1364 IEEE JOURNAL OF QUANTUM ELECTRONICS, VOL. 36, NO. 12, DECEMBER 2000

Atoms and Wires: Toward Atom ChipsMarkus Bartenstein, Donatella Cassettari, Tommaso Calarco, Alexander Chenet, Ron Folman, Karolina Brugger,Albrecht Haase, Eugen Hartungen, Björn Hessmo, Alexander Kasper, Peter Krüger, Thomas Maier, Fritz Payr,

Stephan Schneider, and Jörg Schmiedmayer

Abstract—Atoms can be trapped and guided using nanofabri-cated wires on surfaces, achieving the scales required by quantuminformation proposals. Theseatom chipsform the basis for robustand widespread application of cold atoms ranging from atom op-tics to fundamental questions in mesoscopic physics, and possiblyquantum information systems.

Index Terms—Cold atoms, manipulation, miniaturisation, trap-ping.

I. INTRODUCTION

I N MESOSCOPIC quantum electronics, electrons moveinsidesemiconductor structures and are manipulated using

potentials where at least one dimension is comparable tothe de-Broglie wavelength of the electrons [1], [2]. In suchpotentials, the energy-level spacing is large compared to thetransverse particle temperature and hence excitations aresuppressed and coherent transport is possible. Similar poten-tials can be created for neutral atoms moving micronsabovesurfaces, using nanofabricated charged and current-caryingstructures [3]–[5]. Surfaces carrying such structures formatomchips which, for coherent matter wave optics, may form thebasis for a variety of novel applications and research tools,similar to what integrated circuits are for electronics.

Recent achievements in atom cooling and trapping, as well asrealized novel applications of atom optics including sensors andclocks, have made the above vision both feasible and attractive.More so, recent theoretical proposals regarding quantum com-puting and communications have made the idea of achievingcoherent manipulation of neutral atoms even more appealing.Finally, it is expected that such progress would involve newinsight into fundamental issues such as decoherence. Conse-

Manuscript received February 28, 2000; revised June 19, 2000. This work wassupported by the Austrian Science Foundation (FWF) under Project S065-05and SFB F15-07, by the Jubiläums Fonds der Österreichischen Nationalbankunder Project 6400, and by the European Union under Contract TMRX-CT96-0002. The work of T. Calarco was supported by the Instituto Trentino Di Cultura(ECT). The work of B. Hessmo was supported by the Svenska Institutet.

M. Bartenstein, A. Haase, and P. Krüger are with the Institut für Experimen-talphysik, Universität Innsbruck, A-6020 Innsbruck, Austria, and also with theInstitut für Experimentalphysik, Freie Universität Berlin, D-14195 Berlin, Ger-many.

D. Cassettari, A. Chenet, R. Folman, K. Brugger, E. Hartungen, A. Kasper,F. Payr, S. Schneider, and J. Schmiedmayer are with the Institut für Experimen-talphysik (ECT), Universität Innsbruck, A-6020 Innsbruck, Austria.

T. Calarco is with the Institut für Theoretische Physik, Universität Innsbruck,A-6020 Innsbruck, Austria, and also with the Instituto Trentino Di Cultura(ECT), Villa Tambosi, I-38050 Villazzano, Italy.

B. Hessmo is with the Institut für Experimentalphysik, Universität Innsbruck,A-6020 Innsbruck, Austria, and also with the Department of Quantum Chem-istry, Uppsala University, S-75120 Uppsala, Sweden.

T. Maier is with Institut für Festkörperelektronik, A-1040 Wien, Austria.Publisher Item Identifier S 0018-9197(00)10556-1.

quently, basic research also motivates this theoretical and ex-perimental effort.

In this work, we make use of the magnetic interactionbased on the coupling of the atomic magnetic moment

to the magnetic field to trap and manipulate atoms close to thesurface of anatom chip. The trapping potentials are created bysuperposing a homogeneous magnetic bias field with the fieldgenerated by thin current-carying wires. The trap depth is givenby the homogeneous field, the gradients, and curvature by themagnetic fields from the wire [4], [6].

We have previously reported on the manipulation of neu-tral atoms using thin (down to below 1-m) charged wires [7]and current-carying wires (down to 25m) to form guides [6],[8]–[10], beam splitters [4], and “Z”- or “U”-shaped 3-D traps[11]. These structures were free standing.

The next step was to turn to surface-mounted wires [12]–[14],which was recently achieved for large structures [15]–[18].However, the full potential in surface-mounted atom optics liesin the robust miniaturization down to the mesoscopic scale.Aside from the above-mentioned need for coherent transportand manipulation, such a move is primarily motivated by thetheoretically required scale needed to achieve entanglementwith neutral atoms, through controlled collisions [19]–[21] orcavity QED [22], [23], entanglement being the basic buildingblock for quantum information devices. For a review ofquantum computing in general, see for example, [24]–[29].

Recently [30], we presented such a nanofabricated devicewith which the required ground state size of less than 100 nmwas achieved. This is a first step toward our vision, the realiza-tion of a fully integratedatom chip. Here, we discuss the de-tails of that vision and elaborate on recent experimental real-izations. First, we describe the theoretical foundation of atommagnetic and electric trapping followed by examples from pre-vious experimental studies. This is followed by a review of pos-sible loss mechanisms for trapped atoms. Next, we describe therecent experiments done with free-standing and miniature non-free-standing traps, namely with theatom chip. We start by de-scribing the chip and the experimental setup, followed by a pre-sentation of the results. Finally, we discuss potential applica-tions and future perspectives on the way to the full realizationof theatom chipvision.

II. M AGNETIC AND ELECTRIC TRAPPING

Microscopic potentials for a neutral atom can be con-structed by using [3], [31]: 1) the electric interaction betweena neutral polarizable atom and a charged nanostructure

and 2) the magnetic interaction betweenthe atoms magnetic moment and a magnetic field ,

0018–9197/00$10.00 © 2000 IEEE

BARTENSTEINet al.: ATOMS AND WIRES: TOWARDS ATOM CHIPS 1365

. A variety of novel atom optical elements fortrapping and guiding can then be constructed at the microscopicscale, e.g., quantum wells, quantum wires, and quantum dotsfor neutral atoms.

There are various magnetic and electric atom optical el-ements with which neutral atoms can be manipulated [5],[33]–[37], e.g., atom mirrors produced by the repulsion ofweak field-seekingatoms from the exponentially decayingmagnetic field produced by alternating magnetic fields on asurface. Such alternating fields could and have been producedby static magnets, magnetic registering on tape and computerdisks, and by alternating currents in a grating of parallel wires.Repulsing atoms by an exponentially decaying field has alsobeen achieved with evanescent blue detuned light coming outof a prism or an optical fiber. Another example would be theattraction of polarizable atoms to charges. Combining theattractive and repulsive potentials just described would leadto the formation of a guide (in the case of a charged wireon a mirror) or a 3-D trap (in the case of a charged dot on amirror) [3], [5]. Fig. 1 presents a beam splitter based on thelatter guide [3]. However, as theatom chipwe describe in thiswork makes use of only the magnetic interaction produced by acurrent-carying wire, we will in the following restrict ourselvesto the description of the latter. The interested reader may findextensive details regarding other possible optical elements, inthe long list of works referenced throughout this paper.

Magnetic manipulation and trapping [33]–[37] of neutral par-ticles is a well-developed tool for experiments with cold atomsand has been used in diverse applications such as Atom Op-tics (for an overview, see [38] and [39]), the study of ultra-coldcollisions [40], and the creation of Bose–Einstein Condensates[41]–[45] (for a complete list of references, see [46] and [47]).An atom with a magnetic moment experiences the potential

. In general, the vector coupling results in avery complicated motion for the atom. However, in most cases,the Larmor precession of the magnetic moment is much fasterthan the apparent change in the direction of the magnetic fieldin the rest frame of the moving atom, and hence, an adiabaticapproximation can be applied. The magnetic moment then fol-lows adiabatically the direction of the field and the atoms canbe described as moving in a scalar potential depending only onthe modulus .

