Karl Suss MA8 machine for 9 s
at 6.5mWcm
−2
. Finally, a hard bake is done for 30 min at
140
◦
C so that the resist will be easier to remove after it has
been subjected to ion-beam milling. This also causes the resist
to develop sloping sides as it reflows a little.
Milling is done on an IONFAB 300+, with 388 V of beam
voltage, 200 mA of current and 276 V of accelerating voltage.
The wafer is cooled to a temperature of 21
◦
C using helium and
is milled for 50 min, resulting in a maximum cutting depth of
4.4 μm. The resist is quite hard to remove after exposure to
the ions, despite the hard bake, so we use a plasma asher for
this purpose run at 110
◦
C with 600 W for 60 min. Once all the
resist has been removed the wafer is cleaned in fuming nitric
acid. The etch rate is not uniform across the wafer, resulting
in over-etching in some places. Where the etch is too deep,
the mill can go through the oxide layer and into the silicon
substrate itself. In that case, re-deposited silicon on the side
walls of the cut makes an electrical short to the wafer. This
debris is removed by a 5 s buffered HF acid dip (7:1) followed
2
J. Phys. B: At. Mol. Opt. Phys. 43 (2010) 051003 Fast Track Communication
Top
Side
(a)
(b)
(c)
Figure 3. SEM images of the gold wires fabricated by e-beam
evaporation followed by ion-beam milling. Image (a), wire cleaved
through the middle to reveal sloping sides. (b) Top surface of the
wire. (c) View facing the sloping side wall of the wire.
by a 5 min KOH etch. Finally, we use a diamond scriber to
cleave the wafer into 16 separate atom chips 24 mm wide and
26 mm long.
Cleaving a chip through the middle allowed us to examine
the cross-sectional profiles of the wires using a scanning
electron microscope (SEM). Figure 3(a) shows the sloping
side walls of the wire, transferred from the resist to the wire by
erosion of the resist during the milling process. One also sees
that the milling was too deep on this wire and penetrated into
the silicon. Figure 3(b) shows an SEM image of the surface of
one of the gold wires, which was found using an atomic force
microscope to have 3 nm RMS roughness. Figure 3(c) shows
an SEM image of the sloping side wall of one of the gold wires.
Some grain structure is evident on the μm scale, but there is
no sign of any layering due to the multi-stage evaporation.
The surface and wire edge are smooth on the scale of this
image.
The maximum usable current density in the chip wires
follows from the temperature rise due to resistive power
dissipation and is limited by thermal conduction. The
insulating SiO
2
layer is the main barrier to heat flow. When
current is turned on, the wire temperature rises rapidly
over some microseconds until the drop across this layer
saturates. Thereafter, the wire temperature rises more slowly,
as determined by thermal conduction into the silicon substrate
and on into the mounting structure, made of oxygen-free
copper embedded in a Shapal-M (AlN) base plate, connected
to an 8 inch stainless steel vacuum flange. The mounting
structure acts as a heat sink.
The wires were tested by passing current through them and
using the change in resistance to monitor the slow temperature
rise. Taking an increase of 150
◦
C (50% increase in resistivity)
as a reasonable working upper limit, we measured maximum
current densities of 8.8 × 10
9
Am
−2
in the 50 μm wide wires
and 6.1 × 10
9
Am
−2
in the 100 μm wide wires with current
pulses 10–20 s long and with the atom chip in vacuum.
6.4 K
1.9 K
0.5
µ
K
500
µ
m
0
100
200
300
(b)
(c)
0 500 1000 1500 2000 2500 3000 3500
10
0
10
n(μm
−1
)
δB
z
(mG)
z
(
μm
)
(a)
µ
µ
Figure 4. Cold atom studies of the chip. (a) Absorption images of
successively colder clouds taken after turning off the trap and
accelerating the clouds away from the chip for 3 ms. At ∼2–3 μK
the cloud begins to sense roughness of the trapping potential, and
the wing to the left becomes distinct from the main cloud. A BEC
begins to form in the largest of these lumps at 500 nK. (b) Linear
number density n(z) of the 1.9 μK cloud. (c) Inferred deviation of
trap from a smooth harmonic potential (ω
z
= 2π × 6.5 Hz),
expressed as the magnetic field δB
z
that causes it.
