Quantum Non-Demolition Detection of Polar Molecule Complexes: Dimers, Trimers, Tetramers
aa r X i v : . [ qu a n t - ph ] N ov Quantum Non-Demolition Detection of Polar Molecule Complexes:Dimers, Trimers, Tetramers
Igor B. Mekhov
University of Oxford, Department of Physics, Clarendon Laboratory, Parks Road, Oxford OX1 3PU, UK
The optical nondestructive method for in situ detection of the bound states of ultracold polarmolecules is developed. It promises a minimally destructive measurement scheme up to a physicallyexciting quantum non-demolition (QND) level. The detection of molecular complexes beyond simplepairs of quantum particles (dimers, known, e.g., from the BEC-BCS theory) is suggested, includingthree-body (trimers) and four-body (tertramers) complexes trapped by one-dimensional tubes. Theintensity of scattered light is sensitive to the molecule number fluctuations beyond the mean-densityapproximation. Such fluctuations are very different for various complexes, which leads to radicallydifferent light scattering. This type of research extends ”quantum optics of quantum gases” to thefield of ultracold molecules. Merging the quantum optical and ultracold gas problems will advancethe experimental efforts towards the study of the light-matter interaction at its ultimate quantumlevel, where the quantizations of both light and matter are equally important.
PACS numbers: 03.75.Lm, 42.50.-p, 05.30.Jp, 32.80.Pj
I. INTRODUCTION
The study of ultracold polar molecules attracts sig-nificant attention because of their long-range anisotropicinteraction, which can lead to the creation of exotic quan-tum phases of ultracold particles. The phase diagram isexpected to be much richer than that for atomic gaseswith the short-range interaction. Recently, the existenceof several few-body bound states of polar molecules hasbeen proved for a low dimensional geometry [1, 2]. Be-ing important in the context of few-body physics, thoseresults can help to get insight into the many-body prob-lems as well [3], where the elementary few-body build-ing blocks can play a crucial role. For example, goingbeyond the two-body complexes and predicting the exis-tence of bound states consisting of more than two parti-cles (as trimers and tetramers), those results can modifythe standard description of the BCS-BEC crossover incertain systems, which is usually based on the picture ofpairs (i.e., the dimers) only. In contrast to extensivelystudied Efimov-type states [4] with short-range contactinteraction, the states appearing due to the anisotropiclong-range dipole-dipole interaction are less investigated.The use of optical methods to detect the states of polarmolecules promises the development of non-destructive insitu measurement schemes, which can be used to probethe system dynamics in real time. Moreover, as has beensuggested in Refs. [1, 2], the optical non-destructive de-tection of ultracold molecules can be developed up tothe physically exciting quantum non-demolition (QND)level. Such an ultimately quantum measurement schemeaffects the quantum state in a minimally destructive wayand triggers the intriguing fundamental questions aboutthe quantum measurement back-action and the entan-glement between the light and ultracold molecules [5–8].Other probing methods such as time-of-flight measure-ments or lattice shaking [9] are usually destructive. Thispaper provides further details about the QND measure- zx y Jj Scattered lightProbe light D d Probe light
Detection
Tube A Tube B
FIG. 1: Setup. The molecules with dipole moment d aretrapped in the potential of two 1D tubes. The probe anddetection are in the plane perpendicular to the tubes. ments in ultracold molecular gases [1, 2]. Focusing on asimple physical picture of the light-matter interaction, weshow how the main characteristics of the light scatteringcan be estimated analytically, using a simple statisticalapproach. Moreover, those results should be valid even inthe many-body systems with a large number of ultracoldmolecules, at least, in the low-density regime. This typeof research extends the field of ”quantum optics of quan-tum gases” [5, 10] for the molecular species. Merging thequantum optical and ultracold gas problems will advancethe experimental efforts [11–17] towards the study of thelight-matter interaction at its ultimate quantum level. II. LIGHT SCATTERING FROM ULTRACOLDMOLECULES IN 1D