Since the Earnshaw theorem forbids a local maximum of themagnetic modulus in free space [48] (the Earnshaw theoremcan be generalized to any combination of electric, magneticand gravitational field, as shown in [49]), the most commonmagnetic traps areweak field-seekingtraps. In these traps, thereis a minimum of the field which forms the center of the trap,to which the atoms are attracted.Strong field-seekingatomsmay also be trapped in special circumstances. For example,in Fig. 2(a), we present Kepler-like trapping where the atomswhich have the right angular momentum orbit around theattractive potential of a wire.

In order to evaluate the relevant quantity when super-posing two magnetic fields, one has to consider both the mag-nitude of the fields and their direction. This gives an additionaldegree of freedom for creating the trapping potential by sepa-rating the structures responsible for the trap depth and the field

Fig. 1. A repulsive atom mirror and an attractive charged wire add to form a2-D trap, namely, a guide. Splitting such a guide allows for the formation of abeam splitter.

gradient. This possibility of creating the trapping potentials bysuperposing different magnetic fields leads to the constructionof microscopic traps and guides with high field gradients andcurvature, or similarly, with a high trap frequency, which in turnmeans small trap sizes ( ) and large energy-level spacing( ). The latter two are simply a consequence of the harmonicoscillator approximation, in which and

.For example, the trap can be built by superposing a homoge-

neous field, which is easy to achieve over a large volume, withthe field generated by a thin current-carrying wire. Such a trap isalso presented in Fig. 2(b). The minimum of the combined fieldwill be located where the homogeneous bias field and the fieldof the wire have the same magnitude but opposite directions.Since magnetic fields change at a length scale characteristic ofthe distance from the current, the typical size of the trap is thengiven by the distance from the wire.

Such traps and guides have interesting properties: for a givenhomogeneous bias field , the trap depthis determined by themagnitudeof the bias field, while the trapsize and its distance from the wire are proportional to the cur-rent through the wire ( and are the projection of thetotal atom angular momentum and the hyperfine structure levelspecific constant, respectively. is the Bohr magneton.) At thecenter of the trap, the magnetic field gradient scales within both transverse directions. Here, the apparent paradoxical sit-uation arises where the trap gets steeper fordecreasingcurrentsin the wire. Having the trap depth fixed by the bias field, a com-pression of the trap can be accomplished bydecreasingthe cur-rent in the wire and an expansion byincreasingthe current inthe wire. This allows to control separately the distance of thetrap from the wire and its steepness.

This use of the superposing of fields can be applied to anymagnetic device; for example. a micro-magnet in a homoge-neous field would show similar characteristics [4], [6], [50].

Because of the similarity to quantum electronics (see, for ex-ample, [51]), we call such structures that provide 1-D confine-ment, i.e., atoms are allowed to move freely along the surface,a quantum well. Structures with two dimensional confinement,

1366 IEEE JOURNAL OF QUANTUM ELECTRONICS, VOL. 36, NO. 12, DECEMBER 2000

Fig. 2. (a) Trappingstrong field-seekingatoms around a current-carying wire. (b) The design principle of a micro trap by superposing a homogeneous field withthe field of a microscopic structure, for the trapping ofweak field-seekers.

i.e., a microscopic atom guide, are referred to as aquantum wire.Finally, 3-D confinement is referred to as aquantum dot.

We will now briefly discuss some options for matter-waveoptics structures close to current-carying wires.

The simplest way to demonstrate these design principles istheside guide[4], [6], [52], formed by adding the constant biasfield in a direction orthogonal to the wire. The bias fieldcancels exactly the circular magnetic field of the wire along aline parallel to the wire at a distance .Around this line, the modulus of the magnetic field increases inall directions and forms a tube with a magnetic field minimum inits center. Atoms in theweak field-seekingstate can be trappedin this 2-D quadrupole field and guided along the side of thewire, hence the nameside guide[Fig. 2(b)].

When mounting the wire on a surface the bias field has to beparallel to the surface. The side-guide potential is then above thesurface, and atoms are guided along the wire. It is interesting tonote that the bias field for the side guide can be formed by twoadditional wires on each side of the guiding wire, each carryinga current in the opposite direction to the guiding wire. This isespecially interesting when the wires can be mounted on a chip,and thus produce a self-contained chip (i.e., no external field isneeded).

Examples of typical guiding parameters are given in Table I.For example, trap frequencies above 100 kHz can be achievedwith a moderate current of about 100 mA and a magnetic offsetfield of 40 G. The guided atoms are then located a fewmabove the surface. These tight traps can obtain gradients ex-ceeding G/cm and curvatures larger than G/cm .

Such guides on anatom chipwith external and on-board biasfields will be reviewed in the next sections.

TABLE ITYPICAL POTENTIAL PARAMETERS FORQUANTUM WIRES. (a) SIDE

GUIDE CREATED BY A THIN CURRENT-CARRYING WIRE MOUNTED ON A

SURFACE WITH AN ADDED BIAS FIELD PARALLEL TO THE SURFACE BUT

ORTHOGONAL TO THEWIRE. (b) TWO-WIRE GUIDE CREATED BY TWO THIN

CURRENT-CARRYING WIRE MOUNTED ON A SURFACE WITH AN ADDED

BIAS FIELD ORTHOGONAL TO THE PLANE OF THE WIRES (THE TWO

WIRES ARE10-�m APART (SEE ALSO FIG. 3)

(a)

(b)

A different way to create a guide is by using two wires with op-posite currents and a bias field orthogonal to the plane in whichthe wires are located. This has the clear advantage that the wireis allowed to take any direction (or turn) on the surface plane

BARTENSTEINet al.: ATOMS AND WIRES: TOWARDS ATOM CHIPS 1367

Fig. 3. Top row: One- and two-wire guides. The two-wire guide configurationin which the bias field is perpendicular to the surface (opposing currents), hasthe advantage of not being limited to a certain direction on the chip (i.e., can take“turns”). Bottom row: Three and four-wire guides in which there is no need foran external bias field (see also [14]).

(a)

(b)

Fig. 4. (a) Wire layout and the potential for a Y-shaped beam splitter. The startof the additional guide is visible in the second potential plot. (b) A two-wirebeamsplitter creating a pure Y-shaped beam splitting potential.

of the chip, whereas a one-wire guide is limited to the directionorthogonal to the bias field which is parallel to the chip surface.This is presented in Fig. 3 (top row). One- and two-wire guidesmay also have an on-board bias field by introducing two addi-tional wires. These self-contained three- and four-wire guidesare presented in Fig. 3 (bottom row).

These wire guides can be easily combined to build more com-plicated atom optical elements for guided atoms. The simplestof these are beam splitters and interferometers. These can thenbe constructed using both the one- and the two-wire guides. Letus briefly elaborate.

The simplest beam splitter can be formed by a “Y”-shapedwire and a bias field parallel to the plane of the “Y,” orthogonalto the incoming wire [see Fig. 4(a)]. Atoms are then guidedabove the wire. By sending a current through one arm of the

(a)

(b)

Fig. 5. Wire layout and the potentials for one- and two-wire 4-port beamsplitters, for a bias field (a) parallel and (b) perpendicular to the plane of theatom chip.

wire, the atoms will then be guided along this arm. By sendingthe current through both arms, the atomic wave function can besplit and atoms are guided in both outgoing guides.

Such a beam splitter, though simple, has its disadvantages.Specifically, the potentials of the in- and out-going guides arequite different in location and compression, which may causereflections. Such reflections introduce efficiency losses and maycause additional interference signals when used in interferome-ters and networks. In addition, as there are only three active ports(a fourth port goes in the direction of the surface and acts as an-other loss mechanism), one cannot use a simple beam-splitterformulation [53] to describe its operation. For such a device,one should resort to a description of a scattering process off asymmetric barrier. The advantage of such a scattering process isthat it is symmetric, and is thus inherently a 50%/50% splitter,where by inherent, we mean that it is independent of bias fieldand current magnitude. Here, another limitation should be men-tioned: by introducing a small bias field along the axis of thebeam splitter in order to turn the guides into a Ioffe configura-tion (without Majorana spin flips—see Section III), the abovesymmetry may be hindered. However, more advanced config-urations [see Fig. 4(b)] where the bias field is vertical to thesurface do not present this problem when the extra Ioffe field isgenerated by the tilting of the bias field. This advanced configu-ration also solves the problem of the reflection described above.

