Atom chip for BEC interferometry
R. J. Sewell, J. Dingjan, F. Baumgartner, I. Llorente-Garcia, S. Eriksson, E. A. Hinds, G. Lewis, P. Srinivasan, Z. Moktadir, C. O. Gollasch, M. Kraft
RRAPID COMMUNICATION
Atom Chip for BEC Interferometry
R J Sewell ‡ , J Dingjan , F Baumg¨artner , I Llorente-Garc´ıa ,S Eriksson § , E A Hinds ,G Lewis , P Srinivasan , Z Moktadir ,C O Gollasch , and M Kraft Centre for Cold Matter, Blackett Laboratory, Imperial College, Prince ConsortRoad, London SW7 2BW, United Kingdom School of Electronics and Computer Science, University of Southampton, Highfield,Southampton, SO17 1BJ,United KingdomE-mail: [email protected]
Abstract.
We have fabricated and tested an atom chip that operates as a matterwave interferometer. In this communication we describe the fabrication of the chip byion-beam milling of gold evaporated onto a silicon substrate. We present data on thequality of the wires, on the current density that can be reached in the wires and on thesmoothness of the magnetic traps that are formed. We demonstrate the operation ofthe interferometer, showing that we can coherently split and recombine a BoseEinsteincondensate with good phase stability.PACS numbers: 03.75.Dg,37.25.+k,81.16.Nd,37.10.Gh
Submitted to:
Journal of Physics B: Atomic, Molecular and Optical Physics ‡ Present address: ICFO-Institut de Ciencies Fotoniques, Mediterranean Technology Park, 08860Castelldefels (Barcelona), Spain. § Present address: Department of Physics, Swansea University, Singleton Park, Swansea SA2 8PP,United Kingdom. a r X i v : . [ qu a n t - ph ] F e b tom Chip for BEC Interferometry µm x z Figure 1.
The atom chip used in our interference experiments. Four parallel Z-wiresoccupy the central region of the chip and there are two additional end wires. Thesurrounding 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 paralleltrapping wires (in gold). The silicon substrate can be seen in the gaps between thewires (in grey). The roughness of the silicon substrate is due to over-etching duringthe ion-beam milling.
The atom chip that we have fabricated is shown in figure 1. Four parallel Z-shapedwires produce the necessary dc and rf fields for trapping and manipulating BECs nearthe surface of the chip. The wires in the outer pair are 100 µ m wide and have a separationof 300 µ m (centre-to-centre). The inner wires are 50 µ m wide with 85 µ m 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 theZ-wires along the x-axis. These are used to provide additional axial trap depth and toadjust the field strength at the trap minimum.In order to load this chip, cold 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 tom Chip for BEC Interferometry µ m tom Chip for BEC Interferometry SiO Au CrSiPhotoresist wet oxidationevaporation of Cr seed layerevaporation of Auspinning of photoresistreflow of resistUV photolithographyion beam millingplasma ash
Figure 2.
Fabrication process flow for electron beam evaporation of a thick gold filmfollowed by ion-beam milling. of gold deposited in five steps of 600 nm to avoid overheating the evaporator. Aftercleaning in fuming nitric acid, a 2 . µ m thick layer of HPR504 photoresist is spun ontothe gold at 500 rpm for 10 seconds, followed by 30 seconds at 1500 rpm. This is givena soft bake at 90 ◦ C for 120 seconds, then it is patterned by UV lithography using aKarl Suss MA8 machine for 9 seconds at 6 . − . Finally, a hard bake is done for30 min at 140 ◦ C so that the resist will be easier to remove after it has been subjectedto ion-beam milling. This also causes the resist to develop sloping sides as it reflows alittle.Milling is done on an IONFAB 300+, with 388 V of beam voltage, 200 mA of currentand 276 V of accelerating voltage. The wafer is cooled to a temperature of 21 ◦ C usinghelium and is milled for 50 minutes, resulting in a maximum cutting depth of 4 . µ m.The resist is quite hard to remove after exposure to the ions, despite the hard bake, sowe use a plasma asher for this purpose run at 110 ◦ C with 600 W for 60 minutes. Onceall the resist has been removed the wafer is cleaned in fuming nitric acid. The etch rateis not uniform across the wafer, resulting in over-etching in some places. Where the etchis 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 shortto the wafer. This debris is removed by a 5 second buffered HF acid dip (7:1) followedby a 5 minute KOH etch. Finally, we use a diamond scriber to cleave the wafer into 16separate atom chips 24 mm wide and 26 mm long.Cleaving a chip through the middle allowed us to examine the cross sectional profilesof the wires using a scanning electron microscope (SEM). Figure 3(a) shows the sloping tom Chip for BEC Interferometry Top
Side (a)(b) (c)
Figure 3.