We consider ultracold dipolar molecules trapped in thepotential of two one-dimensional (1D) tubes (cf. Fig. 1).As described in details in Refs. [1, 2], even for the repul-sive interaction between the molecules within each tube,they can form bound complexes due to the attractivedipole-dipole interaction with the molecules in a differ-ent tube. Thus, several repulsing molecules in one tubecan be bound by the presence of a molecule in anothertube, with which they interact attractively. The asso-ciation of molecules into various stable complexes wasproved: dimers ”1-1” (with one molecule in each tube),trimers ”1-2” (with one molecule in one tube and twomolecules in the other tube) and tetramers ”1-3” (withone molecule in one tube and three molecules in the othertube) and ”2-2” (with two molecules in each tube).The few-body complexes can be detected using lightscattering. Recently, several nondestructive (in the senseof the quantum non-demolition, QND) schemes for mea-suring the properties of the many-body states in ultracoldgases observing scattered light have been proposed [5, 18–23]. Among them, the method developed in Refs. [5, 18–20] is the most relevant to the present system, as it explic-itly uses the sensitivity of light scattering to the relativeposition of the particles forming a complex. This is dueto the constructive or destructive interference of the lightwaves scattered from the different particles. This methodcan be directly applied for extended periodic structures(many equidistantly spaced tubes or layers) and many-body systems, which makes the experimental realizationpromising. In contrast to Refs. [21–23] and experiments[24, 25] with spin ensembles, the original proposal of Refs.[18–20] does not rely on any state-selective (e.g., spin-selective) light scattering, but is sensitive to the particleposition.We consider the scattering of the probe light with theamplitude given by the Rabi frequency Ω p = d E p / ¯ h ( E p is the probe-light electric field amplitude and d is the in-duced dipole moment), cf. Fig. 1. To increase the signal,the scattered light can be collected by a cavity, and thephotons leaking from the cavity are then measured. Al-ternatively, the measurement of photons scattered can bemade in a far-field region without the use of a cavity.Using the approach of the second quantization for themolecule-field operator (as it was formulated for atomsin Refs. [18, 19]), the amplitude of the scattered light(i.e., the annihilation operator of the scattered photon)is given by a s = C Z d r ˆΨ † ( r ) u ∗ s ( r ) u p ( r ) ˆΨ( r ) , (1)where ˆΨ( r ) is the matter-field operator at the point r .For the free space scattering, the value of C correspondsto the Rayleigh scattering [26]. Adding a cavity to thesetup the scattering is increased and C = − ig s Ω p / (∆ a κ ) with κ being the cavity decay rate, g s is the molecule-light coupling constant, and ∆ a is the light detuning fromthe resonance, cf. Refs. [5, 18–20]. In Eq. (1), u p,s ( r ) arethe mode functions of probe and scattered light, whichcontain the information about the propagation directionsof probe and scattered light waves with respect to thetube direction. For the simplest case of two travelinglight waves, the product of two mode functions takes thewell-known form from classical light scattering theory: u ∗ s ( r ) u p ( r ) = exp [ i ( k p − k s ) r ], where k p,s are the probeand scattered light wave vectors.One can express the matter-field operator in the basisof the functions corresponding to the transverse distribu-tion of molecules within two tubes A and B:ˆΨ( r ) = ˆΨ A ( x ) w ( ρ − ρ A ) + ˆΨ B ( x ) w ( ρ − ρ B ) , (2)where ˆΨ A,B ( x ) are the matter-field operators within eachtube with the coordinate x alone the tube, where themolecules can move (cf. Fig. 1); w ( ρ ) gives the distri-bution of a molecule in the transverse direction ( ρ is thetransverse coordinate). Substituting this expression inEq. (1), we can describe the light scattering taking intoaccount the possible overlap of the molecules betweentwo tubes (overlapping w ( ρ − ρ A ) and w ( ρ − ρ B )) andthe nontrivial overlap between the molecule distribution w ( ρ ) and the light modes u p,s ( r ). However, followingRefs. [1, 2], we assume that two tubes do not overlap atall, and they are well localized with respect to the lightwave.Thus, after several assumptions (the small tube radius,far off-resonant light