A better geometry is a beam splitter made from two wirescoming close to each other, but not touching (see Fig. 5). Whenthe wires are separated by a distance large compared to the dis-tance of the guide from the wire, this configuration gives twoseparate guides. When the wires approach each other, the twoguides come closer. Depending on the bias field and the distance

1368 IEEE JOURNAL OF QUANTUM ELECTRONICS, VOL. 36, NO. 12, DECEMBER 2000

between the wires, either a tunneling barrier or a single guideforms. When the wires separate after the closest approach, againtwo distinct guides form. In this beam splitter, the splitting ratioand other parameters can be fine tuned by the magnitude and thedirection of the bias field. Aside from the advantage of constantheight and compression in the potential well (as there is no cur-rent splitting), this beam splitter, via the barrier, is able to intro-duce phase changes between the transmitted and reflected beam,and thus give a close analog to a true Mach–Zehnder interfer-ometer. However, relative to the “Y” configuration, it is compli-cated in the sense that fine tuning is needed in order to achievethe exact barrier height and width to obtain the 50%/50% split-ting [53]. This beam splitter is thus quite sensitive to noise.

We recently performed experiments on such guides andbeam splitters, although not yet in the quantum regime. Theguides were either Kepler-like guides (strong field-seekingatoms) or side-guides (weak field-seekingatoms) [4], [52]. Thebeamsplitter consisted of two free-standing wires that werecombined to form a “Y.” By controlling the current throughthe arms of the “Y”-shaped wire we could send cold Li atomsalong either of the two arms of the “Y” or split the beam intotwo [4], [54]. The above is presented in Fig. 6.

Finally, 3-D traps may be formed in various ways. Forexample, applying a homogeneous magnetic field in a directionopposing the field coming out of a magnetic tip [50]. Thiswould create a zero of the field only in a well-localized area. Asall these schemes have the same principles as those describedabove, we will not elaborate further, but will refer the interestedreader to the many referenced works throughout this paper. Wewill, however, describe in the following a trivial way of movingfrom 2-D traps (i.e., guides) to 3-D ones by simply cutting offthe guide at its ends with a potential wall that may be formed,for example, by perpendicular current-carying wires. This willbe explained and demonstrated in Section IV.

III. L OSSMECHANISMS

For atom chips to work, two main destructive elementshave to be put under control. The first is decoherence; namely,the interaction of the atoms with the uncontrolled environ-ment through radiation. Although neutral atoms have beenconsidered as favorable quantum systems over other possiblecandidates such as ions (for heating rates in ion traps see [55]and [56]), because of their small coupling to the environment,they still might have a strong enough coupling on the shorttime-scales needed to possibly render them inadequate. Suchcoupling could for example cause spin flips into untrappedstates (see, for example, [57] and [58]) or cause the dephasingof the coherent internal state of the atom. Taking into accountthe proximity of the atoms to surfaces, this problem has to bethoroughly studied together with theoretical models of quantumerror correction. Indeed surface heating might pose a limiton how close to the surface coherent manipulation of neutralatoms in guides and traps is possible. Still, much research hasto be done in this area, which we will not deal with here. Thesecond destructive element which has to be put under controlis trap loss mechanisms that have to do with the trap itselfor with particle collisions. It is obviously crucial that we are

Fig. 6. Wire guides and traps for atoms. (a) Kepler guide and side guide fordifferent wire currents. (b) Beamsplitter for guided atoms, using Kepler guides.

able to keep the atoms inside the trap for as long as needed.In the following, we describe some of the more common lossmechanisms in their usual order of dominance.

a) Heating: Heating of atoms due to compression is obvi-ously a very effective loss mechanism once the tempera-ture of the atoms rises above the trap depth. The lowestheating is produced if the compression is adiabatic. Insuch a case, atoms are able to follow the rising energyof their specific level. If the compression is not in all di-mensions, one should also require a sufficient collisionrate so that heat is transferred also to the noncompressedaxis. However, even in the limit of adiabatic heating, thetemperature of the atom gas rises proportionally to therise in the trap frequency. It should be noted that if only afraction of the initial number of atoms is finally needed,one can select atoms such that only the cold tail of thevelocity distribution stays in the trap, and hence main-tains the original pre-heating temperatures. One may dothis by lowering the bias field, and hence the trap depth(see previous section), by opening the trap for a brieftime, or even by bringing the trap closer to the surface(e.g., by increasing the bias field or decreasing the cur-rent) thereby “crashing” the hotter atoms, which occupythe outer regions of the trap, onto the surface. As long asthe traps are extremely asymmetric (e.g., only 2-D tightconfinement with transverse level spacing more thentimes larger then the longitudinal level spacing) the phasespace density of a magneto-optical trap (MOT) wouldallow for reasonable population of the transverse groundstate. One may also use more “traditional” methods, suchas radio-frequency knives to open holes in the trap. Inaddition, allowing time for sufficient re-thermalizationthrough elastic collisions, would turn the above selectioncooling into the much more effective evaporative cooling.

BARTENSTEINet al.: ATOMS AND WIRES: TOWARDS ATOM CHIPS 1369

Fig. 7. Creating wire traps: the left column shows the geometry, the currentsand the bias fields. The right column shows the radial trapping potentials, and in(b) and (c) also the axial potentials. (a) The side guide: adding a homogeneousmagnetic field (B ) perpendicular to a current-carying wire (I ) creates a 2-Dquadrupole guide along the wire. (b) A “U”-shaped wire results in confinementin the axial direction also. The field configuration is similar to a 3-D quadrupolefield with a zero in the trapping center. (c) For a “Z”-shaped wire a Ioffe-Prichardtype trap is obtained.

We will not elaborate further, and simply refer the inter-ested reader to the many available references [41]–[47].

b) Background collisions: Here, collisions between back-ground gas atoms and trapped atoms endow the latterwith sufficient energy to escape the trap. It is, therefore,clear that vacuum requirements will become more strin-gent as the need to suppress loss mechanisms becomesmore extreme.

c) Spin flip[59]–[61]: In this so-called Majorana process, theatom’s magnetic moment does not manage to follow therapid change in the direction of the magnetic field. Conse-quently, the atom’s magnetic moment, which was parallelto the magnetic field (the latter for the case of , andopposite for ), and hence was aweak field-seekingatom, may now find itself anti-parallel to the magneticfield, and hence astrong field-seekerwhich, of course, isa nontrapped state. The criterion to be obeyed if the atomis to avoid such a spin flip is , where is theLarmor precession frequency andis the trap vibrationalfrequency. As the former is proportional to the magnitudeof the magnetic field, it is clear that trap configurations, inwhich a zero exists in the field minimum, will give rise tolarge rates of Majorana spin flips. For some trap configu-rations (e.g., a one-wire guide), this is easily resolved byadding a bias field component along the guide axis (e.g.,by slightly rotating the main bias field from being exactlyperpendicular to the guide axis—see previous section). Inother cases, such as for the “U”-trap (see following sec-tion), such a solution would not work.

d) Spin exchange[62]–[64]: This process in which the hy-perfine spin projection is conserved while the spinitself is not, is caused by 2-body inelastic collisions. Obvi-ously, the rate per atom of these collisions is proportionalto the density of the atom gas, and hence the total rate perunit volume is proportional to the square of the density.For example, a collision between the two trapped states ofthe kind, could lead to the emergenceof two untrapped states of the kind .One should note that an excess of energy is needed forthe above to be feasible. One should also note that polar-ized samples (e.g., consisting only of )cannot undergo spin exchange if the only other availablelevel is of a lower , as such a move will not be able toconserve .

e) Spin relaxation[40]: A weaker process, also resultingfrom 2-body inelastic collisions, is that of spin relax-ation, where is not conserved. Here, following thecollision, an atom may stay in his original state,but his state may change value. For example, a

state is a trappedweak field-seekingstate, while is not (as , whichcontributes to the sign of the potential, is for the

state).f) 3-body recombination[62]–[65]: Here, we refer to the

process in which two atoms combine to form a molecule.As these molecules do not have a definite magnetic mo-ment, they are lost from the trap. The third atom, essentialfor taking away the excess energy, may also now be ener-getic enough to escape the trap.

g) Tunneling: Here we are referring to the tunneling of atomsout of the local minimum of the trap. Obviously, trapsin which the magnetic field magnitude rises for long dis-tances will have very little tunneling.