Atoms are loaded into the chip by passing them from the
MOT to a magnetic trap at a height y 150 μm above one
of the wires. This is formed by passing 2 A through the wire,
with a bias field of B
x
= 24.8 G. We then cool the atoms to
the BEC transition by forced rf evaporation. The absorption
images in figure 4(a) show the cloud at several temperatures
near the end of the evaporation process. At 6 μK the cloud
is roughly 1 mm long and exhibits an extended wing on the
left-hand side. With further cooling down to 1.9 μK that wing
becomes a clearly separate cloud, due to a subsidiary minimum
in the axial potential. As described in [24], we can derive
the variation of the potential from the density distribution of
these atoms, shown in figure 4(b). The result is illustrated in
figure 4(c). This roughness is due to transverse currents δI
x
(z),
which generate fields δB
z
(z) parallel to the wire. Since the
bottom of the trapping potential is set by B
z
(z), these fields
make the trap rough [17, 21, 24, 25]. The angular variation
of the current can be estimated from the ratio of the noise
field to the main field, which is approximately ±10
−4
. Since
the variation takes place over typically ±200 μm, the centre
of the wire need only deviate by 20 nm over this length to
cause the effect that we see. It seems probable that this is due
to slight variations in the edges of the wire, though it could
also be due to minor defects in the homogeneity of the gold.
It would be interesting to see if this could be improved by
omitting the reflow of the resist to achieve better definition
of the edges. The magnitude of the potential variation at this
distance is similar to that seen in electroplated wires of similar
dimensions and larger than that reported in evaporated wires
patterned using a lift-off technique [17].
In order to split the matter wave with our atom chip,
we alter the potential by adding near resonant rf fields as
proposed by [26, 27] and demonstrated by [6, 28, 29]. The
experimental arrangement is shown in figure 5(a). Two wires
3
J. Phys. B: At. Mol. Opt. Phys. 43 (2010) 051003 Fast Track Communication
0 50 100 150
x
µ
m
Number Densitya.u.
(a)
(c)
(b)
(d)
RF Field
DC Trapping Field
External Bias Field
Au wires
Si substrate
DC - RF curr
ents
DC + RF curr
ents
x
y
x
y
y
23°
90° 90°
RelativePhase
Probability Density
Figure 5. (a) Configuration of static and rf fields used to split the
BEC coherently. (b) Absorption image of the atomic cloud showing
interference fringes formed when the trapping potential is turned off
and the two arms of the BEC interferometer are allowed to overlap
in free fall. (c) The relative phase is obtained by fitting a modulated
Gaussian (solid line) to a slice through the centre of the interference
pattern. (d) Histogram of the relative phases extracted from 103
repetitions of the experiment. The solid line is a fit to the data using
a normal distribution. The standard deviation is φ =±23
◦
.
carrying parallel dc currents form a 2D quadrupole with the
help of the bias field. We evaporate to BEC in this trap and
continue the evaporation until no discernible thermal atoms
remain, at which point the BEC has ∼1.5 × 10
4
atoms and
a chemical potential μ = h × 3 kHz. The addition of rf
currents, 180
◦
out of phase, generates a near-resonant rf field
along y that splits the cigar-shaped cloud into two parallel
clouds. The trap can be smoothly deformed from a single to a
double well by ramping the intensity and/or frequency of the
rf. A typical double well used in our interference experiments
has a separation of ∼4 μm between the two trap minima and
a barrier height of ∼10 kHz.
After allowing the two parts to evolve separately for
approximately 1 ms, we read out the relative phase between
them by turning off the trapping potential and allowing them
to overlap in free fall. We then take an absorption image of
the density distribution, which exhibits interference fringes
perpendicular to the splitting axis, as illustrated in figure 5(b).
We analyse the pattern by fitting a modulated Gaussian
n(x) = g(x)
1+α cos
2πx
+ φ
to a slice through the centre,
as shown in figure 5(c), to determine the relative phase φ.In
figure 5(d) we plot a histogram of the phases extracted from
103 repetitions of the experiment. The standard deviation of
these is ±23
◦
, indicating that the splitting produces a well-
defined initial relative phase between the two arms of the
interferometer, as is required for a useful measuring device.
This phase spread is similar to that reported by Schumm et al
[25] for similar experimental parameters and evolution time.
In conclusion, we have fabricated an atom chip by a
process involving electron beam evaporation of a thick gold
layer on a silicon substrate followed by ion-beam milling. The
resulting wires are able to carry high-density dc and rf currents
and are sufficiently smooth and uniform to trap a cold atom
cloud close to the surface of the chip. We have used one of
these atom chips to make a working BEC interferometer with
good phase stability.
Acknowledgments
The authors acknowledge the expert technical assistance of
Jon Dyne. This w
ork was supported by the UK EPSRC, by the
Royal Society and by the European Commission through the
SCALA and AtomChips networks.
References
[1] Reichel J, H
¨
ansel W and H
¨
ansch T W 1999 Phys. Rev. Lett.
83 3398–401
[2] Hinds E A and Hughes I G 2002 J. Phys. D: Appl. Phys.