scattering, detection in the far fieldzone), the light scattering has a simple physical interpre-tation. The scattered light amplitude is given by the sumof the light amplitudes, scattered from each molecule (cf.Fig. 1). Each term has a phase and amplitude coeffi-cient depending on the position of the molecule as wellas on the direction and amplitude of the incoming andoutgoing light waves: a s = C X i = A,B Z dx ˆ n i ( x ) u ∗ s ( x, ρ i ) u p ( x, ρ i ) , (3)where the sum is over two tubes A and B, ˆ n i ( x ) =ˆΨ † i ( x ) ˆΨ i ( x ) is the operator of particle linear density. InEq. (3), u p,s ( x, ρ i ) are the mode functions of probe andscattered light at the tube positions ρ A,B .Equation (3) is valid for any optical geometry and candescribe the angular distribution of the scattered light.However, an important conclusion of Refs. [5, 18–20] wasthat some information about the many-body state canbe obtained even by a simple measurement of the photonnumber scattered at a single particular angle, which isfully enough for our purpose. Moreover, as it was shown,the particularly convenient angle of measurement corre-sponds to the direction of a diffraction minimum, ratherthan Bragg angle (diffraction maximum). At the direc-tions of diffraction minimum any classical (possibly verystrong) scattering is suppressed, and the light signal ex-clusively reflects the quantum fluctuations of the parti-cles.We now fix the optical geometry as follows (cf. Fig.1). The incoming probe light is a traveling or stand-ing wave propagating at the direction perpendicular tothe tubes, which gives u p ( r ) = R ( x ) exp( ik p y ) (for thetraveling wave) or u p ( r ) = R ( x ) cos( k p y ) (for the stand-ing wave) and includes the transverse probe profile R ( x )of an effective width W . To perform the measurementsat the direction of a diffraction minimum, the scat-tered light is measured along z direction. For the freespace detection, or the traveling-wave cavity, this gives u s ( r ) = exp( ik s z ), while for the case of a standing wavecavity, u s ( r ) = cos( k s z ). Without loss of generality, wecan assume u s ( r ) = 1 at the tube position z = 0. Theabsolute values of the wave vectors are equal to theirvacuum quantities k p,s = 2 π/λ light .An important property of such a configuration (illumi-nation and detection at the directions perpendicular tothe tubes), is that all molecules within one tube scatterlight with the same phase independently of their longi-tudinal position x within the tube. Thus, the light scat-tered from the molecules within one tube interferes fullyconstructively. As a consequence, all molecules withintwo different tubes scatter light with a fixed phase dif-ference with respect to each other. Due to this fact, theaveraging over the probabilistic position of the complexdoes not involve the light phase and all complexes of thesame type scatter light identically. Moreover, averagingover the probabilistic relative positions within each com-plex does not involve the dependence on the light phaseas well. At other directions, both those kinds of phaseaveraging are important and would decrease the opticalsignal and the distinguishability of the complex types.The simple scattering picture also allows the generaliza-tion of the model for an array of several tubes.The operator of the light amplitude reduces to a s = C (cid:16) u p ( y A ) ˆ N A ( W ) + u p ( y B ) ˆ N B ( W ) (cid:17) , (4)where ˆ N A,B ( W ) are the operators of the effective par-ticle numbers in the tubes A and B within the regionilluminated by the laser beam,ˆ N A,B ( W ) = Z ∞−∞ ˆ n A,B ( x ) R ( x ) dx. (5)If the laser profile can be approximated by a constant inthe interval ( − W/ , W/ N A,B ( W ) ex-actly correspond to the atom number operators in twotubes within the laser beam.The classical condition of the diffraction minimumis fulfilled, when the expectation value of the light-amplitude operator (4) is zero due to the perfect cancela-tion of the expectation values of two terms in Eq. (4) (i.e.the total destructive interference between the scatterersin two tubes). This is achieved for u p ( y B ) /u p ( y A ) = −h ˆ N A i / h ˆ N B i . We introduce the atom number ratio α = h ˆ N A i / h ˆ N B i . For the equal mean atom numbers (thefew-body complexes 1-1 and 2-2), the optical geometryshould be chosen such that u s ( y B ) /u s ( y A ) = −