Next, we describe recent experiments that have been done inInnsbruck with the 3-D trapping and manipulation of neutralatoms by magnetic fields.

IV. FREE-STANDING TRAPS

As mentioned in the Introduction, we have previouslyreported on various manipulations of neutral atoms usingboth the electric and magnetic interactions, in free-standingwires. These structures included among other thingsstrongfield-seekingatom guiding,weak field-seekingatom guiding,and beam splitters. In the following, we present as an exampleof free-standing atom manipulation, two types of 3-D magnetictraps that we have realized; the “U” and the “Z” configurations.

These traps are different from the straight wire configurationswhich form guides, simply by the fact that the wire is bentat some points. The magnetic field from the bent leads creates“endcaps” for the wire guide which confine the atoms along thewire [4], [11], [16], [30]. The size of the trap along this axis isthen given by the distance between the endcaps.

1) Bending the wire into a “U” shape [Fig. 7(b)] creates amagnetic field, which rotates in direction over ineach plane parallel to the “U.” Superposing a homoge-neous bias field one can always find a point where the two

1370 IEEE JOURNAL OF QUANTUM ELECTRONICS, VOL. 36, NO. 12, DECEMBER 2000

Fig. 8. The magnetic field of a “Z”-trap in planes through the potentialminimum. (a) The plane is orthogonal to the central wire (the strong direction).(b) The plane contains the central wire (the weak direction). The graphs showthe direction of the magnetic field vectors and contours for the equipotentialcurves for the absolute value of the magnetic field. The inserts show typicalCCD pictures of trapped atoms taken from the two sides. The shape of theatom cloud resembles closely the equipotential lines.

magnetic fields cancel exactly. Consequently, there is al-ways a zero point in the trapping potential. If the bias fieldis parallel to the bent leads, one obtains a 3-D quadrupoletrap. The minimum of the “U” trap is displaced from thecentral point of the bar, in a direction opposite to the bentwire leads.

2) One can break the rotation of the magnetic fieldvectors by bending the wire ends into opposite direc-tions forming a “Z” [Fig. 7(c)]. The magnetic field vec-tors in a plane parallel to the “Z” rotate by less thenand then back. Consequently, one can find directions ofthe external bias field where there are no zeros in thetrapping potential. This is the case when the bias fieldis parallel with the leads. This configuration creates anIoffe–Pritchard-type trap. The axial components createdby the magnetic field of the two leads, bent in oppositedirections, add up at the center and the potential min-imum is not zero. This additional feature of the “Z” trapwill prevent the atoms from making Majorana transitionswhen trapped (see Section III). The field configurationsare shown in Fig. 8.

Next, we describe the specific procedure and details withwhich we achieved a “Z”-trap. We investigated its propertieswith cold Li atoms from a MOT1 [66]. The MOT is loadedfrom a thermal beam of Li atoms coming out of an oven anddecelerated by a 100-MHz detuned laser beam (“slower”). Theschematics of our experimental setup are described in the nextsection, with the difference that here we use six MOT beamsrather than the four used there. For a more detailed description ofthe apparatus and the atom-trapping procedure, see [4], [6], [52].

1The atoms are loaded into the MOT for 20 s out of an effusive beam. The redlaser has a detuning of 25 MHz and a total laser power of about 150 mW. Anelectro-optic modulator produces sidebands of 812 MHz (30%) one of whichis used as a repumper. To increase the loading rate we use an additional slowerbeam (20 mW, 100 MHz red detuned) directed through the MOT into the oven.

Alkali atoms are usually used in atom-trapping experiments, asatom cooling techniques have been mainly perfected for alkalaiatoms, and as their recombination into untrapped molecules issmall (see Section III).

The wire trap is made using a 1-mm thick standard copperwire bent into a “Z” shape, as shown in Fig. 7(b). The centralbar is 6-mm long, and the two end wires 25 mm. The wire cancarry 25 A without sign of heating.

The MOT is located a few millimeters above the “Z” to avoidlosses caused by the shadows of the wire in the trapping beams[7]. In this configuration, we load atoms into the MOT as de-termined by a calibrated photodiode. The atoms are transferredto the “Z”-trap in the following procedure:

After loading the MOT, the slower beam is shut off and thefrequency and intensity of the trapping lasers are changed inorder to control the size and temperature of the atom cloud. Typ-ically, the atom cloud is 1 mm in diameter (FWHM) and has atemperature of 200K. An external magnetic field simultane-ously moves the center of the MOT quadrupole and shifts theatoms to the position where they are loaded into the “Z” trap.This shifting and cooling process is approximately 20-ms long.

The atoms are then released from the MOT by switching offthe laser light, the external MOT quadrupole, and the shiftingfields within 0.5 ms. At this point, the current through thewire (typically 20 A) and the magnetic bias field (typically 20G) are switched on within 100s. The magnetic trap now holdsatoms in weak field-seeking states. In our experiment, we couldtransfer up to 25% of the unpolarized sample of atoms in theMOT into the magnetic trap. During the trapping time in the“Z” potential, the atoms can be manipulated by changing thecurrent through the wire and the magnetic bias field. Finally,after a given trapping time, the spatial distribution of the atomsis measured by imaging the fluorescence using CCD cameraswith the laser beams switched on for a short period of time (ms).

We have demonstrated that a planar “Z”-shaped wire andan external bias field can be used to magnetically trap neutralatoms. The trap shows interesting scaling laws; it is easy to gen-erate very high field gradients, either by decreasing the currentin the wire or by increasing the bias field. As explained earlier,because the trap width is proportional to , the gradient to

and the trap depth to , these values can be changedseparately. A typical observation of these scaling laws is pre-sented in Fig. 9. We obtained trap parameters of other magnetictraps used in BEC experiments. A further compelling feature ofthe “Z” trap is that with higher gradients the trap center movesclose to the wire, which allows a miniaturization of the trap. Asthe planar structure favors mounting on a surface, the “Z” trapis perfectly suitable to transfer ultra-cold magnetically trappedatoms to other structures microfabricated on the same surface.

We conclude this section by noting once again thatfree-standing structures have been thoroughly examined. Theyhave, indeed, shown that the compelling features of magnetictrapping, such as simplicity and favorable scaling laws, doindeed exist. In order to further miniaturize these structures,the next step was to move from free-standing structures tosurface-mounted structures—the latter being the subject of thenext section.

BARTENSTEINet al.: ATOMS AND WIRES: TOWARDS ATOM CHIPS 1371

V. SURFACE TRAPS

The next obvious step in order to achieve miniaturization wasto turn to nanofabricated surface-mounted structures [12], [13].The motivation for miniaturization is many-fold.

1) Mesoscopic potentials—As mentioned in the Introduc-tion, potentials with sizes comparable or smaller than theparticle de-Broglie wavelength would exhibit coherenttransportation properties.

2) Quantum information processing—Recent theoreticalproposals regarding quantum computing suggest thatsmall ground state sizes are also required for the re-alization of entanglement procedures, such as throughatom–atom collisions where the small ground state sizeensures a large collision rate, or through cavity QED,where the Lamb–Dicke limit, i.e., small ground state size(compared to the light-wavelength) is required so thatthe phase which the atom “sees” is uniform throughoutits wave function. Also, it goes without saying that thesmaller things get, the faster they can operate.

3) Decoherence times—Tighter traps mean larger en-ergy-level spacing and reduced chance that environ-mental noise can induce unwanted excitations.

4) Atom addressability—A small ground state size simplyimplies that you know to a much better extent where theatom is and can therefore address it (e.g., with laser light)by high resolution means. Consequently, close-by atomscan be addressed individually.

5) Additional motivations—Additional reasons for minia-turization also exist, such as a smaller energy consump-tion, working within the Lamb–Dicke limit for coolingwithout exciting the atom into higher trap vibrational de-grees of freedom, scalability, etc.

Using surface nanofabrication is motivated in numerous waysas well.

1) Robustness—Free-standing structures, as those describedin the previous section, are extremely delicate.