SEM images of the gold wires fabricated by e-beam evaporation followed byion-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. side walls of the wire, transferred from the resist to the wire by erosion of the resistduring the milling process. One also sees that the milling was too deep on this wire andpenetrated into the silicon. Figure 3(b) shows an SEM image of the surface of one ofthe gold wires, which was found using an atomic force microscope to have 3 nm RMSroughness. Figure 3(c) shows an SEM image of the sloping side wall of one of the goldwires. Some grain structure is evident on the µ m scale, but there is no sign of anylayering due to the multi-stage evaporation. The surface and wire edge are smooth onthe scale of this image.The maximum usable current density in the chip wires follows from the temperaturerise due to resistive power dissipation and is limited by thermal conduction. Theinsulating SiO 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 thislayer saturates. Thereafter, the wire temperature rises more slowly, as determined bythermal conduction into the silicon substrate and on into the mounting structure, madeof oxygen-free copper embedded in a Shapal-M (AlN) base plate, connected to an 8 inchstainless steel vacuum flange. The mounting structure acts as a heat sink.The wires were tested by passing current through them and using the change inresistance to monitor the slow temperature rise. Taking an increase of 150 ◦ C (50%increase in resistivity) as a reasonable working upper limit, we measured maximumcurrent densities of 8 . × A m − in the 50 µ m wide wires and 6 . × A m − in the100 µ m wide wires with current pulses ten to twenty seconds long and with the atomchip in vacuum.Atoms are loaded into the chip by passing them from the MOT to a magnetictrap at a height y (cid:39) µ m above one of the wires. This is formed by passing 2 Athrough the wire, with a bias field of B x = 24 . tom Chip for BEC Interferometry Μ K1.9 Μ K0.5 Μ K Μ m (b)(c) n ( µ m − ) δ B z ( m G ) z ( µ m ) (a) Figure 4.
Cold atom studies of the chip. (a) Absorption images of successivelycolder clouds taken after turning off the trap and accelerating the clouds away fromthe chip for 3 ms. At ∼ µ K the cloud begins to sense roughness of the trappingpotential, and the wing to the left becomes distinct from the main cloud. A BECbegins to form in the largest of these lumps at 500 nK. (b) Linear number density n ( z )of the 1 . µ K cloud. (c) Inferred deviation of trap from a smooth harmonic potential( ω z = 2 π × . δB z that causes it. at several temperatures near the end of the evaporation process. At 6 µ K the cloud isroughly 1 mm long and exhibits an extended wing in the left hand side. With furthercooling down to 1 . µ K that wing becomes a clearly separate cloud, due to a subsidiaryminimum in the axial potential. As described in [24], we can derive the variation of thepotential from the density distribution of these atoms, shown in figure 4(b). The resultis illustrated in figure 4(c). This roughness is due to transverse currents δI x ( z ), whichgenerate fields δB z ( z ) parallel to the wire. Since the bottom of the trapping potentialis set by B z ( z ), these fields make the trap rough [21, 25, 17, 24]. The angular variationof the current can be estimated from the ratio of the noise field to the main field, whichis approximately ± − . Since the variation takes place over typically ± µ m, thecentre of the wire need only deviate by 20 nm over this length to cause the effect thatwe 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 wouldbe interesting to see if this could be improved by omitting the reflow of the resist toachieve better definition of the edges. The