1, whichcan be achieved if, e.g., the tube spacing is the half ofthe light wavelength, ∆ = λ light /
2. For the few-bodycomplex 1-2, α = 1 /
2, and the diffraction minimum isachieved if the light wavelength and tube spacing satisfythe condition cos( k p y B ) / cos( k p y A ) = − /
2. This can beachieved, e.g., if the position of the tube A corresponds tothe antinode of the standing wave cos( k p y A ) = 1, whilethat of tube B corresponds to k p y B = 2 π/ π/ λ light / λ light /
3. Similarly, for the1-3 complex, that ratio can be ∆ ≈ . λ light or 0 . λ light .All those example ratios can be indeed larger, taking intoaccount the periodicity of the light wave.The expectation value of number of photons scatteredat the direction of diffraction minimum n Φ is then givenby n Φ = h a † s a s i = | C | | u p ( y A ) | (cid:28)(cid:16) ˆ N A ( W ) − α ˆ N B ( W ) (cid:17) (cid:29) , (6)where u p ( y A ) can be easily chosen as 1. This expressionmanifests that the number of photons scattered in thediffraction minimum is proportional to the second mo-ment of the ”rated” particle number difference betweentwo tubes in the laser-illuminated region. The mean lightamplitude is sensitive to the mean values of the particlenumber and is precisely zero at the diffraction minimum: h a s i ∼ D(cid:16) ˆ N A ( W ) − α ˆ N B ( W ) (cid:17)E = 0. However, in gen-eral, the photon number (6) is non-zero. It directly re-flects the particle number fluctuations and correlationsbetween the tubes. Thus, the number of photons reflectsthe quantum state of ultracold molecules. III. APPLICATIONS FOR DIMERS, TRIMERSAND TETRAMERS
In Ref. [1, 2], we presented the results of numericalsimulations for light scattering from few-body complexesfor particular parameters. We have shown that, whilethe photon number in the diffraction minimum is zerofor a bound state, it immediately increases, when thecomplex dissociates into a smaller complex and a freemolecule. Although after such a dissociation, the meanparticle number stays the same (and the light amplitudewould not change), the fluctuations of the particle num-ber inside the laser beam change strongly after the disso-ciation: instead of one bound complex, one gets anothercomplex and a free particle, whose positions are uncorre-lated. The particle fluctuations increase the intensity ofthe scattered light.In this paper, we demonstrate that the values of lightintensity for stable complexes and free molecules can be free moleculesfree molecules free molecules N /34 N /92 N /903 N /2 NN /202 NN (a) P ho t on nu m be r N /310 N /98 N /9 (c) P ho t on nu m be r (b) P ho t on nu m be r FIG. 2: QND measurement of ultracold polar molecule com-plexes. Intensity of scattered light (i.e. the relative pho-ton number n Φ / | C | ) depending on the existence of variousfew-body complexes. The variable on the horizontal axis isschematic. It can correspond to several parameters, whichallow to scan the system through the regimes, were differentcomplexes exist (e.g., the dipole orientation angle or dipole-dipole interaction strength as shown in Refs. [1, 2]). (a) Disso-ciation of dimers ”1-1” into free molecules corresponds to thechange of light intensity from n Φ / | C | = 0 to n Φ / | C | = 2 N .(b) Dissociation of trimers ”1-2” into dimers ”1-1” and freemolecules, and then into all free molecules corresponds tothe intensity jumps as n Φ / | C | = 0, n Φ / | C | = N/ n Φ / | C | = 3 N/
2. (c) Dissociation of tetramers ”1-3” intotrimers ”1-2” and free molecules, then into dimers ”1-1”and free molecules, and finally into all free molecules corre-ponds to the intensity values n Φ / | C | = 0, n Φ / | C | = 2 N/ n Φ / | C | = 6 N/ n Φ / | C | = 12 N/
9. Inversely, the asso-ciation of those complexes will correspond to the suppressionof light scattered into the diffraction minimum. estimated analytically using the statistical calculations.Such estimations are valid for many molecules in eachtube (at least in the low-density regime) and agree wellwith the numerical simulations made for real systems,but the tiny number of molecules per tube [1, 2]. Theapproach developed in this papers also gives a possibilityto get a deeper physical insight into the problem. Al-though the development of modern trapping techniquestargets the manipulation of ultracold atoms at a single-particle level [27], the many-particle realization is stillmore realistic. Expression (6) can be written in the form n Φ / | C | = h ( ˆ N A − α ˆ N B ) i = h ˆ N A i + h ˆ N B i − α h ˆ N A ˆ N B i , (7)which underlines the correlations between the moleculenumbers in two different tubes.Let us start with the example of dimers ”1-1” andconsider the equal number of molecules in two tubes( h ˆ N A i = h ˆ N B i = N , α = 1). When all molecules arestrongly bound into dimers, they appear within the laserbeam only in pairs, or do not appear there at all. Thus,the fluctuations of the molecule number difference is zero(one can think about the two number operators as identi-cal ones, ˆ N A = ˆ N B , i.e., all their moments coincide) andso does the light intensity: n Φ / | C | = h ( ˆ N A − α ˆ N B ) i = h ( ˆ N A − α ˆ N B ) i = 0. On the other hand, when a dimer dis-sociates into two independent free molecules, the two op-erators are different, and the term with the intertube cor-relation function in Eq. (7) decorrelates into a product: h ˆ N A ˆ N B i = h ˆ N A ih ˆ N B i = N . One can assume that thenumber fluctuations of the independent free moleculesare Poissonian, h ˆ N A,B i = h ˆ N A,B i + h ˆ N A,B i = N + N .Then, the number of scattered photons Eq. (7) gets n Φ / | C | = 2 N .Therefore, we see that the light intensity jumps fromzero to n Φ / | C | = 2 N , when the dimers dissociate intofree molecules. Such a change of light intensity for twodifferent phases of ultracold molecules is schematicallydemonstrated in Fig. 2(a). Physically, the stronglybound complex does not scatter light, because the ge-ometry corresponds to the diffraction minimum. Thus,the fluctuation of the complex number within the laserbeam does not change the light intensity (it is zero if boththe complex is within the beam, and, obviously, outsidethe beam). However, when the complex dissociate intotwo independent species (two free molecules in this ex-ample), the species can be within or outside the beam in-dependently from each other. Thus, the condition of thetotal diffraction minimum is not satisfied anymore, be-cause, probabilistically, the numbers of molecules withinthe beam can be nonequal in two tubes (even though theyare always equal in average) and the complete destructiveinterference of light is not possible anymore. Note, thatthis result agrees very well with the numerical calcula-tions presented in Ref. [2] carried out for two moleculesin two tubes. Those numerical results indeed show notonly the constant values of the light intensity, but alsodescribe the continuous transition between them, whenthe dimer dissociates.Let us now consider the case of trimers ”1-2”, whenthe populations of two tubes are imbalanced: h ˆ N A i = N , h ˆ N B i = 2 N , α = 1 /
2. When the molecules are stronglybound into a trimers, they appear in the laser beamonly all three together, or do not appear at all (Herewe indeed neglect the small effects when the trimer islarge and can overlap with the laser beam only par-tially. This however could be captured by the numeri-cal simulations in Ref. [1, 2], and was shown to intro-duce only small corrections to the result.) Therefore,the fluctuations of the operator ( ˆ N A − α ˆ N B ) are zeroand the number of scattered photons is zero as well: n Φ / | C | = h ( ˆ N A − / N B ) i = h ( ˆ N A − / N B ) i = 0.The trimer can dissociate into a dimer ”1-1” and a freeparticle, which are independent from each other. The op-erator of the number of particles in the tube B can besplit into two parts: ˆ N B = ˆ N DB + ˆ N FB , where the op-erator ˆ N B = ˆ N DB corresponds to the molecules, whichform a dimer with another molecule in the tube A, andˆ N B = ˆ N FB corresponds to the free molecules. To cal-culate the expectation value for the photon number, wecan group the molecule number operators in Eq. (7) suchthat they would correspond to the same species (dimersor free molecules). Then, ˆ N A − / N B = ˆ N A − / N DB − / N FB = 1 /
2( ˆ N D − ˆ N F ), where we introduced the op-erators for the number of dimers, ˆ N D = ˆ N DB = ˆ N A , andnumber of free molecules, ˆ N F = ˆ N FB .After introducing the operators for different indepen-dent species (dimers and free molecules), we can calcu-late the expectation value in Eq. (7), assuming that thespecies are uncorrelated ( h ˆ N D ˆ N F i = h ˆ N D ih ˆ N F i = N )and each of them displays the Poissonian fluctuations( h ( ˆ N D,F ) i = h ˆ N D,F i + h ˆ N D,F i = N + N ). The re-sult reads: n Φ / | C | = N/