2) Production—It is harder and harder to accurately and re-liably produce small structures. Existing nanofabricationtechnology allows to make use of a rich and well estab-lished production procedure.

3) Additional motivations—Efficient cooling of the wires,etc.

We now move on to describe our most recent experiment inwhich atoms have been manipulated in a surface-mounted mag-netic trap. The chip we have used in this work is made of a2.5- m gold layer placed on a 600-m thick GaAs substrate.2

The gold layer is patterned using nanofabrication technology.The scale limit of the process used is well below 100 nm.

In Fig. 10(a), we present the main elements of the chip designused in the work described here. Each of the large “U”-shapedwires, together with a bias field, creates a quadrupole field,which may be used to form a MOT on the chip as well as amagnetic trap. Both “U”-shaped wires together may be used to

2The chip is produced using standard nano-fabrication methods. A detaíledaccount will be published in a forthcoming paper by T. Maier.

Fig. 9. The cloud of trapped atoms monitored by the camera looking at thestrong, radial direction of the trap. By changing the (a) bias field or (b) wirecurrent, the trap size and position changes. In sequence (a), the trap frequenciesincrease from 30 to 1600 Hz.

(a)

(b)

Fig. 10. (a) A schematic of the chip surface design. For simplicity, only wiresused in the experiment are shown. The wide wires are 200-�m wide while thethin wires are 10-�m wide. The insert shows an electron microscope image ofthe surface and its 10-�m wide etchings defining the wires. (b) The mountedchip before it is introduced into the vacuum chamber.

form a strong magnetic trap in order to “load” atoms into thesmaller structures, or as an on-board (i.e., without need for ex-ternal coils) bias field, for guides and traps created by the thinwire running between them. The thin wires are 10-m wide and,depending on the contact used, may form a “U”- or “Z”-shapedmagnetic trap or a magnetic guide. The chip wires are all de-fined by boundaries of 10-m wide etchings in which the con-

1372 IEEE JOURNAL OF QUANTUM ELECTRONICS, VOL. 36, NO. 12, DECEMBER 2000

ductive gold has been removed. This leaves the chip as a goldmirror (with 10 m etchings) and it can thus be used to reflectthe laser beams for the MOT during the cooling and collectingof atoms. Fig. 10(b) presents the mounted chip before it is in-troduced into the vacuum chamber. In addition, a “U”-shaped1-mm thick wire, capable of carrying up to 20 A of current, hasbeen put underneath the chip in order to assist with the loadingof the chip. Its location and shape are identical to those of one ofthe 200 m “U”-shaped wires and it differs only in the amountof current it can carry.

The chip assembly [Fig. 10(b)] is then mounted inside avacuum chamber used for atom trapping experiments, withoptical access for the laser beams and the observation camerasand with the possibility of applying the desired bias fields(Fig. 11).

The experimental procedure for loading cold atoms into thesmall traps on the chip is the following [16], [30]. In the firststep, typically Li atoms are loaded from an effusive atomicbeam into a MOT, as described in the previous section. Becausethe atoms have to be collected a few millimeters away from asurface, we use a “reflection” MOT [67]. Thereby, the six laserbeams needed for the MOT are formed from fourbeams by re-flecting two of them off the chip surface (Fig. 11). Hence, atomsabove the chip actually encounter six light beams. To assure acorrect magnetic field configuration needed for the formation ofa MOT, one of the reflected light beams has to be in the axis ofthe MOT coils. Fig. 12(a) shows a top view of the chip and thereflection MOT sitting above the “U”-shaped wires.

The large external quadrupole coils are then switched offwhile the current in the “U”-shaped wire underneath the chipis switched on (up to 16 A), together with an external biasfield (8 G). This forms a nearly identical, but spatially smaller,quadrupole field as compared to the fields of the large coils.The atoms are thus transferred to a secondary MOT which byconstruction is always well aligned with the chip [Fig. 12(b)].By changing the bias field, the MOT can be shifted close to thechip surface (typically, 2 mm). The laser power and detuningare changed to further cool the atoms, giving us a sample witha temperature below 200K.

In the next step, the laser beams are switched off and thequadrupole field serves as a magnetic trap in which theweakfield-seekingatoms are attracted to the minimum of the field.Without the difficulties of near surface shadows hindering theMOT, the magnetic trap can now be lowered further towardthe surface of the chip [Fig. 12(c)]. This is simply done by in-creasing the bias field. Atoms are now close enough so that theycan be trapped by the chip fields (see Fig. 13). The loading ofthe chip has begun.

Next, 2 A are sent through each of the two 200-m “U”-shaped wires on the chip, and the current in the “U”-shapedwire located underneath the chip is ramped down to zero. Thisprocedure brings the atoms even closer to the chip, compressesthe trap considerably, and transfers the atoms to a magnetictrap formed by the currents in the chip. The distances of theatoms from the surface are now typically a few hundred microns[Fig. 12(d)].

Finally, the 10- m wire trap is loaded in much the sameway. It first receives a current of 300 mA. Then the current in

Fig. 11. Experimental setup: four circularly polarized light beams enter thechamber; two are counter propagating parallel to the surface of the chip, whilethe two others, impinging on the surface of the chip at 45 degrees, are reflectedby the gold layer. The chip, the underlying “U”-wire trap, and the bias field,are oriented in such a manner as to provide a quadrupole field with the sameorientations as the MOT coils. The oven, the effusive beam, and the two camerasobserving the atomic cloud are also shown.

both the “U”-shaped wires is ramped down to zero (Fig. 14).Atoms are now typically a few tens of microns above the sur-face [Fig. 12(e)].

These guides and traps can be further compressed by rampingup the bias magnetic field. In this process, we typically achievegradients of kG/cm. By applying a bias field of 40 G anda current of 200 mA in the 10-m wire, we achieve trap pa-rameters with a transverse ground state size below 100 nm andfrequencies of above 100 kHz (as required by the quantum com-putation proposals [20], [21]). We remind the reader that as thede-Broglie wavelength is of the order of 0.1–1m for tempera-tures of the order of 100–1K, respectively, the ground statesizes achieved are adequate also for the purpose of coherenttransportation, as long as the initial MOT temperature could bemaintained on theatom chip.

By running the current through a longer 10-m section of thethin wire, we turn the magnetic trap into a guide, and atomscould be observed expanding along it [Fig. 12(f)].

In an additional experiment, we used the thick wires on thechip to create anon chipbias field for the trapping. In the exper-iment, this is done by sending current through the two “U”-trapsin the opposite direction, with respect to the current in the 10-mwire, which creates a magnetic field parallel to the chip surface.Hence, we demonstrate trapping of atoms on a self-containedchip.

In these small traps, the atom gas can be compressed to thepoint where direct visual observation is difficult. In such a case,we observe those atoms after guiding or trapping by “pulling”them up from the surface into a less compressed wire trap (byincreasing the wire current or decreasing the bias field).

During the transfer from the large magnetic trap to the small10- m trap, the density of the atomic cloud is increased by up toa factor 350. As the trap is compressed, the temperature of theatoms rises, and if in this case the trapping potential is not deepenough, atoms are lost. In our case, the trap depth is uniquelydetermined by the bias field used, which leads to depths

ranging between MHz ( mK) for the8-G bias field and to MHz ( mK) for the 50-Gbias field and . This adiabatic heating and the finite

BARTENSTEINet al.: ATOMS AND WIRES: TOWARDS ATOM CHIPS 1373

Fig. 12. Experiments with anatom chip. Column (i): The view from the top (camera 1). Column (ii): The front view (camera 2). Column (iii): A schematic ofthe wire configuration. Current-carrying wires are highlighted in red. The front view shows two images: the upper is the actual atom cloud and the lower is thereflection on the gold surface of the chip. The distance between both images is an indication of the distance of the atoms from the chip surface. The pictures of themagnetically trapped atomic cloud are obtained by fluorescence imaging using a short laser pulse (typically 0.5 ms). (a)–(f) The various steps of theexperiments.(a)–(d) The step-wise process of loading atoms onto the chip. (e) and (f) Atoms in a microscopic trap and propagating in a guide.