32 R119–46
[3] Folman R, Kr
¨
uger P, Schmiedmayer J, Denschlag J
and Henkel C 2002 Adv. At. Mol. Opt. Phys. 48 263–356
[4] Fortagh J and Zimmermann C 2007 Rev. Mod. Phys.
79 235–89
[5] H
¨
ansel W, Hommelhoff P, H
¨
ansch T W and Reichel J 2000
Nature 413 498–501
[6] Schumm T, Hofferberth S, Andersson L M, Wildermuth S,
Groth S, Bar-Joseph I, Schmiedmayer J and Kr
¨
uger P 2005
Nature Phys. 1 57–62
[7] Wang Y-J, Anderson D Z, Bright V M, Cornell E A, Diot Q,
Kishimoto T, Prentiss M, Saravanan R A, Segal S
and Saijun W 2005 Phys. Rev. Lett. 94 090405
[8] B
¨
ohl P, Reidel M F, Hoffrogge J, Reichel J, H
¨
ansch T W
and Treutlein P 2009 Nature Phys. 5 592–97
[9] Sinclair C D J, Curtis E A, Llorente Garc
´
ıa I, Retter J A,
Hall B V, Eriksson S, Sauer B E and Hinds E A 2005
Phys. Rev. A 72 031603
[10] Jones M P A, Vale C J, Sahagun Sanchez D, Hall B V
and Hinds E A 2003 Phys. Rev. Lett. 91 080401
[11] Eriksson S et al 2005 Eur. Phys. J. D 35 135–9
[12] Trupke M et al 2006 Appl. Phys. Lett. 88 071116
[13] Trupke M, Goldwin J, Darqui
´
e B, Dutier G, Eriksson S,
Ashmore J P and Hinds E A 2007 Phys. Rev. Lett.
99 063601
[14] Pollock S, Cotter J P, Laliotis A and Hinds E A 2009
Opt. Express 17 14109–14
[15] Reichel J 2002 Appl. Phys. B 75 469–87
[16] Groth S, Kr
¨
uger P, Wildermuth S, Folman R, Fernholz T,
Schmiedmayer J, Mahalu D and Bar-Joseph I 2004
Appl. Phys. Lett. 85 2980–2
[17] Kr
¨
uger P, Andersson L M, Wildermuth S, Hofferberth S,
Haller A, Aigner S, Groth S, Bar-Joseph I
and Schmiedmayer J 2007 Phys. Rev. A 76 063621
[18] Trinker M, Groth S, Haslinger S, Manz S, Betz T, Schneider S,
Bar-Joseph I, Schumm T and Schmiedmayer J 2008 Appl.
Phys. Lett. 92 254102
[19] Ott H, Fortagh J, Schlotterbeck G, Grossmann A
and Zimmermann C 2001 Phys. Rev. Lett. 87 230401
[20] Lev B 2003 Fabrication of micro-magnetic traps for cold
neutral atoms arXiv:quant-ph/0305067v1
[21] Est
`
eve J, Aussibal C, Schumm T, Figl C, Mailly D,
Bouchoule I, Westbrook C I and Aspect A 2004 Phys. Rev.
A 70 043629
[22] Koukharenko E, Moktadir Z, Kraft M, Abdelsalam M E,
Bagnall D M, Vale C, Jones M P A and Hinds E A 2004
Sensors Actuators A 115 600–7
[23] Lewis G, Moktadir Z, Gollasch C, Kraft M, Trupke M,
Eriksson S and Hinds E A 2005 Proceedings of MME
Conference (Sweden, 2005) pp 56–9
4
J. Phys. B: At. Mol. Opt. Phys. 43 (2010) 051003 Fast Track Communication
[24] Jones M, Vale C, Sahagun-Sanchez D, Hall B, Eberlein C,
Sauer B, Furusawa K, Richardson D and Hinds E A 2004
J. Phys. B: At. Mol. Opt. Phys. 37 L15–20
[25] Schumm T, Est
`
eve J, Figl C, Trebbia J-B,
Aussibal C, Nguyen H, Mailly D, Bouchoule I,
Westbrook C I and Aspect A 2005 Eur. Phys. J.
D 32 171–80
[26] Zobay O and Garraway B 2001 Phys. Rev. Lett. 86 1195–98
[27] Zobay O and Garraway B 2004 Phys. Rev. A 69 023605
[28] Colombe Y, Knyazchyan E, Morizot O, Mercier B, Lorent V
and Perrin H 2007 Europhys. Lett. 67 593–9
[29] Jo G-B, Shin Y, Will S, Pasquini T A, Saba M, Ketterle W,
Pritchard D E, Vengalattore M and Prentiss M 2007 Phys.
Rev. Lett. 98 030407
5