2. So, we see, how the lightintensity jumps from zero to this non-zero value, whenthe trimer dissociates into a dimer and a free molecule.Those dimers and free molecules can dissociate fur-ther into three independent molecules. Taking into ac-count the mean values of the free molecules in two tubes, h ˆ N FA i = N and h ˆ N FB i = 2 N , the expectation value of thelight intensity reads n Φ / | C | = 3 N/
2. That is, it jumpsfurther upwards.The consecutive dissociation of the trimers is schemat-ically shown in Fig. 2(b). All three phases can be dis-tinguished by the light intensity: it is zero for boundtrimers, proportional to N/ N/ h ˆ N A i = N , h ˆ N B i = 3 N , α = 1 /
3. The nu-merical simulations for that situation were not reportedin Refs. [1, 2]. As before, when the complex is stronglybound, it does not scatter light into the diffraction min-imum and n Φ / | C | = h ( ˆ N A − / N B ) i = 0. The fol-lowing steps of a tetramer dissociation are possible: 1) atrimer ”1-2” and a free molecule, 2) a dimer ”1-1” andtwo free molecules and 3) three free molecules.The tetramers first dissociate into the trimers ”1-2”and free molecules. Proceeding as before, the numberoperator in the tube B can be split into two statisti-cally independent operators: ˆ N B = ˆ N TB + ˆ N FB , whereˆ N TB corresponds to the molecules in the tube B, whichform trimers with molecules in A, and ˆ N FB corresponds to free molecules. As before, we introduces the operatorof the trimer number ˆ N T . All molecules in the tube Aparticipate in the trimer creation: ˆ N A = ˆ N T , while thenumber of molecules forming the trimer in the tube Bis two times larger: ˆ N TB = 2 ˆ N T . Proceeding as before,assuming that the trimers and free molecules are not cor-related ( h ˆ N T ˆ N F i = h ˆ N T ih ˆ N F i = N ), and obey thePoissonian fluctuations, we arrive to the photon numberas n Φ / | C | = h ( ˆ N A − / N B ) i = 2 N/ N A = ˆ N D , and thenumber of molecules from the tube B forming the dimerswill be the same, ˆ N DB = ˆ N D . In this case, the meanvalues are: h ˆ N D i = N , h ˆ N F i = 2 N . The photon numberjumps upwards: n Φ / | C | = h ( ˆ N A − / N B ) i = 2 N/ n Φ / | C | = 4 N/
3. (To derivethis expression, note that h ˆ N FA i = N , while h ˆ N FB i = 3 N ).The dependence of the light intensity on the moleculestate is schematically shown in Fig. 2(c). The plateauswith four different values are expected: n Φ / | C | = 0 forthe tetramers ”1-3”, n Φ / | C | = 2 N/ n Φ / | C | = 6 N/ n Φ / | C | = 12 N/ θ in Fig. 1), or the strength of the dipole-dipole interac-tion between the molecules. Interestingly, in contrast tothe light intensity in the diffraction minimum, the meanlight amplitude would not change at all and would stayzero for all states considered above. This is an exam-ple of a quantum optical problem, where one has a zerolight amplitude h a s i = 0, but non-zero photon number h a † s a s i 6 = |h a s i| due to the matter-induced photon fluc-tuations. IV. CONCLUSIONS
The optical nondestructive scheme for probing boundstates of ultracold polar molecules is presented. Basedon the off-resonant light scattering it promises the insitu measurement of the molecular dynamics in realtime up to a physically exciting QND level. Thedetection of association and dissociation of molecularpairs (dimers), three-body states (trimers) and four-bodystates (tetramers) has been demonstrated. In contrast toother QND schemes [21–25] requiring the state-selective(e.g. spin-selective) light scattering, this method is orig-inally based on the proposal of Refs. [18–20] and is notsensitive to the internal-level structure, which is its ad-vantage. The light scattering directly reflects the relativespatial positions of the complex parts and measures thequantum fluctuations of the molecule numbers beyondthe mean-density approximation. Development of suchQND techniques opens the field of ”quantum optics ofquantum gases” [5, 10] for ultracold molecular gases andraises intriguing questions about the quantum measure-ment back-action and preparation of the exotic many-body phases using the entanglement between the lightand many-body molecular states [5–8, 28]. Merging the quantum optical and ultracold gas problems will advancethe experimental efforts [11–17] towards the study of thelight-matter interaction at its ultimate quantum level,where the quantum natures of both light and matter playequally important roles.