1374 IEEE JOURNAL OF QUANTUM ELECTRONICS, VOL. 36, NO. 12, DECEMBER 2000

trap depth limited the transfer efficiency for atoms from the largemagnetic trap into the smallest chip trap to50%.

Since we use a trapped atomic sample consisting of three dif-ferent spin states ( , ,and ), the large compression also increasesthe rate for inelastic two body spin exchange collisions dramat-ically (see Section III). This rate is for our Li sample similar tothe elastic collision rate [68]–[70] (a description of the-waveestimate approximation can be found, e.g, in [71]) and is, there-fore, a good estimate of the achievable collision rates in a po-larized sample. From measured decay curves, we estimate thecollision rate to be on the order of 20 for atoms in a typicalsmall chip trap. This estimate of the scattering rate in the smallchip traps is supported by the observation that the atoms expandvery fast into the wire guide, indicating that energy gained fromthe transverse compression of the trap is transformed efficientlyinto longitudinal velocity at a very high rate.

VI. SUMMARY OF RESULTS

The above shows that the concept of anatom chipclearlyworks. We have demonstrated that a wide variety of magneticpotentials may be realized with simple wires on surfaces. Wirestogether with a bias field can produce quadrupole fields for aMOT, 3-D minima for trapping, and 2-D minima for guiding.Furthermore it is very easy to manipulate the center of the trapand its width. We have shown that loading such an atom trap10- m above the surface does not present a major problem, andtrap parameters with a transverse ground state size below 100nm, and frequencies of above 100 kHz have been achieved. Inaddition, we could trap atoms exclusively with the chip fields,creating the required bias fields “on board.” Last, but not least,it has been shown that standard nanofabrication techniques andmaterials may be utilized to build theseatom chips. The wires onthe surface can stand sufficiently high current densities (A/cm ) in vacuum and at room temperature. Together with thescaling laws of these traps [4], [6], [11], this will allow us touse much thinner wires and reach traps with ground state sizesof 10 nm and trap frequencies in the megahertz range (see alsoTable I). This will ensure that not only the particle temperatureis low with respect to the energy-level spacing, but also the ex-citation probability due to environmental heating is suppressed[57], [58]. These tight traps will also enable to study quantumgases in low dimensions [72]. Finally, we add that the splittingof an atom beam on the surface has also been recently achieved[73].

VII. V ISION AND OUTLOOK

We would like to conclude with a long-term outlook.Atom optics has proven itself to be an extremely successful

tool for research into the foundations of quantum theory con-cerning topics such as matter-wave interference, entanglementand the uncertainty principle, and for the development of tech-nological applications such as clocks and acceleration sensors.Most of the present experiments are put together as a macro-scopic apparatus from single-atom optical elements, in a waysimilar to the first electronic devices which were built from partsperforming single tasks.

Fig. 13. Reciprocal distance of the double-U trap from the chip surface versusbias field at constant wire current 2 A. The points are experimental valuesdeduced from the images while the solid line is calculated from the potential.The error on the experimental points comes from the resolution of the camera.

Integrating many elements to control atoms onto a singledevice, anatom chipwill make atom physics experiments muchmore robust and simple. Neutral-atom manipulation usingintegrated micro-devices will combine the best of two worlds:the ability to use cold atoms—a well-controllable quantumsystem (for example as a qubit), and the immense technologicalcapabilities of nanofabrication and microelectronics to buildhighly complex experiments for manipulating the atoms. Thiswill allow much more complicated tasks in atom manipulationto be performed in a way similar to integration of electronicelements allows the developement of new powerful devices.

This is of special interest, since it has been suggested thatthe high degree of control achieved over neutral atoms and theirweak coupling to the environment (long decoherence time) willallow the realization of quantum information processing (QIP)[20], [21], [24]–[29].

In this work, we have successfully realized the first steps ofmany still needed in this direction. A final integratedatom chipshould have many more components integrated on it.

On the atom optical side, this will include miniaturized trapsand guides for neutral atoms by various combinations of magne-tized, current-carying, and charged microstructures. To achievethis goal, we will need to adapt standard nanofabrication tech-niques and develop new methods and materials. This will makeavailable a toolbox which allows for atom manipulation at dis-tances of order 1m above a surface in structures similar to thequantum wells, quantum wires and quantum dots of mesoscopicquantum electronics.

Another integratedatom chipcomponent will include a reli-able source of cold atoms with an efficient loading mechanism.As we have demonstrated in our experiments, traps with param-eters similar to BEC traps can be built on chips. This shouldallow the on-chip creation of a BEC or degenerate Fermi gas asa reservoir of cold atoms for the experiments.

From the reservoir, atoms will be transported coherently tothe designed experiments. Therefore, another component willneed to be single mode guides or movable potentials for con-trolled atom transport. Ideally, one wants routine single-atomstate-selective loading, preparation, and manipulation. For in-stance, by adiabatically turning on a periodic potential within a

BARTENSTEINet al.: ATOMS AND WIRES: TOWARDS ATOM CHIPS 1375

condensate, it is possible to achieve a phase transition from thesuperfluid BEC to a Mott insulator phase with regular filling ofthe sites of the periodic potential, as already proposed for thecase of optical lattices [74].

In a further advance, one can implement state-dependent con-trollable potentials, whose shape can be adiabatically changedor suddenly switched to different configurations to induce col-lisions between atoms in a controlled way in order to create en-tanglement between their internal states [20], [21]. This wouldbe a first important step toward the implementation of QIP pro-posals usingatom chips.

A limiting factor for theatom chipdesign and theatom chipexperiments will be the coherence and decoherence propertiesof neutral atoms in these micro traps and guides. The specificinfluence of the nearbyhot surfaces containing the trappingstructures will be very important. This will limit how close onecan get to the surface, and thereby how small the traps andguides may be made, and consequently, what potential steep-ness may be reached. It will be interesting to identify the mech-anisms through which decoherence limits QIP in atom opticalmicrostructures, as well as strategies to fight decoherence it-self, e.g., by encoding the quantum information into states be-longing to subspaces decoupled from the environment (i.e., de-coherence-free).

Another component will be the integration of light optics ontheatom chip. For example, light guides and micro optics canbe fabricated on theatom chipto bring laser light to the atomsfor manipulation of internal and external degrees of freedom.This can be done using evanescent waves, or direct focusing oflight with high numerical aperture micro optics onto the atomstrapped in the micro traps. Here, the precision of nanofabrica-tion (typical 100 nm) allows for the addressability of specificatoms, as ground state sizes will be extremely small and pre-cisely aligned.

As in QIP proposals for neutral atoms, the internal state isthe qubit; another component of the integrated chip will bethe detection of internal degrees of freedom, as well as themere existence of the atom at a specific location. Here, onecan learn from single-molecule spectroscopy using nano-opticsand cavity QED. As an example, one can imagine havinga light-wave guide interferometer on the chip which willmeasure the phase shift introduced by the atoms in the microtrap, similar to environmental sensors. One can also imaginefabricating a microscopic cavity coupled to two wave guides.An atom in the cavity will influence the transmission and leadto a detectable signal. A different approach to detecting theatoms can be a scanning probe technique, in which the probeis a microsphere cavity with one of its whispering gallerymodes tuned close to the resonance frequency of the atoms.An atom in the evanescent wave of the microsphere cavity willinfluence the transmission/coupling probability of light into themicrosphere and lead to a detectable signal.

All of these, and probably even the light sources, could beon-board a self-contained chip. This would involve sophisti-cated 3-D nanofabrication and the integration of a diversity ofelectronic and optical elements.

Such a robust and easy to useatom chip device wouldmake possible advances in many different fields of quantum

(a) (b) (c) (d)

Fig. 14. Transfer from a large trap formed by two “U”-shaped wires to one thinwire. The position of the surface-mounted wires and equipotential lines for thetrapping potential are shown. (a) The large 200-�m “U”-traps carry a current of2 A and the small 10�m wire 300 mA. (b) An intermediate stage in the transferto the 10-�m trap. The current in the large “U”-traps is decreased to 0.5 A. (c)The large “U”-traps are now switched off and the transfer to the small 10-�mtrap is complete. (d) By increasing the bias field, the trap can be compressedfurther.

optics—from applications such as clocks and sensors , toimplementations of quantum information processing andcommunication.