magnitude of the potential variation at thisdistance is similar to that seen in electroplated wires of similar dimensions and largerthan 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 byadding 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 carrying parallel dccurrents form a 2D quadrupole with the help of the bias field. We evaporate to BECin this trap and continue the evaporation until no discernible thermal atoms remain, atwhich point the BEC has ∼ . × atoms and a chemical potential µ = h × tom Chip for BEC Interferometry Μ m N u m b er D e n s i t y a . u . (a)(c) (b)(d) RF Field DC Trapping FieldExternal Bias Field
Au wiresSi substrateDC - RF currentsDC + RF currents xy x y x y Φ Φ Φ P r o b a b ili t y D e n s i t y 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 thetrapping potential is turned off and the two arms of the BEC interferometer are allowedto 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 ofthe relative phases extracted from 103 repetitions of the experiment. The solid line isa fit to the data using a normal distribution. The standard deviation is ∆ φ = ± ◦ . 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 smoothlydeformed from a single to a double well by ramping the intensity and/or frequency ofthe rf. A typical double well used in our interference experiments has a separation of ∼ µ m between the two trap minima, and a barrier height of ∼
10 kHz.After allowing the two parts to evolve separately for approximately 1 ms, weread out the relative phase between them by turning off the trapping potential andallowing them to overlap in free fall. We then take an absorption image of the densitydistribution, which exhibits interference fringes perpendicular to the splitting axis, asillustrated in figure 5(b). We analyse the pattern by fitting a modulated gaussian n ( x ) = g ( x ) (cid:0) α cos (cid:0) πx Λ + φ (cid:1)(cid:1) 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 phasesextracted from 103 repetitions of the experiment. The standard deviation of these is ± ◦ , indicating that the splitting produces a well-defined initial relative phase betweenthe two arms of the interferometer, as is required for a useful measuring device. Thisphase spread is similar to that reported by Schumm et al. [25] for similar experimentalparameters and evolution time.In conclusion, we have fabricated an atom chip by a process involving electron beamevaporation 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 sufficientlysmooth and uniform to trap a cold atom cloud close to the surface of the chip. We haveused one of these atom chips to make a working BEC interferometer with good phase tom Chip for BEC Interferometry Acknowledgements
The authors acknowledge the expert technical assistance of Jon Dyne. This work wassupported by the UK EPSRC, by the Royal Society, and by the European Commissionthrough the SCALA and AtomChips networks.
References [1] Reichel J, H¨ansel W and H¨ansch T W 1999
Phys. Rev. Lett. J. Phys. D R119–46[3] Folman R, Kr¨uger P, Schmiedmayer J, Denschlag J and Henkel C 2002
Adv. At. Mol. Opt. Phy. Rev. Mod. Phys. Nature
Nature Physics Phys. Rev. Lett. Nat. Phys. Phys. Rev. A Phys. Rev. Lett. Eur. Phys. J. D Appl. Phys. Lett. Phys.Rev. Lett. Opt. Express Appl. Phys. B Appl. Phys. Lett. Phys. Rev. A Appl. Phys. Lett. Phys. Rev. Lett. 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 A2004
Phys. Rev. A Sens. Actuators, A
ProceedingsMME 2005 Conference (Sweden)
J. Phys. B L15–20 tom Chip for BEC Interferometry [25] Schumm T, Est`eve J, Figl C, Trebbia J-B, Aussibal C, Nguyen H, Mailly D, Bouchoule I, WestbrookC I and Aspect A 2005 Eur. Phys. J. D Phys. Rev. Lett. Phys. Rev. A Europhys. Lett. Phys. Rev. Lett.98