Acknowledgement
The stimulating discussions with B. Wunsch, N. Zinnerand E. Demler are appreciated. This work was supportedby the EPSRC project EP/I004394/1. [1] B. Wunsch, N. T. Zinner, I. B. Mekhov, S.J. Huang, D.-W. Wang, and E. Demler, Phys. Rev. Lett. , 073201(2011).[2] N. T. Zinner, B. Wunsch, I. B. Mekhov, S.J. Huang, D.-W. Wang, and E. Demler, arXiv:1110.1552, to be pub-lished in Phys. Rev. A.[3] M. Dalmonte, P. Zoller, and G. Pupillo, Phys. Rev. Lett. , 163202 (2011).[4] L. Tomio, M. T. Yamashita, T. Frederico, and F. Bringas,Laser Phys. , 1464 (2011).[5] I. B. Mekhov, and H. Ritsch, Phys. Rev. A , 013604(2009).[6] I. B. Mekhov and H. Ritsch, Phys. Rev. Lett. , 020403(2009).[7] I. B. Mekhov and H. Ritsch, Laser Phys. , 694 (2010).[8] I. B. Mekhov and H. Ritsch, Laser Phys. , 1486 (2011).[9] A. Zenesini, H. Lignier, C. Sias, O. Morsch, D. Ciampini,and E. Arimondo, Laser Phys. , 1182 (2010).[10] I. B. Mekhov, C. Maschler, and H. Ritsch, Nature Phys. , 319 (2007).[11] F. Brennecke, T. Donner, S. Ritter, T. Bourdel, M. Kohl,and T. Esslinger, Nature , 268 (2007).[12] Y. Colombe, T. Steinmetz, G. Dubois, F. Linke, D.Hunger, and J. Reichel, Nature , 272 (2007).[13] S. Slama, S. Bux, G. Krenz, C. Zimmermann, and Ph.W. Courteille, Phys. Rev. Lett. , 053603 (2007).[14] S. Ritter, F. Brennecke, C. Guerlin, K. Baumann, T.Donner, T. Esslinger, Appl. Phys. B , 213 (2009).[15] K. Baumann, C. Guerlin, F. Brennecke, and T. Esslinger,Nature , 1301 (2010).[16] C. Weitenberg, P. Schau, T. Fukuhara, M. Cheneau, M.Endres, I. Bloch, and S. Kuhr, Phys. Rev. Lett. ,215301 (2011).[17] H. Miyake, G. A. Siviloglou, G. Puentes, D. E. Pritchard,W. Ketterle, D. M. Weld, Rev. Lett. , 175302 (2011).[18] I. B. Mekhov, C. Maschler, and H. Ritsch, Phys. Rev.Lett. , 100402 (2007).[19] I. B. Mekhov, C. Maschler, and H. Ritsch, Phys. Rev. A , 053618 (2007).[20] I. B. Mekhov and H. Ritsch, Laser Phys. , 610 (2009).[21] I. de Vega, J. I. Cirac, and D. Porras, Phys. Rev. A ,051804(R) (2008).[22] G. M. Bruun, B. M. Andersen, E. Demler, and A. S.Sørensen, Phys. Rev. Lett. , 030401 (2009). [23] T. Roscilde, M. Rodrguez, K. Eckert, O. Romero-Isart,M. Lewenstein, E. Polzik, and A. Sanpera, New J. Phys. , 055041 (2009).[24] C. Sanner, E. J. Su, A. Keshet, W. Huang, J. Gillen,R. Gommers, and W. Ketterle, Phys. Rev. Lett. ,010402 (2011).[25] C. Sanner, E. J. Su, W. Huang, A. Keshet, J. Gillen, andW. Ketterle, arXiv:1108.2017.[26] M. O. Scully and M. S. Zubairy, Quantum Optics , Sec.10.1 (Cambridge University Press, Cambridge, 1997).[27] M. McGovern, T. Gr¨unzweig, A.J. Hilliard, and M. F.Andersen, Laser Phys. Lett. , 78 (2012).[28] C. Maschler, I. B. Mekhov, and H. Ritsch, Eur. Phys. J.D46