Finally, we note that a European collaboration named AC-QUIRE (which we are the coordinators of), aiming at buildingtheseatom chips, has just been financed by the European Unionas part of the Fifth-Framework Program. Indeed, we look for-ward to many exciting developments in this field.

ACKNOWLEDGMENT

The authors would like to thank A. Mitterer for help in the ex-periments. Atom chips used in the preparation of this work andin the actual experiments were fabricated at the Institut für Fes-tkörperelektronik, Technische Universität Wien, Austria, andthe Sub-micron Center, Weizmann Institute of Science, Israel.The authors thank E. Gornik, C. Unterrainer, and I. Bar-Josephof these institutions for their assistance. They also thank R. Cotéfor his theoretical insights regarding atom collisions. Last, butnot least, they gratefully acknowledge P. Zoller, who is respon-sible for the theoretical vision.

REFERENCES

[1] Y. Imry, Introduction to Mesoscopic Physics. Oxford, U.K.: OxfordUniv. Press, 1987.

[2] E. Buks, R. Schuster, M. Heilblum, D. Mahalu, and V. Umansky, “De-phasing in electron interference by ‘which-path’ detector,”Nature, vol.391, pp. 871–874, 1998.

[3] J. Schmiedmayer, “Quantum wires and quantum dots for neutral atoms,”Eur. Phys. J. D., vol. 4, pp. 57–62, 1998.

[4] J. Denschlag, D. Cassettari, A. Chenet, S. Schneider, and J. Schmied-mayer, “A neutral atom and a wire: Toward mesoscopic atom optics,”Appl. Phys. B, vol. 69, pp. 291–301, 1999.

[5] E. A. Hinds and I. G. Hughes, “Magnetic atom optic mirrors: Mirrors,guides, traps, and chips for atoms,”J. Phys. D, vol. 32, pp. R119–R146,1999.

[6] J. Denschlag, D. Cassettari, and J. Schmiedmayer, “Guiding neutralatoms with a wire,”Phys. Rev. Lett., vol. 82, pp. 2014–2017, 1999.

[7] J. Denschlag, G. Umshaus, and J. Schmiedmayer, “Probing a singularpotential with cold atoms: A neutral atom and a charged wire,”Phys.Rev. Lett., vol. 81, pp. 737–741, 1998.

[8] J. Schmiedmayer, “Guided matter waves, an atom and a current,” inTech. Dig. XVIII Int. Conf. Quantum Electron.. ser. 1992, G. Magerl,Ed. Vienna, Austria: Technische Univ. Wien, 1992, vol. 9, p. 284.

[9] J. Schmiedmayer, “A wire trap for neutral atoms,”Appl. Phys. B, vol.60, pp. 169–179, 1995.

[10] J. Schmiedmayer, “Guiding and trapping a neutral atom on a wire,”Phys. Rev. A, vol. 52, pp. R13–R16, 1995.

1376 IEEE JOURNAL OF QUANTUM ELECTRONICS, VOL. 36, NO. 12, DECEMBER 2000

[11] A. Haase, D. Cassettari, B. Hessmo, and J. Schmiedmayer, “Trappingneutral atoms with a wire”, submitted for publication.

[12] J. D. Weinstein and K. Libbrecht, “Microscopic magnetic traps for neu-tral atoms,”Phys. Rev. A, vol. 52, pp. 4004–4009, 1995.

[13] M. Drndic, K. S. Johnson, J. H. Thywissen, M. Prentiss, and R. M. West-ervelt, “Micro-electromagnets for atom manipulation,”Appl. Phys. Lett.,vol. 72, pp. 2906–2908, 1998.

[14] J. H. Thywissen, “Microfabricated magnetic waveguides for neutralatoms,”Eur. Phys. J. D, vol. 7, pp. 361–367, 1999.

[15] J. Fortagh, A. Grossmann, and C. Zimmermann, “Miniaturized wire trapfor neutral atoms,”Phys. Rev. Lett., vol. 81, pp. 5310–5313, 1998.

[16] J. Reichel, W. Hänsel , and T. W. Hänsch, “Atomic micromanipulationwith magnetic surface traps,”Phys. Rev. Lett., vol. 83, pp. 3398–3401,1999.

[17] D. Müller , D. Z. Anderson, R. J. Grow, P. D. D. Schwindt, and E. A. Cor-nell, “Guiding neutral atoms around curves with lithographically pat-terned current-carrying wires,”Phys. Rev. Lett., vol. 83, pp. 5194–5197,1999.

[18] N. H. Dekkeret al., “Guiding neutral atoms on a chip,”Phys. Rev. Lett.,vol. 84, pp. 1124–1127, 2000.

[19] D. Jaksch, H.-J. Briegel, J. I. Cirac, C. W. Gardiner, and P. Zoller, “En-tanglement of atoms via cold controlled collisions,”Phys. Rev. Lett., vol.82, pp. 1975–1978, 1999.

[20] T. Calarcoet al., “Quantum gates with neutral atoms: Controlling col-lisional interactions in time-dependent traps,”Phys. Rev. A, vol. 61, p.022 304, 2000.

[21] H.-J. Briegel, T. Calarco, D. Jaksch, J. I. Cirac, and P. Zoller, “Quantumcomputing with neutral atoms,”J. Mod. Opt., vol. 47, p. 415, 2000.

[22] T. Pellizzari, S. A. Gardiner, and P. Zoller, “Decoherence, continuousobservation, and quantum computing: A cavity QED model,”Phys. Rev.Lett., vol. 75, pp. 3788–3791, 1995.

[23] S. J. Van Enk, J. I. Cirac, and P. Zoller, “Photonic channels for quantumcommunication,”Science, vol. 279, pp. 205–208, 1998.

[24] “Special issue on quantum information,”Phys. World, vol. 11, no. 3, p.33, 1998.

[25] M. B. Plenio and V. Vedral, “Entanglement in quantum informationtheory,”Contemp. Phys., vol. 39, pp. 431–466, 1998.

[26] V. Vedral and M. B. Plenio, “Basics of quantum computation,”Prog.Quantum Electron., vol. 22, pp. 1–40, 1998.

[27] A. M. Steane, “Quantum computing,”Rept. Prog. Phys., vol. 61, pp.117–173, 1998.

[28] D. Bouwmeester, A. Ekert, and A. Zeilinger,The Physics of QuantumInformation. New York: Springer-Verlag, 2000.

[29] Ch. Bennet and D. DiVincenzo, “Quantum information and computa-tion,” Nature, vol. 404, pp. 247–255, 2000.

[30] R. Folmanet al., “Controlling cold atoms using nanofabricated surfaces:Atom chips,”Phys. Rev. Lett., vol. 84, pp. 4749–4752, 2000.

[31] J. Schmiedmayer and J. Habilitationsschrift, “Optics, interferometry,and fundamental experiments with massive particles,” Univ. Innsbruck,Innsbruck, Austria, 1996.

[32] J. Denschlag and J. Schmiedmayer,Proc. Int. Quantum Coherence Con-ference. Boston, MA: World Scientific, 1997.

[33] O. Stern, “Ein Weg zur experimentellen Prfung der Richtungsquantelungim Magnetfeld,”Z. Phys., vol. 7, p. 249, 1921.

[34] O. Stern and W. Gerlach, “Der experimentelle Nachweis des magnetis-chen Moments des Silberatoms,”Z. Phys., vol. 8, p. 110, 1921.

[35] V. V. Vladimirskii, “Magnetic mirrors, channels and bottels for cold neu-trons,”Sov. Phys. JETP, vol. 12, pp. 740–746, 1961.

[36] G. I. Opat, S. J. Wark, and A. Cimmino, “Electric and magnetic mir-rors and gratings for slowly moving neutral atoms and molecules,”Appl.Phys. B, vol. 54, pp. 396–401, 1992.

[37] T. Bergman, G. Erez, and H. J. Metcalf, “Magnetostatic trapping fieldsfor neutral atoms,”Phys. Rev. A., vol. 35, pp. 1535–1546, 1987.

[38] C. S. Adams, M. Sigel, and J. Mlynek, “Atom optics,”Phys. Rep., vol.240, pp. 143–210, 1994.

[39] P. Berman,Atom Interferometry. New York: Academic, 1997.[40] J. M. Gerton, C. A. Sackett, B. J. Frew, and R. G. Hulet, “Dipolar relax-

ation collisions in magnetically trappedLi,” Phys. Rev. A, vol. 59, pp.1514–1516, 1999.

[41] M. H. Anderson, J. R. Ensher, M. R. Matthews, C. E. Wieman, and E. A.Cornell, “Observation of Bose–Einstein condensation in a dilute atomicvapor,”Science, vol. 269, pp. 198–201, 1995.

[42] K. B. Davis et al., “Bose–Einstein condensation in a gas of sodiumatoms,”Phys. Rev. Lett., vol. 75, pp. 3969–3973, 1995.

[43] M.-O. Meweset al., “Collective excitations of a Bose–Einstein conden-sate in a magnetic trap,”Phys. Rev. Lett., vol. 77, pp. 988–991, 1996.

[44] C. C. Bradley, C. A. Sacket, and R. G. Hulet, “Bose–Einstein condensa-tion of lithium: Observation of limited condensate number,”Phys. Rev.Lett., vol. 78, pp. 985–989, 1997.

[45] C. C. Bradley, C. A. Sackett, J. J. Tollett, and R. G. Hulet, “Evidenceof Bose–Einstein condensation in an atomic gas with attractive interac-tions,” Phys. Rev. Lett., vol. 75, pp. 1687–1690, 1995.

[46] The Bose–Einstein condensation homepage [Online]. Available:http://amo.phy.gasou.edu/bec.html

[47] The condensed matter Los Alamos Archive [Online]. Available:http://xxx.lanl.gov/

[48] W. H. Wing, “On neutral particle trapping in quasistatic electromagneticfields,” Prog. Quantum Electron., vol. 8, pp. 181–199, 1984.

[49] W. Ketterle and D. E. Pritchard, “Trapping and focusing ground stateatoms with static fields,”Appl. Phys. B, vol. 54, pp. 403–406, 1992.

[50] V. Vultetic, T. Fischer, M. Praeger, T. W. Hänsch , and C. Zimmermann,“Microscopic magnetic quadrupole trap for neutral atoms with extremeadiabatic compression,”Phys. Rev. Lett., vol. 80, pp. 1634–1637, 1998.

[51] B. Kramer, Ed.,Quantum Coherence in Mesoscopic Systems. ser. NATOASI Series B Physics. New York: Plenum, 1991, vol. 254.

[52] J. Denschlag, “Kalte Atome in singulären Potentialen,” PhD disserta-tion, Univ. Innsbruck, Innsbruck, Austria, 1998.

[53] E. Andersson, M. T. Fontenelle, and S. Stenholm, “Quantum statistics ofatoms in microstructures,”Phys. Rev. A, vol. 59, pp. 3841–3850, 1999.

[54] D. Cassettari, A. Chenet, J. Denschlag, St. Schneider, and J. Schmied-mayer, “Towards mesoscopic physics with cold atoms,” inICOLS 99,Innisbruck, Austria, June 1999.

[55] Q. A. Turchetteet al., “Heating of trapped ions from the quantum groundstate,”Phys. Rev. A, vol. 61, p. 063 418, 2000.

[56] Ch. Rooset al., “Quantum state engineering on an optical transition anddecoherence in a paul trap,”Phys. Rev. Lett., vol. 83, pp. 4713–4716,1999.

[57] C. Henkel, S. Pötting, and M. Wilkens, “Loss and heating of particles insmall and noisy traps,”Appl. Phys. B, vol. 69, pp. 379–387, 1999.

[58] C. Henkel and M. Wilkens, “Heating of trapped atoms near thermal sur-faces,”EuroPhys. Lett., vol. 47, pp. 414–420, 1999.

[59] S. Gov, S. Shtrikman, and H. Thomas, “Magnetic trapping of neutralparticles: Classical and quantum-mechanical study of a ioffe-pritchardtype trap,”, Los Alamos e-print archive: quant-ph/9 907 044, 1999.

[60] C. V. Sukumar and D. M. Brink, “Spin-flip transitions in a magnetictrap,” Phys. Rev. A, vol. 56, pp. 2451–2453, 1997.

[61] E. A. Hinds and C. Eberlein, “Quantum propagation of neutral atoms ina magnetic quadrupole guide,”Phys. Rev. A, vol. 61, p. 033 614, 2000.

[62] J. Södinget al., “Three-body decay of a rubidium Bose–Einstein con-densate,”Appl. Phys. B, vol. 69, pp. 257–261, 1999.

[63] J. Söding , D. Guéry-Odelin , P. Desbiolles, G. Ferrari, and J. Dalibard,“Giant spin relaxation of an ultracold cesium gas,”Phys. Rev. Lett., vol.80, pp. 1869–1872, 1998.

[64] D. Guéry-Odelin , J. Söding , P. Desbiolles, and J. Dalibard, “IsBose–Einstein condensation of atomic cesium possible?,”EuroPhys.Lett., vol. 44, pp. 25–30, 1998.

[65] E. A. Burt et al., “Coherence, correlations, and collisions: What onelearns about Bose–Einstein condensates from their decay,”Phys. Rev.Lett., vol. 79, pp. 337–340, 1997.

[66] E. L. Raab, M. Prentiss, A. Cable, S. Chu, and D. E. Pritchard, “Trappingof neutral sodium atoms with radiation pressure,”Phys. Rev. Lett., vol.59, pp. 2631–2634, 1987.

[67] K. I. Lee, J. A. Kim, H. R. Noh, and W. Jhe, “Single-beam atom trapin a pyramidial and conical hollow mirror,”Opt. Lett., vol. 21, pp.1177–1179, 1996.

[68] R. Côté , private communication.[69] R. Côté , E. Timmermans, and P. Tommasini, “Superfluidity and fes-

hbach resonances in BEC,” inNew Directions in Atomic Physics, P.Whelanet al., Eds. New York: Kluwer/Plenum, 1999, p. 71.

[70] E. Timmermans and R. Côté , “Superfluidity in sympathetic coolingwith atomic Bose–Einstein condensates,”Phys. Rev. Lett., vol. 80, pp.3419–3423, 1998.

[71] R. Côté , A. Dalgarno, H. Wang, and W. C. Stwalley, “Potassium scat-tering lengths and prospects for Bose–Einstein condensation and sym-pathetic cooling,”Phys. Rev A, vol. 57, pp. R4118–R4121, 1998.

[72] M. Olshanii, “Atomic scattering in the presence of an external confine-ment and a gas of impenetrable bosons,”Phys. Rev. Lett., vol. 81, pp.938–941, 1998.

[73] D. Cassettari, B. Hessmo, R. Folman, T. Maier, and J. Schmiedmayer,“A beam splitter for guided atoms,”, submitted for publication.

[74] D. Jaksch, C. Bruder, J. I. Cirac, C. W. Gardiner, and P. Zoller,“Cold bosonic atoms in optical lattices,”Phys. Rev. Lett., vol. 81, pp.3108–3111, 1998.

BARTENSTEINet al.: ATOMS AND WIRES: TOWARDS ATOM CHIPS 1377

Markus Bartenstein, photograph and biography not available at the time ofpublication.

Donatella Cassettari, photograph and biography not available at time of pub-lication.

Tommaso Calarcophotograph and biography not available at the time of pub-lication.

Alexander Chenet, photograph and biography not available at the time of pub-lication.

Ron Folman, photograph and biography not available at the time of publication.

Karolina Brugger , photograph and biography not available at the time of pub-lication.

Albrecht Haase, photograph and biography not available at time of publication.

Eugen Hartungen, photograph and biography not available at time of publica-tion.

Björn Hessmo, photograph and biography not available at the time of publica-tion.

Alexander Kasper, photograph and biography not available at the time of pub-lication.

Peter Krüger, photograph and biography not available at the time of publica-tion.

Thomas Maier, photograph and biography not available at time of publication.

Fritz Payr , photograph and biography not available at the time of publication.

Stephan Schneider, photograph and biography not available at the time of pub-lication.

Jörg Schmiedmayer, photograph and biography not available at the time ofpublication.